From 8c9dcc251221deb99eda1a63fc0689cbb2887d9a Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Tue, 15 Nov 2022 11:08:35 +0530 Subject: [PATCH 001/141] Druid automated quickstart --- examples/bin/druid-automated-quickstart.py | 299 ++++++++++++++++++ examples/bin/run-druid | 33 +- examples/bin/start-druid | 24 ++ .../quickstart/_common/common.jvm.config | 7 + .../_common/common.runtime.properties | 158 +++++++++ .../quickstart/_common/log4j2.xml | 86 +++++ .../quickstart/broker/runtime.properties | 21 ++ .../coordinator-overlord/runtime.properties | 21 ++ .../quickstart/historical/jvm.config | 0 .../quickstart/historical/runtime.properties | 24 ++ .../middleManager/runtime.properties | 21 ++ .../quickstart/router/runtime.properties | 21 ++ 12 files changed, 710 insertions(+), 5 deletions(-) create mode 100644 examples/bin/druid-automated-quickstart.py create mode 100644 examples/bin/start-druid create mode 100644 examples/conf/druid/single-server/quickstart/_common/common.jvm.config create mode 100644 examples/conf/druid/single-server/quickstart/_common/common.runtime.properties create mode 100644 examples/conf/druid/single-server/quickstart/_common/log4j2.xml create mode 100644 examples/conf/druid/single-server/quickstart/broker/runtime.properties create mode 100644 examples/conf/druid/single-server/quickstart/coordinator-overlord/runtime.properties create mode 100644 examples/conf/druid/single-server/quickstart/historical/jvm.config create mode 100644 examples/conf/druid/single-server/quickstart/historical/runtime.properties create mode 100644 examples/conf/druid/single-server/quickstart/middleManager/runtime.properties create mode 100644 examples/conf/druid/single-server/quickstart/router/runtime.properties diff --git a/examples/bin/druid-automated-quickstart.py b/examples/bin/druid-automated-quickstart.py new file mode 100644 index 000000000000..24bee48fbb96 --- /dev/null +++ b/examples/bin/druid-automated-quickstart.py @@ -0,0 +1,299 @@ +import sys +import os +import psutil +from pathlib import Path + +SUPERVISE_CONFIG_FILE_PATH = "../conf/supervise/single-server/quickstart.conf" +QUICKSTART_BASE_CONFIG_PATH = "conf/druid/single-server/quickstart" +HELP_ARG_IDENTIFIER = "help" +COMPUTE_ONLY_ARG_IDENTIFIER = "computeOnly" +RUN_ZK_IDENTIFIER = "runZk" +BASE_CONFIG_PATH_IDENTIFIER = "baseConfigPath" +MEMORY_ARG_IDENTIFIER = "totalMemory" +MEMORY_GIGABYTES_IDENTIFIER = "g" +MEMORY_MEGABYTES_IDENTIFIER = "m" +ARG_SEPARATOR = "=" + +BROKER_SERVICE_NAME = "broker" +ROUTER_SERVICE_NAME = "router" +COORDINATOR_SERVICE_NAME = "coordinator-overlord" +HISTORICAL_SERVICE_NAME = "historical" +MIDDLE_MANAGER_SERVICE_NAME = "middleManager" + +DEFAULT_SERVICES = [ + BROKER_SERVICE_NAME, + ROUTER_SERVICE_NAME, + COORDINATOR_SERVICE_NAME, + HISTORICAL_SERVICE_NAME, + MIDDLE_MANAGER_SERVICE_NAME +] + +SERVICE_MEMORY_DISTRIBUTION_WEIGHT = { + MIDDLE_MANAGER_SERVICE_NAME: 0.25, + ROUTER_SERVICE_NAME: 1, + COORDINATOR_SERVICE_NAME: 18, + BROKER_SERVICE_NAME: 28, + HISTORICAL_SERVICE_NAME: 50 +} + +SERVICE_MEMORY_HEAP_RATIO = { + MIDDLE_MANAGER_SERVICE_NAME: 1, + ROUTER_SERVICE_NAME: 1, + COORDINATOR_SERVICE_NAME: 1, + BROKER_SERVICE_NAME: 0.60, + HISTORICAL_SERVICE_NAME: 0.40 +} + +def check_argument_type(argument, type): + split_args = argument.split(ARG_SEPARATOR) + return split_args[0] == type + +def get_argument_value(argument): + split_args = argument.split(ARG_SEPARATOR) + return split_args[1] + +def parse_arguments(): + num_args = len(sys.argv) + service_list = [] + service_path_list = [] + base_config_path = "" + total_memory = "" + compute_only = False + run_zk = False + + for argument in sys.argv[1:]: + if check_argument_type(argument, COMPUTE_ONLY_ARG_IDENTIFIER): + compute_only = True + elif check_argument_type(argument, RUN_ZK_IDENTIFIER): + run_zk = True + elif check_argument_type(argument, BASE_CONFIG_PATH_IDENTIFIER): + base_config_path = os.path.join(os.getcwd(), get_argument_value(argument)) + elif (check_argument_type(argument, MEMORY_ARG_IDENTIFIER)): + total_memory = get_argument_value(argument) + else: + split_args = argument.split(ARG_SEPARATOR) + service = split_args[0] + + if service not in DEFAULT_SERVICES: + raise Exception(f'{service} is not a valid service name, should be one of {DEFAULT_SERVICES}') + + path = "" + + if len(split_args) == 2: + path = split_args[1] + complete_path = os.path.join(base_config_path, path) + if os.path.exists(os.path.join(complete_path)) is False: + raise Exception(f'Path `{complete_path}` specified for service `{service}` doesn\'t exist ') + service_list.append(service) + service_path_list.append(path) + + if len(service_list) == 0: + # start all services + service_list = DEFAULT_SERVICES + service_path_list = [""] * len(DEFAULT_SERVICES) + run_zk = True + + return base_config_path, total_memory, list(zip(service_list, service_path_list)), run_zk, compute_only + +def should_compute_memory(base_config_path, total_memory, service_config): + # if jvm file is present for any of the services + # it should be present for all services and totalMemory should not be specified + # if totalMemory is given, jvm file shouldn't be present for any service + + jvm_config_count = 0 + for item in service_config: + if item[1] != "": + if Path(f'{base_config_path}/{item[1]}/jvm.config').is_file(): + jvm_config_count += 1 + elif jvm_config_count > 0: + raise Exception('jvm.config file is missing for service {item[0]}, jvm.config should be specified for all the services or none') + + if jvm_config_count > 0 and (jvm_config_count != len(service_config) or total_memory != ""): + if jvm_config_count != len(service_config): + raise Exception("jvm.config file should be present for all services or none") + if total_memory != "": + raise Exception("If jvm.config is given for services, `totalMemory` argument shouldn't be specified") + + return jvm_config_count == 0 + +def compute_system_memory(): + system_memory = psutil.virtual_memory().total # mem in bytes + memory_for_druid = int((system_memory * 0.8) / (1024 * 1024)) + return memory_for_druid + +def convert_total_memory_string(memory): + if memory == "": + computed_memory = compute_system_memory() + print(f'`{MEMORY_ARG_IDENTIFIER}` argument is not specified Druid will use 80% of system memory: {computed_memory}m') + return computed_memory + elif memory.endswith(MEMORY_MEGABYTES_IDENTIFIER): + return int(memory[:-1]) + elif memory.endswith(MEMORY_GIGABYTES_IDENTIFIER): + return 1024 * int(memory[:-1]) + else: + raise Exception('Incorrect format for totalMemory argument, expected format is m or g') + +def build_memory_config_string(heap_memory, direct_memory): + if direct_memory == 0: + return f'-Xms{heap_memory}m -Xmx{heap_memory}m' + return f'-Xms{heap_memory}m -Xmx{heap_memory}m -XX:MaxDirectMemorySize={direct_memory}m' + +def distribute_memory_over_services(service_config, total_memory): + service_memory_config = {} + service_instance_map = {} + + for item in service_config: + service_instance_map[item[0]] = service_instance_map.get(item[0], 0) + 1 + + memory_weight_sum = 0 + for key, value in service_instance_map.items(): + memory_weight_sum += SERVICE_MEMORY_DISTRIBUTION_WEIGHT.get(key) * value + + multiplier = total_memory / memory_weight_sum + + for key, value in service_instance_map.items(): + allocated_memory = SERVICE_MEMORY_DISTRIBUTION_WEIGHT.get(key) * multiplier + heap_memory = SERVICE_MEMORY_HEAP_RATIO.get(key) * allocated_memory + direct_memory = allocated_memory - heap_memory + service_memory_config[key] = build_memory_config_string(int(heap_memory), int(direct_memory)) + + print(f'\nMemory distribution for services') + for key, value in service_memory_config.items(): + print(f'{key}, memory_config: {value}, instance_count: {service_instance_map[key]}') + print('\n') + + return service_memory_config + +def create_supervise_config_file(service_config, service_memory_config, base_config_path, run_zk): + + with open(SUPERVISE_CONFIG_FILE_PATH, 'w+') as the_file: + the_file.write(":verify bin/verify-java\n") + the_file.write(":verify bin/verify-default-ports\n") + the_file.write(":notify bin/greet\n") + the_file.write(":kill-timeout 10\n") + the_file.write("\n") + + if run_zk: + the_file.write("!p10 zk bin/run-zk conf\n") + + for item in service_config: + service = item[0] + prefix = '' + if service == MIDDLE_MANAGER_SERVICE_NAME: + prefix = '!p90 ' + if item[1] == "": + service_path = item[0] + else: + service_path = item[1] + jvm_args = service_memory_config.get(item[0]) + + if jvm_args == "": + the_file.write(f'{prefix}{service} bin/run-druid {service} {base_config_path} {service_path}\n') + else: + the_file.write(f'{prefix}{service} bin/run-druid {service} {base_config_path} {service_path} \'{jvm_args}\'\n') + +def print_service_config(service_config, base_config_path): + print('Services to start:') + for item in service_config: + if item[1] == "": + print(f'{item[0]}, using default config from {os.getcwd()}/../{QUICKSTART_BASE_CONFIG_PATH}') + else: + print(f'{item[0]}, using config from {base_config_path}/{item[1]}') + print('\n') + +def display_help(): + text = """ + Usage: start-druid [options] + + where options include: + totalMemory= + memory for druid cluster, if totalMemory is not specified + 80 percent of system memory is used. Note, if service + specific jvm config is present, + totalMemory shouldn't be specified + baseConfigPath= + relative path to base directory containing common and service specific + properties to be overridden, this directory must contain `_common` + directory with `common.jvm.config` & `common.runtime.properties` + if `baseConfigPath` is not specified, config from + conf/druid/single-server/quickstart direcotry is used + computeOnly + command dry-run, validates the arguments and + display the memory distribution for services + runZk + specification to run zookeeper, zk config is picked up from conf/zk + =[subdirectory] + service_identifier is the service to be started, multiple services + can be specified, `service_identifier` should be one of + [broker, router, middleManager, historical, coordinator-overlord] + `subdirectory` is optional directory within `baseConfigPath` + containing runtime properties or/and jvm properties + Note, if jvm.config file is present for one service, it must be + present for all other services + If no service is specified, all services and zookeeper are started + + sample usage: + start-druid + start up all the services using the default system memory + start-druid totalMemory=100g + start up all the services using the given memory + start-druid totalMemory=100g computeOnly + compute memory distribution for all the services + start-druid totalMemory=100g broker router historical + starts `broker`, `router` and `historical` services, using `100g` of memory + start-druid totalMemory=100g baseConfigPath=/conf/druid/single-server/large broker router historical + starts `broker`, `router` and `historical` service, using 100g of memory, + use common configs from specified `baseConfigPath` + start-druid totalMemory=100g baseConfigPath=/conf/druid/single-server/large broker=broker router=router historical=historical + starts `broker`, `router` and `historical` services, using 100g of memory, use common configs + from specified `baseConfigPath`, use service specific config from specified directories + if jvm.config is specified for all the services, memory distribution is not computed + start-druid totalMemory=100g baseConfigPath=/conf/druid/single-server/large broker=broker1 broker=broker2 + starts 2 instances of `broker` + config is read from respective directories, depending on whether jvm.config is specified, + memory distribution is computed + start-druid totalMemory=100g baseConfigPath=conf/druid/profile broker=broker1 historical=historical1 + if either of `broker1`, `historical1` subdirectory contains jvm.config, + exception is thrown since `totalMemory` argument is specified + start-druid baseConfigPath=conf/druid/profile broker=broker1 historical=historical1 + exception is thrown if either of `broker1`, `historical1` + subdirectory contains jvm.config but not both + If none of the subdirectory contains jvm.config, memory distribution is computed + """ + + print(text) + + +def main(): + for argument in sys.argv[1:]: + if check_argument_type(argument, HELP_ARG_IDENTIFIER): + display_help() + return + + print("Druid auto tuning quickstart\n") + + base_config_path, total_memory, service_config, run_zk, compute_only = parse_arguments() + + # change directory to bin + os.chdir(os.path.dirname(sys.argv[0])) + + print(f'Arguments passed, baseConfigPath: "{base_config_path}", totalMemory: "{total_memory}"\n') + print_service_config(service_config, base_config_path) + + service_memory_config = {} + if (should_compute_memory(base_config_path, total_memory, service_config)): + memory_in_mega_bytes = convert_total_memory_string(total_memory) + service_memory_config = distribute_memory_over_services(service_config, memory_in_mega_bytes) + + if compute_only: + return + + if base_config_path == "": + base_config_path = QUICKSTART_BASE_CONFIG_PATH + + create_supervise_config_file(service_config, service_memory_config, base_config_path, run_zk) + + os.system(f'exec ./supervise -c {SUPERVISE_CONFIG_FILE_PATH}') + +if __name__ == '__main__': + main() \ No newline at end of file diff --git a/examples/bin/run-druid b/examples/bin/run-druid index c302672209a0..c98c2e53bc3c 100755 --- a/examples/bin/run-druid +++ b/examples/bin/run-druid @@ -17,7 +17,7 @@ # specific language governing permissions and limitations # under the License. -if [ "$#" -gt 2 ] || [ "$#" -eq 0 ] +if [ "$#" -gt 4 ] || [ "$#" -eq 0 ] then >&2 echo "usage: $0 [conf-dir]" exit 1 @@ -47,7 +47,30 @@ if [ ! -d "$LOG_DIR" ]; then mkdir -p $LOG_DIR; fi echo "Running [$1], logging to [$LOG_DIR/$1.log] if no changes made to log4j2.xml" -cd "$WHEREAMI/.." -exec "$WHEREAMI"/run-java -Ddruid.node.type=$1 "-Ddruid.log.path=$LOG_DIR" `cat "$CONFDIR"/"$WHATAMI"/jvm.config | xargs` \ - -cp "$CONFDIR"/"$WHATAMI":"$CONFDIR"/_common:"$CONFDIR"/_common/hadoop-xml:"$CONFDIR"/../_common:"$CONFDIR"/../_common/hadoop-xml:"$WHEREAMI/../lib/*" \ - `cat "$CONFDIR"/$WHATAMI/main.config | xargs` +if [ "$WHATAMI" = 'coordinator-overlord' ] +then + SERVER_NAME=coordinator +else + SERVER_NAME="$WHATAMI" +fi + +if [ "$#" -eq 3 ] +then + cd "$WHEREAMI/.." + exec "$WHEREAMI"/run-java -Ddruid.node.type=$1 "-Ddruid.log.path=$LOG_DIR" `cat "$CONFDIR"/"$3"/jvm.config | xargs` \ + -cp "$CONFDIR"/"$WHATAMI":"$CONFDIR"/_common:"$CONFDIR"/_common/hadoop-xml:"$CONFDIR"/../_common:"$CONFDIR"/../_common/hadoop-xml:"$WHEREAMI/../lib/*" org.apache.druid.cli.Main server $SERVER_NAME +elif [ "$#" -eq 4 ] +then + JVMARGS=`cat "$CONFDIR/_common/common.jvm.config" | xargs` + JVMARGS+=' ' + JVMARGS+=$4 + + cd "$WHEREAMI/.." + exec "$WHEREAMI"/run-java -Ddruid.node.type=$1 "-Ddruid.log.path=$LOG_DIR" $JVMARGS \ + -cp "$CONFDIR"/"$WHATAMI":"$CONFDIR"/_common:"$CONFDIR"/_common/hadoop-xml:"$CONFDIR"/../_common:"$CONFDIR"/../_common/hadoop-xml:"$WHEREAMI/../lib/*" org.apache.druid.cli.Main server $SERVER_NAME +else + cd "$WHEREAMI/.." + exec "$WHEREAMI"/run-java -Ddruid.node.type=$1 "-Ddruid.log.path=$LOG_DIR" `cat "$CONFDIR"/"$WHATAMI"/jvm.config | xargs` \ + -cp "$CONFDIR"/"$WHATAMI":"$CONFDIR"/_common:"$CONFDIR"/_common/hadoop-xml:"$CONFDIR"/../_common:"$CONFDIR"/../_common/hadoop-xml:"$WHEREAMI/../lib/*" \ + `cat "$CONFDIR"/$WHATAMI/main.config | xargs` +fi \ No newline at end of file diff --git a/examples/bin/start-druid b/examples/bin/start-druid new file mode 100644 index 000000000000..13781fca8e9b --- /dev/null +++ b/examples/bin/start-druid @@ -0,0 +1,24 @@ +#!/bin/bash -eu + +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. + +PWD="$(pwd)" +WHEREAMI="$(dirname "$0")" +WHEREAMI="$(cd "$WHEREAMI" && pwd)" + +python3 "$WHEREAMI"/druid-automated-quickstart.py $@ \ No newline at end of file diff --git a/examples/conf/druid/single-server/quickstart/_common/common.jvm.config b/examples/conf/druid/single-server/quickstart/_common/common.jvm.config new file mode 100644 index 000000000000..4db1d94ec38e --- /dev/null +++ b/examples/conf/druid/single-server/quickstart/_common/common.jvm.config @@ -0,0 +1,7 @@ +-server +-XX:+ExitOnOutOfMemoryError +-XX:+UseG1GC +-Duser.timezone=UTC +-Dfile.encoding=UTF-8 +-Djava.io.tmpdir=var/tmp +-Djava.util.logging.manager=org.apache.logging.log4j.jul.LogManager \ No newline at end of file diff --git a/examples/conf/druid/single-server/quickstart/_common/common.runtime.properties b/examples/conf/druid/single-server/quickstart/_common/common.runtime.properties new file mode 100644 index 000000000000..b0adb0695cd7 --- /dev/null +++ b/examples/conf/druid/single-server/quickstart/_common/common.runtime.properties @@ -0,0 +1,158 @@ +# +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. +# + +# Extensions specified in the load list will be loaded by Druid +# We are using local fs for deep storage - not recommended for production - use S3, HDFS, or NFS instead +# We are using local derby for the metadata store - not recommended for production - use MySQL or Postgres instead + +# If you specify `druid.extensions.loadList=[]`, Druid won't load any extension from file system. +# If you don't specify `druid.extensions.loadList`, Druid will load all the extensions under root extension directory. +# More info: https://druid.apache.org/docs/latest/operations/including-extensions.html +druid.extensions.loadList=["druid-hdfs-storage", "druid-kafka-indexing-service", "druid-datasketches", "druid-multi-stage-query"] + +# If you have a different version of Hadoop, place your Hadoop client jar files in your hadoop-dependencies directory +# and uncomment the line below to point to your directory. +#druid.extensions.hadoopDependenciesDir=/my/dir/hadoop-dependencies + + +# +# Hostname +# +druid.host=localhost + +# +# Logging +# + +# Log all runtime properties on startup. Disable to avoid logging properties on startup: +druid.startup.logging.logProperties=true + +# +# Zookeeper +# + +druid.zk.service.host=localhost +druid.zk.paths.base=/druid + +# +# Metadata storage +# + +# For Derby server on your Druid Coordinator (only viable in a cluster with a single Coordinator, no fail-over): +druid.metadata.storage.type=derby +druid.metadata.storage.connector.connectURI=jdbc:derby://localhost:1527/var/druid/metadata.db;create=true +druid.metadata.storage.connector.host=localhost +druid.metadata.storage.connector.port=1527 + +# For MySQL (make sure to include the MySQL JDBC driver on the classpath): +#druid.metadata.storage.type=mysql +#druid.metadata.storage.connector.connectURI=jdbc:mysql://db.example.com:3306/druid +#druid.metadata.storage.connector.user=... +#druid.metadata.storage.connector.password=... + +# For PostgreSQL: +#druid.metadata.storage.type=postgresql +#druid.metadata.storage.connector.connectURI=jdbc:postgresql://db.example.com:5432/druid +#druid.metadata.storage.connector.user=... +#druid.metadata.storage.connector.password=... + +# +# Deep storage +# + +# For local disk (only viable in a cluster if this is a network mount): +druid.storage.type=local +druid.storage.storageDirectory=var/druid/segments + +# For HDFS: +#druid.storage.type=hdfs +#druid.storage.storageDirectory=/druid/segments + +# For S3: +#druid.storage.type=s3 +#druid.storage.bucket=your-bucket +#druid.storage.baseKey=druid/segments +#druid.s3.accessKey=... +#druid.s3.secretKey=... + +# +# Indexing service logs +# + +# For local disk (only viable in a cluster if this is a network mount): +druid.indexer.logs.type=file +druid.indexer.logs.directory=var/druid/indexing-logs + +# For HDFS: +#druid.indexer.logs.type=hdfs +#druid.indexer.logs.directory=/druid/indexing-logs + +# For S3: +#druid.indexer.logs.type=s3 +#druid.indexer.logs.s3Bucket=your-bucket +#druid.indexer.logs.s3Prefix=druid/indexing-logs + +# +# Service discovery +# + +druid.selectors.indexing.serviceName=druid/overlord +druid.selectors.coordinator.serviceName=druid/coordinator + +# +# Monitoring +# + +druid.monitoring.monitors=["org.apache.druid.java.util.metrics.JvmMonitor"] +druid.emitter=noop +druid.emitter.logging.logLevel=info + +# Storage type of double columns +# ommiting this will lead to index double as float at the storage layer + +druid.indexing.doubleStorage=double + +# +# Security +# +druid.server.hiddenProperties=["druid.s3.accessKey","druid.s3.secretKey","druid.metadata.storage.connector.password", "password", "key", "token", "pwd"] + + +# +# SQL +# +druid.sql.enable=true + +# Planning SQL query when there is aggregate distinct in the statement +druid.sql.planner.useGroupingSetForExactDistinct=true + +# +# Lookups +# +druid.lookup.enableLookupSyncOnStartup=false + +# +# Expression processing config +# +druid.expressions.useStrictBooleans=true + +# +# Http client +# +druid.global.http.eagerInitialization=false diff --git a/examples/conf/druid/single-server/quickstart/_common/log4j2.xml b/examples/conf/druid/single-server/quickstart/_common/log4j2.xml new file mode 100644 index 000000000000..66dc13da4c5e --- /dev/null +++ b/examples/conf/druid/single-server/quickstart/_common/log4j2.xml @@ -0,0 +1,86 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/examples/conf/druid/single-server/quickstart/broker/runtime.properties b/examples/conf/druid/single-server/quickstart/broker/runtime.properties new file mode 100644 index 000000000000..4d67149d29e9 --- /dev/null +++ b/examples/conf/druid/single-server/quickstart/broker/runtime.properties @@ -0,0 +1,21 @@ +# +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. +# + +druid.service=druid/broker +druid.plaintextPort=8082 \ No newline at end of file diff --git a/examples/conf/druid/single-server/quickstart/coordinator-overlord/runtime.properties b/examples/conf/druid/single-server/quickstart/coordinator-overlord/runtime.properties new file mode 100644 index 000000000000..855eeeafa7dd --- /dev/null +++ b/examples/conf/druid/single-server/quickstart/coordinator-overlord/runtime.properties @@ -0,0 +1,21 @@ +# +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. +# + +druid.service=druid/coordinator +druid.plaintextPort=8081 \ No newline at end of file diff --git a/examples/conf/druid/single-server/quickstart/historical/jvm.config b/examples/conf/druid/single-server/quickstart/historical/jvm.config new file mode 100644 index 000000000000..e69de29bb2d1 diff --git a/examples/conf/druid/single-server/quickstart/historical/runtime.properties b/examples/conf/druid/single-server/quickstart/historical/runtime.properties new file mode 100644 index 000000000000..4ce14669dbaf --- /dev/null +++ b/examples/conf/druid/single-server/quickstart/historical/runtime.properties @@ -0,0 +1,24 @@ +# +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. +# + +druid.service=druid/historical +druid.plaintextPort=8083 + +# Segment storage +druid.segmentCache.locations=[{"path":"var/druid/segment-cache","maxSize":"300g"}] \ No newline at end of file diff --git a/examples/conf/druid/single-server/quickstart/middleManager/runtime.properties b/examples/conf/druid/single-server/quickstart/middleManager/runtime.properties new file mode 100644 index 000000000000..fdec5cfca425 --- /dev/null +++ b/examples/conf/druid/single-server/quickstart/middleManager/runtime.properties @@ -0,0 +1,21 @@ +# +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. +# + +druid.service=druid/middleManager +druid.plaintextPort=8091 diff --git a/examples/conf/druid/single-server/quickstart/router/runtime.properties b/examples/conf/druid/single-server/quickstart/router/runtime.properties new file mode 100644 index 000000000000..3754ec172210 --- /dev/null +++ b/examples/conf/druid/single-server/quickstart/router/runtime.properties @@ -0,0 +1,21 @@ +# +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. +# + +druid.service=druid/router +druid.plaintextPort=8888 \ No newline at end of file From 342e5d72b1ce5d012622f501649b36320ecc5598 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Tue, 15 Nov 2022 11:35:00 +0530 Subject: [PATCH 002/141] remove conf/druid/single-server/quickstart/_common/historical/jvm.config --- .../conf/druid/single-server/quickstart/historical/jvm.config | 0 1 file changed, 0 insertions(+), 0 deletions(-) delete mode 100644 examples/conf/druid/single-server/quickstart/historical/jvm.config diff --git a/examples/conf/druid/single-server/quickstart/historical/jvm.config b/examples/conf/druid/single-server/quickstart/historical/jvm.config deleted file mode 100644 index e69de29bb2d1..000000000000 From 6a4a48edd6ceed77ee26c40c55625a76f200b131 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Tue, 15 Nov 2022 11:53:25 +0530 Subject: [PATCH 003/141] Minor changes in python script --- examples/bin/druid-automated-quickstart.py | 8 +++----- 1 file changed, 3 insertions(+), 5 deletions(-) diff --git a/examples/bin/druid-automated-quickstart.py b/examples/bin/druid-automated-quickstart.py index 24bee48fbb96..de6827f61f67 100644 --- a/examples/bin/druid-automated-quickstart.py +++ b/examples/bin/druid-automated-quickstart.py @@ -53,7 +53,6 @@ def get_argument_value(argument): return split_args[1] def parse_arguments(): - num_args = len(sys.argv) service_list = [] service_path_list = [] base_config_path = "" @@ -157,7 +156,7 @@ def distribute_memory_over_services(service_config, total_memory): direct_memory = allocated_memory - heap_memory service_memory_config[key] = build_memory_config_string(int(heap_memory), int(direct_memory)) - print(f'\nMemory distribution for services') + print(f'\nMemory distribution for services:') for key, value in service_memory_config.items(): print(f'{key}, memory_config: {value}, instance_count: {service_instance_map[key]}') print('\n') @@ -199,7 +198,6 @@ def print_service_config(service_config, base_config_path): print(f'{item[0]}, using default config from {os.getcwd()}/../{QUICKSTART_BASE_CONFIG_PATH}') else: print(f'{item[0]}, using config from {base_config_path}/{item[1]}') - print('\n') def display_help(): text = """ @@ -270,14 +268,14 @@ def main(): display_help() return - print("Druid auto tuning quickstart\n") + print("Druid automated quickstart\n") base_config_path, total_memory, service_config, run_zk, compute_only = parse_arguments() # change directory to bin os.chdir(os.path.dirname(sys.argv[0])) - print(f'Arguments passed, baseConfigPath: "{base_config_path}", totalMemory: "{total_memory}"\n') + print(f'Arguments passed: baseConfigPath: "{base_config_path}", totalMemory: "{total_memory}"\n') print_service_config(service_config, base_config_path) service_memory_config = {} From 36b86b6336b366fa0ac0135177e6e556c2621974 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Tue, 15 Nov 2022 14:38:12 +0530 Subject: [PATCH 004/141] Add lower bound memory for some services --- examples/bin/druid-automated-quickstart.py | 48 +++++++++++++++++++--- 1 file changed, 42 insertions(+), 6 deletions(-) diff --git a/examples/bin/druid-automated-quickstart.py b/examples/bin/druid-automated-quickstart.py index de6827f61f67..ae4bc6fddfc6 100644 --- a/examples/bin/druid-automated-quickstart.py +++ b/examples/bin/druid-automated-quickstart.py @@ -29,11 +29,16 @@ ] SERVICE_MEMORY_DISTRIBUTION_WEIGHT = { - MIDDLE_MANAGER_SERVICE_NAME: 0.25, + MIDDLE_MANAGER_SERVICE_NAME: 0.5, ROUTER_SERVICE_NAME: 1, COORDINATOR_SERVICE_NAME: 18, BROKER_SERVICE_NAME: 28, - HISTORICAL_SERVICE_NAME: 50 + HISTORICAL_SERVICE_NAME: 45 +} + +SERVICE_MEMORY_LOWER_BOUND = { + MIDDLE_MANAGER_SERVICE_NAME: 64, + ROUTER_SERVICE_NAME: 128 } SERVICE_MEMORY_HEAP_RATIO = { @@ -150,8 +155,34 @@ def distribute_memory_over_services(service_config, total_memory): multiplier = total_memory / memory_weight_sum + lower_bound_memory_allocation = 0 + allocated_services = set() + for key, value in service_instance_map.items(): + allocated_memory = SERVICE_MEMORY_DISTRIBUTION_WEIGHT.get(key) * multiplier + if key in SERVICE_MEMORY_LOWER_BOUND and allocated_memory < SERVICE_MEMORY_LOWER_BOUND.get(key): + allocated_memory = SERVICE_MEMORY_LOWER_BOUND.get(key) + heap_memory = SERVICE_MEMORY_HEAP_RATIO.get(key) * allocated_memory + direct_memory = allocated_memory - heap_memory + service_memory_config[key] = build_memory_config_string(int(heap_memory), int(direct_memory)) + lower_bound_memory_allocation += allocated_memory + allocated_services.add(key) + + if lower_bound_memory_allocation > 0: + # compute the multiplier again for remaing services + memory_weight_sum = 0 + for key, value in service_instance_map.items(): + if key in allocated_services: + continue + memory_weight_sum += SERVICE_MEMORY_DISTRIBUTION_WEIGHT.get(key) * value + multiplier = (total_memory - lower_bound_memory_allocation) / memory_weight_sum + for key, value in service_instance_map.items(): + if key in allocated_services: + continue allocated_memory = SERVICE_MEMORY_DISTRIBUTION_WEIGHT.get(key) * multiplier + if key in SERVICE_MEMORY_LOWER_BOUND and allocated_memory < SERVICE_MEMORY_LOWER_BOUND.get(key): + allocated_memory = SERVICE_MEMORY_LOWER_BOUND.get(key) + heap_memory = SERVICE_MEMORY_HEAP_RATIO.get(key) * allocated_memory direct_memory = allocated_memory - heap_memory service_memory_config[key] = build_memory_config_string(int(heap_memory), int(direct_memory)) @@ -191,13 +222,15 @@ def create_supervise_config_file(service_config, service_memory_config, base_con else: the_file.write(f'{prefix}{service} bin/run-druid {service} {base_config_path} {service_path} \'{jvm_args}\'\n') -def print_service_config(service_config, base_config_path): +def print_service_config(service_config, base_config_path, run_zk): print('Services to start:') for item in service_config: if item[1] == "": print(f'{item[0]}, using default config from {os.getcwd()}/../{QUICKSTART_BASE_CONFIG_PATH}') else: print(f'{item[0]}, using config from {base_config_path}/{item[1]}') + if run_zk: + print(f'zk, using default config from {os.getcwd()}/../conf/zk') def display_help(): text = """ @@ -206,9 +239,12 @@ def display_help(): where options include: totalMemory= memory for druid cluster, if totalMemory is not specified - 80 percent of system memory is used. Note, if service - specific jvm config is present, + 80 percent of system memory is used. + Note, if service specific jvm config is present, totalMemory shouldn't be specified + Integer value is supported with `m` or `g` suffix, + denoting memory in mb or gb + Memory should be greater than equals 2g baseConfigPath= relative path to base directory containing common and service specific properties to be overridden, this directory must contain `_common` @@ -276,7 +312,7 @@ def main(): os.chdir(os.path.dirname(sys.argv[0])) print(f'Arguments passed: baseConfigPath: "{base_config_path}", totalMemory: "{total_memory}"\n') - print_service_config(service_config, base_config_path) + print_service_config(service_config, base_config_path, run_zk) service_memory_config = {} if (should_compute_memory(base_config_path, total_memory, service_config)): From dbfb465610897af715307135f80cd7065526d3ae Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Tue, 15 Nov 2022 16:41:41 +0530 Subject: [PATCH 005/141] Additional runtime properties for services --- .../quickstart/broker/runtime.properties | 18 +++++++++++++++++- .../coordinator-overlord/runtime.properties | 14 +++++++++++++- .../quickstart/historical/runtime.properties | 17 ++++++++++++++++- .../middleManager/runtime.properties | 19 +++++++++++++++++++ .../quickstart/router/runtime.properties | 15 ++++++++++++++- 5 files changed, 79 insertions(+), 4 deletions(-) diff --git a/examples/conf/druid/single-server/quickstart/broker/runtime.properties b/examples/conf/druid/single-server/quickstart/broker/runtime.properties index 4d67149d29e9..ac42accd8576 100644 --- a/examples/conf/druid/single-server/quickstart/broker/runtime.properties +++ b/examples/conf/druid/single-server/quickstart/broker/runtime.properties @@ -18,4 +18,20 @@ # druid.service=druid/broker -druid.plaintextPort=8082 \ No newline at end of file +druid.plaintextPort=8082 + +# HTTP server settings +druid.server.http.numThreads=12 + +# HTTP client settings +druid.broker.http.numConnections=10 +druid.broker.http.maxQueuedBytes=5MiB + +# Processing threads and buffers +druid.processing.buffer.sizeBytes=100MiB +druid.processing.numMergeBuffers=2 +druid.processing.tmpDir=var/druid/processing + +# Query cache disabled -- push down caching and merging instead +druid.broker.cache.useCache=false +druid.broker.cache.populateCache=false \ No newline at end of file diff --git a/examples/conf/druid/single-server/quickstart/coordinator-overlord/runtime.properties b/examples/conf/druid/single-server/quickstart/coordinator-overlord/runtime.properties index 855eeeafa7dd..c808bb6d2e4b 100644 --- a/examples/conf/druid/single-server/quickstart/coordinator-overlord/runtime.properties +++ b/examples/conf/druid/single-server/quickstart/coordinator-overlord/runtime.properties @@ -18,4 +18,16 @@ # druid.service=druid/coordinator -druid.plaintextPort=8081 \ No newline at end of file +druid.plaintextPort=8081 + +druid.coordinator.startDelay=PT10S +druid.coordinator.period=PT5S +druid.manager.segments.pollDuration=PT5S + +# Run the overlord service in the coordinator process +druid.coordinator.asOverlord.enabled=true +druid.coordinator.asOverlord.overlordService=druid/overlord + +druid.indexer.queue.startDelay=PT5S + +druid.indexer.storage.type=metadata \ No newline at end of file diff --git a/examples/conf/druid/single-server/quickstart/historical/runtime.properties b/examples/conf/druid/single-server/quickstart/historical/runtime.properties index 4ce14669dbaf..76eaa5d4b0cd 100644 --- a/examples/conf/druid/single-server/quickstart/historical/runtime.properties +++ b/examples/conf/druid/single-server/quickstart/historical/runtime.properties @@ -20,5 +20,20 @@ druid.service=druid/historical druid.plaintextPort=8083 +# HTTP server threads +druid.server.http.numThreads=12 + +# Processing threads and buffers +druid.processing.buffer.sizeBytes=200MiB +druid.processing.numMergeBuffers=2 +druid.processing.numThreads=2 +druid.processing.tmpDir=var/druid/processing + # Segment storage -druid.segmentCache.locations=[{"path":"var/druid/segment-cache","maxSize":"300g"}] \ No newline at end of file +druid.segmentCache.locations=[{"path":"var/druid/segment-cache","maxSize":"300g"}] + +# Query cache +druid.historical.cache.useCache=true +druid.historical.cache.populateCache=true +druid.cache.type=caffeine +druid.cache.sizeInBytes=10MiB diff --git a/examples/conf/druid/single-server/quickstart/middleManager/runtime.properties b/examples/conf/druid/single-server/quickstart/middleManager/runtime.properties index fdec5cfca425..e7262930f498 100644 --- a/examples/conf/druid/single-server/quickstart/middleManager/runtime.properties +++ b/examples/conf/druid/single-server/quickstart/middleManager/runtime.properties @@ -19,3 +19,22 @@ druid.service=druid/middleManager druid.plaintextPort=8091 + +# Number of tasks per middleManager +druid.worker.capacity=2 + +# Task launch parameters +druid.indexer.runner.javaCommand=bin/run-java +druid.indexer.runner.javaOptsArray=["-server","-Xms1g","-Xmx1g","-XX:MaxDirectMemorySize=1g","-Duser.timezone=UTC","-Dfile.encoding=UTF-8","-XX:+ExitOnOutOfMemoryError","-Djava.util.logging.manager=org.apache.logging.log4j.jul.LogManager"] +druid.indexer.task.baseTaskDir=var/druid/task + +# HTTP server threads +druid.server.http.numThreads=12 + +# Processing threads and buffers on Peons +druid.indexer.fork.property.druid.processing.numMergeBuffers=2 +druid.indexer.fork.property.druid.processing.buffer.sizeBytes=100MiB +druid.indexer.fork.property.druid.processing.numThreads=1 + +# Hadoop indexing +druid.indexer.task.hadoopWorkingPath=var/druid/hadoop-tmp diff --git a/examples/conf/druid/single-server/quickstart/router/runtime.properties b/examples/conf/druid/single-server/quickstart/router/runtime.properties index 3754ec172210..e5448b38d0cc 100644 --- a/examples/conf/druid/single-server/quickstart/router/runtime.properties +++ b/examples/conf/druid/single-server/quickstart/router/runtime.properties @@ -18,4 +18,17 @@ # druid.service=druid/router -druid.plaintextPort=8888 \ No newline at end of file +druid.plaintextPort=8888 + +# HTTP proxy +druid.router.http.numConnections=50 +druid.router.http.readTimeout=PT5M +druid.router.http.numMaxThreads=100 +druid.server.http.numThreads=100 + +# Service discovery +druid.router.defaultBrokerServiceName=druid/broker +druid.router.coordinatorServiceName=druid/coordinator + +# Management proxy to coordinator / overlord: required for unified web console. +druid.router.managementProxy.enabled=true \ No newline at end of file From 7324252be7ab4fd5c13adcdb494d9b015cf73186 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Fri, 18 Nov 2022 13:26:17 +0530 Subject: [PATCH 006/141] Update supervise script to accept command arguments, corresponding changes in druid-quickstart.py --- ...ated-quickstart.py => druid-quickstart.py} | 114 ++++++++++-------- 1 file changed, 62 insertions(+), 52 deletions(-) rename examples/bin/{druid-automated-quickstart.py => druid-quickstart.py} (79%) diff --git a/examples/bin/druid-automated-quickstart.py b/examples/bin/druid-quickstart.py similarity index 79% rename from examples/bin/druid-automated-quickstart.py rename to examples/bin/druid-quickstart.py index ae4bc6fddfc6..8058315b3783 100644 --- a/examples/bin/druid-automated-quickstart.py +++ b/examples/bin/druid-quickstart.py @@ -3,7 +3,6 @@ import psutil from pathlib import Path -SUPERVISE_CONFIG_FILE_PATH = "../conf/supervise/single-server/quickstart.conf" QUICKSTART_BASE_CONFIG_PATH = "conf/druid/single-server/quickstart" HELP_ARG_IDENTIFIER = "help" COMPUTE_ONLY_ARG_IDENTIFIER = "computeOnly" @@ -29,11 +28,11 @@ ] SERVICE_MEMORY_DISTRIBUTION_WEIGHT = { - MIDDLE_MANAGER_SERVICE_NAME: 0.5, - ROUTER_SERVICE_NAME: 1, - COORDINATOR_SERVICE_NAME: 18, - BROKER_SERVICE_NAME: 28, - HISTORICAL_SERVICE_NAME: 45 + MIDDLE_MANAGER_SERVICE_NAME: 1, + ROUTER_SERVICE_NAME: 2, + COORDINATOR_SERVICE_NAME: 36, + BROKER_SERVICE_NAME: 56, + HISTORICAL_SERVICE_NAME: 90 } SERVICE_MEMORY_LOWER_BOUND = { @@ -41,7 +40,7 @@ ROUTER_SERVICE_NAME: 128 } -SERVICE_MEMORY_HEAP_RATIO = { +SERVICE_MEMORY_HEAP_PERCENTAGE = { MIDDLE_MANAGER_SERVICE_NAME: 1, ROUTER_SERVICE_NAME: 1, COORDINATOR_SERVICE_NAME: 1, @@ -81,15 +80,18 @@ def parse_arguments(): if service not in DEFAULT_SERVICES: raise Exception(f'{service} is not a valid service name, should be one of {DEFAULT_SERVICES}') - path = "" + subdirectory = "" if len(split_args) == 2: - path = split_args[1] - complete_path = os.path.join(base_config_path, path) - if os.path.exists(os.path.join(complete_path)) is False: - raise Exception(f'Path `{complete_path}` specified for service `{service}` doesn\'t exist ') + subdirectory = split_args[1] + + if subdirectory != "": + complete_path = os.path.join(base_config_path, subdirectory) + if os.path.exists(os.path.join(complete_path)) is False: + raise Exception(f'Path `{complete_path}` specified for service `{service}` doesn\'t exist') + service_list.append(service) - service_path_list.append(path) + service_path_list.append(subdirectory) if len(service_list) == 0: # start all services @@ -161,7 +163,7 @@ def distribute_memory_over_services(service_config, total_memory): allocated_memory = SERVICE_MEMORY_DISTRIBUTION_WEIGHT.get(key) * multiplier if key in SERVICE_MEMORY_LOWER_BOUND and allocated_memory < SERVICE_MEMORY_LOWER_BOUND.get(key): allocated_memory = SERVICE_MEMORY_LOWER_BOUND.get(key) - heap_memory = SERVICE_MEMORY_HEAP_RATIO.get(key) * allocated_memory + heap_memory = SERVICE_MEMORY_HEAP_PERCENTAGE.get(key) * allocated_memory direct_memory = allocated_memory - heap_memory service_memory_config[key] = build_memory_config_string(int(heap_memory), int(direct_memory)) lower_bound_memory_allocation += allocated_memory @@ -183,7 +185,7 @@ def distribute_memory_over_services(service_config, total_memory): if key in SERVICE_MEMORY_LOWER_BOUND and allocated_memory < SERVICE_MEMORY_LOWER_BOUND.get(key): allocated_memory = SERVICE_MEMORY_LOWER_BOUND.get(key) - heap_memory = SERVICE_MEMORY_HEAP_RATIO.get(key) * allocated_memory + heap_memory = SERVICE_MEMORY_HEAP_PERCENTAGE.get(key) * allocated_memory direct_memory = allocated_memory - heap_memory service_memory_config[key] = build_memory_config_string(int(heap_memory), int(direct_memory)) @@ -194,33 +196,40 @@ def distribute_memory_over_services(service_config, total_memory): return service_memory_config -def create_supervise_config_file(service_config, service_memory_config, base_config_path, run_zk): - - with open(SUPERVISE_CONFIG_FILE_PATH, 'w+') as the_file: - the_file.write(":verify bin/verify-java\n") - the_file.write(":verify bin/verify-default-ports\n") - the_file.write(":notify bin/greet\n") - the_file.write(":kill-timeout 10\n") - the_file.write("\n") - - if run_zk: - the_file.write("!p10 zk bin/run-zk conf\n") - - for item in service_config: - service = item[0] - prefix = '' - if service == MIDDLE_MANAGER_SERVICE_NAME: - prefix = '!p90 ' - if item[1] == "": - service_path = item[0] - else: - service_path = item[1] - jvm_args = service_memory_config.get(item[0]) - - if jvm_args == "": - the_file.write(f'{prefix}{service} bin/run-druid {service} {base_config_path} {service_path}\n') - else: - the_file.write(f'{prefix}{service} bin/run-druid {service} {base_config_path} {service_path} \'{jvm_args}\'\n') +def build_supervise_script_arguments(service_config, service_memory_config, base_config_path, run_zk): + argument_list = [] + + argument_list.append("\":verify bin/verify-java\"") + argument_list.append("\":verify bin/verify-default-ports\"") + argument_list.append("\":notify bin/greet\"") + argument_list.append("\":kill-timeout 10\"") + + if run_zk: + argument_list.append("\"!p10 zk bin/run-zk conf\"") + + for item in service_config: + service = item[0] + prefix = '' + if service == MIDDLE_MANAGER_SERVICE_NAME: + prefix = '!p90 ' + if item[1] == "": + service_path = item[0] + else: + service_path = item[1] + jvm_args = service_memory_config.get(item[0]) + + if jvm_args is None: + argument_list.append(f'\"{prefix}{service} bin/run-druid {service} {base_config_path} {service_path}\"') + else: + argument_list.append(f'\"{prefix}{service} bin/run-druid {service} {base_config_path} {service_path} \'{jvm_args}\'\"') + + print('Commands for supervise script:') + for item in argument_list: + print(item) + + print('\n') + + return ",".join(argument_list) def print_service_config(service_config, base_config_path, run_zk): print('Services to start:') @@ -231,6 +240,7 @@ def print_service_config(service_config, base_config_path, run_zk): print(f'{item[0]}, using config from {base_config_path}/{item[1]}') if run_zk: print(f'zk, using default config from {os.getcwd()}/../conf/zk') + print('\n') def display_help(): text = """ @@ -246,11 +256,11 @@ def display_help(): denoting memory in mb or gb Memory should be greater than equals 2g baseConfigPath= - relative path to base directory containing common and service specific + relative path to base directory, containing common and service specific properties to be overridden, this directory must contain `_common` directory with `common.jvm.config` & `common.runtime.properties` if `baseConfigPath` is not specified, config from - conf/druid/single-server/quickstart direcotry is used + conf/druid/single-server/quickstart directory is used computeOnly command dry-run, validates the arguments and display the memory distribution for services @@ -275,21 +285,21 @@ def display_help(): compute memory distribution for all the services start-druid totalMemory=100g broker router historical starts `broker`, `router` and `historical` services, using `100g` of memory - start-druid totalMemory=100g baseConfigPath=/conf/druid/single-server/large broker router historical + start-druid totalMemory=100g baseConfigPath=../conf/druid/single-server/large broker router historical starts `broker`, `router` and `historical` service, using 100g of memory, use common configs from specified `baseConfigPath` - start-druid totalMemory=100g baseConfigPath=/conf/druid/single-server/large broker=broker router=router historical=historical + start-druid totalMemory=100g baseConfigPath=../conf/druid/single-server/large broker=broker router=router historical=historical starts `broker`, `router` and `historical` services, using 100g of memory, use common configs from specified `baseConfigPath`, use service specific config from specified directories if jvm.config is specified for all the services, memory distribution is not computed - start-druid totalMemory=100g baseConfigPath=/conf/druid/single-server/large broker=broker1 broker=broker2 + start-druid totalMemory=100g baseConfigPath=../conf/druid/single-server/large broker=broker1 broker=broker2 starts 2 instances of `broker` config is read from respective directories, depending on whether jvm.config is specified, memory distribution is computed - start-druid totalMemory=100g baseConfigPath=conf/druid/profile broker=broker1 historical=historical1 + start-druid totalMemory=100g baseConfigPath=../conf/druid/profile broker=broker1 historical=historical1 if either of `broker1`, `historical1` subdirectory contains jvm.config, exception is thrown since `totalMemory` argument is specified - start-druid baseConfigPath=conf/druid/profile broker=broker1 historical=historical1 + start-druid baseConfigPath=../conf/druid/profile broker=broker1 historical=historical1 exception is thrown if either of `broker1`, `historical1` subdirectory contains jvm.config but not both If none of the subdirectory contains jvm.config, memory distribution is computed @@ -304,7 +314,7 @@ def main(): display_help() return - print("Druid automated quickstart\n") + print("Druid quickstart\n") base_config_path, total_memory, service_config, run_zk, compute_only = parse_arguments() @@ -325,9 +335,9 @@ def main(): if base_config_path == "": base_config_path = QUICKSTART_BASE_CONFIG_PATH - create_supervise_config_file(service_config, service_memory_config, base_config_path, run_zk) + script_arguments = build_supervise_script_arguments(service_config, service_memory_config, base_config_path, run_zk) - os.system(f'exec ./supervise -c {SUPERVISE_CONFIG_FILE_PATH}') + os.system(f'exec ./supervise -a {script_arguments}') if __name__ == '__main__': main() \ No newline at end of file From 72494e939341110b3db5558d663402503bcffd68 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Fri, 18 Nov 2022 13:29:42 +0530 Subject: [PATCH 007/141] File end newline --- examples/bin/druid-quickstart.py | 2 +- examples/bin/run-druid | 2 +- examples/bin/start-druid | 2 +- examples/bin/supervise | 99 ++++++++++++------- .../quickstart/_common/common.jvm.config | 2 +- .../quickstart/broker/runtime.properties | 2 +- .../coordinator-overlord/runtime.properties | 2 +- .../quickstart/router/runtime.properties | 2 +- 8 files changed, 70 insertions(+), 43 deletions(-) diff --git a/examples/bin/druid-quickstart.py b/examples/bin/druid-quickstart.py index 8058315b3783..136dcf9ccd15 100644 --- a/examples/bin/druid-quickstart.py +++ b/examples/bin/druid-quickstart.py @@ -340,4 +340,4 @@ def main(): os.system(f'exec ./supervise -a {script_arguments}') if __name__ == '__main__': - main() \ No newline at end of file + main() diff --git a/examples/bin/run-druid b/examples/bin/run-druid index c98c2e53bc3c..005c116e6100 100755 --- a/examples/bin/run-druid +++ b/examples/bin/run-druid @@ -73,4 +73,4 @@ else exec "$WHEREAMI"/run-java -Ddruid.node.type=$1 "-Ddruid.log.path=$LOG_DIR" `cat "$CONFDIR"/"$WHATAMI"/jvm.config | xargs` \ -cp "$CONFDIR"/"$WHATAMI":"$CONFDIR"/_common:"$CONFDIR"/_common/hadoop-xml:"$CONFDIR"/../_common:"$CONFDIR"/../_common/hadoop-xml:"$WHEREAMI/../lib/*" \ `cat "$CONFDIR"/$WHATAMI/main.config | xargs` -fi \ No newline at end of file +fi diff --git a/examples/bin/start-druid b/examples/bin/start-druid index 13781fca8e9b..41639d8cdce6 100644 --- a/examples/bin/start-druid +++ b/examples/bin/start-druid @@ -21,4 +21,4 @@ PWD="$(pwd)" WHEREAMI="$(dirname "$0")" WHEREAMI="$(cd "$WHEREAMI" && pwd)" -python3 "$WHEREAMI"/druid-automated-quickstart.py $@ \ No newline at end of file +python3 "$WHEREAMI"/druid-quickstart.py $@ diff --git a/examples/bin/supervise b/examples/bin/supervise index 81b7e57b049e..507795733c34 100755 --- a/examples/bin/supervise +++ b/examples/bin/supervise @@ -46,6 +46,52 @@ sub usage die "usage: $0 -c [-d ] [-t ] [--svlogd ]\n"; } +sub process_config +{ + my @lines = @_; + my @commands; + my @verify; + my @notify; + my $kill_timeout; + for my $line (@lines) + { + if ($line =~ /^(:verify|:notify|:kill-timeout|(?:\!p[0-9]+\s+)?[^:]\S+)\s+(.+)$/) { + + my $name = $1; + my $order = 50; + my $command = $2; + + if ($name =~ /^(?:\!p([0-9]+)\s+)(.*)$/) { + $order = $1; + $name = $2; + } + + if ($name eq ':verify') { + push @verify, $command; + } elsif ($name eq ':notify') { + push @notify, $command; + } elsif ($name eq ':kill-timeout') { + $kill_timeout = int($command); + } else { + die "Duplicate command: $line\n" if grep { $_->{name} eq $name } @commands; + push @commands, { + name => $name, + command => $command, + order => $order, # Stop order for this command + pid => 0, # Current pid, or 0 if not running + down => 0, # Time the proc should be down until + killed => 0, # Signal we sent to this process + restarting => 0, # True if this command is currently restarting + }; + } + } else { + die "Syntax error: $line\n"; + } + } + + return { commands => \@commands, verify => \@verify, notify => \@notify, 'kill-timeout' => $kill_timeout }; +} + sub read_config_file { my ($config_file) = @_; @@ -53,49 +99,20 @@ sub read_config_file open my $config_fh, "<", $config_file or die "open $config_file: $!"; - my @commands; - my @verify; - my @notify; - my $kill_timeout; + my @lines; while (my $line = <$config_fh>) { chomp $line; next if $line =~ /^(\s*\#.*|\s*)$/; if ($line =~ /^(:verify|:notify|:kill-timeout|(?:\!p[0-9]+\s+)?[^:]\S+)\s+(.+)$/) { - my $name = $1; - my $order = 50; - my $command = $2; - - if ($name =~ /^(?:\!p([0-9]+)\s+)(.*)$/) { - $order = $1; - $name = $2; - } - - if ($name eq ':verify') { - push @verify, $command; - } elsif ($name eq ':notify') { - push @notify, $command; - } elsif ($name eq ':kill-timeout') { - $kill_timeout = int($command); - } else { - die "Duplicate command: $line\n" if grep { $_->{name} eq $name } @commands; - push @commands, { - name => $name, - command => $command, - order => $order, # Stop order for this command - pid => 0, # Current pid, or 0 if not running - down => 0, # Time the proc should be down until - killed => 0, # Signal we sent to this process - restarting => 0, # True if this command is currently restarting - }; - } + push @lines, $line } else { die "Syntax error: $line\n"; } } close $config_fh; - return { commands => \@commands, verify => \@verify, notify => \@notify, 'kill-timeout' => $kill_timeout }; + return @lines; } sub stringify_exit_status @@ -179,13 +196,23 @@ usage() unless GetOptions( 'vardir|d=s', 'kill-timeout|t=i', 'chdir=s', - 'svlogd:s' + 'svlogd:s', + 'array|a=s{,}' ); -usage() unless $opt{'conf'} && $opt{'vardir'}; +usage() unless (($opt{'array'} || $opt{'conf'}) && $opt{'vardir'}); + +my @config_lines; + +# get commands to execute either from reading the config file or command line +if (not defined $opt{'conf'}) { + @config_lines = split(',', $opt{'array'}); +} else { + @config_lines = read_config_file($opt{'conf'}); +} + +my $config = process_config(@config_lines); -# Read config file -my $config = read_config_file($opt{'conf'}); @commands = @{$config->{commands}}; if (!@commands) { diff --git a/examples/conf/druid/single-server/quickstart/_common/common.jvm.config b/examples/conf/druid/single-server/quickstart/_common/common.jvm.config index 4db1d94ec38e..fd74cf358979 100644 --- a/examples/conf/druid/single-server/quickstart/_common/common.jvm.config +++ b/examples/conf/druid/single-server/quickstart/_common/common.jvm.config @@ -4,4 +4,4 @@ -Duser.timezone=UTC -Dfile.encoding=UTF-8 -Djava.io.tmpdir=var/tmp --Djava.util.logging.manager=org.apache.logging.log4j.jul.LogManager \ No newline at end of file +-Djava.util.logging.manager=org.apache.logging.log4j.jul.LogManager diff --git a/examples/conf/druid/single-server/quickstart/broker/runtime.properties b/examples/conf/druid/single-server/quickstart/broker/runtime.properties index ac42accd8576..92b85f8bc754 100644 --- a/examples/conf/druid/single-server/quickstart/broker/runtime.properties +++ b/examples/conf/druid/single-server/quickstart/broker/runtime.properties @@ -34,4 +34,4 @@ druid.processing.tmpDir=var/druid/processing # Query cache disabled -- push down caching and merging instead druid.broker.cache.useCache=false -druid.broker.cache.populateCache=false \ No newline at end of file +druid.broker.cache.populateCache=false diff --git a/examples/conf/druid/single-server/quickstart/coordinator-overlord/runtime.properties b/examples/conf/druid/single-server/quickstart/coordinator-overlord/runtime.properties index c808bb6d2e4b..c053823f8ed8 100644 --- a/examples/conf/druid/single-server/quickstart/coordinator-overlord/runtime.properties +++ b/examples/conf/druid/single-server/quickstart/coordinator-overlord/runtime.properties @@ -30,4 +30,4 @@ druid.coordinator.asOverlord.overlordService=druid/overlord druid.indexer.queue.startDelay=PT5S -druid.indexer.storage.type=metadata \ No newline at end of file +druid.indexer.storage.type=metadata diff --git a/examples/conf/druid/single-server/quickstart/router/runtime.properties b/examples/conf/druid/single-server/quickstart/router/runtime.properties index e5448b38d0cc..497d3b4d3aa1 100644 --- a/examples/conf/druid/single-server/quickstart/router/runtime.properties +++ b/examples/conf/druid/single-server/quickstart/router/runtime.properties @@ -31,4 +31,4 @@ druid.router.defaultBrokerServiceName=druid/broker druid.router.coordinatorServiceName=druid/coordinator # Management proxy to coordinator / overlord: required for unified web console. -druid.router.managementProxy.enabled=true \ No newline at end of file +druid.router.managementProxy.enabled=true From e3505879770b227a087828e28a67654c3699037e Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Fri, 18 Nov 2022 18:49:11 +0530 Subject: [PATCH 008/141] Limit the ability to start multiple instances of a service, documentation changes --- examples/bin/druid-quickstart.py | 150 +++++++++--------- examples/bin/run-druid | 2 + .../medium/broker/runtime.properties | 10 -- .../medium/historical/runtime.properties | 8 - .../medium/router/runtime.properties | 6 - 5 files changed, 76 insertions(+), 100 deletions(-) diff --git a/examples/bin/druid-quickstart.py b/examples/bin/druid-quickstart.py index 136dcf9ccd15..30ca0f60aad0 100644 --- a/examples/bin/druid-quickstart.py +++ b/examples/bin/druid-quickstart.py @@ -7,7 +7,7 @@ HELP_ARG_IDENTIFIER = "help" COMPUTE_ONLY_ARG_IDENTIFIER = "computeOnly" RUN_ZK_IDENTIFIER = "runZk" -BASE_CONFIG_PATH_IDENTIFIER = "baseConfigPath" +ROOT_CONFIG_PATH_IDENTIFIER = "rootConfigPath" MEMORY_ARG_IDENTIFIER = "totalMemory" MEMORY_GIGABYTES_IDENTIFIER = "g" MEMORY_MEGABYTES_IDENTIFIER = "m" @@ -59,7 +59,7 @@ def get_argument_value(argument): def parse_arguments(): service_list = [] service_path_list = [] - base_config_path = "" + root_config_path = "" total_memory = "" compute_only = False run_zk = False @@ -69,8 +69,8 @@ def parse_arguments(): compute_only = True elif check_argument_type(argument, RUN_ZK_IDENTIFIER): run_zk = True - elif check_argument_type(argument, BASE_CONFIG_PATH_IDENTIFIER): - base_config_path = os.path.join(os.getcwd(), get_argument_value(argument)) + elif check_argument_type(argument, ROOT_CONFIG_PATH_IDENTIFIER): + root_config_path = os.path.join(os.getcwd(), get_argument_value(argument)) elif (check_argument_type(argument, MEMORY_ARG_IDENTIFIER)): total_memory = get_argument_value(argument) else: @@ -80,13 +80,16 @@ def parse_arguments(): if service not in DEFAULT_SERVICES: raise Exception(f'{service} is not a valid service name, should be one of {DEFAULT_SERVICES}') + if service in service_list: + raise Exception(f'{service} is specified multiple times') + subdirectory = "" if len(split_args) == 2: subdirectory = split_args[1] if subdirectory != "": - complete_path = os.path.join(base_config_path, subdirectory) + complete_path = os.path.join(root_config_path, subdirectory) if os.path.exists(os.path.join(complete_path)) is False: raise Exception(f'Path `{complete_path}` specified for service `{service}` doesn\'t exist') @@ -99,9 +102,9 @@ def parse_arguments(): service_path_list = [""] * len(DEFAULT_SERVICES) run_zk = True - return base_config_path, total_memory, list(zip(service_list, service_path_list)), run_zk, compute_only + return root_config_path, total_memory, list(zip(service_list, service_path_list)), run_zk, compute_only -def should_compute_memory(base_config_path, total_memory, service_config): +def should_compute_memory(root_config_path, total_memory, service_config): # if jvm file is present for any of the services # it should be present for all services and totalMemory should not be specified # if totalMemory is given, jvm file shouldn't be present for any service @@ -109,7 +112,7 @@ def should_compute_memory(base_config_path, total_memory, service_config): jvm_config_count = 0 for item in service_config: if item[1] != "": - if Path(f'{base_config_path}/{item[1]}/jvm.config').is_file(): + if Path(f'{root_config_path}/{item[1]}/jvm.config').is_file(): jvm_config_count += 1 elif jvm_config_count > 0: raise Exception('jvm.config file is missing for service {item[0]}, jvm.config should be specified for all the services or none') @@ -130,7 +133,7 @@ def compute_system_memory(): def convert_total_memory_string(memory): if memory == "": computed_memory = compute_system_memory() - print(f'`{MEMORY_ARG_IDENTIFIER}` argument is not specified Druid will use 80% of system memory: {computed_memory}m') + print(f'`{MEMORY_ARG_IDENTIFIER}` argument is not specified, Druid will use 80% of system memory: {computed_memory}m') return computed_memory elif memory.endswith(MEMORY_MEGABYTES_IDENTIFIER): return int(memory[:-1]) @@ -146,57 +149,57 @@ def build_memory_config_string(heap_memory, direct_memory): def distribute_memory_over_services(service_config, total_memory): service_memory_config = {} - service_instance_map = {} - - for item in service_config: - service_instance_map[item[0]] = service_instance_map.get(item[0], 0) + 1 memory_weight_sum = 0 - for key, value in service_instance_map.items(): - memory_weight_sum += SERVICE_MEMORY_DISTRIBUTION_WEIGHT.get(key) * value + for item in service_config: + memory_weight_sum += SERVICE_MEMORY_DISTRIBUTION_WEIGHT.get(item[0]) multiplier = total_memory / memory_weight_sum lower_bound_memory_allocation = 0 allocated_services = set() - for key, value in service_instance_map.items(): - allocated_memory = SERVICE_MEMORY_DISTRIBUTION_WEIGHT.get(key) * multiplier - if key in SERVICE_MEMORY_LOWER_BOUND and allocated_memory < SERVICE_MEMORY_LOWER_BOUND.get(key): - allocated_memory = SERVICE_MEMORY_LOWER_BOUND.get(key) - heap_memory = SERVICE_MEMORY_HEAP_PERCENTAGE.get(key) * allocated_memory + + for item in service_config: + service = item[0] + allocated_memory = SERVICE_MEMORY_DISTRIBUTION_WEIGHT.get(service) * multiplier + if service in SERVICE_MEMORY_LOWER_BOUND and allocated_memory < SERVICE_MEMORY_LOWER_BOUND.get(service): + allocated_memory = SERVICE_MEMORY_LOWER_BOUND.get(service) + heap_memory = SERVICE_MEMORY_HEAP_PERCENTAGE.get(service) * allocated_memory direct_memory = allocated_memory - heap_memory - service_memory_config[key] = build_memory_config_string(int(heap_memory), int(direct_memory)) + service_memory_config[service] = build_memory_config_string(int(heap_memory), int(direct_memory)) lower_bound_memory_allocation += allocated_memory - allocated_services.add(key) + allocated_services.add(service) if lower_bound_memory_allocation > 0: # compute the multiplier again for remaing services memory_weight_sum = 0 - for key, value in service_instance_map.items(): - if key in allocated_services: + for item in service_config: + service = item[0] + if service in allocated_services: continue - memory_weight_sum += SERVICE_MEMORY_DISTRIBUTION_WEIGHT.get(key) * value + memory_weight_sum += SERVICE_MEMORY_DISTRIBUTION_WEIGHT.get(service) * value multiplier = (total_memory - lower_bound_memory_allocation) / memory_weight_sum - for key, value in service_instance_map.items(): - if key in allocated_services: + for item in service_config: + service = item[0] + if service in allocated_services: continue - allocated_memory = SERVICE_MEMORY_DISTRIBUTION_WEIGHT.get(key) * multiplier - if key in SERVICE_MEMORY_LOWER_BOUND and allocated_memory < SERVICE_MEMORY_LOWER_BOUND.get(key): - allocated_memory = SERVICE_MEMORY_LOWER_BOUND.get(key) + allocated_memory = SERVICE_MEMORY_DISTRIBUTION_WEIGHT.get(service) * multiplier + if service in SERVICE_MEMORY_LOWER_BOUND and allocated_memory < SERVICE_MEMORY_LOWER_BOUND.get(service): + allocated_memory = SERVICE_MEMORY_LOWER_BOUND.get(service) - heap_memory = SERVICE_MEMORY_HEAP_PERCENTAGE.get(key) * allocated_memory + heap_memory = SERVICE_MEMORY_HEAP_PERCENTAGE.get(service) * allocated_memory direct_memory = allocated_memory - heap_memory - service_memory_config[key] = build_memory_config_string(int(heap_memory), int(direct_memory)) + service_memory_config[service] = build_memory_config_string(int(heap_memory), int(direct_memory)) print(f'\nMemory distribution for services:') for key, value in service_memory_config.items(): - print(f'{key}, memory_config: {value}, instance_count: {service_instance_map[key]}') + print(f'{key}, memory_config: {value}') print('\n') return service_memory_config -def build_supervise_script_arguments(service_config, service_memory_config, base_config_path, run_zk): +def build_supervise_script_arguments(service_config, service_memory_config, root_config_path, run_zk): argument_list = [] argument_list.append("\":verify bin/verify-java\"") @@ -219,11 +222,11 @@ def build_supervise_script_arguments(service_config, service_memory_config, base jvm_args = service_memory_config.get(item[0]) if jvm_args is None: - argument_list.append(f'\"{prefix}{service} bin/run-druid {service} {base_config_path} {service_path}\"') + argument_list.append(f'\"{prefix}{service} bin/run-druid {service} {root_config_path} {service_path}\"') else: - argument_list.append(f'\"{prefix}{service} bin/run-druid {service} {base_config_path} {service_path} \'{jvm_args}\'\"') + argument_list.append(f'\"{prefix}{service} bin/run-druid {service} {root_config_path} {service_path} \'{jvm_args}\'\"') - print('Commands for supervise script:') + print('Command for supervise script:') for item in argument_list: print(item) @@ -231,13 +234,13 @@ def build_supervise_script_arguments(service_config, service_memory_config, base return ",".join(argument_list) -def print_service_config(service_config, base_config_path, run_zk): +def print_service_config(service_config, root_config_path, run_zk): print('Services to start:') for item in service_config: if item[1] == "": print(f'{item[0]}, using default config from {os.getcwd()}/../{QUICKSTART_BASE_CONFIG_PATH}') else: - print(f'{item[0]}, using config from {base_config_path}/{item[1]}') + print(f'{item[0]}, using config from {root_config_path}/{item[1]}') if run_zk: print(f'zk, using default config from {os.getcwd()}/../conf/zk') print('\n') @@ -251,58 +254,53 @@ def display_help(): memory for druid cluster, if totalMemory is not specified 80 percent of system memory is used. Note, if service specific jvm config is present, - totalMemory shouldn't be specified - Integer value is supported with `m` or `g` suffix, - denoting memory in mb or gb + totalMemory shouldn't be specified. + Integer value is supported with `m` or `g` suffix. Memory should be greater than equals 2g - baseConfigPath= - relative path to base directory, containing common and service specific + rootConfigPath= + directory containing common and service specific properties to be overridden, this directory must contain `_common` directory with `common.jvm.config` & `common.runtime.properties` - if `baseConfigPath` is not specified, config from + if `rootConfigPath` is not specified, config from conf/druid/single-server/quickstart directory is used computeOnly command dry-run, validates the arguments and display the memory distribution for services runZk specification to run zookeeper, zk config is picked up from conf/zk - =[subdirectory] + [=subdirectory] service_identifier is the service to be started, multiple services can be specified, `service_identifier` should be one of [broker, router, middleManager, historical, coordinator-overlord] - `subdirectory` is optional directory within `baseConfigPath` + `subdirectory` is optional directory within `rootConfigPath` containing runtime properties or/and jvm properties - Note, if jvm.config file is present for one service, it must be - present for all other services - If no service is specified, all services and zookeeper are started + Note, if jvm.config file is present for one of the service, + it must be present for all other services. + If service is not explicitly specified, all services + alongwith zookeeper is started + Note each service should be specified at most once sample usage: start-druid - start up all the services using the default system memory + start up all the services (including zk) using the default system memory start-druid totalMemory=100g start up all the services using the given memory start-druid totalMemory=100g computeOnly compute memory distribution for all the services start-druid totalMemory=100g broker router historical starts `broker`, `router` and `historical` services, using `100g` of memory - start-druid totalMemory=100g baseConfigPath=../conf/druid/single-server/large broker router historical - starts `broker`, `router` and `historical` service, using 100g of memory, - use common configs from specified `baseConfigPath` - start-druid totalMemory=100g baseConfigPath=../conf/druid/single-server/large broker=broker router=router historical=historical + start-druid totalMemory=100g rootConfigPath=../conf/druid/single-server/large broker router historical + start `broker`, `router` and `historical` service, using 100g of memory, + use common configs from specified `rootConfigPath` + start-druid totalMemory=100g rootConfigPath=../conf/druid/single-server/large broker=broker router=router historical=historical starts `broker`, `router` and `historical` services, using 100g of memory, use common configs - from specified `baseConfigPath`, use service specific config from specified directories - if jvm.config is specified for all the services, memory distribution is not computed - start-druid totalMemory=100g baseConfigPath=../conf/druid/single-server/large broker=broker1 broker=broker2 - starts 2 instances of `broker` - config is read from respective directories, depending on whether jvm.config is specified, + from specified `rootConfigPath`, use service specific config from specified sub directories + Since totalMemory is specific, the service specific folder shouldn't contain jvm.config + start-druid rootConfigPath=../conf/druid/profile broker=broker historical=historical + exception is thrown if either of `broker`, `historical` + subdirectory contains jvm.config but not both. + If none of the service specific subdirectory contains jvm.config, memory distribution is computed - start-druid totalMemory=100g baseConfigPath=../conf/druid/profile broker=broker1 historical=historical1 - if either of `broker1`, `historical1` subdirectory contains jvm.config, - exception is thrown since `totalMemory` argument is specified - start-druid baseConfigPath=../conf/druid/profile broker=broker1 historical=historical1 - exception is thrown if either of `broker1`, `historical1` - subdirectory contains jvm.config but not both - If none of the subdirectory contains jvm.config, memory distribution is computed """ print(text) @@ -316,26 +314,26 @@ def main(): print("Druid quickstart\n") - base_config_path, total_memory, service_config, run_zk, compute_only = parse_arguments() + root_config_path, total_memory, service_config, run_zk, compute_only = parse_arguments() # change directory to bin os.chdir(os.path.dirname(sys.argv[0])) - print(f'Arguments passed: baseConfigPath: "{base_config_path}", totalMemory: "{total_memory}"\n') - print_service_config(service_config, base_config_path, run_zk) + print(f'Arguments passed: rootConfigPath: "{root_config_path}", totalMemory: "{total_memory}"\n') + print_service_config(service_config, root_config_path, run_zk) service_memory_config = {} - if (should_compute_memory(base_config_path, total_memory, service_config)): + if (should_compute_memory(root_config_path, total_memory, service_config)): memory_in_mega_bytes = convert_total_memory_string(total_memory) service_memory_config = distribute_memory_over_services(service_config, memory_in_mega_bytes) - if compute_only: - return + if root_config_path == "": + root_config_path = QUICKSTART_BASE_CONFIG_PATH - if base_config_path == "": - base_config_path = QUICKSTART_BASE_CONFIG_PATH + script_arguments = build_supervise_script_arguments(service_config, service_memory_config, root_config_path, run_zk) - script_arguments = build_supervise_script_arguments(service_config, service_memory_config, base_config_path, run_zk) + if compute_only: + return os.system(f'exec ./supervise -a {script_arguments}') diff --git a/examples/bin/run-druid b/examples/bin/run-druid index 005c116e6100..2726477684df 100755 --- a/examples/bin/run-druid +++ b/examples/bin/run-druid @@ -56,11 +56,13 @@ fi if [ "$#" -eq 3 ] then + # arguments for this case is cd "$WHEREAMI/.." exec "$WHEREAMI"/run-java -Ddruid.node.type=$1 "-Ddruid.log.path=$LOG_DIR" `cat "$CONFDIR"/"$3"/jvm.config | xargs` \ -cp "$CONFDIR"/"$WHATAMI":"$CONFDIR"/_common:"$CONFDIR"/_common/hadoop-xml:"$CONFDIR"/../_common:"$CONFDIR"/../_common/hadoop-xml:"$WHEREAMI/../lib/*" org.apache.druid.cli.Main server $SERVER_NAME elif [ "$#" -eq 4 ] then + # arguments for this case is JVMARGS=`cat "$CONFDIR/_common/common.jvm.config" | xargs` JVMARGS+=' ' JVMARGS+=$4 diff --git a/examples/conf/druid/single-server/medium/broker/runtime.properties b/examples/conf/druid/single-server/medium/broker/runtime.properties index d46d6de70514..32dff403fa10 100644 --- a/examples/conf/druid/single-server/medium/broker/runtime.properties +++ b/examples/conf/druid/single-server/medium/broker/runtime.properties @@ -20,16 +20,6 @@ druid.service=druid/broker druid.plaintextPort=8082 -# HTTP server settings -druid.server.http.numThreads=60 - -# HTTP client settings -druid.broker.http.numConnections=50 -druid.broker.http.maxQueuedBytes=10MiB - -# Processing threads and buffers -druid.processing.buffer.sizeBytes=500MiB -druid.processing.numMergeBuffers=4 druid.processing.tmpDir=var/druid/processing # Query cache disabled -- push down caching and merging instead diff --git a/examples/conf/druid/single-server/medium/historical/runtime.properties b/examples/conf/druid/single-server/medium/historical/runtime.properties index a7a640b8061d..82443c1c43c2 100644 --- a/examples/conf/druid/single-server/medium/historical/runtime.properties +++ b/examples/conf/druid/single-server/medium/historical/runtime.properties @@ -20,13 +20,6 @@ druid.service=druid/historical druid.plaintextPort=8083 -# HTTP server threads -druid.server.http.numThreads=60 - -# Processing threads and buffers -druid.processing.buffer.sizeBytes=500MiB -druid.processing.numMergeBuffers=4 -druid.processing.numThreads=15 druid.processing.tmpDir=var/druid/processing # Segment storage @@ -36,4 +29,3 @@ druid.segmentCache.locations=[{"path":"var/druid/segment-cache","maxSize":"300g" druid.historical.cache.useCache=true druid.historical.cache.populateCache=true druid.cache.type=caffeine -druid.cache.sizeInBytes=256MiB diff --git a/examples/conf/druid/single-server/medium/router/runtime.properties b/examples/conf/druid/single-server/medium/router/runtime.properties index 497d3b4d3aa1..3858dec044bd 100644 --- a/examples/conf/druid/single-server/medium/router/runtime.properties +++ b/examples/conf/druid/single-server/medium/router/runtime.properties @@ -20,12 +20,6 @@ druid.service=druid/router druid.plaintextPort=8888 -# HTTP proxy -druid.router.http.numConnections=50 -druid.router.http.readTimeout=PT5M -druid.router.http.numMaxThreads=100 -druid.server.http.numThreads=100 - # Service discovery druid.router.defaultBrokerServiceName=druid/broker druid.router.coordinatorServiceName=druid/coordinator From 68a1e664b225f750a42b57b5ae27683960404418 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Sat, 19 Nov 2022 01:05:25 +0530 Subject: [PATCH 009/141] simplify script arguments --- examples/bin/druid-quickstart.py | 228 +++++++++++++++---------------- examples/bin/run-druid | 25 ++-- 2 files changed, 121 insertions(+), 132 deletions(-) diff --git a/examples/bin/druid-quickstart.py b/examples/bin/druid-quickstart.py index 30ca0f60aad0..0b2da6658771 100644 --- a/examples/bin/druid-quickstart.py +++ b/examples/bin/druid-quickstart.py @@ -1,6 +1,7 @@ import sys import os import psutil +import pathlib from pathlib import Path QUICKSTART_BASE_CONFIG_PATH = "conf/druid/single-server/quickstart" @@ -8,10 +9,12 @@ COMPUTE_ONLY_ARG_IDENTIFIER = "computeOnly" RUN_ZK_IDENTIFIER = "runZk" ROOT_CONFIG_PATH_IDENTIFIER = "rootConfigPath" +SERVICES_IDENTIFIER = "services" MEMORY_ARG_IDENTIFIER = "totalMemory" MEMORY_GIGABYTES_IDENTIFIER = "g" MEMORY_MEGABYTES_IDENTIFIER = "m" ARG_SEPARATOR = "=" +SERVICE_SEPARATOR = "," BROKER_SERVICE_NAME = "broker" ROUTER_SERVICE_NAME = "router" @@ -58,7 +61,6 @@ def get_argument_value(argument): def parse_arguments(): service_list = [] - service_path_list = [] root_config_path = "" total_memory = "" compute_only = False @@ -71,54 +73,52 @@ def parse_arguments(): run_zk = True elif check_argument_type(argument, ROOT_CONFIG_PATH_IDENTIFIER): root_config_path = os.path.join(os.getcwd(), get_argument_value(argument)) + if os.path.exists(root_config_path) is False: + raise Exception(f'rootConfigPath `{root_config_path}` doesn\'t exist') elif (check_argument_type(argument, MEMORY_ARG_IDENTIFIER)): total_memory = get_argument_value(argument) - else: + elif (check_argument_type(argument, SERVICES_IDENTIFIER)): split_args = argument.split(ARG_SEPARATOR) - service = split_args[0] - - if service not in DEFAULT_SERVICES: - raise Exception(f'{service} is not a valid service name, should be one of {DEFAULT_SERVICES}') - - if service in service_list: - raise Exception(f'{service} is specified multiple times') + services_arg = split_args[1] + services = services_arg.split(SERVICE_SEPARATOR) - subdirectory = "" + for service in services: + if service not in DEFAULT_SERVICES: + raise Exception(f'{service} is not a valid service name, should be one of {DEFAULT_SERVICES}') - if len(split_args) == 2: - subdirectory = split_args[1] + if service in service_list: + raise Exception(f'{service} is specified multiple times') - if subdirectory != "": - complete_path = os.path.join(root_config_path, subdirectory) - if os.path.exists(os.path.join(complete_path)) is False: - raise Exception(f'Path `{complete_path}` specified for service `{service}` doesn\'t exist') - - service_list.append(service) - service_path_list.append(subdirectory) + service_list.append(service) if len(service_list) == 0: # start all services service_list = DEFAULT_SERVICES - service_path_list = [""] * len(DEFAULT_SERVICES) run_zk = True - return root_config_path, total_memory, list(zip(service_list, service_path_list)), run_zk, compute_only + return root_config_path, total_memory, service_list, run_zk, compute_only + +def print_startup_config(service_list, root_config_path, run_zk): + print(f'starting {service_list}, using config from {root_config_path}') + if run_zk: + zk_config_path = pathlib.Path(f'{os.getcwd()}/../conf/zk').resolve() + print(f'starting zk, using default config from {zk_config_path}') + print('\n') -def should_compute_memory(root_config_path, total_memory, service_config): +def should_compute_memory(root_config_path, total_memory, service_list): # if jvm file is present for any of the services # it should be present for all services and totalMemory should not be specified # if totalMemory is given, jvm file shouldn't be present for any service jvm_config_count = 0 - for item in service_config: - if item[1] != "": - if Path(f'{root_config_path}/{item[1]}/jvm.config').is_file(): - jvm_config_count += 1 - elif jvm_config_count > 0: - raise Exception('jvm.config file is missing for service {item[0]}, jvm.config should be specified for all the services or none') - - if jvm_config_count > 0 and (jvm_config_count != len(service_config) or total_memory != ""): - if jvm_config_count != len(service_config): + for service in service_list: + if Path(f'{root_config_path}/{service}/jvm.config').is_file(): + jvm_config_count += 1 + elif jvm_config_count > 0: + raise Exception('jvm.config file is missing for service {service}, jvm.config should be specified for all the services or none') + + if jvm_config_count > 0 and (jvm_config_count != len(service_list) or total_memory != ""): + if jvm_config_count != len(service_list): raise Exception("jvm.config file should be present for all services or none") if total_memory != "": raise Exception("If jvm.config is given for services, `totalMemory` argument shouldn't be specified") @@ -140,27 +140,26 @@ def convert_total_memory_string(memory): elif memory.endswith(MEMORY_GIGABYTES_IDENTIFIER): return 1024 * int(memory[:-1]) else: - raise Exception('Incorrect format for totalMemory argument, expected format is m or g') + raise Exception('Incorrect format for totalMemory argument, expected format is ') def build_memory_config_string(heap_memory, direct_memory): if direct_memory == 0: return f'-Xms{heap_memory}m -Xmx{heap_memory}m' return f'-Xms{heap_memory}m -Xmx{heap_memory}m -XX:MaxDirectMemorySize={direct_memory}m' -def distribute_memory_over_services(service_config, total_memory): +def distribute_memory_over_services(service_list, total_memory): service_memory_config = {} memory_weight_sum = 0 - for item in service_config: - memory_weight_sum += SERVICE_MEMORY_DISTRIBUTION_WEIGHT.get(item[0]) + for service in service_list: + memory_weight_sum += SERVICE_MEMORY_DISTRIBUTION_WEIGHT.get(service) multiplier = total_memory / memory_weight_sum lower_bound_memory_allocation = 0 allocated_services = set() - for item in service_config: - service = item[0] + for service in service_list: allocated_memory = SERVICE_MEMORY_DISTRIBUTION_WEIGHT.get(service) * multiplier if service in SERVICE_MEMORY_LOWER_BOUND and allocated_memory < SERVICE_MEMORY_LOWER_BOUND.get(service): allocated_memory = SERVICE_MEMORY_LOWER_BOUND.get(service) @@ -173,15 +172,13 @@ def distribute_memory_over_services(service_config, total_memory): if lower_bound_memory_allocation > 0: # compute the multiplier again for remaing services memory_weight_sum = 0 - for item in service_config: - service = item[0] + for service in service_list: if service in allocated_services: continue - memory_weight_sum += SERVICE_MEMORY_DISTRIBUTION_WEIGHT.get(service) * value + memory_weight_sum += SERVICE_MEMORY_DISTRIBUTION_WEIGHT.get(service) multiplier = (total_memory - lower_bound_memory_allocation) / memory_weight_sum - for item in service_config: - service = item[0] + for service in service_list: if service in allocated_services: continue allocated_memory = SERVICE_MEMORY_DISTRIBUTION_WEIGHT.get(service) * multiplier @@ -199,7 +196,7 @@ def distribute_memory_over_services(service_config, total_memory): return service_memory_config -def build_supervise_script_arguments(service_config, service_memory_config, root_config_path, run_zk): +def build_supervise_script_arguments(service_list, service_memory_config, root_config_path, run_zk): argument_list = [] argument_list.append("\":verify bin/verify-java\"") @@ -210,23 +207,19 @@ def build_supervise_script_arguments(service_config, service_memory_config, root if run_zk: argument_list.append("\"!p10 zk bin/run-zk conf\"") - for item in service_config: - service = item[0] + for service in service_list: prefix = '' if service == MIDDLE_MANAGER_SERVICE_NAME: prefix = '!p90 ' - if item[1] == "": - service_path = item[0] - else: - service_path = item[1] - jvm_args = service_memory_config.get(item[0]) + + jvm_args = service_memory_config.get(service) if jvm_args is None: - argument_list.append(f'\"{prefix}{service} bin/run-druid {service} {root_config_path} {service_path}\"') + argument_list.append(f'\"{prefix}{service} bin/run-druid {service} {root_config_path}\"') else: - argument_list.append(f'\"{prefix}{service} bin/run-druid {service} {root_config_path} {service_path} \'{jvm_args}\'\"') + argument_list.append(f'\"{prefix}{service} bin/run-druid {service} {root_config_path} \'{jvm_args}\'\"') - print('Command for supervise script:') + print('Supervise script command:') for item in argument_list: print(item) @@ -234,73 +227,75 @@ def build_supervise_script_arguments(service_config, service_memory_config, root return ",".join(argument_list) -def print_service_config(service_config, root_config_path, run_zk): - print('Services to start:') - for item in service_config: - if item[1] == "": - print(f'{item[0]}, using default config from {os.getcwd()}/../{QUICKSTART_BASE_CONFIG_PATH}') - else: - print(f'{item[0]}, using config from {root_config_path}/{item[1]}') - if run_zk: - print(f'zk, using default config from {os.getcwd()}/../conf/zk') - print('\n') - def display_help(): text = """ Usage: start-druid [options] where options include: - totalMemory= - memory for druid cluster, if totalMemory is not specified + totalMemory= + Integer value is supported with 'm' or 'g' suffix. + Memory for druid services, if totalMemory is not specified 80 percent of system memory is used. Note, if service specific jvm config is present, - totalMemory shouldn't be specified. - Integer value is supported with `m` or `g` suffix. - Memory should be greater than equals 2g - rootConfigPath= - directory containing common and service specific - properties to be overridden, this directory must contain `_common` - directory with `common.jvm.config` & `common.runtime.properties` - if `rootConfigPath` is not specified, config from - conf/druid/single-server/quickstart directory is used - computeOnly - command dry-run, validates the arguments and - display the memory distribution for services + shouldn't be specified. + should be greater than equals 3g + rootConfigPath= + Directory containing common and service specific + properties to be overridden, this directory must contain '_common' + directory with 'common.jvm.config' & 'common.runtime.properties' + files. + If is not specified, config from + 'conf/druid/single-server/quickstart' directory is used. + This path is relative to current working directory + services= + Value is comma separated string. + List of services to be started, should be a subset of + [broker, router, middleManager, historical, coordinator-overlord]. + If runtime or jvm properties are to be overridden, they should be + kept within /. + If is not specified config files can also be + placed within `conf/druid/single-server/quickstart/` + directory. + Note, if jvm.config file is present for one of the services, + it must be present for all services. + If argument is not given, all services + alongwith zookeeper is started. runZk - specification to run zookeeper, zk config is picked up from conf/zk - [=subdirectory] - service_identifier is the service to be started, multiple services - can be specified, `service_identifier` should be one of - [broker, router, middleManager, historical, coordinator-overlord] - `subdirectory` is optional directory within `rootConfigPath` - containing runtime properties or/and jvm properties - Note, if jvm.config file is present for one of the service, - it must be present for all other services. - If service is not explicitly specified, all services - alongwith zookeeper is started - Note each service should be specified at most once + Specification to run zookeeper, zk config is picked up from conf/zk. + computeOnly + Validate the arguments and display memory distribution for services. sample usage: start-druid - start up all the services (including zk) using the default system memory + Start up all the services (including zk) + using 80% of system memory. start-druid totalMemory=100g - start up all the services using the given memory + Start up all the services (including zk) + using the given memory. start-druid totalMemory=100g computeOnly - compute memory distribution for all the services - start-druid totalMemory=100g broker router historical - starts `broker`, `router` and `historical` services, using `100g` of memory - start-druid totalMemory=100g rootConfigPath=../conf/druid/single-server/large broker router historical - start `broker`, `router` and `historical` service, using 100g of memory, - use common configs from specified `rootConfigPath` - start-druid totalMemory=100g rootConfigPath=../conf/druid/single-server/large broker=broker router=router historical=historical - starts `broker`, `router` and `historical` services, using 100g of memory, use common configs - from specified `rootConfigPath`, use service specific config from specified sub directories - Since totalMemory is specific, the service specific folder shouldn't contain jvm.config - start-druid rootConfigPath=../conf/druid/profile broker=broker historical=historical - exception is thrown if either of `broker`, `historical` - subdirectory contains jvm.config but not both. - If none of the service specific subdirectory contains jvm.config, - memory distribution is computed + Compute memory distribution and validate + arguments for starting all the services. + start-druid totalMemory=100g services=broker,router + Start `broker` & `router` service, using `100g` of memory. + Read config from conf/druid/single-server/quickstart. + start-druid totalMemory=100g rootConfigPath=conf/druid/single-server/custom services=broker,router + Start `broker` & `router` service, using 100g of memory. + Read config from . + Since is specified, exception is thrown if + jvm.config is present for any of the services. + start-druid rootConfigPath=conf/druid/single-server/custom services=broker,router + Start `broker` & `router` service, + using 80% of system memory. + If jvm.config is specified for both the + services within /, + memory distribution is not calculated. + If jvm.config is present for either of the services, + exception is thrown. + If jvm.config is not present for both of the services, + memory distribution is calculated. + start-druid totalMemory=100g rootConfigPath=conf/druid/single-server/custom services=broker,router runZk + Start zookeeper alongwith other services. + zk config is read from conf/zk. """ print(text) @@ -314,23 +309,24 @@ def main(): print("Druid quickstart\n") - root_config_path, total_memory, service_config, run_zk, compute_only = parse_arguments() + root_config_path, total_memory, service_list, run_zk, compute_only = parse_arguments() # change directory to bin os.chdir(os.path.dirname(sys.argv[0])) - print(f'Arguments passed: rootConfigPath: "{root_config_path}", totalMemory: "{total_memory}"\n') - print_service_config(service_config, root_config_path, run_zk) + if root_config_path == "": + root_config_path = pathlib.Path(f'{os.getcwd()}/../{QUICKSTART_BASE_CONFIG_PATH}').resolve() + + print_startup_config(service_list, root_config_path, run_zk) service_memory_config = {} - if (should_compute_memory(root_config_path, total_memory, service_config)): + if (should_compute_memory(root_config_path, total_memory, service_list)): memory_in_mega_bytes = convert_total_memory_string(total_memory) - service_memory_config = distribute_memory_over_services(service_config, memory_in_mega_bytes) - - if root_config_path == "": - root_config_path = QUICKSTART_BASE_CONFIG_PATH + service_memory_config = distribute_memory_over_services(service_list, memory_in_mega_bytes) + else: + print('not computing memory distribution, reading memory specification from service jvm.config\n') - script_arguments = build_supervise_script_arguments(service_config, service_memory_config, root_config_path, run_zk) + script_arguments = build_supervise_script_arguments(service_list, service_memory_config, root_config_path, run_zk) if compute_only: return diff --git a/examples/bin/run-druid b/examples/bin/run-druid index 2726477684df..4d050d3d5712 100755 --- a/examples/bin/run-druid +++ b/examples/bin/run-druid @@ -17,7 +17,7 @@ # specific language governing permissions and limitations # under the License. -if [ "$#" -gt 4 ] || [ "$#" -eq 0 ] +if [ "$#" -gt 3 ] || [ "$#" -eq 0 ] then >&2 echo "usage: $0 [conf-dir]" exit 1 @@ -47,25 +47,18 @@ if [ ! -d "$LOG_DIR" ]; then mkdir -p $LOG_DIR; fi echo "Running [$1], logging to [$LOG_DIR/$1.log] if no changes made to log4j2.xml" -if [ "$WHATAMI" = 'coordinator-overlord' ] -then - SERVER_NAME=coordinator -else - SERVER_NAME="$WHATAMI" -fi - if [ "$#" -eq 3 ] then - # arguments for this case is - cd "$WHEREAMI/.." - exec "$WHEREAMI"/run-java -Ddruid.node.type=$1 "-Ddruid.log.path=$LOG_DIR" `cat "$CONFDIR"/"$3"/jvm.config | xargs` \ - -cp "$CONFDIR"/"$WHATAMI":"$CONFDIR"/_common:"$CONFDIR"/_common/hadoop-xml:"$CONFDIR"/../_common:"$CONFDIR"/../_common/hadoop-xml:"$WHEREAMI/../lib/*" org.apache.druid.cli.Main server $SERVER_NAME -elif [ "$#" -eq 4 ] -then - # arguments for this case is JVMARGS=`cat "$CONFDIR/_common/common.jvm.config" | xargs` JVMARGS+=' ' - JVMARGS+=$4 + JVMARGS+=$3 + + if [ "$WHATAMI" = 'coordinator-overlord' ] + then + SERVER_NAME=coordinator + else + SERVER_NAME="$WHATAMI" + fi cd "$WHEREAMI/.." exec "$WHEREAMI"/run-java -Ddruid.node.type=$1 "-Ddruid.log.path=$LOG_DIR" $JVMARGS \ From e5706066c56527759c814a832eb226828cad90d9 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Sat, 19 Nov 2022 01:40:57 +0530 Subject: [PATCH 010/141] restore changes in medium profile --- .../single-server/medium/broker/runtime.properties | 10 ++++++++++ .../single-server/medium/historical/runtime.properties | 8 ++++++++ .../single-server/medium/router/runtime.properties | 6 ++++++ 3 files changed, 24 insertions(+) diff --git a/examples/conf/druid/single-server/medium/broker/runtime.properties b/examples/conf/druid/single-server/medium/broker/runtime.properties index 32dff403fa10..d46d6de70514 100644 --- a/examples/conf/druid/single-server/medium/broker/runtime.properties +++ b/examples/conf/druid/single-server/medium/broker/runtime.properties @@ -20,6 +20,16 @@ druid.service=druid/broker druid.plaintextPort=8082 +# HTTP server settings +druid.server.http.numThreads=60 + +# HTTP client settings +druid.broker.http.numConnections=50 +druid.broker.http.maxQueuedBytes=10MiB + +# Processing threads and buffers +druid.processing.buffer.sizeBytes=500MiB +druid.processing.numMergeBuffers=4 druid.processing.tmpDir=var/druid/processing # Query cache disabled -- push down caching and merging instead diff --git a/examples/conf/druid/single-server/medium/historical/runtime.properties b/examples/conf/druid/single-server/medium/historical/runtime.properties index 82443c1c43c2..a7a640b8061d 100644 --- a/examples/conf/druid/single-server/medium/historical/runtime.properties +++ b/examples/conf/druid/single-server/medium/historical/runtime.properties @@ -20,6 +20,13 @@ druid.service=druid/historical druid.plaintextPort=8083 +# HTTP server threads +druid.server.http.numThreads=60 + +# Processing threads and buffers +druid.processing.buffer.sizeBytes=500MiB +druid.processing.numMergeBuffers=4 +druid.processing.numThreads=15 druid.processing.tmpDir=var/druid/processing # Segment storage @@ -29,3 +36,4 @@ druid.segmentCache.locations=[{"path":"var/druid/segment-cache","maxSize":"300g" druid.historical.cache.useCache=true druid.historical.cache.populateCache=true druid.cache.type=caffeine +druid.cache.sizeInBytes=256MiB diff --git a/examples/conf/druid/single-server/medium/router/runtime.properties b/examples/conf/druid/single-server/medium/router/runtime.properties index 3858dec044bd..497d3b4d3aa1 100644 --- a/examples/conf/druid/single-server/medium/router/runtime.properties +++ b/examples/conf/druid/single-server/medium/router/runtime.properties @@ -20,6 +20,12 @@ druid.service=druid/router druid.plaintextPort=8888 +# HTTP proxy +druid.router.http.numConnections=50 +druid.router.http.readTimeout=PT5M +druid.router.http.numMaxThreads=100 +druid.server.http.numThreads=100 + # Service discovery druid.router.defaultBrokerServiceName=druid/broker druid.router.coordinatorServiceName=druid/coordinator From b413d159ef0ec9b9cba3d01518bffbbd622da7a5 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Mon, 21 Nov 2022 11:53:52 +0530 Subject: [PATCH 011/141] run-druid refactor --- examples/bin/run-druid | 34 ++++++++++++++++++++++------------ 1 file changed, 22 insertions(+), 12 deletions(-) diff --git a/examples/bin/run-druid b/examples/bin/run-druid index 4d050d3d5712..258f8376d0d2 100755 --- a/examples/bin/run-druid +++ b/examples/bin/run-druid @@ -47,25 +47,35 @@ if [ ! -d "$LOG_DIR" ]; then mkdir -p $LOG_DIR; fi echo "Running [$1], logging to [$LOG_DIR/$1.log] if no changes made to log4j2.xml" +if [ "$WHATAMI" = 'coordinator-overlord' ] +then + SERVER_NAME=coordinator +else + SERVER_NAME="$WHATAMI" +fi + + +if [ ! -f "$CONFDIR"/$WHATAMI/main.config ]; + then + MAIN_CLASS="org.apache.druid.cli.Main server $SERVER_NAME" + else + MAIN_CLASS=`cat "$CONFDIR"/$WHATAMI/main.config | xargs` +fi + +cd "$WHEREAMI/.." + if [ "$#" -eq 3 ] +then + +elif [ "$#" -eq 4 ] then JVMARGS=`cat "$CONFDIR/_common/common.jvm.config" | xargs` JVMARGS+=' ' JVMARGS+=$3 - if [ "$WHATAMI" = 'coordinator-overlord' ] - then - SERVER_NAME=coordinator - else - SERVER_NAME="$WHATAMI" - fi - - cd "$WHEREAMI/.." exec "$WHEREAMI"/run-java -Ddruid.node.type=$1 "-Ddruid.log.path=$LOG_DIR" $JVMARGS \ - -cp "$CONFDIR"/"$WHATAMI":"$CONFDIR"/_common:"$CONFDIR"/_common/hadoop-xml:"$CONFDIR"/../_common:"$CONFDIR"/../_common/hadoop-xml:"$WHEREAMI/../lib/*" org.apache.druid.cli.Main server $SERVER_NAME + -cp "$CONFDIR"/"$WHATAMI":"$CONFDIR"/_common:"$CONFDIR"/_common/hadoop-xml:"$CONFDIR"/../_common:"$CONFDIR"/../_common/hadoop-xml:"$WHEREAMI/../lib/*" $MAIN_CLASS else - cd "$WHEREAMI/.." exec "$WHEREAMI"/run-java -Ddruid.node.type=$1 "-Ddruid.log.path=$LOG_DIR" `cat "$CONFDIR"/"$WHATAMI"/jvm.config | xargs` \ - -cp "$CONFDIR"/"$WHATAMI":"$CONFDIR"/_common:"$CONFDIR"/_common/hadoop-xml:"$CONFDIR"/../_common:"$CONFDIR"/../_common/hadoop-xml:"$WHEREAMI/../lib/*" \ - `cat "$CONFDIR"/$WHATAMI/main.config | xargs` + -cp "$CONFDIR"/"$WHATAMI":"$CONFDIR"/_common:"$CONFDIR"/_common/hadoop-xml:"$CONFDIR"/../_common:"$CONFDIR"/../_common/hadoop-xml:"$WHEREAMI/../lib/*" $MAIN_CLASS fi From 68018a808fff8d04e5875ec366b7d3a7a0a16d9e Mon Sep 17 00:00:00 2001 From: Jill Osborne Date: Mon, 21 Nov 2022 19:17:12 +0000 Subject: [PATCH 012/141] Firehose migration doc (#12981) * Firehose migration doc * Update migrate-from-firehose-ingestion.md * Updated with review comments and suggestions * Update migrate-from-firehose-ingestion.md * Update migrate-from-firehose-ingestion.md * Update migrate-from-firehose-ingestion.md --- .../migrate-from-firehose-ingestion.md | 209 ++++++++++++++++++ docs/ingestion/native-batch-firehose.md | 11 +- website/sidebars.json | 2 + 3 files changed, 216 insertions(+), 6 deletions(-) create mode 100644 docs/ingestion/migrate-from-firehose-ingestion.md diff --git a/docs/ingestion/migrate-from-firehose-ingestion.md b/docs/ingestion/migrate-from-firehose-ingestion.md new file mode 100644 index 000000000000..2aa77772652c --- /dev/null +++ b/docs/ingestion/migrate-from-firehose-ingestion.md @@ -0,0 +1,209 @@ +--- +id: migrate-from-firehose +title: "Migrate from firehose to input source ingestion" +sidebar_label: "Migrate from firehose" +--- + + + +Apache deprecated support for Druid firehoses in version 0.17. Support for firehose ingestion was removed in version 24.0. + +If you're using a firehose for batch ingestion, we strongly recommend that you follow the instructions on this page to transition to using native batch ingestion input sources as soon as possible. + +Firehose ingestion doesn't work with newer Druid versions, so you must be using an ingestion spec with a defined input source before you upgrade. + +## Migrate from firehose ingestion to an input source + +To migrate from firehose ingestion, you can use the Druid console to update your ingestion spec, or you can update it manually. + +### Use the Druid console + +To update your ingestion spec using the Druid console, open the console and copy your spec into the **Edit spec** stage of the data loader. + +Druid converts the spec into one with a defined input source. For example, it converts the [example firehose ingestion spec](#example-firehose-ingestion-spec) below into the [example ingestion spec after migration](#example-ingestion-spec-after-migration). + +If you're unable to use the console or you have problems with the console method, the alternative is to update your ingestion spec manually. + +### Update your ingestion spec manually + +To update your ingestion spec manually, copy your existing spec into a new file. Refer to [Native batch ingestion with firehose (Deprecated)](./native-batch-firehose.md) for a description of firehose properties. + +Edit the new file as follows: + +1. In the `ioConfig` component, replace the `firehose` definition with an `inputSource` definition for your chosen input source. See [Native batch input sources](./native-batch-input-source.md) for details. +2. Move the `timeStampSpec` definition from `parser.parseSpec` to the `dataSchema` component. +3. Move the `dimensionsSpec` definition from `parser.parseSpec` to the `dataSchema` component. +4. Move the `format` definition from `parser.parseSpec` to an `inputFormat` definition in `ioConfig`. +5. Delete the `parser` definition. +6. Save the file. +
You can check the format of your new ingestion file against the [migrated example](#example-ingestion-spec-after-migration) below. +7. Test the new ingestion spec with a temporary data source. +8. Once you've successfully ingested sample data with the new spec, stop firehose ingestion and switch to the new spec. + +When the transition is complete, you can upgrade Druid to the latest version. See the [Druid release notes](https://druid.apache.org/downloads.html) for upgrade instructions. + +### Example firehose ingestion spec + +An example firehose ingestion spec is as follows: + +```json +{ + "type" : "index", + "spec" : { + "dataSchema" : { + "dataSource" : "wikipedia", + "metricsSpec" : [ + { + "type" : "count", + "name" : "count" + }, + { + "type" : "doubleSum", + "name" : "added", + "fieldName" : "added" + }, + { + "type" : "doubleSum", + "name" : "deleted", + "fieldName" : "deleted" + }, + { + "type" : "doubleSum", + "name" : "delta", + "fieldName" : "delta" + } + ], + "granularitySpec" : { + "type" : "uniform", + "segmentGranularity" : "DAY", + "queryGranularity" : "NONE", + "intervals" : [ "2013-08-31/2013-09-01" ] + }, + "parser": { + "type": "string", + "parseSpec": { + "format": "json", + "timestampSpec" : { + "column" : "timestamp", + "format" : "auto" + }, + "dimensionsSpec" : { + "dimensions": ["country", "page","language","user","unpatrolled","newPage","robot","anonymous","namespace","continent","region","city"], + "dimensionExclusions" : [] + } + } + } + }, + "ioConfig" : { + "type" : "index", + "firehose" : { + "type" : "local", + "baseDir" : "examples/indexing/", + "filter" : "wikipedia_data.json" + } + }, + "tuningConfig" : { + "type" : "index", + "partitionsSpec": { + "type": "single_dim", + "partitionDimension": "country", + "targetRowsPerSegment": 5000000 + } + } + } +} +``` + +### Example ingestion spec after migration + +The following example illustrates the result of migrating the [example firehose ingestion spec](#example-firehose-ingestion-spec) to a spec with an input source: + +```json +{ + "type" : "index", + "spec" : { + "dataSchema" : { + "dataSource" : "wikipedia", + "timestampSpec" : { + "column" : "timestamp", + "format" : "auto" + }, + "dimensionsSpec" : { + "dimensions": ["country", "page","language","user","unpatrolled","newPage","robot","anonymous","namespace","continent","region","city"], + "dimensionExclusions" : [] + }, + "metricsSpec" : [ + { + "type" : "count", + "name" : "count" + }, + { + "type" : "doubleSum", + "name" : "added", + "fieldName" : "added" + }, + { + "type" : "doubleSum", + "name" : "deleted", + "fieldName" : "deleted" + }, + { + "type" : "doubleSum", + "name" : "delta", + "fieldName" : "delta" + } + ], + "granularitySpec" : { + "type" : "uniform", + "segmentGranularity" : "DAY", + "queryGranularity" : "NONE", + "intervals" : [ "2013-08-31/2013-09-01" ] + } + }, + "ioConfig" : { + "type" : "index", + "inputSource" : { + "type" : "local", + "baseDir" : "examples/indexing/", + "filter" : "wikipedia_data.json" + }, + "inputFormat": { + "type": "json" + } + }, + "tuningConfig" : { + "type" : "index", + "partitionsSpec": { + "type": "single_dim", + "partitionDimension": "country", + "targetRowsPerSegment": 5000000 + } + } + } +} +``` + +## Learn more + +For more information, see the following pages: + +- [Ingestion](./index.md): Overview of the Druid ingestion process. +- [Native batch ingestion](./native-batch.md): Description of the supported native batch indexing tasks. +- [Ingestion spec reference](./ingestion-spec.md): Description of the components and properties in the ingestion spec. diff --git a/docs/ingestion/native-batch-firehose.md b/docs/ingestion/native-batch-firehose.md index 4e2cad97fc1a..ca848e725c87 100644 --- a/docs/ingestion/native-batch-firehose.md +++ b/docs/ingestion/native-batch-firehose.md @@ -1,6 +1,6 @@ --- id: native-batch-firehose -title: "Native batch ingestion with firehose" +title: "Native batch ingestion with firehose (Deprecated)" sidebar_label: "Firehose (deprecated)" --- @@ -23,14 +23,13 @@ sidebar_label: "Firehose (deprecated)" ~ under the License. --> - -Firehoses are deprecated in 0.17.0. It's highly recommended to use the [Native batch ingestion input sources](./native-batch-input-source.md) instead. +> Firehose ingestion is deprecated. See [Migrate from firehose to input source ingestion](./migrate-from-firehose-ingestion.md) for instructions on migrating from firehose ingestion to using native batch ingestion input sources. There are several firehoses readily available in Druid, some are meant for examples, others can be used directly in a production environment. ## StaticS3Firehose -> You need to include the [`druid-s3-extensions`](../development/extensions-core/s3.md) as an extension to use the StaticS3Firehose. +You need to include the [`druid-s3-extensions`](../development/extensions-core/s3.md) as an extension to use the StaticS3Firehose. This firehose ingests events from a predefined list of S3 objects. This firehose is _splittable_ and can be used by the [Parallel task](./native-batch.md). @@ -62,7 +61,7 @@ Note that prefetching or caching isn't that useful in the Parallel task. ## StaticGoogleBlobStoreFirehose -> You need to include the [`druid-google-extensions`](../development/extensions-core/google.md) as an extension to use the StaticGoogleBlobStoreFirehose. +You need to include the [`druid-google-extensions`](../development/extensions-core/google.md) as an extension to use the StaticGoogleBlobStoreFirehose. This firehose ingests events, similar to the StaticS3Firehose, but from an Google Cloud Store. @@ -112,7 +111,7 @@ Google Blobs: ## HDFSFirehose -> You need to include the [`druid-hdfs-storage`](../development/extensions-core/hdfs.md) as an extension to use the HDFSFirehose. +You need to include the [`druid-hdfs-storage`](../development/extensions-core/hdfs.md) as an extension to use the HDFSFirehose. This firehose ingests events from a predefined list of files from the HDFS storage. This firehose is _splittable_ and can be used by the [Parallel task](./native-batch.md). diff --git a/website/sidebars.json b/website/sidebars.json index a398f9fd3642..1ebc2140279f 100644 --- a/website/sidebars.json +++ b/website/sidebars.json @@ -57,6 +57,8 @@ "ids": [ "ingestion/native-batch", "ingestion/native-batch-input-sources", + "ingestion/migrate-from-firehose", + "ingestion/native-batch-firehose", "ingestion/hadoop" ] }, From fe34ecc5e35d5091c3ca6c9043d6eac0db1dba56 Mon Sep 17 00:00:00 2001 From: Vadim Ogievetsky Date: Mon, 21 Nov 2022 12:50:44 -0800 Subject: [PATCH 013/141] add ability to make inputFormat part of the example datasets (#13402) --- .../input-source-step/example-inputs.ts | 67 ++++++++++++++++++- .../input-source-step/input-source-step.tsx | 36 ++++++---- 2 files changed, 86 insertions(+), 17 deletions(-) diff --git a/web-console/src/views/workbench-view/input-source-step/example-inputs.ts b/web-console/src/views/workbench-view/input-source-step/example-inputs.ts index e58dfacca35e..a74f1754b179 100644 --- a/web-console/src/views/workbench-view/input-source-step/example-inputs.ts +++ b/web-console/src/views/workbench-view/input-source-step/example-inputs.ts @@ -16,15 +16,74 @@ * limitations under the License. */ -import { InputSource } from '../../../druid-models'; +import { InputFormat, InputSource } from '../../../druid-models'; -export interface ExampleInputSource { +export interface ExampleInput { name: string; description: string; inputSource: InputSource; + inputFormat?: InputFormat; } -export const EXAMPLE_INPUT_SOURCES: ExampleInputSource[] = [ +const TRIPS_INPUT_FORMAT: InputFormat = { + type: 'csv', + findColumnsFromHeader: false, + columns: [ + 'trip_id', + 'vendor_id', + 'pickup_datetime', + 'dropoff_datetime', + 'store_and_fwd_flag', + 'rate_code_id', + 'pickup_longitude', + 'pickup_latitude', + 'dropoff_longitude', + 'dropoff_latitude', + 'passenger_count', + 'trip_distance', + 'fare_amount', + 'extra', + 'mta_tax', + 'tip_amount', + 'tolls_amount', + 'ehail_fee', + 'improvement_surcharge', + 'total_amount', + 'payment_type', + 'trip_type', + 'pickup', + 'dropoff', + 'cab_type', + 'precipitation', + 'snow_depth', + 'snowfall', + 'max_temperature', + 'min_temperature', + 'average_wind_speed', + 'pickup_nyct2010_gid', + 'pickup_ctlabel', + 'pickup_borocode', + 'pickup_boroname', + 'pickup_ct2010', + 'pickup_boroct2010', + 'pickup_cdeligibil', + 'pickup_ntacode', + 'pickup_ntaname', + 'pickup_puma', + 'dropoff_nyct2010_gid', + 'dropoff_ctlabel', + 'dropoff_borocode', + 'dropoff_boroname', + 'dropoff_ct2010', + 'dropoff_boroct2010', + 'dropoff_cdeligibil', + 'dropoff_ntacode', + 'dropoff_ntaname', + 'dropoff_puma', + ], +}; + +export const EXAMPLE_INPUTS: ExampleInput[] = [ { name: 'Wikipedia', description: 'One day of wikipedia edits (JSON)', @@ -62,6 +121,7 @@ export const EXAMPLE_INPUT_SOURCES: ExampleInputSource[] = [ 'https://static.imply.io/example-data/trips/trips_xac.csv.gz', ], }, + inputFormat: TRIPS_INPUT_FORMAT, }, { name: 'NYC Taxi cabs (all files)', @@ -145,6 +205,7 @@ export const EXAMPLE_INPUT_SOURCES: ExampleInputSource[] = [ 'https://static.imply.io/example-data/trips/trips_xcv.csv.gz', ], }, + inputFormat: TRIPS_INPUT_FORMAT, }, { name: 'FlightCarrierOnTime (1 month)', diff --git a/web-console/src/views/workbench-view/input-source-step/input-source-step.tsx b/web-console/src/views/workbench-view/input-source-step/input-source-step.tsx index 211271c62c3d..f144e8f975d5 100644 --- a/web-console/src/views/workbench-view/input-source-step/input-source-step.tsx +++ b/web-console/src/views/workbench-view/input-source-step/input-source-step.tsx @@ -55,7 +55,7 @@ import { UrlBaser } from '../../../singletons'; import { filterMap, IntermediateQueryState } from '../../../utils'; import { postToSampler, SampleSpec } from '../../../utils/sampler'; -import { EXAMPLE_INPUT_SOURCES } from './example-inputs'; +import { EXAMPLE_INPUTS } from './example-inputs'; import { InputSourceInfo } from './input-source-info'; import './input-source-step.scss'; @@ -81,16 +81,15 @@ export const InputSourceStep = React.memo(function InputSourceStep(props: InputS const [inputSource, setInputSource] = useState | string | undefined>( initInputSource, ); - const exampleInputSource = EXAMPLE_INPUT_SOURCES.find( - ({ name }) => name === inputSource, - )?.inputSource; + const exampleInput = EXAMPLE_INPUTS.find(({ name }) => name === inputSource); const [guessedInputFormatState, connectQueryManager] = useQueryManager< - InputSource, + { inputSource: InputSource; suggestedInputFormat?: InputFormat }, InputFormat, Execution >({ - processQuery: async (inputSource: InputSource, cancelToken) => { + processQuery: async ({ inputSource, suggestedInputFormat }, cancelToken) => { + let guessedInputFormat: InputFormat | undefined; if (mode === 'sampler') { const sampleSpec: SampleSpec = { type: 'index_parallel', @@ -127,7 +126,7 @@ export const InputSourceStep = React.memo(function InputSourceStep(props: InputS ); if (!sampleLines.length) throw new Error('No data returned from sampler'); - return guessInputFormat(sampleLines); + guessedInputFormat = guessInputFormat(sampleLines); } else { const tableExpression = externalConfigToTableExpression({ inputSource, @@ -151,8 +150,14 @@ export const InputSourceStep = React.memo(function InputSourceStep(props: InputS ); if (result instanceof IntermediateQueryState) return result; - return resultToInputFormat(result); + guessedInputFormat = resultToInputFormat(result); } + + if (suggestedInputFormat?.type === guessedInputFormat.type) { + return suggestedInputFormat; + } + + return guessedInputFormat; }, backgroundStatusCheck: async (execution, query, cancelToken) => { const result = await executionBackgroundResultStatusCheck(execution, query, cancelToken); @@ -164,7 +169,7 @@ export const InputSourceStep = React.memo(function InputSourceStep(props: InputS useEffect(() => { const guessedInputFormat = guessedInputFormatState.data; if (!guessedInputFormat) return; - onSet(exampleInputSource || (inputSource as any), guessedInputFormat); + onSet(exampleInput?.inputSource || (inputSource as any), guessedInputFormat); // eslint-disable-next-line react-hooks/exhaustive-deps }, [guessedInputFormatState]); @@ -217,7 +222,7 @@ export const InputSourceStep = React.memo(function InputSourceStep(props: InputS selectedValue={inputSource} onChange={e => setInputSource(e.currentTarget.value)} > - {EXAMPLE_INPUT_SOURCES.map((e, i) => ( + {EXAMPLE_INPUTS.map((e, i) => ( { - if (!exampleInputSource) return; - connectQueryManager.runQuery(exampleInputSource); + if (!exampleInput) return; + connectQueryManager.runQuery({ + inputSource: exampleInput.inputSource, + suggestedInputFormat: exampleInput.inputFormat, + }); }} /> ) : inputSource ? ( @@ -324,7 +332,7 @@ export const InputSourceStep = React.memo(function InputSourceStep(props: InputS } onClick={() => { if (!AutoForm.isValidModel(inputSource, INPUT_SOURCE_FIELDS)) return; - connectQueryManager.runQuery(inputSource); + connectQueryManager.runQuery({ inputSource }); }} /> ) : undefined} From 280a0f7158eba2a5a5bf2ab0a9fd0ccc00656fcb Mon Sep 17 00:00:00 2001 From: Adarsh Sanjeev Date: Tue, 22 Nov 2022 09:56:32 +0530 Subject: [PATCH 014/141] Add sequential sketch merging to MSQ (#13205) * Add sketch fetching framework * Refactor code to support sequential merge * Update worker sketch fetcher * Refactor sketch fetcher * Refactor sketch fetcher * Add context parameter and threshold to trigger sequential merge * Fix test * Add integration test for non sequential merge * Address review comments * Address review comments * Address review comments * Resolve maxRetainedBytes * Add new classes * Renamed key statistics information class * Rename fetchStatisticsSnapshotForTimeChunk function * Address review comments * Address review comments * Update documentation and add comments * Resolve build issues * Resolve build issues * Change worker APIs to async * Address review comments * Resolve build issues * Add null time check * Update integration tests * Address review comments * Add log messages and comments * Resolve build issues * Add unit tests * Add unit tests * Fix timing issue in tests --- docs/multi-stage-query/reference.md | 22 ++ .../msq/exec/ClusterStatisticsMergeMode.java | 44 +++ .../org/apache/druid/msq/exec/Controller.java | 8 +- .../druid/msq/exec/ControllerClient.java | 10 +- .../apache/druid/msq/exec/ControllerImpl.java | 63 +++- .../exec/ExceptionWrappingWorkerClient.java | 18 + .../org/apache/druid/msq/exec/Worker.java | 13 + .../apache/druid/msq/exec/WorkerClient.java | 22 ++ .../org/apache/druid/msq/exec/WorkerImpl.java | 26 +- .../druid/msq/exec/WorkerSketchFetcher.java | 340 ++++++++++++++++++ .../msq/indexing/ControllerChatHandler.java | 16 +- .../msq/indexing/IndexerControllerClient.java | 10 +- .../msq/indexing/IndexerWorkerClient.java | 43 +++ .../druid/msq/indexing/WorkerChatHandler.java | 40 +++ .../controller/ControllerQueryKernel.java | 46 ++- .../controller/ControllerStagePhase.java | 13 +- .../controller/ControllerStageTracker.java | 127 ++++--- .../ClusterByStatisticsCollectorImpl.java | 20 +- .../ClusterByStatisticsSnapshot.java | 31 +- .../CompleteKeyStatisticsInformation.java | 82 +++++ .../DelegateOrMinKeyCollectorSnapshot.java | 3 + .../msq/statistics/DistinctKeySnapshot.java | 3 + .../msq/statistics/KeyCollectorSnapshot.java | 9 + .../PartialKeyStatisticsInformation.java | 67 ++++ .../QuantilesSketchKeyCollectorSnapshot.java | 3 + .../msq/util/MultiStageQueryContext.java | 15 + .../exec/WorkerSketchFetcherAutoModeTest.java | 139 +++++++ .../msq/exec/WorkerSketchFetcherTest.java | 295 +++++++++++++++ .../BaseControllerQueryKernelTest.java | 27 +- .../ControllerQueryKernelTests.java | 55 ++- ...tialKeyStatisticsInformationSerdeTest.java | 62 ++++ .../msq/test/MSQTestControllerClient.java | 10 +- .../druid/msq/test/MSQTestWorkerClient.java | 24 ++ integration-tests-ex/README.md | 2 +- integration-tests-ex/cases/pom.xml | 5 + .../msq/ITKeyStatisticsSketchMergeMode.java | 206 +++++++++++ .../apache/druid/frame/key/RowKeyReader.java | 22 ++ .../druid/frame/key/RowKeyReaderTest.java | 38 ++ website/.spelling | 3 + 39 files changed, 1850 insertions(+), 132 deletions(-) create mode 100644 extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/ClusterStatisticsMergeMode.java create mode 100644 extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/WorkerSketchFetcher.java create mode 100644 extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/statistics/CompleteKeyStatisticsInformation.java create mode 100644 extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/statistics/PartialKeyStatisticsInformation.java create mode 100644 extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/WorkerSketchFetcherAutoModeTest.java create mode 100644 extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/WorkerSketchFetcherTest.java create mode 100644 extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/statistics/PartialKeyStatisticsInformationSerdeTest.java create mode 100644 integration-tests-ex/cases/src/test/java/org/apache/druid/testsEx/msq/ITKeyStatisticsSketchMergeMode.java diff --git a/docs/multi-stage-query/reference.md b/docs/multi-stage-query/reference.md index 3550566874b8..a4bcbfc27b1f 100644 --- a/docs/multi-stage-query/reference.md +++ b/docs/multi-stage-query/reference.md @@ -203,6 +203,28 @@ The following table lists the context parameters for the MSQ task engine: | `maxParseExceptions`| SELECT, INSERT, REPLACE

Maximum number of parse exceptions that are ignored while executing the query before it stops with `TooManyWarningsFault`. To ignore all the parse exceptions, set the value to -1.| 0 | | `rowsPerSegment` | INSERT or REPLACE

The number of rows per segment to target. The actual number of rows per segment may be somewhat higher or lower than this number. In most cases, use the default. For general information about sizing rows per segment, see [Segment Size Optimization](../operations/segment-optimization.md). | 3,000,000 | | `indexSpec` | INSERT or REPLACE

An [`indexSpec`](../ingestion/ingestion-spec.md#indexspec) to use when generating segments. May be a JSON string or object. See [Front coding](../ingestion/ingestion-spec.md#front-coding) for details on configuring an `indexSpec` with front coding. | See [`indexSpec`](../ingestion/ingestion-spec.md#indexspec). | +| `clusterStatisticsMergeMode` | Whether to use parallel or sequential mode for merging of the worker sketches. Can be `PARALLEL`, `SEQUENTIAL` or `AUTO`. See [Sketch Merging Mode](#sketch-merging-mode) for more information. | `AUTO` | + +## Sketch Merging Mode +This section details the advantages and performance of various Cluster By Statistics Merge Modes. + +If a query requires key statistics to generate partition boundaries, key statistics are gathered by the workers while +reading rows from the datasource. These statistics must be transferred to the controller to be merged together. +`clusterStatisticsMergeMode` configures the way in which this happens. + +`PARALLEL` mode fetches the key statistics for all time chunks from all workers together and the controller then downsamples +the sketch if it does not fit in memory. This is faster than `SEQUENTIAL` mode as there is less over head in fetching sketches +for all time chunks together. This is good for small sketches which won't be downsampled even if merged together or if +accuracy in segment sizing for the ingestion is not very important. + +`SEQUENTIAL` mode fetches the sketches in ascending order of time and generates the partition boundaries for one time +chunk at a time. This gives more working memory to the controller for merging sketches, which results in less +downsampling and thus, more accuracy. There is, however, a time overhead on fetching sketches in sequential order. This is +good for cases where accuracy is important. + +`AUTO` mode tries to find the best approach based on number of workers and size of input rows. If there are more +than 100 workers or if the combined sketch size among all workers is more than 1GB, `SEQUENTIAL` is chosen, otherwise, +`PARALLEL` is chosen. ## Durable Storage This section enumerates the advantages and performance implications of enabling durable storage while executing MSQ tasks. diff --git a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/ClusterStatisticsMergeMode.java b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/ClusterStatisticsMergeMode.java new file mode 100644 index 000000000000..92ed82ff5e11 --- /dev/null +++ b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/ClusterStatisticsMergeMode.java @@ -0,0 +1,44 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.msq.exec; + +/** + * Mode which dictates how {@link WorkerSketchFetcher} gets sketches for the partition boundaries from workers. + */ +public enum ClusterStatisticsMergeMode +{ + /** + * Fetches sketch in sequential order based on time. Slower due to overhead, but more accurate. + */ + SEQUENTIAL, + + /** + * Fetch all sketches from the worker at once. Faster to generate partitions, but less accurate. + */ + PARALLEL, + + /** + * Tries to decide between sequential and parallel modes based on the number of workers and size of the input + * + * If there are more than 100 workers or if the combined sketch size among all workers is more than + * 1,000,000,000 bytes, SEQUENTIAL mode is chosen, otherwise, PARALLEL mode is chosen. + */ + AUTO +} diff --git a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/Controller.java b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/Controller.java index 07730de45e7a..fe010b219e49 100644 --- a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/Controller.java +++ b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/Controller.java @@ -27,7 +27,7 @@ import org.apache.druid.msq.counters.CounterSnapshotsTree; import org.apache.druid.msq.indexing.MSQControllerTask; import org.apache.druid.msq.indexing.error.MSQErrorReport; -import org.apache.druid.msq.statistics.ClusterByStatisticsSnapshot; +import org.apache.druid.msq.statistics.PartialKeyStatisticsInformation; import javax.annotation.Nullable; import java.util.List; @@ -81,9 +81,11 @@ public String getId() // Worker-to-controller messages /** - * Provide a {@link ClusterByStatisticsSnapshot} for shuffling stages. + * Accepts a {@link PartialKeyStatisticsInformation} and updates the controller key statistics information. If all key + * statistics have been gathered, enqueues the task with the {@link WorkerSketchFetcher} to generate partiton boundaries. + * This is intended to be called by the {@link org.apache.druid.msq.indexing.ControllerChatHandler}. */ - void updateStatus(int stageNumber, int workerNumber, Object keyStatisticsObject); + void updatePartialKeyStatisticsInformation(int stageNumber, int workerNumber, Object partialKeyStatisticsInformationObject); /** * System error reported by a subtask. Note that the errors are organized by diff --git a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/ControllerClient.java b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/ControllerClient.java index faf1c3ff5e98..9d197b76dd17 100644 --- a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/ControllerClient.java +++ b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/ControllerClient.java @@ -22,7 +22,7 @@ import org.apache.druid.msq.counters.CounterSnapshotsTree; import org.apache.druid.msq.indexing.error.MSQErrorReport; import org.apache.druid.msq.kernel.StageId; -import org.apache.druid.msq.statistics.ClusterByStatisticsSnapshot; +import org.apache.druid.msq.statistics.PartialKeyStatisticsInformation; import javax.annotation.Nullable; import java.io.IOException; @@ -34,13 +34,13 @@ public interface ControllerClient extends AutoCloseable { /** - * Client side method to update the controller with key statistics for a particular stage and worker. - * Controller's implementation collates all the key statistics for a stage to generate the partition boundaries. + * Client side method to update the controller with partial key statistics information for a particular stage and worker. + * Controller's implementation collates all the information for a stage to fetch key statistics from workers. */ - void postKeyStatistics( + void postPartialKeyStatistics( StageId stageId, int workerNumber, - ClusterByStatisticsSnapshot keyStatistics + PartialKeyStatisticsInformation partialKeyStatisticsInformation ) throws IOException; /** diff --git a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/ControllerImpl.java b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/ControllerImpl.java index a4812292fb15..cafc0f389258 100644 --- a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/ControllerImpl.java +++ b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/ControllerImpl.java @@ -64,6 +64,7 @@ import org.apache.druid.indexing.overlord.SegmentPublishResult; import org.apache.druid.indexing.overlord.Segments; import org.apache.druid.java.util.common.DateTimes; +import org.apache.druid.java.util.common.Either; import org.apache.druid.java.util.common.IAE; import org.apache.druid.java.util.common.ISE; import org.apache.druid.java.util.common.Intervals; @@ -107,6 +108,7 @@ import org.apache.druid.msq.indexing.error.MSQWarningReportLimiterPublisher; import org.apache.druid.msq.indexing.error.MSQWarnings; import org.apache.druid.msq.indexing.error.QueryNotSupportedFault; +import org.apache.druid.msq.indexing.error.TooManyPartitionsFault; import org.apache.druid.msq.indexing.error.TooManyWarningsFault; import org.apache.druid.msq.indexing.error.UnknownFault; import org.apache.druid.msq.indexing.report.MSQResultsReport; @@ -149,7 +151,8 @@ import org.apache.druid.msq.shuffle.DurableStorageInputChannelFactory; import org.apache.druid.msq.shuffle.DurableStorageUtils; import org.apache.druid.msq.shuffle.WorkerInputChannelFactory; -import org.apache.druid.msq.statistics.ClusterByStatisticsSnapshot; +import org.apache.druid.msq.statistics.CompleteKeyStatisticsInformation; +import org.apache.druid.msq.statistics.PartialKeyStatisticsInformation; import org.apache.druid.msq.util.DimensionSchemaUtils; import org.apache.druid.msq.util.IntervalUtils; import org.apache.druid.msq.util.MSQFutureUtils; @@ -201,6 +204,7 @@ import java.util.Set; import java.util.concurrent.ArrayBlockingQueue; import java.util.concurrent.BlockingQueue; +import java.util.concurrent.CompletableFuture; import java.util.concurrent.ConcurrentHashMap; import java.util.concurrent.ConcurrentLinkedQueue; import java.util.concurrent.ThreadLocalRandom; @@ -259,6 +263,7 @@ public class ControllerImpl implements Controller // For live reports. Written by the main controller thread, read by HTTP threads. private final ConcurrentHashMap stagePartitionCountsForLiveReports = new ConcurrentHashMap<>(); + private WorkerSketchFetcher workerSketchFetcher; // Time at which the query started. // For live reports. Written by the main controller thread, read by HTTP threads. private volatile DateTime queryStartTime = null; @@ -519,6 +524,15 @@ private QueryDefinition initializeQueryDefAndState(final Closer closer) context.registerController(this, closer); this.netClient = new ExceptionWrappingWorkerClient(context.taskClientFor(this)); + ClusterStatisticsMergeMode clusterStatisticsMergeMode = + MultiStageQueryContext.getClusterStatisticsMergeMode(task.getQuerySpec().getQuery().context()); + + log.debug("Query [%s] cluster statistics merge mode is set to %s.", id(), clusterStatisticsMergeMode); + + int statisticsMaxRetainedBytes = WorkerMemoryParameters.createProductionInstanceForController(context.injector()) + .getPartitionStatisticsMaxRetainedBytes(); + this.workerSketchFetcher = new WorkerSketchFetcher(netClient, clusterStatisticsMergeMode, statisticsMaxRetainedBytes); + closer.register(netClient::close); final boolean isDurableStorageEnabled = @@ -565,10 +579,12 @@ private QueryDefinition initializeQueryDefAndState(final Closer closer) } /** - * Provide a {@link ClusterByStatisticsSnapshot} for shuffling stages. + * Accepts a {@link PartialKeyStatisticsInformation} and updates the controller key statistics information. If all key + * statistics information has been gathered, enqueues the task with the {@link WorkerSketchFetcher} to generate + * partiton boundaries. This is intended to be called by the {@link org.apache.druid.msq.indexing.ControllerChatHandler}. */ @Override - public void updateStatus(int stageNumber, int workerNumber, Object keyStatisticsObject) + public void updatePartialKeyStatisticsInformation(int stageNumber, int workerNumber, Object partialKeyStatisticsInformationObject) { addToKernelManipulationQueue( queryKernel -> { @@ -582,9 +598,9 @@ public void updateStatus(int stageNumber, int workerNumber, Object keyStatistics stageDef.getShuffleSpec().get().doesAggregateByClusterKey() ); - final ClusterByStatisticsSnapshot keyStatistics; + final PartialKeyStatisticsInformation partialKeyStatisticsInformation; try { - keyStatistics = mapper.convertValue(keyStatisticsObject, ClusterByStatisticsSnapshot.class); + partialKeyStatisticsInformation = mapper.convertValue(partialKeyStatisticsInformationObject, PartialKeyStatisticsInformation.class); } catch (IllegalArgumentException e) { throw new IAE( @@ -595,7 +611,36 @@ public void updateStatus(int stageNumber, int workerNumber, Object keyStatistics ); } - queryKernel.addResultKeyStatisticsForStageAndWorker(stageId, workerNumber, keyStatistics); + queryKernel.addPartialKeyStatisticsForStageAndWorker(stageId, workerNumber, partialKeyStatisticsInformation); + + if (queryKernel.getStagePhase(stageId).equals(ControllerStagePhase.MERGING_STATISTICS)) { + List workerTaskIds = workerTaskLauncher.getTaskList(); + CompleteKeyStatisticsInformation completeKeyStatisticsInformation = + queryKernel.getCompleteKeyStatisticsInformation(stageId); + + // Queue the sketch fetching task into the worker sketch fetcher. + CompletableFuture> clusterByPartitionsCompletableFuture = + workerSketchFetcher.submitFetcherTask( + completeKeyStatisticsInformation, + workerTaskIds, + stageDef + ); + + // Add the listener to handle completion. + clusterByPartitionsCompletableFuture.whenComplete((clusterByPartitionsEither, throwable) -> { + addToKernelManipulationQueue(holder -> { + if (throwable != null) { + holder.failStageForReason(stageId, UnknownFault.forException(throwable)); + } else if (clusterByPartitionsEither.isError()) { + holder.failStageForReason(stageId, new TooManyPartitionsFault(stageDef.getMaxPartitionCount())); + } else { + log.debug("Query [%s] Partition boundaries generated for stage %s", id(), stageId); + holder.setClusterByPartitionBoundaries(stageId, clusterByPartitionsEither.valueOrThrow()); + } + holder.transitionStageKernel(stageId, queryKernel.getStagePhase(stageId)); + }); + }); + } } ); } @@ -1959,11 +2004,7 @@ public RunQueryUntilDone( this.queryDef = queryDef; this.inputSpecSlicerFactory = inputSpecSlicerFactory; this.closer = closer; - this.queryKernel = new ControllerQueryKernel( - queryDef, - WorkerMemoryParameters.createProductionInstanceForController(context.injector()) - .getPartitionStatisticsMaxRetainedBytes() - ); + this.queryKernel = new ControllerQueryKernel(queryDef); } /** diff --git a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/ExceptionWrappingWorkerClient.java b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/ExceptionWrappingWorkerClient.java index 1c3cc39987c7..3d78b7c9ced4 100644 --- a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/ExceptionWrappingWorkerClient.java +++ b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/ExceptionWrappingWorkerClient.java @@ -31,6 +31,7 @@ import org.apache.druid.msq.indexing.error.WorkerRpcFailedFault; import org.apache.druid.msq.kernel.StageId; import org.apache.druid.msq.kernel.WorkOrder; +import org.apache.druid.msq.statistics.ClusterByStatisticsSnapshot; import javax.annotation.Nullable; import java.io.IOException; @@ -55,6 +56,23 @@ public ListenableFuture postWorkOrder(String workerTaskId, WorkOrder workO return wrap(workerTaskId, client, c -> c.postWorkOrder(workerTaskId, workOrder)); } + @Override + public ListenableFuture fetchClusterByStatisticsSnapshot(String workerTaskId, String queryId, int stageNumber) + { + return client.fetchClusterByStatisticsSnapshot(workerTaskId, queryId, stageNumber); + } + + @Override + public ListenableFuture fetchClusterByStatisticsSnapshotForTimeChunk( + String workerTaskId, + String queryId, + int stageNumber, + long timeChunk + ) + { + return client.fetchClusterByStatisticsSnapshotForTimeChunk(workerTaskId, queryId, stageNumber, timeChunk); + } + @Override public ListenableFuture postResultPartitionBoundaries( final String workerTaskId, diff --git a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/Worker.java b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/Worker.java index f069c91e145a..cc5f0fae1732 100644 --- a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/Worker.java +++ b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/Worker.java @@ -25,6 +25,7 @@ import org.apache.druid.msq.indexing.MSQWorkerTask; import org.apache.druid.msq.kernel.StageId; import org.apache.druid.msq.kernel.WorkOrder; +import org.apache.druid.msq.statistics.ClusterByStatisticsSnapshot; import javax.annotation.Nullable; import java.io.IOException; @@ -67,6 +68,18 @@ public interface Worker */ void postWorkOrder(WorkOrder workOrder); + /** + * Returns the statistics snapshot for the given stageId. This is called from {@link WorkerSketchFetcher} under + * PARALLEL OR AUTO modes. + */ + ClusterByStatisticsSnapshot fetchStatisticsSnapshot(StageId stageId); + + /** + * Returns the statistics snapshot for the given stageId which contains only the sketch for the specified timeChunk. + * This is called from {@link WorkerSketchFetcher} under SEQUENTIAL OR AUTO modes. + */ + ClusterByStatisticsSnapshot fetchStatisticsSnapshotForTimeChunk(StageId stageId, long timeChunk); + /** * Called when the worker chat handler recieves the result partition boundaries for a particular stageNumber * and queryId diff --git a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/WorkerClient.java b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/WorkerClient.java index 8f4c7bac239a..5c02a79f89a3 100644 --- a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/WorkerClient.java +++ b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/WorkerClient.java @@ -25,6 +25,7 @@ import org.apache.druid.msq.counters.CounterSnapshotsTree; import org.apache.druid.msq.kernel.StageId; import org.apache.druid.msq.kernel.WorkOrder; +import org.apache.druid.msq.statistics.ClusterByStatisticsSnapshot; import java.io.IOException; @@ -38,6 +39,27 @@ public interface WorkerClient extends AutoCloseable */ ListenableFuture postWorkOrder(String workerTaskId, WorkOrder workOrder); + /** + * Fetches the {@link ClusterByStatisticsSnapshot} from a worker. This is intended to be used by the + * {@link WorkerSketchFetcher} under PARALLEL or AUTO modes. + */ + ListenableFuture fetchClusterByStatisticsSnapshot( + String workerTaskId, + String queryId, + int stageNumber + ); + + /** + * Fetches a {@link ClusterByStatisticsSnapshot} which contains only the sketch of the specified timeChunk. + * This is intended to be used by the {@link WorkerSketchFetcher} under SEQUENTIAL or AUTO modes. + */ + ListenableFuture fetchClusterByStatisticsSnapshotForTimeChunk( + String workerTaskId, + String queryId, + int stageNumber, + long timeChunk + ); + /** * Worker's client method to inform it of the partition boundaries for the given stage. This is usually invoked by the * controller after collating the result statistics from all the workers processing the query diff --git a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/WorkerImpl.java b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/WorkerImpl.java index 5b68041d0e97..49d6f9080d7c 100644 --- a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/WorkerImpl.java +++ b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/WorkerImpl.java @@ -106,6 +106,7 @@ import org.apache.druid.msq.shuffle.WorkerInputChannelFactory; import org.apache.druid.msq.statistics.ClusterByStatisticsCollector; import org.apache.druid.msq.statistics.ClusterByStatisticsSnapshot; +import org.apache.druid.msq.statistics.PartialKeyStatisticsInformation; import org.apache.druid.msq.util.DecoratedExecutorService; import org.apache.druid.msq.util.MultiStageQueryContext; import org.apache.druid.query.PrioritizedCallable; @@ -159,6 +160,7 @@ public class WorkerImpl implements Worker private final BlockingQueue> kernelManipulationQueue = new LinkedBlockingDeque<>(); private final ConcurrentHashMap> stageOutputs = new ConcurrentHashMap<>(); private final ConcurrentHashMap stageCounters = new ConcurrentHashMap<>(); + private final ConcurrentHashMap stageKernelMap = new ConcurrentHashMap<>(); private final boolean durableStageStorageEnabled; /** @@ -365,10 +367,14 @@ public Optional runTask(final Closer closer) throws Exception if (kernel.getPhase() == WorkerStagePhase.READING_INPUT && kernel.hasResultKeyStatisticsSnapshot()) { if (controllerAlive) { - controllerClient.postKeyStatistics( + PartialKeyStatisticsInformation partialKeyStatisticsInformation = + kernel.getResultKeyStatisticsSnapshot() + .partialKeyStatistics(); + + controllerClient.postPartialKeyStatistics( stageDefinition.getId(), kernel.getWorkOrder().getWorkerNumber(), - kernel.getResultKeyStatisticsSnapshot() + partialKeyStatisticsInformation ); } kernel.startPreshuffleWaitingForResultPartitionBoundaries(); @@ -562,6 +568,19 @@ public void postFinish() kernelManipulationQueue.add(KernelHolder::setDone); } + @Override + public ClusterByStatisticsSnapshot fetchStatisticsSnapshot(StageId stageId) + { + return stageKernelMap.get(stageId).getResultKeyStatisticsSnapshot(); + } + + @Override + public ClusterByStatisticsSnapshot fetchStatisticsSnapshotForTimeChunk(StageId stageId, long timeChunk) + { + ClusterByStatisticsSnapshot snapshot = stageKernelMap.get(stageId).getResultKeyStatisticsSnapshot(); + return snapshot.getSnapshotForTimeChunk(timeChunk); + } + @Override public CounterSnapshotsTree getCounters() { @@ -1273,9 +1292,8 @@ public ReadableFrameChannel openChannel(StageId stageId, int workerNumber, int p } } - private static class KernelHolder + private class KernelHolder { - private final Map stageKernelMap = new HashMap<>(); private boolean done = false; public Map getStageKernelMap() diff --git a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/WorkerSketchFetcher.java b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/WorkerSketchFetcher.java new file mode 100644 index 000000000000..3482b50daaff --- /dev/null +++ b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/WorkerSketchFetcher.java @@ -0,0 +1,340 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.msq.exec; + +import com.google.common.util.concurrent.ListenableFuture; +import org.apache.druid.frame.key.ClusterBy; +import org.apache.druid.frame.key.ClusterByPartition; +import org.apache.druid.frame.key.ClusterByPartitions; +import org.apache.druid.java.util.common.Either; +import org.apache.druid.java.util.common.ISE; +import org.apache.druid.java.util.common.logger.Logger; +import org.apache.druid.msq.kernel.StageDefinition; +import org.apache.druid.msq.statistics.ClusterByStatisticsCollector; +import org.apache.druid.msq.statistics.ClusterByStatisticsSnapshot; +import org.apache.druid.msq.statistics.CompleteKeyStatisticsInformation; + +import java.util.ArrayList; +import java.util.HashSet; +import java.util.Iterator; +import java.util.List; +import java.util.Map; +import java.util.Set; +import java.util.concurrent.CompletableFuture; +import java.util.concurrent.ExecutorService; +import java.util.concurrent.Executors; +import java.util.stream.IntStream; + +/** + * Queues up fetching sketches from workers and progressively generates partitions boundaries. + */ +public class WorkerSketchFetcher +{ + private static final Logger log = new Logger(WorkerSketchFetcher.class); + private static final int DEFAULT_THREAD_COUNT = 4; + // If the combined size of worker sketches is more than this threshold, SEQUENTIAL merging mode is used. + static final long BYTES_THRESHOLD = 1_000_000_000L; + // If there are more workers than this threshold, SEQUENTIAL merging mode is used. + static final long WORKER_THRESHOLD = 100; + + private final ClusterStatisticsMergeMode clusterStatisticsMergeMode; + private final int statisticsMaxRetainedBytes; + private final WorkerClient workerClient; + private final ExecutorService executorService; + + public WorkerSketchFetcher(WorkerClient workerClient, ClusterStatisticsMergeMode clusterStatisticsMergeMode, int statisticsMaxRetainedBytes) + { + this.workerClient = workerClient; + this.clusterStatisticsMergeMode = clusterStatisticsMergeMode; + this.executorService = Executors.newFixedThreadPool(DEFAULT_THREAD_COUNT); + this.statisticsMaxRetainedBytes = statisticsMaxRetainedBytes; + } + + /** + * Submits a request to fetch and generate partitions for the given worker statistics and returns a future for it. It + * decides based on the statistics if it should fetch sketches one by one or together. + */ + public CompletableFuture> submitFetcherTask( + CompleteKeyStatisticsInformation completeKeyStatisticsInformation, + List workerTaskIds, + StageDefinition stageDefinition + ) + { + ClusterBy clusterBy = stageDefinition.getClusterBy(); + + switch (clusterStatisticsMergeMode) { + case SEQUENTIAL: + return sequentialTimeChunkMerging(completeKeyStatisticsInformation, stageDefinition, workerTaskIds); + case PARALLEL: + return inMemoryFullSketchMerging(stageDefinition, workerTaskIds); + case AUTO: + if (clusterBy.getBucketByCount() == 0) { + log.debug("Query [%s] AUTO mode: chose PARALLEL mode to merge key statistics", stageDefinition.getId().getQueryId()); + // If there is no time clustering, there is no scope for sequential merge + return inMemoryFullSketchMerging(stageDefinition, workerTaskIds); + } else if (stageDefinition.getMaxWorkerCount() > WORKER_THRESHOLD || completeKeyStatisticsInformation.getBytesRetained() > BYTES_THRESHOLD) { + log.debug("Query [%s] AUTO mode: chose SEQUENTIAL mode to merge key statistics", stageDefinition.getId().getQueryId()); + return sequentialTimeChunkMerging(completeKeyStatisticsInformation, stageDefinition, workerTaskIds); + } + log.debug("Query [%s] AUTO mode: chose PARALLEL mode to merge key statistics", stageDefinition.getId().getQueryId()); + return inMemoryFullSketchMerging(stageDefinition, workerTaskIds); + default: + throw new IllegalStateException("No fetching strategy found for mode: " + clusterStatisticsMergeMode); + } + } + + /** + * Fetches the full {@link ClusterByStatisticsCollector} from all workers and generates partition boundaries from them. + * This is faster than fetching them timechunk by timechunk but the collector will be downsampled till it can fit + * on the controller, resulting in less accurate partition boundries. + */ + CompletableFuture> inMemoryFullSketchMerging( + StageDefinition stageDefinition, + List workerTaskIds + ) + { + CompletableFuture> partitionFuture = new CompletableFuture<>(); + + // Create a new key statistics collector to merge worker sketches into + final ClusterByStatisticsCollector mergedStatisticsCollector = + stageDefinition.createResultKeyStatisticsCollector(statisticsMaxRetainedBytes); + final int workerCount = workerTaskIds.size(); + // Guarded by synchronized mergedStatisticsCollector + final Set finishedWorkers = new HashSet<>(); + + // Submit a task for each worker to fetch statistics + IntStream.range(0, workerCount).forEach(workerNo -> { + executorService.submit(() -> { + ListenableFuture snapshotFuture = + workerClient.fetchClusterByStatisticsSnapshot( + workerTaskIds.get(workerNo), + stageDefinition.getId().getQueryId(), + stageDefinition.getStageNumber() + ); + partitionFuture.whenComplete((result, exception) -> { + if (exception != null || (result != null && result.isError())) { + snapshotFuture.cancel(true); + } + }); + + try { + ClusterByStatisticsSnapshot clusterByStatisticsSnapshot = snapshotFuture.get(); + if (clusterByStatisticsSnapshot == null) { + throw new ISE("Worker %s returned null sketch, this should never happen", workerNo); + } + synchronized (mergedStatisticsCollector) { + mergedStatisticsCollector.addAll(clusterByStatisticsSnapshot); + finishedWorkers.add(workerNo); + + if (finishedWorkers.size() == workerCount) { + log.debug("Query [%s] Received all statistics, generating partitions", stageDefinition.getId().getQueryId()); + partitionFuture.complete(stageDefinition.generatePartitionsForShuffle(mergedStatisticsCollector)); + } + } + } + catch (Exception e) { + synchronized (mergedStatisticsCollector) { + partitionFuture.completeExceptionally(e); + mergedStatisticsCollector.clear(); + } + } + }); + }); + return partitionFuture; + } + + /** + * Fetches cluster statistics from all workers and generates partition boundaries from them one time chunk at a time. + * This takes longer due to the overhead of fetching sketches, however, this prevents any loss in accuracy from + * downsampling on the controller. + */ + CompletableFuture> sequentialTimeChunkMerging( + CompleteKeyStatisticsInformation completeKeyStatisticsInformation, + StageDefinition stageDefinition, + List workerTaskIds + ) + { + SequentialFetchStage sequentialFetchStage = new SequentialFetchStage( + stageDefinition, + workerTaskIds, + completeKeyStatisticsInformation.getTimeSegmentVsWorkerMap().entrySet().iterator() + ); + sequentialFetchStage.submitFetchingTasksForNextTimeChunk(); + return sequentialFetchStage.getPartitionFuture(); + } + + private class SequentialFetchStage + { + private final StageDefinition stageDefinition; + private final List workerTaskIds; + private final Iterator>> timeSegmentVsWorkerIdIterator; + private final CompletableFuture> partitionFuture; + // Final sorted list of partition boundaries. This is appended to after statistics for each time chunk are gathered. + private final List finalPartitionBoundries; + + public SequentialFetchStage( + StageDefinition stageDefinition, + List workerTaskIds, + Iterator>> timeSegmentVsWorkerIdIterator + ) + { + this.finalPartitionBoundries = new ArrayList<>(); + this.stageDefinition = stageDefinition; + this.workerTaskIds = workerTaskIds; + this.timeSegmentVsWorkerIdIterator = timeSegmentVsWorkerIdIterator; + this.partitionFuture = new CompletableFuture<>(); + } + + /** + * Submits the tasks to fetch key statistics for the time chunk pointed to by {@link #timeSegmentVsWorkerIdIterator}. + * Once the statistics have been gathered from all workers which have them, generates partitions and adds it to + * {@link #finalPartitionBoundries}, stiching the partitions between time chunks using + * {@link #abutAndAppendPartitionBoundries(List, List)} to make them continuous. + * + * The time chunks returned by {@link #timeSegmentVsWorkerIdIterator} should be in ascending order for the partitions + * to be generated correctly. + * + * If {@link #timeSegmentVsWorkerIdIterator} doesn't have any more values, assumes that partition boundaries have + * been successfully generated and completes {@link #partitionFuture} with the result. + * + * Completes the future with an error as soon as the number of partitions exceed max partition count for the stage + * definition. + */ + public void submitFetchingTasksForNextTimeChunk() + { + if (!timeSegmentVsWorkerIdIterator.hasNext()) { + partitionFuture.complete(Either.value(new ClusterByPartitions(finalPartitionBoundries))); + } else { + Map.Entry> entry = timeSegmentVsWorkerIdIterator.next(); + // Time chunk for which partition boundries are going to be generated for + Long timeChunk = entry.getKey(); + Set workerIdsWithTimeChunk = entry.getValue(); + // Create a new key statistics collector to merge worker sketches into + ClusterByStatisticsCollector mergedStatisticsCollector = + stageDefinition.createResultKeyStatisticsCollector(statisticsMaxRetainedBytes); + // Guarded by synchronized mergedStatisticsCollector + Set finishedWorkers = new HashSet<>(); + + log.debug("Query [%s]. Submitting request for statistics for time chunk %s to %s workers", + stageDefinition.getId().getQueryId(), + timeChunk, + workerIdsWithTimeChunk.size()); + + // Submits a task for every worker which has a certain time chunk + for (int workerNo : workerIdsWithTimeChunk) { + executorService.submit(() -> { + ListenableFuture snapshotFuture = + workerClient.fetchClusterByStatisticsSnapshotForTimeChunk( + workerTaskIds.get(workerNo), + stageDefinition.getId().getQueryId(), + stageDefinition.getStageNumber(), + timeChunk + ); + partitionFuture.whenComplete((result, exception) -> { + if (exception != null || (result != null && result.isError())) { + snapshotFuture.cancel(true); + } + }); + + try { + ClusterByStatisticsSnapshot snapshotForTimeChunk = snapshotFuture.get(); + if (snapshotForTimeChunk == null) { + throw new ISE("Worker %s returned null sketch for %s, this should never happen", workerNo, timeChunk); + } + synchronized (mergedStatisticsCollector) { + mergedStatisticsCollector.addAll(snapshotForTimeChunk); + finishedWorkers.add(workerNo); + + if (finishedWorkers.size() == workerIdsWithTimeChunk.size()) { + Either longClusterByPartitionsEither = + stageDefinition.generatePartitionsForShuffle(mergedStatisticsCollector); + + log.debug("Query [%s]. Received all statistics for time chunk %s, generating partitions", + stageDefinition.getId().getQueryId(), + timeChunk); + + long totalPartitionCount = finalPartitionBoundries.size() + getPartitionCountFromEither(longClusterByPartitionsEither); + if (totalPartitionCount > stageDefinition.getMaxPartitionCount()) { + // Fail fast if more partitions than the maximum have been reached. + partitionFuture.complete(Either.error(totalPartitionCount)); + mergedStatisticsCollector.clear(); + } else { + List timeSketchPartitions = longClusterByPartitionsEither.valueOrThrow().ranges(); + abutAndAppendPartitionBoundries(finalPartitionBoundries, timeSketchPartitions); + log.debug("Query [%s]. Finished generating partitions for time chunk %s, total count so far %s", + stageDefinition.getId().getQueryId(), + timeChunk, + finalPartitionBoundries.size()); + submitFetchingTasksForNextTimeChunk(); + } + } + } + } + catch (Exception e) { + synchronized (mergedStatisticsCollector) { + partitionFuture.completeExceptionally(e); + mergedStatisticsCollector.clear(); + } + } + }); + } + } + } + + /** + * Takes a list of sorted {@link ClusterByPartitions} {@param timeSketchPartitions} and adds it to a sorted list + * {@param finalPartitionBoundries}. If {@param finalPartitionBoundries} is not empty, the end time of the last + * partition of {@param finalPartitionBoundries} is changed to abut with the starting time of the first partition + * of {@param timeSketchPartitions}. + * + * This is used to make the partitions generated continuous. + */ + private void abutAndAppendPartitionBoundries( + List finalPartitionBoundries, + List timeSketchPartitions + ) + { + if (!finalPartitionBoundries.isEmpty()) { + // Stitch up the end time of the last partition with the start time of the first partition. + ClusterByPartition clusterByPartition = finalPartitionBoundries.remove(finalPartitionBoundries.size() - 1); + finalPartitionBoundries.add(new ClusterByPartition(clusterByPartition.getStart(), timeSketchPartitions.get(0).getStart())); + } + finalPartitionBoundries.addAll(timeSketchPartitions); + } + + public CompletableFuture> getPartitionFuture() + { + return partitionFuture; + } + } + + /** + * Gets the partition size from an {@link Either}. If it is an error, the long denotes the number of partitions + * (in the case of creating too many partitions), otherwise checks the size of the list. + */ + private static long getPartitionCountFromEither(Either either) + { + if (either.isError()) { + return either.error(); + } else { + return either.valueOrThrow().size(); + } + } +} diff --git a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/indexing/ControllerChatHandler.java b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/indexing/ControllerChatHandler.java index 1857d83708c7..8fa04ce6d90b 100644 --- a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/indexing/ControllerChatHandler.java +++ b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/indexing/ControllerChatHandler.java @@ -26,7 +26,8 @@ import org.apache.druid.msq.exec.Controller; import org.apache.druid.msq.exec.ControllerClient; import org.apache.druid.msq.indexing.error.MSQErrorReport; -import org.apache.druid.msq.statistics.ClusterByStatisticsSnapshot; +import org.apache.druid.msq.kernel.StageId; +import org.apache.druid.msq.statistics.PartialKeyStatisticsInformation; import org.apache.druid.segment.realtime.firehose.ChatHandler; import org.apache.druid.segment.realtime.firehose.ChatHandlers; import org.apache.druid.server.security.Action; @@ -58,16 +59,17 @@ public ControllerChatHandler(TaskToolbox toolbox, Controller controller) } /** - * Used by subtasks to post {@link ClusterByStatisticsSnapshot} for shuffling stages. + * Used by subtasks to post {@link PartialKeyStatisticsInformation} for shuffling stages. * - * See {@link ControllerClient#postKeyStatistics} for the client-side code that calls this API. + * See {@link ControllerClient#postPartialKeyStatistics(StageId, int, PartialKeyStatisticsInformation)} + * for the client-side code that calls this API. */ @POST - @Path("/keyStatistics/{queryId}/{stageNumber}/{workerNumber}") + @Path("/partialKeyStatisticsInformation/{queryId}/{stageNumber}/{workerNumber}") @Produces(MediaType.APPLICATION_JSON) @Consumes(MediaType.APPLICATION_JSON) - public Response httpPostKeyStatistics( - final Object keyStatisticsObject, + public Response httpPostPartialKeyStatistics( + final Object partialKeyStatisticsObject, @PathParam("queryId") final String queryId, @PathParam("stageNumber") final int stageNumber, @PathParam("workerNumber") final int workerNumber, @@ -75,7 +77,7 @@ public Response httpPostKeyStatistics( ) { ChatHandlers.authorizationCheck(req, Action.WRITE, task.getDataSource(), toolbox.getAuthorizerMapper()); - controller.updateStatus(stageNumber, workerNumber, keyStatisticsObject); + controller.updatePartialKeyStatisticsInformation(stageNumber, workerNumber, partialKeyStatisticsObject); return Response.status(Response.Status.ACCEPTED).build(); } diff --git a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/indexing/IndexerControllerClient.java b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/indexing/IndexerControllerClient.java index 3a6c9e78797b..1ff5952a8958 100644 --- a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/indexing/IndexerControllerClient.java +++ b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/indexing/IndexerControllerClient.java @@ -29,7 +29,7 @@ import org.apache.druid.msq.exec.ControllerClient; import org.apache.druid.msq.indexing.error.MSQErrorReport; import org.apache.druid.msq.kernel.StageId; -import org.apache.druid.msq.statistics.ClusterByStatisticsSnapshot; +import org.apache.druid.msq.statistics.PartialKeyStatisticsInformation; import org.apache.druid.rpc.IgnoreHttpResponseHandler; import org.apache.druid.rpc.RequestBuilder; import org.apache.druid.rpc.ServiceClient; @@ -59,14 +59,14 @@ public IndexerControllerClient( } @Override - public void postKeyStatistics( + public void postPartialKeyStatistics( StageId stageId, int workerNumber, - ClusterByStatisticsSnapshot keyStatistics + PartialKeyStatisticsInformation partialKeyStatisticsInformation ) throws IOException { final String path = StringUtils.format( - "/keyStatistics/%s/%s/%d", + "/partialKeyStatisticsInformation/%s/%d/%d", StringUtils.urlEncode(stageId.getQueryId()), stageId.getStageNumber(), workerNumber @@ -74,7 +74,7 @@ public void postKeyStatistics( doRequest( new RequestBuilder(HttpMethod.POST, path) - .jsonContent(jsonMapper, keyStatistics), + .jsonContent(jsonMapper, partialKeyStatisticsInformation), IgnoreHttpResponseHandler.INSTANCE ); } diff --git a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/indexing/IndexerWorkerClient.java b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/indexing/IndexerWorkerClient.java index 430c264291f8..980c7f97bee5 100644 --- a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/indexing/IndexerWorkerClient.java +++ b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/indexing/IndexerWorkerClient.java @@ -41,6 +41,7 @@ import org.apache.druid.msq.exec.WorkerClient; import org.apache.druid.msq.kernel.StageId; import org.apache.druid.msq.kernel.WorkOrder; +import org.apache.druid.msq.statistics.ClusterByStatisticsSnapshot; import org.apache.druid.rpc.IgnoreHttpResponseHandler; import org.apache.druid.rpc.RequestBuilder; import org.apache.druid.rpc.ServiceClient; @@ -103,6 +104,48 @@ public ListenableFuture postWorkOrder(String workerTaskId, WorkOrder workO ); } + @Override + public ListenableFuture fetchClusterByStatisticsSnapshot( + String workerTaskId, + String queryId, + int stageNumber + ) + { + String path = StringUtils.format("/keyStatistics/%s/%d", + StringUtils.urlEncode(queryId), + stageNumber); + + return FutureUtils.transform( + getClient(workerTaskId).asyncRequest( + new RequestBuilder(HttpMethod.POST, path), + new BytesFullResponseHandler() + ), + holder -> deserialize(holder, new TypeReference() {}) + ); + } + + @Override + public ListenableFuture fetchClusterByStatisticsSnapshotForTimeChunk( + String workerTaskId, + String queryId, + int stageNumber, + long timeChunk + ) + { + String path = StringUtils.format("/keyStatisticsForTimeChunk/%s/%d/%d", + StringUtils.urlEncode(queryId), + stageNumber, + timeChunk); + + return FutureUtils.transform( + getClient(workerTaskId).asyncRequest( + new RequestBuilder(HttpMethod.POST, path), + new BytesFullResponseHandler() + ), + holder -> deserialize(holder, new TypeReference() {}) + ); + } + @Override public ListenableFuture postResultPartitionBoundaries( String workerTaskId, diff --git a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/indexing/WorkerChatHandler.java b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/indexing/WorkerChatHandler.java index 06fd8f36caf7..dd6ea7cb7124 100644 --- a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/indexing/WorkerChatHandler.java +++ b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/indexing/WorkerChatHandler.java @@ -28,6 +28,7 @@ import org.apache.druid.msq.exec.Worker; import org.apache.druid.msq.kernel.StageId; import org.apache.druid.msq.kernel.WorkOrder; +import org.apache.druid.msq.statistics.ClusterByStatisticsSnapshot; import org.apache.druid.segment.realtime.firehose.ChatHandler; import org.apache.druid.segment.realtime.firehose.ChatHandlers; import org.apache.druid.server.security.Action; @@ -179,6 +180,45 @@ public Response httpPostResultPartitionBoundaries( } } + @POST + @Path("/keyStatistics/{queryId}/{stageNumber}") + @Produces(MediaType.APPLICATION_JSON) + @Consumes(MediaType.APPLICATION_JSON) + public Response httpFetchKeyStatistics( + @PathParam("queryId") final String queryId, + @PathParam("stageNumber") final int stageNumber, + @Context final HttpServletRequest req + ) + { + ChatHandlers.authorizationCheck(req, Action.READ, task.getDataSource(), toolbox.getAuthorizerMapper()); + ClusterByStatisticsSnapshot clusterByStatisticsSnapshot; + StageId stageId = new StageId(queryId, stageNumber); + clusterByStatisticsSnapshot = worker.fetchStatisticsSnapshot(stageId); + return Response.status(Response.Status.ACCEPTED) + .entity(clusterByStatisticsSnapshot) + .build(); + } + + @POST + @Path("/keyStatisticsForTimeChunk/{queryId}/{stageNumber}/{timeChunk}") + @Produces(MediaType.APPLICATION_JSON) + @Consumes(MediaType.APPLICATION_JSON) + public Response httpSketch( + @PathParam("queryId") final String queryId, + @PathParam("stageNumber") final int stageNumber, + @PathParam("timeChunk") final long timeChunk, + @Context final HttpServletRequest req + ) + { + ChatHandlers.authorizationCheck(req, Action.READ, task.getDataSource(), toolbox.getAuthorizerMapper()); + ClusterByStatisticsSnapshot snapshotForTimeChunk; + StageId stageId = new StageId(queryId, stageNumber); + snapshotForTimeChunk = worker.fetchStatisticsSnapshotForTimeChunk(stageId, timeChunk); + return Response.status(Response.Status.ACCEPTED) + .entity(snapshotForTimeChunk) + .build(); + } + /** * See {@link org.apache.druid.msq.exec.WorkerClient#postCleanupStage} for the client-side code that calls this API. */ diff --git a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/kernel/controller/ControllerQueryKernel.java b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/kernel/controller/ControllerQueryKernel.java index 9de5c692c990..d2e9f198f04b 100644 --- a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/kernel/controller/ControllerQueryKernel.java +++ b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/kernel/controller/ControllerQueryKernel.java @@ -41,7 +41,8 @@ import org.apache.druid.msq.kernel.StageId; import org.apache.druid.msq.kernel.WorkOrder; import org.apache.druid.msq.kernel.WorkerAssignmentStrategy; -import org.apache.druid.msq.statistics.ClusterByStatisticsSnapshot; +import org.apache.druid.msq.statistics.CompleteKeyStatisticsInformation; +import org.apache.druid.msq.statistics.PartialKeyStatisticsInformation; import javax.annotation.Nullable; import java.util.HashMap; @@ -65,7 +66,6 @@ public class ControllerQueryKernel { private final QueryDefinition queryDef; - private final int partitionStatisticsMaxRetainedBytes; /** * Stage ID -> tracker for that stage. An extension of the state of this kernel. @@ -107,10 +107,9 @@ public class ControllerQueryKernel */ private final Set effectivelyFinishedStages = new HashSet<>(); - public ControllerQueryKernel(final QueryDefinition queryDef, final int partitionStatisticsMaxRetainedBytes) + public ControllerQueryKernel(final QueryDefinition queryDef) { this.queryDef = queryDef; - this.partitionStatisticsMaxRetainedBytes = partitionStatisticsMaxRetainedBytes; this.inflowMap = ImmutableMap.copyOf(computeStageInflowMap(queryDef)); this.outflowMap = ImmutableMap.copyOf(computeStageOutflowMap(queryDef)); @@ -266,8 +265,7 @@ private void createNewKernels( stageDef, stageWorkerCountMap, slicer, - assignmentStrategy, - partitionStatisticsMaxRetainedBytes + assignmentStrategy ); stageTracker.put(nextStage, stageKernel); } @@ -334,6 +332,22 @@ public ClusterByPartitions getResultPartitionBoundariesForStage(final StageId st return getStageKernelOrThrow(stageId).getResultPartitionBoundaries(); } + /** + * Delegates call to {@link ControllerStageTracker#getCompleteKeyStatisticsInformation()} + */ + public CompleteKeyStatisticsInformation getCompleteKeyStatisticsInformation(final StageId stageId) + { + return getStageKernelOrThrow(stageId).getCompleteKeyStatisticsInformation(); + } + + /** + * Delegates call to {@link ControllerStageTracker#setClusterByPartitionBoundaries(ClusterByPartitions)} ()} + */ + public void setClusterByPartitionBoundaries(final StageId stageId, ClusterByPartitions clusterByPartitions) + { + getStageKernelOrThrow(stageId).setClusterByPartitionBoundaries(clusterByPartitions); + } + /** * Delegates call to {@link ControllerStageTracker#collectorEncounteredAnyMultiValueField()} */ @@ -390,22 +404,24 @@ public WorkerInputs getWorkerInputsForStage(final StageId stageId) } /** - * Delegates call to {@link ControllerStageTracker#addResultKeyStatisticsForWorker(int, ClusterByStatisticsSnapshot)}. + * Delegates call to {@link ControllerStageTracker#addPartialKeyStatisticsForWorker(int, PartialKeyStatisticsInformation)}. * If calling this causes transition for the stage kernel, then this gets registered in this query kernel */ - public void addResultKeyStatisticsForStageAndWorker( + public void addPartialKeyStatisticsForStageAndWorker( final StageId stageId, final int workerNumber, - final ClusterByStatisticsSnapshot snapshot + final PartialKeyStatisticsInformation partialKeyStatisticsInformation ) { - ControllerStagePhase newPhase = getStageKernelOrThrow(stageId).addResultKeyStatisticsForWorker( + ControllerStageTracker stageKernel = getStageKernelOrThrow(stageId); + ControllerStagePhase newPhase = stageKernel.addPartialKeyStatisticsForWorker( workerNumber, - snapshot + partialKeyStatisticsInformation ); - // If the phase is POST_READING or FAILED, that implies the kernel has transitioned. We need to account for that + // If the kernel phase has transitioned, we need to account for that. switch (newPhase) { + case MERGING_STATISTICS: case POST_READING: case FAILED: transitionStageKernel(stageId, newPhase); @@ -436,6 +452,12 @@ public MSQFault getFailureReasonForStage(final StageId stageId) return getStageKernelOrThrow(stageId).getFailureReason(); } + public void failStageForReason(final StageId stageId, MSQFault fault) + { + getStageKernelOrThrow(stageId).failForReason(fault); + transitionStageKernel(stageId, ControllerStagePhase.FAILED); + } + /** * Delegates call to {@link ControllerStageTracker#fail()} and registers this transition to FAILED in this query kernel */ diff --git a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/kernel/controller/ControllerStagePhase.java b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/kernel/controller/ControllerStagePhase.java index 6686292c1d85..896b345b4d22 100644 --- a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/kernel/controller/ControllerStagePhase.java +++ b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/kernel/controller/ControllerStagePhase.java @@ -48,6 +48,17 @@ public boolean canTransitionFrom(final ControllerStagePhase priorPhase) } }, + // Waiting to fetch key statistics in the background from the workers and incrementally generate partitions. + // This phase is only transitioned to once all partialKeyInformation are recieved from workers. + // Transitioning to this phase should also enqueue the task to fetch key statistics to WorkerSketchFetcher. + MERGING_STATISTICS { + @Override + public boolean canTransitionFrom(final ControllerStagePhase priorPhase) + { + return priorPhase == READING_INPUT; + } + }, + // Post the inputs have been read and mapped to frames, in the `POST_READING` stage, we pre-shuffle and determing the partition boundaries. // This step for a stage spits out the statistics of the data as a whole (and not just the individual records). This // phase is not required in non-pre shuffle contexts. @@ -55,7 +66,7 @@ public boolean canTransitionFrom(final ControllerStagePhase priorPhase) @Override public boolean canTransitionFrom(final ControllerStagePhase priorPhase) { - return priorPhase == READING_INPUT; + return priorPhase == MERGING_STATISTICS; } }, diff --git a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/kernel/controller/ControllerStageTracker.java b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/kernel/controller/ControllerStageTracker.java index 3ad01a513c80..bf53239335c5 100644 --- a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/kernel/controller/ControllerStageTracker.java +++ b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/kernel/controller/ControllerStageTracker.java @@ -28,6 +28,7 @@ import org.apache.druid.java.util.common.Either; import org.apache.druid.java.util.common.IAE; import org.apache.druid.java.util.common.ISE; +import org.apache.druid.msq.indexing.error.InsertTimeNullFault; import org.apache.druid.msq.indexing.error.MSQFault; import org.apache.druid.msq.indexing.error.TooManyPartitionsFault; import org.apache.druid.msq.indexing.error.UnknownFault; @@ -38,11 +39,12 @@ import org.apache.druid.msq.input.stage.StageInputSlice; import org.apache.druid.msq.kernel.StageDefinition; import org.apache.druid.msq.kernel.WorkerAssignmentStrategy; -import org.apache.druid.msq.statistics.ClusterByStatisticsCollector; -import org.apache.druid.msq.statistics.ClusterByStatisticsSnapshot; +import org.apache.druid.msq.statistics.CompleteKeyStatisticsInformation; +import org.apache.druid.msq.statistics.PartialKeyStatisticsInformation; import javax.annotation.Nullable; import java.util.List; +import java.util.TreeMap; /** * Controller-side state machine for each stage. Used by {@link ControllerQueryKernel} to form the overall state @@ -57,13 +59,13 @@ class ControllerStageTracker private final int workerCount; private final WorkerInputs workerInputs; - private final IntSet workersWithResultKeyStatistics = new IntAVLTreeSet(); + private final IntSet workersWithReportedKeyStatistics = new IntAVLTreeSet(); private final IntSet workersWithResultsComplete = new IntAVLTreeSet(); private ControllerStagePhase phase = ControllerStagePhase.NEW; @Nullable - private final ClusterByStatisticsCollector resultKeyStatisticsCollector; + public final CompleteKeyStatisticsInformation completeKeyStatisticsInformation; // Result partitions and where they can be read from. @Nullable @@ -81,8 +83,7 @@ class ControllerStageTracker private ControllerStageTracker( final StageDefinition stageDef, - final WorkerInputs workerInputs, - final int partitionStatisticsMaxRetainedBytes + final WorkerInputs workerInputs ) { this.stageDef = stageDef; @@ -90,11 +91,11 @@ private ControllerStageTracker( this.workerInputs = workerInputs; if (stageDef.mustGatherResultKeyStatistics()) { - this.resultKeyStatisticsCollector = - stageDef.createResultKeyStatisticsCollector(partitionStatisticsMaxRetainedBytes); + this.completeKeyStatisticsInformation = + new CompleteKeyStatisticsInformation(new TreeMap<>(), false, 0); } else { - this.resultKeyStatisticsCollector = null; - generateResultPartitionsAndBoundaries(); + this.completeKeyStatisticsInformation = null; + generateResultPartitionsAndBoundariesWithoutKeyStatistics(); } } @@ -107,12 +108,11 @@ static ControllerStageTracker create( final StageDefinition stageDef, final Int2IntMap stageWorkerCountMap, final InputSpecSlicer slicer, - final WorkerAssignmentStrategy assignmentStrategy, - final int partitionStatisticsMaxRetainedBytes + final WorkerAssignmentStrategy assignmentStrategy ) { final WorkerInputs workerInputs = WorkerInputs.create(stageDef, stageWorkerCountMap, slicer, assignmentStrategy); - return new ControllerStageTracker(stageDef, workerInputs, partitionStatisticsMaxRetainedBytes); + return new ControllerStageTracker(stageDef, workerInputs); } /** @@ -175,18 +175,12 @@ ClusterByPartitions getResultPartitionBoundaries() */ boolean collectorEncounteredAnyMultiValueField() { - if (resultKeyStatisticsCollector == null) { + if (completeKeyStatisticsInformation == null) { throw new ISE("Stage does not gather result key statistics"); - } else if (resultPartitions == null) { + } else if (workersWithReportedKeyStatistics.size() != workerCount) { throw new ISE("Result key statistics are not ready"); } else { - for (int i = 0; i < resultKeyStatisticsCollector.getClusterBy().getColumns().size(); i++) { - if (resultKeyStatisticsCollector.hasMultipleValues(i)) { - return true; - } - } - - return false; + return completeKeyStatisticsInformation.hasMultipleValues(); } } @@ -219,10 +213,6 @@ void start() */ void finish() { - if (resultKeyStatisticsCollector != null) { - resultKeyStatisticsCollector.clear(); - } - transitionTo(ControllerStagePhase.FINISHED); } @@ -234,23 +224,31 @@ WorkerInputs getWorkerInputs() return workerInputs; } + /** + * Returns the merged key statistics. + */ + @Nullable + public CompleteKeyStatisticsInformation getCompleteKeyStatisticsInformation() + { + return completeKeyStatisticsInformation; + } + /** * Adds result key statistics for a particular worker number. If statistics have already been added for this worker, * then this call ignores the new ones and does nothing. * * @param workerNumber the worker - * @param snapshot worker statistics + * @param partialKeyStatisticsInformation partial key statistics */ - ControllerStagePhase addResultKeyStatisticsForWorker( + ControllerStagePhase addPartialKeyStatisticsForWorker( final int workerNumber, - final ClusterByStatisticsSnapshot snapshot + final PartialKeyStatisticsInformation partialKeyStatisticsInformation ) { if (phase != ControllerStagePhase.READING_INPUT) { throw new ISE("Cannot add result key statistics from stage [%s]", phase); } - - if (resultKeyStatisticsCollector == null) { + if (!stageDef.mustGatherResultKeyStatistics() || !stageDef.doesShuffle() || completeKeyStatisticsInformation == null) { throw new ISE("Stage does not gather result key statistics"); } @@ -259,16 +257,21 @@ ControllerStagePhase addResultKeyStatisticsForWorker( } try { - if (workersWithResultKeyStatistics.add(workerNumber)) { - resultKeyStatisticsCollector.addAll(snapshot); + if (workersWithReportedKeyStatistics.add(workerNumber)) { - if (workersWithResultKeyStatistics.size() == workerCount) { - generateResultPartitionsAndBoundaries(); + if (partialKeyStatisticsInformation.getTimeSegments().contains(null)) { + // Time should not contain null value + failForReason(InsertTimeNullFault.instance()); + return getPhase(); + } + + completeKeyStatisticsInformation.mergePartialInformation(workerNumber, partialKeyStatisticsInformation); + + if (workersWithReportedKeyStatistics.size() == workerCount) { + // All workers have sent the partial key statistics information. + // Transition to MERGING_STATISTICS state to queue fetch clustering statistics from workers. + transitionTo(ControllerStagePhase.MERGING_STATISTICS); - // Phase can become FAILED after generateResultPartitionsAndBoundaries, if there were too many partitions. - if (phase != ControllerStagePhase.FAILED) { - transitionTo(ControllerStagePhase.POST_READING); - } } } } @@ -280,6 +283,33 @@ ControllerStagePhase addResultKeyStatisticsForWorker( return getPhase(); } + /** + * Sets the {@link #resultPartitions} and {@link #resultPartitionBoundaries} and transitions the phase to POST_READING. + */ + void setClusterByPartitionBoundaries(ClusterByPartitions clusterByPartitions) + { + if (resultPartitions != null) { + throw new ISE("Result partitions have already been generated"); + } + + if (!stageDef.mustGatherResultKeyStatistics()) { + throw new ISE("Result partitions does not require key statistics, should not have set partition boundries here"); + } + + if (!ControllerStagePhase.MERGING_STATISTICS.equals(getPhase())) { + throw new ISE("Cannot set partition boundires from key statistics from stage [%s]", getPhase()); + } + + this.resultPartitionBoundaries = clusterByPartitions; + this.resultPartitions = ReadablePartitions.striped( + stageDef.getStageNumber(), + workerCount, + clusterByPartitions.size() + ); + + transitionTo(ControllerStagePhase.POST_READING); + } + /** * Accepts and sets the results that each worker produces for this particular stage * @@ -339,12 +369,11 @@ void fail() } /** - * Sets {@link #resultPartitions} (always) and {@link #resultPartitionBoundaries}. + * Sets {@link #resultPartitions} (always) and {@link #resultPartitionBoundaries} without using key statistics. * - * If {@link StageDefinition#mustGatherResultKeyStatistics()} is true, this method cannot be called until after - * statistics have been provided to {@link #addResultKeyStatisticsForWorker} for all workers. + * If {@link StageDefinition#mustGatherResultKeyStatistics()} is true, this method should not be called. */ - private void generateResultPartitionsAndBoundaries() + private void generateResultPartitionsAndBoundariesWithoutKeyStatistics() { if (resultPartitions != null) { throw new ISE("Result partitions have already been generated"); @@ -353,12 +382,12 @@ private void generateResultPartitionsAndBoundaries() final int stageNumber = stageDef.getStageNumber(); if (stageDef.doesShuffle()) { - if (stageDef.mustGatherResultKeyStatistics() && workersWithResultKeyStatistics.size() != workerCount) { - throw new ISE("Cannot generate result partitions without all worker statistics"); + if (stageDef.mustGatherResultKeyStatistics()) { + throw new ISE("Cannot generate result partitions without key statistics"); } final Either maybeResultPartitionBoundaries = - stageDef.generatePartitionsForShuffle(resultKeyStatisticsCollector); + stageDef.generatePartitionsForShuffle(null); if (maybeResultPartitionBoundaries.isError()) { failForReason(new TooManyPartitionsFault(stageDef.getMaxPartitionCount())); @@ -397,15 +426,11 @@ private void generateResultPartitionsAndBoundaries() * * @param fault reason why this stage has failed */ - private void failForReason(final MSQFault fault) + void failForReason(final MSQFault fault) { transitionTo(ControllerStagePhase.FAILED); this.failureReason = fault; - - if (resultKeyStatisticsCollector != null) { - resultKeyStatisticsCollector.clear(); - } } void transitionTo(final ControllerStagePhase newPhase) diff --git a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/statistics/ClusterByStatisticsCollectorImpl.java b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/statistics/ClusterByStatisticsCollectorImpl.java index 13ab5dc01cee..bd010c555f17 100644 --- a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/statistics/ClusterByStatisticsCollectorImpl.java +++ b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/statistics/ClusterByStatisticsCollectorImpl.java @@ -35,6 +35,7 @@ import java.util.ArrayList; import java.util.Collections; import java.util.Comparator; +import java.util.HashMap; import java.util.List; import java.util.Map; import java.util.SortedMap; @@ -56,7 +57,7 @@ public class ClusterByStatisticsCollectorImpl implements ClusterByStatisticsColl private final boolean[] hasMultipleValues; - private final int maxRetainedBytes; + private final long maxRetainedBytes; private final int maxBuckets; private long totalRetainedBytes; @@ -64,7 +65,7 @@ private ClusterByStatisticsCollectorImpl( final ClusterBy clusterBy, final RowKeyReader keyReader, final KeyCollectorFactory keyCollectorFactory, - final int maxRetainedBytes, + final long maxRetainedBytes, final int maxBuckets, final boolean checkHasMultipleValues ) @@ -86,7 +87,7 @@ private ClusterByStatisticsCollectorImpl( public static ClusterByStatisticsCollector create( final ClusterBy clusterBy, final RowSignature signature, - final int maxRetainedBytes, + final long maxRetainedBytes, final int maxBuckets, final boolean aggregate, final boolean checkHasMultipleValues @@ -167,7 +168,7 @@ public ClusterByStatisticsCollector addAll(final ClusterByStatisticsCollector ot public ClusterByStatisticsCollector addAll(final ClusterByStatisticsSnapshot snapshot) { // Add all key collectors from the other collector. - for (ClusterByStatisticsSnapshot.Bucket otherBucket : snapshot.getBuckets()) { + for (ClusterByStatisticsSnapshot.Bucket otherBucket : snapshot.getBuckets().values()) { //noinspection rawtypes, unchecked final KeyCollector otherKeyCollector = ((KeyCollectorFactory) keyCollectorFactory).fromSnapshot(otherBucket.getKeyCollectorSnapshot()); @@ -315,13 +316,20 @@ public ClusterByStatisticsSnapshot snapshot() { assertRetainedByteCountsAreTrackedCorrectly(); - final List bucketSnapshots = new ArrayList<>(); + final Map bucketSnapshots = new HashMap<>(); + final RowKeyReader trimmedRowReader = keyReader.trimmedKeyReader(clusterBy.getBucketByCount()); for (final Map.Entry bucketEntry : buckets.entrySet()) { //noinspection rawtypes, unchecked final KeyCollectorSnapshot keyCollectorSnapshot = ((KeyCollectorFactory) keyCollectorFactory).toSnapshot(bucketEntry.getValue().keyCollector); - bucketSnapshots.add(new ClusterByStatisticsSnapshot.Bucket(bucketEntry.getKey(), keyCollectorSnapshot)); + Long bucketKey = Long.MIN_VALUE; + + // If there is a clustering on time, read the first field from each bucket and add it to the snapshots. + if (clusterBy.getBucketByCount() == 1) { + bucketKey = (Long) trimmedRowReader.read(bucketEntry.getKey(), 0); + } + bucketSnapshots.put(bucketKey, new ClusterByStatisticsSnapshot.Bucket(bucketEntry.getKey(), keyCollectorSnapshot, totalRetainedBytes)); } final IntSet hasMultipleValuesSet; diff --git a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/statistics/ClusterByStatisticsSnapshot.java b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/statistics/ClusterByStatisticsSnapshot.java index 6b16890254a0..e54253ad2186 100644 --- a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/statistics/ClusterByStatisticsSnapshot.java +++ b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/statistics/ClusterByStatisticsSnapshot.java @@ -23,22 +23,23 @@ import com.fasterxml.jackson.annotation.JsonInclude; import com.fasterxml.jackson.annotation.JsonProperty; import com.google.common.base.Preconditions; +import com.google.common.collect.ImmutableMap; import org.apache.druid.frame.key.RowKey; import javax.annotation.Nullable; import java.util.Collections; -import java.util.List; +import java.util.Map; import java.util.Objects; import java.util.Set; public class ClusterByStatisticsSnapshot { - private final List buckets; + private final Map buckets; private final Set hasMultipleValues; @JsonCreator ClusterByStatisticsSnapshot( - @JsonProperty("buckets") final List buckets, + @JsonProperty("buckets") final Map buckets, @JsonProperty("hasMultipleValues") @Nullable final Set hasMultipleValues ) { @@ -48,15 +49,21 @@ public class ClusterByStatisticsSnapshot public static ClusterByStatisticsSnapshot empty() { - return new ClusterByStatisticsSnapshot(Collections.emptyList(), null); + return new ClusterByStatisticsSnapshot(Collections.emptyMap(), null); } @JsonProperty("buckets") - List getBuckets() + Map getBuckets() { return buckets; } + public ClusterByStatisticsSnapshot getSnapshotForTimeChunk(long timeChunk) + { + Bucket bucket = buckets.get(timeChunk); + return new ClusterByStatisticsSnapshot(ImmutableMap.of(timeChunk, bucket), null); + } + @JsonProperty("hasMultipleValues") @JsonInclude(JsonInclude.Include.NON_EMPTY) Set getHasMultipleValues() @@ -64,6 +71,15 @@ Set getHasMultipleValues() return hasMultipleValues; } + public PartialKeyStatisticsInformation partialKeyStatistics() + { + double bytesRetained = 0; + for (ClusterByStatisticsSnapshot.Bucket bucket : buckets.values()) { + bytesRetained += bucket.bytesRetained; + } + return new PartialKeyStatisticsInformation(buckets.keySet(), !getHasMultipleValues().isEmpty(), bytesRetained); + } + @Override public boolean equals(Object o) { @@ -86,16 +102,19 @@ public int hashCode() static class Bucket { private final RowKey bucketKey; + private final double bytesRetained; private final KeyCollectorSnapshot keyCollectorSnapshot; @JsonCreator Bucket( @JsonProperty("bucketKey") RowKey bucketKey, - @JsonProperty("data") KeyCollectorSnapshot keyCollectorSnapshot + @JsonProperty("data") KeyCollectorSnapshot keyCollectorSnapshot, + @JsonProperty("bytesRetained") double bytesRetained ) { this.bucketKey = Preconditions.checkNotNull(bucketKey, "bucketKey"); this.keyCollectorSnapshot = Preconditions.checkNotNull(keyCollectorSnapshot, "data"); + this.bytesRetained = bytesRetained; } @JsonProperty diff --git a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/statistics/CompleteKeyStatisticsInformation.java b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/statistics/CompleteKeyStatisticsInformation.java new file mode 100644 index 000000000000..8a915fc8cd8e --- /dev/null +++ b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/statistics/CompleteKeyStatisticsInformation.java @@ -0,0 +1,82 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.msq.statistics; + +import com.google.common.collect.ImmutableSortedMap; + +import java.util.HashSet; +import java.util.Set; +import java.util.SortedMap; + +/** + * Class maintained by the controller to merge {@link PartialKeyStatisticsInformation} sent by the worker. + */ +public class CompleteKeyStatisticsInformation +{ + private final SortedMap> timeSegmentVsWorkerMap; + + private boolean multipleValues; + + private double bytesRetained; + + public CompleteKeyStatisticsInformation( + final SortedMap> timeChunks, + boolean multipleValues, + double bytesRetained + ) + { + this.timeSegmentVsWorkerMap = timeChunks; + this.multipleValues = multipleValues; + this.bytesRetained = bytesRetained; + } + + /** + * Merges the {@link PartialKeyStatisticsInformation} into the complete key statistics information object. + * {@link #timeSegmentVsWorkerMap} is updated in sorted order with the timechunks from + * {@param partialKeyStatisticsInformation}, {@link #multipleValues} is set to true if + * {@param partialKeyStatisticsInformation} contains multipleValues and the bytes retained by the partial sketch + * is added to {@link #bytesRetained}. + */ + public void mergePartialInformation(int workerNumber, PartialKeyStatisticsInformation partialKeyStatisticsInformation) + { + for (Long timeSegment : partialKeyStatisticsInformation.getTimeSegments()) { + this.timeSegmentVsWorkerMap + .computeIfAbsent(timeSegment, key -> new HashSet<>()) + .add(workerNumber); + } + this.multipleValues = this.multipleValues || partialKeyStatisticsInformation.hasMultipleValues(); + this.bytesRetained += bytesRetained; + } + + public SortedMap> getTimeSegmentVsWorkerMap() + { + return ImmutableSortedMap.copyOfSorted(timeSegmentVsWorkerMap); + } + + public boolean hasMultipleValues() + { + return multipleValues; + } + + public double getBytesRetained() + { + return bytesRetained; + } +} diff --git a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/statistics/DelegateOrMinKeyCollectorSnapshot.java b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/statistics/DelegateOrMinKeyCollectorSnapshot.java index f0191da126b2..5e8da8df0c22 100644 --- a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/statistics/DelegateOrMinKeyCollectorSnapshot.java +++ b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/statistics/DelegateOrMinKeyCollectorSnapshot.java @@ -22,16 +22,19 @@ import com.fasterxml.jackson.annotation.JsonCreator; import com.fasterxml.jackson.annotation.JsonInclude; import com.fasterxml.jackson.annotation.JsonProperty; +import com.fasterxml.jackson.annotation.JsonTypeName; import org.apache.druid.frame.key.RowKey; import org.apache.druid.java.util.common.ISE; import javax.annotation.Nullable; import java.util.Objects; +@JsonTypeName(DelegateOrMinKeyCollectorSnapshot.TYPE) public class DelegateOrMinKeyCollectorSnapshot implements KeyCollectorSnapshot { static final String FIELD_SNAPSHOT = "snapshot"; static final String FIELD_MIN_KEY = "minKey"; + static final String TYPE = "delegate"; private final T snapshot; private final RowKey minKey; diff --git a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/statistics/DistinctKeySnapshot.java b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/statistics/DistinctKeySnapshot.java index 44b95d3874f6..47d46f72cabb 100644 --- a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/statistics/DistinctKeySnapshot.java +++ b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/statistics/DistinctKeySnapshot.java @@ -22,6 +22,7 @@ import com.fasterxml.jackson.annotation.JsonCreator; import com.fasterxml.jackson.annotation.JsonInclude; import com.fasterxml.jackson.annotation.JsonProperty; +import com.fasterxml.jackson.annotation.JsonTypeName; import com.google.common.base.Preconditions; import org.apache.druid.collections.SerializablePair; import org.apache.druid.frame.key.RowKey; @@ -31,8 +32,10 @@ import java.util.Map; import java.util.Objects; +@JsonTypeName(DistinctKeySnapshot.TYPE) public class DistinctKeySnapshot implements KeyCollectorSnapshot { + static final String TYPE = "distinct"; private final List> keys; private final int spaceReductionFactor; diff --git a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/statistics/KeyCollectorSnapshot.java b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/statistics/KeyCollectorSnapshot.java index ad359cb1e60e..d1e717c37ec9 100644 --- a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/statistics/KeyCollectorSnapshot.java +++ b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/statistics/KeyCollectorSnapshot.java @@ -19,9 +19,18 @@ package org.apache.druid.msq.statistics; +import com.fasterxml.jackson.annotation.JsonSubTypes; +import com.fasterxml.jackson.annotation.JsonTypeInfo; + /** * Marker interface for deserialization. */ +@JsonTypeInfo(use = JsonTypeInfo.Id.NAME, property = "collectorType") +@JsonSubTypes(value = { + @JsonSubTypes.Type(name = DelegateOrMinKeyCollectorSnapshot.TYPE, value = DelegateOrMinKeyCollectorSnapshot.class), + @JsonSubTypes.Type(name = QuantilesSketchKeyCollectorSnapshot.TYPE, value = QuantilesSketchKeyCollectorSnapshot.class), + @JsonSubTypes.Type(name = DistinctKeySnapshot.TYPE, value = DistinctKeySnapshot.class), +}) public interface KeyCollectorSnapshot { } diff --git a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/statistics/PartialKeyStatisticsInformation.java b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/statistics/PartialKeyStatisticsInformation.java new file mode 100644 index 000000000000..535af8dafb0a --- /dev/null +++ b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/statistics/PartialKeyStatisticsInformation.java @@ -0,0 +1,67 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.msq.statistics; + +import com.fasterxml.jackson.annotation.JsonCreator; +import com.fasterxml.jackson.annotation.JsonProperty; + +import java.util.Set; + +/** + * Class sent by worker to controller after reading input to generate partition boundries. + */ +public class PartialKeyStatisticsInformation +{ + private final Set timeSegments; + + private final boolean multipleValues; + + private final double bytesRetained; + + @JsonCreator + public PartialKeyStatisticsInformation( + @JsonProperty("timeSegments") Set timeSegments, + @JsonProperty("multipleValues") boolean hasMultipleValues, + @JsonProperty("bytesRetained") double bytesRetained + ) + { + this.timeSegments = timeSegments; + this.multipleValues = hasMultipleValues; + this.bytesRetained = bytesRetained; + } + + @JsonProperty("timeSegments") + public Set getTimeSegments() + { + return timeSegments; + } + + @JsonProperty("multipleValues") + public boolean hasMultipleValues() + { + return multipleValues; + } + + @JsonProperty("bytesRetained") + public double getBytesRetained() + { + return bytesRetained; + } +} diff --git a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/statistics/QuantilesSketchKeyCollectorSnapshot.java b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/statistics/QuantilesSketchKeyCollectorSnapshot.java index 1b555ac3f944..5784afd49a0f 100644 --- a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/statistics/QuantilesSketchKeyCollectorSnapshot.java +++ b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/statistics/QuantilesSketchKeyCollectorSnapshot.java @@ -21,11 +21,14 @@ import com.fasterxml.jackson.annotation.JsonCreator; import com.fasterxml.jackson.annotation.JsonProperty; +import com.fasterxml.jackson.annotation.JsonTypeName; import java.util.Objects; +@JsonTypeName(QuantilesSketchKeyCollectorSnapshot.TYPE) public class QuantilesSketchKeyCollectorSnapshot implements KeyCollectorSnapshot { + static final String TYPE = "quantile"; private final String encodedSketch; private final double averageKeyLength; diff --git a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/util/MultiStageQueryContext.java b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/util/MultiStageQueryContext.java index c7c61fd30c76..7c589f2326f1 100644 --- a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/util/MultiStageQueryContext.java +++ b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/util/MultiStageQueryContext.java @@ -25,6 +25,7 @@ import com.google.common.annotations.VisibleForTesting; import com.opencsv.RFC4180Parser; import com.opencsv.RFC4180ParserBuilder; +import org.apache.druid.msq.exec.ClusterStatisticsMergeMode; import org.apache.druid.msq.kernel.WorkerAssignmentStrategy; import org.apache.druid.msq.sql.MSQMode; import org.apache.druid.query.QueryContext; @@ -58,6 +59,8 @@ public class MultiStageQueryContext private static final boolean DEFAULT_FINALIZE_AGGREGATIONS = true; public static final String CTX_ENABLE_DURABLE_SHUFFLE_STORAGE = "durableShuffleStorage"; + public static final String CTX_CLUSTER_STATISTICS_MERGE_MODE = "clusterStatisticsMergeMode"; + public static final String DEFAULT_CLUSTER_STATISTICS_MERGE_MODE = ClusterStatisticsMergeMode.AUTO.toString(); private static final boolean DEFAULT_ENABLE_DURABLE_SHUFFLE_STORAGE = false; public static final String CTX_DESTINATION = "destination"; @@ -93,6 +96,18 @@ public static boolean isDurableStorageEnabled(final QueryContext queryContext) ); } + public static ClusterStatisticsMergeMode getClusterStatisticsMergeMode(QueryContext queryContext) + { + return ClusterStatisticsMergeMode.valueOf( + String.valueOf( + queryContext.getString( + CTX_CLUSTER_STATISTICS_MERGE_MODE, + DEFAULT_CLUSTER_STATISTICS_MERGE_MODE + ) + ) + ); + } + public static boolean isFinalizeAggregations(final QueryContext queryContext) { return queryContext.getBoolean( diff --git a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/WorkerSketchFetcherAutoModeTest.java b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/WorkerSketchFetcherAutoModeTest.java new file mode 100644 index 000000000000..42f6f0437f59 --- /dev/null +++ b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/WorkerSketchFetcherAutoModeTest.java @@ -0,0 +1,139 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.msq.exec; + +import org.apache.druid.frame.key.ClusterBy; +import org.apache.druid.msq.kernel.StageDefinition; +import org.apache.druid.msq.kernel.StageId; +import org.apache.druid.msq.statistics.CompleteKeyStatisticsInformation; +import org.junit.After; +import org.junit.Before; +import org.junit.Test; +import org.mockito.Mock; +import org.mockito.MockitoAnnotations; + +import java.util.Collections; + +import static org.mockito.ArgumentMatchers.any; +import static org.mockito.Mockito.doReturn; +import static org.mockito.Mockito.mock; +import static org.mockito.Mockito.spy; +import static org.mockito.Mockito.times; +import static org.mockito.Mockito.verify; + +public class WorkerSketchFetcherAutoModeTest +{ + @Mock + private CompleteKeyStatisticsInformation completeKeyStatisticsInformation; + @Mock + private StageDefinition stageDefinition; + @Mock + private ClusterBy clusterBy; + private AutoCloseable mocks; + private WorkerSketchFetcher target; + + @Before + public void setUp() + { + mocks = MockitoAnnotations.openMocks(this); + + target = spy(new WorkerSketchFetcher(mock(WorkerClient.class), ClusterStatisticsMergeMode.AUTO, 300_000_000)); + // Don't actually try to fetch sketches + doReturn(null).when(target).inMemoryFullSketchMerging(any(), any()); + doReturn(null).when(target).sequentialTimeChunkMerging(any(), any(), any()); + + doReturn(StageId.fromString("1_1")).when(stageDefinition).getId(); + doReturn(clusterBy).when(stageDefinition).getClusterBy(); + } + + @After + public void tearDown() throws Exception + { + mocks.close(); + } + + @Test + public void test_submitFetcherTask_belowThresholds_ShouldBeParallel() + { + // Bytes below threshold + doReturn(10.0).when(completeKeyStatisticsInformation).getBytesRetained(); + + // Cluster by bucket count not 0 + doReturn(1).when(clusterBy).getBucketByCount(); + + // Worker count below threshold + doReturn(1).when(stageDefinition).getMaxWorkerCount(); + + target.submitFetcherTask(completeKeyStatisticsInformation, Collections.emptyList(), stageDefinition); + verify(target, times(1)).inMemoryFullSketchMerging(any(), any()); + verify(target, times(0)).sequentialTimeChunkMerging(any(), any(), any()); + } + + @Test + public void test_submitFetcherTask_workerCountAboveThreshold_shouldBeSequential() + { + // Bytes below threshold + doReturn(10.0).when(completeKeyStatisticsInformation).getBytesRetained(); + + // Cluster by bucket count not 0 + doReturn(1).when(clusterBy).getBucketByCount(); + + // Worker count below threshold + doReturn((int) WorkerSketchFetcher.WORKER_THRESHOLD + 1).when(stageDefinition).getMaxWorkerCount(); + + target.submitFetcherTask(completeKeyStatisticsInformation, Collections.emptyList(), stageDefinition); + verify(target, times(0)).inMemoryFullSketchMerging(any(), any()); + verify(target, times(1)).sequentialTimeChunkMerging(any(), any(), any()); + } + + @Test + public void test_submitFetcherTask_noClusterByColumns_shouldBeParallel() + { + // Bytes above threshold + doReturn(WorkerSketchFetcher.BYTES_THRESHOLD + 10.0).when(completeKeyStatisticsInformation).getBytesRetained(); + + // Cluster by bucket count 0 + doReturn(ClusterBy.none()).when(stageDefinition).getClusterBy(); + + // Worker count above threshold + doReturn((int) WorkerSketchFetcher.WORKER_THRESHOLD + 1).when(stageDefinition).getMaxWorkerCount(); + + target.submitFetcherTask(completeKeyStatisticsInformation, Collections.emptyList(), stageDefinition); + verify(target, times(1)).inMemoryFullSketchMerging(any(), any()); + verify(target, times(0)).sequentialTimeChunkMerging(any(), any(), any()); + } + + @Test + public void test_submitFetcherTask_bytesRetainedAboveThreshold_shouldBeSequential() + { + // Bytes above threshold + doReturn(WorkerSketchFetcher.BYTES_THRESHOLD + 10.0).when(completeKeyStatisticsInformation).getBytesRetained(); + + // Cluster by bucket count not 0 + doReturn(1).when(clusterBy).getBucketByCount(); + + // Worker count below threshold + doReturn(1).when(stageDefinition).getMaxWorkerCount(); + + target.submitFetcherTask(completeKeyStatisticsInformation, Collections.emptyList(), stageDefinition); + verify(target, times(0)).inMemoryFullSketchMerging(any(), any()); + verify(target, times(1)).sequentialTimeChunkMerging(any(), any(), any()); + } +} diff --git a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/WorkerSketchFetcherTest.java b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/WorkerSketchFetcherTest.java new file mode 100644 index 000000000000..54c9a792e558 --- /dev/null +++ b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/WorkerSketchFetcherTest.java @@ -0,0 +1,295 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.msq.exec; + +import com.google.common.collect.ImmutableList; +import com.google.common.collect.ImmutableSet; +import com.google.common.collect.ImmutableSortedMap; +import com.google.common.util.concurrent.Futures; +import com.google.common.util.concurrent.ListenableFuture; +import org.apache.druid.frame.key.ClusterBy; +import org.apache.druid.frame.key.ClusterByPartition; +import org.apache.druid.frame.key.ClusterByPartitions; +import org.apache.druid.frame.key.RowKey; +import org.apache.druid.java.util.common.Either; +import org.apache.druid.msq.kernel.StageDefinition; +import org.apache.druid.msq.kernel.StageId; +import org.apache.druid.msq.statistics.ClusterByStatisticsCollector; +import org.apache.druid.msq.statistics.ClusterByStatisticsSnapshot; +import org.apache.druid.msq.statistics.CompleteKeyStatisticsInformation; +import org.junit.After; +import org.junit.Assert; +import org.junit.Before; +import org.junit.Test; +import org.mockito.Mock; +import org.mockito.MockitoAnnotations; + +import java.util.List; +import java.util.Queue; +import java.util.Set; +import java.util.SortedMap; +import java.util.concurrent.CompletableFuture; +import java.util.concurrent.CompletionException; +import java.util.concurrent.ConcurrentLinkedQueue; +import java.util.concurrent.CountDownLatch; +import java.util.concurrent.CyclicBarrier; +import java.util.concurrent.ExecutionException; + +import static org.easymock.EasyMock.mock; +import static org.mockito.ArgumentMatchers.any; +import static org.mockito.ArgumentMatchers.anyInt; +import static org.mockito.ArgumentMatchers.anyLong; +import static org.mockito.ArgumentMatchers.anyString; +import static org.mockito.ArgumentMatchers.eq; +import static org.mockito.Mockito.doAnswer; +import static org.mockito.Mockito.doReturn; +import static org.mockito.Mockito.spy; +import static org.mockito.Mockito.times; +import static org.mockito.Mockito.verify; + +public class WorkerSketchFetcherTest +{ + @Mock + private CompleteKeyStatisticsInformation completeKeyStatisticsInformation; + @Mock + private StageDefinition stageDefinition; + @Mock + private ClusterBy clusterBy; + @Mock + private ClusterByStatisticsCollector mergedClusterByStatisticsCollector1; + @Mock + private ClusterByStatisticsCollector mergedClusterByStatisticsCollector2; + @Mock + private WorkerClient workerClient; + private ClusterByPartitions expectedPartitions1; + private ClusterByPartitions expectedPartitions2; + private AutoCloseable mocks; + private WorkerSketchFetcher target; + + @Before + public void setUp() + { + mocks = MockitoAnnotations.openMocks(this); + doReturn(StageId.fromString("1_1")).when(stageDefinition).getId(); + doReturn(clusterBy).when(stageDefinition).getClusterBy(); + doReturn(25_000).when(stageDefinition).getMaxPartitionCount(); + + expectedPartitions1 = new ClusterByPartitions(ImmutableList.of(new ClusterByPartition(mock(RowKey.class), mock(RowKey.class)))); + expectedPartitions2 = new ClusterByPartitions(ImmutableList.of(new ClusterByPartition(mock(RowKey.class), mock(RowKey.class)))); + + doReturn(Either.value(expectedPartitions1)).when(stageDefinition).generatePartitionsForShuffle(eq(mergedClusterByStatisticsCollector1)); + doReturn(Either.value(expectedPartitions2)).when(stageDefinition).generatePartitionsForShuffle(eq(mergedClusterByStatisticsCollector2)); + + doReturn( + mergedClusterByStatisticsCollector1, + mergedClusterByStatisticsCollector2 + ).when(stageDefinition).createResultKeyStatisticsCollector(anyInt()); + } + + @After + public void tearDown() throws Exception + { + mocks.close(); + } + + @Test + public void test_submitFetcherTask_parallelFetch_workerThrowsException_shouldCancelOtherTasks() throws Exception + { + // Store futures in a queue + final Queue> futureQueue = new ConcurrentLinkedQueue<>(); + final List workerIds = ImmutableList.of("0", "1", "2", "3"); + final CountDownLatch latch = new CountDownLatch(workerIds.size()); + + target = spy(new WorkerSketchFetcher(workerClient, ClusterStatisticsMergeMode.PARALLEL, 300_000_000)); + + // When fetching snapshots, return a mock and add future to queue + doAnswer(invocation -> { + ListenableFuture snapshotListenableFuture = + spy(Futures.immediateFuture(mock(ClusterByStatisticsSnapshot.class))); + futureQueue.add(snapshotListenableFuture); + latch.countDown(); + latch.await(); + return snapshotListenableFuture; + }).when(workerClient).fetchClusterByStatisticsSnapshot(any(), any(), anyInt()); + + // Cause a worker to fail instead of returning the result + doAnswer(invocation -> { + latch.countDown(); + latch.await(); + return Futures.immediateFailedFuture(new InterruptedException("interrupted")); + }).when(workerClient).fetchClusterByStatisticsSnapshot(eq("2"), any(), anyInt()); + + CompletableFuture> eitherCompletableFuture = target.submitFetcherTask( + completeKeyStatisticsInformation, + workerIds, + stageDefinition + ); + + // Assert that the final result is failed and all other task futures are also cancelled. + Assert.assertThrows(CompletionException.class, eitherCompletableFuture::join); + Thread.sleep(1000); + + Assert.assertTrue(eitherCompletableFuture.isCompletedExceptionally()); + // Verify that the statistics collector was cleared due to the error. + verify(mergedClusterByStatisticsCollector1, times(1)).clear(); + // Verify that other task futures were requested to be cancelled. + Assert.assertFalse(futureQueue.isEmpty()); + for (ListenableFuture snapshotFuture : futureQueue) { + verify(snapshotFuture, times(1)).cancel(eq(true)); + } + } + + @Test + public void test_submitFetcherTask_parallelFetch_mergePerformedCorrectly() + throws ExecutionException, InterruptedException + { + // Store snapshots in a queue + final Queue snapshotQueue = new ConcurrentLinkedQueue<>(); + final List workerIds = ImmutableList.of("0", "1", "2", "3", "4"); + final CountDownLatch latch = new CountDownLatch(workerIds.size()); + + target = spy(new WorkerSketchFetcher(workerClient, ClusterStatisticsMergeMode.PARALLEL, 300_000_000)); + + // When fetching snapshots, return a mock and add it to queue + doAnswer(invocation -> { + ClusterByStatisticsSnapshot snapshot = mock(ClusterByStatisticsSnapshot.class); + snapshotQueue.add(snapshot); + latch.countDown(); + return Futures.immediateFuture(snapshot); + }).when(workerClient).fetchClusterByStatisticsSnapshot(any(), any(), anyInt()); + + CompletableFuture> eitherCompletableFuture = target.submitFetcherTask( + completeKeyStatisticsInformation, + workerIds, + stageDefinition + ); + + // Assert that the final result is complete and all other sketches returned have been merged. + eitherCompletableFuture.join(); + Thread.sleep(1000); + Assert.assertTrue(eitherCompletableFuture.isDone() && !eitherCompletableFuture.isCompletedExceptionally()); + Assert.assertFalse(snapshotQueue.isEmpty()); + // Verify that all statistics were added to controller. + for (ClusterByStatisticsSnapshot snapshot : snapshotQueue) { + verify(mergedClusterByStatisticsCollector1, times(1)).addAll(eq(snapshot)); + } + // Check that the partitions returned by the merged collector is returned by the final future. + Assert.assertEquals(expectedPartitions1, eitherCompletableFuture.get().valueOrThrow()); + } + + @Test + public void test_submitFetcherTask_sequentialFetch_workerThrowsException_shouldCancelOtherTasks() throws Exception + { + // Store futures in a queue + final Queue> futureQueue = new ConcurrentLinkedQueue<>(); + + SortedMap> timeSegmentVsWorkerMap = ImmutableSortedMap.of(1L, ImmutableSet.of(0, 1, 2), 2L, ImmutableSet.of(0, 1, 4)); + doReturn(timeSegmentVsWorkerMap).when(completeKeyStatisticsInformation).getTimeSegmentVsWorkerMap(); + + final CyclicBarrier barrier = new CyclicBarrier(3); + target = spy(new WorkerSketchFetcher(workerClient, ClusterStatisticsMergeMode.SEQUENTIAL, 300_000_000)); + + // When fetching snapshots, return a mock and add future to queue + doAnswer(invocation -> { + ListenableFuture snapshotListenableFuture = + spy(Futures.immediateFuture(mock(ClusterByStatisticsSnapshot.class))); + futureQueue.add(snapshotListenableFuture); + barrier.await(); + return snapshotListenableFuture; + }).when(workerClient).fetchClusterByStatisticsSnapshotForTimeChunk(anyString(), anyString(), anyInt(), anyLong()); + + // Cause a worker in the second time chunk to fail instead of returning the result + doAnswer(invocation -> { + barrier.await(); + return Futures.immediateFailedFuture(new InterruptedException("interrupted")); + }).when(workerClient).fetchClusterByStatisticsSnapshotForTimeChunk(eq("4"), any(), anyInt(), eq(2L)); + + CompletableFuture> eitherCompletableFuture = target.submitFetcherTask( + completeKeyStatisticsInformation, + ImmutableList.of("0", "1", "2", "3", "4"), + stageDefinition + ); + + // Assert that the final result is failed and all other task futures are also cancelled. + Assert.assertThrows(CompletionException.class, eitherCompletableFuture::join); + Thread.sleep(1000); + + Assert.assertTrue(eitherCompletableFuture.isCompletedExceptionally()); + // Verify that the correct statistics collector was cleared due to the error. + verify(mergedClusterByStatisticsCollector1, times(0)).clear(); + verify(mergedClusterByStatisticsCollector2, times(1)).clear(); + // Verify that other task futures were requested to be cancelled. + Assert.assertFalse(futureQueue.isEmpty()); + for (ListenableFuture snapshotFuture : futureQueue) { + verify(snapshotFuture, times(1)).cancel(eq(true)); + } + } + + @Test + public void test_submitFetcherTask_sequentialFetch_mergePerformedCorrectly() + throws ExecutionException, InterruptedException + { + // Store snapshots in a queue + final Queue snapshotQueue = new ConcurrentLinkedQueue<>(); + + SortedMap> timeSegmentVsWorkerMap = ImmutableSortedMap.of(1L, ImmutableSet.of(0, 1, 2), 2L, ImmutableSet.of(0, 1, 4)); + doReturn(timeSegmentVsWorkerMap).when(completeKeyStatisticsInformation).getTimeSegmentVsWorkerMap(); + + final CyclicBarrier barrier = new CyclicBarrier(3); + target = spy(new WorkerSketchFetcher(workerClient, ClusterStatisticsMergeMode.SEQUENTIAL, 300_000_000)); + + // When fetching snapshots, return a mock and add it to queue + doAnswer(invocation -> { + ClusterByStatisticsSnapshot snapshot = mock(ClusterByStatisticsSnapshot.class); + snapshotQueue.add(snapshot); + barrier.await(); + return Futures.immediateFuture(snapshot); + }).when(workerClient).fetchClusterByStatisticsSnapshotForTimeChunk(any(), any(), anyInt(), anyLong()); + + CompletableFuture> eitherCompletableFuture = target.submitFetcherTask( + completeKeyStatisticsInformation, + ImmutableList.of("0", "1", "2", "3", "4"), + stageDefinition + ); + + // Assert that the final result is complete and all other sketches returned have been merged. + eitherCompletableFuture.join(); + Thread.sleep(1000); + + Assert.assertTrue(eitherCompletableFuture.isDone() && !eitherCompletableFuture.isCompletedExceptionally()); + Assert.assertFalse(snapshotQueue.isEmpty()); + // Verify that all statistics were added to controller. + snapshotQueue.stream().limit(3).forEach(snapshot -> { + verify(mergedClusterByStatisticsCollector1, times(1)).addAll(eq(snapshot)); + }); + snapshotQueue.stream().skip(3).limit(3).forEach(snapshot -> { + verify(mergedClusterByStatisticsCollector2, times(1)).addAll(eq(snapshot)); + }); + ClusterByPartitions expectedResult = + new ClusterByPartitions( + ImmutableList.of( + new ClusterByPartition(expectedPartitions1.get(0).getStart(), expectedPartitions2.get(0).getStart()), + new ClusterByPartition(expectedPartitions2.get(0).getStart(), expectedPartitions2.get(0).getEnd()) + ) + ); + // Check that the partitions returned by the merged collector is returned by the final future. + Assert.assertEquals(expectedResult, eitherCompletableFuture.get().valueOrThrow()); + } +} diff --git a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/kernel/controller/BaseControllerQueryKernelTest.java b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/kernel/controller/BaseControllerQueryKernelTest.java index c31d8c69fa99..f6b0ea9f08bb 100644 --- a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/kernel/controller/BaseControllerQueryKernelTest.java +++ b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/kernel/controller/BaseControllerQueryKernelTest.java @@ -21,6 +21,7 @@ import com.google.common.base.Preconditions; import com.google.common.collect.ImmutableMap; +import org.apache.druid.frame.key.ClusterByPartitions; import org.apache.druid.frame.key.KeyTestUtils; import org.apache.druid.frame.key.RowKey; import org.apache.druid.java.util.common.IAE; @@ -31,6 +32,7 @@ import org.apache.druid.msq.input.stage.StageInputSpec; import org.apache.druid.msq.input.stage.StageInputSpecSlicer; import org.apache.druid.msq.kernel.QueryDefinition; +import org.apache.druid.msq.kernel.StageDefinition; import org.apache.druid.msq.kernel.StageId; import org.apache.druid.msq.kernel.WorkerAssignmentStrategy; import org.apache.druid.msq.statistics.ClusterByStatisticsCollector; @@ -80,7 +82,7 @@ private ControllerQueryKernelTester(int numWorkers) public ControllerQueryKernelTester queryDefinition(QueryDefinition queryDefinition) { this.queryDefinition = Preconditions.checkNotNull(queryDefinition); - this.controllerQueryKernel = new ControllerQueryKernel(queryDefinition, 10_000_000); + this.controllerQueryKernel = new ControllerQueryKernel(queryDefinition); return this; } @@ -121,10 +123,10 @@ public ControllerQueryKernelTester setupStage( if (queryDefinition.getStageDefinition(stageNumber).mustGatherResultKeyStatistics()) { for (int i = 0; i < numWorkers; ++i) { - controllerQueryKernel.addResultKeyStatisticsForStageAndWorker( + controllerQueryKernel.addPartialKeyStatisticsForStageAndWorker( new StageId(queryDefinition.getQueryId(), stageNumber), i, - ClusterByStatisticsSnapshot.empty() + ClusterByStatisticsSnapshot.empty().partialKeyStatistics() ); } } else { @@ -238,7 +240,7 @@ public void finishStage(int stageNumber, boolean strict) controllerQueryKernel.finishStage(new StageId(queryDefinition.getQueryId(), stageNumber), strict); } - public void addResultKeyStatisticsForStageAndWorker(int stageNumber, int workerNumber) + public ClusterByStatisticsCollector addResultKeyStatisticsForStageAndWorker(int stageNumber, int workerNumber) { Preconditions.checkArgument(initialized); @@ -254,11 +256,12 @@ public void addResultKeyStatisticsForStageAndWorker(int stageNumber, int workerN keyStatsCollector.add(key, 1); } - controllerQueryKernel.addResultKeyStatisticsForStageAndWorker( + controllerQueryKernel.addPartialKeyStatisticsForStageAndWorker( new StageId(queryDefinition.getQueryId(), stageNumber), workerNumber, - keyStatsCollector.snapshot() + keyStatsCollector.snapshot().partialKeyStatistics() ); + return keyStatsCollector; } public void setResultsCompleteForStageAndWorker(int stageNumber, int workerNumber) @@ -271,6 +274,18 @@ public void setResultsCompleteForStageAndWorker(int stageNumber, int workerNumbe ); } + public void setPartitionBoundaries(int stageNumber, ClusterByStatisticsCollector clusterByStatisticsCollector) + { + Preconditions.checkArgument(initialized); + StageId stageId = new StageId(queryDefinition.getQueryId(), stageNumber); + StageDefinition stageDefinition = controllerQueryKernel.getStageDefinition(stageId); + ClusterByPartitions clusterByPartitions = + stageDefinition + .generatePartitionsForShuffle(clusterByStatisticsCollector) + .valueOrThrow(); + controllerQueryKernel.setClusterByPartitionBoundaries(stageId, clusterByPartitions); + } + public void failStage(int stageNumber) { Preconditions.checkArgument(initialized); diff --git a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/kernel/controller/ControllerQueryKernelTests.java b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/kernel/controller/ControllerQueryKernelTests.java index 3094bf344d2f..f2b6139e9985 100644 --- a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/kernel/controller/ControllerQueryKernelTests.java +++ b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/kernel/controller/ControllerQueryKernelTests.java @@ -20,6 +20,7 @@ package org.apache.druid.msq.kernel.controller; import com.google.common.collect.ImmutableSet; +import org.apache.druid.msq.statistics.ClusterByStatisticsCollector; import org.junit.Assert; import org.junit.Test; @@ -146,8 +147,13 @@ public void testCompleteDAGExecutionForMultipleWorkers() Assert.assertEquals(ImmutableSet.of(0), newStageNumbers); Assert.assertEquals(ImmutableSet.of(), effectivelyFinishedStageNumbers); controllerQueryKernelTester.startStage(0); - controllerQueryKernelTester.addResultKeyStatisticsForStageAndWorker(0, 0); - controllerQueryKernelTester.assertStagePhase(0, ControllerStagePhase.POST_READING); + ClusterByStatisticsCollector clusterByStatisticsCollector = + controllerQueryKernelTester.addResultKeyStatisticsForStageAndWorker( + 0, + 0 + ); + controllerQueryKernelTester.assertStagePhase(0, ControllerStagePhase.MERGING_STATISTICS); + controllerQueryKernelTester.setPartitionBoundaries(0, clusterByStatisticsCollector); controllerQueryKernelTester.setResultsCompleteForStageAndWorker(0, 0); controllerQueryKernelTester.assertStagePhase(0, ControllerStagePhase.RESULTS_READY); @@ -156,9 +162,20 @@ public void testCompleteDAGExecutionForMultipleWorkers() Assert.assertEquals(ImmutableSet.of(1), newStageNumbers); Assert.assertEquals(ImmutableSet.of(), effectivelyFinishedStageNumbers); controllerQueryKernelTester.startStage(1); - controllerQueryKernelTester.addResultKeyStatisticsForStageAndWorker(1, 0); + clusterByStatisticsCollector = + controllerQueryKernelTester.addResultKeyStatisticsForStageAndWorker( + 1, + 0 + ); controllerQueryKernelTester.assertStagePhase(1, ControllerStagePhase.READING_INPUT); - controllerQueryKernelTester.addResultKeyStatisticsForStageAndWorker(1, 1); + clusterByStatisticsCollector.addAll( + controllerQueryKernelTester.addResultKeyStatisticsForStageAndWorker( + 1, + 1 + ) + ); + controllerQueryKernelTester.assertStagePhase(1, ControllerStagePhase.MERGING_STATISTICS); + controllerQueryKernelTester.setPartitionBoundaries(1, clusterByStatisticsCollector); controllerQueryKernelTester.assertStagePhase(1, ControllerStagePhase.POST_READING); controllerQueryKernelTester.setResultsCompleteForStageAndWorker(1, 0); controllerQueryKernelTester.assertStagePhase(1, ControllerStagePhase.POST_READING); @@ -182,9 +199,19 @@ public void testCompleteDAGExecutionForMultipleWorkers() Assert.assertEquals(ImmutableSet.of(1), effectivelyFinishedStageNumbers); controllerQueryKernelTester.startStage(3); controllerQueryKernelTester.assertStagePhase(3, ControllerStagePhase.READING_INPUT); - controllerQueryKernelTester.addResultKeyStatisticsForStageAndWorker(3, 0); + ClusterByStatisticsCollector clusterByStatisticsCollector3 = + controllerQueryKernelTester.addResultKeyStatisticsForStageAndWorker( + 3, + 0 + ); controllerQueryKernelTester.assertStagePhase(3, ControllerStagePhase.READING_INPUT); - controllerQueryKernelTester.addResultKeyStatisticsForStageAndWorker(3, 1); + ClusterByStatisticsCollector clusterByStatisticsCollector4 = + controllerQueryKernelTester.addResultKeyStatisticsForStageAndWorker( + 3, + 1 + ); + controllerQueryKernelTester.assertStagePhase(3, ControllerStagePhase.MERGING_STATISTICS); + controllerQueryKernelTester.setPartitionBoundaries(3, clusterByStatisticsCollector3.addAll(clusterByStatisticsCollector4)); controllerQueryKernelTester.assertStagePhase(3, ControllerStagePhase.POST_READING); controllerQueryKernelTester.setResultsCompleteForStageAndWorker(3, 0); controllerQueryKernelTester.assertStagePhase(3, ControllerStagePhase.POST_READING); @@ -217,11 +244,21 @@ public void testTransitionsInShufflingStagesAndMultipleWorkers() controllerQueryKernelTester.createAndGetNewStageNumbers(); controllerQueryKernelTester.startStage(0); - controllerQueryKernelTester.addResultKeyStatisticsForStageAndWorker(0, 0); + ClusterByStatisticsCollector clusterByStatisticsCollector = + controllerQueryKernelTester.addResultKeyStatisticsForStageAndWorker( + 0, + 0 + ); controllerQueryKernelTester.assertStagePhase(0, ControllerStagePhase.READING_INPUT); - controllerQueryKernelTester.addResultKeyStatisticsForStageAndWorker(0, 1); - controllerQueryKernelTester.assertStagePhase(0, ControllerStagePhase.POST_READING); + clusterByStatisticsCollector.addAll( + controllerQueryKernelTester.addResultKeyStatisticsForStageAndWorker( + 0, + 1 + ) + ); + controllerQueryKernelTester.assertStagePhase(0, ControllerStagePhase.MERGING_STATISTICS); + controllerQueryKernelTester.setPartitionBoundaries(0, clusterByStatisticsCollector); controllerQueryKernelTester.setResultsCompleteForStageAndWorker(0, 0); controllerQueryKernelTester.assertStagePhase(0, ControllerStagePhase.POST_READING); diff --git a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/statistics/PartialKeyStatisticsInformationSerdeTest.java b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/statistics/PartialKeyStatisticsInformationSerdeTest.java new file mode 100644 index 000000000000..c23106c0299d --- /dev/null +++ b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/statistics/PartialKeyStatisticsInformationSerdeTest.java @@ -0,0 +1,62 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.msq.statistics; + +import com.fasterxml.jackson.core.JsonParser; +import com.fasterxml.jackson.core.JsonProcessingException; +import com.fasterxml.jackson.databind.ObjectMapper; +import com.google.common.collect.ImmutableSet; +import org.apache.druid.msq.guice.MSQIndexingModule; +import org.apache.druid.segment.TestHelper; +import org.junit.Assert; +import org.junit.Before; +import org.junit.Test; + +public class PartialKeyStatisticsInformationSerdeTest +{ + private ObjectMapper objectMapper; + + @Before + public void setUp() + { + objectMapper = TestHelper.makeJsonMapper(); + objectMapper.registerModules(new MSQIndexingModule().getJacksonModules()); + objectMapper.enable(JsonParser.Feature.STRICT_DUPLICATE_DETECTION); + } + + @Test + public void testSerde() throws JsonProcessingException + { + PartialKeyStatisticsInformation partialInformation = new PartialKeyStatisticsInformation( + ImmutableSet.of(2L, 3L), + false, + 0.0 + ); + + final String json = objectMapper.writeValueAsString(partialInformation); + final PartialKeyStatisticsInformation deserializedKeyStatistics = objectMapper.readValue( + json, + PartialKeyStatisticsInformation.class + ); + Assert.assertEquals(json, partialInformation.getTimeSegments(), deserializedKeyStatistics.getTimeSegments()); + Assert.assertEquals(json, partialInformation.hasMultipleValues(), deserializedKeyStatistics.hasMultipleValues()); + Assert.assertEquals(json, partialInformation.getBytesRetained(), deserializedKeyStatistics.getBytesRetained(), 0); + } +} diff --git a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/test/MSQTestControllerClient.java b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/test/MSQTestControllerClient.java index 5b088b71d5bb..06b0bae29690 100644 --- a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/test/MSQTestControllerClient.java +++ b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/test/MSQTestControllerClient.java @@ -25,7 +25,7 @@ import org.apache.druid.msq.exec.ControllerClient; import org.apache.druid.msq.indexing.error.MSQErrorReport; import org.apache.druid.msq.kernel.StageId; -import org.apache.druid.msq.statistics.ClusterByStatisticsSnapshot; +import org.apache.druid.msq.statistics.PartialKeyStatisticsInformation; import javax.annotation.Nullable; import java.util.List; @@ -40,17 +40,17 @@ public MSQTestControllerClient(Controller controller) } @Override - public void postKeyStatistics( + public void postPartialKeyStatistics( StageId stageId, int workerNumber, - ClusterByStatisticsSnapshot keyStatistics + PartialKeyStatisticsInformation partialKeyStatisticsInformation ) { try { - controller.updateStatus(stageId.getStageNumber(), workerNumber, keyStatistics); + controller.updatePartialKeyStatisticsInformation(stageId.getStageNumber(), workerNumber, partialKeyStatisticsInformation); } catch (Exception e) { - throw new ISE(e, "unable to post key statistics"); + throw new ISE(e, "unable to post partial key statistics"); } } diff --git a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/test/MSQTestWorkerClient.java b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/test/MSQTestWorkerClient.java index 094c1bd16cf7..ae892c34500a 100644 --- a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/test/MSQTestWorkerClient.java +++ b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/test/MSQTestWorkerClient.java @@ -29,6 +29,7 @@ import org.apache.druid.msq.exec.WorkerClient; import org.apache.druid.msq.kernel.StageId; import org.apache.druid.msq.kernel.WorkOrder; +import org.apache.druid.msq.statistics.ClusterByStatisticsSnapshot; import java.io.InputStream; import java.util.Arrays; @@ -50,6 +51,29 @@ public ListenableFuture postWorkOrder(String workerTaskId, WorkOrder workO return Futures.immediateFuture(null); } + @Override + public ListenableFuture fetchClusterByStatisticsSnapshot( + String workerTaskId, + String queryId, + int stageNumber + ) + { + StageId stageId = new StageId(queryId, stageNumber); + return Futures.immediateFuture(inMemoryWorkers.get(workerTaskId).fetchStatisticsSnapshot(stageId)); + } + + @Override + public ListenableFuture fetchClusterByStatisticsSnapshotForTimeChunk( + String workerTaskId, + String queryId, + int stageNumber, + long timeChunk + ) + { + StageId stageId = new StageId(queryId, stageNumber); + return Futures.immediateFuture(inMemoryWorkers.get(workerTaskId).fetchStatisticsSnapshotForTimeChunk(stageId, timeChunk)); + } + @Override public ListenableFuture postResultPartitionBoundaries( String workerTaskId, diff --git a/integration-tests-ex/README.md b/integration-tests-ex/README.md index 9c29ec101fe9..3a17c6e7c700 100644 --- a/integration-tests-ex/README.md +++ b/integration-tests-ex/README.md @@ -71,7 +71,7 @@ Start the cluster: ```bash cd $DRUID_DEV/integration-tests-ex/cases -./cluster.sh up +./cluster.sh up ``` Where `` is one of the test categories. Then launch the diff --git a/integration-tests-ex/cases/pom.xml b/integration-tests-ex/cases/pom.xml index cf781f6f8885..92a632d1f14c 100644 --- a/integration-tests-ex/cases/pom.xml +++ b/integration-tests-ex/cases/pom.xml @@ -178,6 +178,11 @@ + + org.apache.druid + druid-sql + 25.0.0-SNAPSHOT + org.apache.druid.extensions druid-multi-stage-query diff --git a/integration-tests-ex/cases/src/test/java/org/apache/druid/testsEx/msq/ITKeyStatisticsSketchMergeMode.java b/integration-tests-ex/cases/src/test/java/org/apache/druid/testsEx/msq/ITKeyStatisticsSketchMergeMode.java new file mode 100644 index 000000000000..c0f22f2c762d --- /dev/null +++ b/integration-tests-ex/cases/src/test/java/org/apache/druid/testsEx/msq/ITKeyStatisticsSketchMergeMode.java @@ -0,0 +1,206 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.testsEx.msq; + +import com.fasterxml.jackson.databind.ObjectMapper; +import com.google.common.collect.ImmutableMap; +import com.google.inject.Inject; +import org.apache.druid.java.util.common.StringUtils; +import org.apache.druid.msq.exec.ClusterStatisticsMergeMode; +import org.apache.druid.msq.sql.SqlTaskStatus; +import org.apache.druid.msq.util.MultiStageQueryContext; +import org.apache.druid.sql.http.SqlQuery; +import org.apache.druid.testing.IntegrationTestingConfig; +import org.apache.druid.testing.clients.CoordinatorResourceTestClient; +import org.apache.druid.testing.clients.SqlResourceTestClient; +import org.apache.druid.testing.utils.DataLoaderHelper; +import org.apache.druid.testing.utils.MsqTestQueryHelper; +import org.apache.druid.testsEx.categories.MultiStageQuery; +import org.apache.druid.testsEx.config.DruidTestRunner; +import org.junit.Assert; +import org.junit.Test; +import org.junit.experimental.categories.Category; +import org.junit.runner.RunWith; + +@RunWith(DruidTestRunner.class) +@Category(MultiStageQuery.class) +public class ITKeyStatisticsSketchMergeMode +{ + @Inject + private MsqTestQueryHelper msqHelper; + + @Inject + private SqlResourceTestClient msqClient; + + @Inject + private IntegrationTestingConfig config; + + @Inject + private ObjectMapper jsonMapper; + + @Inject + private DataLoaderHelper dataLoaderHelper; + + @Inject + private CoordinatorResourceTestClient coordinatorClient; + + private static final String QUERY_FILE = "/multi-stage-query/wikipedia_msq_select_query1.json"; + + + @Test + public void testMsqIngestionParallelMerging() throws Exception + { + String datasource = "dst"; + + // Clear up the datasource from the previous runs + coordinatorClient.unloadSegmentsForDataSource(datasource); + + String queryLocal = + StringUtils.format( + "INSERT INTO %s\n" + + "SELECT\n" + + " TIME_PARSE(\"timestamp\") AS __time,\n" + + " isRobot,\n" + + " diffUrl,\n" + + " added,\n" + + " countryIsoCode,\n" + + " regionName,\n" + + " channel,\n" + + " flags,\n" + + " delta,\n" + + " isUnpatrolled,\n" + + " isNew,\n" + + " deltaBucket,\n" + + " isMinor,\n" + + " isAnonymous,\n" + + " deleted,\n" + + " cityName,\n" + + " metroCode,\n" + + " namespace,\n" + + " comment,\n" + + " page,\n" + + " commentLength,\n" + + " countryName,\n" + + " user,\n" + + " regionIsoCode\n" + + "FROM TABLE(\n" + + " EXTERN(\n" + + " '{\"type\":\"local\",\"files\":[\"/resources/data/batch_index/json/wikipedia_index_data1.json\"]}',\n" + + " '{\"type\":\"json\"}',\n" + + " '[{\"type\":\"string\",\"name\":\"timestamp\"},{\"type\":\"string\",\"name\":\"isRobot\"},{\"type\":\"string\",\"name\":\"diffUrl\"},{\"type\":\"long\",\"name\":\"added\"},{\"type\":\"string\",\"name\":\"countryIsoCode\"},{\"type\":\"string\",\"name\":\"regionName\"},{\"type\":\"string\",\"name\":\"channel\"},{\"type\":\"string\",\"name\":\"flags\"},{\"type\":\"long\",\"name\":\"delta\"},{\"type\":\"string\",\"name\":\"isUnpatrolled\"},{\"type\":\"string\",\"name\":\"isNew\"},{\"type\":\"double\",\"name\":\"deltaBucket\"},{\"type\":\"string\",\"name\":\"isMinor\"},{\"type\":\"string\",\"name\":\"isAnonymous\"},{\"type\":\"long\",\"name\":\"deleted\"},{\"type\":\"string\",\"name\":\"cityName\"},{\"type\":\"long\",\"name\":\"metroCode\"},{\"type\":\"string\",\"name\":\"namespace\"},{\"type\":\"string\",\"name\":\"comment\"},{\"type\":\"string\",\"name\":\"page\"},{\"type\":\"long\",\"name\":\"commentLength\"},{\"type\":\"string\",\"name\":\"countryName\"},{\"type\":\"string\",\"name\":\"user\"},{\"type\":\"string\",\"name\":\"regionIsoCode\"}]'\n" + + " )\n" + + ")\n" + + "PARTITIONED BY DAY\n" + + "CLUSTERED BY \"__time\"", + datasource + ); + + ImmutableMap context = ImmutableMap.of( + MultiStageQueryContext.CTX_CLUSTER_STATISTICS_MERGE_MODE, + ClusterStatisticsMergeMode.PARALLEL + ); + + // Submit the task and wait for the datasource to get loaded + SqlQuery sqlQuery = new SqlQuery(queryLocal, null, false, false, false, context, null); + SqlTaskStatus sqlTaskStatus = msqHelper.submitMsqTask(sqlQuery); + + if (sqlTaskStatus.getState().isFailure()) { + Assert.fail(StringUtils.format( + "Unable to start the task successfully.\nPossible exception: %s", + sqlTaskStatus.getError() + )); + } + + msqHelper.pollTaskIdForCompletion(sqlTaskStatus.getTaskId()); + dataLoaderHelper.waitUntilDatasourceIsReady(datasource); + + msqHelper.testQueriesFromFile(QUERY_FILE, datasource); + } + + @Test + public void testMsqIngestionSequentialMerging() throws Exception + { + String datasource = "dst"; + + // Clear up the datasource from the previous runs + coordinatorClient.unloadSegmentsForDataSource(datasource); + + String queryLocal = + StringUtils.format( + "INSERT INTO %s\n" + + "SELECT\n" + + " TIME_PARSE(\"timestamp\") AS __time,\n" + + " isRobot,\n" + + " diffUrl,\n" + + " added,\n" + + " countryIsoCode,\n" + + " regionName,\n" + + " channel,\n" + + " flags,\n" + + " delta,\n" + + " isUnpatrolled,\n" + + " isNew,\n" + + " deltaBucket,\n" + + " isMinor,\n" + + " isAnonymous,\n" + + " deleted,\n" + + " cityName,\n" + + " metroCode,\n" + + " namespace,\n" + + " comment,\n" + + " page,\n" + + " commentLength,\n" + + " countryName,\n" + + " user,\n" + + " regionIsoCode\n" + + "FROM TABLE(\n" + + " EXTERN(\n" + + " '{\"type\":\"local\",\"files\":[\"/resources/data/batch_index/json/wikipedia_index_data1.json\"]}',\n" + + " '{\"type\":\"json\"}',\n" + + " '[{\"type\":\"string\",\"name\":\"timestamp\"},{\"type\":\"string\",\"name\":\"isRobot\"},{\"type\":\"string\",\"name\":\"diffUrl\"},{\"type\":\"long\",\"name\":\"added\"},{\"type\":\"string\",\"name\":\"countryIsoCode\"},{\"type\":\"string\",\"name\":\"regionName\"},{\"type\":\"string\",\"name\":\"channel\"},{\"type\":\"string\",\"name\":\"flags\"},{\"type\":\"long\",\"name\":\"delta\"},{\"type\":\"string\",\"name\":\"isUnpatrolled\"},{\"type\":\"string\",\"name\":\"isNew\"},{\"type\":\"double\",\"name\":\"deltaBucket\"},{\"type\":\"string\",\"name\":\"isMinor\"},{\"type\":\"string\",\"name\":\"isAnonymous\"},{\"type\":\"long\",\"name\":\"deleted\"},{\"type\":\"string\",\"name\":\"cityName\"},{\"type\":\"long\",\"name\":\"metroCode\"},{\"type\":\"string\",\"name\":\"namespace\"},{\"type\":\"string\",\"name\":\"comment\"},{\"type\":\"string\",\"name\":\"page\"},{\"type\":\"long\",\"name\":\"commentLength\"},{\"type\":\"string\",\"name\":\"countryName\"},{\"type\":\"string\",\"name\":\"user\"},{\"type\":\"string\",\"name\":\"regionIsoCode\"}]'\n" + + " )\n" + + ")\n" + + "PARTITIONED BY DAY\n" + + "CLUSTERED BY \"__time\"", + datasource + ); + + ImmutableMap context = ImmutableMap.of( + MultiStageQueryContext.CTX_CLUSTER_STATISTICS_MERGE_MODE, + ClusterStatisticsMergeMode.SEQUENTIAL + ); + + // Submit the task and wait for the datasource to get loaded + SqlQuery sqlQuery = new SqlQuery(queryLocal, null, false, false, false, context, null); + SqlTaskStatus sqlTaskStatus = msqHelper.submitMsqTask(sqlQuery); + + if (sqlTaskStatus.getState().isFailure()) { + Assert.fail(StringUtils.format( + "Unable to start the task successfully.\nPossible exception: %s", + sqlTaskStatus.getError() + )); + } + + msqHelper.pollTaskIdForCompletion(sqlTaskStatus.getTaskId()); + dataLoaderHelper.waitUntilDatasourceIsReady(datasource); + + msqHelper.testQueriesFromFile(QUERY_FILE, datasource); + } +} diff --git a/processing/src/main/java/org/apache/druid/frame/key/RowKeyReader.java b/processing/src/main/java/org/apache/druid/frame/key/RowKeyReader.java index d3ca6d035c49..3102b103c9c7 100644 --- a/processing/src/main/java/org/apache/druid/frame/key/RowKeyReader.java +++ b/processing/src/main/java/org/apache/druid/frame/key/RowKeyReader.java @@ -130,6 +130,28 @@ public boolean hasMultipleValues(final RowKey key, final int fieldNumber) } } + /** + * Trims the key reader to a particular fieldCount. Used to read keys trimmed by {@link #trim(RowKey, int)}. + */ + public RowKeyReader trimmedKeyReader(int trimmedFieldCount) + { + final RowSignature.Builder newSignature = RowSignature.builder(); + + if (trimmedFieldCount > signature.size()) { + throw new IAE("Cannot trim to [%,d] fields, only have [%,d] fields", trimmedFieldCount, signature); + } + + for (int i = 0; i < trimmedFieldCount; i++) { + final String columnName = signature.getColumnName(i); + final ColumnType columnType = + Preconditions.checkNotNull(signature.getColumnType(i).orElse(null), "Type for column [%s]", columnName); + + newSignature.add(columnName, columnType); + } + + return RowKeyReader.create(newSignature.build()); + } + /** * Trim a key to a particular fieldCount. The returned key may be a copy, but is not guaranteed to be. */ diff --git a/processing/src/test/java/org/apache/druid/frame/key/RowKeyReaderTest.java b/processing/src/test/java/org/apache/druid/frame/key/RowKeyReaderTest.java index 7a9131729539..48d940384270 100644 --- a/processing/src/test/java/org/apache/druid/frame/key/RowKeyReaderTest.java +++ b/processing/src/test/java/org/apache/druid/frame/key/RowKeyReaderTest.java @@ -30,6 +30,7 @@ import org.junit.internal.matchers.ThrowableMessageMatcher; import java.util.Arrays; +import java.util.Collections; import java.util.List; import java.util.stream.IntStream; @@ -144,4 +145,41 @@ public void test_trim_beyondFullLength() MatcherAssert.assertThat(e, ThrowableMessageMatcher.hasMessage(CoreMatchers.containsString("Cannot trim"))); } + + @Test + public void test_trimmedKeyReader_zero() + { + RowKey trimmedKey = keyReader.trim(key, 0); + RowKeyReader trimmedKeyReader = keyReader.trimmedKeyReader(0); + + Assert.assertEquals( + Collections.emptyList(), + trimmedKeyReader.read(trimmedKey) + ); + } + + @Test + public void test_trimmedKeyReader_one() + { + RowKey trimmedKey = keyReader.trim(key, 1); + RowKeyReader trimmedKeyReader = keyReader.trimmedKeyReader(1); + + Assert.assertEquals( + objects.subList(0, 1), + trimmedKeyReader.read(trimmedKey) + ); + } + + @Test + public void test_trimmedKeyReader_oneLessThanFullLength() + { + final int numFields = signature.size() - 1; + RowKey trimmedKey = keyReader.trim(key, numFields); + RowKeyReader trimmedKeyReader = keyReader.trimmedKeyReader(numFields); + + Assert.assertEquals( + objects.subList(0, numFields), + trimmedKeyReader.read(trimmedKey) + ); + } } diff --git a/website/.spelling b/website/.spelling index c27c4d8ccc80..c2eb66104ee8 100644 --- a/website/.spelling +++ b/website/.spelling @@ -68,6 +68,9 @@ Double.NEGATIVE_INFINITY Double.NEGATIVE_INFINITY. Double.POSITIVE_INFINITY Double.POSITIVE_INFINITY. +downsampled +downsamples +downsampling Dropwizard dropwizard DruidInputSource From 9b8cdc6197f9aeec7008c880e64f7d8916c10c9a Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Tue, 22 Nov 2022 11:09:37 +0530 Subject: [PATCH 015/141] compute and pass middle manager runtime properties to run-druid supervise script changes to process java opts array use argparse, leave free memory, logging --- examples/bin/druid-quickstart.py | 337 -------------- examples/bin/run-druid | 27 +- examples/bin/start-druid | 2 +- examples/bin/start-druid-main | 439 ++++++++++++++++++ examples/bin/supervise | 6 +- .../middleManager/runtime.properties | 6 +- 6 files changed, 467 insertions(+), 350 deletions(-) delete mode 100644 examples/bin/druid-quickstart.py create mode 100644 examples/bin/start-druid-main diff --git a/examples/bin/druid-quickstart.py b/examples/bin/druid-quickstart.py deleted file mode 100644 index 0b2da6658771..000000000000 --- a/examples/bin/druid-quickstart.py +++ /dev/null @@ -1,337 +0,0 @@ -import sys -import os -import psutil -import pathlib -from pathlib import Path - -QUICKSTART_BASE_CONFIG_PATH = "conf/druid/single-server/quickstart" -HELP_ARG_IDENTIFIER = "help" -COMPUTE_ONLY_ARG_IDENTIFIER = "computeOnly" -RUN_ZK_IDENTIFIER = "runZk" -ROOT_CONFIG_PATH_IDENTIFIER = "rootConfigPath" -SERVICES_IDENTIFIER = "services" -MEMORY_ARG_IDENTIFIER = "totalMemory" -MEMORY_GIGABYTES_IDENTIFIER = "g" -MEMORY_MEGABYTES_IDENTIFIER = "m" -ARG_SEPARATOR = "=" -SERVICE_SEPARATOR = "," - -BROKER_SERVICE_NAME = "broker" -ROUTER_SERVICE_NAME = "router" -COORDINATOR_SERVICE_NAME = "coordinator-overlord" -HISTORICAL_SERVICE_NAME = "historical" -MIDDLE_MANAGER_SERVICE_NAME = "middleManager" - -DEFAULT_SERVICES = [ - BROKER_SERVICE_NAME, - ROUTER_SERVICE_NAME, - COORDINATOR_SERVICE_NAME, - HISTORICAL_SERVICE_NAME, - MIDDLE_MANAGER_SERVICE_NAME -] - -SERVICE_MEMORY_DISTRIBUTION_WEIGHT = { - MIDDLE_MANAGER_SERVICE_NAME: 1, - ROUTER_SERVICE_NAME: 2, - COORDINATOR_SERVICE_NAME: 36, - BROKER_SERVICE_NAME: 56, - HISTORICAL_SERVICE_NAME: 90 -} - -SERVICE_MEMORY_LOWER_BOUND = { - MIDDLE_MANAGER_SERVICE_NAME: 64, - ROUTER_SERVICE_NAME: 128 -} - -SERVICE_MEMORY_HEAP_PERCENTAGE = { - MIDDLE_MANAGER_SERVICE_NAME: 1, - ROUTER_SERVICE_NAME: 1, - COORDINATOR_SERVICE_NAME: 1, - BROKER_SERVICE_NAME: 0.60, - HISTORICAL_SERVICE_NAME: 0.40 -} - -def check_argument_type(argument, type): - split_args = argument.split(ARG_SEPARATOR) - return split_args[0] == type - -def get_argument_value(argument): - split_args = argument.split(ARG_SEPARATOR) - return split_args[1] - -def parse_arguments(): - service_list = [] - root_config_path = "" - total_memory = "" - compute_only = False - run_zk = False - - for argument in sys.argv[1:]: - if check_argument_type(argument, COMPUTE_ONLY_ARG_IDENTIFIER): - compute_only = True - elif check_argument_type(argument, RUN_ZK_IDENTIFIER): - run_zk = True - elif check_argument_type(argument, ROOT_CONFIG_PATH_IDENTIFIER): - root_config_path = os.path.join(os.getcwd(), get_argument_value(argument)) - if os.path.exists(root_config_path) is False: - raise Exception(f'rootConfigPath `{root_config_path}` doesn\'t exist') - elif (check_argument_type(argument, MEMORY_ARG_IDENTIFIER)): - total_memory = get_argument_value(argument) - elif (check_argument_type(argument, SERVICES_IDENTIFIER)): - split_args = argument.split(ARG_SEPARATOR) - services_arg = split_args[1] - services = services_arg.split(SERVICE_SEPARATOR) - - for service in services: - if service not in DEFAULT_SERVICES: - raise Exception(f'{service} is not a valid service name, should be one of {DEFAULT_SERVICES}') - - if service in service_list: - raise Exception(f'{service} is specified multiple times') - - service_list.append(service) - - if len(service_list) == 0: - # start all services - service_list = DEFAULT_SERVICES - run_zk = True - - return root_config_path, total_memory, service_list, run_zk, compute_only - -def print_startup_config(service_list, root_config_path, run_zk): - print(f'starting {service_list}, using config from {root_config_path}') - if run_zk: - zk_config_path = pathlib.Path(f'{os.getcwd()}/../conf/zk').resolve() - print(f'starting zk, using default config from {zk_config_path}') - print('\n') - -def should_compute_memory(root_config_path, total_memory, service_list): - # if jvm file is present for any of the services - # it should be present for all services and totalMemory should not be specified - # if totalMemory is given, jvm file shouldn't be present for any service - - jvm_config_count = 0 - for service in service_list: - if Path(f'{root_config_path}/{service}/jvm.config').is_file(): - jvm_config_count += 1 - elif jvm_config_count > 0: - raise Exception('jvm.config file is missing for service {service}, jvm.config should be specified for all the services or none') - - if jvm_config_count > 0 and (jvm_config_count != len(service_list) or total_memory != ""): - if jvm_config_count != len(service_list): - raise Exception("jvm.config file should be present for all services or none") - if total_memory != "": - raise Exception("If jvm.config is given for services, `totalMemory` argument shouldn't be specified") - - return jvm_config_count == 0 - -def compute_system_memory(): - system_memory = psutil.virtual_memory().total # mem in bytes - memory_for_druid = int((system_memory * 0.8) / (1024 * 1024)) - return memory_for_druid - -def convert_total_memory_string(memory): - if memory == "": - computed_memory = compute_system_memory() - print(f'`{MEMORY_ARG_IDENTIFIER}` argument is not specified, Druid will use 80% of system memory: {computed_memory}m') - return computed_memory - elif memory.endswith(MEMORY_MEGABYTES_IDENTIFIER): - return int(memory[:-1]) - elif memory.endswith(MEMORY_GIGABYTES_IDENTIFIER): - return 1024 * int(memory[:-1]) - else: - raise Exception('Incorrect format for totalMemory argument, expected format is ') - -def build_memory_config_string(heap_memory, direct_memory): - if direct_memory == 0: - return f'-Xms{heap_memory}m -Xmx{heap_memory}m' - return f'-Xms{heap_memory}m -Xmx{heap_memory}m -XX:MaxDirectMemorySize={direct_memory}m' - -def distribute_memory_over_services(service_list, total_memory): - service_memory_config = {} - - memory_weight_sum = 0 - for service in service_list: - memory_weight_sum += SERVICE_MEMORY_DISTRIBUTION_WEIGHT.get(service) - - multiplier = total_memory / memory_weight_sum - - lower_bound_memory_allocation = 0 - allocated_services = set() - - for service in service_list: - allocated_memory = SERVICE_MEMORY_DISTRIBUTION_WEIGHT.get(service) * multiplier - if service in SERVICE_MEMORY_LOWER_BOUND and allocated_memory < SERVICE_MEMORY_LOWER_BOUND.get(service): - allocated_memory = SERVICE_MEMORY_LOWER_BOUND.get(service) - heap_memory = SERVICE_MEMORY_HEAP_PERCENTAGE.get(service) * allocated_memory - direct_memory = allocated_memory - heap_memory - service_memory_config[service] = build_memory_config_string(int(heap_memory), int(direct_memory)) - lower_bound_memory_allocation += allocated_memory - allocated_services.add(service) - - if lower_bound_memory_allocation > 0: - # compute the multiplier again for remaing services - memory_weight_sum = 0 - for service in service_list: - if service in allocated_services: - continue - memory_weight_sum += SERVICE_MEMORY_DISTRIBUTION_WEIGHT.get(service) - multiplier = (total_memory - lower_bound_memory_allocation) / memory_weight_sum - - for service in service_list: - if service in allocated_services: - continue - allocated_memory = SERVICE_MEMORY_DISTRIBUTION_WEIGHT.get(service) * multiplier - if service in SERVICE_MEMORY_LOWER_BOUND and allocated_memory < SERVICE_MEMORY_LOWER_BOUND.get(service): - allocated_memory = SERVICE_MEMORY_LOWER_BOUND.get(service) - - heap_memory = SERVICE_MEMORY_HEAP_PERCENTAGE.get(service) * allocated_memory - direct_memory = allocated_memory - heap_memory - service_memory_config[service] = build_memory_config_string(int(heap_memory), int(direct_memory)) - - print(f'\nMemory distribution for services:') - for key, value in service_memory_config.items(): - print(f'{key}, memory_config: {value}') - print('\n') - - return service_memory_config - -def build_supervise_script_arguments(service_list, service_memory_config, root_config_path, run_zk): - argument_list = [] - - argument_list.append("\":verify bin/verify-java\"") - argument_list.append("\":verify bin/verify-default-ports\"") - argument_list.append("\":notify bin/greet\"") - argument_list.append("\":kill-timeout 10\"") - - if run_zk: - argument_list.append("\"!p10 zk bin/run-zk conf\"") - - for service in service_list: - prefix = '' - if service == MIDDLE_MANAGER_SERVICE_NAME: - prefix = '!p90 ' - - jvm_args = service_memory_config.get(service) - - if jvm_args is None: - argument_list.append(f'\"{prefix}{service} bin/run-druid {service} {root_config_path}\"') - else: - argument_list.append(f'\"{prefix}{service} bin/run-druid {service} {root_config_path} \'{jvm_args}\'\"') - - print('Supervise script command:') - for item in argument_list: - print(item) - - print('\n') - - return ",".join(argument_list) - -def display_help(): - text = """ - Usage: start-druid [options] - - where options include: - totalMemory= - Integer value is supported with 'm' or 'g' suffix. - Memory for druid services, if totalMemory is not specified - 80 percent of system memory is used. - Note, if service specific jvm config is present, - shouldn't be specified. - should be greater than equals 3g - rootConfigPath= - Directory containing common and service specific - properties to be overridden, this directory must contain '_common' - directory with 'common.jvm.config' & 'common.runtime.properties' - files. - If is not specified, config from - 'conf/druid/single-server/quickstart' directory is used. - This path is relative to current working directory - services= - Value is comma separated string. - List of services to be started, should be a subset of - [broker, router, middleManager, historical, coordinator-overlord]. - If runtime or jvm properties are to be overridden, they should be - kept within /. - If is not specified config files can also be - placed within `conf/druid/single-server/quickstart/` - directory. - Note, if jvm.config file is present for one of the services, - it must be present for all services. - If argument is not given, all services - alongwith zookeeper is started. - runZk - Specification to run zookeeper, zk config is picked up from conf/zk. - computeOnly - Validate the arguments and display memory distribution for services. - - sample usage: - start-druid - Start up all the services (including zk) - using 80% of system memory. - start-druid totalMemory=100g - Start up all the services (including zk) - using the given memory. - start-druid totalMemory=100g computeOnly - Compute memory distribution and validate - arguments for starting all the services. - start-druid totalMemory=100g services=broker,router - Start `broker` & `router` service, using `100g` of memory. - Read config from conf/druid/single-server/quickstart. - start-druid totalMemory=100g rootConfigPath=conf/druid/single-server/custom services=broker,router - Start `broker` & `router` service, using 100g of memory. - Read config from . - Since is specified, exception is thrown if - jvm.config is present for any of the services. - start-druid rootConfigPath=conf/druid/single-server/custom services=broker,router - Start `broker` & `router` service, - using 80% of system memory. - If jvm.config is specified for both the - services within /, - memory distribution is not calculated. - If jvm.config is present for either of the services, - exception is thrown. - If jvm.config is not present for both of the services, - memory distribution is calculated. - start-druid totalMemory=100g rootConfigPath=conf/druid/single-server/custom services=broker,router runZk - Start zookeeper alongwith other services. - zk config is read from conf/zk. - """ - - print(text) - - -def main(): - for argument in sys.argv[1:]: - if check_argument_type(argument, HELP_ARG_IDENTIFIER): - display_help() - return - - print("Druid quickstart\n") - - root_config_path, total_memory, service_list, run_zk, compute_only = parse_arguments() - - # change directory to bin - os.chdir(os.path.dirname(sys.argv[0])) - - if root_config_path == "": - root_config_path = pathlib.Path(f'{os.getcwd()}/../{QUICKSTART_BASE_CONFIG_PATH}').resolve() - - print_startup_config(service_list, root_config_path, run_zk) - - service_memory_config = {} - if (should_compute_memory(root_config_path, total_memory, service_list)): - memory_in_mega_bytes = convert_total_memory_string(total_memory) - service_memory_config = distribute_memory_over_services(service_list, memory_in_mega_bytes) - else: - print('not computing memory distribution, reading memory specification from service jvm.config\n') - - script_arguments = build_supervise_script_arguments(service_list, service_memory_config, root_config_path, run_zk) - - if compute_only: - return - - os.system(f'exec ./supervise -a {script_arguments}') - -if __name__ == '__main__': - main() diff --git a/examples/bin/run-druid b/examples/bin/run-druid index 258f8376d0d2..9e9f93d90deb 100755 --- a/examples/bin/run-druid +++ b/examples/bin/run-druid @@ -17,7 +17,7 @@ # specific language governing permissions and limitations # under the License. -if [ "$#" -gt 3 ] || [ "$#" -eq 0 ] +if [ "$#" -gt 5 ] || [ "$#" -eq 0 ] then >&2 echo "usage: $0 [conf-dir]" exit 1 @@ -64,18 +64,29 @@ fi cd "$WHEREAMI/.." -if [ "$#" -eq 3 ] -then - -elif [ "$#" -eq 4 ] +if [ "$#" -eq 3 ] || [ "$#" -eq 5 ] then + # args: or JVMARGS=`cat "$CONFDIR/_common/common.jvm.config" | xargs` JVMARGS+=' ' JVMARGS+=$3 - exec "$WHEREAMI"/run-java -Ddruid.node.type=$1 "-Ddruid.log.path=$LOG_DIR" $JVMARGS \ - -cp "$CONFDIR"/"$WHATAMI":"$CONFDIR"/_common:"$CONFDIR"/_common/hadoop-xml:"$CONFDIR"/../_common:"$CONFDIR"/../_common/hadoop-xml:"$WHEREAMI/../lib/*" $MAIN_CLASS + if [ "$#" -eq 3 ] + then + exec "$WHEREAMI"/run-java -Ddruid.node.type=$1 $3 "-Ddruid.log.path=$LOG_DIR" $JVMARGS \ + -cp "$CONFDIR"/"$WHATAMI":"$CONFDIR"/_common:"$CONFDIR"/_common/hadoop-xml:"$CONFDIR"/../_common:"$CONFDIR"/../_common/hadoop-xml:"$WHEREAMI/../lib/*" org.apache.druid.cli.Main server $SERVER_NAME + else + exec "$WHEREAMI"/run-java -Ddruid.node.type=$1 $3 "-Ddruid.log.path=$LOG_DIR" $4 $5 $JVMARGS \ + -cp "$CONFDIR"/"$WHATAMI":"$CONFDIR"/_common:"$CONFDIR"/_common/hadoop-xml:"$CONFDIR"/../_common:"$CONFDIR"/../_common/hadoop-xml:"$WHEREAMI/../lib/*" org.apache.druid.cli.Main server $SERVER_NAME + fi +elif [ "$#" -eq 4 ] +then + # args: + exec "$WHEREAMI"/run-java -Ddruid.node.type=$1 $3 "-Ddruid.log.path=$LOG_DIR" `cat "$CONFDIR"/"$WHATAMI"/jvm.config | xargs` \ + -cp "$CONFDIR"/"$WHATAMI":"$CONFDIR"/_common:"$CONFDIR"/_common/hadoop-xml:"$CONFDIR"/../_common:"$CONFDIR"/../_common/hadoop-xml:"$WHEREAMI/../lib/*" org.apache.druid.cli.Main server $SERVER_NAME else + # args: exec "$WHEREAMI"/run-java -Ddruid.node.type=$1 "-Ddruid.log.path=$LOG_DIR" `cat "$CONFDIR"/"$WHATAMI"/jvm.config | xargs` \ - -cp "$CONFDIR"/"$WHATAMI":"$CONFDIR"/_common:"$CONFDIR"/_common/hadoop-xml:"$CONFDIR"/../_common:"$CONFDIR"/../_common/hadoop-xml:"$WHEREAMI/../lib/*" $MAIN_CLASS + -cp "$CONFDIR"/"$WHATAMI":"$CONFDIR"/_common:"$CONFDIR"/_common/hadoop-xml:"$CONFDIR"/../_common:"$CONFDIR"/../_common/hadoop-xml:"$WHEREAMI/../lib/*" \ + $MAIN_CLASS fi diff --git a/examples/bin/start-druid b/examples/bin/start-druid index 41639d8cdce6..601df55ffef5 100644 --- a/examples/bin/start-druid +++ b/examples/bin/start-druid @@ -21,4 +21,4 @@ PWD="$(pwd)" WHEREAMI="$(dirname "$0")" WHEREAMI="$(cd "$WHEREAMI" && pwd)" -python3 "$WHEREAMI"/druid-quickstart.py $@ +exec "$WHEREAMI/start-druid-main" "$@" diff --git a/examples/bin/start-druid-main b/examples/bin/start-druid-main new file mode 100644 index 000000000000..b74d43336c98 --- /dev/null +++ b/examples/bin/start-druid-main @@ -0,0 +1,439 @@ +#!/usr/bin/env python3 + +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. + +import sys +import os +import psutil +import pathlib +import multiprocessing +import argparse + +QUICKSTART_ROOT_CONFIG_PATH = "conf/druid/single-server/quickstart" + +MEMORY_GIGABYTES_IDENTIFIER = "g" +MEMORY_MEGABYTES_IDENTIFIER = "m" +SERVICE_SEPARATOR = "," + +MM_TASK_JAVAOPTS_ARRAY = ["-server", "-Duser.timezone=UTC","-Dfile.encoding=UTF-8","-XX:+ExitOnOutOfMemoryError","-Djava.util.logging.manager=org.apache.logging.log4j.jul.LogManager"] +MM_TASK_MEMORY_TYPE_LOW = "low" +MM_TASK_MEMORY_TYPE_HIGH = "high" +MM_TASK_MEM_MAP = { + MM_TASK_MEMORY_TYPE_LOW: ["-Xms256m", "-Xmx256m", "-XX:MaxDirectMemorySize=256g"], + MM_TASK_MEMORY_TYPE_HIGH: ["-Xms2g", "-Xmx2g", "-XX:MaxDirectMemorySize=2g"] +} + +MM_TASK_JAVAOPTS_PROP = "-Ddruid.indexer.runner.javaOptsArray" +MM_TASK_WORKER_CAPACITY_PROP = "-Ddruid.worker.capacity" + +BROKER = "broker" +ROUTER = "router" +COORDINATOR = "coordinator-overlord" +HISTORICAL = "historical" +MM = "middleManager" +MM_TASK = "middleManager-task" +MM_TASK_COUNT = "task-count" + +DEFAULT_SERVICES = [ + BROKER, + ROUTER, + COORDINATOR, + HISTORICAL, + MM +] + +SERVICE_MEMORY_DISTRIBUTION_WEIGHT = { + MM: 0.5, + ROUTER: 1, + COORDINATOR: 15, + BROKER: 23, + HISTORICAL: 40, + MM_TASK: 15 +} + +SERVICE_MEMORY_LOWER_BOUND = { + MM: 64, + ROUTER: 128, + MM_TASK: 1024, + BROKER: 900, + COORDINATOR: 256, + HISTORICAL: 900 +} + +SERVICE_MEMORY_HEAP_PERCENTAGE = { + MM: 1, + ROUTER: 1, + COORDINATOR: 1, + BROKER: 0.60, + HISTORICAL: 0.40, + MM_TASK: 0.50 +} + +LOGGING_ENABLED = False + +def custom_print(message): + if LOGGING_ENABLED: + print(message) + +def error_and_exit(message): + sys.stderr.write(message + '\n') + sys.exit(1) + +def configure_parser(): + parser = argparse.ArgumentParser( + prog='Druid quickstart', + formatter_class=argparse.RawTextHelpFormatter, + epilog= +""" +sample usage: + start-druid + Start up all the services (including zk). + 50 - 80 percent of system memory is used. + start-druid -m=100g + Start up all the services (including zk) + using the given memory. + start-druid -m=100g --compute_only + Compute memory distribution and validate + arguments. + start-druid -m=100g -sl=broker,router + Start broker & router service, using 100g of memory. + Read config from conf/druid/single-server/quickstart. + start-druid -m=100g --sl=broker,router \\ + -cp=conf/druid/single-server/custom + Start broker & router service, using 100g of memory. + Read config from . + Since is specified, exception is thrown if + jvm.config is present for any of the services. + start-druid -sl=broker,router \\ + -cp=conf/druid/single-server/custom + Start broker & router service, using system memory. + If jvm.config is specified for both the + services within /, + memory distribution is not calculated. + If jvm.config is present for either of the services, + exception is thrown. + If jvm.config is not present for both of the services, + memory distribution is calculated. + start-druid -m=100g \\ + -cp=conf/druid/single-server/custom \\ + -sl=broker,router \\ + --run_zk + Start broker, router and zookeeper. + zk config is read from conf/zk. +""" + ) + parser.add_argument('--memory', '-m', type=str, required=False, + help='Total memory for all processes (services and tasks, if any). \n' + 'This parameter is ignored if each service already has a jvm.config \n' + 'in the given conf directory. e.g. 500m, 4g, 6g\n') + parser.add_argument('--service_list', '-sl', type=str, required=False, + help='List of services to be started, subset of \n' + '{broker, router, middleManager, historical, coordinator-overlord}. \n' + 'If the argument is not given, all services \n' + 'and zookeeper is started. e.g. -sl=broker,historical') + parser.add_argument('--config_path', '-cp', type=str, required=False, + help='Relative path to the directory containing common and service \n' + 'specific properties to be overridden. \n' + 'This directory must contain \'_common\' directory with \n' + '\'common.jvm.config\' & \'common.runtime.properties\' files. \n' + 'If this argument is not given, config from \n' + '\'conf/druid/single-server/quickstart\' directory is used.\n') + parser.add_argument('--compute_only', action='store_true', + help='Validate the arguments and display memory distribution') + parser.add_argument('--run_zk', action='store_true', + help='Specification to run zookeeper, \n' + 'zk config is picked up from conf/zk.') + parser.add_argument('--verbose', action='store_true', help='Log details.') + + parser.set_defaults(run_zk=False) + parser.set_defaults(compute_only=False) + parser.set_defaults(verbose=False) + + return parser + +def parse_arguments(args): + service_list = [] + config_path = "" + total_memory = "" + compute_only = False + run_zk = False + + if args.compute_only: + compute_only = True + if args.run_zk: + run_zk = True + if args.config_path is not None: + config_path = os.path.join(os.getcwd(), args.config_path) + if os.path.exists(config_path) is False: + error_and_exit(f'config_path {config_path} doesn\'t exist') + if args.memory is not None: + total_memory = args.memory + if args.service_list is not None: + services = args.service_list.split(SERVICE_SEPARATOR) + + for service in services: + if service not in DEFAULT_SERVICES: + error_and_exit(f'Invalid service name {service}, should be one of {DEFAULT_SERVICES}') + + if service in service_list: + error_and_exit(f'{service} is specified multiple times') + + service_list.append(service) + + if len(service_list) == 0: + # start all services + service_list = DEFAULT_SERVICES + run_zk = True + + return config_path, total_memory, service_list, run_zk, compute_only + +def print_startup_config(service_list, config_path, run_zk): + custom_print(f'starting {service_list}') + custom_print(f'reading config from {config_path}') + if run_zk: + zk_config_path = pathlib.Path(f'{os.getcwd()}/../conf/zk').resolve() + custom_print(f'starting zk, reading default config from {zk_config_path}') + custom_print('\n') + +def should_compute_memory(config_path, total_memory, service_list): + # if jvm file is present for any of the services + # it should be present for all services and memory should not be specified + # if memory is given, jvm file shouldn't be present for any service + + jvm_config_count = 0 + for service in service_list: + if pathlib.Path(f'{config_path}/{service}/jvm.config').is_file(): + jvm_config_count += 1 + elif jvm_config_count > 0: + error_and_exit(f'jvm.config file is missing for service {service}, jvm.config should be specified for all the services or none') + + if jvm_config_count > 0 and (jvm_config_count != len(service_list) or total_memory != ""): + if jvm_config_count != len(service_list): + error_and_exit("jvm.config file should be present for all services or none") + if total_memory != "": + error_and_exit("If jvm.config is given for services, memory argument shouldn't be specified") + + return jvm_config_count == 0 + +def check_memory_constraint(total_memory, service_list): + # sum of lower bound service memory should be >= 80% of total memory + lower_bound_memory = 0 + + for service in service_list: + lower_bound_memory += SERVICE_MEMORY_LOWER_BOUND.get(service) + + required_memory = int(lower_bound_memory / 0.8) + + if total_memory < required_memory: + error_and_exit(f'Minimum memory required for starting services is {required_memory}m') + + if total_memory > 2 * required_memory: + return int(total_memory / 2) + else: + return required_memory + +def compute_system_memory(): + system_memory = psutil.virtual_memory().total # mem in bytes + memory_for_druid = int(system_memory / (1024 * 1024)) + return memory_for_druid + +def convert_total_memory_string(memory): + if memory == "": + computed_memory = compute_system_memory() + return computed_memory + elif memory.endswith(MEMORY_MEGABYTES_IDENTIFIER): + return int(memory[:-1]) + elif memory.endswith(MEMORY_GIGABYTES_IDENTIFIER): + return 1024 * int(memory[:-1]) + else: + error_and_exit('Incorrect format for memory argument, expected format is ') + +def build_mm_task_javaopts_array(memory_type): + task_memory = f'{MM_TASK_JAVAOPTS_PROP}=[' + + MEM_ARRAY = MM_TASK_MEM_MAP.get(memory_type) + + javaopts_list = MM_TASK_JAVAOPTS_ARRAY + MEM_ARRAY + + for item in javaopts_list: + task_memory += f'"\"{item}\";' + + task_memory = task_memory[:-1] + task_memory += ']' + return task_memory + +def build_memory_config_string(service, allocated_memory): + if service == MM_TASK: + if allocated_memory >= 2048: + task_count = int(allocated_memory / 2048) + memory_type = MM_TASK_MEMORY_TYPE_HIGH + task_memory = 2048 + else: + task_count = int(allocated_memory / 512) + memory_type = MM_TASK_MEMORY_TYPE_LOW + task_memory = 512 + task_count = min(task_count, multiprocessing.cpu_count()) + + javaopts_array = build_mm_task_javaopts_array(memory_type) + return [f'{MM_TASK_WORKER_CAPACITY_PROP}={task_count}', javaopts_array], task_memory * task_count + else: + heap_memory = SERVICE_MEMORY_HEAP_PERCENTAGE.get(service) * allocated_memory + direct_memory = int(allocated_memory - heap_memory) + heap_memory = int(heap_memory) + + if direct_memory == 0: + return f'-Xms{heap_memory}m -Xmx{heap_memory}m', allocated_memory + + return f'-Xms{heap_memory}m -Xmx{heap_memory}m -XX:MaxDirectMemorySize={direct_memory}m', allocated_memory + +def distribute_memory_over_services(service_list, total_memory): + service_memory_config = {} + + memory_weight_sum = 0 + for service in service_list: + memory_weight_sum += SERVICE_MEMORY_DISTRIBUTION_WEIGHT.get(service) + + multiplier = total_memory / memory_weight_sum + + lower_bound_memory_allocation = 0 + allocated_services = set() + + for service in service_list: + allocated_memory = SERVICE_MEMORY_DISTRIBUTION_WEIGHT.get(service) * multiplier + if service in SERVICE_MEMORY_LOWER_BOUND and allocated_memory < SERVICE_MEMORY_LOWER_BOUND.get(service): + allocated_memory = SERVICE_MEMORY_LOWER_BOUND.get(service) + service_memory_config[service], allocated_memory = build_memory_config_string(service, allocated_memory) + lower_bound_memory_allocation += allocated_memory + allocated_services.add(service) + + if lower_bound_memory_allocation > 0: + # compute the multiplier again for remaining services + memory_weight_sum = 0 + for service in service_list: + if service in allocated_services: + continue + memory_weight_sum += SERVICE_MEMORY_DISTRIBUTION_WEIGHT.get(service) + multiplier = (total_memory - lower_bound_memory_allocation) / memory_weight_sum + + for service in service_list: + if service in allocated_services: + continue + allocated_memory = SERVICE_MEMORY_DISTRIBUTION_WEIGHT.get(service) * multiplier + if service in SERVICE_MEMORY_LOWER_BOUND and allocated_memory < SERVICE_MEMORY_LOWER_BOUND.get(service): + allocated_memory = SERVICE_MEMORY_LOWER_BOUND.get(service) + + service_memory_config[service], allocated_memory = build_memory_config_string(service, allocated_memory) + + custom_print(f'\nMemory distribution for services:') + for key, value in service_memory_config.items(): + custom_print(f'{key}, memory_config: {value}') + custom_print('\n') + + return service_memory_config + +def build_supervise_script_arguments(service_list, service_memory_config, config_path, run_zk): + args = [] + commands = [] + args.append('supervise') + args.append('-a') + + commands.append(":verify bin/verify-java") + commands.append(":verify bin/verify-default-ports") + commands.append(":notify bin/greet") + commands.append(":kill-timeout 10") + + if run_zk: + commands.append("!p10 zk bin/run-zk conf") + + for service in service_list: + if service == MM_TASK: + continue + + prefix = '' + if service == MM: + prefix = '!p90 ' + + jvm_args = service_memory_config.get(service) + + if service == MM: + task_config = service_memory_config.get(MM_TASK) + task_count = task_config[0] + task_memory = task_config[1] + if jvm_args is None: + commands.append( + f'{prefix}{service} bin/run-druid {service} {config_path} \'{task_count}\' \'{task_memory}\'') + else: + commands.append( + f'{prefix}{service} bin/run-druid {service} {config_path} \'{task_count}\' \'{task_memory}\' \'{jvm_args}\'') + else: + if jvm_args is None: + commands.append(f'{prefix}{service} bin/run-druid {service} {config_path}') + else: + commands.append(f'{prefix}{service} bin/run-druid {service} {config_path} \'{jvm_args}\'') + + custom_print('Supervise script args:') + for item in commands: + custom_print(item) + + custom_print('\n') + + args.append(",".join(commands)) + return args + +def main(): + parser = configure_parser() + args = parser.parse_args() + + global LOGGING_ENABLED + LOGGING_ENABLED = args.verbose + + config_path, total_memory, service_list, run_zk, compute_only = parse_arguments(args) + + # change directory to bin + os.chdir(os.path.dirname(sys.argv[0])) + + if config_path == "": + config_path = pathlib.Path(f'{os.getcwd()}/../{QUICKSTART_ROOT_CONFIG_PATH}').resolve() + + print_startup_config(service_list, config_path, run_zk) + + service_memory_config = {} + + if MM in service_list: + service_list.append(MM_TASK) + + if (should_compute_memory(config_path, total_memory, service_list)): + memory_in_mega_bytes = convert_total_memory_string(total_memory) + custom_print(f'total memory is {memory_in_mega_bytes}m') + memory_to_be_used = check_memory_constraint(memory_in_mega_bytes, service_list) + custom_print(f'memory used for services & tasks {memory_to_be_used}m') + service_memory_config = distribute_memory_over_services(service_list, memory_to_be_used) + else: + custom_print('not computing memory distribution, reading memory specification from service jvm.config\n') + + script_arguments = build_supervise_script_arguments(service_list, service_memory_config, config_path, run_zk) + + if compute_only: + return + + os.execv('./supervise', script_arguments) + +try: + main() +except KeyboardInterrupt: + sys.exit(1) diff --git a/examples/bin/supervise b/examples/bin/supervise index 507795733c34..9a6809bacdc8 100755 --- a/examples/bin/supervise +++ b/examples/bin/supervise @@ -56,7 +56,6 @@ sub process_config for my $line (@lines) { if ($line =~ /^(:verify|:notify|:kill-timeout|(?:\!p[0-9]+\s+)?[^:]\S+)\s+(.+)$/) { - my $name = $1; my $order = 50; my $command = $2; @@ -211,6 +210,11 @@ if (not defined $opt{'conf'}) { @config_lines = read_config_file($opt{'conf'}); } +for my $alpha (@config_lines) +{ + $alpha =~ s/;/,/g; +} + my $config = process_config(@config_lines); @commands = @{$config->{commands}}; diff --git a/examples/conf/druid/single-server/quickstart/middleManager/runtime.properties b/examples/conf/druid/single-server/quickstart/middleManager/runtime.properties index e7262930f498..b0cf3ca3c43e 100644 --- a/examples/conf/druid/single-server/quickstart/middleManager/runtime.properties +++ b/examples/conf/druid/single-server/quickstart/middleManager/runtime.properties @@ -21,15 +21,15 @@ druid.service=druid/middleManager druid.plaintextPort=8091 # Number of tasks per middleManager -druid.worker.capacity=2 +# druid.worker.capacity # Task launch parameters druid.indexer.runner.javaCommand=bin/run-java -druid.indexer.runner.javaOptsArray=["-server","-Xms1g","-Xmx1g","-XX:MaxDirectMemorySize=1g","-Duser.timezone=UTC","-Dfile.encoding=UTF-8","-XX:+ExitOnOutOfMemoryError","-Djava.util.logging.manager=org.apache.logging.log4j.jul.LogManager"] +#druid.indexer.runner.javaOptsArray=["-server","-Xms1g","-Xmx1g","-XX:MaxDirectMemorySize=1g","-Duser.timezone=UTC","-Dfile.encoding=UTF-8","-XX:+ExitOnOutOfMemoryError","-Djava.util.logging.manager=org.apache.logging.log4j.jul.LogManager"] druid.indexer.task.baseTaskDir=var/druid/task # HTTP server threads -druid.server.http.numThreads=12 +#druid.server.http.numThreads=12 # Processing threads and buffers on Peons druid.indexer.fork.property.druid.processing.numMergeBuffers=2 From c6054b7cb78a9aaed0763465730434d2c9d6f6c9 Mon Sep 17 00:00:00 2001 From: Gian Merlino Date: Mon, 21 Nov 2022 22:20:26 -0800 Subject: [PATCH 016/141] Attach IO error to parse error when we can't contact Avro schema registry. (#13403) * Attach IO error to parse error when we can't contact Avro schema registry. The change in #12080 lost the original exception context. This patch adds it back. * Add hamcrest-core. * Fix format string. --- extensions-core/avro-extensions/pom.xml | 5 +++++ .../avro/SchemaRegistryBasedAvroBytesDecoder.java | 6 +++--- .../avro/SchemaRegistryBasedAvroBytesDecoderTest.java | 11 +++++++++-- 3 files changed, 17 insertions(+), 5 deletions(-) diff --git a/extensions-core/avro-extensions/pom.xml b/extensions-core/avro-extensions/pom.xml index fcabcd7c9daa..1c2dddb9902d 100644 --- a/extensions-core/avro-extensions/pom.xml +++ b/extensions-core/avro-extensions/pom.xml @@ -254,6 +254,11 @@ mockito-core test + + org.hamcrest + hamcrest-core + test + org.apache.druid druid-core diff --git a/extensions-core/avro-extensions/src/main/java/org/apache/druid/data/input/avro/SchemaRegistryBasedAvroBytesDecoder.java b/extensions-core/avro-extensions/src/main/java/org/apache/druid/data/input/avro/SchemaRegistryBasedAvroBytesDecoder.java index f7006cb02c21..05c5c9fec2fe 100644 --- a/extensions-core/avro-extensions/src/main/java/org/apache/druid/data/input/avro/SchemaRegistryBasedAvroBytesDecoder.java +++ b/extensions-core/avro-extensions/src/main/java/org/apache/druid/data/input/avro/SchemaRegistryBasedAvroBytesDecoder.java @@ -140,17 +140,17 @@ public GenericRecord parse(ByteBuffer bytes) schema = parsedSchema instanceof AvroSchema ? ((AvroSchema) parsedSchema).rawSchema() : null; } catch (IOException | RestClientException ex) { - throw new ParseException(null, "Failed to get Avro schema: %s", id); + throw new ParseException(null, ex, "Failed to fetch Avro schema from registry: %s", id); } if (schema == null) { - throw new ParseException(null, "Failed to find Avro schema: %s", id); + throw new ParseException(null, "No Avro schema in registry: %s", id); } DatumReader reader = new GenericDatumReader<>(schema); try { return reader.read(null, DecoderFactory.get().binaryDecoder(bytes.array(), offset, length, null)); } catch (Exception e) { - throw new ParseException(null, e, "Fail to decode Avro message for schema: %s!", id); + throw new ParseException(null, e, "Failed to decode Avro message for schema: %s", id); } } diff --git a/extensions-core/avro-extensions/src/test/java/org/apache/druid/data/input/avro/SchemaRegistryBasedAvroBytesDecoderTest.java b/extensions-core/avro-extensions/src/test/java/org/apache/druid/data/input/avro/SchemaRegistryBasedAvroBytesDecoderTest.java index 9348e9486aef..7644b61bb106 100644 --- a/extensions-core/avro-extensions/src/test/java/org/apache/druid/data/input/avro/SchemaRegistryBasedAvroBytesDecoderTest.java +++ b/extensions-core/avro-extensions/src/test/java/org/apache/druid/data/input/avro/SchemaRegistryBasedAvroBytesDecoderTest.java @@ -35,6 +35,8 @@ import org.apache.druid.jackson.DefaultObjectMapper; import org.apache.druid.java.util.common.parsers.ParseException; import org.apache.druid.utils.DynamicConfigProviderUtils; +import org.hamcrest.CoreMatchers; +import org.hamcrest.MatcherAssert; import org.junit.Assert; import org.junit.Before; import org.junit.Test; @@ -158,7 +160,7 @@ public void testParseWrongSchemaType() throws Exception new SchemaRegistryBasedAvroBytesDecoder(registry).parse(bb); } - @Test(expected = ParseException.class) + @Test public void testParseWrongId() throws Exception { // Given @@ -166,7 +168,12 @@ public void testParseWrongId() throws Exception ByteBuffer bb = ByteBuffer.allocate(5).put((byte) 0).putInt(1234); bb.rewind(); // When - new SchemaRegistryBasedAvroBytesDecoder(registry).parse(bb); + final ParseException e = Assert.assertThrows( + ParseException.class, + () -> new SchemaRegistryBasedAvroBytesDecoder(registry).parse(bb) + ); + MatcherAssert.assertThat(e.getCause(), CoreMatchers.instanceOf(IOException.class)); + MatcherAssert.assertThat(e.getCause().getMessage(), CoreMatchers.containsString("no pasaran")); } private byte[] getAvroDatum(Schema schema, GenericRecord someAvroDatum) throws IOException From 75d169f15cbcd0a6fdc68f4068f14050c40909fd Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Tue, 22 Nov 2022 11:50:59 +0530 Subject: [PATCH 017/141] Remove extra quotes from mm task javaopts array --- examples/bin/start-druid-main | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/examples/bin/start-druid-main b/examples/bin/start-druid-main index b74d43336c98..a1fe6af4985d 100644 --- a/examples/bin/start-druid-main +++ b/examples/bin/start-druid-main @@ -35,7 +35,7 @@ MM_TASK_MEMORY_TYPE_LOW = "low" MM_TASK_MEMORY_TYPE_HIGH = "high" MM_TASK_MEM_MAP = { MM_TASK_MEMORY_TYPE_LOW: ["-Xms256m", "-Xmx256m", "-XX:MaxDirectMemorySize=256g"], - MM_TASK_MEMORY_TYPE_HIGH: ["-Xms2g", "-Xmx2g", "-XX:MaxDirectMemorySize=2g"] + MM_TASK_MEMORY_TYPE_HIGH: ["-Xms1g", "-Xmx1g", "-XX:MaxDirectMemorySize=1g"] } MM_TASK_JAVAOPTS_PROP = "-Ddruid.indexer.runner.javaOptsArray" @@ -242,7 +242,7 @@ def check_memory_constraint(total_memory, service_list): if total_memory < required_memory: error_and_exit(f'Minimum memory required for starting services is {required_memory}m') - if total_memory > 2 * required_memory: + if total_memory >= 2 * required_memory: return int(total_memory / 2) else: return required_memory @@ -271,7 +271,7 @@ def build_mm_task_javaopts_array(memory_type): javaopts_list = MM_TASK_JAVAOPTS_ARRAY + MEM_ARRAY for item in javaopts_list: - task_memory += f'"\"{item}\";' + task_memory += f'\"{item}\";' task_memory = task_memory[:-1] task_memory += ']' From a891319e005eac2d09bbe9af8887536286ba9b60 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Tue, 22 Nov 2022 12:37:00 +0530 Subject: [PATCH 018/141] Update logic to compute minimum memory --- examples/bin/start-druid-main | 34 +++++++++++++++++----------------- 1 file changed, 17 insertions(+), 17 deletions(-) diff --git a/examples/bin/start-druid-main b/examples/bin/start-druid-main index a1fe6af4985d..ecd630c10501 100644 --- a/examples/bin/start-druid-main +++ b/examples/bin/start-druid-main @@ -230,23 +230,6 @@ def should_compute_memory(config_path, total_memory, service_list): return jvm_config_count == 0 -def check_memory_constraint(total_memory, service_list): - # sum of lower bound service memory should be >= 80% of total memory - lower_bound_memory = 0 - - for service in service_list: - lower_bound_memory += SERVICE_MEMORY_LOWER_BOUND.get(service) - - required_memory = int(lower_bound_memory / 0.8) - - if total_memory < required_memory: - error_and_exit(f'Minimum memory required for starting services is {required_memory}m') - - if total_memory >= 2 * required_memory: - return int(total_memory / 2) - else: - return required_memory - def compute_system_memory(): system_memory = psutil.virtual_memory().total # mem in bytes memory_for_druid = int(system_memory / (1024 * 1024)) @@ -263,6 +246,23 @@ def convert_total_memory_string(memory): else: error_and_exit('Incorrect format for memory argument, expected format is ') +def check_memory_constraint(total_memory, service_list): + # 80% of total memory >= sum of lower bound service memory should be + lower_bound_memory = 0 + + for service in service_list: + lower_bound_memory += SERVICE_MEMORY_LOWER_BOUND.get(service) + + required_memory = int(lower_bound_memory / 0.8) + + if total_memory < required_memory: + error_and_exit(f'Minimum memory required for starting services is {required_memory}m') + + if total_memory >= 2 * lower_bound_memory: + return int(total_memory / 2) + else: + return lower_bound_memory + def build_mm_task_javaopts_array(memory_type): task_memory = f'{MM_TASK_JAVAOPTS_PROP}=[' From 569275378487ddf6dd516a448f0c8f3edeb11a64 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Tue, 22 Nov 2022 15:29:04 +0530 Subject: [PATCH 019/141] simplify run-druid --- examples/bin/run-druid | 18 +++++------ examples/bin/start-druid-main | 59 +++++++++++++++++------------------ 2 files changed, 36 insertions(+), 41 deletions(-) diff --git a/examples/bin/run-druid b/examples/bin/run-druid index 9e9f93d90deb..73ce559ef481 100755 --- a/examples/bin/run-druid +++ b/examples/bin/run-druid @@ -17,7 +17,7 @@ # specific language governing permissions and limitations # under the License. -if [ "$#" -gt 5 ] || [ "$#" -eq 0 ] +if [ "$#" -gt 4 ] || [ "$#" -eq 0 ] then >&2 echo "usage: $0 [conf-dir]" exit 1 @@ -64,28 +64,26 @@ fi cd "$WHEREAMI/.." -if [ "$#" -eq 3 ] || [ "$#" -eq 5 ] +if [ "$#" -eq 3 ] || [ "$#" -eq 4 ] then - # args: or + # args: or JVMARGS=`cat "$CONFDIR/_common/common.jvm.config" | xargs` JVMARGS+=' ' JVMARGS+=$3 if [ "$#" -eq 3 ] then - exec "$WHEREAMI"/run-java -Ddruid.node.type=$1 $3 "-Ddruid.log.path=$LOG_DIR" $JVMARGS \ + # args: + exec "$WHEREAMI"/run-java -Ddruid.node.type=$1 $4 "-Ddruid.log.path=$LOG_DIR" $JVMARGS \ -cp "$CONFDIR"/"$WHATAMI":"$CONFDIR"/_common:"$CONFDIR"/_common/hadoop-xml:"$CONFDIR"/../_common:"$CONFDIR"/../_common/hadoop-xml:"$WHEREAMI/../lib/*" org.apache.druid.cli.Main server $SERVER_NAME else - exec "$WHEREAMI"/run-java -Ddruid.node.type=$1 $3 "-Ddruid.log.path=$LOG_DIR" $4 $5 $JVMARGS \ + # args: + exec "$WHEREAMI"/run-java -Ddruid.node.type=$1 $4"-Ddruid.log.path=$LOG_DIR" -agentlib:jdwp=transport=dt_socket,server=y,suspend=n,address=5005 $JVMARGS \ -cp "$CONFDIR"/"$WHATAMI":"$CONFDIR"/_common:"$CONFDIR"/_common/hadoop-xml:"$CONFDIR"/../_common:"$CONFDIR"/../_common/hadoop-xml:"$WHEREAMI/../lib/*" org.apache.druid.cli.Main server $SERVER_NAME fi -elif [ "$#" -eq 4 ] -then - # args: - exec "$WHEREAMI"/run-java -Ddruid.node.type=$1 $3 "-Ddruid.log.path=$LOG_DIR" `cat "$CONFDIR"/"$WHATAMI"/jvm.config | xargs` \ - -cp "$CONFDIR"/"$WHATAMI":"$CONFDIR"/_common:"$CONFDIR"/_common/hadoop-xml:"$CONFDIR"/../_common:"$CONFDIR"/../_common/hadoop-xml:"$WHEREAMI/../lib/*" org.apache.druid.cli.Main server $SERVER_NAME else # args: + echo "case parameters count 2, $#" exec "$WHEREAMI"/run-java -Ddruid.node.type=$1 "-Ddruid.log.path=$LOG_DIR" `cat "$CONFDIR"/"$WHATAMI"/jvm.config | xargs` \ -cp "$CONFDIR"/"$WHATAMI":"$CONFDIR"/_common:"$CONFDIR"/_common/hadoop-xml:"$CONFDIR"/../_common:"$CONFDIR"/../_common/hadoop-xml:"$WHEREAMI/../lib/*" \ $MAIN_CLASS diff --git a/examples/bin/start-druid-main b/examples/bin/start-druid-main index ecd630c10501..86eec4b8eddf 100644 --- a/examples/bin/start-druid-main +++ b/examples/bin/start-druid-main @@ -158,7 +158,7 @@ sample usage: parser.add_argument('--run_zk', action='store_true', help='Specification to run zookeeper, \n' 'zk config is picked up from conf/zk.') - parser.add_argument('--verbose', action='store_true', help='Log details.') + parser.add_argument('--verbose', action='store_true', help='Log details') parser.set_defaults(run_zk=False) parser.set_defaults(compute_only=False) @@ -214,7 +214,6 @@ def should_compute_memory(config_path, total_memory, service_list): # if jvm file is present for any of the services # it should be present for all services and memory should not be specified # if memory is given, jvm file shouldn't be present for any service - jvm_config_count = 0 for service in service_list: if pathlib.Path(f'{config_path}/{service}/jvm.config').is_file(): @@ -236,14 +235,17 @@ def compute_system_memory(): return memory_for_druid def convert_total_memory_string(memory): - if memory == "": - computed_memory = compute_system_memory() - return computed_memory - elif memory.endswith(MEMORY_MEGABYTES_IDENTIFIER): - return int(memory[:-1]) - elif memory.endswith(MEMORY_GIGABYTES_IDENTIFIER): - return 1024 * int(memory[:-1]) - else: + try: + if memory == "": + computed_memory = compute_system_memory() + return computed_memory + elif memory.endswith(MEMORY_MEGABYTES_IDENTIFIER): + return int(memory[:-1]) + elif memory.endswith(MEMORY_GIGABYTES_IDENTIFIER): + return 1024 * int(memory[:-1]) + else: + error_and_exit('Incorrect format for memory argument, expected format is ') + except Exception: error_and_exit('Incorrect format for memory argument, expected format is ') def check_memory_constraint(total_memory, service_list): @@ -301,10 +303,15 @@ def build_memory_config_string(service, allocated_memory): return f'-Xms{heap_memory}m -Xmx{heap_memory}m -XX:MaxDirectMemorySize={direct_memory}m', allocated_memory -def distribute_memory_over_services(service_list, total_memory): +def distribute_memory_over_services(services, total_memory): service_memory_config = {} memory_weight_sum = 0 + + service_list = services.copy() + if MM in services: + service_list.append(MM_TASK) + for service in service_list: memory_weight_sum += SERVICE_MEMORY_DISTRIBUTION_WEIGHT.get(service) @@ -361,28 +368,21 @@ def build_supervise_script_arguments(service_list, service_memory_config, config commands.append("!p10 zk bin/run-zk conf") for service in service_list: - if service == MM_TASK: - continue + jvm_args = service_memory_config.get(service) prefix = '' if service == MM: prefix = '!p90 ' - jvm_args = service_memory_config.get(service) - - if service == MM: - task_config = service_memory_config.get(MM_TASK) - task_count = task_config[0] - task_memory = task_config[1] - if jvm_args is None: - commands.append( - f'{prefix}{service} bin/run-druid {service} {config_path} \'{task_count}\' \'{task_memory}\'') - else: - commands.append( - f'{prefix}{service} bin/run-druid {service} {config_path} \'{task_count}\' \'{task_memory}\' \'{jvm_args}\'') + if jvm_args is None: + commands.append(f'{prefix}{service} bin/run-druid {service} {config_path}') else: - if jvm_args is None: - commands.append(f'{prefix}{service} bin/run-druid {service} {config_path}') + if service == MM: + task_config = service_memory_config.get(MM_TASK) + task_count = task_config[0] + task_memory = task_config[1] + commands.append( + f'{prefix}{service} bin/run-druid {service} {config_path} \'{jvm_args}\' \'{task_count} {task_memory}\'') else: commands.append(f'{prefix}{service} bin/run-druid {service} {config_path} \'{jvm_args}\'') @@ -400,7 +400,7 @@ def main(): args = parser.parse_args() global LOGGING_ENABLED - LOGGING_ENABLED = args.verbose + LOGGING_ENABLED = args.verbose or args.compute_only config_path, total_memory, service_list, run_zk, compute_only = parse_arguments(args) @@ -414,9 +414,6 @@ def main(): service_memory_config = {} - if MM in service_list: - service_list.append(MM_TASK) - if (should_compute_memory(config_path, total_memory, service_list)): memory_in_mega_bytes = convert_total_memory_string(total_memory) custom_print(f'total memory is {memory_in_mega_bytes}m') From d111012eaee512cda1b92fb5621d21cfa12f69c5 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Tue, 22 Nov 2022 15:30:01 +0530 Subject: [PATCH 020/141] remove debug options from run-druid --- examples/bin/run-druid | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/examples/bin/run-druid b/examples/bin/run-druid index 73ce559ef481..f6fd1cfd85bb 100755 --- a/examples/bin/run-druid +++ b/examples/bin/run-druid @@ -78,12 +78,11 @@ then -cp "$CONFDIR"/"$WHATAMI":"$CONFDIR"/_common:"$CONFDIR"/_common/hadoop-xml:"$CONFDIR"/../_common:"$CONFDIR"/../_common/hadoop-xml:"$WHEREAMI/../lib/*" org.apache.druid.cli.Main server $SERVER_NAME else # args: - exec "$WHEREAMI"/run-java -Ddruid.node.type=$1 $4"-Ddruid.log.path=$LOG_DIR" -agentlib:jdwp=transport=dt_socket,server=y,suspend=n,address=5005 $JVMARGS \ + exec "$WHEREAMI"/run-java -Ddruid.node.type=$1 $4"-Ddruid.log.path=$LOG_DIR" $JVMARGS \ -cp "$CONFDIR"/"$WHATAMI":"$CONFDIR"/_common:"$CONFDIR"/_common/hadoop-xml:"$CONFDIR"/../_common:"$CONFDIR"/../_common/hadoop-xml:"$WHEREAMI/../lib/*" org.apache.druid.cli.Main server $SERVER_NAME fi else # args: - echo "case parameters count 2, $#" exec "$WHEREAMI"/run-java -Ddruid.node.type=$1 "-Ddruid.log.path=$LOG_DIR" `cat "$CONFDIR"/"$WHATAMI"/jvm.config | xargs` \ -cp "$CONFDIR"/"$WHATAMI":"$CONFDIR"/_common:"$CONFDIR"/_common/hadoop-xml:"$CONFDIR"/../_common:"$CONFDIR"/../_common/hadoop-xml:"$WHEREAMI/../lib/*" \ $MAIN_CLASS From 7cf761cee42dae8c21e415b7ce3fd4191f199a38 Mon Sep 17 00:00:00 2001 From: Kashif Faraz Date: Tue, 22 Nov 2022 15:31:01 +0530 Subject: [PATCH 021/141] Prepare master branch for next release, 26.0.0 (#13401) * Prepare master branch for next release, 26.0.0 * Use docker image for druid 24.0.1 * Fix version in druid-it-cases pom.xml --- benchmarks/pom.xml | 2 +- cloud/aws-common/pom.xml | 2 +- cloud/gcp-common/pom.xml | 2 +- core/pom.xml | 2 +- distribution/docker/docker-compose.yml | 2 +- distribution/pom.xml | 2 +- extendedset/pom.xml | 2 +- extensions-contrib/aliyun-oss-extensions/pom.xml | 2 +- extensions-contrib/ambari-metrics-emitter/pom.xml | 2 +- extensions-contrib/cassandra-storage/pom.xml | 2 +- extensions-contrib/cloudfiles-extensions/pom.xml | 2 +- extensions-contrib/compressed-bigdecimal/pom.xml | 2 +- extensions-contrib/distinctcount/pom.xml | 2 +- extensions-contrib/dropwizard-emitter/pom.xml | 2 +- extensions-contrib/gce-extensions/pom.xml | 2 +- extensions-contrib/graphite-emitter/pom.xml | 2 +- extensions-contrib/influx-extensions/pom.xml | 2 +- extensions-contrib/influxdb-emitter/pom.xml | 2 +- extensions-contrib/kafka-emitter/pom.xml | 2 +- extensions-contrib/kubernetes-overlord-extensions/pom.xml | 2 +- extensions-contrib/materialized-view-maintenance/pom.xml | 2 +- extensions-contrib/materialized-view-selection/pom.xml | 2 +- extensions-contrib/momentsketch/pom.xml | 2 +- extensions-contrib/moving-average-query/pom.xml | 2 +- extensions-contrib/opentelemetry-emitter/pom.xml | 2 +- extensions-contrib/opentsdb-emitter/pom.xml | 2 +- extensions-contrib/prometheus-emitter/pom.xml | 2 +- extensions-contrib/redis-cache/pom.xml | 2 +- extensions-contrib/sqlserver-metadata-storage/pom.xml | 2 +- extensions-contrib/statsd-emitter/pom.xml | 2 +- extensions-contrib/tdigestsketch/pom.xml | 2 +- extensions-contrib/thrift-extensions/pom.xml | 2 +- extensions-contrib/time-min-max/pom.xml | 2 +- extensions-contrib/virtual-columns/pom.xml | 2 +- extensions-core/avro-extensions/pom.xml | 2 +- extensions-core/azure-extensions/pom.xml | 2 +- extensions-core/datasketches/pom.xml | 2 +- extensions-core/druid-aws-rds-extensions/pom.xml | 2 +- extensions-core/druid-basic-security/pom.xml | 2 +- extensions-core/druid-bloom-filter/pom.xml | 2 +- extensions-core/druid-catalog/pom.xml | 2 +- extensions-core/druid-kerberos/pom.xml | 2 +- extensions-core/druid-pac4j/pom.xml | 2 +- extensions-core/druid-ranger-security/pom.xml | 2 +- extensions-core/ec2-extensions/pom.xml | 2 +- extensions-core/google-extensions/pom.xml | 2 +- extensions-core/hdfs-storage/pom.xml | 2 +- extensions-core/histogram/pom.xml | 2 +- extensions-core/kafka-extraction-namespace/pom.xml | 2 +- extensions-core/kafka-indexing-service/pom.xml | 2 +- extensions-core/kinesis-indexing-service/pom.xml | 2 +- extensions-core/kubernetes-extensions/pom.xml | 2 +- extensions-core/lookups-cached-global/pom.xml | 2 +- extensions-core/lookups-cached-single/pom.xml | 2 +- extensions-core/multi-stage-query/pom.xml | 2 +- extensions-core/mysql-metadata-storage/pom.xml | 2 +- extensions-core/orc-extensions/pom.xml | 2 +- extensions-core/parquet-extensions/pom.xml | 2 +- extensions-core/postgresql-metadata-storage/pom.xml | 2 +- extensions-core/protobuf-extensions/pom.xml | 2 +- extensions-core/s3-extensions/pom.xml | 2 +- extensions-core/simple-client-sslcontext/pom.xml | 2 +- extensions-core/stats/pom.xml | 2 +- extensions-core/testing-tools/pom.xml | 2 +- hll/pom.xml | 2 +- indexing-hadoop/pom.xml | 2 +- indexing-service/pom.xml | 2 +- integration-tests-ex/cases/pom.xml | 4 ++-- integration-tests-ex/image/pom.xml | 2 +- integration-tests-ex/tools/pom.xml | 2 +- integration-tests/pom.xml | 2 +- pom.xml | 2 +- processing/pom.xml | 2 +- server/pom.xml | 2 +- services/pom.xml | 2 +- sql/pom.xml | 2 +- web-console/package-lock.json | 4 ++-- web-console/package.json | 2 +- web-console/pom.xml | 2 +- web-console/unified-console.html | 2 +- website/pom.xml | 2 +- 81 files changed, 83 insertions(+), 83 deletions(-) diff --git a/benchmarks/pom.xml b/benchmarks/pom.xml index 842fe65f86ce..39590175727e 100644 --- a/benchmarks/pom.xml +++ b/benchmarks/pom.xml @@ -27,7 +27,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT diff --git a/cloud/aws-common/pom.xml b/cloud/aws-common/pom.xml index 73c0288dfad3..ebfe95bcfe84 100644 --- a/cloud/aws-common/pom.xml +++ b/cloud/aws-common/pom.xml @@ -28,7 +28,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/cloud/gcp-common/pom.xml b/cloud/gcp-common/pom.xml index d755b7cf844e..2bfa4ec33165 100644 --- a/cloud/gcp-common/pom.xml +++ b/cloud/gcp-common/pom.xml @@ -28,7 +28,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/core/pom.xml b/core/pom.xml index 2f8b9dc0b1d2..6727aa4ca0e3 100644 --- a/core/pom.xml +++ b/core/pom.xml @@ -29,7 +29,7 @@ druid org.apache.druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT diff --git a/distribution/docker/docker-compose.yml b/distribution/docker/docker-compose.yml index 58b7a47a4b4a..ea9adcea0d70 100644 --- a/distribution/docker/docker-compose.yml +++ b/distribution/docker/docker-compose.yml @@ -49,7 +49,7 @@ services: - ZOO_MY_ID=1 coordinator: - image: apache/druid:0.24.0 + image: apache/druid:24.0.1 container_name: coordinator volumes: - druid_shared:/opt/shared diff --git a/distribution/pom.xml b/distribution/pom.xml index e2b7773b09d4..7b92e3436b45 100644 --- a/distribution/pom.xml +++ b/distribution/pom.xml @@ -30,7 +30,7 @@ druid org.apache.druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT diff --git a/extendedset/pom.xml b/extendedset/pom.xml index bdd52ead1656..1927d59cff52 100755 --- a/extendedset/pom.xml +++ b/extendedset/pom.xml @@ -31,7 +31,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT diff --git a/extensions-contrib/aliyun-oss-extensions/pom.xml b/extensions-contrib/aliyun-oss-extensions/pom.xml index 4fdde743a947..9336122ed5ff 100644 --- a/extensions-contrib/aliyun-oss-extensions/pom.xml +++ b/extensions-contrib/aliyun-oss-extensions/pom.xml @@ -28,7 +28,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-contrib/ambari-metrics-emitter/pom.xml b/extensions-contrib/ambari-metrics-emitter/pom.xml index a267edc55485..fc7f73be4a6c 100644 --- a/extensions-contrib/ambari-metrics-emitter/pom.xml +++ b/extensions-contrib/ambari-metrics-emitter/pom.xml @@ -24,7 +24,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-contrib/cassandra-storage/pom.xml b/extensions-contrib/cassandra-storage/pom.xml index d9ed4ccff56b..2dd7d409143e 100644 --- a/extensions-contrib/cassandra-storage/pom.xml +++ b/extensions-contrib/cassandra-storage/pom.xml @@ -29,7 +29,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-contrib/cloudfiles-extensions/pom.xml b/extensions-contrib/cloudfiles-extensions/pom.xml index 3bd9c3220371..e3c13d1962cc 100644 --- a/extensions-contrib/cloudfiles-extensions/pom.xml +++ b/extensions-contrib/cloudfiles-extensions/pom.xml @@ -29,7 +29,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-contrib/compressed-bigdecimal/pom.xml b/extensions-contrib/compressed-bigdecimal/pom.xml index a9929c3c949c..f4309fa2d26a 100644 --- a/extensions-contrib/compressed-bigdecimal/pom.xml +++ b/extensions-contrib/compressed-bigdecimal/pom.xml @@ -25,7 +25,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-contrib/distinctcount/pom.xml b/extensions-contrib/distinctcount/pom.xml index 7dd0b77f69be..35ffdd9510d2 100644 --- a/extensions-contrib/distinctcount/pom.xml +++ b/extensions-contrib/distinctcount/pom.xml @@ -29,7 +29,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-contrib/dropwizard-emitter/pom.xml b/extensions-contrib/dropwizard-emitter/pom.xml index 714ac2090d9a..0ebe49754660 100644 --- a/extensions-contrib/dropwizard-emitter/pom.xml +++ b/extensions-contrib/dropwizard-emitter/pom.xml @@ -24,7 +24,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-contrib/gce-extensions/pom.xml b/extensions-contrib/gce-extensions/pom.xml index 3a3fb73b1be0..aebb755c0b1d 100644 --- a/extensions-contrib/gce-extensions/pom.xml +++ b/extensions-contrib/gce-extensions/pom.xml @@ -21,7 +21,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml 4.0.0 diff --git a/extensions-contrib/graphite-emitter/pom.xml b/extensions-contrib/graphite-emitter/pom.xml index e44036db513f..44759f8f4653 100644 --- a/extensions-contrib/graphite-emitter/pom.xml +++ b/extensions-contrib/graphite-emitter/pom.xml @@ -24,7 +24,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-contrib/influx-extensions/pom.xml b/extensions-contrib/influx-extensions/pom.xml index 326cea69f81e..aab1f8c05cca 100644 --- a/extensions-contrib/influx-extensions/pom.xml +++ b/extensions-contrib/influx-extensions/pom.xml @@ -29,7 +29,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-contrib/influxdb-emitter/pom.xml b/extensions-contrib/influxdb-emitter/pom.xml index 01381b799c32..12dc1b34bba8 100644 --- a/extensions-contrib/influxdb-emitter/pom.xml +++ b/extensions-contrib/influxdb-emitter/pom.xml @@ -28,7 +28,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-contrib/kafka-emitter/pom.xml b/extensions-contrib/kafka-emitter/pom.xml index 38ef7efcf5ba..b1f575d9020d 100644 --- a/extensions-contrib/kafka-emitter/pom.xml +++ b/extensions-contrib/kafka-emitter/pom.xml @@ -24,7 +24,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-contrib/kubernetes-overlord-extensions/pom.xml b/extensions-contrib/kubernetes-overlord-extensions/pom.xml index ffd94574111d..9503eeeb6ced 100644 --- a/extensions-contrib/kubernetes-overlord-extensions/pom.xml +++ b/extensions-contrib/kubernetes-overlord-extensions/pom.xml @@ -30,7 +30,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-contrib/materialized-view-maintenance/pom.xml b/extensions-contrib/materialized-view-maintenance/pom.xml index 449d2324aa7c..c8057bc78c57 100644 --- a/extensions-contrib/materialized-view-maintenance/pom.xml +++ b/extensions-contrib/materialized-view-maintenance/pom.xml @@ -22,7 +22,7 @@ druid org.apache.druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml 4.0.0 diff --git a/extensions-contrib/materialized-view-selection/pom.xml b/extensions-contrib/materialized-view-selection/pom.xml index c65bccd0bc1b..d23344f1fd02 100644 --- a/extensions-contrib/materialized-view-selection/pom.xml +++ b/extensions-contrib/materialized-view-selection/pom.xml @@ -22,7 +22,7 @@ druid org.apache.druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml 4.0.0 diff --git a/extensions-contrib/momentsketch/pom.xml b/extensions-contrib/momentsketch/pom.xml index 1d6afba02d1a..08843ed1669f 100644 --- a/extensions-contrib/momentsketch/pom.xml +++ b/extensions-contrib/momentsketch/pom.xml @@ -22,7 +22,7 @@ druid org.apache.druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml 4.0.0 diff --git a/extensions-contrib/moving-average-query/pom.xml b/extensions-contrib/moving-average-query/pom.xml index d53c79b3f963..f686cdfc6a4e 100644 --- a/extensions-contrib/moving-average-query/pom.xml +++ b/extensions-contrib/moving-average-query/pom.xml @@ -24,7 +24,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-contrib/opentelemetry-emitter/pom.xml b/extensions-contrib/opentelemetry-emitter/pom.xml index 5a1c4027e7be..8bdb7b3dd5d8 100644 --- a/extensions-contrib/opentelemetry-emitter/pom.xml +++ b/extensions-contrib/opentelemetry-emitter/pom.xml @@ -23,7 +23,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml 4.0.0 diff --git a/extensions-contrib/opentsdb-emitter/pom.xml b/extensions-contrib/opentsdb-emitter/pom.xml index 18fd9c61442d..2d68ecf1bca9 100644 --- a/extensions-contrib/opentsdb-emitter/pom.xml +++ b/extensions-contrib/opentsdb-emitter/pom.xml @@ -29,7 +29,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-contrib/prometheus-emitter/pom.xml b/extensions-contrib/prometheus-emitter/pom.xml index 63e7fba6d528..8d5b13aa7281 100644 --- a/extensions-contrib/prometheus-emitter/pom.xml +++ b/extensions-contrib/prometheus-emitter/pom.xml @@ -23,7 +23,7 @@ druid org.apache.druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml 4.0.0 diff --git a/extensions-contrib/redis-cache/pom.xml b/extensions-contrib/redis-cache/pom.xml index f7ec74e8b28b..e6551e6b792b 100644 --- a/extensions-contrib/redis-cache/pom.xml +++ b/extensions-contrib/redis-cache/pom.xml @@ -29,7 +29,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-contrib/sqlserver-metadata-storage/pom.xml b/extensions-contrib/sqlserver-metadata-storage/pom.xml index 67c78fbb81c6..388f40cb2955 100644 --- a/extensions-contrib/sqlserver-metadata-storage/pom.xml +++ b/extensions-contrib/sqlserver-metadata-storage/pom.xml @@ -28,7 +28,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-contrib/statsd-emitter/pom.xml b/extensions-contrib/statsd-emitter/pom.xml index e4aad827e271..ea0f5ba231e9 100644 --- a/extensions-contrib/statsd-emitter/pom.xml +++ b/extensions-contrib/statsd-emitter/pom.xml @@ -21,7 +21,7 @@ druid org.apache.druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml 4.0.0 diff --git a/extensions-contrib/tdigestsketch/pom.xml b/extensions-contrib/tdigestsketch/pom.xml index 7d271fb06708..76cf7b4055d0 100644 --- a/extensions-contrib/tdigestsketch/pom.xml +++ b/extensions-contrib/tdigestsketch/pom.xml @@ -22,7 +22,7 @@ druid org.apache.druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml 4.0.0 diff --git a/extensions-contrib/thrift-extensions/pom.xml b/extensions-contrib/thrift-extensions/pom.xml index 70aef1f61a93..5d61693a2868 100644 --- a/extensions-contrib/thrift-extensions/pom.xml +++ b/extensions-contrib/thrift-extensions/pom.xml @@ -28,7 +28,7 @@ druid org.apache.druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml 4.0.0 diff --git a/extensions-contrib/time-min-max/pom.xml b/extensions-contrib/time-min-max/pom.xml index a11c79367930..d4767a94f30a 100644 --- a/extensions-contrib/time-min-max/pom.xml +++ b/extensions-contrib/time-min-max/pom.xml @@ -21,7 +21,7 @@ druid org.apache.druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml 4.0.0 diff --git a/extensions-contrib/virtual-columns/pom.xml b/extensions-contrib/virtual-columns/pom.xml index abdc8606e45d..7ae88be798cf 100644 --- a/extensions-contrib/virtual-columns/pom.xml +++ b/extensions-contrib/virtual-columns/pom.xml @@ -29,7 +29,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-core/avro-extensions/pom.xml b/extensions-core/avro-extensions/pom.xml index 1c2dddb9902d..53dff82a980c 100644 --- a/extensions-core/avro-extensions/pom.xml +++ b/extensions-core/avro-extensions/pom.xml @@ -29,7 +29,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-core/azure-extensions/pom.xml b/extensions-core/azure-extensions/pom.xml index 6a01cea8a316..e34bacbe69ef 100644 --- a/extensions-core/azure-extensions/pom.xml +++ b/extensions-core/azure-extensions/pom.xml @@ -29,7 +29,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-core/datasketches/pom.xml b/extensions-core/datasketches/pom.xml index a37efd2dc215..83101d2e7c01 100644 --- a/extensions-core/datasketches/pom.xml +++ b/extensions-core/datasketches/pom.xml @@ -29,7 +29,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-core/druid-aws-rds-extensions/pom.xml b/extensions-core/druid-aws-rds-extensions/pom.xml index b18e1aea7214..f242e1e485c5 100644 --- a/extensions-core/druid-aws-rds-extensions/pom.xml +++ b/extensions-core/druid-aws-rds-extensions/pom.xml @@ -30,7 +30,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-core/druid-basic-security/pom.xml b/extensions-core/druid-basic-security/pom.xml index c8a3ab20e8ee..4bce2588c574 100644 --- a/extensions-core/druid-basic-security/pom.xml +++ b/extensions-core/druid-basic-security/pom.xml @@ -30,7 +30,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-core/druid-bloom-filter/pom.xml b/extensions-core/druid-bloom-filter/pom.xml index 0bc86e8e271e..4a457f92a129 100644 --- a/extensions-core/druid-bloom-filter/pom.xml +++ b/extensions-core/druid-bloom-filter/pom.xml @@ -29,7 +29,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-core/druid-catalog/pom.xml b/extensions-core/druid-catalog/pom.xml index 0678f95ff047..6a40adb17bc8 100644 --- a/extensions-core/druid-catalog/pom.xml +++ b/extensions-core/druid-catalog/pom.xml @@ -31,7 +31,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-core/druid-kerberos/pom.xml b/extensions-core/druid-kerberos/pom.xml index a2d248348222..9a36a98ba6eb 100644 --- a/extensions-core/druid-kerberos/pom.xml +++ b/extensions-core/druid-kerberos/pom.xml @@ -29,7 +29,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-core/druid-pac4j/pom.xml b/extensions-core/druid-pac4j/pom.xml index e3e675800438..5e4807def663 100644 --- a/extensions-core/druid-pac4j/pom.xml +++ b/extensions-core/druid-pac4j/pom.xml @@ -29,7 +29,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-core/druid-ranger-security/pom.xml b/extensions-core/druid-ranger-security/pom.xml index 39d42ccb2685..edb31fe29336 100644 --- a/extensions-core/druid-ranger-security/pom.xml +++ b/extensions-core/druid-ranger-security/pom.xml @@ -30,7 +30,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-core/ec2-extensions/pom.xml b/extensions-core/ec2-extensions/pom.xml index 57f41e3ce8f6..85e95f53a624 100644 --- a/extensions-core/ec2-extensions/pom.xml +++ b/extensions-core/ec2-extensions/pom.xml @@ -29,7 +29,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-core/google-extensions/pom.xml b/extensions-core/google-extensions/pom.xml index c29738a0cba3..ea6ab401f34c 100644 --- a/extensions-core/google-extensions/pom.xml +++ b/extensions-core/google-extensions/pom.xml @@ -29,7 +29,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-core/hdfs-storage/pom.xml b/extensions-core/hdfs-storage/pom.xml index 83a6fd8b1849..587474b5475f 100644 --- a/extensions-core/hdfs-storage/pom.xml +++ b/extensions-core/hdfs-storage/pom.xml @@ -29,7 +29,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-core/histogram/pom.xml b/extensions-core/histogram/pom.xml index 9dd821478071..67fdf2003d45 100644 --- a/extensions-core/histogram/pom.xml +++ b/extensions-core/histogram/pom.xml @@ -28,7 +28,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-core/kafka-extraction-namespace/pom.xml b/extensions-core/kafka-extraction-namespace/pom.xml index 29f39f4dc2e8..1fa023dd3e64 100644 --- a/extensions-core/kafka-extraction-namespace/pom.xml +++ b/extensions-core/kafka-extraction-namespace/pom.xml @@ -29,7 +29,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-core/kafka-indexing-service/pom.xml b/extensions-core/kafka-indexing-service/pom.xml index 61102acc2bee..05a5bd840595 100644 --- a/extensions-core/kafka-indexing-service/pom.xml +++ b/extensions-core/kafka-indexing-service/pom.xml @@ -29,7 +29,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-core/kinesis-indexing-service/pom.xml b/extensions-core/kinesis-indexing-service/pom.xml index 166b0291e00c..28533af40191 100644 --- a/extensions-core/kinesis-indexing-service/pom.xml +++ b/extensions-core/kinesis-indexing-service/pom.xml @@ -29,7 +29,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-core/kubernetes-extensions/pom.xml b/extensions-core/kubernetes-extensions/pom.xml index 80e110f231a9..cb25eee2abf6 100644 --- a/extensions-core/kubernetes-extensions/pom.xml +++ b/extensions-core/kubernetes-extensions/pom.xml @@ -30,7 +30,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-core/lookups-cached-global/pom.xml b/extensions-core/lookups-cached-global/pom.xml index b9538f7cacbb..79bff849cbf5 100644 --- a/extensions-core/lookups-cached-global/pom.xml +++ b/extensions-core/lookups-cached-global/pom.xml @@ -28,7 +28,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-core/lookups-cached-single/pom.xml b/extensions-core/lookups-cached-single/pom.xml index de98aaa38bd1..37fe9bf0e10d 100644 --- a/extensions-core/lookups-cached-single/pom.xml +++ b/extensions-core/lookups-cached-single/pom.xml @@ -28,7 +28,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-core/multi-stage-query/pom.xml b/extensions-core/multi-stage-query/pom.xml index f8382a967b09..443b792cca65 100644 --- a/extensions-core/multi-stage-query/pom.xml +++ b/extensions-core/multi-stage-query/pom.xml @@ -31,7 +31,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-core/mysql-metadata-storage/pom.xml b/extensions-core/mysql-metadata-storage/pom.xml index de6512d44d3c..1238d601dbd7 100644 --- a/extensions-core/mysql-metadata-storage/pom.xml +++ b/extensions-core/mysql-metadata-storage/pom.xml @@ -30,7 +30,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-core/orc-extensions/pom.xml b/extensions-core/orc-extensions/pom.xml index 57812e840d0d..9de5c3a91fd6 100644 --- a/extensions-core/orc-extensions/pom.xml +++ b/extensions-core/orc-extensions/pom.xml @@ -26,7 +26,7 @@ druid org.apache.druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml 4.0.0 diff --git a/extensions-core/parquet-extensions/pom.xml b/extensions-core/parquet-extensions/pom.xml index 2870b5c1dbef..ab834af8fa7a 100644 --- a/extensions-core/parquet-extensions/pom.xml +++ b/extensions-core/parquet-extensions/pom.xml @@ -27,7 +27,7 @@ druid org.apache.druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml 4.0.0 diff --git a/extensions-core/postgresql-metadata-storage/pom.xml b/extensions-core/postgresql-metadata-storage/pom.xml index f6c3b8534348..78aa3c0ac4ae 100644 --- a/extensions-core/postgresql-metadata-storage/pom.xml +++ b/extensions-core/postgresql-metadata-storage/pom.xml @@ -30,7 +30,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-core/protobuf-extensions/pom.xml b/extensions-core/protobuf-extensions/pom.xml index c7b7fc6e8b8c..0085f21aa564 100644 --- a/extensions-core/protobuf-extensions/pom.xml +++ b/extensions-core/protobuf-extensions/pom.xml @@ -29,7 +29,7 @@ druid org.apache.druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-core/s3-extensions/pom.xml b/extensions-core/s3-extensions/pom.xml index 14b6ff9f6c59..0b1eacf79bdb 100644 --- a/extensions-core/s3-extensions/pom.xml +++ b/extensions-core/s3-extensions/pom.xml @@ -29,7 +29,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-core/simple-client-sslcontext/pom.xml b/extensions-core/simple-client-sslcontext/pom.xml index 9dd441647558..ef5fb1bb836b 100644 --- a/extensions-core/simple-client-sslcontext/pom.xml +++ b/extensions-core/simple-client-sslcontext/pom.xml @@ -22,7 +22,7 @@ druid org.apache.druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml 4.0.0 diff --git a/extensions-core/stats/pom.xml b/extensions-core/stats/pom.xml index caddef0821da..7eddbc61ef32 100644 --- a/extensions-core/stats/pom.xml +++ b/extensions-core/stats/pom.xml @@ -29,7 +29,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/extensions-core/testing-tools/pom.xml b/extensions-core/testing-tools/pom.xml index f20c090969c9..c9a7c88d5600 100644 --- a/extensions-core/testing-tools/pom.xml +++ b/extensions-core/testing-tools/pom.xml @@ -29,7 +29,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/hll/pom.xml b/hll/pom.xml index 876257a84469..f39a6d33c461 100644 --- a/hll/pom.xml +++ b/hll/pom.xml @@ -24,7 +24,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT druid-hll diff --git a/indexing-hadoop/pom.xml b/indexing-hadoop/pom.xml index d93a8c698247..7f56e7f7562d 100644 --- a/indexing-hadoop/pom.xml +++ b/indexing-hadoop/pom.xml @@ -28,7 +28,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT diff --git a/indexing-service/pom.xml b/indexing-service/pom.xml index cba1cace7081..1ea333873843 100644 --- a/indexing-service/pom.xml +++ b/indexing-service/pom.xml @@ -29,7 +29,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT diff --git a/integration-tests-ex/cases/pom.xml b/integration-tests-ex/cases/pom.xml index 92a632d1f14c..5a35b1300055 100644 --- a/integration-tests-ex/cases/pom.xml +++ b/integration-tests-ex/cases/pom.xml @@ -31,7 +31,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml @@ -181,7 +181,7 @@ org.apache.druid druid-sql - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT org.apache.druid.extensions diff --git a/integration-tests-ex/image/pom.xml b/integration-tests-ex/image/pom.xml index 649367f056af..2fbeba3935a4 100644 --- a/integration-tests-ex/image/pom.xml +++ b/integration-tests-ex/image/pom.xml @@ -46,7 +46,7 @@ Reference: https://dzone.com/articles/build-docker-image-from-maven druid org.apache.druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/integration-tests-ex/tools/pom.xml b/integration-tests-ex/tools/pom.xml index fe7777ebf11f..aa79b786f80b 100644 --- a/integration-tests-ex/tools/pom.xml +++ b/integration-tests-ex/tools/pom.xml @@ -31,7 +31,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT ../../pom.xml diff --git a/integration-tests/pom.xml b/integration-tests/pom.xml index 4dc88f517f1a..ea80d69d0707 100644 --- a/integration-tests/pom.xml +++ b/integration-tests/pom.xml @@ -28,7 +28,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT diff --git a/pom.xml b/pom.xml index 9cddaf49cb3b..5e44487413d2 100644 --- a/pom.xml +++ b/pom.xml @@ -29,7 +29,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT pom Druid diff --git a/processing/pom.xml b/processing/pom.xml index c05e15499853..99d1b404e560 100644 --- a/processing/pom.xml +++ b/processing/pom.xml @@ -28,7 +28,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT diff --git a/server/pom.xml b/server/pom.xml index d4bafbb23b9f..c62533763bbb 100644 --- a/server/pom.xml +++ b/server/pom.xml @@ -28,7 +28,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT diff --git a/services/pom.xml b/services/pom.xml index d53a7229dd6b..f5b007ba1b50 100644 --- a/services/pom.xml +++ b/services/pom.xml @@ -27,7 +27,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT diff --git a/sql/pom.xml b/sql/pom.xml index 28f7eecffc48..9d260eddf4f7 100644 --- a/sql/pom.xml +++ b/sql/pom.xml @@ -28,7 +28,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT diff --git a/web-console/package-lock.json b/web-console/package-lock.json index 1821eb62bd7d..900bee41578a 100644 --- a/web-console/package-lock.json +++ b/web-console/package-lock.json @@ -1,12 +1,12 @@ { "name": "web-console", - "version": "25.0.0", + "version": "26.0.0", "lockfileVersion": 2, "requires": true, "packages": { "": { "name": "web-console", - "version": "25.0.0", + "version": "26.0.0", "license": "Apache-2.0", "dependencies": { "@blueprintjs/core": "^4.0.3", diff --git a/web-console/package.json b/web-console/package.json index 8e3f0f1b515e..97a03efface0 100644 --- a/web-console/package.json +++ b/web-console/package.json @@ -1,6 +1,6 @@ { "name": "web-console", - "version": "25.0.0", + "version": "26.0.0", "description": "A web console for Apache Druid", "author": "Apache Druid Developers ", "license": "Apache-2.0", diff --git a/web-console/pom.xml b/web-console/pom.xml index 6e8f48f1a142..b624919ee41c 100644 --- a/web-console/pom.xml +++ b/web-console/pom.xml @@ -28,7 +28,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT diff --git a/web-console/unified-console.html b/web-console/unified-console.html index 0f49d63d7024..e2b33fef8ada 100644 --- a/web-console/unified-console.html +++ b/web-console/unified-console.html @@ -71,6 +71,6 @@ }; - + diff --git a/website/pom.xml b/website/pom.xml index a8b7bb659bdb..a071223aeeff 100644 --- a/website/pom.xml +++ b/website/pom.xml @@ -28,7 +28,7 @@ org.apache.druid druid - 25.0.0-SNAPSHOT + 26.0.0-SNAPSHOT From a9b24e28ba1ace4a4abefa8fc1a4252d9b1a48fa Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Tue, 22 Nov 2022 15:49:21 +0530 Subject: [PATCH 022/141] resolve the config_path provided --- examples/bin/run-druid | 4 ++-- examples/bin/start-druid-main | 2 +- 2 files changed, 3 insertions(+), 3 deletions(-) diff --git a/examples/bin/run-druid b/examples/bin/run-druid index f6fd1cfd85bb..c2fd8056ba79 100755 --- a/examples/bin/run-druid +++ b/examples/bin/run-druid @@ -74,11 +74,11 @@ then if [ "$#" -eq 3 ] then # args: - exec "$WHEREAMI"/run-java -Ddruid.node.type=$1 $4 "-Ddruid.log.path=$LOG_DIR" $JVMARGS \ + exec "$WHEREAMI"/run-java -Ddruid.node.type=$1 "-Ddruid.log.path=$LOG_DIR" $JVMARGS \ -cp "$CONFDIR"/"$WHATAMI":"$CONFDIR"/_common:"$CONFDIR"/_common/hadoop-xml:"$CONFDIR"/../_common:"$CONFDIR"/../_common/hadoop-xml:"$WHEREAMI/../lib/*" org.apache.druid.cli.Main server $SERVER_NAME else # args: - exec "$WHEREAMI"/run-java -Ddruid.node.type=$1 $4"-Ddruid.log.path=$LOG_DIR" $JVMARGS \ + exec "$WHEREAMI"/run-java -Ddruid.node.type=$1 $4 "-Ddruid.log.path=$LOG_DIR" $JVMARGS \ -cp "$CONFDIR"/"$WHATAMI":"$CONFDIR"/_common:"$CONFDIR"/_common/hadoop-xml:"$CONFDIR"/../_common:"$CONFDIR"/../_common/hadoop-xml:"$WHEREAMI/../lib/*" org.apache.druid.cli.Main server $SERVER_NAME fi else diff --git a/examples/bin/start-druid-main b/examples/bin/start-druid-main index 86eec4b8eddf..e4d1b956883f 100644 --- a/examples/bin/start-druid-main +++ b/examples/bin/start-druid-main @@ -178,7 +178,7 @@ def parse_arguments(args): if args.run_zk: run_zk = True if args.config_path is not None: - config_path = os.path.join(os.getcwd(), args.config_path) + config_path = pathlib.Path(os.path.join(os.getcwd(), args.config_path)).resolve() if os.path.exists(config_path) is False: error_and_exit(f'config_path {config_path} doesn\'t exist') if args.memory is not None: From 71bfca1da0cc1d30103b14a4aee4e27b7bcd96d3 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Tue, 22 Nov 2022 16:02:11 +0530 Subject: [PATCH 023/141] comment out service specific runtime properties which are computed in the code --- .../single-server/quickstart/broker/runtime.properties | 10 +++++----- .../quickstart/historical/runtime.properties | 10 +++++----- .../quickstart/middleManager/runtime.properties | 6 +++--- .../single-server/quickstart/router/runtime.properties | 8 ++++---- 4 files changed, 17 insertions(+), 17 deletions(-) diff --git a/examples/conf/druid/single-server/quickstart/broker/runtime.properties b/examples/conf/druid/single-server/quickstart/broker/runtime.properties index 92b85f8bc754..549117efa50d 100644 --- a/examples/conf/druid/single-server/quickstart/broker/runtime.properties +++ b/examples/conf/druid/single-server/quickstart/broker/runtime.properties @@ -21,15 +21,15 @@ druid.service=druid/broker druid.plaintextPort=8082 # HTTP server settings -druid.server.http.numThreads=12 +#druid.server.http.numThreads=12 # HTTP client settings -druid.broker.http.numConnections=10 -druid.broker.http.maxQueuedBytes=5MiB +#druid.broker.http.numConnections=10 +#druid.broker.http.maxQueuedBytes=5MiB # Processing threads and buffers -druid.processing.buffer.sizeBytes=100MiB -druid.processing.numMergeBuffers=2 +#druid.processing.buffer.sizeBytes=100MiB +#druid.processing.numMergeBuffers=2 druid.processing.tmpDir=var/druid/processing # Query cache disabled -- push down caching and merging instead diff --git a/examples/conf/druid/single-server/quickstart/historical/runtime.properties b/examples/conf/druid/single-server/quickstart/historical/runtime.properties index 76eaa5d4b0cd..cf3e966cc30e 100644 --- a/examples/conf/druid/single-server/quickstart/historical/runtime.properties +++ b/examples/conf/druid/single-server/quickstart/historical/runtime.properties @@ -21,12 +21,12 @@ druid.service=druid/historical druid.plaintextPort=8083 # HTTP server threads -druid.server.http.numThreads=12 +#druid.server.http.numThreads=12 # Processing threads and buffers -druid.processing.buffer.sizeBytes=200MiB -druid.processing.numMergeBuffers=2 -druid.processing.numThreads=2 +#druid.processing.buffer.sizeBytes=200MiB +#druid.processing.numMergeBuffers=2 +#druid.processing.numThreads=2 druid.processing.tmpDir=var/druid/processing # Segment storage @@ -36,4 +36,4 @@ druid.segmentCache.locations=[{"path":"var/druid/segment-cache","maxSize":"300g" druid.historical.cache.useCache=true druid.historical.cache.populateCache=true druid.cache.type=caffeine -druid.cache.sizeInBytes=10MiB +#druid.cache.sizeInBytes=10MiB diff --git a/examples/conf/druid/single-server/quickstart/middleManager/runtime.properties b/examples/conf/druid/single-server/quickstart/middleManager/runtime.properties index b0cf3ca3c43e..df9b790a69b0 100644 --- a/examples/conf/druid/single-server/quickstart/middleManager/runtime.properties +++ b/examples/conf/druid/single-server/quickstart/middleManager/runtime.properties @@ -32,9 +32,9 @@ druid.indexer.task.baseTaskDir=var/druid/task #druid.server.http.numThreads=12 # Processing threads and buffers on Peons -druid.indexer.fork.property.druid.processing.numMergeBuffers=2 -druid.indexer.fork.property.druid.processing.buffer.sizeBytes=100MiB -druid.indexer.fork.property.druid.processing.numThreads=1 +#druid.indexer.fork.property.druid.processing.numMergeBuffers=2 +#druid.indexer.fork.property.druid.processing.buffer.sizeBytes=100MiB +#druid.indexer.fork.property.druid.processing.numThreads=1 # Hadoop indexing druid.indexer.task.hadoopWorkingPath=var/druid/hadoop-tmp diff --git a/examples/conf/druid/single-server/quickstart/router/runtime.properties b/examples/conf/druid/single-server/quickstart/router/runtime.properties index 497d3b4d3aa1..4c38ba88baab 100644 --- a/examples/conf/druid/single-server/quickstart/router/runtime.properties +++ b/examples/conf/druid/single-server/quickstart/router/runtime.properties @@ -21,10 +21,10 @@ druid.service=druid/router druid.plaintextPort=8888 # HTTP proxy -druid.router.http.numConnections=50 -druid.router.http.readTimeout=PT5M -druid.router.http.numMaxThreads=100 -druid.server.http.numThreads=100 +#druid.router.http.numConnections=50 +#druid.router.http.readTimeout=PT5M +#druid.router.http.numMaxThreads=100 +#druid.server.http.numThreads=100 # Service discovery druid.router.defaultBrokerServiceName=druid/broker From 032bae0f20042f0ed4038345d878cd6e52fd2dc3 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Tue, 22 Nov 2022 16:13:31 +0530 Subject: [PATCH 024/141] simplify run-druid --- examples/bin/run-druid | 12 +++++++----- 1 file changed, 7 insertions(+), 5 deletions(-) diff --git a/examples/bin/run-druid b/examples/bin/run-druid index c2fd8056ba79..4be0afc5854b 100755 --- a/examples/bin/run-druid +++ b/examples/bin/run-druid @@ -64,6 +64,8 @@ fi cd "$WHEREAMI/.." +CLASS_PATH="$CONFDIR"/"$WHATAMI":"$CONFDIR"/_common:"$CONFDIR"/_common/hadoop-xml:"$CONFDIR"/../_common:"$CONFDIR"/../_common/hadoop-xml:"$WHEREAMI/../lib/*" + if [ "$#" -eq 3 ] || [ "$#" -eq 4 ] then # args: or @@ -75,15 +77,15 @@ then then # args: exec "$WHEREAMI"/run-java -Ddruid.node.type=$1 "-Ddruid.log.path=$LOG_DIR" $JVMARGS \ - -cp "$CONFDIR"/"$WHATAMI":"$CONFDIR"/_common:"$CONFDIR"/_common/hadoop-xml:"$CONFDIR"/../_common:"$CONFDIR"/../_common/hadoop-xml:"$WHEREAMI/../lib/*" org.apache.druid.cli.Main server $SERVER_NAME + -cp $CLASS_PATH $MAIN_CLASS else # args: exec "$WHEREAMI"/run-java -Ddruid.node.type=$1 $4 "-Ddruid.log.path=$LOG_DIR" $JVMARGS \ - -cp "$CONFDIR"/"$WHATAMI":"$CONFDIR"/_common:"$CONFDIR"/_common/hadoop-xml:"$CONFDIR"/../_common:"$CONFDIR"/../_common/hadoop-xml:"$WHEREAMI/../lib/*" org.apache.druid.cli.Main server $SERVER_NAME + -cp $CLASS_PATH $MAIN_CLASS fi else # args: - exec "$WHEREAMI"/run-java -Ddruid.node.type=$1 "-Ddruid.log.path=$LOG_DIR" `cat "$CONFDIR"/"$WHATAMI"/jvm.config | xargs` \ - -cp "$CONFDIR"/"$WHATAMI":"$CONFDIR"/_common:"$CONFDIR"/_common/hadoop-xml:"$CONFDIR"/../_common:"$CONFDIR"/../_common/hadoop-xml:"$WHEREAMI/../lib/*" \ - $MAIN_CLASS + exec "$WHEREAMI"/run-java -Ddruid.node.type=$1 "-Ddruid.log.path=$LOG_DIR" \ + `cat "$CONFDIR"/"$WHATAMI"/jvm.config | xargs` \ + -cp $CLASS_PATH $MAIN_CLASS fi From be4914dcd9bea9e56f8c8ef5c966eb34b84739a1 Mon Sep 17 00:00:00 2001 From: Clint Wylie Date: Tue, 22 Nov 2022 12:46:06 -0800 Subject: [PATCH 025/141] fix off by one error in nested column range index (#13405) --- ...NestedFieldLiteralColumnIndexSupplier.java | 5 +- ...edFieldLiteralColumnIndexSupplierTest.java | 118 ++++++++++++++---- 2 files changed, 94 insertions(+), 29 deletions(-) diff --git a/processing/src/main/java/org/apache/druid/segment/nested/NestedFieldLiteralColumnIndexSupplier.java b/processing/src/main/java/org/apache/druid/segment/nested/NestedFieldLiteralColumnIndexSupplier.java index 373dcc822f2e..36c39f251b79 100644 --- a/processing/src/main/java/org/apache/druid/segment/nested/NestedFieldLiteralColumnIndexSupplier.java +++ b/processing/src/main/java/org/apache/druid/segment/nested/NestedFieldLiteralColumnIndexSupplier.java @@ -228,10 +228,10 @@ private IntIntPair getLocalRangeFromDictionary( // valid global index in local dictionary, start here localStartIndex = localFound; } - // global end index is exclusive already, so we don't adjust local end index even for missing values + int localEndFound = localDictionary.indexOf(globalEndIndex); if (localEndFound < 0) { - localEndIndex = -localEndFound; + localEndIndex = -(localEndFound + 1); } else { localEndIndex = localEndFound; } @@ -239,6 +239,7 @@ private IntIntPair getLocalRangeFromDictionary( localStartIndex = Math.min(localStartIndex, localDictionary.size()); localEndIndex = Math.max(localStartIndex, Math.min(localDictionary.size(), localEndIndex)); + return new IntIntImmutablePair(localStartIndex, localEndIndex); } diff --git a/processing/src/test/java/org/apache/druid/segment/nested/NestedFieldLiteralColumnIndexSupplierTest.java b/processing/src/test/java/org/apache/druid/segment/nested/NestedFieldLiteralColumnIndexSupplierTest.java index 15b40b729f07..25fec68bb82b 100644 --- a/processing/src/test/java/org/apache/druid/segment/nested/NestedFieldLiteralColumnIndexSupplierTest.java +++ b/processing/src/test/java/org/apache/druid/segment/nested/NestedFieldLiteralColumnIndexSupplierTest.java @@ -89,6 +89,9 @@ public void setup() throws IOException stringWriter.write("fo"); stringWriter.write("foo"); stringWriter.write("fooo"); + stringWriter.write("g"); + stringWriter.write("gg"); + stringWriter.write("ggg"); stringWriter.write("z"); writeToBuffer(stringBuffer, stringWriter); @@ -197,13 +200,44 @@ public void testSingleTypeStringColumnRangeIndex() throws IOException Assert.assertNotNull(rangeIndex); // 10 rows + // global: [null, a, b, fo, foo, fooo, g, gg, ggg, z] // local: [b, foo, fooo, z] // column: [foo, b, fooo, b, z, fooo, z, b, b, foo] - BitmapColumnIndex forRange = rangeIndex.forRange("f", true, "g", true); + BitmapColumnIndex forRange = rangeIndex.forRange(null, false, "a", false); Assert.assertNotNull(forRange); - Assert.assertEquals(0.4, forRange.estimateSelectivity(10), 0.0); + Assert.assertEquals(0.0, forRange.estimateSelectivity(10), 0.0); ImmutableBitmap bitmap = forRange.computeBitmapResult(bitmapResultFactory); + checkBitmap(bitmap); + + forRange = rangeIndex.forRange(null, false, "b", false); + Assert.assertNotNull(forRange); + Assert.assertEquals(0.4, forRange.estimateSelectivity(10), 0.0); + bitmap = forRange.computeBitmapResult(bitmapResultFactory); + checkBitmap(bitmap, 1, 3, 7, 8); + + forRange = rangeIndex.forRange("b", false, "fon", false); + Assert.assertNotNull(forRange); + Assert.assertEquals(0.4, forRange.estimateSelectivity(10), 0.0); + bitmap = forRange.computeBitmapResult(bitmapResultFactory); + checkBitmap(bitmap, 1, 3, 7, 8); + + forRange = rangeIndex.forRange("bb", false, "fon", false); + Assert.assertNotNull(forRange); + Assert.assertEquals(0.0, forRange.estimateSelectivity(10), 0.0); + bitmap = forRange.computeBitmapResult(bitmapResultFactory); + checkBitmap(bitmap); + + forRange = rangeIndex.forRange("b", true, "foo", false); + Assert.assertNotNull(forRange); + Assert.assertEquals(0.2, forRange.estimateSelectivity(10), 0.0); + bitmap = forRange.computeBitmapResult(bitmapResultFactory); + checkBitmap(bitmap, 0, 9); + + forRange = rangeIndex.forRange("f", true, "g", true); + Assert.assertNotNull(forRange); + Assert.assertEquals(0.4, forRange.estimateSelectivity(10), 0.0); + bitmap = forRange.computeBitmapResult(bitmapResultFactory); checkBitmap(bitmap, 0, 2, 5, 9); forRange = rangeIndex.forRange(null, false, "g", true); @@ -257,6 +291,36 @@ public void testSingleTypeStringColumnRangeIndex() throws IOException Assert.assertEquals(1.0, forRange.estimateSelectivity(10), 0.0); bitmap = forRange.computeBitmapResult(bitmapResultFactory); checkBitmap(bitmap, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9); + + forRange = rangeIndex.forRange(null, true, "foa", false); + Assert.assertEquals(0.4, forRange.estimateSelectivity(10), 0.0); + bitmap = forRange.computeBitmapResult(bitmapResultFactory); + checkBitmap(bitmap, 1, 3, 7, 8); + + forRange = rangeIndex.forRange(null, true, "foooa", false); + Assert.assertEquals(0.8, forRange.estimateSelectivity(10), 0.0); + bitmap = forRange.computeBitmapResult(bitmapResultFactory); + checkBitmap(bitmap, 0, 1, 2, 3, 5, 7, 8, 9); + + forRange = rangeIndex.forRange("foooa", true, "ggg", false); + Assert.assertEquals(0.0, forRange.estimateSelectivity(10), 0.0); + bitmap = forRange.computeBitmapResult(bitmapResultFactory); + checkBitmap(bitmap); + + forRange = rangeIndex.forRange("g", true, "gg", false); + Assert.assertEquals(0.0, forRange.estimateSelectivity(10), 0.0); + bitmap = forRange.computeBitmapResult(bitmapResultFactory); + checkBitmap(bitmap); + + forRange = rangeIndex.forRange("z", true, "zz", false); + Assert.assertEquals(0.0, forRange.estimateSelectivity(10), 0.0); + bitmap = forRange.computeBitmapResult(bitmapResultFactory); + checkBitmap(bitmap); + + forRange = rangeIndex.forRange("z", false, "zz", false); + Assert.assertEquals(0.2, forRange.estimateSelectivity(10), 0.0); + bitmap = forRange.computeBitmapResult(bitmapResultFactory); + checkBitmap(bitmap, 4, 6); } @Test @@ -1123,7 +1187,7 @@ private NestedFieldLiteralColumnIndexSupplier makeSingleTypeStringSupplier() // 10 rows // globals: [ - // [null, a, b, fo, foo, fooo, z], + // [null, a, b, fo, foo, fooo, g, gg, ggg, z], // [1, 2, 3, 5, 100, 300, 9000], // [1.0, 1.1, 1.2, 2.0, 2.5, 3.3, 6.6, 9.9] // ] @@ -1143,7 +1207,7 @@ private NestedFieldLiteralColumnIndexSupplier makeSingleTypeStringSupplier() bitmapWriter.write(fillBitmap(2, 5)); // z - localDictionaryWriter.write(6); + localDictionaryWriter.write(9); bitmapWriter.write(fillBitmap(4, 6)); writeToBuffer(localDictionaryBuffer, localDictionaryWriter); @@ -1193,7 +1257,7 @@ private NestedFieldLiteralColumnIndexSupplier makeSingleTypeStringWithNullsSu bitmapWriter.open(); // 10 rows // globals: [ - // [null, a, b, fo, foo, fooo, z], + // [null, a, b, fo, foo, fooo, g, gg, ggg, z], // [1, 2, 3, 5, 100, 300, 9000], // [1.0, 1.1, 1.2, 2.0, 2.5, 3.3, 6.6, 9.9] // ] @@ -1217,7 +1281,7 @@ private NestedFieldLiteralColumnIndexSupplier makeSingleTypeStringWithNullsSu bitmapWriter.write(fillBitmap(2, 5)); // z - localDictionaryWriter.write(6); + localDictionaryWriter.write(9); bitmapWriter.write(fillBitmap(4, 6)); writeToBuffer(localDictionaryBuffer, localDictionaryWriter); @@ -1276,19 +1340,19 @@ private NestedFieldLiteralColumnIndexSupplier makeSingleTypeLongSupplier() th // column: [100, 1, 300, 1, 3, 3, 100, 300, 300, 1] // 1 - localDictionaryWriter.write(7); + localDictionaryWriter.write(10); bitmapWriter.write(fillBitmap(1, 3, 9)); // 3 - localDictionaryWriter.write(9); + localDictionaryWriter.write(12); bitmapWriter.write(fillBitmap(4, 5)); // 100 - localDictionaryWriter.write(11); + localDictionaryWriter.write(14); bitmapWriter.write(fillBitmap(0, 6)); // 300 - localDictionaryWriter.write(12); + localDictionaryWriter.write(15); bitmapWriter.write(fillBitmap(2, 7, 8)); writeToBuffer(localDictionaryBuffer, localDictionaryWriter); @@ -1351,19 +1415,19 @@ private NestedFieldLiteralColumnIndexSupplier makeSingleTypeLongSupplierWithN bitmapWriter.write(fillBitmap(2, 5, 8)); // 1 - localDictionaryWriter.write(7); + localDictionaryWriter.write(10); bitmapWriter.write(fillBitmap(1, 3, 9)); // 3 - localDictionaryWriter.write(9); + localDictionaryWriter.write(12); bitmapWriter.write(fillBitmap(4)); // 100 - localDictionaryWriter.write(11); + localDictionaryWriter.write(14); bitmapWriter.write(fillBitmap(0, 6)); // 300 - localDictionaryWriter.write(12); + localDictionaryWriter.write(15); bitmapWriter.write(fillBitmap(7)); writeToBuffer(localDictionaryBuffer, localDictionaryWriter); @@ -1422,19 +1486,19 @@ private NestedFieldLiteralColumnIndexSupplier makeSingleTypeDoubleSupplier() // column: [1.1, 1.1, 1.2, 3.3, 1.2, 6.6, 3.3, 1.2, 1.1, 3.3] // 1.1 - localDictionaryWriter.write(15); + localDictionaryWriter.write(18); bitmapWriter.write(fillBitmap(0, 1, 8)); // 1.2 - localDictionaryWriter.write(16); + localDictionaryWriter.write(19); bitmapWriter.write(fillBitmap(2, 4, 7)); // 3.3 - localDictionaryWriter.write(19); + localDictionaryWriter.write(22); bitmapWriter.write(fillBitmap(3, 6, 9)); // 6.6 - localDictionaryWriter.write(20); + localDictionaryWriter.write(23); bitmapWriter.write(fillBitmap(5)); writeToBuffer(localDictionaryBuffer, localDictionaryWriter); @@ -1497,19 +1561,19 @@ private NestedFieldLiteralColumnIndexSupplier makeSingleTypeDoubleSupplierWit bitmapWriter.write(fillBitmap(1, 3, 6)); // 1.1 - localDictionaryWriter.write(15); + localDictionaryWriter.write(18); bitmapWriter.write(fillBitmap(0, 8)); // 1.2 - localDictionaryWriter.write(16); + localDictionaryWriter.write(19); bitmapWriter.write(fillBitmap(2, 4, 7)); // 3.3 - localDictionaryWriter.write(19); + localDictionaryWriter.write(22); bitmapWriter.write(fillBitmap(9)); // 6.6 - localDictionaryWriter.write(20); + localDictionaryWriter.write(23); bitmapWriter.write(fillBitmap(5)); writeToBuffer(localDictionaryBuffer, localDictionaryWriter); @@ -1576,23 +1640,23 @@ private NestedFieldLiteralColumnIndexSupplier makeVariantSupplierWithNull() t bitmapWriter.write(fillBitmap(1, 9)); // z - localDictionaryWriter.write(6); + localDictionaryWriter.write(9); bitmapWriter.write(fillBitmap(6)); // 1 - localDictionaryWriter.write(7); + localDictionaryWriter.write(10); bitmapWriter.write(fillBitmap(0, 5)); // 300 - localDictionaryWriter.write(12); + localDictionaryWriter.write(15); bitmapWriter.write(fillBitmap(4)); // 1.1 - localDictionaryWriter.write(15); + localDictionaryWriter.write(18); bitmapWriter.write(fillBitmap(8)); // 9.9 - localDictionaryWriter.write(21); + localDictionaryWriter.write(24); bitmapWriter.write(fillBitmap(3)); writeToBuffer(localDictionaryBuffer, localDictionaryWriter); From f55c74929ab72965d0b772e8572edecf96f6b788 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Wed, 23 Nov 2022 09:56:20 +0530 Subject: [PATCH 026/141] clean up docs, naming changes --- examples/bin/start-druid-main | 200 +++++++++++++++++----------------- 1 file changed, 98 insertions(+), 102 deletions(-) diff --git a/examples/bin/start-druid-main b/examples/bin/start-druid-main index e4d1b956883f..1447d8bfba4b 100644 --- a/examples/bin/start-druid-main +++ b/examples/bin/start-druid-main @@ -26,67 +26,66 @@ import argparse QUICKSTART_ROOT_CONFIG_PATH = "conf/druid/single-server/quickstart" -MEMORY_GIGABYTES_IDENTIFIER = "g" -MEMORY_MEGABYTES_IDENTIFIER = "m" +MEM_GB_SUFFIX = "g" +MEM_MB_SUFFIX = "m" SERVICE_SEPARATOR = "," -MM_TASK_JAVAOPTS_ARRAY = ["-server", "-Duser.timezone=UTC","-Dfile.encoding=UTF-8","-XX:+ExitOnOutOfMemoryError","-Djava.util.logging.manager=org.apache.logging.log4j.jul.LogManager"] -MM_TASK_MEMORY_TYPE_LOW = "low" -MM_TASK_MEMORY_TYPE_HIGH = "high" -MM_TASK_MEM_MAP = { - MM_TASK_MEMORY_TYPE_LOW: ["-Xms256m", "-Xmx256m", "-XX:MaxDirectMemorySize=256g"], - MM_TASK_MEMORY_TYPE_HIGH: ["-Xms1g", "-Xmx1g", "-XX:MaxDirectMemorySize=1g"] +TASK_JAVA_OPTS_ARRAY = ["-server", "-Duser.timezone=UTC", "-Dfile.encoding=UTF-8", "-XX:+ExitOnOutOfMemoryError", "-Djava.util.logging.manager=org.apache.logging.log4j.jul.LogManager"] +TASK_JAVA_OPTS_PROPERTY = "-Ddruid.indexer.runner.javaOptsArray" +TASK_WORKER_CAPACITY_PROPERTY = "-Ddruid.worker.capacity" +TASK_COUNT = "task-count" +TASK_MEM_TYPE_LOW = "low" +TASK_MEM_TYPE_HIGH = "high" +TASK_MEM_MAP = { + TASK_MEM_TYPE_LOW: ["-Xms256m", "-Xmx256m", "-XX:MaxDirectMemorySize=256g"], + TASK_MEM_TYPE_HIGH: ["-Xms1g", "-Xmx1g", "-XX:MaxDirectMemorySize=1g"] } -MM_TASK_JAVAOPTS_PROP = "-Ddruid.indexer.runner.javaOptsArray" -MM_TASK_WORKER_CAPACITY_PROP = "-Ddruid.worker.capacity" - BROKER = "broker" ROUTER = "router" COORDINATOR = "coordinator-overlord" HISTORICAL = "historical" -MM = "middleManager" -MM_TASK = "middleManager-task" -MM_TASK_COUNT = "task-count" +MIDDLE_MANAGER = "middleManager" +TASK = "task" DEFAULT_SERVICES = [ BROKER, ROUTER, COORDINATOR, HISTORICAL, - MM + MIDDLE_MANAGER ] -SERVICE_MEMORY_DISTRIBUTION_WEIGHT = { - MM: 0.5, - ROUTER: 1, - COORDINATOR: 15, - BROKER: 23, - HISTORICAL: 40, - MM_TASK: 15 +SERVICE_MEMORY_RATIO = { + MIDDLE_MANAGER: 1, + ROUTER: 2, + COORDINATOR: 30, + BROKER: 46, + HISTORICAL: 80, + TASK: 30 } -SERVICE_MEMORY_LOWER_BOUND = { - MM: 64, +MINIMUM_MEMORY_MB = { + MIDDLE_MANAGER: 64, ROUTER: 128, - MM_TASK: 1024, + TASK: 1024, BROKER: 900, COORDINATOR: 256, HISTORICAL: 900 } -SERVICE_MEMORY_HEAP_PERCENTAGE = { - MM: 1, +HEAP_TO_TOTAL_MEM_RATIO = { + MIDDLE_MANAGER: 1, ROUTER: 1, COORDINATOR: 1, BROKER: 0.60, HISTORICAL: 0.40, - MM_TASK: 0.50 + TASK: 0.50 } LOGGING_ENABLED = False -def custom_print(message): +def print_if_verbose(message): if LOGGING_ENABLED: print(message) @@ -103,58 +102,55 @@ def configure_parser(): sample usage: start-druid Start up all the services (including zk). - 50 - 80 percent of system memory is used. start-druid -m=100g Start up all the services (including zk) - using the given memory. + using a total memory of 100GB. start-druid -m=100g --compute_only - Compute memory distribution and validate - arguments. - start-druid -m=100g -sl=broker,router - Start broker & router service, using 100g of memory. - Read config from conf/druid/single-server/quickstart. - start-druid -m=100g --sl=broker,router \\ - -cp=conf/druid/single-server/custom - Start broker & router service, using 100g of memory. - Read config from . - Since is specified, exception is thrown if - jvm.config is present for any of the services. - start-druid -sl=broker,router \\ + Compute memory distribution and validate arguments. + start-druid -m=100g -s=broker,router + Starts a broker and a router, using a total memory of 100GB. + Reads config from conf/druid/single-server/quickstart. + start-druid -m=100g --s=broker,router \\ -cp=conf/druid/single-server/custom - Start broker & router service, using system memory. - If jvm.config is specified for both the - services within /, - memory distribution is not calculated. - If jvm.config is present for either of the services, - exception is thrown. - If jvm.config is not present for both of the services, - memory distribution is calculated. + Starts a broker and a router, using a total memory of 100GB. + Reads configs from the given config path. + Throws an exception if there is a jvm.config present in any + of the service directories inside the given root conf directory. + start-druid -s=broker,router \\ + -c=conf/druid/single-server/custom + Starts a broker and a router service, using upto 80% of the + total system memory. Calculates the memory requirements for + each service if no jvm.config is present for any of the + services. Does not calculate memory requirements if jvm.config + is present in all of the service directories. Throws an exception + if jvm.config is present in only some of the service directories. start-druid -m=100g \\ - -cp=conf/druid/single-server/custom \\ - -sl=broker,router \\ + -s=broker,router \\ + -c=conf/druid/single-server/custom \\ --run_zk - Start broker, router and zookeeper. - zk config is read from conf/zk. + Starts broker, router and zookeeper. + zookeeper config is read from conf/zk. """ ) parser.add_argument('--memory', '-m', type=str, required=False, help='Total memory for all processes (services and tasks, if any). \n' 'This parameter is ignored if each service already has a jvm.config \n' 'in the given conf directory. e.g. 500m, 4g, 6g\n') - parser.add_argument('--service_list', '-sl', type=str, required=False, + parser.add_argument('--services', '-s', type=str, required=False, help='List of services to be started, subset of \n' '{broker, router, middleManager, historical, coordinator-overlord}. \n' 'If the argument is not given, all services \n' 'and zookeeper is started. e.g. -sl=broker,historical') - parser.add_argument('--config_path', '-cp', type=str, required=False, + parser.add_argument('--config', '-c', type=str, required=False, help='Relative path to the directory containing common and service \n' 'specific properties to be overridden. \n' 'This directory must contain \'_common\' directory with \n' '\'common.jvm.config\' & \'common.runtime.properties\' files. \n' 'If this argument is not given, config from \n' - '\'conf/druid/single-server/quickstart\' directory is used.\n') + 'conf/druid/single-server/quickstart directory is used.\n') parser.add_argument('--compute_only', action='store_true', - help='Validate the arguments and display memory distribution') + help='Does not start Druid, only displays the memory allocated \n' + 'to each service if started with the given total memory.\n') parser.add_argument('--run_zk', action='store_true', help='Specification to run zookeeper, \n' 'zk config is picked up from conf/zk.') @@ -177,14 +173,14 @@ def parse_arguments(args): compute_only = True if args.run_zk: run_zk = True - if args.config_path is not None: - config_path = pathlib.Path(os.path.join(os.getcwd(), args.config_path)).resolve() - if os.path.exists(config_path) is False: - error_and_exit(f'config_path {config_path} doesn\'t exist') + if args.config is not None: + config = pathlib.Path(os.path.join(os.getcwd(), args.config)).resolve() + if os.path.exists(config) is False: + error_and_exit(f'config {config} doesn\'t exist') if args.memory is not None: total_memory = args.memory - if args.service_list is not None: - services = args.service_list.split(SERVICE_SEPARATOR) + if args.services is not None: + services = args.services.split(SERVICE_SEPARATOR) for service in services: if service not in DEFAULT_SERVICES: @@ -203,12 +199,12 @@ def parse_arguments(args): return config_path, total_memory, service_list, run_zk, compute_only def print_startup_config(service_list, config_path, run_zk): - custom_print(f'starting {service_list}') - custom_print(f'reading config from {config_path}') + print_if_verbose(f'starting {service_list}') + print_if_verbose(f'reading config from {config_path}') if run_zk: zk_config_path = pathlib.Path(f'{os.getcwd()}/../conf/zk').resolve() - custom_print(f'starting zk, reading default config from {zk_config_path}') - custom_print('\n') + print_if_verbose(f'starting zk, reading default config from {zk_config_path}') + print_if_verbose('\n') def should_compute_memory(config_path, total_memory, service_list): # if jvm file is present for any of the services @@ -239,9 +235,9 @@ def convert_total_memory_string(memory): if memory == "": computed_memory = compute_system_memory() return computed_memory - elif memory.endswith(MEMORY_MEGABYTES_IDENTIFIER): + elif memory.endswith(MEM_MB_SUFFIX): return int(memory[:-1]) - elif memory.endswith(MEMORY_GIGABYTES_IDENTIFIER): + elif memory.endswith(MEM_GB_SUFFIX): return 1024 * int(memory[:-1]) else: error_and_exit('Incorrect format for memory argument, expected format is ') @@ -253,7 +249,7 @@ def check_memory_constraint(total_memory, service_list): lower_bound_memory = 0 for service in service_list: - lower_bound_memory += SERVICE_MEMORY_LOWER_BOUND.get(service) + lower_bound_memory += MINIMUM_MEMORY_MB.get(service) required_memory = int(lower_bound_memory / 0.8) @@ -266,11 +262,11 @@ def check_memory_constraint(total_memory, service_list): return lower_bound_memory def build_mm_task_javaopts_array(memory_type): - task_memory = f'{MM_TASK_JAVAOPTS_PROP}=[' + task_memory = f'{TASK_JAVA_OPTS_PROPERTY}=[' - MEM_ARRAY = MM_TASK_MEM_MAP.get(memory_type) + MEM_ARRAY = TASK_MEM_MAP.get(memory_type) - javaopts_list = MM_TASK_JAVAOPTS_ARRAY + MEM_ARRAY + javaopts_list = TASK_JAVA_OPTS_ARRAY + MEM_ARRAY for item in javaopts_list: task_memory += f'\"{item}\";' @@ -280,21 +276,21 @@ def build_mm_task_javaopts_array(memory_type): return task_memory def build_memory_config_string(service, allocated_memory): - if service == MM_TASK: + if service == TASK: if allocated_memory >= 2048: task_count = int(allocated_memory / 2048) - memory_type = MM_TASK_MEMORY_TYPE_HIGH + memory_type = TASK_MEM_TYPE_HIGH task_memory = 2048 else: task_count = int(allocated_memory / 512) - memory_type = MM_TASK_MEMORY_TYPE_LOW + memory_type = TASK_MEM_TYPE_LOW task_memory = 512 task_count = min(task_count, multiprocessing.cpu_count()) javaopts_array = build_mm_task_javaopts_array(memory_type) - return [f'{MM_TASK_WORKER_CAPACITY_PROP}={task_count}', javaopts_array], task_memory * task_count + return [f'{TASK_WORKER_CAPACITY_PROPERTY}={task_count}', javaopts_array], task_memory * task_count else: - heap_memory = SERVICE_MEMORY_HEAP_PERCENTAGE.get(service) * allocated_memory + heap_memory = HEAP_TO_TOTAL_MEM_RATIO.get(service) * allocated_memory direct_memory = int(allocated_memory - heap_memory) heap_memory = int(heap_memory) @@ -309,11 +305,11 @@ def distribute_memory_over_services(services, total_memory): memory_weight_sum = 0 service_list = services.copy() - if MM in services: - service_list.append(MM_TASK) + if MIDDLE_MANAGER in services: + service_list.append(TASK) for service in service_list: - memory_weight_sum += SERVICE_MEMORY_DISTRIBUTION_WEIGHT.get(service) + memory_weight_sum += SERVICE_MEMORY_RATIO.get(service) multiplier = total_memory / memory_weight_sum @@ -321,9 +317,9 @@ def distribute_memory_over_services(services, total_memory): allocated_services = set() for service in service_list: - allocated_memory = SERVICE_MEMORY_DISTRIBUTION_WEIGHT.get(service) * multiplier - if service in SERVICE_MEMORY_LOWER_BOUND and allocated_memory < SERVICE_MEMORY_LOWER_BOUND.get(service): - allocated_memory = SERVICE_MEMORY_LOWER_BOUND.get(service) + allocated_memory = SERVICE_MEMORY_RATIO.get(service) * multiplier + if service in MINIMUM_MEMORY_MB and allocated_memory < MINIMUM_MEMORY_MB.get(service): + allocated_memory = MINIMUM_MEMORY_MB.get(service) service_memory_config[service], allocated_memory = build_memory_config_string(service, allocated_memory) lower_bound_memory_allocation += allocated_memory allocated_services.add(service) @@ -334,22 +330,22 @@ def distribute_memory_over_services(services, total_memory): for service in service_list: if service in allocated_services: continue - memory_weight_sum += SERVICE_MEMORY_DISTRIBUTION_WEIGHT.get(service) + memory_weight_sum += SERVICE_MEMORY_RATIO.get(service) multiplier = (total_memory - lower_bound_memory_allocation) / memory_weight_sum for service in service_list: if service in allocated_services: continue - allocated_memory = SERVICE_MEMORY_DISTRIBUTION_WEIGHT.get(service) * multiplier - if service in SERVICE_MEMORY_LOWER_BOUND and allocated_memory < SERVICE_MEMORY_LOWER_BOUND.get(service): - allocated_memory = SERVICE_MEMORY_LOWER_BOUND.get(service) + allocated_memory = SERVICE_MEMORY_RATIO.get(service) * multiplier + if service in MINIMUM_MEMORY_MB and allocated_memory < MINIMUM_MEMORY_MB.get(service): + allocated_memory = MINIMUM_MEMORY_MB.get(service) service_memory_config[service], allocated_memory = build_memory_config_string(service, allocated_memory) - custom_print(f'\nMemory distribution for services:') + print_if_verbose(f'\nMemory distribution for services:') for key, value in service_memory_config.items(): - custom_print(f'{key}, memory_config: {value}') - custom_print('\n') + print_if_verbose(f'{key}, memory_config: {value}') + print_if_verbose('\n') return service_memory_config @@ -371,14 +367,14 @@ def build_supervise_script_arguments(service_list, service_memory_config, config jvm_args = service_memory_config.get(service) prefix = '' - if service == MM: + if service == MIDDLE_MANAGER: prefix = '!p90 ' if jvm_args is None: commands.append(f'{prefix}{service} bin/run-druid {service} {config_path}') else: - if service == MM: - task_config = service_memory_config.get(MM_TASK) + if service == MIDDLE_MANAGER: + task_config = service_memory_config.get(TASK) task_count = task_config[0] task_memory = task_config[1] commands.append( @@ -386,11 +382,11 @@ def build_supervise_script_arguments(service_list, service_memory_config, config else: commands.append(f'{prefix}{service} bin/run-druid {service} {config_path} \'{jvm_args}\'') - custom_print('Supervise script args:') + print_if_verbose('Supervise script args:') for item in commands: - custom_print(item) + print_if_verbose(item) - custom_print('\n') + print_if_verbose('\n') args.append(",".join(commands)) return args @@ -416,12 +412,12 @@ def main(): if (should_compute_memory(config_path, total_memory, service_list)): memory_in_mega_bytes = convert_total_memory_string(total_memory) - custom_print(f'total memory is {memory_in_mega_bytes}m') + print_if_verbose(f'total memory is {memory_in_mega_bytes}m') memory_to_be_used = check_memory_constraint(memory_in_mega_bytes, service_list) - custom_print(f'memory used for services & tasks {memory_to_be_used}m') + print_if_verbose(f'memory used for services & tasks {memory_to_be_used}m') service_memory_config = distribute_memory_over_services(service_list, memory_to_be_used) else: - custom_print('not computing memory distribution, reading memory specification from service jvm.config\n') + print_if_verbose('not computing memory distribution, reading memory specification from service jvm.config\n') script_arguments = build_supervise_script_arguments(service_list, service_memory_config, config_path, run_zk) From 2a0dd346cdaaf44754f6b7f50205a8214498bb0d Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Wed, 23 Nov 2022 10:03:22 +0530 Subject: [PATCH 027/141] Throw ValueError exception on illegal state --- examples/bin/start-druid-main | 29 +++++++++++++---------------- 1 file changed, 13 insertions(+), 16 deletions(-) diff --git a/examples/bin/start-druid-main b/examples/bin/start-druid-main index 1447d8bfba4b..11c06aa6bc3d 100644 --- a/examples/bin/start-druid-main +++ b/examples/bin/start-druid-main @@ -89,10 +89,6 @@ def print_if_verbose(message): if LOGGING_ENABLED: print(message) -def error_and_exit(message): - sys.stderr.write(message + '\n') - sys.exit(1) - def configure_parser(): parser = argparse.ArgumentParser( prog='Druid quickstart', @@ -176,7 +172,7 @@ def parse_arguments(args): if args.config is not None: config = pathlib.Path(os.path.join(os.getcwd(), args.config)).resolve() if os.path.exists(config) is False: - error_and_exit(f'config {config} doesn\'t exist') + raise ValueError(f'config {config} doesn\'t exist') if args.memory is not None: total_memory = args.memory if args.services is not None: @@ -184,10 +180,10 @@ def parse_arguments(args): for service in services: if service not in DEFAULT_SERVICES: - error_and_exit(f'Invalid service name {service}, should be one of {DEFAULT_SERVICES}') + raise ValueError(f'Invalid service name {service}, should be one of {DEFAULT_SERVICES}') if service in service_list: - error_and_exit(f'{service} is specified multiple times') + raise ValueError(f'{service} is specified multiple times') service_list.append(service) @@ -215,13 +211,13 @@ def should_compute_memory(config_path, total_memory, service_list): if pathlib.Path(f'{config_path}/{service}/jvm.config').is_file(): jvm_config_count += 1 elif jvm_config_count > 0: - error_and_exit(f'jvm.config file is missing for service {service}, jvm.config should be specified for all the services or none') + raise ValueError(f'jvm.config file is missing for service {service}, jvm.config should be specified for all the services or none') if jvm_config_count > 0 and (jvm_config_count != len(service_list) or total_memory != ""): if jvm_config_count != len(service_list): - error_and_exit("jvm.config file should be present for all services or none") + raise ValueError("jvm.config file should be present for all services or none") if total_memory != "": - error_and_exit("If jvm.config is given for services, memory argument shouldn't be specified") + raise ValueError("If jvm.config is given for services, memory argument shouldn't be specified") return jvm_config_count == 0 @@ -240,9 +236,9 @@ def convert_total_memory_string(memory): elif memory.endswith(MEM_GB_SUFFIX): return 1024 * int(memory[:-1]) else: - error_and_exit('Incorrect format for memory argument, expected format is ') + raise ValueError('Incorrect format for memory argument, expected format is ') except Exception: - error_and_exit('Incorrect format for memory argument, expected format is ') + raise ValueError('Incorrect format for memory argument, expected format is ') def check_memory_constraint(total_memory, service_list): # 80% of total memory >= sum of lower bound service memory should be @@ -254,7 +250,7 @@ def check_memory_constraint(total_memory, service_list): required_memory = int(lower_bound_memory / 0.8) if total_memory < required_memory: - error_and_exit(f'Minimum memory required for starting services is {required_memory}m') + raise ValueError(f'Minimum memory required for starting services is {required_memory}m') if total_memory >= 2 * lower_bound_memory: return int(total_memory / 2) @@ -427,6 +423,7 @@ def main(): os.execv('./supervise', script_arguments) try: - main() -except KeyboardInterrupt: - sys.exit(1) + main() +except (KeyboardInterrupt, ValueError) as error: + print(error) + sys.exit(1) From c35cc1e79dfcad3699b9a1ef1835c870e9af08ea Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Wed, 23 Nov 2022 10:25:34 +0530 Subject: [PATCH 028/141] update docs --- examples/bin/start-druid-main | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/examples/bin/start-druid-main b/examples/bin/start-druid-main index 11c06aa6bc3d..6165a0e91e7f 100644 --- a/examples/bin/start-druid-main +++ b/examples/bin/start-druid-main @@ -241,7 +241,7 @@ def convert_total_memory_string(memory): raise ValueError('Incorrect format for memory argument, expected format is ') def check_memory_constraint(total_memory, service_list): - # 80% of total memory >= sum of lower bound service memory should be + # 80% of total memory >= sum of lower bound service memory lower_bound_memory = 0 for service in service_list: From c26b18c953e33b3bc886695e45836182357b593e Mon Sep 17 00:00:00 2001 From: Rohan Garg <7731512+rohangarg@users.noreply.github.com> Date: Wed, 23 Nov 2022 11:35:33 +0530 Subject: [PATCH 029/141] Port CVE suppressions from 24.0.1 (#13415) * Suppress jackson-databind CVE-2022-42003 and CVE-2022-42004 (cherry picked from commit 1f4d892c9a2dbc3ce6df1481fd4c6d242ba0ea8d) * Suppress CVEs (cherry picked from commit ed55baa8fa7d7f914a0addabb072d9ed47e1cd9f) * Suppress vulnerabilities from druid-website package (cherry picked from commit c0fb364f8049d53cd704e414e2ffeab6c49b012e) * Add more suppressions for website package (cherry picked from commit 9bba569ebd52c5480bf4219c420ed78eb053701f) --- owasp-dependency-check-suppressions.xml | 100 +++++++++++++++++++++++- 1 file changed, 97 insertions(+), 3 deletions(-) diff --git a/owasp-dependency-check-suppressions.xml b/owasp-dependency-check-suppressions.xml index 6ffb3b9f2e1f..a09ed507cc83 100644 --- a/owasp-dependency-check-suppressions.xml +++ b/owasp-dependency-check-suppressions.xml @@ -88,6 +88,17 @@ ^pkg:maven/net\.minidev/accessors\-smart@.*$ CVE-2021-27568 + + + + ^pkg:maven/com\.fasterxml\.jackson\.core/jackson\-databind@.*$ + CVE-2022-42003 + CVE-2022-42004 + @@ -209,6 +220,15 @@ CVE-2018-1320 CVE-2019-0205 + + + + ^pkg:maven/org\.codehaus\.jettison/jettison@1.*$ + CVE-2022-40149 + CVE-2022-40150 + CVE-2019-12399 CVE-2018-17196 + + + ^pkg:maven/org\.apache\.kafka/kafka\-clients@.*$ + CVE-2022-34917 + + + ^pkg:maven/org\.ini4j/ini4j@.*$ + CVE-2022-41404 - CVE-2022-36364 + CVE-2022-39135 + + + + ^pkg:maven/org\.apache\.calcite/calcite\-core@.*$ + CVE-2020-13955 - CVE-2022-31197 - 1084597 + + + ^pkg:npm/d3\-color@.*$ + 1084597 + + + + ^pkg:maven/com\.google\.protobuf/protobuf\-java@.*$ + CVE-2022-3171 + + + + ^pkg:maven/com\.google\.protobuf/protobuf\-java\-util@.*$ + CVE-2022-3171 + + + + ^pkg:npm/ansi\-regex@.*$ + 1084697 + CVE-2021-3807 + + + + ^pkg:npm/glob\-parent@.*$ + 1081884 + CVE-2020-28469 + + + + ^pkg:npm/minimatch@.*$ + 1084765 + + + + ^pkg:npm/y18n@.*$ + 1070209 + CVE-2020-7774 + From 891eaaa1308b249acbb235d0d17309ca559eeed6 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Wed, 23 Nov 2022 11:36:40 +0530 Subject: [PATCH 030/141] rename args, compute_only -> compute, run_zk -> zk --- examples/bin/start-druid-main | 74 +++++++++++++++++------------------ 1 file changed, 37 insertions(+), 37 deletions(-) diff --git a/examples/bin/start-druid-main b/examples/bin/start-druid-main index 6165a0e91e7f..5444ef4bab3c 100644 --- a/examples/bin/start-druid-main +++ b/examples/bin/start-druid-main @@ -101,13 +101,13 @@ sample usage: start-druid -m=100g Start up all the services (including zk) using a total memory of 100GB. - start-druid -m=100g --compute_only + start-druid -m=100g --compute Compute memory distribution and validate arguments. start-druid -m=100g -s=broker,router Starts a broker and a router, using a total memory of 100GB. Reads config from conf/druid/single-server/quickstart. start-druid -m=100g --s=broker,router \\ - -cp=conf/druid/single-server/custom + -c=conf/druid/single-server/custom Starts a broker and a router, using a total memory of 100GB. Reads configs from the given config path. Throws an exception if there is a jvm.config present in any @@ -123,7 +123,7 @@ sample usage: start-druid -m=100g \\ -s=broker,router \\ -c=conf/druid/single-server/custom \\ - --run_zk + --zk Starts broker, router and zookeeper. zookeeper config is read from conf/zk. """ @@ -144,31 +144,31 @@ sample usage: '\'common.jvm.config\' & \'common.runtime.properties\' files. \n' 'If this argument is not given, config from \n' 'conf/druid/single-server/quickstart directory is used.\n') - parser.add_argument('--compute_only', action='store_true', + parser.add_argument('--compute', action='store_true', help='Does not start Druid, only displays the memory allocated \n' 'to each service if started with the given total memory.\n') - parser.add_argument('--run_zk', action='store_true', + parser.add_argument('--zk', '-zk', action='store_true', help='Specification to run zookeeper, \n' 'zk config is picked up from conf/zk.') parser.add_argument('--verbose', action='store_true', help='Log details') - parser.set_defaults(run_zk=False) - parser.set_defaults(compute_only=False) + parser.set_defaults(zk=False) + parser.set_defaults(compute=False) parser.set_defaults(verbose=False) return parser def parse_arguments(args): service_list = [] - config_path = "" + config = "" total_memory = "" - compute_only = False - run_zk = False + compute = False + zk = False - if args.compute_only: - compute_only = True - if args.run_zk: - run_zk = True + if args.compute: + compute = True + if args.zk: + zk = True if args.config is not None: config = pathlib.Path(os.path.join(os.getcwd(), args.config)).resolve() if os.path.exists(config) is False: @@ -190,25 +190,25 @@ def parse_arguments(args): if len(service_list) == 0: # start all services service_list = DEFAULT_SERVICES - run_zk = True + zk = True - return config_path, total_memory, service_list, run_zk, compute_only + return config, total_memory, service_list, zk, compute -def print_startup_config(service_list, config_path, run_zk): +def print_startup_config(service_list, config, zk): print_if_verbose(f'starting {service_list}') - print_if_verbose(f'reading config from {config_path}') - if run_zk: - zk_config_path = pathlib.Path(f'{os.getcwd()}/../conf/zk').resolve() - print_if_verbose(f'starting zk, reading default config from {zk_config_path}') + print_if_verbose(f'reading config from {config}') + if zk: + zk_config = pathlib.Path(f'{os.getcwd()}/../conf/zk').resolve() + print_if_verbose(f'starting zk, reading default config from {zk_config}') print_if_verbose('\n') -def should_compute_memory(config_path, total_memory, service_list): +def should_compute_memory(config, total_memory, service_list): # if jvm file is present for any of the services # it should be present for all services and memory should not be specified # if memory is given, jvm file shouldn't be present for any service jvm_config_count = 0 for service in service_list: - if pathlib.Path(f'{config_path}/{service}/jvm.config').is_file(): + if pathlib.Path(f'{config}/{service}/jvm.config').is_file(): jvm_config_count += 1 elif jvm_config_count > 0: raise ValueError(f'jvm.config file is missing for service {service}, jvm.config should be specified for all the services or none') @@ -241,7 +241,7 @@ def convert_total_memory_string(memory): raise ValueError('Incorrect format for memory argument, expected format is ') def check_memory_constraint(total_memory, service_list): - # 80% of total memory >= sum of lower bound service memory + # 80% of total memory >= sum of lower bound service memory should be lower_bound_memory = 0 for service in service_list: @@ -345,7 +345,7 @@ def distribute_memory_over_services(services, total_memory): return service_memory_config -def build_supervise_script_arguments(service_list, service_memory_config, config_path, run_zk): +def build_supervise_script_arguments(service_list, service_memory_config, config, zk): args = [] commands = [] args.append('supervise') @@ -356,7 +356,7 @@ def build_supervise_script_arguments(service_list, service_memory_config, config commands.append(":notify bin/greet") commands.append(":kill-timeout 10") - if run_zk: + if zk: commands.append("!p10 zk bin/run-zk conf") for service in service_list: @@ -367,16 +367,16 @@ def build_supervise_script_arguments(service_list, service_memory_config, config prefix = '!p90 ' if jvm_args is None: - commands.append(f'{prefix}{service} bin/run-druid {service} {config_path}') + commands.append(f'{prefix}{service} bin/run-druid {service} {config}') else: if service == MIDDLE_MANAGER: task_config = service_memory_config.get(TASK) task_count = task_config[0] task_memory = task_config[1] commands.append( - f'{prefix}{service} bin/run-druid {service} {config_path} \'{jvm_args}\' \'{task_count} {task_memory}\'') + f'{prefix}{service} bin/run-druid {service} {config} \'{jvm_args}\' \'{task_count} {task_memory}\'') else: - commands.append(f'{prefix}{service} bin/run-druid {service} {config_path} \'{jvm_args}\'') + commands.append(f'{prefix}{service} bin/run-druid {service} {config} \'{jvm_args}\'') print_if_verbose('Supervise script args:') for item in commands: @@ -392,21 +392,21 @@ def main(): args = parser.parse_args() global LOGGING_ENABLED - LOGGING_ENABLED = args.verbose or args.compute_only + LOGGING_ENABLED = args.verbose or args.compute - config_path, total_memory, service_list, run_zk, compute_only = parse_arguments(args) + config, total_memory, service_list, zk, compute = parse_arguments(args) # change directory to bin os.chdir(os.path.dirname(sys.argv[0])) - if config_path == "": - config_path = pathlib.Path(f'{os.getcwd()}/../{QUICKSTART_ROOT_CONFIG_PATH}').resolve() + if config == "": + config = pathlib.Path(f'{os.getcwd()}/../{QUICKSTART_ROOT_CONFIG_PATH}').resolve() - print_startup_config(service_list, config_path, run_zk) + print_startup_config(service_list, config, zk) service_memory_config = {} - if (should_compute_memory(config_path, total_memory, service_list)): + if (should_compute_memory(config, total_memory, service_list)): memory_in_mega_bytes = convert_total_memory_string(total_memory) print_if_verbose(f'total memory is {memory_in_mega_bytes}m') memory_to_be_used = check_memory_constraint(memory_in_mega_bytes, service_list) @@ -415,9 +415,9 @@ def main(): else: print_if_verbose('not computing memory distribution, reading memory specification from service jvm.config\n') - script_arguments = build_supervise_script_arguments(service_list, service_memory_config, config_path, run_zk) + script_arguments = build_supervise_script_arguments(service_list, service_memory_config, config, zk) - if compute_only: + if compute: return os.execv('./supervise', script_arguments) From f524c68f08eb58047064e8234d320b51e83499ce Mon Sep 17 00:00:00 2001 From: Clint Wylie Date: Wed, 23 Nov 2022 00:25:22 -0800 Subject: [PATCH 031/141] Add mechanism for 'safe' memory reads for complex types (#13361) * we can read where we want to we can leave your bounds behind 'cause if the memory is not there we really don't care and we'll crash this process of mine --- .../hll/HllSketchMergeComplexMetricSerde.java | 15 +- .../hll/HllSketchObjectStrategy.java | 10 + .../KllDoublesSketchComplexMetricSerde.java | 2 +- .../kll/KllDoublesSketchObjectStrategy.java | 13 + .../kll/KllDoublesSketchOperations.java | 21 + .../KllFloatsSketchComplexMetricSerde.java | 2 +- .../kll/KllFloatsSketchObjectStrategy.java | 13 + .../kll/KllFloatsSketchOperations.java | 21 + .../DoublesSketchComplexMetricSerde.java | 2 +- .../DoublesSketchObjectStrategy.java | 13 + .../quantiles/DoublesSketchOperations.java | 20 + .../theta/SketchConstantPostAggregator.java | 2 +- .../datasketches/theta/SketchHolder.java | 22 + .../theta/SketchHolderObjectStrategy.java | 14 + .../theta/SketchMergeComplexMetricSerde.java | 2 +- ...fDoublesSketchMergeComplexMetricSerde.java | 2 +- .../ArrayOfDoublesSketchObjectStrategy.java | 13 +- .../tuple/ArrayOfDoublesSketchOperations.java | 24 +- .../hll/HllSketchObjectStrategyTest.java | 77 +++ ...llDoublesSketchComplexMetricSerdeTest.java | 44 ++ .../kll/KllDoublesSketchOperationsTest.java | 51 ++ ...KllFloatsSketchComplexMetricSerdeTest.java | 44 ++ .../kll/KllFloatsSketchOperationsTest.java | 51 ++ .../DoublesSketchComplexMetricSerdeTest.java | 43 ++ .../DoublesSketchOperationsTest.java | 50 ++ .../theta/SketchHolderObjectStrategyTest.java | 79 +++ .../datasketches/theta/SketchHolderTest.java | 52 ++ ...rrayOfDoublesSketchObjectStrategyTest.java | 70 +++ .../ArrayOfDoublesSketchOperationsTest.java | 55 ++ .../ObjectStrategyComplexTypeStrategy.java | 2 +- .../druid/segment/data/ObjectStrategy.java | 27 + .../druid/segment/data/SafeWritableBase.java | 450 ++++++++++++++++ .../segment/data/SafeWritableBuffer.java | 501 ++++++++++++++++++ .../segment/data/SafeWritableMemory.java | 417 +++++++++++++++ .../segment/data/SafeWritableBufferTest.java | 224 ++++++++ .../segment/data/SafeWritableMemoryTest.java | 359 +++++++++++++ 36 files changed, 2796 insertions(+), 11 deletions(-) create mode 100644 extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/hll/HllSketchObjectStrategyTest.java create mode 100644 extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/kll/KllDoublesSketchOperationsTest.java create mode 100644 extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/kll/KllFloatsSketchOperationsTest.java create mode 100644 extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/quantiles/DoublesSketchOperationsTest.java create mode 100644 extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/theta/SketchHolderObjectStrategyTest.java create mode 100644 extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/theta/SketchHolderTest.java create mode 100644 extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/tuple/ArrayOfDoublesSketchObjectStrategyTest.java create mode 100644 extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/tuple/ArrayOfDoublesSketchOperationsTest.java create mode 100644 processing/src/main/java/org/apache/druid/segment/data/SafeWritableBase.java create mode 100644 processing/src/main/java/org/apache/druid/segment/data/SafeWritableBuffer.java create mode 100644 processing/src/main/java/org/apache/druid/segment/data/SafeWritableMemory.java create mode 100644 processing/src/test/java/org/apache/druid/segment/data/SafeWritableBufferTest.java create mode 100644 processing/src/test/java/org/apache/druid/segment/data/SafeWritableMemoryTest.java diff --git a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/hll/HllSketchMergeComplexMetricSerde.java b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/hll/HllSketchMergeComplexMetricSerde.java index c8ac48ab186e..1063bbdfec1a 100644 --- a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/hll/HllSketchMergeComplexMetricSerde.java +++ b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/hll/HllSketchMergeComplexMetricSerde.java @@ -28,6 +28,7 @@ import org.apache.druid.segment.column.ColumnBuilder; import org.apache.druid.segment.data.GenericIndexed; import org.apache.druid.segment.data.ObjectStrategy; +import org.apache.druid.segment.data.SafeWritableMemory; import org.apache.druid.segment.serde.ComplexColumnPartSupplier; import org.apache.druid.segment.serde.ComplexMetricExtractor; import org.apache.druid.segment.serde.ComplexMetricSerde; @@ -70,7 +71,7 @@ public HllSketch extractValue(final InputRow inputRow, final String metricName) if (object == null) { return null; } - return deserializeSketch(object); + return deserializeSketchSafe(object); } }; } @@ -98,6 +99,18 @@ static HllSketch deserializeSketch(final Object object) throw new IAE("Object is not of a type that can be deserialized to an HllSketch:" + object.getClass().getName()); } + static HllSketch deserializeSketchSafe(final Object object) + { + if (object instanceof String) { + return HllSketch.wrap(SafeWritableMemory.wrap(StringUtils.decodeBase64(((String) object).getBytes(StandardCharsets.UTF_8)))); + } else if (object instanceof byte[]) { + return HllSketch.wrap(SafeWritableMemory.wrap((byte[]) object)); + } else if (object instanceof HllSketch) { + return (HllSketch) object; + } + throw new IAE("Object is not of a type that can be deserialized to an HllSketch:" + object.getClass().getName()); + } + // support large columns @Override public GenericColumnSerializer getSerializer(final SegmentWriteOutMedium segmentWriteOutMedium, final String column) diff --git a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/hll/HllSketchObjectStrategy.java b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/hll/HllSketchObjectStrategy.java index 34145863fdf8..65257b22b796 100644 --- a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/hll/HllSketchObjectStrategy.java +++ b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/hll/HllSketchObjectStrategy.java @@ -22,7 +22,9 @@ import org.apache.datasketches.hll.HllSketch; import org.apache.datasketches.memory.Memory; import org.apache.druid.segment.data.ObjectStrategy; +import org.apache.druid.segment.data.SafeWritableMemory; +import javax.annotation.Nullable; import java.nio.ByteBuffer; import java.nio.ByteOrder; @@ -55,4 +57,12 @@ public byte[] toBytes(final HllSketch sketch) return sketch.toCompactByteArray(); } + @Nullable + @Override + public HllSketch fromByteBufferSafe(ByteBuffer buffer, int numBytes) + { + return HllSketch.wrap( + SafeWritableMemory.wrap(buffer, ByteOrder.LITTLE_ENDIAN).region(buffer.position(), numBytes) + ); + } } diff --git a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/kll/KllDoublesSketchComplexMetricSerde.java b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/kll/KllDoublesSketchComplexMetricSerde.java index 4c18a9785607..e5249853ac34 100644 --- a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/kll/KllDoublesSketchComplexMetricSerde.java +++ b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/kll/KllDoublesSketchComplexMetricSerde.java @@ -91,7 +91,7 @@ public Object extractValue(final InputRow inputRow, final String metricName) if (object == null || object instanceof KllDoublesSketch || object instanceof Memory) { return object; } - return KllDoublesSketchOperations.deserialize(object); + return KllDoublesSketchOperations.deserializeSafe(object); } }; } diff --git a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/kll/KllDoublesSketchObjectStrategy.java b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/kll/KllDoublesSketchObjectStrategy.java index 97e670a625a5..17cb94e2fcf8 100644 --- a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/kll/KllDoublesSketchObjectStrategy.java +++ b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/kll/KllDoublesSketchObjectStrategy.java @@ -23,7 +23,9 @@ import org.apache.datasketches.kll.KllDoublesSketch; import org.apache.datasketches.memory.Memory; import org.apache.druid.segment.data.ObjectStrategy; +import org.apache.druid.segment.data.SafeWritableMemory; +import javax.annotation.Nullable; import java.nio.ByteBuffer; import java.nio.ByteOrder; @@ -60,4 +62,15 @@ public byte[] toBytes(final KllDoublesSketch sketch) return sketch.toByteArray(); } + @Nullable + @Override + public KllDoublesSketch fromByteBufferSafe(ByteBuffer buffer, int numBytes) + { + if (numBytes == 0) { + return KllDoublesSketchOperations.EMPTY_SKETCH; + } + return KllDoublesSketch.wrap( + SafeWritableMemory.wrap(buffer, ByteOrder.LITTLE_ENDIAN).region(buffer.position(), numBytes) + ); + } } diff --git a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/kll/KllDoublesSketchOperations.java b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/kll/KllDoublesSketchOperations.java index 57cb51747101..6da454d7f868 100644 --- a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/kll/KllDoublesSketchOperations.java +++ b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/kll/KllDoublesSketchOperations.java @@ -23,6 +23,7 @@ import org.apache.datasketches.memory.Memory; import org.apache.druid.java.util.common.ISE; import org.apache.druid.java.util.common.StringUtils; +import org.apache.druid.segment.data.SafeWritableMemory; import java.nio.charset.StandardCharsets; @@ -46,6 +47,16 @@ public static KllDoublesSketch deserialize(final Object serializedSketch) ); } + public static KllDoublesSketch deserializeSafe(final Object serializedSketch) + { + if (serializedSketch instanceof String) { + return deserializeFromBase64EncodedStringSafe((String) serializedSketch); + } else if (serializedSketch instanceof byte[]) { + return deserializeFromByteArraySafe((byte[]) serializedSketch); + } + return deserialize(serializedSketch); + } + public static KllDoublesSketch deserializeFromBase64EncodedString(final String str) { return deserializeFromByteArray(StringUtils.decodeBase64(str.getBytes(StandardCharsets.UTF_8))); @@ -56,4 +67,14 @@ public static KllDoublesSketch deserializeFromByteArray(final byte[] data) return KllDoublesSketch.wrap(Memory.wrap(data)); } + public static KllDoublesSketch deserializeFromBase64EncodedStringSafe(final String str) + { + return deserializeFromByteArraySafe(StringUtils.decodeBase64(str.getBytes(StandardCharsets.UTF_8))); + } + + public static KllDoublesSketch deserializeFromByteArraySafe(final byte[] data) + { + return KllDoublesSketch.wrap(SafeWritableMemory.wrap(data)); + } + } diff --git a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/kll/KllFloatsSketchComplexMetricSerde.java b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/kll/KllFloatsSketchComplexMetricSerde.java index 4a71befe0c66..175b307ec33e 100644 --- a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/kll/KllFloatsSketchComplexMetricSerde.java +++ b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/kll/KllFloatsSketchComplexMetricSerde.java @@ -91,7 +91,7 @@ public Object extractValue(final InputRow inputRow, final String metricName) if (object == null || object instanceof KllFloatsSketch || object instanceof Memory) { return object; } - return KllFloatsSketchOperations.deserialize(object); + return KllFloatsSketchOperations.deserializeSafe(object); } }; } diff --git a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/kll/KllFloatsSketchObjectStrategy.java b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/kll/KllFloatsSketchObjectStrategy.java index ff177a2f54f1..93ff0a7dba52 100644 --- a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/kll/KllFloatsSketchObjectStrategy.java +++ b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/kll/KllFloatsSketchObjectStrategy.java @@ -23,7 +23,9 @@ import org.apache.datasketches.kll.KllFloatsSketch; import org.apache.datasketches.memory.Memory; import org.apache.druid.segment.data.ObjectStrategy; +import org.apache.druid.segment.data.SafeWritableMemory; +import javax.annotation.Nullable; import java.nio.ByteBuffer; import java.nio.ByteOrder; @@ -60,4 +62,15 @@ public byte[] toBytes(final KllFloatsSketch sketch) return sketch.toByteArray(); } + @Nullable + @Override + public KllFloatsSketch fromByteBufferSafe(ByteBuffer buffer, int numBytes) + { + if (numBytes == 0) { + return KllFloatsSketchOperations.EMPTY_SKETCH; + } + return KllFloatsSketch.wrap( + SafeWritableMemory.wrap(buffer, ByteOrder.LITTLE_ENDIAN).region(buffer.position(), numBytes) + ); + } } diff --git a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/kll/KllFloatsSketchOperations.java b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/kll/KllFloatsSketchOperations.java index e32b67b2549a..02fb615da4dc 100644 --- a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/kll/KllFloatsSketchOperations.java +++ b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/kll/KllFloatsSketchOperations.java @@ -23,6 +23,7 @@ import org.apache.datasketches.memory.Memory; import org.apache.druid.java.util.common.ISE; import org.apache.druid.java.util.common.StringUtils; +import org.apache.druid.segment.data.SafeWritableMemory; import java.nio.charset.StandardCharsets; @@ -46,6 +47,16 @@ public static KllFloatsSketch deserialize(final Object serializedSketch) ); } + public static KllFloatsSketch deserializeSafe(final Object serializedSketch) + { + if (serializedSketch instanceof String) { + return deserializeFromBase64EncodedStringSafe((String) serializedSketch); + } else if (serializedSketch instanceof byte[]) { + return deserializeFromByteArraySafe((byte[]) serializedSketch); + } + return deserialize(serializedSketch); + } + public static KllFloatsSketch deserializeFromBase64EncodedString(final String str) { return deserializeFromByteArray(StringUtils.decodeBase64(str.getBytes(StandardCharsets.UTF_8))); @@ -56,4 +67,14 @@ public static KllFloatsSketch deserializeFromByteArray(final byte[] data) return KllFloatsSketch.wrap(Memory.wrap(data)); } + public static KllFloatsSketch deserializeFromBase64EncodedStringSafe(final String str) + { + return deserializeFromByteArraySafe(StringUtils.decodeBase64(str.getBytes(StandardCharsets.UTF_8))); + } + + public static KllFloatsSketch deserializeFromByteArraySafe(final byte[] data) + { + return KllFloatsSketch.wrap(SafeWritableMemory.wrap(data)); + } + } diff --git a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/quantiles/DoublesSketchComplexMetricSerde.java b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/quantiles/DoublesSketchComplexMetricSerde.java index d97b5f8c6dd3..3614f214c708 100644 --- a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/quantiles/DoublesSketchComplexMetricSerde.java +++ b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/quantiles/DoublesSketchComplexMetricSerde.java @@ -92,7 +92,7 @@ public Object extractValue(final InputRow inputRow, final String metricName) if (object == null || object instanceof DoublesSketch || object instanceof Memory) { return object; } - return DoublesSketchOperations.deserialize(object); + return DoublesSketchOperations.deserializeSafe(object); } }; } diff --git a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/quantiles/DoublesSketchObjectStrategy.java b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/quantiles/DoublesSketchObjectStrategy.java index 826de9378f8c..569b60bf03f7 100644 --- a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/quantiles/DoublesSketchObjectStrategy.java +++ b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/quantiles/DoublesSketchObjectStrategy.java @@ -23,7 +23,9 @@ import org.apache.datasketches.memory.Memory; import org.apache.datasketches.quantiles.DoublesSketch; import org.apache.druid.segment.data.ObjectStrategy; +import org.apache.druid.segment.data.SafeWritableMemory; +import javax.annotation.Nullable; import java.nio.ByteBuffer; import java.nio.ByteOrder; @@ -60,4 +62,15 @@ public byte[] toBytes(final DoublesSketch sketch) return sketch.toByteArray(true); } + @Nullable + @Override + public DoublesSketch fromByteBufferSafe(ByteBuffer buffer, int numBytes) + { + if (numBytes == 0) { + return DoublesSketchOperations.EMPTY_SKETCH; + } + return DoublesSketch.wrap( + SafeWritableMemory.wrap(buffer, ByteOrder.LITTLE_ENDIAN).region(buffer.position(), numBytes) + ); + } } diff --git a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/quantiles/DoublesSketchOperations.java b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/quantiles/DoublesSketchOperations.java index e30fb9bdae83..a2ca197c11ac 100644 --- a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/quantiles/DoublesSketchOperations.java +++ b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/quantiles/DoublesSketchOperations.java @@ -23,6 +23,7 @@ import org.apache.datasketches.quantiles.DoublesSketch; import org.apache.druid.java.util.common.ISE; import org.apache.druid.java.util.common.StringUtils; +import org.apache.druid.segment.data.SafeWritableMemory; import java.nio.charset.StandardCharsets; @@ -46,6 +47,16 @@ public static DoublesSketch deserialize(final Object serializedSketch) ); } + public static DoublesSketch deserializeSafe(final Object serializedSketch) + { + if (serializedSketch instanceof String) { + return deserializeFromBase64EncodedStringSafe((String) serializedSketch); + } else if (serializedSketch instanceof byte[]) { + return deserializeFromByteArraySafe((byte[]) serializedSketch); + } + return deserialize(serializedSketch); + } + public static DoublesSketch deserializeFromBase64EncodedString(final String str) { return deserializeFromByteArray(StringUtils.decodeBase64(str.getBytes(StandardCharsets.UTF_8))); @@ -56,4 +67,13 @@ public static DoublesSketch deserializeFromByteArray(final byte[] data) return DoublesSketch.wrap(Memory.wrap(data)); } + public static DoublesSketch deserializeFromBase64EncodedStringSafe(final String str) + { + return deserializeFromByteArraySafe(StringUtils.decodeBase64(str.getBytes(StandardCharsets.UTF_8))); + } + + public static DoublesSketch deserializeFromByteArraySafe(final byte[] data) + { + return DoublesSketch.wrap(SafeWritableMemory.wrap(data)); + } } diff --git a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/theta/SketchConstantPostAggregator.java b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/theta/SketchConstantPostAggregator.java index b3541bd50649..64c182a0d62a 100644 --- a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/theta/SketchConstantPostAggregator.java +++ b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/theta/SketchConstantPostAggregator.java @@ -51,7 +51,7 @@ public SketchConstantPostAggregator(@JsonProperty("name") String name, @JsonProp Preconditions.checkArgument(value != null && !value.isEmpty(), "Constant value cannot be null or empty, expecting base64 encoded sketch string"); this.value = value; - this.sketchValue = SketchHolder.deserialize(value); + this.sketchValue = SketchHolder.deserializeSafe(value); } @Override diff --git a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/theta/SketchHolder.java b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/theta/SketchHolder.java index 59ca453bb25a..838b4ae91f0f 100644 --- a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/theta/SketchHolder.java +++ b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/theta/SketchHolder.java @@ -34,6 +34,7 @@ import org.apache.druid.java.util.common.IAE; import org.apache.druid.java.util.common.ISE; import org.apache.druid.java.util.common.StringUtils; +import org.apache.druid.segment.data.SafeWritableMemory; import javax.annotation.Nullable; @@ -224,6 +225,17 @@ public static SketchHolder deserialize(Object serializedSketch) ); } + public static SketchHolder deserializeSafe(Object serializedSketch) + { + if (serializedSketch instanceof String) { + return SketchHolder.of(deserializeFromBase64EncodedStringSafe((String) serializedSketch)); + } else if (serializedSketch instanceof byte[]) { + return SketchHolder.of(deserializeFromByteArraySafe((byte[]) serializedSketch)); + } + + return deserialize(serializedSketch); + } + private static Sketch deserializeFromBase64EncodedString(String str) { return deserializeFromByteArray(StringUtils.decodeBase64(StringUtils.toUtf8(str))); @@ -234,6 +246,16 @@ private static Sketch deserializeFromByteArray(byte[] data) return deserializeFromMemory(Memory.wrap(data)); } + private static Sketch deserializeFromBase64EncodedStringSafe(String str) + { + return deserializeFromByteArraySafe(StringUtils.decodeBase64(StringUtils.toUtf8(str))); + } + + private static Sketch deserializeFromByteArraySafe(byte[] data) + { + return deserializeFromMemory(SafeWritableMemory.wrap(data)); + } + private static Sketch deserializeFromMemory(Memory mem) { if (Sketch.getSerializationVersion(mem) < 3) { diff --git a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/theta/SketchHolderObjectStrategy.java b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/theta/SketchHolderObjectStrategy.java index e98bc3d95a04..96fafe826211 100644 --- a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/theta/SketchHolderObjectStrategy.java +++ b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/theta/SketchHolderObjectStrategy.java @@ -23,6 +23,7 @@ import org.apache.datasketches.memory.Memory; import org.apache.datasketches.theta.Sketch; import org.apache.druid.segment.data.ObjectStrategy; +import org.apache.druid.segment.data.SafeWritableMemory; import javax.annotation.Nullable; import java.nio.ByteBuffer; @@ -66,4 +67,17 @@ public byte[] toBytes(@Nullable SketchHolder obj) return ByteArrays.EMPTY_ARRAY; } } + + @Nullable + @Override + public SketchHolder fromByteBufferSafe(ByteBuffer buffer, int numBytes) + { + if (numBytes == 0) { + return SketchHolder.EMPTY; + } + + return SketchHolder.of( + SafeWritableMemory.wrap(buffer, ByteOrder.LITTLE_ENDIAN).region(buffer.position(), numBytes) + ); + } } diff --git a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/theta/SketchMergeComplexMetricSerde.java b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/theta/SketchMergeComplexMetricSerde.java index a824312c0ef7..4f3ecfae2915 100644 --- a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/theta/SketchMergeComplexMetricSerde.java +++ b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/theta/SketchMergeComplexMetricSerde.java @@ -59,7 +59,7 @@ public Class extractedClass() public SketchHolder extractValue(InputRow inputRow, String metricName) { final Object object = inputRow.getRaw(metricName); - return object == null ? null : SketchHolder.deserialize(object); + return object == null ? null : SketchHolder.deserializeSafe(object); } }; } diff --git a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/tuple/ArrayOfDoublesSketchMergeComplexMetricSerde.java b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/tuple/ArrayOfDoublesSketchMergeComplexMetricSerde.java index 19c8da292b4e..028bcdc35491 100644 --- a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/tuple/ArrayOfDoublesSketchMergeComplexMetricSerde.java +++ b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/tuple/ArrayOfDoublesSketchMergeComplexMetricSerde.java @@ -60,7 +60,7 @@ public Object extractValue(final InputRow inputRow, final String metricName) if (object == null || object instanceof ArrayOfDoublesSketch) { return object; } - return ArrayOfDoublesSketchOperations.deserialize(object); + return ArrayOfDoublesSketchOperations.deserializeSafe(object); } }; } diff --git a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/tuple/ArrayOfDoublesSketchObjectStrategy.java b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/tuple/ArrayOfDoublesSketchObjectStrategy.java index 1ae950e068f8..f893c83b5707 100644 --- a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/tuple/ArrayOfDoublesSketchObjectStrategy.java +++ b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/tuple/ArrayOfDoublesSketchObjectStrategy.java @@ -23,6 +23,7 @@ import org.apache.datasketches.tuple.arrayofdoubles.ArrayOfDoublesSketch; import org.apache.datasketches.tuple.arrayofdoubles.ArrayOfDoublesSketches; import org.apache.druid.segment.data.ObjectStrategy; +import org.apache.druid.segment.data.SafeWritableMemory; import javax.annotation.Nullable; import java.nio.ByteBuffer; @@ -48,7 +49,9 @@ public Class getClazz() @Override public ArrayOfDoublesSketch fromByteBuffer(final ByteBuffer buffer, final int numBytes) { - return ArrayOfDoublesSketches.wrapSketch(Memory.wrap(buffer, ByteOrder.LITTLE_ENDIAN).region(buffer.position(), numBytes)); + return ArrayOfDoublesSketches.wrapSketch( + Memory.wrap(buffer, ByteOrder.LITTLE_ENDIAN).region(buffer.position(), numBytes) + ); } @Override @@ -61,4 +64,12 @@ public byte[] toBytes(@Nullable final ArrayOfDoublesSketch sketch) return sketch.toByteArray(); } + @Nullable + @Override + public ArrayOfDoublesSketch fromByteBufferSafe(ByteBuffer buffer, int numBytes) + { + return ArrayOfDoublesSketches.wrapSketch( + SafeWritableMemory.wrap(buffer, ByteOrder.LITTLE_ENDIAN).region(buffer.position(), numBytes) + ); + } } diff --git a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/tuple/ArrayOfDoublesSketchOperations.java b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/tuple/ArrayOfDoublesSketchOperations.java index b1658a995714..2768858ffe32 100644 --- a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/tuple/ArrayOfDoublesSketchOperations.java +++ b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/tuple/ArrayOfDoublesSketchOperations.java @@ -30,6 +30,7 @@ import org.apache.druid.java.util.common.IAE; import org.apache.druid.java.util.common.ISE; import org.apache.druid.java.util.common.StringUtils; +import org.apache.druid.segment.data.SafeWritableMemory; import java.nio.charset.StandardCharsets; @@ -115,6 +116,17 @@ public static ArrayOfDoublesSketch deserialize(final Object serializedSketch) throw new ISE("Object is not of a type that can deserialize to sketch: %s", serializedSketch.getClass()); } + public static ArrayOfDoublesSketch deserializeSafe(final Object serializedSketch) + { + if (serializedSketch instanceof String) { + return deserializeFromBase64EncodedStringSafe((String) serializedSketch); + } else if (serializedSketch instanceof byte[]) { + return deserializeFromByteArraySafe((byte[]) serializedSketch); + } + + return deserialize(serializedSketch); + } + public static ArrayOfDoublesSketch deserializeFromBase64EncodedString(final String str) { return deserializeFromByteArray(StringUtils.decodeBase64(str.getBytes(StandardCharsets.UTF_8))); @@ -122,8 +134,16 @@ public static ArrayOfDoublesSketch deserializeFromBase64EncodedString(final Stri public static ArrayOfDoublesSketch deserializeFromByteArray(final byte[] data) { - final Memory mem = Memory.wrap(data); - return ArrayOfDoublesSketches.wrapSketch(mem); + return ArrayOfDoublesSketches.wrapSketch(Memory.wrap(data)); + } + + public static ArrayOfDoublesSketch deserializeFromBase64EncodedStringSafe(final String str) + { + return deserializeFromByteArraySafe(StringUtils.decodeBase64(str.getBytes(StandardCharsets.UTF_8))); } + public static ArrayOfDoublesSketch deserializeFromByteArraySafe(final byte[] data) + { + return ArrayOfDoublesSketches.wrapSketch(SafeWritableMemory.wrap(data)); + } } diff --git a/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/hll/HllSketchObjectStrategyTest.java b/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/hll/HllSketchObjectStrategyTest.java new file mode 100644 index 000000000000..ff1eb947408e --- /dev/null +++ b/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/hll/HllSketchObjectStrategyTest.java @@ -0,0 +1,77 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.aggregation.datasketches.hll; + +import org.apache.datasketches.SketchesArgumentException; +import org.apache.datasketches.hll.HllSketch; +import org.apache.druid.java.util.common.StringUtils; +import org.junit.Assert; +import org.junit.Test; + +import java.nio.ByteBuffer; +import java.nio.ByteOrder; + +public class HllSketchObjectStrategyTest +{ + @Test + public void testSafeRead() + { + HllSketch sketch = new HllSketch(); + sketch.update(new int[]{1, 2, 3}); + + final byte[] bytes = sketch.toCompactByteArray(); + + ByteBuffer buf = ByteBuffer.wrap(bytes).order(ByteOrder.LITTLE_ENDIAN); + HllSketchObjectStrategy objectStrategy = new HllSketchObjectStrategy(); + + // valid sketch should not explode when copied, which reads the memory + objectStrategy.fromByteBufferSafe(buf, bytes.length).copy(); + + // corrupted sketch should fail with a regular java buffer exception + for (int subset = 3; subset < bytes.length - 1; subset++) { + final byte[] garbage2 = new byte[subset]; + for (int i = 0; i < garbage2.length; i++) { + garbage2[i] = buf.get(i); + } + + final ByteBuffer buf2 = ByteBuffer.wrap(garbage2).order(ByteOrder.LITTLE_ENDIAN); + Assert.assertThrows( + IndexOutOfBoundsException.class, + () -> objectStrategy.fromByteBufferSafe(buf2, garbage2.length).copy() + ); + } + + // non sketch that is too short to contain header should fail with regular java buffer exception + final byte[] garbage = new byte[]{0x01, 0x02}; + final ByteBuffer buf3 = ByteBuffer.wrap(garbage).order(ByteOrder.LITTLE_ENDIAN); + Assert.assertThrows( + IndexOutOfBoundsException.class, + () -> objectStrategy.fromByteBufferSafe(buf3, garbage.length).copy() + ); + + // non sketch that is long enough to check (this one doesn't actually need 'safe' read) + final byte[] garbageLonger = StringUtils.toUtf8("notasketch"); + final ByteBuffer buf4 = ByteBuffer.wrap(garbageLonger).order(ByteOrder.LITTLE_ENDIAN); + Assert.assertThrows( + SketchesArgumentException.class, + () -> objectStrategy.fromByteBufferSafe(buf4, garbageLonger.length).copy() + ); + } +} diff --git a/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/kll/KllDoublesSketchComplexMetricSerdeTest.java b/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/kll/KllDoublesSketchComplexMetricSerdeTest.java index 3628c5e6212f..0ae46bef496b 100644 --- a/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/kll/KllDoublesSketchComplexMetricSerdeTest.java +++ b/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/kll/KllDoublesSketchComplexMetricSerdeTest.java @@ -23,10 +23,14 @@ import com.google.common.collect.ImmutableMap; import org.apache.datasketches.kll.KllDoublesSketch; import org.apache.druid.data.input.MapBasedInputRow; +import org.apache.druid.segment.data.ObjectStrategy; import org.apache.druid.segment.serde.ComplexMetricExtractor; import org.junit.Assert; import org.junit.Test; +import java.nio.ByteBuffer; +import java.nio.ByteOrder; + public class KllDoublesSketchComplexMetricSerdeTest { @Test @@ -92,4 +96,44 @@ public void testExtractorOnLeadingDecimalNumber() Assert.assertEquals(1, sketch.getNumRetained()); Assert.assertEquals(0.1d, sketch.getMaxValue(), 0.01d); } + + @Test + public void testSafeRead() + { + final KllDoublesSketchComplexMetricSerde serde = new KllDoublesSketchComplexMetricSerde(); + final ObjectStrategy objectStrategy = serde.getObjectStrategy(); + + KllDoublesSketch sketch = KllDoublesSketch.newHeapInstance(); + sketch.update(1.1); + sketch.update(1.2); + final byte[] bytes = sketch.toByteArray(); + + ByteBuffer buf = ByteBuffer.wrap(bytes).order(ByteOrder.LITTLE_ENDIAN); + + // valid sketch should not explode when converted to byte array, which reads the memory + objectStrategy.fromByteBufferSafe(buf, bytes.length).toByteArray(); + + // corrupted sketch should fail with a regular java buffer exception, not all subsets actually fail with the same + // index out of bounds exceptions, but at least this many do + for (int subset = 3; subset < 24; subset++) { + final byte[] garbage2 = new byte[subset]; + for (int i = 0; i < garbage2.length; i++) { + garbage2[i] = buf.get(i); + } + + final ByteBuffer buf2 = ByteBuffer.wrap(garbage2).order(ByteOrder.LITTLE_ENDIAN); + Assert.assertThrows( + IndexOutOfBoundsException.class, + () -> objectStrategy.fromByteBufferSafe(buf2, garbage2.length).toByteArray() + ); + } + + // non sketch that is too short to contain header should fail with regular java buffer exception + final byte[] garbage = new byte[]{0x01, 0x02}; + final ByteBuffer buf3 = ByteBuffer.wrap(garbage).order(ByteOrder.LITTLE_ENDIAN); + Assert.assertThrows( + IndexOutOfBoundsException.class, + () -> objectStrategy.fromByteBufferSafe(buf3, garbage.length).toByteArray() + ); + } } diff --git a/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/kll/KllDoublesSketchOperationsTest.java b/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/kll/KllDoublesSketchOperationsTest.java new file mode 100644 index 000000000000..d2b0e3839848 --- /dev/null +++ b/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/kll/KllDoublesSketchOperationsTest.java @@ -0,0 +1,51 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.aggregation.datasketches.kll; + +import org.apache.datasketches.kll.KllDoublesSketch; +import org.apache.druid.java.util.common.StringUtils; +import org.junit.Assert; +import org.junit.Test; + +import java.util.Arrays; + +public class KllDoublesSketchOperationsTest +{ + @Test + public void testDeserializeSafe() + { + KllDoublesSketch sketch = KllDoublesSketch.newHeapInstance(); + sketch.update(1.1); + sketch.update(1.2); + final byte[] bytes = sketch.toByteArray(); + final String base64 = StringUtils.encodeBase64String(bytes); + + Assert.assertArrayEquals(bytes, KllDoublesSketchOperations.deserializeSafe(sketch).toByteArray()); + Assert.assertArrayEquals(bytes, KllDoublesSketchOperations.deserializeSafe(bytes).toByteArray()); + Assert.assertArrayEquals(bytes, KllDoublesSketchOperations.deserializeSafe(base64).toByteArray()); + + final byte[] trunacted = Arrays.copyOfRange(bytes, 0, 20); + Assert.assertThrows(IndexOutOfBoundsException.class, () -> KllDoublesSketchOperations.deserializeSafe(trunacted)); + Assert.assertThrows( + IndexOutOfBoundsException.class, + () -> KllDoublesSketchOperations.deserializeSafe(StringUtils.encodeBase64String(trunacted)) + ); + } +} diff --git a/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/kll/KllFloatsSketchComplexMetricSerdeTest.java b/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/kll/KllFloatsSketchComplexMetricSerdeTest.java index 5ff441df1c14..c6b8c310221c 100644 --- a/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/kll/KllFloatsSketchComplexMetricSerdeTest.java +++ b/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/kll/KllFloatsSketchComplexMetricSerdeTest.java @@ -23,10 +23,14 @@ import com.google.common.collect.ImmutableMap; import org.apache.datasketches.kll.KllFloatsSketch; import org.apache.druid.data.input.MapBasedInputRow; +import org.apache.druid.segment.data.ObjectStrategy; import org.apache.druid.segment.serde.ComplexMetricExtractor; import org.junit.Assert; import org.junit.Test; +import java.nio.ByteBuffer; +import java.nio.ByteOrder; + public class KllFloatsSketchComplexMetricSerdeTest { @Test @@ -92,4 +96,44 @@ public void testExtractorOnLeadingDecimalNumber() Assert.assertEquals(1, sketch.getNumRetained()); Assert.assertEquals(0.1d, sketch.getMaxValue(), 0.01d); } + + @Test + public void testSafeRead() + { + final KllFloatsSketchComplexMetricSerde serde = new KllFloatsSketchComplexMetricSerde(); + final ObjectStrategy objectStrategy = serde.getObjectStrategy(); + + KllFloatsSketch sketch = KllFloatsSketch.newHeapInstance(); + sketch.update(1.1f); + sketch.update(1.2f); + final byte[] bytes = sketch.toByteArray(); + + ByteBuffer buf = ByteBuffer.wrap(bytes).order(ByteOrder.LITTLE_ENDIAN); + + // valid sketch should not explode when converted to byte array, which reads the memory + objectStrategy.fromByteBufferSafe(buf, bytes.length).toByteArray(); + + // corrupted sketch should fail with a regular java buffer exception, not all subsets actually fail with the same + // index out of bounds exceptions, but at least this many do + for (int subset = 3; subset < 24; subset++) { + final byte[] garbage2 = new byte[subset]; + for (int i = 0; i < garbage2.length; i++) { + garbage2[i] = buf.get(i); + } + + final ByteBuffer buf2 = ByteBuffer.wrap(garbage2).order(ByteOrder.LITTLE_ENDIAN); + Assert.assertThrows( + IndexOutOfBoundsException.class, + () -> objectStrategy.fromByteBufferSafe(buf2, garbage2.length).toByteArray() + ); + } + + // non sketch that is too short to contain header should fail with regular java buffer exception + final byte[] garbage = new byte[]{0x01, 0x02}; + final ByteBuffer buf3 = ByteBuffer.wrap(garbage).order(ByteOrder.LITTLE_ENDIAN); + Assert.assertThrows( + IndexOutOfBoundsException.class, + () -> objectStrategy.fromByteBufferSafe(buf3, garbage.length).toByteArray() + ); + } } diff --git a/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/kll/KllFloatsSketchOperationsTest.java b/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/kll/KllFloatsSketchOperationsTest.java new file mode 100644 index 000000000000..613b38c6601a --- /dev/null +++ b/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/kll/KllFloatsSketchOperationsTest.java @@ -0,0 +1,51 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.aggregation.datasketches.kll; + +import org.apache.datasketches.kll.KllFloatsSketch; +import org.apache.druid.java.util.common.StringUtils; +import org.junit.Assert; +import org.junit.Test; + +import java.util.Arrays; + +public class KllFloatsSketchOperationsTest +{ + @Test + public void testDeserializeSafe() + { + KllFloatsSketch sketch = KllFloatsSketch.newHeapInstance(); + sketch.update(1.1f); + sketch.update(1.2f); + final byte[] bytes = sketch.toByteArray(); + final String base64 = StringUtils.encodeBase64String(bytes); + + Assert.assertArrayEquals(bytes, KllFloatsSketchOperations.deserializeSafe(sketch).toByteArray()); + Assert.assertArrayEquals(bytes, KllFloatsSketchOperations.deserializeSafe(bytes).toByteArray()); + Assert.assertArrayEquals(bytes, KllFloatsSketchOperations.deserializeSafe(base64).toByteArray()); + + final byte[] trunacted = Arrays.copyOfRange(bytes, 0, 20); + Assert.assertThrows(IndexOutOfBoundsException.class, () -> KllFloatsSketchOperations.deserializeSafe(trunacted)); + Assert.assertThrows( + IndexOutOfBoundsException.class, + () -> KllFloatsSketchOperations.deserializeSafe(StringUtils.encodeBase64String(trunacted)) + ); + } +} diff --git a/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/quantiles/DoublesSketchComplexMetricSerdeTest.java b/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/quantiles/DoublesSketchComplexMetricSerdeTest.java index e198c7704257..7dc82baee925 100644 --- a/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/quantiles/DoublesSketchComplexMetricSerdeTest.java +++ b/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/quantiles/DoublesSketchComplexMetricSerdeTest.java @@ -22,11 +22,16 @@ import com.google.common.collect.ImmutableList; import com.google.common.collect.ImmutableMap; import org.apache.datasketches.quantiles.DoublesSketch; +import org.apache.datasketches.quantiles.DoublesUnion; import org.apache.druid.data.input.MapBasedInputRow; +import org.apache.druid.segment.data.ObjectStrategy; import org.apache.druid.segment.serde.ComplexMetricExtractor; import org.junit.Assert; import org.junit.Test; +import java.nio.ByteBuffer; +import java.nio.ByteOrder; + public class DoublesSketchComplexMetricSerdeTest { @Test @@ -92,4 +97,42 @@ public void testExtractorOnLeadingDecimalNumber() Assert.assertEquals(1, sketch.getRetainedItems()); Assert.assertEquals(0.1d, sketch.getMaxValue(), 0.01d); } + + @Test + public void testSafeRead() + { + final DoublesSketchComplexMetricSerde serde = new DoublesSketchComplexMetricSerde(); + DoublesUnion union = DoublesUnion.builder().setMaxK(1024).build(); + union.update(1.1); + final byte[] bytes = union.toByteArray(); + + ByteBuffer buf = ByteBuffer.wrap(bytes).order(ByteOrder.LITTLE_ENDIAN); + ObjectStrategy objectStrategy = serde.getObjectStrategy(); + + // valid sketch should not explode when copied, which reads the memory + objectStrategy.fromByteBufferSafe(buf, bytes.length).toByteArray(true); + + // corrupted sketch should fail with a regular java buffer exception + for (int subset = 3; subset < 15; subset++) { + final byte[] garbage2 = new byte[subset]; + for (int i = 0; i < garbage2.length; i++) { + garbage2[i] = buf.get(i); + } + + final ByteBuffer buf2 = ByteBuffer.wrap(garbage2).order(ByteOrder.LITTLE_ENDIAN); + Assert.assertThrows( + "i " + subset, + IndexOutOfBoundsException.class, + () -> objectStrategy.fromByteBufferSafe(buf2, garbage2.length).toByteArray(true) + ); + } + + // non sketch that is too short to contain header should fail with regular java buffer exception + final byte[] garbage = new byte[]{0x01, 0x02}; + final ByteBuffer buf3 = ByteBuffer.wrap(garbage).order(ByteOrder.LITTLE_ENDIAN); + Assert.assertThrows( + IndexOutOfBoundsException.class, + () -> objectStrategy.fromByteBufferSafe(buf3, garbage.length).toByteArray(true) + ); + } } diff --git a/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/quantiles/DoublesSketchOperationsTest.java b/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/quantiles/DoublesSketchOperationsTest.java new file mode 100644 index 000000000000..38e5d39a91b8 --- /dev/null +++ b/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/quantiles/DoublesSketchOperationsTest.java @@ -0,0 +1,50 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.aggregation.datasketches.quantiles; + +import org.apache.datasketches.quantiles.DoublesUnion; +import org.apache.druid.java.util.common.StringUtils; +import org.junit.Assert; +import org.junit.Test; + +import java.util.Arrays; + +public class DoublesSketchOperationsTest +{ + @Test + public void testDeserializeSafe() + { + DoublesUnion union = DoublesUnion.builder().setMaxK(1024).build(); + union.update(1.1); + final byte[] bytes = union.getResult().toByteArray(); + final String base64 = StringUtils.encodeBase64String(bytes); + + Assert.assertArrayEquals(bytes, DoublesSketchOperations.deserializeSafe(union.getResult()).toByteArray()); + Assert.assertArrayEquals(bytes, DoublesSketchOperations.deserializeSafe(bytes).toByteArray()); + Assert.assertArrayEquals(bytes, DoublesSketchOperations.deserializeSafe(base64).toByteArray()); + + final byte[] trunacted = Arrays.copyOfRange(bytes, 0, 4); + Assert.assertThrows(IndexOutOfBoundsException.class, () -> DoublesSketchOperations.deserializeSafe(trunacted)); + Assert.assertThrows( + IndexOutOfBoundsException.class, + () -> DoublesSketchOperations.deserializeSafe(StringUtils.encodeBase64(trunacted)) + ); + } +} diff --git a/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/theta/SketchHolderObjectStrategyTest.java b/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/theta/SketchHolderObjectStrategyTest.java new file mode 100644 index 000000000000..5619facd5f62 --- /dev/null +++ b/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/theta/SketchHolderObjectStrategyTest.java @@ -0,0 +1,79 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.aggregation.datasketches.theta; + +import org.apache.datasketches.Family; +import org.apache.datasketches.SketchesArgumentException; +import org.apache.datasketches.theta.SetOperation; +import org.apache.datasketches.theta.Union; +import org.apache.druid.java.util.common.StringUtils; +import org.junit.Assert; +import org.junit.Test; + +import java.nio.ByteBuffer; +import java.nio.ByteOrder; + +public class SketchHolderObjectStrategyTest +{ + @Test + public void testSafeRead() + { + SketchHolderObjectStrategy objectStrategy = new SketchHolderObjectStrategy(); + Union union = (Union) SetOperation.builder().setNominalEntries(1024).build(Family.UNION); + union.update(1234L); + + final byte[] bytes = union.getResult().toByteArray(); + + ByteBuffer buf = ByteBuffer.wrap(bytes).order(ByteOrder.LITTLE_ENDIAN); + + // valid sketch should not explode when copied, which reads the memory + objectStrategy.fromByteBufferSafe(buf, bytes.length).getSketch().compact().getCompactBytes(); + + // corrupted sketch should fail with a regular java buffer exception + for (int subset = 3; subset < bytes.length - 1; subset++) { + final byte[] garbage2 = new byte[subset]; + for (int i = 0; i < garbage2.length; i++) { + garbage2[i] = buf.get(i); + } + + final ByteBuffer buf2 = ByteBuffer.wrap(garbage2).order(ByteOrder.LITTLE_ENDIAN); + Assert.assertThrows( + IndexOutOfBoundsException.class, + () -> objectStrategy.fromByteBufferSafe(buf2, garbage2.length).getSketch().compact().getCompactBytes() + ); + } + + // non sketch that is too short to contain header should fail with regular java buffer exception + final byte[] garbage = new byte[]{0x01, 0x02}; + final ByteBuffer buf3 = ByteBuffer.wrap(garbage).order(ByteOrder.LITTLE_ENDIAN); + Assert.assertThrows( + IndexOutOfBoundsException.class, + () -> objectStrategy.fromByteBufferSafe(buf3, garbage.length).getSketch().compact().getCompactBytes() + ); + + // non sketch that is long enough to check (this one doesn't actually need 'safe' read) + final byte[] garbageLonger = StringUtils.toUtf8("notasketch"); + final ByteBuffer buf4 = ByteBuffer.wrap(garbageLonger).order(ByteOrder.LITTLE_ENDIAN); + Assert.assertThrows( + SketchesArgumentException.class, + () -> objectStrategy.fromByteBufferSafe(buf4, garbageLonger.length).getSketch().compact().getCompactBytes() + ); + } +} diff --git a/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/theta/SketchHolderTest.java b/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/theta/SketchHolderTest.java new file mode 100644 index 000000000000..ef68fdeb8c58 --- /dev/null +++ b/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/theta/SketchHolderTest.java @@ -0,0 +1,52 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.aggregation.datasketches.theta; + +import org.apache.datasketches.Family; +import org.apache.datasketches.theta.SetOperation; +import org.apache.datasketches.theta.Union; +import org.apache.druid.java.util.common.StringUtils; +import org.junit.Assert; +import org.junit.Test; + +import java.util.Arrays; + +public class SketchHolderTest +{ + @Test + public void testDeserializeSafe() + { + Union union = (Union) SetOperation.builder().setNominalEntries(1024).build(Family.UNION); + union.update(1234L); + final byte[] bytes = union.getResult().toByteArray(); + final String base64 = StringUtils.encodeBase64String(bytes); + + Assert.assertArrayEquals(bytes, SketchHolder.deserializeSafe(union.getResult()).getSketch().toByteArray()); + Assert.assertArrayEquals(bytes, SketchHolder.deserializeSafe(bytes).getSketch().toByteArray()); + Assert.assertArrayEquals(bytes, SketchHolder.deserializeSafe(base64).getSketch().toByteArray()); + + final byte[] trunacted = Arrays.copyOfRange(bytes, 0, 10); + Assert.assertThrows(IndexOutOfBoundsException.class, () -> SketchHolder.deserializeSafe(trunacted)); + Assert.assertThrows( + IndexOutOfBoundsException.class, + () -> SketchHolder.deserializeSafe(StringUtils.encodeBase64String(trunacted)) + ); + } +} diff --git a/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/tuple/ArrayOfDoublesSketchObjectStrategyTest.java b/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/tuple/ArrayOfDoublesSketchObjectStrategyTest.java new file mode 100644 index 000000000000..ee59ddf57645 --- /dev/null +++ b/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/tuple/ArrayOfDoublesSketchObjectStrategyTest.java @@ -0,0 +1,70 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.aggregation.datasketches.tuple; + +import org.apache.datasketches.tuple.arrayofdoubles.ArrayOfDoublesUpdatableSketch; +import org.apache.datasketches.tuple.arrayofdoubles.ArrayOfDoublesUpdatableSketchBuilder; +import org.junit.Assert; +import org.junit.Test; + +import java.nio.ByteBuffer; +import java.nio.ByteOrder; + +public class ArrayOfDoublesSketchObjectStrategyTest +{ + @Test + public void testSafeRead() + { + ArrayOfDoublesSketchObjectStrategy objectStrategy = new ArrayOfDoublesSketchObjectStrategy(); + ArrayOfDoublesUpdatableSketch sketch = new ArrayOfDoublesUpdatableSketchBuilder().setNominalEntries(1024) + .setNumberOfValues(4) + .build(); + sketch.update(1L, new double[]{1.0, 2.0, 3.0, 4.0}); + + final byte[] bytes = sketch.compact().toByteArray(); + + ByteBuffer buf = ByteBuffer.wrap(bytes).order(ByteOrder.LITTLE_ENDIAN); + + // valid sketch should not explode when copied, which reads the memory + objectStrategy.fromByteBufferSafe(buf, bytes.length).compact().toByteArray(); + + // corrupted sketch should fail with a regular java buffer exception + for (int subset = 3; subset < bytes.length - 1; subset++) { + final byte[] garbage2 = new byte[subset]; + for (int i = 0; i < garbage2.length; i++) { + garbage2[i] = buf.get(i); + } + + final ByteBuffer buf2 = ByteBuffer.wrap(garbage2).order(ByteOrder.LITTLE_ENDIAN); + Assert.assertThrows( + IndexOutOfBoundsException.class, + () -> objectStrategy.fromByteBufferSafe(buf2, garbage2.length).compact().toByteArray() + ); + } + + // non sketch that is too short to contain header should fail with regular java buffer exception + final byte[] garbage = new byte[]{0x01, 0x02}; + final ByteBuffer buf3 = ByteBuffer.wrap(garbage).order(ByteOrder.LITTLE_ENDIAN); + Assert.assertThrows( + IndexOutOfBoundsException.class, + () -> objectStrategy.fromByteBufferSafe(buf3, garbage.length).compact().toByteArray() + ); + } +} diff --git a/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/tuple/ArrayOfDoublesSketchOperationsTest.java b/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/tuple/ArrayOfDoublesSketchOperationsTest.java new file mode 100644 index 000000000000..415f3acab970 --- /dev/null +++ b/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/tuple/ArrayOfDoublesSketchOperationsTest.java @@ -0,0 +1,55 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.aggregation.datasketches.tuple; + +import org.apache.datasketches.tuple.arrayofdoubles.ArrayOfDoublesUpdatableSketch; +import org.apache.datasketches.tuple.arrayofdoubles.ArrayOfDoublesUpdatableSketchBuilder; +import org.apache.druid.java.util.common.StringUtils; +import org.junit.Assert; +import org.junit.Test; + +import java.util.Arrays; + +public class ArrayOfDoublesSketchOperationsTest +{ + @Test + public void testDeserializeSafe() + { + ArrayOfDoublesSketchObjectStrategy objectStrategy = new ArrayOfDoublesSketchObjectStrategy(); + ArrayOfDoublesUpdatableSketch sketch = new ArrayOfDoublesUpdatableSketchBuilder().setNominalEntries(1024) + .setNumberOfValues(4) + .build(); + sketch.update(1L, new double[]{1.0, 2.0, 3.0, 4.0}); + + final byte[] bytes = sketch.toByteArray(); + final String base64 = StringUtils.encodeBase64String(bytes); + + Assert.assertArrayEquals(bytes, ArrayOfDoublesSketchOperations.deserializeSafe(sketch).toByteArray()); + Assert.assertArrayEquals(bytes, ArrayOfDoublesSketchOperations.deserializeSafe(bytes).toByteArray()); + Assert.assertArrayEquals(bytes, ArrayOfDoublesSketchOperations.deserializeSafe(base64).toByteArray()); + + final byte[] trunacted = Arrays.copyOfRange(bytes, 0, 10); + Assert.assertThrows(IndexOutOfBoundsException.class, () -> ArrayOfDoublesSketchOperations.deserializeSafe(trunacted)); + Assert.assertThrows( + IndexOutOfBoundsException.class, + () -> ArrayOfDoublesSketchOperations.deserializeSafe(StringUtils.encodeBase64String(trunacted)) + ); + } +} diff --git a/processing/src/main/java/org/apache/druid/segment/column/ObjectStrategyComplexTypeStrategy.java b/processing/src/main/java/org/apache/druid/segment/column/ObjectStrategyComplexTypeStrategy.java index 351f2665d05e..d05ba208585c 100644 --- a/processing/src/main/java/org/apache/druid/segment/column/ObjectStrategyComplexTypeStrategy.java +++ b/processing/src/main/java/org/apache/druid/segment/column/ObjectStrategyComplexTypeStrategy.java @@ -90,6 +90,6 @@ public int compare(T o1, T o2) @Override public T fromBytes(byte[] value) { - return objectStrategy.fromByteBuffer(ByteBuffer.wrap(value), value.length); + return objectStrategy.fromByteBufferSafe(ByteBuffer.wrap(value), value.length); } } diff --git a/processing/src/main/java/org/apache/druid/segment/data/ObjectStrategy.java b/processing/src/main/java/org/apache/druid/segment/data/ObjectStrategy.java index 8a53fc57a7d7..eba97d04bbbe 100644 --- a/processing/src/main/java/org/apache/druid/segment/data/ObjectStrategy.java +++ b/processing/src/main/java/org/apache/druid/segment/data/ObjectStrategy.java @@ -79,4 +79,31 @@ default void writeTo(T val, WriteOutBytes out) throws IOException out.write(bytes); } } + + /** + * Convert values from their underlying byte representation, when the underlying bytes might be corrupted or + * maliciously constructed + * + * Implementations of this method absolutely must never perform any sun.misc.Unsafe based memory read or write + * operations from instructions contained in the data read from this buffer without first validating the data. If the + * data cannot be validated, all read and write operations from instructions in this data must be done directly with + * the {@link ByteBuffer} methods, or using {@link SafeWritableMemory} if + * {@link org.apache.datasketches.memory.Memory} is employed to materialize the value. + * + * Implementations of this method may change the given buffer's mark, or limit, and position. + * + * Implementations of this method may not store the given buffer in a field of the "deserialized" object, + * need to use {@link ByteBuffer#slice()}, {@link ByteBuffer#asReadOnlyBuffer()} or {@link ByteBuffer#duplicate()} in + * this case. + * + * + * @param buffer buffer to read value from + * @param numBytes number of bytes used to store the value, starting at buffer.position() + * @return an object created from the given byte buffer representation + */ + @Nullable + default T fromByteBufferSafe(ByteBuffer buffer, int numBytes) + { + return fromByteBuffer(buffer, numBytes); + } } diff --git a/processing/src/main/java/org/apache/druid/segment/data/SafeWritableBase.java b/processing/src/main/java/org/apache/druid/segment/data/SafeWritableBase.java new file mode 100644 index 000000000000..df2fc14d0535 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/segment/data/SafeWritableBase.java @@ -0,0 +1,450 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.segment.data; + +import com.google.common.base.Preconditions; +import com.google.common.primitives.Ints; +import org.apache.datasketches.memory.BaseState; +import org.apache.datasketches.memory.MemoryRequestServer; +import org.apache.datasketches.memory.WritableMemory; +import org.apache.datasketches.memory.internal.BaseStateImpl; +import org.apache.datasketches.memory.internal.UnsafeUtil; +import org.apache.datasketches.memory.internal.XxHash64; +import org.apache.druid.java.util.common.StringUtils; + +import java.nio.ByteBuffer; +import java.nio.ByteOrder; + +/** + * Base class for making a regular {@link ByteBuffer} look like a {@link org.apache.datasketches.memory.Memory} or + * {@link org.apache.datasketches.memory.Buffer}. All methods delegate directly to the {@link ByteBuffer} rather + * than using 'unsafe' reads. + * + * @see SafeWritableMemory + * @see SafeWritableBuffer + */ + +@SuppressWarnings("unused") +public abstract class SafeWritableBase implements BaseState +{ + static final MemoryRequestServer SAFE_HEAP_REQUEST_SERVER = new HeapByteBufferMemoryRequestServer(); + + final ByteBuffer buffer; + + public SafeWritableBase(ByteBuffer buffer) + { + this.buffer = buffer; + } + + public MemoryRequestServer getMemoryRequestServer() + { + return SAFE_HEAP_REQUEST_SERVER; + } + + public boolean getBoolean(long offsetBytes) + { + return getByte(Ints.checkedCast(offsetBytes)) != 0; + } + + public byte getByte(long offsetBytes) + { + return buffer.get(Ints.checkedCast(offsetBytes)); + } + + public char getChar(long offsetBytes) + { + return buffer.getChar(Ints.checkedCast(offsetBytes)); + } + + public double getDouble(long offsetBytes) + { + return buffer.getDouble(Ints.checkedCast(offsetBytes)); + } + + public float getFloat(long offsetBytes) + { + return buffer.getFloat(Ints.checkedCast(offsetBytes)); + } + + public int getInt(long offsetBytes) + { + return buffer.getInt(Ints.checkedCast(offsetBytes)); + } + + public long getLong(long offsetBytes) + { + return buffer.getLong(Ints.checkedCast(offsetBytes)); + } + + public short getShort(long offsetBytes) + { + return buffer.getShort(Ints.checkedCast(offsetBytes)); + } + + public void putBoolean(long offsetBytes, boolean value) + { + buffer.put(Ints.checkedCast(offsetBytes), (byte) (value ? 1 : 0)); + } + + public void putByte(long offsetBytes, byte value) + { + buffer.put(Ints.checkedCast(offsetBytes), value); + } + + public void putChar(long offsetBytes, char value) + { + buffer.putChar(Ints.checkedCast(offsetBytes), value); + } + + public void putDouble(long offsetBytes, double value) + { + buffer.putDouble(Ints.checkedCast(offsetBytes), value); + } + + public void putFloat(long offsetBytes, float value) + { + buffer.putFloat(Ints.checkedCast(offsetBytes), value); + } + + public void putInt(long offsetBytes, int value) + { + buffer.putInt(Ints.checkedCast(offsetBytes), value); + } + + public void putLong(long offsetBytes, long value) + { + buffer.putLong(Ints.checkedCast(offsetBytes), value); + } + + public void putShort(long offsetBytes, short value) + { + buffer.putShort(Ints.checkedCast(offsetBytes), value); + } + + @Override + public ByteOrder getTypeByteOrder() + { + return buffer.order(); + } + + @Override + public boolean isByteOrderCompatible(ByteOrder byteOrder) + { + return buffer.order().equals(byteOrder); + } + + @Override + public ByteBuffer getByteBuffer() + { + return buffer; + } + + @Override + public long getCapacity() + { + return buffer.capacity(); + } + + @Override + public long getCumulativeOffset() + { + return 0; + } + + @Override + public long getCumulativeOffset(long offsetBytes) + { + return offsetBytes; + } + + @Override + public long getRegionOffset() + { + return 0; + } + + @Override + public long getRegionOffset(long offsetBytes) + { + return offsetBytes; + } + + @Override + public boolean hasArray() + { + return false; + } + + @Override + public long xxHash64(long offsetBytes, long lengthBytes, long seed) + { + return hash(buffer, offsetBytes, lengthBytes, seed); + } + + @Override + public long xxHash64(long in, long seed) + { + return XxHash64.hash(in, seed); + } + + @Override + public boolean hasByteBuffer() + { + return true; + } + + @Override + public boolean isDirect() + { + return false; + } + + @Override + public boolean isReadOnly() + { + return false; + } + + @Override + public boolean isSameResource(Object that) + { + return this.equals(that); + } + + @Override + public boolean isValid() + { + return true; + } + + @Override + public void checkValidAndBounds(long offsetBytes, long lengthBytes) + { + Preconditions.checkArgument( + Ints.checkedCast(offsetBytes) < buffer.limit(), + "start offset %s is greater than buffer limit %s", + offsetBytes, + buffer.limit() + ); + Preconditions.checkArgument( + Ints.checkedCast(offsetBytes + lengthBytes) < buffer.limit(), + "end offset %s is greater than buffer limit %s", + offsetBytes + lengthBytes, + buffer.limit() + ); + } + + /** + * Adapted from {@link BaseStateImpl#toHexString(String, long, int)} + */ + @Override + public String toHexString(String header, long offsetBytes, int lengthBytes) + { + final String klass = this.getClass().getSimpleName(); + final String s1 = StringUtils.format("(..., %d, %d)", offsetBytes, lengthBytes); + final long hcode = hashCode() & 0XFFFFFFFFL; + final String call = ".toHexString" + s1 + ", hashCode: " + hcode; + String sb = "### " + klass + " SUMMARY ###" + UnsafeUtil.LS + + "Header Comment : " + header + UnsafeUtil.LS + + "Call Parameters : " + call; + return toHex(this, sb, offsetBytes, lengthBytes); + } + + /** + * Adapted from {@link BaseStateImpl#toHex(BaseStateImpl, String, long, int)} + */ + static String toHex( + final SafeWritableBase state, + final String preamble, + final long offsetBytes, + final int lengthBytes + ) + { + final String lineSeparator = UnsafeUtil.LS; + final long capacity = state.getCapacity(); + UnsafeUtil.checkBounds(offsetBytes, lengthBytes, capacity); + final StringBuilder sb = new StringBuilder(); + final String uObjStr; + final long uObjHeader; + uObjStr = "null"; + uObjHeader = 0; + final ByteBuffer bb = state.getByteBuffer(); + final String bbStr = bb == null ? "null" + : bb.getClass().getSimpleName() + ", " + (bb.hashCode() & 0XFFFFFFFFL); + final MemoryRequestServer memReqSvr = state.getMemoryRequestServer(); + final String memReqStr = memReqSvr != null + ? memReqSvr.getClass().getSimpleName() + ", " + (memReqSvr.hashCode() & 0XFFFFFFFFL) + : "null"; + final long cumBaseOffset = state.getCumulativeOffset(); + sb.append(preamble).append(lineSeparator); + sb.append("UnsafeObj, hashCode : ").append(uObjStr).append(lineSeparator); + sb.append("UnsafeObjHeader : ").append(uObjHeader).append(lineSeparator); + sb.append("ByteBuf, hashCode : ").append(bbStr).append(lineSeparator); + sb.append("RegionOffset : ").append(state.getRegionOffset()).append(lineSeparator); + sb.append("Capacity : ").append(capacity).append(lineSeparator); + sb.append("CumBaseOffset : ").append(cumBaseOffset).append(lineSeparator); + sb.append("MemReq, hashCode : ").append(memReqStr).append(lineSeparator); + sb.append("Valid : ").append(state.isValid()).append(lineSeparator); + sb.append("Read Only : ").append(state.isReadOnly()).append(lineSeparator); + sb.append("Type Byte Order : ").append(state.getTypeByteOrder()).append(lineSeparator); + sb.append("Native Byte Order : ").append(ByteOrder.nativeOrder()).append(lineSeparator); + sb.append("JDK Runtime Version : ").append(UnsafeUtil.JDK).append(lineSeparator); + //Data detail + sb.append("Data, littleEndian : 0 1 2 3 4 5 6 7"); + + for (long i = 0; i < lengthBytes; i++) { + final int b = state.getByte(cumBaseOffset + offsetBytes + i) & 0XFF; + if (i % 8 == 0) { //row header + sb.append(StringUtils.format("%n%20s: ", offsetBytes + i)); + } + sb.append(StringUtils.format("%02x ", b)); + } + sb.append(lineSeparator); + + return sb.toString(); + } + + // copied from datasketches-memory XxHash64.java + private static final long P1 = -7046029288634856825L; + private static final long P2 = -4417276706812531889L; + private static final long P3 = 1609587929392839161L; + private static final long P4 = -8796714831421723037L; + private static final long P5 = 2870177450012600261L; + + /** + * Adapted from {@link XxHash64#hash(Object, long, long, long)} to work with {@link ByteBuffer} + */ + static long hash(ByteBuffer memory, long cumOffsetBytes, final long lengthBytes, final long seed) + { + long hash; + long remaining = lengthBytes; + int offset = Ints.checkedCast(cumOffsetBytes); + + if (remaining >= 32) { + long v1 = seed + P1 + P2; + long v2 = seed + P2; + long v3 = seed; + long v4 = seed - P1; + + do { + v1 += memory.getLong(offset) * P2; + v1 = Long.rotateLeft(v1, 31); + v1 *= P1; + + v2 += memory.getLong(offset + 8) * P2; + v2 = Long.rotateLeft(v2, 31); + v2 *= P1; + + v3 += memory.getLong(offset + 16) * P2; + v3 = Long.rotateLeft(v3, 31); + v3 *= P1; + + v4 += memory.getLong(offset + 24) * P2; + v4 = Long.rotateLeft(v4, 31); + v4 *= P1; + + offset += 32; + remaining -= 32; + } while (remaining >= 32); + + hash = Long.rotateLeft(v1, 1) + + Long.rotateLeft(v2, 7) + + Long.rotateLeft(v3, 12) + + Long.rotateLeft(v4, 18); + + v1 *= P2; + v1 = Long.rotateLeft(v1, 31); + v1 *= P1; + hash ^= v1; + hash = (hash * P1) + P4; + + v2 *= P2; + v2 = Long.rotateLeft(v2, 31); + v2 *= P1; + hash ^= v2; + hash = (hash * P1) + P4; + + v3 *= P2; + v3 = Long.rotateLeft(v3, 31); + v3 *= P1; + hash ^= v3; + hash = (hash * P1) + P4; + + v4 *= P2; + v4 = Long.rotateLeft(v4, 31); + v4 *= P1; + hash ^= v4; + hash = (hash * P1) + P4; + } else { //end remaining >= 32 + hash = seed + P5; + } + + hash += lengthBytes; + + while (remaining >= 8) { + long k1 = memory.getLong(offset); + k1 *= P2; + k1 = Long.rotateLeft(k1, 31); + k1 *= P1; + hash ^= k1; + hash = (Long.rotateLeft(hash, 27) * P1) + P4; + offset += 8; + remaining -= 8; + } + + if (remaining >= 4) { //treat as unsigned ints + hash ^= (memory.getInt(offset) & 0XFFFF_FFFFL) * P1; + hash = (Long.rotateLeft(hash, 23) * P2) + P3; + offset += 4; + remaining -= 4; + } + + while (remaining != 0) { //treat as unsigned bytes + hash ^= (memory.get(offset) & 0XFFL) * P5; + hash = Long.rotateLeft(hash, 11) * P1; + --remaining; + ++offset; + } + + hash ^= hash >>> 33; + hash *= P2; + hash ^= hash >>> 29; + hash *= P3; + hash ^= hash >>> 32; + return hash; + } + + private static class HeapByteBufferMemoryRequestServer implements MemoryRequestServer + { + @Override + public WritableMemory request(WritableMemory currentWritableMemory, long capacityBytes) + { + ByteBuffer newBuffer = ByteBuffer.allocate(Ints.checkedCast(capacityBytes)); + newBuffer.order(currentWritableMemory.getTypeByteOrder()); + return new SafeWritableMemory(newBuffer); + } + + @Override + public void requestClose(WritableMemory memToClose, WritableMemory newMemory) + { + // do nothing + } + } +} diff --git a/processing/src/main/java/org/apache/druid/segment/data/SafeWritableBuffer.java b/processing/src/main/java/org/apache/druid/segment/data/SafeWritableBuffer.java new file mode 100644 index 000000000000..3da7e70b457c --- /dev/null +++ b/processing/src/main/java/org/apache/druid/segment/data/SafeWritableBuffer.java @@ -0,0 +1,501 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.segment.data; + +import com.google.common.primitives.Ints; +import org.apache.datasketches.memory.BaseBuffer; +import org.apache.datasketches.memory.Buffer; +import org.apache.datasketches.memory.Memory; +import org.apache.datasketches.memory.WritableBuffer; +import org.apache.datasketches.memory.WritableMemory; + +import java.nio.ByteBuffer; +import java.nio.ByteOrder; + +/** + * Safety first! Don't trust something whose contents you locations to read and write stuff to, but need a + * {@link Buffer} or {@link WritableBuffer}? use this! + *

+ * Delegates everything to an underlying {@link ByteBuffer} so all read and write operations will have bounds checks + * built in rather than using 'unsafe'. + */ +public class SafeWritableBuffer extends SafeWritableBase implements WritableBuffer +{ + private int start; + private int end; + + public SafeWritableBuffer(ByteBuffer buffer) + { + super(buffer); + this.start = 0; + this.buffer.position(0); + this.end = buffer.capacity(); + } + + @Override + public WritableBuffer writableDuplicate() + { + return writableDuplicate(buffer.order()); + } + + @Override + public WritableBuffer writableDuplicate(ByteOrder byteOrder) + { + ByteBuffer dupe = buffer.duplicate(); + dupe.order(byteOrder); + WritableBuffer duplicate = new SafeWritableBuffer(dupe); + duplicate.setStartPositionEnd(start, buffer.position(), end); + return duplicate; + } + + @Override + public WritableBuffer writableRegion() + { + ByteBuffer dupe = buffer.duplicate().order(buffer.order()); + dupe.position(start); + dupe.limit(end); + ByteBuffer remaining = buffer.slice(); + remaining.order(dupe.order()); + return new SafeWritableBuffer(remaining); + } + + @Override + public WritableBuffer writableRegion(long offsetBytes, long capacityBytes, ByteOrder byteOrder) + { + ByteBuffer dupe = buffer.duplicate(); + dupe.position(Ints.checkedCast(offsetBytes)); + dupe.limit(dupe.position() + Ints.checkedCast(capacityBytes)); + return new SafeWritableBuffer(dupe.slice().order(byteOrder)); + } + + @Override + public WritableMemory asWritableMemory(ByteOrder byteOrder) + { + ByteBuffer dupe = buffer.duplicate(); + dupe.order(byteOrder); + return new SafeWritableMemory(dupe); + } + + @Override + public void putBoolean(boolean value) + { + buffer.put((byte) (value ? 1 : 0)); + } + + @Override + public void putBooleanArray(boolean[] srcArray, int srcOffsetBooleans, int lengthBooleans) + { + for (int i = 0; i < lengthBooleans; i++) { + putBoolean(srcArray[srcOffsetBooleans + i]); + } + } + + @Override + public void putByte(byte value) + { + buffer.put(value); + } + + @Override + public void putByteArray(byte[] srcArray, int srcOffsetBytes, int lengthBytes) + { + buffer.put(srcArray, srcOffsetBytes, lengthBytes); + } + + @Override + public void putChar(char value) + { + buffer.putChar(value); + } + + @Override + public void putCharArray(char[] srcArray, int srcOffsetChars, int lengthChars) + { + for (int i = 0; i < lengthChars; i++) { + buffer.putChar(srcArray[srcOffsetChars + i]); + } + } + + @Override + public void putDouble(double value) + { + buffer.putDouble(value); + } + + @Override + public void putDoubleArray(double[] srcArray, int srcOffsetDoubles, int lengthDoubles) + { + for (int i = 0; i < lengthDoubles; i++) { + buffer.putDouble(srcArray[srcOffsetDoubles + i]); + } + } + + @Override + public void putFloat(float value) + { + buffer.putFloat(value); + } + + @Override + public void putFloatArray(float[] srcArray, int srcOffsetFloats, int lengthFloats) + { + for (int i = 0; i < lengthFloats; i++) { + buffer.putFloat(srcArray[srcOffsetFloats + i]); + } + } + + @Override + public void putInt(int value) + { + buffer.putInt(value); + } + + @Override + public void putIntArray(int[] srcArray, int srcOffsetInts, int lengthInts) + { + for (int i = 0; i < lengthInts; i++) { + buffer.putInt(srcArray[srcOffsetInts + i]); + } + } + + @Override + public void putLong(long value) + { + buffer.putLong(value); + } + + @Override + public void putLongArray(long[] srcArray, int srcOffsetLongs, int lengthLongs) + { + for (int i = 0; i < lengthLongs; i++) { + buffer.putLong(srcArray[srcOffsetLongs + i]); + } + } + + @Override + public void putShort(short value) + { + buffer.putShort(value); + } + + @Override + public void putShortArray(short[] srcArray, int srcOffsetShorts, int lengthShorts) + { + for (int i = 0; i < lengthShorts; i++) { + buffer.putShort(srcArray[srcOffsetShorts + i]); + } + } + + @Override + public Object getArray() + { + return null; + } + + @Override + public void clear() + { + fill((byte) 0); + } + + @Override + public void fill(byte value) + { + while (buffer.hasRemaining() && buffer.position() < end) { + buffer.put(value); + } + } + + @Override + public Buffer duplicate() + { + return writableDuplicate(); + } + + @Override + public Buffer duplicate(ByteOrder byteOrder) + { + return writableDuplicate(byteOrder); + } + + @Override + public Buffer region() + { + return writableRegion(); + } + + @Override + public Buffer region(long offsetBytes, long capacityBytes, ByteOrder byteOrder) + { + return writableRegion(offsetBytes, capacityBytes, byteOrder); + } + + @Override + public Memory asMemory(ByteOrder byteOrder) + { + return asWritableMemory(byteOrder); + } + + @Override + public boolean getBoolean() + { + return buffer.get() == 0 ? false : true; + } + + @Override + public void getBooleanArray(boolean[] dstArray, int dstOffsetBooleans, int lengthBooleans) + { + for (int i = 0; i < lengthBooleans; i++) { + dstArray[dstOffsetBooleans + i] = getBoolean(); + } + } + + @Override + public byte getByte() + { + return buffer.get(); + } + + @Override + public void getByteArray(byte[] dstArray, int dstOffsetBytes, int lengthBytes) + { + for (int i = 0; i < lengthBytes; i++) { + dstArray[dstOffsetBytes + i] = buffer.get(); + } + } + + @Override + public char getChar() + { + return buffer.getChar(); + } + + @Override + public void getCharArray(char[] dstArray, int dstOffsetChars, int lengthChars) + { + for (int i = 0; i < lengthChars; i++) { + dstArray[dstOffsetChars + i] = buffer.getChar(); + } + } + + @Override + public double getDouble() + { + return buffer.getDouble(); + } + + @Override + public void getDoubleArray(double[] dstArray, int dstOffsetDoubles, int lengthDoubles) + { + for (int i = 0; i < lengthDoubles; i++) { + dstArray[dstOffsetDoubles + i] = buffer.getDouble(); + } + } + + @Override + public float getFloat() + { + return buffer.getFloat(); + } + + @Override + public void getFloatArray(float[] dstArray, int dstOffsetFloats, int lengthFloats) + { + for (int i = 0; i < lengthFloats; i++) { + dstArray[dstOffsetFloats + i] = buffer.getFloat(); + } + } + + @Override + public int getInt() + { + return buffer.getInt(); + } + + @Override + public void getIntArray(int[] dstArray, int dstOffsetInts, int lengthInts) + { + for (int i = 0; i < lengthInts; i++) { + dstArray[dstOffsetInts + i] = buffer.getInt(); + } + } + + @Override + public long getLong() + { + return buffer.getLong(); + } + + @Override + public void getLongArray(long[] dstArray, int dstOffsetLongs, int lengthLongs) + { + for (int i = 0; i < lengthLongs; i++) { + dstArray[dstOffsetLongs + i] = buffer.getLong(); + } + } + + @Override + public short getShort() + { + return buffer.getShort(); + } + + @Override + public void getShortArray(short[] dstArray, int dstOffsetShorts, int lengthShorts) + { + for (int i = 0; i < lengthShorts; i++) { + dstArray[dstOffsetShorts + i] = buffer.getShort(); + } + } + + @Override + public int compareTo( + long thisOffsetBytes, + long thisLengthBytes, + Buffer that, + long thatOffsetBytes, + long thatLengthBytes + ) + { + final int thisLength = Ints.checkedCast(thisLengthBytes); + final int thatLength = Ints.checkedCast(thatLengthBytes); + + final int commonLength = Math.min(thisLength, thatLength); + + for (int i = 0; i < commonLength; i++) { + final int cmp = Byte.compare(getByte(thisOffsetBytes + i), that.getByte(thatOffsetBytes + i)); + if (cmp != 0) { + return cmp; + } + } + + return Integer.compare(thisLength, thatLength); + } + + @Override + public BaseBuffer incrementPosition(long increment) + { + buffer.position(buffer.position() + Ints.checkedCast(increment)); + return this; + } + + @Override + public BaseBuffer incrementAndCheckPosition(long increment) + { + checkInvariants(start, buffer.position() + increment, end, buffer.capacity()); + return incrementPosition(increment); + } + + @Override + public long getEnd() + { + return end; + } + + @Override + public long getPosition() + { + return buffer.position(); + } + + @Override + public long getStart() + { + return start; + } + + @Override + public long getRemaining() + { + return buffer.remaining(); + } + + @Override + public boolean hasRemaining() + { + return buffer.hasRemaining(); + } + + @Override + public BaseBuffer resetPosition() + { + buffer.position(start); + return this; + } + + @Override + public BaseBuffer setPosition(long position) + { + buffer.position(Ints.checkedCast(position)); + return this; + } + + @Override + public BaseBuffer setAndCheckPosition(long position) + { + checkInvariants(start, position, end, buffer.capacity()); + return setPosition(position); + } + + @Override + public BaseBuffer setStartPositionEnd(long start, long position, long end) + { + this.start = Ints.checkedCast(start); + this.end = Ints.checkedCast(end); + buffer.position(Ints.checkedCast(position)); + buffer.limit(this.end); + return this; + } + + @Override + public BaseBuffer setAndCheckStartPositionEnd(long start, long position, long end) + { + checkInvariants(start, position, end, buffer.capacity()); + return setStartPositionEnd(start, position, end); + } + + @Override + public boolean equalTo(long thisOffsetBytes, Object that, long thatOffsetBytes, long lengthBytes) + { + if (!(that instanceof SafeWritableBuffer)) { + return false; + } + return compareTo(thisOffsetBytes, lengthBytes, (SafeWritableBuffer) that, thatOffsetBytes, lengthBytes) == 0; + } + + /** + * Adapted from {@link org.apache.datasketches.memory.internal.BaseBufferImpl#checkInvariants(long, long, long, long)} + */ + static void checkInvariants(final long start, final long pos, final long end, final long cap) + { + if ((start | pos | end | cap | (pos - start) | (end - pos) | (cap - end)) < 0L) { + throw new IllegalArgumentException( + "Violation of Invariants: " + + "start: " + start + + " <= pos: " + pos + + " <= end: " + end + + " <= cap: " + cap + + "; (pos - start): " + (pos - start) + + ", (end - pos): " + (end - pos) + + ", (cap - end): " + (cap - end) + ); + } + } +} diff --git a/processing/src/main/java/org/apache/druid/segment/data/SafeWritableMemory.java b/processing/src/main/java/org/apache/druid/segment/data/SafeWritableMemory.java new file mode 100644 index 000000000000..9006ac5cec9e --- /dev/null +++ b/processing/src/main/java/org/apache/druid/segment/data/SafeWritableMemory.java @@ -0,0 +1,417 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.segment.data; + +import com.google.common.primitives.Ints; +import org.apache.datasketches.memory.Buffer; +import org.apache.datasketches.memory.Memory; +import org.apache.datasketches.memory.Utf8CodingException; +import org.apache.datasketches.memory.WritableBuffer; +import org.apache.datasketches.memory.WritableMemory; +import org.apache.druid.java.util.common.StringUtils; + +import java.io.IOException; +import java.nio.ByteBuffer; +import java.nio.ByteOrder; +import java.nio.channels.WritableByteChannel; + +/** + * Safety first! Don't trust something whose contents you locations to read and write stuff to, but need a + * {@link Memory} or {@link WritableMemory}? use this! + *

+ * Delegates everything to an underlying {@link ByteBuffer} so all read and write operations will have bounds checks + * built in rather than using 'unsafe'. + */ +public class SafeWritableMemory extends SafeWritableBase implements WritableMemory +{ + public static SafeWritableMemory wrap(byte[] bytes) + { + return wrap(ByteBuffer.wrap(bytes).order(ByteOrder.nativeOrder()), 0, bytes.length); + } + + public static SafeWritableMemory wrap(ByteBuffer buffer) + { + return wrap(buffer.duplicate().order(buffer.order()), 0, buffer.capacity()); + } + + public static SafeWritableMemory wrap(ByteBuffer buffer, ByteOrder byteOrder) + { + return wrap(buffer.duplicate().order(byteOrder), 0, buffer.capacity()); + } + + public static SafeWritableMemory wrap(ByteBuffer buffer, int offset, int size) + { + final ByteBuffer dupe = buffer.duplicate().order(buffer.order()); + dupe.position(offset); + dupe.limit(offset + size); + return new SafeWritableMemory(dupe.slice().order(buffer.order())); + } + + public SafeWritableMemory(ByteBuffer buffer) + { + super(buffer); + } + + @Override + public Memory region(long offsetBytes, long capacityBytes, ByteOrder byteOrder) + { + return writableRegion(offsetBytes, capacityBytes, byteOrder); + } + + @Override + public Buffer asBuffer(ByteOrder byteOrder) + { + return asWritableBuffer(byteOrder); + } + + @Override + public void getBooleanArray(long offsetBytes, boolean[] dstArray, int dstOffsetBooleans, int lengthBooleans) + { + final int offset = Ints.checkedCast(offsetBytes); + for (int j = 0; j < lengthBooleans; j++) { + dstArray[dstOffsetBooleans + j] = buffer.get(offset + j) != 0; + } + } + + @Override + public void getByteArray(long offsetBytes, byte[] dstArray, int dstOffsetBytes, int lengthBytes) + { + final int offset = Ints.checkedCast(offsetBytes); + for (int j = 0; j < lengthBytes; j++) { + dstArray[dstOffsetBytes + j] = buffer.get(offset + j); + } + } + + @Override + public void getCharArray(long offsetBytes, char[] dstArray, int dstOffsetChars, int lengthChars) + { + final int offset = Ints.checkedCast(offsetBytes); + for (int j = 0; j < lengthChars; j++) { + dstArray[dstOffsetChars + j] = buffer.getChar(offset + (j * Character.BYTES)); + } + } + + @Override + public int getCharsFromUtf8(long offsetBytes, int utf8LengthBytes, Appendable dst) + throws IOException, Utf8CodingException + { + ByteBuffer dupe = buffer.asReadOnlyBuffer().order(buffer.order()); + dupe.position(Ints.checkedCast(offsetBytes)); + String s = StringUtils.fromUtf8(dupe, utf8LengthBytes); + dst.append(s); + return s.length(); + } + + @Override + public int getCharsFromUtf8(long offsetBytes, int utf8LengthBytes, StringBuilder dst) throws Utf8CodingException + { + ByteBuffer dupe = buffer.asReadOnlyBuffer().order(buffer.order()); + dupe.position(Ints.checkedCast(offsetBytes)); + String s = StringUtils.fromUtf8(dupe, utf8LengthBytes); + dst.append(s); + return s.length(); + } + + @Override + public void getDoubleArray(long offsetBytes, double[] dstArray, int dstOffsetDoubles, int lengthDoubles) + { + final int offset = Ints.checkedCast(offsetBytes); + for (int j = 0; j < lengthDoubles; j++) { + dstArray[dstOffsetDoubles + j] = buffer.getDouble(offset + (j * Double.BYTES)); + } + } + + @Override + public void getFloatArray(long offsetBytes, float[] dstArray, int dstOffsetFloats, int lengthFloats) + { + final int offset = Ints.checkedCast(offsetBytes); + for (int j = 0; j < lengthFloats; j++) { + dstArray[dstOffsetFloats + j] = buffer.getFloat(offset + (j * Float.BYTES)); + } + } + + @Override + public void getIntArray(long offsetBytes, int[] dstArray, int dstOffsetInts, int lengthInts) + { + final int offset = Ints.checkedCast(offsetBytes); + for (int j = 0; j < lengthInts; j++) { + dstArray[dstOffsetInts + j] = buffer.getInt(offset + (j * Integer.BYTES)); + } + } + + @Override + public void getLongArray(long offsetBytes, long[] dstArray, int dstOffsetLongs, int lengthLongs) + { + final int offset = Ints.checkedCast(offsetBytes); + for (int j = 0; j < lengthLongs; j++) { + dstArray[dstOffsetLongs + j] = buffer.getLong(offset + (j * Long.BYTES)); + } + } + + @Override + public void getShortArray(long offsetBytes, short[] dstArray, int dstOffsetShorts, int lengthShorts) + { + final int offset = Ints.checkedCast(offsetBytes); + for (int j = 0; j < lengthShorts; j++) { + dstArray[dstOffsetShorts + j] = buffer.getShort(offset + (j * Short.BYTES)); + } + } + + @Override + public int compareTo( + long thisOffsetBytes, + long thisLengthBytes, + Memory that, + long thatOffsetBytes, + long thatLengthBytes + ) + { + final int thisLength = Ints.checkedCast(thisLengthBytes); + final int thatLength = Ints.checkedCast(thatLengthBytes); + + final int commonLength = Math.min(thisLength, thatLength); + + for (int i = 0; i < commonLength; i++) { + final int cmp = Byte.compare(getByte(thisOffsetBytes + i), that.getByte(thatOffsetBytes + i)); + if (cmp != 0) { + return cmp; + } + } + + return Integer.compare(thisLength, thatLength); + } + + @Override + public void copyTo(long srcOffsetBytes, WritableMemory destination, long dstOffsetBytes, long lengthBytes) + { + int offset = Ints.checkedCast(srcOffsetBytes); + for (int i = 0; i < lengthBytes; i++) { + destination.putByte(dstOffsetBytes + i, buffer.get(offset + i)); + } + } + + @Override + public void writeTo(long offsetBytes, long lengthBytes, WritableByteChannel out) throws IOException + { + ByteBuffer dupe = buffer.duplicate(); + dupe.position(Ints.checkedCast(offsetBytes)); + dupe.limit(dupe.position() + Ints.checkedCast(lengthBytes)); + ByteBuffer view = dupe.slice(); + view.order(buffer.order()); + out.write(view); + } + + @Override + public boolean equalTo(long thisOffsetBytes, Object that, long thatOffsetBytes, long lengthBytes) + { + if (!(that instanceof SafeWritableMemory)) { + return false; + } + return compareTo(thisOffsetBytes, lengthBytes, (SafeWritableMemory) that, thatOffsetBytes, lengthBytes) == 0; + } + + + @Override + public WritableMemory writableRegion(long offsetBytes, long capacityBytes, ByteOrder byteOrder) + { + final ByteBuffer dupe = buffer.duplicate().order(buffer.order()); + final int sizeBytes = Ints.checkedCast(capacityBytes); + dupe.position(Ints.checkedCast(offsetBytes)); + dupe.limit(dupe.position() + sizeBytes); + final ByteBuffer view = dupe.slice(); + view.order(byteOrder); + return new SafeWritableMemory(view); + } + + @Override + public WritableBuffer asWritableBuffer(ByteOrder byteOrder) + { + return new SafeWritableBuffer(buffer.duplicate().order(byteOrder)); + } + + @Override + public void putBooleanArray(long offsetBytes, boolean[] srcArray, int srcOffsetBooleans, int lengthBooleans) + { + final int offset = Ints.checkedCast(offsetBytes); + for (int i = 0; i < lengthBooleans; i++) { + buffer.put(offset + i, (byte) (srcArray[i + srcOffsetBooleans] ? 1 : 0)); + } + } + + @Override + public void putByteArray(long offsetBytes, byte[] srcArray, int srcOffsetBytes, int lengthBytes) + { + final int offset = Ints.checkedCast(offsetBytes); + for (int i = 0; i < lengthBytes; i++) { + buffer.put(offset + i, srcArray[srcOffsetBytes + i]); + } + } + + @Override + public void putCharArray(long offsetBytes, char[] srcArray, int srcOffsetChars, int lengthChars) + { + final int offset = Ints.checkedCast(offsetBytes); + for (int i = 0; i < lengthChars; i++) { + buffer.putChar(offset + (i * Character.BYTES), srcArray[srcOffsetChars + i]); + } + } + + @Override + public long putCharsToUtf8(long offsetBytes, CharSequence src) + { + final byte[] bytes = StringUtils.toUtf8(src.toString()); + putByteArray(offsetBytes, bytes, 0, bytes.length); + return bytes.length; + } + + @Override + public void putDoubleArray(long offsetBytes, double[] srcArray, int srcOffsetDoubles, int lengthDoubles) + { + final int offset = Ints.checkedCast(offsetBytes); + for (int i = 0; i < lengthDoubles; i++) { + buffer.putDouble(offset + (i * Double.BYTES), srcArray[srcOffsetDoubles + i]); + } + } + + @Override + public void putFloatArray(long offsetBytes, float[] srcArray, int srcOffsetFloats, int lengthFloats) + { + final int offset = Ints.checkedCast(offsetBytes); + for (int i = 0; i < lengthFloats; i++) { + buffer.putFloat(offset + (i * Float.BYTES), srcArray[srcOffsetFloats + i]); + } + } + + @Override + public void putIntArray(long offsetBytes, int[] srcArray, int srcOffsetInts, int lengthInts) + { + final int offset = Ints.checkedCast(offsetBytes); + for (int i = 0; i < lengthInts; i++) { + buffer.putInt(offset + (i * Integer.BYTES), srcArray[srcOffsetInts + i]); + } + } + + @Override + public void putLongArray(long offsetBytes, long[] srcArray, int srcOffsetLongs, int lengthLongs) + { + final int offset = Ints.checkedCast(offsetBytes); + for (int i = 0; i < lengthLongs; i++) { + buffer.putLong(offset + (i * Long.BYTES), srcArray[srcOffsetLongs + i]); + } + } + + @Override + public void putShortArray(long offsetBytes, short[] srcArray, int srcOffsetShorts, int lengthShorts) + { + final int offset = Ints.checkedCast(offsetBytes); + for (int i = 0; i < lengthShorts; i++) { + buffer.putShort(offset + (i * Short.BYTES), srcArray[srcOffsetShorts + i]); + } + } + + @Override + public long getAndAddLong(long offsetBytes, long delta) + { + final int offset = Ints.checkedCast(offsetBytes); + final long currentValue; + synchronized (buffer) { + currentValue = buffer.getLong(offset); + buffer.putLong(offset, currentValue + delta); + } + return currentValue; + } + + @Override + public boolean compareAndSwapLong(long offsetBytes, long expect, long update) + { + final int offset = Ints.checkedCast(offsetBytes); + synchronized (buffer) { + final long actual = buffer.getLong(offset); + if (expect == actual) { + buffer.putLong(offset, update); + return true; + } + } + return false; + } + + @Override + public long getAndSetLong(long offsetBytes, long newValue) + { + int offset = Ints.checkedCast(offsetBytes); + synchronized (buffer) { + long l = buffer.getLong(offset); + buffer.putLong(offset, newValue); + return l; + } + } + + @Override + public Object getArray() + { + return null; + } + + @Override + public void clear() + { + fill((byte) 0); + } + + @Override + public void clear(long offsetBytes, long lengthBytes) + { + fill(offsetBytes, lengthBytes, (byte) 0); + } + + @Override + public void clearBits(long offsetBytes, byte bitMask) + { + final int offset = Ints.checkedCast(offsetBytes); + int value = buffer.get(offset) & 0XFF; + value &= ~bitMask; + buffer.put(offset, (byte) value); + } + + @Override + public void fill(byte value) + { + for (int i = 0; i < buffer.capacity(); i++) { + buffer.put(i, value); + } + } + + @Override + public void fill(long offsetBytes, long lengthBytes, byte value) + { + int offset = Ints.checkedCast(offsetBytes); + int length = Ints.checkedCast(lengthBytes); + for (int i = 0; i < length; i++) { + buffer.put(offset + i, value); + } + } + + @Override + public void setBits(long offsetBytes, byte bitMask) + { + final int offset = Ints.checkedCast(offsetBytes); + buffer.put(offset, (byte) (buffer.get(offset) | bitMask)); + } +} diff --git a/processing/src/test/java/org/apache/druid/segment/data/SafeWritableBufferTest.java b/processing/src/test/java/org/apache/druid/segment/data/SafeWritableBufferTest.java new file mode 100644 index 000000000000..f432b7c167ca --- /dev/null +++ b/processing/src/test/java/org/apache/druid/segment/data/SafeWritableBufferTest.java @@ -0,0 +1,224 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.segment.data; + +import org.apache.datasketches.memory.Buffer; +import org.apache.datasketches.memory.WritableBuffer; +import org.junit.Assert; +import org.junit.Test; + +import java.nio.ByteBuffer; +import java.nio.ByteOrder; + +public class SafeWritableBufferTest +{ + private static final int CAPACITY = 1024; + + @Test + public void testPutAndGet() + { + WritableBuffer b1 = getBuffer(); + Assert.assertEquals(0, b1.getPosition()); + b1.putByte((byte) 0x01); + Assert.assertEquals(1, b1.getPosition()); + b1.putBoolean(true); + Assert.assertEquals(2, b1.getPosition()); + b1.putBoolean(false); + Assert.assertEquals(3, b1.getPosition()); + b1.putChar('c'); + Assert.assertEquals(5, b1.getPosition()); + b1.putDouble(1.1); + Assert.assertEquals(13, b1.getPosition()); + b1.putFloat(1.1f); + Assert.assertEquals(17, b1.getPosition()); + b1.putInt(100); + Assert.assertEquals(21, b1.getPosition()); + b1.putLong(1000L); + Assert.assertEquals(29, b1.getPosition()); + b1.putShort((short) 15); + Assert.assertEquals(31, b1.getPosition()); + b1.resetPosition(); + + Assert.assertEquals(0x01, b1.getByte()); + Assert.assertTrue(b1.getBoolean()); + Assert.assertFalse(b1.getBoolean()); + Assert.assertEquals('c', b1.getChar()); + Assert.assertEquals(1.1, b1.getDouble(), 0.0); + Assert.assertEquals(1.1f, b1.getFloat(), 0.0); + Assert.assertEquals(100, b1.getInt()); + Assert.assertEquals(1000L, b1.getLong()); + Assert.assertEquals(15, b1.getShort()); + } + + @Test + public void testPutAndGetArrays() + { + WritableBuffer buffer = getBuffer(); + final byte[] b1 = new byte[]{0x01, 0x02, 0x08, 0x08}; + final byte[] b2 = new byte[b1.length]; + + final boolean[] bool1 = new boolean[]{true, false, false, true}; + final boolean[] bool2 = new boolean[bool1.length]; + + final char[] chars1 = new char[]{'a', 'b', 'c', 'd'}; + final char[] chars2 = new char[chars1.length]; + + final double[] double1 = new double[]{1.1, -2.2, 3.3, 4.4}; + final double[] double2 = new double[double1.length]; + + final float[] float1 = new float[]{1.1f, 2.2f, -3.3f, 4.4f}; + final float[] float2 = new float[float1.length]; + + final int[] ints1 = new int[]{1, 2, -3, 4}; + final int[] ints2 = new int[ints1.length]; + + final long[] longs1 = new long[]{1L, -2L, 3L, -14L}; + final long[] longs2 = new long[ints1.length]; + + final short[] shorts1 = new short[]{1, -2, 3, -14}; + final short[] shorts2 = new short[ints1.length]; + + buffer.putByteArray(b1, 0, 2); + buffer.putByteArray(b1, 2, b1.length - 2); + buffer.putBooleanArray(bool1, 0, bool1.length); + buffer.putCharArray(chars1, 0, chars1.length); + buffer.putDoubleArray(double1, 0, double1.length); + buffer.putFloatArray(float1, 0, float1.length); + buffer.putIntArray(ints1, 0, ints1.length); + buffer.putLongArray(longs1, 0, longs1.length); + buffer.putShortArray(shorts1, 0, shorts1.length); + long pos = buffer.getPosition(); + buffer.resetPosition(); + buffer.getByteArray(b2, 0, b1.length); + buffer.getBooleanArray(bool2, 0, bool1.length); + buffer.getCharArray(chars2, 0, chars1.length); + buffer.getDoubleArray(double2, 0, double1.length); + buffer.getFloatArray(float2, 0, float1.length); + buffer.getIntArray(ints2, 0, ints1.length); + buffer.getLongArray(longs2, 0, longs1.length); + buffer.getShortArray(shorts2, 0, shorts1.length); + + Assert.assertArrayEquals(b1, b2); + Assert.assertArrayEquals(bool1, bool2); + Assert.assertArrayEquals(chars1, chars2); + for (int i = 0; i < double1.length; i++) { + Assert.assertEquals(double1[i], double2[i], 0.0); + } + for (int i = 0; i < float1.length; i++) { + Assert.assertEquals(float1[i], float2[i], 0.0); + } + Assert.assertArrayEquals(ints1, ints2); + Assert.assertArrayEquals(longs1, longs2); + Assert.assertArrayEquals(shorts1, shorts2); + + Assert.assertEquals(pos, buffer.getPosition()); + } + + @Test + public void testStartEndRegionAndDuplicate() + { + WritableBuffer buffer = getBuffer(); + Assert.assertEquals(0, buffer.getPosition()); + Assert.assertEquals(0, buffer.getStart()); + Assert.assertEquals(CAPACITY, buffer.getEnd()); + Assert.assertEquals(CAPACITY, buffer.getRemaining()); + Assert.assertEquals(CAPACITY, buffer.getCapacity()); + Assert.assertTrue(buffer.hasRemaining()); + buffer.fill((byte) 0x07); + buffer.setAndCheckStartPositionEnd(10L, 15L, 100L); + Assert.assertEquals(15L, buffer.getPosition()); + Assert.assertEquals(10L, buffer.getStart()); + Assert.assertEquals(100L, buffer.getEnd()); + Assert.assertEquals(85L, buffer.getRemaining()); + Assert.assertEquals(CAPACITY, buffer.getCapacity()); + buffer.fill((byte) 0x70); + buffer.resetPosition(); + Assert.assertEquals(10L, buffer.getPosition()); + for (int i = 0; i < 90; i++) { + if (i < 5) { + Assert.assertEquals(0x07, buffer.getByte()); + } else { + Assert.assertEquals(0x70, buffer.getByte()); + } + } + buffer.setAndCheckPosition(50); + + Buffer duplicate = buffer.duplicate(); + Assert.assertEquals(buffer.getStart(), duplicate.getStart()); + Assert.assertEquals(buffer.getPosition(), duplicate.getPosition()); + Assert.assertEquals(buffer.getEnd(), duplicate.getEnd()); + Assert.assertEquals(buffer.getRemaining(), duplicate.getRemaining()); + Assert.assertEquals(buffer.getCapacity(), duplicate.getCapacity()); + + duplicate.resetPosition(); + for (int i = 0; i < 90; i++) { + if (i < 5) { + Assert.assertEquals(0x07, duplicate.getByte()); + } else { + Assert.assertEquals(0x70, duplicate.getByte()); + } + } + + Buffer region = buffer.region(5L, 105L, buffer.getTypeByteOrder()); + Assert.assertEquals(0, region.getStart()); + Assert.assertEquals(0, region.getPosition()); + Assert.assertEquals(105L, region.getEnd()); + Assert.assertEquals(105L, region.getRemaining()); + Assert.assertEquals(105L, region.getCapacity()); + + for (int i = 0; i < 105; i++) { + if (i < 10) { + Assert.assertEquals(0x07, region.getByte()); + } else if (i < 95) { + Assert.assertEquals(0x70, region.getByte()); + } else { + Assert.assertEquals(0x07, region.getByte()); + } + } + } + + @Test + public void testFill() + { + WritableBuffer buffer = getBuffer(); + WritableBuffer anotherBuffer = getBuffer(); + + buffer.fill((byte) 0x0F); + anotherBuffer.fill((byte) 0x0F); + Assert.assertTrue(buffer.equalTo(0L, anotherBuffer, 0L, CAPACITY)); + + anotherBuffer.setPosition(100); + anotherBuffer.clear(); + Assert.assertFalse(buffer.equalTo(0L, anotherBuffer, 0L, CAPACITY)); + Assert.assertTrue(buffer.equalTo(0L, anotherBuffer, 0L, 100L)); + } + + private WritableBuffer getBuffer() + { + return getBuffer(CAPACITY); + } + + private WritableBuffer getBuffer(int capacity) + { + final ByteBuffer aBuffer = ByteBuffer.allocate(capacity).order(ByteOrder.LITTLE_ENDIAN); + SafeWritableBuffer memory = new SafeWritableBuffer(aBuffer); + return memory; + } +} diff --git a/processing/src/test/java/org/apache/druid/segment/data/SafeWritableMemoryTest.java b/processing/src/test/java/org/apache/druid/segment/data/SafeWritableMemoryTest.java new file mode 100644 index 000000000000..786443f43ed6 --- /dev/null +++ b/processing/src/test/java/org/apache/druid/segment/data/SafeWritableMemoryTest.java @@ -0,0 +1,359 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.segment.data; + +import org.apache.datasketches.memory.Memory; +import org.apache.datasketches.memory.WritableMemory; +import org.apache.datasketches.memory.internal.UnsafeUtil; +import org.junit.Assert; +import org.junit.Test; + +import java.io.CharArrayWriter; +import java.io.IOException; +import java.nio.ByteBuffer; +import java.nio.ByteOrder; + +public class SafeWritableMemoryTest +{ + private static final int CAPACITY = 1024; + + @Test + public void testPutAndGet() + { + final WritableMemory memory = getMemory(); + memory.putByte(3L, (byte) 0x01); + Assert.assertEquals(memory.getByte(3L), 0x01); + + memory.putBoolean(1L, true); + Assert.assertTrue(memory.getBoolean(1L)); + memory.putBoolean(1L, false); + Assert.assertFalse(memory.getBoolean(1L)); + + memory.putChar(10L, 'c'); + Assert.assertEquals('c', memory.getChar(10L)); + + memory.putDouble(14L, 3.3); + Assert.assertEquals(3.3, memory.getDouble(14L), 0.0); + + memory.putFloat(27L, 3.3f); + Assert.assertEquals(3.3f, memory.getFloat(27L), 0.0); + + memory.putInt(11L, 1234); + Assert.assertEquals(1234, memory.getInt(11L)); + + memory.putLong(500L, 500L); + Assert.assertEquals(500L, memory.getLong(500L)); + + memory.putShort(11L, (short) 15); + Assert.assertEquals(15, memory.getShort(11L)); + + long l = memory.getAndSetLong(900L, 10L); + Assert.assertEquals(0L, l); + l = memory.getAndSetLong(900L, 100L); + Assert.assertEquals(10L, l); + l = memory.getAndAddLong(900L, 10L); + Assert.assertEquals(100L, l); + Assert.assertEquals(110L, memory.getLong(900L)); + Assert.assertTrue(memory.compareAndSwapLong(900L, 110L, 120L)); + Assert.assertFalse(memory.compareAndSwapLong(900L, 110L, 120L)); + Assert.assertEquals(120L, memory.getLong(900L)); + } + + @Test + public void testPutAndGetArrays() + { + final WritableMemory memory = getMemory(); + final byte[] b1 = new byte[]{0x01, 0x02, 0x08, 0x08}; + final byte[] b2 = new byte[b1.length]; + memory.putByteArray(12L, b1, 0, 3); + memory.putByteArray(15L, b1, 3, 1); + memory.getByteArray(12L, b2, 0, 3); + memory.getByteArray(15L, b2, 3, 1); + Assert.assertArrayEquals(b1, b2); + + final boolean[] bool1 = new boolean[]{true, false, false, true}; + final boolean[] bool2 = new boolean[bool1.length]; + memory.putBooleanArray(100L, bool1, 0, 2); + memory.putBooleanArray(102L, bool1, 2, 2); + memory.getBooleanArray(100L, bool2, 0, 2); + memory.getBooleanArray(102L, bool2, 2, 2); + Assert.assertArrayEquals(bool1, bool2); + + final char[] chars1 = new char[]{'a', 'b', 'c', 'd'}; + final char[] chars2 = new char[chars1.length]; + memory.putCharArray(10L, chars1, 0, 4); + memory.getCharArray(10L, chars2, 0, chars1.length); + Assert.assertArrayEquals(chars1, chars2); + + final double[] double1 = new double[]{1.1, -2.2, 3.3, 4.4}; + final double[] double2 = new double[double1.length]; + memory.putDoubleArray(100L, double1, 0, 1); + memory.putDoubleArray(100L + Double.BYTES, double1, 1, 3); + memory.getDoubleArray(100L, double2, 0, 2); + memory.getDoubleArray(100L + (2 * Double.BYTES), double2, 2, 2); + for (int i = 0; i < double1.length; i++) { + Assert.assertEquals(double1[i], double2[i], 0.0); + } + + final float[] float1 = new float[]{1.1f, 2.2f, -3.3f, 4.4f}; + final float[] float2 = new float[float1.length]; + memory.putFloatArray(100L, float1, 0, 1); + memory.putFloatArray(100L + Float.BYTES, float1, 1, 3); + memory.getFloatArray(100L, float2, 0, 2); + memory.getFloatArray(100L + (2 * Float.BYTES), float2, 2, 2); + for (int i = 0; i < float1.length; i++) { + Assert.assertEquals(float1[i], float2[i], 0.0); + } + + final int[] ints1 = new int[]{1, 2, -3, 4}; + final int[] ints2 = new int[ints1.length]; + memory.putIntArray(100L, ints1, 0, 1); + memory.putIntArray(100L + Integer.BYTES, ints1, 1, 3); + memory.getIntArray(100L, ints2, 0, 2); + memory.getIntArray(100L + (2 * Integer.BYTES), ints2, 2, 2); + Assert.assertArrayEquals(ints1, ints2); + + final long[] longs1 = new long[]{1L, -2L, 3L, -14L}; + final long[] longs2 = new long[ints1.length]; + memory.putLongArray(100L, longs1, 0, 1); + memory.putLongArray(100L + Long.BYTES, longs1, 1, 3); + memory.getLongArray(100L, longs2, 0, 2); + memory.getLongArray(100L + (2 * Long.BYTES), longs2, 2, 2); + Assert.assertArrayEquals(longs1, longs2); + + final short[] shorts1 = new short[]{1, -2, 3, -14}; + final short[] shorts2 = new short[ints1.length]; + memory.putShortArray(100L, shorts1, 0, 1); + memory.putShortArray(100L + Short.BYTES, shorts1, 1, 3); + memory.getShortArray(100L, shorts2, 0, 2); + memory.getShortArray(100L + (2 * Short.BYTES), shorts2, 2, 2); + Assert.assertArrayEquals(shorts1, shorts2); + } + + @Test + public void testFill() + { + final byte theByte = 0x01; + final byte anotherByte = 0x02; + final WritableMemory memory = getMemory(); + final int halfWay = (int) (memory.getCapacity() / 2); + + memory.fill(theByte); + for (int i = 0; i < memory.getCapacity(); i++) { + Assert.assertEquals(theByte, memory.getByte(i)); + } + + memory.fill(halfWay, memory.getCapacity() - halfWay, anotherByte); + for (int i = 0; i < memory.getCapacity(); i++) { + if (i < halfWay) { + Assert.assertEquals(theByte, memory.getByte(i)); + } else { + Assert.assertEquals(anotherByte, memory.getByte(i)); + } + } + + memory.clear(halfWay, memory.getCapacity() - halfWay); + for (int i = 0; i < memory.getCapacity(); i++) { + if (i < halfWay) { + Assert.assertEquals(theByte, memory.getByte(i)); + } else { + Assert.assertEquals(0, memory.getByte(i)); + } + } + + memory.setBits(halfWay - 1, anotherByte); + Assert.assertEquals(0x03, memory.getByte(halfWay - 1)); + memory.clearBits(halfWay - 1, theByte); + Assert.assertEquals(anotherByte, memory.getByte(halfWay - 1)); + + memory.clear(); + for (int i = 0; i < memory.getCapacity(); i++) { + Assert.assertEquals(0, memory.getByte(i)); + } + } + + @Test + public void testStringStuff() throws IOException + { + WritableMemory memory = getMemory(); + String s1 = "hello "; + memory.putCharsToUtf8(10L, s1); + + StringBuilder builder = new StringBuilder(); + memory.getCharsFromUtf8(10L, s1.length(), builder); + Assert.assertEquals(s1, builder.toString()); + + CharArrayWriter someAppendable = new CharArrayWriter(); + memory.getCharsFromUtf8(10L, s1.length(), someAppendable); + Assert.assertEquals(s1, someAppendable.toString()); + } + + @Test + public void testRegion() + { + WritableMemory memory = getMemory(); + Assert.assertEquals(CAPACITY, memory.getCapacity()); + Assert.assertEquals(0, memory.getCumulativeOffset()); + Assert.assertEquals(10L, memory.getCumulativeOffset(10L)); + Assert.assertThrows( + IllegalArgumentException.class, + () -> memory.checkValidAndBounds(CAPACITY - 10, 11L) + ); + + final byte[] someBytes = new byte[]{0x01, 0x02, 0x03, 0x04}; + memory.putByteArray(10L, someBytes, 0, someBytes.length); + + Memory region = memory.region(10L, someBytes.length); + Assert.assertEquals(someBytes.length, region.getCapacity()); + Assert.assertEquals(0, region.getCumulativeOffset()); + Assert.assertEquals(2L, region.getCumulativeOffset(2L)); + Assert.assertThrows( + IllegalArgumentException.class, + () -> region.checkValidAndBounds(2L, 4L) + ); + + final byte[] andBack = new byte[someBytes.length]; + region.getByteArray(0L, andBack, 0, someBytes.length); + Assert.assertArrayEquals(someBytes, andBack); + + Memory differentOrderRegion = memory.region(10L, someBytes.length, ByteOrder.BIG_ENDIAN); + // different order + Assert.assertFalse(region.isByteOrderCompatible(differentOrderRegion.getTypeByteOrder())); + // contents are equal tho + Assert.assertTrue(region.equalTo(0L, differentOrderRegion, 0L, someBytes.length)); + } + + @Test + public void testCompareAndEquals() + { + WritableMemory memory = getMemory(); + final byte[] someBytes = new byte[]{0x01, 0x02, 0x03, 0x04}; + final byte[] shorterSameBytes = new byte[]{0x01, 0x02, 0x03}; + final byte[] differentBytes = new byte[]{0x02, 0x02, 0x03, 0x04}; + memory.putByteArray(10L, someBytes, 0, someBytes.length); + memory.putByteArray(400L, someBytes, 0, someBytes.length); + memory.putByteArray(200L, shorterSameBytes, 0, shorterSameBytes.length); + memory.putByteArray(500L, differentBytes, 0, differentBytes.length); + + Assert.assertEquals(0, memory.compareTo(10L, someBytes.length, memory, 400L, someBytes.length)); + Assert.assertEquals(4, memory.compareTo(10L, someBytes.length, memory, 200L, someBytes.length)); + Assert.assertEquals(-1, memory.compareTo(10L, someBytes.length, memory, 500L, differentBytes.length)); + + WritableMemory memory2 = getMemory(); + memory2.putByteArray(0L, someBytes, 0, someBytes.length); + + Assert.assertEquals(0, memory.compareTo(10L, someBytes.length, memory2, 0L, someBytes.length)); + + Assert.assertTrue(memory.equalTo(10L, memory2, 0L, someBytes.length)); + + WritableMemory memory3 = getMemory(); + memory2.copyTo(0L, memory3, 0L, CAPACITY); + Assert.assertTrue(memory2.equalTo(0L, memory3, 0L, CAPACITY)); + } + + @Test + public void testHash() + { + WritableMemory memory = getMemory(); + final long[] someLongs = new long[]{1L, 10L, 100L, 1000L, 10000L}; + final int[] someInts = new int[]{1, 2, 3}; + final byte[] someBytes = new byte[]{0x01, 0x02, 0x03}; + final int longsLength = Long.BYTES * someLongs.length; + final int someIntsLength = Integer.BYTES * someInts.length; + final int totalLength = longsLength + someIntsLength + someBytes.length; + memory.putLongArray(2L, someLongs, 0, someLongs.length); + memory.putIntArray(2L + longsLength, someInts, 0, someInts.length); + memory.putByteArray(2L + longsLength + someIntsLength, someBytes, 0, someBytes.length); + Memory memory2 = Memory.wrap(memory.getByteBuffer(), ByteOrder.LITTLE_ENDIAN); + Assert.assertEquals( + memory2.xxHash64(2L, totalLength, 0), + memory.xxHash64(2L, totalLength, 0) + ); + + Assert.assertEquals( + memory2.xxHash64(2L, 0), + memory.xxHash64(2L, 0) + ); + } + + @Test + public void testToHexString() + { + + final byte[] bytes = new byte[]{0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07}; + final WritableMemory memory = getMemory(bytes.length); + memory.putByteArray(0L, bytes, 0, bytes.length); + final long hcode = memory.hashCode() & 0XFFFFFFFFL; + final long bufferhcode = memory.getByteBuffer().hashCode() & 0XFFFFFFFFL; + final long reqhcode = memory.getMemoryRequestServer().hashCode() & 0XFFFFFFFFL; + Assert.assertEquals( + "### SafeWritableMemory SUMMARY ###\n" + + "Header Comment : test memory dump\n" + + "Call Parameters : .toHexString(..., 0, 8), hashCode: " + hcode + "\n" + + "UnsafeObj, hashCode : null\n" + + "UnsafeObjHeader : 0\n" + + "ByteBuf, hashCode : HeapByteBuffer, " + bufferhcode + "\n" + + "RegionOffset : 0\n" + + "Capacity : 8\n" + + "CumBaseOffset : 0\n" + + "MemReq, hashCode : HeapByteBufferMemoryRequestServer, " + reqhcode + "\n" + + "Valid : true\n" + + "Read Only : false\n" + + "Type Byte Order : LITTLE_ENDIAN\n" + + "Native Byte Order : LITTLE_ENDIAN\n" + + "JDK Runtime Version : " + UnsafeUtil.JDK + "\n" + + "Data, littleEndian : 0 1 2 3 4 5 6 7\n" + + " 0: 00 01 02 03 04 05 06 07 \n", + memory.toHexString("test memory dump", 0, bytes.length) + ); + } + + @Test + public void testMisc() + { + WritableMemory memory = getMemory(10); + WritableMemory memory2 = memory.getMemoryRequestServer().request(memory, 20); + Assert.assertEquals(20, memory2.getCapacity()); + + Assert.assertFalse(memory2.hasArray()); + + Assert.assertFalse(memory2.isReadOnly()); + Assert.assertFalse(memory2.isDirect()); + Assert.assertTrue(memory2.isValid()); + Assert.assertTrue(memory2.hasByteBuffer()); + + Assert.assertFalse(memory2.isSameResource(memory)); + Assert.assertTrue(memory2.isSameResource(memory2)); + + // does nothing + memory.getMemoryRequestServer().requestClose(memory, memory2); + } + + private WritableMemory getMemory() + { + return getMemory(CAPACITY); + } + + private WritableMemory getMemory(int capacity) + { + final ByteBuffer aBuffer = ByteBuffer.allocate(capacity).order(ByteOrder.LITTLE_ENDIAN); + return SafeWritableMemory.wrap(aBuffer); + } +} From a06e86bcfa12bcec8b788239f62cbe96581f2020 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Wed, 23 Nov 2022 14:50:04 +0530 Subject: [PATCH 032/141] update help documentation --- examples/bin/start-druid-main | 27 ++++++++++++--------------- 1 file changed, 12 insertions(+), 15 deletions(-) diff --git a/examples/bin/start-druid-main b/examples/bin/start-druid-main index 5444ef4bab3c..081c90d67211 100644 --- a/examples/bin/start-druid-main +++ b/examples/bin/start-druid-main @@ -46,7 +46,7 @@ ROUTER = "router" COORDINATOR = "coordinator-overlord" HISTORICAL = "historical" MIDDLE_MANAGER = "middleManager" -TASK = "task" +TASKS = "tasks" DEFAULT_SERVICES = [ BROKER, @@ -62,13 +62,13 @@ SERVICE_MEMORY_RATIO = { COORDINATOR: 30, BROKER: 46, HISTORICAL: 80, - TASK: 30 + TASKS: 30 } MINIMUM_MEMORY_MB = { MIDDLE_MANAGER: 64, ROUTER: 128, - TASK: 1024, + TASKS: 1024, BROKER: 900, COORDINATOR: 256, HISTORICAL: 900 @@ -80,7 +80,7 @@ HEAP_TO_TOTAL_MEM_RATIO = { COORDINATOR: 1, BROKER: 0.60, HISTORICAL: 0.40, - TASK: 0.50 + TASKS: 0.50 } LOGGING_ENABLED = False @@ -105,21 +105,18 @@ sample usage: Compute memory distribution and validate arguments. start-druid -m=100g -s=broker,router Starts a broker and a router, using a total memory of 100GB. - Reads config from conf/druid/single-server/quickstart. start-druid -m=100g --s=broker,router \\ -c=conf/druid/single-server/custom Starts a broker and a router, using a total memory of 100GB. - Reads configs from the given config path. + Reads configs for each service (jvm.config, runtime.properties) + from respective folders inside the given root config path. Throws an exception if there is a jvm.config present in any of the service directories inside the given root conf directory. start-druid -s=broker,router \\ -c=conf/druid/single-server/custom - Starts a broker and a router service, using upto 80% of the - total system memory. Calculates the memory requirements for - each service if no jvm.config is present for any of the - services. Does not calculate memory requirements if jvm.config - is present in all of the service directories. Throws an exception - if jvm.config is present in only some of the service directories. + Starts a broker and a router service, reading service configs + from the given root directory. Calculates memory requirements for + each service, if required, using upto 80% of the total system memory. start-druid -m=100g \\ -s=broker,router \\ -c=conf/druid/single-server/custom \\ @@ -272,7 +269,7 @@ def build_mm_task_javaopts_array(memory_type): return task_memory def build_memory_config_string(service, allocated_memory): - if service == TASK: + if service == TASKS: if allocated_memory >= 2048: task_count = int(allocated_memory / 2048) memory_type = TASK_MEM_TYPE_HIGH @@ -302,7 +299,7 @@ def distribute_memory_over_services(services, total_memory): service_list = services.copy() if MIDDLE_MANAGER in services: - service_list.append(TASK) + service_list.append(TASKS) for service in service_list: memory_weight_sum += SERVICE_MEMORY_RATIO.get(service) @@ -370,7 +367,7 @@ def build_supervise_script_arguments(service_list, service_memory_config, config commands.append(f'{prefix}{service} bin/run-druid {service} {config}') else: if service == MIDDLE_MANAGER: - task_config = service_memory_config.get(TASK) + task_config = service_memory_config.get(TASKS) task_count = task_config[0] task_memory = task_config[1] commands.append( From dc8f2b23b04eed8a89376764754e6a90e7787b5a Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Wed, 23 Nov 2022 14:52:09 +0530 Subject: [PATCH 033/141] update help documentation --- examples/bin/start-druid-main | 2 -- 1 file changed, 2 deletions(-) diff --git a/examples/bin/start-druid-main b/examples/bin/start-druid-main index 081c90d67211..312557125934 100644 --- a/examples/bin/start-druid-main +++ b/examples/bin/start-druid-main @@ -110,8 +110,6 @@ sample usage: Starts a broker and a router, using a total memory of 100GB. Reads configs for each service (jvm.config, runtime.properties) from respective folders inside the given root config path. - Throws an exception if there is a jvm.config present in any - of the service directories inside the given root conf directory. start-druid -s=broker,router \\ -c=conf/druid/single-server/custom Starts a broker and a router service, reading service configs From 589586f8d0edcacf241452dd0cc068863ea875b9 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Wed, 23 Nov 2022 15:09:17 +0530 Subject: [PATCH 034/141] move task memory computation into separate method --- examples/bin/start-druid-main | 72 +++++++++++++++++++++++------------ 1 file changed, 47 insertions(+), 25 deletions(-) diff --git a/examples/bin/start-druid-main b/examples/bin/start-druid-main index 312557125934..7605ee8a0097 100644 --- a/examples/bin/start-druid-main +++ b/examples/bin/start-druid-main @@ -30,7 +30,8 @@ MEM_GB_SUFFIX = "g" MEM_MB_SUFFIX = "m" SERVICE_SEPARATOR = "," -TASK_JAVA_OPTS_ARRAY = ["-server", "-Duser.timezone=UTC", "-Dfile.encoding=UTF-8", "-XX:+ExitOnOutOfMemoryError", "-Djava.util.logging.manager=org.apache.logging.log4j.jul.LogManager"] +TASK_JAVA_OPTS_ARRAY = ["-server", "-Duser.timezone=UTC", "-Dfile.encoding=UTF-8", "-XX:+ExitOnOutOfMemoryError", + "-Djava.util.logging.manager=org.apache.logging.log4j.jul.LogManager"] TASK_JAVA_OPTS_PROPERTY = "-Ddruid.indexer.runner.javaOptsArray" TASK_WORKER_CAPACITY_PROPERTY = "-Ddruid.worker.capacity" TASK_COUNT = "task-count" @@ -85,16 +86,18 @@ HEAP_TO_TOTAL_MEM_RATIO = { LOGGING_ENABLED = False + def print_if_verbose(message): if LOGGING_ENABLED: print(message) + def configure_parser(): parser = argparse.ArgumentParser( prog='Druid quickstart', formatter_class=argparse.RawTextHelpFormatter, epilog= -""" + """ sample usage: start-druid Start up all the services (including zk). @@ -153,6 +156,7 @@ sample usage: return parser + def parse_arguments(args): service_list = [] config = "" @@ -167,7 +171,7 @@ def parse_arguments(args): if args.config is not None: config = pathlib.Path(os.path.join(os.getcwd(), args.config)).resolve() if os.path.exists(config) is False: - raise ValueError(f'config {config} doesn\'t exist') + raise ValueError(f'config {config} not found') if args.memory is not None: total_memory = args.memory if args.services is not None: @@ -189,6 +193,7 @@ def parse_arguments(args): return config, total_memory, service_list, zk, compute + def print_startup_config(service_list, config, zk): print_if_verbose(f'starting {service_list}') print_if_verbose(f'reading config from {config}') @@ -197,6 +202,7 @@ def print_startup_config(service_list, config, zk): print_if_verbose(f'starting zk, reading default config from {zk_config}') print_if_verbose('\n') + def should_compute_memory(config, total_memory, service_list): # if jvm file is present for any of the services # it should be present for all services and memory should not be specified @@ -206,7 +212,9 @@ def should_compute_memory(config, total_memory, service_list): if pathlib.Path(f'{config}/{service}/jvm.config').is_file(): jvm_config_count += 1 elif jvm_config_count > 0: - raise ValueError(f'jvm.config file is missing for service {service}, jvm.config should be specified for all the services or none') + raise ValueError( + f'jvm.config file is missing for service {service}, jvm.config should be specified for all the ' + f'services or none') if jvm_config_count > 0 and (jvm_config_count != len(service_list) or total_memory != ""): if jvm_config_count != len(service_list): @@ -216,11 +224,13 @@ def should_compute_memory(config, total_memory, service_list): return jvm_config_count == 0 + def compute_system_memory(): - system_memory = psutil.virtual_memory().total # mem in bytes + system_memory = psutil.virtual_memory().total # mem in bytes memory_for_druid = int(system_memory / (1024 * 1024)) return memory_for_druid + def convert_total_memory_string(memory): try: if memory == "": @@ -235,6 +245,7 @@ def convert_total_memory_string(memory): except Exception: raise ValueError('Incorrect format for memory argument, expected format is ') + def check_memory_constraint(total_memory, service_list): # 80% of total memory >= sum of lower bound service memory should be lower_bound_memory = 0 @@ -252,34 +263,41 @@ def check_memory_constraint(total_memory, service_list): else: return lower_bound_memory -def build_mm_task_javaopts_array(memory_type): + +def build_mm_task_java_opts_array(memory_type): task_memory = f'{TASK_JAVA_OPTS_PROPERTY}=[' - MEM_ARRAY = TASK_MEM_MAP.get(memory_type) + mem_array = TASK_MEM_MAP.get(memory_type) - javaopts_list = TASK_JAVA_OPTS_ARRAY + MEM_ARRAY + java_opts_list = TASK_JAVA_OPTS_ARRAY + mem_array - for item in javaopts_list: + for item in java_opts_list: task_memory += f'\"{item}\";' task_memory = task_memory[:-1] task_memory += ']' return task_memory -def build_memory_config_string(service, allocated_memory): - if service == TASKS: - if allocated_memory >= 2048: - task_count = int(allocated_memory / 2048) - memory_type = TASK_MEM_TYPE_HIGH - task_memory = 2048 - else: - task_count = int(allocated_memory / 512) - memory_type = TASK_MEM_TYPE_LOW - task_memory = 512 - task_count = min(task_count, multiprocessing.cpu_count()) - javaopts_array = build_mm_task_javaopts_array(memory_type) - return [f'{TASK_WORKER_CAPACITY_PROPERTY}={task_count}', javaopts_array], task_memory * task_count +def compute_tasks_memory(allocated_memory): + if allocated_memory >= 2048: + task_count = int(allocated_memory / 2048) + memory_type = TASK_MEM_TYPE_HIGH + task_memory_mb = 2048 + else: + task_count = int(allocated_memory / 512) + memory_type = TASK_MEM_TYPE_LOW + task_memory_mb = 512 + task_count = min(task_count, multiprocessing.cpu_count()) + + return memory_type, task_count, task_memory_mb + + +def build_memory_config(service, allocated_memory): + if service == TASKS: + memory_type, task_count, task_memory = compute_tasks_memory(allocated_memory) + java_opts_array = build_mm_task_java_opts_array(memory_type) + return [f'{TASK_WORKER_CAPACITY_PROPERTY}={task_count}', java_opts_array], task_memory * task_count else: heap_memory = HEAP_TO_TOTAL_MEM_RATIO.get(service) * allocated_memory direct_memory = int(allocated_memory - heap_memory) @@ -290,6 +308,7 @@ def build_memory_config_string(service, allocated_memory): return f'-Xms{heap_memory}m -Xmx{heap_memory}m -XX:MaxDirectMemorySize={direct_memory}m', allocated_memory + def distribute_memory_over_services(services, total_memory): service_memory_config = {} @@ -311,7 +330,7 @@ def distribute_memory_over_services(services, total_memory): allocated_memory = SERVICE_MEMORY_RATIO.get(service) * multiplier if service in MINIMUM_MEMORY_MB and allocated_memory < MINIMUM_MEMORY_MB.get(service): allocated_memory = MINIMUM_MEMORY_MB.get(service) - service_memory_config[service], allocated_memory = build_memory_config_string(service, allocated_memory) + service_memory_config[service], allocated_memory = build_memory_config(service, allocated_memory) lower_bound_memory_allocation += allocated_memory allocated_services.add(service) @@ -331,7 +350,7 @@ def distribute_memory_over_services(services, total_memory): if service in MINIMUM_MEMORY_MB and allocated_memory < MINIMUM_MEMORY_MB.get(service): allocated_memory = MINIMUM_MEMORY_MB.get(service) - service_memory_config[service], allocated_memory = build_memory_config_string(service, allocated_memory) + service_memory_config[service], allocated_memory = build_memory_config(service, allocated_memory) print_if_verbose(f'\nMemory distribution for services:') for key, value in service_memory_config.items(): @@ -340,6 +359,7 @@ def distribute_memory_over_services(services, total_memory): return service_memory_config + def build_supervise_script_arguments(service_list, service_memory_config, config, zk): args = [] commands = [] @@ -382,6 +402,7 @@ def build_supervise_script_arguments(service_list, service_memory_config, config args.append(",".join(commands)) return args + def main(): parser = configure_parser() args = parser.parse_args() @@ -401,7 +422,7 @@ def main(): service_memory_config = {} - if (should_compute_memory(config, total_memory, service_list)): + if should_compute_memory(config, total_memory, service_list): memory_in_mega_bytes = convert_total_memory_string(total_memory) print_if_verbose(f'total memory is {memory_in_mega_bytes}m') memory_to_be_used = check_memory_constraint(memory_in_mega_bytes, service_list) @@ -417,6 +438,7 @@ def main(): os.execv('./supervise', script_arguments) + try: main() except (KeyboardInterrupt, ValueError) as error: From b78b1164d81b9c15f814b902712bd5b1afedc47f Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Wed, 23 Nov 2022 19:05:01 +0530 Subject: [PATCH 035/141] Add validation checks --- examples/bin/start-druid-main | 71 ++++++++++++++++--- .../apache/druid/guice/StorageNodeModule.java | 2 +- 2 files changed, 63 insertions(+), 10 deletions(-) diff --git a/examples/bin/start-druid-main b/examples/bin/start-druid-main index 7605ee8a0097..07a23586aacc 100644 --- a/examples/bin/start-druid-main +++ b/examples/bin/start-druid-main @@ -28,12 +28,15 @@ QUICKSTART_ROOT_CONFIG_PATH = "conf/druid/single-server/quickstart" MEM_GB_SUFFIX = "g" MEM_MB_SUFFIX = "m" +XMX_PARAMETER = "-Xmx" +XMS_PARAMETER = "-Xms" +DIRECT_MEM_PARAMETER = "-XX:MaxDirectMemorySize" SERVICE_SEPARATOR = "," TASK_JAVA_OPTS_ARRAY = ["-server", "-Duser.timezone=UTC", "-Dfile.encoding=UTF-8", "-XX:+ExitOnOutOfMemoryError", "-Djava.util.logging.manager=org.apache.logging.log4j.jul.LogManager"] -TASK_JAVA_OPTS_PROPERTY = "-Ddruid.indexer.runner.javaOptsArray" -TASK_WORKER_CAPACITY_PROPERTY = "-Ddruid.worker.capacity" +TASK_JAVA_OPTS_PROPERTY = "druid.indexer.runner.javaOptsArray" +TASK_WORKER_CAPACITY_PROPERTY = "druid.worker.capacity" TASK_COUNT = "task-count" TASK_MEM_TYPE_LOW = "low" TASK_MEM_TYPE_HIGH = "high" @@ -157,6 +160,18 @@ sample usage: return parser +def validate_common_directory(config): + if pathlib.Path(f'{config}/_common').is_dir() is False: + raise ValueError( + f'_common directory is missing in the root config, check {QUICKSTART_ROOT_CONFIG_PATH}/_common directory') + + if pathlib.Path(f'{config}/_common/common.runtime.properties').is_file() is False: + raise ValueError(f'_common/common.runtime.properties file is missing in the root config, check {QUICKSTART_ROOT_CONFIG_PATH}/_common directory') + + if pathlib.Path(f'{config}/_common/common.jvm.config').is_file() is False: + raise ValueError(f'_common/jvm.config file is missing in the root config, check {QUICKSTART_ROOT_CONFIG_PATH}/_common directory') + + def parse_arguments(args): service_list = [] config = "" @@ -172,6 +187,7 @@ def parse_arguments(args): config = pathlib.Path(os.path.join(os.getcwd(), args.config)).resolve() if os.path.exists(config) is False: raise ValueError(f'config {config} not found') + validate_common_directory(config) if args.memory is not None: total_memory = args.memory if args.services is not None: @@ -203,6 +219,42 @@ def print_startup_config(service_list, config, zk): print_if_verbose('\n') +def verify_service_config(service, config): + path = f'{config}/{service}/jvm.config' + + required_parameters = [XMX_PARAMETER, XMS_PARAMETER] + + if HEAP_TO_TOTAL_MEM_RATIO.get(service) != 1: + required_parameters.append(DIRECT_MEM_PARAMETER) + + with open(path) as file: + for line in file: + if line.startswith(XMX_PARAMETER): + required_parameters.remove(XMX_PARAMETER) + if line.startswith(XMS_PARAMETER): + required_parameters.remove(XMS_PARAMETER) + if line.startswith(DIRECT_MEM_PARAMETER) and DIRECT_MEM_PARAMETER in required_parameters: + required_parameters.remove(DIRECT_MEM_PARAMETER) + + if len(required_parameters) > 0: + params = ",".join(required_parameters) + raise ValueError(f'{params} missing in {service}/jvm.config') + + if service == MIDDLE_MANAGER: + if pathlib.Path(f'{config}/{service}/runtime.properties').is_file() is False: + raise ValueError(f'{service}/runtime.properties file is missing in the root config') + + java_opts_property_present = False + print(config) + with open(f'{config}/{service}/runtime.properties') as file: + for line in file: + if line.startswith(TASK_JAVA_OPTS_PROPERTY): + java_opts_property_present = True + + if java_opts_property_present is False: + raise ValueError(f'{TASK_JAVA_OPTS_PROPERTY} property missing in {service}/runtime.properties') + + def should_compute_memory(config, total_memory, service_list): # if jvm file is present for any of the services # it should be present for all services and memory should not be specified @@ -216,11 +268,13 @@ def should_compute_memory(config, total_memory, service_list): f'jvm.config file is missing for service {service}, jvm.config should be specified for all the ' f'services or none') - if jvm_config_count > 0 and (jvm_config_count != len(service_list) or total_memory != ""): + if jvm_config_count > 0: if jvm_config_count != len(service_list): raise ValueError("jvm.config file should be present for all services or none") if total_memory != "": raise ValueError("If jvm.config is given for services, memory argument shouldn't be specified") + for service in service_list: + verify_service_config(service, config) return jvm_config_count == 0 @@ -265,7 +319,7 @@ def check_memory_constraint(total_memory, service_list): def build_mm_task_java_opts_array(memory_type): - task_memory = f'{TASK_JAVA_OPTS_PROPERTY}=[' + task_memory = f'-D{TASK_JAVA_OPTS_PROPERTY}=[' mem_array = TASK_MEM_MAP.get(memory_type) @@ -297,7 +351,7 @@ def build_memory_config(service, allocated_memory): if service == TASKS: memory_type, task_count, task_memory = compute_tasks_memory(allocated_memory) java_opts_array = build_mm_task_java_opts_array(memory_type) - return [f'{TASK_WORKER_CAPACITY_PROPERTY}={task_count}', java_opts_array], task_memory * task_count + return [f'-D{TASK_WORKER_CAPACITY_PROPERTY}={task_count}', java_opts_array], task_memory * task_count else: heap_memory = HEAP_TO_TOTAL_MEM_RATIO.get(service) * allocated_memory direct_memory = int(allocated_memory - heap_memory) @@ -309,7 +363,7 @@ def build_memory_config(service, allocated_memory): return f'-Xms{heap_memory}m -Xmx{heap_memory}m -XX:MaxDirectMemorySize={direct_memory}m', allocated_memory -def distribute_memory_over_services(services, total_memory): +def distribute_memory(services, total_memory): service_memory_config = {} memory_weight_sum = 0 @@ -354,7 +408,7 @@ def distribute_memory_over_services(services, total_memory): print_if_verbose(f'\nMemory distribution for services:') for key, value in service_memory_config.items(): - print_if_verbose(f'{key}, memory_config: {value}') + print_if_verbose(f'{key}, {value}') print_if_verbose('\n') return service_memory_config @@ -427,7 +481,7 @@ def main(): print_if_verbose(f'total memory is {memory_in_mega_bytes}m') memory_to_be_used = check_memory_constraint(memory_in_mega_bytes, service_list) print_if_verbose(f'memory used for services & tasks {memory_to_be_used}m') - service_memory_config = distribute_memory_over_services(service_list, memory_to_be_used) + service_memory_config = distribute_memory(service_list, memory_to_be_used) else: print_if_verbose('not computing memory distribution, reading memory specification from service jvm.config\n') @@ -438,7 +492,6 @@ def main(): os.execv('./supervise', script_arguments) - try: main() except (KeyboardInterrupt, ValueError) as error: diff --git a/server/src/main/java/org/apache/druid/guice/StorageNodeModule.java b/server/src/main/java/org/apache/druid/guice/StorageNodeModule.java index a232f6d01aac..842abd0e5347 100644 --- a/server/src/main/java/org/apache/druid/guice/StorageNodeModule.java +++ b/server/src/main/java/org/apache/druid/guice/StorageNodeModule.java @@ -101,7 +101,7 @@ public DataNodeService getDataNodeService( serverTypeConfig.getServerType() ); if (ServerType.HISTORICAL.equals(serverTypeConfig.getServerType())) { - throw new ProvisionException("Segment cache locations must be set on historicals."); + throw new ProvisionException("druid.segmentCache.locations must be set on historicals."); } } From a2887049dfbbef8266cf31e93e9aa4972b9af689 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Wed, 23 Nov 2022 19:12:21 +0530 Subject: [PATCH 036/141] remove print --- examples/bin/start-druid-main | 1 - 1 file changed, 1 deletion(-) diff --git a/examples/bin/start-druid-main b/examples/bin/start-druid-main index 07a23586aacc..3146a06a1290 100644 --- a/examples/bin/start-druid-main +++ b/examples/bin/start-druid-main @@ -245,7 +245,6 @@ def verify_service_config(service, config): raise ValueError(f'{service}/runtime.properties file is missing in the root config') java_opts_property_present = False - print(config) with open(f'{config}/{service}/runtime.properties') as file: for line in file: if line.startswith(TASK_JAVA_OPTS_PROPERTY): From 16385c71011aaf7568345657491d4350c26438ea Mon Sep 17 00:00:00 2001 From: "dependabot[bot]" <49699333+dependabot[bot]@users.noreply.github.com> Date: Wed, 23 Nov 2022 12:16:00 -0800 Subject: [PATCH 037/141] Bump minimatch and replace in /web-console (#13396) Bumps [minimatch](https://github.com/isaacs/minimatch) to 3.0.5 and updates ancestor dependency [replace](https://github.com/ALMaclaine/replace). These dependencies need to be updated together. Updates `minimatch` from 3.0.4 to 3.0.5 - [Release notes](https://github.com/isaacs/minimatch/releases) - [Commits](https://github.com/isaacs/minimatch/compare/v3.0.4...v3.0.5) Updates `replace` from 1.2.1 to 1.2.2 - [Release notes](https://github.com/ALMaclaine/replace/releases) - [Commits](https://github.com/ALMaclaine/replace/commits) --- updated-dependencies: - dependency-name: minimatch dependency-type: indirect - dependency-name: replace dependency-type: direct:development ... Signed-off-by: dependabot[bot] Signed-off-by: dependabot[bot] Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> --- web-console/package-lock.json | 30 +++++++++++++++--------------- web-console/package.json | 2 +- 2 files changed, 16 insertions(+), 16 deletions(-) diff --git a/web-console/package-lock.json b/web-console/package-lock.json index 900bee41578a..8776b62e3b55 100644 --- a/web-console/package-lock.json +++ b/web-console/package-lock.json @@ -103,7 +103,7 @@ "postcss-loader": "^5.3.0", "postcss-preset-env": "^6.7.0", "prettier": "^2.5.1", - "replace": "^1.2.1", + "replace": "^1.2.2", "sass-loader": "^11.1.0", "snarkdown": "^2.0.0", "style-loader": "^2.0.0", @@ -18238,9 +18238,9 @@ "dev": true }, "node_modules/minimatch": { - "version": "3.0.4", - "resolved": "https://registry.npmjs.org/minimatch/-/minimatch-3.0.4.tgz", - "integrity": "sha512-yJHVQEhyqPLUTgt9B83PXu6W3rx4MvvHvSUvToogpwoGDOUQ+yDrR0HRot+yOCdCO7u4hX3pWft6kWBBcqh0UA==", + "version": "3.0.5", + "resolved": "https://registry.npmjs.org/minimatch/-/minimatch-3.0.5.tgz", + "integrity": "sha512-tUpxzX0VAzJHjLu0xUfFv1gwVp9ba3IOuRAVH2EGuRW8a5emA2FlACLqiT/lDVtS1W+TGNwqz3sWaNyLgDJWuw==", "dev": true, "dependencies": { "brace-expansion": "^1.1.7" @@ -22276,13 +22276,13 @@ } }, "node_modules/replace": { - "version": "1.2.1", - "resolved": "https://registry.npmjs.org/replace/-/replace-1.2.1.tgz", - "integrity": "sha512-KZCBe/tPanwBlbjSMQby4l+zjSiFi3CLEP/6VLClnRYgJ46DZ5u9tmA6ceWeFS8coaUnU4ZdGNb/puUGMHNSRg==", + "version": "1.2.2", + "resolved": "https://registry.npmjs.org/replace/-/replace-1.2.2.tgz", + "integrity": "sha512-C4EDifm22XZM2b2JOYe6Mhn+lBsLBAvLbK8drfUQLTfD1KYl/n3VaW/CDju0Ny4w3xTtegBpg8YNSpFJPUDSjA==", "dev": true, "dependencies": { "chalk": "2.4.2", - "minimatch": "3.0.4", + "minimatch": "3.0.5", "yargs": "^15.3.1" }, "bin": { @@ -41739,9 +41739,9 @@ "dev": true }, "minimatch": { - "version": "3.0.4", - "resolved": "https://registry.npmjs.org/minimatch/-/minimatch-3.0.4.tgz", - "integrity": "sha512-yJHVQEhyqPLUTgt9B83PXu6W3rx4MvvHvSUvToogpwoGDOUQ+yDrR0HRot+yOCdCO7u4hX3pWft6kWBBcqh0UA==", + "version": "3.0.5", + "resolved": "https://registry.npmjs.org/minimatch/-/minimatch-3.0.5.tgz", + "integrity": "sha512-tUpxzX0VAzJHjLu0xUfFv1gwVp9ba3IOuRAVH2EGuRW8a5emA2FlACLqiT/lDVtS1W+TGNwqz3sWaNyLgDJWuw==", "dev": true, "requires": { "brace-expansion": "^1.1.7" @@ -44807,13 +44807,13 @@ "dev": true }, "replace": { - "version": "1.2.1", - "resolved": "https://registry.npmjs.org/replace/-/replace-1.2.1.tgz", - "integrity": "sha512-KZCBe/tPanwBlbjSMQby4l+zjSiFi3CLEP/6VLClnRYgJ46DZ5u9tmA6ceWeFS8coaUnU4ZdGNb/puUGMHNSRg==", + "version": "1.2.2", + "resolved": "https://registry.npmjs.org/replace/-/replace-1.2.2.tgz", + "integrity": "sha512-C4EDifm22XZM2b2JOYe6Mhn+lBsLBAvLbK8drfUQLTfD1KYl/n3VaW/CDju0Ny4w3xTtegBpg8YNSpFJPUDSjA==", "dev": true, "requires": { "chalk": "2.4.2", - "minimatch": "3.0.4", + "minimatch": "3.0.5", "yargs": "^15.3.1" }, "dependencies": { diff --git a/web-console/package.json b/web-console/package.json index 97a03efface0..b8c33c6fdade 100644 --- a/web-console/package.json +++ b/web-console/package.json @@ -160,7 +160,7 @@ "postcss-loader": "^5.3.0", "postcss-preset-env": "^6.7.0", "prettier": "^2.5.1", - "replace": "^1.2.1", + "replace": "^1.2.2", "sass-loader": "^11.1.0", "snarkdown": "^2.0.0", "style-loader": "^2.0.0", From 81134a6c0741aaa2c850aca3c374d5af0e8e0f21 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Thu, 24 Nov 2022 14:05:27 +0530 Subject: [PATCH 038/141] Add validations --- examples/bin/start-druid-main | 100 ++++++++++++++++++++++++---------- 1 file changed, 71 insertions(+), 29 deletions(-) diff --git a/examples/bin/start-druid-main b/examples/bin/start-druid-main index 3146a06a1290..f5acbc5df6b9 100644 --- a/examples/bin/start-druid-main +++ b/examples/bin/start-druid-main @@ -160,16 +160,20 @@ sample usage: return parser +def validate_common_jvm_args(config): + if pathlib.Path(f'{config}/_common/common.jvm.config').is_file() is False: + raise ValueError(f'_common/common.jvm.config file is missing in the root config, ' + f'check {QUICKSTART_ROOT_CONFIG_PATH}/_common directory') + + def validate_common_directory(config): if pathlib.Path(f'{config}/_common').is_dir() is False: raise ValueError( f'_common directory is missing in the root config, check {QUICKSTART_ROOT_CONFIG_PATH}/_common directory') if pathlib.Path(f'{config}/_common/common.runtime.properties').is_file() is False: - raise ValueError(f'_common/common.runtime.properties file is missing in the root config, check {QUICKSTART_ROOT_CONFIG_PATH}/_common directory') - - if pathlib.Path(f'{config}/_common/common.jvm.config').is_file() is False: - raise ValueError(f'_common/jvm.config file is missing in the root config, check {QUICKSTART_ROOT_CONFIG_PATH}/_common directory') + raise ValueError(f'_common/common.runtime.properties file is missing in the root config, ' + f'check {QUICKSTART_ROOT_CONFIG_PATH}/_common directory') def parse_arguments(args): @@ -211,14 +215,29 @@ def parse_arguments(args): def print_startup_config(service_list, config, zk): - print_if_verbose(f'starting {service_list}') - print_if_verbose(f'reading config from {config}') + print_if_verbose(f'Starting {service_list}') + print_if_verbose(f'Reading config from {config}') if zk: zk_config = pathlib.Path(f'{os.getcwd()}/../conf/zk').resolve() - print_if_verbose(f'starting zk, reading default config from {zk_config}') + print_if_verbose(f'Starting zk, reading default config from {zk_config}') print_if_verbose('\n') +def middle_manager_task_memory_params_present(config): + java_opts_property_present = False + worker_capacity_property_present = False + + if pathlib.Path(f'{config}/middleManager/runtime.properties').is_file(): + with open(f'{config}/middleManager/runtime.properties') as file: + for line in file: + if line.startswith(TASK_JAVA_OPTS_PROPERTY): + java_opts_property_present = True + elif line.startswith(TASK_WORKER_CAPACITY_PROPERTY): + worker_capacity_property_present = True + + return java_opts_property_present, worker_capacity_property_present + + def verify_service_config(service, config): path = f'{config}/{service}/jvm.config' @@ -229,9 +248,9 @@ def verify_service_config(service, config): with open(path) as file: for line in file: - if line.startswith(XMX_PARAMETER): + if line.startswith(XMX_PARAMETER) and XMX_PARAMETER in required_parameters: required_parameters.remove(XMX_PARAMETER) - if line.startswith(XMS_PARAMETER): + if line.startswith(XMS_PARAMETER) and XMS_PARAMETER in required_parameters: required_parameters.remove(XMS_PARAMETER) if line.startswith(DIRECT_MEM_PARAMETER) and DIRECT_MEM_PARAMETER in required_parameters: required_parameters.remove(DIRECT_MEM_PARAMETER) @@ -244,38 +263,57 @@ def verify_service_config(service, config): if pathlib.Path(f'{config}/{service}/runtime.properties').is_file() is False: raise ValueError(f'{service}/runtime.properties file is missing in the root config') - java_opts_property_present = False - with open(f'{config}/{service}/runtime.properties') as file: - for line in file: - if line.startswith(TASK_JAVA_OPTS_PROPERTY): - java_opts_property_present = True + mm_task_java_opts_property, mm_task_worker_capacity_prop = middle_manager_task_memory_params_present(config) - if java_opts_property_present is False: + if mm_task_java_opts_property is False: raise ValueError(f'{TASK_JAVA_OPTS_PROPERTY} property missing in {service}/runtime.properties') def should_compute_memory(config, total_memory, service_list): - # if jvm file is present for any of the services - # it should be present for all services and memory should not be specified - # if memory is given, jvm file shouldn't be present for any service + """ + if memory argument is given, memory for services and tasks is computed, jvm.config file + or runtime.properties with task memory specification shouldn't be present + Alternatively, all memory related parameters are specified + which implies following should be present: + jvm.config file for all services with -Xmx=***, Xms=*** parameters + -XX:MaxDirectMemorySize=** in jvm.config for broker and historical + druid.indexer.runner.javaOptsArray (optionally druid.worker.capacity) in + rootDirectory/middleManager/runtime.properties + """ + jvm_config_count = 0 for service in service_list: if pathlib.Path(f'{config}/{service}/jvm.config').is_file(): jvm_config_count += 1 - elif jvm_config_count > 0: - raise ValueError( - f'jvm.config file is missing for service {service}, jvm.config should be specified for all the ' - f'services or none') - if jvm_config_count > 0: - if jvm_config_count != len(service_list): - raise ValueError("jvm.config file should be present for all services or none") + mm_task_property_present = False + if MIDDLE_MANAGER in service_list: + mm_task_java_opts_property, mm_task_worker_capacity_prop = middle_manager_task_memory_params_present(config) + mm_task_property_present = mm_task_java_opts_property or mm_task_worker_capacity_prop + + # possible error states + # 1. memory argument is specified, also jvm.config or middleManger/runtime.properties having + # druid.indexer.runner.javaOptsArray or druid.worker.capacity parameters is present + # 2. jvm.config is not present for any service, but middleManger/runtime.properties has + # druid.indexer.runner.javaOptsArray or druid.worker.capacity parameters + # 3. jvm.config present for some but not all services + # 4. jvm.config file is present for all services, but it doesn't contain required parameters + # 5. lastly, if middleManager is to be started, and it is missing task memory properties + if jvm_config_count > 0 or mm_task_property_present: if total_memory != "": raise ValueError("If jvm.config is given for services, memory argument shouldn't be specified") + if jvm_config_count == 0: + raise ValueError("druid.indexer.runner.javaOptsArray or druid.worker.capacity is present in " + "middleManager/runtime.properties, \n " + "add jvm.config for all other services") + if jvm_config_count != len(service_list): + raise ValueError("jvm.config file should be present for all services or none") for service in service_list: verify_service_config(service, config) - return jvm_config_count == 0 + # compute memory only when none of the specified services contains jvm.config, + # if middleManager is to be started it doesn't contain task memory properties + return jvm_config_count == 0 and mm_task_property_present is False def compute_system_memory(): @@ -476,13 +514,16 @@ def main(): service_memory_config = {} if should_compute_memory(config, total_memory, service_list): + # if memory is to be computed, _common directory should contain common.jvm.config + validate_common_jvm_args(config) memory_in_mega_bytes = convert_total_memory_string(total_memory) - print_if_verbose(f'total memory is {memory_in_mega_bytes}m') + print_if_verbose(f'Total memory is {memory_in_mega_bytes}m\n') memory_to_be_used = check_memory_constraint(memory_in_mega_bytes, service_list) - print_if_verbose(f'memory used for services & tasks {memory_to_be_used}m') + print_if_verbose(f'Memory used for services & tasks {memory_to_be_used}m\n') service_memory_config = distribute_memory(service_list, memory_to_be_used) else: - print_if_verbose('not computing memory distribution, reading memory specification from service jvm.config\n') + print_if_verbose('Not computing memory distribution, reading memory specification from service jvm.config & ' + 'middleManager/runtime.properties\n') script_arguments = build_supervise_script_arguments(service_list, service_memory_config, config, zk) @@ -491,6 +532,7 @@ def main(): os.execv('./supervise', script_arguments) + try: main() except (KeyboardInterrupt, ValueError) as error: From b8f01cc60caaa29423a864ad9ce2f8b61fc632b2 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Thu, 24 Nov 2022 14:06:26 +0530 Subject: [PATCH 039/141] remove start-druid bash script, rename start-druid-main --- examples/bin/start-druid | 526 ++++++++++++++++++++++++++++++++- examples/bin/start-druid-main | 540 ---------------------------------- 2 files changed, 521 insertions(+), 545 deletions(-) delete mode 100644 examples/bin/start-druid-main diff --git a/examples/bin/start-druid b/examples/bin/start-druid index 601df55ffef5..f5acbc5df6b9 100644 --- a/examples/bin/start-druid +++ b/examples/bin/start-druid @@ -1,4 +1,4 @@ -#!/bin/bash -eu +#!/usr/bin/env python3 # Licensed to the Apache Software Foundation (ASF) under one # or more contributor license agreements. See the NOTICE file @@ -17,8 +17,524 @@ # specific language governing permissions and limitations # under the License. -PWD="$(pwd)" -WHEREAMI="$(dirname "$0")" -WHEREAMI="$(cd "$WHEREAMI" && pwd)" +import sys +import os +import psutil +import pathlib +import multiprocessing +import argparse -exec "$WHEREAMI/start-druid-main" "$@" +QUICKSTART_ROOT_CONFIG_PATH = "conf/druid/single-server/quickstart" + +MEM_GB_SUFFIX = "g" +MEM_MB_SUFFIX = "m" +XMX_PARAMETER = "-Xmx" +XMS_PARAMETER = "-Xms" +DIRECT_MEM_PARAMETER = "-XX:MaxDirectMemorySize" +SERVICE_SEPARATOR = "," + +TASK_JAVA_OPTS_ARRAY = ["-server", "-Duser.timezone=UTC", "-Dfile.encoding=UTF-8", "-XX:+ExitOnOutOfMemoryError", + "-Djava.util.logging.manager=org.apache.logging.log4j.jul.LogManager"] +TASK_JAVA_OPTS_PROPERTY = "druid.indexer.runner.javaOptsArray" +TASK_WORKER_CAPACITY_PROPERTY = "druid.worker.capacity" +TASK_COUNT = "task-count" +TASK_MEM_TYPE_LOW = "low" +TASK_MEM_TYPE_HIGH = "high" +TASK_MEM_MAP = { + TASK_MEM_TYPE_LOW: ["-Xms256m", "-Xmx256m", "-XX:MaxDirectMemorySize=256g"], + TASK_MEM_TYPE_HIGH: ["-Xms1g", "-Xmx1g", "-XX:MaxDirectMemorySize=1g"] +} + +BROKER = "broker" +ROUTER = "router" +COORDINATOR = "coordinator-overlord" +HISTORICAL = "historical" +MIDDLE_MANAGER = "middleManager" +TASKS = "tasks" + +DEFAULT_SERVICES = [ + BROKER, + ROUTER, + COORDINATOR, + HISTORICAL, + MIDDLE_MANAGER +] + +SERVICE_MEMORY_RATIO = { + MIDDLE_MANAGER: 1, + ROUTER: 2, + COORDINATOR: 30, + BROKER: 46, + HISTORICAL: 80, + TASKS: 30 +} + +MINIMUM_MEMORY_MB = { + MIDDLE_MANAGER: 64, + ROUTER: 128, + TASKS: 1024, + BROKER: 900, + COORDINATOR: 256, + HISTORICAL: 900 +} + +HEAP_TO_TOTAL_MEM_RATIO = { + MIDDLE_MANAGER: 1, + ROUTER: 1, + COORDINATOR: 1, + BROKER: 0.60, + HISTORICAL: 0.40, + TASKS: 0.50 +} + +LOGGING_ENABLED = False + + +def print_if_verbose(message): + if LOGGING_ENABLED: + print(message) + + +def configure_parser(): + parser = argparse.ArgumentParser( + prog='Druid quickstart', + formatter_class=argparse.RawTextHelpFormatter, + epilog= + """ +sample usage: + start-druid + Start up all the services (including zk). + start-druid -m=100g + Start up all the services (including zk) + using a total memory of 100GB. + start-druid -m=100g --compute + Compute memory distribution and validate arguments. + start-druid -m=100g -s=broker,router + Starts a broker and a router, using a total memory of 100GB. + start-druid -m=100g --s=broker,router \\ + -c=conf/druid/single-server/custom + Starts a broker and a router, using a total memory of 100GB. + Reads configs for each service (jvm.config, runtime.properties) + from respective folders inside the given root config path. + start-druid -s=broker,router \\ + -c=conf/druid/single-server/custom + Starts a broker and a router service, reading service configs + from the given root directory. Calculates memory requirements for + each service, if required, using upto 80% of the total system memory. + start-druid -m=100g \\ + -s=broker,router \\ + -c=conf/druid/single-server/custom \\ + --zk + Starts broker, router and zookeeper. + zookeeper config is read from conf/zk. +""" + ) + parser.add_argument('--memory', '-m', type=str, required=False, + help='Total memory for all processes (services and tasks, if any). \n' + 'This parameter is ignored if each service already has a jvm.config \n' + 'in the given conf directory. e.g. 500m, 4g, 6g\n') + parser.add_argument('--services', '-s', type=str, required=False, + help='List of services to be started, subset of \n' + '{broker, router, middleManager, historical, coordinator-overlord}. \n' + 'If the argument is not given, all services \n' + 'and zookeeper is started. e.g. -sl=broker,historical') + parser.add_argument('--config', '-c', type=str, required=False, + help='Relative path to the directory containing common and service \n' + 'specific properties to be overridden. \n' + 'This directory must contain \'_common\' directory with \n' + '\'common.jvm.config\' & \'common.runtime.properties\' files. \n' + 'If this argument is not given, config from \n' + 'conf/druid/single-server/quickstart directory is used.\n') + parser.add_argument('--compute', action='store_true', + help='Does not start Druid, only displays the memory allocated \n' + 'to each service if started with the given total memory.\n') + parser.add_argument('--zk', '-zk', action='store_true', + help='Specification to run zookeeper, \n' + 'zk config is picked up from conf/zk.') + parser.add_argument('--verbose', action='store_true', help='Log details') + + parser.set_defaults(zk=False) + parser.set_defaults(compute=False) + parser.set_defaults(verbose=False) + + return parser + + +def validate_common_jvm_args(config): + if pathlib.Path(f'{config}/_common/common.jvm.config').is_file() is False: + raise ValueError(f'_common/common.jvm.config file is missing in the root config, ' + f'check {QUICKSTART_ROOT_CONFIG_PATH}/_common directory') + + +def validate_common_directory(config): + if pathlib.Path(f'{config}/_common').is_dir() is False: + raise ValueError( + f'_common directory is missing in the root config, check {QUICKSTART_ROOT_CONFIG_PATH}/_common directory') + + if pathlib.Path(f'{config}/_common/common.runtime.properties').is_file() is False: + raise ValueError(f'_common/common.runtime.properties file is missing in the root config, ' + f'check {QUICKSTART_ROOT_CONFIG_PATH}/_common directory') + + +def parse_arguments(args): + service_list = [] + config = "" + total_memory = "" + compute = False + zk = False + + if args.compute: + compute = True + if args.zk: + zk = True + if args.config is not None: + config = pathlib.Path(os.path.join(os.getcwd(), args.config)).resolve() + if os.path.exists(config) is False: + raise ValueError(f'config {config} not found') + validate_common_directory(config) + if args.memory is not None: + total_memory = args.memory + if args.services is not None: + services = args.services.split(SERVICE_SEPARATOR) + + for service in services: + if service not in DEFAULT_SERVICES: + raise ValueError(f'Invalid service name {service}, should be one of {DEFAULT_SERVICES}') + + if service in service_list: + raise ValueError(f'{service} is specified multiple times') + + service_list.append(service) + + if len(service_list) == 0: + # start all services + service_list = DEFAULT_SERVICES + zk = True + + return config, total_memory, service_list, zk, compute + + +def print_startup_config(service_list, config, zk): + print_if_verbose(f'Starting {service_list}') + print_if_verbose(f'Reading config from {config}') + if zk: + zk_config = pathlib.Path(f'{os.getcwd()}/../conf/zk').resolve() + print_if_verbose(f'Starting zk, reading default config from {zk_config}') + print_if_verbose('\n') + + +def middle_manager_task_memory_params_present(config): + java_opts_property_present = False + worker_capacity_property_present = False + + if pathlib.Path(f'{config}/middleManager/runtime.properties').is_file(): + with open(f'{config}/middleManager/runtime.properties') as file: + for line in file: + if line.startswith(TASK_JAVA_OPTS_PROPERTY): + java_opts_property_present = True + elif line.startswith(TASK_WORKER_CAPACITY_PROPERTY): + worker_capacity_property_present = True + + return java_opts_property_present, worker_capacity_property_present + + +def verify_service_config(service, config): + path = f'{config}/{service}/jvm.config' + + required_parameters = [XMX_PARAMETER, XMS_PARAMETER] + + if HEAP_TO_TOTAL_MEM_RATIO.get(service) != 1: + required_parameters.append(DIRECT_MEM_PARAMETER) + + with open(path) as file: + for line in file: + if line.startswith(XMX_PARAMETER) and XMX_PARAMETER in required_parameters: + required_parameters.remove(XMX_PARAMETER) + if line.startswith(XMS_PARAMETER) and XMS_PARAMETER in required_parameters: + required_parameters.remove(XMS_PARAMETER) + if line.startswith(DIRECT_MEM_PARAMETER) and DIRECT_MEM_PARAMETER in required_parameters: + required_parameters.remove(DIRECT_MEM_PARAMETER) + + if len(required_parameters) > 0: + params = ",".join(required_parameters) + raise ValueError(f'{params} missing in {service}/jvm.config') + + if service == MIDDLE_MANAGER: + if pathlib.Path(f'{config}/{service}/runtime.properties').is_file() is False: + raise ValueError(f'{service}/runtime.properties file is missing in the root config') + + mm_task_java_opts_property, mm_task_worker_capacity_prop = middle_manager_task_memory_params_present(config) + + if mm_task_java_opts_property is False: + raise ValueError(f'{TASK_JAVA_OPTS_PROPERTY} property missing in {service}/runtime.properties') + + +def should_compute_memory(config, total_memory, service_list): + """ + if memory argument is given, memory for services and tasks is computed, jvm.config file + or runtime.properties with task memory specification shouldn't be present + Alternatively, all memory related parameters are specified + which implies following should be present: + jvm.config file for all services with -Xmx=***, Xms=*** parameters + -XX:MaxDirectMemorySize=** in jvm.config for broker and historical + druid.indexer.runner.javaOptsArray (optionally druid.worker.capacity) in + rootDirectory/middleManager/runtime.properties + """ + + jvm_config_count = 0 + for service in service_list: + if pathlib.Path(f'{config}/{service}/jvm.config').is_file(): + jvm_config_count += 1 + + mm_task_property_present = False + if MIDDLE_MANAGER in service_list: + mm_task_java_opts_property, mm_task_worker_capacity_prop = middle_manager_task_memory_params_present(config) + mm_task_property_present = mm_task_java_opts_property or mm_task_worker_capacity_prop + + # possible error states + # 1. memory argument is specified, also jvm.config or middleManger/runtime.properties having + # druid.indexer.runner.javaOptsArray or druid.worker.capacity parameters is present + # 2. jvm.config is not present for any service, but middleManger/runtime.properties has + # druid.indexer.runner.javaOptsArray or druid.worker.capacity parameters + # 3. jvm.config present for some but not all services + # 4. jvm.config file is present for all services, but it doesn't contain required parameters + # 5. lastly, if middleManager is to be started, and it is missing task memory properties + if jvm_config_count > 0 or mm_task_property_present: + if total_memory != "": + raise ValueError("If jvm.config is given for services, memory argument shouldn't be specified") + if jvm_config_count == 0: + raise ValueError("druid.indexer.runner.javaOptsArray or druid.worker.capacity is present in " + "middleManager/runtime.properties, \n " + "add jvm.config for all other services") + if jvm_config_count != len(service_list): + raise ValueError("jvm.config file should be present for all services or none") + for service in service_list: + verify_service_config(service, config) + + # compute memory only when none of the specified services contains jvm.config, + # if middleManager is to be started it doesn't contain task memory properties + return jvm_config_count == 0 and mm_task_property_present is False + + +def compute_system_memory(): + system_memory = psutil.virtual_memory().total # mem in bytes + memory_for_druid = int(system_memory / (1024 * 1024)) + return memory_for_druid + + +def convert_total_memory_string(memory): + try: + if memory == "": + computed_memory = compute_system_memory() + return computed_memory + elif memory.endswith(MEM_MB_SUFFIX): + return int(memory[:-1]) + elif memory.endswith(MEM_GB_SUFFIX): + return 1024 * int(memory[:-1]) + else: + raise ValueError('Incorrect format for memory argument, expected format is ') + except Exception: + raise ValueError('Incorrect format for memory argument, expected format is ') + + +def check_memory_constraint(total_memory, service_list): + # 80% of total memory >= sum of lower bound service memory should be + lower_bound_memory = 0 + + for service in service_list: + lower_bound_memory += MINIMUM_MEMORY_MB.get(service) + + required_memory = int(lower_bound_memory / 0.8) + + if total_memory < required_memory: + raise ValueError(f'Minimum memory required for starting services is {required_memory}m') + + if total_memory >= 2 * lower_bound_memory: + return int(total_memory / 2) + else: + return lower_bound_memory + + +def build_mm_task_java_opts_array(memory_type): + task_memory = f'-D{TASK_JAVA_OPTS_PROPERTY}=[' + + mem_array = TASK_MEM_MAP.get(memory_type) + + java_opts_list = TASK_JAVA_OPTS_ARRAY + mem_array + + for item in java_opts_list: + task_memory += f'\"{item}\";' + + task_memory = task_memory[:-1] + task_memory += ']' + return task_memory + + +def compute_tasks_memory(allocated_memory): + if allocated_memory >= 2048: + task_count = int(allocated_memory / 2048) + memory_type = TASK_MEM_TYPE_HIGH + task_memory_mb = 2048 + else: + task_count = int(allocated_memory / 512) + memory_type = TASK_MEM_TYPE_LOW + task_memory_mb = 512 + task_count = min(task_count, multiprocessing.cpu_count()) + + return memory_type, task_count, task_memory_mb + + +def build_memory_config(service, allocated_memory): + if service == TASKS: + memory_type, task_count, task_memory = compute_tasks_memory(allocated_memory) + java_opts_array = build_mm_task_java_opts_array(memory_type) + return [f'-D{TASK_WORKER_CAPACITY_PROPERTY}={task_count}', java_opts_array], task_memory * task_count + else: + heap_memory = HEAP_TO_TOTAL_MEM_RATIO.get(service) * allocated_memory + direct_memory = int(allocated_memory - heap_memory) + heap_memory = int(heap_memory) + + if direct_memory == 0: + return f'-Xms{heap_memory}m -Xmx{heap_memory}m', allocated_memory + + return f'-Xms{heap_memory}m -Xmx{heap_memory}m -XX:MaxDirectMemorySize={direct_memory}m', allocated_memory + + +def distribute_memory(services, total_memory): + service_memory_config = {} + + memory_weight_sum = 0 + + service_list = services.copy() + if MIDDLE_MANAGER in services: + service_list.append(TASKS) + + for service in service_list: + memory_weight_sum += SERVICE_MEMORY_RATIO.get(service) + + multiplier = total_memory / memory_weight_sum + + lower_bound_memory_allocation = 0 + allocated_services = set() + + for service in service_list: + allocated_memory = SERVICE_MEMORY_RATIO.get(service) * multiplier + if service in MINIMUM_MEMORY_MB and allocated_memory < MINIMUM_MEMORY_MB.get(service): + allocated_memory = MINIMUM_MEMORY_MB.get(service) + service_memory_config[service], allocated_memory = build_memory_config(service, allocated_memory) + lower_bound_memory_allocation += allocated_memory + allocated_services.add(service) + + if lower_bound_memory_allocation > 0: + # compute the multiplier again for remaining services + memory_weight_sum = 0 + for service in service_list: + if service in allocated_services: + continue + memory_weight_sum += SERVICE_MEMORY_RATIO.get(service) + multiplier = (total_memory - lower_bound_memory_allocation) / memory_weight_sum + + for service in service_list: + if service in allocated_services: + continue + allocated_memory = SERVICE_MEMORY_RATIO.get(service) * multiplier + if service in MINIMUM_MEMORY_MB and allocated_memory < MINIMUM_MEMORY_MB.get(service): + allocated_memory = MINIMUM_MEMORY_MB.get(service) + + service_memory_config[service], allocated_memory = build_memory_config(service, allocated_memory) + + print_if_verbose(f'\nMemory distribution for services:') + for key, value in service_memory_config.items(): + print_if_verbose(f'{key}, {value}') + print_if_verbose('\n') + + return service_memory_config + + +def build_supervise_script_arguments(service_list, service_memory_config, config, zk): + args = [] + commands = [] + args.append('supervise') + args.append('-a') + + commands.append(":verify bin/verify-java") + commands.append(":verify bin/verify-default-ports") + commands.append(":notify bin/greet") + commands.append(":kill-timeout 10") + + if zk: + commands.append("!p10 zk bin/run-zk conf") + + for service in service_list: + jvm_args = service_memory_config.get(service) + + prefix = '' + if service == MIDDLE_MANAGER: + prefix = '!p90 ' + + if jvm_args is None: + commands.append(f'{prefix}{service} bin/run-druid {service} {config}') + else: + if service == MIDDLE_MANAGER: + task_config = service_memory_config.get(TASKS) + task_count = task_config[0] + task_memory = task_config[1] + commands.append( + f'{prefix}{service} bin/run-druid {service} {config} \'{jvm_args}\' \'{task_count} {task_memory}\'') + else: + commands.append(f'{prefix}{service} bin/run-druid {service} {config} \'{jvm_args}\'') + + print_if_verbose('Supervise script args:') + for item in commands: + print_if_verbose(item) + + print_if_verbose('\n') + + args.append(",".join(commands)) + return args + + +def main(): + parser = configure_parser() + args = parser.parse_args() + + global LOGGING_ENABLED + LOGGING_ENABLED = args.verbose or args.compute + + config, total_memory, service_list, zk, compute = parse_arguments(args) + + # change directory to bin + os.chdir(os.path.dirname(sys.argv[0])) + + if config == "": + config = pathlib.Path(f'{os.getcwd()}/../{QUICKSTART_ROOT_CONFIG_PATH}').resolve() + + print_startup_config(service_list, config, zk) + + service_memory_config = {} + + if should_compute_memory(config, total_memory, service_list): + # if memory is to be computed, _common directory should contain common.jvm.config + validate_common_jvm_args(config) + memory_in_mega_bytes = convert_total_memory_string(total_memory) + print_if_verbose(f'Total memory is {memory_in_mega_bytes}m\n') + memory_to_be_used = check_memory_constraint(memory_in_mega_bytes, service_list) + print_if_verbose(f'Memory used for services & tasks {memory_to_be_used}m\n') + service_memory_config = distribute_memory(service_list, memory_to_be_used) + else: + print_if_verbose('Not computing memory distribution, reading memory specification from service jvm.config & ' + 'middleManager/runtime.properties\n') + + script_arguments = build_supervise_script_arguments(service_list, service_memory_config, config, zk) + + if compute: + return + + os.execv('./supervise', script_arguments) + + +try: + main() +except (KeyboardInterrupt, ValueError) as error: + print(error) + sys.exit(1) diff --git a/examples/bin/start-druid-main b/examples/bin/start-druid-main deleted file mode 100644 index f5acbc5df6b9..000000000000 --- a/examples/bin/start-druid-main +++ /dev/null @@ -1,540 +0,0 @@ -#!/usr/bin/env python3 - -# Licensed to the Apache Software Foundation (ASF) under one -# or more contributor license agreements. See the NOTICE file -# distributed with this work for additional information -# regarding copyright ownership. The ASF licenses this file -# to you under the Apache License, Version 2.0 (the -# "License"); you may not use this file except in compliance -# with the License. You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, -# software distributed under the License is distributed on an -# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY -# KIND, either express or implied. See the License for the -# specific language governing permissions and limitations -# under the License. - -import sys -import os -import psutil -import pathlib -import multiprocessing -import argparse - -QUICKSTART_ROOT_CONFIG_PATH = "conf/druid/single-server/quickstart" - -MEM_GB_SUFFIX = "g" -MEM_MB_SUFFIX = "m" -XMX_PARAMETER = "-Xmx" -XMS_PARAMETER = "-Xms" -DIRECT_MEM_PARAMETER = "-XX:MaxDirectMemorySize" -SERVICE_SEPARATOR = "," - -TASK_JAVA_OPTS_ARRAY = ["-server", "-Duser.timezone=UTC", "-Dfile.encoding=UTF-8", "-XX:+ExitOnOutOfMemoryError", - "-Djava.util.logging.manager=org.apache.logging.log4j.jul.LogManager"] -TASK_JAVA_OPTS_PROPERTY = "druid.indexer.runner.javaOptsArray" -TASK_WORKER_CAPACITY_PROPERTY = "druid.worker.capacity" -TASK_COUNT = "task-count" -TASK_MEM_TYPE_LOW = "low" -TASK_MEM_TYPE_HIGH = "high" -TASK_MEM_MAP = { - TASK_MEM_TYPE_LOW: ["-Xms256m", "-Xmx256m", "-XX:MaxDirectMemorySize=256g"], - TASK_MEM_TYPE_HIGH: ["-Xms1g", "-Xmx1g", "-XX:MaxDirectMemorySize=1g"] -} - -BROKER = "broker" -ROUTER = "router" -COORDINATOR = "coordinator-overlord" -HISTORICAL = "historical" -MIDDLE_MANAGER = "middleManager" -TASKS = "tasks" - -DEFAULT_SERVICES = [ - BROKER, - ROUTER, - COORDINATOR, - HISTORICAL, - MIDDLE_MANAGER -] - -SERVICE_MEMORY_RATIO = { - MIDDLE_MANAGER: 1, - ROUTER: 2, - COORDINATOR: 30, - BROKER: 46, - HISTORICAL: 80, - TASKS: 30 -} - -MINIMUM_MEMORY_MB = { - MIDDLE_MANAGER: 64, - ROUTER: 128, - TASKS: 1024, - BROKER: 900, - COORDINATOR: 256, - HISTORICAL: 900 -} - -HEAP_TO_TOTAL_MEM_RATIO = { - MIDDLE_MANAGER: 1, - ROUTER: 1, - COORDINATOR: 1, - BROKER: 0.60, - HISTORICAL: 0.40, - TASKS: 0.50 -} - -LOGGING_ENABLED = False - - -def print_if_verbose(message): - if LOGGING_ENABLED: - print(message) - - -def configure_parser(): - parser = argparse.ArgumentParser( - prog='Druid quickstart', - formatter_class=argparse.RawTextHelpFormatter, - epilog= - """ -sample usage: - start-druid - Start up all the services (including zk). - start-druid -m=100g - Start up all the services (including zk) - using a total memory of 100GB. - start-druid -m=100g --compute - Compute memory distribution and validate arguments. - start-druid -m=100g -s=broker,router - Starts a broker and a router, using a total memory of 100GB. - start-druid -m=100g --s=broker,router \\ - -c=conf/druid/single-server/custom - Starts a broker and a router, using a total memory of 100GB. - Reads configs for each service (jvm.config, runtime.properties) - from respective folders inside the given root config path. - start-druid -s=broker,router \\ - -c=conf/druid/single-server/custom - Starts a broker and a router service, reading service configs - from the given root directory. Calculates memory requirements for - each service, if required, using upto 80% of the total system memory. - start-druid -m=100g \\ - -s=broker,router \\ - -c=conf/druid/single-server/custom \\ - --zk - Starts broker, router and zookeeper. - zookeeper config is read from conf/zk. -""" - ) - parser.add_argument('--memory', '-m', type=str, required=False, - help='Total memory for all processes (services and tasks, if any). \n' - 'This parameter is ignored if each service already has a jvm.config \n' - 'in the given conf directory. e.g. 500m, 4g, 6g\n') - parser.add_argument('--services', '-s', type=str, required=False, - help='List of services to be started, subset of \n' - '{broker, router, middleManager, historical, coordinator-overlord}. \n' - 'If the argument is not given, all services \n' - 'and zookeeper is started. e.g. -sl=broker,historical') - parser.add_argument('--config', '-c', type=str, required=False, - help='Relative path to the directory containing common and service \n' - 'specific properties to be overridden. \n' - 'This directory must contain \'_common\' directory with \n' - '\'common.jvm.config\' & \'common.runtime.properties\' files. \n' - 'If this argument is not given, config from \n' - 'conf/druid/single-server/quickstart directory is used.\n') - parser.add_argument('--compute', action='store_true', - help='Does not start Druid, only displays the memory allocated \n' - 'to each service if started with the given total memory.\n') - parser.add_argument('--zk', '-zk', action='store_true', - help='Specification to run zookeeper, \n' - 'zk config is picked up from conf/zk.') - parser.add_argument('--verbose', action='store_true', help='Log details') - - parser.set_defaults(zk=False) - parser.set_defaults(compute=False) - parser.set_defaults(verbose=False) - - return parser - - -def validate_common_jvm_args(config): - if pathlib.Path(f'{config}/_common/common.jvm.config').is_file() is False: - raise ValueError(f'_common/common.jvm.config file is missing in the root config, ' - f'check {QUICKSTART_ROOT_CONFIG_PATH}/_common directory') - - -def validate_common_directory(config): - if pathlib.Path(f'{config}/_common').is_dir() is False: - raise ValueError( - f'_common directory is missing in the root config, check {QUICKSTART_ROOT_CONFIG_PATH}/_common directory') - - if pathlib.Path(f'{config}/_common/common.runtime.properties').is_file() is False: - raise ValueError(f'_common/common.runtime.properties file is missing in the root config, ' - f'check {QUICKSTART_ROOT_CONFIG_PATH}/_common directory') - - -def parse_arguments(args): - service_list = [] - config = "" - total_memory = "" - compute = False - zk = False - - if args.compute: - compute = True - if args.zk: - zk = True - if args.config is not None: - config = pathlib.Path(os.path.join(os.getcwd(), args.config)).resolve() - if os.path.exists(config) is False: - raise ValueError(f'config {config} not found') - validate_common_directory(config) - if args.memory is not None: - total_memory = args.memory - if args.services is not None: - services = args.services.split(SERVICE_SEPARATOR) - - for service in services: - if service not in DEFAULT_SERVICES: - raise ValueError(f'Invalid service name {service}, should be one of {DEFAULT_SERVICES}') - - if service in service_list: - raise ValueError(f'{service} is specified multiple times') - - service_list.append(service) - - if len(service_list) == 0: - # start all services - service_list = DEFAULT_SERVICES - zk = True - - return config, total_memory, service_list, zk, compute - - -def print_startup_config(service_list, config, zk): - print_if_verbose(f'Starting {service_list}') - print_if_verbose(f'Reading config from {config}') - if zk: - zk_config = pathlib.Path(f'{os.getcwd()}/../conf/zk').resolve() - print_if_verbose(f'Starting zk, reading default config from {zk_config}') - print_if_verbose('\n') - - -def middle_manager_task_memory_params_present(config): - java_opts_property_present = False - worker_capacity_property_present = False - - if pathlib.Path(f'{config}/middleManager/runtime.properties').is_file(): - with open(f'{config}/middleManager/runtime.properties') as file: - for line in file: - if line.startswith(TASK_JAVA_OPTS_PROPERTY): - java_opts_property_present = True - elif line.startswith(TASK_WORKER_CAPACITY_PROPERTY): - worker_capacity_property_present = True - - return java_opts_property_present, worker_capacity_property_present - - -def verify_service_config(service, config): - path = f'{config}/{service}/jvm.config' - - required_parameters = [XMX_PARAMETER, XMS_PARAMETER] - - if HEAP_TO_TOTAL_MEM_RATIO.get(service) != 1: - required_parameters.append(DIRECT_MEM_PARAMETER) - - with open(path) as file: - for line in file: - if line.startswith(XMX_PARAMETER) and XMX_PARAMETER in required_parameters: - required_parameters.remove(XMX_PARAMETER) - if line.startswith(XMS_PARAMETER) and XMS_PARAMETER in required_parameters: - required_parameters.remove(XMS_PARAMETER) - if line.startswith(DIRECT_MEM_PARAMETER) and DIRECT_MEM_PARAMETER in required_parameters: - required_parameters.remove(DIRECT_MEM_PARAMETER) - - if len(required_parameters) > 0: - params = ",".join(required_parameters) - raise ValueError(f'{params} missing in {service}/jvm.config') - - if service == MIDDLE_MANAGER: - if pathlib.Path(f'{config}/{service}/runtime.properties').is_file() is False: - raise ValueError(f'{service}/runtime.properties file is missing in the root config') - - mm_task_java_opts_property, mm_task_worker_capacity_prop = middle_manager_task_memory_params_present(config) - - if mm_task_java_opts_property is False: - raise ValueError(f'{TASK_JAVA_OPTS_PROPERTY} property missing in {service}/runtime.properties') - - -def should_compute_memory(config, total_memory, service_list): - """ - if memory argument is given, memory for services and tasks is computed, jvm.config file - or runtime.properties with task memory specification shouldn't be present - Alternatively, all memory related parameters are specified - which implies following should be present: - jvm.config file for all services with -Xmx=***, Xms=*** parameters - -XX:MaxDirectMemorySize=** in jvm.config for broker and historical - druid.indexer.runner.javaOptsArray (optionally druid.worker.capacity) in - rootDirectory/middleManager/runtime.properties - """ - - jvm_config_count = 0 - for service in service_list: - if pathlib.Path(f'{config}/{service}/jvm.config').is_file(): - jvm_config_count += 1 - - mm_task_property_present = False - if MIDDLE_MANAGER in service_list: - mm_task_java_opts_property, mm_task_worker_capacity_prop = middle_manager_task_memory_params_present(config) - mm_task_property_present = mm_task_java_opts_property or mm_task_worker_capacity_prop - - # possible error states - # 1. memory argument is specified, also jvm.config or middleManger/runtime.properties having - # druid.indexer.runner.javaOptsArray or druid.worker.capacity parameters is present - # 2. jvm.config is not present for any service, but middleManger/runtime.properties has - # druid.indexer.runner.javaOptsArray or druid.worker.capacity parameters - # 3. jvm.config present for some but not all services - # 4. jvm.config file is present for all services, but it doesn't contain required parameters - # 5. lastly, if middleManager is to be started, and it is missing task memory properties - if jvm_config_count > 0 or mm_task_property_present: - if total_memory != "": - raise ValueError("If jvm.config is given for services, memory argument shouldn't be specified") - if jvm_config_count == 0: - raise ValueError("druid.indexer.runner.javaOptsArray or druid.worker.capacity is present in " - "middleManager/runtime.properties, \n " - "add jvm.config for all other services") - if jvm_config_count != len(service_list): - raise ValueError("jvm.config file should be present for all services or none") - for service in service_list: - verify_service_config(service, config) - - # compute memory only when none of the specified services contains jvm.config, - # if middleManager is to be started it doesn't contain task memory properties - return jvm_config_count == 0 and mm_task_property_present is False - - -def compute_system_memory(): - system_memory = psutil.virtual_memory().total # mem in bytes - memory_for_druid = int(system_memory / (1024 * 1024)) - return memory_for_druid - - -def convert_total_memory_string(memory): - try: - if memory == "": - computed_memory = compute_system_memory() - return computed_memory - elif memory.endswith(MEM_MB_SUFFIX): - return int(memory[:-1]) - elif memory.endswith(MEM_GB_SUFFIX): - return 1024 * int(memory[:-1]) - else: - raise ValueError('Incorrect format for memory argument, expected format is ') - except Exception: - raise ValueError('Incorrect format for memory argument, expected format is ') - - -def check_memory_constraint(total_memory, service_list): - # 80% of total memory >= sum of lower bound service memory should be - lower_bound_memory = 0 - - for service in service_list: - lower_bound_memory += MINIMUM_MEMORY_MB.get(service) - - required_memory = int(lower_bound_memory / 0.8) - - if total_memory < required_memory: - raise ValueError(f'Minimum memory required for starting services is {required_memory}m') - - if total_memory >= 2 * lower_bound_memory: - return int(total_memory / 2) - else: - return lower_bound_memory - - -def build_mm_task_java_opts_array(memory_type): - task_memory = f'-D{TASK_JAVA_OPTS_PROPERTY}=[' - - mem_array = TASK_MEM_MAP.get(memory_type) - - java_opts_list = TASK_JAVA_OPTS_ARRAY + mem_array - - for item in java_opts_list: - task_memory += f'\"{item}\";' - - task_memory = task_memory[:-1] - task_memory += ']' - return task_memory - - -def compute_tasks_memory(allocated_memory): - if allocated_memory >= 2048: - task_count = int(allocated_memory / 2048) - memory_type = TASK_MEM_TYPE_HIGH - task_memory_mb = 2048 - else: - task_count = int(allocated_memory / 512) - memory_type = TASK_MEM_TYPE_LOW - task_memory_mb = 512 - task_count = min(task_count, multiprocessing.cpu_count()) - - return memory_type, task_count, task_memory_mb - - -def build_memory_config(service, allocated_memory): - if service == TASKS: - memory_type, task_count, task_memory = compute_tasks_memory(allocated_memory) - java_opts_array = build_mm_task_java_opts_array(memory_type) - return [f'-D{TASK_WORKER_CAPACITY_PROPERTY}={task_count}', java_opts_array], task_memory * task_count - else: - heap_memory = HEAP_TO_TOTAL_MEM_RATIO.get(service) * allocated_memory - direct_memory = int(allocated_memory - heap_memory) - heap_memory = int(heap_memory) - - if direct_memory == 0: - return f'-Xms{heap_memory}m -Xmx{heap_memory}m', allocated_memory - - return f'-Xms{heap_memory}m -Xmx{heap_memory}m -XX:MaxDirectMemorySize={direct_memory}m', allocated_memory - - -def distribute_memory(services, total_memory): - service_memory_config = {} - - memory_weight_sum = 0 - - service_list = services.copy() - if MIDDLE_MANAGER in services: - service_list.append(TASKS) - - for service in service_list: - memory_weight_sum += SERVICE_MEMORY_RATIO.get(service) - - multiplier = total_memory / memory_weight_sum - - lower_bound_memory_allocation = 0 - allocated_services = set() - - for service in service_list: - allocated_memory = SERVICE_MEMORY_RATIO.get(service) * multiplier - if service in MINIMUM_MEMORY_MB and allocated_memory < MINIMUM_MEMORY_MB.get(service): - allocated_memory = MINIMUM_MEMORY_MB.get(service) - service_memory_config[service], allocated_memory = build_memory_config(service, allocated_memory) - lower_bound_memory_allocation += allocated_memory - allocated_services.add(service) - - if lower_bound_memory_allocation > 0: - # compute the multiplier again for remaining services - memory_weight_sum = 0 - for service in service_list: - if service in allocated_services: - continue - memory_weight_sum += SERVICE_MEMORY_RATIO.get(service) - multiplier = (total_memory - lower_bound_memory_allocation) / memory_weight_sum - - for service in service_list: - if service in allocated_services: - continue - allocated_memory = SERVICE_MEMORY_RATIO.get(service) * multiplier - if service in MINIMUM_MEMORY_MB and allocated_memory < MINIMUM_MEMORY_MB.get(service): - allocated_memory = MINIMUM_MEMORY_MB.get(service) - - service_memory_config[service], allocated_memory = build_memory_config(service, allocated_memory) - - print_if_verbose(f'\nMemory distribution for services:') - for key, value in service_memory_config.items(): - print_if_verbose(f'{key}, {value}') - print_if_verbose('\n') - - return service_memory_config - - -def build_supervise_script_arguments(service_list, service_memory_config, config, zk): - args = [] - commands = [] - args.append('supervise') - args.append('-a') - - commands.append(":verify bin/verify-java") - commands.append(":verify bin/verify-default-ports") - commands.append(":notify bin/greet") - commands.append(":kill-timeout 10") - - if zk: - commands.append("!p10 zk bin/run-zk conf") - - for service in service_list: - jvm_args = service_memory_config.get(service) - - prefix = '' - if service == MIDDLE_MANAGER: - prefix = '!p90 ' - - if jvm_args is None: - commands.append(f'{prefix}{service} bin/run-druid {service} {config}') - else: - if service == MIDDLE_MANAGER: - task_config = service_memory_config.get(TASKS) - task_count = task_config[0] - task_memory = task_config[1] - commands.append( - f'{prefix}{service} bin/run-druid {service} {config} \'{jvm_args}\' \'{task_count} {task_memory}\'') - else: - commands.append(f'{prefix}{service} bin/run-druid {service} {config} \'{jvm_args}\'') - - print_if_verbose('Supervise script args:') - for item in commands: - print_if_verbose(item) - - print_if_verbose('\n') - - args.append(",".join(commands)) - return args - - -def main(): - parser = configure_parser() - args = parser.parse_args() - - global LOGGING_ENABLED - LOGGING_ENABLED = args.verbose or args.compute - - config, total_memory, service_list, zk, compute = parse_arguments(args) - - # change directory to bin - os.chdir(os.path.dirname(sys.argv[0])) - - if config == "": - config = pathlib.Path(f'{os.getcwd()}/../{QUICKSTART_ROOT_CONFIG_PATH}').resolve() - - print_startup_config(service_list, config, zk) - - service_memory_config = {} - - if should_compute_memory(config, total_memory, service_list): - # if memory is to be computed, _common directory should contain common.jvm.config - validate_common_jvm_args(config) - memory_in_mega_bytes = convert_total_memory_string(total_memory) - print_if_verbose(f'Total memory is {memory_in_mega_bytes}m\n') - memory_to_be_used = check_memory_constraint(memory_in_mega_bytes, service_list) - print_if_verbose(f'Memory used for services & tasks {memory_to_be_used}m\n') - service_memory_config = distribute_memory(service_list, memory_to_be_used) - else: - print_if_verbose('Not computing memory distribution, reading memory specification from service jvm.config & ' - 'middleManager/runtime.properties\n') - - script_arguments = build_supervise_script_arguments(service_list, service_memory_config, config, zk) - - if compute: - return - - os.execv('./supervise', script_arguments) - - -try: - main() -except (KeyboardInterrupt, ValueError) as error: - print(error) - sys.exit(1) From 6e6305f5ec165e0dc26bcb1312c7ff7f7b7be80f Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Thu, 24 Nov 2022 14:56:34 +0530 Subject: [PATCH 040/141] Include tasks in lower bound memory calculation --- examples/bin/start-druid | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) diff --git a/examples/bin/start-druid b/examples/bin/start-druid index f5acbc5df6b9..07f3fdee673e 100644 --- a/examples/bin/start-druid +++ b/examples/bin/start-druid @@ -191,7 +191,6 @@ def parse_arguments(args): config = pathlib.Path(os.path.join(os.getcwd(), args.config)).resolve() if os.path.exists(config) is False: raise ValueError(f'config {config} not found') - validate_common_directory(config) if args.memory is not None: total_memory = args.memory if args.services is not None: @@ -337,10 +336,14 @@ def convert_total_memory_string(memory): raise ValueError('Incorrect format for memory argument, expected format is ') -def check_memory_constraint(total_memory, service_list): +def check_memory_constraint(total_memory, services): # 80% of total memory >= sum of lower bound service memory should be lower_bound_memory = 0 + service_list = services.copy() + if MIDDLE_MANAGER in services: + service_list.append(TASKS) + for service in service_list: lower_bound_memory += MINIMUM_MEMORY_MB.get(service) @@ -509,6 +512,8 @@ def main(): if config == "": config = pathlib.Path(f'{os.getcwd()}/../{QUICKSTART_ROOT_CONFIG_PATH}').resolve() + validate_common_directory(config) + print_startup_config(service_list, config, zk) service_memory_config = {} From a14af0f3cc19bef189b977d548b58eef35419125 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Thu, 24 Nov 2022 15:02:21 +0530 Subject: [PATCH 041/141] Fix test --- .../test/java/org/apache/druid/guice/StorageNodeModuleTest.java | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/server/src/test/java/org/apache/druid/guice/StorageNodeModuleTest.java b/server/src/test/java/org/apache/druid/guice/StorageNodeModuleTest.java index e849bb6910e8..082361a1c885 100644 --- a/server/src/test/java/org/apache/druid/guice/StorageNodeModuleTest.java +++ b/server/src/test/java/org/apache/druid/guice/StorageNodeModuleTest.java @@ -112,7 +112,7 @@ public void getDataNodeServiceWithNoServerTypeConfigShouldThrowProvisionExceptio public void getDataNodeServiceWithNoSegmentCacheConfiguredThrowProvisionException() { exceptionRule.expect(ProvisionException.class); - exceptionRule.expectMessage("Segment cache locations must be set on historicals."); + exceptionRule.expectMessage("druid.segmentCache.locations must be set on historicals."); Mockito.doReturn(ServerType.HISTORICAL).when(serverTypeConfig).getServerType(); mockSegmentCacheNotConfigured(); injector.getInstance(DataNodeService.class); From b091b32f21f22cdb8daa48de06d18116d76f6780 Mon Sep 17 00:00:00 2001 From: Tejaswini Bandlamudi <96047043+tejaswini-imply@users.noreply.github.com> Date: Fri, 25 Nov 2022 19:22:48 +0530 Subject: [PATCH 042/141] Fixes reindexing bug with filter on long column (#13386) * fixes BlockLayoutColumnarLongs close method to nullify internal buffer. * fixes other BlockLayoutColumnar supplier close methods to nullify internal buffers. * fix spotbugs --- .../input/DruidSegmentInputFormatTest.java | 3 +- .../input/DruidSegmentReaderTest.java | 355 +++++++++++------- .../BlockLayoutColumnarDoublesSupplier.java | 6 + .../BlockLayoutColumnarFloatsSupplier.java | 6 + .../BlockLayoutColumnarLongsSupplier.java | 8 + 5 files changed, 245 insertions(+), 133 deletions(-) diff --git a/indexing-service/src/test/java/org/apache/druid/indexing/input/DruidSegmentInputFormatTest.java b/indexing-service/src/test/java/org/apache/druid/indexing/input/DruidSegmentInputFormatTest.java index 910371a56699..e4faea1c069c 100644 --- a/indexing-service/src/test/java/org/apache/druid/indexing/input/DruidSegmentInputFormatTest.java +++ b/indexing-service/src/test/java/org/apache/druid/indexing/input/DruidSegmentInputFormatTest.java @@ -19,6 +19,7 @@ package org.apache.druid.indexing.input; +import com.google.common.collect.ImmutableList; import org.apache.druid.data.input.ColumnsFilter; import org.apache.druid.data.input.InputEntityReader; import org.apache.druid.data.input.InputRowSchema; @@ -50,7 +51,7 @@ public void testDruidSegmentInputEntityReader() DruidSegmentInputFormat format = new DruidSegmentInputFormat(null, null); InputEntityReader reader = format.createReader( INPUT_ROW_SCHEMA, - DruidSegmentReaderTest.makeInputEntity(Intervals.of("2000/P1D"), null), + DruidSegmentReaderTest.makeInputEntity(Intervals.of("2000/P1D"), null, ImmutableList.of("s", "d"), ImmutableList.of("cnt", "met_s")), null ); Assert.assertTrue(reader instanceof DruidSegmentReader); diff --git a/indexing-service/src/test/java/org/apache/druid/indexing/input/DruidSegmentReaderTest.java b/indexing-service/src/test/java/org/apache/druid/indexing/input/DruidSegmentReaderTest.java index 1638e79e9a70..ff4b50cee481 100644 --- a/indexing-service/src/test/java/org/apache/druid/indexing/input/DruidSegmentReaderTest.java +++ b/indexing-service/src/test/java/org/apache/druid/indexing/input/DruidSegmentReaderTest.java @@ -31,6 +31,7 @@ import org.apache.druid.data.input.impl.DimensionsSpec; import org.apache.druid.data.input.impl.DoubleDimensionSchema; import org.apache.druid.data.input.impl.FileEntity; +import org.apache.druid.data.input.impl.LongDimensionSchema; import org.apache.druid.data.input.impl.StringDimensionSchema; import org.apache.druid.data.input.impl.TimestampSpec; import org.apache.druid.hll.HyperLogLogCollector; @@ -41,8 +42,11 @@ import org.apache.druid.java.util.common.guava.BaseSequence.IteratorMaker; import org.apache.druid.java.util.common.guava.Sequence; import org.apache.druid.java.util.common.parsers.CloseableIterator; +import org.apache.druid.query.aggregation.AggregatorFactory; import org.apache.druid.query.aggregation.CountAggregatorFactory; import org.apache.druid.query.aggregation.hyperloglog.HyperUniquesAggregatorFactory; +import org.apache.druid.query.filter.NotDimFilter; +import org.apache.druid.query.filter.OrDimFilter; import org.apache.druid.query.filter.SelectorDimFilter; import org.apache.druid.segment.IndexBuilder; import org.apache.druid.segment.IndexIO; @@ -80,67 +84,44 @@ public class DruidSegmentReaderTest extends NullHandlingTest private File segmentDirectory; private final IndexIO indexIO = TestHelper.getTestIndexIO(); + private DimensionsSpec dimensionsSpec; + private List metrics; + private List rows; @Before public void setUp() throws IOException { // Write a segment with two rows in it, with columns: s (string), d (double), cnt (long), met_s (complex). - final IncrementalIndex incrementalIndex = - IndexBuilder.create() - .schema( - new IncrementalIndexSchema.Builder() - .withDimensionsSpec( - new DimensionsSpec( - ImmutableList.of( - StringDimensionSchema.create("s"), - new DoubleDimensionSchema("d") - ) - ) - ) - .withMetrics( - new CountAggregatorFactory("cnt"), - new HyperUniquesAggregatorFactory("met_s", "s") - ) - .withRollup(false) - .build() - ) - .rows( - ImmutableList.of( - new MapBasedInputRow( - DateTimes.of("2000"), - ImmutableList.of("s", "d"), - ImmutableMap.builder() - .put("s", "foo") - .put("d", 1.23) - .build() - ), - new MapBasedInputRow( - DateTimes.of("2000T01"), - ImmutableList.of("s", "d"), - ImmutableMap.builder() - .put("s", "bar") - .put("d", 4.56) - .build() - ) - ) - ) - .buildIncrementalIndex(); - - segmentDirectory = temporaryFolder.newFolder(); + dimensionsSpec = new DimensionsSpec( + ImmutableList.of( + StringDimensionSchema.create("strCol"), + new DoubleDimensionSchema("dblCol") + ) + ); + metrics = ImmutableList.of( + new CountAggregatorFactory("cnt"), + new HyperUniquesAggregatorFactory("met_s", "strCol") + ); + rows = ImmutableList.of( + new MapBasedInputRow( + DateTimes.of("2000"), + ImmutableList.of("strCol", "dblCol"), + ImmutableMap.builder() + .put("strCol", "foo") + .put("dblCol", 1.23) + .build() + ), + new MapBasedInputRow( + DateTimes.of("2000T01"), + ImmutableList.of("strCol", "dblCol"), + ImmutableMap.builder() + .put("strCol", "bar") + .put("dblCol", 4.56) + .build() + ) + ); - try { - TestHelper.getTestIndexMergerV9( - OnHeapMemorySegmentWriteOutMediumFactory.instance() - ).persist( - incrementalIndex, - segmentDirectory, - new IndexSpec(), - null - ); - } - finally { - incrementalIndex.close(); - } + createTestSetup(); } @Test @@ -152,8 +133,8 @@ public void testReader() throws IOException new TimestampSpec("__time", "millis", DateTimes.of("1971")), new DimensionsSpec( ImmutableList.of( - StringDimensionSchema.create("s"), - new DoubleDimensionSchema("d") + StringDimensionSchema.create("strCol"), + new DoubleDimensionSchema("dblCol") ) ), ColumnsFilter.all(), @@ -165,22 +146,22 @@ public void testReader() throws IOException ImmutableList.of( new MapBasedInputRow( DateTimes.of("2000"), - ImmutableList.of("s", "d"), + ImmutableList.of("strCol", "dblCol"), ImmutableMap.builder() .put("__time", DateTimes.of("2000T").getMillis()) - .put("s", "foo") - .put("d", 1.23d) + .put("strCol", "foo") + .put("dblCol", 1.23d) .put("cnt", 1L) .put("met_s", makeHLLC("foo")) .build() ), new MapBasedInputRow( DateTimes.of("2000T01"), - ImmutableList.of("s", "d"), + ImmutableList.of("strCol", "dblCol"), ImmutableMap.builder() .put("__time", DateTimes.of("2000T01").getMillis()) - .put("s", "bar") - .put("d", 4.56d) + .put("strCol", "bar") + .put("dblCol", 4.56d) .put("cnt", 1L) .put("met_s", makeHLLC("bar")) .build() @@ -190,6 +171,74 @@ public void testReader() throws IOException ); } + @Test + public void testReaderWhenFilteringOnLongColumn() throws IOException + { + dimensionsSpec = new DimensionsSpec( + ImmutableList.of( + new LongDimensionSchema("longCol"), + StringDimensionSchema.create("a"), + StringDimensionSchema.create("b") + ) + ); + metrics = ImmutableList.of(); + + List columnNames = ImmutableList.of("longCol", "a", "b"); + rows = ImmutableList.of( + new MapBasedInputRow( + DateTimes.utc(1667115726217L), + columnNames, + ImmutableMap.builder() + .put("__time", 1667115726217L) + .put("longCol", 0L) + .put("a", "foo1") + .put("b", "bar1") + .build() + ), + new MapBasedInputRow( + DateTimes.utc(1667115726224L), + columnNames, + ImmutableMap.builder() + .put("__time", 1667115726224L) + .put("longCol", 0L) + .put("a", "foo2") + .put("b", "bar2") + .build() + ), + new MapBasedInputRow( + DateTimes.utc(1667115726128L), + columnNames, + ImmutableMap.builder() + .put("__time", 1667115726128L) + .put("longCol", 5L) + .put("a", "foo3") + .put("b", "bar3") + .build() + ) + ); + + createTestSetup(); + + final DruidSegmentReader reader = new DruidSegmentReader( + makeInputEntityWithParams(Intervals.of("2022-10-30/2022-10-31"), columnNames, null), + indexIO, + new TimestampSpec("__time", "iso", null), + dimensionsSpec, + ColumnsFilter.all(), + new OrDimFilter( + new SelectorDimFilter("longCol", "5", null), + new NotDimFilter(new SelectorDimFilter("a", "foo1", null)), + new NotDimFilter(new SelectorDimFilter("b", "bar1", null)) + ), + temporaryFolder.newFolder() + ); + + List expectedRows = new ArrayList<>(); + expectedRows.add(rows.get(2)); + expectedRows.add(rows.get(1)); + Assert.assertEquals(expectedRows, readRows(reader)); + } + @Test public void testDruidTombstoneSegmentReader() throws IOException { @@ -235,8 +284,8 @@ public void testReaderAutoTimestampFormat() throws IOException new TimestampSpec("__time", "auto", DateTimes.of("1971")), new DimensionsSpec( ImmutableList.of( - StringDimensionSchema.create("s"), - new DoubleDimensionSchema("d") + StringDimensionSchema.create("strCol"), + new DoubleDimensionSchema("dblCol") ) ), ColumnsFilter.all(), @@ -248,22 +297,22 @@ public void testReaderAutoTimestampFormat() throws IOException ImmutableList.of( new MapBasedInputRow( DateTimes.of("2000"), - ImmutableList.of("s", "d"), + ImmutableList.of("strCol", "dblCol"), ImmutableMap.builder() .put("__time", DateTimes.of("2000T").getMillis()) - .put("s", "foo") - .put("d", 1.23d) + .put("strCol", "foo") + .put("dblCol", 1.23d) .put("cnt", 1L) .put("met_s", makeHLLC("foo")) .build() ), new MapBasedInputRow( DateTimes.of("2000T01"), - ImmutableList.of("s", "d"), + ImmutableList.of("strCol", "dblCol"), ImmutableMap.builder() .put("__time", DateTimes.of("2000T01").getMillis()) - .put("s", "bar") - .put("d", 4.56d) + .put("strCol", "bar") + .put("dblCol", 4.56d) .put("cnt", 1L) .put("met_s", makeHLLC("bar")) .build() @@ -280,7 +329,7 @@ public void testReaderWithDimensionExclusions() throws IOException makeInputEntity(Intervals.of("2000/P1D")), indexIO, new TimestampSpec("__time", "millis", DateTimes.of("1971")), - DimensionsSpec.builder().setDimensionExclusions(ImmutableList.of("__time", "s", "cnt", "met_s")).build(), + DimensionsSpec.builder().setDimensionExclusions(ImmutableList.of("__time", "strCol", "cnt", "met_s")).build(), ColumnsFilter.all(), null, temporaryFolder.newFolder() @@ -290,22 +339,22 @@ public void testReaderWithDimensionExclusions() throws IOException ImmutableList.of( new MapBasedInputRow( DateTimes.of("2000"), - ImmutableList.of("d"), + ImmutableList.of("dblCol"), ImmutableMap.builder() .put("__time", DateTimes.of("2000T").getMillis()) - .put("s", "foo") - .put("d", 1.23d) + .put("strCol", "foo") + .put("dblCol", 1.23d) .put("cnt", 1L) .put("met_s", makeHLLC("foo")) .build() ), new MapBasedInputRow( DateTimes.of("2000T01"), - ImmutableList.of("d"), + ImmutableList.of("dblCol"), ImmutableMap.builder() .put("__time", DateTimes.of("2000T01").getMillis()) - .put("s", "bar") - .put("d", 4.56d) + .put("strCol", "bar") + .put("dblCol", 4.56d) .put("cnt", 1L) .put("met_s", makeHLLC("bar")) .build() @@ -324,11 +373,11 @@ public void testReaderWithInclusiveColumnsFilter() throws IOException new TimestampSpec("__time", "millis", DateTimes.of("1971")), new DimensionsSpec( ImmutableList.of( - StringDimensionSchema.create("s"), - new DoubleDimensionSchema("d") + StringDimensionSchema.create("strCol"), + new DoubleDimensionSchema("dblCol") ) ), - ColumnsFilter.inclusionBased(ImmutableSet.of("__time", "s", "d")), + ColumnsFilter.inclusionBased(ImmutableSet.of("__time", "strCol", "dblCol")), null, temporaryFolder.newFolder() ); @@ -337,20 +386,20 @@ public void testReaderWithInclusiveColumnsFilter() throws IOException ImmutableList.of( new MapBasedInputRow( DateTimes.of("2000"), - ImmutableList.of("s", "d"), + ImmutableList.of("strCol", "dblCol"), ImmutableMap.builder() .put("__time", DateTimes.of("2000T").getMillis()) - .put("s", "foo") - .put("d", 1.23d) + .put("strCol", "foo") + .put("dblCol", 1.23d) .build() ), new MapBasedInputRow( DateTimes.of("2000T01"), - ImmutableList.of("s", "d"), + ImmutableList.of("strCol", "dblCol"), ImmutableMap.builder() .put("__time", DateTimes.of("2000T01").getMillis()) - .put("s", "bar") - .put("d", 4.56d) + .put("strCol", "bar") + .put("dblCol", 4.56d) .build() ) ), @@ -367,11 +416,11 @@ public void testReaderWithInclusiveColumnsFilterNoTimestamp() throws IOException new TimestampSpec("__time", "millis", DateTimes.of("1971")), new DimensionsSpec( ImmutableList.of( - StringDimensionSchema.create("s"), - new DoubleDimensionSchema("d") + StringDimensionSchema.create("strCol"), + new DoubleDimensionSchema("dblCol") ) ), - ColumnsFilter.inclusionBased(ImmutableSet.of("s", "d")), + ColumnsFilter.inclusionBased(ImmutableSet.of("strCol", "dblCol")), null, temporaryFolder.newFolder() ); @@ -380,18 +429,18 @@ public void testReaderWithInclusiveColumnsFilterNoTimestamp() throws IOException ImmutableList.of( new MapBasedInputRow( DateTimes.of("1971"), - ImmutableList.of("s", "d"), + ImmutableList.of("strCol", "dblCol"), ImmutableMap.builder() - .put("s", "foo") - .put("d", 1.23d) + .put("strCol", "foo") + .put("dblCol", 1.23d) .build() ), new MapBasedInputRow( DateTimes.of("1971"), - ImmutableList.of("s", "d"), + ImmutableList.of("strCol", "dblCol"), ImmutableMap.builder() - .put("s", "bar") - .put("d", 4.56d) + .put("strCol", "bar") + .put("dblCol", 4.56d) .build() ) ), @@ -408,12 +457,12 @@ public void testReaderWithFilter() throws IOException new TimestampSpec("__time", "millis", DateTimes.of("1971")), new DimensionsSpec( ImmutableList.of( - StringDimensionSchema.create("s"), - new DoubleDimensionSchema("d") + StringDimensionSchema.create("strCol"), + new DoubleDimensionSchema("dblCol") ) ), ColumnsFilter.all(), - new SelectorDimFilter("d", "1.23", null), + new SelectorDimFilter("dblCol", "1.23", null), temporaryFolder.newFolder() ); @@ -421,11 +470,11 @@ public void testReaderWithFilter() throws IOException ImmutableList.of( new MapBasedInputRow( DateTimes.of("2000"), - ImmutableList.of("s", "d"), + ImmutableList.of("strCol", "dblCol"), ImmutableMap.builder() .put("__time", DateTimes.of("2000T").getMillis()) - .put("s", "foo") - .put("d", 1.23d) + .put("strCol", "foo") + .put("dblCol", 1.23d) .put("cnt", 1L) .put("met_s", makeHLLC("foo")) .build() @@ -441,11 +490,11 @@ public void testReaderTimestampFromDouble() throws IOException final DruidSegmentReader reader = new DruidSegmentReader( makeInputEntity(Intervals.of("2000/P1D")), indexIO, - new TimestampSpec("d", "posix", null), + new TimestampSpec("dblCol", "posix", null), new DimensionsSpec( ImmutableList.of( - StringDimensionSchema.create("s"), - new DoubleDimensionSchema("d") + StringDimensionSchema.create("strCol"), + new DoubleDimensionSchema("dblCol") ) ), ColumnsFilter.all(), @@ -457,22 +506,22 @@ public void testReaderTimestampFromDouble() throws IOException ImmutableList.of( new MapBasedInputRow( DateTimes.of("1970-01-01T00:00:01.000Z"), - ImmutableList.of("s", "d"), + ImmutableList.of("strCol", "dblCol"), ImmutableMap.builder() .put("__time", DateTimes.of("2000T").getMillis()) - .put("s", "foo") - .put("d", 1.23d) + .put("strCol", "foo") + .put("dblCol", 1.23d) .put("cnt", 1L) .put("met_s", makeHLLC("foo")) .build() ), new MapBasedInputRow( DateTimes.of("1970-01-01T00:00:04.000Z"), - ImmutableList.of("s", "d"), + ImmutableList.of("strCol", "dblCol"), ImmutableMap.builder() .put("__time", DateTimes.of("2000T01").getMillis()) - .put("s", "bar") - .put("d", 4.56d) + .put("strCol", "bar") + .put("dblCol", 4.56d) .put("cnt", 1L) .put("met_s", makeHLLC("bar")) .build() @@ -491,8 +540,8 @@ public void testReaderTimestampAsPosixIncorrectly() throws IOException new TimestampSpec("__time", "posix", null), new DimensionsSpec( ImmutableList.of( - StringDimensionSchema.create("s"), - new DoubleDimensionSchema("d") + StringDimensionSchema.create("strCol"), + new DoubleDimensionSchema("dblCol") ) ), ColumnsFilter.all(), @@ -504,22 +553,22 @@ public void testReaderTimestampAsPosixIncorrectly() throws IOException ImmutableList.of( new MapBasedInputRow( DateTimes.of("31969-04-01T00:00:00.000Z"), - ImmutableList.of("s", "d"), + ImmutableList.of("strCol", "dblCol"), ImmutableMap.builder() .put("__time", DateTimes.of("2000T").getMillis()) - .put("s", "foo") - .put("d", 1.23d) + .put("strCol", "foo") + .put("dblCol", 1.23d) .put("cnt", 1L) .put("met_s", makeHLLC("foo")) .build() ), new MapBasedInputRow( DateTimes.of("31969-05-12T16:00:00.000Z"), - ImmutableList.of("s", "d"), + ImmutableList.of("strCol", "dblCol"), ImmutableMap.builder() .put("__time", DateTimes.of("2000T01").getMillis()) - .put("s", "bar") - .put("d", 4.56d) + .put("strCol", "bar") + .put("dblCol", 4.56d) .put("cnt", 1L) .put("met_s", makeHLLC("bar")) .build() @@ -538,8 +587,8 @@ public void testReaderTimestampSpecDefault() throws IOException new TimestampSpec(null, null, DateTimes.of("1971")), new DimensionsSpec( ImmutableList.of( - StringDimensionSchema.create("s"), - new DoubleDimensionSchema("d") + StringDimensionSchema.create("strCol"), + new DoubleDimensionSchema("dblCol") ) ), ColumnsFilter.all(), @@ -551,22 +600,22 @@ public void testReaderTimestampSpecDefault() throws IOException ImmutableList.of( new MapBasedInputRow( DateTimes.of("1971"), - ImmutableList.of("s", "d"), + ImmutableList.of("strCol", "dblCol"), ImmutableMap.builder() .put("__time", DateTimes.of("2000T").getMillis()) - .put("s", "foo") - .put("d", 1.23d) + .put("strCol", "foo") + .put("dblCol", 1.23d) .put("cnt", 1L) .put("met_s", makeHLLC("foo")) .build() ), new MapBasedInputRow( DateTimes.of("1971"), - ImmutableList.of("s", "d"), + ImmutableList.of("strCol", "dblCol"), ImmutableMap.builder() .put("__time", DateTimes.of("2000T01").getMillis()) - .put("s", "bar") - .put("d", 4.56d) + .put("strCol", "bar") + .put("dblCol", 4.56d) .put("cnt", 1L) .put("met_s", makeHLLC("bar")) .build() @@ -623,10 +672,20 @@ public void close() private DruidSegmentInputEntity makeInputEntity(final Interval interval) { - return makeInputEntity(interval, segmentDirectory); + return makeInputEntity(interval, segmentDirectory, ImmutableList.of("strCol", "dblCol"), ImmutableList.of("cnt", "met_s")); + } + + private DruidSegmentInputEntity makeInputEntityWithParams(final Interval interval, final List dimensions, final List metrics) + { + return makeInputEntity(interval, segmentDirectory, dimensions, metrics); } - public static DruidSegmentInputEntity makeInputEntity(final Interval interval, final File segmentDirectory) + public static DruidSegmentInputEntity makeInputEntity( + final Interval interval, + final File segmentDirectory, + final List dimensions, + final List metrics + ) { return new DruidSegmentInputEntity( new SegmentCacheManager() @@ -669,9 +728,9 @@ public void loadSegmentIntoPageCache(DataSegment segment, ExecutorService exec) }, DataSegment.builder() .dataSource("ds") - .dimensions(ImmutableList.of("s", "d")) - .metrics(ImmutableList.of("cnt", "met_s")) - .interval(Intervals.of("2000/P1D")) + .dimensions(dimensions) + .metrics(metrics) + .interval(interval) .version("1") .size(0) .build(), @@ -765,4 +824,36 @@ private static HyperLogLogCollector makeHLLC(final String... values) } return collector; } + + private void createTestSetup() throws IOException + { + final IncrementalIndex incrementalIndex = + IndexBuilder.create() + .schema( + new IncrementalIndexSchema.Builder() + .withDimensionsSpec(dimensionsSpec) + .withMetrics(metrics.toArray(new AggregatorFactory[0])) + .withRollup(false) + .build() + ) + .rows(rows) + .buildIncrementalIndex(); + + segmentDirectory = temporaryFolder.newFolder(); + + try { + TestHelper.getTestIndexMergerV9( + OnHeapMemorySegmentWriteOutMediumFactory.instance() + ).persist( + incrementalIndex, + segmentDirectory, + new IndexSpec(), + null + ); + } + finally { + incrementalIndex.close(); + } + } + } diff --git a/processing/src/main/java/org/apache/druid/segment/data/BlockLayoutColumnarDoublesSupplier.java b/processing/src/main/java/org/apache/druid/segment/data/BlockLayoutColumnarDoublesSupplier.java index 28b3c5d2b6c3..98a7ab51f987 100644 --- a/processing/src/main/java/org/apache/druid/segment/data/BlockLayoutColumnarDoublesSupplier.java +++ b/processing/src/main/java/org/apache/druid/segment/data/BlockLayoutColumnarDoublesSupplier.java @@ -22,6 +22,7 @@ import com.google.common.base.Supplier; import org.apache.druid.collections.ResourceHolder; +import javax.annotation.Nullable; import java.nio.ByteBuffer; import java.nio.ByteOrder; import java.nio.DoubleBuffer; @@ -82,10 +83,12 @@ private class BlockLayoutColumnarDoubles implements ColumnarDoubles final Indexed> singleThreadedDoubleBuffers = baseDoubleBuffers.singleThreaded(); int currBufferNum = -1; + @Nullable ResourceHolder holder; /** * doubleBuffer's position must be 0 */ + @Nullable DoubleBuffer doubleBuffer; @Override @@ -180,7 +183,10 @@ protected void loadBuffer(int bufferNum) public void close() { if (holder != null) { + currBufferNum = -1; holder.close(); + holder = null; + doubleBuffer = null; } } diff --git a/processing/src/main/java/org/apache/druid/segment/data/BlockLayoutColumnarFloatsSupplier.java b/processing/src/main/java/org/apache/druid/segment/data/BlockLayoutColumnarFloatsSupplier.java index c11ba18ec3df..26d7c798c79e 100644 --- a/processing/src/main/java/org/apache/druid/segment/data/BlockLayoutColumnarFloatsSupplier.java +++ b/processing/src/main/java/org/apache/druid/segment/data/BlockLayoutColumnarFloatsSupplier.java @@ -22,6 +22,7 @@ import com.google.common.base.Supplier; import org.apache.druid.collections.ResourceHolder; +import javax.annotation.Nullable; import java.nio.ByteBuffer; import java.nio.ByteOrder; import java.nio.FloatBuffer; @@ -82,10 +83,12 @@ private class BlockLayoutColumnarFloats implements ColumnarFloats final Indexed> singleThreadedFloatBuffers = baseFloatBuffers.singleThreaded(); int currBufferNum = -1; + @Nullable ResourceHolder holder; /** * floatBuffer's position must be 0 */ + @Nullable FloatBuffer floatBuffer; @Override @@ -180,7 +183,10 @@ protected void loadBuffer(int bufferNum) public void close() { if (holder != null) { + currBufferNum = -1; holder.close(); + holder = null; + floatBuffer = null; } } diff --git a/processing/src/main/java/org/apache/druid/segment/data/BlockLayoutColumnarLongsSupplier.java b/processing/src/main/java/org/apache/druid/segment/data/BlockLayoutColumnarLongsSupplier.java index 05473320553f..29a0748bcee5 100644 --- a/processing/src/main/java/org/apache/druid/segment/data/BlockLayoutColumnarLongsSupplier.java +++ b/processing/src/main/java/org/apache/druid/segment/data/BlockLayoutColumnarLongsSupplier.java @@ -22,6 +22,7 @@ import com.google.common.base.Supplier; import org.apache.druid.collections.ResourceHolder; +import javax.annotation.Nullable; import java.nio.ByteBuffer; import java.nio.ByteOrder; import java.nio.LongBuffer; @@ -123,11 +124,14 @@ private class BlockLayoutColumnarLongs implements ColumnarLongs final Indexed> singleThreadedLongBuffers = baseLongBuffers.singleThreaded(); int currBufferNum = -1; + @Nullable ResourceHolder holder; + @Nullable ByteBuffer buffer; /** * longBuffer's position must be 0 */ + @Nullable LongBuffer longBuffer; @Override @@ -204,7 +208,11 @@ protected void loadBuffer(int bufferNum) public void close() { if (holder != null) { + currBufferNum = -1; holder.close(); + holder = null; + buffer = null; + longBuffer = null; } } From db7c29c6f9e39ccd482b3f5374e40ae31bfb7d0e Mon Sep 17 00:00:00 2001 From: Jill Osborne Date: Mon, 28 Nov 2022 04:54:27 +0000 Subject: [PATCH 043/141] Correction to firehose migration doc (#13423) --- docs/ingestion/migrate-from-firehose-ingestion.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/docs/ingestion/migrate-from-firehose-ingestion.md b/docs/ingestion/migrate-from-firehose-ingestion.md index 2aa77772652c..c5d3f6946f21 100644 --- a/docs/ingestion/migrate-from-firehose-ingestion.md +++ b/docs/ingestion/migrate-from-firehose-ingestion.md @@ -23,7 +23,7 @@ sidebar_label: "Migrate from firehose" ~ under the License. --> -Apache deprecated support for Druid firehoses in version 0.17. Support for firehose ingestion was removed in version 24.0. +Apache deprecated support for Druid firehoses in version 0.17. Support for firehose ingestion will be removed in version 26.0. If you're using a firehose for batch ingestion, we strongly recommend that you follow the instructions on this page to transition to using native batch ingestion input sources as soon as possible. From 656b6cdf62a66aac6ff8fb7b7823d6f056b37d13 Mon Sep 17 00:00:00 2001 From: Kashif Faraz Date: Mon, 28 Nov 2022 19:32:37 +0530 Subject: [PATCH 044/141] Add MetricsVerifier to simplify verification of metric values in tests (#13442) --- .../java/util/metrics/MetricsVerifier.java | 98 +++++++++++++++++++ .../java/util/metrics/StubServiceEmitter.java | 38 +++++-- .../query/groupby/GroupByQueryRunnerTest.java | 3 +- .../druid/query/scan/ScanQueryRunnerTest.java | 3 +- .../timeseries/TimeseriesQueryRunnerTest.java | 6 +- .../simulate/BalancingStrategiesTest.java | 12 +-- .../simulate/CoordinatorSimulation.java | 7 +- .../CoordinatorSimulationBaseTest.java | 88 ++++------------- .../CoordinatorSimulationBuilder.java | 6 +- .../simulate/SegmentLoadingTest.java | 4 +- .../metrics/TaskCountStatsMonitorTest.java | 15 +-- .../TaskSlotCountStatsMonitorTest.java | 15 +-- .../WorkerTaskCountStatsMonitorTest.java | 59 ++++++----- .../druid/sql/http/SqlResourceTest.java | 12 +-- 14 files changed, 205 insertions(+), 161 deletions(-) create mode 100644 core/src/test/java/org/apache/druid/java/util/metrics/MetricsVerifier.java diff --git a/core/src/test/java/org/apache/druid/java/util/metrics/MetricsVerifier.java b/core/src/test/java/org/apache/druid/java/util/metrics/MetricsVerifier.java new file mode 100644 index 000000000000..c0f2eaeaf15f --- /dev/null +++ b/core/src/test/java/org/apache/druid/java/util/metrics/MetricsVerifier.java @@ -0,0 +1,98 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.java.util.metrics; + +import org.junit.Assert; + +import java.util.List; +import java.util.Map; + +/** + * Test utility to extract and verify metric values. + */ +public interface MetricsVerifier +{ + /** + * Verifies that no event has been emitted for the given metric. + */ + default void verifyNotEmitted(String metricName) + { + verifyEmitted(metricName, 0); + } + + /** + * Verifies that the metric was emitted the expected number of times. + */ + default void verifyEmitted(String metricName, int times) + { + verifyEmitted(metricName, null, times); + } + + /** + * Verifies that the metric was emitted for the given dimension filters the + * expected number of times. + */ + default void verifyEmitted(String metricName, Map dimensionFilters, int times) + { + Assert.assertEquals( + "Metric was emitted unexpected number of times.", + times, + getMetricValues(metricName, dimensionFilters).size() + ); + } + + /** + * Verifies the value of the specified metric emitted in the previous run. + */ + default void verifyValue(String metricName, Number expectedValue) + { + verifyValue(metricName, null, expectedValue); + } + + /** + * Verifies the value of the event corresponding to the specified metric and + * dimensionFilters emitted in the previous run. + */ + default void verifyValue(String metricName, Map dimensionFilters, Number expectedValue) + { + Assert.assertEquals(expectedValue, getValue(metricName, dimensionFilters)); + } + + /** + * Gets the value of the event corresponding to the specified metric and + * dimensionFilters. + */ + default Number getValue(String metricName, Map dimensionFilters) + { + List values = getMetricValues(metricName, dimensionFilters); + Assert.assertEquals( + "Metric must have been emitted exactly once for the given dimensions.", + 1, + values.size() + ); + return values.get(0); + } + + /** + * Gets the metric values for the specified dimension filters. + */ + List getMetricValues(String metricName, Map dimensionFilters); + +} diff --git a/core/src/test/java/org/apache/druid/java/util/metrics/StubServiceEmitter.java b/core/src/test/java/org/apache/druid/java/util/metrics/StubServiceEmitter.java index 653dc8a08aae..395245815792 100644 --- a/core/src/test/java/org/apache/druid/java/util/metrics/StubServiceEmitter.java +++ b/core/src/test/java/org/apache/druid/java/util/metrics/StubServiceEmitter.java @@ -24,12 +24,15 @@ import org.apache.druid.java.util.emitter.service.ServiceMetricEvent; import java.util.ArrayList; +import java.util.Collections; +import java.util.HashMap; import java.util.List; +import java.util.Map; -public class StubServiceEmitter extends ServiceEmitter +public class StubServiceEmitter extends ServiceEmitter implements MetricsVerifier { private final List events = new ArrayList<>(); - private final List metricEvents = new ArrayList<>(); + private final Map> metricEvents = new HashMap<>(); public StubServiceEmitter(String service, String host) { @@ -40,7 +43,9 @@ public StubServiceEmitter(String service, String host) public void emit(Event event) { if (event instanceof ServiceMetricEvent) { - metricEvents.add((ServiceMetricEvent) event); + ServiceMetricEvent metricEvent = (ServiceMetricEvent) event; + metricEvents.computeIfAbsent(metricEvent.getMetric(), name -> new ArrayList<>()) + .add(metricEvent); } events.add(event); } @@ -53,12 +58,29 @@ public List getEvents() return events; } - /** - * Gets all the metric events emitted since the previous {@link #flush()}. - */ - public List getMetricEvents() + @Override + public List getMetricValues( + String metricName, + Map dimensionFilters + ) { - return metricEvents; + final List values = new ArrayList<>(); + final List events = + metricEvents.getOrDefault(metricName, Collections.emptyList()); + final Map filters = + dimensionFilters == null ? Collections.emptyMap() : dimensionFilters; + for (ServiceMetricEvent event : events) { + final Map userDims = event.getUserDims(); + boolean match = filters.keySet().stream() + .map(d -> filters.get(d).equals(userDims.get(d))) + .reduce((a, b) -> a && b) + .orElse(true); + if (match) { + values.add(event.getValue()); + } + } + + return values; } @Override diff --git a/processing/src/test/java/org/apache/druid/query/groupby/GroupByQueryRunnerTest.java b/processing/src/test/java/org/apache/druid/query/groupby/GroupByQueryRunnerTest.java index f9060db4c94c..47be50997f23 100644 --- a/processing/src/test/java/org/apache/druid/query/groupby/GroupByQueryRunnerTest.java +++ b/processing/src/test/java/org/apache/druid/query/groupby/GroupByQueryRunnerTest.java @@ -757,8 +757,7 @@ public void testGroupBy() query, serviceEmitter ); - Assert.assertEquals(1, serviceEmitter.getEvents().size()); - Assert.assertEquals(vectorize, serviceEmitter.getEvents().get(0).toMap().getOrDefault("vectorized", null)); + serviceEmitter.verifyEmitted("query/wait/time", ImmutableMap.of("vectorized", vectorize), 1); TestHelper.assertExpectedObjects(expectedResults, results, "groupBy"); } diff --git a/processing/src/test/java/org/apache/druid/query/scan/ScanQueryRunnerTest.java b/processing/src/test/java/org/apache/druid/query/scan/ScanQueryRunnerTest.java index 31a7a454f374..4840a1ccfd18 100644 --- a/processing/src/test/java/org/apache/druid/query/scan/ScanQueryRunnerTest.java +++ b/processing/src/test/java/org/apache/druid/query/scan/ScanQueryRunnerTest.java @@ -230,8 +230,7 @@ public void testFullOnSelect() 0, 3 ); - Assert.assertEquals(1, stubServiceEmitter.getEvents().size()); - Assert.assertEquals(false, stubServiceEmitter.getEvents().get(0).toMap().getOrDefault("vectorized", null)); + stubServiceEmitter.verifyEmitted("query/wait/time", ImmutableMap.of("vectorized", false), 1); verify(expectedResults, populateNullColumnAtLastForQueryableIndexCase(results, "null_column")); } diff --git a/processing/src/test/java/org/apache/druid/query/timeseries/TimeseriesQueryRunnerTest.java b/processing/src/test/java/org/apache/druid/query/timeseries/TimeseriesQueryRunnerTest.java index 74155ce51c54..d9d73e67effb 100644 --- a/processing/src/test/java/org/apache/druid/query/timeseries/TimeseriesQueryRunnerTest.java +++ b/processing/src/test/java/org/apache/druid/query/timeseries/TimeseriesQueryRunnerTest.java @@ -317,11 +317,7 @@ public void testFullOnTimeseries() ++count; } - Assert.assertEquals(1, stubServiceEmitter.getEvents().size()); - Assert.assertEquals( - vectorize, - stubServiceEmitter.getEvents().get(0).toMap().getOrDefault("vectorized", null) - ); + stubServiceEmitter.verifyEmitted("query/wait/time", ImmutableMap.of("vectorized", vectorize), 1); Assert.assertEquals(lastResult.toString(), expectedLast, lastResult.getTimestamp()); } diff --git a/server/src/test/java/org/apache/druid/server/coordinator/simulate/BalancingStrategiesTest.java b/server/src/test/java/org/apache/druid/server/coordinator/simulate/BalancingStrategiesTest.java index fd1c6c4d0f57..f135280825e9 100644 --- a/server/src/test/java/org/apache/druid/server/coordinator/simulate/BalancingStrategiesTest.java +++ b/server/src/test/java/org/apache/druid/server/coordinator/simulate/BalancingStrategiesTest.java @@ -76,8 +76,8 @@ public void testFreshClusterGetsBalanced() runCoordinatorCycle(); loadQueuedSegments(); verifyValue(Metric.ASSIGNED_COUNT, 1000L); - verifyNoEvent(Metric.MOVED_COUNT); - verifyNoEvent(Metric.UNMOVED_COUNT); + verifyNotEmitted(Metric.MOVED_COUNT); + verifyNotEmitted(Metric.UNMOVED_COUNT); for (DruidServer historical : historicals) { Assert.assertEquals(200, historical.getTotalSegments()); @@ -113,8 +113,8 @@ public void testClusterGetsBalancedWhenServerIsAdded() runCoordinatorCycle(); loadQueuedSegments(); verifyValue(Metric.ASSIGNED_COUNT, 1000L); - verifyNoEvent(Metric.MOVED_COUNT); - verifyNoEvent(Metric.UNMOVED_COUNT); + verifyNotEmitted(Metric.MOVED_COUNT); + verifyNotEmitted(Metric.UNMOVED_COUNT); // Verify that each server is equally loaded for (DruidServer historical : historicals) { @@ -161,8 +161,8 @@ public void testClusterGetsBalancedWhenServerIsRemoved() runCoordinatorCycle(); loadQueuedSegments(); verifyValue(Metric.ASSIGNED_COUNT, 1000L); - verifyNoEvent(Metric.MOVED_COUNT); - verifyNoEvent(Metric.UNMOVED_COUNT); + verifyNotEmitted(Metric.MOVED_COUNT); + verifyNotEmitted(Metric.UNMOVED_COUNT); // Verify that each server is equally loaded for (DruidServer historical : historicals) { diff --git a/server/src/test/java/org/apache/druid/server/coordinator/simulate/CoordinatorSimulation.java b/server/src/test/java/org/apache/druid/server/coordinator/simulate/CoordinatorSimulation.java index f79a9bf48041..c0565c19ba9d 100644 --- a/server/src/test/java/org/apache/druid/server/coordinator/simulate/CoordinatorSimulation.java +++ b/server/src/test/java/org/apache/druid/server/coordinator/simulate/CoordinatorSimulation.java @@ -20,7 +20,7 @@ package org.apache.druid.server.coordinator.simulate; import org.apache.druid.client.DruidServer; -import org.apache.druid.java.util.emitter.service.ServiceMetricEvent; +import org.apache.druid.java.util.metrics.MetricsVerifier; import org.apache.druid.server.coordinator.CoordinatorDynamicConfig; import org.apache.druid.timeline.DataSegment; @@ -81,9 +81,10 @@ interface CoordinatorState DruidServer getInventoryView(String serverName); /** - * Returns the metric events emitted in the previous coordinator run. + * Returns a MetricsVerifier which can be used to extract and verify the + * metric values emitted in the previous coordinator run. */ - List getMetricEvents(); + MetricsVerifier getMetricsVerifier(); /** * Gets the load percentage of the specified datasource as seen by the coordinator. diff --git a/server/src/test/java/org/apache/druid/server/coordinator/simulate/CoordinatorSimulationBaseTest.java b/server/src/test/java/org/apache/druid/server/coordinator/simulate/CoordinatorSimulationBaseTest.java index d4e8b7760de0..84c6886d9d40 100644 --- a/server/src/test/java/org/apache/druid/server/coordinator/simulate/CoordinatorSimulationBaseTest.java +++ b/server/src/test/java/org/apache/druid/server/coordinator/simulate/CoordinatorSimulationBaseTest.java @@ -21,7 +21,7 @@ import org.apache.druid.client.DruidServer; import org.apache.druid.java.util.common.granularity.Granularities; -import org.apache.druid.java.util.emitter.service.ServiceMetricEvent; +import org.apache.druid.java.util.metrics.MetricsVerifier; import org.apache.druid.server.coordination.ServerType; import org.apache.druid.server.coordinator.CoordinatorDynamicConfig; import org.apache.druid.server.coordinator.CreateDataSegments; @@ -32,8 +32,6 @@ import org.junit.Assert; import org.junit.Before; -import java.util.ArrayList; -import java.util.Collections; import java.util.HashMap; import java.util.List; import java.util.Map; @@ -51,13 +49,15 @@ * leading to flakiness in the tests. The simulation sets this field to true by * default. */ -public abstract class CoordinatorSimulationBaseTest - implements CoordinatorSimulation.CoordinatorState, CoordinatorSimulation.ClusterState +public abstract class CoordinatorSimulationBaseTest implements + CoordinatorSimulation.CoordinatorState, + CoordinatorSimulation.ClusterState, + MetricsVerifier { static final double DOUBLE_DELTA = 10e-9; private CoordinatorSimulation sim; - private final Map> latestMetricEvents = new HashMap<>(); + private MetricsVerifier metricsVerifier; @Before public abstract void setUp(); @@ -78,25 +78,19 @@ void startSimulation(CoordinatorSimulation simulation) { this.sim = simulation; simulation.start(); + this.metricsVerifier = this.sim.coordinator().getMetricsVerifier(); } @Override public void runCoordinatorCycle() { - latestMetricEvents.clear(); sim.coordinator().runCoordinatorCycle(); - - // Extract the metric values of this run - for (ServiceMetricEvent event : sim.coordinator().getMetricEvents()) { - latestMetricEvents.computeIfAbsent(event.getMetric(), m -> new ArrayList<>()) - .add(event); - } } @Override - public List getMetricEvents() + public MetricsVerifier getMetricsVerifier() { - return sim.coordinator().getMetricEvents(); + return null; } @Override @@ -153,61 +147,13 @@ void verifyDatasourceIsFullyLoaded(String datasource) Assert.assertEquals(100.0, getLoadPercentage(datasource), DOUBLE_DELTA); } - void verifyNoEvent(String metricName) - { - Assert.assertTrue(getMetricValues(metricName, null).isEmpty()); - } - - /** - * Verifies the value of the specified metric emitted in the previous run. - */ - void verifyValue(String metricName, Number expectedValue) - { - verifyValue(metricName, null, expectedValue); - } - - /** - * Verifies the value of the event corresponding to the specified metric and - * dimensionFilters emitted in the previous run. - */ - void verifyValue(String metricName, Map dimensionFilters, Number expectedValue) - { - Assert.assertEquals(expectedValue, getValue(metricName, dimensionFilters)); - } - - /** - * Gets the value of the event corresponding to the specified metric and - * dimensionFilters emitted in the previous run. - */ - Number getValue(String metricName, Map dimensionFilters) - { - List values = getMetricValues(metricName, dimensionFilters); - Assert.assertEquals( - "Metric must have been emitted exactly once for the given dimensions.", - 1, - values.size() - ); - return values.get(0); - } - - private List getMetricValues(String metricName, Map dimensionFilters) + @Override + public List getMetricValues( + String metricName, + Map dimensionFilters + ) { - final List values = new ArrayList<>(); - final List events = latestMetricEvents.getOrDefault(metricName, Collections.emptyList()); - final Map filters = dimensionFilters == null - ? Collections.emptyMap() : dimensionFilters; - for (ServiceMetricEvent event : events) { - final Map userDims = event.getUserDims(); - boolean match = filters.keySet().stream() - .map(d -> filters.get(d).equals(userDims.get(d))) - .reduce((a, b) -> a && b) - .orElse(true); - if (match) { - values.add(event.getValue()); - } - } - - return values; + return metricsVerifier.getMetricValues(metricName, dimensionFilters); } // Utility methods @@ -237,13 +183,13 @@ static CoordinatorDynamicConfig createDynamicConfig( /** * Creates a map containing dimension key-values to filter out metric events. */ - static Map filter(String... dimensionValues) + static Map filter(String... dimensionValues) { if (dimensionValues.length < 2 || dimensionValues.length % 2 == 1) { throw new IllegalArgumentException("Dimension key-values must be specified in pairs."); } - final Map filters = new HashMap<>(); + final Map filters = new HashMap<>(); for (int i = 0; i < dimensionValues.length; ) { filters.put(dimensionValues[i], dimensionValues[i + 1]); i += 2; diff --git a/server/src/test/java/org/apache/druid/server/coordinator/simulate/CoordinatorSimulationBuilder.java b/server/src/test/java/org/apache/druid/server/coordinator/simulate/CoordinatorSimulationBuilder.java index 01308d82e73f..7066183c3ac6 100644 --- a/server/src/test/java/org/apache/druid/server/coordinator/simulate/CoordinatorSimulationBuilder.java +++ b/server/src/test/java/org/apache/druid/server/coordinator/simulate/CoordinatorSimulationBuilder.java @@ -32,8 +32,8 @@ import org.apache.druid.java.util.common.concurrent.ScheduledExecutorFactory; import org.apache.druid.java.util.common.lifecycle.Lifecycle; import org.apache.druid.java.util.emitter.EmittingLogger; -import org.apache.druid.java.util.emitter.service.ServiceMetricEvent; import org.apache.druid.java.util.http.client.HttpClient; +import org.apache.druid.java.util.metrics.MetricsVerifier; import org.apache.druid.java.util.metrics.StubServiceEmitter; import org.apache.druid.server.coordinator.BalancerStrategyFactory; import org.apache.druid.server.coordinator.CachingCostBalancerStrategyConfig; @@ -392,9 +392,9 @@ public double getLoadPercentage(String datasource) } @Override - public List getMetricEvents() + public MetricsVerifier getMetricsVerifier() { - return new ArrayList<>(env.serviceEmitter.getMetricEvents()); + return env.serviceEmitter; } } diff --git a/server/src/test/java/org/apache/druid/server/coordinator/simulate/SegmentLoadingTest.java b/server/src/test/java/org/apache/druid/server/coordinator/simulate/SegmentLoadingTest.java index 1edeab8a370f..e04ebe532df5 100644 --- a/server/src/test/java/org/apache/druid/server/coordinator/simulate/SegmentLoadingTest.java +++ b/server/src/test/java/org/apache/druid/server/coordinator/simulate/SegmentLoadingTest.java @@ -148,7 +148,7 @@ public void testDropHappensAfterTargetReplicationOnEveryTier() startSimulation(sim); runCoordinatorCycle(); - verifyNoEvent(Metric.DROPPED_COUNT); + verifyNotEmitted(Metric.DROPPED_COUNT); int totalAssignedInRun1 = getValue(Metric.ASSIGNED_COUNT, filter(DruidMetrics.TIER, Tier.T2)).intValue() + getValue(Metric.ASSIGNED_COUNT, filter(DruidMetrics.TIER, Tier.T3)).intValue(); @@ -158,7 +158,7 @@ public void testDropHappensAfterTargetReplicationOnEveryTier() runCoordinatorCycle(); loadQueuedSegments(); - verifyNoEvent(Metric.DROPPED_COUNT); + verifyNotEmitted(Metric.DROPPED_COUNT); int totalLoadedAfterRun2 = historicalT21.getTotalSegments() + historicalT22.getTotalSegments() + historicalT31.getTotalSegments() + historicalT32.getTotalSegments(); diff --git a/server/src/test/java/org/apache/druid/server/metrics/TaskCountStatsMonitorTest.java b/server/src/test/java/org/apache/druid/server/metrics/TaskCountStatsMonitorTest.java index 24d89ab957a2..c6a47d56c427 100644 --- a/server/src/test/java/org/apache/druid/server/metrics/TaskCountStatsMonitorTest.java +++ b/server/src/test/java/org/apache/druid/server/metrics/TaskCountStatsMonitorTest.java @@ -75,15 +75,10 @@ public void testMonitor() final StubServiceEmitter emitter = new StubServiceEmitter("service", "host"); monitor.doMonitor(emitter); Assert.assertEquals(5, emitter.getEvents().size()); - Assert.assertEquals("task/success/count", emitter.getEvents().get(0).toMap().get("metric")); - Assert.assertEquals(1L, emitter.getEvents().get(0).toMap().get("value")); - Assert.assertEquals("task/failed/count", emitter.getEvents().get(1).toMap().get("metric")); - Assert.assertEquals(1L, emitter.getEvents().get(1).toMap().get("value")); - Assert.assertEquals("task/running/count", emitter.getEvents().get(2).toMap().get("metric")); - Assert.assertEquals(1L, emitter.getEvents().get(2).toMap().get("value")); - Assert.assertEquals("task/pending/count", emitter.getEvents().get(3).toMap().get("metric")); - Assert.assertEquals(1L, emitter.getEvents().get(3).toMap().get("value")); - Assert.assertEquals("task/waiting/count", emitter.getEvents().get(4).toMap().get("metric")); - Assert.assertEquals(1L, emitter.getEvents().get(4).toMap().get("value")); + emitter.verifyValue("task/success/count", 1L); + emitter.verifyValue("task/failed/count", 1L); + emitter.verifyValue("task/running/count", 1L); + emitter.verifyValue("task/pending/count", 1L); + emitter.verifyValue("task/waiting/count", 1L); } } diff --git a/server/src/test/java/org/apache/druid/server/metrics/TaskSlotCountStatsMonitorTest.java b/server/src/test/java/org/apache/druid/server/metrics/TaskSlotCountStatsMonitorTest.java index be33a4c98b68..0fed1c9b6bc1 100644 --- a/server/src/test/java/org/apache/druid/server/metrics/TaskSlotCountStatsMonitorTest.java +++ b/server/src/test/java/org/apache/druid/server/metrics/TaskSlotCountStatsMonitorTest.java @@ -75,15 +75,10 @@ public void testMonitor() final StubServiceEmitter emitter = new StubServiceEmitter("service", "host"); monitor.doMonitor(emitter); Assert.assertEquals(5, emitter.getEvents().size()); - Assert.assertEquals("taskSlot/total/count", emitter.getEvents().get(0).toMap().get("metric")); - Assert.assertEquals(1L, emitter.getEvents().get(0).toMap().get("value")); - Assert.assertEquals("taskSlot/idle/count", emitter.getEvents().get(1).toMap().get("metric")); - Assert.assertEquals(1L, emitter.getEvents().get(1).toMap().get("value")); - Assert.assertEquals("taskSlot/used/count", emitter.getEvents().get(2).toMap().get("metric")); - Assert.assertEquals(1L, emitter.getEvents().get(2).toMap().get("value")); - Assert.assertEquals("taskSlot/lazy/count", emitter.getEvents().get(3).toMap().get("metric")); - Assert.assertEquals(1L, emitter.getEvents().get(3).toMap().get("value")); - Assert.assertEquals("taskSlot/blacklisted/count", emitter.getEvents().get(4).toMap().get("metric")); - Assert.assertEquals(1L, emitter.getEvents().get(4).toMap().get("value")); + emitter.verifyValue("taskSlot/total/count", 1L); + emitter.verifyValue("taskSlot/idle/count", 1L); + emitter.verifyValue("taskSlot/used/count", 1L); + emitter.verifyValue("taskSlot/lazy/count", 1L); + emitter.verifyValue("taskSlot/blacklisted/count", 1L); } } diff --git a/server/src/test/java/org/apache/druid/server/metrics/WorkerTaskCountStatsMonitorTest.java b/server/src/test/java/org/apache/druid/server/metrics/WorkerTaskCountStatsMonitorTest.java index 01a97b7a9a03..fadb2f988269 100644 --- a/server/src/test/java/org/apache/druid/server/metrics/WorkerTaskCountStatsMonitorTest.java +++ b/server/src/test/java/org/apache/druid/server/metrics/WorkerTaskCountStatsMonitorTest.java @@ -20,10 +20,10 @@ package org.apache.druid.server.metrics; import com.google.common.collect.ImmutableList; +import com.google.common.collect.ImmutableMap; import com.google.common.collect.ImmutableSet; import com.google.inject.Guice; import com.google.inject.Injector; -import com.google.inject.Module; import org.apache.druid.discovery.NodeRole; import org.apache.druid.java.util.metrics.StubServiceEmitter; import org.junit.Assert; @@ -143,24 +143,18 @@ public String getWorkerVersion() injectorForMiddleManager = Guice.createInjector( ImmutableList.of( - (Module) binder -> { - binder.bind(WorkerTaskCountStatsProvider.class).toInstance(statsProvider); - } + binder -> binder.bind(WorkerTaskCountStatsProvider.class).toInstance(statsProvider) ) ); injectorForMiddleManagerNullStats = Guice.createInjector( ImmutableList.of( - (Module) binder -> { - binder.bind(WorkerTaskCountStatsProvider.class).toInstance(nullStatsProvider); - } + binder -> binder.bind(WorkerTaskCountStatsProvider.class).toInstance(nullStatsProvider) ) ); injectorForPeon = Guice.createInjector( - ImmutableList.of( - (Module) binder -> {} - ) + ImmutableList.of(binder -> {}) ); } @@ -172,26 +166,31 @@ public void testMonitor() final StubServiceEmitter emitter = new StubServiceEmitter("service", "host"); monitor.doMonitor(emitter); Assert.assertEquals(5, emitter.getEvents().size()); - Assert.assertEquals("worker/task/failed/count", emitter.getEvents().get(0).toMap().get("metric")); - Assert.assertEquals("workerCategory", emitter.getEvents().get(0).toMap().get("category")); - Assert.assertEquals("workerVersion", emitter.getEvents().get(0).toMap().get("workerVersion")); - Assert.assertEquals(4L, emitter.getEvents().get(0).toMap().get("value")); - Assert.assertEquals("worker/task/success/count", emitter.getEvents().get(1).toMap().get("metric")); - Assert.assertEquals("workerCategory", emitter.getEvents().get(1).toMap().get("category")); - Assert.assertEquals("workerVersion", emitter.getEvents().get(1).toMap().get("workerVersion")); - Assert.assertEquals(2L, emitter.getEvents().get(1).toMap().get("value")); - Assert.assertEquals("worker/taskSlot/idle/count", emitter.getEvents().get(2).toMap().get("metric")); - Assert.assertEquals("workerCategory", emitter.getEvents().get(2).toMap().get("category")); - Assert.assertEquals("workerVersion", emitter.getEvents().get(2).toMap().get("workerVersion")); - Assert.assertEquals(3L, emitter.getEvents().get(2).toMap().get("value")); - Assert.assertEquals("worker/taskSlot/total/count", emitter.getEvents().get(3).toMap().get("metric")); - Assert.assertEquals("workerCategory", emitter.getEvents().get(3).toMap().get("category")); - Assert.assertEquals("workerVersion", emitter.getEvents().get(3).toMap().get("workerVersion")); - Assert.assertEquals(5L, emitter.getEvents().get(3).toMap().get("value")); - Assert.assertEquals("worker/taskSlot/used/count", emitter.getEvents().get(4).toMap().get("metric")); - Assert.assertEquals("workerCategory", emitter.getEvents().get(4).toMap().get("category")); - Assert.assertEquals("workerVersion", emitter.getEvents().get(4).toMap().get("workerVersion")); - Assert.assertEquals(1L, emitter.getEvents().get(4).toMap().get("value")); + emitter.verifyValue( + "worker/task/failed/count", + ImmutableMap.of("category", "workerCategory", "workerVersion", "workerVersion"), + 4L + ); + emitter.verifyValue( + "worker/task/success/count", + ImmutableMap.of("category", "workerCategory", "workerVersion", "workerVersion"), + 2L + ); + emitter.verifyValue( + "worker/taskSlot/idle/count", + ImmutableMap.of("category", "workerCategory", "workerVersion", "workerVersion"), + 3L + ); + emitter.verifyValue( + "worker/taskSlot/total/count", + ImmutableMap.of("category", "workerCategory", "workerVersion", "workerVersion"), + 5L + ); + emitter.verifyValue( + "worker/taskSlot/used/count", + ImmutableMap.of("category", "workerCategory", "workerVersion", "workerVersion"), + 1L + ); } @Test diff --git a/sql/src/test/java/org/apache/druid/sql/http/SqlResourceTest.java b/sql/src/test/java/org/apache/druid/sql/http/SqlResourceTest.java index d3287530c55b..42fa66cbaa71 100644 --- a/sql/src/test/java/org/apache/druid/sql/http/SqlResourceTest.java +++ b/sql/src/test/java/org/apache/druid/sql/http/SqlResourceTest.java @@ -425,15 +425,9 @@ public void testSqlLifecycleMetrics() throws Exception ); checkSqlRequestLog(true); Assert.assertTrue(lifecycleManager.getAll("id").isEmpty()); - Set metricNames = ImmutableSet.of("sqlQuery/time", "sqlQuery/bytes", "sqlQuery/planningTimeMs"); - Assert.assertEquals(3, stubServiceEmitter.getEvents().size()); - for (String metricName : metricNames) { - Assert.assertTrue( - stubServiceEmitter.getEvents() - .stream() - .anyMatch(event -> event.toMap().containsValue(metricName)) - ); - } + stubServiceEmitter.verifyEmitted("sqlQuery/time", 1); + stubServiceEmitter.verifyValue("sqlQuery/bytes", 27L); + stubServiceEmitter.verifyEmitted("sqlQuery/planningTimeMs", 1); } From 58c896ea0b7c9896ed1cada416e872e5dfee7ff3 Mon Sep 17 00:00:00 2001 From: Gian Merlino Date: Mon, 28 Nov 2022 08:54:46 -0800 Subject: [PATCH 045/141] ServiceClient: More robust redirect handling. (#13413) Detects self-redirects, redirect loops, long redirect chains, and redirects to unknown servers. Treat all of these cases as an unavailable service, retrying if the retry policy allows it. Previously, some of these cases would lead to a prompt, unretryable error. This caused clients contacting an Overlord during a leader change to fail with error messages like: org.apache.druid.rpc.RpcException: Service [overlord] redirected too many times Additionally, a slight refactor of callbacks in ServiceClientImpl improves readability of the flow through onSuccess. --- .../org/apache/druid/rpc/ServiceClient.java | 6 +- .../apache/druid/rpc/ServiceClientImpl.java | 237 ++++++++++++------ .../rpc/ServiceNotAvailableException.java | 9 +- .../druid/rpc/ServiceClientImplTest.java | 102 +++++++- 4 files changed, 272 insertions(+), 82 deletions(-) diff --git a/server/src/main/java/org/apache/druid/rpc/ServiceClient.java b/server/src/main/java/org/apache/druid/rpc/ServiceClient.java index cb33f713d988..1c416f88a1dc 100644 --- a/server/src/main/java/org/apache/druid/rpc/ServiceClient.java +++ b/server/src/main/java/org/apache/druid/rpc/ServiceClient.java @@ -49,8 +49,10 @@ public interface ServiceClient * encountered error. * * Redirects from 3xx responses are followed up to a chain length of {@link #MAX_REDIRECTS} and do not consume - * attempts. Redirects are validated against the targets returned by {@link ServiceLocator}: the client will not - * follow a redirect to a target that does not appear in the returned {@link ServiceLocations}. + * attempts. Redirects are validated against the targets returned by {@link ServiceLocator}: the client will only + * follow redirects to targets that appear in {@link ServiceLocations}. If the client encounters a redirect to an + * unknown target, or if a redirect loop or self-redirect is detected, it is treated as an unavailable service and + * an attempt is consumed. * * If the service is unavailable at the time an attempt is made, the client will automatically retry based on * {@link ServiceRetryPolicy#retryNotAvailable()}. If true, an attempt is consumed and the client will try to locate diff --git a/server/src/main/java/org/apache/druid/rpc/ServiceClientImpl.java b/server/src/main/java/org/apache/druid/rpc/ServiceClientImpl.java index 1445b943e96e..eca2cfdc5a2e 100644 --- a/server/src/main/java/org/apache/druid/rpc/ServiceClientImpl.java +++ b/server/src/main/java/org/apache/druid/rpc/ServiceClientImpl.java @@ -21,6 +21,7 @@ import com.google.common.annotations.VisibleForTesting; import com.google.common.base.Preconditions; +import com.google.common.collect.ImmutableSet; import com.google.common.collect.Iterables; import com.google.common.util.concurrent.FutureCallback; import com.google.common.util.concurrent.Futures; @@ -88,7 +89,7 @@ public ListenableFuture asyncRequest( ) { final SettableFuture retVal = SettableFuture.create(); - tryRequest(requestBuilder, handler, retVal, 0, 0); + tryRequest(requestBuilder, handler, retVal, 0, ImmutableSet.of()); return retVal; } @@ -98,21 +99,35 @@ public ServiceClientImpl withRetryPolicy(ServiceRetryPolicy newRetryPolicy) return new ServiceClientImpl(serviceName, httpClient, serviceLocator, newRetryPolicy, connectExec); } + /** + * Internal helper used by {@link #asyncRequest(RequestBuilder, HttpResponseHandler)}. + * + * Handles retries by calling itself back in {@link #connectExec} with an incremented {@code attemptNumber}. + * + * @param requestBuilder request builder from call to {@link #asyncRequest} + * @param handler handler from call to {@link #asyncRequest} + * @param retVal return future generated by {@link #asyncRequest} + * @param attemptNumber attempt number; starts at 0 and is incremented on each retry + * @param redirectLocations redirect locations observed from the server on this attempt; used for detecting redirect + * loops and for limiting redirect chain length to {@link #MAX_REDIRECTS}. Cleared when + * a new attempt is issued. + */ private void tryRequest( final RequestBuilder requestBuilder, final HttpResponseHandler handler, final SettableFuture retVal, final long attemptNumber, - final int redirectCount + final ImmutableSet redirectLocations ) { whenServiceReady( - serviceLocation -> { + serviceLocations -> { if (retVal.isCancelled()) { // Return early if the caller canceled the return future. return; } + final ServiceLocation serviceLocation = pick(serviceLocations); final long nextAttemptNumber = attemptNumber + 1; if (serviceLocation == null) { @@ -128,7 +143,7 @@ private void tryRequest( ); connectExec.schedule( - () -> tryRequest(requestBuilder, handler, retVal, attemptNumber + 1, redirectCount), + () -> tryRequest(requestBuilder, handler, retVal, nextAttemptNumber, ImmutableSet.of()), backoffMs, TimeUnit.MILLISECONDS ); @@ -172,71 +187,15 @@ public void onSuccess(@Nullable final Either 1) { - // There were retries. Log at INFO level to provide the user some closure. - log.info( - "Service [%s] request [%s %s] completed.", - serviceName, - request.getMethod(), - request.getUrl() - ); - } else { - // No retries. Log at debug level to avoid cluttering the logs. - log.debug( - "Service [%s] request [%s %s] completed.", - serviceName, - request.getMethod(), - request.getUrl() - ); - } - - // Will not throw, because we checked result.isValue() earlier. - retVal.set(result.valueOrThrow()); + handleResultValue(result.valueOrThrow()); } else { final StringFullResponseHolder errorHolder = result != null ? result.error() : null; if (errorHolder != null && isRedirect(errorHolder.getResponse().getStatus())) { - // Redirect. Update preferredLocationNoPath if appropriate, then reissue. - final String newUri = result.error().getResponse().headers().get("Location"); - - if (redirectCount >= MAX_REDIRECTS) { - retVal.setException(new RpcException( - "Service [%s] redirected too many times [%d] to invalid url %s", - serviceName, - redirectCount, - newUri - )); - } else { - // Update preferredLocationNoPath if we got a redirect. - final ServiceLocation redirectLocationNoPath = serviceLocationNoPathFromUri(newUri); - - if (redirectLocationNoPath != null) { - preferredLocationNoPath.set(redirectLocationNoPath); - connectExec.submit( - () -> tryRequest(requestBuilder, handler, retVal, attemptNumber, redirectCount + 1) - ); - } else { - retVal.setException( - new RpcException( - "Service [%s] redirected [%d] times to invalid URL [%s]", - serviceName, - redirectCount, - newUri - ) - ); - } - } + handleRedirect(errorHolder); } else if (shouldTry(nextAttemptNumber) && (errorHolder == null || retryPolicy.retryHttpResponse(errorHolder.getResponse()))) { - // Retryable server response (or null errorHolder, which means null result, which can happen - // if the HttpClient encounters an exception in the midst of response processing). - final long backoffMs = computeBackoffMs(retryPolicy, attemptNumber); - log.noStackTrace().info(buildErrorMessage(request, errorHolder, backoffMs, nextAttemptNumber)); - connectExec.schedule( - () -> tryRequest(requestBuilder, handler, retVal, attemptNumber + 1, redirectCount), - backoffMs, - TimeUnit.MILLISECONDS - ); + handleRetryableErrorResponse(errorHolder); } else if (errorHolder != null) { // Nonretryable server response. retVal.setException(new HttpResponseException(errorHolder)); @@ -264,7 +223,7 @@ public void onFailure(final Throwable t) log.noStackTrace().info(t, buildErrorMessage(request, null, backoffMs, nextAttemptNumber)); connectExec.schedule( - () -> tryRequest(requestBuilder, handler, retVal, attemptNumber + 1, redirectCount), + () -> tryRequest(requestBuilder, handler, retVal, nextAttemptNumber, ImmutableSet.of()), backoffMs, TimeUnit.MILLISECONDS ); @@ -277,6 +236,135 @@ public void onFailure(final Throwable t) retVal.setException(new RpcException(t, "Service [%s] handler exited unexpectedly", serviceName)); } } + + /** + * Handles HTTP 2xx responses from the server. + */ + private void handleResultValue(final FinalType value) + { + if (nextAttemptNumber > 1) { + // There were retries. Log at INFO level to provide the user some closure. + log.info( + "Service [%s] request [%s %s] completed.", + serviceName, + request.getMethod(), + request.getUrl() + ); + } else { + // No retries. Log at debug level to avoid cluttering the logs. + log.debug( + "Service [%s] request [%s %s] completed.", + serviceName, + request.getMethod(), + request.getUrl() + ); + } + + // Will not throw, because we checked result.isValue() earlier. + retVal.set(value); + } + + /** + * Handles retryable HTTP error responses from the server. + */ + private void handleRetryableErrorResponse(final StringFullResponseHolder errorHolder) + { + // Retryable server response (or null errorHolder, which means null result, which can happen + // if the HttpClient encounters an exception in the midst of response processing). + final long backoffMs = computeBackoffMs(retryPolicy, attemptNumber); + log.info(buildErrorMessage(request, errorHolder, backoffMs, nextAttemptNumber)); + connectExec.schedule( + () -> tryRequest(requestBuilder, handler, retVal, nextAttemptNumber, ImmutableSet.of()), + backoffMs, + TimeUnit.MILLISECONDS + ); + } + + /** + * Handles HTTP redirect responses from the server. + */ + private void handleRedirect(final StringFullResponseHolder errorHolder) + { + // Redirect. Update preferredLocationNoPath if appropriate, then reissue. + final String newUri = errorHolder.getResponse().headers().get("Location"); + final ServiceLocation redirectLocationNoPath = serviceLocationNoPathFromUri(newUri); + + if (redirectLocationNoPath == null) { + // Redirect to invalid URL. Something is wrong with the server: fail immediately + // without retries. + retVal.setException( + new RpcException( + "Service [%s] redirected to invalid URL [%s]", + serviceName, + newUri + ) + ); + } else if (serviceLocations.getLocations() + .stream() + .anyMatch(loc -> serviceLocationNoPath(loc) + .equals(redirectLocationNoPath))) { + // Valid redirect, to a server that is one of the known locations. + final boolean isRedirectLoop = redirectLocations.contains(newUri); + final boolean isRedirectChainTooLong = redirectLocations.size() >= MAX_REDIRECTS; + + if (isRedirectLoop || isRedirectChainTooLong) { + // Treat redirect loops, or too-long redirect chains, as unavailable services. + if (retryPolicy.retryNotAvailable() && shouldTry(nextAttemptNumber)) { + final long backoffMs = computeBackoffMs(retryPolicy, attemptNumber); + + log.info( + "Service [%s] issued too many redirects on attempt #%d; retrying in %,d ms.", + serviceName, + nextAttemptNumber, + backoffMs + ); + + connectExec.schedule( + () -> tryRequest(requestBuilder, handler, retVal, nextAttemptNumber, ImmutableSet.of()), + backoffMs, + TimeUnit.MILLISECONDS + ); + } else { + retVal.setException(new ServiceNotAvailableException(serviceName, "issued too many redirects")); + } + } else { + // Valid redirect. Follow it without incrementing the attempt number. + preferredLocationNoPath.set(redirectLocationNoPath); + final ImmutableSet newRedirectLocations = + ImmutableSet.builder().addAll(redirectLocations).add(newUri).build(); + connectExec.submit( + () -> tryRequest(requestBuilder, handler, retVal, attemptNumber, newRedirectLocations) + ); + } + } else { + // Redirect to a server that is not one of the known locations. Treat service as unavailable. + if (retryPolicy.retryNotAvailable() && shouldTry(nextAttemptNumber)) { + final long backoffMs = computeBackoffMs(retryPolicy, attemptNumber); + + log.info( + "Service [%s] issued redirect to unknown URL [%s] on attempt #%d; retrying in %,d ms.", + serviceName, + newUri, + nextAttemptNumber, + backoffMs + ); + + connectExec.schedule( + () -> tryRequest(requestBuilder, handler, retVal, nextAttemptNumber, ImmutableSet.of()), + backoffMs, + TimeUnit.MILLISECONDS + ); + } else { + retVal.setException( + new ServiceNotAvailableException( + serviceName, + "issued redirect to unknown URL [%s]", + newUri + ) + ); + } + } + } }, connectExec ); @@ -285,7 +373,7 @@ public void onFailure(final Throwable t) ); } - private void whenServiceReady(final Consumer callback, final SettableFuture retVal) + private void whenServiceReady(final Consumer callback, final SettableFuture retVal) { Futures.addCallback( serviceLocator.locate(), @@ -300,8 +388,7 @@ public void onSuccess(final ServiceLocations locations) } try { - final ServiceLocation location = pick(locations); - callback.accept(location); + callback.accept(locations); } catch (Throwable t) { // It's a bug if this happens. The purpose of this line is to help us debug what went wrong. @@ -328,10 +415,7 @@ private ServiceLocation pick(final ServiceLocations locations) if (preferred != null) { // Preferred location is set. Use it if it's one of known locations. for (final ServiceLocation location : locations.getLocations()) { - final ServiceLocation locationNoPath = - new ServiceLocation(location.getHost(), location.getPlaintextPort(), location.getTlsPort(), ""); - - if (locationNoPath.equals(preferred)) { + if (serviceLocationNoPath(location).equals(preferred)) { return location; } } @@ -392,6 +476,9 @@ static long computeBackoffMs(final ServiceRetryPolicy retryPolicy, final long at ); } + /** + * Returns a {@link ServiceLocation} without a path component, based on a URI. + */ @Nullable @VisibleForTesting static ServiceLocation serviceLocationNoPathFromUri(@Nullable final String uriString) @@ -423,6 +510,14 @@ static ServiceLocation serviceLocationNoPathFromUri(@Nullable final String uriSt } } + /** + * Returns a {@link ServiceLocation} without its path. + */ + static ServiceLocation serviceLocationNoPath(final ServiceLocation location) + { + return new ServiceLocation(location.getHost(), location.getPlaintextPort(), location.getTlsPort(), ""); + } + @VisibleForTesting static boolean isRedirect(final HttpResponseStatus responseStatus) { diff --git a/server/src/main/java/org/apache/druid/rpc/ServiceNotAvailableException.java b/server/src/main/java/org/apache/druid/rpc/ServiceNotAvailableException.java index 290fa2cf7ab7..843ac2dfa54f 100644 --- a/server/src/main/java/org/apache/druid/rpc/ServiceNotAvailableException.java +++ b/server/src/main/java/org/apache/druid/rpc/ServiceNotAvailableException.java @@ -19,13 +19,20 @@ package org.apache.druid.rpc; +import org.apache.druid.java.util.common.StringUtils; + /** * Returned by {@link ServiceClient#asyncRequest} when a request has failed because the service is not available. */ public class ServiceNotAvailableException extends RpcException { + public ServiceNotAvailableException(final String serviceName, final String reason, final Object... reasonArgs) + { + super("Service [%s] %s", serviceName, StringUtils.format(reason, reasonArgs)); + } + public ServiceNotAvailableException(final String serviceName) { - super("Service [%s] is not available", serviceName); + this(serviceName, "is not available"); } } diff --git a/server/src/test/java/org/apache/druid/rpc/ServiceClientImplTest.java b/server/src/test/java/org/apache/druid/rpc/ServiceClientImplTest.java index 48f922e6ab6b..dc8bba87d139 100644 --- a/server/src/test/java/org/apache/druid/rpc/ServiceClientImplTest.java +++ b/server/src/test/java/org/apache/druid/rpc/ServiceClientImplTest.java @@ -67,6 +67,9 @@ public class ServiceClientImplTest private static final String SERVICE_NAME = "test-service"; private static final ServiceLocation SERVER1 = new ServiceLocation("example.com", -1, 8888, "/q"); private static final ServiceLocation SERVER2 = new ServiceLocation("example.com", -1, 9999, "/q"); + private static final ServiceLocation SERVER3 = new ServiceLocation("example.com", -1, 1111, "/q"); + private static final ServiceLocation SERVER4 = new ServiceLocation("example.com", -1, 2222, "/q"); + private static final ServiceLocation SERVER5 = new ServiceLocation("example.com", -1, 3333, "/q"); private ScheduledExecutorService exec; @@ -270,7 +273,63 @@ public void test_request_followRedirect() throws Exception } @Test - public void test_request_tooManyRedirects() + public void test_request_tooLongRedirectChain() + { + final RequestBuilder requestBuilder = new RequestBuilder(HttpMethod.GET, "/foo"); + + // Redirect chain longer than max length. + stubLocatorCall(locations(SERVER1, SERVER2, SERVER3, SERVER4, SERVER5)); + expectHttpCall(requestBuilder, SERVER1) + .thenReturn(redirectResponse(requestBuilder.build(SERVER2).getUrl().toString())); + expectHttpCall(requestBuilder, SERVER2) + .thenReturn(redirectResponse(requestBuilder.build(SERVER3).getUrl().toString())); + expectHttpCall(requestBuilder, SERVER3) + .thenReturn(redirectResponse(requestBuilder.build(SERVER4).getUrl().toString())); + expectHttpCall(requestBuilder, SERVER4) + .thenReturn(redirectResponse(requestBuilder.build(SERVER5).getUrl().toString())); + + serviceClient = makeServiceClient(StandardRetryPolicy.noRetries()); + + final ExecutionException e = Assert.assertThrows( + ExecutionException.class, + () -> doRequest(serviceClient, requestBuilder) + ); + + MatcherAssert.assertThat(e.getCause(), CoreMatchers.instanceOf(ServiceNotAvailableException.class)); + MatcherAssert.assertThat( + e.getCause(), + ThrowableMessageMatcher.hasMessage(CoreMatchers.containsString("issued too many redirects")) + ); + } + + @Test + public void test_request_tooLongRedirectChainRetry() throws Exception + { + final RequestBuilder requestBuilder = new RequestBuilder(HttpMethod.GET, "/foo"); + final ImmutableMap expectedResponseObject = ImmutableMap.of("foo", "bar"); + + // Redirect chain longer than max length. Can be followed across retries. + stubLocatorCall(locations(SERVER1, SERVER2, SERVER3, SERVER4, SERVER5)); + expectHttpCall(requestBuilder, SERVER1) + .thenReturn(redirectResponse(requestBuilder.build(SERVER2).getUrl().toString())); + expectHttpCall(requestBuilder, SERVER2) + .thenReturn(redirectResponse(requestBuilder.build(SERVER3).getUrl().toString())); + expectHttpCall(requestBuilder, SERVER3) + .thenReturn(redirectResponse(requestBuilder.build(SERVER4).getUrl().toString())); + expectHttpCall(requestBuilder, SERVER4) + .thenReturn(redirectResponse(requestBuilder.build(SERVER5).getUrl().toString())); + expectHttpCall(requestBuilder, SERVER5) + .thenReturn(valueResponse(expectedResponseObject)); + + serviceClient = makeServiceClient(StandardRetryPolicy.builder().maxAttempts(2).build()); + + final Map response = doRequest(serviceClient, requestBuilder); + + Assert.assertEquals(expectedResponseObject, response); + } + + @Test + public void test_request_selfRedirectLoop() { final RequestBuilder requestBuilder = new RequestBuilder(HttpMethod.GET, "/foo"); @@ -279,17 +338,43 @@ public void test_request_tooManyRedirects() expectHttpCall(requestBuilder, SERVER1) .thenReturn(redirectResponse(requestBuilder.build(SERVER1).getUrl().toString())); - serviceClient = makeServiceClient(StandardRetryPolicy.unlimited()); + serviceClient = makeServiceClient(StandardRetryPolicy.builder().maxAttempts(10).build()); final ExecutionException e = Assert.assertThrows( ExecutionException.class, () -> doRequest(serviceClient, requestBuilder) ); - MatcherAssert.assertThat(e.getCause(), CoreMatchers.instanceOf(RpcException.class)); + MatcherAssert.assertThat(e.getCause(), CoreMatchers.instanceOf(ServiceNotAvailableException.class)); MatcherAssert.assertThat( e.getCause(), - ThrowableMessageMatcher.hasMessage(CoreMatchers.containsString("redirected too many times")) + ThrowableMessageMatcher.hasMessage(CoreMatchers.containsString("issued too many redirects")) + ); + } + + @Test + public void test_request_twoServerRedirectLoop() + { + final RequestBuilder requestBuilder = new RequestBuilder(HttpMethod.GET, "/foo"); + + // Endless redirects between the same two servers. + stubLocatorCall(locations(SERVER1, SERVER2)); + expectHttpCall(requestBuilder, SERVER1) + .thenReturn(redirectResponse(requestBuilder.build(SERVER2).getUrl().toString())); + expectHttpCall(requestBuilder, SERVER2) + .thenReturn(redirectResponse(requestBuilder.build(SERVER1).getUrl().toString())); + + serviceClient = makeServiceClient(StandardRetryPolicy.builder().maxAttempts(10).build()); + + final ExecutionException e = Assert.assertThrows( + ExecutionException.class, + () -> doRequest(serviceClient, requestBuilder) + ); + + MatcherAssert.assertThat(e.getCause(), CoreMatchers.instanceOf(ServiceNotAvailableException.class)); + MatcherAssert.assertThat( + e.getCause(), + ThrowableMessageMatcher.hasMessage(CoreMatchers.containsString("issued too many redirects")) ); } @@ -314,7 +399,7 @@ public void test_request_redirectInvalid() MatcherAssert.assertThat( e.getCause(), ThrowableMessageMatcher.hasMessage( - CoreMatchers.containsString("redirected [0] times to invalid URL [invalid-url]")) + CoreMatchers.containsString("redirected to invalid URL [invalid-url]")) ); } @@ -338,7 +423,7 @@ public void test_request_redirectNil() MatcherAssert.assertThat(e.getCause(), CoreMatchers.instanceOf(RpcException.class)); MatcherAssert.assertThat( e.getCause(), - ThrowableMessageMatcher.hasMessage(CoreMatchers.containsString("redirected [0] times to invalid URL [null]")) + ThrowableMessageMatcher.hasMessage(CoreMatchers.containsString("redirected to invalid URL [null]")) ); } @@ -359,10 +444,11 @@ public void test_request_dontFollowRedirectToUnknownServer() () -> doRequest(serviceClient, requestBuilder) ); - MatcherAssert.assertThat(e.getCause(), CoreMatchers.instanceOf(RpcException.class)); + MatcherAssert.assertThat(e.getCause(), CoreMatchers.instanceOf(ServiceNotAvailableException.class)); MatcherAssert.assertThat( e.getCause(), - ThrowableMessageMatcher.hasMessage(CoreMatchers.containsString("redirected too many times")) + ThrowableMessageMatcher.hasMessage(CoreMatchers.containsString( + "issued redirect to unknown URL [https://example.com:9999/q/foo]")) ); } From b12e5f300e7c2795ba3d9c7ef17fb64f4925b9c0 Mon Sep 17 00:00:00 2001 From: Tejaswini Bandlamudi <96047043+tejaswini-imply@users.noreply.github.com> Date: Mon, 28 Nov 2022 23:04:33 +0530 Subject: [PATCH 046/141] Add filter in cloud object input source for backward compatibility (#13437) https://github.com/apache/druid/pull/13027 PR replaces `filter` parameter with `objectGlob` in ingestion input source. However, this will cause existing ingestion jobs to fail if they are using a filter already. This PR adds old filter functionality alongside objectGlob to preserve backward compatibility. --- .../input/impl/CloudObjectInputSource.java | 32 ++++++++++- .../impl/CloudObjectInputSourceTest.java | 45 +++++++++++---- .../data/input/aliyun/OssInputSource.java | 4 +- .../data/input/aliyun/OssInputSourceTest.java | 17 ++++++ .../data/input/azure/AzureInputSource.java | 4 +- .../input/azure/AzureInputSourceTest.java | 55 +++++++++++++++++++ .../google/GoogleCloudStorageInputSource.java | 5 +- .../GoogleCloudStorageInputSourceTest.java | 36 ++++++++++-- .../druid/data/input/s3/S3InputSource.java | 6 +- 9 files changed, 181 insertions(+), 23 deletions(-) diff --git a/core/src/main/java/org/apache/druid/data/input/impl/CloudObjectInputSource.java b/core/src/main/java/org/apache/druid/data/input/impl/CloudObjectInputSource.java index 3c3e3c6b72d4..716e5b611ceb 100644 --- a/core/src/main/java/org/apache/druid/data/input/impl/CloudObjectInputSource.java +++ b/core/src/main/java/org/apache/druid/data/input/impl/CloudObjectInputSource.java @@ -21,7 +21,9 @@ import com.fasterxml.jackson.annotation.JsonInclude; import com.fasterxml.jackson.annotation.JsonProperty; +import com.google.common.base.Preconditions; import com.google.common.primitives.Ints; +import org.apache.commons.io.FilenameUtils; import org.apache.commons.lang.StringUtils; import org.apache.druid.data.input.AbstractInputSource; import org.apache.druid.data.input.InputEntity; @@ -50,6 +52,13 @@ public abstract class CloudObjectInputSource extends AbstractInputSource private final List uris; private final List prefixes; private final List objects; + + /** + * Preserved filter for backward compatibility, should be removed on next major release; + * use objectGlob instead. + */ + @Deprecated + private final String filter; private final String objectGlob; public CloudObjectInputSource( @@ -63,6 +72,7 @@ public CloudObjectInputSource( this.uris = uris; this.prefixes = prefixes; this.objects = objects; + this.filter = null; this.objectGlob = null; illegalArgsChecker(); @@ -73,6 +83,7 @@ public CloudObjectInputSource( @Nullable List uris, @Nullable List prefixes, @Nullable List objects, + @Deprecated @Nullable String filter, @Nullable String objectGlob ) { @@ -80,8 +91,12 @@ public CloudObjectInputSource( this.uris = uris; this.prefixes = prefixes; this.objects = objects; + this.filter = filter; this.objectGlob = objectGlob; - + Preconditions.checkArgument( + filter == null || objectGlob == null, + "Cannot use filter and objectGlob together. Try using objectGlob instead of filter." + ); illegalArgsChecker(); } @@ -107,6 +122,14 @@ public List getObjects() return objects; } + @Nullable + @JsonProperty + @JsonInclude(JsonInclude.Include.NON_NULL) + public String getFilter() + { + return filter; + } + @Nullable @JsonProperty @JsonInclude(JsonInclude.Include.NON_NULL) @@ -144,6 +167,8 @@ public Stream>> createSplits( if (StringUtils.isNotBlank(objectGlob)) { PathMatcher m = FileSystems.getDefault().getPathMatcher("glob:" + getObjectGlob()); objectStream = objectStream.filter(object -> m.matches(Paths.get(object.getPath()))); + } else if (StringUtils.isNotBlank(filter)) { + objectStream = objectStream.filter(object -> FilenameUtils.wildcardMatch(object.getPath(), getFilter())); } return objectStream.map(object -> new InputSplit<>(Collections.singletonList(object))); @@ -155,6 +180,8 @@ public Stream>> createSplits( if (StringUtils.isNotBlank(objectGlob)) { PathMatcher m = FileSystems.getDefault().getPathMatcher("glob:" + getObjectGlob()); uriStream = uriStream.filter(uri -> m.matches(Paths.get(uri.toString()))); + } else if (StringUtils.isNotBlank(filter)) { + uriStream = uriStream.filter(uri -> FilenameUtils.wildcardMatch(uri.toString(), filter)); } return uriStream.map(CloudObjectLocation::new).map(object -> new InputSplit<>(Collections.singletonList(object))); @@ -212,13 +239,14 @@ public boolean equals(Object o) Objects.equals(uris, that.uris) && Objects.equals(prefixes, that.prefixes) && Objects.equals(objects, that.objects) && + Objects.equals(filter, that.filter) && Objects.equals(objectGlob, that.objectGlob); } @Override public int hashCode() { - return Objects.hash(scheme, uris, prefixes, objects, objectGlob); + return Objects.hash(scheme, uris, prefixes, objects, filter, objectGlob); } private void illegalArgsChecker() throws IllegalArgumentException diff --git a/core/src/test/java/org/apache/druid/data/input/impl/CloudObjectInputSourceTest.java b/core/src/test/java/org/apache/druid/data/input/impl/CloudObjectInputSourceTest.java index 539581403989..da421b1b1676 100644 --- a/core/src/test/java/org/apache/druid/data/input/impl/CloudObjectInputSourceTest.java +++ b/core/src/test/java/org/apache/druid/data/input/impl/CloudObjectInputSourceTest.java @@ -72,7 +72,7 @@ public class CloudObjectInputSourceTest public void testGetUris() { CloudObjectInputSource inputSource = Mockito.mock(CloudObjectInputSource.class, Mockito.withSettings() - .useConstructor(SCHEME, URIS, null, null, null) + .useConstructor(SCHEME, URIS, null, null, null, null) .defaultAnswer(Mockito.CALLS_REAL_METHODS) ); @@ -86,7 +86,7 @@ public void testGetUris() public void testGetPrefixes() { CloudObjectInputSource inputSource = Mockito.mock(CloudObjectInputSource.class, Mockito.withSettings() - .useConstructor(SCHEME, null, PREFIXES, null, null) + .useConstructor(SCHEME, null, PREFIXES, null, null, null) .defaultAnswer(Mockito.CALLS_REAL_METHODS) ); @@ -100,7 +100,7 @@ public void testGetPrefixes() public void testGetObjectGlob() { CloudObjectInputSource inputSource = Mockito.mock(CloudObjectInputSource.class, Mockito.withSettings() - .useConstructor(SCHEME, URIS, null, null, "**.parquet") + .useConstructor(SCHEME, URIS, null, null, null, "**.parquet") .defaultAnswer(Mockito.CALLS_REAL_METHODS) ); @@ -111,12 +111,12 @@ public void testGetObjectGlob() public void testInequality() { CloudObjectInputSource inputSource1 = Mockito.mock(CloudObjectInputSource.class, Mockito.withSettings() - .useConstructor(SCHEME, URIS, null, null, "**.parquet") + .useConstructor(SCHEME, URIS, null, null, null, "**.parquet") .defaultAnswer(Mockito.CALLS_REAL_METHODS) ); CloudObjectInputSource inputSource2 = Mockito.mock(CloudObjectInputSource.class, Mockito.withSettings() - .useConstructor(SCHEME, URIS, null, null, "**.csv") + .useConstructor(SCHEME, URIS, null, null, null, "**.csv") .defaultAnswer(Mockito.CALLS_REAL_METHODS) ); @@ -128,9 +128,32 @@ public void testInequality() @Test public void testWithUrisFilter() { - CloudObjectInputSource inputSource = Mockito.mock(CloudObjectInputSource.class, Mockito.withSettings() - .useConstructor(SCHEME, URIS2, null, null, "**.csv") - .defaultAnswer(Mockito.CALLS_REAL_METHODS) + CloudObjectInputSource inputSource = Mockito.mock(CloudObjectInputSource.class, + Mockito.withSettings() + .useConstructor(SCHEME, URIS2, null, null, "*.csv", null) + .defaultAnswer(Mockito.CALLS_REAL_METHODS) + ); + + Stream>> splits = inputSource.createSplits( + new JsonInputFormat(JSONPathSpec.DEFAULT, null, null, null, null), + new MaxSizeSplitHintSpec(null, 1) + ); + + List returnedLocations = splits.map(InputSplit::get).collect(Collectors.toList()).get(0); + + List returnedLocationUris = returnedLocations.stream().map(object -> object.toUri(SCHEME)).collect(Collectors.toList()); + + Assert.assertEquals("*.csv", inputSource.getFilter()); + Assert.assertEquals(URIS, returnedLocationUris); + } + + @Test + public void testWithUrisObjectGlob() + { + CloudObjectInputSource inputSource = Mockito.mock(CloudObjectInputSource.class, + Mockito.withSettings() + .useConstructor(SCHEME, URIS2, null, null, null, "**.csv") + .defaultAnswer(Mockito.CALLS_REAL_METHODS) ); Stream>> splits = inputSource.createSplits( @@ -150,7 +173,7 @@ public void testWithUrisFilter() public void testWithUris() { CloudObjectInputSource inputSource = Mockito.mock(CloudObjectInputSource.class, Mockito.withSettings() - .useConstructor(SCHEME, URIS, null, null, null) + .useConstructor(SCHEME, URIS, null, null, null, null) .defaultAnswer(Mockito.CALLS_REAL_METHODS) ); @@ -171,7 +194,7 @@ public void testWithUris() public void testWithObjectsFilter() { CloudObjectInputSource inputSource = Mockito.mock(CloudObjectInputSource.class, Mockito.withSettings() - .useConstructor(SCHEME, null, null, OBJECTS_BEFORE_GLOB, "**.csv") + .useConstructor(SCHEME, null, null, OBJECTS_BEFORE_GLOB, null, "**.csv") .defaultAnswer(Mockito.CALLS_REAL_METHODS) ); @@ -192,7 +215,7 @@ public void testWithObjectsFilter() public void testWithObjects() { CloudObjectInputSource inputSource = Mockito.mock(CloudObjectInputSource.class, Mockito.withSettings() - .useConstructor(SCHEME, null, null, OBJECTS, null) + .useConstructor(SCHEME, null, null, OBJECTS, null, null) .defaultAnswer(Mockito.CALLS_REAL_METHODS) ); diff --git a/extensions-contrib/aliyun-oss-extensions/src/main/java/org/apache/druid/data/input/aliyun/OssInputSource.java b/extensions-contrib/aliyun-oss-extensions/src/main/java/org/apache/druid/data/input/aliyun/OssInputSource.java index 71edd26b590b..a90e96f047b8 100644 --- a/extensions-contrib/aliyun-oss-extensions/src/main/java/org/apache/druid/data/input/aliyun/OssInputSource.java +++ b/extensions-contrib/aliyun-oss-extensions/src/main/java/org/apache/druid/data/input/aliyun/OssInputSource.java @@ -79,11 +79,12 @@ public OssInputSource( @JsonProperty("uris") @Nullable List uris, @JsonProperty("prefixes") @Nullable List prefixes, @JsonProperty("objects") @Nullable List objects, + @Deprecated @JsonProperty("filter") @Nullable String filter, @JsonProperty("objectGlob") @Nullable String objectGlob, @JsonProperty("properties") @Nullable OssClientConfig inputSourceConfig ) { - super(OssStorageDruidModule.SCHEME, uris, prefixes, objects, objectGlob); + super(OssStorageDruidModule.SCHEME, uris, prefixes, objects, filter, objectGlob); this.inputDataConfig = Preconditions.checkNotNull(inputDataConfig, "inputDataConfig"); Preconditions.checkNotNull(client, "client"); this.inputSourceConfig = inputSourceConfig; @@ -136,6 +137,7 @@ public SplittableInputSource> withSplit(InputSplit uris, @JsonProperty("prefixes") @Nullable List prefixes, @JsonProperty("objects") @Nullable List objects, + @Deprecated @JsonProperty("filter") @Nullable String filter, @JsonProperty("objectGlob") @Nullable String objectGlob ) { - super(SCHEME, uris, prefixes, objects, objectGlob); + super(SCHEME, uris, prefixes, objects, filter, objectGlob); this.storage = Preconditions.checkNotNull(storage, "AzureStorage"); this.entityFactory = Preconditions.checkNotNull(entityFactory, "AzureEntityFactory"); this.azureCloudBlobIterableFactory = Preconditions.checkNotNull( @@ -103,6 +104,7 @@ public SplittableInputSource> withSplit(InputSplit>> cloudObjectStream = azureInputSource.getPrefixesSplitStream( + new MaxSizeSplitHintSpec(null, 1) + ); + + List> actualCloudLocationList = cloudObjectStream.map(InputSplit::get) + .collect(Collectors.toList()); + verifyAll(); + Assert.assertEquals(expectedCloudLocations, actualCloudLocationList); + } + + @Test + public void test_getPrefixesSplitStream_withFilter_successfullyCreatesCloudLocation_returnsExpectedLocations() + { + List prefixes = ImmutableList.of(PREFIX_URI); + List> expectedCloudLocations = ImmutableList.of(ImmutableList.of(CLOUD_OBJECT_LOCATION_1)); + List expectedCloudBlobs = ImmutableList.of(cloudBlobDruid1); + Iterator expectedCloudBlobsIterator = expectedCloudBlobs.iterator(); + String filter = "*.csv"; + + expectedCloudBlobsIterator = Iterators.filter( + expectedCloudBlobsIterator, + object -> FilenameUtils.wildcardMatch(object.getName(), filter) + ); + + EasyMock.expect(inputDataConfig.getMaxListingLength()).andReturn(MAX_LISTING_LENGTH); + EasyMock.expect(azureCloudBlobIterableFactory.create(prefixes, MAX_LISTING_LENGTH)).andReturn( + azureCloudBlobIterable); + EasyMock.expect(azureCloudBlobIterable.iterator()).andReturn(expectedCloudBlobsIterator); + EasyMock.expect(azureCloudBlobToLocationConverter.createCloudObjectLocation(cloudBlobDruid1)) + .andReturn(CLOUD_OBJECT_LOCATION_1); + EasyMock.expect(cloudBlobDruid1.getBlobLength()).andReturn(100L).anyTimes(); + EasyMock.expect(cloudBlobDruid1.getName()).andReturn(BLOB_PATH).anyTimes(); + + replayAll(); + + azureInputSource = new AzureInputSource( + storage, + entityFactory, + azureCloudBlobIterableFactory, + azureCloudBlobToLocationConverter, + inputDataConfig, + EMPTY_URIS, + prefixes, + EMPTY_OBJECTS, + filter, null ); @@ -215,6 +267,7 @@ public void test_getPrefixesSplitStream_withObjectGlob_successfullyCreatesCloudL EMPTY_URIS, prefixes, EMPTY_OBJECTS, + null, objectGlob ); @@ -244,6 +297,7 @@ public void test_withSplit_constructsExpectedInputSource() EMPTY_URIS, prefixes, EMPTY_OBJECTS, + null, null ); @@ -265,6 +319,7 @@ public void test_toString_returnsExpectedString() EMPTY_URIS, prefixes, EMPTY_OBJECTS, + null, null ); diff --git a/extensions-core/google-extensions/src/main/java/org/apache/druid/data/input/google/GoogleCloudStorageInputSource.java b/extensions-core/google-extensions/src/main/java/org/apache/druid/data/input/google/GoogleCloudStorageInputSource.java index a4d175aab2c0..009bae7dbe2e 100644 --- a/extensions-core/google-extensions/src/main/java/org/apache/druid/data/input/google/GoogleCloudStorageInputSource.java +++ b/extensions-core/google-extensions/src/main/java/org/apache/druid/data/input/google/GoogleCloudStorageInputSource.java @@ -65,10 +65,11 @@ public GoogleCloudStorageInputSource( @JsonProperty("uris") @Nullable List uris, @JsonProperty("prefixes") @Nullable List prefixes, @JsonProperty("objects") @Nullable List objects, + @Deprecated @JsonProperty("filter") @Nullable String filter, @JsonProperty("objectGlob") @Nullable String objectGlob ) { - super(GoogleStorageDruidModule.SCHEME_GS, uris, prefixes, objects, objectGlob); + super(GoogleStorageDruidModule.SCHEME_GS, uris, prefixes, objects, filter, objectGlob); this.storage = storage; this.inputDataConfig = inputDataConfig; } @@ -117,7 +118,7 @@ protected Stream>> getPrefixesSplitStream(@ @Override public SplittableInputSource> withSplit(InputSplit> split) { - return new GoogleCloudStorageInputSource(storage, inputDataConfig, null, null, split.get(), getObjectGlob()); + return new GoogleCloudStorageInputSource(storage, inputDataConfig, null, null, split.get(), getFilter(), getObjectGlob()); } private CloudObjectLocation byteSourceFromStorageObject(final StorageObject storageObject) diff --git a/extensions-core/google-extensions/src/test/java/org/apache/druid/data/input/google/GoogleCloudStorageInputSourceTest.java b/extensions-core/google-extensions/src/test/java/org/apache/druid/data/input/google/GoogleCloudStorageInputSourceTest.java index c0648e7f4c85..f15b6280cc85 100644 --- a/extensions-core/google-extensions/src/test/java/org/apache/druid/data/input/google/GoogleCloudStorageInputSourceTest.java +++ b/extensions-core/google-extensions/src/test/java/org/apache/druid/data/input/google/GoogleCloudStorageInputSourceTest.java @@ -116,7 +116,7 @@ public void testSerde() throws Exception { final ObjectMapper mapper = createGoogleObjectMapper(); final GoogleCloudStorageInputSource withUris = - new GoogleCloudStorageInputSource(STORAGE, INPUT_DATA_CONFIG, EXPECTED_URIS, ImmutableList.of(), null, null); + new GoogleCloudStorageInputSource(STORAGE, INPUT_DATA_CONFIG, EXPECTED_URIS, ImmutableList.of(), null, null, null); final GoogleCloudStorageInputSource serdeWithUris = mapper.readValue(mapper.writeValueAsString(withUris), GoogleCloudStorageInputSource.class); Assert.assertEquals(withUris, serdeWithUris); @@ -127,7 +127,7 @@ public void testSerdePrefixes() throws Exception { final ObjectMapper mapper = createGoogleObjectMapper(); final GoogleCloudStorageInputSource withPrefixes = - new GoogleCloudStorageInputSource(STORAGE, INPUT_DATA_CONFIG, ImmutableList.of(), PREFIXES, null, null); + new GoogleCloudStorageInputSource(STORAGE, INPUT_DATA_CONFIG, ImmutableList.of(), PREFIXES, null, null, null); final GoogleCloudStorageInputSource serdeWithPrefixes = mapper.readValue(mapper.writeValueAsString(withPrefixes), GoogleCloudStorageInputSource.class); Assert.assertEquals(withPrefixes, serdeWithPrefixes); @@ -144,6 +144,7 @@ public void testSerdeObjects() throws Exception null, null, ImmutableList.of(new CloudObjectLocation("foo", "bar/file.gz")), + null, null ); final GoogleCloudStorageInputSource serdeWithObjects = @@ -156,7 +157,7 @@ public void testWithUrisSplit() { GoogleCloudStorageInputSource inputSource = - new GoogleCloudStorageInputSource(STORAGE, INPUT_DATA_CONFIG, EXPECTED_URIS, ImmutableList.of(), null, null); + new GoogleCloudStorageInputSource(STORAGE, INPUT_DATA_CONFIG, EXPECTED_URIS, ImmutableList.of(), null, null, null); Stream>> splits = inputSource.createSplits( new JsonInputFormat(JSONPathSpec.DEFAULT, null, null, null, null), @@ -174,6 +175,7 @@ public void testWithUrisGlob() URIS_BEFORE_GLOB, null, null, + null, "**.csv" ); @@ -184,6 +186,26 @@ public void testWithUrisGlob() Assert.assertEquals(EXPECTED_OBJECTS, splits.map(InputSplit::get).collect(Collectors.toList())); } + @Test + public void testWithUrisFilter() + { + GoogleCloudStorageInputSource inputSource = new GoogleCloudStorageInputSource( + STORAGE, + INPUT_DATA_CONFIG, + URIS_BEFORE_GLOB, + null, + null, + "*.csv", + null + ); + + Stream>> splits = inputSource.createSplits( + new JsonInputFormat(JSONPathSpec.DEFAULT, null, null, null, null), + null + ); + Assert.assertEquals(EXPECTED_OBJECTS, splits.map(InputSplit::get).collect(Collectors.toList())); + } + @Test public void testIllegalObjectsAndPrefixes() { @@ -195,6 +217,7 @@ public void testIllegalObjectsAndPrefixes() null, PREFIXES, EXPECTED_OBJECTS.get(0), + null, "**.csv" ); } @@ -210,6 +233,7 @@ public void testIllegalUrisAndPrefixes() URIS_BEFORE_GLOB, PREFIXES, null, + null, "**.csv" ); } @@ -226,7 +250,7 @@ public void testWithPrefixesSplit() throws IOException EasyMock.replay(INPUT_DATA_CONFIG); GoogleCloudStorageInputSource inputSource = - new GoogleCloudStorageInputSource(STORAGE, INPUT_DATA_CONFIG, null, PREFIXES, null, null); + new GoogleCloudStorageInputSource(STORAGE, INPUT_DATA_CONFIG, null, PREFIXES, null, null, null); Stream>> splits = inputSource.createSplits( new JsonInputFormat(JSONPathSpec.DEFAULT, null, null, null, null), @@ -248,7 +272,7 @@ public void testCreateSplitsWithSplitHintSpecRespectingHint() throws IOException EasyMock.replay(INPUT_DATA_CONFIG); GoogleCloudStorageInputSource inputSource = - new GoogleCloudStorageInputSource(STORAGE, INPUT_DATA_CONFIG, null, PREFIXES, null, null); + new GoogleCloudStorageInputSource(STORAGE, INPUT_DATA_CONFIG, null, PREFIXES, null, null, null); Stream>> splits = inputSource.createSplits( new JsonInputFormat(JSONPathSpec.DEFAULT, null, null, null, null), @@ -280,6 +304,7 @@ public void testReader() throws IOException null, PREFIXES, null, + null, null ); @@ -324,6 +349,7 @@ public void testCompressedReader() throws IOException null, PREFIXES, null, + null, null ); diff --git a/extensions-core/s3-extensions/src/main/java/org/apache/druid/data/input/s3/S3InputSource.java b/extensions-core/s3-extensions/src/main/java/org/apache/druid/data/input/s3/S3InputSource.java index 4d0d4f0a9472..c0ee349ec77a 100644 --- a/extensions-core/s3-extensions/src/main/java/org/apache/druid/data/input/s3/S3InputSource.java +++ b/extensions-core/s3-extensions/src/main/java/org/apache/druid/data/input/s3/S3InputSource.java @@ -111,6 +111,7 @@ public S3InputSource( @JsonProperty("uris") @Nullable List uris, @JsonProperty("prefixes") @Nullable List prefixes, @JsonProperty("objects") @Nullable List objects, + @Deprecated @JsonProperty("filter") @Nullable String filter, @JsonProperty("objectGlob") @Nullable String objectGlob, @JsonProperty("properties") @Nullable S3InputSourceConfig s3InputSourceConfig, @JsonProperty("proxyConfig") @Nullable AWSProxyConfig awsProxyConfig, @@ -118,7 +119,7 @@ public S3InputSource( @JsonProperty("clientConfig") @Nullable AWSClientConfig awsClientConfig ) { - super(S3StorageDruidModule.SCHEME, uris, prefixes, objects, objectGlob); + super(S3StorageDruidModule.SCHEME, uris, prefixes, objects, filter, objectGlob); this.inputDataConfig = Preconditions.checkNotNull(inputDataConfig, "S3DataSegmentPusherConfig"); Preconditions.checkNotNull(s3Client, "s3Client"); this.s3InputSourceConfig = s3InputSourceConfig; @@ -202,6 +203,7 @@ public S3InputSource( uris, prefixes, objects, + null, objectGlob, s3InputSourceConfig, awsProxyConfig, @@ -234,6 +236,7 @@ public S3InputSource( uris, prefixes, objects, + null, objectGlob, s3InputSourceConfig, awsProxyConfig, @@ -343,6 +346,7 @@ public SplittableInputSource> withSplit(InputSplit Date: Mon, 28 Nov 2022 23:13:19 +0530 Subject: [PATCH 047/141] 256m instead of 256g --- examples/bin/start-druid | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/examples/bin/start-druid b/examples/bin/start-druid index 07f3fdee673e..1bdd0c58a249 100644 --- a/examples/bin/start-druid +++ b/examples/bin/start-druid @@ -41,7 +41,7 @@ TASK_COUNT = "task-count" TASK_MEM_TYPE_LOW = "low" TASK_MEM_TYPE_HIGH = "high" TASK_MEM_MAP = { - TASK_MEM_TYPE_LOW: ["-Xms256m", "-Xmx256m", "-XX:MaxDirectMemorySize=256g"], + TASK_MEM_TYPE_LOW: ["-Xms256m", "-Xmx256m", "-XX:MaxDirectMemorySize=256m"], TASK_MEM_TYPE_HIGH: ["-Xms1g", "-Xmx1g", "-XX:MaxDirectMemorySize=1g"] } From a2d5e335f338fd58a84950633bbc3232210091c9 Mon Sep 17 00:00:00 2001 From: Vadim Ogievetsky Date: Mon, 28 Nov 2022 11:40:45 -0800 Subject: [PATCH 048/141] Web console: Index spec dialog (#13425) * add index spec dialog * add sanpshot --- .../src/components/auto-form/auto-form.scss | 4 + .../src/components/auto-form/auto-form.tsx | 55 ++- .../index-spec-dialog.spec.tsx.snap | 317 ++++++++++++++++++ .../index-spec-dialog/index-spec-dialog.scss | 36 ++ .../index-spec-dialog.spec.tsx | 44 +++ .../index-spec-dialog/index-spec-dialog.tsx | 88 +++++ .../druid-models/index-spec/index-spec.tsx | 158 +++++++++ web-console/src/druid-models/index.ts | 1 + .../ingestion-spec/ingestion-spec.tsx | 120 ++----- .../workbench-query/workbench-query-part.ts | 4 +- .../workbench-query/workbench-query.spec.ts | 2 +- .../src/helpers/spec-conversion.spec.ts | 6 + web-console/src/helpers/spec-conversion.ts | 5 + .../workbench-view/run-panel/run-panel.tsx | 130 ++++--- 14 files changed, 798 insertions(+), 172 deletions(-) create mode 100644 web-console/src/dialogs/index-spec-dialog/__snapshots__/index-spec-dialog.spec.tsx.snap create mode 100644 web-console/src/dialogs/index-spec-dialog/index-spec-dialog.scss create mode 100644 web-console/src/dialogs/index-spec-dialog/index-spec-dialog.spec.tsx create mode 100644 web-console/src/dialogs/index-spec-dialog/index-spec-dialog.tsx create mode 100644 web-console/src/druid-models/index-spec/index-spec.tsx diff --git a/web-console/src/components/auto-form/auto-form.scss b/web-console/src/components/auto-form/auto-form.scss index 5523f0f8173f..36f8c34d92d2 100644 --- a/web-console/src/components/auto-form/auto-form.scss +++ b/web-console/src/components/auto-form/auto-form.scss @@ -28,4 +28,8 @@ right: 0; } } + + .custom-input input { + cursor: pointer; + } } diff --git a/web-console/src/components/auto-form/auto-form.tsx b/web-console/src/components/auto-form/auto-form.tsx index 1e56ef2b725d..146de61b6272 100644 --- a/web-console/src/components/auto-form/auto-form.tsx +++ b/web-console/src/components/auto-form/auto-form.tsx @@ -16,7 +16,14 @@ * limitations under the License. */ -import { Button, ButtonGroup, FormGroup, Intent, NumericInput } from '@blueprintjs/core'; +import { + Button, + ButtonGroup, + FormGroup, + InputGroup, + Intent, + NumericInput, +} from '@blueprintjs/core'; import { IconNames } from '@blueprintjs/icons'; import React from 'react'; @@ -46,7 +53,8 @@ export interface Field { | 'boolean' | 'string-array' | 'json' - | 'interval'; + | 'interval' + | 'custom'; defaultValue?: any; emptyValue?: any; suggestions?: Functor; @@ -64,6 +72,13 @@ export interface Field { valueAdjustment?: (value: any) => any; adjustment?: (model: Partial) => Partial; issueWithValue?: (value: any) => string | undefined; + + customSummary?: (v: any) => string; + customDialog?: (o: { + value: any; + onValueChange: (v: any) => void; + onClose: () => void; + }) => JSX.Element; } interface ComputedFieldValues { @@ -84,6 +99,7 @@ export interface AutoFormProps { export interface AutoFormState { showMore: boolean; + customDialog?: JSX.Element; } export class AutoForm> extends React.PureComponent< @@ -395,6 +411,36 @@ export class AutoForm> extends React.PureComponent ); } + private renderCustomInput(field: Field): JSX.Element { + const { model } = this.props; + const { required, defaultValue, modelValue } = AutoForm.computeFieldValues(model, field); + const effectiveValue = modelValue || defaultValue; + + const onEdit = () => { + this.setState({ + customDialog: field.customDialog?.({ + value: effectiveValue, + onValueChange: v => this.fieldChange(field, v), + onClose: () => { + this.setState({ customDialog: undefined }); + }, + }), + }); + }; + + return ( + } + onClick={onEdit} + /> + ); + } + renderFieldInput(field: Field) { switch (field.type) { case 'number': @@ -413,6 +459,8 @@ export class AutoForm> extends React.PureComponent return this.renderJsonInput(field); case 'interval': return this.renderIntervalInput(field); + case 'custom': + return this.renderCustomInput(field); default: throw new Error(`unknown field type '${field.type}'`); } @@ -464,7 +512,7 @@ export class AutoForm> extends React.PureComponent render(): JSX.Element { const { fields, model, showCustom } = this.props; - const { showMore } = this.state; + const { showMore, customDialog } = this.state; let shouldShowMore = false; const shownFields = fields.filter(field => { @@ -489,6 +537,7 @@ export class AutoForm> extends React.PureComponent {model && shownFields.map(this.renderField)} {model && showCustom && showCustom(model) && this.renderCustom()} {shouldShowMore && this.renderMoreOrLess()} + {customDialog} ); } diff --git a/web-console/src/dialogs/index-spec-dialog/__snapshots__/index-spec-dialog.spec.tsx.snap b/web-console/src/dialogs/index-spec-dialog/__snapshots__/index-spec-dialog.spec.tsx.snap new file mode 100644 index 000000000000..57d989621b70 --- /dev/null +++ b/web-console/src/dialogs/index-spec-dialog/__snapshots__/index-spec-dialog.spec.tsx.snap @@ -0,0 +1,317 @@ +// Jest Snapshot v1, https://goo.gl/fbAQLP + +exports[`IndexSpecDialog matches snapshot with indexSpec 1`] = ` + + +

+ + Encoding format for STRING value dictionaries used by STRING and COMPLEX<json> columns. + , + "label": "String dictionary encoding", + "name": "stringDictionaryEncoding.type", + "suggestions": Array [ + "utf8", + "frontCoded", + ], + "type": "string", + }, + Object { + "defaultValue": 4, + "defined": [Function], + "info": + The number of values to place in a bucket to perform delta encoding. Must be a power of 2, maximum is 128. + , + "label": "String dictionary encoding bucket size", + "max": 128, + "min": 1, + "name": "stringDictionaryEncoding.bucketSize", + "type": "number", + }, + Object { + "defaultValue": "roaring", + "info": + Compression format for bitmap indexes. + , + "label": "Bitmap type", + "name": "bitmap.type", + "suggestions": Array [ + "roaring", + "concise", + ], + "type": "string", + }, + Object { + "defaultValue": true, + "defined": [Function], + "info": + Controls whether or not run-length encoding will be used when it is determined to be more space-efficient. + , + "label": "Bitmap compress run on serialization", + "name": "bitmap.compressRunOnSerialization", + "type": "boolean", + }, + Object { + "defaultValue": "lz4", + "info": + Compression format for dimension columns. + , + "name": "dimensionCompression", + "suggestions": Array [ + "lz4", + "lzf", + "zstd", + "uncompressed", + ], + "type": "string", + }, + Object { + "defaultValue": "longs", + "info": + Encoding format for long-typed columns. Applies regardless of whether they are dimensions or metrics. + + auto + + encodes the values using offset or lookup table depending on column cardinality, and store them with variable size. + + longs + + stores the value as-is with 8 bytes each. + , + "name": "longEncoding", + "suggestions": Array [ + "longs", + "auto", + ], + "type": "string", + }, + Object { + "defaultValue": "lz4", + "info": + Compression format for primitive type metric columns. + , + "name": "metricCompression", + "suggestions": Array [ + "lz4", + "lzf", + "zstd", + "uncompressed", + ], + "type": "string", + }, + Object { + "defaultValue": "lz4", + "info": + Compression format to use for nested column raw data. + , + "label": "JSON compression", + "name": "jsonCompression", + "suggestions": Array [ + "lz4", + "lzf", + "zstd", + "uncompressed", + ], + "type": "string", + }, + ] + } + model={ + Object { + "dimensionCompression": "lzf", + } + } + onChange={[Function]} + /> +
+
+
+ + +
+
+ +`; + +exports[`IndexSpecDialog matches snapshot without compactionConfig 1`] = ` + + +
+ + Encoding format for STRING value dictionaries used by STRING and COMPLEX<json> columns. + , + "label": "String dictionary encoding", + "name": "stringDictionaryEncoding.type", + "suggestions": Array [ + "utf8", + "frontCoded", + ], + "type": "string", + }, + Object { + "defaultValue": 4, + "defined": [Function], + "info": + The number of values to place in a bucket to perform delta encoding. Must be a power of 2, maximum is 128. + , + "label": "String dictionary encoding bucket size", + "max": 128, + "min": 1, + "name": "stringDictionaryEncoding.bucketSize", + "type": "number", + }, + Object { + "defaultValue": "roaring", + "info": + Compression format for bitmap indexes. + , + "label": "Bitmap type", + "name": "bitmap.type", + "suggestions": Array [ + "roaring", + "concise", + ], + "type": "string", + }, + Object { + "defaultValue": true, + "defined": [Function], + "info": + Controls whether or not run-length encoding will be used when it is determined to be more space-efficient. + , + "label": "Bitmap compress run on serialization", + "name": "bitmap.compressRunOnSerialization", + "type": "boolean", + }, + Object { + "defaultValue": "lz4", + "info": + Compression format for dimension columns. + , + "name": "dimensionCompression", + "suggestions": Array [ + "lz4", + "lzf", + "zstd", + "uncompressed", + ], + "type": "string", + }, + Object { + "defaultValue": "longs", + "info": + Encoding format for long-typed columns. Applies regardless of whether they are dimensions or metrics. + + auto + + encodes the values using offset or lookup table depending on column cardinality, and store them with variable size. + + longs + + stores the value as-is with 8 bytes each. + , + "name": "longEncoding", + "suggestions": Array [ + "longs", + "auto", + ], + "type": "string", + }, + Object { + "defaultValue": "lz4", + "info": + Compression format for primitive type metric columns. + , + "name": "metricCompression", + "suggestions": Array [ + "lz4", + "lzf", + "zstd", + "uncompressed", + ], + "type": "string", + }, + Object { + "defaultValue": "lz4", + "info": + Compression format to use for nested column raw data. + , + "label": "JSON compression", + "name": "jsonCompression", + "suggestions": Array [ + "lz4", + "lzf", + "zstd", + "uncompressed", + ], + "type": "string", + }, + ] + } + model={Object {}} + onChange={[Function]} + /> +
+
+
+ + +
+
+ +`; diff --git a/web-console/src/dialogs/index-spec-dialog/index-spec-dialog.scss b/web-console/src/dialogs/index-spec-dialog/index-spec-dialog.scss new file mode 100644 index 000000000000..e7cc53ee47dd --- /dev/null +++ b/web-console/src/dialogs/index-spec-dialog/index-spec-dialog.scss @@ -0,0 +1,36 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +@import '../../variables'; + +.index-spec-dialog { + &.#{$bp-ns}-dialog { + height: 70vh; + } + + .form-json-selector { + margin: 15px; + } + + .content { + margin: 0 15px 10px 0; + padding: 0 5px 0 15px; + flex: 1; + overflow: auto; + } +} diff --git a/web-console/src/dialogs/index-spec-dialog/index-spec-dialog.spec.tsx b/web-console/src/dialogs/index-spec-dialog/index-spec-dialog.spec.tsx new file mode 100644 index 000000000000..68f7f56b885b --- /dev/null +++ b/web-console/src/dialogs/index-spec-dialog/index-spec-dialog.spec.tsx @@ -0,0 +1,44 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +import { shallow } from 'enzyme'; +import React from 'react'; + +import { IndexSpecDialog } from './index-spec-dialog'; + +describe('IndexSpecDialog', () => { + it('matches snapshot without compactionConfig', () => { + const compactionDialog = shallow( + {}} onSave={() => {}} indexSpec={undefined} />, + ); + expect(compactionDialog).toMatchSnapshot(); + }); + + it('matches snapshot with indexSpec', () => { + const compactionDialog = shallow( + {}} + onSave={() => {}} + indexSpec={{ + dimensionCompression: 'lzf', + }} + />, + ); + expect(compactionDialog).toMatchSnapshot(); + }); +}); diff --git a/web-console/src/dialogs/index-spec-dialog/index-spec-dialog.tsx b/web-console/src/dialogs/index-spec-dialog/index-spec-dialog.tsx new file mode 100644 index 000000000000..4c870df45af2 --- /dev/null +++ b/web-console/src/dialogs/index-spec-dialog/index-spec-dialog.tsx @@ -0,0 +1,88 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +import { Button, Classes, Dialog, Intent } from '@blueprintjs/core'; +import React, { useState } from 'react'; + +import { AutoForm, FormJsonSelector, FormJsonTabs, JsonInput } from '../../components'; +import { INDEX_SPEC_FIELDS, IndexSpec } from '../../druid-models'; + +import './index-spec-dialog.scss'; + +export interface IndexSpecDialogProps { + title?: string; + onClose: () => void; + onSave: (indexSpec: IndexSpec) => void; + indexSpec: IndexSpec | undefined; +} + +export const IndexSpecDialog = React.memo(function IndexSpecDialog(props: IndexSpecDialogProps) { + const { title, indexSpec, onSave, onClose } = props; + + const [currentTab, setCurrentTab] = useState('form'); + const [currentIndexSpec, setCurrentIndexSpec] = useState(indexSpec || {}); + const [jsonError, setJsonError] = useState(); + + const issueWithCurrentIndexSpec = AutoForm.issueWithModel(currentIndexSpec, INDEX_SPEC_FIELDS); + + return ( + + +
+ {currentTab === 'form' ? ( + setCurrentIndexSpec(m)} + /> + ) : ( + { + setCurrentIndexSpec(v); + setJsonError(undefined); + }} + onError={setJsonError} + issueWithValue={value => AutoForm.issueWithModel(value, INDEX_SPEC_FIELDS)} + height="100%" + /> + )} +
+
+
+
+
+ + ); +}); diff --git a/web-console/src/druid-models/index-spec/index-spec.tsx b/web-console/src/druid-models/index-spec/index-spec.tsx new file mode 100644 index 000000000000..1a4246299661 --- /dev/null +++ b/web-console/src/druid-models/index-spec/index-spec.tsx @@ -0,0 +1,158 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +import { Code } from '@blueprintjs/core'; +import React from 'react'; + +import { Field } from '../../components'; +import { deepGet } from '../../utils'; + +export interface IndexSpec { + bitmap?: Bitmap; + dimensionCompression?: string; + stringDictionaryEncoding?: { type: 'utf8' | 'frontCoded'; bucketSize: number }; + metricCompression?: string; + longEncoding?: string; + jsonCompression?: string; +} + +export interface Bitmap { + type: string; + compressRunOnSerialization?: boolean; +} + +export function summarizeIndexSpec(indexSpec: IndexSpec | undefined): string { + if (!indexSpec) return ''; + + const { stringDictionaryEncoding, bitmap, longEncoding } = indexSpec; + + const ret: string[] = []; + if (stringDictionaryEncoding) { + switch (stringDictionaryEncoding.type) { + case 'frontCoded': + ret.push(`frontCoded(${stringDictionaryEncoding.bucketSize || 4})`); + break; + + default: + ret.push(stringDictionaryEncoding.type); + break; + } + } + + if (bitmap) { + ret.push(bitmap.type); + } + + if (longEncoding) { + ret.push(longEncoding); + } + + return ret.join('; '); +} + +export const INDEX_SPEC_FIELDS: Field[] = [ + { + name: 'stringDictionaryEncoding.type', + label: 'String dictionary encoding', + type: 'string', + defaultValue: 'utf8', + suggestions: ['utf8', 'frontCoded'], + info: ( + <> + Encoding format for STRING value dictionaries used by STRING and COMPLEX<json> + columns. + + ), + }, + { + name: 'stringDictionaryEncoding.bucketSize', + label: 'String dictionary encoding bucket size', + type: 'number', + defaultValue: 4, + min: 1, + max: 128, + defined: spec => deepGet(spec, 'stringDictionaryEncoding.type') === 'frontCoded', + info: ( + <> + The number of values to place in a bucket to perform delta encoding. Must be a power of 2, + maximum is 128. + + ), + }, + + { + name: 'bitmap.type', + label: 'Bitmap type', + type: 'string', + defaultValue: 'roaring', + suggestions: ['roaring', 'concise'], + info: <>Compression format for bitmap indexes., + }, + { + name: 'bitmap.compressRunOnSerialization', + label: 'Bitmap compress run on serialization', + type: 'boolean', + defaultValue: true, + defined: spec => (deepGet(spec, 'bitmap.type') || 'roaring') === 'roaring', + info: ( + <> + Controls whether or not run-length encoding will be used when it is determined to be more + space-efficient. + + ), + }, + + { + name: 'dimensionCompression', + type: 'string', + defaultValue: 'lz4', + suggestions: ['lz4', 'lzf', 'zstd', 'uncompressed'], + info: <>Compression format for dimension columns., + }, + + { + name: 'longEncoding', + type: 'string', + defaultValue: 'longs', + suggestions: ['longs', 'auto'], + info: ( + <> + Encoding format for long-typed columns. Applies regardless of whether they are dimensions or + metrics. auto encodes the values using offset or lookup table depending on + column cardinality, and store them with variable size. longs stores the value + as-is with 8 bytes each. + + ), + }, + { + name: 'metricCompression', + type: 'string', + defaultValue: 'lz4', + suggestions: ['lz4', 'lzf', 'zstd', 'uncompressed'], + info: <>Compression format for primitive type metric columns., + }, + + { + name: 'jsonCompression', + label: 'JSON compression', + type: 'string', + defaultValue: 'lz4', + suggestions: ['lz4', 'lzf', 'zstd', 'uncompressed'], + info: <>Compression format to use for nested column raw data. , + }, +]; diff --git a/web-console/src/druid-models/index.ts b/web-console/src/druid-models/index.ts index 359ba70440b8..0b4ad6b65f70 100644 --- a/web-console/src/druid-models/index.ts +++ b/web-console/src/druid-models/index.ts @@ -25,6 +25,7 @@ export * from './execution/execution'; export * from './external-config/external-config'; export * from './filter/filter'; export * from './flatten-spec/flatten-spec'; +export * from './index-spec/index-spec'; export * from './ingest-query-pattern/ingest-query-pattern'; export * from './ingestion-spec/ingestion-spec'; export * from './input-format/input-format'; diff --git a/web-console/src/druid-models/ingestion-spec/ingestion-spec.tsx b/web-console/src/druid-models/ingestion-spec/ingestion-spec.tsx index b68052abd471..f144f4c2ad29 100644 --- a/web-console/src/druid-models/ingestion-spec/ingestion-spec.tsx +++ b/web-console/src/druid-models/ingestion-spec/ingestion-spec.tsx @@ -21,6 +21,7 @@ import { range } from 'd3-array'; import React from 'react'; import { AutoForm, ExternalLink, Field } from '../../components'; +import { IndexSpecDialog } from '../../dialogs/index-spec-dialog/index-spec-dialog'; import { getLink } from '../../links'; import { allowKeys, @@ -44,6 +45,7 @@ import { getDimensionSpecs, getDimensionSpecType, } from '../dimension-spec/dimension-spec'; +import { IndexSpec, summarizeIndexSpec } from '../index-spec/index-spec'; import { InputFormat, issueWithInputFormat } from '../input-format/input-format'; import { FILTER_SUGGESTIONS, @@ -1379,6 +1381,7 @@ export interface TuningConfig { partitionsSpec?: PartitionsSpec; maxPendingPersists?: number; indexSpec?: IndexSpec; + indexSpecForIntermediatePersists?: IndexSpec; forceExtendableShardSpecs?: boolean; forceGuaranteedRollup?: boolean; reportParseExceptions?: boolean; @@ -1869,103 +1872,38 @@ const TUNING_FORM_FIELDS: Field[] = [ }, { - name: 'spec.tuningConfig.indexSpec.bitmap.type', - label: 'Index bitmap type', - type: 'string', - defaultValue: 'roaring', - suggestions: ['concise', 'roaring'], + name: 'spec.tuningConfig.indexSpec', + type: 'custom', hideInMore: true, - info: <>Compression format for bitmap indexes., - }, - { - name: 'spec.tuningConfig.indexSpec.bitmap.compressRunOnSerialization', - type: 'boolean', - defaultValue: true, - defined: spec => deepGet(spec, 'spec.tuningConfig.indexSpec.bitmap.type') === 'roaring', - info: ( - <> - Controls whether or not run-length encoding will be used when it is determined to be more - space-efficient. - + info: <>Defines segment storage format options to use at indexing time., + placeholder: 'Default index spec', + customSummary: summarizeIndexSpec, + customDialog: ({ value, onValueChange, onClose }) => ( + ), }, - - { - name: 'spec.tuningConfig.indexSpec.dimensionCompression', - label: 'Index dimension compression', - type: 'string', - defaultValue: 'lz4', - suggestions: ['lz4', 'lzf', 'zstd', 'uncompressed'], - hideInMore: true, - info: <>Compression format for dimension columns., - }, - { - name: 'spec.tuningConfig.indexSpec.stringDictionaryEncoding.type', - label: 'Index string dictionary encoding', - type: 'string', - defaultValue: 'utf8', - suggestions: ['utf8', 'frontCoded'], + name: 'spec.tuningConfig.indexSpecForIntermediatePersists', + type: 'custom', hideInMore: true, info: ( <> - Encoding format for STRING value dictionaries used by STRING and COMPLEX<json> - columns. + Defines segment storage format options to use at indexing time for intermediate persisted + temporary segments. ), - }, - { - name: 'spec.tuningConfig.indexSpec.stringDictionaryEncoding.bucketSize', - label: 'Index string dictionary encoding bucket size', - type: 'number', - defaultValue: 4, - min: 1, - max: 128, - defined: spec => - deepGet(spec, 'spec.tuningConfig.indexSpec.stringDictionaryEncoding.type') === 'frontCoded', - hideInMore: true, - info: ( - <> - The number of values to place in a bucket to perform delta encoding. Must be a power of 2, - maximum is 128. - + placeholder: 'Default index spec', + customSummary: summarizeIndexSpec, + customDialog: ({ value, onValueChange, onClose }) => ( + ), }, - { - name: 'spec.tuningConfig.indexSpec.metricCompression', - label: 'Index metric compression', - type: 'string', - defaultValue: 'lz4', - suggestions: ['lz4', 'lzf', 'zstd', 'uncompressed'], - hideInMore: true, - info: <>Compression format for primitive type metric columns., - }, - { - name: 'spec.tuningConfig.indexSpec.longEncoding', - label: 'Index long encoding', - type: 'string', - defaultValue: 'longs', - suggestions: ['longs', 'auto'], - hideInMore: true, - info: ( - <> - Encoding format for long-typed columns. Applies regardless of whether they are dimensions or - metrics. auto encodes the values using offset or lookup table depending on - column cardinality, and store them with variable size. longs stores the value - as-is with 8 bytes each. - - ), - }, - { - name: 'spec.tuningConfig.indexSpec.jsonCompression', - label: 'Index JSON compression', - type: 'string', - defaultValue: 'lz4', - suggestions: ['lz4', 'lzf', 'zstd', 'uncompressed'], - hideInMore: true, - info: <>Compression format to use for nested column raw data. , - }, { name: 'spec.tuningConfig.splitHintSpec.maxSplitSize', type: 'number', @@ -2172,18 +2110,6 @@ export function getTuningFormFields() { return TUNING_FORM_FIELDS; } -export interface IndexSpec { - bitmap?: Bitmap; - dimensionCompression?: string; - metricCompression?: string; - longEncoding?: string; -} - -export interface Bitmap { - type: string; - compressRunOnSerialization?: boolean; -} - // -------------- export function updateIngestionType( diff --git a/web-console/src/druid-models/workbench-query/workbench-query-part.ts b/web-console/src/druid-models/workbench-query/workbench-query-part.ts index 604cfb013213..5e4afb453f09 100644 --- a/web-console/src/druid-models/workbench-query/workbench-query-part.ts +++ b/web-console/src/druid-models/workbench-query/workbench-query-part.ts @@ -62,10 +62,10 @@ export class WorkbenchQueryPart { static getIngestDatasourceFromQueryFragment(queryFragment: string): string | undefined { // Assuming the queryFragment is no parsable find the prefix that look like: // REPLACEINTOSELECT - const matchInsertReplaceIndex = queryFragment.match(/(?:INSERT|REPLACE)\s+INTO/)?.index; + const matchInsertReplaceIndex = queryFragment.match(/(?:INSERT|REPLACE)\s+INTO/i)?.index; if (typeof matchInsertReplaceIndex !== 'number') return; - const matchEnd = queryFragment.match(/\b(?:SELECT|WITH)\b|$/); + const matchEnd = queryFragment.match(/\b(?:SELECT|WITH)\b|$/i); const fragmentQuery = SqlQuery.maybeParse( queryFragment.substring(matchInsertReplaceIndex, matchEnd?.index) + ' SELECT * FROM t', ); diff --git a/web-console/src/druid-models/workbench-query/workbench-query.spec.ts b/web-console/src/druid-models/workbench-query/workbench-query.spec.ts index 5d7e2615098c..9af0fb240700 100644 --- a/web-console/src/druid-models/workbench-query/workbench-query.spec.ts +++ b/web-console/src/druid-models/workbench-query/workbench-query.spec.ts @@ -465,7 +465,7 @@ describe('WorkbenchQuery', () => { it('works with INSERT (unparsable)', () => { const sql = sane` -- Some comment - INSERT INTO trips2 + INSERT into trips2 SELECT TIME_PARSE(pickup_datetime) AS __time, * diff --git a/web-console/src/helpers/spec-conversion.spec.ts b/web-console/src/helpers/spec-conversion.spec.ts index 0239da185415..2f6aa59f51e2 100644 --- a/web-console/src/helpers/spec-conversion.spec.ts +++ b/web-console/src/helpers/spec-conversion.spec.ts @@ -106,6 +106,9 @@ describe('spec conversion', () => { partitionDimension: 'isRobot', targetRowsPerSegment: 150000, }, + indexSpec: { + dimensionCompression: 'lzf', + }, forceGuaranteedRollup: true, maxNumConcurrentSubTasks: 4, maxParseExceptions: 3, @@ -159,6 +162,9 @@ describe('spec conversion', () => { maxParseExceptions: 3, finalizeAggregations: false, maxNumTasks: 5, + indexSpec: { + dimensionCompression: 'lzf', + }, }); }); diff --git a/web-console/src/helpers/spec-conversion.ts b/web-console/src/helpers/spec-conversion.ts index d35433917f59..9b76e787dd61 100644 --- a/web-console/src/helpers/spec-conversion.ts +++ b/web-console/src/helpers/spec-conversion.ts @@ -70,6 +70,11 @@ export function convertSpecToSql(spec: any): QueryWithContext { groupByEnableMultiValueUnnesting: false, }; + const indexSpec = deepGet(spec, 'spec.tuningConfig.indexSpec'); + if (indexSpec) { + context.indexSpec = indexSpec; + } + const lines: string[] = []; const rollup = deepGet(spec, 'spec.dataSchema.granularitySpec.rollup') ?? true; diff --git a/web-console/src/views/workbench-view/run-panel/run-panel.tsx b/web-console/src/views/workbench-view/run-panel/run-panel.tsx index 572120c9e730..7299760b4621 100644 --- a/web-console/src/views/workbench-view/run-panel/run-panel.tsx +++ b/web-console/src/views/workbench-view/run-panel/run-panel.tsx @@ -33,6 +33,7 @@ import React, { useCallback, useMemo, useState } from 'react'; import { MenuCheckbox, MenuTristate } from '../../../components'; import { EditContextDialog, StringInputDialog } from '../../../dialogs'; +import { IndexSpecDialog } from '../../../dialogs/index-spec-dialog/index-spec-dialog'; import { changeDurableShuffleStorage, changeFinalizeAggregations, @@ -51,9 +52,12 @@ import { getUseApproximateCountDistinct, getUseApproximateTopN, getUseCache, + IndexSpec, + QueryContext, + summarizeIndexSpec, WorkbenchQuery, } from '../../../druid-models'; -import { pluralIfNeeded, tickIcon } from '../../../utils'; +import { deepGet, pluralIfNeeded, tickIcon } from '../../../utils'; import { MaxTasksButton } from '../max-tasks-button/max-tasks-button'; import './run-panel.scss'; @@ -94,6 +98,7 @@ export const RunPanel = React.memo(function RunPanel(props: RunPanelProps) { const { query, onQueryChange, onRun, moreMenu, loading, small, queryEngines } = props; const [editContextDialogOpen, setEditContextDialogOpen] = useState(false); const [customTimezoneDialogOpen, setCustomTimezoneDialogOpen] = useState(false); + const [indexSpecDialogSpec, setIndexSpecDialogSpec] = useState(); const emptyQuery = query.isEmptyQuery(); const ingestMode = query.isIngestQuery(); @@ -104,6 +109,7 @@ export const RunPanel = React.memo(function RunPanel(props: RunPanelProps) { const finalizeAggregations = getFinalizeAggregations(queryContext); const groupByEnableMultiValueUnnesting = getGroupByEnableMultiValueUnnesting(queryContext); const durableShuffleStorage = getDurableShuffleStorage(queryContext); + const indexSpec: IndexSpec | undefined = deepGet(queryContext, 'indexSpec'); const useApproximateCountDistinct = getUseApproximateCountDistinct(queryContext); const useApproximateTopN = getUseApproximateTopN(queryContext); const useCache = getUseCache(queryContext); @@ -157,6 +163,10 @@ export const RunPanel = React.memo(function RunPanel(props: RunPanelProps) { ); } + function changeQueryContext(queryContext: QueryContext) { + onQueryChange(query.changeQueryContext(queryContext)); + } + const availableEngines = ([undefined] as (DruidEngine | undefined)[]).concat(queryEngines); function offsetOptions(): JSX.Element[] { @@ -170,9 +180,7 @@ export const RunPanel = React.memo(function RunPanel(props: RunPanelProps) { icon={tickIcon(offset === timezone)} text={offset} shouldDismissPopover={false} - onClick={() => { - onQueryChange(query.changeQueryContext(changeTimezone(queryContext, offset))); - }} + onClick={() => changeQueryContext(changeTimezone(queryContext, offset))} />, ); } @@ -233,11 +241,7 @@ export const RunPanel = React.memo(function RunPanel(props: RunPanelProps) { icon={tickIcon(!timezone)} text="Default" shouldDismissPopover={false} - onClick={() => { - onQueryChange( - query.changeQueryContext(changeTimezone(queryContext, undefined)), - ); - }} + onClick={() => changeQueryContext(changeTimezone(queryContext, undefined))} /> {NAMED_TIMEZONES.map(namedTimezone => ( @@ -246,11 +250,9 @@ export const RunPanel = React.memo(function RunPanel(props: RunPanelProps) { icon={tickIcon(namedTimezone === timezone)} text={namedTimezone} shouldDismissPopover={false} - onClick={() => { - onQueryChange( - query.changeQueryContext(changeTimezone(queryContext, namedTimezone)), - ); - }} + onClick={() => + changeQueryContext(changeTimezone(queryContext, namedTimezone)) + } /> ))} @@ -276,11 +278,9 @@ export const RunPanel = React.memo(function RunPanel(props: RunPanelProps) { key={String(v)} icon={tickIcon(v === maxParseExceptions)} text={v === -1 ? '∞ (-1)' : String(v)} - onClick={() => { - onQueryChange( - query.changeQueryContext(changeMaxParseExceptions(queryContext, v)), - ); - }} + onClick={() => + changeQueryContext(changeMaxParseExceptions(queryContext, v)) + } shouldDismissPopover={false} /> ))} @@ -290,35 +290,36 @@ export const RunPanel = React.memo(function RunPanel(props: RunPanelProps) { text="Finalize aggregations" value={finalizeAggregations} undefinedEffectiveValue={!ingestMode} - onValueChange={v => { - onQueryChange( - query.changeQueryContext(changeFinalizeAggregations(queryContext, v)), - ); - }} + onValueChange={v => + changeQueryContext(changeFinalizeAggregations(queryContext, v)) + } /> { - onQueryChange( - query.changeQueryContext( - changeGroupByEnableMultiValueUnnesting(queryContext, v), - ), - ); + onValueChange={v => + changeQueryContext(changeGroupByEnableMultiValueUnnesting(queryContext, v)) + } + /> + { + setIndexSpecDialogSpec(indexSpec || {}); }} /> { - onQueryChange( - query.changeQueryContext( - changeDurableShuffleStorage(queryContext, !durableShuffleStorage), - ), - ); - }} + onChange={() => + changeQueryContext( + changeDurableShuffleStorage(queryContext, !durableShuffleStorage), + ) + } /> ) : ( @@ -326,22 +327,16 @@ export const RunPanel = React.memo(function RunPanel(props: RunPanelProps) { { - onQueryChange( - query.changeQueryContext(changeUseCache(queryContext, !useCache)), - ); - }} + onChange={() => changeQueryContext(changeUseCache(queryContext, !useCache))} /> { - onQueryChange( - query.changeQueryContext( - changeUseApproximateTopN(queryContext, !useApproximateTopN), - ), - ); - }} + onChange={() => + changeQueryContext( + changeUseApproximateTopN(queryContext, !useApproximateTopN), + ) + } /> )} @@ -349,16 +344,14 @@ export const RunPanel = React.memo(function RunPanel(props: RunPanelProps) { { - onQueryChange( - query.changeQueryContext( - changeUseApproximateCountDistinct( - queryContext, - !useApproximateCountDistinct, - ), + onChange={() => + changeQueryContext( + changeUseApproximateCountDistinct( + queryContext, + !useApproximateCountDistinct, ), - ); - }} + ) + } /> )} {effectiveEngine === 'sql-msq-task' && ( - - onQueryChange(query.changeQueryContext(queryContext)) - } - /> + )} )} @@ -399,10 +387,7 @@ export const RunPanel = React.memo(function RunPanel(props: RunPanelProps) { {editContextDialogOpen && ( { - if (!onQueryChange) return; - onQueryChange(query.changeQueryContext(newContext)); - }} + onQueryContextChange={changeQueryContext} onClose={() => { setEditContextDialogOpen(false); }} @@ -413,10 +398,17 @@ export const RunPanel = React.memo(function RunPanel(props: RunPanelProps) { title="Custom timezone" placeholder="Etc/UTC" maxLength={50} - onSubmit={tz => onQueryChange(query.changeQueryContext(changeTimezone(queryContext, tz)))} + onSubmit={tz => changeQueryContext(changeTimezone(queryContext, tz))} onClose={() => setCustomTimezoneDialogOpen(false)} /> )} + {indexSpecDialogSpec && ( + setIndexSpecDialogSpec(undefined)} + onSave={indexSpec => changeQueryContext({ ...queryContext, indexSpec })} + indexSpec={indexSpecDialogSpec} + /> + )} ); }); From 4b58f5f23c2a1d23bb9dcb90ff13916659539ee3 Mon Sep 17 00:00:00 2001 From: Clint Wylie Date: Mon, 28 Nov 2022 12:28:07 -0800 Subject: [PATCH 049/141] fix KafkaInputFormat with nested columns by delegating to underlying inputRow map instead of eagerly copying (#13406) --- .../input/kafkainput/KafkaInputReader.java | 177 ++++++++++++------ .../kafkainput/KafkaInputFormatTest.java | 7 +- 2 files changed, 124 insertions(+), 60 deletions(-) diff --git a/extensions-core/kafka-indexing-service/src/main/java/org/apache/druid/data/input/kafkainput/KafkaInputReader.java b/extensions-core/kafka-indexing-service/src/main/java/org/apache/druid/data/input/kafkainput/KafkaInputReader.java index 458955e58070..56e94215940b 100644 --- a/extensions-core/kafka-indexing-service/src/main/java/org/apache/druid/data/input/kafkainput/KafkaInputReader.java +++ b/extensions-core/kafka-indexing-service/src/main/java/org/apache/druid/data/input/kafkainput/KafkaInputReader.java @@ -30,24 +30,24 @@ import org.apache.druid.indexing.seekablestream.SettableByteEntity; import org.apache.druid.java.util.common.CloseableIterators; import org.apache.druid.java.util.common.Pair; -import org.apache.druid.java.util.common.logger.Logger; import org.apache.druid.java.util.common.parsers.CloseableIterator; import org.apache.druid.java.util.common.parsers.ParseException; import javax.annotation.Nullable; - import java.io.IOException; +import java.util.AbstractMap; import java.util.Collections; import java.util.HashMap; import java.util.HashSet; +import java.util.LinkedHashSet; import java.util.List; import java.util.Map; +import java.util.Set; import java.util.function.Function; +import java.util.stream.Collectors; public class KafkaInputReader implements InputEntityReader { - private static final Logger log = new Logger(KafkaInputReader.class); - private final InputRowSchema inputRowSchema; private final SettableByteEntity source; private final Function headerParserSupplier; @@ -85,7 +85,60 @@ public KafkaInputReader( this.timestampColumnName = timestampColumnName; } - private List getFinalDimensionList(HashSet newDimensions) + @Override + public CloseableIterator read() throws IOException + { + final KafkaRecordEntity record = source.getEntity(); + final Map mergedHeaderMap = new HashMap<>(); + if (headerParserSupplier != null) { + KafkaHeaderReader headerParser = headerParserSupplier.apply(record); + List> headerList = headerParser.read(); + for (Pair ele : headerList) { + mergedHeaderMap.put(ele.lhs, ele.rhs); + } + } + + // Add kafka record timestamp to the mergelist, we will skip record timestamp if the same key exists already in + // the header list + mergedHeaderMap.putIfAbsent(timestampColumnName, record.getRecord().timestamp()); + + InputEntityReader keyParser = (keyParserSupplier == null) ? null : keyParserSupplier.apply(record); + if (keyParser != null) { + try (CloseableIterator keyIterator = keyParser.read()) { + // Key currently only takes the first row and ignores the rest. + if (keyIterator.hasNext()) { + // Return type for the key parser should be of type MapBasedInputRow + // Parsers returning other types are not compatible currently. + MapBasedInputRow keyRow = (MapBasedInputRow) keyIterator.next(); + // Add the key to the mergeList only if the key string is not already present + mergedHeaderMap.putIfAbsent( + keyColumnName, + keyRow.getEvent().entrySet().stream().findFirst().get().getValue() + ); + } + } + catch (ClassCastException e) { + throw new IOException( + "Unsupported keyFormat. KafkaInputformat only supports input format that return MapBasedInputRow rows" + ); + } + } + + // Ignore tombstone records that have null values. + if (record.getRecord().value() != null) { + return buildBlendedRows(valueParser, mergedHeaderMap); + } else { + return buildRowsWithoutValuePayload(mergedHeaderMap); + } + } + + @Override + public CloseableIterator sample() throws IOException + { + return read().map(row -> InputRowListPlusRawValues.of(row, ((MapBasedInputRow) row).getEvent())); + } + + private List getFinalDimensionList(Set newDimensions) { final List schemaDimensions = inputRowSchema.getDimensionsSpec().getDimensionNames(); if (!schemaDimensions.isEmpty()) { @@ -97,11 +150,14 @@ private List getFinalDimensionList(HashSet newDimensions) } } - private CloseableIterator buildBlendedRows(InputEntityReader valueParser, Map headerKeyList) throws IOException + private CloseableIterator buildBlendedRows( + InputEntityReader valueParser, + Map headerKeyList + ) throws IOException { return valueParser.read().map( r -> { - MapBasedInputRow valueRow; + final MapBasedInputRow valueRow; try { // Return type for the value parser should be of type MapBasedInputRow // Parsers returning other types are not compatible currently. @@ -113,14 +169,9 @@ private CloseableIterator buildBlendedRows(InputEntityReader valuePars "Unsupported input format in valueFormat. KafkaInputFormat only supports input format that return MapBasedInputRow rows" ); } - Map event = new HashMap<>(headerKeyList); - /* Currently we prefer payload attributes if there is a collision in names. - We can change this beahvior in later changes with a config knob. This default - behavior lets easy porting of existing inputFormats to the new one without any changes. - */ - event.putAll(valueRow.getEvent()); - - HashSet newDimensions = new HashSet(valueRow.getDimensions()); + + final Map event = buildBlendedEventMap(valueRow.getEvent(), headerKeyList); + final HashSet newDimensions = new HashSet<>(valueRow.getDimensions()); newDimensions.addAll(headerKeyList.keySet()); // Remove the dummy timestamp added in KafkaInputFormat newDimensions.remove(KafkaInputFormat.DEFAULT_AUTO_TIMESTAMP_STRING); @@ -136,60 +187,70 @@ private CloseableIterator buildBlendedRows(InputEntityReader valuePars private CloseableIterator buildRowsWithoutValuePayload(Map headerKeyList) { - HashSet newDimensions = new HashSet(headerKeyList.keySet()); - InputRow row = new MapBasedInputRow( + final InputRow row = new MapBasedInputRow( inputRowSchema.getTimestampSpec().extractTimestamp(headerKeyList), - getFinalDimensionList(newDimensions), + getFinalDimensionList(headerKeyList.keySet()), headerKeyList ); - List rows = Collections.singletonList(row); + final List rows = Collections.singletonList(row); return CloseableIterators.withEmptyBaggage(rows.iterator()); } - @Override - public CloseableIterator read() throws IOException + /** + * Builds a map that blends two {@link Map}, presenting the combined keyset of both maps, and preferring to read + * from the first map and falling back to the second map if the value is not present. + * + * This strategy is used rather than just copying the values of the keyset into a new map so that any 'flattening' + * machinery (such as {@link Map} created by {@link org.apache.druid.java.util.common.parsers.ObjectFlatteners}) is + * still in place to be lazily evaluated instead of eagerly copying. + */ + private static Map buildBlendedEventMap(Map map, Map fallback) { - KafkaRecordEntity record = source.getEntity(); - Map mergeMap = new HashMap<>(); - if (headerParserSupplier != null) { - KafkaHeaderReader headerParser = headerParserSupplier.apply(record); - List> headerList = headerParser.read(); - for (Pair ele : headerList) { - mergeMap.put(ele.lhs, ele.rhs); - } - } - - // Add kafka record timestamp to the mergelist, we will skip record timestamp if the same key exists already in the header list - mergeMap.putIfAbsent(timestampColumnName, record.getRecord().timestamp()); + final Set keySet = new HashSet<>(fallback.keySet()); + keySet.addAll(map.keySet()); - InputEntityReader keyParser = (keyParserSupplier == null) ? null : keyParserSupplier.apply(record); - if (keyParser != null) { - try (CloseableIterator keyIterator = keyParser.read()) { - // Key currently only takes the first row and ignores the rest. - if (keyIterator.hasNext()) { - // Return type for the key parser should be of type MapBasedInputRow - // Parsers returning other types are not compatible currently. - MapBasedInputRow keyRow = (MapBasedInputRow) keyIterator.next(); - // Add the key to the mergeList only if the key string is not already present - mergeMap.putIfAbsent(keyColumnName, keyRow.getEvent().entrySet().stream().findFirst().get().getValue()); - } + return new AbstractMap() + { + @Override + public Object get(Object key) + { + return map.getOrDefault((String) key, fallback.get(key)); } - catch (ClassCastException e) { - throw new IOException("Unsupported input format in keyFormat. KafkaInputformat only supports input format that return MapBasedInputRow rows"); + + @Override + public Set keySet() + { + return keySet; } - } - // Ignore tombstone records that have null values. - if (record.getRecord().value() != null) { - return buildBlendedRows(valueParser, mergeMap); - } else { - return buildRowsWithoutValuePayload(mergeMap); - } - } + @Override + public Set> entrySet() + { + return keySet().stream() + .map( + field -> new Entry() + { + @Override + public String getKey() + { + return field; + } - @Override - public CloseableIterator sample() throws IOException - { - return read().map(row -> InputRowListPlusRawValues.of(row, ((MapBasedInputRow) row).getEvent())); + @Override + public Object getValue() + { + return get(field); + } + + @Override + public Object setValue(final Object value) + { + throw new UnsupportedOperationException(); + } + } + ) + .collect(Collectors.toCollection(LinkedHashSet::new)); + } + }; } } diff --git a/extensions-core/kafka-indexing-service/src/test/java/org/apache/druid/data/input/kafkainput/KafkaInputFormatTest.java b/extensions-core/kafka-indexing-service/src/test/java/org/apache/druid/data/input/kafkainput/KafkaInputFormatTest.java index fe0b89e996f8..a45730005a9c 100644 --- a/extensions-core/kafka-indexing-service/src/test/java/org/apache/druid/data/input/kafkainput/KafkaInputFormatTest.java +++ b/extensions-core/kafka-indexing-service/src/test/java/org/apache/druid/data/input/kafkainput/KafkaInputFormatTest.java @@ -22,12 +22,12 @@ import com.fasterxml.jackson.core.JsonProcessingException; import com.fasterxml.jackson.databind.ObjectMapper; import com.google.common.collect.ImmutableList; +import com.google.common.collect.ImmutableMap; import com.google.common.collect.Iterables; import org.apache.druid.data.input.ColumnsFilter; import org.apache.druid.data.input.InputEntityReader; import org.apache.druid.data.input.InputRow; import org.apache.druid.data.input.InputRowSchema; -import org.apache.druid.data.input.MapBasedInputRow; import org.apache.druid.data.input.impl.DimensionsSpec; import org.apache.druid.data.input.impl.JsonInputFormat; import org.apache.druid.data.input.impl.TimestampSpec; @@ -205,6 +205,7 @@ public void testWithHeaderKeyAndValue() throws IOException Assert.assertEquals("4", Iterables.getOnlyElement(row.getDimension("root_baz"))); Assert.assertEquals("1", Iterables.getOnlyElement(row.getDimension("path_omg"))); Assert.assertEquals("1", Iterables.getOnlyElement(row.getDimension("jq_omg"))); + Assert.assertEquals(ImmutableMap.of("mg", 1L), row.getRaw("o")); // Header verification Assert.assertEquals("application/json", Iterables.getOnlyElement(row.getDimension("kafka.newheader.encoding"))); @@ -342,7 +343,6 @@ public byte[] value() while (iterator.hasNext()) { final InputRow row = iterator.next(); - final MapBasedInputRow mrow = (MapBasedInputRow) row; // Payload verifications Assert.assertEquals(DateTimes.of("2021-06-24"), row.getTimestamp()); Assert.assertEquals("x", Iterables.getOnlyElement(row.getDimension("foo"))); @@ -350,6 +350,7 @@ public byte[] value() Assert.assertEquals("4", Iterables.getOnlyElement(row.getDimension("root_baz"))); Assert.assertEquals("1", Iterables.getOnlyElement(row.getDimension("path_omg"))); Assert.assertEquals("1", Iterables.getOnlyElement(row.getDimension("jq_omg"))); + Assert.assertEquals(ImmutableMap.of("mg", 1L), row.getRaw("o")); // Header verification Assert.assertEquals("application/json", Iterables.getOnlyElement(row.getDimension("kafka.newheader.encoding"))); @@ -444,6 +445,7 @@ public void testWithOutKeyAndHeaderSpecs() throws IOException Assert.assertEquals("4", Iterables.getOnlyElement(row.getDimension("root_baz"))); Assert.assertEquals("1", Iterables.getOnlyElement(row.getDimension("path_omg"))); Assert.assertEquals("1", Iterables.getOnlyElement(row.getDimension("jq_omg"))); + Assert.assertEquals(ImmutableMap.of("mg", 1L), row.getRaw("o")); numActualIterations++; } @@ -521,6 +523,7 @@ public void testWithMultipleMixedRecords() throws IOException Assert.assertEquals("4", Iterables.getOnlyElement(row.getDimension("root_baz"))); Assert.assertEquals("1", Iterables.getOnlyElement(row.getDimension("path_omg"))); Assert.assertEquals("1", Iterables.getOnlyElement(row.getDimension("jq_omg"))); + Assert.assertEquals(ImmutableMap.of("mg", 1L), row.getRaw("o")); Assert.assertEquals(String.valueOf(i), Iterables.getOnlyElement(row.getDimension("index"))); From 37b8d4861c928a9b9e5cbde74df7f8a8a5562e45 Mon Sep 17 00:00:00 2001 From: Clint Wylie Date: Mon, 28 Nov 2022 12:29:43 -0800 Subject: [PATCH 050/141] fix issues with nested data conversion (#13407) --- .../util/common/parsers/ObjectFlatteners.java | 13 +++++- .../common/parsers/ObjectFlattenersTest.java | 14 ++++++- .../orc-extensions/example/test_simple.orc | Bin 0 -> 506 bytes .../druid/data/input/orc/OrcReaderTest.java | 39 ++++++++++++++++++ 4 files changed, 63 insertions(+), 3 deletions(-) create mode 100644 extensions-core/orc-extensions/example/test_simple.orc diff --git a/core/src/main/java/org/apache/druid/java/util/common/parsers/ObjectFlatteners.java b/core/src/main/java/org/apache/druid/java/util/common/parsers/ObjectFlatteners.java index 4e955a4d50b8..98f087333a55 100644 --- a/core/src/main/java/org/apache/druid/java/util/common/parsers/ObjectFlatteners.java +++ b/core/src/main/java/org/apache/druid/java/util/common/parsers/ObjectFlatteners.java @@ -29,6 +29,7 @@ import java.util.AbstractMap; import java.util.ArrayList; import java.util.Collection; +import java.util.Collections; import java.util.HashMap; import java.util.LinkedHashMap; import java.util.LinkedHashSet; @@ -250,15 +251,23 @@ default Function makeJsonTreeExtractor(List nodes) */ default Map toMap(T obj) { - return (Map) toPlainJavaType(obj); + final Object mapOrNull = toPlainJavaType(obj); + if (mapOrNull == null) { + return Collections.emptyMap(); + } + return (Map) mapOrNull; } /** * Recursively traverse "json" object using a {@link JsonProvider}, converting to Java {@link Map} and {@link List}, * potentially transforming via {@link #finalizeConversionForMap} as we go */ + @Nullable default Object toPlainJavaType(Object o) { + if (o == null) { + return null; + } final JsonProvider jsonProvider = getJsonProvider(); if (jsonProvider.isMap(o)) { Map actualMap = new HashMap<>(); @@ -287,7 +296,7 @@ default Object toPlainJavaType(Object o) return finalizeConversionForMap(actualList); } // unknown, just pass it through - return o; + return finalizeConversionForMap(o); } /** diff --git a/core/src/test/java/org/apache/druid/java/util/common/parsers/ObjectFlattenersTest.java b/core/src/test/java/org/apache/druid/java/util/common/parsers/ObjectFlattenersTest.java index 2b610690db0a..e0b0fbcc510c 100644 --- a/core/src/test/java/org/apache/druid/java/util/common/parsers/ObjectFlattenersTest.java +++ b/core/src/test/java/org/apache/druid/java/util/common/parsers/ObjectFlattenersTest.java @@ -23,6 +23,7 @@ import com.fasterxml.jackson.databind.JsonNode; import com.fasterxml.jackson.databind.ObjectMapper; import com.google.common.collect.ImmutableList; +import com.google.common.collect.ImmutableMap; import com.google.common.collect.ImmutableSet; import org.junit.Assert; import org.junit.Test; @@ -32,12 +33,14 @@ public class ObjectFlattenersTest { private static final String SOME_JSON = "{\"foo\": null, \"bar\": 1}"; + + private static final ObjectFlatteners.FlattenerMaker FLATTENER_MAKER = new JSONFlattenerMaker(true); private static final ObjectFlattener FLATTENER = ObjectFlatteners.create( new JSONPathSpec( true, ImmutableList.of(new JSONPathFieldSpec(JSONPathFieldType.PATH, "extract", "$.bar")) ), - new JSONFlattenerMaker(true) + FLATTENER_MAKER ); private static final ObjectMapper OBJECT_MAPPER = new ObjectMapper(); @@ -62,4 +65,13 @@ public void testToMap() throws JsonProcessingException Assert.assertNull(flat.get("foo")); Assert.assertEquals(1, flat.get("bar")); } + + @Test + public void testToMapNull() throws JsonProcessingException + { + JsonNode node = OBJECT_MAPPER.readTree("null"); + Map flat = FLATTENER.toMap(node); + Assert.assertNull(FLATTENER_MAKER.toPlainJavaType(node)); + Assert.assertEquals(ImmutableMap.of(), flat); + } } diff --git a/extensions-core/orc-extensions/example/test_simple.orc b/extensions-core/orc-extensions/example/test_simple.orc new file mode 100644 index 0000000000000000000000000000000000000000..31c5391fb687961c1837637c7954841c96235b02 GIT binary patch literal 506 zcmeYda+YOa;Nsz8VE_Ul77nHWh5!bJ$D5u%N@-wJROVAtmDNjBW>!{?Wp-BJGBnu0 zq^#JoKrklYka`0ngBnn&(9=f=4NQuDd~UjFdTz%U+!)M&qCro<1_<)GsoLpHdo)SO zS881S%MxY5S4l!>dv>mXClL9Pw|wPN^O9;eT>H}$Xd zi^#fO;Bj1y+1htb|D`3DoqC;iYHvS(WOup#gsJSOtiITM=}=w^&P({Qq@wOO;_;Hi#gPtSJm8z&B&@VI_Jpu>rWXUzc*4#f-> z@hKk|Up;)1a3$$kQVLJP7NLg;A&(v>Y3gYji7-tTl40!O6aHf(F*$zS`-7&-bG b1sEk78d#K=714cMB#Ff;oHIZFTlJD96o literal 0 HcmV?d00001 diff --git a/extensions-core/orc-extensions/src/test/java/org/apache/druid/data/input/orc/OrcReaderTest.java b/extensions-core/orc-extensions/src/test/java/org/apache/druid/data/input/orc/OrcReaderTest.java index d0f012245b03..03083f61ebc8 100644 --- a/extensions-core/orc-extensions/src/test/java/org/apache/druid/data/input/orc/OrcReaderTest.java +++ b/extensions-core/orc-extensions/src/test/java/org/apache/druid/data/input/orc/OrcReaderTest.java @@ -29,6 +29,7 @@ import org.apache.druid.data.input.InputRowSchema; import org.apache.druid.data.input.impl.DimensionsSpec; import org.apache.druid.data.input.impl.FileEntity; +import org.apache.druid.data.input.impl.StringDimensionSchema; import org.apache.druid.data.input.impl.TimestampSpec; import org.apache.druid.java.util.common.DateTimes; import org.apache.druid.java.util.common.parsers.CloseableIterator; @@ -569,6 +570,44 @@ public void testNestedArray() throws IOException } } + @Test + public void testSimpleNullValues() throws IOException + { + final InputFormat inputFormat = new OrcInputFormat( + new JSONPathSpec( + true, + ImmutableList.of() + ), + null, + new Configuration() + ); + final InputEntityReader reader = createReader( + new TimestampSpec("timestamp", "auto", null), + new DimensionsSpec( + ImmutableList.of( + new StringDimensionSchema("c1"), + new StringDimensionSchema("c2") + ) + ), + inputFormat, + "example/test_simple.orc" + ); + try (CloseableIterator iterator = reader.read()) { + Assert.assertTrue(iterator.hasNext()); + InputRow row = iterator.next(); + + Assert.assertEquals(DateTimes.of("2022-01-01T00:00:00.000Z"), row.getTimestamp()); + Assert.assertEquals("true", Iterables.getOnlyElement(row.getDimension("c1"))); + Assert.assertEquals("str1", Iterables.getOnlyElement(row.getDimension("c2"))); + + row = iterator.next(); + Assert.assertEquals(DateTimes.of("2022-01-02T00:00:00.000Z"), row.getTimestamp()); + Assert.assertEquals(ImmutableList.of(), row.getDimension("c1")); + Assert.assertEquals(ImmutableList.of(), row.getDimension("c2")); + Assert.assertFalse(iterator.hasNext()); + } + } + private InputEntityReader createReader( TimestampSpec timestampSpec, DimensionsSpec dimensionsSpec, From d8f4353c43ccc2578b2c58488e10ff8dfb38b6c5 Mon Sep 17 00:00:00 2001 From: Vadim Ogievetsky Date: Mon, 28 Nov 2022 16:50:38 -0800 Subject: [PATCH 051/141] Web console: be more robust to aux queries failing and improve kill tasks (#13431) * be more robust to aux queries failing * feedback fixes * remove empty block * fix spelling * remove killAllDataSources from the console --- .../src/components/auto-form/auto-form.scss | 12 - .../form-group-with-info.scss | 14 + .../table-clickable-cell.scss | 4 + .../table-clickable-cell.tsx | 11 +- .../warning-checklist/warning-checklist.tsx | 18 +- .../async-action-dialog.tsx | 2 +- web-console/src/dialogs/index.ts | 1 + .../kill-datasource-dialog.tsx | 110 +++++ .../compaction-status.spec.ts | 41 +- .../compaction-status/compaction-status.ts | 22 +- .../coordinator-dynamic-config.tsx | 13 +- .../datasources-view/datasources-view.tsx | 390 ++++++++++-------- .../src/views/services-view/services-view.tsx | 215 ++++++---- 13 files changed, 524 insertions(+), 329 deletions(-) create mode 100644 web-console/src/dialogs/kill-datasource-dialog/kill-datasource-dialog.tsx diff --git a/web-console/src/components/auto-form/auto-form.scss b/web-console/src/components/auto-form/auto-form.scss index 36f8c34d92d2..c303abc294f5 100644 --- a/web-console/src/components/auto-form/auto-form.scss +++ b/web-console/src/components/auto-form/auto-form.scss @@ -16,19 +16,7 @@ * limitations under the License. */ -@import '../../variables'; - .auto-form { - // Popover in info label - label.#{$bp-ns}-label { - position: relative; - - .#{$bp-ns}-text-muted { - position: absolute; - right: 0; - } - } - .custom-input input { cursor: pointer; } diff --git a/web-console/src/components/form-group-with-info/form-group-with-info.scss b/web-console/src/components/form-group-with-info/form-group-with-info.scss index c9587cb088fa..a64c6d29274c 100644 --- a/web-console/src/components/form-group-with-info/form-group-with-info.scss +++ b/web-console/src/components/form-group-with-info/form-group-with-info.scss @@ -19,6 +19,20 @@ @import '../../variables'; .form-group-with-info { + label.#{$bp-ns}-label { + position: relative; + + .#{$bp-ns}-text-muted { + position: absolute; + right: 0; + + // This is only needed because BP4 alerts are too agro in setting CSS on icons + .#{$bp-ns}-icon { + margin-right: 0; + } + } + } + .#{$bp-ns}-text-muted .#{$bp-ns}-popover2-target { margin-top: 0; } diff --git a/web-console/src/components/table-clickable-cell/table-clickable-cell.scss b/web-console/src/components/table-clickable-cell/table-clickable-cell.scss index d6f6f8b2d7f7..5c5991df54e2 100644 --- a/web-console/src/components/table-clickable-cell/table-clickable-cell.scss +++ b/web-console/src/components/table-clickable-cell/table-clickable-cell.scss @@ -24,6 +24,10 @@ overflow: hidden; text-overflow: ellipsis; + &.disabled { + cursor: not-allowed; + } + .hover-icon { position: absolute; top: $table-cell-v-padding; diff --git a/web-console/src/components/table-clickable-cell/table-clickable-cell.tsx b/web-console/src/components/table-clickable-cell/table-clickable-cell.tsx index cc8cfd71e56d..7e4c66fdd5e5 100644 --- a/web-console/src/components/table-clickable-cell/table-clickable-cell.tsx +++ b/web-console/src/components/table-clickable-cell/table-clickable-cell.tsx @@ -27,18 +27,23 @@ export interface TableClickableCellProps { onClick: MouseEventHandler; hoverIcon?: IconName; title?: string; + disabled?: boolean; children?: ReactNode; } export const TableClickableCell = React.memo(function TableClickableCell( props: TableClickableCellProps, ) { - const { className, onClick, hoverIcon, title, children } = props; + const { className, onClick, hoverIcon, title, disabled, children } = props; return ( -
+
{children} - {hoverIcon && } + {hoverIcon && !disabled && }
); }); diff --git a/web-console/src/components/warning-checklist/warning-checklist.tsx b/web-console/src/components/warning-checklist/warning-checklist.tsx index 449ad970045d..5c74cbdb08a0 100644 --- a/web-console/src/components/warning-checklist/warning-checklist.tsx +++ b/web-console/src/components/warning-checklist/warning-checklist.tsx @@ -17,29 +17,31 @@ */ import { Switch } from '@blueprintjs/core'; -import React, { useState } from 'react'; +import React, { ReactNode, useState } from 'react'; export interface WarningChecklistProps { - checks: string[]; - onChange: (allChecked: boolean) => void; + checks: ReactNode[]; + onChange(allChecked: boolean): void; } export const WarningChecklist = React.memo(function WarningChecklist(props: WarningChecklistProps) { const { checks, onChange } = props; - const [checked, setChecked] = useState>({}); + const [checked, setChecked] = useState>({}); - function doCheck(check: string) { + function doCheck(checkIndex: number) { const newChecked = { ...checked }; - newChecked[check] = !newChecked[check]; + newChecked[checkIndex] = !newChecked[checkIndex]; setChecked(newChecked); - onChange(checks.every(check => newChecked[check])); + onChange(checks.every((_, i) => newChecked[i])); } return (
{checks.map((check, i) => ( - doCheck(check)} /> + doCheck(i)}> + {check} + ))}
); diff --git a/web-console/src/dialogs/async-action-dialog/async-action-dialog.tsx b/web-console/src/dialogs/async-action-dialog/async-action-dialog.tsx index f892936bab77..0d8cf385a5b6 100644 --- a/web-console/src/dialogs/async-action-dialog/async-action-dialog.tsx +++ b/web-console/src/dialogs/async-action-dialog/async-action-dialog.tsx @@ -47,7 +47,7 @@ export interface AsyncActionDialogProps { intent?: Intent; successText: string; failText: string; - warningChecks?: string[]; + warningChecks?: ReactNode[]; children?: ReactNode; } diff --git a/web-console/src/dialogs/index.ts b/web-console/src/dialogs/index.ts index 9509442c8bd2..588257c84e74 100644 --- a/web-console/src/dialogs/index.ts +++ b/web-console/src/dialogs/index.ts @@ -24,6 +24,7 @@ export * from './diff-dialog/diff-dialog'; export * from './doctor-dialog/doctor-dialog'; export * from './edit-context-dialog/edit-context-dialog'; export * from './history-dialog/history-dialog'; +export * from './kill-datasource-dialog/kill-datasource-dialog'; export * from './lookup-edit-dialog/lookup-edit-dialog'; export * from './numeric-input-dialog/numeric-input-dialog'; export * from './overlord-dynamic-config-dialog/overlord-dynamic-config-dialog'; diff --git a/web-console/src/dialogs/kill-datasource-dialog/kill-datasource-dialog.tsx b/web-console/src/dialogs/kill-datasource-dialog/kill-datasource-dialog.tsx new file mode 100644 index 000000000000..3eb7e9fdf243 --- /dev/null +++ b/web-console/src/dialogs/kill-datasource-dialog/kill-datasource-dialog.tsx @@ -0,0 +1,110 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +import { Code, Intent } from '@blueprintjs/core'; +import React, { useState } from 'react'; + +import { FormGroupWithInfo, PopoverText } from '../../components'; +import { SuggestibleInput } from '../../components/suggestible-input/suggestible-input'; +import { Api } from '../../singletons'; +import { uniq } from '../../utils'; +import { AsyncActionDialog } from '../async-action-dialog/async-action-dialog'; + +function getSuggestions(): string[] { + // Default to a data 24h ago so as not to cause a conflict between streaming ingestion and kill tasks + const end = new Date(Date.now() - 24 * 60 * 60 * 1000).toISOString(); + const startOfDay = end.slice(0, 10); + const startOfMonth = end.slice(0, 7) + '-01'; + const startOfYear = end.slice(0, 4) + '-01-01'; + + return uniq([ + `1000-01-01/${startOfDay}`, + `1000-01-01/${startOfMonth}`, + `1000-01-01/${startOfYear}`, + '1000-01-01/3000-01-01', + ]); +} + +export interface KillDatasourceDialogProps { + datasource: string; + onClose(): void; + onSuccess(): void; +} + +export const KillDatasourceDialog = function KillDatasourceDialog( + props: KillDatasourceDialogProps, +) { + const { datasource, onClose, onSuccess } = props; + const suggestions = getSuggestions(); + const [interval, setInterval] = useState(suggestions[0]); + + return ( + { + const resp = await Api.instance.delete( + `/druid/coordinator/v1/datasources/${Api.encodePath( + datasource, + )}?kill=true&interval=${Api.encodePath(interval)}`, + {}, + ); + return resp.data; + }} + confirmButtonText="Permanently delete unused segments" + successText="Kill task was issued. Unused segments in datasource will be deleted" + failText="Failed submit kill task" + intent={Intent.DANGER} + onClose={onClose} + onSuccess={onSuccess} + warningChecks={[ + <> + I understand that this operation will delete all metadata about the unused segments of{' '} + {datasource} and removes them from deep storage. + , + 'I understand that this operation cannot be undone.', + ]} + > +

+ Are you sure you want to permanently delete unused segments in {datasource}? +

+

This action is not reversible and the data deleted will be lost.

+ +

+ The range of time over which to delete unused segments specified in ISO8601 interval + format. +

+

+ If you have streaming ingestion running make sure that your interval range doe not + overlap with intervals where streaming data is being added - otherwise the kill task + will not start. +

+ + } + > + setInterval(s || '')} + suggestions={suggestions} + /> + + + ); +}; diff --git a/web-console/src/druid-models/compaction-status/compaction-status.spec.ts b/web-console/src/druid-models/compaction-status/compaction-status.spec.ts index 8ed0c5141366..9d1254090bf5 100644 --- a/web-console/src/druid-models/compaction-status/compaction-status.spec.ts +++ b/web-console/src/druid-models/compaction-status/compaction-status.spec.ts @@ -18,11 +18,7 @@ import { CompactionConfig } from '../compaction-config/compaction-config'; -import { - CompactionStatus, - formatCompactionConfigAndStatus, - zeroCompactionStatus, -} from './compaction-status'; +import { CompactionStatus, formatCompactionInfo, zeroCompactionStatus } from './compaction-status'; describe('compaction status', () => { const BASIC_CONFIG: CompactionConfig = {}; @@ -61,27 +57,30 @@ describe('compaction status', () => { }); it('formatCompactionConfigAndStatus', () => { - expect(formatCompactionConfigAndStatus(undefined, undefined)).toEqual('Not enabled'); + expect(formatCompactionInfo({})).toEqual('Not enabled'); - expect(formatCompactionConfigAndStatus(BASIC_CONFIG, undefined)).toEqual('Awaiting first run'); + expect(formatCompactionInfo({ config: BASIC_CONFIG })).toEqual('Awaiting first run'); - expect(formatCompactionConfigAndStatus(undefined, ZERO_STATUS)).toEqual('Not enabled'); + expect(formatCompactionInfo({ status: ZERO_STATUS })).toEqual('Not enabled'); - expect(formatCompactionConfigAndStatus(BASIC_CONFIG, ZERO_STATUS)).toEqual('Running'); + expect(formatCompactionInfo({ config: BASIC_CONFIG, status: ZERO_STATUS })).toEqual('Running'); expect( - formatCompactionConfigAndStatus(BASIC_CONFIG, { - dataSource: 'tbl', - scheduleStatus: 'RUNNING', - bytesAwaitingCompaction: 0, - bytesCompacted: 100, - bytesSkipped: 0, - segmentCountAwaitingCompaction: 0, - segmentCountCompacted: 10, - segmentCountSkipped: 0, - intervalCountAwaitingCompaction: 0, - intervalCountCompacted: 10, - intervalCountSkipped: 0, + formatCompactionInfo({ + config: BASIC_CONFIG, + status: { + dataSource: 'tbl', + scheduleStatus: 'RUNNING', + bytesAwaitingCompaction: 0, + bytesCompacted: 100, + bytesSkipped: 0, + segmentCountAwaitingCompaction: 0, + segmentCountCompacted: 10, + segmentCountSkipped: 0, + intervalCountAwaitingCompaction: 0, + intervalCountCompacted: 10, + intervalCountSkipped: 0, + }, }), ).toEqual('Fully compacted'); }); diff --git a/web-console/src/druid-models/compaction-status/compaction-status.ts b/web-console/src/druid-models/compaction-status/compaction-status.ts index 2982d9b69e17..d17f2c44fda4 100644 --- a/web-console/src/druid-models/compaction-status/compaction-status.ts +++ b/web-console/src/druid-models/compaction-status/compaction-status.ts @@ -50,19 +50,19 @@ export function zeroCompactionStatus(compactionStatus: CompactionStatus): boolea ); } -export function formatCompactionConfigAndStatus( - compactionConfig: CompactionConfig | undefined, - compactionStatus: CompactionStatus | undefined, -) { - if (compactionConfig) { - if (compactionStatus) { - if ( - compactionStatus.bytesAwaitingCompaction === 0 && - !zeroCompactionStatus(compactionStatus) - ) { +export interface CompactionInfo { + config?: CompactionConfig; + status?: CompactionStatus; +} + +export function formatCompactionInfo(compaction: CompactionInfo) { + const { config, status } = compaction; + if (config) { + if (status) { + if (status.bytesAwaitingCompaction === 0 && !zeroCompactionStatus(status)) { return 'Fully compacted'; } else { - return capitalizeFirst(compactionStatus.scheduleStatus); + return capitalizeFirst(status.scheduleStatus); } } else { return 'Awaiting first run'; diff --git a/web-console/src/druid-models/coordinator-dynamic-config/coordinator-dynamic-config.tsx b/web-console/src/druid-models/coordinator-dynamic-config/coordinator-dynamic-config.tsx index eeb25db09c4b..ca957309ff83 100644 --- a/web-console/src/druid-models/coordinator-dynamic-config/coordinator-dynamic-config.tsx +++ b/web-console/src/druid-models/coordinator-dynamic-config/coordinator-dynamic-config.tsx @@ -69,20 +69,9 @@ export const COORDINATOR_DYNAMIC_CONFIG_FIELDS: Field[ ), }, - { - name: 'killAllDataSources', - type: 'boolean', - defaultValue: false, - info: ( - <> - Send kill tasks for ALL dataSources if property druid.coordinator.kill.on is - true. If this is set to true then killDataSourceWhitelist must not be specified - or be empty list. - - ), - }, { name: 'killDataSourceWhitelist', + label: 'Kill datasource whitelist', type: 'string-array', emptyValue: [], info: ( diff --git a/web-console/src/views/datasources-view/datasources-view.tsx b/web-console/src/views/datasources-view/datasources-view.tsx index 7f384fc8bc2a..c38a42b63938 100644 --- a/web-console/src/views/datasources-view/datasources-view.tsx +++ b/web-console/src/views/datasources-view/datasources-view.tsx @@ -36,12 +36,18 @@ import { TableColumnSelector, ViewControlBar, } from '../../components'; -import { AsyncActionDialog, CompactionDialog, RetentionDialog } from '../../dialogs'; +import { + AsyncActionDialog, + CompactionDialog, + KillDatasourceDialog, + RetentionDialog, +} from '../../dialogs'; import { DatasourceTableActionDialog } from '../../dialogs/datasource-table-action-dialog/datasource-table-action-dialog'; import { CompactionConfig, + CompactionInfo, CompactionStatus, - formatCompactionConfigAndStatus, + formatCompactionInfo, QueryWithContext, zeroCompactionStatus, } from '../../druid-models'; @@ -208,9 +214,8 @@ function segmentGranularityCountsToRank(row: DatasourceQueryResultRow): number { } interface Datasource extends DatasourceQueryResultRow { - readonly rules: Rule[]; - readonly compactionConfig?: CompactionConfig; - readonly compactionStatus?: CompactionStatus; + readonly rules?: Rule[]; + readonly compaction?: CompactionInfo; readonly unused?: boolean; } @@ -220,7 +225,7 @@ function makeUnusedDatasource(datasource: string): Datasource { interface DatasourcesAndDefaultRules { readonly datasources: Datasource[]; - readonly defaultRules: Rule[]; + readonly defaultRules?: Rule[]; } interface RetentionDialogOpenOn { @@ -433,43 +438,85 @@ ORDER BY 1`; let unused: string[] = []; if (showUnused) { - const unusedResp = await Api.instance.get( - '/druid/coordinator/v1/metadata/datasources?includeUnused', + try { + unused = ( + await Api.instance.get( + '/druid/coordinator/v1/metadata/datasources?includeUnused', + ) + ).data.filter(d => !seen[d]); + } catch { + AppToaster.show({ + icon: IconNames.ERROR, + intent: Intent.DANGER, + message: 'Could not get the list of unused datasources', + }); + } + } + + let rules: Record = {}; + try { + rules = (await Api.instance.get>('/druid/coordinator/v1/rules')) + .data; + } catch { + AppToaster.show({ + icon: IconNames.ERROR, + intent: Intent.DANGER, + message: 'Could not get load rules', + }); + } + + let compactionConfigs: Record | undefined; + try { + const compactionConfigsResp = await Api.instance.get<{ + compactionConfigs: CompactionConfig[]; + }>('/druid/coordinator/v1/config/compaction'); + compactionConfigs = lookupBy( + compactionConfigsResp.data.compactionConfigs || [], + c => c.dataSource, ); - unused = unusedResp.data.filter(d => !seen[d]); + } catch { + AppToaster.show({ + icon: IconNames.ERROR, + intent: Intent.DANGER, + message: 'Could not get compaction configs', + }); } - const rulesResp = await Api.instance.get>( - '/druid/coordinator/v1/rules', - ); - const rules = rulesResp.data; - - const compactionConfigsResp = await Api.instance.get<{ - compactionConfigs: CompactionConfig[]; - }>('/druid/coordinator/v1/config/compaction'); - const compactionConfigs = lookupBy( - compactionConfigsResp.data.compactionConfigs || [], - c => c.dataSource, - ); - - const compactionStatusesResp = await Api.instance.get<{ latestStatus: CompactionStatus[] }>( - '/druid/coordinator/v1/compaction/status', - ); - const compactionStatuses = lookupBy( - compactionStatusesResp.data.latestStatus || [], - c => c.dataSource, - ); + let compactionStatuses: Record | undefined; + if (compactionConfigs) { + // Don't bother getting the statuses if we can not even get the configs + try { + const compactionStatusesResp = await Api.instance.get<{ + latestStatus: CompactionStatus[]; + }>('/druid/coordinator/v1/compaction/status'); + compactionStatuses = lookupBy( + compactionStatusesResp.data.latestStatus || [], + c => c.dataSource, + ); + } catch { + AppToaster.show({ + icon: IconNames.ERROR, + intent: Intent.DANGER, + message: 'Could not get compaction statuses', + }); + } + } return { datasources: datasources.concat(unused.map(makeUnusedDatasource)).map(ds => { return { ...ds, - rules: rules[ds.datasource] || [], - compactionConfig: compactionConfigs[ds.datasource], - compactionStatus: compactionStatuses[ds.datasource], + rules: rules[ds.datasource], + compaction: + compactionConfigs && compactionStatuses + ? { + config: compactionConfigs[ds.datasource], + status: compactionStatuses[ds.datasource], + } + : undefined, }; }), - defaultRules: rules[DEFAULT_RULES_KEY] || [], + defaultRules: rules[DEFAULT_RULES_KEY], }; }, onStateChange: datasourcesAndDefaultRulesState => { @@ -633,36 +680,15 @@ ORDER BY 1`; if (!killDatasource) return; return ( - { - const resp = await Api.instance.delete( - `/druid/coordinator/v1/datasources/${Api.encodePath( - killDatasource, - )}?kill=true&interval=1000/3000`, - {}, - ); - return resp.data; - }} - confirmButtonText="Permanently delete unused segments" - successText="Kill task was issued. Unused segments in datasource will be deleted" - failText="Failed submit kill task" - intent={Intent.DANGER} + { this.setState({ killDatasource: undefined }); }} onSuccess={() => { this.fetchDatasourceData(); }} - warningChecks={[ - `I understand that this operation will delete all metadata about the unused segments of ${killDatasource} and removes them from deep storage.`, - 'I understand that this operation cannot be undone.', - ]} - > -

- {`Are you sure you want to permanently delete unused segments in '${killDatasource}'?`} -

-

This action is not reversible and the data deleted will be lost.

- + /> ); } @@ -756,20 +782,20 @@ ORDER BY 1`; this.setState({ retentionDialogOpenOn: undefined }); setTimeout(() => { this.setState(state => { - const datasourcesAndDefaultRules = state.datasourcesAndDefaultRulesState.data; - if (!datasourcesAndDefaultRules) return {}; + const defaultRules = state.datasourcesAndDefaultRulesState.data?.defaultRules; + if (!defaultRules) return {}; return { retentionDialogOpenOn: { datasource: '_default', - rules: datasourcesAndDefaultRules.defaultRules, + rules: defaultRules, }, }; }); }, 50); }; - private readonly saveCompaction = async (compactionConfig: any) => { + private readonly saveCompaction = async (compactionConfig: CompactionConfig) => { if (!compactionConfig) return; try { await Api.instance.post(`/druid/coordinator/v1/config/compaction`, compactionConfig); @@ -819,8 +845,8 @@ ORDER BY 1`; getDatasourceActions( datasource: string, unused: boolean | undefined, - rules: Rule[], - compactionConfig: CompactionConfig | undefined, + rules: Rule[] | undefined, + compactionInfo: CompactionInfo | undefined, ): BasicAction[] { const { goToQuery, goToTask, capabilities } = this.props; @@ -863,82 +889,83 @@ ORDER BY 1`; }, ]; } else { - return goToActions.concat([ - { - icon: IconNames.AUTOMATIC_UPDATES, - title: 'Edit retention rules', - onAction: () => { - this.setState({ - retentionDialogOpenOn: { - datasource, - rules, - }, - }); + return goToActions.concat( + compact([ + { + icon: IconNames.AUTOMATIC_UPDATES, + title: 'Edit retention rules', + onAction: () => { + this.setState({ + retentionDialogOpenOn: { + datasource, + rules: rules || [], + }, + }); + }, }, - }, - { - icon: IconNames.REFRESH, - title: 'Mark as used all segments (will lead to reapplying retention rules)', - onAction: () => - this.setState({ - datasourceToMarkAllNonOvershadowedSegmentsAsUsedIn: datasource, - }), - }, - { - icon: IconNames.COMPRESSED, - title: 'Edit compaction configuration', - onAction: () => { - this.setState({ - compactionDialogOpenOn: { - datasource, - compactionConfig, - }, - }); + { + icon: IconNames.REFRESH, + title: 'Mark as used all segments (will lead to reapplying retention rules)', + onAction: () => + this.setState({ + datasourceToMarkAllNonOvershadowedSegmentsAsUsedIn: datasource, + }), }, - }, - { - icon: IconNames.EXPORT, - title: 'Mark as used segments by interval', - - onAction: () => - this.setState({ - datasourceToMarkSegmentsByIntervalIn: datasource, - useUnuseAction: 'use', - }), - }, - { - icon: IconNames.IMPORT, - title: 'Mark as unused segments by interval', - - onAction: () => - this.setState({ - datasourceToMarkSegmentsByIntervalIn: datasource, - useUnuseAction: 'unuse', - }), - }, - { - icon: IconNames.IMPORT, - title: 'Mark as unused all segments', - intent: Intent.DANGER, - onAction: () => this.setState({ datasourceToMarkAsUnusedAllSegmentsIn: datasource }), - }, - { - icon: IconNames.TRASH, - title: 'Delete unused segments (issue kill task)', - intent: Intent.DANGER, - onAction: () => this.setState({ killDatasource: datasource }), - }, - ]); + compactionInfo + ? { + icon: IconNames.COMPRESSED, + title: 'Edit compaction configuration', + onAction: () => { + this.setState({ + compactionDialogOpenOn: { + datasource, + compactionConfig: compactionInfo.config, + }, + }); + }, + } + : undefined, + { + icon: IconNames.EXPORT, + title: 'Mark as used segments by interval', + + onAction: () => + this.setState({ + datasourceToMarkSegmentsByIntervalIn: datasource, + useUnuseAction: 'use', + }), + }, + { + icon: IconNames.IMPORT, + title: 'Mark as unused segments by interval', + + onAction: () => + this.setState({ + datasourceToMarkSegmentsByIntervalIn: datasource, + useUnuseAction: 'unuse', + }), + }, + { + icon: IconNames.IMPORT, + title: 'Mark as unused all segments', + intent: Intent.DANGER, + onAction: () => this.setState({ datasourceToMarkAsUnusedAllSegmentsIn: datasource }), + }, + { + icon: IconNames.TRASH, + title: 'Delete unused segments (issue kill task)', + intent: Intent.DANGER, + onAction: () => this.setState({ killDatasource: datasource }), + }, + ]), + ); } } private renderRetentionDialog(): JSX.Element | undefined { const { retentionDialogOpenOn, tiersState, datasourcesAndDefaultRulesState } = this.state; - const { defaultRules } = datasourcesAndDefaultRulesState.data || { - datasources: [], - defaultRules: [], - }; - if (!retentionDialogOpenOn) return; + const defaultRules = datasourcesAndDefaultRulesState.data?.defaultRules; + if (!retentionDialogOpenOn || !defaultRules) return; return ( !d.unused); @@ -1009,8 +1034,8 @@ ORDER BY 1`; const replicatedSizeValues = datasources.map(d => formatReplicatedSize(d.replicated_size)); const leftToBeCompactedValues = datasources.map(d => - d.compactionStatus - ? formatLeftToBeCompacted(d.compactionStatus.bytesAwaitingCompaction) + d.compaction?.status + ? formatLeftToBeCompacted(d.compaction?.status.bytesAwaitingCompaction) : '-', ); @@ -1297,24 +1322,26 @@ ORDER BY 1`; Header: 'Compaction', show: capabilities.hasCoordinatorAccess() && visibleColumns.shown('Compaction'), id: 'compactionStatus', - accessor: row => Boolean(row.compactionStatus), + accessor: row => Boolean(row.compaction?.status), filterable: false, width: 150, Cell: ({ original }) => { - const { datasource, compactionConfig, compactionStatus } = original as Datasource; + const { datasource, compaction } = original as Datasource; return ( + disabled={!compaction} + onClick={() => { + if (!compaction) return; this.setState({ compactionDialogOpenOn: { datasource, - compactionConfig, + compactionConfig: compaction.config, }, - }) - } + }); + }} hoverIcon={IconNames.EDIT} > - {formatCompactionConfigAndStatus(compactionConfig, compactionStatus)} + {compaction ? formatCompactionInfo(compaction) : 'Could not get compaction info'} ); }, @@ -1324,17 +1351,22 @@ ORDER BY 1`; show: capabilities.hasCoordinatorAccess() && visibleColumns.shown('% Compacted'), id: 'percentCompacted', width: 200, - accessor: ({ compactionStatus }) => - compactionStatus && compactionStatus.bytesCompacted - ? compactionStatus.bytesCompacted / - (compactionStatus.bytesAwaitingCompaction + compactionStatus.bytesCompacted) - : 0, + accessor: ({ compaction }) => { + const status = compaction?.status; + return status?.bytesCompacted + ? status.bytesCompacted / (status.bytesAwaitingCompaction + status.bytesCompacted) + : 0; + }, filterable: false, className: 'padded', Cell: ({ original }) => { - const { compactionStatus } = original as Datasource; + const { compaction } = original as Datasource; + if (!compaction) { + return 'Could not get compaction info'; + } - if (!compactionStatus || zeroCompactionStatus(compactionStatus)) { + const { status } = compaction; + if (!status || zeroCompactionStatus(status)) { return ( <>  {' '} @@ -1348,20 +1380,14 @@ ORDER BY 1`; <> {' '}  {' '} {' '} @@ -1369,8 +1395,8 @@ ORDER BY 1`; - (compactionStatus && compactionStatus.bytesAwaitingCompaction) || 0, + accessor: ({ compaction }) => { + const status = compaction?.status; + return status?.bytesAwaitingCompaction || 0; + }, filterable: false, className: 'padded', Cell: ({ original }) => { - const { compactionStatus } = original as Datasource; + const { compaction } = original as Datasource; + if (!compaction) { + return 'Could not get compaction info'; + } - if (!compactionStatus) { + const { status } = compaction; + if (!status) { return ; } return ( ); @@ -1408,26 +1440,30 @@ ORDER BY 1`; Header: 'Retention', show: capabilities.hasCoordinatorAccess() && visibleColumns.shown('Retention'), id: 'retention', - accessor: row => row.rules.length, + accessor: row => row.rules?.length || 0, filterable: false, width: 200, Cell: ({ original }) => { const { datasource, rules } = original as Datasource; return ( + disabled={!defaultRules} + onClick={() => { + if (!defaultRules) return; this.setState({ retentionDialogOpenOn: { datasource, - rules, + rules: rules || [], }, - }) - } + }); + }} hoverIcon={IconNames.EDIT} > - {rules.length + {rules?.length ? DatasourcesView.formatRules(rules) - : `Cluster default: ${DatasourcesView.formatRules(defaultRules)}`} + : defaultRules + ? `Cluster default: ${DatasourcesView.formatRules(defaultRules)}` + : 'Could not get default rules'} ); }, @@ -1440,12 +1476,12 @@ ORDER BY 1`; width: ACTION_COLUMN_WIDTH, filterable: false, Cell: ({ value: datasource, original }) => { - const { unused, rules, compactionConfig } = original as Datasource; + const { unused, rules, compaction } = original as Datasource; const datasourceActions = this.getDatasourceActions( datasource, unused, rules, - compactionConfig, + compaction, ); return ( , - Partial {} - export class ServicesView extends React.PureComponent { private readonly serviceQueryManager: QueryManager; @@ -198,7 +196,7 @@ ORDER BY ) DESC, "service" DESC`; - static async getServices(): Promise { + static async getServices(): Promise { const allServiceResp = await Api.instance.get('/druid/coordinator/v1/servers?simple'); const allServices = allServiceResp.data; return allServices.map((s: any) => { @@ -228,7 +226,7 @@ ORDER BY this.serviceQueryManager = new QueryManager({ processQuery: async capabilities => { - let services: ServiceQueryResultRow[]; + let services: ServiceResultRow[]; if (capabilities.hasSql()) { services = await queryDruidSql({ query: ServicesView.SERVICE_SQL }); } else if (capabilities.hasCoordinatorAccess()) { @@ -238,50 +236,49 @@ ORDER BY } if (capabilities.hasCoordinatorAccess()) { - const loadQueueResponse = await Api.instance.get( - '/druid/coordinator/v1/loadqueue?simple', - ); - const loadQueues: Record = loadQueueResponse.data; - services = services.map(s => { - const loadQueueInfo = loadQueues[s.service]; - if (loadQueueInfo) { - s = { ...s, ...loadQueueInfo }; - } - return s; - }); + try { + const loadQueueInfos = ( + await Api.instance.get>( + '/druid/coordinator/v1/loadqueue?simple', + ) + ).data; + services.forEach(s => { + s.loadQueueInfo = loadQueueInfos[s.service]; + }); + } catch { + AppToaster.show({ + icon: IconNames.ERROR, + intent: Intent.DANGER, + message: 'There was an error getting the load queue info', + }); + } } if (capabilities.hasOverlordAccess()) { - let middleManagers: MiddleManagerQueryResultRow[]; try { - const middleManagerResponse = await Api.instance.get('/druid/indexer/v1/workers'); - middleManagers = middleManagerResponse.data; + const workerInfos = (await Api.instance.get('/druid/indexer/v1/workers')) + .data; + + const workerInfoLookup: Record = lookupBy( + workerInfos, + m => m.worker?.host, + ); + + services.forEach(s => { + s.workerInfo = workerInfoLookup[s.service]; + }); } catch (e) { + // Swallow this error because it simply a reflection of a local task runner. if ( - e.response && - typeof e.response.data === 'object' && - e.response.data.error === 'Task Runner does not support worker listing' + deepGet(e, 'response.data.error') !== 'Task Runner does not support worker listing' ) { - // Swallow this error because it simply a reflection of a local task runner. - middleManagers = []; - } else { - // Otherwise re-throw. - throw e; + AppToaster.show({ + icon: IconNames.ERROR, + intent: Intent.DANGER, + message: 'There was an error getting the worker info', + }); } } - - const middleManagersLookup: Record = lookupBy( - middleManagers, - m => m.worker.host, - ); - - services = services.map(s => { - const middleManagerInfo = middleManagersLookup[s.service]; - if (middleManagerInfo) { - s = { ...s, ...middleManagerInfo }; - } - return s; - }); } return services; @@ -372,7 +369,8 @@ ORDER BY id: 'tier', width: 180, accessor: row => { - return row.tier ? row.tier : row.worker ? row.worker.category : null; + if (row.tier) return row.tier; + return deepGet(row, 'workerInfo.worker.category'); }, Cell: this.renderFilterableCell('tier'), }, @@ -451,9 +449,11 @@ ORDER BY className: 'padded', accessor: row => { if (oneOf(row.service_type, 'middle_manager', 'indexer')) { - return row.worker - ? (Number(row.currCapacityUsed) || 0) / Number(row.worker.capacity) - : null; + const { workerInfo } = row; + if (!workerInfo) return 0; + return ( + (Number(workerInfo.currCapacityUsed) || 0) / Number(workerInfo.worker?.capacity) + ); } else { return row.max_size ? Number(row.curr_size) / Number(row.max_size) : null; } @@ -469,15 +469,21 @@ ORDER BY case 'indexer': case 'middle_manager': { - const originalMiddleManagers: ServiceResultRow[] = row.subRows.map( - r => r._original, + const workerInfos: WorkerInfo[] = filterMap( + row.subRows, + r => r._original.workerInfo, ); + + if (!workerInfos.length) { + return 'Could not get worker infos'; + } + const totalCurrCapacityUsed = sum( - originalMiddleManagers, - s => Number(s.currCapacityUsed) || 0, + workerInfos, + w => Number(w.currCapacityUsed) || 0, ); const totalWorkerCapacity = sum( - originalMiddleManagers, + workerInfos, s => deepGet(s, 'worker.capacity') || 0, ); return `${totalCurrCapacityUsed} / ${totalWorkerCapacity} (total slots)`; @@ -496,8 +502,12 @@ ORDER BY case 'indexer': case 'middle_manager': { - const currCapacityUsed = deepGet(row, 'original.currCapacityUsed') || 0; - const capacity = deepGet(row, 'original.worker.capacity'); + if (!deepGet(row, 'original.workerInfo')) { + return 'Could not get capacity info'; + } + const currCapacityUsed = + deepGet(row, 'original.workerInfo.currCapacityUsed') || 0; + const capacity = deepGet(row, 'original.workerInfo.worker.capacity'); if (typeof capacity === 'number') { return `Slots used: ${currCapacityUsed} of ${capacity}`; } else { @@ -518,30 +528,58 @@ ORDER BY filterable: false, className: 'padded', accessor: row => { - if (oneOf(row.service_type, 'middle_manager', 'indexer')) { - if (deepGet(row, 'worker.version') === '') return 'Disabled'; + switch (row.service_type) { + case 'middle_manager': + case 'indexer': { + if (deepGet(row, 'worker.version') === '') return 'Disabled'; + const { workerInfo } = row; + if (!workerInfo) { + return 'Could not get detail info'; + } - const details: string[] = []; - if (row.lastCompletedTaskTime) { - details.push(`Last completed task: ${row.lastCompletedTaskTime}`); + const details: string[] = []; + if (workerInfo.lastCompletedTaskTime) { + details.push(`Last completed task: ${workerInfo.lastCompletedTaskTime}`); + } + if (workerInfo.blacklistedUntil) { + details.push(`Blacklisted until: ${workerInfo.blacklistedUntil}`); + } + return details.join(' '); } - if (row.blacklistedUntil) { - details.push(`Blacklisted until: ${row.blacklistedUntil}`); + + case 'coordinator': + case 'overlord': + return row.is_leader === 1 ? 'Leader' : ''; + + case 'historical': { + const { loadQueueInfo } = row; + if (!loadQueueInfo) return 0; + return ( + (Number(loadQueueInfo.segmentsToLoad) || 0) + + (Number(loadQueueInfo.segmentsToDrop) || 0) + ); } - return details.join(' '); - } else if (oneOf(row.service_type, 'coordinator', 'overlord')) { - return row.is_leader === 1 ? 'Leader' : ''; - } else { - return (Number(row.segmentsToLoad) || 0) + (Number(row.segmentsToDrop) || 0); + + default: + return 0; } }, Cell: row => { if (row.aggregated) return ''; const { service_type } = row.original; switch (service_type) { + case 'middle_manager': + case 'indexer': + case 'coordinator': + case 'overlord': + return row.value; + case 'historical': { + const { loadQueueInfo } = row.original; + if (!loadQueueInfo) return 'Could not get load queue info'; + const { segmentsToLoad, segmentsToLoadSize, segmentsToDrop, segmentsToDropSize } = - row.original; + loadQueueInfo; return formatQueues( segmentsToLoad, segmentsToLoadSize, @@ -550,23 +588,31 @@ ORDER BY ); } - case 'indexer': - case 'middle_manager': - case 'coordinator': - case 'overlord': - return row.value; - default: return ''; } }, Aggregated: row => { if (row.row._pivotVal !== 'historical') return ''; - const originals: ServiceResultRow[] = row.subRows.map(r => r._original); - const segmentsToLoad = sum(originals, s => Number(s.segmentsToLoad) || 0); - const segmentsToLoadSize = sum(originals, s => Number(s.segmentsToLoadSize) || 0); - const segmentsToDrop = sum(originals, s => Number(s.segmentsToDrop) || 0); - const segmentsToDropSize = sum(originals, s => Number(s.segmentsToDropSize) || 0); + const loadQueueInfos: LoadQueueInfo[] = filterMap( + row.subRows, + r => r._original.loadQueueInfo, + ); + + if (!loadQueueInfos.length) { + return 'Could not get load queue infos'; + } + + const segmentsToLoad = sum(loadQueueInfos, s => Number(s.segmentsToLoad) || 0); + const segmentsToLoadSize = sum( + loadQueueInfos, + s => Number(s.segmentsToLoadSize) || 0, + ); + const segmentsToDrop = sum(loadQueueInfos, s => Number(s.segmentsToDrop) || 0); + const segmentsToDropSize = sum( + loadQueueInfos, + s => Number(s.segmentsToDropSize) || 0, + ); return formatQueues( segmentsToLoad, segmentsToLoadSize, @@ -580,13 +626,14 @@ ORDER BY show: capabilities.hasOverlordAccess() && visibleColumns.shown(ACTION_COLUMN_LABEL), id: ACTION_COLUMN_ID, width: ACTION_COLUMN_WIDTH, - accessor: row => row.worker, + accessor: row => row.workerInfo, filterable: false, Cell: ({ value, aggregated }) => { if (aggregated) return ''; if (!value) return null; - const disabled = value.version === ''; - const workerActions = this.getWorkerActions(value.host, disabled); + const { worker } = value; + const disabled = worker.version === ''; + const workerActions = this.getWorkerActions(worker.host, disabled); return ; }, Aggregated: () => '', From 100a2aa4a234f8c0dd84d34a9819484fb55b9bf6 Mon Sep 17 00:00:00 2001 From: Jill Osborne Date: Tue, 29 Nov 2022 02:31:28 +0000 Subject: [PATCH 052/141] Update and document experimental features (#13348) * Update and document experimental features * Updated * Update experimental-features.md * Update docs/development/experimental-features.md Co-authored-by: Abhishek Agarwal <1477457+abhishekagarwal87@users.noreply.github.com> * Updated after review * Updated * Update materialized-view.md * Update experimental-features.md Co-authored-by: Abhishek Agarwal <1477457+abhishekagarwal87@users.noreply.github.com> --- docs/configuration/index.md | 4 +- docs/development/experimental-features.md | 68 +++++++++++++++++++ .../extensions-core/druid-lookups.md | 3 - .../kafka-supervisor-reference.md | 2 - .../extensions-core/kinesis-ingestion.md | 2 - docs/querying/lookups.md | 4 +- 6 files changed, 72 insertions(+), 11 deletions(-) create mode 100644 docs/development/experimental-features.md diff --git a/docs/configuration/index.md b/docs/configuration/index.md index 6fb3201d98f4..8ac9382b7e21 100644 --- a/docs/configuration/index.md +++ b/docs/configuration/index.md @@ -1375,7 +1375,7 @@ For GCE's properties, please refer to the [gce-extensions](../development/extens This section contains the configuration options for the processes that reside on Data servers (MiddleManagers/Peons and Historicals) in the suggested [three-server configuration](../design/processes.md#server-types). -Configuration options for the experimental [Indexer process](../design/indexer.md) are also provided here. +Configuration options for the [Indexer process](../design/indexer.md) are also provided here. ### MiddleManager and Peons @@ -2188,7 +2188,7 @@ Supported query contexts: |Key|Description|Default| |---|-----------|-------| |`druid.expressions.useStrictBooleans`|Controls the behavior of Druid boolean operators and functions, if set to `true` all boolean values will be either a `1` or `0`. See [expression documentation](../misc/math-expr.md#logical-operator-modes)|false| -|`druid.expressions.allowNestedArrays`|If enabled, Druid array expressions can create nested arrays. This is experimental and should be used with caution.|false| +|`druid.expressions.allowNestedArrays`|If enabled, Druid array expressions can create nested arrays.|false| ### Router #### Router Process Configs diff --git a/docs/development/experimental-features.md b/docs/development/experimental-features.md new file mode 100644 index 000000000000..a68fc47411af --- /dev/null +++ b/docs/development/experimental-features.md @@ -0,0 +1,68 @@ +--- +id: experimental-features +title: "Experimental features" +--- + + + +The following features are marked [experimental](./experimental.md) in the Druid docs. + +This document includes each page that mentions an experimental feature. To graduate a feature, remove all mentions of its experimental status on all relevant pages. + +Note that this document does not track the status of contrib extensions, some of which are experimental. + +## SQL-based ingestion + +- [SQL-based ingestion](../multi-stage-query/index.md) +- [SQL-based ingestion concepts](../multi-stage-query/concepts.md) +- [SQL-based ingestion and multi-stage query task API](../multi-stage-query/api.md) + +## Nested columns + +- [Nested columns](../querying/nested-columns.md) + +## Indexer process + +- [Indexer process](../design/indexer.md) +- [Processes and servers](../design/processes.md#indexer-process-optional) + +## Kubernetes + +- [Kubernetes](../development/extensions-core/kubernetes.md) + +## Segment locking + +- [Configuration reference](../configuration/index.md#overlord-operations) +- [Task reference](../ingestion/tasks.md#locking) +- [Design](../design/architecture.md#availability-and-consistency) + +## Moments sketch + +- [Aggregations](../querying/aggregations.md#moments-sketch-experimental) + +## Front coding + +- [Ingestion spec reference](../ingestion/ingestion-spec.md#front-coding) + +## Other configuration properties + +- [Configuration reference](../configuration/index.md) + - `CLOSED_SEGMENTS_SINKS` mode + - Expression processing configuration `druid.expressions.allowNestedArrays` diff --git a/docs/development/extensions-core/druid-lookups.md b/docs/development/extensions-core/druid-lookups.md index b44f9620bd0a..5b19508c2375 100644 --- a/docs/development/extensions-core/druid-lookups.md +++ b/docs/development/extensions-core/druid-lookups.md @@ -22,9 +22,6 @@ title: "Cached Lookup Module" ~ under the License. --> - -> Please note that this is an experimental module and the development/testing still at early stage. Feel free to try it and give us your feedback. - ## Description This Apache Druid module provides a per-lookup caching mechanism for JDBC data sources. The main goal of this cache is to speed up the access to a high latency lookup sources and to provide a caching isolation for every lookup source. diff --git a/docs/development/extensions-core/kafka-supervisor-reference.md b/docs/development/extensions-core/kafka-supervisor-reference.md index 210207302f2d..9dc80d0c6585 100644 --- a/docs/development/extensions-core/kafka-supervisor-reference.md +++ b/docs/development/extensions-core/kafka-supervisor-reference.md @@ -56,8 +56,6 @@ This topic contains configuration reference information for the Apache Kafka sup ## Task Autoscaler Properties -> Note that Task AutoScaler is currently designated as experimental. - | Property | Description | Required | | ------------- | ------------- | ------------- | | `enableTaskAutoScaler` | Enable or disable autoscaling. `false` or blank disables the `autoScaler` even when `autoScalerConfig` is not null| no (default == false) | diff --git a/docs/development/extensions-core/kinesis-ingestion.md b/docs/development/extensions-core/kinesis-ingestion.md index 84036665ec9e..31682b54a291 100644 --- a/docs/development/extensions-core/kinesis-ingestion.md +++ b/docs/development/extensions-core/kinesis-ingestion.md @@ -149,8 +149,6 @@ Where the file `supervisor-spec.json` contains a Kinesis supervisor spec: #### Task Autoscaler Properties -> Note that Task AutoScaler is currently designated as experimental. - | Property | Description | Required | | ------------- | ------------- | ------------- | | `enableTaskAutoScaler` | Enable or disable the auto scaler. When false or absent, Druid disables the `autoScaler` even when `autoScalerConfig` is not null.| no (default == false) | diff --git a/docs/querying/lookups.md b/docs/querying/lookups.md index 57b406afb199..860a3ed2e277 100644 --- a/docs/querying/lookups.md +++ b/docs/querying/lookups.md @@ -115,8 +115,8 @@ will not detect this automatically. Dynamic Configuration --------------------- -> Dynamic lookup configuration is an [experimental](../development/experimental.md) feature. Static -> configuration is no longer supported. +> Static configuration is no longer supported. + The following documents the behavior of the cluster-wide config which is accessible through the Coordinator. The configuration is propagated through the concept of "tier" of servers. A "tier" is defined as a group of services which should receive a set of lookups. From edd076ca696e3ebcb45c4b0219440a50add9e13a Mon Sep 17 00:00:00 2001 From: Karan Kumar Date: Tue, 29 Nov 2022 08:46:59 +0530 Subject: [PATCH 053/141] Remove duplicate FrameRowTooLargeException.java (#13441) * Removing duplicate FrameRowTooLargeException.java * Fixing intellij inspection --- .../write/FrameRowTooLargeException.java | 43 ------------------- .../apache/druid/frame/write/FrameWriter.java | 4 +- .../frame/testutil/FrameSequenceBuilder.java | 2 +- .../write/FrameRowTooLargeExceptionTest.java | 34 --------------- 4 files changed, 3 insertions(+), 80 deletions(-) delete mode 100644 processing/src/main/java/org/apache/druid/frame/write/FrameRowTooLargeException.java delete mode 100644 processing/src/test/java/org/apache/druid/frame/write/FrameRowTooLargeExceptionTest.java diff --git a/processing/src/main/java/org/apache/druid/frame/write/FrameRowTooLargeException.java b/processing/src/main/java/org/apache/druid/frame/write/FrameRowTooLargeException.java deleted file mode 100644 index e08c8ccc2cab..000000000000 --- a/processing/src/main/java/org/apache/druid/frame/write/FrameRowTooLargeException.java +++ /dev/null @@ -1,43 +0,0 @@ -/* - * Licensed to the Apache Software Foundation (ASF) under one - * or more contributor license agreements. See the NOTICE file - * distributed with this work for additional information - * regarding copyright ownership. The ASF licenses this file - * to you under the Apache License, Version 2.0 (the - * "License"); you may not use this file except in compliance - * with the License. You may obtain a copy of the License at - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, - * software distributed under the License is distributed on an - * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY - * KIND, either express or implied. See the License for the - * specific language governing permissions and limitations - * under the License. - */ - -package org.apache.druid.frame.write; - -import org.apache.druid.java.util.common.StringUtils; - -/** - * Exception that is conventionally thrown by workers when they call - * {@link FrameWriter#addSelection} and it returns false on an empty frame, or in - * a situation where allocating a new frame is impractical. - */ -public class FrameRowTooLargeException extends RuntimeException -{ - private final long maxFrameSize; - - public FrameRowTooLargeException(final long maxFrameSize) - { - super(StringUtils.format("Row too large to add to frame (max frame size = %,d)", maxFrameSize)); - this.maxFrameSize = maxFrameSize; - } - - public long getMaxFrameSize() - { - return maxFrameSize; - } -} diff --git a/processing/src/main/java/org/apache/druid/frame/write/FrameWriter.java b/processing/src/main/java/org/apache/druid/frame/write/FrameWriter.java index 5de131946d5e..f37ed8571ac1 100644 --- a/processing/src/main/java/org/apache/druid/frame/write/FrameWriter.java +++ b/processing/src/main/java/org/apache/druid/frame/write/FrameWriter.java @@ -34,9 +34,9 @@ public interface FrameWriter extends Closeable { /** * Write the current row to the frame that is under construction, if there is enough space to do so. - * + *

* If this method returns false on an empty frame, or in a situation where starting a new frame is impractical, - * it is conventional (although not required) for the caller to throw {@link FrameRowTooLargeException}. + * it is conventional (although not required) for the caller to throw {@link org.apache.druid.frame.processor.FrameRowTooLargeException}. * * @return true if the row was written, false if there was not enough space */ diff --git a/processing/src/test/java/org/apache/druid/frame/testutil/FrameSequenceBuilder.java b/processing/src/test/java/org/apache/druid/frame/testutil/FrameSequenceBuilder.java index 3fbd8ba3180d..de7de8dd0da4 100644 --- a/processing/src/test/java/org/apache/druid/frame/testutil/FrameSequenceBuilder.java +++ b/processing/src/test/java/org/apache/druid/frame/testutil/FrameSequenceBuilder.java @@ -24,7 +24,7 @@ import org.apache.druid.frame.allocation.HeapMemoryAllocator; import org.apache.druid.frame.allocation.MemoryAllocator; import org.apache.druid.frame.key.SortColumn; -import org.apache.druid.frame.write.FrameRowTooLargeException; +import org.apache.druid.frame.processor.FrameRowTooLargeException; import org.apache.druid.frame.write.FrameWriter; import org.apache.druid.frame.write.FrameWriterFactory; import org.apache.druid.frame.write.FrameWriters; diff --git a/processing/src/test/java/org/apache/druid/frame/write/FrameRowTooLargeExceptionTest.java b/processing/src/test/java/org/apache/druid/frame/write/FrameRowTooLargeExceptionTest.java deleted file mode 100644 index 7c33ee098962..000000000000 --- a/processing/src/test/java/org/apache/druid/frame/write/FrameRowTooLargeExceptionTest.java +++ /dev/null @@ -1,34 +0,0 @@ -/* - * Licensed to the Apache Software Foundation (ASF) under one - * or more contributor license agreements. See the NOTICE file - * distributed with this work for additional information - * regarding copyright ownership. The ASF licenses this file - * to you under the Apache License, Version 2.0 (the - * "License"); you may not use this file except in compliance - * with the License. You may obtain a copy of the License at - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, - * software distributed under the License is distributed on an - * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY - * KIND, either express or implied. See the License for the - * specific language governing permissions and limitations - * under the License. - */ - -package org.apache.druid.frame.write; - -import org.junit.Assert; -import org.junit.Test; - -public class FrameRowTooLargeExceptionTest -{ - @Test - public void testBasic() - { - final int maxFrameSize = 1000; - final FrameRowTooLargeException e = new FrameRowTooLargeException(maxFrameSize); - Assert.assertEquals(maxFrameSize, e.getMaxFrameSize()); - } -} From af0d4458664ebcb0bbefef57953e3414d4dd6179 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Tue, 29 Nov 2022 15:37:39 +0530 Subject: [PATCH 054/141] caffeine cache uses 5% of heap --- .../main/java/org/apache/druid/client/cache/CaffeineCache.java | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/server/src/main/java/org/apache/druid/client/cache/CaffeineCache.java b/server/src/main/java/org/apache/druid/client/cache/CaffeineCache.java index b7a1ea826c58..362a387231a6 100644 --- a/server/src/main/java/org/apache/druid/client/cache/CaffeineCache.java +++ b/server/src/main/java/org/apache/druid/client/cache/CaffeineCache.java @@ -70,7 +70,7 @@ public static CaffeineCache create(final CaffeineCacheConfig config, final Execu if (config.getSizeInBytes() >= 0) { builder.maximumWeight(config.getSizeInBytes()); } else { - builder.maximumWeight(Math.min(MAX_DEFAULT_BYTES, JvmUtils.getRuntimeInfo().getMaxHeapSizeBytes() / 10)); + builder.maximumWeight(Math.min(MAX_DEFAULT_BYTES, JvmUtils.getRuntimeInfo().getMaxHeapSizeBytes() / 20)); } builder .weigher((NamedKey key, byte[] value) -> value.length From 6397eefcd235009ebbdb9d9a32f163d2f4fc1601 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Tue, 29 Nov 2022 15:38:02 +0530 Subject: [PATCH 055/141] ensure min task count is 2, task count is monotonic --- examples/bin/start-druid | 13 ++++++++----- 1 file changed, 8 insertions(+), 5 deletions(-) diff --git a/examples/bin/start-druid b/examples/bin/start-druid index 1bdd0c58a249..25c282f461c5 100644 --- a/examples/bin/start-druid +++ b/examples/bin/start-druid @@ -97,7 +97,7 @@ def print_if_verbose(message): def configure_parser(): parser = argparse.ArgumentParser( - prog='Druid quickstart', + prog='start-druid', formatter_class=argparse.RawTextHelpFormatter, epilog= """ @@ -300,9 +300,12 @@ def should_compute_memory(config, total_memory, service_list): # 5. lastly, if middleManager is to be started, and it is missing task memory properties if jvm_config_count > 0 or mm_task_property_present: if total_memory != "": - raise ValueError("If jvm.config is given for services, memory argument shouldn't be specified") + raise ValueError( + "If jvm.config for services and/or middleManager configs " + "(druid.worker.capacity, druid.indexer.runner.javaOptsArray) is present, " + "memory argument shouldn't be specified") if jvm_config_count == 0: - raise ValueError("druid.indexer.runner.javaOptsArray or druid.worker.capacity is present in " + raise ValueError("middleManger configs (druid.indexer.runner.javaOptsArray or druid.worker.capacity) is present in " "middleManager/runtime.properties, \n " "add jvm.config for all other services") if jvm_config_count != len(service_list): @@ -374,12 +377,12 @@ def build_mm_task_java_opts_array(memory_type): def compute_tasks_memory(allocated_memory): - if allocated_memory >= 2048: + if allocated_memory >= 4096: task_count = int(allocated_memory / 2048) memory_type = TASK_MEM_TYPE_HIGH task_memory_mb = 2048 else: - task_count = int(allocated_memory / 512) + task_count = 2 memory_type = TASK_MEM_TYPE_LOW task_memory_mb = 512 task_count = min(task_count, multiprocessing.cpu_count()) From a3d442bbaf3de1d71df938546853281ae1ed3ec8 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Tue, 29 Nov 2022 15:57:49 +0530 Subject: [PATCH 056/141] update configs and documentation for runtime props in conf/druid/single-server/quickstart --- .../quickstart/broker/runtime.properties | 14 +++++++++----- .../quickstart/historical/runtime.properties | 12 ++++++------ .../quickstart/middleManager/runtime.properties | 9 +++------ .../quickstart/router/runtime.properties | 6 ------ 4 files changed, 18 insertions(+), 23 deletions(-) diff --git a/examples/conf/druid/single-server/quickstart/broker/runtime.properties b/examples/conf/druid/single-server/quickstart/broker/runtime.properties index 549117efa50d..f4c494019933 100644 --- a/examples/conf/druid/single-server/quickstart/broker/runtime.properties +++ b/examples/conf/druid/single-server/quickstart/broker/runtime.properties @@ -21,15 +21,19 @@ druid.service=druid/broker druid.plaintextPort=8082 # HTTP server settings -#druid.server.http.numThreads=12 +# HTTP server thread pool size. Higher values increase peak load on the Broker, but +# may be useful for high-concurrency workloads. +# Default is max(10, (Number of processors * 17) / 16 + 2) + 30. +# druid.server.http.numThreads=N # HTTP client settings -#druid.broker.http.numConnections=10 -#druid.broker.http.maxQueuedBytes=5MiB +# Connection pool size from the Broker to each data server. May be useful to +# raise this for high-concurrency workloads. +# druid.broker.http.numConnections=20 # Processing threads and buffers -#druid.processing.buffer.sizeBytes=100MiB -#druid.processing.numMergeBuffers=2 +# Determined automatically based on available memory. For details on how to manually set parameters: +# https://druid.apache.org/docs/latest/operations/basic-cluster-tuning.html#guidelines-for-processing-threads-and-buffers druid.processing.tmpDir=var/druid/processing # Query cache disabled -- push down caching and merging instead diff --git a/examples/conf/druid/single-server/quickstart/historical/runtime.properties b/examples/conf/druid/single-server/quickstart/historical/runtime.properties index cf3e966cc30e..d9ce1850e4fc 100644 --- a/examples/conf/druid/single-server/quickstart/historical/runtime.properties +++ b/examples/conf/druid/single-server/quickstart/historical/runtime.properties @@ -20,13 +20,14 @@ druid.service=druid/historical druid.plaintextPort=8083 -# HTTP server threads -#druid.server.http.numThreads=12 +# HTTP server thread pool size. Higher values increase peak load on the Broker, but +# may be useful for high-concurrency workloads. +# Default is max(10, (Number of processors * 17) / 16 + 2) + 30. +# druid.server.http.numThreads=N # Processing threads and buffers -#druid.processing.buffer.sizeBytes=200MiB -#druid.processing.numMergeBuffers=2 -#druid.processing.numThreads=2 +# Determined automatically based on available memory. For details on how to manually set parameters: +# https://druid.apache.org/docs/latest/operations/basic-cluster-tuning.html#guidelines-for-processing-threads-and-buffers druid.processing.tmpDir=var/druid/processing # Segment storage @@ -36,4 +37,3 @@ druid.segmentCache.locations=[{"path":"var/druid/segment-cache","maxSize":"300g" druid.historical.cache.useCache=true druid.historical.cache.populateCache=true druid.cache.type=caffeine -#druid.cache.sizeInBytes=10MiB diff --git a/examples/conf/druid/single-server/quickstart/middleManager/runtime.properties b/examples/conf/druid/single-server/quickstart/middleManager/runtime.properties index df9b790a69b0..a90e62579676 100644 --- a/examples/conf/druid/single-server/quickstart/middleManager/runtime.properties +++ b/examples/conf/druid/single-server/quickstart/middleManager/runtime.properties @@ -20,17 +20,14 @@ druid.service=druid/middleManager druid.plaintextPort=8091 -# Number of tasks per middleManager -# druid.worker.capacity +# Number of tasks (druid.worker.capacity) and memory usage per task (druid.indexer.runner.javaOptsArray) is automatically +# determined based on available memory. For details on how to manually set parameters, see: +# https://druid.apache.org/docs/latest/operations/basic-cluster-tuning.html#middlemanager # Task launch parameters druid.indexer.runner.javaCommand=bin/run-java -#druid.indexer.runner.javaOptsArray=["-server","-Xms1g","-Xmx1g","-XX:MaxDirectMemorySize=1g","-Duser.timezone=UTC","-Dfile.encoding=UTF-8","-XX:+ExitOnOutOfMemoryError","-Djava.util.logging.manager=org.apache.logging.log4j.jul.LogManager"] druid.indexer.task.baseTaskDir=var/druid/task -# HTTP server threads -#druid.server.http.numThreads=12 - # Processing threads and buffers on Peons #druid.indexer.fork.property.druid.processing.numMergeBuffers=2 #druid.indexer.fork.property.druid.processing.buffer.sizeBytes=100MiB diff --git a/examples/conf/druid/single-server/quickstart/router/runtime.properties b/examples/conf/druid/single-server/quickstart/router/runtime.properties index 4c38ba88baab..3858dec044bd 100644 --- a/examples/conf/druid/single-server/quickstart/router/runtime.properties +++ b/examples/conf/druid/single-server/quickstart/router/runtime.properties @@ -20,12 +20,6 @@ druid.service=druid/router druid.plaintextPort=8888 -# HTTP proxy -#druid.router.http.numConnections=50 -#druid.router.http.readTimeout=PT5M -#druid.router.http.numMaxThreads=100 -#druid.server.http.numThreads=100 - # Service discovery druid.router.defaultBrokerServiceName=druid/broker druid.router.coordinatorServiceName=druid/coordinator From b332729455dcf2c4b6e636ed4d40ef321240c29e Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Tue, 29 Nov 2022 17:06:02 +0530 Subject: [PATCH 057/141] Update docs --- docs/operations/single-server.md | 36 +++++++++++++++---------- docs/tutorials/index.md | 34 ++++++++++++----------- docs/tutorials/tutorial-batch-hadoop.md | 2 +- docs/tutorials/tutorial-kafka.md | 2 +- 4 files changed, 42 insertions(+), 32 deletions(-) diff --git a/docs/operations/single-server.md b/docs/operations/single-server.md index 35413952e222..7ce252379e9c 100644 --- a/docs/operations/single-server.md +++ b/docs/operations/single-server.md @@ -31,6 +31,7 @@ Druid includes a set of reference configurations and launch scripts for single-m - `medium` - `large` - `xlarge` +- `start-druid` The `micro-quickstart` is sized for small machines like laptops and is intended for quick evaluation use-cases. @@ -42,37 +43,44 @@ The startup scripts for these example configurations run a single ZK instance al The example configurations run the Druid Coordinator and Overlord together in a single process using the optional configuration `druid.coordinator.asOverlord.enabled=true`, described in the [Coordinator configuration documentation](../configuration/index.md#coordinator-operation). +The `start-druid` is a generic launch script for starting druid services on single server, it accepts optional arguments like services, memory and config. +All reference configurations can be acheived by passing appropriate arguments to this script. +If memory argument isn't specified, services will use upto 80% system memory. +Existing launch scripts are deprecated and will be removed in the next release. + + While example configurations are provided for very large single machines, at higher scales we recommend running Druid in a [clustered deployment](../tutorials/cluster.md), for fault-tolerance and reduced resource contention. ## Single server reference configurations -### Nano-Quickstart: 1 CPU, 4GiB RAM +### Nano: 1 CPU, 4GiB RAM -- Launch command: `bin/start-nano-quickstart` -- Configuration directory: `conf/druid/single-server/nano-quickstart` +- Launch command: `bin/start-druid` +- Configuration directory: `conf/druid/single-server/quickstart` -### Micro-Quickstart: 4 CPU, 16GiB RAM +### Micro: 4 CPU, 16GiB RAM -- Launch command: `bin/start-micro-quickstart` -- Configuration directory: `conf/druid/single-server/micro-quickstart` +- Launch command: `bin/start-druid` +- Configuration directory: `conf/druid/single-server/quickstart` ### Small: 8 CPU, 64GiB RAM (~i3.2xlarge) -- Launch command: `bin/start-small` -- Configuration directory: `conf/druid/single-server/small` +- Launch command: `bin/start-druid` +- Configuration directory: `conf/druid/single-server/quickstart` ### Medium: 16 CPU, 128GiB RAM (~i3.4xlarge) -- Launch command: `bin/start-medium` -- Configuration directory: `conf/druid/single-server/medium` +- Launch command: `bin/start-druid` +- Configuration directory: `conf/druid/single-server/quickstart` ### Large: 32 CPU, 256GiB RAM (~i3.8xlarge) -- Launch command: `bin/start-large` -- Configuration directory: `conf/druid/single-server/large` +- Launch command: `bin/start-druid` +- Configuration directory: `conf/druid/single-server/quickstart` ### X-Large: 64 CPU, 512GiB RAM (~i3.16xlarge) -- Launch command: `bin/start-xlarge` -- Configuration directory: `conf/druid/single-server/xlarge` +- Launch command: `bin/start-druid` +- Configuration directory: `conf/druid/single-server/quickstart` +The amount of memory used for Druid can be limited by passing memory argument in the launch command, `bin/start-druid --memory=value` \ No newline at end of file diff --git a/docs/tutorials/index.md b/docs/tutorials/index.md index 7d3d2ab68ec1..8d029d568189 100644 --- a/docs/tutorials/index.md +++ b/docs/tutorials/index.md @@ -23,7 +23,7 @@ title: "Quickstart (local)" --> -This quickstart gets you started with Apache Druid using the [`micro-quickstart`](../operations/single-server.md#micro-quickstart-4-cpu-16gib-ram) configuration, and introduces you to Druid ingestion and query features. +This quickstart gets you started with Apache Druid using the [`micro`](../operations/single-server.md#micro-4-cpu-16gib-ram) configuration, and introduces you to Druid ingestion and query features. In this quickstart, you'll do the following: - install Druid @@ -37,7 +37,8 @@ Druid supports a variety of ingestion options. Once you're done with this tutori You can follow these steps on a relatively modest machine, such as a workstation or virtual server with 16 GiB of RAM. -Druid comes equipped with several [startup configuration profiles](../operations/single-server.md) for a +Druid comes equipped with a single launch script that can be used to run several +[startup configuration profiles](../operations/single-server.md) for a range of machine sizes. These range from `nano` (1 CPU, 4GiB RAM) to `x-large` (64 CPU, 512GiB RAM). For more information, see [Single server deployment](../operations/single-server.md). For information on deploying Druid services across clustered machines, see [Clustered deployment](./cluster.md). @@ -72,30 +73,31 @@ The distribution directory contains `LICENSE` and `NOTICE` files and subdirector ## Start up Druid services -Start up Druid services using the `micro-quickstart` single-machine configuration. +Start up Druid services using the `micro` single-machine configuration. This configuration includes default settings that are appropriate for this tutorial, such as loading the `druid-multi-stage-query` extension by default so that you can use the MSQ task engine. -You can view that setting and others in the configuration files in the `conf/druid/single-server/micro-quickstart/`. +You can view that setting and others in the configuration files in the `conf/druid/single-server/quickstart/`. From the apache-druid-{{DRUIDVERSION}} package root, run the following command: ```bash -./bin/start-micro-quickstart +./bin/start-druid --memory=16g ``` This brings up instances of ZooKeeper and the Druid services: ```bash $ ./bin/start-micro-quickstart -[Thu Sep 8 18:30:00 2022] Starting Apache Druid. -[Thu Sep 8 18:30:00 2022] Open http://localhost:8888/ in your browser to access the web console. -[Thu Sep 8 18:30:00 2022] Or, if you have enabled TLS, use https on port 9088. -[Thu Sep 8 18:30:00 2022] Running command[zk], logging to[/apache-druid-{{DRUIDVERSION}}/var/sv/zk.log]: bin/run-zk conf -[Thu Sep 8 18:30:00 2022] Running command[coordinator-overlord], logging to[/apache-druid-{{DRUIDVERSION}}/var/sv/coordinator-overlord.log]: bin/run-druid coordinator-overlord conf/druid/single-server/micro-quickstart -[Thu Sep 8 18:30:00 2022] Running command[broker], logging to[/apache-druid-{{DRUIDVERSION}}/var/sv/broker.log]: bin/run-druid broker conf/druid/single-server/micro-quickstart -[Thu Sep 8 18:30:00 2022] Running command[router], logging to[/apache-druid-{{DRUIDVERSION}}/var/sv/router.log]: bin/run-druid router conf/druid/single-server/micro-quickstart -[Thu Sep 8 18:30:00 2022] Running command[historical], logging to[/apache-druid-{{DRUIDVERSION}}/var/sv/historical.log]: bin/run-druid historical conf/druid/single-server/micro-quickstart -[Thu Sep 8 18:30:00 2022] Running command[middleManager], logging to[/apache-druid-{{DRUIDVERSION}}/var/sv/middleManager.log]: bin/run-druid middleManager conf/druid/single-server/micro-quickstart +[Tue Nov 29 16:31:06 2022] Starting Apache Druid. +[Tue Nov 29 16:31:06 2022] Open http://localhost:8888/ in your browser to access the web console. +[Tue Nov 29 16:31:06 2022] Or, if you have enabled TLS, use https on port 9088. +[Tue Nov 29 16:31:06 2022] Starting services with log directory [/apache-druid-{{DRUIDVERSION}}/log]. +[Tue Nov 29 16:31:06 2022] Running command[zk]: bin/run-zk conf +[Tue Nov 29 16:31:06 2022] Running command[broker]: bin/run-druid broker /apache-druid-{{DRUIDVERSION}}/conf/druid/single-server/quickstart '-Xms1187m -Xmx1187m -XX:MaxDirectMemorySize=791m' +[Tue Nov 29 16:31:06 2022] Running command[router]: bin/run-druid router /apache-druid-{{DRUIDVERSION}}/conf/druid/single-server/quickstart '-Xms128m -Xmx128m' +[Tue Nov 29 16:31:06 2022] Running command[coordinator-overlord]: bin/run-druid coordinator-overlord /apache-druid-{{DRUIDVERSION}}/conf/druid/single-server/quickstart '-Xms1290m -Xmx1290m' +[Tue Nov 29 16:31:06 2022] Running command[historical]: bin/run-druid historical /apache-druid-{{DRUIDVERSION}}/conf/druid/single-server/quickstart '-Xms1376m -Xmx1376m -XX:MaxDirectMemorySize=2064m' +[Tue Nov 29 16:31:06 2022] Running command[middleManager]: bin/run-druid middleManager /apache-druid-{{DRUIDVERSION}}/conf/druid/single-server/quickstart '-Xms64m -Xmx64m' '-Ddruid.worker.capacity=2 -Ddruid.indexer.runner.javaOptsArray=["-server","-Duser.timezone=UTC","-Dfile.encoding=UTF-8","-XX:+ExitOnOutOfMemoryError","-Djava.util.logging.manager=org.apache.logging.log4j.jul.LogManager","-Xms256m","-Xmx256m","-XX:MaxDirectMemorySize=256m"]' ``` All persistent state, such as the cluster metadata store and segments for the services, are kept in the `var` directory under @@ -103,7 +105,7 @@ the Druid root directory, apache-druid-{{DRUIDVERSION}}. Each service writes to At any time, you can revert Druid to its original, post-installation state by deleting the entire `var` directory. You may want to do this, for example, between Druid tutorials or after experimentation, to start with a fresh instance. -To stop Druid at any time, use CTRL+C in the terminal. This exits the `bin/start-micro-quickstart` script and terminates all Druid processes. +To stop Druid at any time, use CTRL+C in the terminal. This exits the `bin/start-druid` script and terminates all Druid processes. ## Open the web console @@ -222,4 +224,4 @@ See the following topics for more information: * [Tutorial: Load stream data from Apache Kafka](./tutorial-kafka.md) to load streaming data from a Kafka topic. * [Extensions](../development/extensions.md) for details on Druid extensions. -Remember that after stopping Druid services, you can start clean next time by deleting the `var` directory from the Druid root directory and running the `bin/start-micro-quickstart` script again. You may want to do this before using other data ingestion tutorials, since they use the same Wikipedia datasource. +Remember that after stopping Druid services, you can start clean next time by deleting the `var` directory from the Druid root directory and running the `bin/start-druid` script again. You may want to do this before using other data ingestion tutorials, since they use the same Wikipedia datasource. diff --git a/docs/tutorials/tutorial-batch-hadoop.md b/docs/tutorials/tutorial-batch-hadoop.md index 47cd2d6bcbe5..bff033a28c41 100644 --- a/docs/tutorials/tutorial-batch-hadoop.md +++ b/docs/tutorials/tutorial-batch-hadoop.md @@ -28,7 +28,7 @@ This tutorial shows you how to load data files into Apache Druid using a remote For this tutorial, we'll assume that you've already completed the previous [batch ingestion tutorial](tutorial-batch.md) using Druid's native batch ingestion system and are using the -`micro-quickstart` single-machine configuration as described in the [quickstart](index.md). +`micro` single-machine configuration as described in the [quickstart](../operations/single-server.md#micro-4-cpu-16gib-ram). ## Install Docker diff --git a/docs/tutorials/tutorial-kafka.md b/docs/tutorials/tutorial-kafka.md index eb06f4239f95..bbdf6af72ea0 100644 --- a/docs/tutorials/tutorial-kafka.md +++ b/docs/tutorials/tutorial-kafka.md @@ -30,7 +30,7 @@ The tutorial guides you through the steps to load sample nested clickstream data ## Prerequisites -Before you follow the steps in this tutorial, download Druid as described in the [quickstart](index.md) using the [micro-quickstart](../operations/single-server.md#micro-quickstart-4-cpu-16gib-ram) single-machine configuration and have it running on your local machine. You don't need to have loaded any data. +Before you follow the steps in this tutorial, download Druid as described in the [quickstart](index.md) using the [micro](../operations/single-server.md#micro-4-cpu-16gib-ram) single-machine configuration and have it running on your local machine. You don't need to have loaded any data. ## Download and start Kafka From b8d34c423c0e56af21e3197d01c983e65697a8ae Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Tue, 29 Nov 2022 17:11:47 +0530 Subject: [PATCH 058/141] Specify memory argument for each profile in single-server.md --- docs/operations/single-server.md | 17 ++++++++--------- 1 file changed, 8 insertions(+), 9 deletions(-) diff --git a/docs/operations/single-server.md b/docs/operations/single-server.md index 7ce252379e9c..34dd2dc0e006 100644 --- a/docs/operations/single-server.md +++ b/docs/operations/single-server.md @@ -44,8 +44,7 @@ The startup scripts for these example configurations run a single ZK instance al The example configurations run the Druid Coordinator and Overlord together in a single process using the optional configuration `druid.coordinator.asOverlord.enabled=true`, described in the [Coordinator configuration documentation](../configuration/index.md#coordinator-operation). The `start-druid` is a generic launch script for starting druid services on single server, it accepts optional arguments like services, memory and config. -All reference configurations can be acheived by passing appropriate arguments to this script. -If memory argument isn't specified, services will use upto 80% system memory. +All reference configurations can be acheived by passing appropriate arguments to this script. Existing launch scripts are deprecated and will be removed in the next release. @@ -55,32 +54,32 @@ While example configurations are provided for very large single machines, at hig ### Nano: 1 CPU, 4GiB RAM -- Launch command: `bin/start-druid` +- Launch command: `bin/start-druid --memory=4g` - Configuration directory: `conf/druid/single-server/quickstart` ### Micro: 4 CPU, 16GiB RAM -- Launch command: `bin/start-druid` +- Launch command: `bin/start-druid --memory=16g` - Configuration directory: `conf/druid/single-server/quickstart` ### Small: 8 CPU, 64GiB RAM (~i3.2xlarge) -- Launch command: `bin/start-druid` +- Launch command: `bin/start-druid --memory=64g` - Configuration directory: `conf/druid/single-server/quickstart` ### Medium: 16 CPU, 128GiB RAM (~i3.4xlarge) -- Launch command: `bin/start-druid` +- Launch command: `bin/start-druid --memory=128g` - Configuration directory: `conf/druid/single-server/quickstart` ### Large: 32 CPU, 256GiB RAM (~i3.8xlarge) -- Launch command: `bin/start-druid` +- Launch command: `bin/start-druid --memory=256g` - Configuration directory: `conf/druid/single-server/quickstart` ### X-Large: 64 CPU, 512GiB RAM (~i3.16xlarge) -- Launch command: `bin/start-druid` +- Launch command: `bin/start-druid --memory=512g` - Configuration directory: `conf/druid/single-server/quickstart` -The amount of memory used for Druid can be limited by passing memory argument in the launch command, `bin/start-druid --memory=value` \ No newline at end of file +Memory argument (`--memory`) in the above launch command is optional, if not specified Druid will use upto 80% of system memory. \ No newline at end of file From 4ed6255bdf213b15577d645a36e0e341a16e8d9d Mon Sep 17 00:00:00 2001 From: Laksh Singla Date: Tue, 29 Nov 2022 17:19:57 +0530 Subject: [PATCH 059/141] Convert errors based on implicit type conversion in multi value arrays to parse exception in MSQ (#13366) * initial commit * fix test * push the json changes * reduce the area of the try..catch * Trigger Build * review --- .../apache/druid/msq/exec/MSQSelectTest.java | 58 +++++++++++++++++++ .../unparseable-mv-string-array.json | 3 + .../frame/write/RowBasedFrameWriter.java | 10 +++- .../write/columnar/ColumnarFrameWriter.java | 12 +++- 4 files changed, 78 insertions(+), 5 deletions(-) create mode 100644 extensions-core/multi-stage-query/src/test/resources/unparseable-mv-string-array.json diff --git a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/MSQSelectTest.java b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/MSQSelectTest.java index c83481757d8f..26d566b31e8c 100644 --- a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/MSQSelectTest.java +++ b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/MSQSelectTest.java @@ -31,6 +31,7 @@ import org.apache.druid.msq.indexing.ColumnMappings; import org.apache.druid.msq.indexing.MSQSpec; import org.apache.druid.msq.indexing.MSQTuningConfig; +import org.apache.druid.msq.indexing.error.CannotParseExternalDataFault; import org.apache.druid.msq.shuffle.DurableStorageUtils; import org.apache.druid.msq.test.MSQTestBase; import org.apache.druid.query.InlineDataSource; @@ -1253,6 +1254,63 @@ public void testGroupByOnFooWithDurableStoragePathAssertions() throws IOExceptio .write(ArgumentMatchers.endsWith("__success")); } + @Test + public void testMultiValueStringWithIncorrectType() throws IOException + { + final File toRead = getResourceAsTemporaryFile("/unparseable-mv-string-array.json"); + final String toReadAsJson = queryFramework().queryJsonMapper().writeValueAsString(toRead.getAbsolutePath()); + + RowSignature rowSignature = RowSignature.builder() + .add("__time", ColumnType.LONG) + .add("language", ColumnType.STRING_ARRAY) + .build(); + + final GroupByQuery expectedQuery = + GroupByQuery.builder() + .setDataSource(CalciteTests.DATASOURCE1) + .setInterval(querySegmentSpec(Filtration.eternity())) + .setGranularity(Granularities.ALL) + .setDimensions(dimensions(new DefaultDimensionSpec("__time", "d0", ColumnType.LONG))) + .build(); + + + testSelectQuery() + .setSql("WITH\n" + + "kttm_data AS (\n" + + "SELECT * FROM TABLE(\n" + + " EXTERN(\n" + + " '{ \"files\": [" + toReadAsJson + "],\"type\":\"local\"}',\n" + + " '{\"type\":\"json\"}',\n" + + " '[{\"name\":\"timestamp\",\"type\":\"string\"},{\"name\":\"agent_category\",\"type\":\"string\"},{\"name\":\"agent_type\",\"type\":\"string\"},{\"name\":\"browser\",\"type\":\"string\"},{\"name\":\"browser_version\",\"type\":\"string\"},{\"name\":\"city\",\"type\":\"string\"},{\"name\":\"continent\",\"type\":\"string\"},{\"name\":\"country\",\"type\":\"string\"},{\"name\":\"version\",\"type\":\"string\"},{\"name\":\"event_type\",\"type\":\"string\"},{\"name\":\"event_subtype\",\"type\":\"string\"},{\"name\":\"loaded_image\",\"type\":\"string\"},{\"name\":\"adblock_list\",\"type\":\"string\"},{\"name\":\"forwarded_for\",\"type\":\"string\"},{\"name\":\"language\",\"type\":\"string\"},{\"name\":\"number\",\"type\":\"long\"},{\"name\":\"os\",\"type\":\"string\"},{\"name\":\"path\",\"type\":\"string\"},{\"name\":\"platform\",\"type\":\"string\"},{\"name\":\"referrer\",\"type\":\"string\"},{\"name\":\"referrer_host\",\"type\":\"string\"},{\"name\":\"region\",\"type\":\"string\"},{\"name\":\"remote_address\",\"type\":\"string\"},{\"name\":\"screen\",\"type\":\"string\"},{\"name\":\"session\",\"type\":\"string\"},{\"name\":\"session_length\",\"type\":\"long\"},{\"name\":\"timezone\",\"type\":\"string\"},{\"name\":\"timezone_offset\",\"type\":\"long\"},{\"name\":\"window\",\"type\":\"string\"}]'\n" + + " )\n" + + "))\n" + + "\n" + + "SELECT\n" + + " FLOOR(TIME_PARSE(\"timestamp\") TO MINUTE) AS __time,\n" + + " MV_TO_ARRAY(\"language\") AS \"language\"\n" + + "FROM kttm_data") + .setExpectedRowSignature(rowSignature) + .setExpectedResultRows(ImmutableList.of( + new Object[]{1566691200000L, ImmutableList.of("en")}, + new Object[]{1566691200000L, ImmutableList.of("en", "es", "es-419", "es-MX")}, + new Object[]{1566691200000L, ImmutableList.of("en", "es", "es-419", "es-US")} + )) + .setExpectedMSQSpec( + MSQSpec + .builder() + .query(expectedQuery) + .columnMappings(new ColumnMappings( + ImmutableList.of( + new ColumnMapping("d0", "__time"), + new ColumnMapping("a0", "cnt") + ) + )) + .tuningConfig(MSQTuningConfig.defaultConfig()) + .build()) + .setExpectedMSQFault(new CannotParseExternalDataFault( + "Unable to add the row to the frame. Type conversion might be required.")) + .verifyResults(); + } @Nonnull private List expectedMultiValueFooRowsGroup() diff --git a/extensions-core/multi-stage-query/src/test/resources/unparseable-mv-string-array.json b/extensions-core/multi-stage-query/src/test/resources/unparseable-mv-string-array.json new file mode 100644 index 000000000000..57b9a7709b09 --- /dev/null +++ b/extensions-core/multi-stage-query/src/test/resources/unparseable-mv-string-array.json @@ -0,0 +1,3 @@ +{"timestamp":"2019-08-25T00:00:00.031Z","agent_category":"Personal computer","agent_type":"Browser","browser":"Chrome","browser_version":"76.0.3809.100","city":"Rosario","continent":"South America","country":"Argentina","version":"1.9.6","event_type":"PercentClear","event_subtype":"55","loaded_image":"http://www.koalastothemax.com/img/koalas2.jpg","adblock_list":"NoAdblock","forwarded_for":"181.13.41.82","language":[{},{}],"number":"16","os":"Windows 7","path":"http://www.koalastothemax.com/","platform":"Windows","referrer":"Direct","referrer_host":"Direct","region":"Santa Fe","remote_address":"172.31.57.89","screen":"1680x1050","session":"S56194838","session_length":76261,"timezone":"N/A","timezone_offset":"180","window":"1680x939"} +{"timestamp":"2019-08-25T00:00:00.059Z","agent_category":"Smartphone","agent_type":"Mobile Browser","browser":"Chrome Mobile","browser_version":"50.0.2661.89","city":"Nuevo Casas Grandes","continent":"North America","country":"Mexico","version":"1.9.6","event_type":"PercentClear","event_subtype":"85","loaded_image":"https://koalastothemax.com/img/koalas1.jpg","adblock_list":"NoAdblock","forwarded_for":"177.242.100.0","language":["en","es","es-419","es-MX"],"number":"24","os":"Android","path":"https://koalastothemax.com/","platform":"Android","referrer":"https://www.google.com/","referrer_host":"www.google.com","region":"Chihuahua","remote_address":"172.31.11.5","screen":"320x570","session":"S46093731","session_length":252689,"timezone":"CDT","timezone_offset":"300","window":"540x743"} +{"timestamp":"2019-08-25T00:00:00.178Z","agent_category":"Personal computer","agent_type":"Browser","browser":"Chrome","browser_version":"76.0.3809.100","city":"Luis Guillon","continent":"South America","country":"Argentina","version":"1.9.6","event_type":"PercentClear","event_subtype":"90","loaded_image":"http://www.koalastothemax.com/img/koalas.jpg","adblock_list":"NoAdblock","forwarded_for":"181.46.136.44","language":["en","es","es-419","es-US"],"number":"24","os":"Windows 7","path":"http://www.koalastothemax.com/","platform":"Windows","referrer":"Direct","referrer_host":"Direct","region":"Buenos Aires","remote_address":"172.31.11.5","screen":"1366x768","session":"S13352079","session_length":1753602,"timezone":"N/A","timezone_offset":"180","window":"1366x652"} diff --git a/processing/src/main/java/org/apache/druid/frame/write/RowBasedFrameWriter.java b/processing/src/main/java/org/apache/druid/frame/write/RowBasedFrameWriter.java index 2bf4b343080c..2a8af4b9e8f8 100644 --- a/processing/src/main/java/org/apache/druid/frame/write/RowBasedFrameWriter.java +++ b/processing/src/main/java/org/apache/druid/frame/write/RowBasedFrameWriter.java @@ -32,6 +32,7 @@ import org.apache.druid.java.util.common.IAE; import org.apache.druid.java.util.common.ISE; import org.apache.druid.java.util.common.io.Closer; +import org.apache.druid.java.util.common.parsers.ParseException; import org.apache.druid.segment.column.RowSignature; import org.apache.druid.utils.CloseableUtils; @@ -118,8 +119,13 @@ public boolean addSelection() return false; } - if (!writeData()) { - return false; + try { + if (!writeData()) { + return false; + } + } + catch (Exception e) { + throw new ParseException("", e, "Unable to add the row to the frame. Type conversion might be required."); } final MemoryRange rowOffsetCursor = rowOffsetMemory.cursor(); diff --git a/processing/src/main/java/org/apache/druid/frame/write/columnar/ColumnarFrameWriter.java b/processing/src/main/java/org/apache/druid/frame/write/columnar/ColumnarFrameWriter.java index 76f5a17d27af..eff5213ae100 100644 --- a/processing/src/main/java/org/apache/druid/frame/write/columnar/ColumnarFrameWriter.java +++ b/processing/src/main/java/org/apache/druid/frame/write/columnar/ColumnarFrameWriter.java @@ -30,6 +30,7 @@ import org.apache.druid.frame.write.FrameWriter; import org.apache.druid.frame.write.FrameWriterUtils; import org.apache.druid.java.util.common.ISE; +import org.apache.druid.java.util.common.parsers.ParseException; import org.apache.druid.segment.column.RowSignature; import javax.annotation.Nullable; @@ -74,11 +75,16 @@ public boolean addSelection() } int i = 0; - for (; i < columnWriters.size(); i++) { - if (!columnWriters.get(i).addSelection()) { - break; + try { + for (; i < columnWriters.size(); i++) { + if (!columnWriters.get(i).addSelection()) { + break; + } } } + catch (Exception e) { + throw new ParseException("", e, "Unable to add the row to the frame. Type conversion might be required."); + } if (i < columnWriters.size()) { // Add failed, clean up. From 79df11c16cd2fd1c6a9867855c42768a079917ff Mon Sep 17 00:00:00 2001 From: Laksh Singla Date: Tue, 29 Nov 2022 17:27:04 +0530 Subject: [PATCH 060/141] Improve unit test coverage for MSQ (#13398) * add faults tests for the multi stage query * add too many parttiions fault * add toomanyinputfilesfault * programmatically generate the file * refactor * Trigger Build --- .../apache/druid/msq/exec/MSQFaultsTest.java | 215 ++++++++++++++++++ .../apache/druid/msq/exec/MSQInsertTest.java | 55 +---- .../apache/druid/msq/exec/MSQReplaceTest.java | 5 +- .../apache/druid/msq/exec/MSQSelectTest.java | 3 +- .../msq/indexing/error/MSQWarningsTest.java | 3 +- .../apache/druid/msq/test/MSQTestBase.java | 20 -- .../druid/msq/test/MSQTestFileUtils.java | 74 ++++++ 7 files changed, 301 insertions(+), 74 deletions(-) create mode 100644 extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/MSQFaultsTest.java create mode 100644 extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/test/MSQTestFileUtils.java diff --git a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/MSQFaultsTest.java b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/MSQFaultsTest.java new file mode 100644 index 000000000000..7ce0a33578b9 --- /dev/null +++ b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/MSQFaultsTest.java @@ -0,0 +1,215 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.msq.exec; + +import com.google.common.collect.ImmutableMap; +import org.apache.druid.java.util.common.StringUtils; +import org.apache.druid.msq.indexing.error.TooManyClusteredByColumnsFault; +import org.apache.druid.msq.indexing.error.TooManyColumnsFault; +import org.apache.druid.msq.indexing.error.TooManyInputFilesFault; +import org.apache.druid.msq.indexing.error.TooManyPartitionsFault; +import org.apache.druid.msq.indexing.error.UnknownFault; +import org.apache.druid.msq.test.MSQTestBase; +import org.apache.druid.msq.test.MSQTestFileUtils; +import org.apache.druid.segment.column.ColumnType; +import org.apache.druid.segment.column.RowSignature; +import org.junit.Test; + +import java.io.File; +import java.io.IOException; +import java.util.Collections; +import java.util.Map; +import java.util.stream.Collectors; +import java.util.stream.IntStream; + +public class MSQFaultsTest extends MSQTestBase +{ + + @Test + public void testInsertWithTooManySegments() throws IOException + { + Map context = ImmutableMap.builder() + .putAll(DEFAULT_MSQ_CONTEXT) + .put("rowsPerSegment", 1) + .build(); + + + RowSignature rowSignature = RowSignature.builder() + .add("__time", ColumnType.LONG) + .build(); + + File file = MSQTestFileUtils.generateTemporaryNdJsonFile(30000, 1); + String filePathAsJson = queryFramework().queryJsonMapper().writeValueAsString(file.getAbsolutePath()); + + testIngestQuery().setSql(" insert into foo1 SELECT\n" + + " floor(TIME_PARSE(\"timestamp\") to day) AS __time\n" + + "FROM TABLE(\n" + + " EXTERN(\n" + + " '{ \"files\": [" + filePathAsJson + "],\"type\":\"local\"}',\n" + + " '{\"type\": \"json\"}',\n" + + " '[{\"name\": \"timestamp\",\"type\":\"string\"}]'\n" + + " )\n" + + ") PARTITIONED by day") + .setExpectedDataSource("foo1") + .setExpectedRowSignature(rowSignature) + .setQueryContext(context) + .setExpectedMSQFault(new TooManyPartitionsFault(25000)) + .verifyResults(); + + } + + @Test + public void testInsertWithUnsupportedColumnType() + { + RowSignature dummyRowSignature = RowSignature.builder().add("__time", ColumnType.LONG).build(); + + testIngestQuery() + .setSql(StringUtils.format( + " insert into foo1 SELECT\n" + + " floor(TIME_PARSE(\"timestamp\") to day) AS __time,\n" + + " col1\n" + + "FROM TABLE(\n" + + " EXTERN(\n" + + " '{ \"files\": [\"ignored\"],\"type\":\"local\"}',\n" + + " '{\"type\": \"json\"}',\n" + + " '[{\"name\": \"timestamp\", \"type\": \"string\"},{\"name\": \"col1\", \"type\": \"long_array\"} ]'\n" + + " )\n" + + ") PARTITIONED by day" + )) + .setExpectedDataSource("foo1") + .setExpectedRowSignature(dummyRowSignature) + .setExpectedMSQFault(UnknownFault.forMessage( + "org.apache.druid.java.util.common.ISE: Cannot create dimension for type [ARRAY]")) + .verifyResults(); + } + + @Test + public void testInsertWithManyColumns() + { + RowSignature dummyRowSignature = RowSignature.builder().add("__time", ColumnType.LONG).build(); + + final int numColumns = 2000; + + String columnNames = IntStream.range(1, numColumns) + .mapToObj(i -> "col" + i).collect(Collectors.joining(", ")); + + String externSignature = IntStream.range(1, numColumns) + .mapToObj(i -> StringUtils.format( + "{\"name\": \"col%d\", \"type\": \"string\"}", + i + )) + .collect(Collectors.joining(", ")); + + testIngestQuery() + .setSql(StringUtils.format( + " insert into foo1 SELECT\n" + + " floor(TIME_PARSE(\"timestamp\") to day) AS __time,\n" + + " %s\n" + + "FROM TABLE(\n" + + " EXTERN(\n" + + " '{ \"files\": [\"ignored\"],\"type\":\"local\"}',\n" + + " '{\"type\": \"json\"}',\n" + + " '[{\"name\": \"timestamp\", \"type\": \"string\"}, %s]'\n" + + " )\n" + + ") PARTITIONED by day", + columnNames, + externSignature + )) + .setExpectedDataSource("foo1") + .setExpectedRowSignature(dummyRowSignature) + .setExpectedMSQFault(new TooManyColumnsFault(numColumns + 2, 2000)) + .verifyResults(); + } + + @Test + public void testInsertWithHugeClusteringKeys() + { + RowSignature dummyRowSignature = RowSignature.builder().add("__time", ColumnType.LONG).build(); + + final int numColumns = 1700; + + String columnNames = IntStream.range(1, numColumns) + .mapToObj(i -> "col" + i).collect(Collectors.joining(", ")); + + String clusteredByClause = IntStream.range(1, numColumns + 1) + .mapToObj(String::valueOf) + .collect(Collectors.joining(", ")); + + String externSignature = IntStream.range(1, numColumns) + .mapToObj(i -> StringUtils.format( + "{\"name\": \"col%d\", \"type\": \"string\"}", + i + )) + .collect(Collectors.joining(", ")); + + testIngestQuery() + .setSql(StringUtils.format( + " insert into foo1 SELECT\n" + + " floor(TIME_PARSE(\"timestamp\") to day) AS __time,\n" + + " %s\n" + + "FROM TABLE(\n" + + " EXTERN(\n" + + " '{ \"files\": [\"ignored\"],\"type\":\"local\"}',\n" + + " '{\"type\": \"json\"}',\n" + + " '[{\"name\": \"timestamp\", \"type\": \"string\"}, %s]'\n" + + " )\n" + + ") PARTITIONED by day CLUSTERED BY %s", + columnNames, + externSignature, + clusteredByClause + )) + .setExpectedDataSource("foo1") + .setExpectedRowSignature(dummyRowSignature) + .setExpectedMSQFault(new TooManyClusteredByColumnsFault(numColumns + 2, 1500, 0)) + .verifyResults(); + } + + @Test + public void testTooManyInputFiles() throws IOException + { + RowSignature dummyRowSignature = RowSignature.builder().add("__time", ColumnType.LONG).build(); + + final int numFiles = 20000; + + final File toRead = MSQTestFileUtils.getResourceAsTemporaryFile(this, "/wikipedia-sampled.json"); + final String toReadFileNameAsJson = queryFramework().queryJsonMapper().writeValueAsString(toRead.getAbsolutePath()); + + String externalFiles = String.join(", ", Collections.nCopies(numFiles, toReadFileNameAsJson)); + + testIngestQuery() + .setSql(StringUtils.format( + "insert into foo1 SELECT\n" + + " floor(TIME_PARSE(\"timestamp\") to day) AS __time\n" + + "FROM TABLE(\n" + + " EXTERN(\n" + + " '{ \"files\": [%s],\"type\":\"local\"}',\n" + + " '{\"type\": \"csv\", \"hasHeaderRow\": true}',\n" + + " '[{\"name\": \"timestamp\", \"type\": \"string\"}]'\n" + + " )\n" + + ") PARTITIONED by day", + externalFiles + )) + .setExpectedDataSource("foo1") + .setExpectedRowSignature(dummyRowSignature) + .setExpectedMSQFault(new TooManyInputFilesFault(numFiles, Limits.MAX_INPUT_FILES_PER_WORKER, 2)) + .verifyResults(); + } + +} diff --git a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/MSQInsertTest.java b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/MSQInsertTest.java index 64b9a8067240..940008080436 100644 --- a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/MSQInsertTest.java +++ b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/MSQInsertTest.java @@ -27,13 +27,12 @@ import org.apache.druid.hll.HyperLogLogCollector; import org.apache.druid.java.util.common.ISE; import org.apache.druid.java.util.common.Intervals; -import org.apache.druid.java.util.common.StringUtils; import org.apache.druid.java.util.common.granularity.Granularities; import org.apache.druid.msq.indexing.error.ColumnNameRestrictedFault; import org.apache.druid.msq.indexing.error.InsertTimeNullFault; import org.apache.druid.msq.indexing.error.RowTooLargeFault; -import org.apache.druid.msq.indexing.error.TooManyClusteredByColumnsFault; import org.apache.druid.msq.test.MSQTestBase; +import org.apache.druid.msq.test.MSQTestFileUtils; import org.apache.druid.msq.util.MultiStageQueryContext; import org.apache.druid.query.aggregation.LongSumAggregatorFactory; import org.apache.druid.query.aggregation.hyperloglog.HyperUniquesAggregatorFactory; @@ -55,8 +54,6 @@ import java.util.Map; import java.util.Set; import java.util.TreeSet; -import java.util.stream.Collectors; -import java.util.stream.IntStream; public class MSQInsertTest extends MSQTestBase { @@ -83,7 +80,7 @@ public void testInsertOnFoo1() @Test public void testInsertOnExternalDataSource() throws IOException { - final File toRead = getResourceAsTemporaryFile("/wikipedia-sampled.json"); + final File toRead = MSQTestFileUtils.getResourceAsTemporaryFile(this, "/wikipedia-sampled.json"); final String toReadFileNameAsJson = queryFramework().queryJsonMapper().writeValueAsString(toRead.getAbsolutePath()); RowSignature rowSignature = RowSignature.builder() @@ -308,7 +305,7 @@ public void testRollUpOnFoo1ComplexCol() @Test public void testRollUpOnExternalDataSource() throws IOException { - final File toRead = getResourceAsTemporaryFile("/wikipedia-sampled.json"); + final File toRead = MSQTestFileUtils.getResourceAsTemporaryFile(this, "/wikipedia-sampled.json"); final String toReadFileNameAsJson = queryFramework().queryJsonMapper().writeValueAsString(toRead.getAbsolutePath()); RowSignature rowSignature = RowSignature.builder() @@ -344,7 +341,7 @@ public void testRollUpOnExternalDataSource() throws IOException @Test() public void testRollUpOnExternalDataSourceWithCompositeKey() throws IOException { - final File toRead = getResourceAsTemporaryFile("/wikipedia-sampled.json"); + final File toRead = MSQTestFileUtils.getResourceAsTemporaryFile(this, "/wikipedia-sampled.json"); final String toReadFileNameAsJson = queryFramework().queryJsonMapper().writeValueAsString(toRead.getAbsolutePath()); RowSignature rowSignature = RowSignature.builder() @@ -448,48 +445,6 @@ public void testIncorrectInsertQuery() .verifyPlanningErrors(); } - @Test - public void testInsertWithHugeClusteringKeys() - { - RowSignature dummyRowSignature = RowSignature.builder().add("__time", ColumnType.LONG).build(); - - final int numColumns = 1700; - - String columnNames = IntStream.range(1, numColumns) - .mapToObj(i -> "col" + i).collect(Collectors.joining(", ")); - - String clusteredByClause = IntStream.range(1, numColumns + 1) - .mapToObj(String::valueOf) - .collect(Collectors.joining(", ")); - - String externSignature = IntStream.range(1, numColumns) - .mapToObj(i -> StringUtils.format( - "{\"name\": \"col%d\", \"type\": \"string\"}", - i - )) - .collect(Collectors.joining(", ")); - - testIngestQuery() - .setSql(StringUtils.format( - " insert into foo1 SELECT\n" - + " floor(TIME_PARSE(\"timestamp\") to day) AS __time,\n" - + " %s\n" - + "FROM TABLE(\n" - + " EXTERN(\n" - + " '{ \"files\": [\"ignored\"],\"type\":\"local\"}',\n" - + " '{\"type\": \"json\"}',\n" - + " '[{\"name\": \"timestamp\", \"type\": \"string\"}, %s]'\n" - + " )\n" - + ") PARTITIONED by day CLUSTERED BY %s", - columnNames, - externSignature, - clusteredByClause - )) - .setExpectedDataSource("foo1") - .setExpectedRowSignature(dummyRowSignature) - .setExpectedMSQFault(new TooManyClusteredByColumnsFault(numColumns + 2, 1500, 0)) - .verifyResults(); - } @Test public void testInsertRestrictedColumns() @@ -542,7 +497,7 @@ public void testInsertQueryWithInvalidSubtaskCount() @Test public void testInsertWithTooLargeRowShouldThrowException() throws IOException { - final File toRead = getResourceAsTemporaryFile("/wikipedia-sampled.json"); + final File toRead = MSQTestFileUtils.getResourceAsTemporaryFile(this, "/wikipedia-sampled.json"); final String toReadFileNameAsJson = queryFramework().queryJsonMapper().writeValueAsString(toRead.getAbsolutePath()); Mockito.doReturn(500).when(workerMemoryParameters).getLargeFrameSize(); diff --git a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/MSQReplaceTest.java b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/MSQReplaceTest.java index b268b1dcf0d3..49bb7f6d090d 100644 --- a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/MSQReplaceTest.java +++ b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/MSQReplaceTest.java @@ -23,6 +23,7 @@ import com.google.common.collect.ImmutableSet; import org.apache.druid.java.util.common.Intervals; import org.apache.druid.msq.test.MSQTestBase; +import org.apache.druid.msq.test.MSQTestFileUtils; import org.apache.druid.segment.column.ColumnType; import org.apache.druid.segment.column.RowSignature; import org.apache.druid.sql.SqlPlanningException; @@ -109,7 +110,7 @@ public void testReplaceOnFoo1WithAllExtern() throws IOException .add("__time", ColumnType.LONG) .add("cnt", ColumnType.LONG).build(); - final File toRead = getResourceAsTemporaryFile("/wikipedia-sampled.json"); + final File toRead = MSQTestFileUtils.getResourceAsTemporaryFile(this, "/wikipedia-sampled.json"); final String toReadFileNameAsJson = queryFramework().queryJsonMapper().writeValueAsString(toRead.getAbsolutePath()); testIngestQuery().setSql(" REPLACE INTO foo1 OVERWRITE ALL SELECT " @@ -147,7 +148,7 @@ public void testReplaceOnFoo1WithWhereExtern() throws IOException .add("__time", ColumnType.LONG) .add("user", ColumnType.STRING).build(); - final File toRead = getResourceAsTemporaryFile("/wikipedia-sampled.json"); + final File toRead = MSQTestFileUtils.getResourceAsTemporaryFile(this, "/wikipedia-sampled.json"); final String toReadFileNameAsJson = queryFramework().queryJsonMapper().writeValueAsString(toRead.getAbsolutePath()); testIngestQuery().setSql(" REPLACE INTO foo1 OVERWRITE WHERE __time >= TIMESTAMP '2016-06-27 01:00:00.00' AND __time < TIMESTAMP '2016-06-27 02:00:00.00' " diff --git a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/MSQSelectTest.java b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/MSQSelectTest.java index 26d566b31e8c..dd9bf05d2094 100644 --- a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/MSQSelectTest.java +++ b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/MSQSelectTest.java @@ -34,6 +34,7 @@ import org.apache.druid.msq.indexing.error.CannotParseExternalDataFault; import org.apache.druid.msq.shuffle.DurableStorageUtils; import org.apache.druid.msq.test.MSQTestBase; +import org.apache.druid.msq.test.MSQTestFileUtils; import org.apache.druid.query.InlineDataSource; import org.apache.druid.query.QueryDataSource; import org.apache.druid.query.TableDataSource; @@ -733,7 +734,7 @@ public void testGroupByOrderByAggregationWithLimitAndOffset() @Test public void testExternSelect1() throws IOException { - final File toRead = getResourceAsTemporaryFile("/wikipedia-sampled.json"); + final File toRead = MSQTestFileUtils.getResourceAsTemporaryFile(this, "/wikipedia-sampled.json"); final String toReadAsJson = queryFramework().queryJsonMapper().writeValueAsString(toRead.getAbsolutePath()); RowSignature rowSignature = RowSignature.builder() diff --git a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/indexing/error/MSQWarningsTest.java b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/indexing/error/MSQWarningsTest.java index 5ad2191af3f1..38318bf889b3 100644 --- a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/indexing/error/MSQWarningsTest.java +++ b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/indexing/error/MSQWarningsTest.java @@ -29,6 +29,7 @@ import org.apache.druid.msq.indexing.MSQSpec; import org.apache.druid.msq.indexing.MSQTuningConfig; import org.apache.druid.msq.test.MSQTestBase; +import org.apache.druid.msq.test.MSQTestFileUtils; import org.apache.druid.msq.util.MultiStageQueryContext; import org.apache.druid.query.Query; import org.apache.druid.query.aggregation.CountAggregatorFactory; @@ -65,7 +66,7 @@ public class MSQWarningsTest extends MSQTestBase @Before public void setUp3() throws IOException { - toRead = getResourceAsTemporaryFile("/unparseable.gz"); + toRead = MSQTestFileUtils.getResourceAsTemporaryFile(this, "/unparseable.gz"); toReadFileNameAsJson = queryFramework().queryJsonMapper().writeValueAsString(toRead.getAbsolutePath()); rowSignature = RowSignature.builder() diff --git a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/test/MSQTestBase.java b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/test/MSQTestBase.java index 096d2cba1c69..19b85034ff93 100644 --- a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/test/MSQTestBase.java +++ b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/test/MSQTestBase.java @@ -28,7 +28,6 @@ import com.google.common.collect.ImmutableMap; import com.google.common.collect.ImmutableSet; import com.google.common.collect.Iterables; -import com.google.common.io.ByteStreams; import com.google.inject.Injector; import com.google.inject.Key; import com.google.inject.Module; @@ -56,7 +55,6 @@ import org.apache.druid.indexing.common.task.IndexTask; import org.apache.druid.indexing.common.task.batch.parallel.ParallelIndexTuningConfig; import org.apache.druid.initialization.CoreInjectorBuilder; -import org.apache.druid.java.util.common.IOE; import org.apache.druid.java.util.common.ISE; import org.apache.druid.java.util.common.Pair; import org.apache.druid.java.util.common.StringUtils; @@ -167,8 +165,6 @@ import java.io.Closeable; import java.io.File; import java.io.IOException; -import java.io.InputStream; -import java.nio.file.Files; import java.util.ArrayList; import java.util.Arrays; import java.util.Collection; @@ -446,22 +442,6 @@ protected Map defaultScanQueryContext(final RowSignature signatu } } - /** - * Helper method that copies a resource to a temporary file, then returns it. - */ - protected File getResourceAsTemporaryFile(final String resource) throws IOException - { - final File file = temporaryFolder.newFile(); - final InputStream stream = getClass().getResourceAsStream(resource); - - if (stream == null) { - throw new IOE("No such resource [%s]", resource); - } - - ByteStreams.copy(stream, Files.newOutputStream(file.toPath())); - return file; - } - @Nonnull private Supplier> getSupplierForSegment(SegmentId segmentId) { diff --git a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/test/MSQTestFileUtils.java b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/test/MSQTestFileUtils.java new file mode 100644 index 000000000000..70c2f55ea33f --- /dev/null +++ b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/test/MSQTestFileUtils.java @@ -0,0 +1,74 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.msq.test; + +import com.google.common.collect.ImmutableList; +import com.google.common.io.ByteStreams; +import org.apache.druid.java.util.common.IOE; +import org.apache.druid.java.util.common.StringUtils; +import org.apache.druid.sql.calcite.BaseCalciteQueryTest; + +import java.io.File; +import java.io.IOException; +import java.io.InputStream; +import java.nio.charset.StandardCharsets; +import java.nio.file.Files; +import java.nio.file.StandardOpenOption; + +public class MSQTestFileUtils +{ + + /** + * Helper method that copies a resource to a temporary file, then returns it. + */ + public static File getResourceAsTemporaryFile(Object object, final String resource) throws IOException + { + final File file = BaseCalciteQueryTest.temporaryFolder.newFile(); + final InputStream stream = object.getClass().getResourceAsStream(resource); + + if (stream == null) { + throw new IOE("No such resource [%s]", resource); + } + + ByteStreams.copy(stream, Files.newOutputStream(file.toPath())); + return file; + } + + /** + * Helper method that populates a temporary file with {@code numRows} rows and {@code numColumns} columns where the + * first column is a string 'timestamp' while the rest are string columns with junk value + */ + public static File generateTemporaryNdJsonFile(final int numRows, final int numColumns) throws IOException + { + final File file = BaseCalciteQueryTest.temporaryFolder.newFile(); + for (int currentRow = 0; currentRow < numRows; ++currentRow) { + StringBuilder sb = new StringBuilder(); + sb.append("{"); + sb.append("\"timestamp\":\"2016-06-27T00:00:11.080Z\""); + for (int currentColumn = 1; currentColumn < numColumns; ++currentColumn) { + sb.append(StringUtils.format(",\"column%s\":\"val%s\"", currentColumn, currentRow)); + } + sb.append("}"); + Files.write(file.toPath(), ImmutableList.of(sb.toString()), StandardCharsets.UTF_8, StandardOpenOption.APPEND); + } + file.deleteOnExit(); + return file; + } +} From cafdc83e85c35438fdced27f8756567fe2b37125 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Tue, 29 Nov 2022 17:30:15 +0530 Subject: [PATCH 061/141] Update middleManager runtime.properties --- .../quickstart/middleManager/runtime.properties | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) diff --git a/examples/conf/druid/single-server/quickstart/middleManager/runtime.properties b/examples/conf/druid/single-server/quickstart/middleManager/runtime.properties index a90e62579676..08c58bae6de0 100644 --- a/examples/conf/druid/single-server/quickstart/middleManager/runtime.properties +++ b/examples/conf/druid/single-server/quickstart/middleManager/runtime.properties @@ -29,9 +29,8 @@ druid.indexer.runner.javaCommand=bin/run-java druid.indexer.task.baseTaskDir=var/druid/task # Processing threads and buffers on Peons -#druid.indexer.fork.property.druid.processing.numMergeBuffers=2 -#druid.indexer.fork.property.druid.processing.buffer.sizeBytes=100MiB -#druid.indexer.fork.property.druid.processing.numThreads=1 +# Determined automatically based on available memory. For details on how to manually set parameters: +# https://druid.apache.org/docs/latest/operations/basic-cluster-tuning.html#guidelines-for-processing-threads-and-buffers # Hadoop indexing druid.indexer.task.hadoopWorkingPath=var/druid/hadoop-tmp From 5c520e0cf92fec6a336ff71d1eb36ddbd9503066 Mon Sep 17 00:00:00 2001 From: Jill Osborne Date: Tue, 29 Nov 2022 17:26:32 +0000 Subject: [PATCH 062/141] Update LDAP configuration docs (#13245) * Update LDAP configuration docs * Updated after review * Update auth-ldap.md Updated. * Update auth-ldap.md * Updated spelling file * Update docs/operations/auth-ldap.md Co-authored-by: Charles Smith * Update docs/operations/auth-ldap.md Co-authored-by: Charles Smith * Update docs/operations/auth-ldap.md Co-authored-by: Charles Smith * Update auth-ldap.md Co-authored-by: Charles Smith --- .../extensions-core/druid-basic-security.md | 420 ++++++++++++++---- docs/operations/auth-ldap.md | 295 ++++++++---- docs/operations/security-overview.md | 75 +--- website/.spelling | 3 + 4 files changed, 542 insertions(+), 251 deletions(-) diff --git a/docs/development/extensions-core/druid-basic-security.md b/docs/development/extensions-core/druid-basic-security.md index b6698b4f153d..4e042b4405f3 100644 --- a/docs/development/extensions-core/druid-basic-security.md +++ b/docs/development/extensions-core/druid-basic-security.md @@ -53,12 +53,29 @@ To set the value for the configuration properties, add them to the common runtim ### General properties -|Property|Description|Default|required| -|--------|-----------|-------|--------| -|`druid.auth.basic.common.pollingPeriod`|Defines in milliseconds how often processes should poll the Coordinator for the current Druid metadata store authenticator/authorizer state.|60000|No| -|`druid.auth.basic.common.maxRandomDelay`|Defines in milliseconds the amount of random delay to add to the pollingPeriod, to spread polling requests across time.|6000|No| -|`druid.auth.basic.common.maxSyncRetries`|Determines how many times a service will retry if the authentication/authorization Druid metadata store state sync with the Coordinator fails.|10|No| -|`druid.auth.basic.common.cacheDirectory`|If defined, snapshots of the basic Authenticator and Authorizer Druid metadata store caches will be stored on disk in this directory. If this property is defined, when a service is starting, it will attempt to initialize its caches from these on-disk snapshots, if the service is unable to initialize its state by communicating with the Coordinator.|null|No| +**`druid.auth.basic.common.pollingPeriod`** + +Defines in milliseconds how often processes should poll the Coordinator for the current Druid metadata store authenticator/authorizer state.
+         **Required**: No
+         **Default**: 60000 + +**`druid.auth.basic.common.maxRandomDelay`** + +Defines in milliseconds the amount of random delay to add to the pollingPeriod, to spread polling requests across time.
+         **Required**: No
+         **Default**: 6000 + +**`druid.auth.basic.common.maxSyncRetries`** + +Determines how many times a service will retry if the authentication/authorization Druid metadata store state sync with the Coordinator fails.
+         **Required**: No
+         **Default**: 10 + +**`druid.auth.basic.common.cacheDirectory`** + +If defined, snapshots of the basic Authenticator and Authorizer Druid metadata store caches will be stored on disk in this directory. If this property is defined, when a service is starting, it will attempt to initialize its caches from these on-disk snapshots, if the service is unable to initialize its state by communicating with the Coordinator.
+         **Required**: No
+         **Default**: null ### Authenticator @@ -96,16 +113,55 @@ The remaining examples of authenticator configuration use either `MyBasicMetadat #### Properties for Druid metadata store user authentication -|Property|Description|Default|required| -|--------|-----------|-------|--------| -|`druid.auth.authenticator.MyBasicMetadataAuthenticator.initialAdminPassword`|Initial [Password Provider](../../operations/password-provider.md) for the automatically created default admin user. If no password is specified, the default admin user will not be created. If the default admin user already exists, setting this property will not affect its password.|null|No| -|`druid.auth.authenticator.MyBasicMetadataAuthenticator.initialInternalClientPassword`|Initial [Password Provider](../../operations/password-provider.md) for the default internal system user, used for internal process communication. If no password is specified, the default internal system user will not be created. If the default internal system user already exists, setting this property will not affect its password.|null|No| -|`druid.auth.authenticator.MyBasicMetadataAuthenticator.enableCacheNotifications`|If true, the Coordinator will notify Druid processes whenever a configuration change to this Authenticator occurs, allowing them to immediately update their state without waiting for polling.|true|No| -|`druid.auth.authenticator.MyBasicMetadataAuthenticator.cacheNotificationTimeout`|The timeout in milliseconds for the cache notifications.|5000|No| -|`druid.auth.authenticator.MyBasicMetadataAuthenticator.credentialIterations`|Number of iterations to use for password hashing. See [Credential iterations and API performance](#credential-iterations-and-api-performance)|10000|No| -|`druid.auth.authenticator.MyBasicMetadataAuthenticator.credentialsValidator.type`|The type of credentials store (metadata) to validate requests credentials.|metadata|No| -|`druid.auth.authenticator.MyBasicMetadataAuthenticator.skipOnFailure`|If true and the request credential doesn't exists or isn't fully configured in the credentials store, the request will proceed to next Authenticator in the chain.|false|No| -|`druid.auth.authenticator.MyBasicMetadataAuthenticator.authorizerName`|Authorizer that requests should be directed to|N/A|Yes| + +**`druid.auth.authenticator.MyBasicMetadataAuthenticator.initialAdminPassword`** + +Initial [Password Provider](../../operations/password-provider.md) for the automatically created default admin user. If no password is specified, the default admin user will not be created. If the default admin user already exists, setting this property will not affect its password.
+         **Required**: No
+         **Default**: null + +**`druid.auth.authenticator.MyBasicMetadataAuthenticator.initialInternalClientPassword`** + +Initial [Password Provider](../../operations/password-provider.md) for the default internal system user, used for internal process communication. If no password is specified, the default internal system user will not be created. If the default internal system user already exists, setting this property will not affect its password.
+         **Required**: No
+         **Default**: null + +**`druid.auth.authenticator.MyBasicMetadataAuthenticator.enableCacheNotifications`** + +If true, the Coordinator will notify Druid processes whenever a configuration change to this Authenticator occurs, allowing them to immediately update their state without waiting for polling.
+         **Required**: No
+         **Default**: True + +**`druid.auth.authenticator.MyBasicMetadataAuthenticator.cacheNotificationTimeout`** + +The timeout in milliseconds for the cache notifications.
+         **Required**: No
+         **Default**: 5000 + +**`druid.auth.authenticator.MyBasicMetadataAuthenticator.credentialIterations`** + +Number of iterations to use for password hashing. See [Credential iterations and API performance](#credential-iterations-and-api-performance)
+         **Required**: No
+         **Default**: 10000 + +**`druid.auth.authenticator.MyBasicMetadataAuthenticator.credentialsValidator.type`** + +The type of credentials store (metadata) to validate requests credentials.
+         **Required**: No
+         **Default**: metadata + +**`druid.auth.authenticator.MyBasicMetadataAuthenticator.skipOnFailure`** + +If true and the request credential doesn't exists or isn't fully configured in the credentials store, the request will proceed to next Authenticator in the chain.
+         **Required**: No
+         **Default**: false + +**`druid.auth.authenticator.MyBasicMetadataAuthenticator.authorizerName`** + +Authorizer that requests should be directed to.
+         **Required**: Yes
+         **Default**: N/A + ##### Credential iterations and API performance @@ -121,25 +177,107 @@ If Druid uses the default credentials validator (i.e., `credentialsValidator.typ #### Properties for LDAP user authentication -|Property|Description|Default|required| -|--------|-----------|-------|--------| -|`druid.auth.authenticator.MyBasicLDAPAuthenticator.initialAdminPassword`|Initial [Password Provider](../../operations/password-provider.md) for the automatically created default admin user. If no password is specified, the default admin user will not be created. If the default admin user already exists, setting this property will not affect its password.|null|No| -|`druid.auth.authenticator.MyBasicLDAPAuthenticator.initialInternalClientPassword`|Initial [Password Provider](../../operations/password-provider.md) for the default internal system user, used for internal process communication. If no password is specified, the default internal system user will not be created. If the default internal system user already exists, setting this property will not affect its password.|null|No| -|`druid.auth.authenticator.MyBasicLDAPAuthenticator.enableCacheNotifications`|If true, the Coordinator will notify Druid processes whenever a configuration change to this Authenticator occurs, allowing them to immediately update their state without waiting for polling.|true|No| -|`druid.auth.authenticator.MyBasicLDAPAuthenticator.cacheNotificationTimeout`|The timeout in milliseconds for the cache notifications.|5000|No| -|`druid.auth.authenticator.MyBasicLDAPAuthenticator.credentialIterations`|Number of iterations to use for password hashing.|10000|No| -|`druid.auth.authenticator.MyBasicLDAPAuthenticator.credentialsValidator.type`|The type of credentials store (ldap) to validate requests credentials.|metadata|No| -|`druid.auth.authenticator.MyBasicLDAPAuthenticator.credentialsValidator.url`|URL of the LDAP server.|null|Yes| -|`druid.auth.authenticator.MyBasicLDAPAuthenticator.credentialsValidator.bindUser`|LDAP bind user username.|null|Yes| -|`druid.auth.authenticator.MyBasicLDAPAuthenticator.credentialsValidator.bindPassword`|[Password Provider](../../operations/password-provider.md) LDAP bind user password.|null|Yes| -|`druid.auth.authenticator.MyBasicLDAPAuthenticator.credentialsValidator.baseDn`|The point from where the LDAP server will search for users.|null|Yes| -|`druid.auth.authenticator.MyBasicLDAPAuthenticator.credentialsValidator.userSearch`|The filter/expression to use for the search. For example, (&(sAMAccountName=%s)(objectClass=user))|null|Yes| -|`druid.auth.authenticator.MyBasicLDAPAuthenticator.credentialsValidator.userAttribute`|The attribute id identifying the attribute that will be returned as part of the search. For example, sAMAccountName. |null|Yes| -|`druid.auth.authenticator.MyBasicLDAPAuthenticator.credentialsValidator.credentialVerifyDuration`|The duration in seconds for how long valid credentials are verifiable within the cache when not requested.|600|No| -|`druid.auth.authenticator.MyBasicLDAPAuthenticator.credentialsValidator.credentialMaxDuration`|The max duration in seconds for valid credentials that can reside in cache regardless of how often they are requested.|3600|No| -|`druid.auth.authenticator.MyBasicLDAPAuthenticator.credentialsValidator.credentialCacheSize`|The valid credentials cache size. The cache uses a LRU policy.|100|No| -|`druid.auth.authenticator.MyBasicLDAPAuthenticator.skipOnFailure`|If true and the request credential doesn't exists or isn't fully configured in the credentials store, the request will proceed to next Authenticator in the chain.|false|No| -|`druid.auth.authenticator.MyBasicLDAPAuthenticator.authorizerName`|Authorizer that requests should be directed to.|N/A|Yes| +**`druid.auth.authenticator.MyBasicLDAPAuthenticator.initialAdminPassword`** + +Initial [Password Provider](../../operations/password-provider.md) for the automatically created default admin user. If no password is specified, the default admin user will not be created. If the default admin user already exists, setting this property will not affect its password.
+         **Required**: No
+         **Default**: null + +**`druid.auth.authenticator.MyBasicLDAPAuthenticator.initialInternalClientPassword`** + +Initial [Password Provider](../../operations/password-provider.md) for the default internal system user, used for internal process communication. If no password is specified, the default internal system user will not be created. If the default internal system user already exists, setting this property will not affect its password.
+         **Required**: No
+         **Default**: null + +**`druid.auth.authenticator.MyBasicLDAPAuthenticator.enableCacheNotifications`** + +If true, the Coordinator will notify Druid processes whenever a configuration change to this Authenticator occurs, allowing them to immediately update their state without waiting for polling.
+         **Required**: No
+         **Default**: true + +**`druid.auth.authenticator.MyBasicLDAPAuthenticator.cacheNotificationTimeout`** + +The timeout in milliseconds for the cache notifications.
+         **Required**: No
+         **Default**: 5000 + +**`druid.auth.authenticator.MyBasicLDAPAuthenticator.credentialIterations`** + +Number of iterations to use for password hashing.
+         **Required**: No
+         **Default**: 10000 + +**`druid.auth.authenticator.MyBasicLDAPAuthenticator.credentialsValidator.type`** + +The type of credentials store (ldap) to validate requests credentials.
+         **Required**: No
+         **Default**: metadata + +**`druid.auth.authenticator.MyBasicLDAPAuthenticator.credentialsValidator.url`** + +URL of the LDAP server.
+         **Required**: Yes
+         **Default**: null + +**`druid.auth.authenticator.MyBasicLDAPAuthenticator.credentialsValidator.bindUser`** + +LDAP bind user username.
+         **Required**: Yes
+         **Default**: null + +**`druid.auth.authenticator.MyBasicLDAPAuthenticator.credentialsValidator.bindPassword`** + +[Password Provider](../../operations/password-provider.md) LDAP bind user password.
+         **Required**: Yes
+         **Default**: null + +**`druid.auth.authenticator.MyBasicLDAPAuthenticator.credentialsValidator.baseDn`** + +The point from where the LDAP server will search for users.
+         **Required**: Yes
+         **Default**: null + +**`druid.auth.authenticator.MyBasicLDAPAuthenticator.credentialsValidator.userSearch`** + +The filter/expression to use for the search. For example, (&(sAMAccountName=%s)(objectClass=user))
+         **Required**: Yes
+         **Default**: null + +**`druid.auth.authenticator.MyBasicLDAPAuthenticator.credentialsValidator.userAttribute`** + +The attribute id identifying the attribute that will be returned as part of the search. For example, sAMAccountName.
+         **Required**: Yes
+         **Default**: null + +**`druid.auth.authenticator.MyBasicLDAPAuthenticator.credentialsValidator.credentialVerifyDuration`** + +The duration in seconds for how long valid credentials are verifiable within the cache when not requested.
+         **Required**: No
+         **Default**: 600 + +**`druid.auth.authenticator.MyBasicLDAPAuthenticator.credentialsValidator.credentialMaxDuration`** + +The max duration in seconds for valid credentials that can reside in cache regardless of how often they are requested.
+         **Required**: No
+         **Default**: 3600 + +**`druid.auth.authenticator.MyBasicLDAPAuthenticator.credentialsValidator.credentialCacheSize`** + +The valid credentials cache size. The cache uses a LRU policy.
+         **Required**: No
+         **Default**: 100 + +**`druid.auth.authenticator.MyBasicLDAPAuthenticator.skipOnFailure`** + +If true and the request credential doesn't exists or isn't fully configured in the credentials store, the request will proceed to next Authenticator in the chain.
+         **Required**: No
+         **Default**: false + +**`druid.auth.authenticator.MyBasicLDAPAuthenticator.authorizerName`** + +Authorizer that requests should be directed to.
+         **Required**: Yes
+         **Default**: N/A ### Escalator @@ -155,11 +293,24 @@ druid.escalator.authorizerName=MyBasicMetadataAuthorizer ``` #### Properties -|Property|Description|Default|required| -|--------|-----------|-------|--------| -|`druid.escalator.internalClientUsername`|The escalator will use this username for requests made as the internal system user.|n/a|Yes| -|`druid.escalator.internalClientPassword`|The escalator will use this [Password Provider](../../operations/password-provider.md) for requests made as the internal system user.|n/a|Yes| -|`druid.escalator.authorizerName`|Authorizer that requests should be directed to.|n/a|Yes| + +**`druid.escalator.internalClientUsername`** + +The escalator will use this username for requests made as the internal system user.
+         **Required**: Yes
+         **Default**: N/A + +**`druid.escalator.internalClientPassword`** + +The escalator will use this [Password Provider](../../operations/password-provider.md) for requests made as the internal system user.
+         **Required**: Yes
+         **Default**: N/A + +**`druid.escalator.authorizerName`** + +Authorizer that requests should be directed to.
+         **Required**: Yes
+         **Default**: N/A ### Authorizer @@ -182,24 +333,131 @@ druid.auth.authorizer.MyBasicMetadataAuthorizer.type=basic The examples in the rest of this article use `MyBasicMetadataAuthorizer` or `MyBasicLDAPAuthorizer` as the authorizer name. #### Properties for Druid metadata store user authorization -|Property|Description|Default|required| -|--------|-----------|-------|--------| -|`druid.auth.authorizer.MyBasicMetadataAuthorizer.enableCacheNotifications`|If true, the Coordinator will notify Druid processes whenever a configuration change to this Authorizer occurs, allowing them to immediately update their state without waiting for polling.|true|No| -|`druid.auth.authorizer.MyBasicMetadataAuthorizer.cacheNotificationTimeout`|The timeout in milliseconds for the cache notifications.|5000|No| -|`druid.auth.authorizer.MyBasicMetadataAuthorizer.initialAdminUser`|The initial admin user with role defined in initialAdminRole property if specified, otherwise the default admin role will be assigned.|admin|No| -|`druid.auth.authorizer.MyBasicMetadataAuthorizer.initialAdminRole`|The initial admin role to create if it doesn't already exists.|admin|No| -|`druid.auth.authorizer.MyBasicMetadataAuthorizer.roleProvider.type`|The type of role provider to authorize requests credentials.|metadata|No + +**`druid.auth.authorizer.MyBasicMetadataAuthorizer.enableCacheNotifications`** + +If true, the Coordinator will notify Druid processes whenever a configuration change to this Authorizer occurs, allowing them to immediately update their state without waiting for polling.
+         **Required**: No
+         **Default**: true + +**`druid.auth.authorizer.MyBasicMetadataAuthorizer.cacheNotificationTimeout`** + +The timeout in milliseconds for the cache notifications.
+         **Required**: No
+         **Default**: 5000 + +**`druid.auth.authorizer.MyBasicMetadataAuthorizer.initialAdminUser`** + +The initial admin user with role defined in initialAdminRole property if specified, otherwise the default admin role will be assigned.
+         **Required**: No
+         **Default**: admin + +**`druid.auth.authorizer.MyBasicMetadataAuthorizer.initialAdminRole`** + +The initial admin role to create if it doesn't already exists.
+         **Required**: No
+         **Default**: admin + +**`druid.auth.authorizer.MyBasicMetadataAuthorizer.roleProvider.type`** + +The type of role provider to authorize requests credentials.
+         **Required**: No
+         **Default**: metadata #### Properties for LDAP user authorization -|Property|Description|Default|required| -|--------|-----------|-------|--------| -|`druid.auth.authorizer.MyBasicLDAPAuthorizer.enableCacheNotifications`|If true, the Coordinator will notify Druid processes whenever a configuration change to this Authorizer occurs, allowing them to immediately update their state without waiting for polling.|true|No| -|`druid.auth.authorizer.MyBasicLDAPAuthorizer.cacheNotificationTimeout`|The timeout in milliseconds for the cache notifications.|5000|No| -|`druid.auth.authorizer.MyBasicLDAPAuthorizer.initialAdminUser`|The initial admin user with role defined in initialAdminRole property if specified, otherwise the default admin role will be assigned.|admin|No| -|`druid.auth.authorizer.MyBasicLDAPAuthorizer.initialAdminRole`|The initial admin role to create if it doesn't already exists.|admin|No| -|`druid.auth.authorizer.MyBasicLDAPAuthorizer.initialAdminGroupMapping`|The initial admin group mapping with role defined in initialAdminRole property if specified, otherwise the default admin role will be assigned. The name of this initial admin group mapping will be set to adminGroupMapping|null|No| -|`druid.auth.authorizer.MyBasicLDAPAuthorizer.roleProvider.type`|The type of role provider (ldap) to authorize requests credentials.|metadata|No -|`druid.auth.authorizer.MyBasicLDAPAuthorizer.roleProvider.groupFilters`|Array of LDAP group filters used to filter out the allowed set of groups returned from LDAP search. Filters can be begin with *, or end with ,* to provide configurational flexibility to limit or filter allowed set of groups available to LDAP Authorizer.|null|No| + +**`druid.auth.authorizer.MyBasicLDAPAuthorizer.enableCacheNotifications`** + +If true, the Coordinator will notify Druid processes whenever a configuration change to this Authorizer occurs, allowing them to immediately update their state without waiting for polling.
+         **Required**: No
+         **Default**: true + +**`druid.auth.authorizer.MyBasicLDAPAuthorizer.cacheNotificationTimeout`** + +The timeout in milliseconds for the cache notifications.
+         **Required**: No
+         **Default**: 5000 + +**`druid.auth.authorizer.MyBasicLDAPAuthorizer.initialAdminUser`** + +The initial admin user with role defined in initialAdminRole property if specified, otherwise the default admin role will be assigned.
+         **Required**: No
+         **Default**: admin + +**`druid.auth.authorizer.MyBasicLDAPAuthorizer.initialAdminRole`** + +The initial admin role to create if it doesn't already exists.
+         **Required**: No
+         **Default**: admin + +**`druid.auth.authorizer.MyBasicLDAPAuthorizer.initialAdminGroupMapping`** + +The initial admin group mapping with role defined in initialAdminRole property if specified, otherwise the default admin role will be assigned. The name of this initial admin group mapping will be set to adminGroupMapping
+         **Required**: No
+         **Default**: null + +**`druid.auth.authorizer.MyBasicLDAPAuthorizer.roleProvider.type`** + +The type of role provider (ldap) to authorize requests credentials.
+         **Required**: No
+         **Default**: metadata + +**`druid.auth.authorizer.MyBasicLDAPAuthorizer.roleProvider.groupFilters`** + +Array of LDAP group filters used to filter out the allowed set of groups returned from LDAP search. Filters can be begin with *, or end with ,* to provide configurational flexibility to limit or filter allowed set of groups available to LDAP Authorizer.
+         **Required**: No
+         **Default**: null + +#### Properties for LDAPS + +Use the following properties to configure Druid authentication with LDAP over TLS (LDAPS). See [Configure LDAP authentication](../../operations/auth-ldap.md) for more information. + +**`druid.auth.basic.ssl.protocol`** + +SSL protocol to use. The TLS version is 1.2.
+         **Required**: Yes
+         **Default**: tls + +**`druid.auth.basic.ssl.trustStorePath`** + +Path to the trust store file.
+         **Required**: Yes
+         **Default**: N/A + +**`druid.auth.basic.ssl.trustStorePassword`** + +Password to access the trust store file.
+         **Required**: Yes
+         **Default**: N/A + +**`druid.auth.basic.ssl.trustStoreType`** + +Format of the trust store file. For Java the format is jks.
+         **Required**: No
+         **Default**: jks + +**`druid.auth.basic.ssl.trustStoreAlgorithm`** + +Algorithm used by the trust manager to validate certificate chains.
+         **Required**: No
+         **Default**: N/A + +**`druid.auth.basic.ssl.trustStorePassword`** + +Password details that enable access to the truststore.
+         **Required**: No
+         **Default**: N/A + +Example LDAPS configuration: + +```json +druid.auth.basic.ssl.protocol=tls +druid.auth.basic.ssl.trustStorePath=/usr/local/druid-path/certs/truststore.jks +druid.auth.basic.ssl.trustStorePassword=xxxxx +druid.auth.basic.ssl.trustStoreType=jks +druid.auth.basic.ssl.trustStoreAlgorithm=PKIX +``` +You can configure `druid.auth.basic.ssl.trustStorePassword` to be a plain text password or you can set the password as an environment variable. See [Password providers](../../operations/password-provider.md) for more information. ## Usage @@ -213,19 +471,19 @@ Root path: `/druid-ext/basic-security/authentication` Each API endpoint includes {authenticatorName}, specifying which Authenticator instance is being configured. ##### User/Credential Management -`GET(/druid-ext/basic-security/authentication/db/{authenticatorName}/users)` +`GET(/druid-ext/basic-security/authentication/db/{authenticatorName}/users)`
Return a list of all user names. -`GET(/druid-ext/basic-security/authentication/db/{authenticatorName}/users/{userName})` +`GET(/druid-ext/basic-security/authentication/db/{authenticatorName}/users/{userName})`
Return the name and credentials information of the user with name {userName} -`POST(/druid-ext/basic-security/authentication/db/{authenticatorName}/users/{userName})` +`POST(/druid-ext/basic-security/authentication/db/{authenticatorName}/users/{userName})`
Create a new user with name {userName} -`DELETE(/druid-ext/basic-security/authentication/db/{authenticatorName}/users/{userName})` +`DELETE(/druid-ext/basic-security/authentication/db/{authenticatorName}/users/{userName})`
Delete the user with name {userName} -`POST(/druid-ext/basic-security/authentication/db/{authenticatorName}/users/{userName}/credentials)` +`POST(/druid-ext/basic-security/authentication/db/{authenticatorName}/users/{userName}/credentials)`
Assign a password used for HTTP basic authentication for {userName} Content: JSON password request object @@ -238,20 +496,20 @@ Example request body: ``` ##### Cache Load Status -`GET(/druid-ext/basic-security/authentication/loadStatus)` +`GET(/druid-ext/basic-security/authentication/loadStatus)`
Return the current load status of the local caches of the authentication Druid metadata store. #### Authorization API -Root path: `/druid-ext/basic-security/authorization` +Root path: `/druid-ext/basic-security/authorization`
Each API endpoint includes {authorizerName}, specifying which Authorizer instance is being configured. ##### User Creation/Deletion -`GET(/druid-ext/basic-security/authorization/db/{authorizerName}/users)` +`GET(/druid-ext/basic-security/authorization/db/{authorizerName}/users)`
Return a list of all user names. -`GET(/druid-ext/basic-security/authorization/db/{authorizerName}/users/{userName})` +`GET(/druid-ext/basic-security/authorization/db/{authorizerName}/users/{userName})`
Return the name and role information of the user with name {userName} Example output: @@ -338,20 +596,20 @@ The `resourceNamePattern` is a compiled version of the resource name regex. It i } ``` -`POST(/druid-ext/basic-security/authorization/db/{authorizerName}/users/{userName})` +`POST(/druid-ext/basic-security/authorization/db/{authorizerName}/users/{userName})`
Create a new user with name {userName} -`DELETE(/druid-ext/basic-security/authorization/db/{authorizerName}/users/{userName})` +`DELETE(/druid-ext/basic-security/authorization/db/{authorizerName}/users/{userName})`
Delete the user with name {userName} ##### Group mapping Creation/Deletion -`GET(/druid-ext/basic-security/authorization/db/{authorizerName}/groupMappings)` +`GET(/druid-ext/basic-security/authorization/db/{authorizerName}/groupMappings)`
Return a list of all group mappings. -`GET(/druid-ext/basic-security/authorization/db/{authorizerName}/groupMappings/{groupMappingName})` +`GET(/druid-ext/basic-security/authorization/db/{authorizerName}/groupMappings/{groupMappingName})`
Return the group mapping and role information of the group mapping with name {groupMappingName} -`POST(/druid-ext/basic-security/authorization/db/{authorizerName}/groupMappings/{groupMappingName})` +`POST(/druid-ext/basic-security/authorization/db/{authorizerName}/groupMappings/{groupMappingName})`
Create a new group mapping with name {groupMappingName} Content: JSON group mapping object Example request body: @@ -366,14 +624,14 @@ Example request body: } ``` -`DELETE(/druid-ext/basic-security/authorization/db/{authorizerName}/groupMappings/{groupMappingName})` +`DELETE(/druid-ext/basic-security/authorization/db/{authorizerName}/groupMappings/{groupMappingName})`
Delete the group mapping with name {groupMappingName} #### Role Creation/Deletion -`GET(/druid-ext/basic-security/authorization/db/{authorizerName}/roles)` +`GET(/druid-ext/basic-security/authorization/db/{authorizerName}/roles)`
Return a list of all role names. -`GET(/druid-ext/basic-security/authorization/db/{authorizerName}/roles/{roleName})` +`GET(/druid-ext/basic-security/authorization/db/{authorizerName}/roles/{roleName})`
Return name and permissions for the role named {roleName}. Example output: @@ -427,30 +685,30 @@ Example output: ``` -`POST(/druid-ext/basic-security/authorization/db/{authorizerName}/roles/{roleName})` +`POST(/druid-ext/basic-security/authorization/db/{authorizerName}/roles/{roleName})`
Create a new role with name {roleName}. Content: username string -`DELETE(/druid-ext/basic-security/authorization/db/{authorizerName}/roles/{roleName})` +`DELETE(/druid-ext/basic-security/authorization/db/{authorizerName}/roles/{roleName})`
Delete the role with name {roleName}. #### Role Assignment -`POST(/druid-ext/basic-security/authorization/db/{authorizerName}/users/{userName}/roles/{roleName})` +`POST(/druid-ext/basic-security/authorization/db/{authorizerName}/users/{userName}/roles/{roleName})`
Assign role {roleName} to user {userName}. -`DELETE(/druid-ext/basic-security/authorization/db/{authorizerName}/users/{userName}/roles/{roleName})` +`DELETE(/druid-ext/basic-security/authorization/db/{authorizerName}/users/{userName}/roles/{roleName})`
Unassign role {roleName} from user {userName} -`POST(/druid-ext/basic-security/authorization/db/{authorizerName}/groupMappings/{groupMappingName}/roles/{roleName})` +`POST(/druid-ext/basic-security/authorization/db/{authorizerName}/groupMappings/{groupMappingName}/roles/{roleName})`
Assign role {roleName} to group mapping {groupMappingName}. -`DELETE(/druid-ext/basic-security/authorization/db/{authorizerName}/groupMappings/{groupMappingName}/roles/{roleName})` +`DELETE(/druid-ext/basic-security/authorization/db/{authorizerName}/groupMappings/{groupMappingName}/roles/{roleName})`
Unassign role {roleName} from group mapping {groupMappingName} #### Permissions -`POST(/druid-ext/basic-security/authorization/db/{authorizerName}/roles/{roleName}/permissions)` +`POST(/druid-ext/basic-security/authorization/db/{authorizerName}/roles/{roleName}/permissions)`
Set the permissions of {roleName}. This replaces the previous set of permissions on the role. Content: List of JSON Resource-Action objects, e.g.: @@ -479,5 +737,5 @@ The "name" field for resources in the permission definitions are regexes used to Please see [Defining permissions](../../operations/security-user-auth.md#defining-permissions) for more details. ##### Cache Load Status -`GET(/druid-ext/basic-security/authorization/loadStatus)` +`GET(/druid-ext/basic-security/authorization/loadStatus)`
Return the current load status of the local caches of the authorization Druid metadata store. diff --git a/docs/operations/auth-ldap.md b/docs/operations/auth-ldap.md index b9603a0c38fd..5a2d8b7e292f 100644 --- a/docs/operations/auth-ldap.md +++ b/docs/operations/auth-ldap.md @@ -1,6 +1,7 @@ --- id: auth-ldap -title: "LDAP auth" +title: "Configure LDAP authentication" +sidebar_label: "LDAP auth" --- +You can use [Lightweight Directory Access Protocol (LDAP)](https://en.wikipedia.org/wiki/Lightweight_Directory_Access_Protocol) to secure access to Apache Druid. This topic describes how to set up Druid authentication and authorization with LDAP and LDAP over TLS (LDAPS). The examples on this page show the configuration for an Active Directory LDAP system. -This page describes how to set up Druid user authentication and authorization through LDAP. The first step is to enable LDAP authentication and authorization for Druid. You then map an LDAP group to roles and assign permissions to roles. +The first step is to enable LDAP authentication and authorization for Druid. You then map an LDAP group to Druid roles and assign permissions to those roles. After you've completed this configuration you can optionally choose to enable LDAPS to make LDAP traffic confidential and secure. -## Enable LDAP in Druid +## Prerequisites -Before starting, verify that the active directory is reachable from the Druid Master servers. Command line tools such as `ldapsearch` and `ldapwhoami`, which are included with OpenLDAP, are useful for this testing.  +Before you start to configure LDAP for Druid, test your LDAP connection and perform a sample search. -### Check the connection +### Check your LDAP connection -First test that the basic connection and user credential works. For example, given a user `uuser1@example.com`, try: +Test your LDAP connection to verify it works with user credentials. Later in the process you [configure Druid for LDAP authentication](#configure-druid-for-ldap-authentication) with this user as the `bindUser`. + +The following example command tests the connection for the user `myuser@example.com`. Insert your LDAP server IP address. Modify the port number of your LDAP instance if it listens on a port other than `389`. ```bash -ldapwhoami -vv -H ldap://:389  -D"uuser1@example.com" -W +ldapwhoami -vv -H ldap://ip_address:389 -D "myuser@example.com" -W ``` -Enter the password associated with the user when prompted and verify that the command succeeded. If it didn't, try the following troubleshooting steps: +Enter the password for the user when prompted and verify that the command succeeded. If it failed, check the following: + +- Make sure you're using the correct port for your LDAP instance. +- Check if a network firewall is preventing connections to the LDAP port. +- Review your LDAP implementation details to see whether you need to specifically allow LDAP clients at the LDAP server. If so, add the Druid Coordinator server to the allow list. -* Verify that you've used the correct port for your LDAP instance. By default, the LDAP port is 389, but double-check with your LDAP admin if unable to connect. -* Check whether a network firewall is not preventing connections to the LDAP port. -* Check whether LDAP clients need to be specifically whitelisted at the LDAP server to be able to reach it. If so, add the Druid Coordinator server to the AD whitelist. +### Test your LDAP search +Once your LDAP connection is working, search for a user. For example, the following command searches for the user `myuser` in an Active Directory system. The `sAMAccountName` attribute is specific to Active Directory and contains the authenticated user identity: -### Check the search criteria +```bash +ldapsearch -x -W -H ldap://ip_address:389 -D "cn=admin,dc=example,dc=com" -b "dc=example,dc=com" "(sAMAccountName=myuser)" + +``` -After verifying basic connectivity, check your search criteria. For example, the command for searching for user `uuser1@example.com ` is as follows: +The `memberOf` attribute in the results shows the groups the user belongs to. For example, the following response shows that the user is a member of the `mygroup` group: ```bash -ldapsearch -x -W -H ldap://  -D"uuser1@example.com" -b "dc=example,dc=com" "(sAMAccountName=uuser1)" +memberOf: cn=mygroup,ou=groups,dc=example,dc=com ``` -Note the `memberOf` attribute in the results; it shows the groups that the user belongs to. You will use this value to map the LDAP group to the Druid roles later. This attribute may be implemented differently on different types of LDAP servers. For instance, some LDAP servers may support recursive groupings, and some may not. Some LDAP server implementations may not have any object classes that contain this attribute altogether. If your LDAP server does not use the `memberOf` attribute, then Druid will not be able to determine a user's group membership using LDAP. The sAMAccountName attribute used in this example contains the authenticated user identity. This is an attribute of an object class specific to Microsoft Active Directory. The object classes and attribute used in your LDAP server may be different. +You use this information to map the LDAP group to Druid roles in a later step. + +> Druid uses the `memberOf` attribute to determine a group's membership using LDAP. If your LDAP server implementation doesn't include this attribute, you must complete some additional steps when you [map LDAP groups to Druid roles](#map-ldap-groups-to-druid-roles). + +## Configure Druid for LDAP authentication + +To configure Druid to use LDAP authentication, follow these steps. See [Configuration reference](../configuration/index.md) for the location of the configuration files. + +1. Create a user in your LDAP system that you'll use both for internal communication with Druid and as the LDAP initial admin user. See [Security overview](./security-overview.md) for more information. +In the example below, the LDAP user is `internal@example.com`. -## Configure Druid user authentication with LDAP/Active Directory  +2. Enable the `druid-basic-security` extension in the `common.runtime.properties` file. -1. Enable the `druid-basic-security` extension in the `common.runtime.properties` file. See [Security Overview](security-overview.md) for details. -2. As a best practice, create a user in LDAP to be used for internal communication with Druid. -3. In `common.runtime.properties`, update LDAP-related properties, as shown in the following listing:  - ``` - druid.auth.authenticatorChain=["ldap"] - druid.auth.authenticator.ldap.type=basic - druid.auth.authenticator.ldap.enableCacheNotifications=true - druid.auth.authenticator.ldap.credentialsValidator.type=ldap - druid.auth.authenticator.ldap.credentialsValidator.url=ldap://: - druid.auth.authenticator.ldap.credentialsValidator.bindUser= - druid.auth.authenticator.ldap.credentialsValidator.bindPassword= - druid.auth.authenticator.ldap.credentialsValidator.baseDn= - druid.auth.authenticator.ldap.credentialsValidator.userSearch= - druid.auth.authenticator.ldap.credentialsValidator.userAttribute=sAMAccountName - druid.auth.authenticator.ldap.authorizerName=ldapauth - druid.escalator.type=basic - druid.escalator.internalClientUsername= - druid.escalator.internalClientPassword=Welcome123 - druid.escalator.authorizerName=ldapauth - druid.auth.authorizers=["ldapauth"] - druid.auth.authorizer.ldapauth.type=basic - druid.auth.authorizer.ldapauth.initialAdminUser=AD user who acts as the initial admin user, e.g.: internal@example.com> - druid.auth.authorizer.ldapauth.initialAdminRole=admin - druid.auth.authorizer.ldapauth.roleProvider.type=ldap +3. In the `common.runtime.properties` file, add the following lines for LDAP properties and substitute the values for your own. See [Druid basic security](../development/extensions-core/druid-basic-security.md#properties-for-ldap-user-authentication) for details about these properties. + + ``` + druid.auth.authenticatorChain=["ldap"] + druid.auth.authenticator.ldap.type=basic + druid.auth.authenticator.ldap.enableCacheNotifications=true + druid.auth.authenticator.ldap.credentialsValidator.type=ldap + druid.auth.authenticator.ldap.credentialsValidator.url=ldap://ip_address:port + druid.auth.authenticator.ldap.credentialsValidator.bindUser=administrator@example.com + druid.auth.authenticator.ldap.credentialsValidator.bindPassword=adminpassword + druid.auth.authenticator.ldap.credentialsValidator.baseDn=dc=example,dc=com + druid.auth.authenticator.ldap.credentialsValidator.userSearch=(&(sAMAccountName=%s)(objectClass=user)) + druid.auth.authenticator.ldap.credentialsValidator.userAttribute=sAMAccountName + druid.auth.authenticator.ldap.authorizerName=ldapauth + druid.escalator.type=basic + druid.escalator.internalClientUsername=internal@example.com + druid.escalator.internalClientPassword=internaluserpassword + druid.escalator.authorizerName=ldapauth + druid.auth.authorizers=["ldapauth"] + druid.auth.authorizer.ldapauth.type=basic + druid.auth.authorizer.ldapauth.initialAdminUser=internal@example.com + druid.auth.authorizer.ldapauth.initialAdminRole=admin + druid.auth.authorizer.ldapauth.roleProvider.type=ldap ``` + Note the following: - Notice that the LDAP user created in the previous step, `internal@example.com`, serves as the internal client user and the initial admin user. + - `bindUser`: A user for connecting to LDAP. This should be the same user you used to [test your LDAP search](#test-your-ldap-search). + - `userSearch`: Your LDAP search syntax. + - `userAttribute`: The user search attribute. + - `internal@example.com` is the LDAP user you created in step 1. In the example it serves as both the internal client user and the initial admin user. -## Use LDAP groups to assign roles + > In the above example, the [Druid escalator](../development/extensions-core/druid-basic-security.md#escalator) and LDAP initial admin user are set to the same user - `internal@example.com`. If the escalator is set to a different user, you must follow steps 4 and 5 to create the group mapping and allocate initial roles before the rest of the cluster can function. -You can map LDAP groups to a role in Druid. Members in the group get access to the permissions of the corresponding role.  +4. Save your group mapping to a JSON file. An example file `groupmap.json` looks like this: + + ``` + { + "name": "mygroupmap", + "groupPattern": "CN=mygroup,CN=Users,DC=example,DC=com", + "roles": [ + "readRole" + ] + } + ``` + In the example, the LDAP group `mygroup` maps to Druid role `readRole` and the name of the mapping is `mygroupmap`. +5. Use the Druid API to create the group mapping and allocate initial roles according to your JSON file. The following example uses curl to create the mapping defined in `groupmap.json` for the LDAP group `mygroup`: + + ``` + curl -i -v -H "Content-Type: application/json" -u internal -X POST -d @groupmap.json http://localhost:8081/druid-ext/basic-security/authorization/db/ldapauth/groupMappings/mygroupmap + ``` +6. Check that the group mapping was created successfully. The following example request lists all group mappings: -### Step 1: Create a role + ``` + curl -i -v -H "Content-Type: application/json" -u internal -X GET http://localhost:8081/druid-ext/basic-security/authorization/db/ldapauth/groupMappings + ``` -First create the role in Druid using the Druid REST API. +## Map LDAP groups to Druid roles -Creating a role involves submitting a POST request to the Coordinator process.  +Once you've completed the initial setup and mapping, you can map more LDAP groups to Druid roles. Members of an LDAP group get access to the permissions of the corresponding Druid role. -The following REST APIs to create the role to read access for datasource, config, state. +### Create a Druid role -> As mentioned, the REST API calls need to address the Coordinator node. The examples used below use localhost as the Coordinator host and 8081 as the port. Adjust these settings according to your deployment. +To create a Druid role, you can submit a POST request to the Coordinator process using the Druid REST API or you can use the Druid console. -Call the following API to create role `readRole` .  +The examples below use `localhost` as the Coordinator host and `8081` as the port. Amend these properties according to the details of your deployment. + +Example request to create a role named `readRole`: ``` -curl -i -v  -H "Content-Type: application/json" -u internal -X POST http://localhost:8081/druid-ext/basic-security/authorization/db/ldapauth/roles/readRole +curl -i -v -H "Content-Type: application/json" -u internal -X POST http://localhost:8081/druid-ext/basic-security/authorization/db/ldapauth/roles/readRole ``` -Check that the role has been created successfully by entering the following: +Check that Druid created the role successfully. The following example request lists all roles: ``` -curl -i -v  -H "Content-Type: application/json" -u internal -X GET  http://localhost:8081/druid-ext/basic-security/authorization/db/ldapauth/roles +curl -i -v -H "Content-Type: application/json" -u internal -X GET http://localhost:8081/druid-ext/basic-security/authorization/db/ldapauth/roles ``` +### Add permissions to the Druid role -### Step 2: Add permissions to a role  - -You can now add one or more permission to the role. The following example adds read-only access to a `wikipedia` data source.   +Once you have a Druid role you can add permissions to it. The following example adds read-only access to a `wikipedia` data source. Given the following JSON in a file named `perm.json`: ``` -[{ "resource": { "name": "wikipedia", "type": "DATASOURCE" }, "action": "READ" } -,{ "resource": { "name": ".*", "type": "STATE" }, "action": "READ" }, -{ "resource": {"name": ".*", "type": "CONFIG"}, "action": "READ"}] -``` - -The following command associates the permissions in the JSON file with the role - -``` -curl -i -v  -H "Content-Type: application/json" -u internal -X POST -d@perm.json  http://localhost:8081/druid-ext/basic-security/authorization/db/ldapauth/roles/readRole/permissions +[ + { "resource": { "name": "wikipedia", "type": "DATASOURCE" }, "action": "READ" }, + { "resource": { "name": ".*", "type": "STATE" }, "action": "READ" }, + { "resource": {"name": ".*", "type": "CONFIG"}, "action": "READ"} +] ``` -Note that the STATE and CONFIG permissions in `perm.json` are needed to see the data source in the web console. If only querying permissions are needed, the READ action is sufficient: +The following request associates the permissions in the JSON file with the `readRole` role: ``` -[{ "resource": { "name": "wikipedia", "type": "DATASOURCE" }, "action": "READ" }] +curl -i -v -H "Content-Type: application/json" -u internal -X POST -d@perm.json http://localhost:8081/druid-ext/basic-security/authorization/db/ldapauth/roles/readRole/permissions ``` -You can also provide the name in the form of regular expression. For example, to give access to all data sources starting with `wiki`, specify the name as  `{ "name": "wiki.*", .....`. +Druid users need the `STATE` and `CONFIG` permissions to view the data source in the Druid console. If you only want to assign querying permissions you can apply just the `READ` permission with the first line in the `perm.json` file. +You can also provide the data source name in the form of a regular expression. For example, to give access to all data sources starting with `wiki`, you would specify the data source name as `{ "name": "wiki.*" }` . -### Step 3: Create group Mapping  +### Create the group mapping -The following shows an example of a group to role mapping. It assumes that a group named `group1` exists in the directory. Also assuming the following role mapping in a file named `groupmap.json`: +You can now map an LDAP group to the Druid role. The following example request creates a mapping with name `mygroupmap`. It assumes that a group named `mygroup` exists in the directory. ``` { -    "name": "group1map", -    "groupPattern": "CN=group1,CN=Users,DC=example,DC=com", -    "roles": [ -        "readRole" -    ] + "name": "mygroupmap", + "groupPattern": "CN=mygroup,CN=Users,DC=example,DC=com", + "roles": [ + "readRole" + ] } ``` -You can configure the mapping as follows: +The following example request configures the mapping—the role mapping is in the file `groupmap.json`. See [Configure Druid for LDAP authentication](#configure-druid-for-ldap-authentication) for the contents of an example file. ``` -curl -i -v  -H "Content-Type: application/json" -u internal -X POST -d @groupmap.json http://localhost:8081/druid-ext/basic-security/authorization/db/ldapauth/groupMappings/group1map +curl -i -v -H "Content-Type: application/json" -u internal -X POST -d @groupmap.json http://localhost:8081/druid-ext/basic-security/authorization/db/ldapauth/groupMappings/mygroupmap ``` -To check whether the group mapping was created successfully, run the following command: +To check whether the group mapping was created successfully, the following request lists all group mappings: ``` -curl -i -v  -H "Content-Type: application/json" -u internal -X GET http://localhost:8081/druid-ext/basic-security/authorization/db/ldapauth/groupMappings +curl -i -v -H "Content-Type: application/json" -u internal -X GET http://localhost:8081/druid-ext/basic-security/authorization/db/ldapauth/groupMappings ``` -To check the details of a specific group mapping, use the following: +The following example request returns the details of the `mygroupmap` group: ``` -curl -i -v  -H "Content-Type: application/json" -u internal -X GET http://localhost:8081/druid-ext/basic-security/authorization/db/ldapauth/groupMappings/group1map +curl -i -v -H "Content-Type: application/json" -u internal -X GET http://localhost:8081/druid-ext/basic-security/authorization/db/ldapauth/groupMappings/mygroupmap ``` -To add additional roles to the group mapping, use the following API: +The following example request adds the role `queryRole` to the `mygroupmap` mapping: ``` -curl -i -v  -H "Content-Type: application/json" -u internal -X POST http://localhost:8081/druid-ext/basic-security/authorization/db/ldapauth/groupMappings/group1/roles/  +curl -i -v -H "Content-Type: application/json" -u internal -X POST http://localhost:8081/druid-ext/basic-security/authorization/db/ldapauth/groupMappings/mygroup/roles/queryrole ``` -In the next two steps you will be creating a user, and assigning previously created roles to it. These steps are only needed in the following cases: - - - Your LDAP server does not support the `memberOf` attribute, or - - You want to configure a user with additional roles that are not mapped to the group(s) that the user is a member of - - If this is not the case for your scenario, you can skip these steps. +### Add an LDAP user to Druid and assign a role -### Step 4. Create a user +You only need to complete this step if: +- Your LDAP user doesn't belong to any of your LDAP groups, or +- You want to configure a user with additional Druid roles that are not mapped to the LDAP groups that the user belongs to. -Once LDAP is enabled, only user passwords are verified with LDAP. You add the LDAP user to Druid as follows:  +Example request to add the LDAP user `myuser` to Druid: ``` -curl -i -v  -H "Content-Type: application/json" -u internal -X POST http://localhost:8081/druid-ext/basic-security/authentication/db/ldap/users/  +curl -i -v -H "Content-Type: application/json" -u internal -X POST http://localhost:8081/druid-ext/basic-security/authentication/db/ldap/users/myuser ``` -### Step 5. Assign the role to the user - -The following command shows how to assign a role to a user: +Example request to assign the `myuser` user to the `queryRole` role: ``` -curl -i -v  -H "Content-Type: application/json" -u internal -X POST http://localhost:8081/druid-ext/basic-security/authorization/db/ldapauth/users//roles/  +curl -i -v -H "Content-Type: application/json" -u internal -X POST http://localhost:8081/druid-ext/basic-security/authentication/db/ldap/users/myuser/roles/queryRole ``` -For more information about security and the basic security extension, see [Security Overview](security-overview.md). +## Enable LDAP over TLS (LDAPS) + +Once you've configured LDAP authentication in Druid, you can optionally make LDAP traffic confidential and secure by using Transport Layer Security (TLS)—previously Secure Socket Layer(SSL)—technology. + +Configuring LDAPS establishes trust between Druid and the LDAP server. + +## Prerequisites + +Before you start to set up LDAPS in Druid, you must [configure Druid for LDAP authentication](#configure-druid-for-ldap-authentication). You also need: + +- A certificate issued by a public certificate authority (CA) or a self-signed certificate by an internal CA. +- The root certificate for the CA that signed the certificate for the LDAP server. If you're using a common public CA, the certificate may already be in the Java truststore. Otherwise you need to import the certificate for the CA. + +## Configure Druid for LDAPS + +Complete the following steps to set up LDAPS for Druid. See [Configuration reference](../configuration/index.md) for the location of the configuration files. + +1. Import the CA certificate for your LDAP server or a self-signed certificate into the truststore location saved as `druid.client.https.trustStorePath` in your `common.runtime.properties` file. + + ``` + keytool -import -trustcacerts -keystore path/to/cacerts -storepass truststorepassword -alias aliasName -file path/to/certificate.cer + ``` + + Replace `path/to/cacerts` with the path to your truststore, `truststorepassword` with your truststore password, `aliasName` with an alias name for the keystore, and `path/to/certificate.cer` with the location and name of your certificate. For example: + + ``` + keytool -import -trustcacerts -keystore /Library/Java/JavaVirtualMachines/adoptopenjdk-8.jdk/Contents/Home/jre/lib/security/cacerts -storepass mypassword -alias myAlias -file /etc/ssl/certs/my-certificate.cer + ``` + +2. If the root certificate for the CA isn't already in the Java truststore, import it: + + ``` + keytool -importcert -keystore path/to/cacerts -storepass truststorepassword -alias aliasName -file path/to/certificate.cer + ``` + + Replace `path/to/cacerts` with the path to your truststore, `truststorepassword` with your truststore password, `aliasName` with an alias name for the keystore, and `path/to/certificate.cer` with the location and name of your certificate. For example: + + ``` + keytool -importcert -keystore /Library/Java/JavaVirtualMachines/adoptopenjdk-8.jdk/Contents/Home/jre/lib/security/cacerts -storepass mypassword -alias myAlias -file /etc/ssl/certs/my-certificate.cer + ``` + +3. In your `common.runtime.properties` file, add the following lines to the LDAP configuration section, substituting your own truststore path and password: + + ``` + druid.auth.basic.ssl.trustStorePath=/Library/Java/JavaVirtualMachines/adoptopenjdk-8.jdk/Contents/Home/jre/lib/security/cacerts + druid.auth.basic.ssl.protocol=TLS + druid.auth.basic.ssl.trustStorePassword=xxxxxx + ``` + + See [Druid basic security](../development/extensions-core/druid-basic-security.md#properties-for-ldaps) for details about these properties. + +4. You can optionally configure additional LDAPS properties in the `common.runtime.properties` file. See [Druid basic security](../development/extensions-core/druid-basic-security.md#properties-for-ldaps) for more information. + +5. Restart Druid. + + +## Troubleshooting tips + +The following are some ideas to help you troubleshoot issues with LDAP and LDAPS. + +### Check the coordinator logs + +If your LDAP connection isn't working, check the coordinator logs. See [Logging](../configuration/logging.md) for details. + +### Check the Druid escalator configuration + +If the coordinator is working but the rest of the cluster isn't, check the escalator configuration. See the [Configuration reference](../configuration/index.md) for details. You can also check other service logs to see why the services are unable to fetch authorization details from the coordinator. + +### Check your LDAP server response time + +If a user can log in to the Druid console but the landing page shows a 401 error, check your LDAP server response time. In a large organization with a high number of LDAP users, LDAP may be slow to respond, and this can result in a connection timeout. diff --git a/docs/operations/security-overview.md b/docs/operations/security-overview.md index 028cff04f55a..3fa80e24adc2 100644 --- a/docs/operations/security-overview.md +++ b/docs/operations/security-overview.md @@ -250,80 +250,7 @@ The following steps walk through a sample setup procedure: ## Configuring an LDAP authenticator -As an alternative to using the basic metadata authenticator, you can use LDAP to authenticate users. The following steps provide an overview of the setup procedure. For more information on these settings, see [Properties for LDAP user authentication](../development/extensions-core/druid-basic-security.md#properties-for-ldap-user-authentication). - -1. In `common.runtime.properties`, add LDAP to the authenticator chain in the order in which you want requests to be evaluated. For example: - ``` - # Druid basic security - druid.auth.authenticatorChain=["ldap", "MyBasicMetadataAuthenticator"] - ``` - -2. Configure LDAP settings in `common.runtime.properties` as appropriate for your LDAP scheme and system. For example: - ``` - druid.auth.authenticator.ldap.type=basic - druid.auth.authenticator.ldap.enableCacheNotifications=true - druid.auth.authenticator.ldap.credentialsValidator.type=ldap - druid.auth.authenticator.ldap.credentialsValidator.url=ldap://ad_host:389 - druid.auth.authenticator.ldap.credentialsValidator.bindUser=ad_admin_user - druid.auth.authenticator.ldap.credentialsValidator.bindPassword=ad_admin_password - druid.auth.authenticator.ldap.credentialsValidator.baseDn=dc=example,dc=com - druid.auth.authenticator.ldap.credentialsValidator.userSearch=(&(sAMAccountName=%s)(objectClass=user)) - druid.auth.authenticator.ldap.credentialsValidator.userAttribute=sAMAccountName - druid.auth.authenticator.ldap.authorizerName=ldapauth - druid.escalator.type=basic - druid.escalator.internalClientUsername=ad_interal_user - druid.escalator.internalClientPassword=Welcome123 - druid.escalator.authorizerName=ldapauth - druid.auth.authorizers=["ldapauth"] - druid.auth.authorizer.ldapauth.type=basic - druid.auth.authorizer.ldapauth.initialAdminUser= - druid.auth.authorizer.ldapauth.initialAdminRole=admin - druid.auth.authorizer.ldapauth.roleProvider.type=ldap - ``` - -3. Use the Druid API to create the group mapping and allocate initial roles. For example, using curl and given a group named `group1` in the directory, run: - ``` - curl -i -v -H "Content-Type: application/json" -u internal -X POST -d @groupmap.json http://localhost:8081/druid-ext/basic-security/authorization/db/ldapauth/groupMappings/group1map - ``` - The `groupmap.json` file contents would be something like: - ``` - { - "name": "group1map", - "groupPattern": "CN=group1,CN=Users,DC=example,DC=com", - "roles": [ - "readRole" - ] - } - ``` -4. Check if the group mapping is created successfully by executing the following API. This lists all group mappings. - ``` - curl -i -v -H "Content-Type: application/json" -u internal -X GET http://localhost:8081/druid-ext/basic-security/authorization/db/ldapauth/groupMappings - ``` - - Alternatively, to check the details of a specific group mapping, use the following API: - ``` - curl -i -v -H "Content-Type: application/json" -u internal -X GET http://localhost:8081/druid-ext/basic-security/authorization/db/ldapauth/groupMappings/group1map - ``` - -5. To add additional roles to the group mapping, use the following API: - ``` - curl -i -v -H "Content-Type: application/json" -u internal -X POST http://localhost:8081/druid-ext/basic-security/authorization/db/ldapauth/groupMappings/group1/roles/ - ``` - -6. Add the LDAP user to Druid. To add a user, use the following authentication API: - ``` - curl -i -v -H "Content-Type: application/json" -u internal -X POST http://localhost:8081/druid-ext/basic-security/authentication/db/ldap/users/ - ``` - -7. Use the following command to assign the role to a user: - ``` - curl -i -v -H "Content-Type: application/json" -u internal -X POST http://localhost:8081/druid-ext/basic-security/authorization/db/ldapauth/users//roles/ - ``` - - - -Congratulations, you have configured permissions for user-assigned roles in Druid! - +As an alternative to using the basic metadata authenticator, you can use LDAP to authenticate users. See [Configure LDAP authentication](./auth-ldap.md) for information on configuring Druid for LDAP and LDAPS. ## Druid security trust model Within Druid's trust model there users can have different authorization levels: diff --git a/website/.spelling b/website/.spelling index c2eb66104ee8..d02f4f846799 100644 --- a/website/.spelling +++ b/website/.spelling @@ -129,6 +129,7 @@ JDK JDK7 JDK8 JKS +jks JMX JRE JS @@ -147,6 +148,7 @@ Kerberos KeyStores Kinesis Kubernetes +LDAPS LRU LZ4 LZO @@ -203,6 +205,7 @@ SYSTEM_TABLE TCP TGT TLS +tls TopN TopNs UI From 50963edcae70150f13520b619f167512d951a71b Mon Sep 17 00:00:00 2001 From: Gian Merlino Date: Tue, 29 Nov 2022 15:51:03 -0800 Subject: [PATCH 063/141] Fix compile error in MSQSelectTest. (#13456) --- .../src/test/java/org/apache/druid/msq/exec/MSQSelectTest.java | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/MSQSelectTest.java b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/MSQSelectTest.java index dd9bf05d2094..b6305e13542b 100644 --- a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/MSQSelectTest.java +++ b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/MSQSelectTest.java @@ -1258,7 +1258,7 @@ public void testGroupByOnFooWithDurableStoragePathAssertions() throws IOExceptio @Test public void testMultiValueStringWithIncorrectType() throws IOException { - final File toRead = getResourceAsTemporaryFile("/unparseable-mv-string-array.json"); + final File toRead = MSQTestFileUtils.getResourceAsTemporaryFile(this, "/unparseable-mv-string-array.json"); final String toReadAsJson = queryFramework().queryJsonMapper().writeValueAsString(toRead.getAbsolutePath()); RowSignature rowSignature = RowSignature.builder() From 291ded22d5510148b39b113e7d0a492ba905d787 Mon Sep 17 00:00:00 2001 From: Jill Osborne Date: Wed, 30 Nov 2022 10:44:43 +0000 Subject: [PATCH 064/141] Update experimental features doc (#13452) --- docs/development/experimental-features.md | 10 +--------- 1 file changed, 1 insertion(+), 9 deletions(-) diff --git a/docs/development/experimental-features.md b/docs/development/experimental-features.md index a68fc47411af..30d8c2f77c2b 100644 --- a/docs/development/experimental-features.md +++ b/docs/development/experimental-features.md @@ -26,7 +26,7 @@ The following features are marked [experimental](./experimental.md) in the Druid This document includes each page that mentions an experimental feature. To graduate a feature, remove all mentions of its experimental status on all relevant pages. -Note that this document does not track the status of contrib extensions, some of which are experimental. +Note that this document does not track the status of contrib extensions, all of which are considered experimental. ## SQL-based ingestion @@ -34,10 +34,6 @@ Note that this document does not track the status of contrib extensions, some of - [SQL-based ingestion concepts](../multi-stage-query/concepts.md) - [SQL-based ingestion and multi-stage query task API](../multi-stage-query/api.md) -## Nested columns - -- [Nested columns](../querying/nested-columns.md) - ## Indexer process - [Indexer process](../design/indexer.md) @@ -53,10 +49,6 @@ Note that this document does not track the status of contrib extensions, some of - [Task reference](../ingestion/tasks.md#locking) - [Design](../design/architecture.md#availability-and-consistency) -## Moments sketch - -- [Aggregations](../querying/aggregations.md#moments-sketch-experimental) - ## Front coding - [Ingestion spec reference](../ingestion/ingestion-spec.md#front-coding) From 8ff1b2d5d41752ade0f915d4dac6f5d9b6f45542 Mon Sep 17 00:00:00 2001 From: Kashif Faraz Date: Wed, 30 Nov 2022 16:33:05 +0530 Subject: [PATCH 065/141] Revert "Add filter in cloud object input source for backward compatibility (#13437)" (#13450) This reverts commit b12e5f300e7c2795ba3d9c7ef17fb64f4925b9c0. --- .../input/impl/CloudObjectInputSource.java | 32 +---------- .../impl/CloudObjectInputSourceTest.java | 45 ++++----------- .../data/input/aliyun/OssInputSource.java | 4 +- .../data/input/aliyun/OssInputSourceTest.java | 17 ------ .../data/input/azure/AzureInputSource.java | 4 +- .../input/azure/AzureInputSourceTest.java | 55 ------------------- .../google/GoogleCloudStorageInputSource.java | 5 +- .../GoogleCloudStorageInputSourceTest.java | 36 ++---------- .../druid/data/input/s3/S3InputSource.java | 6 +- 9 files changed, 23 insertions(+), 181 deletions(-) diff --git a/core/src/main/java/org/apache/druid/data/input/impl/CloudObjectInputSource.java b/core/src/main/java/org/apache/druid/data/input/impl/CloudObjectInputSource.java index 716e5b611ceb..3c3e3c6b72d4 100644 --- a/core/src/main/java/org/apache/druid/data/input/impl/CloudObjectInputSource.java +++ b/core/src/main/java/org/apache/druid/data/input/impl/CloudObjectInputSource.java @@ -21,9 +21,7 @@ import com.fasterxml.jackson.annotation.JsonInclude; import com.fasterxml.jackson.annotation.JsonProperty; -import com.google.common.base.Preconditions; import com.google.common.primitives.Ints; -import org.apache.commons.io.FilenameUtils; import org.apache.commons.lang.StringUtils; import org.apache.druid.data.input.AbstractInputSource; import org.apache.druid.data.input.InputEntity; @@ -52,13 +50,6 @@ public abstract class CloudObjectInputSource extends AbstractInputSource private final List uris; private final List prefixes; private final List objects; - - /** - * Preserved filter for backward compatibility, should be removed on next major release; - * use objectGlob instead. - */ - @Deprecated - private final String filter; private final String objectGlob; public CloudObjectInputSource( @@ -72,7 +63,6 @@ public CloudObjectInputSource( this.uris = uris; this.prefixes = prefixes; this.objects = objects; - this.filter = null; this.objectGlob = null; illegalArgsChecker(); @@ -83,7 +73,6 @@ public CloudObjectInputSource( @Nullable List uris, @Nullable List prefixes, @Nullable List objects, - @Deprecated @Nullable String filter, @Nullable String objectGlob ) { @@ -91,12 +80,8 @@ public CloudObjectInputSource( this.uris = uris; this.prefixes = prefixes; this.objects = objects; - this.filter = filter; this.objectGlob = objectGlob; - Preconditions.checkArgument( - filter == null || objectGlob == null, - "Cannot use filter and objectGlob together. Try using objectGlob instead of filter." - ); + illegalArgsChecker(); } @@ -122,14 +107,6 @@ public List getObjects() return objects; } - @Nullable - @JsonProperty - @JsonInclude(JsonInclude.Include.NON_NULL) - public String getFilter() - { - return filter; - } - @Nullable @JsonProperty @JsonInclude(JsonInclude.Include.NON_NULL) @@ -167,8 +144,6 @@ public Stream>> createSplits( if (StringUtils.isNotBlank(objectGlob)) { PathMatcher m = FileSystems.getDefault().getPathMatcher("glob:" + getObjectGlob()); objectStream = objectStream.filter(object -> m.matches(Paths.get(object.getPath()))); - } else if (StringUtils.isNotBlank(filter)) { - objectStream = objectStream.filter(object -> FilenameUtils.wildcardMatch(object.getPath(), getFilter())); } return objectStream.map(object -> new InputSplit<>(Collections.singletonList(object))); @@ -180,8 +155,6 @@ public Stream>> createSplits( if (StringUtils.isNotBlank(objectGlob)) { PathMatcher m = FileSystems.getDefault().getPathMatcher("glob:" + getObjectGlob()); uriStream = uriStream.filter(uri -> m.matches(Paths.get(uri.toString()))); - } else if (StringUtils.isNotBlank(filter)) { - uriStream = uriStream.filter(uri -> FilenameUtils.wildcardMatch(uri.toString(), filter)); } return uriStream.map(CloudObjectLocation::new).map(object -> new InputSplit<>(Collections.singletonList(object))); @@ -239,14 +212,13 @@ public boolean equals(Object o) Objects.equals(uris, that.uris) && Objects.equals(prefixes, that.prefixes) && Objects.equals(objects, that.objects) && - Objects.equals(filter, that.filter) && Objects.equals(objectGlob, that.objectGlob); } @Override public int hashCode() { - return Objects.hash(scheme, uris, prefixes, objects, filter, objectGlob); + return Objects.hash(scheme, uris, prefixes, objects, objectGlob); } private void illegalArgsChecker() throws IllegalArgumentException diff --git a/core/src/test/java/org/apache/druid/data/input/impl/CloudObjectInputSourceTest.java b/core/src/test/java/org/apache/druid/data/input/impl/CloudObjectInputSourceTest.java index da421b1b1676..539581403989 100644 --- a/core/src/test/java/org/apache/druid/data/input/impl/CloudObjectInputSourceTest.java +++ b/core/src/test/java/org/apache/druid/data/input/impl/CloudObjectInputSourceTest.java @@ -72,7 +72,7 @@ public class CloudObjectInputSourceTest public void testGetUris() { CloudObjectInputSource inputSource = Mockito.mock(CloudObjectInputSource.class, Mockito.withSettings() - .useConstructor(SCHEME, URIS, null, null, null, null) + .useConstructor(SCHEME, URIS, null, null, null) .defaultAnswer(Mockito.CALLS_REAL_METHODS) ); @@ -86,7 +86,7 @@ public void testGetUris() public void testGetPrefixes() { CloudObjectInputSource inputSource = Mockito.mock(CloudObjectInputSource.class, Mockito.withSettings() - .useConstructor(SCHEME, null, PREFIXES, null, null, null) + .useConstructor(SCHEME, null, PREFIXES, null, null) .defaultAnswer(Mockito.CALLS_REAL_METHODS) ); @@ -100,7 +100,7 @@ public void testGetPrefixes() public void testGetObjectGlob() { CloudObjectInputSource inputSource = Mockito.mock(CloudObjectInputSource.class, Mockito.withSettings() - .useConstructor(SCHEME, URIS, null, null, null, "**.parquet") + .useConstructor(SCHEME, URIS, null, null, "**.parquet") .defaultAnswer(Mockito.CALLS_REAL_METHODS) ); @@ -111,12 +111,12 @@ public void testGetObjectGlob() public void testInequality() { CloudObjectInputSource inputSource1 = Mockito.mock(CloudObjectInputSource.class, Mockito.withSettings() - .useConstructor(SCHEME, URIS, null, null, null, "**.parquet") + .useConstructor(SCHEME, URIS, null, null, "**.parquet") .defaultAnswer(Mockito.CALLS_REAL_METHODS) ); CloudObjectInputSource inputSource2 = Mockito.mock(CloudObjectInputSource.class, Mockito.withSettings() - .useConstructor(SCHEME, URIS, null, null, null, "**.csv") + .useConstructor(SCHEME, URIS, null, null, "**.csv") .defaultAnswer(Mockito.CALLS_REAL_METHODS) ); @@ -128,32 +128,9 @@ public void testInequality() @Test public void testWithUrisFilter() { - CloudObjectInputSource inputSource = Mockito.mock(CloudObjectInputSource.class, - Mockito.withSettings() - .useConstructor(SCHEME, URIS2, null, null, "*.csv", null) - .defaultAnswer(Mockito.CALLS_REAL_METHODS) - ); - - Stream>> splits = inputSource.createSplits( - new JsonInputFormat(JSONPathSpec.DEFAULT, null, null, null, null), - new MaxSizeSplitHintSpec(null, 1) - ); - - List returnedLocations = splits.map(InputSplit::get).collect(Collectors.toList()).get(0); - - List returnedLocationUris = returnedLocations.stream().map(object -> object.toUri(SCHEME)).collect(Collectors.toList()); - - Assert.assertEquals("*.csv", inputSource.getFilter()); - Assert.assertEquals(URIS, returnedLocationUris); - } - - @Test - public void testWithUrisObjectGlob() - { - CloudObjectInputSource inputSource = Mockito.mock(CloudObjectInputSource.class, - Mockito.withSettings() - .useConstructor(SCHEME, URIS2, null, null, null, "**.csv") - .defaultAnswer(Mockito.CALLS_REAL_METHODS) + CloudObjectInputSource inputSource = Mockito.mock(CloudObjectInputSource.class, Mockito.withSettings() + .useConstructor(SCHEME, URIS2, null, null, "**.csv") + .defaultAnswer(Mockito.CALLS_REAL_METHODS) ); Stream>> splits = inputSource.createSplits( @@ -173,7 +150,7 @@ public void testWithUrisObjectGlob() public void testWithUris() { CloudObjectInputSource inputSource = Mockito.mock(CloudObjectInputSource.class, Mockito.withSettings() - .useConstructor(SCHEME, URIS, null, null, null, null) + .useConstructor(SCHEME, URIS, null, null, null) .defaultAnswer(Mockito.CALLS_REAL_METHODS) ); @@ -194,7 +171,7 @@ public void testWithUris() public void testWithObjectsFilter() { CloudObjectInputSource inputSource = Mockito.mock(CloudObjectInputSource.class, Mockito.withSettings() - .useConstructor(SCHEME, null, null, OBJECTS_BEFORE_GLOB, null, "**.csv") + .useConstructor(SCHEME, null, null, OBJECTS_BEFORE_GLOB, "**.csv") .defaultAnswer(Mockito.CALLS_REAL_METHODS) ); @@ -215,7 +192,7 @@ public void testWithObjectsFilter() public void testWithObjects() { CloudObjectInputSource inputSource = Mockito.mock(CloudObjectInputSource.class, Mockito.withSettings() - .useConstructor(SCHEME, null, null, OBJECTS, null, null) + .useConstructor(SCHEME, null, null, OBJECTS, null) .defaultAnswer(Mockito.CALLS_REAL_METHODS) ); diff --git a/extensions-contrib/aliyun-oss-extensions/src/main/java/org/apache/druid/data/input/aliyun/OssInputSource.java b/extensions-contrib/aliyun-oss-extensions/src/main/java/org/apache/druid/data/input/aliyun/OssInputSource.java index a90e96f047b8..71edd26b590b 100644 --- a/extensions-contrib/aliyun-oss-extensions/src/main/java/org/apache/druid/data/input/aliyun/OssInputSource.java +++ b/extensions-contrib/aliyun-oss-extensions/src/main/java/org/apache/druid/data/input/aliyun/OssInputSource.java @@ -79,12 +79,11 @@ public OssInputSource( @JsonProperty("uris") @Nullable List uris, @JsonProperty("prefixes") @Nullable List prefixes, @JsonProperty("objects") @Nullable List objects, - @Deprecated @JsonProperty("filter") @Nullable String filter, @JsonProperty("objectGlob") @Nullable String objectGlob, @JsonProperty("properties") @Nullable OssClientConfig inputSourceConfig ) { - super(OssStorageDruidModule.SCHEME, uris, prefixes, objects, filter, objectGlob); + super(OssStorageDruidModule.SCHEME, uris, prefixes, objects, objectGlob); this.inputDataConfig = Preconditions.checkNotNull(inputDataConfig, "inputDataConfig"); Preconditions.checkNotNull(client, "client"); this.inputSourceConfig = inputSourceConfig; @@ -137,7 +136,6 @@ public SplittableInputSource> withSplit(InputSplit uris, @JsonProperty("prefixes") @Nullable List prefixes, @JsonProperty("objects") @Nullable List objects, - @Deprecated @JsonProperty("filter") @Nullable String filter, @JsonProperty("objectGlob") @Nullable String objectGlob ) { - super(SCHEME, uris, prefixes, objects, filter, objectGlob); + super(SCHEME, uris, prefixes, objects, objectGlob); this.storage = Preconditions.checkNotNull(storage, "AzureStorage"); this.entityFactory = Preconditions.checkNotNull(entityFactory, "AzureEntityFactory"); this.azureCloudBlobIterableFactory = Preconditions.checkNotNull( @@ -104,7 +103,6 @@ public SplittableInputSource> withSplit(InputSplit>> cloudObjectStream = azureInputSource.getPrefixesSplitStream( - new MaxSizeSplitHintSpec(null, 1) - ); - - List> actualCloudLocationList = cloudObjectStream.map(InputSplit::get) - .collect(Collectors.toList()); - verifyAll(); - Assert.assertEquals(expectedCloudLocations, actualCloudLocationList); - } - - @Test - public void test_getPrefixesSplitStream_withFilter_successfullyCreatesCloudLocation_returnsExpectedLocations() - { - List prefixes = ImmutableList.of(PREFIX_URI); - List> expectedCloudLocations = ImmutableList.of(ImmutableList.of(CLOUD_OBJECT_LOCATION_1)); - List expectedCloudBlobs = ImmutableList.of(cloudBlobDruid1); - Iterator expectedCloudBlobsIterator = expectedCloudBlobs.iterator(); - String filter = "*.csv"; - - expectedCloudBlobsIterator = Iterators.filter( - expectedCloudBlobsIterator, - object -> FilenameUtils.wildcardMatch(object.getName(), filter) - ); - - EasyMock.expect(inputDataConfig.getMaxListingLength()).andReturn(MAX_LISTING_LENGTH); - EasyMock.expect(azureCloudBlobIterableFactory.create(prefixes, MAX_LISTING_LENGTH)).andReturn( - azureCloudBlobIterable); - EasyMock.expect(azureCloudBlobIterable.iterator()).andReturn(expectedCloudBlobsIterator); - EasyMock.expect(azureCloudBlobToLocationConverter.createCloudObjectLocation(cloudBlobDruid1)) - .andReturn(CLOUD_OBJECT_LOCATION_1); - EasyMock.expect(cloudBlobDruid1.getBlobLength()).andReturn(100L).anyTimes(); - EasyMock.expect(cloudBlobDruid1.getName()).andReturn(BLOB_PATH).anyTimes(); - - replayAll(); - - azureInputSource = new AzureInputSource( - storage, - entityFactory, - azureCloudBlobIterableFactory, - azureCloudBlobToLocationConverter, - inputDataConfig, - EMPTY_URIS, - prefixes, - EMPTY_OBJECTS, - filter, null ); @@ -267,7 +215,6 @@ public void test_getPrefixesSplitStream_withObjectGlob_successfullyCreatesCloudL EMPTY_URIS, prefixes, EMPTY_OBJECTS, - null, objectGlob ); @@ -297,7 +244,6 @@ public void test_withSplit_constructsExpectedInputSource() EMPTY_URIS, prefixes, EMPTY_OBJECTS, - null, null ); @@ -319,7 +265,6 @@ public void test_toString_returnsExpectedString() EMPTY_URIS, prefixes, EMPTY_OBJECTS, - null, null ); diff --git a/extensions-core/google-extensions/src/main/java/org/apache/druid/data/input/google/GoogleCloudStorageInputSource.java b/extensions-core/google-extensions/src/main/java/org/apache/druid/data/input/google/GoogleCloudStorageInputSource.java index 009bae7dbe2e..a4d175aab2c0 100644 --- a/extensions-core/google-extensions/src/main/java/org/apache/druid/data/input/google/GoogleCloudStorageInputSource.java +++ b/extensions-core/google-extensions/src/main/java/org/apache/druid/data/input/google/GoogleCloudStorageInputSource.java @@ -65,11 +65,10 @@ public GoogleCloudStorageInputSource( @JsonProperty("uris") @Nullable List uris, @JsonProperty("prefixes") @Nullable List prefixes, @JsonProperty("objects") @Nullable List objects, - @Deprecated @JsonProperty("filter") @Nullable String filter, @JsonProperty("objectGlob") @Nullable String objectGlob ) { - super(GoogleStorageDruidModule.SCHEME_GS, uris, prefixes, objects, filter, objectGlob); + super(GoogleStorageDruidModule.SCHEME_GS, uris, prefixes, objects, objectGlob); this.storage = storage; this.inputDataConfig = inputDataConfig; } @@ -118,7 +117,7 @@ protected Stream>> getPrefixesSplitStream(@ @Override public SplittableInputSource> withSplit(InputSplit> split) { - return new GoogleCloudStorageInputSource(storage, inputDataConfig, null, null, split.get(), getFilter(), getObjectGlob()); + return new GoogleCloudStorageInputSource(storage, inputDataConfig, null, null, split.get(), getObjectGlob()); } private CloudObjectLocation byteSourceFromStorageObject(final StorageObject storageObject) diff --git a/extensions-core/google-extensions/src/test/java/org/apache/druid/data/input/google/GoogleCloudStorageInputSourceTest.java b/extensions-core/google-extensions/src/test/java/org/apache/druid/data/input/google/GoogleCloudStorageInputSourceTest.java index f15b6280cc85..c0648e7f4c85 100644 --- a/extensions-core/google-extensions/src/test/java/org/apache/druid/data/input/google/GoogleCloudStorageInputSourceTest.java +++ b/extensions-core/google-extensions/src/test/java/org/apache/druid/data/input/google/GoogleCloudStorageInputSourceTest.java @@ -116,7 +116,7 @@ public void testSerde() throws Exception { final ObjectMapper mapper = createGoogleObjectMapper(); final GoogleCloudStorageInputSource withUris = - new GoogleCloudStorageInputSource(STORAGE, INPUT_DATA_CONFIG, EXPECTED_URIS, ImmutableList.of(), null, null, null); + new GoogleCloudStorageInputSource(STORAGE, INPUT_DATA_CONFIG, EXPECTED_URIS, ImmutableList.of(), null, null); final GoogleCloudStorageInputSource serdeWithUris = mapper.readValue(mapper.writeValueAsString(withUris), GoogleCloudStorageInputSource.class); Assert.assertEquals(withUris, serdeWithUris); @@ -127,7 +127,7 @@ public void testSerdePrefixes() throws Exception { final ObjectMapper mapper = createGoogleObjectMapper(); final GoogleCloudStorageInputSource withPrefixes = - new GoogleCloudStorageInputSource(STORAGE, INPUT_DATA_CONFIG, ImmutableList.of(), PREFIXES, null, null, null); + new GoogleCloudStorageInputSource(STORAGE, INPUT_DATA_CONFIG, ImmutableList.of(), PREFIXES, null, null); final GoogleCloudStorageInputSource serdeWithPrefixes = mapper.readValue(mapper.writeValueAsString(withPrefixes), GoogleCloudStorageInputSource.class); Assert.assertEquals(withPrefixes, serdeWithPrefixes); @@ -144,7 +144,6 @@ public void testSerdeObjects() throws Exception null, null, ImmutableList.of(new CloudObjectLocation("foo", "bar/file.gz")), - null, null ); final GoogleCloudStorageInputSource serdeWithObjects = @@ -157,7 +156,7 @@ public void testWithUrisSplit() { GoogleCloudStorageInputSource inputSource = - new GoogleCloudStorageInputSource(STORAGE, INPUT_DATA_CONFIG, EXPECTED_URIS, ImmutableList.of(), null, null, null); + new GoogleCloudStorageInputSource(STORAGE, INPUT_DATA_CONFIG, EXPECTED_URIS, ImmutableList.of(), null, null); Stream>> splits = inputSource.createSplits( new JsonInputFormat(JSONPathSpec.DEFAULT, null, null, null, null), @@ -175,7 +174,6 @@ public void testWithUrisGlob() URIS_BEFORE_GLOB, null, null, - null, "**.csv" ); @@ -186,26 +184,6 @@ public void testWithUrisGlob() Assert.assertEquals(EXPECTED_OBJECTS, splits.map(InputSplit::get).collect(Collectors.toList())); } - @Test - public void testWithUrisFilter() - { - GoogleCloudStorageInputSource inputSource = new GoogleCloudStorageInputSource( - STORAGE, - INPUT_DATA_CONFIG, - URIS_BEFORE_GLOB, - null, - null, - "*.csv", - null - ); - - Stream>> splits = inputSource.createSplits( - new JsonInputFormat(JSONPathSpec.DEFAULT, null, null, null, null), - null - ); - Assert.assertEquals(EXPECTED_OBJECTS, splits.map(InputSplit::get).collect(Collectors.toList())); - } - @Test public void testIllegalObjectsAndPrefixes() { @@ -217,7 +195,6 @@ public void testIllegalObjectsAndPrefixes() null, PREFIXES, EXPECTED_OBJECTS.get(0), - null, "**.csv" ); } @@ -233,7 +210,6 @@ public void testIllegalUrisAndPrefixes() URIS_BEFORE_GLOB, PREFIXES, null, - null, "**.csv" ); } @@ -250,7 +226,7 @@ public void testWithPrefixesSplit() throws IOException EasyMock.replay(INPUT_DATA_CONFIG); GoogleCloudStorageInputSource inputSource = - new GoogleCloudStorageInputSource(STORAGE, INPUT_DATA_CONFIG, null, PREFIXES, null, null, null); + new GoogleCloudStorageInputSource(STORAGE, INPUT_DATA_CONFIG, null, PREFIXES, null, null); Stream>> splits = inputSource.createSplits( new JsonInputFormat(JSONPathSpec.DEFAULT, null, null, null, null), @@ -272,7 +248,7 @@ public void testCreateSplitsWithSplitHintSpecRespectingHint() throws IOException EasyMock.replay(INPUT_DATA_CONFIG); GoogleCloudStorageInputSource inputSource = - new GoogleCloudStorageInputSource(STORAGE, INPUT_DATA_CONFIG, null, PREFIXES, null, null, null); + new GoogleCloudStorageInputSource(STORAGE, INPUT_DATA_CONFIG, null, PREFIXES, null, null); Stream>> splits = inputSource.createSplits( new JsonInputFormat(JSONPathSpec.DEFAULT, null, null, null, null), @@ -304,7 +280,6 @@ public void testReader() throws IOException null, PREFIXES, null, - null, null ); @@ -349,7 +324,6 @@ public void testCompressedReader() throws IOException null, PREFIXES, null, - null, null ); diff --git a/extensions-core/s3-extensions/src/main/java/org/apache/druid/data/input/s3/S3InputSource.java b/extensions-core/s3-extensions/src/main/java/org/apache/druid/data/input/s3/S3InputSource.java index c0ee349ec77a..4d0d4f0a9472 100644 --- a/extensions-core/s3-extensions/src/main/java/org/apache/druid/data/input/s3/S3InputSource.java +++ b/extensions-core/s3-extensions/src/main/java/org/apache/druid/data/input/s3/S3InputSource.java @@ -111,7 +111,6 @@ public S3InputSource( @JsonProperty("uris") @Nullable List uris, @JsonProperty("prefixes") @Nullable List prefixes, @JsonProperty("objects") @Nullable List objects, - @Deprecated @JsonProperty("filter") @Nullable String filter, @JsonProperty("objectGlob") @Nullable String objectGlob, @JsonProperty("properties") @Nullable S3InputSourceConfig s3InputSourceConfig, @JsonProperty("proxyConfig") @Nullable AWSProxyConfig awsProxyConfig, @@ -119,7 +118,7 @@ public S3InputSource( @JsonProperty("clientConfig") @Nullable AWSClientConfig awsClientConfig ) { - super(S3StorageDruidModule.SCHEME, uris, prefixes, objects, filter, objectGlob); + super(S3StorageDruidModule.SCHEME, uris, prefixes, objects, objectGlob); this.inputDataConfig = Preconditions.checkNotNull(inputDataConfig, "S3DataSegmentPusherConfig"); Preconditions.checkNotNull(s3Client, "s3Client"); this.s3InputSourceConfig = s3InputSourceConfig; @@ -203,7 +202,6 @@ public S3InputSource( uris, prefixes, objects, - null, objectGlob, s3InputSourceConfig, awsProxyConfig, @@ -236,7 +234,6 @@ public S3InputSource( uris, prefixes, objects, - null, objectGlob, s3InputSourceConfig, awsProxyConfig, @@ -346,7 +343,6 @@ public SplittableInputSource> withSplit(InputSplit Date: Wed, 30 Nov 2022 17:38:59 +0530 Subject: [PATCH 066/141] Move quickstart configs to conf/druid/base, add bash launch script, support python2 --- examples/bin/start-druid | 89 ++++++++++--------- examples/bin/start-druid-base | 32 +++++++ .../_common/common.jvm.config | 0 .../_common/common.runtime.properties | 0 .../quickstart => base}/_common/log4j2.xml | 0 .../broker/runtime.properties | 14 ++- .../coordinator-overlord/runtime.properties | 0 .../historical/runtime.properties | 12 +-- .../middleManager/runtime.properties | 15 ++-- .../router/runtime.properties | 6 ++ 10 files changed, 104 insertions(+), 64 deletions(-) create mode 100644 examples/bin/start-druid-base rename examples/conf/druid/{single-server/quickstart => base}/_common/common.jvm.config (100%) rename examples/conf/druid/{single-server/quickstart => base}/_common/common.runtime.properties (100%) rename examples/conf/druid/{single-server/quickstart => base}/_common/log4j2.xml (100%) rename examples/conf/druid/{single-server/quickstart => base}/broker/runtime.properties (64%) rename examples/conf/druid/{single-server/quickstart => base}/coordinator-overlord/runtime.properties (100%) rename examples/conf/druid/{single-server/quickstart => base}/historical/runtime.properties (71%) rename examples/conf/druid/{single-server/quickstart => base}/middleManager/runtime.properties (57%) rename examples/conf/druid/{single-server/quickstart => base}/router/runtime.properties (87%) diff --git a/examples/bin/start-druid b/examples/bin/start-druid index 25c282f461c5..37fe506adec5 100644 --- a/examples/bin/start-druid +++ b/examples/bin/start-druid @@ -1,4 +1,4 @@ -#!/usr/bin/env python3 +#!/usr/bin/env python # Licensed to the Apache Software Foundation (ASF) under one # or more contributor license agreements. See the NOTICE file @@ -24,7 +24,7 @@ import pathlib import multiprocessing import argparse -QUICKSTART_ROOT_CONFIG_PATH = "conf/druid/single-server/quickstart" +BASE_CONFIG_PATH = "conf/druid/base" MEM_GB_SUFFIX = "g" MEM_MB_SUFFIX = "m" @@ -161,19 +161,19 @@ sample usage: def validate_common_jvm_args(config): - if pathlib.Path(f'{config}/_common/common.jvm.config').is_file() is False: - raise ValueError(f'_common/common.jvm.config file is missing in the root config, ' - f'check {QUICKSTART_ROOT_CONFIG_PATH}/_common directory') + if pathlib.Path('{0}/_common/common.jvm.config'.format(config)).is_file() is False: + raise ValueError('_common/common.jvm.config file is missing in the root config, ' + 'check {0}/_common directory'.format(BASE_CONFIG_PATH)) def validate_common_directory(config): - if pathlib.Path(f'{config}/_common').is_dir() is False: + if pathlib.Path('{0}/_common'.format(config)).is_dir() is False: raise ValueError( - f'_common directory is missing in the root config, check {QUICKSTART_ROOT_CONFIG_PATH}/_common directory') + '_common directory is missing in the root config, check {0}/_common directory'.format(BASE_CONFIG_PATH)) - if pathlib.Path(f'{config}/_common/common.runtime.properties').is_file() is False: - raise ValueError(f'_common/common.runtime.properties file is missing in the root config, ' - f'check {QUICKSTART_ROOT_CONFIG_PATH}/_common directory') + if pathlib.Path('{0}/_common/common.runtime.properties'.format(config)).is_file() is False: + raise ValueError('_common/common.runtime.properties file is missing in the root config, ' + 'check {0}/_common directory'.format(BASE_CONFIG_PATH)) def parse_arguments(args): @@ -189,8 +189,8 @@ def parse_arguments(args): zk = True if args.config is not None: config = pathlib.Path(os.path.join(os.getcwd(), args.config)).resolve() - if os.path.exists(config) is False: - raise ValueError(f'config {config} not found') + if config.is_dir() == False: + raise ValueError('config {0} not found'.format(config)) if args.memory is not None: total_memory = args.memory if args.services is not None: @@ -198,10 +198,10 @@ def parse_arguments(args): for service in services: if service not in DEFAULT_SERVICES: - raise ValueError(f'Invalid service name {service}, should be one of {DEFAULT_SERVICES}') + raise ValueError('Invalid service name {0}, should be one of {1}'.format(service, DEFAULT_SERVICES)) if service in service_list: - raise ValueError(f'{service} is specified multiple times') + raise ValueError('{0} is specified multiple times'.format(service)) service_list.append(service) @@ -214,11 +214,11 @@ def parse_arguments(args): def print_startup_config(service_list, config, zk): - print_if_verbose(f'Starting {service_list}') - print_if_verbose(f'Reading config from {config}') + print_if_verbose('Starting {0}'.format(service_list)) + print_if_verbose('Reading config from {0}'.format(config)) if zk: - zk_config = pathlib.Path(f'{os.getcwd()}/../conf/zk').resolve() - print_if_verbose(f'Starting zk, reading default config from {zk_config}') + zk_config = pathlib.Path('{0}/../conf/zk'.format(os.getcwd())).resolve() + print_if_verbose('Starting zk, reading default config from {0}'.format(zk_config)) print_if_verbose('\n') @@ -226,8 +226,8 @@ def middle_manager_task_memory_params_present(config): java_opts_property_present = False worker_capacity_property_present = False - if pathlib.Path(f'{config}/middleManager/runtime.properties').is_file(): - with open(f'{config}/middleManager/runtime.properties') as file: + if pathlib.Path('{0}/middleManager/runtime.properties'.format(config)).is_file(): + with open('{0}/middleManager/runtime.properties'.format(config)) as file: for line in file: if line.startswith(TASK_JAVA_OPTS_PROPERTY): java_opts_property_present = True @@ -238,7 +238,7 @@ def middle_manager_task_memory_params_present(config): def verify_service_config(service, config): - path = f'{config}/{service}/jvm.config' + path = '{0}/{1}/jvm.config'.format(config, service) required_parameters = [XMX_PARAMETER, XMS_PARAMETER] @@ -256,16 +256,16 @@ def verify_service_config(service, config): if len(required_parameters) > 0: params = ",".join(required_parameters) - raise ValueError(f'{params} missing in {service}/jvm.config') + raise ValueError('{0} missing in {1}/jvm.config'.format(params, service)) if service == MIDDLE_MANAGER: - if pathlib.Path(f'{config}/{service}/runtime.properties').is_file() is False: - raise ValueError(f'{service}/runtime.properties file is missing in the root config') + if pathlib.Path('{0}/{1}/runtime.properties'.format(config, service)).is_file() is False: + raise ValueError('{0}/runtime.properties file is missing in the root config'.format(service)) mm_task_java_opts_property, mm_task_worker_capacity_prop = middle_manager_task_memory_params_present(config) if mm_task_java_opts_property is False: - raise ValueError(f'{TASK_JAVA_OPTS_PROPERTY} property missing in {service}/runtime.properties') + raise ValueError('{0} property missing in {1}/runtime.properties'.format(TASK_JAVA_OPTS_PROPERTY, service)) def should_compute_memory(config, total_memory, service_list): @@ -282,7 +282,7 @@ def should_compute_memory(config, total_memory, service_list): jvm_config_count = 0 for service in service_list: - if pathlib.Path(f'{config}/{service}/jvm.config').is_file(): + if pathlib.Path('{0}/{1}/jvm.config'.format(config, service)).is_file(): jvm_config_count += 1 mm_task_property_present = False @@ -305,8 +305,8 @@ def should_compute_memory(config, total_memory, service_list): "(druid.worker.capacity, druid.indexer.runner.javaOptsArray) is present, " "memory argument shouldn't be specified") if jvm_config_count == 0: - raise ValueError("middleManger configs (druid.indexer.runner.javaOptsArray or druid.worker.capacity) is present in " - "middleManager/runtime.properties, \n " + raise ValueError("middleManger configs (druid.indexer.runner.javaOptsArray or druid.worker.capacity) " + "is present in middleManager/runtime.properties, " "add jvm.config for all other services") if jvm_config_count != len(service_list): raise ValueError("jvm.config file should be present for all services or none") @@ -343,7 +343,7 @@ def check_memory_constraint(total_memory, services): # 80% of total memory >= sum of lower bound service memory should be lower_bound_memory = 0 - service_list = services.copy() + service_list = list(services) if MIDDLE_MANAGER in services: service_list.append(TASKS) @@ -353,7 +353,7 @@ def check_memory_constraint(total_memory, services): required_memory = int(lower_bound_memory / 0.8) if total_memory < required_memory: - raise ValueError(f'Minimum memory required for starting services is {required_memory}m') + raise ValueError('Minimum memory required for starting services is {0}m'.format(required_memory)) if total_memory >= 2 * lower_bound_memory: return int(total_memory / 2) @@ -362,14 +362,14 @@ def check_memory_constraint(total_memory, services): def build_mm_task_java_opts_array(memory_type): - task_memory = f'-D{TASK_JAVA_OPTS_PROPERTY}=[' + task_memory = '-D{0}=['.format(TASK_JAVA_OPTS_PROPERTY) mem_array = TASK_MEM_MAP.get(memory_type) java_opts_list = TASK_JAVA_OPTS_ARRAY + mem_array for item in java_opts_list: - task_memory += f'\"{item}\";' + task_memory += '\"{0}\";'.format(item) task_memory = task_memory[:-1] task_memory += ']' @@ -394,16 +394,16 @@ def build_memory_config(service, allocated_memory): if service == TASKS: memory_type, task_count, task_memory = compute_tasks_memory(allocated_memory) java_opts_array = build_mm_task_java_opts_array(memory_type) - return [f'-D{TASK_WORKER_CAPACITY_PROPERTY}={task_count}', java_opts_array], task_memory * task_count + return ['-D{0}={1}'.format(TASK_WORKER_CAPACITY_PROPERTY, task_count), java_opts_array], task_memory * task_count else: heap_memory = HEAP_TO_TOTAL_MEM_RATIO.get(service) * allocated_memory direct_memory = int(allocated_memory - heap_memory) heap_memory = int(heap_memory) if direct_memory == 0: - return f'-Xms{heap_memory}m -Xmx{heap_memory}m', allocated_memory + return '-Xms{0}m -Xmx{0}m'.format(heap_memory), allocated_memory - return f'-Xms{heap_memory}m -Xmx{heap_memory}m -XX:MaxDirectMemorySize={direct_memory}m', allocated_memory + return '-Xms{0}m -Xmx{0}m -XX:MaxDirectMemorySize={1}m'.format(heap_memory, direct_memory), allocated_memory def distribute_memory(services, total_memory): @@ -411,7 +411,7 @@ def distribute_memory(services, total_memory): memory_weight_sum = 0 - service_list = services.copy() + service_list = list(services) if MIDDLE_MANAGER in services: service_list.append(TASKS) @@ -449,9 +449,9 @@ def distribute_memory(services, total_memory): service_memory_config[service], allocated_memory = build_memory_config(service, allocated_memory) - print_if_verbose(f'\nMemory distribution for services:') + print_if_verbose('\nMemory distribution for services:') for key, value in service_memory_config.items(): - print_if_verbose(f'{key}, {value}') + print_if_verbose('{0}, {1}'.format(key, value)) print_if_verbose('\n') return service_memory_config @@ -479,16 +479,17 @@ def build_supervise_script_arguments(service_list, service_memory_config, config prefix = '!p90 ' if jvm_args is None: - commands.append(f'{prefix}{service} bin/run-druid {service} {config}') + commands.append('{0}{1} bin/run-druid {1} {2}'.format(prefix, service, config)) else: if service == MIDDLE_MANAGER: task_config = service_memory_config.get(TASKS) task_count = task_config[0] task_memory = task_config[1] commands.append( - f'{prefix}{service} bin/run-druid {service} {config} \'{jvm_args}\' \'{task_count} {task_memory}\'') + '{0}{1} bin/run-druid {1} {2} \'{3}\' \'{4} {5}\'' + .format(prefix, service, config, jvm_args, task_count, task_memory)) else: - commands.append(f'{prefix}{service} bin/run-druid {service} {config} \'{jvm_args}\'') + commands.append('{0}{1} bin/run-druid {1} {2} \'{3}\''.format(prefix, service, config, jvm_args)) print_if_verbose('Supervise script args:') for item in commands: @@ -513,7 +514,7 @@ def main(): os.chdir(os.path.dirname(sys.argv[0])) if config == "": - config = pathlib.Path(f'{os.getcwd()}/../{QUICKSTART_ROOT_CONFIG_PATH}').resolve() + config = pathlib.Path('{0}/../{1}'.format(os.getcwd(), BASE_CONFIG_PATH)).resolve() validate_common_directory(config) @@ -525,9 +526,9 @@ def main(): # if memory is to be computed, _common directory should contain common.jvm.config validate_common_jvm_args(config) memory_in_mega_bytes = convert_total_memory_string(total_memory) - print_if_verbose(f'Total memory is {memory_in_mega_bytes}m\n') + print_if_verbose('Total memory is {0}m\n'.format(memory_in_mega_bytes)) memory_to_be_used = check_memory_constraint(memory_in_mega_bytes, service_list) - print_if_verbose(f'Memory used for services & tasks {memory_to_be_used}m\n') + print_if_verbose('Memory used for services & tasks {0}m\n'.format(memory_to_be_used)) service_memory_config = distribute_memory(service_list, memory_to_be_used) else: print_if_verbose('Not computing memory distribution, reading memory specification from service jvm.config & ' diff --git a/examples/bin/start-druid-base b/examples/bin/start-druid-base new file mode 100644 index 000000000000..38f7a5f9b0a5 --- /dev/null +++ b/examples/bin/start-druid-base @@ -0,0 +1,32 @@ +#!/bin/bash -eu + +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. + +PWD="$(pwd)" +WHEREAMI="$(dirname "$0")" +WHEREAMI="$(cd "$WHEREAMI" && pwd)" + +if [ -x "$(command -v python3)" ] +then + exec python3 "$WHEREAMI/start-druid" "$@" +elif [ -x "$(command -v python2)" ] +then + exec python2 "$WHEREAMI/start-druid" "$@" +else + exec "$WHEREAMI/start-druid" "$@" +fi diff --git a/examples/conf/druid/single-server/quickstart/_common/common.jvm.config b/examples/conf/druid/base/_common/common.jvm.config similarity index 100% rename from examples/conf/druid/single-server/quickstart/_common/common.jvm.config rename to examples/conf/druid/base/_common/common.jvm.config diff --git a/examples/conf/druid/single-server/quickstart/_common/common.runtime.properties b/examples/conf/druid/base/_common/common.runtime.properties similarity index 100% rename from examples/conf/druid/single-server/quickstart/_common/common.runtime.properties rename to examples/conf/druid/base/_common/common.runtime.properties diff --git a/examples/conf/druid/single-server/quickstart/_common/log4j2.xml b/examples/conf/druid/base/_common/log4j2.xml similarity index 100% rename from examples/conf/druid/single-server/quickstart/_common/log4j2.xml rename to examples/conf/druid/base/_common/log4j2.xml diff --git a/examples/conf/druid/single-server/quickstart/broker/runtime.properties b/examples/conf/druid/base/broker/runtime.properties similarity index 64% rename from examples/conf/druid/single-server/quickstart/broker/runtime.properties rename to examples/conf/druid/base/broker/runtime.properties index f4c494019933..549117efa50d 100644 --- a/examples/conf/druid/single-server/quickstart/broker/runtime.properties +++ b/examples/conf/druid/base/broker/runtime.properties @@ -21,19 +21,15 @@ druid.service=druid/broker druid.plaintextPort=8082 # HTTP server settings -# HTTP server thread pool size. Higher values increase peak load on the Broker, but -# may be useful for high-concurrency workloads. -# Default is max(10, (Number of processors * 17) / 16 + 2) + 30. -# druid.server.http.numThreads=N +#druid.server.http.numThreads=12 # HTTP client settings -# Connection pool size from the Broker to each data server. May be useful to -# raise this for high-concurrency workloads. -# druid.broker.http.numConnections=20 +#druid.broker.http.numConnections=10 +#druid.broker.http.maxQueuedBytes=5MiB # Processing threads and buffers -# Determined automatically based on available memory. For details on how to manually set parameters: -# https://druid.apache.org/docs/latest/operations/basic-cluster-tuning.html#guidelines-for-processing-threads-and-buffers +#druid.processing.buffer.sizeBytes=100MiB +#druid.processing.numMergeBuffers=2 druid.processing.tmpDir=var/druid/processing # Query cache disabled -- push down caching and merging instead diff --git a/examples/conf/druid/single-server/quickstart/coordinator-overlord/runtime.properties b/examples/conf/druid/base/coordinator-overlord/runtime.properties similarity index 100% rename from examples/conf/druid/single-server/quickstart/coordinator-overlord/runtime.properties rename to examples/conf/druid/base/coordinator-overlord/runtime.properties diff --git a/examples/conf/druid/single-server/quickstart/historical/runtime.properties b/examples/conf/druid/base/historical/runtime.properties similarity index 71% rename from examples/conf/druid/single-server/quickstart/historical/runtime.properties rename to examples/conf/druid/base/historical/runtime.properties index d9ce1850e4fc..cf3e966cc30e 100644 --- a/examples/conf/druid/single-server/quickstart/historical/runtime.properties +++ b/examples/conf/druid/base/historical/runtime.properties @@ -20,14 +20,13 @@ druid.service=druid/historical druid.plaintextPort=8083 -# HTTP server thread pool size. Higher values increase peak load on the Broker, but -# may be useful for high-concurrency workloads. -# Default is max(10, (Number of processors * 17) / 16 + 2) + 30. -# druid.server.http.numThreads=N +# HTTP server threads +#druid.server.http.numThreads=12 # Processing threads and buffers -# Determined automatically based on available memory. For details on how to manually set parameters: -# https://druid.apache.org/docs/latest/operations/basic-cluster-tuning.html#guidelines-for-processing-threads-and-buffers +#druid.processing.buffer.sizeBytes=200MiB +#druid.processing.numMergeBuffers=2 +#druid.processing.numThreads=2 druid.processing.tmpDir=var/druid/processing # Segment storage @@ -37,3 +36,4 @@ druid.segmentCache.locations=[{"path":"var/druid/segment-cache","maxSize":"300g" druid.historical.cache.useCache=true druid.historical.cache.populateCache=true druid.cache.type=caffeine +#druid.cache.sizeInBytes=10MiB diff --git a/examples/conf/druid/single-server/quickstart/middleManager/runtime.properties b/examples/conf/druid/base/middleManager/runtime.properties similarity index 57% rename from examples/conf/druid/single-server/quickstart/middleManager/runtime.properties rename to examples/conf/druid/base/middleManager/runtime.properties index 08c58bae6de0..15fa486c7098 100644 --- a/examples/conf/druid/single-server/quickstart/middleManager/runtime.properties +++ b/examples/conf/druid/base/middleManager/runtime.properties @@ -20,17 +20,22 @@ druid.service=druid/middleManager druid.plaintextPort=8091 -# Number of tasks (druid.worker.capacity) and memory usage per task (druid.indexer.runner.javaOptsArray) is automatically -# determined based on available memory. For details on how to manually set parameters, see: -# https://druid.apache.org/docs/latest/operations/basic-cluster-tuning.html#middlemanager +# Number of tasks per middleManager +# druid.worker.capacity # Task launch parameters druid.indexer.runner.javaCommand=bin/run-java +#druid.indexer.runner.javaOptsArray=["-server","-Xms1g","-Xmx1g","-XX:MaxDirectMemorySize=1g","-Duser.timezone=UTC","-Dfile.encoding=UTF-8","-XX:+ExitOnOutOfMemoryError","-Djava.util.logging.manager=org.apache.logging.log4j.jul.LogManager"] druid.indexer.task.baseTaskDir=var/druid/task +#druid.indexer.runner.javaOptsArray=["-server","-Xms1g","-Xmx1g","-XX:MaxDirectMemorySize=1g","-Duser.timezone=UTC","-Dfile.encoding=UTF-8","-XX:+ExitOnOutOfMemoryError","-Djava.util.logging.manager=org.apache.logging.log4j.jul.LogManager"] + +# HTTP server threads +#druid.server.http.numThreads=12 # Processing threads and buffers on Peons -# Determined automatically based on available memory. For details on how to manually set parameters: -# https://druid.apache.org/docs/latest/operations/basic-cluster-tuning.html#guidelines-for-processing-threads-and-buffers +#druid.indexer.fork.property.druid.processing.numMergeBuffers=2 +#druid.indexer.fork.property.druid.processing.buffer.sizeBytes=100MiB +#druid.indexer.fork.property.druid.processing.numThreads=1 # Hadoop indexing druid.indexer.task.hadoopWorkingPath=var/druid/hadoop-tmp diff --git a/examples/conf/druid/single-server/quickstart/router/runtime.properties b/examples/conf/druid/base/router/runtime.properties similarity index 87% rename from examples/conf/druid/single-server/quickstart/router/runtime.properties rename to examples/conf/druid/base/router/runtime.properties index 3858dec044bd..4c38ba88baab 100644 --- a/examples/conf/druid/single-server/quickstart/router/runtime.properties +++ b/examples/conf/druid/base/router/runtime.properties @@ -20,6 +20,12 @@ druid.service=druid/router druid.plaintextPort=8888 +# HTTP proxy +#druid.router.http.numConnections=50 +#druid.router.http.readTimeout=PT5M +#druid.router.http.numMaxThreads=100 +#druid.server.http.numThreads=100 + # Service discovery druid.router.defaultBrokerServiceName=druid/broker druid.router.coordinatorServiceName=druid/coordinator From af164cbc100185f33388759df9e866b468925c58 Mon Sep 17 00:00:00 2001 From: Adarsh Sanjeev Date: Wed, 30 Nov 2022 21:02:48 +0530 Subject: [PATCH 067/141] Fix an issue with WorkerSketchFetcher not terminating on shutdown (#13459) * Fix an issue with WorkerSketchFetcher not terminating on shutdown * Change threadpool name --- .../org/apache/druid/msq/exec/ControllerImpl.java | 5 +++-- .../apache/druid/msq/exec/WorkerSketchFetcher.java | 12 +++++++++--- 2 files changed, 12 insertions(+), 5 deletions(-) diff --git a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/ControllerImpl.java b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/ControllerImpl.java index cafc0f389258..318c33a759c7 100644 --- a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/ControllerImpl.java +++ b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/ControllerImpl.java @@ -524,6 +524,8 @@ private QueryDefinition initializeQueryDefAndState(final Closer closer) context.registerController(this, closer); this.netClient = new ExceptionWrappingWorkerClient(context.taskClientFor(this)); + closer.register(netClient::close); + ClusterStatisticsMergeMode clusterStatisticsMergeMode = MultiStageQueryContext.getClusterStatisticsMergeMode(task.getQuerySpec().getQuery().context()); @@ -532,8 +534,7 @@ private QueryDefinition initializeQueryDefAndState(final Closer closer) int statisticsMaxRetainedBytes = WorkerMemoryParameters.createProductionInstanceForController(context.injector()) .getPartitionStatisticsMaxRetainedBytes(); this.workerSketchFetcher = new WorkerSketchFetcher(netClient, clusterStatisticsMergeMode, statisticsMaxRetainedBytes); - - closer.register(netClient::close); + closer.register(workerSketchFetcher::close); final boolean isDurableStorageEnabled = MultiStageQueryContext.isDurableStorageEnabled(task.getQuerySpec().getQuery().context()); diff --git a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/WorkerSketchFetcher.java b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/WorkerSketchFetcher.java index 3482b50daaff..c4118a9d38e0 100644 --- a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/WorkerSketchFetcher.java +++ b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/WorkerSketchFetcher.java @@ -25,6 +25,7 @@ import org.apache.druid.frame.key.ClusterByPartitions; import org.apache.druid.java.util.common.Either; import org.apache.druid.java.util.common.ISE; +import org.apache.druid.java.util.common.concurrent.Execs; import org.apache.druid.java.util.common.logger.Logger; import org.apache.druid.msq.kernel.StageDefinition; import org.apache.druid.msq.statistics.ClusterByStatisticsCollector; @@ -39,13 +40,12 @@ import java.util.Set; import java.util.concurrent.CompletableFuture; import java.util.concurrent.ExecutorService; -import java.util.concurrent.Executors; import java.util.stream.IntStream; /** * Queues up fetching sketches from workers and progressively generates partitions boundaries. */ -public class WorkerSketchFetcher +public class WorkerSketchFetcher implements AutoCloseable { private static final Logger log = new Logger(WorkerSketchFetcher.class); private static final int DEFAULT_THREAD_COUNT = 4; @@ -63,7 +63,7 @@ public WorkerSketchFetcher(WorkerClient workerClient, ClusterStatisticsMergeMode { this.workerClient = workerClient; this.clusterStatisticsMergeMode = clusterStatisticsMergeMode; - this.executorService = Executors.newFixedThreadPool(DEFAULT_THREAD_COUNT); + this.executorService = Execs.multiThreaded(DEFAULT_THREAD_COUNT, "SketchFetcherThreadPool-%d"); this.statisticsMaxRetainedBytes = statisticsMaxRetainedBytes; } @@ -337,4 +337,10 @@ private static long getPartitionCountFromEither(Either Date: Thu, 1 Dec 2022 00:27:38 +0800 Subject: [PATCH 068/141] update org.bouncycastle:bcprov-jdk15on 1.68 to 1.69 (#13440) --- extensions-core/kubernetes-extensions/pom.xml | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/extensions-core/kubernetes-extensions/pom.xml b/extensions-core/kubernetes-extensions/pom.xml index cb25eee2abf6..aec1de9b8ea1 100644 --- a/extensions-core/kubernetes-extensions/pom.xml +++ b/extensions-core/kubernetes-extensions/pom.xml @@ -90,7 +90,7 @@ org.bouncycastle bcprov-jdk15on - 1.68 + 1.69 runtime From b03d873fe3e0a0e2c73f9bb6ce285a0a75b7fc54 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Wed, 30 Nov 2022 23:50:50 +0530 Subject: [PATCH 069/141] Update supervise script --- examples/bin/supervise | 11 +++-------- 1 file changed, 3 insertions(+), 8 deletions(-) diff --git a/examples/bin/supervise b/examples/bin/supervise index 9a6809bacdc8..e9cdca5ce925 100755 --- a/examples/bin/supervise +++ b/examples/bin/supervise @@ -196,25 +196,20 @@ usage() unless GetOptions( 'kill-timeout|t=i', 'chdir=s', 'svlogd:s', - 'array|a=s{,}' + 'command=s@' => \@commands ); -usage() unless (($opt{'array'} || $opt{'conf'}) && $opt{'vardir'}); +usage() unless ((@commands || $opt{'conf'}) && $opt{'vardir'}); my @config_lines; # get commands to execute either from reading the config file or command line if (not defined $opt{'conf'}) { - @config_lines = split(',', $opt{'array'}); + @config_lines = @commands } else { @config_lines = read_config_file($opt{'conf'}); } -for my $alpha (@config_lines) -{ - $alpha =~ s/;/,/g; -} - my $config = process_config(@config_lines); @commands = @{$config->{commands}}; From aabc53c86ea18360ca758bea509e0be098950469 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Thu, 1 Dec 2022 03:21:23 +0530 Subject: [PATCH 070/141] rename base config directory to auto --- .../{start-druid-base => start-druid-auto} | 0 .../bin/{start-druid => start-druid-main} | 0 .../conf/druid/auto/_common/common.jvm.config | 7 + .../auto/_common/common.runtime.properties | 158 ++++++++++++++++++ examples/conf/druid/auto/_common/log4j2.xml | 86 ++++++++++ .../conf/druid/auto/broker/runtime.properties | 37 ++++ .../coordinator-overlord/runtime.properties | 33 ++++ .../druid/auto/historical/runtime.properties | 39 +++++ .../auto/middleManager/runtime.properties | 41 +++++ .../conf/druid/auto/router/runtime.properties | 34 ++++ 10 files changed, 435 insertions(+) rename examples/bin/{start-druid-base => start-druid-auto} (100%) rename examples/bin/{start-druid => start-druid-main} (100%) create mode 100644 examples/conf/druid/auto/_common/common.jvm.config create mode 100644 examples/conf/druid/auto/_common/common.runtime.properties create mode 100644 examples/conf/druid/auto/_common/log4j2.xml create mode 100644 examples/conf/druid/auto/broker/runtime.properties create mode 100644 examples/conf/druid/auto/coordinator-overlord/runtime.properties create mode 100644 examples/conf/druid/auto/historical/runtime.properties create mode 100644 examples/conf/druid/auto/middleManager/runtime.properties create mode 100644 examples/conf/druid/auto/router/runtime.properties diff --git a/examples/bin/start-druid-base b/examples/bin/start-druid-auto similarity index 100% rename from examples/bin/start-druid-base rename to examples/bin/start-druid-auto diff --git a/examples/bin/start-druid b/examples/bin/start-druid-main similarity index 100% rename from examples/bin/start-druid rename to examples/bin/start-druid-main diff --git a/examples/conf/druid/auto/_common/common.jvm.config b/examples/conf/druid/auto/_common/common.jvm.config new file mode 100644 index 000000000000..fd74cf358979 --- /dev/null +++ b/examples/conf/druid/auto/_common/common.jvm.config @@ -0,0 +1,7 @@ +-server +-XX:+ExitOnOutOfMemoryError +-XX:+UseG1GC +-Duser.timezone=UTC +-Dfile.encoding=UTF-8 +-Djava.io.tmpdir=var/tmp +-Djava.util.logging.manager=org.apache.logging.log4j.jul.LogManager diff --git a/examples/conf/druid/auto/_common/common.runtime.properties b/examples/conf/druid/auto/_common/common.runtime.properties new file mode 100644 index 000000000000..b0adb0695cd7 --- /dev/null +++ b/examples/conf/druid/auto/_common/common.runtime.properties @@ -0,0 +1,158 @@ +# +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. +# + +# Extensions specified in the load list will be loaded by Druid +# We are using local fs for deep storage - not recommended for production - use S3, HDFS, or NFS instead +# We are using local derby for the metadata store - not recommended for production - use MySQL or Postgres instead + +# If you specify `druid.extensions.loadList=[]`, Druid won't load any extension from file system. +# If you don't specify `druid.extensions.loadList`, Druid will load all the extensions under root extension directory. +# More info: https://druid.apache.org/docs/latest/operations/including-extensions.html +druid.extensions.loadList=["druid-hdfs-storage", "druid-kafka-indexing-service", "druid-datasketches", "druid-multi-stage-query"] + +# If you have a different version of Hadoop, place your Hadoop client jar files in your hadoop-dependencies directory +# and uncomment the line below to point to your directory. +#druid.extensions.hadoopDependenciesDir=/my/dir/hadoop-dependencies + + +# +# Hostname +# +druid.host=localhost + +# +# Logging +# + +# Log all runtime properties on startup. Disable to avoid logging properties on startup: +druid.startup.logging.logProperties=true + +# +# Zookeeper +# + +druid.zk.service.host=localhost +druid.zk.paths.base=/druid + +# +# Metadata storage +# + +# For Derby server on your Druid Coordinator (only viable in a cluster with a single Coordinator, no fail-over): +druid.metadata.storage.type=derby +druid.metadata.storage.connector.connectURI=jdbc:derby://localhost:1527/var/druid/metadata.db;create=true +druid.metadata.storage.connector.host=localhost +druid.metadata.storage.connector.port=1527 + +# For MySQL (make sure to include the MySQL JDBC driver on the classpath): +#druid.metadata.storage.type=mysql +#druid.metadata.storage.connector.connectURI=jdbc:mysql://db.example.com:3306/druid +#druid.metadata.storage.connector.user=... +#druid.metadata.storage.connector.password=... + +# For PostgreSQL: +#druid.metadata.storage.type=postgresql +#druid.metadata.storage.connector.connectURI=jdbc:postgresql://db.example.com:5432/druid +#druid.metadata.storage.connector.user=... +#druid.metadata.storage.connector.password=... + +# +# Deep storage +# + +# For local disk (only viable in a cluster if this is a network mount): +druid.storage.type=local +druid.storage.storageDirectory=var/druid/segments + +# For HDFS: +#druid.storage.type=hdfs +#druid.storage.storageDirectory=/druid/segments + +# For S3: +#druid.storage.type=s3 +#druid.storage.bucket=your-bucket +#druid.storage.baseKey=druid/segments +#druid.s3.accessKey=... +#druid.s3.secretKey=... + +# +# Indexing service logs +# + +# For local disk (only viable in a cluster if this is a network mount): +druid.indexer.logs.type=file +druid.indexer.logs.directory=var/druid/indexing-logs + +# For HDFS: +#druid.indexer.logs.type=hdfs +#druid.indexer.logs.directory=/druid/indexing-logs + +# For S3: +#druid.indexer.logs.type=s3 +#druid.indexer.logs.s3Bucket=your-bucket +#druid.indexer.logs.s3Prefix=druid/indexing-logs + +# +# Service discovery +# + +druid.selectors.indexing.serviceName=druid/overlord +druid.selectors.coordinator.serviceName=druid/coordinator + +# +# Monitoring +# + +druid.monitoring.monitors=["org.apache.druid.java.util.metrics.JvmMonitor"] +druid.emitter=noop +druid.emitter.logging.logLevel=info + +# Storage type of double columns +# ommiting this will lead to index double as float at the storage layer + +druid.indexing.doubleStorage=double + +# +# Security +# +druid.server.hiddenProperties=["druid.s3.accessKey","druid.s3.secretKey","druid.metadata.storage.connector.password", "password", "key", "token", "pwd"] + + +# +# SQL +# +druid.sql.enable=true + +# Planning SQL query when there is aggregate distinct in the statement +druid.sql.planner.useGroupingSetForExactDistinct=true + +# +# Lookups +# +druid.lookup.enableLookupSyncOnStartup=false + +# +# Expression processing config +# +druid.expressions.useStrictBooleans=true + +# +# Http client +# +druid.global.http.eagerInitialization=false diff --git a/examples/conf/druid/auto/_common/log4j2.xml b/examples/conf/druid/auto/_common/log4j2.xml new file mode 100644 index 000000000000..66dc13da4c5e --- /dev/null +++ b/examples/conf/druid/auto/_common/log4j2.xml @@ -0,0 +1,86 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/examples/conf/druid/auto/broker/runtime.properties b/examples/conf/druid/auto/broker/runtime.properties new file mode 100644 index 000000000000..549117efa50d --- /dev/null +++ b/examples/conf/druid/auto/broker/runtime.properties @@ -0,0 +1,37 @@ +# +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. +# + +druid.service=druid/broker +druid.plaintextPort=8082 + +# HTTP server settings +#druid.server.http.numThreads=12 + +# HTTP client settings +#druid.broker.http.numConnections=10 +#druid.broker.http.maxQueuedBytes=5MiB + +# Processing threads and buffers +#druid.processing.buffer.sizeBytes=100MiB +#druid.processing.numMergeBuffers=2 +druid.processing.tmpDir=var/druid/processing + +# Query cache disabled -- push down caching and merging instead +druid.broker.cache.useCache=false +druid.broker.cache.populateCache=false diff --git a/examples/conf/druid/auto/coordinator-overlord/runtime.properties b/examples/conf/druid/auto/coordinator-overlord/runtime.properties new file mode 100644 index 000000000000..c053823f8ed8 --- /dev/null +++ b/examples/conf/druid/auto/coordinator-overlord/runtime.properties @@ -0,0 +1,33 @@ +# +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. +# + +druid.service=druid/coordinator +druid.plaintextPort=8081 + +druid.coordinator.startDelay=PT10S +druid.coordinator.period=PT5S +druid.manager.segments.pollDuration=PT5S + +# Run the overlord service in the coordinator process +druid.coordinator.asOverlord.enabled=true +druid.coordinator.asOverlord.overlordService=druid/overlord + +druid.indexer.queue.startDelay=PT5S + +druid.indexer.storage.type=metadata diff --git a/examples/conf/druid/auto/historical/runtime.properties b/examples/conf/druid/auto/historical/runtime.properties new file mode 100644 index 000000000000..cf3e966cc30e --- /dev/null +++ b/examples/conf/druid/auto/historical/runtime.properties @@ -0,0 +1,39 @@ +# +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. +# + +druid.service=druid/historical +druid.plaintextPort=8083 + +# HTTP server threads +#druid.server.http.numThreads=12 + +# Processing threads and buffers +#druid.processing.buffer.sizeBytes=200MiB +#druid.processing.numMergeBuffers=2 +#druid.processing.numThreads=2 +druid.processing.tmpDir=var/druid/processing + +# Segment storage +druid.segmentCache.locations=[{"path":"var/druid/segment-cache","maxSize":"300g"}] + +# Query cache +druid.historical.cache.useCache=true +druid.historical.cache.populateCache=true +druid.cache.type=caffeine +#druid.cache.sizeInBytes=10MiB diff --git a/examples/conf/druid/auto/middleManager/runtime.properties b/examples/conf/druid/auto/middleManager/runtime.properties new file mode 100644 index 000000000000..15fa486c7098 --- /dev/null +++ b/examples/conf/druid/auto/middleManager/runtime.properties @@ -0,0 +1,41 @@ +# +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. +# + +druid.service=druid/middleManager +druid.plaintextPort=8091 + +# Number of tasks per middleManager +# druid.worker.capacity + +# Task launch parameters +druid.indexer.runner.javaCommand=bin/run-java +#druid.indexer.runner.javaOptsArray=["-server","-Xms1g","-Xmx1g","-XX:MaxDirectMemorySize=1g","-Duser.timezone=UTC","-Dfile.encoding=UTF-8","-XX:+ExitOnOutOfMemoryError","-Djava.util.logging.manager=org.apache.logging.log4j.jul.LogManager"] +druid.indexer.task.baseTaskDir=var/druid/task +#druid.indexer.runner.javaOptsArray=["-server","-Xms1g","-Xmx1g","-XX:MaxDirectMemorySize=1g","-Duser.timezone=UTC","-Dfile.encoding=UTF-8","-XX:+ExitOnOutOfMemoryError","-Djava.util.logging.manager=org.apache.logging.log4j.jul.LogManager"] + +# HTTP server threads +#druid.server.http.numThreads=12 + +# Processing threads and buffers on Peons +#druid.indexer.fork.property.druid.processing.numMergeBuffers=2 +#druid.indexer.fork.property.druid.processing.buffer.sizeBytes=100MiB +#druid.indexer.fork.property.druid.processing.numThreads=1 + +# Hadoop indexing +druid.indexer.task.hadoopWorkingPath=var/druid/hadoop-tmp diff --git a/examples/conf/druid/auto/router/runtime.properties b/examples/conf/druid/auto/router/runtime.properties new file mode 100644 index 000000000000..4c38ba88baab --- /dev/null +++ b/examples/conf/druid/auto/router/runtime.properties @@ -0,0 +1,34 @@ +# +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. +# + +druid.service=druid/router +druid.plaintextPort=8888 + +# HTTP proxy +#druid.router.http.numConnections=50 +#druid.router.http.readTimeout=PT5M +#druid.router.http.numMaxThreads=100 +#druid.server.http.numThreads=100 + +# Service discovery +druid.router.defaultBrokerServiceName=druid/broker +druid.router.coordinatorServiceName=druid/coordinator + +# Management proxy to coordinator / overlord: required for unified web console. +druid.router.managementProxy.enabled=true From 2d19c54cccd0b82c66c295c1f3306f90844255f4 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Thu, 1 Dec 2022 03:22:48 +0530 Subject: [PATCH 071/141] rename python script, changes to pass repeated args to supervise --- examples/bin/start-druid-main | 29 +++++++++++++++-------------- 1 file changed, 15 insertions(+), 14 deletions(-) diff --git a/examples/bin/start-druid-main b/examples/bin/start-druid-main index 37fe506adec5..e4e727e04748 100644 --- a/examples/bin/start-druid-main +++ b/examples/bin/start-druid-main @@ -24,7 +24,7 @@ import pathlib import multiprocessing import argparse -BASE_CONFIG_PATH = "conf/druid/base" +BASE_CONFIG_PATH = "conf/druid/auto" MEM_GB_SUFFIX = "g" MEM_MB_SUFFIX = "m" @@ -456,20 +456,21 @@ def distribute_memory(services, total_memory): return service_memory_config +def append_command(commands, command): + commands.append('--command') + commands.append(command) def build_supervise_script_arguments(service_list, service_memory_config, config, zk): - args = [] commands = [] - args.append('supervise') - args.append('-a') + commands.append('supervise') - commands.append(":verify bin/verify-java") - commands.append(":verify bin/verify-default-ports") - commands.append(":notify bin/greet") - commands.append(":kill-timeout 10") + append_command(commands, ":verify bin/verify-java") + append_command(commands, ":verify bin/verify-default-ports") + append_command(commands, ":notify bin/greet") + append_command(commands, ":kill-timeout 10") if zk: - commands.append("!p10 zk bin/run-zk conf") + append_command(commands, "!p10 zk bin/run-zk conf") for service in service_list: jvm_args = service_memory_config.get(service) @@ -479,17 +480,18 @@ def build_supervise_script_arguments(service_list, service_memory_config, config prefix = '!p90 ' if jvm_args is None: - commands.append('{0}{1} bin/run-druid {1} {2}'.format(prefix, service, config)) + append_command(commands, '{0}{1} bin/run-druid {1} {2}'.format(prefix, service, config)) else: if service == MIDDLE_MANAGER: task_config = service_memory_config.get(TASKS) task_count = task_config[0] task_memory = task_config[1] - commands.append( + append_command( + commands, '{0}{1} bin/run-druid {1} {2} \'{3}\' \'{4} {5}\'' .format(prefix, service, config, jvm_args, task_count, task_memory)) else: - commands.append('{0}{1} bin/run-druid {1} {2} \'{3}\''.format(prefix, service, config, jvm_args)) + append_command(commands, '{0}{1} bin/run-druid {1} {2} \'{3}\''.format(prefix, service, config, jvm_args)) print_if_verbose('Supervise script args:') for item in commands: @@ -497,8 +499,7 @@ def build_supervise_script_arguments(service_list, service_memory_config, config print_if_verbose('\n') - args.append(",".join(commands)) - return args + return commands def main(): From 657b5135e46a2b73cf1b86298f7f9004ec552cb6 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Thu, 1 Dec 2022 03:23:42 +0530 Subject: [PATCH 072/141] remove exmaples/conf/druid/base dir --- .../conf/druid/base/_common/common.jvm.config | 7 - .../base/_common/common.runtime.properties | 158 ------------------ examples/conf/druid/base/_common/log4j2.xml | 86 ---------- .../conf/druid/base/broker/runtime.properties | 37 ---- .../coordinator-overlord/runtime.properties | 33 ---- .../druid/base/historical/runtime.properties | 39 ----- .../base/middleManager/runtime.properties | 41 ----- .../conf/druid/base/router/runtime.properties | 34 ---- 8 files changed, 435 deletions(-) delete mode 100644 examples/conf/druid/base/_common/common.jvm.config delete mode 100644 examples/conf/druid/base/_common/common.runtime.properties delete mode 100644 examples/conf/druid/base/_common/log4j2.xml delete mode 100644 examples/conf/druid/base/broker/runtime.properties delete mode 100644 examples/conf/druid/base/coordinator-overlord/runtime.properties delete mode 100644 examples/conf/druid/base/historical/runtime.properties delete mode 100644 examples/conf/druid/base/middleManager/runtime.properties delete mode 100644 examples/conf/druid/base/router/runtime.properties diff --git a/examples/conf/druid/base/_common/common.jvm.config b/examples/conf/druid/base/_common/common.jvm.config deleted file mode 100644 index fd74cf358979..000000000000 --- a/examples/conf/druid/base/_common/common.jvm.config +++ /dev/null @@ -1,7 +0,0 @@ --server --XX:+ExitOnOutOfMemoryError --XX:+UseG1GC --Duser.timezone=UTC --Dfile.encoding=UTF-8 --Djava.io.tmpdir=var/tmp --Djava.util.logging.manager=org.apache.logging.log4j.jul.LogManager diff --git a/examples/conf/druid/base/_common/common.runtime.properties b/examples/conf/druid/base/_common/common.runtime.properties deleted file mode 100644 index b0adb0695cd7..000000000000 --- a/examples/conf/druid/base/_common/common.runtime.properties +++ /dev/null @@ -1,158 +0,0 @@ -# -# Licensed to the Apache Software Foundation (ASF) under one -# or more contributor license agreements. See the NOTICE file -# distributed with this work for additional information -# regarding copyright ownership. The ASF licenses this file -# to you under the Apache License, Version 2.0 (the -# "License"); you may not use this file except in compliance -# with the License. You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, -# software distributed under the License is distributed on an -# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY -# KIND, either express or implied. See the License for the -# specific language governing permissions and limitations -# under the License. -# - -# Extensions specified in the load list will be loaded by Druid -# We are using local fs for deep storage - not recommended for production - use S3, HDFS, or NFS instead -# We are using local derby for the metadata store - not recommended for production - use MySQL or Postgres instead - -# If you specify `druid.extensions.loadList=[]`, Druid won't load any extension from file system. -# If you don't specify `druid.extensions.loadList`, Druid will load all the extensions under root extension directory. -# More info: https://druid.apache.org/docs/latest/operations/including-extensions.html -druid.extensions.loadList=["druid-hdfs-storage", "druid-kafka-indexing-service", "druid-datasketches", "druid-multi-stage-query"] - -# If you have a different version of Hadoop, place your Hadoop client jar files in your hadoop-dependencies directory -# and uncomment the line below to point to your directory. -#druid.extensions.hadoopDependenciesDir=/my/dir/hadoop-dependencies - - -# -# Hostname -# -druid.host=localhost - -# -# Logging -# - -# Log all runtime properties on startup. Disable to avoid logging properties on startup: -druid.startup.logging.logProperties=true - -# -# Zookeeper -# - -druid.zk.service.host=localhost -druid.zk.paths.base=/druid - -# -# Metadata storage -# - -# For Derby server on your Druid Coordinator (only viable in a cluster with a single Coordinator, no fail-over): -druid.metadata.storage.type=derby -druid.metadata.storage.connector.connectURI=jdbc:derby://localhost:1527/var/druid/metadata.db;create=true -druid.metadata.storage.connector.host=localhost -druid.metadata.storage.connector.port=1527 - -# For MySQL (make sure to include the MySQL JDBC driver on the classpath): -#druid.metadata.storage.type=mysql -#druid.metadata.storage.connector.connectURI=jdbc:mysql://db.example.com:3306/druid -#druid.metadata.storage.connector.user=... -#druid.metadata.storage.connector.password=... - -# For PostgreSQL: -#druid.metadata.storage.type=postgresql -#druid.metadata.storage.connector.connectURI=jdbc:postgresql://db.example.com:5432/druid -#druid.metadata.storage.connector.user=... -#druid.metadata.storage.connector.password=... - -# -# Deep storage -# - -# For local disk (only viable in a cluster if this is a network mount): -druid.storage.type=local -druid.storage.storageDirectory=var/druid/segments - -# For HDFS: -#druid.storage.type=hdfs -#druid.storage.storageDirectory=/druid/segments - -# For S3: -#druid.storage.type=s3 -#druid.storage.bucket=your-bucket -#druid.storage.baseKey=druid/segments -#druid.s3.accessKey=... -#druid.s3.secretKey=... - -# -# Indexing service logs -# - -# For local disk (only viable in a cluster if this is a network mount): -druid.indexer.logs.type=file -druid.indexer.logs.directory=var/druid/indexing-logs - -# For HDFS: -#druid.indexer.logs.type=hdfs -#druid.indexer.logs.directory=/druid/indexing-logs - -# For S3: -#druid.indexer.logs.type=s3 -#druid.indexer.logs.s3Bucket=your-bucket -#druid.indexer.logs.s3Prefix=druid/indexing-logs - -# -# Service discovery -# - -druid.selectors.indexing.serviceName=druid/overlord -druid.selectors.coordinator.serviceName=druid/coordinator - -# -# Monitoring -# - -druid.monitoring.monitors=["org.apache.druid.java.util.metrics.JvmMonitor"] -druid.emitter=noop -druid.emitter.logging.logLevel=info - -# Storage type of double columns -# ommiting this will lead to index double as float at the storage layer - -druid.indexing.doubleStorage=double - -# -# Security -# -druid.server.hiddenProperties=["druid.s3.accessKey","druid.s3.secretKey","druid.metadata.storage.connector.password", "password", "key", "token", "pwd"] - - -# -# SQL -# -druid.sql.enable=true - -# Planning SQL query when there is aggregate distinct in the statement -druid.sql.planner.useGroupingSetForExactDistinct=true - -# -# Lookups -# -druid.lookup.enableLookupSyncOnStartup=false - -# -# Expression processing config -# -druid.expressions.useStrictBooleans=true - -# -# Http client -# -druid.global.http.eagerInitialization=false diff --git a/examples/conf/druid/base/_common/log4j2.xml b/examples/conf/druid/base/_common/log4j2.xml deleted file mode 100644 index 66dc13da4c5e..000000000000 --- a/examples/conf/druid/base/_common/log4j2.xml +++ /dev/null @@ -1,86 +0,0 @@ - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - diff --git a/examples/conf/druid/base/broker/runtime.properties b/examples/conf/druid/base/broker/runtime.properties deleted file mode 100644 index 549117efa50d..000000000000 --- a/examples/conf/druid/base/broker/runtime.properties +++ /dev/null @@ -1,37 +0,0 @@ -# -# Licensed to the Apache Software Foundation (ASF) under one -# or more contributor license agreements. See the NOTICE file -# distributed with this work for additional information -# regarding copyright ownership. The ASF licenses this file -# to you under the Apache License, Version 2.0 (the -# "License"); you may not use this file except in compliance -# with the License. You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, -# software distributed under the License is distributed on an -# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY -# KIND, either express or implied. See the License for the -# specific language governing permissions and limitations -# under the License. -# - -druid.service=druid/broker -druid.plaintextPort=8082 - -# HTTP server settings -#druid.server.http.numThreads=12 - -# HTTP client settings -#druid.broker.http.numConnections=10 -#druid.broker.http.maxQueuedBytes=5MiB - -# Processing threads and buffers -#druid.processing.buffer.sizeBytes=100MiB -#druid.processing.numMergeBuffers=2 -druid.processing.tmpDir=var/druid/processing - -# Query cache disabled -- push down caching and merging instead -druid.broker.cache.useCache=false -druid.broker.cache.populateCache=false diff --git a/examples/conf/druid/base/coordinator-overlord/runtime.properties b/examples/conf/druid/base/coordinator-overlord/runtime.properties deleted file mode 100644 index c053823f8ed8..000000000000 --- a/examples/conf/druid/base/coordinator-overlord/runtime.properties +++ /dev/null @@ -1,33 +0,0 @@ -# -# Licensed to the Apache Software Foundation (ASF) under one -# or more contributor license agreements. See the NOTICE file -# distributed with this work for additional information -# regarding copyright ownership. The ASF licenses this file -# to you under the Apache License, Version 2.0 (the -# "License"); you may not use this file except in compliance -# with the License. You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, -# software distributed under the License is distributed on an -# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY -# KIND, either express or implied. See the License for the -# specific language governing permissions and limitations -# under the License. -# - -druid.service=druid/coordinator -druid.plaintextPort=8081 - -druid.coordinator.startDelay=PT10S -druid.coordinator.period=PT5S -druid.manager.segments.pollDuration=PT5S - -# Run the overlord service in the coordinator process -druid.coordinator.asOverlord.enabled=true -druid.coordinator.asOverlord.overlordService=druid/overlord - -druid.indexer.queue.startDelay=PT5S - -druid.indexer.storage.type=metadata diff --git a/examples/conf/druid/base/historical/runtime.properties b/examples/conf/druid/base/historical/runtime.properties deleted file mode 100644 index cf3e966cc30e..000000000000 --- a/examples/conf/druid/base/historical/runtime.properties +++ /dev/null @@ -1,39 +0,0 @@ -# -# Licensed to the Apache Software Foundation (ASF) under one -# or more contributor license agreements. See the NOTICE file -# distributed with this work for additional information -# regarding copyright ownership. The ASF licenses this file -# to you under the Apache License, Version 2.0 (the -# "License"); you may not use this file except in compliance -# with the License. You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, -# software distributed under the License is distributed on an -# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY -# KIND, either express or implied. See the License for the -# specific language governing permissions and limitations -# under the License. -# - -druid.service=druid/historical -druid.plaintextPort=8083 - -# HTTP server threads -#druid.server.http.numThreads=12 - -# Processing threads and buffers -#druid.processing.buffer.sizeBytes=200MiB -#druid.processing.numMergeBuffers=2 -#druid.processing.numThreads=2 -druid.processing.tmpDir=var/druid/processing - -# Segment storage -druid.segmentCache.locations=[{"path":"var/druid/segment-cache","maxSize":"300g"}] - -# Query cache -druid.historical.cache.useCache=true -druid.historical.cache.populateCache=true -druid.cache.type=caffeine -#druid.cache.sizeInBytes=10MiB diff --git a/examples/conf/druid/base/middleManager/runtime.properties b/examples/conf/druid/base/middleManager/runtime.properties deleted file mode 100644 index 15fa486c7098..000000000000 --- a/examples/conf/druid/base/middleManager/runtime.properties +++ /dev/null @@ -1,41 +0,0 @@ -# -# Licensed to the Apache Software Foundation (ASF) under one -# or more contributor license agreements. See the NOTICE file -# distributed with this work for additional information -# regarding copyright ownership. The ASF licenses this file -# to you under the Apache License, Version 2.0 (the -# "License"); you may not use this file except in compliance -# with the License. You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, -# software distributed under the License is distributed on an -# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY -# KIND, either express or implied. See the License for the -# specific language governing permissions and limitations -# under the License. -# - -druid.service=druid/middleManager -druid.plaintextPort=8091 - -# Number of tasks per middleManager -# druid.worker.capacity - -# Task launch parameters -druid.indexer.runner.javaCommand=bin/run-java -#druid.indexer.runner.javaOptsArray=["-server","-Xms1g","-Xmx1g","-XX:MaxDirectMemorySize=1g","-Duser.timezone=UTC","-Dfile.encoding=UTF-8","-XX:+ExitOnOutOfMemoryError","-Djava.util.logging.manager=org.apache.logging.log4j.jul.LogManager"] -druid.indexer.task.baseTaskDir=var/druid/task -#druid.indexer.runner.javaOptsArray=["-server","-Xms1g","-Xmx1g","-XX:MaxDirectMemorySize=1g","-Duser.timezone=UTC","-Dfile.encoding=UTF-8","-XX:+ExitOnOutOfMemoryError","-Djava.util.logging.manager=org.apache.logging.log4j.jul.LogManager"] - -# HTTP server threads -#druid.server.http.numThreads=12 - -# Processing threads and buffers on Peons -#druid.indexer.fork.property.druid.processing.numMergeBuffers=2 -#druid.indexer.fork.property.druid.processing.buffer.sizeBytes=100MiB -#druid.indexer.fork.property.druid.processing.numThreads=1 - -# Hadoop indexing -druid.indexer.task.hadoopWorkingPath=var/druid/hadoop-tmp diff --git a/examples/conf/druid/base/router/runtime.properties b/examples/conf/druid/base/router/runtime.properties deleted file mode 100644 index 4c38ba88baab..000000000000 --- a/examples/conf/druid/base/router/runtime.properties +++ /dev/null @@ -1,34 +0,0 @@ -# -# Licensed to the Apache Software Foundation (ASF) under one -# or more contributor license agreements. See the NOTICE file -# distributed with this work for additional information -# regarding copyright ownership. The ASF licenses this file -# to you under the Apache License, Version 2.0 (the -# "License"); you may not use this file except in compliance -# with the License. You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, -# software distributed under the License is distributed on an -# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY -# KIND, either express or implied. See the License for the -# specific language governing permissions and limitations -# under the License. -# - -druid.service=druid/router -druid.plaintextPort=8888 - -# HTTP proxy -#druid.router.http.numConnections=50 -#druid.router.http.readTimeout=PT5M -#druid.router.http.numMaxThreads=100 -#druid.server.http.numThreads=100 - -# Service discovery -druid.router.defaultBrokerServiceName=druid/broker -druid.router.coordinatorServiceName=druid/coordinator - -# Management proxy to coordinator / overlord: required for unified web console. -druid.router.managementProxy.enabled=true From a4c34e7dffa5d0f0a934d9aa17ebde2b3b22f838 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Thu, 1 Dec 2022 03:23:51 +0530 Subject: [PATCH 073/141] add docs --- docs/operations/python.md | 54 +++++++++++++++++++++ docs/operations/single-server.md | 64 ++++++++++++++----------- docs/tutorials/cluster.md | 5 +- docs/tutorials/index.md | 25 +++++----- docs/tutorials/tutorial-batch-hadoop.md | 8 ++-- docs/tutorials/tutorial-kafka.md | 2 +- 6 files changed, 112 insertions(+), 46 deletions(-) create mode 100644 docs/operations/python.md diff --git a/docs/operations/python.md b/docs/operations/python.md new file mode 100644 index 000000000000..2bfb8f0a0615 --- /dev/null +++ b/docs/operations/python.md @@ -0,0 +1,54 @@ +--- +id: python +title: "Python Installation" +--- + + + +Apache Druid startup script requires python. This page has steps to install python and additional libraries required by the startup script. + +## Python installation instruction + +### MacOS +MacOS comes with a version of Python2, which has been deprecated. Python2 and Python3 can coexist on the same machine without problems. + +#### Install the official Python release +* Browse to the [Python Downloads Page](https://www.python.org/downloads/) and download the latest version (3.x.x) +* Verify if python3 and pip3 (python package manager) is installed by issuing `python3` and `pip3 -V` commands. + +#### Install with Homebrew +Refer [Installing Python 3 on Mac OS X](https://docs.python-guide.org/starting/install3/osx/) + +### Linux + +#### Debian + - `sudo apt update` + - `sudo apt install -y python3-pip` +#### RHEL + - `sudo yum install -y epel-release` + - `sudo yum install -y python3-pip` + +## Additional libraries + +### psutil +- pip3 install psutil + +### pathlib +- pip3 install pathlib \ No newline at end of file diff --git a/docs/operations/single-server.md b/docs/operations/single-server.md index 34dd2dc0e006..852e1597a99f 100644 --- a/docs/operations/single-server.md +++ b/docs/operations/single-server.md @@ -25,13 +25,15 @@ title: "Single server deployment" Druid includes a set of reference configurations and launch scripts for single-machine deployments: -- `nano-quickstart` -- `micro-quickstart` -- `small` -- `medium` -- `large` -- `xlarge` -- `start-druid` +- `auto` (run script: `bin/start-druid-auto`) +- `nano-quickstart` (run script: `bin/start-nano-quickstart`) +- `micro-quickstart` (run script: `bin/start-micro-quickstart`) +- `small` (run script: `bin/start-single-server-small`) +- `medium` (run script: `bin/start-single-server-medium`) +- `large` (run script: `bin/start-single-server-large`) +- `xlarge` (run script: `bin/start-single-server-xlarge`) + +The `auto` can be used to spin up Druid cluster on any machine. Read more details about this configuration [below](#Druid-auto-start). The `micro-quickstart` is sized for small machines like laptops and is intended for quick evaluation use-cases. @@ -43,43 +45,51 @@ The startup scripts for these example configurations run a single ZK instance al The example configurations run the Druid Coordinator and Overlord together in a single process using the optional configuration `druid.coordinator.asOverlord.enabled=true`, described in the [Coordinator configuration documentation](../configuration/index.md#coordinator-operation). -The `start-druid` is a generic launch script for starting druid services on single server, it accepts optional arguments like services, memory and config. -All reference configurations can be acheived by passing appropriate arguments to this script. -Existing launch scripts are deprecated and will be removed in the next release. +While example configurations are provided for very large single machines, at higher scales we recommend running Druid in a [clustered deployment](../tutorials/cluster.md), for fault-tolerance and reduced resource contention. + +## Druid auto start + +Setting up a new Druid cluster can sometimes be complicated as there are several runtime properties required by each service. Each process must also be given the appropriate jvm arguments so that the system can perform optimally. + +`start-druid-auto` is a generic launch script capable of starting any set of Druid services on a server. +It accepts optional arguments such as list of services, total memory and a config directory to override default jvm arguments and service-specific runtime properties. +All other reference configurations (e.g. `micro`, `small`, `xlarge`) can be obtained by passing the +correct memory to this script. +Druid services will use all processors and upto 80% memory on the system. +For details about possible arguments, run `bin/start-druid-auto --help`. + +The corresponding launch scripts (e.g. `start-micro-quickstart`) are now deprecated. -While example configurations are provided for very large single machines, at higher scales we recommend running Druid in a [clustered deployment](../tutorials/cluster.md), for fault-tolerance and reduced resource contention. ## Single server reference configurations -### Nano: 1 CPU, 4GiB RAM +### Nano-Quickstart: 1 CPU, 4GiB RAM -- Launch command: `bin/start-druid --memory=4g` -- Configuration directory: `conf/druid/single-server/quickstart` +- Launch command: `bin/start-nano-quickstart` +- Configuration directory: `conf/druid/single-server/nano-quickstart` -### Micro: 4 CPU, 16GiB RAM +### Micro-Quickstart: 4 CPU, 16GiB RAM -- Launch command: `bin/start-druid --memory=16g` -- Configuration directory: `conf/druid/single-server/quickstart` +- Launch command: `bin/start-micro-quickstart` +- Configuration directory: `conf/druid/single-server/micro-quickstart` ### Small: 8 CPU, 64GiB RAM (~i3.2xlarge) -- Launch command: `bin/start-druid --memory=64g` -- Configuration directory: `conf/druid/single-server/quickstart` +- Launch command: `bin/start-small` +- Configuration directory: `conf/druid/single-server/small` ### Medium: 16 CPU, 128GiB RAM (~i3.4xlarge) -- Launch command: `bin/start-druid --memory=128g` -- Configuration directory: `conf/druid/single-server/quickstart` +- Launch command: `bin/start-medium` +- Configuration directory: `conf/druid/single-server/medium` ### Large: 32 CPU, 256GiB RAM (~i3.8xlarge) -- Launch command: `bin/start-druid --memory=256g` -- Configuration directory: `conf/druid/single-server/quickstart` +- Launch command: `bin/start-large` +- Configuration directory: `conf/druid/single-server/large` ### X-Large: 64 CPU, 512GiB RAM (~i3.16xlarge) -- Launch command: `bin/start-druid --memory=512g` -- Configuration directory: `conf/druid/single-server/quickstart` - -Memory argument (`--memory`) in the above launch command is optional, if not specified Druid will use upto 80% of system memory. \ No newline at end of file +- Launch command: `bin/start-xlarge` +- Configuration directory: `conf/druid/single-server/xlarge` \ No newline at end of file diff --git a/docs/tutorials/cluster.md b/docs/tutorials/cluster.md index b61953c2f427..5ee68e2cd13e 100644 --- a/docs/tutorials/cluster.md +++ b/docs/tutorials/cluster.md @@ -130,7 +130,10 @@ The [basic cluster tuning guide](../operations/basic-cluster-tuning.md) has info ## Select OS -We recommend running your favorite Linux distribution. You will also need [Java 8 or 11](../operations/java.md). +We recommend running your favorite Linux distribution. You will also need + +* [Java 8 or 11](../operations/java.md). +* [Python2 or Python3](../operations/python.md) > If needed, you can specify where to find Java using the environment variables > `DRUID_JAVA_HOME` or `JAVA_HOME`. For more details run the `bin/verify-java` script. diff --git a/docs/tutorials/index.md b/docs/tutorials/index.md index 8d029d568189..c611fde67cc7 100644 --- a/docs/tutorials/index.md +++ b/docs/tutorials/index.md @@ -22,8 +22,7 @@ title: "Quickstart (local)" ~ under the License. --> - -This quickstart gets you started with Apache Druid using the [`micro`](../operations/single-server.md#micro-4-cpu-16gib-ram) configuration, and introduces you to Druid ingestion and query features. +This quickstart gets you started with Apache Druid and introduces you to Druid ingestion and query features. For this tutorial, we recommend a machine with at least 6 GB of RAM. In this quickstart, you'll do the following: - install Druid @@ -37,16 +36,16 @@ Druid supports a variety of ingestion options. Once you're done with this tutori You can follow these steps on a relatively modest machine, such as a workstation or virtual server with 16 GiB of RAM. -Druid comes equipped with a single launch script that can be used to run several -[startup configuration profiles](../operations/single-server.md) for a -range of machine sizes. These range from `nano` (1 CPU, 4GiB RAM) to `x-large` (64 CPU, 512GiB RAM). For more -information, see [Single server deployment](../operations/single-server.md). For information on deploying Druid services -across clustered machines, see [Clustered deployment](./cluster.md). +Druid comes equipped with launch scripts that can be used to start all processes on a single server. Here, we will use [`auto`](../operations/single-server.md#Druid-auto-start), which automatically sets various runtime properties based on available processors and memory. + +In addition, Druid includes several [bundled non-automatic profiles](../operations/single-server.md) for a range of machine sizes. These range from nano (1 CPU, 4GiB RAM) to x-large (64 CPU, 512GiB RAM). +We won't use those here, but for more information, see [Single server deployment](../operations/single-server.md). For additional information on deploying Druid services across clustered machines, see [Clustered deployment](./cluster.md). The software requirements for the installation machine are: * Linux, Mac OS X, or other Unix-like OS. (Windows is not supported.) * Java 8u92+ or Java 11. +* [Python2 or Python3](../operations/python.md) > Druid relies on the environment variables `JAVA_HOME` or `DRUID_JAVA_HOME` to find Java on the machine. You can set `DRUID_JAVA_HOME` if there is more than one instance of Java. To verify Java requirements for your environment, run the @@ -73,21 +72,21 @@ The distribution directory contains `LICENSE` and `NOTICE` files and subdirector ## Start up Druid services -Start up Druid services using the `micro` single-machine configuration. +Start up Druid services using the `auto` single-machine configuration. This configuration includes default settings that are appropriate for this tutorial, such as loading the `druid-multi-stage-query` extension by default so that you can use the MSQ task engine. -You can view that setting and others in the configuration files in the `conf/druid/single-server/quickstart/`. +You can view that setting and others in the configuration files in the `conf/druid/auto`. From the apache-druid-{{DRUIDVERSION}} package root, run the following command: ```bash -./bin/start-druid --memory=16g +./bin/start-druid-main-auto ``` This brings up instances of ZooKeeper and the Druid services: ```bash -$ ./bin/start-micro-quickstart +$ ./bin/start-druid-main-auto [Tue Nov 29 16:31:06 2022] Starting Apache Druid. [Tue Nov 29 16:31:06 2022] Open http://localhost:8888/ in your browser to access the web console. [Tue Nov 29 16:31:06 2022] Or, if you have enabled TLS, use https on port 9088. @@ -105,7 +104,7 @@ the Druid root directory, apache-druid-{{DRUIDVERSION}}. Each service writes to At any time, you can revert Druid to its original, post-installation state by deleting the entire `var` directory. You may want to do this, for example, between Druid tutorials or after experimentation, to start with a fresh instance. -To stop Druid at any time, use CTRL+C in the terminal. This exits the `bin/start-druid` script and terminates all Druid processes. +To stop Druid at any time, use CTRL+C in the terminal. This exits the `bin/start-druid-auto` script and terminates all Druid processes. ## Open the web console @@ -224,4 +223,4 @@ See the following topics for more information: * [Tutorial: Load stream data from Apache Kafka](./tutorial-kafka.md) to load streaming data from a Kafka topic. * [Extensions](../development/extensions.md) for details on Druid extensions. -Remember that after stopping Druid services, you can start clean next time by deleting the `var` directory from the Druid root directory and running the `bin/start-druid` script again. You may want to do this before using other data ingestion tutorials, since they use the same Wikipedia datasource. +Remember that after stopping Druid services, you can start clean next time by deleting the `var` directory from the Druid root directory and running the `bin/start-druid-auto` script again. You may want to do this before using other data ingestion tutorials, since they use the same Wikipedia datasource. diff --git a/docs/tutorials/tutorial-batch-hadoop.md b/docs/tutorials/tutorial-batch-hadoop.md index bff033a28c41..e8ea8a19052f 100644 --- a/docs/tutorials/tutorial-batch-hadoop.md +++ b/docs/tutorials/tutorial-batch-hadoop.md @@ -28,7 +28,7 @@ This tutorial shows you how to load data files into Apache Druid using a remote For this tutorial, we'll assume that you've already completed the previous [batch ingestion tutorial](tutorial-batch.md) using Druid's native batch ingestion system and are using the -`micro` single-machine configuration as described in the [quickstart](../operations/single-server.md#micro-4-cpu-16gib-ram). +`auto` single-machine configuration as described in the [quickstart](../operations/single-server.md#Druid-auto-start). ## Install Docker @@ -156,7 +156,7 @@ cp /tmp/shared/hadoop_xml/*.xml {PATH_TO_DRUID}/conf/druid/single-server/micro-q ### Update Druid segment and log storage -In your favorite text editor, open `conf/druid/single-server/micro-quickstart/_common/common.runtime.properties`, and make the following edits: +In your favorite text editor, open `conf/druid/auto/_common/common.runtime.properties`, and make the following edits: #### Disable local deep storage and enable HDFS deep storage @@ -196,7 +196,7 @@ druid.indexer.logs.directory=/druid/indexing-logs Once the Hadoop .xml files have been copied to the Druid cluster and the segment/log storage configuration has been updated to use HDFS, the Druid cluster needs to be restarted for the new configurations to take effect. -If the cluster is still running, CTRL-C to terminate the `bin/start-micro-quickstart` script, and re-run it to bring the Druid services back up. +If the cluster is still running, CTRL-C to terminate the `bin/start-druid-auto` script, and re-run it to bring the Druid services back up. ## Load batch data @@ -221,7 +221,7 @@ This tutorial is only meant to be used together with the [query tutorial](../tut If you wish to go through any of the other tutorials, you will need to: * Shut down the cluster and reset the cluster state by removing the contents of the `var` directory under the druid package. -* Revert the deep storage and task storage config back to local types in `conf/druid/single-server/micro-quickstart/_common/common.runtime.properties` +* Revert the deep storage and task storage config back to local types in `conf/druid/auto/_common/common.runtime.properties` * Restart the cluster This is necessary because the other ingestion tutorials will write to the same "wikipedia" datasource, and later tutorials expect the cluster to use local deep storage. diff --git a/docs/tutorials/tutorial-kafka.md b/docs/tutorials/tutorial-kafka.md index bbdf6af72ea0..73feeaf65901 100644 --- a/docs/tutorials/tutorial-kafka.md +++ b/docs/tutorials/tutorial-kafka.md @@ -30,7 +30,7 @@ The tutorial guides you through the steps to load sample nested clickstream data ## Prerequisites -Before you follow the steps in this tutorial, download Druid as described in the [quickstart](index.md) using the [micro](../operations/single-server.md#micro-4-cpu-16gib-ram) single-machine configuration and have it running on your local machine. You don't need to have loaded any data. +Before you follow the steps in this tutorial, download Druid as described in the [quickstart](index.md) using the [auto](../operations/single-server.md#Druid-auto-start) single-machine configuration and have it running on your local machine. You don't need to have loaded any data. ## Download and start Kafka From 0fd9b09f61acce6104364a5cbda2207257b4569c Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Thu, 1 Dec 2022 03:33:14 +0530 Subject: [PATCH 074/141] restore changes in conf dir --- .../conf/druid/auto/broker/runtime.properties | 14 +++++++++----- .../druid/auto/historical/runtime.properties | 16 ++++++---------- .../druid/auto/middleManager/runtime.properties | 15 +++++---------- .../conf/druid/auto/router/runtime.properties | 6 ------ 4 files changed, 20 insertions(+), 31 deletions(-) diff --git a/examples/conf/druid/auto/broker/runtime.properties b/examples/conf/druid/auto/broker/runtime.properties index 549117efa50d..f4c494019933 100644 --- a/examples/conf/druid/auto/broker/runtime.properties +++ b/examples/conf/druid/auto/broker/runtime.properties @@ -21,15 +21,19 @@ druid.service=druid/broker druid.plaintextPort=8082 # HTTP server settings -#druid.server.http.numThreads=12 +# HTTP server thread pool size. Higher values increase peak load on the Broker, but +# may be useful for high-concurrency workloads. +# Default is max(10, (Number of processors * 17) / 16 + 2) + 30. +# druid.server.http.numThreads=N # HTTP client settings -#druid.broker.http.numConnections=10 -#druid.broker.http.maxQueuedBytes=5MiB +# Connection pool size from the Broker to each data server. May be useful to +# raise this for high-concurrency workloads. +# druid.broker.http.numConnections=20 # Processing threads and buffers -#druid.processing.buffer.sizeBytes=100MiB -#druid.processing.numMergeBuffers=2 +# Determined automatically based on available memory. For details on how to manually set parameters: +# https://druid.apache.org/docs/latest/operations/basic-cluster-tuning.html#guidelines-for-processing-threads-and-buffers druid.processing.tmpDir=var/druid/processing # Query cache disabled -- push down caching and merging instead diff --git a/examples/conf/druid/auto/historical/runtime.properties b/examples/conf/druid/auto/historical/runtime.properties index cf3e966cc30e..6c241aa7cf06 100644 --- a/examples/conf/druid/auto/historical/runtime.properties +++ b/examples/conf/druid/auto/historical/runtime.properties @@ -17,23 +17,19 @@ # under the License. # -druid.service=druid/historical -druid.plaintextPort=8083 - -# HTTP server threads -#druid.server.http.numThreads=12 +# HTTP server thread pool size. Higher values increase peak load on the Broker, but +# may be useful for high-concurrency workloads. +# Default is max(10, (Number of processors * 17) / 16 + 2) + 30. +# druid.server.http.numThreads=N # Processing threads and buffers -#druid.processing.buffer.sizeBytes=200MiB -#druid.processing.numMergeBuffers=2 -#druid.processing.numThreads=2 +# Determined automatically based on available memory. For details on how to manually set parameters: +# https://druid.apache.org/docs/latest/operations/basic-cluster-tuning.html#guidelines-for-processing-threads-and-buffers druid.processing.tmpDir=var/druid/processing # Segment storage druid.segmentCache.locations=[{"path":"var/druid/segment-cache","maxSize":"300g"}] - # Query cache druid.historical.cache.useCache=true druid.historical.cache.populateCache=true druid.cache.type=caffeine -#druid.cache.sizeInBytes=10MiB diff --git a/examples/conf/druid/auto/middleManager/runtime.properties b/examples/conf/druid/auto/middleManager/runtime.properties index 15fa486c7098..08c58bae6de0 100644 --- a/examples/conf/druid/auto/middleManager/runtime.properties +++ b/examples/conf/druid/auto/middleManager/runtime.properties @@ -20,22 +20,17 @@ druid.service=druid/middleManager druid.plaintextPort=8091 -# Number of tasks per middleManager -# druid.worker.capacity +# Number of tasks (druid.worker.capacity) and memory usage per task (druid.indexer.runner.javaOptsArray) is automatically +# determined based on available memory. For details on how to manually set parameters, see: +# https://druid.apache.org/docs/latest/operations/basic-cluster-tuning.html#middlemanager # Task launch parameters druid.indexer.runner.javaCommand=bin/run-java -#druid.indexer.runner.javaOptsArray=["-server","-Xms1g","-Xmx1g","-XX:MaxDirectMemorySize=1g","-Duser.timezone=UTC","-Dfile.encoding=UTF-8","-XX:+ExitOnOutOfMemoryError","-Djava.util.logging.manager=org.apache.logging.log4j.jul.LogManager"] druid.indexer.task.baseTaskDir=var/druid/task -#druid.indexer.runner.javaOptsArray=["-server","-Xms1g","-Xmx1g","-XX:MaxDirectMemorySize=1g","-Duser.timezone=UTC","-Dfile.encoding=UTF-8","-XX:+ExitOnOutOfMemoryError","-Djava.util.logging.manager=org.apache.logging.log4j.jul.LogManager"] - -# HTTP server threads -#druid.server.http.numThreads=12 # Processing threads and buffers on Peons -#druid.indexer.fork.property.druid.processing.numMergeBuffers=2 -#druid.indexer.fork.property.druid.processing.buffer.sizeBytes=100MiB -#druid.indexer.fork.property.druid.processing.numThreads=1 +# Determined automatically based on available memory. For details on how to manually set parameters: +# https://druid.apache.org/docs/latest/operations/basic-cluster-tuning.html#guidelines-for-processing-threads-and-buffers # Hadoop indexing druid.indexer.task.hadoopWorkingPath=var/druid/hadoop-tmp diff --git a/examples/conf/druid/auto/router/runtime.properties b/examples/conf/druid/auto/router/runtime.properties index 4c38ba88baab..3858dec044bd 100644 --- a/examples/conf/druid/auto/router/runtime.properties +++ b/examples/conf/druid/auto/router/runtime.properties @@ -20,12 +20,6 @@ druid.service=druid/router druid.plaintextPort=8888 -# HTTP proxy -#druid.router.http.numConnections=50 -#druid.router.http.readTimeout=PT5M -#druid.router.http.numMaxThreads=100 -#druid.server.http.numThreads=100 - # Service discovery druid.router.defaultBrokerServiceName=druid/broker druid.router.coordinatorServiceName=druid/coordinator From c0da3cdbed645aef569d912fd44fcb3682eeb13a Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Thu, 1 Dec 2022 03:44:54 +0530 Subject: [PATCH 075/141] update start-druid-auto --- examples/bin/start-druid-auto | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/examples/bin/start-druid-auto b/examples/bin/start-druid-auto index 38f7a5f9b0a5..b6dc9cfbf57b 100644 --- a/examples/bin/start-druid-auto +++ b/examples/bin/start-druid-auto @@ -23,10 +23,10 @@ WHEREAMI="$(cd "$WHEREAMI" && pwd)" if [ -x "$(command -v python3)" ] then - exec python3 "$WHEREAMI/start-druid" "$@" + exec python3 "$WHEREAMI/start-druid-main" "$@" elif [ -x "$(command -v python2)" ] then - exec python2 "$WHEREAMI/start-druid" "$@" + exec python2 "$WHEREAMI/start-druid-main" "$@" else - exec "$WHEREAMI/start-druid" "$@" + exec "$WHEREAMI/start-druid-main" "$@" fi From cc2e4a80ff485a15a9bba1ea94bcf0915a35b264 Mon Sep 17 00:00:00 2001 From: 317brian <53799971+317brian@users.noreply.github.com> Date: Wed, 30 Nov 2022 16:25:35 -0800 Subject: [PATCH 076/141] doc: add a basic JDBC tutorial (#13343) * initial commit for jdbc tutorial (cherry picked from commit 04c4adad71e5436b76c3425fe369df03aaaf0acb) * add commentary * address comments from charles * add query context to example * fix typo * add links * Apply suggestions from code review Co-authored-by: Frank Chen * fix datatype * address feedback * add parameterize to spelling file. the past tense version was already there Co-authored-by: Frank Chen --- docs/querying/sql-jdbc.md | 121 +++++++++++++++++++++++++++++++- docs/tutorials/tutorial-jdbc.md | 31 ++++++++ website/.spelling | 1 + website/sidebars.json | 5 +- 4 files changed, 154 insertions(+), 4 deletions(-) create mode 100644 docs/tutorials/tutorial-jdbc.md diff --git a/docs/querying/sql-jdbc.md b/docs/querying/sql-jdbc.md index 0f041c6fad38..a558637bc110 100644 --- a/docs/querying/sql-jdbc.md +++ b/docs/querying/sql-jdbc.md @@ -29,6 +29,12 @@ sidebar_label: "JDBC driver API" You can make [Druid SQL](./sql.md) queries using the [Avatica JDBC driver](https://calcite.apache.org/avatica/downloads/). We recommend using Avatica JDBC driver version 1.17.0 or later. Note that as of the time of this writing, Avatica 1.17.0, the latest version, does not support passing connection string parameters from the URL to Druid, so you must pass them using a `Properties` object. Once you've downloaded the Avatica client jar, add it to your classpath and use the connect string `jdbc:avatica:remote:url=http://BROKER:8082/druid/v2/sql/avatica/`. +When using the JDBC connector for the [examples](#examples) or in general, it's helpful to understand the parts of the connect string stored in the `url` variable: + + - `jdbc:avatica:remote:url=` is prepended to the hostname and port. + - The hostname and port number for your Druid deployment depends on whether you want to connect to the Router or a specific Broker. For more information, see [Connection stickiness](#connection-stickiness). In the case of the quickstart deployment, the hostname and port are `http://localhost:8888`, which connects to the Router running on your local machine. + - The SQL endpoint in Druid for the Avatica driver is `/druid/v2/sql/avatica/`. + Example code: ```java @@ -51,6 +57,8 @@ try (Connection connection = DriverManager.getConnection(url, connectionProperti } ``` +For a runnable example that includes a query that you might run, see [Examples](#examples). + It is also possible to use a protocol buffers JDBC connection with Druid, this offer reduced bloat and potential performance improvements for larger result sets. To use it apply the following connection url instead, everything else remains the same ``` @@ -60,12 +68,12 @@ String url = "jdbc:avatica:remote:url=http://localhost:8082/druid/v2/sql/avatica > The protobuf endpoint is also known to work with the official [Golang Avatica driver](https://github.com/apache/calcite-avatica-go) Table metadata is available over JDBC using `connection.getMetaData()` or by querying the -["INFORMATION_SCHEMA" tables](sql-metadata-tables.md). +["INFORMATION_SCHEMA" tables](sql-metadata-tables.md). For an example of this, see [Get the metadata for a datasource](#get-the-metadata-for-a-datasource). ## Connection stickiness Druid's JDBC server does not share connection state between Brokers. This means that if you're using JDBC and have -multiple Druid Brokers, you should either connect to a specific Broker, or use a load balancer with sticky sessions +multiple Druid Brokers, you should either connect to a specific Broker or use a load balancer with sticky sessions enabled. The Druid Router process provides connection stickiness when balancing JDBC requests, and can be used to achieve the necessary stickiness even with a normal non-sticky load balancer. Please see the [Router](../design/router.md) documentation for more details. @@ -82,3 +90,112 @@ statement.setString(1, "abc"); statement.setString(2, "def"); final ResultSet resultSet = statement.executeQuery(); ``` + +## Examples + + + +The following section contains two complete samples that use the JDBC connector: + +- [Get the metadata for a datasource](#get-the-metadata-for-a-datasource) shows you how to query the `INFORMATION_SCHEMA` to get metadata like column names. +- [Query data](#query-data) runs a select query against the datasource. + +You can try out these examples after verifying that you meet the [prerequisites](#prerequisites). + +For more information about the connection options, see [Client Reference](https://calcite.apache.org/avatica/docs/client_reference.html). + +### Prerequisites + +Make sure you meet the following requirements before trying these examples: + +- A supported Java version, such as Java 8 + +- [Avatica JDBC driver](https://calcite.apache.org/avatica/downloads/). You can add the JAR to your `CLASSPATH` directly or manage it externally, such as through Maven and a `pom.xml` file. + +- An available Druid instance. You can use the `micro-quickstart` configuration described in [Quickstart (local)](../tutorials/index.md). The examples assume that you are using the quickstart, so no authentication or authorization is expected unless explicitly mentioned. + +- The example `wikipedia` datasource from the quickstart is loaded on your Druid instance. If you have a different datasource loaded, you can still try these examples. You'll have to update the table name and column names to match your datasource. + +### Get the metadata for a datasource + +Metadata, such as column names, is available either through the [`INFORMATION_SCHEMA`](../querying/sql-metadata-tables.md) table or through `connect.getMetaData()`. The following example uses the `INFORMATION_SCHEMA` table to retrieve and print the list of column names for the `wikipedia` datasource that you loaded during a previous tutorial. + +```java +import java.sql.*; +import java.util.Properties; + +public class JdbcListColumns { + + public static void main(String args[]) throws SQLException + { + // Connect to /druid/v2/sql/avatica/ on your Router. + // You can connect to a Broker but must configure connection stickiness if you do. + String url = "jdbc:avatica:remote:url=http://localhost:8888/druid/v2/sql/avatica/"; + + String query = "SELECT COLUMN_NAME,* FROM INFORMATION_SCHEMA.COLUMNS WHERE TABLE_NAME = 'wikipedia' and TABLE_SCHEMA='druid'"; + // Set any connection context parameters you need here + // Or leave empty for default behavior. + Properties connectionProperties = new Properties(); + + try (Connection connection = DriverManager.getConnection(url, connectionProperties)) { + try ( + final Statement statement = connection.createStatement(); + final ResultSet rs = statement.executeQuery(query) + ) { + while (rs.next()) { + String columnName = rs.getString("COLUMN_NAME"); + System.out.println(columnName); + } + } + } catch (SQLException e) { + throw new RuntimeException(e); + } + + } +} +``` + +### Query data + +Now that you know what columns are available, you can start querying the data. The following example queries the datasource named `wikipedia` for the timestamps and comments from Japan. It also sets the [query context parameter](../querying/sql-query-context.md) `sqlTimeZone` . Optionally, you can also parameterize queries by using [dynamic parameters](#dynamic-parameters). + +```java +import java.sql.*; +import java.util.Properties; + +public class JdbcCountryAndTime { + + public static void main(String args[]) throws SQLException + { + // Connect to /druid/v2/sql/avatica/ on your Router. + // You can connect to a Broker but must configure connection stickiness if you do. + String url = "jdbc:avatica:remote:url=http://localhost:8888/druid/v2/sql/avatica/"; + + //The query you want to run. + String query = "SELECT __time, isRobot, countryName, comment FROM wikipedia WHERE countryName='Japan'"; + // Set any connection context parameters you need here + // Or leave empty for default behavior. + Properties connectionProperties = new Properties(); + connectionProperties.setProperty("sqlTimeZone", "America/Los_Angeles"); + + try (Connection connection = DriverManager.getConnection(url, connectionProperties)) { + try ( + final Statement statement = connection.createStatement(); + final ResultSet rs = statement.executeQuery(query) + ) { + while (rs.next()) { + Timestamp timeStamp = rs.getTimestamp("__time"); + String comment = rs.getString("comment"); + System.out.println(timeStamp); + System.out.println(comment); + } + } + } catch (SQLException e) { + throw new RuntimeException(e); + } + + } +} +````` + + diff --git a/docs/tutorials/tutorial-jdbc.md b/docs/tutorials/tutorial-jdbc.md new file mode 100644 index 000000000000..28cee144c3b0 --- /dev/null +++ b/docs/tutorials/tutorial-jdbc.md @@ -0,0 +1,31 @@ +--- +id: tutorial-jdbc +title: "Tutorial: Using the JDBC driver to query Druid" +sidebar_label: JDBC connector +--- + + + +Redirecting you to the JDBC connector examples... + + + +Click here if you are not redirected. + diff --git a/website/.spelling b/website/.spelling index d02f4f846799..7c4d471aafce 100644 --- a/website/.spelling +++ b/website/.spelling @@ -379,6 +379,7 @@ non-nullable noop numerics numShards +parameterize objectGlob parameterized parse_json diff --git a/website/sidebars.json b/website/sidebars.json index 1ebc2140279f..e0fcbf040358 100644 --- a/website/sidebars.json +++ b/website/sidebars.json @@ -13,7 +13,7 @@ "tutorials/tutorial-batch-hadoop", "tutorials/tutorial-query", "tutorials/tutorial-rollup", - "tutorials/tutorial-sketches-theta", + "tutorials/tutorial-sketches-theta", "tutorials/tutorial-retention", "tutorials/tutorial-update-data", "tutorials/tutorial-compaction", @@ -22,7 +22,8 @@ "tutorials/tutorial-transform-spec", "tutorials/docker", "tutorials/tutorial-kerberos-hadoop", - "tutorials/tutorial-msq-convert-spec" + "tutorials/tutorial-msq-convert-spec", + "tutorials/tutorial-jdbc" ], "Design": [ "design/architecture", From 2fdcfffe40cb400db4015fd4ae37478b4779512e Mon Sep 17 00:00:00 2001 From: Vadim Ogievetsky Date: Wed, 30 Nov 2022 19:21:07 -0800 Subject: [PATCH 077/141] don't render duration if aggregated (#13455) --- web-console/src/views/ingestion-view/ingestion-view.tsx | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/web-console/src/views/ingestion-view/ingestion-view.tsx b/web-console/src/views/ingestion-view/ingestion-view.tsx index e25d9be28def..e350ad498578 100644 --- a/web-console/src/views/ingestion-view/ingestion-view.tsx +++ b/web-console/src/views/ingestion-view/ingestion-view.tsx @@ -917,7 +917,8 @@ ORDER BY width: 80, filterable: false, className: 'padded', - Cell({ value, original }) { + Cell({ value, original, aggregated }) { + if (aggregated) return ''; if (value > 0) { return formatDuration(value); } From 2f3b97194fb579d33bd6063c3d3c2bba1bb6f84f Mon Sep 17 00:00:00 2001 From: Adarsh Sanjeev Date: Thu, 1 Dec 2022 10:10:04 +0530 Subject: [PATCH 078/141] Fix harcoded version in pom file (#13460) --- integration-tests-ex/cases/pom.xml | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/integration-tests-ex/cases/pom.xml b/integration-tests-ex/cases/pom.xml index 5a35b1300055..57fa522974cc 100644 --- a/integration-tests-ex/cases/pom.xml +++ b/integration-tests-ex/cases/pom.xml @@ -181,7 +181,7 @@ org.apache.druid druid-sql - 26.0.0-SNAPSHOT + ${project.parent.version} org.apache.druid.extensions From 8395273099e82da9cbc1d8b25ed6971fc0085624 Mon Sep 17 00:00:00 2001 From: Adarsh Sanjeev Date: Thu, 1 Dec 2022 10:11:49 +0530 Subject: [PATCH 079/141] Add unit tests for MSQ ingestion faults (#13439) * Add unit tests for MSQ ingestion faults * Resolve build failure * Move test to MSQFaultTest * Rename test --- .../apache/druid/msq/exec/MSQFaultsTest.java | 138 +++++++++++++++++- .../apache/druid/msq/exec/MSQInsertTest.java | 24 --- .../apache/druid/msq/test/MSQTestBase.java | 4 +- 3 files changed, 140 insertions(+), 26 deletions(-) diff --git a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/MSQFaultsTest.java b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/MSQFaultsTest.java index 7ce0a33578b9..51d0c1b83d76 100644 --- a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/MSQFaultsTest.java +++ b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/MSQFaultsTest.java @@ -20,7 +20,17 @@ package org.apache.druid.msq.exec; import com.google.common.collect.ImmutableMap; +import com.google.common.collect.ImmutableSet; +import org.apache.druid.indexing.common.actions.RetrieveUsedSegmentsAction; +import org.apache.druid.indexing.common.actions.SegmentAllocateAction; +import org.apache.druid.java.util.common.Intervals; import org.apache.druid.java.util.common.StringUtils; +import org.apache.druid.msq.indexing.error.InsertCannotAllocateSegmentFault; +import org.apache.druid.msq.indexing.error.InsertCannotBeEmptyFault; +import org.apache.druid.msq.indexing.error.InsertCannotOrderByDescendingFault; +import org.apache.druid.msq.indexing.error.InsertCannotReplaceExistingSegmentFault; +import org.apache.druid.msq.indexing.error.InsertTimeNullFault; +import org.apache.druid.msq.indexing.error.InsertTimeOutOfBoundsFault; import org.apache.druid.msq.indexing.error.TooManyClusteredByColumnsFault; import org.apache.druid.msq.indexing.error.TooManyColumnsFault; import org.apache.druid.msq.indexing.error.TooManyInputFilesFault; @@ -30,7 +40,10 @@ import org.apache.druid.msq.test.MSQTestFileUtils; import org.apache.druid.segment.column.ColumnType; import org.apache.druid.segment.column.RowSignature; +import org.apache.druid.timeline.DataSegment; +import org.apache.druid.timeline.SegmentId; import org.junit.Test; +import org.mockito.Mockito; import java.io.File; import java.io.IOException; @@ -39,8 +52,132 @@ import java.util.stream.Collectors; import java.util.stream.IntStream; +import static org.mockito.ArgumentMatchers.isA; + public class MSQFaultsTest extends MSQTestBase { + @Test + public void testInsertCannotAllocateSegmentFault() + { + RowSignature rowSignature = RowSignature.builder() + .add("__time", ColumnType.LONG) + .add("dim1", ColumnType.STRING) + .add("cnt", ColumnType.LONG).build(); + + // If there is some problem allocating the segment,task action client will return a null value. + Mockito.doReturn(null).when(testTaskActionClient).submit(isA(SegmentAllocateAction.class)); + + testIngestQuery().setSql( + "insert into foo1 select __time, dim1 , count(*) as cnt from foo where dim1 is not null and __time >= TIMESTAMP '2000-01-02 00:00:00' and __time < TIMESTAMP '2000-01-03 00:00:00' group by 1, 2 PARTITIONED by day clustered by dim1") + .setExpectedDataSource("foo1") + .setExpectedRowSignature(rowSignature) + .setExpectedMSQFault( + new InsertCannotAllocateSegmentFault( + "foo1", + Intervals.of("2000-01-02T00:00:00.000Z/2000-01-03T00:00:00.000Z") + ) + ) + .verifyResults(); + } + + @Test + public void testInsertCannotBeEmptyFault() + { + RowSignature rowSignature = RowSignature.builder() + .add("__time", ColumnType.LONG) + .add("dim1", ColumnType.STRING) + .add("cnt", ColumnType.LONG).build(); + + //Insert with a condition which results in 0 rows being inserted + testIngestQuery().setSql( + "insert into foo1 select __time, dim1 , count(*) as cnt from foo where dim1 is not null and __time < TIMESTAMP '1971-01-01 00:00:00' group by 1, 2 PARTITIONED by day clustered by dim1") + .setExpectedDataSource("foo1") + .setExpectedRowSignature(rowSignature) + .setExpectedMSQFault(new InsertCannotBeEmptyFault("foo1")) + .verifyResults(); + } + + @Test + public void testInsertCannotOrderByDescendingFault() + { + RowSignature rowSignature = RowSignature.builder() + .add("__time", ColumnType.LONG) + .add("dim1", ColumnType.STRING) + .add("cnt", ColumnType.LONG).build(); + + // Add an DESC clustered by column, which should not be allowed + testIngestQuery().setSql( + "insert into foo1 select __time, dim1 , count(*) as cnt from foo where dim1 is not null and __time < TIMESTAMP '2000-01-02 00:00:00' group by 1, 2 PARTITIONED by day clustered by dim1 DESC") + .setExpectedDataSource("foo1") + .setExpectedRowSignature(rowSignature) + .setExpectedMSQFault(new InsertCannotOrderByDescendingFault("d1")) + .verifyResults(); + } + + @Test + public void testInsertCannotReplaceExistingSegmentFault() + { + RowSignature rowSignature = RowSignature.builder() + .add("__time", ColumnType.LONG) + .add("dim1", ColumnType.STRING) + .add("cnt", ColumnType.LONG).build(); + + // Create a datasegment which lies partially outside the generated segment + DataSegment existingDataSegment = DataSegment.builder() + .interval(Intervals.of("2001-01-01T/2003-01-04T")) + .size(50) + .version("1").dataSource("foo1") + .build(); + Mockito.doReturn(ImmutableSet.of(existingDataSegment)).when(testTaskActionClient).submit(isA(RetrieveUsedSegmentsAction.class)); + + testIngestQuery().setSql( + "replace into foo1 overwrite where __time >= TIMESTAMP '2000-01-01 00:00:00' and __time < TIMESTAMP '2002-01-03 00:00:00' select __time, dim1 , count(*) as cnt from foo where dim1 is not null group by 1, 2 PARTITIONED by day clustered by dim1") + .setExpectedDataSource("foo1") + .setExpectedRowSignature(rowSignature) + .setExpectedMSQFault(new InsertCannotReplaceExistingSegmentFault(existingDataSegment.getId())) + .verifyResults(); + } + + @Test + public void testInsertTimeOutOfBoundsFault() + { + RowSignature rowSignature = RowSignature.builder() + .add("__time", ColumnType.LONG) + .add("dim1", ColumnType.STRING) + .add("cnt", ColumnType.LONG).build(); + + // Add a REPLACE statement which replaces a different partition than the ones which rows are present for. The generated + // partition will be outside the replace interval which should throw an InsertTimeOutOfBoundsFault. + testIngestQuery().setSql( + "replace into foo1 overwrite where __time >= TIMESTAMP '2002-01-02 00:00:00' and __time < TIMESTAMP '2002-01-03 00:00:00' select __time, dim1 , count(*) as cnt from foo where dim1 is not null group by 1, 2 PARTITIONED by day clustered by dim1") + .setExpectedDataSource("foo1") + .setExpectedRowSignature(rowSignature) + .setExpectedMSQFault(new InsertTimeOutOfBoundsFault(Intervals.of("2000-01-02T00:00:00.000Z/2000-01-03T00:00:00.000Z"))) + .verifyResults(); + } + + @Test + public void testInsertTimeNullFault() + { + final RowSignature rowSignature = + RowSignature.builder() + .add("__time", ColumnType.LONG) + .add("dim1", ColumnType.STRING) + .build(); + + testIngestQuery() + .setSql( + "INSERT INTO foo1\n" + + "SELECT TIME_PARSE(dim1) AS __time, dim1 as cnt\n" + + "FROM foo\n" + + "PARTITIONED BY DAY\n" + + "CLUSTERED BY dim1") + .setExpectedDataSource("foo1") + .setExpectedRowSignature(rowSignature) + .setExpectedSegment(ImmutableSet.of(SegmentId.of("foo", Intervals.of("2000-01-01T/P1M"), "test", 0))) + .setExpectedMSQFault(InsertTimeNullFault.instance()) + .verifyResults(); + } @Test public void testInsertWithTooManySegments() throws IOException @@ -211,5 +348,4 @@ public void testTooManyInputFiles() throws IOException .setExpectedMSQFault(new TooManyInputFilesFault(numFiles, Limits.MAX_INPUT_FILES_PER_WORKER, 2)) .verifyResults(); } - } diff --git a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/MSQInsertTest.java b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/MSQInsertTest.java index 940008080436..f54d2fa880cb 100644 --- a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/MSQInsertTest.java +++ b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/exec/MSQInsertTest.java @@ -29,7 +29,6 @@ import org.apache.druid.java.util.common.Intervals; import org.apache.druid.java.util.common.granularity.Granularities; import org.apache.druid.msq.indexing.error.ColumnNameRestrictedFault; -import org.apache.druid.msq.indexing.error.InsertTimeNullFault; import org.apache.druid.msq.indexing.error.RowTooLargeFault; import org.apache.druid.msq.test.MSQTestBase; import org.apache.druid.msq.test.MSQTestFileUtils; @@ -383,29 +382,6 @@ public void testRollUpOnExternalDataSourceWithCompositeKey() throws IOException } - @Test - public void testInsertNullTimestamp() - { - final RowSignature rowSignature = - RowSignature.builder() - .add("__time", ColumnType.LONG) - .add("dim1", ColumnType.STRING) - .build(); - - testIngestQuery() - .setSql( - "INSERT INTO foo1\n" - + "SELECT TIME_PARSE(dim1) AS __time, dim1 as cnt\n" - + "FROM foo\n" - + "PARTITIONED BY DAY\n" - + "CLUSTERED BY dim1") - .setExpectedDataSource("foo1") - .setExpectedRowSignature(rowSignature) - .setExpectedSegment(ImmutableSet.of(SegmentId.of("foo", Intervals.of("2000-01-01T/P1M"), "test", 0))) - .setExpectedMSQFault(InsertTimeNullFault.instance()) - .verifyResults(); - } - @Test public void testInsertWrongTypeTimestamp() { diff --git a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/test/MSQTestBase.java b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/test/MSQTestBase.java index 19b85034ff93..133517dd760e 100644 --- a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/test/MSQTestBase.java +++ b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/test/MSQTestBase.java @@ -223,6 +223,7 @@ public class MSQTestBase extends BaseCalciteQueryTest private static final Logger log = new Logger(MSQTestBase.class); private ObjectMapper objectMapper; private MSQTestOverlordServiceClient indexingServiceClient; + protected MSQTestTaskActionClient testTaskActionClient; private SqlStatementFactory sqlStatementFactory; private IndexIO indexIO; @@ -385,10 +386,11 @@ public String getFormatString() objectMapper = setupObjectMapper(injector); objectMapper.registerModules(sqlModule.getJacksonModules()); + testTaskActionClient = Mockito.spy(new MSQTestTaskActionClient(objectMapper)); indexingServiceClient = new MSQTestOverlordServiceClient( objectMapper, injector, - new MSQTestTaskActionClient(objectMapper), + testTaskActionClient, workerMemoryParameters ); final InProcessViewManager viewManager = new InProcessViewManager(SqlTestFramework.DRUID_VIEW_MACRO_FACTORY); From 7b2132373ffb0ee3c8ba5c578d85e3d0bf34d60c Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Thu, 1 Dec 2022 10:51:53 +0530 Subject: [PATCH 080/141] remove hashref for commands in supervise script --- examples/bin/supervise | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/examples/bin/supervise b/examples/bin/supervise index e9cdca5ce925..29aca9eeddb8 100755 --- a/examples/bin/supervise +++ b/examples/bin/supervise @@ -196,16 +196,16 @@ usage() unless GetOptions( 'kill-timeout|t=i', 'chdir=s', 'svlogd:s', - 'command=s@' => \@commands + 'command=s@' ); -usage() unless ((@commands || $opt{'conf'}) && $opt{'vardir'}); +usage() unless ((@{$opt{'command'}} || $opt{'conf'}) && $opt{'vardir'}); my @config_lines; # get commands to execute either from reading the config file or command line if (not defined $opt{'conf'}) { - @config_lines = @commands + @config_lines = @{$opt{'command'}} } else { @config_lines = read_config_file($opt{'conf'}); } From 7535ec87e90c138322b2d7a8a8eb1a292523fb05 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Thu, 1 Dec 2022 10:53:30 +0530 Subject: [PATCH 081/141] start-druid-main java_opts array is comma separated --- examples/bin/start-druid-main | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/examples/bin/start-druid-main b/examples/bin/start-druid-main index e4e727e04748..17e74ff9d8f5 100644 --- a/examples/bin/start-druid-main +++ b/examples/bin/start-druid-main @@ -369,7 +369,7 @@ def build_mm_task_java_opts_array(memory_type): java_opts_list = TASK_JAVA_OPTS_ARRAY + mem_array for item in java_opts_list: - task_memory += '\"{0}\";'.format(item) + task_memory += '\"{0}\",'.format(item) task_memory = task_memory[:-1] task_memory += ']' From 144042f7c0bf1a645f944cfcbb4779a23f3755a6 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Thu, 1 Dec 2022 11:04:46 +0530 Subject: [PATCH 082/141] update entry point script name in python script --- examples/bin/start-druid-main | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) diff --git a/examples/bin/start-druid-main b/examples/bin/start-druid-main index 17e74ff9d8f5..6889fdb5e8f8 100644 --- a/examples/bin/start-druid-main +++ b/examples/bin/start-druid-main @@ -97,31 +97,31 @@ def print_if_verbose(message): def configure_parser(): parser = argparse.ArgumentParser( - prog='start-druid', + prog='start-druid-auto', formatter_class=argparse.RawTextHelpFormatter, epilog= """ sample usage: - start-druid + start-druid-auto Start up all the services (including zk). - start-druid -m=100g + start-druid-auto -m=100g Start up all the services (including zk) using a total memory of 100GB. - start-druid -m=100g --compute + start-druid-auto -m=100g --compute Compute memory distribution and validate arguments. - start-druid -m=100g -s=broker,router + start-druid-auto -m=100g -s=broker,router Starts a broker and a router, using a total memory of 100GB. - start-druid -m=100g --s=broker,router \\ + start-druid-auto -m=100g --s=broker,router \\ -c=conf/druid/single-server/custom Starts a broker and a router, using a total memory of 100GB. Reads configs for each service (jvm.config, runtime.properties) from respective folders inside the given root config path. - start-druid -s=broker,router \\ + start-druid-auto -s=broker,router \\ -c=conf/druid/single-server/custom Starts a broker and a router service, reading service configs from the given root directory. Calculates memory requirements for each service, if required, using upto 80% of the total system memory. - start-druid -m=100g \\ + start-druid-auto -m=100g \\ -s=broker,router \\ -c=conf/druid/single-server/custom \\ --zk From f5b225f836cc63e87b95ab2d2eff36402931208a Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Thu, 1 Dec 2022 11:10:43 +0530 Subject: [PATCH 083/141] Update help docs --- examples/bin/start-druid-main | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/examples/bin/start-druid-main b/examples/bin/start-druid-main index 6889fdb5e8f8..9e80fd22bb98 100644 --- a/examples/bin/start-druid-main +++ b/examples/bin/start-druid-main @@ -144,7 +144,7 @@ sample usage: 'This directory must contain \'_common\' directory with \n' '\'common.jvm.config\' & \'common.runtime.properties\' files. \n' 'If this argument is not given, config from \n' - 'conf/druid/single-server/quickstart directory is used.\n') + 'conf/druid/auto directory is used.\n') parser.add_argument('--compute', action='store_true', help='Does not start Druid, only displays the memory allocated \n' 'to each service if started with the given total memory.\n') From f6f625ee088e955af01974c3850e095d7d52496a Mon Sep 17 00:00:00 2001 From: abhagraw <99210446+abhagraw@users.noreply.github.com> Date: Thu, 1 Dec 2022 12:13:23 +0530 Subject: [PATCH 084/141] MSQ Reindex IT (#13433) * MSQ Reindex IT * Fixing checkstyle errors * Addressing comments * Addressing comments --- .../msq/AbstractITSQLBasedIngestion.java | 75 +++++++++- .../druid/testsEx/msq/ITMSQReindexTest.java | 83 +++++++++++ .../testsEx/msq/ITSQLBasedBatchIngestion.java | 20 +-- .../json_path_index_queries.json | 0 .../{batch-index => }/msq_inline.sql | 0 .../{batch-index => }/sparse_column_msq.json | 0 .../{batch-index => }/sparse_column_msq.sql | 0 ...edia_combining_firehose_index_queries.json | 141 ++++++++++++++++++ ... wikipedia_combining_input_source_msq.sql} | 27 ++-- .../wikipedia_http_inputsource_msq.sql | 0 .../wikipedia_http_inputsource_queries.json | 0 .../{batch-index => }/wikipedia_index_msq.sql | 0 .../wikipedia_index_queries.json | 0 ...ikipedia_index_queries_with_transform.json | 0 .../wikipedia_index_task_with_transform.sql | 0 .../wikipedia_merge_index_msq.sql | 20 +++ .../wikipedia_merge_index_queries.json | 42 ++++++ .../wikipedia_merge_reindex_msq.sql | 17 +++ .../wikipedia_reindex_msq.sql | 24 +++ .../wikipedia_reindex_queries.json | 66 ++++++++ ...pedia_reindex_queries_with_transforms.json | 79 ++++++++++ .../wikipedia_reindex_with_transform_msq.sql | 27 ++++ .../druid/testing/utils/DataLoaderHelper.java | 6 + .../testing/utils/SqlTestQueryHelper.java | 2 +- 24 files changed, 599 insertions(+), 30 deletions(-) create mode 100644 integration-tests-ex/cases/src/test/java/org/apache/druid/testsEx/msq/ITMSQReindexTest.java rename integration-tests-ex/cases/src/test/resources/multi-stage-query/{batch-index => }/json_path_index_queries.json (100%) rename integration-tests-ex/cases/src/test/resources/multi-stage-query/{batch-index => }/msq_inline.sql (100%) rename integration-tests-ex/cases/src/test/resources/multi-stage-query/{batch-index => }/sparse_column_msq.json (100%) rename integration-tests-ex/cases/src/test/resources/multi-stage-query/{batch-index => }/sparse_column_msq.sql (100%) create mode 100644 integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_combining_firehose_index_queries.json rename integration-tests-ex/cases/src/test/resources/multi-stage-query/{batch-index/wikipedia_merge_index_task.sql => wikipedia_combining_input_source_msq.sql} (65%) rename integration-tests-ex/cases/src/test/resources/multi-stage-query/{batch-index => }/wikipedia_http_inputsource_msq.sql (100%) rename integration-tests-ex/cases/src/test/resources/multi-stage-query/{batch-index => }/wikipedia_http_inputsource_queries.json (100%) rename integration-tests-ex/cases/src/test/resources/multi-stage-query/{batch-index => }/wikipedia_index_msq.sql (100%) rename integration-tests-ex/cases/src/test/resources/multi-stage-query/{batch-index => }/wikipedia_index_queries.json (100%) rename integration-tests-ex/cases/src/test/resources/multi-stage-query/{batch-index => }/wikipedia_index_queries_with_transform.json (100%) rename integration-tests-ex/cases/src/test/resources/multi-stage-query/{batch-index => }/wikipedia_index_task_with_transform.sql (100%) create mode 100644 integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_merge_index_msq.sql create mode 100644 integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_merge_index_queries.json create mode 100644 integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_merge_reindex_msq.sql create mode 100644 integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_reindex_msq.sql create mode 100644 integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_reindex_queries.json create mode 100644 integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_reindex_queries_with_transforms.json create mode 100644 integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_reindex_with_transform_msq.sql diff --git a/integration-tests-ex/cases/src/test/java/org/apache/druid/testsEx/msq/AbstractITSQLBasedIngestion.java b/integration-tests-ex/cases/src/test/java/org/apache/druid/testsEx/msq/AbstractITSQLBasedIngestion.java index 4bb1cdc4783b..66df19864796 100644 --- a/integration-tests-ex/cases/src/test/java/org/apache/druid/testsEx/msq/AbstractITSQLBasedIngestion.java +++ b/integration-tests-ex/cases/src/test/java/org/apache/druid/testsEx/msq/AbstractITSQLBasedIngestion.java @@ -28,6 +28,9 @@ import org.apache.druid.testing.utils.MsqTestQueryHelper; import org.apache.druid.testing.utils.TestQueryHelper; import org.apache.druid.testsEx.indexer.AbstractITBatchIndexTest; +import org.junit.Rule; +import org.junit.rules.TestWatcher; +import org.junit.runner.Description; import java.io.IOException; import java.io.InputStream; @@ -36,6 +39,9 @@ public class AbstractITSQLBasedIngestion { + String DATASOURCE_STRING_IN_TASK = "%%DATASOURCE%%"; + String REINDEX_DATASOURCE_STRING_IN_TASK = "%%REINDEX_DATASOURCE%%"; + public static final Logger LOG = new Logger(TestQueryHelper.class); @Inject private MsqTestQueryHelper msqHelper; @@ -46,6 +52,28 @@ public class AbstractITSQLBasedIngestion @Inject private DataLoaderHelper dataLoaderHelper; + @Rule + public TestWatcher watchman = new TestWatcher() + { + @Override + public void starting(Description d) + { + LOG.info("RUNNING %s", d.getDisplayName()); + } + + @Override + public void failed(Throwable e, Description d) + { + LOG.error("FAILED %s", d.getDisplayName()); + } + + @Override + public void finished(Description d) + { + LOG.info("FINISHED %s", d.getDisplayName()); + } + }; + /** * Reads file as utf-8 string and replace %%DATASOURCE%% with the provide datasource value. */ @@ -62,7 +90,7 @@ protected String getStringFromFileAndReplaceDatasource(String filePath, String d fileString = StringUtils.replace( fileString, - "%%DATASOURCE%%", + DATASOURCE_STRING_IN_TASK, datasource ); @@ -85,10 +113,9 @@ protected void doTestQuery(String queryFilePath, String dataSource) } /** - * Sumits a sqlTask, waits for task completion and then runs test queries on ingested datasource. + * Sumits a sqlTask, waits for task completion. */ - protected void submitTaskAnddoTestQuery(String sqlTask, String queryFilePath, String datasource, - Map msqContext) throws Exception + protected void submitTask(String sqlTask, String datasource, Map msqContext) throws Exception { LOG.info("SqlTask - \n %s", sqlTask); @@ -99,23 +126,55 @@ protected void submitTaskAnddoTestQuery(String sqlTask, String queryFilePath, St ); dataLoaderHelper.waitUntilDatasourceIsReady(datasource); - doTestQuery(queryFilePath, datasource); } /** - * Runs a MSQ ingest sql test. + * Sumits a sqlTask, waits for task completion. + */ + protected void submitTaskFromFile(String sqlFilePath, String datasource, Map msqContext) throws Exception + { + String sqlTask = getStringFromFileAndReplaceDatasource(sqlFilePath, datasource); + submitTask(sqlTask, datasource, msqContext); + } + + /** + * Runs a SQL ingest test. * * @param sqlFilePath path of file containing the sql query. * @param queryFilePath path of file containing the native test queries to be run on the ingested datasource. * @param datasource name of the datasource. %%DATASOURCE%% in the sql and queries will be replaced with this value. * @param msqContext context parameters to be passed with MSQ API call. */ - protected void runMSQTaskandTestQueries(String sqlFilePath, String queryFilePath, String datasource, + protected void runMSQTaskandTestQueries(String sqlFilePath, + String queryFilePath, + String datasource, Map msqContext) throws Exception { LOG.info("Starting MSQ test for [%s, %s]", sqlFilePath, queryFilePath); + submitTaskFromFile(sqlFilePath, datasource, msqContext); + doTestQuery(queryFilePath, datasource); + } + + /** + * Runs a reindex SQL ingest test. + * Same as runMSQTaskandTestQueries, but replaces both %%DATASOURCE%% and %%REINDEX_DATASOURCE%% in the SQL Task. + */ + protected void runReindexMSQTaskandTestQueries(String sqlFilePath, + String queryFilePath, + String datasource, + String reindexDatasource, + Map msqContext) throws Exception + { + LOG.info("Starting Reindex MSQ test for [%s, %s]", sqlFilePath, queryFilePath); + String sqlTask = getStringFromFileAndReplaceDatasource(sqlFilePath, datasource); - submitTaskAnddoTestQuery(sqlTask, queryFilePath, datasource, msqContext); + sqlTask = StringUtils.replace( + sqlTask, + REINDEX_DATASOURCE_STRING_IN_TASK, + reindexDatasource + ); + submitTask(sqlTask, reindexDatasource, msqContext); + doTestQuery(queryFilePath, reindexDatasource); } } diff --git a/integration-tests-ex/cases/src/test/java/org/apache/druid/testsEx/msq/ITMSQReindexTest.java b/integration-tests-ex/cases/src/test/java/org/apache/druid/testsEx/msq/ITMSQReindexTest.java new file mode 100644 index 000000000000..26a317ccf148 --- /dev/null +++ b/integration-tests-ex/cases/src/test/java/org/apache/druid/testsEx/msq/ITMSQReindexTest.java @@ -0,0 +1,83 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.testsEx.msq; + +import com.google.inject.Inject; +import junitparams.Parameters; +import junitparams.naming.TestCaseName; +import org.apache.commons.io.FilenameUtils; +import org.apache.curator.shaded.com.google.common.collect.ImmutableMap; +import org.apache.druid.testing.utils.MsqTestQueryHelper; +import org.apache.druid.testsEx.categories.MultiStageQuery; +import org.apache.druid.testsEx.config.DruidTestRunner; +import org.junit.Test; +import org.junit.experimental.categories.Category; +import org.junit.runner.RunWith; + +import java.util.Arrays; +import java.util.List; +import java.util.Map; + +@RunWith(DruidTestRunner.class) +@Category(MultiStageQuery.class) +public class ITMSQReindexTest extends AbstractITSQLBasedIngestion +{ + private static final String MSQ_TASKS_DIR = "/multi-stage-query/"; + + @Inject + private MsqTestQueryHelper msqHelper; + + public static List> test_cases() + { + return Arrays.asList( + Arrays.asList("wikipedia_index_msq.sql", "wikipedia_reindex_msq.sql", "wikipedia_reindex_queries.json"), + Arrays.asList("wikipedia_merge_index_msq.sql", "wikipedia_merge_reindex_msq.sql", "wikipedia_merge_index_queries.json"), + Arrays.asList("wikipedia_index_task_with_transform.sql", "wikipedia_reindex_with_transform_msq.sql", "wikipedia_reindex_queries_with_transforms.json") + ); + + } + + @Test + @Parameters(method = "test_cases") + @TestCaseName("Test_{index} ({0}, {1}, {2})") + public void testMSQDruidInputSource(String sqlFileName, String reIndexSqlFileName, String reIndexQueryFileName) + { + String indexDatasource = FilenameUtils.removeExtension(sqlFileName); + String reindexDatasource = FilenameUtils.removeExtension(reIndexSqlFileName); + Map context = ImmutableMap.of("finalizeAggregations", false, + "maxNumTasks", 5, + "groupByEnableMultiValueUnnesting", false); + try { + submitTaskFromFile(MSQ_TASKS_DIR + sqlFileName, + indexDatasource, + context); + + runReindexMSQTaskandTestQueries(MSQ_TASKS_DIR + reIndexSqlFileName, + MSQ_TASKS_DIR + reIndexQueryFileName, + indexDatasource, + reindexDatasource, + context); + } + catch (Exception e) { + LOG.error(e, "Error while testing [%s, %s, %s]", sqlFileName, reIndexSqlFileName, reIndexQueryFileName); + throw new RuntimeException(e); + } + } +} diff --git a/integration-tests-ex/cases/src/test/java/org/apache/druid/testsEx/msq/ITSQLBasedBatchIngestion.java b/integration-tests-ex/cases/src/test/java/org/apache/druid/testsEx/msq/ITSQLBasedBatchIngestion.java index dbc26d7d4085..9f7754b2e8df 100644 --- a/integration-tests-ex/cases/src/test/java/org/apache/druid/testsEx/msq/ITSQLBasedBatchIngestion.java +++ b/integration-tests-ex/cases/src/test/java/org/apache/druid/testsEx/msq/ITSQLBasedBatchIngestion.java @@ -20,6 +20,7 @@ package org.apache.druid.testsEx.msq; import junitparams.Parameters; +import junitparams.naming.TestCaseName; import org.apache.commons.io.FilenameUtils; import org.apache.curator.shaded.com.google.common.collect.ImmutableMap; import org.apache.druid.testsEx.categories.MultiStageQuery; @@ -35,7 +36,7 @@ @Category(MultiStageQuery.class) public class ITSQLBasedBatchIngestion extends AbstractITSQLBasedIngestion { - private static final String BATCH_INDEX_TASKS_DIR = "/multi-stage-query/batch-index/"; + private static final String MSQ_TASKS_DIR = "/multi-stage-query/"; public static List> test_cases() { @@ -44,7 +45,7 @@ public static List> test_cases() Arrays.asList("sparse_column_msq.sql", "sparse_column_msq.json"), Arrays.asList("wikipedia_http_inputsource_msq.sql", "wikipedia_http_inputsource_queries.json"), Arrays.asList("wikipedia_index_msq.sql", "wikipedia_index_queries.json"), - Arrays.asList("wikipedia_merge_index_task.sql", "wikipedia_index_queries.json"), + Arrays.asList("wikipedia_merge_index_msq.sql", "wikipedia_merge_index_queries.json"), Arrays.asList("wikipedia_index_task_with_transform.sql", "wikipedia_index_queries_with_transform.json") ); @@ -52,16 +53,17 @@ public static List> test_cases() @Test @Parameters(method = "test_cases") + @TestCaseName("Test_{index} ({0}, {1})") public void testSQLBasedBatchIngestion(String sqlFileName, String queryFileName) { try { - runMSQTaskandTestQueries(BATCH_INDEX_TASKS_DIR + sqlFileName, - BATCH_INDEX_TASKS_DIR + queryFileName, - FilenameUtils.removeExtension(sqlFileName), - ImmutableMap.of("finalizeAggregations", false, - "maxNumTasks", 5, - "groupByEnableMultiValueUnnesting", false - )); + runMSQTaskandTestQueries( + MSQ_TASKS_DIR + sqlFileName, + MSQ_TASKS_DIR + queryFileName, + FilenameUtils.removeExtension(sqlFileName), + ImmutableMap.of("finalizeAggregations", false, + "maxNumTasks", 5, + "groupByEnableMultiValueUnnesting", false)); } catch (Exception e) { LOG.error(e, "Error while testing [%s, %s]", sqlFileName, queryFileName); diff --git a/integration-tests-ex/cases/src/test/resources/multi-stage-query/batch-index/json_path_index_queries.json b/integration-tests-ex/cases/src/test/resources/multi-stage-query/json_path_index_queries.json similarity index 100% rename from integration-tests-ex/cases/src/test/resources/multi-stage-query/batch-index/json_path_index_queries.json rename to integration-tests-ex/cases/src/test/resources/multi-stage-query/json_path_index_queries.json diff --git a/integration-tests-ex/cases/src/test/resources/multi-stage-query/batch-index/msq_inline.sql b/integration-tests-ex/cases/src/test/resources/multi-stage-query/msq_inline.sql similarity index 100% rename from integration-tests-ex/cases/src/test/resources/multi-stage-query/batch-index/msq_inline.sql rename to integration-tests-ex/cases/src/test/resources/multi-stage-query/msq_inline.sql diff --git a/integration-tests-ex/cases/src/test/resources/multi-stage-query/batch-index/sparse_column_msq.json b/integration-tests-ex/cases/src/test/resources/multi-stage-query/sparse_column_msq.json similarity index 100% rename from integration-tests-ex/cases/src/test/resources/multi-stage-query/batch-index/sparse_column_msq.json rename to integration-tests-ex/cases/src/test/resources/multi-stage-query/sparse_column_msq.json diff --git a/integration-tests-ex/cases/src/test/resources/multi-stage-query/batch-index/sparse_column_msq.sql b/integration-tests-ex/cases/src/test/resources/multi-stage-query/sparse_column_msq.sql similarity index 100% rename from integration-tests-ex/cases/src/test/resources/multi-stage-query/batch-index/sparse_column_msq.sql rename to integration-tests-ex/cases/src/test/resources/multi-stage-query/sparse_column_msq.sql diff --git a/integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_combining_firehose_index_queries.json b/integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_combining_firehose_index_queries.json new file mode 100644 index 000000000000..302d2fea284e --- /dev/null +++ b/integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_combining_firehose_index_queries.json @@ -0,0 +1,141 @@ +[ + { + "description": "timeseries, 1 agg, all", + "query": { + "queryType": "timeBoundary", + "dataSource": "%%DATASOURCE%%" + }, + "expectedResults": [ + { + "timestamp": "2013-08-31T01:02:33.000Z", + "result": { + "minTime": "2013-08-31T01:02:33.000Z", + "maxTime": "2013-09-01T18:22:39.000Z" + } + } + ] + }, + { + "description": "timeseries, datasketch aggs, all", + "query": { + "queryType": "timeseries", + "dataSource": "%%DATASOURCE%%", + "granularity": "day", + "intervals": [ + "2013-09-01T00:00/2013-09-02T00:00" + ], + "filter": null, + "aggregations": [ + { + "type": "HLLSketchMerge", + "name": "approxCountHLL", + "fieldName": "HLLSketchBuild", + "lgK": 12, + "tgtHllType": "HLL_4", + "round": true + }, + { + "type": "thetaSketch", + "name": "approxCountTheta", + "fieldName": "thetaSketch", + "size": 16384, + "shouldFinalize": true, + "isInputThetaSketch": false, + "errorBoundsStdDev": null + }, + { + "type": "quantilesDoublesSketch", + "name": "quantilesSketch", + "fieldName": "quantilesDoublesSketch", + "k": 128 + } + ] + }, + "expectedResults": [ + { + "timestamp": "2013-09-01T00:00:00.000Z", + "result": { + "quantilesSketch": 6, + "approxCountTheta": 6.0, + "approxCountHLL": 6 + } + } + ] + }, + { + "description": "having spec on post aggregation", + "query": { + "queryType": "groupBy", + "dataSource": "%%DATASOURCE%%", + "granularity": "day", + "dimensions": [ + "page" + ], + "filter": { + "type": "selector", + "dimension": "language", + "value": "zh" + }, + "aggregations": [ + { + "type": "count", + "name": "rows" + }, + { + "type": "longSum", + "fieldName": "added", + "name": "added_count" + } + ], + "postAggregations": [ + { + "type": "arithmetic", + "name": "added_count_times_ten", + "fn": "*", + "fields": [ + { + "type": "fieldAccess", + "name": "added_count", + "fieldName": "added_count" + }, + { + "type": "constant", + "name": "const", + "value": 10 + } + ] + } + ], + "having": { + "type": "greaterThan", + "aggregation": "added_count_times_ten", + "value": 9000 + }, + "intervals": [ + "2013-08-31T00:00/2013-09-01T00:00" + ] + }, + "expectedResults": [ + { + "version": "v1", + "timestamp": "2013-08-31T00:00:00.000Z", + "event": { + "added_count_times_ten": 9050.0, + "page": "Crimson Typhoon", + "added_count": 905, + "rows": 1 + } + }, + { + "version": "v1", + "timestamp": "2013-08-31T00:00:00.000Z", + "event": { + "added_count_times_ten": 9770.0, + "page": "Gypsy Danger", + "added_count": 977, + "rows": 1 + } + } + ] + } +] \ No newline at end of file diff --git a/integration-tests-ex/cases/src/test/resources/multi-stage-query/batch-index/wikipedia_merge_index_task.sql b/integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_combining_input_source_msq.sql similarity index 65% rename from integration-tests-ex/cases/src/test/resources/multi-stage-query/batch-index/wikipedia_merge_index_task.sql rename to integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_combining_input_source_msq.sql index a8160aa9055b..212487d5bf03 100644 --- a/integration-tests-ex/cases/src/test/resources/multi-stage-query/batch-index/wikipedia_merge_index_task.sql +++ b/integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_combining_input_source_msq.sql @@ -1,13 +1,8 @@ -REPLACE INTO "%%DATASOURCE%%" OVERWRITE ALL -WITH "source" AS (SELECT * FROM TABLE( - EXTERN( - '{"type":"local","baseDir":"/resources/data/batch_index/json","filter":"wikipedia_index_data*"}', - '{"type":"json"}', - '[{"name":"timestamp","type":"string"},{"name":"page","type":"string"},{"name":"language","type":"string"},{"name":"user","type":"string"},{"name":"unpatrolled","type":"string"},{"name":"newPage","type":"string"},{"name":"robot","type":"string"},{"name":"anonymous","type":"string"},{"name":"namespace","type":"string"},{"name":"continent","type":"string"},{"name":"country","type":"string"},{"name":"region","type":"string"},{"name":"city","type":"string"},{"name":"added","type":"double"},{"name":"deleted","type":"double"},{"name":"delta","type":"double"}]' - ) -)) +INSERT INTO "%%REINDEX_DATASOURCE%%" + +with externalSource as ( SELECT - TIME_FLOOR(CASE WHEN CAST("timestamp" AS BIGINT) > 0 THEN MILLIS_TO_TIMESTAMP(CAST("timestamp" AS BIGINT)) ELSE TIME_PARSE("timestamp") END, 'PT1S') AS __time, +TIME_FLOOR(CASE WHEN CAST("timestamp" AS BIGINT) > 0 THEN MILLIS_TO_TIMESTAMP(CAST("timestamp" AS BIGINT)) ELSE TIME_PARSE("timestamp") END, 'PT1S') AS __time, "page", "language", "user", @@ -20,7 +15,6 @@ SELECT "country", "region", "city", - "timestamp", COUNT(*) AS "count", SUM("added") AS "added", SUM("deleted") AS "deleted", @@ -28,6 +22,15 @@ SELECT APPROX_COUNT_DISTINCT_DS_THETA("user") AS "thetaSketch", DS_QUANTILES_SKETCH("delta") AS "quantilesDoublesSketch", APPROX_COUNT_DISTINCT_DS_HLL("user") AS "HLLSketchBuild" -FROM "source" -GROUP BY 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14 +FROM TABLE( + EXTERN( + '{"type":"local","baseDir":"/resources/indexer","filter":"wikipedia_combining_index_data.json"}', + '{"type":"json"}', + '[{"name":"timestamp","type":"string"},{"name":"page","type":"string"},{"name":"language","type":"string"},{"name":"user","type":"string"},{"name":"unpatrolled","type":"string"},{"name":"newPage","type":"string"},{"name":"robot","type":"string"},{"name":"anonymous","type":"string"},{"name":"namespace","type":"string"},{"name":"continent","type":"string"},{"name":"country","type":"string"},{"name":"region","type":"string"},{"name":"city","type":"string"},{"name":"added","type":"double"},{"name":"deleted","type":"double"},{"name":"delta","type":"double"}]' + )) GROUP BY 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13 + + UNION ALL (SELECT * FROM "%%DATASOURCE%%" WHERE TIMESTAMP '2013-08-31' <= "__time" AND "__time" < TIMESTAMP '2013-09-01') +) + +SELECT * FROM externalSource PARTITIONED BY DAY \ No newline at end of file diff --git a/integration-tests-ex/cases/src/test/resources/multi-stage-query/batch-index/wikipedia_http_inputsource_msq.sql b/integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_http_inputsource_msq.sql similarity index 100% rename from integration-tests-ex/cases/src/test/resources/multi-stage-query/batch-index/wikipedia_http_inputsource_msq.sql rename to integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_http_inputsource_msq.sql diff --git a/integration-tests-ex/cases/src/test/resources/multi-stage-query/batch-index/wikipedia_http_inputsource_queries.json b/integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_http_inputsource_queries.json similarity index 100% rename from integration-tests-ex/cases/src/test/resources/multi-stage-query/batch-index/wikipedia_http_inputsource_queries.json rename to integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_http_inputsource_queries.json diff --git a/integration-tests-ex/cases/src/test/resources/multi-stage-query/batch-index/wikipedia_index_msq.sql b/integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_index_msq.sql similarity index 100% rename from integration-tests-ex/cases/src/test/resources/multi-stage-query/batch-index/wikipedia_index_msq.sql rename to integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_index_msq.sql diff --git a/integration-tests-ex/cases/src/test/resources/multi-stage-query/batch-index/wikipedia_index_queries.json b/integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_index_queries.json similarity index 100% rename from integration-tests-ex/cases/src/test/resources/multi-stage-query/batch-index/wikipedia_index_queries.json rename to integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_index_queries.json diff --git a/integration-tests-ex/cases/src/test/resources/multi-stage-query/batch-index/wikipedia_index_queries_with_transform.json b/integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_index_queries_with_transform.json similarity index 100% rename from integration-tests-ex/cases/src/test/resources/multi-stage-query/batch-index/wikipedia_index_queries_with_transform.json rename to integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_index_queries_with_transform.json diff --git a/integration-tests-ex/cases/src/test/resources/multi-stage-query/batch-index/wikipedia_index_task_with_transform.sql b/integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_index_task_with_transform.sql similarity index 100% rename from integration-tests-ex/cases/src/test/resources/multi-stage-query/batch-index/wikipedia_index_task_with_transform.sql rename to integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_index_task_with_transform.sql diff --git a/integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_merge_index_msq.sql b/integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_merge_index_msq.sql new file mode 100644 index 000000000000..1e04e38436b5 --- /dev/null +++ b/integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_merge_index_msq.sql @@ -0,0 +1,20 @@ +REPLACE INTO "%%DATASOURCE%%" OVERWRITE ALL +WITH "source" AS (SELECT * FROM TABLE( + EXTERN( + '{"type":"local","baseDir":"/resources/data/batch_index/json","filter":"wikipedia_index_data*"}', + '{"type":"json"}', + '[{"name":"timestamp","type":"string"},{"name":"continent","type":"string"},{"name":"added","type":"double"},{"name":"deleted","type":"double"},{"name":"delta","type":"double"},{"name":"user","type":"string"}]' + ) +)) +SELECT + TIME_FLOOR(CASE WHEN CAST("timestamp" AS BIGINT) > 0 THEN MILLIS_TO_TIMESTAMP(CAST("timestamp" AS BIGINT)) ELSE TIME_PARSE("timestamp") END, 'P1D') AS __time, + "continent", + COUNT(*) AS "count", + SUM("added") AS "added", + SUM("deleted") AS "deleted", + SUM("delta") AS "delta", + EARLIEST("user", 128) AS "first_user", + LATEST("user", 128) AS "last_user" +FROM "source" +GROUP BY 1, 2 +PARTITIONED BY DAY \ No newline at end of file diff --git a/integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_merge_index_queries.json b/integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_merge_index_queries.json new file mode 100644 index 000000000000..0439b5fdca14 --- /dev/null +++ b/integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_merge_index_queries.json @@ -0,0 +1,42 @@ +[ + { + "description": "groupby, stringFirst/stringLast rollup aggs, all", + "query":{ + "queryType" : "groupBy", + "dataSource": "%%DATASOURCE%%", + "granularity":"day", + "dimensions":[ + "continent" + ], + "intervals":[ + "2013-08-31T00:00/2013-09-01T00:00" + ], + "filter":{ + "type":"selector", + "dimension":"continent", + "value":"Asia" + }, + "aggregations":[ + { + "type": "stringFirst", + "name": "earliest_user", + "fieldName": "first_user" + }, + { + "type":"stringLast", + "name":"latest_user", + "fieldName":"last_user" + } + ] + }, + "expectedResults":[ { + "version" : "v1", + "timestamp" : "2013-08-31T00:00:00.000Z", + "event" : { + "continent":"Asia", + "earliest_user":"masterYi", + "latest_user":"stringer" + } + } ] + } +] \ No newline at end of file diff --git a/integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_merge_reindex_msq.sql b/integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_merge_reindex_msq.sql new file mode 100644 index 000000000000..67417409ca97 --- /dev/null +++ b/integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_merge_reindex_msq.sql @@ -0,0 +1,17 @@ +REPLACE INTO "%%REINDEX_DATASOURCE%%" OVERWRITE ALL +WITH "source" AS ( + SELECT * + FROM "%%DATASOURCE%%" + WHERE TIMESTAMP '2013-08-31' <= "__time" AND "__time" < TIMESTAMP '2013-09-01' +) +SELECT + "__time", + "continent", + SUM("added") AS "added", + SUM("deleted") AS "deleted", + SUM("delta") AS "delta", + EARLIEST("first_user", 128) AS "first_user", + LATEST("last_user", 128) AS "last_user" +FROM "source" +GROUP BY 1, 2 +PARTITIONED BY DAY \ No newline at end of file diff --git a/integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_reindex_msq.sql b/integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_reindex_msq.sql new file mode 100644 index 000000000000..e4685c63bff9 --- /dev/null +++ b/integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_reindex_msq.sql @@ -0,0 +1,24 @@ +REPLACE INTO "%%REINDEX_DATASOURCE%%" OVERWRITE ALL +WITH "source" AS ( + SELECT * + FROM "%%DATASOURCE%%" + WHERE TIMESTAMP '2013-08-31' <= "__time" AND "__time" < TIMESTAMP '2013-09-01' +) +SELECT + "__time", + "page", + "language", + "user", + "unpatrolled", + "newPage", + "anonymous", + "namespace", + "country", + "region", + "city", + SUM("added") AS "added", + SUM("deleted") AS "deleted", + SUM("delta") AS "delta" +FROM "source" +GROUP BY 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 +PARTITIONED BY DAY \ No newline at end of file diff --git a/integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_reindex_queries.json b/integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_reindex_queries.json new file mode 100644 index 000000000000..38d09adfa965 --- /dev/null +++ b/integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_reindex_queries.json @@ -0,0 +1,66 @@ +[ + { + "description": "timeseries, 1 agg, all", + "query":{ + "queryType" : "timeBoundary", + "dataSource": "%%DATASOURCE%%" + }, + "expectedResults":[ + { + "timestamp" : "2013-08-31T01:02:33.000Z", + "result" : { + "minTime" : "2013-08-31T01:02:33.000Z", + "maxTime" : "2013-08-31T12:41:27.000Z" + } + } + ] + }, + + { + "description":"having spec on post aggregation", + "query":{ + "queryType":"groupBy", + "dataSource":"%%DATASOURCE%%", + "granularity":"day", + "dimensions":[ + "page" + ], + "filter":{ + "type":"selector", + "dimension":"language", + "value":"zh" + }, + "aggregations":[ + { + "type":"longSum", + "fieldName":"added", + "name":"added_count" + } + ], + "postAggregations": [ + { + "type":"arithmetic", + "name":"added_count_times_ten", + "fn":"*", + "fields":[ + {"type":"fieldAccess", "name":"added_count", "fieldName":"added_count"}, + {"type":"constant", "name":"const", "value":10} + ] + } + ], + "having":{"type":"greaterThan", "aggregation":"added_count_times_ten", "value":9000}, + "intervals":[ + "2013-08-31T00:00/2013-09-01T00:00" + ] + }, + "expectedResults":[ { + "version" : "v1", + "timestamp" : "2013-08-31T00:00:00.000Z", + "event" : { + "added_count_times_ten" : 9050.0, + "page" : "Crimson Typhoon", + "added_count" : 905 + } + } ] + } +] diff --git a/integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_reindex_queries_with_transforms.json b/integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_reindex_queries_with_transforms.json new file mode 100644 index 000000000000..fc8a181e7249 --- /dev/null +++ b/integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_reindex_queries_with_transforms.json @@ -0,0 +1,79 @@ +[ + { + "description": "timeseries, 1 agg, all", + "query":{ + "queryType" : "timeBoundary", + "dataSource": "%%DATASOURCE%%" + }, + "expectedResults":[ + { + "timestamp" : "2013-08-31T01:02:33.000Z", + "result" : { + "minTime" : "2013-08-31T01:02:33.000Z", + "maxTime" : "2013-08-31T12:41:27.000Z" + } + } + ] + }, + { + "description":"having spec on post aggregation", + "query":{ + "queryType":"groupBy", + "dataSource":"%%DATASOURCE%%", + "granularity":"day", + "dimensions":[ + "newPage", + "city" + ], + "filter":{ + "type":"selector", + "dimension":"language", + "value":"language-zh" + }, + "aggregations":[ + { + "type":"longSum", + "fieldName":"one-plus-triple-added", + "name":"added_count" + }, + { + "type":"longSum", + "fieldName":"double-deleted", + "name":"double_deleted_count" + }, + { + "type":"longSum", + "fieldName":"delta", + "name":"delta_overshadowed" + } + ], + "postAggregations": [ + { + "type":"arithmetic", + "name":"added_count_times_ten", + "fn":"*", + "fields":[ + {"type":"fieldAccess", "name":"added_count", "fieldName":"added_count"}, + {"type":"constant", "name":"const", "value":10} + ] + } + ], + "having":{"type":"greaterThan", "aggregation":"added_count_times_ten", "value":9000}, + "intervals":[ + "2013-08-31T00:00/2013-09-01T00:00" + ] + }, + "expectedResults":[ { + "version" : "v1", + "timestamp" : "2013-08-31T00:00:00.000Z", + "event" : { + "added_count_times_ten" : 27160.0, + "newPage" : "Crimson Typhoon", + "city" : "city-Taiyuan", + "double_deleted_count" : 10, + "delta_overshadowed" : 450, + "added_count" : 2716 + } + } ] + } +] diff --git a/integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_reindex_with_transform_msq.sql b/integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_reindex_with_transform_msq.sql new file mode 100644 index 000000000000..565a9a2bbcbb --- /dev/null +++ b/integration-tests-ex/cases/src/test/resources/multi-stage-query/wikipedia_reindex_with_transform_msq.sql @@ -0,0 +1,27 @@ +REPLACE INTO "%%REINDEX_DATASOURCE%%" OVERWRITE ALL +WITH "source" AS ( + SELECT * + FROM "%%DATASOURCE%%" + WHERE TIMESTAMP '2013-08-31' <= "__time" AND "__time" < TIMESTAMP '2013-09-01' +) +SELECT + "__time", + "language", + "user", + "unpatrolled", + "page", + "page" AS "newPage", + "anonymous", + "namespace", + "country", + "region", + concat('city-', city) AS "city", + SUM("added") AS "added", + SUM("triple-added") AS "triple-added", + SUM("triple-added" + 1) AS "one-plus-triple-added", + SUM("deleted") AS "deleted", + SUM("deleted" * 2) AS "double-deleted", + SUM("delta" / 2) AS "delta" +FROM "source" +GROUP BY 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 +PARTITIONED BY DAY \ No newline at end of file diff --git a/integration-tests/src/main/java/org/apache/druid/testing/utils/DataLoaderHelper.java b/integration-tests/src/main/java/org/apache/druid/testing/utils/DataLoaderHelper.java index 76b4300067ea..692ab962e62d 100644 --- a/integration-tests/src/main/java/org/apache/druid/testing/utils/DataLoaderHelper.java +++ b/integration-tests/src/main/java/org/apache/druid/testing/utils/DataLoaderHelper.java @@ -21,10 +21,12 @@ import com.google.inject.Inject; import org.apache.druid.java.util.common.StringUtils; +import org.apache.druid.java.util.common.logger.Logger; import org.apache.druid.testing.clients.CoordinatorResourceTestClient; public final class DataLoaderHelper { + private static final Logger LOG = new Logger(SqlTestQueryHelper.class); @Inject private SqlTestQueryHelper sqlTestQueryHelper; @@ -33,6 +35,7 @@ public final class DataLoaderHelper public void waitUntilDatasourceIsReady(String datasource) { + LOG.info("Waiting for Segments to load for datasource [%s]", datasource); ITRetryUtil.retryUntilTrue( () -> coordinator.areSegmentsLoaded(datasource), StringUtils.format( @@ -40,10 +43,13 @@ public void waitUntilDatasourceIsReady(String datasource) datasource ) ); + LOG.info("Segments loaded for datasource [%s]", datasource); + LOG.info("Waiting for datasource [%s] to be ready for SQL queries", datasource); ITRetryUtil.retryUntilTrue( () -> sqlTestQueryHelper.isDatasourceLoadedInSQL(datasource), StringUtils.format("Waiting for [%s] to be ready for SQL queries", datasource) ); + LOG.info("Datasource [%s] ready for SQL queries", datasource); } } diff --git a/integration-tests/src/main/java/org/apache/druid/testing/utils/SqlTestQueryHelper.java b/integration-tests/src/main/java/org/apache/druid/testing/utils/SqlTestQueryHelper.java index fe9a3243c5f3..962b4a103d06 100644 --- a/integration-tests/src/main/java/org/apache/druid/testing/utils/SqlTestQueryHelper.java +++ b/integration-tests/src/main/java/org/apache/druid/testing/utils/SqlTestQueryHelper.java @@ -58,7 +58,7 @@ public boolean isDatasourceLoadedInSQL(String datasource) null, null ); - + try { //noinspection unchecked queryClient.query(getQueryURL(broker), query); From 42be20aef9ab9f790820c4c6d664285dac6002fd Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Thu, 1 Dec 2022 12:43:24 +0530 Subject: [PATCH 085/141] documentation changes --- docs/operations/single-server.md | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/docs/operations/single-server.md b/docs/operations/single-server.md index 852e1597a99f..ce7bfe5752c1 100644 --- a/docs/operations/single-server.md +++ b/docs/operations/single-server.md @@ -23,7 +23,10 @@ title: "Single server deployment" --> -Druid includes a set of reference configurations and launch scripts for single-machine deployments: +Druid includes a set of reference configurations and launch scripts for single-machine deployments. +These configuration bundles are located in `conf/druid/single-server/`. + +The `auto` configuration sizes runtime parameters based on available processors and memory. Other configurations include hard-coded runtime parameters for various server sizes. Most users should stick with `auto`. - `auto` (run script: `bin/start-druid-auto`) - `nano-quickstart` (run script: `bin/start-nano-quickstart`) @@ -33,8 +36,6 @@ Druid includes a set of reference configurations and launch scripts for single-m - `large` (run script: `bin/start-single-server-large`) - `xlarge` (run script: `bin/start-single-server-xlarge`) -The `auto` can be used to spin up Druid cluster on any machine. Read more details about this configuration [below](#Druid-auto-start). - The `micro-quickstart` is sized for small machines like laptops and is intended for quick evaluation use-cases. The `nano-quickstart` is an even smaller configuration, targeting a machine with 1 CPU and 4GiB memory. It is meant for limited evaluations in resource constrained environments, such as small Docker containers. From eed55bed984a5789ed269ac1294d55f8037e90cb Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Thu, 1 Dec 2022 12:43:35 +0530 Subject: [PATCH 086/141] docs changes --- docs/operations/single-server.md | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) diff --git a/docs/operations/single-server.md b/docs/operations/single-server.md index ce7bfe5752c1..8ea6e8b5ae55 100644 --- a/docs/operations/single-server.md +++ b/docs/operations/single-server.md @@ -26,8 +26,7 @@ title: "Single server deployment" Druid includes a set of reference configurations and launch scripts for single-machine deployments. These configuration bundles are located in `conf/druid/single-server/`. -The `auto` configuration sizes runtime parameters based on available processors and memory. Other configurations include hard-coded runtime parameters for various server sizes. Most users should stick with `auto`. - +The `auto` configuration sizes runtime parameters based on available processors and memory. Other configurations include hard-coded runtime parameters for various server sizes. Most users should stick with `auto`. Refer below [Druid auto start](#Druid-auto-start) - `auto` (run script: `bin/start-druid-auto`) - `nano-quickstart` (run script: `bin/start-nano-quickstart`) - `micro-quickstart` (run script: `bin/start-micro-quickstart`) @@ -48,7 +47,7 @@ The example configurations run the Druid Coordinator and Overlord together in a While example configurations are provided for very large single machines, at higher scales we recommend running Druid in a [clustered deployment](../tutorials/cluster.md), for fault-tolerance and reduced resource contention. -## Druid auto start +## Druid auto start Setting up a new Druid cluster can sometimes be complicated as there are several runtime properties required by each service. Each process must also be given the appropriate jvm arguments so that the system can perform optimally. From 9063884bf9c874331b23b9ac7906b707a22199b9 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Thu, 1 Dec 2022 15:05:59 +0530 Subject: [PATCH 087/141] update docs --- docs/tutorials/index.md | 4 ++-- .../{start-druid-main => start-druid-main.py} | 24 +++++++++++++++---- 2 files changed, 21 insertions(+), 7 deletions(-) rename examples/bin/{start-druid-main => start-druid-main.py} (95%) diff --git a/docs/tutorials/index.md b/docs/tutorials/index.md index c611fde67cc7..ed749c7efcfe 100644 --- a/docs/tutorials/index.md +++ b/docs/tutorials/index.md @@ -80,13 +80,13 @@ You can view that setting and others in the configuration files in the `conf/dru From the apache-druid-{{DRUIDVERSION}} package root, run the following command: ```bash -./bin/start-druid-main-auto +./bin/start-druid-auto ``` This brings up instances of ZooKeeper and the Druid services: ```bash -$ ./bin/start-druid-main-auto +$ ./bin/start-druid-auto [Tue Nov 29 16:31:06 2022] Starting Apache Druid. [Tue Nov 29 16:31:06 2022] Open http://localhost:8888/ in your browser to access the web console. [Tue Nov 29 16:31:06 2022] Or, if you have enabled TLS, use https on port 9088. diff --git a/examples/bin/start-druid-main b/examples/bin/start-druid-main.py similarity index 95% rename from examples/bin/start-druid-main rename to examples/bin/start-druid-main.py index 9e80fd22bb98..1201437479da 100644 --- a/examples/bin/start-druid-main +++ b/examples/bin/start-druid-main.py @@ -1,5 +1,3 @@ -#!/usr/bin/env python - # Licensed to the Apache Software Foundation (ASF) under one # or more contributor license agreements. See the NOTICE file # distributed with this work for additional information @@ -51,6 +49,7 @@ HISTORICAL = "historical" MIDDLE_MANAGER = "middleManager" TASKS = "tasks" +INDEXER = "indexer" DEFAULT_SERVICES = [ BROKER, @@ -135,8 +134,8 @@ def configure_parser(): 'in the given conf directory. e.g. 500m, 4g, 6g\n') parser.add_argument('--services', '-s', type=str, required=False, help='List of services to be started, subset of \n' - '{broker, router, middleManager, historical, coordinator-overlord}. \n' - 'If the argument is not given, all services \n' + '{broker, router, middleManager, historical, coordinator-overlord, indexer}. \n' + 'If the argument is not given, broker, router, middleManager, historical, coordinator-overlord \n' 'and zookeeper is started. e.g. -sl=broker,historical') parser.add_argument('--config', '-c', type=str, required=False, help='Relative path to the directory containing common and service \n' @@ -290,6 +289,21 @@ def should_compute_memory(config, total_memory, service_list): mm_task_java_opts_property, mm_task_worker_capacity_prop = middle_manager_task_memory_params_present(config) mm_task_property_present = mm_task_java_opts_property or mm_task_worker_capacity_prop + # if indexer has to be run, all the memory related parameters need to be specified + if INDEXER in service_list: + if MIDDLE_MANAGER in service_list: + raise ValueError("one of indexer or middleManager can run") + if total_memory != "": + raise ValueError( + "If service list includes indexer, jvm.config should be specified for " + "each service and memory argument shouldn't be specified") + if jvm_config_count != len(service_list): + raise ValueError("If service list includes indexer, jvm.config should be specified for each service") + for service in service_list: + verify_service_config(service, config) + + return False + # possible error states # 1. memory argument is specified, also jvm.config or middleManger/runtime.properties having # druid.indexer.runner.javaOptsArray or druid.worker.capacity parameters is present @@ -298,7 +312,7 @@ def should_compute_memory(config, total_memory, service_list): # 3. jvm.config present for some but not all services # 4. jvm.config file is present for all services, but it doesn't contain required parameters # 5. lastly, if middleManager is to be started, and it is missing task memory properties - if jvm_config_count > 0 or mm_task_property_present: + if INDEXER in service_list or jvm_config_count > 0 or mm_task_property_present: if total_memory != "": raise ValueError( "If jvm.config for services and/or middleManager configs " From fe5c7733b6ea7bc7c07a76bde817c359ba75edf3 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Thu, 1 Dec 2022 15:07:22 +0530 Subject: [PATCH 088/141] add support for running indexer --- examples/bin/start-druid-auto | 6 +-- examples/conf/druid/auto/indexer/jvm.config | 9 +++++ .../druid/auto/indexer/runtime.properties | 38 +++++++++++++++++++ examples/conf/druid/indexer/jvm.config | 9 +++++ examples/conf/druid/indexer/main.config | 1 + .../conf/druid/indexer/runtime.properties | 38 +++++++++++++++++++ 6 files changed, 98 insertions(+), 3 deletions(-) create mode 100644 examples/conf/druid/auto/indexer/jvm.config create mode 100644 examples/conf/druid/auto/indexer/runtime.properties create mode 100644 examples/conf/druid/indexer/jvm.config create mode 100644 examples/conf/druid/indexer/main.config create mode 100644 examples/conf/druid/indexer/runtime.properties diff --git a/examples/bin/start-druid-auto b/examples/bin/start-druid-auto index b6dc9cfbf57b..c0447e6eb939 100644 --- a/examples/bin/start-druid-auto +++ b/examples/bin/start-druid-auto @@ -23,10 +23,10 @@ WHEREAMI="$(cd "$WHEREAMI" && pwd)" if [ -x "$(command -v python3)" ] then - exec python3 "$WHEREAMI/start-druid-main" "$@" + exec python3 "$WHEREAMI/start-druid-main.py" "$@" elif [ -x "$(command -v python2)" ] then - exec python2 "$WHEREAMI/start-druid-main" "$@" + exec python2 "$WHEREAMI/start-druid-main.py" "$@" else - exec "$WHEREAMI/start-druid-main" "$@" + exec "$WHEREAMI/start-druid-main.py" "$@" fi diff --git a/examples/conf/druid/auto/indexer/jvm.config b/examples/conf/druid/auto/indexer/jvm.config new file mode 100644 index 000000000000..4611a65196a2 --- /dev/null +++ b/examples/conf/druid/auto/indexer/jvm.config @@ -0,0 +1,9 @@ +-server +-Xms4g +-Xmx4g +-XX:MaxDirectMemorySize=4g +-XX:+ExitOnOutOfMemoryError +-Duser.timezone=UTC +-Dfile.encoding=UTF-8 +-Djava.io.tmpdir=var/tmp +-Djava.util.logging.manager=org.apache.logging.log4j.jul.LogManager diff --git a/examples/conf/druid/auto/indexer/runtime.properties b/examples/conf/druid/auto/indexer/runtime.properties new file mode 100644 index 000000000000..b36c9eb42872 --- /dev/null +++ b/examples/conf/druid/auto/indexer/runtime.properties @@ -0,0 +1,38 @@ +# +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. +# + +druid.service=druid/indexer +druid.plaintextPort=8091 + +# Number of tasks per indexer +druid.worker.capacity=4 + +# Task launch parameters +druid.indexer.task.baseTaskDir=var/druid/task + +# HTTP server threads +druid.server.http.numThreads=60 + +# Processing threads and buffers on Indexer +druid.processing.numMergeBuffers=2 +druid.processing.buffer.sizeBytes=100MiB +druid.processing.numThreads=4 + +# Hadoop indexing +druid.indexer.task.hadoopWorkingPath=var/druid/hadoop-tmp diff --git a/examples/conf/druid/indexer/jvm.config b/examples/conf/druid/indexer/jvm.config new file mode 100644 index 000000000000..4611a65196a2 --- /dev/null +++ b/examples/conf/druid/indexer/jvm.config @@ -0,0 +1,9 @@ +-server +-Xms4g +-Xmx4g +-XX:MaxDirectMemorySize=4g +-XX:+ExitOnOutOfMemoryError +-Duser.timezone=UTC +-Dfile.encoding=UTF-8 +-Djava.io.tmpdir=var/tmp +-Djava.util.logging.manager=org.apache.logging.log4j.jul.LogManager diff --git a/examples/conf/druid/indexer/main.config b/examples/conf/druid/indexer/main.config new file mode 100644 index 000000000000..5183399415cf --- /dev/null +++ b/examples/conf/druid/indexer/main.config @@ -0,0 +1 @@ +org.apache.druid.cli.Main server indexer diff --git a/examples/conf/druid/indexer/runtime.properties b/examples/conf/druid/indexer/runtime.properties new file mode 100644 index 000000000000..b36c9eb42872 --- /dev/null +++ b/examples/conf/druid/indexer/runtime.properties @@ -0,0 +1,38 @@ +# +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. +# + +druid.service=druid/indexer +druid.plaintextPort=8091 + +# Number of tasks per indexer +druid.worker.capacity=4 + +# Task launch parameters +druid.indexer.task.baseTaskDir=var/druid/task + +# HTTP server threads +druid.server.http.numThreads=60 + +# Processing threads and buffers on Indexer +druid.processing.numMergeBuffers=2 +druid.processing.buffer.sizeBytes=100MiB +druid.processing.numThreads=4 + +# Hadoop indexing +druid.indexer.task.hadoopWorkingPath=var/druid/hadoop-tmp From 62912bb4f71173fac8eb0caecce0b1bbaa054da6 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Thu, 1 Dec 2022 15:17:24 +0530 Subject: [PATCH 089/141] update supported services list --- examples/bin/start-druid-main.py | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) diff --git a/examples/bin/start-druid-main.py b/examples/bin/start-druid-main.py index 1201437479da..b617786d0125 100644 --- a/examples/bin/start-druid-main.py +++ b/examples/bin/start-druid-main.py @@ -59,6 +59,15 @@ MIDDLE_MANAGER ] +SUPPORTED_SERVICES = [ + BROKER, + ROUTER, + COORDINATOR, + HISTORICAL, + MIDDLE_MANAGER, + INDEXER +] + SERVICE_MEMORY_RATIO = { MIDDLE_MANAGER: 1, ROUTER: 2, @@ -196,7 +205,7 @@ def parse_arguments(args): services = args.services.split(SERVICE_SEPARATOR) for service in services: - if service not in DEFAULT_SERVICES: + if service not in SUPPORTED_SERVICES: raise ValueError('Invalid service name {0}, should be one of {1}'.format(service, DEFAULT_SERVICES)) if service in service_list: From 9b08b04941e947a532e11a4747eb70d28f1e9fbc Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Thu, 1 Dec 2022 15:20:12 +0530 Subject: [PATCH 090/141] update help --- examples/bin/start-druid-main.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/examples/bin/start-druid-main.py b/examples/bin/start-druid-main.py index b617786d0125..d6e26e8aaabe 100644 --- a/examples/bin/start-druid-main.py +++ b/examples/bin/start-druid-main.py @@ -145,7 +145,7 @@ def configure_parser(): help='List of services to be started, subset of \n' '{broker, router, middleManager, historical, coordinator-overlord, indexer}. \n' 'If the argument is not given, broker, router, middleManager, historical, coordinator-overlord \n' - 'and zookeeper is started. e.g. -sl=broker,historical') + 'and zookeeper is started. e.g. -s=broker,historical') parser.add_argument('--config', '-c', type=str, required=False, help='Relative path to the directory containing common and service \n' 'specific properties to be overridden. \n' From 622a7ffb39716d5e3b08ec1211b5b3637ec034ea Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Thu, 1 Dec 2022 15:37:34 +0530 Subject: [PATCH 091/141] Update python.md --- docs/operations/python.md | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/docs/operations/python.md b/docs/operations/python.md index 2bfb8f0a0615..ecbfafa4e7c7 100644 --- a/docs/operations/python.md +++ b/docs/operations/python.md @@ -22,9 +22,9 @@ title: "Python Installation" ~ under the License. --> -Apache Druid startup script requires python. This page has steps to install python and additional libraries required by the startup script. +Apache Druid startup script requires a python interpreter. This page has steps to install python interpreter and required libraries. -## Python installation instruction +## Python interpreter installation instructions ### MacOS MacOS comes with a version of Python2, which has been deprecated. Python2 and Python3 can coexist on the same machine without problems. @@ -45,7 +45,7 @@ Refer [Installing Python 3 on Mac OS X](https://docs.python-guide.org/starting/i - `sudo yum install -y epel-release` - `sudo yum install -y python3-pip` -## Additional libraries +## Required libraries ### psutil - pip3 install psutil From 6da787c67766cc4e32ae2a072ba9104c04443023 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Thu, 1 Dec 2022 15:51:04 +0530 Subject: [PATCH 092/141] remove dir --- examples/conf/druid/indexer/jvm.config | 9 ----- examples/conf/druid/indexer/main.config | 1 - .../conf/druid/indexer/runtime.properties | 38 ------------------- 3 files changed, 48 deletions(-) delete mode 100644 examples/conf/druid/indexer/jvm.config delete mode 100644 examples/conf/druid/indexer/main.config delete mode 100644 examples/conf/druid/indexer/runtime.properties diff --git a/examples/conf/druid/indexer/jvm.config b/examples/conf/druid/indexer/jvm.config deleted file mode 100644 index 4611a65196a2..000000000000 --- a/examples/conf/druid/indexer/jvm.config +++ /dev/null @@ -1,9 +0,0 @@ --server --Xms4g --Xmx4g --XX:MaxDirectMemorySize=4g --XX:+ExitOnOutOfMemoryError --Duser.timezone=UTC --Dfile.encoding=UTF-8 --Djava.io.tmpdir=var/tmp --Djava.util.logging.manager=org.apache.logging.log4j.jul.LogManager diff --git a/examples/conf/druid/indexer/main.config b/examples/conf/druid/indexer/main.config deleted file mode 100644 index 5183399415cf..000000000000 --- a/examples/conf/druid/indexer/main.config +++ /dev/null @@ -1 +0,0 @@ -org.apache.druid.cli.Main server indexer diff --git a/examples/conf/druid/indexer/runtime.properties b/examples/conf/druid/indexer/runtime.properties deleted file mode 100644 index b36c9eb42872..000000000000 --- a/examples/conf/druid/indexer/runtime.properties +++ /dev/null @@ -1,38 +0,0 @@ -# -# Licensed to the Apache Software Foundation (ASF) under one -# or more contributor license agreements. See the NOTICE file -# distributed with this work for additional information -# regarding copyright ownership. The ASF licenses this file -# to you under the Apache License, Version 2.0 (the -# "License"); you may not use this file except in compliance -# with the License. You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, -# software distributed under the License is distributed on an -# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY -# KIND, either express or implied. See the License for the -# specific language governing permissions and limitations -# under the License. -# - -druid.service=druid/indexer -druid.plaintextPort=8091 - -# Number of tasks per indexer -druid.worker.capacity=4 - -# Task launch parameters -druid.indexer.task.baseTaskDir=var/druid/task - -# HTTP server threads -druid.server.http.numThreads=60 - -# Processing threads and buffers on Indexer -druid.processing.numMergeBuffers=2 -druid.processing.buffer.sizeBytes=100MiB -druid.processing.numThreads=4 - -# Hadoop indexing -druid.indexer.task.hadoopWorkingPath=var/druid/hadoop-tmp From cc307e4c29a8c73a4253667db02170809fd5c7ff Mon Sep 17 00:00:00 2001 From: AmatyaAvadhanula Date: Thu, 1 Dec 2022 18:31:08 +0530 Subject: [PATCH 093/141] Fix needless task shutdown on leader switch (#13411) * Fix needless task shutdown on leader switch * Add unit test * Fix style * Fix UTs --- .../kafka/supervisor/KafkaSupervisorTest.java | 48 ++++++++++++++++++- .../supervisor/KinesisSupervisorTest.java | 5 ++ .../supervisor/SeekableStreamSupervisor.java | 17 ++++++- .../SeekableStreamSupervisorStateTest.java | 2 + 4 files changed, 69 insertions(+), 3 deletions(-) diff --git a/extensions-core/kafka-indexing-service/src/test/java/org/apache/druid/indexing/kafka/supervisor/KafkaSupervisorTest.java b/extensions-core/kafka-indexing-service/src/test/java/org/apache/druid/indexing/kafka/supervisor/KafkaSupervisorTest.java index ab14f4dac7d6..28b915d22cd4 100644 --- a/extensions-core/kafka-indexing-service/src/test/java/org/apache/druid/indexing/kafka/supervisor/KafkaSupervisorTest.java +++ b/extensions-core/kafka-indexing-service/src/test/java/org/apache/druid/indexing/kafka/supervisor/KafkaSupervisorTest.java @@ -4150,7 +4150,7 @@ public void testIsTaskCurrent() @Test public void testResumeAllActivelyReadingTasks() throws Exception { - supervisor = getTestableSupervisor(2, 2, true, "PT1H", null, null); + supervisor = getTestableSupervisor(2, 3, true, "PT1H", null, null); // Mock with task based setup for resumeAsync EasyMock.reset(taskClient); addSomeEvents(100); @@ -4195,7 +4195,27 @@ public void testResumeAllActivelyReadingTasks() throws Exception supervisor.getTuningConfig() ); - List tasks = ImmutableList.of(readingTask, publishingTask, pausedTask, failsToResumePausedTask); + KafkaIndexTask waitingTask = createKafkaIndexTask("waitingTask", + DATASOURCE, + 2, + new SeekableStreamStartSequenceNumbers<>(topic, ImmutableMap.of(2, 0L), Collections.emptySet()), + new SeekableStreamEndSequenceNumbers<>(topic, ImmutableMap.of(2, Long.MAX_VALUE)), + null, + null, + supervisor.getTuningConfig() + ); + + KafkaIndexTask pendingTask = createKafkaIndexTask("pendingTask", + DATASOURCE, + 2, + new SeekableStreamStartSequenceNumbers<>(topic, ImmutableMap.of(2, 0L), Collections.emptySet()), + new SeekableStreamEndSequenceNumbers<>(topic, ImmutableMap.of(2, Long.MAX_VALUE)), + null, + null, + supervisor.getTuningConfig() + ); + + List tasks = ImmutableList.of(readingTask, publishingTask, pausedTask, failsToResumePausedTask, waitingTask, pendingTask); Collection taskRunnerWorkItems = ImmutableList.of( new TestTaskRunnerWorkItem(readingTask, null, new TaskLocation("testHost", 1001, -1)), new TestTaskRunnerWorkItem(publishingTask, null, new TaskLocation("testHost", 1002, -1)), @@ -4220,6 +4240,10 @@ public void testResumeAllActivelyReadingTasks() throws Exception .andReturn(Optional.of(TaskStatus.running(pausedTask.getId()))).anyTimes(); EasyMock.expect(taskStorage.getStatus(failsToResumePausedTask.getId())) .andReturn(Optional.of(TaskStatus.running(failsToResumePausedTask.getId()))).anyTimes(); + EasyMock.expect(taskStorage.getStatus(waitingTask.getId())) + .andReturn(Optional.of(TaskStatus.running(waitingTask.getId()))).anyTimes(); + EasyMock.expect(taskStorage.getStatus(pendingTask.getId())) + .andReturn(Optional.of(TaskStatus.running(pendingTask.getId()))).anyTimes(); EasyMock.expect(taskStorage.getTask(readingTask.getId())) .andReturn(Optional.of(readingTask)).anyTimes(); @@ -4229,6 +4253,10 @@ public void testResumeAllActivelyReadingTasks() throws Exception .andReturn(Optional.of(pausedTask)).anyTimes(); EasyMock.expect(taskStorage.getTask(failsToResumePausedTask.getId())) .andReturn(Optional.of(failsToResumePausedTask)).anyTimes(); + EasyMock.expect(taskStorage.getTask(waitingTask.getId())) + .andReturn(Optional.of(waitingTask)).anyTimes(); + EasyMock.expect(taskStorage.getTask(pendingTask.getId())) + .andReturn(Optional.of(pendingTask)).anyTimes(); EasyMock.expect(indexerMetadataStorageCoordinator.retrieveDataSourceMetadata(DATASOURCE)).andReturn( new KafkaDataSourceMetadata( @@ -4237,6 +4265,12 @@ public void testResumeAllActivelyReadingTasks() throws Exception ).anyTimes(); EasyMock.expect(taskQueue.add(EasyMock.anyObject(Task.class))).andReturn(true); + + EasyMock.expect(taskClient.getStartTimeAsync(waitingTask.getId())) + .andReturn(Futures.immediateFuture(null)); + EasyMock.expect(taskClient.getStartTimeAsync(pendingTask.getId())) + .andReturn(Futures.immediateFuture(null)); + EasyMock.expect(taskClient.getStatusAsync(readingTask.getId())) .andReturn(Futures.immediateFuture(Status.READING)); EasyMock.expect(taskClient.getStatusAsync(publishingTask.getId())) @@ -4245,6 +4279,10 @@ public void testResumeAllActivelyReadingTasks() throws Exception .andReturn(Futures.immediateFuture(Status.PAUSED)); EasyMock.expect(taskClient.getStatusAsync(failsToResumePausedTask.getId())) .andReturn(Futures.immediateFuture(Status.PAUSED)); + EasyMock.expect(taskClient.getStatusAsync(waitingTask.getId())) + .andReturn(Futures.immediateFuture(Status.NOT_STARTED)); + EasyMock.expect(taskClient.getStatusAsync(pendingTask.getId())) + .andReturn(Futures.immediateFuture(Status.NOT_STARTED)); EasyMock.expect(taskClient.getEndOffsetsAsync(publishingTask.getId())) .andReturn(Futures.immediateFuture(ImmutableMap.of(0, 0L))); @@ -4258,6 +4296,12 @@ public void testResumeAllActivelyReadingTasks() throws Exception EasyMock.expect(taskClient.getCheckpointsAsync(failsToResumePausedTask.getId(), true)) .andReturn(Futures.immediateFuture(new TreeMap<>())); + EasyMock.expect(taskClient.getCheckpointsAsync(waitingTask.getId(), true)) + .andReturn(Futures.immediateFuture(null)); + + EasyMock.expect(taskClient.getCheckpointsAsync(pendingTask.getId(), true)) + .andReturn(Futures.immediateFuture(null)); + taskRunner.registerListener(EasyMock.anyObject(TaskRunnerListener.class), EasyMock.anyObject(Executor.class)); // Only the active i.e non-publishing tasks are resumed diff --git a/extensions-core/kinesis-indexing-service/src/test/java/org/apache/druid/indexing/kinesis/supervisor/KinesisSupervisorTest.java b/extensions-core/kinesis-indexing-service/src/test/java/org/apache/druid/indexing/kinesis/supervisor/KinesisSupervisorTest.java index 50fbe2f721db..265a9fc144f8 100644 --- a/extensions-core/kinesis-indexing-service/src/test/java/org/apache/druid/indexing/kinesis/supervisor/KinesisSupervisorTest.java +++ b/extensions-core/kinesis-indexing-service/src/test/java/org/apache/druid/indexing/kinesis/supervisor/KinesisSupervisorTest.java @@ -3102,6 +3102,11 @@ public void testNoDataIngestionTasks() throws Exception EasyMock.expect(taskMaster.getTaskQueue()).andReturn(Optional.of(taskQueue)).anyTimes(); EasyMock.expect(taskMaster.getTaskRunner()).andReturn(Optional.of(taskRunner)).anyTimes(); + final Collection workItems = new ArrayList(); + workItems.add(new TestTaskRunnerWorkItem(id1, null, new TaskLocation(id1.getId(), 8100, 8100))); + workItems.add(new TestTaskRunnerWorkItem(id2, null, new TaskLocation(id2.getId(), 8100, 8100))); + workItems.add(new TestTaskRunnerWorkItem(id3, null, new TaskLocation(id3.getId(), 8100, 8100))); + EasyMock.expect(taskRunner.getRunningTasks()).andReturn(workItems); EasyMock.expect(taskStorage.getActiveTasksByDatasource(DATASOURCE)).andReturn(ImmutableList.of(id1, id2, id3)).anyTimes(); EasyMock.expect(taskStorage.getStatus("id1")).andReturn(Optional.of(TaskStatus.running("id1"))).anyTimes(); EasyMock.expect(taskStorage.getStatus("id2")).andReturn(Optional.of(TaskStatus.running("id2"))).anyTimes(); diff --git a/indexing-service/src/main/java/org/apache/druid/indexing/seekablestream/supervisor/SeekableStreamSupervisor.java b/indexing-service/src/main/java/org/apache/druid/indexing/seekablestream/supervisor/SeekableStreamSupervisor.java index 53769f84985a..68cf3bf79646 100644 --- a/indexing-service/src/main/java/org/apache/druid/indexing/seekablestream/supervisor/SeekableStreamSupervisor.java +++ b/indexing-service/src/main/java/org/apache/druid/indexing/seekablestream/supervisor/SeekableStreamSupervisor.java @@ -2014,9 +2014,24 @@ private void resumeAllActivelyReadingTasks() } Map> tasksToResume = new HashMap<>(); + if (activelyReadingTaskGroups.isEmpty()) { + return; + } + // Resume only running tasks and not pending / waiting ones. + if (!taskMaster.getTaskRunner().isPresent()) { + return; + } + Set runningTaskIds = taskMaster.getTaskRunner() + .get() + .getRunningTasks() + .stream() + .map(TaskRunnerWorkItem::getTaskId) + .collect(Collectors.toSet()); for (TaskGroup taskGroup : activelyReadingTaskGroups.values()) { for (String taskId : taskGroup.tasks.keySet()) { - tasksToResume.put(taskId, taskClient.resumeAsync(taskId)); + if (runningTaskIds.contains(taskId)) { + tasksToResume.put(taskId, taskClient.resumeAsync(taskId)); + } } } diff --git a/indexing-service/src/test/java/org/apache/druid/indexing/seekablestream/supervisor/SeekableStreamSupervisorStateTest.java b/indexing-service/src/test/java/org/apache/druid/indexing/seekablestream/supervisor/SeekableStreamSupervisorStateTest.java index ab3f9c5f9f1f..61260221bbee 100644 --- a/indexing-service/src/test/java/org/apache/druid/indexing/seekablestream/supervisor/SeekableStreamSupervisorStateTest.java +++ b/indexing-service/src/test/java/org/apache/druid/indexing/seekablestream/supervisor/SeekableStreamSupervisorStateTest.java @@ -331,6 +331,7 @@ public void testConnectingToStreamFailRecoveryFailRecovery() throws Exception EasyMock.expect(recordSupplier.getPartitionIds(STREAM)).andReturn(ImmutableSet.of(SHARD_ID)).times(3); EasyMock.expect(taskStorage.getActiveTasksByDatasource(DATASOURCE)).andReturn(ImmutableList.of()).anyTimes(); EasyMock.expect(taskQueue.add(EasyMock.anyObject())).andReturn(true).anyTimes(); + EasyMock.expect(taskRunner.getRunningTasks()).andReturn(ImmutableList.of()).anyTimes(); replayAll(); @@ -564,6 +565,7 @@ public void testCreatingTasksFailRecoveryFail() throws Exception EasyMock.expect(taskQueue.add(EasyMock.anyObject())).andThrow(new IllegalStateException(EXCEPTION_MSG)).times(3); EasyMock.expect(taskQueue.add(EasyMock.anyObject())).andReturn(true).times(3); EasyMock.expect(taskQueue.add(EasyMock.anyObject())).andThrow(new IllegalStateException(EXCEPTION_MSG)).times(3); + EasyMock.expect(taskRunner.getRunningTasks()).andReturn(ImmutableList.of()).anyTimes(); replayAll(); From 0677d0ca070effef52b146009f300340f54bec6a Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Thu, 1 Dec 2022 21:11:54 +0530 Subject: [PATCH 094/141] update .spelling --- docs/operations/python.md | 4 ++-- website/.spelling | 11 +++++++++++ 2 files changed, 13 insertions(+), 2 deletions(-) diff --git a/docs/operations/python.md b/docs/operations/python.md index ecbfafa4e7c7..4a4a065b0d24 100644 --- a/docs/operations/python.md +++ b/docs/operations/python.md @@ -31,7 +31,7 @@ MacOS comes with a version of Python2, which has been deprecated. Python2 and Py #### Install the official Python release * Browse to the [Python Downloads Page](https://www.python.org/downloads/) and download the latest version (3.x.x) -* Verify if python3 and pip3 (python package manager) is installed by issuing `python3` and `pip3 -V` commands. +* Verify if Python3 and Pip3 (python package manager) is installed by issuing `python3` and `pip3 -V` commands. #### Install with Homebrew Refer [Installing Python 3 on Mac OS X](https://docs.python-guide.org/starting/install3/osx/) @@ -51,4 +51,4 @@ Refer [Installing Python 3 on Mac OS X](https://docs.python-guide.org/starting/i - pip3 install psutil ### pathlib -- pip3 install pathlib \ No newline at end of file +- pip3 install pathlib diff --git a/website/.spelling b/website/.spelling index c2eb66104ee8..91fd5ff36234 100644 --- a/website/.spelling +++ b/website/.spelling @@ -2294,3 +2294,14 @@ Czechia Zeelund - ../docs/tutorials/docker.md nano +MacOS +python2 +python3 +Python2 +Python3 +pip3 +RHEL +psutil +pathlib +jvm +upto \ No newline at end of file From 138a6de5074e787788d682eb5522775558c9e6a7 Mon Sep 17 00:00:00 2001 From: Jill Osborne Date: Thu, 1 Dec 2022 18:47:32 +0000 Subject: [PATCH 095/141] Update nested columns docs (#13461) * Update nested columns docs (cherry picked from commit 04206c5179e0eb46a30d4113c7332daee46c390d) * Update nested-columns.md (cherry picked from commit 8085ee7217d90e0e3f133985a52ec2e0b0552992) --- docs/ingestion/data-formats.md | 4 ++-- docs/ingestion/schema-design.md | 9 ++++----- docs/querying/nested-columns.md | 20 ++++++++++---------- 3 files changed, 16 insertions(+), 17 deletions(-) diff --git a/docs/ingestion/data-formats.md b/docs/ingestion/data-formats.md index eb08df0cf7a8..557060a5e663 100644 --- a/docs/ingestion/data-formats.md +++ b/docs/ingestion/data-formats.md @@ -606,9 +606,9 @@ For example: ### FlattenSpec -The `flattenSpec` object bridges the gap between potentially nested input data, such as Avro or ORC, and Druid's flat data model. It is an object within the `inputFormat` object. +You can use the `flattenSpec` object to flatten nested data, as an alternative to the Druid [nested columns](../querying/nested-columns.md) feature, and for nested input formats unsupported by the feature. It is an object within the `inputFormat` object. -> If you have nested JSON data, you can ingest and store JSON in an Apache Druid column as a `COMPLEX` data type. See [Nested columns](../querying/nested-columns.md) for more information. +See [Nested columns](../querying/nested-columns.md) for information on ingesting and storing nested data in an Apache Druid column as a `COMPLEX` data type. Configure your `flattenSpec` as follows: diff --git a/docs/ingestion/schema-design.md b/docs/ingestion/schema-design.md index 10e6ea82cd2b..f006e792bc44 100644 --- a/docs/ingestion/schema-design.md +++ b/docs/ingestion/schema-design.md @@ -116,14 +116,13 @@ naturally emitted. It is also useful if you want to combine timeseries and non-t Similar to log aggregation systems, Druid offers inverted indexes for fast searching and filtering. Druid's search capabilities are generally less developed than these systems, and its analytical capabilities are generally more developed. The main data modeling differences between Druid and these systems are that when ingesting data into Druid, -you must be more explicit. Druid columns have types specific upfront and Druid does not, at this time, natively support -nested data. +you must be more explicit. Druid columns have types specific upfront. Tips for modeling log data in Druid: * If you don't know ahead of time what columns you'll want to ingest, use an empty dimensions list to trigger [automatic detection of dimension columns](#schema-less-dimensions). -* If you have nested data, flatten it using a [`flattenSpec`](./ingestion-spec.md#flattenspec). +* If you have nested data, you can ingest it using the [nested columns](../querying/nested-columns.md) feature or flatten it using a [`flattenSpec`](./ingestion-spec.md#flattenspec). * Consider enabling [rollup](./rollup.md) if you have mainly analytical use cases for your log data. This will mean you lose the ability to retrieve individual events from Druid, but you potentially gain substantial compression and query performance boosts. @@ -198,9 +197,9 @@ like `MILLIS_TO_TIMESTAMP`, `TIME_FLOOR`, and others. If you're using native Dru ### Nested dimensions -You can ingest and store nested JSON in a Druid column as a `COMPLEX` data type. See [Nested columns](../querying/nested-columns.md) for more information. +You can ingest and store nested data in a Druid column as a `COMPLEX` data type. See [Nested columns](../querying/nested-columns.md) for more information. -If you want to ingest nested data in a format other than JSON—for example Avro, ORC, and Parquet—you must use the `flattenSpec` object to flatten it. For example, if you have data of the following form: +If you want to ingest nested data in a format unsupported by the nested columns feature, you must use the `flattenSpec` object to flatten it. For example, if you have data of the following form: ```json { "foo": { "bar": 3 } } diff --git a/docs/querying/nested-columns.md b/docs/querying/nested-columns.md index e8dc628c8fc5..77af91ddff89 100644 --- a/docs/querying/nested-columns.md +++ b/docs/querying/nested-columns.md @@ -23,17 +23,17 @@ sidebar_label: Nested columns ~ under the License. --> -> Nested columns is an experimental feature available starting in Apache Druid 24.0. Like most experimental features, functionality documented on this page is subject to change in future releases. However, the COMPLEX column type includes versioning to provide backward compatible support in future releases. We strongly encourage you to experiment with nested columns in your development environment to evaluate that they meet your use case. If so, you can use them in production scenarios. Review the release notes and this page to stay up to date with changes. - Apache Druid supports directly storing nested data structures in `COMPLEX` columns. `COMPLEX` columns store a copy of the structured data in JSON format and specialized internal columns and indexes for nested literal values—STRING, LONG, and DOUBLE types. An optimized [virtual column](./virtual-columns.md#nested-field-virtual-column) allows Druid to read and filter these values at speeds consistent with standard Druid LONG, DOUBLE, and STRING columns. Druid [SQL JSON functions](./sql-json-functions.md) allow you to extract, transform, and create `COMPLEX` values in SQL queries, using the specialized virtual columns where appropriate. You can use the [JSON nested columns functions](../misc/math-expr.md#json-functions) in [native queries](./querying.md) using [expression virtual columns](./virtual-columns.md#expression-virtual-column), and in native ingestion with a [`transformSpec`](../ingestion/ingestion-spec.md#transformspec). You can use the JSON functions in INSERT and REPLACE statements in SQL-based ingestion, or in a `transformSpec` in native ingestion as an alternative to using a [`flattenSpec`](../ingestion/data-formats.md#flattenspec) object to "flatten" nested data for ingestion. +Druid supports directly ingesting nested data with the following formats: JSON, Parquet, Avro, ORC. + ## Example nested data -The examples in this topic use the data in [`nested_example_data.json`](https://static.imply.io/data/nested_example_data.json). The file contains a simple facsimile of an order tracking and shipping table. +The examples in this topic use the JSON data in [`nested_example_data.json`](https://static.imply.io/data/nested_example_data.json). The file contains a simple facsimile of an order tracking and shipping table. When pretty-printed, a sample row in `nested_example_data` looks like this: @@ -63,7 +63,7 @@ When pretty-printed, a sample row in `nested_example_data` looks like this: ## Native batch ingestion -For native batch ingestion, you can use the [JSON nested columns functions](./sql-json-functions.md) to extract nested data as an alternative to using the [`flattenSpec`](../ingestion/data-formats.md#flattenspec) input format. +For native batch ingestion, you can use the [SQL JSON functions](./sql-json-functions.md) to extract nested data as an alternative to using the [`flattenSpec`](../ingestion/data-formats.md#flattenspec) input format. To configure a dimension as a nested data type, specify the `json` type for the dimension in the `dimensions` list in the `dimensionsSpec` property of your ingestion spec. @@ -124,7 +124,7 @@ For example, the following ingestion spec instructs Druid to ingest `shipTo` and ### Transform data during batch ingestion -You can use the [JSON nested columns functions](./sql-json-functions.md) to transform JSON data and reference the transformed data in your ingestion spec. +You can use the [SQL JSON functions](./sql-json-functions.md) to transform nested data and reference the transformed data in your ingestion spec. To do this, define the output name and expression in the `transforms` list in the `transformSpec` object of your ingestion spec. @@ -192,7 +192,7 @@ For example, the following ingestion spec extracts `firstName`, `lastName` and ` ## SQL-based ingestion -To ingest nested data using multi-stage query architecture, specify `COMPLEX` as the value for `type` when you define the row signature—`shipTo` and `details` in the following example ingestion spec: +To ingest nested data using SQL-based ingestion, specify `COMPLEX` as the value for `type` when you define the row signature—`shipTo` and `details` in the following example ingestion spec: ![SQL-based ingestion](../assets/nested-msq-ingestion.png) @@ -297,7 +297,7 @@ The [Kafka tutorial](../tutorials/tutorial-kafka.md) guides you through the step ### Transform data during SQL-based ingestion -You can use the [JSON nested columns functions](./sql-json-functions.md) to transform JSON data in your ingestion query. +You can use the [SQL JSON functions](./sql-json-functions.md) to transform nested data in your ingestion query. For example, the following ingestion query is the SQL-based version of the [previous batch example](#transform-data-during-batch-ingestion)—it extracts `firstName`, `lastName`, and `address` from `shipTo` and creates a composite JSON object containing `product`, `details`, and `department`. @@ -326,7 +326,7 @@ PARTITIONED BY ALL ## Ingest a JSON string as COMPLEX -If your source data uses a string representation of your JSON column, you can still ingest the data as `COMPLEX` as follows: +If your source data contains serialized JSON strings, you can ingest the data as `COMPLEX` as follows: - During native batch ingestion, call the `parse_json` function in a `transform` object in the `transformSpec`. - During SQL-based ingestion, use the PARSE_JSON keyword within your SELECT statement to transform the string values to JSON. - If you are concerned that your data may not contain valid JSON, you can use `try_parse_json` for native batch or `TRY_PARSE_JSON` for SQL-based ingestion. For cases where the column does not contain valid JSON, Druid inserts a null value. @@ -563,7 +563,7 @@ In addition to `JSON_VALUE`, Druid offers a number of operators that focus on tr - `PARSE_JSON` - `TO_JSON_STRING` -These functions are primarily intended for use with the multi-stage query architecture to transform data during insert operations, but they also work in traditional Druid SQL queries. Because most of these functions output JSON objects, they have the same limitations when used in traditional Druid queries as interacting with the JSON objects directly. +These functions are primarily intended for use with SQL-based ingestion to transform data during insert operations, but they also work in traditional Druid SQL queries. Because most of these functions output JSON objects, they have the same limitations when used in traditional Druid queries as interacting with the JSON objects directly. #### Example query: Return results in a JSON object @@ -663,7 +663,7 @@ Before you start using the nested columns feature, consider the following known - Directly using `COMPLEX` columns and expressions is not well integrated into the Druid query engine. It can result in errors or undefined behavior when grouping and filtering, and when you use `COMPLEX` objects as inputs to aggregators. As a workaround, consider using `TO_JSON_STRING` to coerce the values to strings before you perform these operations. - Directly using array-typed outputs from `JSON_KEYS` and `JSON_PATHS` is moderately supported by the Druid query engine. You can group on these outputs, and there are a number of array expressions that can operate on these values, such as `ARRAY_CONCAT_AGG`. However, some operations are not well defined for use outside array-specific functions, such as filtering using `=` or `IS NULL`. - Input validation for JSON SQL operators is currently incomplete, which sometimes results in undefined behavior or unhelpful error messages. -- Ingesting JSON columns with a very complex nested structure is potentially an expensive operation and may require you to tune ingestion tasks and/or cluster parameters to account for increased memory usage or overall task run time. When you tune your ingestion configuration, treat each nested literal field inside a JSON object as a flattened top-level Druid column. +- Ingesting data with a very complex nested structure is potentially an expensive operation and may require you to tune ingestion tasks and/or cluster parameters to account for increased memory usage or overall task run time. When you tune your ingestion configuration, treat each nested literal field inside an object as a flattened top-level Druid column. ## Further reading From 43802354a6e4a92348f981af2d0d67c9cc06d94d Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Fri, 2 Dec 2022 01:32:01 +0530 Subject: [PATCH 096/141] Remove dependency on psutil and pathlib --- examples/bin/start-druid-main.py | 103 ++++++++++++++++++++++++------- 1 file changed, 80 insertions(+), 23 deletions(-) diff --git a/examples/bin/start-druid-main.py b/examples/bin/start-druid-main.py index d6e26e8aaabe..657d71f6fdaa 100644 --- a/examples/bin/start-druid-main.py +++ b/examples/bin/start-druid-main.py @@ -17,10 +17,10 @@ import sys import os -import psutil -import pathlib import multiprocessing import argparse +import subprocess +import platform BASE_CONFIG_PATH = "conf/druid/auto" @@ -168,18 +168,30 @@ def configure_parser(): return parser +def is_file(path): + return os.path.isfile(path) + + +def is_dir(path): + return os.path.isdir(path) + + +def resolve_path(path): + return os.path.abspath(path) + + def validate_common_jvm_args(config): - if pathlib.Path('{0}/_common/common.jvm.config'.format(config)).is_file() is False: + if is_file('{0}/_common/common.jvm.config'.format(config)) is False: raise ValueError('_common/common.jvm.config file is missing in the root config, ' 'check {0}/_common directory'.format(BASE_CONFIG_PATH)) def validate_common_directory(config): - if pathlib.Path('{0}/_common'.format(config)).is_dir() is False: + if is_dir('{0}/_common'.format(config)) is False: raise ValueError( '_common directory is missing in the root config, check {0}/_common directory'.format(BASE_CONFIG_PATH)) - if pathlib.Path('{0}/_common/common.runtime.properties'.format(config)).is_file() is False: + if is_file('{0}/_common/common.runtime.properties'.format(config)) is False: raise ValueError('_common/common.runtime.properties file is missing in the root config, ' 'check {0}/_common directory'.format(BASE_CONFIG_PATH)) @@ -196,8 +208,8 @@ def parse_arguments(args): if args.zk: zk = True if args.config is not None: - config = pathlib.Path(os.path.join(os.getcwd(), args.config)).resolve() - if config.is_dir() == False: + config = resolve_path(os.path.join(os.getcwd(), args.config)) + if is_dir(config) is False: raise ValueError('config {0} not found'.format(config)) if args.memory is not None: total_memory = args.memory @@ -225,7 +237,7 @@ def print_startup_config(service_list, config, zk): print_if_verbose('Starting {0}'.format(service_list)) print_if_verbose('Reading config from {0}'.format(config)) if zk: - zk_config = pathlib.Path('{0}/../conf/zk'.format(os.getcwd())).resolve() + zk_config = resolve_path('{0}/../conf/zk'.format(os.getcwd())) print_if_verbose('Starting zk, reading default config from {0}'.format(zk_config)) print_if_verbose('\n') @@ -234,7 +246,7 @@ def middle_manager_task_memory_params_present(config): java_opts_property_present = False worker_capacity_property_present = False - if pathlib.Path('{0}/middleManager/runtime.properties'.format(config)).is_file(): + if is_file('{0}/middleManager/runtime.properties'.format(config)): with open('{0}/middleManager/runtime.properties'.format(config)) as file: for line in file: if line.startswith(TASK_JAVA_OPTS_PROPERTY): @@ -267,7 +279,7 @@ def verify_service_config(service, config): raise ValueError('{0} missing in {1}/jvm.config'.format(params, service)) if service == MIDDLE_MANAGER: - if pathlib.Path('{0}/{1}/runtime.properties'.format(config, service)).is_file() is False: + if is_file('{0}/{1}/runtime.properties'.format(config, service)) is False: raise ValueError('{0}/runtime.properties file is missing in the root config'.format(service)) mm_task_java_opts_property, mm_task_worker_capacity_prop = middle_manager_task_memory_params_present(config) @@ -290,7 +302,7 @@ def should_compute_memory(config, total_memory, service_list): jvm_config_count = 0 for service in service_list: - if pathlib.Path('{0}/{1}/jvm.config'.format(config, service)).is_file(): + if is_file('{0}/{1}/jvm.config'.format(config, service)): jvm_config_count += 1 mm_task_property_present = False @@ -304,8 +316,8 @@ def should_compute_memory(config, total_memory, service_list): raise ValueError("one of indexer or middleManager can run") if total_memory != "": raise ValueError( - "If service list includes indexer, jvm.config should be specified for " - "each service and memory argument shouldn't be specified") + "If service list includes indexer, jvm.config should be specified for " + "each service and memory argument shouldn't be specified") if jvm_config_count != len(service_list): raise ValueError("If service list includes indexer, jvm.config should be specified for each service") for service in service_list: @@ -324,9 +336,9 @@ def should_compute_memory(config, total_memory, service_list): if INDEXER in service_list or jvm_config_count > 0 or mm_task_property_present: if total_memory != "": raise ValueError( - "If jvm.config for services and/or middleManager configs " - "(druid.worker.capacity, druid.indexer.runner.javaOptsArray) is present, " - "memory argument shouldn't be specified") + "If jvm.config for services and/or middleManager configs " + "(druid.worker.capacity, druid.indexer.runner.javaOptsArray) is present, " + "memory argument shouldn't be specified") if jvm_config_count == 0: raise ValueError("middleManger configs (druid.indexer.runner.javaOptsArray or druid.worker.capacity) " "is present in middleManager/runtime.properties, " @@ -341,15 +353,56 @@ def should_compute_memory(config, total_memory, service_list): return jvm_config_count == 0 and mm_task_property_present is False +def estimate_memory_linux(): + mems = {} + + def get_procfs_path(): + return sys.modules['psutil'].PROCFS_PATH + + def open_binary(fname): + FILE_READ_BUFFER_SIZE = 32 * 1024 + return open(fname, "rb", buffering=FILE_READ_BUFFER_SIZE) + + with open_binary('%s/meminfo' % get_procfs_path()) as f: + for line in f: + fields = line.split() + mems[fields[0]] = int(fields[1]) * 1024 + + return mems[b'MemTotal:'] + + +def estimate_memory_osx(): + p1 = subprocess.Popen(['sysctl', '-a'], stdout=subprocess.PIPE) + p2 = subprocess.check_output(['grep', 'hw.memsize'], stdin=p1.stdout) + p2 = p2.decode('utf-8') + fields = p2.split(':') + + mem = int(fields[1]) / (1024 * 1024) + + return mem + + def compute_system_memory(): - system_memory = psutil.virtual_memory().total # mem in bytes - memory_for_druid = int(system_memory / (1024 * 1024)) - return memory_for_druid + operating_system = platform.system() + print_if_verbose('operating system is {0}'.format(operating_system)) + + system_memory = None + + try: + if operating_system == 'Darwin': + system_memory = estimate_memory_osx() + elif operating_system == 'Linux': + system_memory = estimate_memory_linux() + except (Exception) as error: + print(error) + raise ValueError('Please specify memory argument') + + return system_memory def convert_total_memory_string(memory): try: - if memory == "": + if memory == '': computed_memory = compute_system_memory() return computed_memory elif memory.endswith(MEM_MB_SUFFIX): @@ -417,7 +470,8 @@ def build_memory_config(service, allocated_memory): if service == TASKS: memory_type, task_count, task_memory = compute_tasks_memory(allocated_memory) java_opts_array = build_mm_task_java_opts_array(memory_type) - return ['-D{0}={1}'.format(TASK_WORKER_CAPACITY_PROPERTY, task_count), java_opts_array], task_memory * task_count + return ['-D{0}={1}'.format(TASK_WORKER_CAPACITY_PROPERTY, task_count), + java_opts_array], task_memory * task_count else: heap_memory = HEAP_TO_TOTAL_MEM_RATIO.get(service) * allocated_memory direct_memory = int(allocated_memory - heap_memory) @@ -479,10 +533,12 @@ def distribute_memory(services, total_memory): return service_memory_config + def append_command(commands, command): commands.append('--command') commands.append(command) + def build_supervise_script_arguments(service_list, service_memory_config, config, zk): commands = [] commands.append('supervise') @@ -514,7 +570,8 @@ def build_supervise_script_arguments(service_list, service_memory_config, config '{0}{1} bin/run-druid {1} {2} \'{3}\' \'{4} {5}\'' .format(prefix, service, config, jvm_args, task_count, task_memory)) else: - append_command(commands, '{0}{1} bin/run-druid {1} {2} \'{3}\''.format(prefix, service, config, jvm_args)) + append_command(commands, + '{0}{1} bin/run-druid {1} {2} \'{3}\''.format(prefix, service, config, jvm_args)) print_if_verbose('Supervise script args:') for item in commands: @@ -538,7 +595,7 @@ def main(): os.chdir(os.path.dirname(sys.argv[0])) if config == "": - config = pathlib.Path('{0}/../{1}'.format(os.getcwd(), BASE_CONFIG_PATH)).resolve() + config = resolve_path('{0}/../{1}'.format(os.getcwd(), BASE_CONFIG_PATH)) validate_common_directory(config) From 77febf8a3e85276f18247cdb6e94314cd6f40ec7 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Fri, 2 Dec 2022 01:36:41 +0530 Subject: [PATCH 097/141] update docs --- docs/operations/python.md | 10 +--------- docs/operations/single-server.md | 6 +++--- docs/tutorials/index.md | 4 ++-- examples/bin/{start-druid-auto => start-druid} | 0 4 files changed, 6 insertions(+), 14 deletions(-) rename examples/bin/{start-druid-auto => start-druid} (100%) mode change 100644 => 100755 diff --git a/docs/operations/python.md b/docs/operations/python.md index 4a4a065b0d24..14a0f5e7bcd8 100644 --- a/docs/operations/python.md +++ b/docs/operations/python.md @@ -22,7 +22,7 @@ title: "Python Installation" ~ under the License. --> -Apache Druid startup script requires a python interpreter. This page has steps to install python interpreter and required libraries. +Apache Druid startup script requires a python interpreter. ## Python interpreter installation instructions @@ -44,11 +44,3 @@ Refer [Installing Python 3 on Mac OS X](https://docs.python-guide.org/starting/i #### RHEL - `sudo yum install -y epel-release` - `sudo yum install -y python3-pip` - -## Required libraries - -### psutil -- pip3 install psutil - -### pathlib -- pip3 install pathlib diff --git a/docs/operations/single-server.md b/docs/operations/single-server.md index 8ea6e8b5ae55..01a3f5c7d3f6 100644 --- a/docs/operations/single-server.md +++ b/docs/operations/single-server.md @@ -27,7 +27,7 @@ Druid includes a set of reference configurations and launch scripts for single-m These configuration bundles are located in `conf/druid/single-server/`. The `auto` configuration sizes runtime parameters based on available processors and memory. Other configurations include hard-coded runtime parameters for various server sizes. Most users should stick with `auto`. Refer below [Druid auto start](#Druid-auto-start) -- `auto` (run script: `bin/start-druid-auto`) +- `auto` (run script: `bin/start-druid`) - `nano-quickstart` (run script: `bin/start-nano-quickstart`) - `micro-quickstart` (run script: `bin/start-micro-quickstart`) - `small` (run script: `bin/start-single-server-small`) @@ -51,12 +51,12 @@ While example configurations are provided for very large single machines, at hig Setting up a new Druid cluster can sometimes be complicated as there are several runtime properties required by each service. Each process must also be given the appropriate jvm arguments so that the system can perform optimally. -`start-druid-auto` is a generic launch script capable of starting any set of Druid services on a server. +`start-druid` is a generic launch script capable of starting any set of Druid services on a server. It accepts optional arguments such as list of services, total memory and a config directory to override default jvm arguments and service-specific runtime properties. All other reference configurations (e.g. `micro`, `small`, `xlarge`) can be obtained by passing the correct memory to this script. Druid services will use all processors and upto 80% memory on the system. -For details about possible arguments, run `bin/start-druid-auto --help`. +For details about possible arguments, run `bin/start-druid --help`. The corresponding launch scripts (e.g. `start-micro-quickstart`) are now deprecated. diff --git a/docs/tutorials/index.md b/docs/tutorials/index.md index ed749c7efcfe..9f31be652660 100644 --- a/docs/tutorials/index.md +++ b/docs/tutorials/index.md @@ -80,13 +80,13 @@ You can view that setting and others in the configuration files in the `conf/dru From the apache-druid-{{DRUIDVERSION}} package root, run the following command: ```bash -./bin/start-druid-auto +./bin/start-druid ``` This brings up instances of ZooKeeper and the Druid services: ```bash -$ ./bin/start-druid-auto +$ ./bin/start-druid [Tue Nov 29 16:31:06 2022] Starting Apache Druid. [Tue Nov 29 16:31:06 2022] Open http://localhost:8888/ in your browser to access the web console. [Tue Nov 29 16:31:06 2022] Or, if you have enabled TLS, use https on port 9088. diff --git a/examples/bin/start-druid-auto b/examples/bin/start-druid old mode 100644 new mode 100755 similarity index 100% rename from examples/bin/start-druid-auto rename to examples/bin/start-druid From 1c8bd246ccfb9170589f6839137601468a30523a Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Fri, 2 Dec 2022 01:51:03 +0530 Subject: [PATCH 098/141] Update get_physical_memory method --- examples/bin/start-druid-main.py | 27 ++++++++++++++++----------- 1 file changed, 16 insertions(+), 11 deletions(-) diff --git a/examples/bin/start-druid-main.py b/examples/bin/start-druid-main.py index 657d71f6fdaa..35bbd4b427d4 100644 --- a/examples/bin/start-druid-main.py +++ b/examples/bin/start-druid-main.py @@ -353,7 +353,7 @@ def should_compute_memory(config, total_memory, service_list): return jvm_config_count == 0 and mm_task_property_present is False -def estimate_memory_linux(): +def get_physical_memory_linux(): mems = {} def get_procfs_path(): @@ -371,18 +371,18 @@ def open_binary(fname): return mems[b'MemTotal:'] -def estimate_memory_osx(): +def get_physical_memory_osx(): p1 = subprocess.Popen(['sysctl', '-a'], stdout=subprocess.PIPE) p2 = subprocess.check_output(['grep', 'hw.memsize'], stdin=p1.stdout) p2 = p2.decode('utf-8') fields = p2.split(':') - mem = int(fields[1]) / (1024 * 1024) + mem = int(int(fields[1]) / (1024 * 1024)) return mem -def compute_system_memory(): +def get_physical_memory(): operating_system = platform.system() print_if_verbose('operating system is {0}'.format(operating_system)) @@ -390,12 +390,11 @@ def compute_system_memory(): try: if operating_system == 'Darwin': - system_memory = estimate_memory_osx() + system_memory = get_physical_memory_osx() elif operating_system == 'Linux': - system_memory = estimate_memory_linux() - except (Exception) as error: - print(error) - raise ValueError('Please specify memory argument') + system_memory = get_physical_memory_linux() + except Exception: + pass return system_memory @@ -403,14 +402,20 @@ def compute_system_memory(): def convert_total_memory_string(memory): try: if memory == '': - computed_memory = compute_system_memory() - return computed_memory + physical_memory = get_physical_memory() + + if physical_memory == None: + raise ValueError('Please specify memory argument') + + return physical_memory elif memory.endswith(MEM_MB_SUFFIX): return int(memory[:-1]) elif memory.endswith(MEM_GB_SUFFIX): return 1024 * int(memory[:-1]) else: raise ValueError('Incorrect format for memory argument, expected format is ') + except ValueError as e: + raise e except Exception: raise ValueError('Incorrect format for memory argument, expected format is ') From 9e495b36e016b8088bc6f9e71ee87cf6ae6d024a Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Fri, 2 Dec 2022 01:53:42 +0530 Subject: [PATCH 099/141] Update help docs --- examples/bin/start-druid-main.py | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) diff --git a/examples/bin/start-druid-main.py b/examples/bin/start-druid-main.py index 35bbd4b427d4..e47b92c77b4b 100644 --- a/examples/bin/start-druid-main.py +++ b/examples/bin/start-druid-main.py @@ -105,31 +105,31 @@ def print_if_verbose(message): def configure_parser(): parser = argparse.ArgumentParser( - prog='start-druid-auto', + prog='start-druid', formatter_class=argparse.RawTextHelpFormatter, epilog= """ sample usage: - start-druid-auto + start-druid Start up all the services (including zk). - start-druid-auto -m=100g + start-druid -m=100g Start up all the services (including zk) using a total memory of 100GB. - start-druid-auto -m=100g --compute + start-druid -m=100g --compute Compute memory distribution and validate arguments. - start-druid-auto -m=100g -s=broker,router + start-druid -m=100g -s=broker,router Starts a broker and a router, using a total memory of 100GB. - start-druid-auto -m=100g --s=broker,router \\ + start-druid -m=100g --s=broker,router \\ -c=conf/druid/single-server/custom Starts a broker and a router, using a total memory of 100GB. Reads configs for each service (jvm.config, runtime.properties) from respective folders inside the given root config path. - start-druid-auto -s=broker,router \\ + start-druid -s=broker,router \\ -c=conf/druid/single-server/custom Starts a broker and a router service, reading service configs from the given root directory. Calculates memory requirements for each service, if required, using upto 80% of the total system memory. - start-druid-auto -m=100g \\ + start-druid -m=100g \\ -s=broker,router \\ -c=conf/druid/single-server/custom \\ --zk From e04af93a7964b9e864e4a18f550edd24e365a5dc Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Fri, 2 Dec 2022 01:58:05 +0530 Subject: [PATCH 100/141] update docs --- docs/tutorials/index.md | 4 ++-- docs/tutorials/tutorial-batch-hadoop.md | 2 +- website/.spelling | 2 +- 3 files changed, 4 insertions(+), 4 deletions(-) diff --git a/docs/tutorials/index.md b/docs/tutorials/index.md index 9f31be652660..5672fe4e3e6d 100644 --- a/docs/tutorials/index.md +++ b/docs/tutorials/index.md @@ -104,7 +104,7 @@ the Druid root directory, apache-druid-{{DRUIDVERSION}}. Each service writes to At any time, you can revert Druid to its original, post-installation state by deleting the entire `var` directory. You may want to do this, for example, between Druid tutorials or after experimentation, to start with a fresh instance. -To stop Druid at any time, use CTRL+C in the terminal. This exits the `bin/start-druid-auto` script and terminates all Druid processes. +To stop Druid at any time, use CTRL+C in the terminal. This exits the `bin/start-druid` script and terminates all Druid processes. ## Open the web console @@ -223,4 +223,4 @@ See the following topics for more information: * [Tutorial: Load stream data from Apache Kafka](./tutorial-kafka.md) to load streaming data from a Kafka topic. * [Extensions](../development/extensions.md) for details on Druid extensions. -Remember that after stopping Druid services, you can start clean next time by deleting the `var` directory from the Druid root directory and running the `bin/start-druid-auto` script again. You may want to do this before using other data ingestion tutorials, since they use the same Wikipedia datasource. +Remember that after stopping Druid services, you can start clean next time by deleting the `var` directory from the Druid root directory and running the `bin/start-druid` script again. You may want to do this before using other data ingestion tutorials, since they use the same Wikipedia datasource. diff --git a/docs/tutorials/tutorial-batch-hadoop.md b/docs/tutorials/tutorial-batch-hadoop.md index e8ea8a19052f..ee0861b210d2 100644 --- a/docs/tutorials/tutorial-batch-hadoop.md +++ b/docs/tutorials/tutorial-batch-hadoop.md @@ -196,7 +196,7 @@ druid.indexer.logs.directory=/druid/indexing-logs Once the Hadoop .xml files have been copied to the Druid cluster and the segment/log storage configuration has been updated to use HDFS, the Druid cluster needs to be restarted for the new configurations to take effect. -If the cluster is still running, CTRL-C to terminate the `bin/start-druid-auto` script, and re-run it to bring the Druid services back up. +If the cluster is still running, CTRL-C to terminate the `bin/start-druid` script, and re-run it to bring the Druid services back up. ## Load batch data diff --git a/website/.spelling b/website/.spelling index 91fd5ff36234..fc88d5f84af2 100644 --- a/website/.spelling +++ b/website/.spelling @@ -2304,4 +2304,4 @@ RHEL psutil pathlib jvm -upto \ No newline at end of file +upto From ca9d5ea5cb98882a8f3462d1c8c78813df52a783 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Fri, 2 Dec 2022 12:41:58 +0530 Subject: [PATCH 101/141] update method to get physical memory on python --- examples/bin/start-druid-main.py | 22 +++++----------------- 1 file changed, 5 insertions(+), 17 deletions(-) diff --git a/examples/bin/start-druid-main.py b/examples/bin/start-druid-main.py index e47b92c77b4b..51ca126713f5 100644 --- a/examples/bin/start-druid-main.py +++ b/examples/bin/start-druid-main.py @@ -354,21 +354,9 @@ def should_compute_memory(config, total_memory, service_list): def get_physical_memory_linux(): - mems = {} - - def get_procfs_path(): - return sys.modules['psutil'].PROCFS_PATH - - def open_binary(fname): - FILE_READ_BUFFER_SIZE = 32 * 1024 - return open(fname, "rb", buffering=FILE_READ_BUFFER_SIZE) - - with open_binary('%s/meminfo' % get_procfs_path()) as f: - for line in f: - fields = line.split() - mems[fields[0]] = int(fields[1]) * 1024 - - return mems[b'MemTotal:'] + mem_bytes = os.sysconf('SC_PAGE_SIZE') * os.sysconf('SC_PHYS_PAGES') + mem_mbs = int(mem_bytes / (1024 * 1024)) + return mem_mbs def get_physical_memory_osx(): @@ -377,9 +365,9 @@ def get_physical_memory_osx(): p2 = p2.decode('utf-8') fields = p2.split(':') - mem = int(int(fields[1]) / (1024 * 1024)) + mem_mbs = int(int(fields[1]) / (1024 * 1024)) - return mem + return mem_mbs def get_physical_memory(): From 01fb4b20573d6f2e75f7bfc382f5d73bf19c225d Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Fri, 2 Dec 2022 12:42:06 +0530 Subject: [PATCH 102/141] udpate spelling --- website/.spelling | 1 + 1 file changed, 1 insertion(+) diff --git a/website/.spelling b/website/.spelling index fc88d5f84af2..451797e6862b 100644 --- a/website/.spelling +++ b/website/.spelling @@ -2300,6 +2300,7 @@ python3 Python2 Python3 pip3 +Pip3 RHEL psutil pathlib From d0c3b21859a811a6da1d5b8f39c7bb29a14fae76 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Fri, 2 Dec 2022 12:49:09 +0530 Subject: [PATCH 103/141] update .spelling --- website/.spelling | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) diff --git a/website/.spelling b/website/.spelling index 451797e6862b..2637f62b1ca8 100644 --- a/website/.spelling +++ b/website/.spelling @@ -421,6 +421,10 @@ programmatically proto proxied proxyConfig +python2 +python3 +Python2 +Python3 QPS quantile quantiles @@ -2294,15 +2298,13 @@ Czechia Zeelund - ../docs/tutorials/docker.md nano +- ../docs/operations/python.md MacOS -python2 -python3 -Python2 -Python3 pip3 Pip3 RHEL psutil pathlib +- ../docs/operatons/single-server.md jvm upto From c516320d08d0ea8f5381fcdb4d055c88165569f1 Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Fri, 2 Dec 2022 14:41:49 +0530 Subject: [PATCH 104/141] minor change --- examples/bin/start-druid-main.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/examples/bin/start-druid-main.py b/examples/bin/start-druid-main.py index 51ca126713f5..03d5c51de90e 100644 --- a/examples/bin/start-druid-main.py +++ b/examples/bin/start-druid-main.py @@ -392,7 +392,7 @@ def convert_total_memory_string(memory): if memory == '': physical_memory = get_physical_memory() - if physical_memory == None: + if physical_memory is None: raise ValueError('Please specify memory argument') return physical_memory From 30498c1f98346050a9121bae113ace4fd4059302 Mon Sep 17 00:00:00 2001 From: Tejaswini Bandlamudi <96047043+tejaswini-imply@users.noreply.github.com> Date: Fri, 2 Dec 2022 15:06:31 +0530 Subject: [PATCH 105/141] Update gha & travis checks (#13412) * update static-checks GHA to run sequentially remove static-checks from travis.yml move docs, web-console, packaging checks from travis to GHA * nit * nit * groups all checks, runs on 8, 11, 17 jdks * nit * adds license info * update permissions on scripts folder * nit * nit * fix packaging check * changes naming, cleans repo before license checks * simulate failure * bump up license checks * test license checks failure * test license checks failure * test license checks failure * verify gha script run exit code * fail fast in case of shell script * verified fail fast in case of shell script --- ...sage.sh => analyze_dependencies_script.sh} | 3 + .github/scripts/license_checks_script.sh | 27 ++ .github/scripts/setup_generate_license.sh | 4 + .github/workflows/static-checks.yml | 253 ++++++++---------- .travis.yml | 144 ---------- 5 files changed, 150 insertions(+), 281 deletions(-) rename .github/scripts/{display_dependency_analysis_error_message.sh => analyze_dependencies_script.sh} (92%) create mode 100755 .github/scripts/license_checks_script.sh diff --git a/.github/scripts/display_dependency_analysis_error_message.sh b/.github/scripts/analyze_dependencies_script.sh similarity index 92% rename from .github/scripts/display_dependency_analysis_error_message.sh rename to .github/scripts/analyze_dependencies_script.sh index 3899d5917b15..c92d90030f71 100755 --- a/.github/scripts/display_dependency_analysis_error_message.sh +++ b/.github/scripts/analyze_dependencies_script.sh @@ -13,6 +13,9 @@ # See the License for the specific language governing permissions and # limitations under the License. +#!bin/bash + +${MVN} ${MAVEN_SKIP} dependency:analyze -DoutputXML=true -DignoreNonCompile=true -DfailOnWarning=true ${HADOOP_PROFILE} || { echo " The dependency analysis has found a dependency that is either: diff --git a/.github/scripts/license_checks_script.sh b/.github/scripts/license_checks_script.sh new file mode 100755 index 000000000000..410ac60375fa --- /dev/null +++ b/.github/scripts/license_checks_script.sh @@ -0,0 +1,27 @@ +# Licensed to the Apache Software Foundation (ASF) under one or more +# contributor license agreements. See the NOTICE file distributed with +# this work for additional information regarding copyright ownership. +# The ASF licenses this file to You under the Apache License, Version 2.0 +# (the "License"); you may not use this file except in compliance with +# the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +#!/bin/bash + +set -e + +./.github/scripts/setup_generate_license.sh +${MVN} apache-rat:check -Prat --fail-at-end \ +-Dorg.slf4j.simpleLogger.log.org.apache.maven.cli.transfer.Slf4jMavenTransferListener=warn \ +-Drat.consoleOutput=true ${HADOOP_PROFILE} +# Generate dependency reports and checks they are valid. +mkdir -p target +distribution/bin/generate-license-dependency-reports.py . target --clean-maven-artifact-transfer --parallel 2 +distribution/bin/check-licenses.py licenses.yaml target/license-reports diff --git a/.github/scripts/setup_generate_license.sh b/.github/scripts/setup_generate_license.sh index 9e6c2eae68c6..71583bfb2b26 100755 --- a/.github/scripts/setup_generate_license.sh +++ b/.github/scripts/setup_generate_license.sh @@ -13,6 +13,10 @@ # See the License for the specific language governing permissions and # limitations under the License. +#!/bin/bash + +set -e + sudo apt-get update && sudo apt-get install python3 -y curl https://bootstrap.pypa.io/pip/3.5/get-pip.py | sudo -H python3 pip3 install wheel # install wheel first explicitly diff --git a/.github/workflows/static-checks.yml b/.github/workflows/static-checks.yml index 7f4437f0129e..a3f2d7c0e3d5 100644 --- a/.github/workflows/static-checks.yml +++ b/.github/workflows/static-checks.yml @@ -35,160 +35,139 @@ env: MAVEN_OPTS: -Xmx3000m jobs: - build: + static-checks: + strategy: + matrix: + java: [ 'jdk8', 'jdk11', 'jdk17' ] runs-on: ubuntu-latest steps: - - uses: actions/checkout@v3 - - name: setup java 8 + - name: checkout branch + uses: actions/checkout@v3 + + - name: set java version + run: | + export jdk=${{ matrix.java }} + echo "java_version=${jdk:3}" >> $GITHUB_ENV + + - name: setup ${{ matrix.java }} uses: actions/setup-java@v3 with: distribution: 'zulu' - java-version: '8' + java-version: ${{ env.java_version }} cache: 'maven' - - run: | + + - name: license checks + if: ${{ matrix.java == 'jdk8' }} + run: ./.github/scripts/license_checks_script.sh + + - name: license checks for hadoop3 + if: ${{ matrix.java == 'jdk8' }} + env: + HADOOP_PROFILE: -Phadoop3 + run: ./.github/scripts/license_checks_script.sh + + - name: analyze dependencies + if: ${{ matrix.java == 'jdk8' }} + run: | + ./.github/scripts/analyze_dependencies_script.sh + + - name: analyze dependencies for hadoop3 + if: ${{ matrix.java == 'jdk8' }} + env: + HADOOP_PROFILE: -Phadoop3 + run: | + ./.github/scripts/analyze_dependencies_script.sh + + - name: packaging check + run: | + ./.github/scripts/setup_generate_license.sh + ${MVN} clean install -Prat -Pdist -Pbundle-contrib-exts --fail-at-end \ + -pl '!benchmarks' ${MAVEN_SKIP} ${MAVEN_SKIP_TESTS} -Dweb.console.skip=false -T1C + + - name: script checks + # who watches the watchers? + if: ${{ matrix.java == 'jdk8' }} + run: ./check_test_suite_test.py + + - name: (openjdk11) strict compilation + if: ${{ matrix.java == 'jdk11' }} + # errorprone requires JDK 11 + # Strict compilation requires more than 2 GB + run: ${MVN} clean -DstrictCompile compile test-compile --fail-at-end ${MAVEN_SKIP} ${MAVEN_SKIP_TESTS} + + - name: maven install + if: ${{ matrix.java == 'jdk8' }} + run: | echo 'Running Maven install...' && ${MVN} clean install -q -ff -pl '!distribution,!:druid-it-image,!:druid-it-cases' ${MAVEN_SKIP} ${MAVEN_SKIP_TESTS} -T1C && ${MVN} install -q -ff -pl 'distribution' ${MAVEN_SKIP} ${MAVEN_SKIP_TESTS} - animal_sniffer_checks: - runs-on: ubuntu-latest - needs: [build] - steps: - - uses: actions/checkout@v3 - - name: setup java 8 - uses: actions/setup-java@v3 - with: - distribution: 'zulu' - java-version: '8' - cache: 'maven' - - run: ${MVN} animal-sniffer:check --fail-at-end + - name: animal sniffer checks + if: ${{ matrix.java == 'jdk8' }} + run: ${MVN} animal-sniffer:check --fail-at-end - checkstyle: - runs-on: ubuntu-latest - needs: [build] - steps: - - uses: actions/checkout@v3 - - name: setup java 8 - uses: actions/setup-java@v3 - with: - distribution: 'zulu' - java-version: '8' - cache: 'maven' - - run: ${MVN} checkstyle:checkstyle --fail-at-end + - name: checkstyle + if: ${{ matrix.java == 'jdk8' }} + run: ${MVN} checkstyle:checkstyle --fail-at-end - enforcer_checks: - runs-on: ubuntu-latest - needs: [build] - steps: - - uses: actions/checkout@v3 - - name: setup java 8 - uses: actions/setup-java@v3 - with: - distribution: 'zulu' - java-version: '8' - cache: 'maven' - - run: ${MVN} enforcer:enforce --fail-at-end + - name: enforcer checks + if: ${{ matrix.java == 'jdk8' }} + run: ${MVN} enforcer:enforce --fail-at-end - forbidden_api_checks: - runs-on: ubuntu-latest - needs: [build] - steps: - - uses: actions/checkout@v3 - - name: setup java 8 - uses: actions/setup-java@v3 - with: - distribution: 'zulu' - java-version: '8' - cache: 'maven' - - run: ${MVN} forbiddenapis:check forbiddenapis:testCheck --fail-at-end + - name: forbidden api checks + if: ${{ matrix.java == 'jdk8' }} + run: ${MVN} forbiddenapis:check forbiddenapis:testCheck --fail-at-end - pmd_checks: - runs-on: ubuntu-latest - needs: [build] - steps: - - uses: actions/checkout@v3 - - name: setup java 8 - uses: actions/setup-java@v3 - with: - distribution: 'zulu' - java-version: '8' - cache: 'maven' - - run: ${MVN} pmd:check --fail-at-end # TODO: consider adding pmd:cpd-check + - name: pmd checks + if: ${{ matrix.java == 'jdk8' }} + run: ${MVN} pmd:check --fail-at-end # TODO: consider adding pmd:cpd-check - spotbugs_checks: - runs-on: ubuntu-latest - needs: [build] - steps: - - uses: actions/checkout@v3 - - name: setup java 8 - uses: actions/setup-java@v3 - with: - distribution: 'zulu' - java-version: '8' - cache: 'maven' - - run: ${MVN} spotbugs:check --fail-at-end -pl '!benchmarks' + - name: spotbugs checks + if: ${{ matrix.java == 'jdk8' }} + run: ${MVN} spotbugs:check --fail-at-end -pl '!benchmarks' - license_checks: - runs-on: ubuntu-latest - needs: [build] - strategy: - matrix: - HADOOP_PROFILE: ['', '-Phadoop3'] - steps: - - uses: actions/checkout@v3 - - name: setup java 8 - uses: actions/setup-java@v3 - with: - distribution: 'zulu' - java-version: '8' - cache: 'maven' - - run: | - ./.github/scripts/setup_generate_license.sh - ${MVN} apache-rat:check -Prat --fail-at-end \ - -Dorg.slf4j.simpleLogger.log.org.apache.maven.cli.transfer.Slf4jMavenTransferListener=warn \ - -Drat.consoleOutput=true ${{ matrix.HADOOP_PROFILE }} - # Generate dependency reports and checks they are valid. - mkdir -p target - distribution/bin/generate-license-dependency-reports.py . target --clean-maven-artifact-transfer --parallel 2 - distribution/bin/check-licenses.py licenses.yaml target/license-reports - - script_checks: - runs-on: ubuntu-latest - steps: - - uses: actions/checkout@v3 - # who watches the watchers? - - run: ./check_test_suite_test.py + - name: intellij inspections + if: ${{ matrix.java == 'jdk8' }} + run: | + docker run --rm \ + -v $(pwd):/project \ + -v ~/.m2:/home/inspect/.m2 \ + ccaominh/intellij-inspect:1.0.0 \ + /project/pom.xml \ + /project/.idea/inspectionProfiles/Druid.xml \ + --levels ERROR \ + --scope JavaInspectionsScope - analyze_dependencies: - runs-on: ubuntu-latest - needs: [build] - strategy: - matrix: - HADOOP_PROFILE: [ '', '-Phadoop3' ] - steps: - - uses: actions/checkout@v3 - - name: setup java 8 - uses: actions/setup-java@v3 + - name: setup node + if: ${{ matrix.java == 'jdk8' }} + uses: actions/setup-node@v3 with: - distribution: 'zulu' - java-version: '8' - cache: 'maven' - - run: |- - ${MVN} ${MAVEN_SKIP} dependency:analyze -DoutputXML=true -DignoreNonCompile=true -DfailOnWarning=true ${{ matrix.HADOOP_PROFILE }} || - ./.github/scripts/display_dependency_analysis_error_message.sh + node-version: 16.17.0 - openjdk11_strict_compilation: - runs-on: ubuntu-latest - steps: - - uses: actions/checkout@v3 - - name: setup java 11 - uses: actions/setup-java@v3 - with: - distribution: 'zulu' - java-version: '11' - cache: 'maven' + - name: docs + if: ${{ matrix.java == 'jdk8' }} + run: | + (cd website && npm install) + cd website + npm run link-lint + npm run spellcheck - # errorprone requires JDK 11 - # Strict compilation requires more than 2 GB - - run: | - ${MVN} clean -DstrictCompile compile test-compile --fail-at-end ${MAVEN_SKIP} ${MAVEN_SKIP_TESTS} + - name: web console + if: ${{ matrix.java == 'jdk8' }} + run: | + ${MVN} test -pl 'web-console' + cd web-console + { for i in 1 2 3; do npm run codecov && break || sleep 15; done } + + - name: web console end-to-end test + if: ${{ matrix.java == 'jdk8' }} + run: | + ./.github/scripts/setup_generate_license.sh + sudo apt-get update && sudo apt-get install python3 -y + curl https://bootstrap.pypa.io/pip/3.5/get-pip.py | sudo -H python3 + pip3 install wheel # install wheel first explicitly + pip3 install --upgrade pyyaml + web-console/script/druid build + web-console/script/druid start + (cd web-console && npm run test-e2e) + web-console/script/druid stop diff --git a/.travis.yml b/.travis.yml index 750d9e7ce3b6..0c7a73290b1d 100644 --- a/.travis.yml +++ b/.travis.yml @@ -70,120 +70,6 @@ stages: jobs: include: - - name: "animal sniffer checks" - stage: Tests - phase 1 - script: ${MVN} animal-sniffer:check --fail-at-end - - - name: "checkstyle" - script: ${MVN} checkstyle:checkstyle --fail-at-end - - - name: "enforcer checks" - script: ${MVN} enforcer:enforce --fail-at-end - - - name: "forbidden api checks" - script: ${MVN} forbiddenapis:check forbiddenapis:testCheck --fail-at-end - - - name: "pmd checks" - script: ${MVN} pmd:check --fail-at-end # TODO: consider adding pmd:cpd-check - - - name: "spotbugs checks" - script: ${MVN} spotbugs:check --fail-at-end -pl '!benchmarks' - - - &license_checks - name: "license checks" - before_script: &setup_generate_license - - sudo apt-get update && sudo apt-get install python3 -y - - curl https://bootstrap.pypa.io/pip/3.5/get-pip.py | sudo -H python3 - - ./check_test_suite.py && travis_terminate 0 || echo 'Continuing setup' - - pip3 install wheel # install wheel first explicitly - - pip3 install --upgrade pyyaml - script: - - > - ${MVN} apache-rat:check -Prat --fail-at-end - -Dorg.slf4j.simpleLogger.log.org.apache.maven.cli.transfer.Slf4jMavenTransferListener=warn - -Drat.consoleOutput=true ${HADOOP_PROFILE} - # Generate dependency reports and checks they are valid. When running on Travis CI, 2 cores are available - # (https://docs.travis-ci.com/user/reference/overview/#virtualisation-environment-vs-operating-system). - - mkdir -p target - - distribution/bin/generate-license-dependency-reports.py . target --clean-maven-artifact-transfer --parallel 2 - - distribution/bin/check-licenses.py licenses.yaml target/license-reports - - - <<: *license_checks - name: "license checks with Hadoop3" - env: - - HADOOP_PROFILE='-Phadoop3' - - - name: "script checks" - install: skip - # who watches the watchers? - script: ./check_test_suite_test.py - - - name: "(openjdk11) strict compilation" - install: skip - # errorprone requires JDK 11 - jdk: openjdk11 - # Strict compilation requires more than 2 GB - script: > - ./check_test_suite.py && travis_terminate 0 || MAVEN_OPTS='-Xmx3000m' ${MVN} clean -DstrictCompile compile test-compile --fail-at-end - ${MAVEN_SKIP} ${MAVEN_SKIP_TESTS} - - - &analyze_dependencies - name: "analyze dependencies" - script: |- - MAVEN_OPTS='-Xmx3000m' ${MVN} ${MAVEN_SKIP} dependency:analyze -DoutputXML=true -DignoreNonCompile=true -DfailOnWarning=true ${HADOOP_PROFILE} || { echo " - - The dependency analysis has found a dependency that is either: - - 1) Used and undeclared: These are available as a transitive dependency but should be explicitly - added to the POM to ensure the dependency version. The XML to add the dependencies to the POM is - shown above. - - 2) Unused and declared: These are not needed and removing them from the POM will speed up the build - and reduce the artifact size. The dependencies to remove are shown above. - - If there are false positive dependency analysis warnings, they can be suppressed: - https://maven.apache.org/plugins/maven-dependency-plugin/analyze-mojo.html#usedDependencies - https://maven.apache.org/plugins/maven-dependency-plugin/examples/exclude-dependencies-from-dependency-analysis.html - - For more information, refer to: - https://maven.apache.org/plugins/maven-dependency-plugin/analyze-mojo.html - - " && false; } - - - <<: *analyze_dependencies - name: "analyze hadoop 3 dependencies" - env: - - HADOOP_PROFILE='-Phadoop3' - - - name: "intellij inspections" - script: > - ./check_test_suite.py && travis_terminate 0 || docker run --rm - -v $(pwd):/project - -v ~/.m2:/home/inspect/.m2 - ccaominh/intellij-inspect:1.0.0 - /project/pom.xml - /project/.idea/inspectionProfiles/Druid.xml - --levels ERROR - --scope JavaInspectionsScope - - - &package - name: "(openjdk8) packaging check" - install: skip - before_script: *setup_generate_license - script: > - MAVEN_OPTS='-Xmx3000m' ${MVN} clean install -Prat -Pdist -Pbundle-contrib-exts --fail-at-end - -pl '!benchmarks' ${MAVEN_SKIP} ${MAVEN_SKIP_TESTS} -Dweb.console.skip=false -T1C - - - <<: *package - name: "(openjdk11) packaging check" - stage: Tests - phase 2 - jdk: openjdk11 - - - <<: *package - name: "(openjdk17) packaging check" - stage: Tests - phase 2 - jdk: openjdk17 - - &test_processing_module name: "(openjdk8) processing module test" stage: Tests - phase 1 @@ -366,36 +252,6 @@ jobs: stage: Tests - phase 2 jdk: openjdk17 - - name: "web console" - install: skip - stage: Tests - phase 1 - script: - - ./check_test_suite.py && travis_terminate 0 || ${MVN} test -pl 'web-console' - after_success: - - (cd web-console && travis_retry npm run codecov) # retry in case of network error - - - name: "web console end-to-end test" - stage: Tests - phase 1 - before_install: *setup_generate_license - install: web-console/script/druid build - before_script: - - ./check_test_suite.py && travis_terminate 0 || echo 'Starting nvm install...' - - nvm install 16.17.0 - - web-console/script/druid start - script: (cd web-console && npm run test-e2e) - after_script: web-console/script/druid stop - - - name: "docs" - stage: Tests - phase 1 - before_script: - - ./check_test_suite.py && travis_terminate 0 || echo 'Starting nvm install...' - - nvm install 16.17.0 - - (cd website && npm install) - script: - - cd website - - npm run link-lint - - npm run spellcheck - - name: "Build and test on ARM64 CPU architecture (1)" stage: Tests - phase 2 arch: arm64-graviton2 From 9aeb55152dc42e224caa0bec1ff978fdf0149a3a Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Fri, 2 Dec 2022 15:29:22 +0530 Subject: [PATCH 106/141] Minor change --- examples/bin/start-druid-main.py | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) diff --git a/examples/bin/start-druid-main.py b/examples/bin/start-druid-main.py index 03d5c51de90e..25d7e93afc59 100644 --- a/examples/bin/start-druid-main.py +++ b/examples/bin/start-druid-main.py @@ -333,7 +333,7 @@ def should_compute_memory(config, total_memory, service_list): # 3. jvm.config present for some but not all services # 4. jvm.config file is present for all services, but it doesn't contain required parameters # 5. lastly, if middleManager is to be started, and it is missing task memory properties - if INDEXER in service_list or jvm_config_count > 0 or mm_task_property_present: + if jvm_config_count > 0 or mm_task_property_present: if total_memory != "": raise ValueError( "If jvm.config for services and/or middleManager configs " @@ -348,9 +348,11 @@ def should_compute_memory(config, total_memory, service_list): for service in service_list: verify_service_config(service, config) + return False + # compute memory only when none of the specified services contains jvm.config, - # if middleManager is to be started it doesn't contain task memory properties - return jvm_config_count == 0 and mm_task_property_present is False + # if middleManager is to be started it shouldn't contain task memory properties + return True def get_physical_memory_linux(): From b76ff16d00b8a77539af18a36ab940fe31a5817b Mon Sep 17 00:00:00 2001 From: Paul Rogers Date: Fri, 2 Dec 2022 09:11:59 -0800 Subject: [PATCH 107/141] SQL test framework extensions (#13426) SQL test framework extensions * Capture planner artifacts: logical plan, etc. * Planner test builder validates the logical plan * Validation for the SQL resut schema (we already have validation for the Druid row signature) * Better Guice integration: properties, reuse Guice modules * Avoid need for hand-coded expr, macro tables * Retire some of the test-specific query component creation * Fix query log hook race condition --- .../druid/guice/DruidSecondaryModule.java | 1 + .../druid/java/util/common/StringUtils.java | 18 + .../org/apache/druid/utils/RuntimeInfo.java | 2 + ...pressedBigDecimalMaxSqlAggregatorTest.java | 10 - ...pressedBigDecimalMinSqlAggregatorTest.java | 9 - ...ressedBigDecimalSqlAggregatorTestBase.java | 18 +- ...pressedBigDecimalSumSqlAggregatorTest.java | 10 - .../sql/TDigestSketchSqlAggregatorTest.java | 25 +- .../datasketches/hll/HllSketchModule.java | 2 - .../quantiles/DoublesSketchModule.java | 1 - .../hll/sql/HllSketchSqlAggregatorTest.java | 51 +-- .../sql/DoublesSketchSqlAggregatorTest.java | 35 +- .../sql/ThetaSketchSqlAggregatorTest.java | 74 +---- .../sql/BloomFilterSqlAggregatorTest.java | 54 ++- .../filter/sql/BloomDimFilterSqlTest.java | 45 +-- ...etsHistogramQuantileSqlAggregatorTest.java | 27 +- .../sql/QuantileSqlAggregatorTest.java | 25 +- .../apache/druid/msq/guice/MSQSqlModule.java | 6 - .../msq/indexing/error/MSQWarningsTest.java | 1 - .../apache/druid/msq/test/MSQTestBase.java | 54 ++- .../test/MSQTestOverlordServiceClient.java | 1 - .../sql/VarianceSqlAggregatorTest.java | 32 +- .../apache/druid/query/sql/SleepSqlTest.java | 35 +- .../apache/druid/guice/ExtensionsConfig.java | 2 + .../apache/druid/guice/ExtensionsLoader.java | 1 + .../apache/druid/guice/ExtensionsModule.java | 4 +- .../org/apache/druid/guice/ModulesConfig.java | 2 + .../druid/guice/StartupInjectorBuilder.java | 17 + .../druid/jackson/DefaultObjectMapper.java | 2 + .../guice/StartupInjectorBuilderTest.java | 138 ++++++++ .../catalog/model/TableDefnRegistry.java | 3 + .../druid/guice/DruidInjectorBuilder.java | 26 +- .../druid/guice/DruidInjectorBuilderTest.java | 308 ++++++++++++++++++ .../druid/server/StatusResourceTest.java | 7 +- .../apache/druid/sql/AbstractStatement.java | 11 + .../org/apache/druid/sql/DirectStatement.java | 3 +- .../apache/druid/sql/PreparedStatement.java | 9 +- .../sql/calcite/planner/DruidPlanner.java | 16 +- .../sql/calcite/planner/IngestHandler.java | 1 + .../sql/calcite/planner/NoOpPlannerHook.java | 61 ++++ .../calcite/planner/PlannerCaptureHook.java | 78 +++++ .../sql/calcite/planner/PlannerFactory.java | 7 +- .../sql/calcite/planner/PlannerHook.java | 43 +++ .../sql/calcite/planner/QueryHandler.java | 5 + .../calcite/planner/SqlStatementHandler.java | 1 + .../sql/calcite/view/DruidViewMacro.java | 5 +- .../sql/calcite/BaseCalciteQueryTest.java | 126 +++---- .../sql/calcite/CalciteIngestionDmlTest.java | 74 +++++ .../sql/calcite/CalciteInsertDmlTest.java | 5 +- .../calcite/CalciteNestedDataQueryTest.java | 34 +- .../druid/sql/calcite/QueryTestBuilder.java | 29 +- .../druid/sql/calcite/QueryTestRunner.java | 265 ++++++++++++--- .../apache/druid/sql/calcite/SqlSchema.java | 145 +++++++++ .../sql/calcite/util/BasicTestModule.java | 65 ++++ .../util/CalciteTestInjectorBuilder.java | 46 --- .../sql/calcite/util/QueryFrameworkUtils.java | 1 + .../druid/sql/calcite/util/QueryLogHook.java | 13 +- .../sql/calcite/util/SqlTestFramework.java | 148 +++++---- .../ingest/httpExtern-logicalPlan.txt | 3 + .../ingest/insertFromExternal-logicalPlan.txt | 3 + .../insertWithClusteredBy-logicalPlan.txt | 4 + .../insertWithPartitionedBy-logicalPlan.txt | 3 + .../ingest/localExtern-logicalPlan.txt | 3 + 63 files changed, 1595 insertions(+), 658 deletions(-) create mode 100644 processing/src/test/java/org/apache/druid/guice/StartupInjectorBuilderTest.java create mode 100644 server/src/test/java/org/apache/druid/guice/DruidInjectorBuilderTest.java create mode 100644 sql/src/main/java/org/apache/druid/sql/calcite/planner/NoOpPlannerHook.java create mode 100644 sql/src/main/java/org/apache/druid/sql/calcite/planner/PlannerCaptureHook.java create mode 100644 sql/src/main/java/org/apache/druid/sql/calcite/planner/PlannerHook.java create mode 100644 sql/src/test/java/org/apache/druid/sql/calcite/SqlSchema.java create mode 100644 sql/src/test/java/org/apache/druid/sql/calcite/util/BasicTestModule.java create mode 100644 sql/src/test/resources/calcite/expected/ingest/httpExtern-logicalPlan.txt create mode 100644 sql/src/test/resources/calcite/expected/ingest/insertFromExternal-logicalPlan.txt create mode 100644 sql/src/test/resources/calcite/expected/ingest/insertWithClusteredBy-logicalPlan.txt create mode 100644 sql/src/test/resources/calcite/expected/ingest/insertWithPartitionedBy-logicalPlan.txt create mode 100644 sql/src/test/resources/calcite/expected/ingest/localExtern-logicalPlan.txt diff --git a/core/src/main/java/org/apache/druid/guice/DruidSecondaryModule.java b/core/src/main/java/org/apache/druid/guice/DruidSecondaryModule.java index d2a3fd0d52fb..a22ef051457b 100644 --- a/core/src/main/java/org/apache/druid/guice/DruidSecondaryModule.java +++ b/core/src/main/java/org/apache/druid/guice/DruidSecondaryModule.java @@ -36,6 +36,7 @@ import javax.validation.Validator; import java.util.Properties; +@LazySingleton public class DruidSecondaryModule implements Module { private final Properties properties; diff --git a/core/src/main/java/org/apache/druid/java/util/common/StringUtils.java b/core/src/main/java/org/apache/druid/java/util/common/StringUtils.java index 41078961a0c8..26d6f0c80dc4 100644 --- a/core/src/main/java/org/apache/druid/java/util/common/StringUtils.java +++ b/core/src/main/java/org/apache/druid/java/util/common/StringUtils.java @@ -20,9 +20,13 @@ package org.apache.druid.java.util.common; import com.google.common.base.Strings; +import org.apache.commons.io.IOUtils; import javax.annotation.Nonnull; import javax.annotation.Nullable; + +import java.io.IOException; +import java.io.InputStream; import java.io.UnsupportedEncodingException; import java.net.URLDecoder; import java.net.URLEncoder; @@ -797,4 +801,18 @@ public static String fastLooseChop(@Nullable final String s, final int maxBytes) return s.substring(0, maxBytes); } } + + public static String getResource(Object ref, String resource) + { + try { + InputStream is = ref.getClass().getResourceAsStream(resource); + if (is == null) { + throw new ISE("Resource not found: [%s]", resource); + } + return IOUtils.toString(is, StandardCharsets.UTF_8); + } + catch (IOException e) { + throw new ISE(e, "Cannot load resource: [%s]", resource); + } + } } diff --git a/core/src/main/java/org/apache/druid/utils/RuntimeInfo.java b/core/src/main/java/org/apache/druid/utils/RuntimeInfo.java index c8dadfd44d19..86571eff9456 100644 --- a/core/src/main/java/org/apache/druid/utils/RuntimeInfo.java +++ b/core/src/main/java/org/apache/druid/utils/RuntimeInfo.java @@ -19,10 +19,12 @@ package org.apache.druid.utils; +import org.apache.druid.guice.LazySingleton; import org.apache.druid.java.util.common.UOE; import java.lang.reflect.InvocationTargetException; +@LazySingleton public class RuntimeInfo { public int getAvailableProcessors() diff --git a/extensions-contrib/compressed-bigdecimal/src/test/java/org/apache/druid/compressedbigdecimal/CompressedBigDecimalMaxSqlAggregatorTest.java b/extensions-contrib/compressed-bigdecimal/src/test/java/org/apache/druid/compressedbigdecimal/CompressedBigDecimalMaxSqlAggregatorTest.java index fec14699d40c..06709040b645 100644 --- a/extensions-contrib/compressed-bigdecimal/src/test/java/org/apache/druid/compressedbigdecimal/CompressedBigDecimalMaxSqlAggregatorTest.java +++ b/extensions-contrib/compressed-bigdecimal/src/test/java/org/apache/druid/compressedbigdecimal/CompressedBigDecimalMaxSqlAggregatorTest.java @@ -19,23 +19,13 @@ package org.apache.druid.compressedbigdecimal; -import com.google.common.collect.ImmutableSet; -import org.apache.druid.sql.calcite.planner.DruidOperatorTable; - public class CompressedBigDecimalMaxSqlAggregatorTest extends CompressedBigDecimalSqlAggregatorTestBase { private static final String FUNCTION_NAME = CompressedBigDecimalMaxSqlAggregator.NAME; - @Override - public DruidOperatorTable createOperatorTable() - { - return new DruidOperatorTable(ImmutableSet.of(new CompressedBigDecimalMaxSqlAggregator()), ImmutableSet.of()); - } - @Override public void testCompressedBigDecimalAggWithNumberParse() { - testCompressedBigDecimalAggWithNumberParseHelper( FUNCTION_NAME, new Object[]{"6.000000000", "6.000000000", "10.100000000"}, diff --git a/extensions-contrib/compressed-bigdecimal/src/test/java/org/apache/druid/compressedbigdecimal/CompressedBigDecimalMinSqlAggregatorTest.java b/extensions-contrib/compressed-bigdecimal/src/test/java/org/apache/druid/compressedbigdecimal/CompressedBigDecimalMinSqlAggregatorTest.java index 0031780ee300..3a18dd49652b 100644 --- a/extensions-contrib/compressed-bigdecimal/src/test/java/org/apache/druid/compressedbigdecimal/CompressedBigDecimalMinSqlAggregatorTest.java +++ b/extensions-contrib/compressed-bigdecimal/src/test/java/org/apache/druid/compressedbigdecimal/CompressedBigDecimalMinSqlAggregatorTest.java @@ -19,19 +19,10 @@ package org.apache.druid.compressedbigdecimal; -import com.google.common.collect.ImmutableSet; -import org.apache.druid.sql.calcite.planner.DruidOperatorTable; - public class CompressedBigDecimalMinSqlAggregatorTest extends CompressedBigDecimalSqlAggregatorTestBase { private static final String FUNCTION_NAME = CompressedBigDecimalMinSqlAggregator.NAME; - @Override - public DruidOperatorTable createOperatorTable() - { - return new DruidOperatorTable(ImmutableSet.of(new CompressedBigDecimalMinSqlAggregator()), ImmutableSet.of()); - } - @Override public void testCompressedBigDecimalAggWithNumberParse() { diff --git a/extensions-contrib/compressed-bigdecimal/src/test/java/org/apache/druid/compressedbigdecimal/CompressedBigDecimalSqlAggregatorTestBase.java b/extensions-contrib/compressed-bigdecimal/src/test/java/org/apache/druid/compressedbigdecimal/CompressedBigDecimalSqlAggregatorTestBase.java index 9f661b7d3987..5d1c67c2f38a 100644 --- a/extensions-contrib/compressed-bigdecimal/src/test/java/org/apache/druid/compressedbigdecimal/CompressedBigDecimalSqlAggregatorTestBase.java +++ b/extensions-contrib/compressed-bigdecimal/src/test/java/org/apache/druid/compressedbigdecimal/CompressedBigDecimalSqlAggregatorTestBase.java @@ -20,17 +20,17 @@ package org.apache.druid.compressedbigdecimal; import com.fasterxml.jackson.databind.MapperFeature; -import com.fasterxml.jackson.databind.Module; import com.fasterxml.jackson.databind.ObjectMapper; import com.fasterxml.jackson.databind.SerializationFeature; import com.google.common.collect.ImmutableList; -import com.google.common.collect.Iterables; +import com.google.inject.Injector; import org.apache.druid.data.input.InputRow; import org.apache.druid.data.input.impl.DimensionsSpec; import org.apache.druid.data.input.impl.InputRowParser; import org.apache.druid.data.input.impl.MapInputRowParser; import org.apache.druid.data.input.impl.TimeAndDimsParseSpec; import org.apache.druid.data.input.impl.TimestampSpec; +import org.apache.druid.guice.DruidInjectorBuilder; import org.apache.druid.java.util.common.StringUtils; import org.apache.druid.java.util.common.granularity.Granularities; import org.apache.druid.query.Druids; @@ -45,7 +45,6 @@ import org.apache.druid.segment.writeout.OffHeapMemorySegmentWriteOutMediumFactory; import org.apache.druid.sql.calcite.BaseCalciteQueryTest; import org.apache.druid.sql.calcite.filtration.Filtration; -import org.apache.druid.sql.calcite.planner.DruidOperatorTable; import org.apache.druid.sql.calcite.util.CalciteTests; import org.apache.druid.sql.calcite.util.SpecificSegmentsQuerySegmentWalker; import org.apache.druid.sql.calcite.util.TestDataBuilder; @@ -74,16 +73,17 @@ public abstract class CompressedBigDecimalSqlAggregatorTestBase extends BaseCalc TestDataBuilder.RAW_ROWS1.stream().map(m -> TestDataBuilder.createRow(m, PARSER)).collect(Collectors.toList()); @Override - public Iterable getJacksonModules() + public void configureGuice(DruidInjectorBuilder builder) { - CompressedBigDecimalModule bigDecimalModule = new CompressedBigDecimalModule(); - return Iterables.concat(super.getJacksonModules(), bigDecimalModule.getJacksonModules()); + super.configureGuice(builder); + builder.addModule(new CompressedBigDecimalModule()); } @Override public SpecificSegmentsQuerySegmentWalker createQuerySegmentWalker( final QueryRunnerFactoryConglomerate conglomerate, - final JoinableFactoryWrapper joinableFactory + final JoinableFactoryWrapper joinableFactory, + final Injector injector ) throws IOException { QueryableIndex index = @@ -121,9 +121,6 @@ public void configureJsonMapper(ObjectMapper objectMapper) objectMapper.configure(MapperFeature.SORT_PROPERTIES_ALPHABETICALLY, true); } - @Override - public abstract DruidOperatorTable createOperatorTable(); - @Test public abstract void testCompressedBigDecimalAggWithNumberParse(); @@ -279,5 +276,4 @@ public void testCompressedBigDecimalAggDefaultSizeAndScaleHelper( ImmutableList.of(expectedResults) ); } - } diff --git a/extensions-contrib/compressed-bigdecimal/src/test/java/org/apache/druid/compressedbigdecimal/CompressedBigDecimalSumSqlAggregatorTest.java b/extensions-contrib/compressed-bigdecimal/src/test/java/org/apache/druid/compressedbigdecimal/CompressedBigDecimalSumSqlAggregatorTest.java index c7bc86d15438..db248ef67f0e 100644 --- a/extensions-contrib/compressed-bigdecimal/src/test/java/org/apache/druid/compressedbigdecimal/CompressedBigDecimalSumSqlAggregatorTest.java +++ b/extensions-contrib/compressed-bigdecimal/src/test/java/org/apache/druid/compressedbigdecimal/CompressedBigDecimalSumSqlAggregatorTest.java @@ -19,19 +19,10 @@ package org.apache.druid.compressedbigdecimal; -import com.google.common.collect.ImmutableSet; -import org.apache.druid.sql.calcite.planner.DruidOperatorTable; - public class CompressedBigDecimalSumSqlAggregatorTest extends CompressedBigDecimalSqlAggregatorTestBase { private static final String FUNCTION_NAME = CompressedBigDecimalSumSqlAggregator.NAME; - @Override - public DruidOperatorTable createOperatorTable() - { - return new DruidOperatorTable(ImmutableSet.of(new CompressedBigDecimalSumSqlAggregator()), ImmutableSet.of()); - } - @Override public void testCompressedBigDecimalAggWithNumberParse() { @@ -39,7 +30,6 @@ public void testCompressedBigDecimalAggWithNumberParse() FUNCTION_NAME, new Object[]{"21.000000000", "21.000000000", "13.100000000"}, CompressedBigDecimalSumAggregatorFactory::new - ); } diff --git a/extensions-contrib/tdigestsketch/src/test/java/org/apache/druid/query/aggregation/tdigestsketch/sql/TDigestSketchSqlAggregatorTest.java b/extensions-contrib/tdigestsketch/src/test/java/org/apache/druid/query/aggregation/tdigestsketch/sql/TDigestSketchSqlAggregatorTest.java index ae8c5bfe4683..b304dba196a3 100644 --- a/extensions-contrib/tdigestsketch/src/test/java/org/apache/druid/query/aggregation/tdigestsketch/sql/TDigestSketchSqlAggregatorTest.java +++ b/extensions-contrib/tdigestsketch/src/test/java/org/apache/druid/query/aggregation/tdigestsketch/sql/TDigestSketchSqlAggregatorTest.java @@ -19,11 +19,10 @@ package org.apache.druid.query.aggregation.tdigestsketch.sql; -import com.fasterxml.jackson.databind.Module; import com.google.common.collect.ImmutableList; -import com.google.common.collect.ImmutableSet; -import com.google.common.collect.Iterables; +import com.google.inject.Injector; import org.apache.druid.common.config.NullHandling; +import org.apache.druid.guice.DruidInjectorBuilder; import org.apache.druid.java.util.common.granularity.Granularities; import org.apache.druid.math.expr.ExprMacroTable; import org.apache.druid.query.Druids; @@ -50,7 +49,6 @@ import org.apache.druid.segment.writeout.OffHeapMemorySegmentWriteOutMediumFactory; import org.apache.druid.sql.calcite.BaseCalciteQueryTest; import org.apache.druid.sql.calcite.filtration.Filtration; -import org.apache.druid.sql.calcite.planner.DruidOperatorTable; import org.apache.druid.sql.calcite.util.CalciteTests; import org.apache.druid.sql.calcite.util.SpecificSegmentsQuerySegmentWalker; import org.apache.druid.sql.calcite.util.TestDataBuilder; @@ -64,21 +62,18 @@ public class TDigestSketchSqlAggregatorTest extends BaseCalciteQueryTest { - private static final DruidOperatorTable OPERATOR_TABLE = new DruidOperatorTable( - ImmutableSet.of(new TDigestSketchQuantileSqlAggregator(), new TDigestGenerateSketchSqlAggregator()), - ImmutableSet.of() - ); - @Override - public Iterable getJacksonModules() + public void configureGuice(DruidInjectorBuilder builder) { - return Iterables.concat(super.getJacksonModules(), new TDigestSketchModule().getJacksonModules()); + super.configureGuice(builder); + builder.addModule(new TDigestSketchModule()); } @Override public SpecificSegmentsQuerySegmentWalker createQuerySegmentWalker( final QueryRunnerFactoryConglomerate conglomerate, - final JoinableFactoryWrapper joinableFactory + final JoinableFactoryWrapper joinableFactory, + final Injector injector ) throws IOException { TDigestSketchModule.registerSerde(); @@ -116,12 +111,6 @@ public SpecificSegmentsQuerySegmentWalker createQuerySegmentWalker( ); } - @Override - public DruidOperatorTable createOperatorTable() - { - return OPERATOR_TABLE; - } - @Test public void testComputingSketchOnNumericValues() { diff --git a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/hll/HllSketchModule.java b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/hll/HllSketchModule.java index 5044068262d8..2a07c70ec639 100644 --- a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/hll/HllSketchModule.java +++ b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/hll/HllSketchModule.java @@ -44,7 +44,6 @@ */ public class HllSketchModule implements DruidModule { - public static final String TYPE_NAME = "HLLSketch"; // common type name to be associated with segment data public static final String BUILD_TYPE_NAME = "HLLSketchBuild"; public static final String MERGE_TYPE_NAME = "HLLSketchMerge"; @@ -53,7 +52,6 @@ public class HllSketchModule implements DruidModule public static final String ESTIMATE_WITH_BOUNDS_TYPE_NAME = "HLLSketchEstimateWithBounds"; public static final String ESTIMATE_TYPE_NAME = "HLLSketchEstimate"; - @Override public void configure(final Binder binder) { diff --git a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/quantiles/DoublesSketchModule.java b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/quantiles/DoublesSketchModule.java index cd047695f45d..7896e1c55c10 100644 --- a/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/quantiles/DoublesSketchModule.java +++ b/extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/quantiles/DoublesSketchModule.java @@ -43,7 +43,6 @@ public class DoublesSketchModule implements DruidModule { - public static final String DOUBLES_SKETCH = "quantilesDoublesSketch"; public static final String DOUBLES_SKETCH_MERGE = "quantilesDoublesSketchMerge"; public static final ColumnType TYPE = ColumnType.ofComplex(DOUBLES_SKETCH); diff --git a/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/hll/sql/HllSketchSqlAggregatorTest.java b/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/hll/sql/HllSketchSqlAggregatorTest.java index 86e76ca8c219..aa0e9afd13a1 100644 --- a/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/hll/sql/HllSketchSqlAggregatorTest.java +++ b/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/hll/sql/HllSketchSqlAggregatorTest.java @@ -19,12 +19,11 @@ package org.apache.druid.query.aggregation.datasketches.hll.sql; -import com.fasterxml.jackson.databind.Module; import com.google.common.collect.ImmutableList; import com.google.common.collect.ImmutableMap; -import com.google.common.collect.ImmutableSet; -import com.google.common.collect.Iterables; +import com.google.inject.Injector; import org.apache.druid.common.config.NullHandling; +import org.apache.druid.guice.DruidInjectorBuilder; import org.apache.druid.java.util.common.StringUtils; import org.apache.druid.java.util.common.granularity.Granularities; import org.apache.druid.java.util.common.granularity.PeriodGranularity; @@ -61,13 +60,11 @@ import org.apache.druid.segment.virtual.ExpressionVirtualColumn; import org.apache.druid.segment.writeout.OffHeapMemorySegmentWriteOutMediumFactory; import org.apache.druid.sql.calcite.BaseCalciteQueryTest; -import org.apache.druid.sql.calcite.aggregation.ApproxCountDistinctSqlAggregator; -import org.apache.druid.sql.calcite.aggregation.builtin.CountSqlAggregator; import org.apache.druid.sql.calcite.filtration.Filtration; -import org.apache.druid.sql.calcite.planner.DruidOperatorTable; import org.apache.druid.sql.calcite.util.CalciteTests; import org.apache.druid.sql.calcite.util.SpecificSegmentsQuerySegmentWalker; import org.apache.druid.sql.calcite.util.TestDataBuilder; +import org.apache.druid.sql.guice.SqlModule; import org.apache.druid.timeline.DataSegment; import org.apache.druid.timeline.partition.LinearShardSpec; import org.joda.time.DateTimeZone; @@ -78,22 +75,34 @@ import java.util.Arrays; import java.util.Collections; import java.util.List; +import java.util.Properties; public class HllSketchSqlAggregatorTest extends BaseCalciteQueryTest { private static final boolean ROUND = true; @Override - public Iterable getJacksonModules() + public void gatherProperties(Properties properties) { - return Iterables.concat(super.getJacksonModules(), new HllSketchModule().getJacksonModules()); + super.gatherProperties(properties); + + // Use APPROX_COUNT_DISTINCT_DS_HLL as APPROX_COUNT_DISTINCT impl for these tests. + properties.put(SqlModule.PROPERTY_SQL_APPROX_COUNT_DISTINCT_CHOICE, HllSketchApproxCountDistinctSqlAggregator.NAME); + } + + @Override + public void configureGuice(DruidInjectorBuilder builder) + { + super.configureGuice(builder); + builder.addModule(new HllSketchModule()); } @SuppressWarnings("resource") @Override public SpecificSegmentsQuerySegmentWalker createQuerySegmentWalker( final QueryRunnerFactoryConglomerate conglomerate, - final JoinableFactoryWrapper joinableFactory + final JoinableFactoryWrapper joinableFactory, + final Injector injector ) throws IOException { HllSketchModule.registerSerde(); @@ -132,30 +141,6 @@ public SpecificSegmentsQuerySegmentWalker createQuerySegmentWalker( ); } - @Override - public DruidOperatorTable createOperatorTable() - { - final HllSketchApproxCountDistinctSqlAggregator approxCountDistinctSqlAggregator = - new HllSketchApproxCountDistinctSqlAggregator(); - - return new DruidOperatorTable( - ImmutableSet.of( - approxCountDistinctSqlAggregator, - new HllSketchObjectSqlAggregator(), - - // Use APPROX_COUNT_DISTINCT_DS_HLL as APPROX_COUNT_DISTINCT impl for these tests. - new CountSqlAggregator(new ApproxCountDistinctSqlAggregator(approxCountDistinctSqlAggregator)), - new ApproxCountDistinctSqlAggregator(approxCountDistinctSqlAggregator) - ), - ImmutableSet.of( - new HllSketchSetUnionOperatorConversion(), - new HllSketchEstimateOperatorConversion(), - new HllSketchToStringOperatorConversion(), - new HllSketchEstimateWithErrorBoundsOperatorConversion() - ) - ); - } - @Test public void testApproxCountDistinctHllSketch() { diff --git a/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/quantiles/sql/DoublesSketchSqlAggregatorTest.java b/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/quantiles/sql/DoublesSketchSqlAggregatorTest.java index e4a66a9a56c8..f2253462bd01 100644 --- a/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/quantiles/sql/DoublesSketchSqlAggregatorTest.java +++ b/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/quantiles/sql/DoublesSketchSqlAggregatorTest.java @@ -19,12 +19,11 @@ package org.apache.druid.query.aggregation.datasketches.quantiles.sql; -import com.fasterxml.jackson.databind.Module; import com.google.common.collect.ImmutableList; import com.google.common.collect.ImmutableMap; -import com.google.common.collect.ImmutableSet; -import com.google.common.collect.Iterables; +import com.google.inject.Injector; import org.apache.druid.common.config.NullHandling; +import org.apache.druid.guice.DruidInjectorBuilder; import org.apache.druid.java.util.common.granularity.Granularities; import org.apache.druid.math.expr.ExprMacroTable; import org.apache.druid.query.Druids; @@ -63,7 +62,6 @@ import org.apache.druid.segment.writeout.OffHeapMemorySegmentWriteOutMediumFactory; import org.apache.druid.sql.calcite.BaseCalciteQueryTest; import org.apache.druid.sql.calcite.filtration.Filtration; -import org.apache.druid.sql.calcite.planner.DruidOperatorTable; import org.apache.druid.sql.calcite.util.CalciteTests; import org.apache.druid.sql.calcite.util.SpecificSegmentsQuerySegmentWalker; import org.apache.druid.sql.calcite.util.TestDataBuilder; @@ -79,31 +77,18 @@ public class DoublesSketchSqlAggregatorTest extends BaseCalciteQueryTest { - private static final DruidOperatorTable OPERATOR_TABLE = new DruidOperatorTable( - ImmutableSet.of( - new DoublesSketchApproxQuantileSqlAggregator(), - new DoublesSketchObjectSqlAggregator() - ), - ImmutableSet.of( - new DoublesSketchQuantileOperatorConversion(), - new DoublesSketchQuantilesOperatorConversion(), - new DoublesSketchToHistogramOperatorConversion(), - new DoublesSketchRankOperatorConversion(), - new DoublesSketchCDFOperatorConversion(), - new DoublesSketchSummaryOperatorConversion() - ) - ); - @Override - public Iterable getJacksonModules() + public void configureGuice(DruidInjectorBuilder builder) { - return Iterables.concat(super.getJacksonModules(), new DoublesSketchModule().getJacksonModules()); + super.configureGuice(builder); + builder.addModule(new DoublesSketchModule()); } @Override public SpecificSegmentsQuerySegmentWalker createQuerySegmentWalker( final QueryRunnerFactoryConglomerate conglomerate, - final JoinableFactoryWrapper joinableFactory + final JoinableFactoryWrapper joinableFactory, + final Injector injector ) throws IOException { DoublesSketchModule.registerSerde(); @@ -141,12 +126,6 @@ public SpecificSegmentsQuerySegmentWalker createQuerySegmentWalker( ); } - @Override - public DruidOperatorTable createOperatorTable() - { - return OPERATOR_TABLE; - } - @Test public void testQuantileOnFloatAndLongs() { diff --git a/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/theta/sql/ThetaSketchSqlAggregatorTest.java b/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/theta/sql/ThetaSketchSqlAggregatorTest.java index 128ffd9419e3..ac7ae090558f 100644 --- a/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/theta/sql/ThetaSketchSqlAggregatorTest.java +++ b/extensions-core/datasketches/src/test/java/org/apache/druid/query/aggregation/datasketches/theta/sql/ThetaSketchSqlAggregatorTest.java @@ -19,17 +19,14 @@ package org.apache.druid.query.aggregation.datasketches.theta.sql; -import com.fasterxml.jackson.databind.Module; import com.google.common.collect.ImmutableList; import com.google.common.collect.ImmutableMap; -import com.google.common.collect.ImmutableSet; -import com.google.common.collect.Iterables; +import com.google.inject.Injector; import org.apache.druid.common.config.NullHandling; -import org.apache.druid.guice.ExpressionModule; +import org.apache.druid.guice.DruidInjectorBuilder; import org.apache.druid.java.util.common.StringUtils; import org.apache.druid.java.util.common.granularity.Granularities; import org.apache.druid.java.util.common.granularity.PeriodGranularity; -import org.apache.druid.math.expr.ExprMacroTable; import org.apache.druid.query.Druids; import org.apache.druid.query.QueryDataSource; import org.apache.druid.query.QueryRunnerFactoryConglomerate; @@ -57,58 +54,48 @@ import org.apache.druid.segment.virtual.ExpressionVirtualColumn; import org.apache.druid.segment.writeout.OffHeapMemorySegmentWriteOutMediumFactory; import org.apache.druid.sql.calcite.BaseCalciteQueryTest; -import org.apache.druid.sql.calcite.aggregation.ApproxCountDistinctSqlAggregator; -import org.apache.druid.sql.calcite.aggregation.builtin.CountSqlAggregator; import org.apache.druid.sql.calcite.filtration.Filtration; -import org.apache.druid.sql.calcite.planner.DruidOperatorTable; import org.apache.druid.sql.calcite.util.CalciteTests; import org.apache.druid.sql.calcite.util.SpecificSegmentsQuerySegmentWalker; import org.apache.druid.sql.calcite.util.TestDataBuilder; +import org.apache.druid.sql.guice.SqlModule; import org.apache.druid.timeline.DataSegment; import org.apache.druid.timeline.partition.LinearShardSpec; import org.joda.time.DateTimeZone; import org.joda.time.Period; -import org.junit.Before; import org.junit.Test; import java.io.IOException; -import java.util.ArrayList; import java.util.Arrays; import java.util.Collections; import java.util.List; +import java.util.Properties; public class ThetaSketchSqlAggregatorTest extends BaseCalciteQueryTest { private static final String DATA_SOURCE = "foo"; - private ExprMacroTable macroTable; - - @Before - public void setUp() - { - macroTable = createMacroTable(); - } - @Override - public Iterable getJacksonModules() + public void gatherProperties(Properties properties) { - return Iterables.concat(super.getJacksonModules(), new SketchModule().getJacksonModules()); + super.gatherProperties(properties); + + // Use APPROX_COUNT_DISTINCT_DS_THETA as APPROX_COUNT_DISTINCT impl for these tests. + properties.put(SqlModule.PROPERTY_SQL_APPROX_COUNT_DISTINCT_CHOICE, ThetaSketchApproxCountDistinctSqlAggregator.NAME); } @Override - public ExprMacroTable createMacroTable() + public void configureGuice(DruidInjectorBuilder builder) { - final List exprMacros = new ArrayList<>(); - for (Class clazz : ExpressionModule.EXPR_MACROS) { - exprMacros.add(CalciteTests.INJECTOR.getInstance(clazz)); - } - return new ExprMacroTable(exprMacros); + super.configureGuice(builder); + builder.addModule(new SketchModule()); } @Override public SpecificSegmentsQuerySegmentWalker createQuerySegmentWalker( final QueryRunnerFactoryConglomerate conglomerate, - final JoinableFactoryWrapper joinableFactory + final JoinableFactoryWrapper joinableFactory, + final Injector injector ) throws IOException { SketchModule.registerSerde(); @@ -148,31 +135,6 @@ public SpecificSegmentsQuerySegmentWalker createQuerySegmentWalker( ); } - @Override - public DruidOperatorTable createOperatorTable() - { - final ThetaSketchApproxCountDistinctSqlAggregator approxCountDistinctSqlAggregator = - new ThetaSketchApproxCountDistinctSqlAggregator(); - - return new DruidOperatorTable( - ImmutableSet.of( - new ThetaSketchApproxCountDistinctSqlAggregator(), - new ThetaSketchObjectSqlAggregator(), - - // Use APPROX_COUNT_DISTINCT_DS_THETA as APPROX_COUNT_DISTINCT impl for these tests. - new CountSqlAggregator(new ApproxCountDistinctSqlAggregator(approxCountDistinctSqlAggregator)), - new ApproxCountDistinctSqlAggregator(approxCountDistinctSqlAggregator) - ), - ImmutableSet.of( - new ThetaSketchEstimateOperatorConversion(), - new ThetaSketchEstimateWithErrorBoundsOperatorConversion(), - new ThetaSketchSetIntersectOperatorConversion(), - new ThetaSketchSetUnionOperatorConversion(), - new ThetaSketchSetNotOperatorConversion() - ) - ); - } - @Test public void testApproxCountDistinctThetaSketch() { @@ -235,13 +197,13 @@ public void testApproxCountDistinctThetaSketch() "v0", "substring(\"dim2\", 0, 1)", ColumnType.STRING, - macroTable + queryFramework().macroTable() ), new ExpressionVirtualColumn( "v1", "concat(substring(\"dim2\", 0, 1),'x')", ColumnType.STRING, - macroTable + queryFramework().macroTable() ) ) .aggregators( @@ -444,7 +406,7 @@ public void testThetaSketchPostAggs() "v0", "concat(\"dim2\",'hello')", ColumnType.STRING, - macroTable + queryFramework().macroTable() ) ) .aggregators( @@ -621,7 +583,7 @@ public void testThetaSketchPostAggsFinalizeOuterSketches() "v0", "concat(\"dim2\",'hello')", ColumnType.STRING, - macroTable + queryFramework().macroTable() ) ) .aggregators( diff --git a/extensions-core/druid-bloom-filter/src/test/java/org/apache/druid/query/aggregation/bloom/sql/BloomFilterSqlAggregatorTest.java b/extensions-core/druid-bloom-filter/src/test/java/org/apache/druid/query/aggregation/bloom/sql/BloomFilterSqlAggregatorTest.java index 26f1ce108fee..c1426048faf3 100644 --- a/extensions-core/druid-bloom-filter/src/test/java/org/apache/druid/query/aggregation/bloom/sql/BloomFilterSqlAggregatorTest.java +++ b/extensions-core/druid-bloom-filter/src/test/java/org/apache/druid/query/aggregation/bloom/sql/BloomFilterSqlAggregatorTest.java @@ -19,10 +19,9 @@ package org.apache.druid.query.aggregation.bloom.sql; -import com.fasterxml.jackson.databind.Module; +import com.fasterxml.jackson.databind.ObjectMapper; import com.google.common.collect.ImmutableList; -import com.google.common.collect.ImmutableSet; -import com.google.common.collect.Iterables; +import com.google.inject.Injector; import org.apache.druid.common.config.NullHandling; import org.apache.druid.data.input.InputRow; import org.apache.druid.data.input.impl.DimensionSchema; @@ -35,6 +34,7 @@ import org.apache.druid.data.input.impl.TimeAndDimsParseSpec; import org.apache.druid.data.input.impl.TimestampSpec; import org.apache.druid.guice.BloomFilterExtensionModule; +import org.apache.druid.guice.DruidInjectorBuilder; import org.apache.druid.java.util.common.granularity.Granularities; import org.apache.druid.query.Druids; import org.apache.druid.query.QueryRunnerFactoryConglomerate; @@ -59,7 +59,6 @@ import org.apache.druid.segment.writeout.OffHeapMemorySegmentWriteOutMediumFactory; import org.apache.druid.sql.calcite.BaseCalciteQueryTest; import org.apache.druid.sql.calcite.filtration.Filtration; -import org.apache.druid.sql.calcite.planner.DruidOperatorTable; import org.apache.druid.sql.calcite.util.CalciteTests; import org.apache.druid.sql.calcite.util.SpecificSegmentsQuerySegmentWalker; import org.apache.druid.sql.calcite.util.TestDataBuilder; @@ -76,21 +75,18 @@ public class BloomFilterSqlAggregatorTest extends BaseCalciteQueryTest private static final String DATA_SOURCE = "numfoo"; - private static final DruidOperatorTable OPERATOR_TABLE = new DruidOperatorTable( - ImmutableSet.of(new BloomFilterSqlAggregator()), - ImmutableSet.of() - ); - @Override - public Iterable getJacksonModules() + public void configureGuice(DruidInjectorBuilder builder) { - return Iterables.concat(super.getJacksonModules(), new BloomFilterExtensionModule().getJacksonModules()); + super.configureGuice(builder); + builder.addModule(new BloomFilterExtensionModule()); } @Override public SpecificSegmentsQuerySegmentWalker createQuerySegmentWalker( final QueryRunnerFactoryConglomerate conglomerate, - final JoinableFactoryWrapper joinableFactory + final JoinableFactoryWrapper joinableFactory, + final Injector injector ) throws IOException { InputRowParser parser = new MapInputRowParser( @@ -135,12 +131,6 @@ public SpecificSegmentsQuerySegmentWalker createQuerySegmentWalker( ); } - @Override - public DruidOperatorTable createOperatorTable() - { - return OPERATOR_TABLE; - } - @Test public void testBloomFilterAgg() throws Exception { @@ -178,7 +168,7 @@ public void testBloomFilterAgg() throws Exception .build() ), ImmutableList.of( - new Object[]{CalciteTests.getJsonMapper().writeValueAsString(expected1)} + new Object[]{queryFramework().queryJsonMapper().writeValueAsString(expected1)} ) ); } @@ -211,6 +201,7 @@ public void testBloomFilterTwoAggs() throws Exception } } + ObjectMapper jsonMapper = queryFramework().queryJsonMapper(); testQuery( "SELECT\n" + "BLOOM_FILTER(dim1, 1000),\n" @@ -239,8 +230,8 @@ public void testBloomFilterTwoAggs() throws Exception ), ImmutableList.of( new Object[] { - CalciteTests.getJsonMapper().writeValueAsString(expected1), - CalciteTests.getJsonMapper().writeValueAsString(expected2) + jsonMapper.writeValueAsString(expected1), + jsonMapper.writeValueAsString(expected2) } ) ); @@ -288,10 +279,9 @@ public void testBloomFilterAggExtractionFn() throws Exception .build() ), ImmutableList.of( - new Object[]{CalciteTests.getJsonMapper().writeValueAsString(expected1)} + new Object[]{queryFramework().queryJsonMapper().writeValueAsString(expected1)} ) ); - } @Test @@ -335,7 +325,7 @@ public void testBloomFilterAggLong() throws Exception .build() ), ImmutableList.of( - new Object[]{CalciteTests.getJsonMapper().writeValueAsString(expected3)} + new Object[]{queryFramework().queryJsonMapper().writeValueAsString(expected3)} ) ); } @@ -388,7 +378,7 @@ public void testBloomFilterAggLongVirtualColumn() throws Exception .build() ), ImmutableList.of( - new Object[]{CalciteTests.getJsonMapper().writeValueAsString(expected1)} + new Object[]{queryFramework().queryJsonMapper().writeValueAsString(expected1)} ) ); } @@ -442,7 +432,7 @@ public void testBloomFilterAggFloatVirtualColumn() throws Exception .build() ), ImmutableList.of( - new Object[]{CalciteTests.getJsonMapper().writeValueAsString(expected1)} + new Object[]{queryFramework().queryJsonMapper().writeValueAsString(expected1)} ) ); } @@ -496,7 +486,7 @@ public void testBloomFilterAggDoubleVirtualColumn() throws Exception .build() ), ImmutableList.of( - new Object[]{CalciteTests.getJsonMapper().writeValueAsString(expected1)} + new Object[]{queryFramework().queryJsonMapper().writeValueAsString(expected1)} ) ); } @@ -510,6 +500,7 @@ public void testEmptyTimeseriesResults() throws Exception BloomKFilter expected1 = new BloomKFilter(TEST_NUM_ENTRIES); BloomKFilter expected2 = new BloomKFilter(TEST_NUM_ENTRIES); + ObjectMapper jsonMapper = queryFramework().queryJsonMapper(); testQuery( "SELECT\n" + "BLOOM_FILTER(dim1, 1000),\n" @@ -540,8 +531,8 @@ public void testEmptyTimeseriesResults() throws Exception ), ImmutableList.of( new Object[] { - CalciteTests.getJsonMapper().writeValueAsString(expected1), - CalciteTests.getJsonMapper().writeValueAsString(expected2) + jsonMapper.writeValueAsString(expected1), + jsonMapper.writeValueAsString(expected2) } ) ); @@ -556,6 +547,7 @@ public void testGroupByAggregatorDefaultValues() throws Exception BloomKFilter expected1 = new BloomKFilter(TEST_NUM_ENTRIES); BloomKFilter expected2 = new BloomKFilter(TEST_NUM_ENTRIES); + ObjectMapper jsonMapper = queryFramework().queryJsonMapper(); testQuery( "SELECT\n" + "dim2,\n" @@ -596,8 +588,8 @@ public void testGroupByAggregatorDefaultValues() throws Exception ImmutableList.of( new Object[] { "a", - CalciteTests.getJsonMapper().writeValueAsString(expected1), - CalciteTests.getJsonMapper().writeValueAsString(expected2) + jsonMapper.writeValueAsString(expected1), + jsonMapper.writeValueAsString(expected2) } ) ); diff --git a/extensions-core/druid-bloom-filter/src/test/java/org/apache/druid/query/filter/sql/BloomDimFilterSqlTest.java b/extensions-core/druid-bloom-filter/src/test/java/org/apache/druid/query/filter/sql/BloomDimFilterSqlTest.java index 4e1659c228fb..c5d6e631c410 100644 --- a/extensions-core/druid-bloom-filter/src/test/java/org/apache/druid/query/filter/sql/BloomDimFilterSqlTest.java +++ b/extensions-core/druid-bloom-filter/src/test/java/org/apache/druid/query/filter/sql/BloomDimFilterSqlTest.java @@ -19,22 +19,16 @@ package org.apache.druid.query.filter.sql; -import com.fasterxml.jackson.databind.Module; import com.google.common.collect.ImmutableList; -import com.google.common.collect.ImmutableSet; -import com.google.common.collect.Iterables; import org.apache.calcite.avatica.SqlType; import org.apache.druid.common.config.NullHandling; import org.apache.druid.guice.BloomFilterExtensionModule; import org.apache.druid.guice.BloomFilterSerializersModule; -import org.apache.druid.guice.ExpressionModule; +import org.apache.druid.guice.DruidInjectorBuilder; import org.apache.druid.java.util.common.StringUtils; import org.apache.druid.java.util.common.granularity.Granularities; -import org.apache.druid.math.expr.ExprMacroTable; import org.apache.druid.query.Druids; import org.apache.druid.query.aggregation.CountAggregatorFactory; -import org.apache.druid.query.expression.LookupExprMacro; -import org.apache.druid.query.expressions.BloomFilterExpressions; import org.apache.druid.query.filter.BloomDimFilter; import org.apache.druid.query.filter.BloomKFilter; import org.apache.druid.query.filter.BloomKFilterHolder; @@ -42,50 +36,21 @@ import org.apache.druid.query.filter.OrDimFilter; import org.apache.druid.segment.column.ColumnType; import org.apache.druid.sql.calcite.BaseCalciteQueryTest; -import org.apache.druid.sql.calcite.aggregation.ApproxCountDistinctSqlAggregator; -import org.apache.druid.sql.calcite.aggregation.builtin.BuiltinApproxCountDistinctSqlAggregator; -import org.apache.druid.sql.calcite.aggregation.builtin.CountSqlAggregator; import org.apache.druid.sql.calcite.filtration.Filtration; -import org.apache.druid.sql.calcite.planner.DruidOperatorTable; import org.apache.druid.sql.calcite.util.CalciteTests; import org.apache.druid.sql.http.SqlParameter; import org.junit.Ignore; import org.junit.Test; import java.io.IOException; -import java.util.ArrayList; -import java.util.List; public class BloomDimFilterSqlTest extends BaseCalciteQueryTest { @Override - public DruidOperatorTable createOperatorTable() + public void configureGuice(DruidInjectorBuilder builder) { - CalciteTests.getJsonMapper().registerModule(new BloomFilterSerializersModule()); - return new DruidOperatorTable( - ImmutableSet.of( - new CountSqlAggregator(new ApproxCountDistinctSqlAggregator(new BuiltinApproxCountDistinctSqlAggregator())) - ), - ImmutableSet.of(new BloomFilterOperatorConversion()) - ); - } - - @Override - public ExprMacroTable createMacroTable() - { - final List exprMacros = new ArrayList<>(); - for (Class clazz : ExpressionModule.EXPR_MACROS) { - exprMacros.add(CalciteTests.INJECTOR.getInstance(clazz)); - } - exprMacros.add(CalciteTests.INJECTOR.getInstance(LookupExprMacro.class)); - exprMacros.add(new BloomFilterExpressions.TestExprMacro()); - return new ExprMacroTable(exprMacros); - } - - @Override - public Iterable getJacksonModules() - { - return Iterables.concat(super.getJacksonModules(), new BloomFilterExtensionModule().getJacksonModules()); + super.configureGuice(builder); + builder.addModule(new BloomFilterExtensionModule()); } @Test @@ -144,7 +109,7 @@ public void testBloomFilterExprFilter() throws IOException base64 ), null, - createMacroTable() + queryFramework().macroTable() ) ) .aggregators(aggregators(new CountAggregatorFactory("a0"))) diff --git a/extensions-core/histogram/src/test/java/org/apache/druid/query/aggregation/histogram/sql/FixedBucketsHistogramQuantileSqlAggregatorTest.java b/extensions-core/histogram/src/test/java/org/apache/druid/query/aggregation/histogram/sql/FixedBucketsHistogramQuantileSqlAggregatorTest.java index dd6399038c3e..a4eae57756ce 100644 --- a/extensions-core/histogram/src/test/java/org/apache/druid/query/aggregation/histogram/sql/FixedBucketsHistogramQuantileSqlAggregatorTest.java +++ b/extensions-core/histogram/src/test/java/org/apache/druid/query/aggregation/histogram/sql/FixedBucketsHistogramQuantileSqlAggregatorTest.java @@ -19,12 +19,11 @@ package org.apache.druid.query.aggregation.histogram.sql; -import com.fasterxml.jackson.databind.Module; import com.google.common.collect.ImmutableList; import com.google.common.collect.ImmutableMap; -import com.google.common.collect.ImmutableSet; -import com.google.common.collect.Iterables; +import com.google.inject.Injector; import org.apache.druid.common.config.NullHandling; +import org.apache.druid.guice.DruidInjectorBuilder; import org.apache.druid.java.util.common.granularity.Granularities; import org.apache.druid.query.Druids; import org.apache.druid.query.QueryContexts; @@ -55,7 +54,6 @@ import org.apache.druid.segment.writeout.OffHeapMemorySegmentWriteOutMediumFactory; import org.apache.druid.sql.calcite.BaseCalciteQueryTest; import org.apache.druid.sql.calcite.filtration.Filtration; -import org.apache.druid.sql.calcite.planner.DruidOperatorTable; import org.apache.druid.sql.calcite.util.CalciteTests; import org.apache.druid.sql.calcite.util.SpecificSegmentsQuerySegmentWalker; import org.apache.druid.sql.calcite.util.TestDataBuilder; @@ -68,21 +66,18 @@ public class FixedBucketsHistogramQuantileSqlAggregatorTest extends BaseCalciteQueryTest { - private static final DruidOperatorTable OPERATOR_TABLE = new DruidOperatorTable( - ImmutableSet.of(new QuantileSqlAggregator(), new FixedBucketsHistogramQuantileSqlAggregator()), - ImmutableSet.of() - ); - @Override - public Iterable getJacksonModules() + public void configureGuice(DruidInjectorBuilder builder) { - return Iterables.concat(super.getJacksonModules(), new ApproximateHistogramDruidModule().getJacksonModules()); + super.configureGuice(builder); + builder.addModule(new ApproximateHistogramDruidModule()); } @Override public SpecificSegmentsQuerySegmentWalker createQuerySegmentWalker( final QueryRunnerFactoryConglomerate conglomerate, - final JoinableFactoryWrapper joinableFactory + final JoinableFactoryWrapper joinableFactory, + final Injector injector ) throws IOException { ApproximateHistogramDruidModule.registerSerde(); @@ -123,13 +118,6 @@ public SpecificSegmentsQuerySegmentWalker createQuerySegmentWalker( ); } - @Override - public DruidOperatorTable createOperatorTable() - { - return OPERATOR_TABLE; - } - - @Test public void testQuantileOnFloatAndLongs() { @@ -568,7 +556,6 @@ public void testEmptyTimeseriesResults() ); } - @Test public void testGroupByAggregatorDefaultValues() { diff --git a/extensions-core/histogram/src/test/java/org/apache/druid/query/aggregation/histogram/sql/QuantileSqlAggregatorTest.java b/extensions-core/histogram/src/test/java/org/apache/druid/query/aggregation/histogram/sql/QuantileSqlAggregatorTest.java index 6ed4c7e45106..4a67833b5249 100644 --- a/extensions-core/histogram/src/test/java/org/apache/druid/query/aggregation/histogram/sql/QuantileSqlAggregatorTest.java +++ b/extensions-core/histogram/src/test/java/org/apache/druid/query/aggregation/histogram/sql/QuantileSqlAggregatorTest.java @@ -19,11 +19,10 @@ package org.apache.druid.query.aggregation.histogram.sql; -import com.fasterxml.jackson.databind.Module; import com.google.common.collect.ImmutableList; -import com.google.common.collect.ImmutableSet; -import com.google.common.collect.Iterables; +import com.google.inject.Injector; import org.apache.druid.common.config.NullHandling; +import org.apache.druid.guice.DruidInjectorBuilder; import org.apache.druid.java.util.common.granularity.Granularities; import org.apache.druid.math.expr.ExprMacroTable; import org.apache.druid.query.Druids; @@ -54,7 +53,6 @@ import org.apache.druid.segment.writeout.OffHeapMemorySegmentWriteOutMediumFactory; import org.apache.druid.sql.calcite.BaseCalciteQueryTest; import org.apache.druid.sql.calcite.filtration.Filtration; -import org.apache.druid.sql.calcite.planner.DruidOperatorTable; import org.apache.druid.sql.calcite.util.CalciteTests; import org.apache.druid.sql.calcite.util.SpecificSegmentsQuerySegmentWalker; import org.apache.druid.sql.calcite.util.TestDataBuilder; @@ -67,21 +65,18 @@ public class QuantileSqlAggregatorTest extends BaseCalciteQueryTest { - private static final DruidOperatorTable OPERATOR_TABLE = new DruidOperatorTable( - ImmutableSet.of(new QuantileSqlAggregator()), - ImmutableSet.of() - ); - @Override - public Iterable getJacksonModules() + public void configureGuice(DruidInjectorBuilder builder) { - return Iterables.concat(super.getJacksonModules(), new ApproximateHistogramDruidModule().getJacksonModules()); + super.configureGuice(builder); + builder.addModule(new ApproximateHistogramDruidModule()); } @Override public SpecificSegmentsQuerySegmentWalker createQuerySegmentWalker( final QueryRunnerFactoryConglomerate conglomerate, - final JoinableFactoryWrapper joinableFactory + final JoinableFactoryWrapper joinableFactory, + final Injector injector ) throws IOException { ApproximateHistogramDruidModule.registerSerde(); @@ -122,12 +117,6 @@ public SpecificSegmentsQuerySegmentWalker createQuerySegmentWalker( ); } - @Override - public DruidOperatorTable createOperatorTable() - { - return OPERATOR_TABLE; - } - @Test public void testQuantileOnFloatAndLongs() { diff --git a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/guice/MSQSqlModule.java b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/guice/MSQSqlModule.java index e2c47c2a2dda..8e381e50bd01 100644 --- a/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/guice/MSQSqlModule.java +++ b/extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/guice/MSQSqlModule.java @@ -21,7 +21,6 @@ import com.fasterxml.jackson.databind.Module; import com.google.inject.Binder; -import com.google.inject.Inject; import com.google.inject.Provides; import org.apache.druid.discovery.NodeRole; import org.apache.druid.guice.LazySingleton; @@ -36,7 +35,6 @@ import org.apache.druid.sql.guice.SqlBindings; import java.util.List; -import java.util.Properties; /** * Module for providing the {@code EXTERN} operator. @@ -44,9 +42,6 @@ @LoadScope(roles = NodeRole.BROKER_JSON_NAME) public class MSQSqlModule implements DruidModule { - @Inject - Properties properties = null; - @Override public List getJacksonModules() { @@ -66,7 +61,6 @@ public void configure(Binder binder) SqlBindings.addOperatorConversion(binder, ExternalOperatorConversion.class); } - @Provides @MSQ @LazySingleton diff --git a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/indexing/error/MSQWarningsTest.java b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/indexing/error/MSQWarningsTest.java index 38318bf889b3..7ec2766b5967 100644 --- a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/indexing/error/MSQWarningsTest.java +++ b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/indexing/error/MSQWarningsTest.java @@ -55,7 +55,6 @@ */ public class MSQWarningsTest extends MSQTestBase { - private File toRead; private RowSignature rowSignature; private String toReadFileNameAsJson; diff --git a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/test/MSQTestBase.java b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/test/MSQTestBase.java index 133517dd760e..43ae5e0b0c97 100644 --- a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/test/MSQTestBase.java +++ b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/test/MSQTestBase.java @@ -28,6 +28,7 @@ import com.google.common.collect.ImmutableMap; import com.google.common.collect.ImmutableSet; import com.google.common.collect.Iterables; +import com.google.inject.Binder; import com.google.inject.Injector; import com.google.inject.Key; import com.google.inject.Module; @@ -40,6 +41,7 @@ import org.apache.druid.data.input.impl.StringDimensionSchema; import org.apache.druid.discovery.NodeRole; import org.apache.druid.frame.testutil.FrameTestUtil; +import org.apache.druid.guice.DruidInjectorBuilder; import org.apache.druid.guice.DruidSecondaryModule; import org.apache.druid.guice.GuiceInjectors; import org.apache.druid.guice.IndexingServiceTuningConfigModule; @@ -55,6 +57,7 @@ import org.apache.druid.indexing.common.task.IndexTask; import org.apache.druid.indexing.common.task.batch.parallel.ParallelIndexTuningConfig; import org.apache.druid.initialization.CoreInjectorBuilder; +import org.apache.druid.initialization.DruidModule; import org.apache.druid.java.util.common.ISE; import org.apache.druid.java.util.common.Pair; import org.apache.druid.java.util.common.StringUtils; @@ -65,6 +68,7 @@ import org.apache.druid.java.util.common.io.Closer; import org.apache.druid.java.util.common.logger.Logger; import org.apache.druid.math.expr.ExprMacroTable; +import org.apache.druid.metadata.input.InputSourceModule; import org.apache.druid.msq.exec.Controller; import org.apache.druid.msq.exec.WorkerMemoryParameters; import org.apache.druid.msq.guice.MSQDurableStorageModule; @@ -129,6 +133,7 @@ import org.apache.druid.sql.SqlToolbox; import org.apache.druid.sql.calcite.BaseCalciteQueryTest; import org.apache.druid.sql.calcite.external.ExternalDataSource; +import org.apache.druid.sql.calcite.external.ExternalOperatorConversion; import org.apache.druid.sql.calcite.planner.CalciteRulesManager; import org.apache.druid.sql.calcite.planner.PlannerConfig; import org.apache.druid.sql.calcite.planner.PlannerFactory; @@ -141,6 +146,7 @@ import org.apache.druid.sql.calcite.util.SpecificSegmentsQuerySegmentWalker; import org.apache.druid.sql.calcite.util.SqlTestFramework; import org.apache.druid.sql.calcite.view.InProcessViewManager; +import org.apache.druid.sql.guice.SqlBindings; import org.apache.druid.storage.StorageConnector; import org.apache.druid.storage.StorageConnectorProvider; import org.apache.druid.storage.local.LocalFileStorageConnector; @@ -162,6 +168,7 @@ import javax.annotation.Nonnull; import javax.annotation.Nullable; + import java.io.Closeable; import java.io.File; import java.io.IOException; @@ -243,12 +250,54 @@ public class MSQTestBase extends BaseCalciteQueryTest ) ); + @Override + public void configureGuice(DruidInjectorBuilder builder) + { + super.configureGuice(builder); + + builder.addModule(new DruidModule() { + + // Small subset of MsqSqlModule + @Override + public void configure(Binder binder) + { + // We want this module to bring InputSourceModule along for the ride. + binder.install(new InputSourceModule()); + SqlBindings.addOperatorConversion(binder, ExternalOperatorConversion.class); + } + + @Override + public List getJacksonModules() + { + // We want this module to bring input sources along for the ride. + return new InputSourceModule().getJacksonModules(); + } + }); + } + @After public void tearDown2() { groupByBuffers.close(); } + // This test is a Frankenstein creation: it uses the injector set up by the + // SqlTestFramework to pull items from that are then used to create another + // injector that has the MSQ dependencies. This allows the test to create a + // "shadow" statement factory that is used for tests. It works... kinda. + // + // Better would be to sort through the Guice stuff and move it into the + // configureGuice() method above: use the SQL test framework injector so + // that everything is coordinated. Use the planner factory provided by that + // framework. + // + // Leaving well enough alone for now because any change should be done by + // someone familiar with the rather complex setup code below. + // + // One brute-force attempt ran afoul of circular dependencies: the SQL engine + // is created in the main injector, but it depends on the SegmentCacheManagerFactory + // which depends on the object mapper that the injector will provide, once it + // is built, but has not yet been build while we build the SQL engine. @Before public void setUp2() { @@ -291,7 +340,6 @@ public String getFormatString() GroupByQueryConfig groupByQueryConfig = new GroupByQueryConfig(); - binder.bind(DruidProcessingConfig.class).toInstance(druidProcessingConfig); binder.bind(new TypeLiteral>() { @@ -411,8 +459,8 @@ public String getFormatString() PlannerFactory plannerFactory = new PlannerFactory( rootSchema, - CalciteTests.createOperatorTable(), - CalciteTests.createExprMacroTable(), + qf.operatorTable(), + qf.macroTable(), PLANNER_CONFIG_DEFAULT, AuthTestUtils.TEST_AUTHORIZER_MAPPER, objectMapper, diff --git a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/test/MSQTestOverlordServiceClient.java b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/test/MSQTestOverlordServiceClient.java index 87a6f86ee85b..fc5ef645f7ad 100644 --- a/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/test/MSQTestOverlordServiceClient.java +++ b/extensions-core/multi-stage-query/src/test/java/org/apache/druid/msq/test/MSQTestOverlordServiceClient.java @@ -115,5 +115,4 @@ MSQSpec getQuerySpecForTask(String id) { return msqSpec.get(id); } - } diff --git a/extensions-core/stats/src/test/java/org/apache/druid/query/aggregation/variance/sql/VarianceSqlAggregatorTest.java b/extensions-core/stats/src/test/java/org/apache/druid/query/aggregation/variance/sql/VarianceSqlAggregatorTest.java index 03678a603cfc..bc1ef6816930 100644 --- a/extensions-core/stats/src/test/java/org/apache/druid/query/aggregation/variance/sql/VarianceSqlAggregatorTest.java +++ b/extensions-core/stats/src/test/java/org/apache/druid/query/aggregation/variance/sql/VarianceSqlAggregatorTest.java @@ -19,10 +19,8 @@ package org.apache.druid.query.aggregation.variance.sql; -import com.fasterxml.jackson.databind.Module; import com.google.common.collect.ImmutableList; -import com.google.common.collect.ImmutableSet; -import com.google.common.collect.Iterables; +import com.google.inject.Injector; import org.apache.druid.common.config.NullHandling; import org.apache.druid.data.input.InputRow; import org.apache.druid.data.input.impl.DimensionSchema; @@ -30,6 +28,7 @@ import org.apache.druid.data.input.impl.DoubleDimensionSchema; import org.apache.druid.data.input.impl.FloatDimensionSchema; import org.apache.druid.data.input.impl.LongDimensionSchema; +import org.apache.druid.guice.DruidInjectorBuilder; import org.apache.druid.java.util.common.granularity.Granularities; import org.apache.druid.math.expr.ExprMacroTable; import org.apache.druid.query.Druids; @@ -56,7 +55,6 @@ import org.apache.druid.segment.writeout.OffHeapMemorySegmentWriteOutMediumFactory; import org.apache.druid.sql.calcite.BaseCalciteQueryTest; import org.apache.druid.sql.calcite.filtration.Filtration; -import org.apache.druid.sql.calcite.planner.DruidOperatorTable; import org.apache.druid.sql.calcite.util.CalciteTests; import org.apache.druid.sql.calcite.util.SpecificSegmentsQuerySegmentWalker; import org.apache.druid.sql.calcite.util.TestDataBuilder; @@ -70,28 +68,18 @@ public class VarianceSqlAggregatorTest extends BaseCalciteQueryTest { - private static final DruidOperatorTable OPERATOR_TABLE = new DruidOperatorTable( - ImmutableSet.of( - new BaseVarianceSqlAggregator.VarPopSqlAggregator(), - new BaseVarianceSqlAggregator.VarSampSqlAggregator(), - new BaseVarianceSqlAggregator.VarianceSqlAggregator(), - new BaseVarianceSqlAggregator.StdDevPopSqlAggregator(), - new BaseVarianceSqlAggregator.StdDevSampSqlAggregator(), - new BaseVarianceSqlAggregator.StdDevSqlAggregator() - ), - ImmutableSet.of() - ); - @Override - public Iterable getJacksonModules() + public void configureGuice(DruidInjectorBuilder builder) { - return Iterables.concat(super.getJacksonModules(), new DruidStatsModule().getJacksonModules()); + super.configureGuice(builder); + builder.addModule(new DruidStatsModule()); } @Override public SpecificSegmentsQuerySegmentWalker createQuerySegmentWalker( final QueryRunnerFactoryConglomerate conglomerate, - final JoinableFactoryWrapper joinableFactory + final JoinableFactoryWrapper joinableFactory, + final Injector injector ) throws IOException { final QueryableIndex index = @@ -132,12 +120,6 @@ public SpecificSegmentsQuerySegmentWalker createQuerySegmentWalker( ); } - @Override - public DruidOperatorTable createOperatorTable() - { - return OPERATOR_TABLE; - } - public void addToHolder(VarianceAggregatorCollector holder, Object raw) { addToHolder(holder, raw, 1); diff --git a/extensions-core/testing-tools/src/test/java/org/apache/druid/query/sql/SleepSqlTest.java b/extensions-core/testing-tools/src/test/java/org/apache/druid/query/sql/SleepSqlTest.java index cb3e9efdf070..c3b1a102f3de 100644 --- a/extensions-core/testing-tools/src/test/java/org/apache/druid/query/sql/SleepSqlTest.java +++ b/extensions-core/testing-tools/src/test/java/org/apache/druid/query/sql/SleepSqlTest.java @@ -20,15 +20,11 @@ package org.apache.druid.query.sql; import com.google.common.collect.ImmutableList; -import com.google.common.collect.ImmutableSet; import org.apache.druid.common.config.NullHandling; -import org.apache.druid.guice.ExpressionModule; -import org.apache.druid.math.expr.ExprMacroTable; -import org.apache.druid.math.expr.ExprMacroTable.ExprMacro; +import org.apache.druid.guice.DruidInjectorBuilder; +import org.apache.druid.guice.SleepModule; import org.apache.druid.query.Druids; import org.apache.druid.query.TableDataSource; -import org.apache.druid.query.expression.LookupExprMacro; -import org.apache.druid.query.expressions.SleepExprMacro; import org.apache.druid.query.filter.BoundDimFilter; import org.apache.druid.query.ordering.StringComparators; import org.apache.druid.query.scan.ScanQuery.ResultFormat; @@ -36,34 +32,15 @@ import org.apache.druid.segment.virtual.ExpressionVirtualColumn; import org.apache.druid.sql.calcite.BaseCalciteQueryTest; import org.apache.druid.sql.calcite.filtration.Filtration; -import org.apache.druid.sql.calcite.planner.DruidOperatorTable; -import org.apache.druid.sql.calcite.util.CalciteTests; import org.junit.Test; -import java.util.ArrayList; -import java.util.List; - public class SleepSqlTest extends BaseCalciteQueryTest { @Override - public DruidOperatorTable createOperatorTable() - { - return new DruidOperatorTable( - ImmutableSet.of(), - ImmutableSet.of(new SleepOperatorConversion()) - ); - } - - @Override - public ExprMacroTable createMacroTable() + public void configureGuice(DruidInjectorBuilder builder) { - final List exprMacros = new ArrayList<>(); - for (Class clazz : ExpressionModule.EXPR_MACROS) { - exprMacros.add(CalciteTests.INJECTOR.getInstance(clazz)); - } - exprMacros.add(CalciteTests.INJECTOR.getInstance(LookupExprMacro.class)); - exprMacros.add(new SleepExprMacro()); - return new ExprMacroTable(exprMacros); + super.configureGuice(builder); + builder.addModule(new SleepModule()); } @Test @@ -80,7 +57,7 @@ public void testSleepFunction() "v0", "sleep(\"m1\")", ColumnType.STRING, - createMacroTable() + queryFramework().macroTable() ) ) .columns("v0") diff --git a/processing/src/main/java/org/apache/druid/guice/ExtensionsConfig.java b/processing/src/main/java/org/apache/druid/guice/ExtensionsConfig.java index c48d27eb588e..3af3db347aa1 100644 --- a/processing/src/main/java/org/apache/druid/guice/ExtensionsConfig.java +++ b/processing/src/main/java/org/apache/druid/guice/ExtensionsConfig.java @@ -28,6 +28,8 @@ */ public class ExtensionsConfig { + public static final String PROPERTY_BASE = "druid.extensions"; + @JsonProperty @NotNull private boolean searchCurrentClassloader = true; diff --git a/processing/src/main/java/org/apache/druid/guice/ExtensionsLoader.java b/processing/src/main/java/org/apache/druid/guice/ExtensionsLoader.java index b76bcb9a2cb4..0d3b035e7a6a 100644 --- a/processing/src/main/java/org/apache/druid/guice/ExtensionsLoader.java +++ b/processing/src/main/java/org/apache/druid/guice/ExtensionsLoader.java @@ -55,6 +55,7 @@ * any kind of extension that may be needed in the future. * The extensions are cached so that they can be reported by various REST APIs. */ +@LazySingleton public class ExtensionsLoader { private static final Logger log = new Logger(ExtensionsLoader.class); diff --git a/processing/src/main/java/org/apache/druid/guice/ExtensionsModule.java b/processing/src/main/java/org/apache/druid/guice/ExtensionsModule.java index 070bbc152ebf..74be479c2beb 100644 --- a/processing/src/main/java/org/apache/druid/guice/ExtensionsModule.java +++ b/processing/src/main/java/org/apache/druid/guice/ExtensionsModule.java @@ -35,8 +35,8 @@ public class ExtensionsModule implements Module public void configure(Binder binder) { binder.bind(ExtensionsLoader.class).in(LazySingleton.class); - JsonConfigProvider.bind(binder, "druid.extensions", ExtensionsConfig.class); - JsonConfigProvider.bind(binder, "druid.modules", ModulesConfig.class); + JsonConfigProvider.bind(binder, ExtensionsConfig.PROPERTY_BASE, ExtensionsConfig.class); + JsonConfigProvider.bind(binder, ModulesConfig.PROPERTY_BASE, ModulesConfig.class); } /** diff --git a/processing/src/main/java/org/apache/druid/guice/ModulesConfig.java b/processing/src/main/java/org/apache/druid/guice/ModulesConfig.java index 5ff8dee917bf..e7b69c4960a9 100644 --- a/processing/src/main/java/org/apache/druid/guice/ModulesConfig.java +++ b/processing/src/main/java/org/apache/druid/guice/ModulesConfig.java @@ -26,6 +26,8 @@ public class ModulesConfig { + public static final String PROPERTY_BASE = "druid.modules"; + /** * Canonical class names of modules, which should not be loaded despite they are founded in extensions from {@link * ExtensionsConfig#loadList} or the standard list of modules loaded by some node type, e. g. {@code diff --git a/processing/src/main/java/org/apache/druid/guice/StartupInjectorBuilder.java b/processing/src/main/java/org/apache/druid/guice/StartupInjectorBuilder.java index 13d116b60071..e091e6eb5229 100644 --- a/processing/src/main/java/org/apache/druid/guice/StartupInjectorBuilder.java +++ b/processing/src/main/java/org/apache/druid/guice/StartupInjectorBuilder.java @@ -19,8 +19,10 @@ package org.apache.druid.guice; +import com.google.inject.util.Providers; import org.apache.druid.jackson.JacksonModule; import org.apache.druid.math.expr.ExpressionProcessingModule; +import org.apache.druid.utils.RuntimeInfo; import java.util.Arrays; import java.util.Properties; @@ -80,4 +82,19 @@ public StartupInjectorBuilder forServer() ); return this; } + + /** + * Configure the injector to not load server-only classes by binding those + * classes to providers of null values. Avoids accidental dependencies of + * test code on classes not intended for classes by preventing Guice from + * helpfully providing implicit instances. + */ + public StartupInjectorBuilder forTests() + { + add(binder -> { + binder.bind(ExtensionsLoader.class).toProvider(Providers.of(null)); + binder.bind(RuntimeInfo.class).toProvider(Providers.of(null)); + }); + return this; + } } diff --git a/processing/src/main/java/org/apache/druid/jackson/DefaultObjectMapper.java b/processing/src/main/java/org/apache/druid/jackson/DefaultObjectMapper.java index ac9387b11ef6..7e530836c2f7 100644 --- a/processing/src/main/java/org/apache/druid/jackson/DefaultObjectMapper.java +++ b/processing/src/main/java/org/apache/druid/jackson/DefaultObjectMapper.java @@ -42,6 +42,8 @@ */ public class DefaultObjectMapper extends ObjectMapper { + public static final DefaultObjectMapper INSTANCE = new DefaultObjectMapper(); + public DefaultObjectMapper() { this((JsonFactory) null, null); diff --git a/processing/src/test/java/org/apache/druid/guice/StartupInjectorBuilderTest.java b/processing/src/test/java/org/apache/druid/guice/StartupInjectorBuilderTest.java new file mode 100644 index 000000000000..a20cf33c7c06 --- /dev/null +++ b/processing/src/test/java/org/apache/druid/guice/StartupInjectorBuilderTest.java @@ -0,0 +1,138 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.guice; + +import com.google.inject.Injector; +import org.apache.druid.utils.RuntimeInfo; +import org.junit.Test; + +import java.util.Collections; +import java.util.Properties; + +import static org.junit.Assert.assertEquals; +import static org.junit.Assert.assertNotNull; +import static org.junit.Assert.assertNotSame; +import static org.junit.Assert.assertNull; +import static org.junit.Assert.assertSame; +import static org.junit.Assert.assertTrue; + +public class StartupInjectorBuilderTest +{ + @Test + public void testEmpty() + { + Injector injector = new StartupInjectorBuilder().build(); + + // Empty properties come along for free + Properties props = injector.getInstance(Properties.class); + assertNotNull(props); + assertTrue(props.isEmpty()); + + // Since we didn't configure this item, we get a new instance every time. + assertNotSame(props, injector.getInstance(Properties.class)); + + // Runtime info is available, though not configured, because Guice can create + // one when requested. Our class, so marked singleton. + assertNotNull(injector.getInstance(RuntimeInfo.class)); + assertSame(injector.getInstance(RuntimeInfo.class), injector.getInstance(RuntimeInfo.class)); + + // The extension loader is available, again via implicit creation. + // Since it is our class, we marked it as a lazy singleton. + assertNotNull(injector.getInstance(ExtensionsLoader.class)); + assertSame(injector.getInstance(ExtensionsLoader.class), injector.getInstance(ExtensionsLoader.class)); + + // Does have the basics. Sample one such entry. + assertNotNull(injector.getInstance(DruidSecondaryModule.class)); + assertSame(injector.getInstance(DruidSecondaryModule.class), injector.getInstance(DruidSecondaryModule.class)); + } + + @Test + public void testEmptyTestInjector() + { + Injector injector = new StartupInjectorBuilder().forTests().build(); + + // Empty properties come along for free + Properties props = injector.getInstance(Properties.class); + assertNotNull(props); + assertTrue(props.isEmpty()); + + // Since we didn't configure this item, we get a new instance every time. + assertNotSame(props, injector.getInstance(Properties.class)); + + // Runtime info bound to null. + assertNull(injector.getInstance(RuntimeInfo.class)); + + // The extension loader bound to null. + assertNull(injector.getInstance(ExtensionsLoader.class)); + + // Does have the basics. Sample one such entry. + assertNotNull(injector.getInstance(DruidSecondaryModule.class)); + assertSame(injector.getInstance(DruidSecondaryModule.class), injector.getInstance(DruidSecondaryModule.class)); + } + + @Test + public void testEmptyProperties() + { + Injector injector = new StartupInjectorBuilder() + .withEmptyProperties() + .build(); + + // Single empty properties instance + Properties props = injector.getInstance(Properties.class); + assertNotNull(props); + assertTrue(props.isEmpty()); + + // Since we didn't configure this item, we get a new instance every time. + assertSame(props, injector.getInstance(Properties.class)); + } + + @Test + public void testExplicitProperties() + { + Properties props = new Properties(); + props.put("foo", "bar"); + Injector injector = new StartupInjectorBuilder() + .forTests() + .withProperties(props) + .build(); + + // Returns explicit properties + Properties propsInstance = injector.getInstance(Properties.class); + assertSame(props, propsInstance); + } + + @Test + public void testExtensionsOption() + { + Properties props = new Properties(); + props.put(ExtensionsConfig.PROPERTY_BASE + ".directory", "bogus"); + props.put(ModulesConfig.PROPERTY_BASE + ".excludeList", "[\"excluded\"]"); + Injector injector = new StartupInjectorBuilder() + .withExtensions() + .withProperties(props) + .build(); + + // Extensions config is populated. (Can't tests extensions themselves.) + assertEquals("bogus", injector.getInstance(ExtensionsConfig.class).getDirectory()); + assertEquals(Collections.singletonList("excluded"), injector.getInstance(ModulesConfig.class).getExcludeList()); + } + + // Can't test the server option here: there are no actual property files to read. +} diff --git a/server/src/main/java/org/apache/druid/catalog/model/TableDefnRegistry.java b/server/src/main/java/org/apache/druid/catalog/model/TableDefnRegistry.java index 7e952a4393a3..4aa9c46aa16a 100644 --- a/server/src/main/java/org/apache/druid/catalog/model/TableDefnRegistry.java +++ b/server/src/main/java/org/apache/druid/catalog/model/TableDefnRegistry.java @@ -28,6 +28,8 @@ import org.apache.druid.catalog.model.table.LocalTableDefn; import org.apache.druid.java.util.common.IAE; +import javax.inject.Inject; + import java.util.Map; /** @@ -75,6 +77,7 @@ public TableDefnRegistry( this.jsonMapper = jsonMapper; } + @Inject public TableDefnRegistry( final ObjectMapper jsonMapper ) diff --git a/server/src/main/java/org/apache/druid/guice/DruidInjectorBuilder.java b/server/src/main/java/org/apache/druid/guice/DruidInjectorBuilder.java index 81e28431e2db..b650c100669a 100644 --- a/server/src/main/java/org/apache/druid/guice/DruidInjectorBuilder.java +++ b/server/src/main/java/org/apache/druid/guice/DruidInjectorBuilder.java @@ -61,6 +61,7 @@ public class DruidInjectorBuilder private final ObjectMapper smileMapper; private final Set nodeRoles; private final ModulesConfig modulesConfig; + private boolean ignoreLoadScopes; public DruidInjectorBuilder(final Injector baseInjector) { @@ -83,6 +84,18 @@ public DruidInjectorBuilder(final DruidInjectorBuilder from) this.modulesConfig = from.modulesConfig; this.jsonMapper = from.jsonMapper; this.smileMapper = from.smileMapper; + this.ignoreLoadScopes = from.ignoreLoadScopes; + } + + /** + * Ignore load scope annotations on modules. Primarily for testing where a unit + * test is not any Druid node, and may wish to load a module that is annotated + * with a load scope. + */ + public DruidInjectorBuilder ignoreLoadScopes() + { + this.ignoreLoadScopes = true; + return this; } /** @@ -127,7 +140,7 @@ public DruidInjectorBuilder addInput(Object input) } else if (input instanceof Class) { return addClass((Class) input); } else { - throw new ISE("Unknown module type[%s]", input.getClass()); + throw new ISE("Unknown module type [%s]", input.getClass()); } } @@ -176,6 +189,12 @@ private boolean acceptModule(Class moduleClass) log.info("Not loading module [%s] because it is present in excludeList", moduleClassName); return false; } + + // Tests don't have node roles, and so want to load the given modules + // regardless of the node roles provided. + if (ignoreLoadScopes) { + return true; + } LoadScope loadScope = moduleClass.getAnnotation(LoadScope.class); if (loadScope == null) { // always load if annotation is not specified @@ -204,4 +223,9 @@ public Injector build() { return Guice.createInjector(modules); } + + public Injector baseInjector() + { + return baseInjector; + } } diff --git a/server/src/test/java/org/apache/druid/guice/DruidInjectorBuilderTest.java b/server/src/test/java/org/apache/druid/guice/DruidInjectorBuilderTest.java new file mode 100644 index 000000000000..d55ebe8f16b1 --- /dev/null +++ b/server/src/test/java/org/apache/druid/guice/DruidInjectorBuilderTest.java @@ -0,0 +1,308 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.guice; + +import com.fasterxml.jackson.annotation.JsonTypeInfo; +import com.fasterxml.jackson.annotation.JsonTypeName; +import com.fasterxml.jackson.databind.Module; +import com.fasterxml.jackson.databind.ObjectMapper; +import com.fasterxml.jackson.databind.module.SimpleModule; +import com.google.common.collect.ImmutableList; +import com.google.common.collect.ImmutableSet; +import com.google.inject.Binder; +import com.google.inject.Injector; +import com.google.inject.Key; +import org.apache.druid.discovery.NodeRole; +import org.apache.druid.guice.annotations.Json; +import org.apache.druid.guice.annotations.LoadScope; +import org.apache.druid.initialization.CoreInjectorBuilder; +import org.apache.druid.initialization.DruidModule; +import org.apache.druid.java.util.common.ISE; +import org.junit.Test; + +import javax.inject.Inject; +import java.io.IOException; +import java.util.Arrays; +import java.util.List; +import java.util.Properties; + +import static org.junit.Assert.assertSame; +import static org.junit.Assert.assertThrows; +import static org.junit.Assert.assertTrue; + +public class DruidInjectorBuilderTest +{ + @JsonTypeInfo(use = JsonTypeInfo.Id.NAME, property = "type") + public static class MockObject + { + } + + @JsonTypeName("extn") + public static class MockObjectExtension extends MockObject + { + } + + public interface MockInterface + { + } + + public static class MockComponent implements MockInterface + { + } + + private static class MockGuiceModule implements com.google.inject.Module + { + @Inject + public Properties properties; + + @Override + public void configure(Binder binder) + { + binder.bind(MockInterface.class).to(MockComponent.class).in(LazySingleton.class); + } + } + + private static class MockDruidModule implements DruidModule + { + @Inject + public Properties properties; + + @Override + public void configure(Binder binder) + { + } + + @Override + public List getJacksonModules() + { + return ImmutableList.of( + new SimpleModule("MockModule").registerSubtypes(MockObjectExtension.class) + ); + } + } + + @LoadScope(roles = NodeRole.BROKER_JSON_NAME) + private static class MockRoleModule extends MockDruidModule + { + } + + @Test + public void testEmpty() + { + Properties props = new Properties(); + props.put("foo", "bar"); + Injector injector = new CoreInjectorBuilder( + new StartupInjectorBuilder() + .forTests() + .withProperties(props) + .build() + ).build(); + + // Returns explicit properties + Properties propsInstance = injector.getInstance(Properties.class); + assertSame(props, propsInstance); + } + + /** + * Test the most generic form: addInput. Calls addModule() internally. + */ + @Test + public void testAddInputModules() throws IOException + { + Properties props = new Properties(); + props.put("foo", "bar"); + MockGuiceModule guiceModule = new MockGuiceModule(); + MockDruidModule druidModule = new MockDruidModule(); + Injector injector = new CoreInjectorBuilder( + new StartupInjectorBuilder() + .forTests() + .withProperties(props) + .build() + ) + .addInput(guiceModule) + .addInput(druidModule) + .build(); + + // Verify injection occurred + assertSame(props, guiceModule.properties); + assertSame(props, druidModule.properties); + verifyInjector(injector); + } + + private void verifyInjector(Injector injector) throws IOException + { + // Guice module did its thing + assertTrue(injector.getInstance(MockInterface.class) instanceof MockComponent); + + // And that the Druid module set up Jackson. + String json = "{\"type\": \"extn\"}"; + ObjectMapper om = injector.getInstance(Key.get(ObjectMapper.class, Json.class)); + MockObject obj = om.readValue(json, MockObject.class); + assertTrue(obj instanceof MockObjectExtension); + } + + /** + * Test the ability to pass module classes rather than instances. + */ + @Test + public void testAddInputClasses() throws IOException + { + Properties props = new Properties(); + props.put("foo", "bar"); + Injector injector = new CoreInjectorBuilder( + new StartupInjectorBuilder() + .forTests() + .withProperties(props) + .build() + ) + .addInput(MockGuiceModule.class) + .addInput(MockDruidModule.class) + .build(); + + // Can't verify injection here, sadly + + verifyInjector(injector); + } + + @Test + public void testBadModule() + { + DruidInjectorBuilder builder = new CoreInjectorBuilder( + new StartupInjectorBuilder() + .forTests() + .withEmptyProperties() + .build() + ); + assertThrows(ISE.class, () -> builder.addInput("I'm not a module")); + } + + @Test + public void testBadModuleClass() + { + DruidInjectorBuilder builder = new CoreInjectorBuilder( + new StartupInjectorBuilder() + .forTests() + .withEmptyProperties() + .build() + ); + assertThrows(ISE.class, () -> builder.addInput(Object.class)); + } + + @Test + public void testAddModules() throws IOException + { + Injector injector = new CoreInjectorBuilder( + new StartupInjectorBuilder() + .forTests() + .withEmptyProperties() + .build() + ) + .addModules(new MockGuiceModule(), new MockDruidModule()) + .build(); + + verifyInjector(injector); + } + + @Test + public void testAddAll() throws IOException + { + Injector injector = new CoreInjectorBuilder( + new StartupInjectorBuilder() + .forTests() + .withEmptyProperties() + .build() + ) + .addAll(Arrays.asList(new MockGuiceModule(), new MockDruidModule())) + .build(); + + verifyInjector(injector); + } + + /** + * Enable extensions. Then, exclude our JSON test module. As a result, the + * JSON object will fail to deserialize. + */ + @Test + public void testExclude() + { + Properties props = new Properties(); + props.put(ModulesConfig.PROPERTY_BASE + ".excludeList", "[\"" + MockDruidModule.class.getName() + "\"]"); + Injector injector = new CoreInjectorBuilder( + new StartupInjectorBuilder() + .withExtensions() + .withProperties(props) + .build() + ) + .addInput(MockGuiceModule.class) + .addInput(MockDruidModule.class) + .build(); + + assertThrows(IOException.class, () -> verifyInjector(injector)); + } + + @Test + public void testMatchingNodeRole() throws IOException + { + Injector injector = new CoreInjectorBuilder( + new StartupInjectorBuilder() + .forTests() + .withEmptyProperties() + .build(), + ImmutableSet.of(NodeRole.BROKER) + ) + .addModules(new MockGuiceModule(), new MockRoleModule()) + .build(); + + verifyInjector(injector); + } + + @Test + public void testNotMatchingNodeRole() + { + Injector injector = new CoreInjectorBuilder( + new StartupInjectorBuilder() + .forTests() + .withEmptyProperties() + .build(), + ImmutableSet.of(NodeRole.COORDINATOR) + ) + .addModules(new MockGuiceModule(), new MockRoleModule()) + .build(); + + assertThrows(IOException.class, () -> verifyInjector(injector)); + } + + @Test + public void testIgnoreNodeRole() throws IOException + { + Injector injector = new CoreInjectorBuilder( + new StartupInjectorBuilder() + .forTests() + .withEmptyProperties() + .build(), + ImmutableSet.of(NodeRole.COORDINATOR) + ) + .ignoreLoadScopes() + .addModules(new MockGuiceModule(), new MockRoleModule()) + .build(); + + verifyInjector(injector); + } +} diff --git a/server/src/test/java/org/apache/druid/server/StatusResourceTest.java b/server/src/test/java/org/apache/druid/server/StatusResourceTest.java index 52fb0c80d532..8ea8da4b78c7 100644 --- a/server/src/test/java/org/apache/druid/server/StatusResourceTest.java +++ b/server/src/test/java/org/apache/druid/server/StatusResourceTest.java @@ -22,9 +22,9 @@ import com.fasterxml.jackson.core.type.TypeReference; import com.fasterxml.jackson.databind.ObjectMapper; import com.google.common.collect.ImmutableList; -import com.google.inject.Guice; import com.google.inject.Injector; import org.apache.druid.guice.PropertiesModule; +import org.apache.druid.guice.StartupInjectorBuilder; import org.apache.druid.initialization.DruidModule; import org.apache.druid.initialization.ServerInjectorBuilderTest; import org.apache.druid.java.util.common.StringUtils; @@ -78,8 +78,9 @@ public void testHiddenPropertiesContain() throws Exception private void testHiddenPropertiesWithPropertyFileName(String fileName) throws Exception { - Injector injector = Guice.createInjector(Collections.singletonList(new PropertiesModule(Collections.singletonList( - fileName)))); + Injector injector = new StartupInjectorBuilder() + .add(new PropertiesModule(Collections.singletonList(fileName))) + .build(); Map returnedProperties = injector.getInstance(StatusResource.class).getProperties(); Set lowerCasePropertyNames = returnedProperties.keySet() .stream() diff --git a/sql/src/main/java/org/apache/druid/sql/AbstractStatement.java b/sql/src/main/java/org/apache/druid/sql/AbstractStatement.java index 1956b353b770..05152625d8c3 100644 --- a/sql/src/main/java/org/apache/druid/sql/AbstractStatement.java +++ b/sql/src/main/java/org/apache/druid/sql/AbstractStatement.java @@ -32,6 +32,7 @@ import org.apache.druid.server.security.ResourceType; import org.apache.druid.sql.calcite.planner.DruidPlanner; import org.apache.druid.sql.calcite.planner.PlannerContext; +import org.apache.druid.sql.calcite.planner.PlannerHook; import java.io.Closeable; import java.util.HashMap; @@ -71,6 +72,7 @@ public abstract class AbstractStatement implements Closeable protected final Map queryContext; protected PlannerContext plannerContext; protected DruidPlanner.AuthResult authResult; + protected PlannerHook hook; public AbstractStatement( final SqlToolbox sqlToolbox, @@ -109,6 +111,15 @@ public Map context() return queryContext; } + /** + * Set the hook which can capture planner artifacts during planning. Primarily used + * for testing. Defaults to a "no op" hook that does nothing. + */ + public void setHook(PlannerHook hook) + { + this.hook = hook; + } + /** * Validate SQL query and authorize against any datasources or views which * will take part in the query. Must be called by the API methods, not diff --git a/sql/src/main/java/org/apache/druid/sql/DirectStatement.java b/sql/src/main/java/org/apache/druid/sql/DirectStatement.java index d3d9371a3968..ad24274ceb09 100644 --- a/sql/src/main/java/org/apache/druid/sql/DirectStatement.java +++ b/sql/src/main/java/org/apache/druid/sql/DirectStatement.java @@ -206,7 +206,8 @@ public ResultSet plan() try (DruidPlanner planner = sqlToolbox.plannerFactory.createPlanner( sqlToolbox.engine, queryPlus.sql(), - queryContext + queryContext, + hook )) { validate(planner); authorize(planner, authorizer()); diff --git a/sql/src/main/java/org/apache/druid/sql/PreparedStatement.java b/sql/src/main/java/org/apache/druid/sql/PreparedStatement.java index 1774aa31feda..bdbccb1ff6f1 100644 --- a/sql/src/main/java/org/apache/druid/sql/PreparedStatement.java +++ b/sql/src/main/java/org/apache/druid/sql/PreparedStatement.java @@ -66,9 +66,12 @@ public PreparedStatement( public PrepareResult prepare() { try (DruidPlanner planner = sqlToolbox.plannerFactory.createPlanner( - sqlToolbox.engine, - queryPlus.sql(), - queryContext)) { + sqlToolbox.engine, + queryPlus.sql(), + queryContext, + hook + ) + ) { validate(planner); authorize(planner, authorizer()); diff --git a/sql/src/main/java/org/apache/druid/sql/calcite/planner/DruidPlanner.java b/sql/src/main/java/org/apache/druid/sql/calcite/planner/DruidPlanner.java index 1d34713c9e86..b0912507a278 100644 --- a/sql/src/main/java/org/apache/druid/sql/calcite/planner/DruidPlanner.java +++ b/sql/src/main/java/org/apache/druid/sql/calcite/planner/DruidPlanner.java @@ -94,6 +94,7 @@ public AuthResult( private final CalcitePlanner planner; private final PlannerContext plannerContext; private final SqlEngine engine; + private final PlannerHook hook; private State state = State.START; private SqlStatementHandler handler; private boolean authorized; @@ -101,13 +102,15 @@ public AuthResult( DruidPlanner( final FrameworkConfig frameworkConfig, final PlannerContext plannerContext, - final SqlEngine engine + final SqlEngine engine, + final PlannerHook hook ) { this.frameworkConfig = frameworkConfig; this.planner = new CalcitePlanner(frameworkConfig); this.plannerContext = plannerContext; this.engine = engine; + this.hook = hook == null ? NoOpPlannerHook.INSTANCE : hook; } /** @@ -124,7 +127,9 @@ public void validate() throws SqlParseException, ValidationException engine.validateContext(plannerContext.queryContextMap()); // Parse the query string. - SqlNode root = planner.parse(plannerContext.getSql()); + String sql = plannerContext.getSql(); + hook.captureSql(sql); + SqlNode root = planner.parse(sql); handler = createHandler(root); try { @@ -162,7 +167,6 @@ private SqlStatementHandler createHandler(final SqlNode node) throws ValidationE throw new ValidationException(StringUtils.format("Cannot execute [%s].", node.getKind())); } - /** * Prepare a SQL query for execution, including some initial parsing and * validation and any dynamic parameter type resolution, to support prepared @@ -295,5 +299,11 @@ public DateTimeZone timeZone() { return plannerContext.getTimeZone(); } + + @Override + public PlannerHook hook() + { + return hook; + } } } diff --git a/sql/src/main/java/org/apache/druid/sql/calcite/planner/IngestHandler.java b/sql/src/main/java/org/apache/druid/sql/calcite/planner/IngestHandler.java index 80dad64f89f2..f2d98fb70cde 100644 --- a/sql/src/main/java/org/apache/druid/sql/calcite/planner/IngestHandler.java +++ b/sql/src/main/java/org/apache/druid/sql/calcite/planner/IngestHandler.java @@ -74,6 +74,7 @@ public abstract class IngestHandler extends QueryHandler { super(handlerContext, queryNode, explain); this.ingestionGranularity = ingestNode.getPartitionedBy(); + handlerContext.hook().captureInsert(ingestNode); } protected static SqlNode convertQuery(DruidSqlIngest sqlNode) throws ValidationException diff --git a/sql/src/main/java/org/apache/druid/sql/calcite/planner/NoOpPlannerHook.java b/sql/src/main/java/org/apache/druid/sql/calcite/planner/NoOpPlannerHook.java new file mode 100644 index 000000000000..65b0dccef0d5 --- /dev/null +++ b/sql/src/main/java/org/apache/druid/sql/calcite/planner/NoOpPlannerHook.java @@ -0,0 +1,61 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.sql.calcite.planner; + +import org.apache.calcite.interpreter.BindableRel; +import org.apache.calcite.rel.RelRoot; +import org.apache.calcite.rel.type.RelDataType; +import org.apache.calcite.sql.SqlInsert; +import org.apache.druid.sql.calcite.rel.DruidRel; + +public class NoOpPlannerHook implements PlannerHook +{ + public static final NoOpPlannerHook INSTANCE = new NoOpPlannerHook(); + + @Override + public void captureSql(String sql) + { + } + + @Override + public void captureQueryRel(RelRoot rootQueryRel) + { + } + + @Override + public void captureDruidRel(DruidRel druidRel) + { + } + + @Override + public void captureBindableRel(BindableRel bindableRel) + { + } + + @Override + public void captureParameterTypes(RelDataType parameterTypes) + { + } + + @Override + public void captureInsert(SqlInsert insert) + { + } +} diff --git a/sql/src/main/java/org/apache/druid/sql/calcite/planner/PlannerCaptureHook.java b/sql/src/main/java/org/apache/druid/sql/calcite/planner/PlannerCaptureHook.java new file mode 100644 index 000000000000..bdf50a8a0c58 --- /dev/null +++ b/sql/src/main/java/org/apache/druid/sql/calcite/planner/PlannerCaptureHook.java @@ -0,0 +1,78 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.sql.calcite.planner; + +import org.apache.calcite.interpreter.BindableRel; +import org.apache.calcite.rel.RelRoot; +import org.apache.calcite.rel.type.RelDataType; +import org.apache.calcite.sql.SqlInsert; +import org.apache.druid.sql.calcite.rel.DruidRel; + +public class PlannerCaptureHook implements PlannerHook +{ + private RelRoot relRoot; + private SqlInsert insertNode; + + @Override + public void captureSql(String sql) + { + // Not used at present. Add a field to capture this if you need it. + } + + @Override + public void captureQueryRel(RelRoot rootQueryRel) + { + this.relRoot = rootQueryRel; + } + + @Override + public void captureDruidRel(DruidRel druidRel) + { + // Not used at present. Add a field to capture this if you need it. + } + + @Override + public void captureBindableRel(BindableRel bindableRel) + { + // Not used at present. Add a field to capture this if you need it. + } + + @Override + public void captureParameterTypes(RelDataType parameterTypes) + { + // Not used at present. Add a field to capture this if you need it. + } + + @Override + public void captureInsert(SqlInsert insert) + { + this.insertNode = insert; + } + + public RelRoot relRoot() + { + return relRoot; + } + + public SqlInsert insertNode() + { + return insertNode; + } +} diff --git a/sql/src/main/java/org/apache/druid/sql/calcite/planner/PlannerFactory.java b/sql/src/main/java/org/apache/druid/sql/calcite/planner/PlannerFactory.java index 1e8a6c57a029..3cbc20dd05dd 100644 --- a/sql/src/main/java/org/apache/druid/sql/calcite/planner/PlannerFactory.java +++ b/sql/src/main/java/org/apache/druid/sql/calcite/planner/PlannerFactory.java @@ -104,7 +104,8 @@ public PlannerFactory( public DruidPlanner createPlanner( final SqlEngine engine, final String sql, - final Map queryContext + final Map queryContext, + final PlannerHook hook ) { final PlannerContext context = PlannerContext.create( @@ -119,7 +120,7 @@ public DruidPlanner createPlanner( joinableFactoryWrapper ); - return new DruidPlanner(buildFrameworkConfig(context), context, engine); + return new DruidPlanner(buildFrameworkConfig(context), context, engine, hook); } /** @@ -129,7 +130,7 @@ public DruidPlanner createPlanner( @VisibleForTesting public DruidPlanner createPlannerForTesting(final SqlEngine engine, final String sql, final Map queryContext) { - final DruidPlanner thePlanner = createPlanner(engine, sql, queryContext); + final DruidPlanner thePlanner = createPlanner(engine, sql, queryContext, null); thePlanner.getPlannerContext() .setAuthenticationResult(NoopEscalator.getInstance().createEscalatedAuthenticationResult()); try { diff --git a/sql/src/main/java/org/apache/druid/sql/calcite/planner/PlannerHook.java b/sql/src/main/java/org/apache/druid/sql/calcite/planner/PlannerHook.java new file mode 100644 index 000000000000..a65f59d4d1c7 --- /dev/null +++ b/sql/src/main/java/org/apache/druid/sql/calcite/planner/PlannerHook.java @@ -0,0 +1,43 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.sql.calcite.planner; + +import org.apache.calcite.interpreter.BindableRel; +import org.apache.calcite.rel.RelRoot; +import org.apache.calcite.rel.type.RelDataType; +import org.apache.calcite.sql.SqlInsert; +import org.apache.druid.sql.calcite.rel.DruidRel; + +/** + * Druid-specific version of Calcite's {@link org.apache.calcite.runtime.Hook Hook} + * class. Captures artifacts of interest in the Druid planning process, generally + * for test validation. Calcite's hook has multiple low-level events, but, sadly, + * none at the points where tests want to verify, except for the opportunity to + * capture the native query. + */ +public interface PlannerHook +{ + void captureSql(String sql); + void captureQueryRel(RelRoot rootQueryRel); + void captureDruidRel(DruidRel druidRel); + void captureBindableRel(BindableRel bindableRel); + void captureParameterTypes(RelDataType parameterTypes); + void captureInsert(SqlInsert insert); +} diff --git a/sql/src/main/java/org/apache/druid/sql/calcite/planner/QueryHandler.java b/sql/src/main/java/org/apache/druid/sql/calcite/planner/QueryHandler.java index eb1bd43ff625..13daa80da129 100644 --- a/sql/src/main/java/org/apache/druid/sql/calcite/planner/QueryHandler.java +++ b/sql/src/main/java/org/apache/druid/sql/calcite/planner/QueryHandler.java @@ -146,9 +146,11 @@ public void prepare() } isPrepared = true; rootQueryRel = handlerContext.planner().rel(validatedQueryNode); + handlerContext.hook().captureQueryRel(rootQueryRel); final RelDataTypeFactory typeFactory = rootQueryRel.rel.getCluster().getTypeFactory(); final SqlValidator validator = handlerContext.planner().getValidator(); final RelDataType parameterTypes = validator.getParameterRowType(validatedQueryNode); + handlerContext.hook().captureParameterTypes(parameterTypes); final RelDataType returnedRowType; if (explain != null) { @@ -291,6 +293,7 @@ private PlannerResult planWithBindableConvention() ); } + handlerContext.hook().captureBindableRel(bindableRel); PlannerContext plannerContext = handlerContext.plannerContext(); if (explain != null) { return planExplanation(bindableRel, false); @@ -466,6 +469,7 @@ private void flattenOutermostRel(DruidRel druidRel, List> flatten protected PlannerResult planWithDruidConvention() throws ValidationException { final RelRoot possiblyLimitedRoot = possiblyWrapRootWithOuterLimitFromContext(rootQueryRel); + handlerContext.hook().captureQueryRel(possiblyLimitedRoot); final QueryMaker queryMaker = buildQueryMaker(possiblyLimitedRoot); PlannerContext plannerContext = handlerContext.plannerContext(); plannerContext.setQueryMaker(queryMaker); @@ -490,6 +494,7 @@ protected PlannerResult planWithDruidConvention() throws ValidationException .plus(rootQueryRel.collation), parameterized ); + handlerContext.hook().captureDruidRel(druidRel); if (explain != null) { return planExplanation(druidRel, true); diff --git a/sql/src/main/java/org/apache/druid/sql/calcite/planner/SqlStatementHandler.java b/sql/src/main/java/org/apache/druid/sql/calcite/planner/SqlStatementHandler.java index 9185b9862f33..af28ceb3b979 100644 --- a/sql/src/main/java/org/apache/druid/sql/calcite/planner/SqlStatementHandler.java +++ b/sql/src/main/java/org/apache/druid/sql/calcite/planner/SqlStatementHandler.java @@ -57,6 +57,7 @@ interface HandlerContext SchemaPlus defaultSchema(); ObjectMapper jsonMapper(); DateTimeZone timeZone(); + PlannerHook hook(); } abstract class BaseStatementHandler implements SqlStatementHandler diff --git a/sql/src/main/java/org/apache/druid/sql/calcite/view/DruidViewMacro.java b/sql/src/main/java/org/apache/druid/sql/calcite/view/DruidViewMacro.java index 75d82fa03fc9..e584108f9d6e 100644 --- a/sql/src/main/java/org/apache/druid/sql/calcite/view/DruidViewMacro.java +++ b/sql/src/main/java/org/apache/druid/sql/calcite/view/DruidViewMacro.java @@ -61,7 +61,10 @@ public TranslatableTable apply(final List arguments) plannerFactory.createPlanner( ViewSqlEngine.INSTANCE, viewSql, - Collections.emptyMap())) { + Collections.emptyMap(), + null + ) + ) { planner.validate(); rowType = planner.prepare().getValidatedRowType(); } diff --git a/sql/src/test/java/org/apache/druid/sql/calcite/BaseCalciteQueryTest.java b/sql/src/test/java/org/apache/druid/sql/calcite/BaseCalciteQueryTest.java index 150309e374b7..fada41896009 100644 --- a/sql/src/test/java/org/apache/druid/sql/calcite/BaseCalciteQueryTest.java +++ b/sql/src/test/java/org/apache/druid/sql/calcite/BaseCalciteQueryTest.java @@ -20,11 +20,10 @@ package org.apache.druid.sql.calcite; import com.fasterxml.jackson.core.JsonProcessingException; -import com.fasterxml.jackson.databind.Module; import com.fasterxml.jackson.databind.ObjectMapper; -import com.google.common.base.Preconditions; import com.google.common.collect.ImmutableList; import com.google.common.collect.ImmutableMap; +import com.google.inject.Injector; import org.apache.calcite.plan.RelOptPlanner; import org.apache.druid.annotations.UsedByJUnitParamsRunner; import org.apache.druid.common.config.NullHandling; @@ -37,7 +36,6 @@ import org.apache.druid.java.util.common.granularity.Granularity; import org.apache.druid.java.util.common.io.Closer; import org.apache.druid.java.util.common.logger.Logger; -import org.apache.druid.math.expr.ExprMacroTable; import org.apache.druid.query.DataSource; import org.apache.druid.query.Druids; import org.apache.druid.query.JoinDataSource; @@ -84,7 +82,6 @@ import org.apache.druid.sql.SqlStatementFactory; import org.apache.druid.sql.calcite.expression.DruidExpression; import org.apache.druid.sql.calcite.planner.Calcites; -import org.apache.druid.sql.calcite.planner.DruidOperatorTable; import org.apache.druid.sql.calcite.planner.PlannerConfig; import org.apache.druid.sql.calcite.planner.PlannerContext; import org.apache.druid.sql.calcite.planner.PlannerFactory; @@ -119,11 +116,11 @@ import javax.annotation.Nullable; import java.io.IOException; -import java.util.ArrayList; import java.util.Arrays; import java.util.HashMap; import java.util.List; import java.util.Map; +import java.util.Properties; import java.util.Set; import java.util.function.Consumer; import java.util.stream.Collectors; @@ -488,7 +485,12 @@ protected static void resetFramework() @Rule public QueryLogHook getQueryLogHook() { - return queryLogHook = QueryLogHook.create(queryFramework().queryJsonMapper()); + // Indirection for the JSON mapper. Otherwise, this rule method is called + // before Setup is called, causing the query framework to be built before + // tests have done their setup. The indirection means we access the query + // framework only when we log the first query. By then, the query framework + // will have been created via the normal path. + return queryLogHook = new QueryLogHook(() -> queryFramework().queryJsonMapper()); } public SqlTestFramework queryFramework() @@ -512,25 +514,31 @@ private void createFramework(int mergeBufferCount) resetFramework(); try { baseComponentSupplier = new StandardComponentSupplier( - CalciteTests.INJECTOR, - temporaryFolder.newFolder()); + temporaryFolder.newFolder() + ); } catch (IOException e) { throw new RE(e); } - queryFramework = new SqlTestFramework.Builder(this) + SqlTestFramework.Builder builder = new SqlTestFramework.Builder(this) .minTopNThreshold(minTopNThreshold) - .mergeBufferCount(mergeBufferCount) - .build(); + .mergeBufferCount(mergeBufferCount); + configureBuilder(builder); + queryFramework = builder.build(); + } + + protected void configureBuilder(Builder builder) + { } @Override public SpecificSegmentsQuerySegmentWalker createQuerySegmentWalker( final QueryRunnerFactoryConglomerate conglomerate, - final JoinableFactoryWrapper joinableFactory + final JoinableFactoryWrapper joinableFactory, + final Injector injector ) throws IOException { - return baseComponentSupplier.createQuerySegmentWalker(conglomerate, joinableFactory); + return baseComponentSupplier.createQuerySegmentWalker(conglomerate, joinableFactory, injector); } @Override @@ -547,45 +555,39 @@ public SqlEngine createEngine( } @Override - public QueryRunnerFactoryConglomerate createCongolmerate(Builder builder, Closer closer) + public void gatherProperties(Properties properties) { - return baseComponentSupplier.createCongolmerate(builder, closer); + baseComponentSupplier.gatherProperties(properties); } @Override - public void configureJsonMapper(ObjectMapper mapper) + public void configureGuice(DruidInjectorBuilder builder) { - baseComponentSupplier.configureJsonMapper(mapper); + baseComponentSupplier.configureGuice(builder); } @Override - public DruidOperatorTable createOperatorTable() + public QueryRunnerFactoryConglomerate createCongolmerate(Builder builder, Closer closer) { - return baseComponentSupplier.createOperatorTable(); + return baseComponentSupplier.createCongolmerate(builder, closer); } @Override - public ExprMacroTable createMacroTable() + public void configureJsonMapper(ObjectMapper mapper) { - return baseComponentSupplier.createMacroTable(); + baseComponentSupplier.configureJsonMapper(mapper); } @Override - public Map getJacksonInjectables() + public JoinableFactoryWrapper createJoinableFactoryWrapper(LookupExtractorFactoryContainerProvider lookupProvider) { - return baseComponentSupplier.getJacksonInjectables(); + return baseComponentSupplier.createJoinableFactoryWrapper(lookupProvider); } @Override - public Iterable getJacksonModules() + public void finalizeTestFramework(SqlTestFramework sqlTestFramework) { - return baseComponentSupplier.getJacksonModules(); - } - - @Override - public JoinableFactoryWrapper createJoinableFactoryWrapper(LookupExtractorFactoryContainerProvider lookupProvider) - { - return baseComponentSupplier.createJoinableFactoryWrapper(lookupProvider); + baseComponentSupplier.finalizeTestFramework(sqlTestFramework); } @Override @@ -618,11 +620,6 @@ public void finalizePlanner(PlannerFixture plannerFixture) basePlannerComponentSupplier.finalizePlanner(plannerFixture); } - @Override - public void configureGuice(DruidInjectorBuilder builder) - { - } - public void assertQueryIsUnplannable(final String sql, String expectedError) { assertQueryIsUnplannable(PLANNER_CONFIG_DEFAULT, sql, expectedError); @@ -839,49 +836,6 @@ protected QueryTestBuilder testBuilder() public class CalciteTestConfig implements QueryTestBuilder.QueryTestConfig { - @Override - public QueryTestRunner analyze(QueryTestBuilder builder) - { - if (builder.expectedResultsVerifier == null && builder.expectedResults != null) { - builder.expectedResultsVerifier = defaultResultsVerifier( - builder.expectedResults, - builder.expectedResultSignature - ); - } - final List runSteps = new ArrayList<>(); - final List verifySteps = new ArrayList<>(); - - // Historically, a test either prepares the query (to check resources), or - // runs the query (to check the native query and results.) In the future we - // may want to do both in a single test; but we have no such tests today. - if (builder.expectedResources != null) { - Preconditions.checkArgument( - builder.expectedResultsVerifier == null, - "Cannot check both results and resources" - ); - QueryTestRunner.PrepareQuery execStep = new QueryTestRunner.PrepareQuery(builder); - runSteps.add(execStep); - verifySteps.add(new QueryTestRunner.VerifyResources(execStep)); - } else { - QueryTestRunner.ExecuteQuery execStep = new QueryTestRunner.ExecuteQuery(builder); - runSteps.add(execStep); - - // Verify native queries before results. (Note: change from prior pattern - // that reversed the steps. - if (builder.expectedQueries != null) { - verifySteps.add(new QueryTestRunner.VerifyNativeQueries(execStep)); - } - if (builder.expectedResultsVerifier != null) { - verifySteps.add(new QueryTestRunner.VerifyResults(execStep)); - } - - // The exception is always verified: either there should be no exception - // (the other steps ran), or there should be the defined exception. - verifySteps.add(new QueryTestRunner.VerifyExpectedException(execStep)); - } - return new QueryTestRunner(runSteps, verifySteps); - } - @Override public QueryLogHook queryLogHook() { @@ -897,7 +851,7 @@ public ExpectedException expectedException() @Override public PlannerFixture plannerFixture(PlannerConfig plannerConfig, AuthConfig authConfig) { - return queryFramework.plannerFixture(BaseCalciteQueryTest.this, plannerConfig, authConfig); + return queryFramework().plannerFixture(BaseCalciteQueryTest.this, plannerConfig, authConfig); } @Override @@ -905,6 +859,18 @@ public ObjectMapper jsonMapper() { return queryFramework().queryJsonMapper(); } + + @Override + public ResultsVerifier defaultResultsVerifier( + List expectedResults, + RowSignature expectedResultSignature + ) + { + return BaseCalciteQueryTest.this.defaultResultsVerifier( + expectedResults, + expectedResultSignature + ); + } } public Set analyzeResources( diff --git a/sql/src/test/java/org/apache/druid/sql/calcite/CalciteIngestionDmlTest.java b/sql/src/test/java/org/apache/druid/sql/calcite/CalciteIngestionDmlTest.java index 4df3e8e18d3f..f082291242e4 100644 --- a/sql/src/test/java/org/apache/druid/sql/calcite/CalciteIngestionDmlTest.java +++ b/sql/src/test/java/org/apache/druid/sql/calcite/CalciteIngestionDmlTest.java @@ -20,15 +20,21 @@ package org.apache.druid.sql.calcite; import com.fasterxml.jackson.core.JsonProcessingException; +import com.fasterxml.jackson.databind.Module; import com.fasterxml.jackson.databind.ObjectMapper; +import com.fasterxml.jackson.databind.module.SimpleModule; import com.google.common.base.Preconditions; import com.google.common.collect.ImmutableList; import com.google.common.collect.ImmutableMap; +import com.google.inject.Binder; import org.apache.druid.data.input.impl.CsvInputFormat; import org.apache.druid.data.input.impl.InlineInputSource; +import org.apache.druid.guice.DruidInjectorBuilder; +import org.apache.druid.initialization.DruidModule; import org.apache.druid.java.util.common.ISE; import org.apache.druid.java.util.common.StringUtils; import org.apache.druid.java.util.common.granularity.Granularity; +import org.apache.druid.metadata.input.InputSourceModule; import org.apache.druid.query.Query; import org.apache.druid.query.QueryContexts; import org.apache.druid.query.aggregation.hyperloglog.HyperUniquesAggregatorFactory; @@ -41,10 +47,12 @@ import org.apache.druid.server.security.ResourceType; import org.apache.druid.sql.SqlQueryPlus; import org.apache.druid.sql.calcite.external.ExternalDataSource; +import org.apache.druid.sql.calcite.external.ExternalOperatorConversion; import org.apache.druid.sql.calcite.parser.DruidSqlInsert; import org.apache.druid.sql.calcite.planner.Calcites; import org.apache.druid.sql.calcite.planner.PlannerConfig; import org.apache.druid.sql.calcite.util.CalciteTests; +import org.apache.druid.sql.guice.SqlBindings; import org.hamcrest.CoreMatchers; import org.hamcrest.Matcher; import org.hamcrest.MatcherAssert; @@ -93,6 +101,55 @@ public CalciteIngestionDmlTest() super(IngestionTestSqlEngine.INSTANCE); } + @Override + public void configureGuice(DruidInjectorBuilder builder) + { + super.configureGuice(builder); + + builder.addModule(new DruidModule() { + + // Clone of MSQExternalDataSourceModule since it is not + // visible here. + @Override + public List getJacksonModules() + { + return Collections.singletonList( + new SimpleModule(getClass().getSimpleName()) + .registerSubtypes(ExternalDataSource.class) + ); + } + + @Override + public void configure(Binder binder) + { + // Nothing to do. + } + }); + + builder.addModule(new DruidModule() { + + // Partial clone of MsqSqlModule, since that module is not + // visible to this one. + + @Override + public List getJacksonModules() + { + // We want this module to bring input sources along for the ride. + return new InputSourceModule().getJacksonModules(); + } + + @Override + public void configure(Binder binder) + { + // We want this module to bring InputSourceModule along for the ride. + binder.install(new InputSourceModule()); + + // Set up the EXTERN macro. + SqlBindings.addOperatorConversion(binder, ExternalOperatorConversion.class); + } + }); + } + @After public void tearDown() { @@ -147,6 +204,7 @@ public class IngestionDmlTester private List expectedResources; private Query expectedQuery; private Matcher validationErrorMatcher; + private String expectedLogicalPlanResource; private IngestionDmlTester() { @@ -224,6 +282,12 @@ public IngestionDmlTester expectValidationError(Class clazz ); } + public IngestionDmlTester expectLogicalPlanFrom(String resource) + { + this.expectedLogicalPlanResource = resource; + return this; + } + public void verify() { if (didTest) { @@ -299,6 +363,15 @@ private void verifySuccess() .expectedResources(expectedResources) .run(); + String expectedLogicalPlan; + if (expectedLogicalPlanResource != null) { + expectedLogicalPlan = StringUtils.getResource( + this, + "/calcite/expected/ingest/" + expectedLogicalPlanResource + "-logicalPlan.txt" + ); + } else { + expectedLogicalPlan = null; + } testBuilder() .sql(sql) .queryContext(queryContext) @@ -306,6 +379,7 @@ private void verifySuccess() .plannerConfig(plannerConfig) .expectedQuery(expectedQuery) .expectedResults(Collections.singletonList(new Object[]{expectedTargetDataSource, expectedTargetSignature})) + .expectedLogicalPlan(expectedLogicalPlan) .run(); } diff --git a/sql/src/test/java/org/apache/druid/sql/calcite/CalciteInsertDmlTest.java b/sql/src/test/java/org/apache/druid/sql/calcite/CalciteInsertDmlTest.java index 965f66b423b2..a45ab546ecf8 100644 --- a/sql/src/test/java/org/apache/druid/sql/calcite/CalciteInsertDmlTest.java +++ b/sql/src/test/java/org/apache/druid/sql/calcite/CalciteInsertDmlTest.java @@ -60,7 +60,7 @@ public class CalciteInsertDmlTest extends CalciteIngestionDmlTest { - private static final Map PARTITIONED_BY_ALL_TIME_QUERY_CONTEXT = ImmutableMap.of( + public static final Map PARTITIONED_BY_ALL_TIME_QUERY_CONTEXT = ImmutableMap.of( DruidSqlInsert.SQL_INSERT_SEGMENT_GRANULARITY, "{\"type\":\"all\"}" ); @@ -302,6 +302,7 @@ public void testInsertFromExternal() .context(PARTITIONED_BY_ALL_TIME_QUERY_CONTEXT) .build() ) + .expectLogicalPlanFrom("insertFromExternal") .verify(); } @@ -329,6 +330,7 @@ public void testInsertWithPartitionedBy() .context(queryContextWithGranularity(Granularities.HOUR)) .build() ) + .expectLogicalPlanFrom("insertWithPartitionedBy") .verify(); } @@ -423,6 +425,7 @@ public void testInsertWithClusteredBy() .context(queryContextWithGranularity(Granularities.DAY)) .build() ) + .expectLogicalPlanFrom("insertWithClusteredBy") .verify(); } diff --git a/sql/src/test/java/org/apache/druid/sql/calcite/CalciteNestedDataQueryTest.java b/sql/src/test/java/org/apache/druid/sql/calcite/CalciteNestedDataQueryTest.java index 9b557ad0bfc3..af32ac68575c 100644 --- a/sql/src/test/java/org/apache/druid/sql/calcite/CalciteNestedDataQueryTest.java +++ b/sql/src/test/java/org/apache/druid/sql/calcite/CalciteNestedDataQueryTest.java @@ -19,11 +19,10 @@ package org.apache.druid.sql.calcite; -import com.fasterxml.jackson.databind.Module; import com.google.common.collect.ImmutableList; import com.google.common.collect.ImmutableMap; import com.google.common.collect.ImmutableSet; -import com.google.common.collect.Iterables; +import com.google.inject.Injector; import org.apache.druid.common.config.NullHandling; import org.apache.druid.data.input.InputRow; import org.apache.druid.data.input.impl.DimensionSchema; @@ -34,7 +33,7 @@ import org.apache.druid.data.input.impl.StringDimensionSchema; import org.apache.druid.data.input.impl.TimeAndDimsParseSpec; import org.apache.druid.data.input.impl.TimestampSpec; -import org.apache.druid.guice.ExpressionModule; +import org.apache.druid.guice.DruidInjectorBuilder; import org.apache.druid.guice.NestedDataModule; import org.apache.druid.java.util.common.granularity.Granularities; import org.apache.druid.math.expr.ExprMacroTable; @@ -45,7 +44,6 @@ import org.apache.druid.query.aggregation.FilteredAggregatorFactory; import org.apache.druid.query.aggregation.LongSumAggregatorFactory; import org.apache.druid.query.dimension.DefaultDimensionSpec; -import org.apache.druid.query.expression.LookupExprMacro; import org.apache.druid.query.filter.InDimFilter; import org.apache.druid.query.filter.LikeDimFilter; import org.apache.druid.query.groupby.GroupByQuery; @@ -61,13 +59,11 @@ import org.apache.druid.segment.incremental.IncrementalIndexSchema; import org.apache.druid.segment.join.JoinableFactoryWrapper; import org.apache.druid.segment.nested.NestedDataComplexTypeSerde; -import org.apache.druid.segment.serde.ComplexMetrics; import org.apache.druid.segment.virtual.ExpressionVirtualColumn; import org.apache.druid.segment.virtual.NestedFieldVirtualColumn; import org.apache.druid.segment.writeout.OffHeapMemorySegmentWriteOutMediumFactory; import org.apache.druid.sql.calcite.filtration.Filtration; import org.apache.druid.sql.calcite.planner.UnsupportedSQLQueryException; -import org.apache.druid.sql.calcite.util.CalciteTests; import org.apache.druid.sql.calcite.util.SpecificSegmentsQuerySegmentWalker; import org.apache.druid.sql.calcite.util.TestDataBuilder; import org.apache.druid.timeline.DataSegment; @@ -75,7 +71,6 @@ import org.junit.Test; import java.io.IOException; -import java.util.ArrayList; import java.util.List; import java.util.Map; import java.util.stream.Collectors; @@ -150,19 +145,18 @@ public class CalciteNestedDataQueryTest extends BaseCalciteQueryTest RAW_ROWS.stream().map(raw -> TestDataBuilder.createRow(raw, PARSER)).collect(Collectors.toList()); @Override - public Iterable getJacksonModules() + public void configureGuice(DruidInjectorBuilder builder) { - return Iterables.concat( - super.getJacksonModules(), - NestedDataModule.getJacksonModulesList() - ); + super.configureGuice(builder); + builder.addModule(new NestedDataModule()); } @SuppressWarnings("resource") @Override public SpecificSegmentsQuerySegmentWalker createQuerySegmentWalker( final QueryRunnerFactoryConglomerate conglomerate, - final JoinableFactoryWrapper joinableFactory + final JoinableFactoryWrapper joinableFactory, + final Injector injector ) throws IOException { NestedDataModule.registerHandlersAndSerde(); @@ -194,18 +188,6 @@ public SpecificSegmentsQuerySegmentWalker createQuerySegmentWalker( ); } - @Override - public ExprMacroTable createMacroTable() - { - ComplexMetrics.registerSerde(NestedDataComplexTypeSerde.TYPE_NAME, NestedDataComplexTypeSerde.INSTANCE); - final List exprMacros = new ArrayList<>(); - for (Class clazz : ExpressionModule.EXPR_MACROS) { - exprMacros.add(CalciteTests.INJECTOR.getInstance(clazz)); - } - exprMacros.add(CalciteTests.INJECTOR.getInstance(LookupExprMacro.class)); - return new ExprMacroTable(exprMacros); - } - @Test public void testGroupByPath() { @@ -2428,7 +2410,7 @@ public void testCompositionTyping() "v0", "json_value(json_object('x',\"v1\"),'$.x', 'LONG')", ColumnType.LONG, - createMacroTable() + queryFramework().macroTable() ), new NestedFieldVirtualColumn( "nest", diff --git a/sql/src/test/java/org/apache/druid/sql/calcite/QueryTestBuilder.java b/sql/src/test/java/org/apache/druid/sql/calcite/QueryTestBuilder.java index ef8050b6d54e..f9b843d5d8b3 100644 --- a/sql/src/test/java/org/apache/druid/sql/calcite/QueryTestBuilder.java +++ b/sql/src/test/java/org/apache/druid/sql/calcite/QueryTestBuilder.java @@ -25,6 +25,7 @@ import org.apache.druid.server.security.AuthConfig; import org.apache.druid.server.security.AuthenticationResult; import org.apache.druid.server.security.ResourceAction; +import org.apache.druid.sql.SqlStatementFactory; import org.apache.druid.sql.calcite.BaseCalciteQueryTest.ResultsVerifier; import org.apache.druid.sql.calcite.QueryTestRunner.QueryResults; import org.apache.druid.sql.calcite.planner.PlannerConfig; @@ -61,17 +62,17 @@ public class QueryTestBuilder { /** * Implement to provide the execution framework that the tests require. - * The {@link #analyze(QueryTestBuilder)} method builds up the classes that + * The constructor builds up the classes that * will run the test, since some verification depends on context, such as that * provided by {@link BaseCalciteQueryTest}. */ public interface QueryTestConfig { - QueryTestRunner analyze(QueryTestBuilder builder); QueryLogHook queryLogHook(); ExpectedException expectedException(); ObjectMapper jsonMapper(); PlannerFixture plannerFixture(PlannerConfig plannerConfig, AuthConfig authConfig); + ResultsVerifier defaultResultsVerifier(List expectedResults, RowSignature expectedResultSignature); } protected final QueryTestConfig config; @@ -90,6 +91,8 @@ public interface QueryTestConfig protected boolean queryCannotVectorize; protected AuthConfig authConfig = new AuthConfig(); protected PlannerFixture plannerFixture; + protected String expectedLogicalPlan; + protected SqlSchema expectedSqlSchema; public QueryTestBuilder(final QueryTestConfig config) { @@ -215,23 +218,35 @@ public QueryTestBuilder plannerFixture(PlannerFixture plannerFixture) return this; } + public QueryTestBuilder expectedLogicalPlan(String expectedLogicalPlan) + { + this.expectedLogicalPlan = expectedLogicalPlan; + return this; + } + + public QueryTestBuilder expectedSqlSchema(SqlSchema querySchema) + { + this.expectedSqlSchema = querySchema; + return this; + } + public QueryTestRunner build() { - return config.analyze(this); + return new QueryTestRunner(this); } /** - * Internal method to return the cached planner config, or create a new one + * Internal method to return the cached statement factory, or create a new one * based on the configs provided. Note: does not cache the newly created * config: doing so would confuse the "please use mine" vs. "create a new * one each time" semantics. */ - protected PlannerFixture plannerFixture() + protected SqlStatementFactory statementFactory() { if (plannerFixture != null) { - return plannerFixture; + return plannerFixture.statementFactory(); } else { - return config.plannerFixture(plannerConfig, authConfig); + return config.plannerFixture(plannerConfig, authConfig).statementFactory(); } } diff --git a/sql/src/test/java/org/apache/druid/sql/calcite/QueryTestRunner.java b/sql/src/test/java/org/apache/druid/sql/calcite/QueryTestRunner.java index fc60bc4c61aa..f3e1661b09ac 100644 --- a/sql/src/test/java/org/apache/druid/sql/calcite/QueryTestRunner.java +++ b/sql/src/test/java/org/apache/druid/sql/calcite/QueryTestRunner.java @@ -21,8 +21,13 @@ import com.fasterxml.jackson.core.JsonProcessingException; import com.fasterxml.jackson.databind.ObjectMapper; +import com.google.common.base.Preconditions; import com.google.common.collect.ImmutableSet; +import org.apache.calcite.plan.RelOptUtil; import org.apache.calcite.rel.type.RelDataType; +import org.apache.calcite.sql.SqlExplainFormat; +import org.apache.calcite.sql.SqlExplainLevel; +import org.apache.calcite.sql.SqlInsert; import org.apache.druid.java.util.common.ISE; import org.apache.druid.java.util.common.Pair; import org.apache.druid.java.util.common.StringUtils; @@ -35,6 +40,10 @@ import org.apache.druid.sql.PreparedStatement; import org.apache.druid.sql.SqlQueryPlus; import org.apache.druid.sql.SqlStatementFactory; +import org.apache.druid.sql.calcite.QueryTestBuilder.QueryTestConfig; +import org.apache.druid.sql.calcite.parser.DruidSqlIngest; +import org.apache.druid.sql.calcite.planner.PlannerCaptureHook; +import org.apache.druid.sql.calcite.planner.PrepareResult; import org.apache.druid.sql.calcite.table.RowSignatures; import org.apache.druid.sql.calcite.util.QueryLogHook; import org.junit.Assert; @@ -48,7 +57,7 @@ import java.util.Set; /** - * Runs a test build up by {@link QueryTestBuilder}. Running a SQL query test + * Runs a test built up by {@link QueryTestBuilder}. Running a SQL query test * is somewhat complex; with different modes and items to verify. To manage the * complexity, test execution is done in two steps: *
    @@ -95,27 +104,32 @@ public static class QueryResults { public final Map queryContext; public final String vectorizeOption; + public final RelDataType sqlSignature; public final RowSignature signature; public final List results; public final List> recordedQueries; public final Set resourceActions; public final RuntimeException exception; + public final PlannerCaptureHook capture; public QueryResults( final Map queryContext, final String vectorizeOption, - final RowSignature signature, + final RelDataType sqlSignature, final List results, - final List> recordedQueries + final List> recordedQueries, + final PlannerCaptureHook capture ) { this.queryContext = queryContext; this.vectorizeOption = vectorizeOption; - this.signature = signature; + this.sqlSignature = sqlSignature; + this.signature = RowSignatures.fromRelDataType(sqlSignature.getFieldNames(), sqlSignature); this.results = results; this.recordedQueries = recordedQueries; this.resourceActions = null; this.exception = null; + this.capture = capture; } public QueryResults( @@ -131,21 +145,8 @@ public QueryResults( this.recordedQueries = null; this.resourceActions = null; this.exception = exception; - } - - public QueryResults( - final Map queryContext, - final String vectorizeOption, - final Set resourceActions - ) - { - this.queryContext = queryContext; - this.vectorizeOption = vectorizeOption; - this.signature = null; - this.results = null; - this.recordedQueries = null; - this.resourceActions = resourceActions; - this.exception = null; + this.capture = null; + this.sqlSignature = null; } } @@ -156,7 +157,8 @@ public QueryResults( */ public static class PrepareQuery extends QueryRunStep { - private Set resourceActions; + public Set resourceActions; + public RelDataType sqlSignature; public PrepareQuery(QueryTestBuilder builder) { @@ -177,10 +179,11 @@ public void run() .sqlParameters(builder.parameters) .auth(builder.authenticationResult) .build(); - final SqlStatementFactory sqlStatementFactory = builder.plannerFixture().statementFactory(); + final SqlStatementFactory sqlStatementFactory = builder.statementFactory(); final PreparedStatement stmt = sqlStatementFactory.preparedStatement(sqlQuery); - stmt.prepare(); + final PrepareResult prepareResult = stmt.prepare(); resourceActions = stmt.allResources(); + sqlSignature = prepareResult.getReturnedRowType(); } } @@ -191,10 +194,12 @@ public void run() public static class ExecuteQuery extends QueryRunStep { private final List results = new ArrayList<>(); + private final boolean doCapture; public ExecuteQuery(QueryTestBuilder builder) { super(builder); + doCapture = builder.expectedLogicalPlan != null; } public List results() @@ -209,7 +214,7 @@ public void run() BaseCalciteQueryTest.log.info("SQL: %s", builder.sql); - final SqlStatementFactory sqlStatementFactory = builder.plannerFixture().statementFactory(); + final SqlStatementFactory sqlStatementFactory = builder.statementFactory(); final SqlQueryPlus sqlQuery = SqlQueryPlus.builder(builder.sql) .sqlParameters(builder.parameters) .auth(builder.authenticationResult) @@ -233,25 +238,40 @@ public void run() theQueryContext.put(QueryContexts.VECTOR_SIZE_KEY, 2); // Small vector size to ensure we use more than one. } - try { - final Pair> plannerResults = getResults( - sqlStatementFactory, - sqlQuery.withContext(theQueryContext)); - results.add(new QueryResults( - theQueryContext, - vectorize, - plannerResults.lhs, - plannerResults.rhs, - queryLogHook.getRecordedQueries() - )); - } - catch (RuntimeException e) { - results.add(new QueryResults( - theQueryContext, - vectorize, - e - )); - } + results.add(runQuery( + sqlStatementFactory, + sqlQuery.withContext(theQueryContext), + vectorize + )); + } + } + + public QueryResults runQuery( + final SqlStatementFactory sqlStatementFactory, + final SqlQueryPlus query, + final String vectorize + ) + { + try { + final PlannerCaptureHook capture = doCapture ? new PlannerCaptureHook() : null; + final DirectStatement stmt = sqlStatementFactory.directStatement(query); + stmt.setHook(capture); + final Sequence results = stmt.execute().getResults(); + return new QueryResults( + query.context(), + vectorize, + stmt.prepareResult().getReturnedRowType(), + results.toList(), + builder().config.queryLogHook().getRecordedQueries(), + capture + ); + } + catch (RuntimeException e) { + return new QueryResults( + query.context(), + vectorize, + e + ); } } @@ -379,7 +399,7 @@ private void verifyQuery(QueryResults queryResults) } /** - * Verify rsources for a prepared query against the expected list. + * Verify resources for a prepared query against the expected list. */ public static class VerifyResources implements QueryVerifyStep { @@ -402,7 +422,102 @@ public void verify() } /** - * Verify the exception thrown by a query using a jUnit expected + * Verify resources for a prepared query against the expected list. + */ + public static class VerifyPrepareSignature implements QueryVerifyStep + { + private final PrepareQuery prepareStep; + + public VerifyPrepareSignature(PrepareQuery prepareStep) + { + this.prepareStep = prepareStep; + } + + @Override + public void verify() + { + QueryTestBuilder builder = prepareStep.builder(); + Assert.assertEquals( + builder.expectedSqlSchema, + SqlSchema.of(prepareStep.sqlSignature) + ); + } + } + + /** + * Verify resources for a prepared query against the expected list. + */ + public static class VerifyExecuteSignature extends VerifyExecStep + { + public VerifyExecuteSignature(ExecuteQuery execStep) + { + super(execStep); + } + + @Override + public void verify() + { + QueryTestBuilder builder = execStep.builder(); + for (QueryResults queryResults : execStep.results()) { + Assert.assertEquals( + builder.expectedSqlSchema, + SqlSchema.of(queryResults.sqlSignature) + ); + } + } + } + + public static class VerifyLogicalPlan extends VerifyExecStep + { + public VerifyLogicalPlan(ExecuteQuery execStep) + { + super(execStep); + } + + @Override + public void verify() + { + for (QueryResults queryResults : execStep.results()) { + verifyLogicalPlan(queryResults); + } + } + + private void verifyLogicalPlan(QueryResults queryResults) + { + String expectedPlan = execStep.builder().expectedLogicalPlan; + String actualPlan = visualizePlan(queryResults.capture); + Assert.assertEquals(expectedPlan, actualPlan); + } + + private String visualizePlan(PlannerCaptureHook hook) + { + // Do-it-ourselves plan since the actual plan omits insert. + String queryPlan = RelOptUtil.dumpPlan( + "", + hook.relRoot().rel, + SqlExplainFormat.TEXT, + SqlExplainLevel.DIGEST_ATTRIBUTES); + String plan; + SqlInsert insertNode = hook.insertNode(); + if (insertNode == null) { + plan = queryPlan; + } else { + DruidSqlIngest druidInsertNode = (DruidSqlIngest) insertNode; + // The target is a SQLIdentifier literal, pre-resolution, so does + // not include the schema. + plan = StringUtils.format( + "LogicalInsert(target=[%s], partitionedBy=[%s], clusteredBy=[%s])\n", + druidInsertNode.getTargetTable(), + druidInsertNode.getPartitionedBy() == null ? "" : druidInsertNode.getPartitionedBy(), + druidInsertNode.getClusteredBy() == null ? "" : druidInsertNode.getClusteredBy() + ) + " " + StringUtils.replace(queryPlan, "\n ", "\n "); + } + return plan; + } + } + + /** + * Verify the exception thrown by a query using a JUnit expected * exception. This is actually an awkward way to to the job, but it is * what the Calcite queries have long used. There are three modes. * In the first, the exception is simply thrown and the expected @@ -455,16 +570,62 @@ public void verify() } } - private final List runSteps; - private final List verifySteps; + private final List runSteps = new ArrayList<>(); + private final List verifySteps = new ArrayList<>(); - QueryTestRunner( - final List runSteps, - final List verifySteps - ) + /** + * Create a test runner based on the options set in the builder. + */ + public QueryTestRunner(QueryTestBuilder builder) { - this.runSteps = runSteps; - this.verifySteps = verifySteps; + QueryTestConfig config = builder.config; + if (builder.expectedResultsVerifier == null && builder.expectedResults != null) { + builder.expectedResultsVerifier = config.defaultResultsVerifier( + builder.expectedResults, + builder.expectedResultSignature + ); + } + + // Historically, a test either prepares the query (to check resources), or + // runs the query (to check the native query and results.) In the future we + // may want to do both in a single test; but we have no such tests today. + if (builder.expectedResources != null) { + Preconditions.checkArgument( + builder.expectedResultsVerifier == null, + "Cannot check both results and resources" + ); + QueryTestRunner.PrepareQuery execStep = new QueryTestRunner.PrepareQuery(builder); + runSteps.add(execStep); + verifySteps.add(new QueryTestRunner.VerifyResources(execStep)); + if (builder.expectedSqlSchema != null) { + verifySteps.add(new VerifyPrepareSignature(execStep)); + } + } else { + QueryTestRunner.ExecuteQuery execStep = new QueryTestRunner.ExecuteQuery(builder); + runSteps.add(execStep); + + // Verify the logical plan, if requested. + if (builder.expectedLogicalPlan != null) { + verifySteps.add(new QueryTestRunner.VerifyLogicalPlan(execStep)); + } + + if (builder.expectedSqlSchema != null) { + verifySteps.add(new VerifyExecuteSignature(execStep)); + } + + // Verify native queries before results. (Note: change from prior pattern + // that reversed the steps. + if (builder.expectedQueries != null) { + verifySteps.add(new QueryTestRunner.VerifyNativeQueries(execStep)); + } + if (builder.expectedResultsVerifier != null) { + verifySteps.add(new QueryTestRunner.VerifyResults(execStep)); + } + + // The exception is always verified: either there should be no exception + // (the other steps ran), or there should be the defined exception. + verifySteps.add(new QueryTestRunner.VerifyExpectedException(execStep)); + } } /** diff --git a/sql/src/test/java/org/apache/druid/sql/calcite/SqlSchema.java b/sql/src/test/java/org/apache/druid/sql/calcite/SqlSchema.java new file mode 100644 index 000000000000..b5ee0c1198ba --- /dev/null +++ b/sql/src/test/java/org/apache/druid/sql/calcite/SqlSchema.java @@ -0,0 +1,145 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.sql.calcite; + +import org.apache.calcite.rel.type.RelDataType; +import org.apache.calcite.rel.type.RelDataTypeField; +import org.apache.druid.java.util.common.StringUtils; + +import java.util.ArrayList; +import java.util.List; +import java.util.Objects; +import java.util.stream.Collectors; + +/** + * Simple representation of an SQL schema used to validate the row type of + * a SQL query against the SQL types. See {@code RowSignature} when the goal + * is to validate the Druid native signature. + */ +public class SqlSchema +{ + public static class ColumnSignature + { + private final String name; + private final String type; + + public ColumnSignature(final String name, final String type) + { + this.name = name; + this.type = type; + } + + public String name() + { + return name; + } + + public String type() + { + return type; + } + + @Override + public String toString() + { + return StringUtils.format("%s %s", name, type); + } + + @Override + public boolean equals(Object o) + { + if (o == null || o.getClass() != getClass()) { + return false; + } + ColumnSignature other = (ColumnSignature) o; + return Objects.equals(name, other.name) + && Objects.equals(type, other.type); + } + + @Override + public int hashCode() + { + return Objects.hash(name, type); + } + } + + public static class Builder + { + private final List columns = new ArrayList<>(); + + public Builder column(String name, String type) + { + columns.add(new ColumnSignature(name, type)); + return this; + } + + public SqlSchema build() + { + return new SqlSchema(columns); + } + } + + private final List columns; + + public SqlSchema(final List columns) + { + this.columns = columns; + } + + public static Builder builder() + { + return new Builder(); + } + + public static SqlSchema of(RelDataType rowType) + { + final Builder builder = new Builder(); + List fields = rowType.getFieldList(); + for (RelDataTypeField field : fields) { + builder.column(field.getName(), field.getType().getFullTypeString()); + } + return builder.build(); + } + + @Override + public String toString() + { + return "(" + + columns.stream().map( + c -> c.toString()).collect(Collectors.joining(", ")) + + ")"; + } + + @Override + public boolean equals(Object o) + { + if (o == null || o.getClass() != getClass()) { + return false; + } + final SqlSchema other = (SqlSchema) o; + return Objects.equals(columns, other.columns); + } + + @Override + public int hashCode() + { + return Objects.hash(columns); + } +} diff --git a/sql/src/test/java/org/apache/druid/sql/calcite/util/BasicTestModule.java b/sql/src/test/java/org/apache/druid/sql/calcite/util/BasicTestModule.java new file mode 100644 index 000000000000..375352e22c79 --- /dev/null +++ b/sql/src/test/java/org/apache/druid/sql/calcite/util/BasicTestModule.java @@ -0,0 +1,65 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.sql.calcite.util; + +import com.fasterxml.jackson.databind.Module; +import com.google.common.collect.ImmutableMap; +import com.google.inject.Binder; +import org.apache.druid.guice.ExpressionModule; +import org.apache.druid.initialization.DruidModule; +import org.apache.druid.query.expression.LookupEnabledTestExprMacroTable; +import org.apache.druid.query.expression.LookupExprMacro; +import org.apache.druid.query.lookup.LookupExtractorFactoryContainerProvider; +import org.apache.druid.query.lookup.LookupSerdeModule; +import org.apache.druid.sql.calcite.expression.builtin.QueryLookupOperatorConversion; +import org.apache.druid.sql.guice.SqlBindings; +import org.apache.druid.timeline.DataSegment; + +import java.util.List; + +class BasicTestModule implements DruidModule +{ + @Override + public void configure(Binder binder) + { + final LookupExtractorFactoryContainerProvider lookupProvider = + LookupEnabledTestExprMacroTable.createTestLookupProvider( + ImmutableMap.of( + "a", "xa", + "abc", "xabc", + "nosuchkey", "mysteryvalue", + "6", "x6" + ) + ); + + binder.bind(DataSegment.PruneSpecsHolder.class).toInstance(DataSegment.PruneSpecsHolder.DEFAULT); + + // This Module is just to get a LookupExtractorFactoryContainerProvider with a usable "lookyloo" lookup. + binder.bind(LookupExtractorFactoryContainerProvider.class).toInstance(lookupProvider); + SqlBindings.addOperatorConversion(binder, QueryLookupOperatorConversion.class); + ExpressionModule.addExprMacro(binder, LookupExprMacro.class); + } + + @Override + public List getJacksonModules() + { + return new LookupSerdeModule().getJacksonModules(); + } +} diff --git a/sql/src/test/java/org/apache/druid/sql/calcite/util/CalciteTestInjectorBuilder.java b/sql/src/test/java/org/apache/druid/sql/calcite/util/CalciteTestInjectorBuilder.java index 902c61ddb44a..f9e723e36311 100644 --- a/sql/src/test/java/org/apache/druid/sql/calcite/util/CalciteTestInjectorBuilder.java +++ b/sql/src/test/java/org/apache/druid/sql/calcite/util/CalciteTestInjectorBuilder.java @@ -19,25 +19,12 @@ package org.apache.druid.sql.calcite.util; -import com.fasterxml.jackson.databind.Module; -import com.google.common.collect.ImmutableMap; -import com.google.inject.Binder; import com.google.inject.Injector; import org.apache.druid.guice.StartupInjectorBuilder; import org.apache.druid.initialization.CoreInjectorBuilder; -import org.apache.druid.initialization.DruidModule; import org.apache.druid.math.expr.ExprMacroTable; -import org.apache.druid.query.expression.LookupEnabledTestExprMacroTable; import org.apache.druid.query.expression.TestExprMacroTable; -import org.apache.druid.query.lookup.LookupExtractorFactoryContainerProvider; -import org.apache.druid.query.lookup.LookupSerdeModule; import org.apache.druid.sql.calcite.aggregation.SqlAggregationModule; -import org.apache.druid.sql.calcite.expression.builtin.QueryLookupOperatorConversion; -import org.apache.druid.sql.calcite.external.ExternalOperatorConversion; -import org.apache.druid.sql.guice.SqlBindings; -import org.apache.druid.timeline.DataSegment; - -import java.util.List; /** * Create the injector used for {@link CalciteTests#INJECTOR}, but in a way @@ -77,37 +64,4 @@ public Injector build() throw e; } } - - private static class BasicTestModule implements DruidModule - { - @Override - public void configure(Binder binder) - { - final LookupExtractorFactoryContainerProvider lookupProvider = - LookupEnabledTestExprMacroTable.createTestLookupProvider( - ImmutableMap.of( - "a", "xa", - "abc", "xabc", - "nosuchkey", "mysteryvalue", - "6", "x6" - ) - ); - - binder.bind(DataSegment.PruneSpecsHolder.class).toInstance(DataSegment.PruneSpecsHolder.DEFAULT); - binder.bind(LookupExtractorFactoryContainerProvider.class).toInstance(lookupProvider); - - // This Module is just to get a LookupExtractorFactoryContainerProvider with a usable "lookyloo" lookup. - binder.bind(LookupExtractorFactoryContainerProvider.class).toInstance(lookupProvider); - SqlBindings.addOperatorConversion(binder, QueryLookupOperatorConversion.class); - - // Add "EXTERN" table macro, for CalciteInsertDmlTest. - SqlBindings.addOperatorConversion(binder, ExternalOperatorConversion.class); - } - - @Override - public List getJacksonModules() - { - return new LookupSerdeModule().getJacksonModules(); - } - } } diff --git a/sql/src/test/java/org/apache/druid/sql/calcite/util/QueryFrameworkUtils.java b/sql/src/test/java/org/apache/druid/sql/calcite/util/QueryFrameworkUtils.java index f019c37b62cd..495f8a912b3f 100644 --- a/sql/src/test/java/org/apache/druid/sql/calcite/util/QueryFrameworkUtils.java +++ b/sql/src/test/java/org/apache/druid/sql/calcite/util/QueryFrameworkUtils.java @@ -127,6 +127,7 @@ public static SqlStatementFactory createSqlStatementFactory( ); return new SqlStatementFactory(toolbox); } + public static DruidSchemaCatalog createMockRootSchema( final Injector injector, final QueryRunnerFactoryConglomerate conglomerate, diff --git a/sql/src/test/java/org/apache/druid/sql/calcite/util/QueryLogHook.java b/sql/src/test/java/org/apache/druid/sql/calcite/util/QueryLogHook.java index 7a58503717a9..967926681fb2 100644 --- a/sql/src/test/java/org/apache/druid/sql/calcite/util/QueryLogHook.java +++ b/sql/src/test/java/org/apache/druid/sql/calcite/util/QueryLogHook.java @@ -32,6 +32,7 @@ import java.util.List; import java.util.function.Consumer; +import java.util.function.Supplier; /** * JUnit Rule that adds a Calcite hook to log and remember Druid queries. @@ -40,22 +41,22 @@ public class QueryLogHook implements TestRule { private static final Logger log = new Logger(QueryLogHook.class); - private final ObjectMapper objectMapper; + private final Supplier objectMapperSupplier; private final List> recordedQueries = Lists.newCopyOnWriteArrayList(); - public QueryLogHook(final ObjectMapper objectMapper) + public QueryLogHook(final Supplier objectMapperSupplier) { - this.objectMapper = objectMapper; + this.objectMapperSupplier = objectMapperSupplier; } public static QueryLogHook create() { - return new QueryLogHook(new DefaultObjectMapper()); + return new QueryLogHook(() -> DefaultObjectMapper.INSTANCE); } public static QueryLogHook create(final ObjectMapper objectMapper) { - return new QueryLogHook(objectMapper); + return new QueryLogHook(() -> objectMapper); } public void clearRecordedQueries() @@ -83,7 +84,7 @@ public void evaluate() throws Throwable recordedQueries.add((Query) query); log.info( "Issued query: %s", - objectMapper.writerWithDefaultPrettyPrinter().writeValueAsString(query) + objectMapperSupplier.get().writerWithDefaultPrettyPrinter().writeValueAsString(query) ); } catch (Exception e) { diff --git a/sql/src/test/java/org/apache/druid/sql/calcite/util/SqlTestFramework.java b/sql/src/test/java/org/apache/druid/sql/calcite/util/SqlTestFramework.java index 7d7a0b8fd5b7..cace9647e96a 100644 --- a/sql/src/test/java/org/apache/druid/sql/calcite/util/SqlTestFramework.java +++ b/sql/src/test/java/org/apache/druid/sql/calcite/util/SqlTestFramework.java @@ -19,16 +19,17 @@ package org.apache.druid.sql.calcite.util; -import com.fasterxml.jackson.databind.Module; import com.fasterxml.jackson.databind.ObjectMapper; -import com.fasterxml.jackson.databind.module.SimpleModule; import com.google.common.collect.ImmutableMap; import com.google.common.collect.ImmutableSet; -import com.google.common.collect.Lists; import com.google.inject.Binder; import com.google.inject.Injector; import com.google.inject.Provides; import org.apache.druid.guice.DruidInjectorBuilder; +import org.apache.druid.guice.ExpressionModule; +import org.apache.druid.guice.LazySingleton; +import org.apache.druid.guice.StartupInjectorBuilder; +import org.apache.druid.initialization.CoreInjectorBuilder; import org.apache.druid.initialization.DruidModule; import org.apache.druid.java.util.common.RE; import org.apache.druid.java.util.common.io.Closer; @@ -36,7 +37,6 @@ import org.apache.druid.query.GlobalTableDataSource; import org.apache.druid.query.QueryRunnerFactoryConglomerate; import org.apache.druid.query.lookup.LookupExtractorFactoryContainerProvider; -import org.apache.druid.query.lookup.LookupSerdeModule; import org.apache.druid.query.topn.TopNQueryConfig; import org.apache.druid.segment.join.JoinableFactoryWrapper; import org.apache.druid.server.QueryLifecycle; @@ -45,7 +45,7 @@ import org.apache.druid.server.security.AuthConfig; import org.apache.druid.server.security.AuthorizerMapper; import org.apache.druid.sql.SqlStatementFactory; -import org.apache.druid.sql.calcite.external.ExternalDataSource; +import org.apache.druid.sql.calcite.aggregation.SqlAggregationModule; import org.apache.druid.sql.calcite.planner.CalciteRulesManager; import org.apache.druid.sql.calcite.planner.DruidOperatorTable; import org.apache.druid.sql.calcite.planner.PlannerConfig; @@ -61,12 +61,11 @@ import org.apache.druid.sql.calcite.view.ViewManager; import org.apache.druid.timeline.DataSegment; +import javax.inject.Singleton; + import java.io.File; import java.io.IOException; -import java.util.ArrayList; -import java.util.HashMap; -import java.util.List; -import java.util.Map; +import java.util.Properties; import java.util.Set; /** @@ -123,6 +122,20 @@ public class SqlTestFramework */ public interface QueryComponentSupplier { + /** + * Gather properties to be used within tests. Particularly useful when choosing + * among aggregator implementations: avoids the need to copy/paste code to select + * the desired implementation. + */ + void gatherProperties(Properties properties); + + /** + * Configure modules needed for tests. This is the preferred way to configure + * Jackson: include the production module in this method that includes the + * required Jackson configuration. + */ + void configureGuice(DruidInjectorBuilder builder); + QueryRunnerFactoryConglomerate createCongolmerate( Builder builder, Closer closer @@ -130,7 +143,8 @@ QueryRunnerFactoryConglomerate createCongolmerate( SpecificSegmentsQuerySegmentWalker createQuerySegmentWalker( QueryRunnerFactoryConglomerate conglomerate, - JoinableFactoryWrapper joinableFactory + JoinableFactoryWrapper joinableFactory, + Injector injector ) throws IOException; SqlEngine createEngine( @@ -138,19 +152,16 @@ SqlEngine createEngine( ObjectMapper objectMapper ); - DruidOperatorTable createOperatorTable(); - - ExprMacroTable createMacroTable(); - - Iterable getJacksonModules(); - - Map getJacksonInjectables(); - + /** + * Configure the JSON mapper. + * + * @see {@link #configureGuice(DruidInjectorBuilder)} for the preferred solution. + */ void configureJsonMapper(ObjectMapper mapper); - void configureGuice(DruidInjectorBuilder builder); - JoinableFactoryWrapper createJoinableFactoryWrapper(LookupExtractorFactoryContainerProvider lookupProvider); + + void finalizeTestFramework(SqlTestFramework sqlTestFramework); } public interface PlannerComponentSupplier @@ -174,18 +185,25 @@ public interface PlannerComponentSupplier */ public static class StandardComponentSupplier implements QueryComponentSupplier { - private final Injector injector; private final File temporaryFolder; public StandardComponentSupplier( - final Injector injector, final File temporaryFolder ) { - this.injector = injector; this.temporaryFolder = temporaryFolder; } + @Override + public void gatherProperties(Properties properties) + { + } + + @Override + public void configureGuice(DruidInjectorBuilder builder) + { + } + @Override public QueryRunnerFactoryConglomerate createCongolmerate( Builder builder, @@ -208,7 +226,8 @@ public QueryRunnerFactoryConglomerate createCongolmerate( @Override public SpecificSegmentsQuerySegmentWalker createQuerySegmentWalker( final QueryRunnerFactoryConglomerate conglomerate, - final JoinableFactoryWrapper joinableFactory + final JoinableFactoryWrapper joinableFactory, + final Injector injector ) { return TestDataBuilder.createMockWalker( @@ -229,42 +248,11 @@ public SqlEngine createEngine(QueryLifecycleFactory qlf, ObjectMapper objectMapp ); } - @Override - public DruidOperatorTable createOperatorTable() - { - return QueryFrameworkUtils.createOperatorTable(injector); - } - - @Override - public ExprMacroTable createMacroTable() - { - return QueryFrameworkUtils.createExprMacroTable(injector); - } - - @Override - public Iterable getJacksonModules() - { - final List modules = new ArrayList<>(new LookupSerdeModule().getJacksonModules()); - modules.add(new SimpleModule().registerSubtypes(ExternalDataSource.class)); - return modules; - } - - @Override - public Map getJacksonInjectables() - { - return new HashMap<>(); - } - @Override public void configureJsonMapper(ObjectMapper mapper) { } - @Override - public void configureGuice(DruidInjectorBuilder builder) - { - } - @Override public JoinableFactoryWrapper createJoinableFactoryWrapper(LookupExtractorFactoryContainerProvider lookupProvider) { @@ -276,6 +264,11 @@ public JoinableFactoryWrapper createJoinableFactoryWrapper(LookupExtractorFactor ) ); } + + @Override + public void finalizeTestFramework(SqlTestFramework sqlTestFramework) + { + } } public static class StandardPlannerComponentSupplier implements PlannerComponentSupplier @@ -476,24 +469,17 @@ public TestSetupModule(Builder builder) @Override public void configure(Binder binder) { - binder.bind(DruidOperatorTable.class).toInstance(componentSupplier.createOperatorTable()); - binder.bind(ExprMacroTable.class).toInstance(componentSupplier.createMacroTable()); + binder.bind(DruidOperatorTable.class).in(LazySingleton.class); binder.bind(DataSegment.PruneSpecsHolder.class).toInstance(DataSegment.PruneSpecsHolder.DEFAULT); } - @Override - public List getJacksonModules() - { - return Lists.newArrayList(componentSupplier.getJacksonModules()); - } - - @Provides + @Provides @Singleton public QueryRunnerFactoryConglomerate conglomerate() { return componentSupplier.createCongolmerate(builder, resourceCloser); } - @Provides + @Provides @Singleton public JoinableFactoryWrapper joinableFactoryWrapper(final Injector injector) { return builder.componentSupplier.createJoinableFactoryWrapper( @@ -501,13 +487,14 @@ public JoinableFactoryWrapper joinableFactoryWrapper(final Injector injector) ); } - @Provides + @Provides @Singleton public SpecificSegmentsQuerySegmentWalker segmentsQuerySegmentWalker(final Injector injector) { try { SpecificSegmentsQuerySegmentWalker walker = componentSupplier.createQuerySegmentWalker( injector.getInstance(QueryRunnerFactoryConglomerate.class), - injector.getInstance(JoinableFactoryWrapper.class) + injector.getInstance(JoinableFactoryWrapper.class), + injector ); resourceCloser.register(walker); return walker; @@ -517,7 +504,7 @@ public SpecificSegmentsQuerySegmentWalker segmentsQuerySegmentWalker(final Injec } } - @Provides + @Provides @Singleton public QueryLifecycleFactory queryLifecycleFactory(final Injector injector) { return QueryFrameworkUtils.createMockQueryLifecycleFactory( @@ -538,11 +525,30 @@ public QueryLifecycleFactory queryLifecycleFactory(final Injector injector) private SqlTestFramework(Builder builder) { this.componentSupplier = builder.componentSupplier; - this.injector = new CalciteTestInjectorBuilder() - .addModule(new TestSetupModule(builder)) - .build(); + Properties properties = new Properties(); + this.componentSupplier.gatherProperties(properties); + Injector startupInjector = new StartupInjectorBuilder() + .withProperties(properties) + .build(); + DruidInjectorBuilder injectorBuilder = new CoreInjectorBuilder(startupInjector) + // Ignore load scopes. This is a unit test, not a Druid node. If a + // test pulls in a module, then pull in that module, even though we are + // not the Druid node to which the module is scoped. + .ignoreLoadScopes() + .addModule(new BasicTestModule()) + .addModule(new SqlAggregationModule()) + .addModule(new ExpressionModule()) + .addModule(new TestSetupModule(builder)); + builder.componentSupplier.configureGuice(injectorBuilder); + this.injector = injectorBuilder.build(); this.engine = builder.componentSupplier.createEngine(queryLifecycleFactory(), queryJsonMapper()); componentSupplier.configureJsonMapper(queryJsonMapper()); + componentSupplier.finalizeTestFramework(this); + } + + public Injector injector() + { + return injector; } public ObjectMapper queryJsonMapper() diff --git a/sql/src/test/resources/calcite/expected/ingest/httpExtern-logicalPlan.txt b/sql/src/test/resources/calcite/expected/ingest/httpExtern-logicalPlan.txt new file mode 100644 index 000000000000..23c84ffd547f --- /dev/null +++ b/sql/src/test/resources/calcite/expected/ingest/httpExtern-logicalPlan.txt @@ -0,0 +1,3 @@ +LogicalInsert(target=[dst], partitionedBy=[AllGranularity], clusteredBy=[]) + LogicalProject(x=[$0], y=[$1], z=[$2]) + ExternalTableScan(dataSource=[{"type":"external","inputSource":{"type":"http","uris":["http:foo.com/bar.csv"],"httpAuthenticationUsername":"bob","httpAuthenticationPassword":{"type":"default","password":"secret"}},"inputFormat":{"type":"csv","columns":["x","y","z"]},"signature":[{"name":"x","type":"STRING"},{"name":"y","type":"STRING"},{"name":"z","type":"LONG"}]}]) diff --git a/sql/src/test/resources/calcite/expected/ingest/insertFromExternal-logicalPlan.txt b/sql/src/test/resources/calcite/expected/ingest/insertFromExternal-logicalPlan.txt new file mode 100644 index 000000000000..01252a96a133 --- /dev/null +++ b/sql/src/test/resources/calcite/expected/ingest/insertFromExternal-logicalPlan.txt @@ -0,0 +1,3 @@ +LogicalInsert(target=[dst], partitionedBy=[AllGranularity], clusteredBy=[]) + LogicalProject(x=[$0], y=[$1], z=[$2]) + ExternalTableScan(dataSource=[{"type":"external","inputSource":{"type":"inline","data":"a,b,1\nc,d,2\n"},"inputFormat":{"type":"csv","columns":["x","y","z"]},"signature":[{"name":"x","type":"STRING"},{"name":"y","type":"STRING"},{"name":"z","type":"LONG"}]}]) diff --git a/sql/src/test/resources/calcite/expected/ingest/insertWithClusteredBy-logicalPlan.txt b/sql/src/test/resources/calcite/expected/ingest/insertWithClusteredBy-logicalPlan.txt new file mode 100644 index 000000000000..eb2d8501d5a1 --- /dev/null +++ b/sql/src/test/resources/calcite/expected/ingest/insertWithClusteredBy-logicalPlan.txt @@ -0,0 +1,4 @@ +LogicalInsert(target=[druid.dst], partitionedBy=[{type=period, period=P1D, timeZone=UTC, origin=null}], clusteredBy=[2, `dim1` DESC, CEIL(`m2`)]) + LogicalSort(sort0=[$1], sort1=[$2], sort2=[$3], dir0=[ASC], dir1=[DESC], dir2=[ASC]) + LogicalProject(__time=[$0], floor_m1=[FLOOR($5)], dim1=[$1], ceil_m2=[CEIL($6)]) + LogicalTableScan(table=[[druid, foo]]) diff --git a/sql/src/test/resources/calcite/expected/ingest/insertWithPartitionedBy-logicalPlan.txt b/sql/src/test/resources/calcite/expected/ingest/insertWithPartitionedBy-logicalPlan.txt new file mode 100644 index 000000000000..730e4adb15a5 --- /dev/null +++ b/sql/src/test/resources/calcite/expected/ingest/insertWithPartitionedBy-logicalPlan.txt @@ -0,0 +1,3 @@ +LogicalInsert(target=[druid.dst], partitionedBy=[{type=period, period=PT1H, timeZone=UTC, origin=null}], clusteredBy=[]) + LogicalProject(__time=[$0], floor_m1=[FLOOR($5)], dim1=[$1]) + LogicalTableScan(table=[[druid, foo]]) diff --git a/sql/src/test/resources/calcite/expected/ingest/localExtern-logicalPlan.txt b/sql/src/test/resources/calcite/expected/ingest/localExtern-logicalPlan.txt new file mode 100644 index 000000000000..759a1587460b --- /dev/null +++ b/sql/src/test/resources/calcite/expected/ingest/localExtern-logicalPlan.txt @@ -0,0 +1,3 @@ +LogicalInsert(target=[dst], partitionedBy=[AllGranularity], clusteredBy=[]) + LogicalProject(x=[$0], y=[$1], z=[$2]) + ExternalTableScan(dataSource=[{"type":"external","inputSource":{"type":"local","files":["/tmp/foo.csv","/tmp/bar.csv"]},"inputFormat":{"type":"csv","columns":["x","y","z"]},"signature":[{"name":"x","type":"STRING"},{"name":"y","type":"STRING"},{"name":"z","type":"LONG"}]}]) From 78c1a2bd665411f7adc93b67e1a3d3995c931f29 Mon Sep 17 00:00:00 2001 From: Katya Macedo <38017980+ektravel@users.noreply.github.com> Date: Fri, 2 Dec 2022 14:19:59 -0600 Subject: [PATCH 108/141] Remove limit from timeseries (#13457) CI build failures seem unrelated to docs --- docs/querying/sql-translation.md | 11 +++++------ 1 file changed, 5 insertions(+), 6 deletions(-) diff --git a/docs/querying/sql-translation.md b/docs/querying/sql-translation.md index d26ae6f0d276..370ad002c1cd 100644 --- a/docs/querying/sql-translation.md +++ b/docs/querying/sql-translation.md @@ -24,7 +24,7 @@ sidebar_label: "SQL query translation" --> > Apache Druid supports two query languages: Druid SQL and [native queries](querying.md). -> This document describes the SQL language. +> This document describes the Druid SQL language. Druid uses [Apache Calcite](https://calcite.apache.org/) to parse and plan SQL queries. Druid translates SQL statements into its [native JSON-based query language](querying.md). @@ -34,8 +34,8 @@ This topic includes best practices and tools to help you achieve good performanc ## Best practices -Consider this (non-exhaustive) list of things to look out for when looking into the performance implications of -how your SQL queries are translated to native queries. +Consider the following non-exhaustive list of best practices when looking into performance implications of +translating Druid SQL queries to native queries. 1. If you wrote a filter on the primary time column `__time`, make sure it is being correctly translated to an `"intervals"` filter, as described in the [Time filters](#time-filters) section below. If not, you may need to change @@ -241,10 +241,9 @@ enabling logging and running this query, we can see that it actually runs as the Druid SQL uses four different native query types. -- [Scan](scan-query.md) is used for queries that do not aggregate (no GROUP BY, no DISTINCT). +- [Scan](scan-query.md) is used for queries that do not aggregate—no GROUP BY, no DISTINCT. -- [Timeseries](timeseriesquery.md) is used for queries that GROUP BY `FLOOR(__time TO unit)` or `TIME_FLOOR(__time, -period)`, have no other grouping expressions, no HAVING or LIMIT clauses, no nesting, and either no ORDER BY, or an +- [Timeseries](timeseriesquery.md) is used for queries that GROUP BY `FLOOR(__time TO unit)` or `TIME_FLOOR(__time, period)`, have no other grouping expressions, no HAVING clause, no nesting, and either no ORDER BY, or an ORDER BY that orders by same expression as present in GROUP BY. It also uses Timeseries for "grand total" queries that have aggregation functions but no GROUP BY. This query type takes advantage of the fact that Druid segments are sorted by time. From 91774196285ea51b823883394ef1bfe8f2417892 Mon Sep 17 00:00:00 2001 From: somu-imply <93540295+somu-imply@users.noreply.github.com> Date: Fri, 2 Dec 2022 18:48:25 -0800 Subject: [PATCH 109/141] Unnest functionality for Druid (#13268) * Moving all unnest cursor code atop refactored code for unnest * Updating unnest cursor * Removing dedup and fixing up some null checks * AllowList changes * Fixing some NPEs * Using bitset for allowlist * Updating the initialization only when cursor is in non-done state * Updating code to skip rows not in allow list * Adding a flag for cases when first element is not in allowed list * Updating for a null in allowList * Splitting unnest cursor into 2 subclasses * Intercepting some apis with columnName for new unnested column * Adding test cases and renaming some stuff * checkstyle fixes * Moving to an interface for Unnest * handling null rows in a dimension * Updating cursors after comments part-1 * Addressing comments and adding some more tests * Reverting a change to ScanQueryRunner and improving a comment * removing an unused function * Updating cursors after comments part 2 * One last fix for review comments * Making some functions private, deleting some comments, adding a test for unnest of unnest with allowList * Adding an exception for a case * Closure for unnest data source * Adding some javadocs * One minor change in makeDimSelector of columnarCursor * Updating an error message * Update processing/src/main/java/org/apache/druid/segment/DimensionUnnestCursor.java Co-authored-by: Abhishek Agarwal <1477457+abhishekagarwal87@users.noreply.github.com> * Unnesting on virtual columns was missing an object array, adding that to support virtual columns unnesting * Updating exceptions to use UOE * Renamed files, added column capability test on adapter, return statement and made unnest datasource not cacheable for the time being * Handling for null values in dim selector * Fixing a NPE for null row * Updating capabilities * Updating capabilities Co-authored-by: Abhishek Agarwal <1477457+abhishekagarwal87@users.noreply.github.com> --- .../org/apache/druid/query/DataSource.java | 3 +- .../apache/druid/query/UnnestDataSource.java | 212 ++++++ .../query/planning/DataSourceAnalysis.java | 39 +- .../UnnestColumnValueSelectorCursor.java | 336 ++++++++++ .../druid/segment/UnnestDimensionCursor.java | 415 ++++++++++++ .../druid/segment/UnnestSegmentReference.java | 115 ++++ .../druid/segment/UnnestStorageAdapter.java | 234 +++++++ .../org/apache/druid/segment/ListCursor.java | 228 +++++++ .../UnnestColumnValueSelectorCursorTest.java | 632 ++++++++++++++++++ .../segment/UnnestStorageAdapterTest.java | 399 +++++++++++ 10 files changed, 2602 insertions(+), 11 deletions(-) create mode 100644 processing/src/main/java/org/apache/druid/query/UnnestDataSource.java create mode 100644 processing/src/main/java/org/apache/druid/segment/UnnestColumnValueSelectorCursor.java create mode 100644 processing/src/main/java/org/apache/druid/segment/UnnestDimensionCursor.java create mode 100644 processing/src/main/java/org/apache/druid/segment/UnnestSegmentReference.java create mode 100644 processing/src/main/java/org/apache/druid/segment/UnnestStorageAdapter.java create mode 100644 processing/src/test/java/org/apache/druid/segment/ListCursor.java create mode 100644 processing/src/test/java/org/apache/druid/segment/UnnestColumnValueSelectorCursorTest.java create mode 100644 processing/src/test/java/org/apache/druid/segment/UnnestStorageAdapterTest.java diff --git a/processing/src/main/java/org/apache/druid/query/DataSource.java b/processing/src/main/java/org/apache/druid/query/DataSource.java index f56a3550a34b..43dfb3be85c8 100644 --- a/processing/src/main/java/org/apache/druid/query/DataSource.java +++ b/processing/src/main/java/org/apache/druid/query/DataSource.java @@ -41,7 +41,8 @@ @JsonSubTypes.Type(value = JoinDataSource.class, name = "join"), @JsonSubTypes.Type(value = LookupDataSource.class, name = "lookup"), @JsonSubTypes.Type(value = InlineDataSource.class, name = "inline"), - @JsonSubTypes.Type(value = GlobalTableDataSource.class, name = "globalTable") + @JsonSubTypes.Type(value = GlobalTableDataSource.class, name = "globalTable"), + @JsonSubTypes.Type(value = UnnestDataSource.class, name = "unnest") }) public interface DataSource { diff --git a/processing/src/main/java/org/apache/druid/query/UnnestDataSource.java b/processing/src/main/java/org/apache/druid/query/UnnestDataSource.java new file mode 100644 index 000000000000..46237016743f --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/UnnestDataSource.java @@ -0,0 +1,212 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query; + +import com.fasterxml.jackson.annotation.JsonCreator; +import com.fasterxml.jackson.annotation.JsonProperty; +import com.google.common.collect.ImmutableList; +import org.apache.druid.java.util.common.IAE; +import org.apache.druid.segment.SegmentReference; +import org.apache.druid.segment.UnnestSegmentReference; +import org.apache.druid.utils.JvmUtils; + +import javax.annotation.Nullable; +import java.util.LinkedHashSet; +import java.util.List; +import java.util.Objects; +import java.util.Set; +import java.util.concurrent.atomic.AtomicLong; +import java.util.function.Function; + +/** + * The data source for representing an unnest operation. + * + * An unnest data source has the following: + * a base data source which is to be unnested + * the column name of the MVD which will be unnested + * the name of the column that will hold the unnested values + * and an allowlist serving as a filter of which values in the MVD will be unnested. + */ +public class UnnestDataSource implements DataSource +{ + private final DataSource base; + private final String column; + private final String outputName; + private final LinkedHashSet allowList; + + private UnnestDataSource( + DataSource dataSource, + String columnName, + String outputName, + LinkedHashSet allowList + ) + { + this.base = dataSource; + this.column = columnName; + this.outputName = outputName; + this.allowList = allowList; + } + + @JsonCreator + public static UnnestDataSource create( + @JsonProperty("base") DataSource base, + @JsonProperty("column") String columnName, + @JsonProperty("outputName") String outputName, + @Nullable @JsonProperty("allowList") LinkedHashSet allowList + ) + { + return new UnnestDataSource(base, columnName, outputName, allowList); + } + + @JsonProperty("base") + public DataSource getBase() + { + return base; + } + + @JsonProperty("column") + public String getColumn() + { + return column; + } + + @JsonProperty("outputName") + public String getOutputName() + { + return outputName; + } + + @JsonProperty("allowList") + public LinkedHashSet getAllowList() + { + return allowList; + } + + @Override + public Set getTableNames() + { + return base.getTableNames(); + } + + @Override + public List getChildren() + { + return ImmutableList.of(base); + } + + @Override + public DataSource withChildren(List children) + { + if (children.size() != 1) { + throw new IAE("Expected [1] child, got [%d]", children.size()); + } + return new UnnestDataSource(children.get(0), column, outputName, allowList); + } + + @Override + public boolean isCacheable(boolean isBroker) + { + return false; + } + + @Override + public boolean isGlobal() + { + return base.isGlobal(); + } + + @Override + public boolean isConcrete() + { + return base.isConcrete(); + } + + @Override + public Function createSegmentMapFunction( + Query query, + AtomicLong cpuTimeAccumulator + ) + { + final Function segmentMapFn = base.createSegmentMapFunction( + query, + cpuTimeAccumulator + ); + return JvmUtils.safeAccumulateThreadCpuTime( + cpuTimeAccumulator, + () -> { + if (column == null) { + return segmentMapFn; + } else if (column.isEmpty()) { + return segmentMapFn; + } else { + return + baseSegment -> + new UnnestSegmentReference( + segmentMapFn.apply(baseSegment), + column, + outputName, + allowList + ); + } + } + ); + + } + + @Override + public DataSource withUpdatedDataSource(DataSource newSource) + { + return new UnnestDataSource(newSource, column, outputName, allowList); + } + + @Override + public byte[] getCacheKey() + { + // The column being unnested would need to be part of the cache key + // as the results are dependent on what column is being unnested. + // Currently, it is not cacheable. + // Future development should use the table name and column came to + // create an appropriate cac + return null; + } + + @Override + public boolean equals(Object o) + { + if (this == o) { + return true; + } + if (o == null || getClass() != o.getClass()) { + return false; + } + UnnestDataSource that = (UnnestDataSource) o; + return column.equals(that.column) + && outputName.equals(that.outputName) + && base.equals(that.base); + } + + @Override + public int hashCode() + { + return Objects.hash(base, column, outputName); + } +} + + diff --git a/processing/src/main/java/org/apache/druid/query/planning/DataSourceAnalysis.java b/processing/src/main/java/org/apache/druid/query/planning/DataSourceAnalysis.java index c329e3a57089..63c2c8b815d5 100644 --- a/processing/src/main/java/org/apache/druid/query/planning/DataSourceAnalysis.java +++ b/processing/src/main/java/org/apache/druid/query/planning/DataSourceAnalysis.java @@ -28,6 +28,7 @@ import org.apache.druid.query.QueryDataSource; import org.apache.druid.query.TableDataSource; import org.apache.druid.query.UnionDataSource; +import org.apache.druid.query.UnnestDataSource; import org.apache.druid.query.filter.DimFilter; import org.apache.druid.query.spec.QuerySegmentSpec; @@ -112,17 +113,29 @@ public static DataSourceAnalysis forDataSource(final DataSource dataSource) Query baseQuery = null; DataSource current = dataSource; - while (current instanceof QueryDataSource) { - final Query subQuery = ((QueryDataSource) current).getQuery(); + // This needs to be an or condition between QueryDataSource and UnnestDataSource + // As queries can have interleaving query and unnest data sources. + // Ideally if each data source generate their own analysis object we can avoid the or here + // and have cleaner code. Especially as we increase the types of data sources in future + // these or checks will be tedious. Future development should move forDataSource method + // into each data source. - if (!(subQuery instanceof BaseQuery)) { - // We must verify that the subQuery is a BaseQuery, because it is required to make "getBaseQuerySegmentSpec" - // work properly. All built-in query types are BaseQuery, so we only expect this with funky extension queries. - throw new IAE("Cannot analyze subquery of class[%s]", subQuery.getClass().getName()); - } + while (current instanceof QueryDataSource || current instanceof UnnestDataSource) { + if (current instanceof QueryDataSource) { + final Query subQuery = ((QueryDataSource) current).getQuery(); + + if (!(subQuery instanceof BaseQuery)) { + // We must verify that the subQuery is a BaseQuery, because it is required to make "getBaseQuerySegmentSpec" + // work properly. All built-in query types are BaseQuery, so we only expect this with funky extension queries. + throw new IAE("Cannot analyze subquery of class[%s]", subQuery.getClass().getName()); + } - baseQuery = subQuery; - current = subQuery.getDataSource(); + baseQuery = subQuery; + current = subQuery.getDataSource(); + } else { + final UnnestDataSource unnestDataSource = (UnnestDataSource) current; + current = unnestDataSource.getBase(); + } } if (current instanceof JoinDataSource) { @@ -276,7 +289,8 @@ public boolean isConcreteBased() /** * Returns true if this datasource is concrete-based (see {@link #isConcreteBased()}, and the base datasource is a - * {@link TableDataSource} or a {@link UnionDataSource} composed entirely of {@link TableDataSource}. This is an + * {@link TableDataSource} or a {@link UnionDataSource} composed entirely of {@link TableDataSource} + * or an {@link UnnestDataSource} composed entirely of {@link TableDataSource} . This is an * important property, because it corresponds to datasources that can be handled by Druid's distributed query stack. */ public boolean isConcreteTableBased() @@ -286,6 +300,10 @@ public boolean isConcreteTableBased() // so check anyway for future-proofing. return isConcreteBased() && (baseDataSource instanceof TableDataSource || (baseDataSource instanceof UnionDataSource && + baseDataSource.getChildren() + .stream() + .allMatch(ds -> ds instanceof TableDataSource)) + || (baseDataSource instanceof UnnestDataSource && baseDataSource.getChildren() .stream() .allMatch(ds -> ds instanceof TableDataSource))); @@ -298,6 +316,7 @@ public boolean isQuery() { return dataSource instanceof QueryDataSource; } + /** * Returns true if this datasource is made out of a join operation diff --git a/processing/src/main/java/org/apache/druid/segment/UnnestColumnValueSelectorCursor.java b/processing/src/main/java/org/apache/druid/segment/UnnestColumnValueSelectorCursor.java new file mode 100644 index 000000000000..db4acb893ea0 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/segment/UnnestColumnValueSelectorCursor.java @@ -0,0 +1,336 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.segment; + +import org.apache.druid.java.util.common.UOE; +import org.apache.druid.query.BaseQuery; +import org.apache.druid.query.dimension.DimensionSpec; +import org.apache.druid.query.monomorphicprocessing.RuntimeShapeInspector; +import org.apache.druid.segment.column.ColumnCapabilities; +import org.apache.druid.segment.column.ColumnCapabilitiesImpl; +import org.joda.time.DateTime; + +import javax.annotation.Nullable; +import java.util.ArrayList; +import java.util.Arrays; +import java.util.LinkedHashSet; +import java.util.List; + +/** + * The cursor to help unnest MVDs without dictionary encoding and ARRAY type selectors. + *

    + * Consider a segment has 2 rows + * ['a', 'b', 'c'] + * ['d', 'e'] + *

    + * The baseCursor points to the row ['a', 'b', 'c'] + * while the unnestCursor with each call of advance() moves over individual elements. + *

    + * unnestCursor.advance() -> 'a' + * unnestCursor.advance() -> 'b' + * unnestCursor.advance() -> 'c' + * unnestCursor.advance() -> 'd' (advances base cursor first) + * unnestCursor.advance() -> 'e' + *

    + *

    + * The allowSet if available helps skip over elements which are not in the allowList by moving the cursor to + * the next available match. + *

    + * The index reference points to the index of each row that the unnest cursor is accessing through currentVal + * The index ranges from 0 to the size of the list in each row which is held in the unnestListForCurrentRow + *

    + * The needInitialization flag sets up the initial values of unnestListForCurrentRow at the beginning of the segment + */ +public class UnnestColumnValueSelectorCursor implements Cursor +{ + private final Cursor baseCursor; + private final ColumnSelectorFactory baseColumnSelectorFactory; + private final ColumnValueSelector columnValueSelector; + private final String columnName; + private final String outputName; + private final LinkedHashSet allowSet; + private int index; + private Object currentVal; + private List unnestListForCurrentRow; + private boolean needInitialization; + + public UnnestColumnValueSelectorCursor( + Cursor cursor, + ColumnSelectorFactory baseColumSelectorFactory, + String columnName, + String outputColumnName, + LinkedHashSet allowSet + ) + { + this.baseCursor = cursor; + this.baseColumnSelectorFactory = baseColumSelectorFactory; + this.columnValueSelector = this.baseColumnSelectorFactory.makeColumnValueSelector(columnName); + this.columnName = columnName; + this.index = 0; + this.outputName = outputColumnName; + this.needInitialization = true; + this.allowSet = allowSet; + } + + @Override + public ColumnSelectorFactory getColumnSelectorFactory() + { + return new ColumnSelectorFactory() + { + @Override + public DimensionSelector makeDimensionSelector(DimensionSpec dimensionSpec) + { + if (!outputName.equals(dimensionSpec.getDimension())) { + return baseColumnSelectorFactory.makeDimensionSelector(dimensionSpec); + } + throw new UOE("Unsupported dimension selector while using column value selector for column [%s]", outputName); + } + + @Override + public ColumnValueSelector makeColumnValueSelector(String columnName) + { + if (!outputName.equals(columnName)) { + return baseColumnSelectorFactory.makeColumnValueSelector(columnName); + } + return new ColumnValueSelector() + { + @Override + public double getDouble() + { + Object value = getObject(); + if (value == null) { + return 0; + } + if (value instanceof Number) { + return ((Number) value).doubleValue(); + } + throw new UOE("Cannot convert object to double"); + } + + @Override + public float getFloat() + { + Object value = getObject(); + if (value == null) { + return 0; + } + if (value instanceof Number) { + return ((Number) value).floatValue(); + } + throw new UOE("Cannot convert object to float"); + } + + @Override + public long getLong() + { + Object value = getObject(); + if (value == null) { + return 0; + } + if (value instanceof Number) { + return ((Number) value).longValue(); + } + throw new UOE("Cannot convert object to long"); + } + + @Override + public void inspectRuntimeShape(RuntimeShapeInspector inspector) + { + columnValueSelector.inspectRuntimeShape(inspector); + } + + @Override + public boolean isNull() + { + return getObject() == null; + } + + @Nullable + @Override + public Object getObject() + { + if (!unnestListForCurrentRow.isEmpty()) { + if (allowSet == null || allowSet.isEmpty()) { + return unnestListForCurrentRow.get(index); + } else if (allowSet.contains((String) unnestListForCurrentRow.get(index))) { + return unnestListForCurrentRow.get(index); + } + } + return null; + } + + @Override + public Class classOfObject() + { + return Object.class; + } + }; + } + + @Nullable + @Override + public ColumnCapabilities getColumnCapabilities(String column) + { + if (!outputName.equals(column)) { + return baseColumnSelectorFactory.getColumnCapabilities(column); + } + final ColumnCapabilities capabilities = baseColumnSelectorFactory.getColumnCapabilities(columnName); + if (capabilities.isArray()) { + return ColumnCapabilitiesImpl.copyOf(capabilities).setType(capabilities.getElementType()); + } + if (capabilities.hasMultipleValues().isTrue()) { + return ColumnCapabilitiesImpl.copyOf(capabilities).setHasMultipleValues(false); + } + return baseColumnSelectorFactory.getColumnCapabilities(columnName); + } + }; + } + + @Override + public DateTime getTime() + { + return baseCursor.getTime(); + } + + @Override + public void advance() + { + advanceUninterruptibly(); + BaseQuery.checkInterrupted(); + } + + @Override + public void advanceUninterruptibly() + { + do { + advanceAndUpdate(); + } while (matchAndProceed()); + } + + @Override + public boolean isDone() + { + if (needInitialization && !baseCursor.isDone()) { + initialize(); + } + return baseCursor.isDone(); + } + + @Override + public boolean isDoneOrInterrupted() + { + if (needInitialization && !baseCursor.isDoneOrInterrupted()) { + initialize(); + } + return baseCursor.isDoneOrInterrupted(); + } + + @Override + public void reset() + { + index = 0; + needInitialization = true; + baseCursor.reset(); + } + + /** + * This method populates the objects when the base cursor moves to the next row + * + * @param firstRun flag to populate one time object references to hold values for unnest cursor + */ + private void getNextRow(boolean firstRun) + { + currentVal = this.columnValueSelector.getObject(); + if (currentVal == null) { + if (!firstRun) { + unnestListForCurrentRow = new ArrayList<>(); + } + unnestListForCurrentRow.add(null); + } else { + if (currentVal instanceof List) { + unnestListForCurrentRow = (List) currentVal; + } else if (currentVal instanceof Object[]) { + unnestListForCurrentRow = Arrays.asList((Object[]) currentVal); + } else if (currentVal.getClass().equals(String.class)) { + if (!firstRun) { + unnestListForCurrentRow = new ArrayList<>(); + } + unnestListForCurrentRow.add(currentVal); + } + } + } + + /** + * This initializes the unnest cursor and creates data structures + * to start iterating over the values to be unnested. + * This would also create a bitset for dictonary encoded columns to + * check for matching values specified in allowedList of UnnestDataSource. + */ + private void initialize() + { + this.unnestListForCurrentRow = new ArrayList<>(); + getNextRow(needInitialization); + if (allowSet != null) { + if (!allowSet.isEmpty()) { + if (!allowSet.contains((String) unnestListForCurrentRow.get(index))) { + advance(); + } + } + } + needInitialization = false; + } + + /** + * This advances the cursor to move to the next element to be unnested. + * When the last element in a row is unnested, it is also responsible + * to move the base cursor to the next row for unnesting and repopulates + * the data structures, created during initialize(), to point to the new row + */ + private void advanceAndUpdate() + { + if (unnestListForCurrentRow.isEmpty() || index >= unnestListForCurrentRow.size() - 1) { + index = 0; + baseCursor.advance(); + if (!baseCursor.isDone()) { + getNextRow(needInitialization); + } + } else { + index++; + } + } + + /** + * This advances the unnest cursor in cases where an allowList is specified + * and the current value at the unnest cursor is not in the allowList. + * The cursor in such cases is moved till the next match is found. + * + * @return a boolean to indicate whether to stay or move cursor + */ + private boolean matchAndProceed() + { + boolean matchStatus; + if (allowSet == null || allowSet.isEmpty()) { + matchStatus = true; + } else { + matchStatus = allowSet.contains((String) unnestListForCurrentRow.get(index)); + } + return !baseCursor.isDone() && !matchStatus; + } +} diff --git a/processing/src/main/java/org/apache/druid/segment/UnnestDimensionCursor.java b/processing/src/main/java/org/apache/druid/segment/UnnestDimensionCursor.java new file mode 100644 index 000000000000..46a2c626caf5 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/segment/UnnestDimensionCursor.java @@ -0,0 +1,415 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.segment; + +import com.google.common.base.Predicate; +import org.apache.druid.query.BaseQuery; +import org.apache.druid.query.dimension.DefaultDimensionSpec; +import org.apache.druid.query.dimension.DimensionSpec; +import org.apache.druid.query.filter.ValueMatcher; +import org.apache.druid.query.monomorphicprocessing.RuntimeShapeInspector; +import org.apache.druid.segment.column.ColumnCapabilities; +import org.apache.druid.segment.column.ColumnCapabilitiesImpl; +import org.apache.druid.segment.data.IndexedInts; +import org.joda.time.DateTime; + +import javax.annotation.Nullable; +import java.util.BitSet; +import java.util.LinkedHashSet; + +/** + * The cursor to help unnest MVDs with dictionary encoding. + * Consider a segment has 2 rows + * ['a', 'b', 'c'] + * ['d', 'c'] + *

    + * Considering dictionary encoding, these are represented as + *

    + * 'a' -> 0 + * 'b' -> 1 + * 'c' -> 2 + * 'd' -> 3 + *

    + * The baseCursor points to the row of IndexedInts [0, 1, 2] + * while the unnestCursor with each call of advance() moves over individual elements. + *

    + * advance() -> 0 -> 'a' + * advance() -> 1 -> 'b' + * advance() -> 2 -> 'c' + * advance() -> 3 -> 'd' (advances base cursor first) + * advance() -> 2 -> 'c' + *

    + * Total 5 advance calls above + *

    + * The allowSet, if available, helps skip over elements that are not in the allowList by moving the cursor to + * the next available match. The hashSet is converted into a bitset (during initialization) for efficiency. + * If allowSet is ['c', 'd'] then the advance moves over to the next available match + *

    + * advance() -> 2 -> 'c' + * advance() -> 3 -> 'd' (advances base cursor first) + * advance() -> 2 -> 'c' + *

    + * Total 3 advance calls in this case + *

    + * The index reference points to the index of each row that the unnest cursor is accessing + * The indexedInts for each row are held in the indexedIntsForCurrentRow object + *

    + * The needInitialization flag sets up the initial values of indexedIntsForCurrentRow at the beginning of the segment + */ +public class UnnestDimensionCursor implements Cursor +{ + private final Cursor baseCursor; + private final DimensionSelector dimSelector; + private final String columnName; + private final String outputName; + private final LinkedHashSet allowSet; + private final BitSet allowedBitSet; + private final ColumnSelectorFactory baseColumnSelectorFactory; + private int index; + private IndexedInts indexedIntsForCurrentRow; + private boolean needInitialization; + private SingleIndexInts indexIntsForRow; + + public UnnestDimensionCursor( + Cursor cursor, + ColumnSelectorFactory baseColumnSelectorFactory, + String columnName, + String outputColumnName, + LinkedHashSet allowSet + ) + { + this.baseCursor = cursor; + this.baseColumnSelectorFactory = baseColumnSelectorFactory; + this.dimSelector = this.baseColumnSelectorFactory.makeDimensionSelector(DefaultDimensionSpec.of(columnName)); + this.columnName = columnName; + this.index = 0; + this.outputName = outputColumnName; + this.needInitialization = true; + this.allowSet = allowSet; + this.allowedBitSet = new BitSet(); + } + + @Override + public ColumnSelectorFactory getColumnSelectorFactory() + { + return new ColumnSelectorFactory() + { + @Override + public DimensionSelector makeDimensionSelector(DimensionSpec dimensionSpec) + { + if (!outputName.equals(dimensionSpec.getDimension())) { + return baseColumnSelectorFactory.makeDimensionSelector(dimensionSpec); + } + + return new DimensionSelector() + { + @Override + public IndexedInts getRow() + { + // This object reference has been created + // during the call to initialize and referenced henceforth + return indexIntsForRow; + } + + @Override + public ValueMatcher makeValueMatcher(@Nullable String value) + { + final int idForLookup = idLookup().lookupId(value); + if (idForLookup < 0) { + return new ValueMatcher() + { + @Override + public boolean matches() + { + return false; + } + + @Override + public void inspectRuntimeShape(RuntimeShapeInspector inspector) + { + + } + }; + } + + return new ValueMatcher() + { + @Override + public boolean matches() + { + return idForLookup == indexedIntsForCurrentRow.get(index); + } + + @Override + public void inspectRuntimeShape(RuntimeShapeInspector inspector) + { + dimSelector.inspectRuntimeShape(inspector); + } + }; + } + + @Override + public ValueMatcher makeValueMatcher(Predicate predicate) + { + return DimensionSelectorUtils.makeValueMatcherGeneric(this, predicate); + } + + @Override + public void inspectRuntimeShape(RuntimeShapeInspector inspector) + { + dimSelector.inspectRuntimeShape(inspector); + } + + @Nullable + @Override + public Object getObject() + { + if (indexedIntsForCurrentRow == null) { + return null; + } + if (allowedBitSet.isEmpty()) { + if (allowSet == null || allowSet.isEmpty()) { + return lookupName(indexedIntsForCurrentRow.get(index)); + } + } else if (allowedBitSet.get(indexedIntsForCurrentRow.get(index))) { + return lookupName(indexedIntsForCurrentRow.get(index)); + } + return null; + } + + @Override + public Class classOfObject() + { + return Object.class; + } + + @Override + public int getValueCardinality() + { + if (!allowedBitSet.isEmpty()) { + return allowedBitSet.cardinality(); + } + return dimSelector.getValueCardinality(); + } + + @Nullable + @Override + public String lookupName(int id) + { + return dimSelector.lookupName(id); + } + + @Override + public boolean nameLookupPossibleInAdvance() + { + return dimSelector.nameLookupPossibleInAdvance(); + } + + @Nullable + @Override + public IdLookup idLookup() + { + return dimSelector.idLookup(); + } + }; + } + + /* + This ideally should not be called. If called delegate using the makeDimensionSelector + */ + @Override + public ColumnValueSelector makeColumnValueSelector(String columnName) + { + if (!outputName.equals(columnName)) { + return baseColumnSelectorFactory.makeColumnValueSelector(columnName); + } + return makeDimensionSelector(DefaultDimensionSpec.of(columnName)); + } + + @Nullable + @Override + public ColumnCapabilities getColumnCapabilities(String column) + { + if (!outputName.equals(column)) { + return baseColumnSelectorFactory.getColumnCapabilities(column); + } + // This currently returns the same type as of the column to be unnested + // This is fine for STRING types + // But going forward if the dimension to be unnested is of type ARRAY, + // this should strip down to the base type of the array + final ColumnCapabilities capabilities = baseColumnSelectorFactory.getColumnCapabilities(columnName); + if (capabilities.isArray()) { + return ColumnCapabilitiesImpl.copyOf(capabilities).setType(capabilities.getElementType()); + } + if (capabilities.hasMultipleValues().isTrue()) { + return ColumnCapabilitiesImpl.copyOf(capabilities).setHasMultipleValues(false); + } + return baseColumnSelectorFactory.getColumnCapabilities(columnName); + } + }; + } + + @Override + public DateTime getTime() + { + return baseCursor.getTime(); + } + + @Override + public void advance() + { + advanceUninterruptibly(); + BaseQuery.checkInterrupted(); + } + + @Override + public void advanceUninterruptibly() + { + do { + advanceAndUpdate(); + } while (matchAndProceed()); + } + + @Override + public boolean isDone() + { + if (needInitialization && !baseCursor.isDone()) { + initialize(); + } + return baseCursor.isDone(); + } + + @Override + public boolean isDoneOrInterrupted() + { + if (needInitialization && !baseCursor.isDoneOrInterrupted()) { + initialize(); + } + return baseCursor.isDoneOrInterrupted(); + } + + @Override + public void reset() + { + index = 0; + needInitialization = true; + baseCursor.reset(); + } + + /** + * This initializes the unnest cursor and creates data structures + * to start iterating over the values to be unnested. + * This would also create a bitset for dictonary encoded columns to + * check for matching values specified in allowedList of UnnestDataSource. + */ + private void initialize() + { + IdLookup idLookup = dimSelector.idLookup(); + this.indexIntsForRow = new SingleIndexInts(); + if (allowSet != null && !allowSet.isEmpty() && idLookup != null) { + for (String s : allowSet) { + if (idLookup.lookupId(s) >= 0) { + allowedBitSet.set(idLookup.lookupId(s)); + } + } + } + if (dimSelector.getObject() != null) { + this.indexedIntsForCurrentRow = dimSelector.getRow(); + } + if (!allowedBitSet.isEmpty()) { + if (!allowedBitSet.get(indexedIntsForCurrentRow.get(index))) { + advance(); + } + } + needInitialization = false; + } + + /** + * This advances the cursor to move to the next element to be unnested. + * When the last element in a row is unnested, it is also responsible + * to move the base cursor to the next row for unnesting and repopulates + * the data structures, created during initialize(), to point to the new row + */ + private void advanceAndUpdate() + { + if (indexedIntsForCurrentRow == null) { + index = 0; + if (!baseCursor.isDone()) { + baseCursor.advanceUninterruptibly(); + } + } else { + if (index >= indexedIntsForCurrentRow.size() - 1) { + if (!baseCursor.isDone()) { + baseCursor.advanceUninterruptibly(); + } + if (!baseCursor.isDone()) { + indexedIntsForCurrentRow = dimSelector.getRow(); + } + index = 0; + } else { + ++index; + } + } + } + + /** + * This advances the unnest cursor in cases where an allowList is specified + * and the current value at the unnest cursor is not in the allowList. + * The cursor in such cases is moved till the next match is found. + * + * @return a boolean to indicate whether to stay or move cursor + */ + private boolean matchAndProceed() + { + boolean matchStatus; + if ((allowSet == null || allowSet.isEmpty()) && allowedBitSet.isEmpty()) { + matchStatus = true; + } else { + matchStatus = allowedBitSet.get(indexedIntsForCurrentRow.get(index)); + } + return !baseCursor.isDone() && !matchStatus; + } + + // Helper class to help in returning + // getRow from the dimensionSelector + // This is set in the initialize method + private class SingleIndexInts implements IndexedInts + { + + @Override + public void inspectRuntimeShape(RuntimeShapeInspector inspector) + { + //nothing to inspect + } + + @Override + public int size() + { + // After unnest each row will have a single element + return 1; + } + + @Override + public int get(int idx) + { + return indexedIntsForCurrentRow.get(index); + } + } +} diff --git a/processing/src/main/java/org/apache/druid/segment/UnnestSegmentReference.java b/processing/src/main/java/org/apache/druid/segment/UnnestSegmentReference.java new file mode 100644 index 000000000000..9da6b8132cbb --- /dev/null +++ b/processing/src/main/java/org/apache/druid/segment/UnnestSegmentReference.java @@ -0,0 +1,115 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.segment; + +import org.apache.druid.java.util.common.io.Closer; +import org.apache.druid.java.util.common.logger.Logger; +import org.apache.druid.timeline.SegmentId; +import org.apache.druid.utils.CloseableUtils; +import org.joda.time.Interval; + +import javax.annotation.Nullable; +import java.io.Closeable; +import java.io.IOException; +import java.util.LinkedHashSet; +import java.util.Optional; + +/** + * The segment reference for the Unnest Data Source. + * The input column name, output name and the allowSet follow from {@link org.apache.druid.query.UnnestDataSource} + */ +public class UnnestSegmentReference implements SegmentReference +{ + private static final Logger log = new Logger(UnnestSegmentReference.class); + + private final SegmentReference baseSegment; + private final String dimension; + private final String renamedOutputDimension; + private final LinkedHashSet allowSet; + + public UnnestSegmentReference(SegmentReference baseSegment, String dimension, String outputName, LinkedHashSet allowList) + { + this.baseSegment = baseSegment; + this.dimension = dimension; + this.renamedOutputDimension = outputName; + this.allowSet = allowList; + } + + @Override + public Optional acquireReferences() + { + Closer closer = Closer.create(); + try { + boolean acquireFailed = baseSegment.acquireReferences().map(closeable -> { + closer.register(closeable); + return false; + }).orElse(true); + + if (acquireFailed) { + CloseableUtils.closeAndWrapExceptions(closer); + return Optional.empty(); + } else { + return Optional.of(closer); + } + } + catch (Throwable e) { + // acquireReferences is not permitted to throw exceptions. + CloseableUtils.closeAndSuppressExceptions(closer, e::addSuppressed); + log.warn(e, "Exception encountered while trying to acquire reference"); + return Optional.empty(); + } + } + + @Override + public SegmentId getId() + { + return baseSegment.getId(); + } + + @Override + public Interval getDataInterval() + { + return baseSegment.getDataInterval(); + } + + @Nullable + @Override + public QueryableIndex asQueryableIndex() + { + return null; + } + + @Override + public StorageAdapter asStorageAdapter() + { + return new UnnestStorageAdapter( + baseSegment.asStorageAdapter(), + dimension, + renamedOutputDimension, + allowSet + ); + } + + @Override + public void close() throws IOException + { + baseSegment.close(); + } +} diff --git a/processing/src/main/java/org/apache/druid/segment/UnnestStorageAdapter.java b/processing/src/main/java/org/apache/druid/segment/UnnestStorageAdapter.java new file mode 100644 index 000000000000..f76ab89270af --- /dev/null +++ b/processing/src/main/java/org/apache/druid/segment/UnnestStorageAdapter.java @@ -0,0 +1,234 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.segment; + +import com.google.common.collect.Lists; +import org.apache.druid.java.util.common.granularity.Granularity; +import org.apache.druid.java.util.common.guava.Sequence; +import org.apache.druid.java.util.common.guava.Sequences; +import org.apache.druid.query.QueryMetrics; +import org.apache.druid.query.filter.Filter; +import org.apache.druid.query.filter.InDimFilter; +import org.apache.druid.segment.column.ColumnCapabilities; +import org.apache.druid.segment.data.Indexed; +import org.apache.druid.segment.data.ListIndexed; +import org.apache.druid.segment.filter.AndFilter; +import org.joda.time.DateTime; +import org.joda.time.Interval; + +import javax.annotation.Nullable; +import java.util.Arrays; +import java.util.LinkedHashSet; +import java.util.Objects; + +/** + * This class serves as the Storage Adapter for the Unnest Segment and is responsible for creating the cursors + * If the column is dictionary encoded it creates {@link UnnestDimensionCursor} else {@link UnnestColumnValueSelectorCursor} + * These cursors help navigate the segments for these cases + */ +public class UnnestStorageAdapter implements StorageAdapter +{ + private final StorageAdapter baseAdapter; + private final String dimensionToUnnest; + private final String outputColumnName; + private final LinkedHashSet allowSet; + + public UnnestStorageAdapter( + final StorageAdapter baseAdapter, + final String dimension, + final String outputColumnName, + final LinkedHashSet allowSet + ) + { + this.baseAdapter = baseAdapter; + this.dimensionToUnnest = dimension; + this.outputColumnName = outputColumnName; + this.allowSet = allowSet; + } + + @Override + public Sequence makeCursors( + @Nullable Filter filter, + Interval interval, + VirtualColumns virtualColumns, + Granularity gran, + boolean descending, + @Nullable QueryMetrics queryMetrics + ) + { + Filter updatedFilter; + if (allowSet != null && !allowSet.isEmpty()) { + final InDimFilter allowListFilters; + allowListFilters = new InDimFilter(dimensionToUnnest, allowSet); + if (filter != null) { + updatedFilter = new AndFilter(Arrays.asList(filter, allowListFilters)); + } else { + updatedFilter = allowListFilters; + } + } else { + updatedFilter = filter; + } + final Sequence baseCursorSequence = baseAdapter.makeCursors( + updatedFilter, + interval, + virtualColumns, + gran, + descending, + queryMetrics + ); + + return Sequences.map( + baseCursorSequence, + cursor -> { + Objects.requireNonNull(cursor); + Cursor retVal = cursor; + ColumnCapabilities capabilities = cursor.getColumnSelectorFactory().getColumnCapabilities(dimensionToUnnest); + if (capabilities != null) { + if (capabilities.isDictionaryEncoded().and(capabilities.areDictionaryValuesUnique()).isTrue()) { + retVal = new UnnestDimensionCursor( + retVal, + retVal.getColumnSelectorFactory(), + dimensionToUnnest, + outputColumnName, + allowSet + ); + } else { + retVal = new UnnestColumnValueSelectorCursor( + retVal, + retVal.getColumnSelectorFactory(), + dimensionToUnnest, + outputColumnName, + allowSet + ); + } + } else { + retVal = new UnnestColumnValueSelectorCursor( + retVal, + retVal.getColumnSelectorFactory(), + dimensionToUnnest, + outputColumnName, + allowSet + ); + } + return retVal; + } + ); + } + + @Override + public Interval getInterval() + { + return baseAdapter.getInterval(); + } + + @Override + public Indexed getAvailableDimensions() + { + final LinkedHashSet availableDimensions = new LinkedHashSet<>(); + + for (String dim : baseAdapter.getAvailableDimensions()) { + availableDimensions.add(dim); + } + availableDimensions.add(outputColumnName); + return new ListIndexed<>(Lists.newArrayList(availableDimensions)); + } + + @Override + public Iterable getAvailableMetrics() + { + return baseAdapter.getAvailableMetrics(); + } + + @Override + public int getDimensionCardinality(String column) + { + if (!outputColumnName.equals(column)) { + return baseAdapter.getDimensionCardinality(column); + } + return baseAdapter.getDimensionCardinality(dimensionToUnnest); + } + + @Override + public DateTime getMinTime() + { + return baseAdapter.getMinTime(); + } + + @Override + public DateTime getMaxTime() + { + return baseAdapter.getMaxTime(); + } + + @Nullable + @Override + public Comparable getMinValue(String column) + { + if (!outputColumnName.equals(column)) { + return baseAdapter.getMinValue(column); + } + return baseAdapter.getMinValue(dimensionToUnnest); + } + + @Nullable + @Override + public Comparable getMaxValue(String column) + { + if (!outputColumnName.equals(column)) { + return baseAdapter.getMaxValue(column); + } + return baseAdapter.getMaxValue(dimensionToUnnest); + } + + @Nullable + @Override + public ColumnCapabilities getColumnCapabilities(String column) + { + if (!outputColumnName.equals(column)) { + return baseAdapter.getColumnCapabilities(column); + } + return baseAdapter.getColumnCapabilities(dimensionToUnnest); + } + + @Override + public int getNumRows() + { + return 0; + } + + @Override + public DateTime getMaxIngestedEventTime() + { + return baseAdapter.getMaxIngestedEventTime(); + } + + @Nullable + @Override + public Metadata getMetadata() + { + return baseAdapter.getMetadata(); + } + + public String getDimensionToUnnest() + { + return dimensionToUnnest; + } +} + diff --git a/processing/src/test/java/org/apache/druid/segment/ListCursor.java b/processing/src/test/java/org/apache/druid/segment/ListCursor.java new file mode 100644 index 000000000000..666bc21be5bb --- /dev/null +++ b/processing/src/test/java/org/apache/druid/segment/ListCursor.java @@ -0,0 +1,228 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.segment; + +import com.google.common.base.Predicate; +import org.apache.druid.query.BaseQuery; +import org.apache.druid.query.dimension.DimensionSpec; +import org.apache.druid.query.filter.ValueMatcher; +import org.apache.druid.query.monomorphicprocessing.RuntimeShapeInspector; +import org.apache.druid.segment.column.ColumnCapabilities; +import org.apache.druid.segment.data.IndexedInts; +import org.joda.time.DateTime; + +import javax.annotation.Nullable; +import java.util.List; + + +/** + * A Cursor that iterates over a user created list. + * This is used to test the base cursor of an UnnestCursor. + * Usages can be found in tests of {@link UnnestColumnValueSelectorCursor} in {@link UnnestColumnValueSelectorCursorTest} + * However this cannot help with {@link UnnestDimensionCursor}. + * Tests for {@link UnnestDimensionCursor} are done alongside tests for {@link UnnestStorageAdapterTest} + */ +public class ListCursor implements Cursor +{ + List baseList; + private int index; + + public ListCursor(List inputList) + { + this.baseList = inputList; + } + + @Override + public ColumnSelectorFactory getColumnSelectorFactory() + { + return new ColumnSelectorFactory() + { + @Override + public DimensionSelector makeDimensionSelector(DimensionSpec dimensionSpec) + { + return new DimensionSelector() + { + @Override + public IndexedInts getRow() + { + return null; + } + + @Override + public ValueMatcher makeValueMatcher(@Nullable String value) + { + return null; + } + + @Override + public ValueMatcher makeValueMatcher(Predicate predicate) + { + return null; + } + + @Override + public void inspectRuntimeShape(RuntimeShapeInspector inspector) + { + + } + + @Nullable + @Override + public Object getObject() + { + if (index < baseList.size()) { + return baseList.get(index); + } + return null; + } + + @Override + public Class classOfObject() + { + return null; + } + + @Override + public int getValueCardinality() + { + return 0; + } + + @Nullable + @Override + public String lookupName(int id) + { + return null; + } + + @Override + public boolean nameLookupPossibleInAdvance() + { + return false; + } + + @Nullable + @Override + public IdLookup idLookup() + { + return null; + } + }; + } + + @Override + public ColumnValueSelector makeColumnValueSelector(String columnName) + { + return new ColumnValueSelector() + { + @Override + public double getDouble() + { + return 0; + } + + @Override + public float getFloat() + { + return 0; + } + + @Override + public long getLong() + { + return 0; + } + + @Override + public void inspectRuntimeShape(RuntimeShapeInspector inspector) + { + + } + + @Override + public boolean isNull() + { + return false; + } + + @Nullable + @Override + public Object getObject() + { + if (index < baseList.size()) { + return baseList.get(index); + } + return null; + } + + @Override + public Class classOfObject() + { + return null; + } + }; + } + + @Nullable + @Override + public ColumnCapabilities getColumnCapabilities(String column) + { + return null; + } + }; + } + + @Override + public DateTime getTime() + { + return null; + } + + @Override + public void advance() + { + advanceUninterruptibly(); + BaseQuery.checkInterrupted(); + } + + @Override + public void advanceUninterruptibly() + { + index++; + } + + @Override + public boolean isDone() + { + return index > baseList.size() - 1; + } + + @Override + public boolean isDoneOrInterrupted() + { + return false; + } + + @Override + public void reset() + { + index = 0; + } +} diff --git a/processing/src/test/java/org/apache/druid/segment/UnnestColumnValueSelectorCursorTest.java b/processing/src/test/java/org/apache/druid/segment/UnnestColumnValueSelectorCursorTest.java new file mode 100644 index 000000000000..b3346e1e5623 --- /dev/null +++ b/processing/src/test/java/org/apache/druid/segment/UnnestColumnValueSelectorCursorTest.java @@ -0,0 +1,632 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.segment; + +import org.apache.druid.java.util.common.UOE; +import org.apache.druid.query.dimension.DefaultDimensionSpec; +import org.apache.druid.testing.InitializedNullHandlingTest; +import org.junit.Assert; +import org.junit.Test; + +import java.util.ArrayList; +import java.util.Arrays; +import java.util.Collections; +import java.util.LinkedHashSet; +import java.util.List; + +public class UnnestColumnValueSelectorCursorTest extends InitializedNullHandlingTest +{ + private static String OUTPUT_NAME = "unnested-column"; + private static LinkedHashSet IGNORE_SET = null; + private static LinkedHashSet IGNORE_SET1 = new LinkedHashSet<>(Arrays.asList("b", "f")); + + + @Test + public void test_list_unnest_cursors() + { + ArrayList baseList = new ArrayList<>(); + for (int i = 0; i < 2; i++) { + List newList = new ArrayList<>(); + for (int j = 0; j < 2; j++) { + newList.add(String.valueOf(i * 2 + j)); + } + baseList.add(newList); + } + ListCursor listCursor = new ListCursor(baseList); + UnnestColumnValueSelectorCursor unnestCursor = new UnnestColumnValueSelectorCursor( + listCursor, + listCursor.getColumnSelectorFactory(), + "dummy", + OUTPUT_NAME, + IGNORE_SET + ); + ColumnValueSelector unnestColumnValueSelector = unnestCursor.getColumnSelectorFactory() + .makeColumnValueSelector(OUTPUT_NAME); + int j = 0; + while (!unnestCursor.isDone()) { + Object colSelectorVal = unnestColumnValueSelector.getObject(); + Assert.assertEquals(colSelectorVal.toString(), String.valueOf(j)); + j++; + unnestCursor.advance(); + } + Assert.assertEquals(j, 4); + } + + @Test + public void test_list_unnest_cursors_user_supplied_list() + { + List inputList = Arrays.asList( + Arrays.asList("a", "b", "c"), + Arrays.asList("e", "f", "g", "h", "i"), + Collections.singletonList("j") + ); + + List expectedResults = Arrays.asList("a", "b", "c", "e", "f", "g", "h", "i", "j"); + + //Create base cursor + ListCursor listCursor = new ListCursor(inputList); + + //Create unnest cursor + UnnestColumnValueSelectorCursor unnestCursor = new UnnestColumnValueSelectorCursor( + listCursor, + listCursor.getColumnSelectorFactory(), + "dummy", + OUTPUT_NAME, + IGNORE_SET + ); + ColumnValueSelector unnestColumnValueSelector = unnestCursor.getColumnSelectorFactory() + .makeColumnValueSelector(OUTPUT_NAME); + int k = 0; + while (!unnestCursor.isDone()) { + Object valueSelectorVal = unnestColumnValueSelector.getObject(); + Assert.assertEquals(valueSelectorVal.toString(), expectedResults.get(k)); + k++; + unnestCursor.advance(); + } + Assert.assertEquals(k, 9); + } + + @Test + public void test_list_unnest_cursors_user_supplied_list_only_nulls() + { + List inputList = Arrays.asList( + Collections.singletonList(null), + Arrays.asList(null, null), + Collections.singletonList(null) + ); + + List expectedResults = Arrays.asList(null, null, null, null); + + //Create base cursor + ListCursor listCursor = new ListCursor(inputList); + + //Create unnest cursor + UnnestColumnValueSelectorCursor unnestCursor = new UnnestColumnValueSelectorCursor( + listCursor, + listCursor.getColumnSelectorFactory(), + "dummy", + OUTPUT_NAME, + IGNORE_SET + ); + ColumnValueSelector unnestColumnValueSelector = unnestCursor.getColumnSelectorFactory() + .makeColumnValueSelector(OUTPUT_NAME); + int k = 0; + while (!unnestCursor.isDone()) { + Object valueSelectorVal = unnestColumnValueSelector.getObject(); + Assert.assertNull(valueSelectorVal); + k++; + unnestCursor.advance(); + } + Assert.assertEquals(k, 4); + } + + @Test + public void test_list_unnest_cursors_user_supplied_list_mixed_with_nulls() + { + List inputList = Arrays.asList( + Arrays.asList("a", "b"), + Arrays.asList("b", "c"), + "d", + null, + null, + null + ); + + List expectedResults = Arrays.asList("a", "b", "b", "c", "d", null, null, null); + + //Create base cursor + ListCursor listCursor = new ListCursor(inputList); + + //Create unnest cursor + UnnestColumnValueSelectorCursor unnestCursor = new UnnestColumnValueSelectorCursor( + listCursor, + listCursor.getColumnSelectorFactory(), + "dummy", + OUTPUT_NAME, + IGNORE_SET + ); + ColumnValueSelector unnestColumnValueSelector = unnestCursor.getColumnSelectorFactory() + .makeColumnValueSelector(OUTPUT_NAME); + int k = 0; + while (!unnestCursor.isDone()) { + Object valueSelectorVal = unnestColumnValueSelector.getObject(); + if (valueSelectorVal == null) { + Assert.assertEquals(null, expectedResults.get(k)); + } else { + Assert.assertEquals(valueSelectorVal.toString(), expectedResults.get(k)); + } + k++; + unnestCursor.advance(); + } + Assert.assertEquals(k, 8); + } + + @Test + public void test_list_unnest_cursors_user_supplied_strings_and_no_lists() + { + List inputList = Arrays.asList("a", "b", "c", "e", "f", "g", "h", "i", "j"); + + List expectedResults = Arrays.asList("a", "b", "c", "e", "f", "g", "h", "i", "j"); + + //Create base cursor + ListCursor listCursor = new ListCursor(inputList); + + //Create unnest cursor + UnnestColumnValueSelectorCursor unnestCursor = new UnnestColumnValueSelectorCursor( + listCursor, + listCursor.getColumnSelectorFactory(), + "dummy", + OUTPUT_NAME, + IGNORE_SET + ); + ColumnValueSelector unnestColumnValueSelector = unnestCursor.getColumnSelectorFactory() + .makeColumnValueSelector(OUTPUT_NAME); + int k = 0; + while (!unnestCursor.isDone()) { + Object valueSelectorVal = unnestColumnValueSelector.getObject(); + Assert.assertEquals(valueSelectorVal.toString(), expectedResults.get(k)); + k++; + unnestCursor.advance(); + } + Assert.assertEquals(k, 9); + } + + @Test + public void test_list_unnest_cursors_user_supplied_strings_mixed_with_list() + { + List inputList = Arrays.asList("a", "b", "c", "e", "f", Arrays.asList("g", "h"), "i", "j"); + + List expectedResults = Arrays.asList("a", "b", "c", "e", "f", "g", "h", "i", "j"); + + //Create base cursor + ListCursor listCursor = new ListCursor(inputList); + + //Create unnest cursor + UnnestColumnValueSelectorCursor unnestCursor = new UnnestColumnValueSelectorCursor( + listCursor, + listCursor.getColumnSelectorFactory(), + "dummy", + OUTPUT_NAME, + IGNORE_SET + ); + ColumnValueSelector unnestColumnValueSelector = unnestCursor.getColumnSelectorFactory() + .makeColumnValueSelector(OUTPUT_NAME); + int k = 0; + while (!unnestCursor.isDone()) { + Object valueSelectorVal = unnestColumnValueSelector.getObject(); + Assert.assertEquals(valueSelectorVal.toString(), expectedResults.get(k)); + k++; + unnestCursor.advance(); + } + Assert.assertEquals(k, 9); + } + + @Test + public void test_list_unnest_cursors_user_supplied_lists_three_levels() + { + List inputList = Arrays.asList( + Arrays.asList("a", "b", "c"), + Arrays.asList("e", "f", "g", "h", "i"), + Arrays.asList("j", Arrays.asList("a", "b")) + ); + + List expectedResults = Arrays.asList("a", "b", "c", "e", "f", "g", "h", "i", "j", Arrays.asList("a", "b")); + + //Create base cursor + ListCursor listCursor = new ListCursor(inputList); + + //Create unnest cursor + UnnestColumnValueSelectorCursor unnestCursor = new UnnestColumnValueSelectorCursor( + listCursor, + listCursor.getColumnSelectorFactory(), + "dummy", + OUTPUT_NAME, + IGNORE_SET + ); + ColumnValueSelector unnestColumnValueSelector = unnestCursor.getColumnSelectorFactory() + .makeColumnValueSelector(OUTPUT_NAME); + int k = 0; + while (!unnestCursor.isDone()) { + Object valueSelectorVal = unnestColumnValueSelector.getObject(); + Assert.assertEquals(valueSelectorVal.toString(), expectedResults.get(k).toString()); + k++; + unnestCursor.advance(); + } + Assert.assertEquals(k, 10); + } + + @Test + public void test_list_unnest_of_unnest_cursors_user_supplied_list_three_levels() + { + List inputList = Arrays.asList( + Arrays.asList("a", "b", "c"), + Arrays.asList("e", "f", "g", "h", "i"), + Arrays.asList("j", Arrays.asList("a", "b")) + ); + + List expectedResults = Arrays.asList("a", "b", "c", "e", "f", "g", "h", "i", "j", "a", "b"); + + //Create base cursor + ListCursor listCursor = new ListCursor(inputList); + + //Create unnest cursor + UnnestColumnValueSelectorCursor childCursor = new UnnestColumnValueSelectorCursor( + listCursor, + listCursor.getColumnSelectorFactory(), + "dummy", + OUTPUT_NAME, + IGNORE_SET + ); + UnnestColumnValueSelectorCursor parentCursor = new UnnestColumnValueSelectorCursor( + childCursor, + childCursor.getColumnSelectorFactory(), + OUTPUT_NAME, + "tmp-out", + IGNORE_SET + ); + ColumnValueSelector unnestColumnValueSelector = parentCursor.getColumnSelectorFactory() + .makeColumnValueSelector("tmp-out"); + int k = 0; + while (!parentCursor.isDone()) { + Object valueSelectorVal = unnestColumnValueSelector.getObject(); + Assert.assertEquals(valueSelectorVal.toString(), expectedResults.get(k).toString()); + k++; + parentCursor.advance(); + } + Assert.assertEquals(k, 11); + } + + @Test + public void test_list_unnest_cursors_user_supplied_list_with_nulls() + { + List inputList = Arrays.asList( + Arrays.asList("a", "b", "c"), + Arrays.asList("e", "f", "g", "h", "i", null), + Collections.singletonList("j") + ); + + List expectedResults = Arrays.asList("a", "b", "c", "e", "f", "g", "h", "i", null, "j"); + + + //Create base cursor + ListCursor listCursor = new ListCursor(inputList); + + //Create unnest cursor + UnnestColumnValueSelectorCursor unnestCursor = new UnnestColumnValueSelectorCursor( + listCursor, + listCursor.getColumnSelectorFactory(), + "dummy", + OUTPUT_NAME, + IGNORE_SET + ); + ColumnValueSelector unnestColumnValueSelector = unnestCursor.getColumnSelectorFactory() + .makeColumnValueSelector(OUTPUT_NAME); + int k = 0; + while (!unnestCursor.isDone()) { + Object valueSelectorVal = unnestColumnValueSelector.getObject(); + if (valueSelectorVal == null) { + Assert.assertEquals(null, expectedResults.get(k)); + } else { + Assert.assertEquals(valueSelectorVal.toString(), expectedResults.get(k)); + } + k++; + unnestCursor.advance(); + } + Assert.assertEquals(k, expectedResults.size()); + } + + @Test + public void test_list_unnest_cursors_user_supplied_list_with_dups() + { + List inputList = Arrays.asList( + Arrays.asList("a", "a", "a"), + Arrays.asList("e", "f", null, "h", "i", null), + Collections.singletonList("j") + ); + + List expectedResults = Arrays.asList("a", "a", "a", "e", "f", null, "h", "i", null, "j"); + + //Create base cursor + ListCursor listCursor = new ListCursor(inputList); + + //Create unnest cursor + UnnestColumnValueSelectorCursor unnestCursor = new UnnestColumnValueSelectorCursor( + listCursor, + listCursor.getColumnSelectorFactory(), + "dummy", + OUTPUT_NAME, + IGNORE_SET + ); + ColumnValueSelector unnestColumnValueSelector = unnestCursor.getColumnSelectorFactory() + .makeColumnValueSelector(OUTPUT_NAME); + int k = 0; + while (!unnestCursor.isDone()) { + Object valueSelectorVal = unnestColumnValueSelector.getObject(); + if (valueSelectorVal == null) { + Assert.assertEquals(null, expectedResults.get(k)); + } else { + Assert.assertEquals(valueSelectorVal.toString(), expectedResults.get(k)); + } + k++; + unnestCursor.advance(); + } + Assert.assertEquals(k, 10); + } + + @Test + public void test_list_unnest_cursors_user_supplied_list_with_ignore_set() + { + List inputList = Arrays.asList( + Arrays.asList("a", "b", "c"), + Arrays.asList("e", "f", "g", "h", "i"), + Collections.singletonList("j") + ); + + List expectedResults = Arrays.asList("b", "f"); + + //Create base cursor + ListCursor listCursor = new ListCursor(inputList); + + //Create unnest cursor + UnnestColumnValueSelectorCursor unnestCursor = new UnnestColumnValueSelectorCursor( + listCursor, + listCursor.getColumnSelectorFactory(), + "dummy", + OUTPUT_NAME, + IGNORE_SET1 + ); + ColumnValueSelector unnestColumnValueSelector = unnestCursor.getColumnSelectorFactory() + .makeColumnValueSelector(OUTPUT_NAME); + int k = 0; + while (!unnestCursor.isDone()) { + Object valueSelectorVal = unnestColumnValueSelector.getObject(); + if (valueSelectorVal == null) { + Assert.assertEquals(null, expectedResults.get(k)); + } else { + Assert.assertEquals(valueSelectorVal.toString(), expectedResults.get(k)); + } + k++; + unnestCursor.advance(); + } + Assert.assertEquals(k, 2); + } + + @Test + public void test_list_unnest_cursors_user_supplied_list_double() + { + List inputList = Arrays.asList( + Arrays.asList(1, 2, 3), + Arrays.asList(4, 5, 6, 7, 8), + Collections.singletonList(9) + ); + + List expectedResults = Arrays.asList(1d, 2d, 3d, 4d, 5d, 6d, 7d, 8d, 9d); + + //Create base cursor + ListCursor listCursor = new ListCursor(inputList); + + //Create unnest cursor + UnnestColumnValueSelectorCursor unnestCursor = new UnnestColumnValueSelectorCursor( + listCursor, + listCursor.getColumnSelectorFactory(), + "dummy", + OUTPUT_NAME, + IGNORE_SET + ); + ColumnValueSelector unnestColumnValueSelector = unnestCursor.getColumnSelectorFactory() + .makeColumnValueSelector(OUTPUT_NAME); + int k = 0; + while (!unnestCursor.isDone()) { + Double valueSelectorVal = unnestColumnValueSelector.getDouble(); + Assert.assertEquals(valueSelectorVal, expectedResults.get(k)); + k++; + unnestCursor.advance(); + } + Assert.assertEquals(k, 9); + } + + @Test + public void test_list_unnest_cursors_user_supplied_list_float() + { + List inputList = Arrays.asList( + Arrays.asList(1, 2, 3), + Arrays.asList(4, 5, 6, 7, 8), + Collections.singletonList(9) + ); + + List expectedResults = Arrays.asList(1f, 2f, 3f, 4f, 5f, 6f, 7f, 8f, 9f); + + //Create base cursor + ListCursor listCursor = new ListCursor(inputList); + + //Create unnest cursor + UnnestColumnValueSelectorCursor unnestCursor = new UnnestColumnValueSelectorCursor( + listCursor, + listCursor.getColumnSelectorFactory(), + "dummy", + OUTPUT_NAME, + IGNORE_SET + ); + ColumnValueSelector unnestColumnValueSelector = unnestCursor.getColumnSelectorFactory() + .makeColumnValueSelector(OUTPUT_NAME); + int k = 0; + while (!unnestCursor.isDone()) { + Float valueSelectorVal = unnestColumnValueSelector.getFloat(); + Assert.assertEquals(valueSelectorVal, expectedResults.get(k)); + k++; + unnestCursor.advance(); + } + Assert.assertEquals(k, 9); + } + + @Test + public void test_list_unnest_cursors_user_supplied_list_long() + { + List inputList = Arrays.asList( + Arrays.asList(1, 2, 3), + Arrays.asList(4, 5, 6, 7, 8), + Collections.singletonList(9) + ); + + List expectedResults = Arrays.asList(1L, 2L, 3L, 4L, 5L, 6L, 7L, 8L, 9L); + + //Create base cursor + ListCursor listCursor = new ListCursor(inputList); + + //Create unnest cursor + UnnestColumnValueSelectorCursor unnestCursor = new UnnestColumnValueSelectorCursor( + listCursor, + listCursor.getColumnSelectorFactory(), + "dummy", + OUTPUT_NAME, + IGNORE_SET + ); + ColumnValueSelector unnestColumnValueSelector = unnestCursor.getColumnSelectorFactory() + .makeColumnValueSelector(OUTPUT_NAME); + + int k = 0; + while (!unnestCursor.isDone()) { + Object obj = unnestColumnValueSelector.getObject(); + Assert.assertNotNull(obj); + Long valueSelectorVal = unnestColumnValueSelector.getLong(); + Assert.assertEquals(valueSelectorVal, expectedResults.get(k)); + k++; + unnestCursor.advance(); + } + Assert.assertEquals(k, 9); + } + + @Test + public void test_list_unnest_cursors_user_supplied_list_three_level_arrays_and_methods() + { + List inputList = Arrays.asList( + Arrays.asList("a", "b", "c"), + Arrays.asList("e", "f", "g", "h", "i"), + Arrays.asList("j", Arrays.asList("a", "b")) + ); + + List expectedResults = Arrays.asList("a", "b", "c", "e", "f", "g", "h", "i", "j", Arrays.asList("a", "b")); + + //Create base cursor + ListCursor listCursor = new ListCursor(inputList); + + //Create unnest cursor + UnnestColumnValueSelectorCursor unnestCursor = new UnnestColumnValueSelectorCursor( + listCursor, + listCursor.getColumnSelectorFactory(), + "dummy", + OUTPUT_NAME, + IGNORE_SET + ); + ColumnValueSelector unnestColumnValueSelector = unnestCursor.getColumnSelectorFactory() + .makeColumnValueSelector(OUTPUT_NAME); + + int k = 0; + while (!unnestCursor.isDone()) { + Object valueSelectorVal = unnestColumnValueSelector.getObject(); + Assert.assertEquals(valueSelectorVal.toString(), expectedResults.get(k).toString()); + k++; + unnestCursor.advance(); + } + Assert.assertEquals(k, 10); + unnestCursor.reset(); + Assert.assertFalse(unnestCursor.isDoneOrInterrupted()); + } + + @Test(expected = UOE.class) + public void test_list_unnest_cursors_dimSelector() + { + List inputList = Arrays.asList( + Arrays.asList("a", "b", "c"), + Arrays.asList("e", "f", "g", "h", "i"), + Collections.singletonList("j") + ); + + List expectedResults = Arrays.asList("a", "b", "c", "e", "f", "g", "h", "i", "j"); + + //Create base cursor + ListCursor listCursor = new ListCursor(inputList); + + //Create unnest cursor + UnnestColumnValueSelectorCursor unnestCursor = new UnnestColumnValueSelectorCursor( + listCursor, + listCursor.getColumnSelectorFactory(), + "dummy", + OUTPUT_NAME, + IGNORE_SET + ); + unnestCursor.getColumnSelectorFactory().makeDimensionSelector(DefaultDimensionSpec.of(OUTPUT_NAME)); + } + + @Test + public void test_list_unnest_cursors_user_supplied_list_of_integers() + { + List inputList = Arrays.asList( + Arrays.asList(1, 2, 3), + Arrays.asList(4, 5, 6, 7, 8), + Collections.singletonList(9) + ); + + List expectedResults = Arrays.asList(1, 2, 3, 4, 5, 6, 7, 8, 9); + + //Create base cursor + ListCursor listCursor = new ListCursor(inputList); + + //Create unnest cursor + UnnestColumnValueSelectorCursor unnestCursor = new UnnestColumnValueSelectorCursor( + listCursor, + listCursor.getColumnSelectorFactory(), + "dummy", + OUTPUT_NAME, + IGNORE_SET + ); + ColumnValueSelector unnestColumnValueSelector = unnestCursor.getColumnSelectorFactory() + .makeColumnValueSelector(OUTPUT_NAME); + int k = 0; + while (!unnestCursor.isDone()) { + Object valueSelectorVal = unnestColumnValueSelector.getObject(); + Assert.assertEquals(valueSelectorVal.toString(), expectedResults.get(k).toString()); + k++; + unnestCursor.advance(); + } + Assert.assertEquals(k, 9); + } +} + diff --git a/processing/src/test/java/org/apache/druid/segment/UnnestStorageAdapterTest.java b/processing/src/test/java/org/apache/druid/segment/UnnestStorageAdapterTest.java new file mode 100644 index 000000000000..35d42b82d4b3 --- /dev/null +++ b/processing/src/test/java/org/apache/druid/segment/UnnestStorageAdapterTest.java @@ -0,0 +1,399 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.segment; + +import com.google.common.collect.ImmutableList; +import org.apache.druid.java.util.common.DateTimes; +import org.apache.druid.java.util.common.granularity.Granularities; +import org.apache.druid.java.util.common.guava.Sequence; +import org.apache.druid.java.util.common.io.Closer; +import org.apache.druid.query.dimension.DefaultDimensionSpec; +import org.apache.druid.segment.column.ColumnCapabilities; +import org.apache.druid.segment.column.ValueType; +import org.apache.druid.segment.generator.GeneratorBasicSchemas; +import org.apache.druid.segment.generator.GeneratorSchemaInfo; +import org.apache.druid.segment.generator.SegmentGenerator; +import org.apache.druid.segment.incremental.IncrementalIndex; +import org.apache.druid.segment.incremental.IncrementalIndexStorageAdapter; +import org.apache.druid.testing.InitializedNullHandlingTest; +import org.apache.druid.timeline.DataSegment; +import org.apache.druid.timeline.partition.LinearShardSpec; +import org.apache.druid.utils.CloseableUtils; +import org.junit.AfterClass; +import org.junit.Assert; +import org.junit.BeforeClass; +import org.junit.Test; + +import java.util.Arrays; +import java.util.LinkedHashSet; +import java.util.List; + +public class UnnestStorageAdapterTest extends InitializedNullHandlingTest +{ + private static Closer CLOSER; + private static IncrementalIndex INCREMENTAL_INDEX; + private static IncrementalIndexStorageAdapter INCREMENTAL_INDEX_STORAGE_ADAPTER; + private static UnnestStorageAdapter UNNEST_STORAGE_ADAPTER; + private static UnnestStorageAdapter UNNEST_STORAGE_ADAPTER1; + private static UnnestStorageAdapter UNNEST_STORAGE_ADAPTER2; + private static UnnestStorageAdapter UNNEST_STORAGE_ADAPTER3; + private static List ADAPTERS; + private static String COLUMNNAME = "multi-string1"; + private static String OUTPUT_COLUMN_NAME = "unnested-multi-string1"; + private static String OUTPUT_COLUMN_NAME1 = "unnested-multi-string1-again"; + private static LinkedHashSet IGNORE_SET = new LinkedHashSet<>(Arrays.asList("1", "3", "5")); + + @BeforeClass + public static void setup() + { + CLOSER = Closer.create(); + final GeneratorSchemaInfo schemaInfo = GeneratorBasicSchemas.SCHEMA_MAP.get("expression-testbench"); + + final DataSegment dataSegment = DataSegment.builder() + .dataSource("foo") + .interval(schemaInfo.getDataInterval()) + .version("1") + .shardSpec(new LinearShardSpec(0)) + .size(0) + .build(); + final SegmentGenerator segmentGenerator = CLOSER.register(new SegmentGenerator()); + + final int numRows = 2; + INCREMENTAL_INDEX = CLOSER.register( + segmentGenerator.generateIncrementalIndex(dataSegment, schemaInfo, Granularities.HOUR, numRows) + ); + INCREMENTAL_INDEX_STORAGE_ADAPTER = new IncrementalIndexStorageAdapter(INCREMENTAL_INDEX); + UNNEST_STORAGE_ADAPTER = new UnnestStorageAdapter( + INCREMENTAL_INDEX_STORAGE_ADAPTER, + COLUMNNAME, + OUTPUT_COLUMN_NAME, + null + ); + UNNEST_STORAGE_ADAPTER1 = new UnnestStorageAdapter( + INCREMENTAL_INDEX_STORAGE_ADAPTER, + COLUMNNAME, + OUTPUT_COLUMN_NAME, + IGNORE_SET + ); + UNNEST_STORAGE_ADAPTER2 = new UnnestStorageAdapter( + UNNEST_STORAGE_ADAPTER, + COLUMNNAME, + OUTPUT_COLUMN_NAME1, + null + ); + UNNEST_STORAGE_ADAPTER3 = new UnnestStorageAdapter( + UNNEST_STORAGE_ADAPTER1, + COLUMNNAME, + OUTPUT_COLUMN_NAME1, + IGNORE_SET + ); + ADAPTERS = ImmutableList.of( + UNNEST_STORAGE_ADAPTER, + UNNEST_STORAGE_ADAPTER1, + UNNEST_STORAGE_ADAPTER2, + UNNEST_STORAGE_ADAPTER3 + ); + } + + @AfterClass + public static void teardown() + { + CloseableUtils.closeAndSuppressExceptions(CLOSER, throwable -> { + }); + } + + @Test + public void test_group_of_unnest_adapters_methods() + { + String colName = "multi-string1"; + for (StorageAdapter adapter : ADAPTERS) { + Assert.assertEquals( + DateTimes.of("2000-01-01T23:00:00.000Z"), + adapter.getMaxTime() + ); + Assert.assertEquals( + DateTimes.of("2000-01-01T12:00:00.000Z"), + adapter.getMinTime() + ); + adapter.getColumnCapabilities(colName); + Assert.assertEquals(adapter.getNumRows(), 0); + Assert.assertNotNull(adapter.getMetadata()); + Assert.assertEquals( + DateTimes.of("2000-01-01T23:59:59.999Z"), + adapter.getMaxIngestedEventTime() + ); + Assert.assertEquals( + adapter.getColumnCapabilities(colName).toColumnType(), + INCREMENTAL_INDEX_STORAGE_ADAPTER.getColumnCapabilities(colName).toColumnType() + ); + Assert.assertEquals(((UnnestStorageAdapter) adapter).getDimensionToUnnest(), colName); + } + } + + @Test + public void test_group_of_unnest_adapters_column_capabilities() + { + String colName = "multi-string1"; + List columnsInTable = Arrays.asList( + "string1", + "long1", + "double1", + "float1", + "multi-string1", + OUTPUT_COLUMN_NAME + ); + List valueTypes = Arrays.asList( + ValueType.STRING, + ValueType.LONG, + ValueType.DOUBLE, + ValueType.FLOAT, + ValueType.STRING, + ValueType.STRING + ); + UnnestStorageAdapter adapter = UNNEST_STORAGE_ADAPTER; + + for (int i = 0; i < columnsInTable.size(); i++) { + ColumnCapabilities capabilities = adapter.getColumnCapabilities(columnsInTable.get(i)); + Assert.assertEquals(capabilities.getType(), valueTypes.get(i)); + } + Assert.assertEquals(adapter.getDimensionToUnnest(), colName); + + } + + @Test + public void test_unnest_adapters_basic() + { + + Sequence cursorSequence = UNNEST_STORAGE_ADAPTER.makeCursors( + null, + UNNEST_STORAGE_ADAPTER.getInterval(), + VirtualColumns.EMPTY, + Granularities.ALL, + false, + null + ); + + cursorSequence.accumulate(null, (accumulated, cursor) -> { + ColumnSelectorFactory factory = cursor.getColumnSelectorFactory(); + + DimensionSelector dimSelector = factory.makeDimensionSelector(DefaultDimensionSpec.of(OUTPUT_COLUMN_NAME)); + int count = 0; + while (!cursor.isDone()) { + Object dimSelectorVal = dimSelector.getObject(); + if (dimSelectorVal == null) { + Assert.assertNull(dimSelectorVal); + } + cursor.advance(); + count++; + } + /* + each row has 8 entries. + unnest 2 rows -> 16 rows after unnest + */ + Assert.assertEquals(count, 16); + return null; + }); + + } + + @Test + public void test_two_levels_of_unnest_adapters() + { + Sequence cursorSequence = UNNEST_STORAGE_ADAPTER2.makeCursors( + null, + UNNEST_STORAGE_ADAPTER2.getInterval(), + VirtualColumns.EMPTY, + Granularities.ALL, + false, + null + ); + + + cursorSequence.accumulate(null, (accumulated, cursor) -> { + ColumnSelectorFactory factory = cursor.getColumnSelectorFactory(); + + DimensionSelector dimSelector = factory.makeDimensionSelector(DefaultDimensionSpec.of(OUTPUT_COLUMN_NAME1)); + ColumnValueSelector valueSelector = factory.makeColumnValueSelector(OUTPUT_COLUMN_NAME1); + + int count = 0; + while (!cursor.isDone()) { + Object dimSelectorVal = dimSelector.getObject(); + Object valueSelectorVal = valueSelector.getObject(); + if (dimSelectorVal == null) { + Assert.assertNull(dimSelectorVal); + } else if (valueSelectorVal == null) { + Assert.assertNull(valueSelectorVal); + } + cursor.advance(); + count++; + } + /* + each row has 8 entries. + unnest 2 rows -> 16 entries also the value cardinality + unnest of unnest -> 16*8 = 128 rows + */ + Assert.assertEquals(count, 128); + Assert.assertEquals(dimSelector.getValueCardinality(), 16); + return null; + }); + } + + @Test + public void test_unnest_adapters_with_allowList() + { + final String columnName = "multi-string1"; + + Sequence cursorSequence = UNNEST_STORAGE_ADAPTER1.makeCursors( + null, + UNNEST_STORAGE_ADAPTER1.getInterval(), + VirtualColumns.EMPTY, + Granularities.ALL, + false, + null + ); + + cursorSequence.accumulate(null, (accumulated, cursor) -> { + ColumnSelectorFactory factory = cursor.getColumnSelectorFactory(); + + DimensionSelector dimSelector = factory.makeDimensionSelector(DefaultDimensionSpec.of(OUTPUT_COLUMN_NAME)); + ColumnValueSelector valueSelector = factory.makeColumnValueSelector(OUTPUT_COLUMN_NAME); + + int count = 0; + while (!cursor.isDone()) { + Object dimSelectorVal = dimSelector.getObject(); + Object valueSelectorVal = valueSelector.getObject(); + if (dimSelectorVal == null) { + Assert.assertNull(dimSelectorVal); + } else if (valueSelectorVal == null) { + Assert.assertNull(valueSelectorVal); + } + cursor.advance(); + count++; + } + /* + each row has 8 distinct entries. + allowlist has 3 entries also the value cardinality + unnest will have 3 distinct entries + */ + Assert.assertEquals(count, 3); + Assert.assertEquals(dimSelector.getValueCardinality(), 3); + return null; + }); + } + + @Test + public void test_two_levels_of_unnest_adapters_with_allowList() + { + final String columnName = "multi-string1"; + + Sequence cursorSequence = UNNEST_STORAGE_ADAPTER3.makeCursors( + null, + UNNEST_STORAGE_ADAPTER3.getInterval(), + VirtualColumns.EMPTY, + Granularities.ALL, + false, + null + ); + UnnestStorageAdapter adapter = UNNEST_STORAGE_ADAPTER3; + Assert.assertEquals(adapter.getDimensionToUnnest(), columnName); + Assert.assertEquals( + adapter.getColumnCapabilities(OUTPUT_COLUMN_NAME).isDictionaryEncoded(), + ColumnCapabilities.Capable.TRUE + ); + Assert.assertEquals(adapter.getMaxValue(columnName), adapter.getMaxValue(OUTPUT_COLUMN_NAME)); + Assert.assertEquals(adapter.getMinValue(columnName), adapter.getMinValue(OUTPUT_COLUMN_NAME)); + + cursorSequence.accumulate(null, (accumulated, cursor) -> { + ColumnSelectorFactory factory = cursor.getColumnSelectorFactory(); + + DimensionSelector dimSelector = factory.makeDimensionSelector(DefaultDimensionSpec.of(OUTPUT_COLUMN_NAME1)); + ColumnValueSelector valueSelector = factory.makeColumnValueSelector(OUTPUT_COLUMN_NAME1); + + int count = 0; + while (!cursor.isDone()) { + Object dimSelectorVal = dimSelector.getObject(); + Object valueSelectorVal = valueSelector.getObject(); + if (dimSelectorVal == null) { + Assert.assertNull(dimSelectorVal); + } else if (valueSelectorVal == null) { + Assert.assertNull(valueSelectorVal); + } + cursor.advance(); + count++; + } + /* + each row has 8 distinct entries. + allowlist has 3 entries also the value cardinality + unnest will have 3 distinct entries + unnest of that unnest will have 3*3 = 9 entries + */ + Assert.assertEquals(count, 9); + Assert.assertEquals(dimSelector.getValueCardinality(), 3); + return null; + }); + } + + @Test + public void test_unnest_adapters_methods_with_allowList() + { + final String columnName = "multi-string1"; + + Sequence cursorSequence = UNNEST_STORAGE_ADAPTER1.makeCursors( + null, + UNNEST_STORAGE_ADAPTER1.getInterval(), + VirtualColumns.EMPTY, + Granularities.ALL, + false, + null + ); + UnnestStorageAdapter adapter = UNNEST_STORAGE_ADAPTER1; + Assert.assertEquals(adapter.getDimensionToUnnest(), columnName); + Assert.assertEquals( + adapter.getColumnCapabilities(OUTPUT_COLUMN_NAME).isDictionaryEncoded(), + ColumnCapabilities.Capable.TRUE + ); + Assert.assertEquals(adapter.getMaxValue(columnName), adapter.getMaxValue(OUTPUT_COLUMN_NAME)); + Assert.assertEquals(adapter.getMinValue(columnName), adapter.getMinValue(OUTPUT_COLUMN_NAME)); + + cursorSequence.accumulate(null, (accumulated, cursor) -> { + ColumnSelectorFactory factory = cursor.getColumnSelectorFactory(); + + DimensionSelector dimSelector = factory.makeDimensionSelector(DefaultDimensionSpec.of(OUTPUT_COLUMN_NAME)); + IdLookup idlookUp = dimSelector.idLookup(); + Assert.assertFalse(dimSelector.isNull()); + int[] indices = new int[]{1, 3, 5}; + int count = 0; + while (!cursor.isDone()) { + Object dimSelectorVal = dimSelector.getObject(); + Assert.assertEquals(idlookUp.lookupId((String) dimSelectorVal), indices[count]); + // after unnest first entry in get row should equal the object + // and the row size will always be 1 + Assert.assertEquals(dimSelector.getRow().get(0), indices[count]); + Assert.assertEquals(dimSelector.getRow().size(), 1); + Assert.assertNotNull(dimSelector.makeValueMatcher(OUTPUT_COLUMN_NAME)); + cursor.advance(); + count++; + } + Assert.assertEquals(dimSelector.getValueCardinality(), 3); + Assert.assertEquals(count, 3); + return null; + }); + } +} From 45a8fa280c1f34f0401cfa43e1d38240d7d41f91 Mon Sep 17 00:00:00 2001 From: Kashif Faraz Date: Mon, 5 Dec 2022 14:00:07 +0530 Subject: [PATCH 110/141] Add SegmentAllocationQueue to batch SegmentAllocateActions (#13369) In a cluster with a large number of streaming tasks (~1000), SegmentAllocateActions on the overlord can often take very long intervals of time to finish thus causing spikes in the `task/action/run/time`. This may result in lag building up while a task waits for a segment to get allocated. The root causes are: - large number of metadata calls made to the segments and pending segments tables - `giant` lock held in `TaskLockbox.tryLock()` to acquire task locks and allocate segments Since the contention typically arises when several tasks of the same datasource try to allocate segments for the same interval/granularity, the allocation run times can be improved by batching the requests together. Changes - Add flags - `druid.indexer.tasklock.batchSegmentAllocation` (default `false`) - `druid.indexer.tasklock.batchAllocationMaxWaitTime` (in millis) (default `1000`) - Add methods `canPerformAsync` and `performAsync` to `TaskAction` - Submit each allocate action to a `SegmentAllocationQueue`, and add to correct batch - Process batch after `batchAllocationMaxWaitTime` - Acquire `giant` lock just once per batch in `TaskLockbox` - Reduce metadata calls by batching statements together and updating query filters - Except for batching, retain the whole behaviour (order of steps, retries, etc.) - Respond to leadership changes and fail items in queue when not leader - Emit batch and request level metrics --- .../druid/java/util/common/Intervals.java | 29 + .../druid/java/util/common/IntervalsTest.java | 81 ++ .../druid/indexing/common/TaskToolbox.java | 10 +- .../common/actions/LocalTaskActionClient.java | 20 +- .../common/actions/SegmentAllocateAction.java | 18 + .../actions/SegmentAllocateRequest.java | 77 ++ .../common/actions/SegmentAllocateResult.java | 49 ++ .../actions/SegmentAllocationQueue.java | 716 ++++++++++++++++++ .../indexing/common/actions/TaskAction.java | 12 + .../common/actions/TaskActionToolbox.java | 32 + .../druid/indexing/overlord/TaskLockbox.java | 324 +++++++- .../druid/indexing/overlord/TaskMaster.java | 20 +- .../overlord/config/TaskLockConfig.java | 16 + .../actions/SegmentAllocateActionBuilder.java | 138 ++++ .../actions/SegmentAllocateActionTest.java | 102 ++- .../actions/SegmentAllocationQueueTest.java | 349 +++++++++ .../common/actions/TaskActionTestKit.java | 28 +- .../indexing/overlord/http/OverlordTest.java | 4 +- ...TestIndexerMetadataStorageCoordinator.java | 14 + .../IndexerMetadataStorageCoordinator.java | 22 + .../overlord/SegmentCreateRequest.java | 78 ++ .../IndexerSQLMetadataStorageCoordinator.java | 594 +++++++++++++-- .../overlord/SegmentCreateRequestTest.java | 46 ++ ...exerSQLMetadataStorageCoordinatorTest.java | 269 +++---- 24 files changed, 2752 insertions(+), 296 deletions(-) create mode 100644 core/src/test/java/org/apache/druid/java/util/common/IntervalsTest.java create mode 100644 indexing-service/src/main/java/org/apache/druid/indexing/common/actions/SegmentAllocateRequest.java create mode 100644 indexing-service/src/main/java/org/apache/druid/indexing/common/actions/SegmentAllocateResult.java create mode 100644 indexing-service/src/main/java/org/apache/druid/indexing/common/actions/SegmentAllocationQueue.java create mode 100644 indexing-service/src/test/java/org/apache/druid/indexing/common/actions/SegmentAllocateActionBuilder.java create mode 100644 indexing-service/src/test/java/org/apache/druid/indexing/common/actions/SegmentAllocationQueueTest.java create mode 100644 server/src/main/java/org/apache/druid/indexing/overlord/SegmentCreateRequest.java create mode 100644 server/src/test/java/org/apache/druid/indexing/overlord/SegmentCreateRequestTest.java diff --git a/core/src/main/java/org/apache/druid/java/util/common/Intervals.java b/core/src/main/java/org/apache/druid/java/util/common/Intervals.java index b7a1f37cf1c3..96f858fd4be2 100644 --- a/core/src/main/java/org/apache/druid/java/util/common/Intervals.java +++ b/core/src/main/java/org/apache/druid/java/util/common/Intervals.java @@ -20,10 +20,13 @@ package org.apache.druid.java.util.common; import com.google.common.collect.ImmutableList; +import org.apache.druid.java.util.common.guava.Comparators; import org.joda.time.DateTime; import org.joda.time.Interval; import org.joda.time.chrono.ISOChronology; +import javax.annotation.Nullable; + public final class Intervals { public static final Interval ETERNITY = utc(JodaUtils.MIN_INSTANT, JodaUtils.MAX_INSTANT); @@ -68,6 +71,32 @@ public static boolean isEternity(final Interval interval) return ETERNITY.equals(interval); } + /** + * Finds an interval from the given set of sortedIntervals which overlaps with + * the searchInterval. If multiple candidate intervals overlap with the + * searchInterval, the "smallest" interval based on the + * {@link Comparators#intervalsByStartThenEnd()} is returned. + * + * @param searchInterval Interval which should overlap with the result + * @param sortedIntervals Candidate overlapping intervals, sorted in ascending + * order, using {@link Comparators#intervalsByStartThenEnd()}. + * @return The first overlapping interval, if one exists, otherwise null. + */ + @Nullable + public static Interval findOverlappingInterval(Interval searchInterval, Interval[] sortedIntervals) + { + for (Interval interval : sortedIntervals) { + if (interval.overlaps(searchInterval)) { + return interval; + } else if (interval.getStart().isAfter(searchInterval.getEnd())) { + // Intervals after this cannot have an overlap + return null; + } + } + + return null; + } + private Intervals() { } diff --git a/core/src/test/java/org/apache/druid/java/util/common/IntervalsTest.java b/core/src/test/java/org/apache/druid/java/util/common/IntervalsTest.java new file mode 100644 index 000000000000..59eac8d5a991 --- /dev/null +++ b/core/src/test/java/org/apache/druid/java/util/common/IntervalsTest.java @@ -0,0 +1,81 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.java.util.common; + +import org.apache.druid.java.util.common.guava.Comparators; +import org.joda.time.Interval; +import org.junit.Assert; +import org.junit.Test; + +import java.util.Arrays; + +public class IntervalsTest +{ + + @Test + public void testFindOverlappingInterval() + { + final Interval[] sortedIntervals = new Interval[]{ + Intervals.of("2019/2020"), + Intervals.of("2021/2022"), + Intervals.of("2021-04-01/2021-05-01"), + Intervals.of("2022/2023") + }; + Arrays.sort(sortedIntervals, Comparators.intervalsByStartThenEnd()); + + // Search interval outside the bounds of the sorted intervals + Assert.assertNull( + Intervals.findOverlappingInterval(Intervals.of("2018/2019"), sortedIntervals) + ); + Assert.assertNull( + Intervals.findOverlappingInterval(Intervals.of("2023/2024"), sortedIntervals) + ); + + // Search interval within bounds, overlap exists + // Fully overlapping interval + Assert.assertEquals( + Intervals.of("2021/2022"), + Intervals.findOverlappingInterval(Intervals.of("2021/2022"), sortedIntervals) + ); + + // Partially overlapping interval + Assert.assertEquals( + Intervals.of("2022/2023"), + Intervals.findOverlappingInterval(Intervals.of("2022-01-01/2022-01-02"), sortedIntervals) + ); + + Assert.assertEquals( + Intervals.of("2021/2022"), + Intervals.findOverlappingInterval(Intervals.of("2021-06-01/2021-07-01"), sortedIntervals) + ); + + // Overlap with multiple intervals, "smallest" one is returned + Assert.assertEquals( + Intervals.of("2021/2022"), + Intervals.findOverlappingInterval(Intervals.of("2021-03-01/2021-04-01"), sortedIntervals) + ); + + // Search interval within bounds, no overlap + Assert.assertNull( + Intervals.findOverlappingInterval(Intervals.of("2020-01-02/2020-03-03"), sortedIntervals) + ); + } + +} diff --git a/indexing-service/src/main/java/org/apache/druid/indexing/common/TaskToolbox.java b/indexing-service/src/main/java/org/apache/druid/indexing/common/TaskToolbox.java index 6b2fad9610b3..897db9750354 100644 --- a/indexing-service/src/main/java/org/apache/druid/indexing/common/TaskToolbox.java +++ b/indexing-service/src/main/java/org/apache/druid/indexing/common/TaskToolbox.java @@ -20,7 +20,6 @@ package org.apache.druid.indexing.common; import com.fasterxml.jackson.databind.ObjectMapper; -import com.google.common.base.Function; import com.google.common.base.Preconditions; import com.google.common.collect.ImmutableSet; import com.google.common.collect.Multimap; @@ -329,14 +328,7 @@ public void publishSegments(Iterable segments) throws IOException // Request segment pushes for each set final Multimap segmentMultimap = Multimaps.index( segments, - new Function() - { - @Override - public Interval apply(DataSegment segment) - { - return segment.getInterval(); - } - } + DataSegment::getInterval ); for (final Collection segmentCollection : segmentMultimap.asMap().values()) { getTaskActionClient().submit(new SegmentInsertAction(ImmutableSet.copyOf(segmentCollection))); diff --git a/indexing-service/src/main/java/org/apache/druid/indexing/common/actions/LocalTaskActionClient.java b/indexing-service/src/main/java/org/apache/druid/indexing/common/actions/LocalTaskActionClient.java index 27e0bcbaa29d..a07977c6a564 100644 --- a/indexing-service/src/main/java/org/apache/druid/indexing/common/actions/LocalTaskActionClient.java +++ b/indexing-service/src/main/java/org/apache/druid/indexing/common/actions/LocalTaskActionClient.java @@ -30,6 +30,7 @@ import javax.annotation.Nullable; import java.util.Map; +import java.util.concurrent.TimeUnit; public class LocalTaskActionClient implements TaskActionClient { @@ -76,11 +77,28 @@ public RetType submit(TaskAction taskAction) } final long performStartTime = System.currentTimeMillis(); - final RetType result = taskAction.perform(task, toolbox); + final RetType result = performAction(taskAction); emitTimerMetric("task/action/run/time", taskAction, System.currentTimeMillis() - performStartTime); return result; } + private R performAction(TaskAction taskAction) + { + try { + final R result; + if (taskAction.canPerformAsync(task, toolbox)) { + result = taskAction.performAsync(task, toolbox).get(5, TimeUnit.MINUTES); + } else { + result = taskAction.perform(task, toolbox); + } + + return result; + } + catch (Throwable t) { + throw new RuntimeException(t); + } + } + private void emitTimerMetric(final String metric, final TaskAction action, final long time) { final ServiceMetricEvent.Builder metricBuilder = ServiceMetricEvent.builder(); diff --git a/indexing-service/src/main/java/org/apache/druid/indexing/common/actions/SegmentAllocateAction.java b/indexing-service/src/main/java/org/apache/druid/indexing/common/actions/SegmentAllocateAction.java index f61dad98168d..f0fae4a8617d 100644 --- a/indexing-service/src/main/java/org/apache/druid/indexing/common/actions/SegmentAllocateAction.java +++ b/indexing-service/src/main/java/org/apache/druid/indexing/common/actions/SegmentAllocateAction.java @@ -47,6 +47,7 @@ import java.util.HashSet; import java.util.List; import java.util.Set; +import java.util.concurrent.Future; import java.util.concurrent.ThreadLocalRandom; import java.util.stream.Collectors; @@ -180,6 +181,23 @@ public TypeReference getReturnTypeReference() }; } + @Override + public boolean canPerformAsync(Task task, TaskActionToolbox toolbox) + { + return toolbox.canBatchSegmentAllocation(); + } + + @Override + public Future performAsync(Task task, TaskActionToolbox toolbox) + { + if (!toolbox.canBatchSegmentAllocation()) { + throw new ISE("Batched segment allocation is disabled"); + } + return toolbox.getSegmentAllocationQueue().add( + new SegmentAllocateRequest(task, this, MAX_ATTEMPTS) + ); + } + @Override public SegmentIdWithShardSpec perform( final Task task, diff --git a/indexing-service/src/main/java/org/apache/druid/indexing/common/actions/SegmentAllocateRequest.java b/indexing-service/src/main/java/org/apache/druid/indexing/common/actions/SegmentAllocateRequest.java new file mode 100644 index 000000000000..adac7523f442 --- /dev/null +++ b/indexing-service/src/main/java/org/apache/druid/indexing/common/actions/SegmentAllocateRequest.java @@ -0,0 +1,77 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.indexing.common.actions; + +import org.apache.druid.indexing.common.task.Task; +import org.joda.time.Interval; +import org.joda.time.chrono.ISOChronology; + +/** + * Request received by the overlord for segment allocation. + */ +public class SegmentAllocateRequest +{ + private final Task task; + private final SegmentAllocateAction action; + private final int maxAttempts; + private final Interval rowInterval; + + private int attempts; + + public SegmentAllocateRequest(Task task, SegmentAllocateAction action, int maxAttempts) + { + this.task = task; + this.action = action; + this.maxAttempts = maxAttempts; + this.rowInterval = action.getQueryGranularity() + .bucket(action.getTimestamp()) + .withChronology(ISOChronology.getInstanceUTC()); + } + + public Task getTask() + { + return task; + } + + public SegmentAllocateAction getAction() + { + return action; + } + + public void incrementAttempts() + { + ++attempts; + } + + public boolean canRetry() + { + return attempts < maxAttempts; + } + + public int getAttempts() + { + return attempts; + } + + public Interval getRowInterval() + { + return rowInterval; + } +} diff --git a/indexing-service/src/main/java/org/apache/druid/indexing/common/actions/SegmentAllocateResult.java b/indexing-service/src/main/java/org/apache/druid/indexing/common/actions/SegmentAllocateResult.java new file mode 100644 index 000000000000..995bfb9d31d0 --- /dev/null +++ b/indexing-service/src/main/java/org/apache/druid/indexing/common/actions/SegmentAllocateResult.java @@ -0,0 +1,49 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.indexing.common.actions; + +import org.apache.druid.segment.realtime.appenderator.SegmentIdWithShardSpec; + +public class SegmentAllocateResult +{ + private final SegmentIdWithShardSpec segmentId; + private final String errorMessage; + + public SegmentAllocateResult(SegmentIdWithShardSpec segmentId, String errorMessage) + { + this.segmentId = segmentId; + this.errorMessage = errorMessage; + } + + public SegmentIdWithShardSpec getSegmentId() + { + return segmentId; + } + + public String getErrorMessage() + { + return errorMessage; + } + + public boolean isSuccess() + { + return segmentId != null; + } +} diff --git a/indexing-service/src/main/java/org/apache/druid/indexing/common/actions/SegmentAllocationQueue.java b/indexing-service/src/main/java/org/apache/druid/indexing/common/actions/SegmentAllocationQueue.java new file mode 100644 index 000000000000..9ed53d99faf5 --- /dev/null +++ b/indexing-service/src/main/java/org/apache/druid/indexing/common/actions/SegmentAllocationQueue.java @@ -0,0 +1,716 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.indexing.common.actions; + +import com.google.inject.Inject; +import org.apache.druid.guice.ManageLifecycle; +import org.apache.druid.indexing.common.LockGranularity; +import org.apache.druid.indexing.common.task.IndexTaskUtils; +import org.apache.druid.indexing.common.task.Task; +import org.apache.druid.indexing.overlord.IndexerMetadataStorageCoordinator; +import org.apache.druid.indexing.overlord.Segments; +import org.apache.druid.indexing.overlord.TaskLockbox; +import org.apache.druid.indexing.overlord.config.TaskLockConfig; +import org.apache.druid.java.util.common.ISE; +import org.apache.druid.java.util.common.Intervals; +import org.apache.druid.java.util.common.concurrent.ScheduledExecutorFactory; +import org.apache.druid.java.util.common.granularity.Granularity; +import org.apache.druid.java.util.common.guava.Comparators; +import org.apache.druid.java.util.common.lifecycle.LifecycleStart; +import org.apache.druid.java.util.common.lifecycle.LifecycleStop; +import org.apache.druid.java.util.common.logger.Logger; +import org.apache.druid.java.util.emitter.service.ServiceEmitter; +import org.apache.druid.java.util.emitter.service.ServiceMetricEvent; +import org.apache.druid.query.DruidMetrics; +import org.apache.druid.segment.realtime.appenderator.SegmentIdWithShardSpec; +import org.apache.druid.timeline.DataSegment; +import org.joda.time.Interval; + +import java.util.ArrayList; +import java.util.HashMap; +import java.util.HashSet; +import java.util.List; +import java.util.Map; +import java.util.Objects; +import java.util.Set; +import java.util.TreeSet; +import java.util.concurrent.BlockingDeque; +import java.util.concurrent.CompletableFuture; +import java.util.concurrent.ConcurrentHashMap; +import java.util.concurrent.Future; +import java.util.concurrent.LinkedBlockingDeque; +import java.util.concurrent.ScheduledExecutorService; +import java.util.concurrent.TimeUnit; +import java.util.concurrent.atomic.AtomicBoolean; +import java.util.concurrent.atomic.AtomicReference; + +/** + * Queue for {@link SegmentAllocateRequest}s. + */ +@ManageLifecycle +public class SegmentAllocationQueue +{ + private static final Logger log = new Logger(SegmentAllocationQueue.class); + + private static final int MAX_QUEUE_SIZE = 2000; + + private final long maxWaitTimeMillis; + + private final TaskLockbox taskLockbox; + private final ScheduledExecutorService executor; + private final IndexerMetadataStorageCoordinator metadataStorage; + private final AtomicBoolean isLeader = new AtomicBoolean(false); + private final ServiceEmitter emitter; + + private final ConcurrentHashMap keyToBatch = new ConcurrentHashMap<>(); + private final BlockingDeque processingQueue = new LinkedBlockingDeque<>(MAX_QUEUE_SIZE); + + @Inject + public SegmentAllocationQueue( + TaskLockbox taskLockbox, + TaskLockConfig taskLockConfig, + IndexerMetadataStorageCoordinator metadataStorage, + ServiceEmitter emitter, + ScheduledExecutorFactory executorFactory + ) + { + this.emitter = emitter; + this.taskLockbox = taskLockbox; + this.metadataStorage = metadataStorage; + this.maxWaitTimeMillis = taskLockConfig.getBatchAllocationMaxWaitTime(); + + this.executor = taskLockConfig.isBatchSegmentAllocation() + ? executorFactory.create(1, "SegmentAllocQueue-%s") : null; + } + + @LifecycleStart + public void start() + { + if (isEnabled()) { + log.info("Initializing segment allocation queue."); + scheduleQueuePoll(maxWaitTimeMillis); + } + } + + @LifecycleStop + public void stop() + { + if (isEnabled()) { + log.info("Tearing down segment allocation queue."); + executor.shutdownNow(); + } + } + + public void becomeLeader() + { + if (!isLeader.compareAndSet(false, true)) { + log.info("Already the leader. Queue processing has started."); + } else if (isEnabled()) { + log.info("Elected leader. Starting queue processing."); + } else { + log.info( + "Elected leader but batched segment allocation is disabled. " + + "Segment allocation queue will not be used." + ); + } + } + + public void stopBeingLeader() + { + if (!isLeader.compareAndSet(true, false)) { + log.info("Already surrendered leadership. Queue processing is stopped."); + } else if (isEnabled()) { + log.info("Not leader anymore. Stopping queue processing."); + } else { + log.info("Not leader anymore. Segment allocation queue is already disabled."); + } + } + + public boolean isEnabled() + { + return executor != null && !executor.isShutdown(); + } + + private void scheduleQueuePoll(long delay) + { + executor.schedule(this::processBatchesDue, delay, TimeUnit.MILLISECONDS); + } + + /** + * Gets the number of batches currently in the queue. + */ + public int size() + { + return processingQueue.size(); + } + + /** + * Queues a SegmentAllocateRequest. The returned future may complete successfully + * with a non-null value or with a non-null value. + */ + public Future add(SegmentAllocateRequest request) + { + if (!isLeader.get()) { + throw new ISE("Cannot allocate segment if not leader."); + } else if (!isEnabled()) { + throw new ISE("Batched segment allocation is disabled."); + } + + final AllocateRequestKey requestKey = new AllocateRequestKey(request, maxWaitTimeMillis); + final AtomicReference> futureReference = new AtomicReference<>(); + + // Possible race condition: + // t1 -> new batch is added to queue or batch already exists in queue + // t2 -> executor pops batch, processes all requests in it + // t1 -> new request is added to dangling batch and is never picked up + // Solution: For existing batch, call keyToBatch.remove() on the key to + // wait on keyToBatch.compute() to finish before proceeding with processBatch(). + // For new batch, keyToBatch.remove() would not wait as key is not in map yet + // but a new batch is unlikely to be due immediately, so it won't get popped right away. + keyToBatch.compute(requestKey, (key, existingBatch) -> { + if (existingBatch == null) { + AllocateRequestBatch newBatch = new AllocateRequestBatch(key); + futureReference.set(newBatch.add(request)); + return addBatchToQueue(newBatch) ? newBatch : null; + } else { + futureReference.set(existingBatch.add(request)); + return existingBatch; + } + }); + + return futureReference.get(); + } + + /** + * Tries to add the given batch to the processing queue. Fails all the pending + * requests in the batch if we are not leader or if the queue is full. + */ + private boolean addBatchToQueue(AllocateRequestBatch batch) + { + batch.key.resetQueueTime(); + if (!isLeader.get()) { + batch.failPendingRequests("Cannot allocate segment if not leader"); + return false; + } else if (processingQueue.offer(batch.key)) { + log.debug("Added a new batch [%s] to queue.", batch.key); + return true; + } else { + batch.failPendingRequests( + "Segment allocation queue is full. Check the metric `task/action/batch/runTime` " + + "to determine if metadata operations are slow." + ); + return false; + } + } + + /** + * Tries to add the given batch to the processing queue. If a batch already + * exists for this key, transfers all the requests from this batch to the + * existing one. + */ + private void requeueBatch(AllocateRequestBatch batch) + { + log.info("Requeueing [%d] failed requests in batch [%s].", batch.size(), batch.key); + keyToBatch.compute(batch.key, (key, existingBatch) -> { + if (existingBatch == null) { + return addBatchToQueue(batch) ? batch : null; + } else { + // Merge requests from this batch to existing one + existingBatch.transferRequestsFrom(batch); + return existingBatch; + } + }); + } + + private void processBatchesDue() + { + clearQueueIfNotLeader(); + + // Process all batches which are due + log.debug("Processing batches which are due. Queue size [%d].", processingQueue.size()); + int numProcessedBatches = 0; + + AllocateRequestKey nextKey = processingQueue.peekFirst(); + while (nextKey != null && nextKey.isDue()) { + processingQueue.pollFirst(); + AllocateRequestBatch nextBatch = keyToBatch.remove(nextKey); + + boolean processed; + try { + processed = processBatch(nextBatch); + } + catch (Throwable t) { + nextBatch.failPendingRequests(t); + processed = true; + log.error(t, "Error while processing batch [%s]", nextKey); + } + + // Requeue if not fully processed yet + if (processed) { + ++numProcessedBatches; + } else { + requeueBatch(nextBatch); + } + + nextKey = processingQueue.peek(); + } + + // Schedule the next round of processing + final long nextScheduleDelay; + if (processingQueue.isEmpty()) { + nextScheduleDelay = maxWaitTimeMillis; + } else { + nextKey = processingQueue.peek(); + long timeElapsed = System.currentTimeMillis() - nextKey.getQueueTime(); + nextScheduleDelay = Math.max(0, maxWaitTimeMillis - timeElapsed); + } + scheduleQueuePoll(nextScheduleDelay); + log.info("Processed [%d] batches, next execution in [%d ms]", numProcessedBatches, nextScheduleDelay); + } + + /** + * Removes items from the queue as long as we are not leader. + */ + private void clearQueueIfNotLeader() + { + int failedBatches = 0; + AllocateRequestKey nextKey = processingQueue.peekFirst(); + while (nextKey != null && !isLeader.get()) { + processingQueue.pollFirst(); + AllocateRequestBatch nextBatch = keyToBatch.remove(nextKey); + nextBatch.failPendingRequests("Cannot allocate segment if not leader"); + ++failedBatches; + + nextKey = processingQueue.peekFirst(); + } + if (failedBatches > 0) { + log.info("Not leader. Failed [%d] batches, remaining in queue [%d].", failedBatches, processingQueue.size()); + } + } + + /** + * Processes the given batch. Returns true if the batch was completely processed + * and should not be requeued. + */ + private boolean processBatch(AllocateRequestBatch requestBatch) + { + final AllocateRequestKey requestKey = requestBatch.key; + if (requestBatch.isEmpty()) { + return true; + } else if (!isLeader.get()) { + requestBatch.failPendingRequests("Cannot allocate segment if not leader"); + return true; + } + + log.debug( + "Processing [%d] requests for batch [%s], queue time [%s].", + requestBatch.size(), + requestKey, + requestKey.getQueueTime() + ); + + final long startTimeMillis = System.currentTimeMillis(); + final int batchSize = requestBatch.size(); + emitBatchMetric("task/action/batch/size", batchSize, requestKey); + emitBatchMetric("task/action/batch/queueTime", (startTimeMillis - requestKey.getQueueTime()), requestKey); + + final Set usedSegments = retrieveUsedSegments(requestKey); + final int successCount = allocateSegmentsForBatch(requestBatch, usedSegments); + + emitBatchMetric("task/action/batch/attempts", 1L, requestKey); + emitBatchMetric("task/action/batch/runTime", (System.currentTimeMillis() - startTimeMillis), requestKey); + log.info("Successfully processed [%d / %d] requests in batch [%s].", successCount, batchSize, requestKey); + + if (requestBatch.isEmpty()) { + log.debug("All requests in batch [%s] have been processed.", requestKey); + return true; + } + + // Requeue the batch only if used segments have changed + log.debug("There are [%d] failed requests in batch [%s].", requestBatch.size(), requestKey); + final Set updatedUsedSegments = retrieveUsedSegments(requestKey); + + if (updatedUsedSegments.equals(usedSegments)) { + requestBatch.failPendingRequests("Allocation failed probably due to conflicting segments."); + return true; + } else { + log.debug("Used segments have changed. Requeuing failed requests."); + return false; + } + } + + private Set retrieveUsedSegments(AllocateRequestKey key) + { + return new HashSet<>( + metadataStorage.retrieveUsedSegmentsForInterval( + key.dataSource, + key.preferredAllocationInterval, + Segments.ONLY_VISIBLE + ) + ); + } + + private int allocateSegmentsForBatch(AllocateRequestBatch requestBatch, Set usedSegments) + { + int successCount = 0; + + // Find requests whose row interval overlaps with an existing used segment + final Set allRequests = requestBatch.getRequests(); + final Set requestsWithNoOverlappingSegment = new HashSet<>(); + + if (usedSegments.isEmpty()) { + requestsWithNoOverlappingSegment.addAll(allRequests); + } else { + final Interval[] sortedUsedSegmentIntervals = getSortedIntervals(usedSegments); + final Map> overlapIntervalToRequests = new HashMap<>(); + + for (SegmentAllocateRequest request : allRequests) { + // If there is an overlapping used segment, the interval of the used segment + // is the only candidate for allocation for this request + final Interval overlappingInterval = Intervals.findOverlappingInterval( + request.getRowInterval(), + sortedUsedSegmentIntervals + ); + + if (overlappingInterval == null) { + requestsWithNoOverlappingSegment.add(request); + } else if (overlappingInterval.contains(request.getRowInterval())) { + // Found an enclosing interval, use this for allocation + overlapIntervalToRequests.computeIfAbsent(overlappingInterval, i -> new ArrayList<>()) + .add(request); + } else { + // There is no valid allocation interval for this request due to a + // partially overlapping used segment. Need not do anything right now. + // The request will be retried upon requeueing the batch. + } + } + + // Try to allocate segments for the identified used segment intervals. + // Do not retry the failed requests with other intervals unless the batch is requeued. + for (Map.Entry> entry : overlapIntervalToRequests.entrySet()) { + successCount += allocateSegmentsForInterval( + entry.getKey(), + entry.getValue(), + requestBatch + ); + } + } + + // For requests that do not overlap with a used segment, first try to allocate + // using the preferred granularity, then smaller granularities + final Set pendingRequests = new HashSet<>(requestsWithNoOverlappingSegment); + for (Granularity granularity : + Granularity.granularitiesFinerThan(requestBatch.key.preferredSegmentGranularity)) { + Map> requestsByInterval = + getRequestsByInterval(pendingRequests, granularity); + + for (Map.Entry> entry : requestsByInterval.entrySet()) { + successCount += allocateSegmentsForInterval( + entry.getKey(), + entry.getValue(), + requestBatch + ); + pendingRequests.retainAll(requestBatch.getRequests()); + } + } + + return successCount; + } + + private Interval[] getSortedIntervals(Set usedSegments) + { + TreeSet sortedSet = new TreeSet<>(Comparators.intervalsByStartThenEnd()); + usedSegments.forEach(segment -> sortedSet.add(segment.getInterval())); + return sortedSet.toArray(new Interval[0]); + } + + /** + * Tries to allocate segments for the given requests over the specified interval. + * Returns the number of requests for which segments were successfully allocated. + */ + private int allocateSegmentsForInterval( + Interval tryInterval, + List requests, + AllocateRequestBatch requestBatch + ) + { + if (requests.isEmpty()) { + return 0; + } + + final AllocateRequestKey requestKey = requestBatch.key; + log.debug( + "Trying allocation for [%d] requests, interval [%s] in batch [%s]", + requests.size(), + tryInterval, + requestKey + ); + + final List results = taskLockbox.allocateSegments( + requests, + requestKey.dataSource, + tryInterval, + requestKey.skipSegmentLineageCheck, + requestKey.lockGranularity + ); + + int successfulRequests = 0; + for (int i = 0; i < requests.size(); ++i) { + SegmentAllocateRequest request = requests.get(i); + SegmentAllocateResult result = results.get(i); + if (result.isSuccess()) { + ++successfulRequests; + } + + requestBatch.handleResult(result, request); + } + + return successfulRequests; + } + + private Map> getRequestsByInterval( + Set requests, + Granularity tryGranularity + ) + { + final Map> tryIntervalToRequests = new HashMap<>(); + for (SegmentAllocateRequest request : requests) { + Interval tryInterval = tryGranularity.bucket(request.getAction().getTimestamp()); + if (tryInterval.contains(request.getRowInterval())) { + tryIntervalToRequests.computeIfAbsent(tryInterval, i -> new ArrayList<>()).add(request); + } + } + return tryIntervalToRequests; + } + + private void emitTaskMetric(String metric, long value, SegmentAllocateRequest request) + { + final ServiceMetricEvent.Builder metricBuilder = ServiceMetricEvent.builder(); + IndexTaskUtils.setTaskDimensions(metricBuilder, request.getTask()); + metricBuilder.setDimension("taskActionType", SegmentAllocateAction.TYPE); + emitter.emit(metricBuilder.build(metric, value)); + } + + private void emitBatchMetric(String metric, long value, AllocateRequestKey key) + { + final ServiceMetricEvent.Builder metricBuilder = ServiceMetricEvent.builder(); + metricBuilder.setDimension("taskActionType", SegmentAllocateAction.TYPE); + metricBuilder.setDimension(DruidMetrics.DATASOURCE, key.dataSource); + metricBuilder.setDimension(DruidMetrics.INTERVAL, key.preferredAllocationInterval.toString()); + emitter.emit(metricBuilder.build(metric, value)); + } + + /** + * A batch of segment allocation requests. + */ + private class AllocateRequestBatch + { + private final AllocateRequestKey key; + + /** + * Map from allocate requests (represents a single SegmentAllocateAction) + * to the future of allocated segment id. + *

    + * This must be accessed through methods synchronized on this batch. + * It is to avoid races between a new request being added just when the batch + * is being processed. + */ + private final Map> + requestToFuture = new HashMap<>(); + + AllocateRequestBatch(AllocateRequestKey key) + { + this.key = key; + } + + synchronized Future add(SegmentAllocateRequest request) + { + log.debug("Adding request to batch [%s]: %s", key, request.getAction()); + return requestToFuture.computeIfAbsent(request, req -> new CompletableFuture<>()); + } + + synchronized void transferRequestsFrom(AllocateRequestBatch batch) + { + requestToFuture.putAll(batch.requestToFuture); + batch.requestToFuture.clear(); + } + + synchronized Set getRequests() + { + return new HashSet<>(requestToFuture.keySet()); + } + + synchronized void failPendingRequests(String reason) + { + failPendingRequests(new ISE(reason)); + } + + synchronized void failPendingRequests(Throwable cause) + { + if (!requestToFuture.isEmpty()) { + log.warn("Failing [%d] requests in batch due to [%s]. Batch key: %s", size(), cause.getMessage(), key); + requestToFuture.values().forEach(future -> future.completeExceptionally(cause)); + requestToFuture.keySet().forEach( + request -> emitTaskMetric("task/action/failed/count", 1L, request) + ); + requestToFuture.clear(); + } + } + + synchronized void handleResult(SegmentAllocateResult result, SegmentAllocateRequest request) + { + request.incrementAttempts(); + + if (result.isSuccess()) { + emitTaskMetric("task/action/success/count", 1L, request); + requestToFuture.remove(request).complete(result.getSegmentId()); + } else if (request.canRetry()) { + log.info( + "Allocation failed in attempt [%d] due to error [%s]. Can still retry. Action: %s", + request.getAttempts(), + result.getErrorMessage(), + request.getAction() + ); + } else { + emitTaskMetric("task/action/failed/count", 1L, request); + log.error( + "Failing allocate action after [%d] attempts. Latest error [%s]. Action: %s", + request.getAttempts(), + result.getErrorMessage(), + request.getAction() + ); + requestToFuture.remove(request).completeExceptionally(new ISE(result.getErrorMessage())); + } + } + + synchronized boolean isEmpty() + { + return requestToFuture.isEmpty(); + } + + synchronized int size() + { + return requestToFuture.size(); + } + } + + /** + * Key to identify a batch of allocation requests. + */ + private static class AllocateRequestKey + { + private long queueTimeMillis; + private final long maxWaitTimeMillis; + + private final String dataSource; + private final String groupId; + private final Interval preferredAllocationInterval; + private final Granularity preferredSegmentGranularity; + + private final boolean skipSegmentLineageCheck; + private final LockGranularity lockGranularity; + + private final boolean useNonRootGenPartitionSpace; + + private final int hash; + + /** + * Creates a new key for the given request. The batch for a unique key will + * always contain a single request. + */ + AllocateRequestKey(SegmentAllocateRequest request, long maxWaitTimeMillis) + { + final SegmentAllocateAction action = request.getAction(); + final Task task = request.getTask(); + + this.dataSource = action.getDataSource(); + this.groupId = task.getGroupId(); + this.skipSegmentLineageCheck = action.isSkipSegmentLineageCheck(); + this.lockGranularity = action.getLockGranularity(); + this.useNonRootGenPartitionSpace = action.getPartialShardSpec() + .useNonRootGenerationPartitionSpace(); + this.preferredSegmentGranularity = action.getPreferredSegmentGranularity(); + this.preferredAllocationInterval = action.getPreferredSegmentGranularity() + .bucket(action.getTimestamp()); + + this.hash = Objects.hash( + skipSegmentLineageCheck, + useNonRootGenPartitionSpace, + dataSource, + groupId, + preferredAllocationInterval, + lockGranularity + ); + + this.maxWaitTimeMillis = maxWaitTimeMillis; + } + + void resetQueueTime() + { + queueTimeMillis = System.currentTimeMillis(); + } + + long getQueueTime() + { + return queueTimeMillis; + } + + boolean isDue() + { + return System.currentTimeMillis() - queueTimeMillis >= maxWaitTimeMillis; + } + + @Override + public boolean equals(Object o) + { + if (this == o) { + return true; + } + if (o == null || getClass() != o.getClass()) { + return false; + } + AllocateRequestKey that = (AllocateRequestKey) o; + return skipSegmentLineageCheck == that.skipSegmentLineageCheck + && useNonRootGenPartitionSpace == that.useNonRootGenPartitionSpace + && dataSource.equals(that.dataSource) + && groupId.equals(that.groupId) + && preferredAllocationInterval.equals(that.preferredAllocationInterval) + && lockGranularity == that.lockGranularity; + } + + @Override + public int hashCode() + { + return hash; + } + + @Override + public String toString() + { + return "{" + + "ds='" + dataSource + '\'' + + ", gr='" + groupId + '\'' + + ", lock=" + lockGranularity + + ", invl=" + preferredAllocationInterval + + ", slc=" + skipSegmentLineageCheck + + '}'; + } + } +} diff --git a/indexing-service/src/main/java/org/apache/druid/indexing/common/actions/TaskAction.java b/indexing-service/src/main/java/org/apache/druid/indexing/common/actions/TaskAction.java index 559039d96ef6..18e373727790 100644 --- a/indexing-service/src/main/java/org/apache/druid/indexing/common/actions/TaskAction.java +++ b/indexing-service/src/main/java/org/apache/druid/indexing/common/actions/TaskAction.java @@ -24,6 +24,8 @@ import com.fasterxml.jackson.core.type.TypeReference; import org.apache.druid.indexing.common.task.Task; +import java.util.concurrent.Future; + @JsonTypeInfo(use = JsonTypeInfo.Id.NAME, property = TaskAction.TYPE_FIELD) @JsonSubTypes(value = { @JsonSubTypes.Type(name = "lockAcquire", value = TimeChunkLockAcquireAction.class), @@ -58,6 +60,16 @@ public interface TaskAction boolean isAudited(); + default boolean canPerformAsync(Task task, TaskActionToolbox toolbox) + { + return false; + } + + default Future performAsync(Task task, TaskActionToolbox toolbox) + { + throw new UnsupportedOperationException(); + } + @Override String toString(); } diff --git a/indexing-service/src/main/java/org/apache/druid/indexing/common/actions/TaskActionToolbox.java b/indexing-service/src/main/java/org/apache/druid/indexing/common/actions/TaskActionToolbox.java index 134a9bf6c72e..7d001ecbcb9a 100644 --- a/indexing-service/src/main/java/org/apache/druid/indexing/common/actions/TaskActionToolbox.java +++ b/indexing-service/src/main/java/org/apache/druid/indexing/common/actions/TaskActionToolbox.java @@ -35,6 +35,7 @@ public class TaskActionToolbox { private final TaskLockbox taskLockbox; private final TaskStorage taskStorage; + private final SegmentAllocationQueue segmentAllocationQueue; private final IndexerMetadataStorageCoordinator indexerMetadataStorageCoordinator; private final ServiceEmitter emitter; private final SupervisorManager supervisorManager; @@ -46,6 +47,7 @@ public TaskActionToolbox( TaskLockbox taskLockbox, TaskStorage taskStorage, IndexerMetadataStorageCoordinator indexerMetadataStorageCoordinator, + SegmentAllocationQueue segmentAllocationQueue, ServiceEmitter emitter, SupervisorManager supervisorManager, @Json ObjectMapper jsonMapper @@ -57,6 +59,27 @@ public TaskActionToolbox( this.emitter = emitter; this.supervisorManager = supervisorManager; this.jsonMapper = jsonMapper; + this.segmentAllocationQueue = segmentAllocationQueue; + } + + public TaskActionToolbox( + TaskLockbox taskLockbox, + TaskStorage taskStorage, + IndexerMetadataStorageCoordinator indexerMetadataStorageCoordinator, + ServiceEmitter emitter, + SupervisorManager supervisorManager, + @Json ObjectMapper jsonMapper + ) + { + this( + taskLockbox, + taskStorage, + indexerMetadataStorageCoordinator, + null, + emitter, + supervisorManager, + jsonMapper + ); } public TaskLockbox getTaskLockbox() @@ -103,4 +126,13 @@ public Optional getTaskRunner() return Optional.absent(); } + public SegmentAllocationQueue getSegmentAllocationQueue() + { + return segmentAllocationQueue; + } + + public boolean canBatchSegmentAllocation() + { + return segmentAllocationQueue != null && segmentAllocationQueue.isEnabled(); + } } diff --git a/indexing-service/src/main/java/org/apache/druid/indexing/overlord/TaskLockbox.java b/indexing-service/src/main/java/org/apache/druid/indexing/overlord/TaskLockbox.java index 24eebec94471..ca4d4526557b 100644 --- a/indexing-service/src/main/java/org/apache/druid/indexing/overlord/TaskLockbox.java +++ b/indexing-service/src/main/java/org/apache/druid/indexing/overlord/TaskLockbox.java @@ -20,7 +20,6 @@ package org.apache.druid.indexing.overlord; import com.google.common.annotations.VisibleForTesting; -import com.google.common.base.Function; import com.google.common.base.Objects; import com.google.common.base.Preconditions; import com.google.common.collect.ComparisonChain; @@ -34,9 +33,13 @@ import org.apache.druid.indexing.common.TaskLock; import org.apache.druid.indexing.common.TaskLockType; import org.apache.druid.indexing.common.TimeChunkLock; +import org.apache.druid.indexing.common.actions.SegmentAllocateAction; +import org.apache.druid.indexing.common.actions.SegmentAllocateRequest; +import org.apache.druid.indexing.common.actions.SegmentAllocateResult; import org.apache.druid.indexing.common.task.Task; import org.apache.druid.java.util.common.ISE; import org.apache.druid.java.util.common.Pair; +import org.apache.druid.java.util.common.StringUtils; import org.apache.druid.java.util.common.guava.Comparators; import org.apache.druid.java.util.emitter.EmittingLogger; import org.apache.druid.segment.realtime.appenderator.SegmentIdWithShardSpec; @@ -446,6 +449,140 @@ public LockResult tryLock(final Task task, final LockRequest request) } } + /** + * Attempts to allocate segments for the given requests. Each request contains + * a {@link Task} and a {@link SegmentAllocateAction}. This method tries to + * acquire the task locks on the required intervals/segments and then performs + * a batch allocation of segments. It is possible that some requests succeed + * successfully and others failed. In that case, only the failed ones should be + * retried. + * + * @param requests List of allocation requests + * @param dataSource Datasource for which segment is to be allocated. + * @param interval Interval for which segment is to be allocated. + * @param skipSegmentLineageCheck Whether lineage check is to be skipped + * (this is true for streaming ingestion) + * @param lockGranularity Granularity of task lock + * @return List of allocation results in the same order as the requests. + */ + public List allocateSegments( + List requests, + String dataSource, + Interval interval, + boolean skipSegmentLineageCheck, + LockGranularity lockGranularity + ) + { + log.info("Allocating [%d] segments for datasource [%s], interval [%s]", requests.size(), dataSource, interval); + final boolean isTimeChunkLock = lockGranularity == LockGranularity.TIME_CHUNK; + + final AllocationHolderList holderList = new AllocationHolderList(requests, interval); + holderList.getPending().forEach(this::verifyTaskIsActive); + + giant.lock(); + try { + if (isTimeChunkLock) { + // For time-chunk locking, segment must be allocated only after acquiring the lock + holderList.getPending().forEach(holder -> acquireTaskLock(holder, true)); + allocateSegmentIds(dataSource, interval, skipSegmentLineageCheck, holderList.getPending()); + } else { + allocateSegmentIds(dataSource, interval, skipSegmentLineageCheck, holderList.getPending()); + holderList.getPending().forEach(holder -> acquireTaskLock(holder, false)); + } + + holderList.getPending().forEach(holder -> addTaskAndPersistLocks(holder, isTimeChunkLock)); + } + finally { + giant.unlock(); + } + + return holderList.getResults(); + } + + /** + * Marks the segment allocation as failed if the underlying task is not active. + */ + private void verifyTaskIsActive(SegmentAllocationHolder holder) + { + final String taskId = holder.task.getId(); + if (!activeTasks.contains(taskId)) { + holder.markFailed("Unable to grant lock to inactive Task [%s]", taskId); + } + } + + /** + * Creates a task lock request and creates or finds the lock for that request. + * Marks the segment allocation as failed if the lock could not be acquired or + * was revoked. + */ + private void acquireTaskLock(SegmentAllocationHolder holder, boolean isTimeChunkLock) + { + final LockRequest lockRequest; + if (isTimeChunkLock) { + lockRequest = new TimeChunkLockRequest(holder.lockRequest); + } else { + lockRequest = new SpecificSegmentLockRequest(holder.lockRequest, holder.allocatedSegment); + } + + // Create or find the task lock for the created lock request + final TaskLockPosse posseToUse = createOrFindLockPosse(lockRequest); + final TaskLock acquiredLock = posseToUse == null ? null : posseToUse.getTaskLock(); + if (posseToUse == null) { + holder.markFailed("Could not find or create lock posse."); + } else if (acquiredLock.isRevoked()) { + holder.markFailed("Lock was revoked."); + } else { + holder.setAcquiredLock(posseToUse, lockRequest.getInterval()); + } + } + + /** + * Adds the task to the found lock posse if not already added and updates + * in the metadata store. Marks the segment allocation as failed if the update + * did not succeed. + */ + private void addTaskAndPersistLocks(SegmentAllocationHolder holder, boolean isTimeChunkLock) + { + final Task task = holder.task; + final TaskLock acquiredLock = holder.acquiredLock; + + if (holder.taskLockPosse.addTask(task)) { + log.info("Added task [%s] to TaskLock [%s]", task.getId(), acquiredLock); + + // This can also be batched later + boolean success = updateLockInStorage(task, acquiredLock); + if (success) { + holder.markSucceeded(); + } else { + final Integer partitionId = isTimeChunkLock + ? null : ((SegmentLock) acquiredLock).getPartitionId(); + unlock(task, holder.lockRequestInterval, partitionId); + holder.markFailed("Could not update task lock in metadata store."); + } + } else { + log.info("Task [%s] already present in TaskLock [%s]", task.getId(), acquiredLock.getGroupId()); + holder.markSucceeded(); + } + } + + private boolean updateLockInStorage(Task task, TaskLock taskLock) + { + try { + taskStorage.addLock(task.getId(), taskLock); + return true; + } + catch (Exception e) { + log.makeAlert("Failed to persist lock in storage") + .addData("task", task.getId()) + .addData("dataSource", taskLock.getDataSource()) + .addData("interval", taskLock.getInterval()) + .addData("version", taskLock.getVersion()) + .emit(); + + return false; + } + } + private TaskLockPosse createOrFindLockPosse(LockRequest request) { Preconditions.checkState(!(request instanceof LockRequestForNewSegment), "Can't handle LockRequestForNewSegment"); @@ -541,7 +678,6 @@ && areAllEqualOrHigherPriorityLocksSharedOrRevoked(conflictPosses, request.getPr * monotonicity and that callers specifying {@code preferredVersion} are doing the right thing. * * @param request request to lock - * * @return a new {@link TaskLockPosse} */ private TaskLockPosse createNewTaskLockPosse(LockRequest request) @@ -550,7 +686,10 @@ private TaskLockPosse createNewTaskLockPosse(LockRequest request) try { final TaskLockPosse posseToUse = new TaskLockPosse(request.toLock()); running.computeIfAbsent(request.getDataSource(), k -> new TreeMap<>()) - .computeIfAbsent(request.getInterval().getStart(), k -> new TreeMap<>(Comparators.intervalsByStartThenEnd())) + .computeIfAbsent( + request.getInterval().getStart(), + k -> new TreeMap<>(Comparators.intervalsByStartThenEnd()) + ) .computeIfAbsent(request.getInterval(), k -> new ArrayList<>()) .add(posseToUse); @@ -561,6 +700,45 @@ private TaskLockPosse createNewTaskLockPosse(LockRequest request) } } + /** + * Makes a call to the {@link #metadataStorageCoordinator} to allocate segments + * for the given requests. Updates the holder with the allocated segment if + * the allocation succeeds, otherwise marks it as failed. + */ + private void allocateSegmentIds( + String dataSource, + Interval interval, + boolean skipSegmentLineageCheck, + Collection holders + ) + { + if (holders.isEmpty()) { + return; + } + + final List createRequests = + holders.stream() + .map(SegmentAllocationHolder::getSegmentRequest) + .collect(Collectors.toList()); + + Map allocatedSegments = + metadataStorageCoordinator.allocatePendingSegments( + dataSource, + interval, + skipSegmentLineageCheck, + createRequests + ); + + for (SegmentAllocationHolder holder : holders) { + SegmentIdWithShardSpec segmentId = allocatedSegments.get(holder.getSegmentRequest()); + if (segmentId == null) { + holder.markFailed("Storage coordinator could not allocate segment."); + } else { + holder.setAllocatedSegment(segmentId); + } + } + } + private SegmentIdWithShardSpec allocateSegmentId(LockRequestForNewSegment request, String version) { return metadataStorageCoordinator.allocatePendingSegment( @@ -577,7 +755,7 @@ private SegmentIdWithShardSpec allocateSegmentId(LockRequestForNewSegment reques /** * Perform the given action with a guarantee that the locks of the task are not revoked in the middle of action. This * method first checks that all locks for the given task and intervals are valid and perform the right action. - * + *

    * The given action should be finished as soon as possible because all other methods in this class are blocked until * this method is finished. * @@ -611,7 +789,7 @@ private boolean isTaskLocksValid(Task task, List intervals) .allMatch(interval -> { final List lockPosses = getOnlyTaskLockPosseContainingInterval(task, interval); return lockPosses.stream().map(TaskLockPosse::getTaskLock).noneMatch( - lock -> lock.isRevoked() + TaskLock::isRevoked ); }); } @@ -664,7 +842,9 @@ protected void revokeLock(String taskId, TaskLock lock) final TaskLock revokedLock = lock.revokedCopy(); taskStorage.replaceLock(taskId, lock, revokedLock); - final List possesHolder = running.get(task.getDataSource()).get(lock.getInterval().getStart()).get(lock.getInterval()); + final List possesHolder = running.get(task.getDataSource()) + .get(lock.getInterval().getStart()) + .get(lock.getInterval()); final TaskLockPosse foundPosse = possesHolder.stream() .filter(posse -> posse.getTaskLock().equals(lock)) .findFirst() @@ -692,16 +872,7 @@ public List findLocksForTask(final Task task) giant.lock(); try { - return Lists.transform( - findLockPossesForTask(task), new Function() - { - @Override - public TaskLock apply(TaskLockPosse taskLockPosse) - { - return taskLockPosse.getTaskLock(); - } - } - ); + return Lists.transform(findLockPossesForTask(task), TaskLockPosse::getTaskLock); } finally { giant.unlock(); @@ -778,7 +949,7 @@ public void unlock(final Task task, final Interval interval) * Release lock held for a task on a particular interval. Does nothing if the task does not currently * hold the mentioned lock. * - * @param task task to unlock + * @param task task to unlock * @param interval interval to unlock */ public void unlock(final Task task, final Interval interval, @Nullable Integer partitionId) @@ -1233,7 +1404,7 @@ public boolean equals(Object o) TaskLockPosse that = (TaskLockPosse) o; return java.util.Objects.equals(taskLock, that.taskLock) && - java.util.Objects.equals(taskIds, that.taskIds); + java.util.Objects.equals(taskIds, that.taskIds); } @Override @@ -1251,4 +1422,121 @@ public String toString() .toString(); } } + + /** + * Maintains a list of pending allocation holders. + */ + private static class AllocationHolderList + { + final List all = new ArrayList<>(); + final Set pending = new HashSet<>(); + final Set recentlyCompleted = new HashSet<>(); + + AllocationHolderList(List requests, Interval interval) + { + for (SegmentAllocateRequest request : requests) { + SegmentAllocationHolder holder = new SegmentAllocationHolder(request, interval, this); + all.add(holder); + pending.add(holder); + } + } + + void markCompleted(SegmentAllocationHolder holder) + { + recentlyCompleted.add(holder); + } + + Set getPending() + { + pending.removeAll(recentlyCompleted); + recentlyCompleted.clear(); + return pending; + } + + + List getResults() + { + return all.stream().map(holder -> holder.result).collect(Collectors.toList()); + } + } + + /** + * Contains the task, request, lock and final result for a segment allocation. + */ + private static class SegmentAllocationHolder + { + final AllocationHolderList list; + + final Task task; + final Interval allocateInterval; + final SegmentAllocateAction action; + final LockRequestForNewSegment lockRequest; + SegmentCreateRequest segmentRequest; + + TaskLock acquiredLock; + TaskLockPosse taskLockPosse; + Interval lockRequestInterval; + SegmentIdWithShardSpec allocatedSegment; + SegmentAllocateResult result; + + SegmentAllocationHolder(SegmentAllocateRequest request, Interval allocateInterval, AllocationHolderList list) + { + this.list = list; + this.allocateInterval = allocateInterval; + this.task = request.getTask(); + this.action = request.getAction(); + + this.lockRequest = new LockRequestForNewSegment( + action.getLockGranularity(), + action.getTaskLockType(), + task.getGroupId(), + action.getDataSource(), + allocateInterval, + action.getPartialShardSpec(), + task.getPriority(), + action.getSequenceName(), + action.getPreviousSegmentId(), + action.isSkipSegmentLineageCheck() + ); + } + + SegmentCreateRequest getSegmentRequest() + { + // Initialize the first time this is requested + if (segmentRequest == null) { + segmentRequest = new SegmentCreateRequest( + action.getSequenceName(), + action.getPreviousSegmentId(), + acquiredLock == null ? lockRequest.getVersion() : acquiredLock.getVersion(), + action.getPartialShardSpec() + ); + } + + return segmentRequest; + } + + void markFailed(String msgFormat, Object... args) + { + list.markCompleted(this); + result = new SegmentAllocateResult(null, StringUtils.format(msgFormat, args)); + } + + void markSucceeded() + { + list.markCompleted(this); + result = new SegmentAllocateResult(allocatedSegment, null); + } + + void setAllocatedSegment(SegmentIdWithShardSpec segmentId) + { + this.allocatedSegment = segmentId; + } + + void setAcquiredLock(TaskLockPosse lockPosse, Interval lockRequestInterval) + { + this.taskLockPosse = lockPosse; + this.acquiredLock = lockPosse == null ? null : lockPosse.getTaskLock(); + this.lockRequestInterval = lockRequestInterval; + } + } } diff --git a/indexing-service/src/main/java/org/apache/druid/indexing/overlord/TaskMaster.java b/indexing-service/src/main/java/org/apache/druid/indexing/overlord/TaskMaster.java index 7b9101cf1f22..28c623fdadb4 100644 --- a/indexing-service/src/main/java/org/apache/druid/indexing/overlord/TaskMaster.java +++ b/indexing-service/src/main/java/org/apache/druid/indexing/overlord/TaskMaster.java @@ -26,6 +26,7 @@ import org.apache.druid.discovery.DruidLeaderSelector; import org.apache.druid.discovery.DruidLeaderSelector.Listener; import org.apache.druid.guice.annotations.Self; +import org.apache.druid.indexing.common.actions.SegmentAllocationQueue; import org.apache.druid.indexing.common.actions.TaskActionClient; import org.apache.druid.indexing.common.actions.TaskActionClientFactory; import org.apache.druid.indexing.common.task.Task; @@ -91,7 +92,8 @@ public TaskMaster( final ServiceEmitter emitter, final SupervisorManager supervisorManager, final OverlordHelperManager overlordHelperManager, - @IndexingService final DruidLeaderSelector overlordLeaderSelector + @IndexingService final DruidLeaderSelector overlordLeaderSelector, + final SegmentAllocationQueue segmentAllocationQueue ) { this.supervisorManager = supervisorManager; @@ -136,6 +138,22 @@ public void becomeLeader() leaderLifecycle.addManagedInstance(taskQueue); leaderLifecycle.addManagedInstance(supervisorManager); leaderLifecycle.addManagedInstance(overlordHelperManager); + leaderLifecycle.addHandler( + new Lifecycle.Handler() + { + @Override + public void start() + { + segmentAllocationQueue.becomeLeader(); + } + + @Override + public void stop() + { + segmentAllocationQueue.stopBeingLeader(); + } + } + ); leaderLifecycle.addHandler( new Lifecycle.Handler() diff --git a/indexing-service/src/main/java/org/apache/druid/indexing/overlord/config/TaskLockConfig.java b/indexing-service/src/main/java/org/apache/druid/indexing/overlord/config/TaskLockConfig.java index 96f14759f78a..acbc318baabc 100644 --- a/indexing-service/src/main/java/org/apache/druid/indexing/overlord/config/TaskLockConfig.java +++ b/indexing-service/src/main/java/org/apache/druid/indexing/overlord/config/TaskLockConfig.java @@ -30,8 +30,24 @@ public class TaskLockConfig @JsonProperty private boolean forceTimeChunkLock = true; + @JsonProperty + private boolean batchSegmentAllocation = false; + + @JsonProperty + private long batchAllocationMaxWaitTime = 500L; + public boolean isForceTimeChunkLock() { return forceTimeChunkLock; } + + public boolean isBatchSegmentAllocation() + { + return batchSegmentAllocation; + } + + public long getBatchAllocationMaxWaitTime() + { + return batchAllocationMaxWaitTime; + } } diff --git a/indexing-service/src/test/java/org/apache/druid/indexing/common/actions/SegmentAllocateActionBuilder.java b/indexing-service/src/test/java/org/apache/druid/indexing/common/actions/SegmentAllocateActionBuilder.java new file mode 100644 index 000000000000..2e9cd70bbdbc --- /dev/null +++ b/indexing-service/src/test/java/org/apache/druid/indexing/common/actions/SegmentAllocateActionBuilder.java @@ -0,0 +1,138 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.indexing.common.actions; + +import org.apache.druid.indexing.common.LockGranularity; +import org.apache.druid.indexing.common.TaskLockType; +import org.apache.druid.indexing.common.task.Task; +import org.apache.druid.java.util.common.DateTimes; +import org.apache.druid.java.util.common.granularity.Granularity; +import org.apache.druid.timeline.partition.PartialShardSpec; +import org.joda.time.DateTime; + +public class SegmentAllocateActionBuilder +{ + private String dataSource; + private DateTime timestamp; + private Granularity queryGranularity; + private Granularity preferredSegmentGranularity; + private String sequenceName; + private String previousSegmentId; + private boolean skipSegmentLineageCheck; + private PartialShardSpec partialShardSpec; + private LockGranularity lockGranularity; + private TaskLockType taskLockType; + private Task task; + + public SegmentAllocateActionBuilder forDatasource(String dataSource) + { + this.dataSource = dataSource; + return this; + } + + public SegmentAllocateActionBuilder forTimestamp(DateTime timestamp) + { + this.timestamp = timestamp; + return this; + } + + public SegmentAllocateActionBuilder forTimestamp(String instant) + { + this.timestamp = DateTimes.of(instant); + return this; + } + + public SegmentAllocateActionBuilder withQueryGranularity(Granularity queryGranularity) + { + this.queryGranularity = queryGranularity; + return this; + } + + public SegmentAllocateActionBuilder withSegmentGranularity(Granularity segmentGranularity) + { + this.preferredSegmentGranularity = segmentGranularity; + return this; + } + + public SegmentAllocateActionBuilder withSequenceName(String sequenceName) + { + this.sequenceName = sequenceName; + return this; + } + + public SegmentAllocateActionBuilder withPreviousSegmentId(String previousSegmentId) + { + this.previousSegmentId = previousSegmentId; + return this; + } + + public SegmentAllocateActionBuilder withSkipLineageCheck(boolean skipLineageCheck) + { + this.skipSegmentLineageCheck = skipLineageCheck; + return this; + } + + public SegmentAllocateActionBuilder withPartialShardSpec(PartialShardSpec partialShardSpec) + { + this.partialShardSpec = partialShardSpec; + return this; + } + + public SegmentAllocateActionBuilder withLockGranularity(LockGranularity lockGranularity) + { + this.lockGranularity = lockGranularity; + return this; + } + + public SegmentAllocateActionBuilder withTaskLockType(TaskLockType taskLockType) + { + this.taskLockType = taskLockType; + return this; + } + + public SegmentAllocateActionBuilder forTask(Task task) + { + this.dataSource = task.getDataSource(); + this.sequenceName = task.getId(); + this.task = task; + return this; + } + + public SegmentAllocateRequest build() + { + return new SegmentAllocateRequest(task, buildAction(), 1); + } + + public SegmentAllocateAction buildAction() + { + return new SegmentAllocateAction( + dataSource, + timestamp, + queryGranularity, + preferredSegmentGranularity, + sequenceName, + previousSegmentId, + skipSegmentLineageCheck, + partialShardSpec, + lockGranularity, + taskLockType + ); + } +} diff --git a/indexing-service/src/test/java/org/apache/druid/indexing/common/actions/SegmentAllocateActionTest.java b/indexing-service/src/test/java/org/apache/druid/indexing/common/actions/SegmentAllocateActionTest.java index a7e85a027e61..c8861a92cd3f 100644 --- a/indexing-service/src/test/java/org/apache/druid/indexing/common/actions/SegmentAllocateActionTest.java +++ b/indexing-service/src/test/java/org/apache/druid/indexing/common/actions/SegmentAllocateActionTest.java @@ -20,7 +20,6 @@ package org.apache.druid.indexing.common.actions; import com.fasterxml.jackson.databind.ObjectMapper; -import com.google.common.base.Predicate; import com.google.common.collect.FluentIterable; import com.google.common.collect.ImmutableList; import com.google.common.collect.ImmutableSet; @@ -51,6 +50,7 @@ import org.easymock.EasyMock; import org.joda.time.DateTime; import org.joda.time.Period; +import org.junit.After; import org.junit.Assert; import org.junit.Before; import org.junit.Rule; @@ -64,6 +64,8 @@ import java.util.List; import java.util.Map; import java.util.Map.Entry; +import java.util.concurrent.ExecutionException; +import java.util.concurrent.TimeUnit; import java.util.stream.Collectors; @RunWith(Parameterized.class) @@ -79,20 +81,26 @@ public class SegmentAllocateActionTest private static final DateTime PARTY_TIME = DateTimes.of("1999"); private static final DateTime THE_DISTANT_FUTURE = DateTimes.of("3000"); + private final boolean useBatch; private final LockGranularity lockGranularity; - @Parameterized.Parameters(name = "{0}") + private SegmentAllocationQueue allocationQueue; + + @Parameterized.Parameters(name = "granularity = {0}, useBatch = {1}") public static Iterable constructorFeeder() { return ImmutableList.of( - new Object[]{LockGranularity.SEGMENT}, - new Object[]{LockGranularity.TIME_CHUNK} + new Object[]{LockGranularity.SEGMENT, true}, + new Object[]{LockGranularity.SEGMENT, false}, + new Object[]{LockGranularity.TIME_CHUNK, true}, + new Object[]{LockGranularity.TIME_CHUNK, false} ); } - public SegmentAllocateActionTest(LockGranularity lockGranularity) + public SegmentAllocateActionTest(LockGranularity lockGranularity, boolean useBatch) { this.lockGranularity = lockGranularity; + this.useBatch = useBatch; } @Before @@ -101,6 +109,19 @@ public void setUp() ServiceEmitter emitter = EasyMock.createMock(ServiceEmitter.class); EmittingLogger.registerEmitter(emitter); EasyMock.replay(emitter); + allocationQueue = taskActionTestKit.getTaskActionToolbox().getSegmentAllocationQueue(); + if (allocationQueue != null) { + allocationQueue.start(); + allocationQueue.becomeLeader(); + } + } + + @After + public void tearDown() + { + if (allocationQueue != null) { + allocationQueue.stop(); + } } @Test @@ -288,29 +309,11 @@ public void testResumeSequence() if (lockGranularity == LockGranularity.TIME_CHUNK) { final TaskLock partyLock = Iterables.getOnlyElement( FluentIterable.from(taskActionTestKit.getTaskLockbox().findLocksForTask(task)) - .filter( - new Predicate() - { - @Override - public boolean apply(TaskLock input) - { - return input.getInterval().contains(PARTY_TIME); - } - } - ) + .filter(input -> input.getInterval().contains(PARTY_TIME)) ); final TaskLock futureLock = Iterables.getOnlyElement( FluentIterable.from(taskActionTestKit.getTaskLockbox().findLocksForTask(task)) - .filter( - new Predicate() - { - @Override - public boolean apply(TaskLock input) - { - return input.getInterval().contains(THE_DISTANT_FUTURE); - } - } - ) + .filter(input -> input.getInterval().contains(THE_DISTANT_FUTURE)) ); assertSameIdentifier( @@ -446,29 +449,11 @@ public void testMultipleSequences() if (lockGranularity == LockGranularity.TIME_CHUNK) { final TaskLock partyLock = Iterables.getOnlyElement( FluentIterable.from(taskActionTestKit.getTaskLockbox().findLocksForTask(task)) - .filter( - new Predicate() - { - @Override - public boolean apply(TaskLock input) - { - return input.getInterval().contains(PARTY_TIME); - } - } - ) + .filter(input -> input.getInterval().contains(PARTY_TIME)) ); final TaskLock futureLock = Iterables.getOnlyElement( FluentIterable.from(taskActionTestKit.getTaskLockbox().findLocksForTask(task)) - .filter( - new Predicate() - { - @Override - public boolean apply(TaskLock input) - { - return input.getInterval().contains(THE_DISTANT_FUTURE); - } - } - ) + .filter(input -> input.getInterval().contains(THE_DISTANT_FUTURE)) ); assertSameIdentifier( @@ -990,21 +975,26 @@ private SegmentIdWithShardSpec allocate( lockGranularity, null ); - return action.perform(task, taskActionTestKit.getTaskActionToolbox()); + + try { + if (useBatch) { + return action.performAsync(task, taskActionTestKit.getTaskActionToolbox()) + .get(5, TimeUnit.SECONDS); + } else { + return action.perform(task, taskActionTestKit.getTaskActionToolbox()); + } + } + catch (ExecutionException e) { + return null; + } + catch (Exception e) { + throw new RuntimeException(e); + } } private void assertSameIdentifier(final SegmentIdWithShardSpec expected, final SegmentIdWithShardSpec actual) { Assert.assertEquals(expected, actual); - Assert.assertEquals(expected.getShardSpec().getPartitionNum(), actual.getShardSpec().getPartitionNum()); - Assert.assertEquals(expected.getShardSpec().getClass(), actual.getShardSpec().getClass()); - - if (expected.getShardSpec().getClass() == NumberedShardSpec.class - && actual.getShardSpec().getClass() == NumberedShardSpec.class) { - Assert.assertEquals(expected.getShardSpec().getNumCorePartitions(), actual.getShardSpec().getNumCorePartitions()); - } else if (expected.getShardSpec().getClass() == LinearShardSpec.class - && actual.getShardSpec().getClass() == LinearShardSpec.class) { - // do nothing - } + Assert.assertEquals(expected.getShardSpec(), actual.getShardSpec()); } } diff --git a/indexing-service/src/test/java/org/apache/druid/indexing/common/actions/SegmentAllocationQueueTest.java b/indexing-service/src/test/java/org/apache/druid/indexing/common/actions/SegmentAllocationQueueTest.java new file mode 100644 index 000000000000..536e9ffac24c --- /dev/null +++ b/indexing-service/src/test/java/org/apache/druid/indexing/common/actions/SegmentAllocationQueueTest.java @@ -0,0 +1,349 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.indexing.common.actions; + +import org.apache.druid.indexing.common.LockGranularity; +import org.apache.druid.indexing.common.TaskLockType; +import org.apache.druid.indexing.common.task.NoopTask; +import org.apache.druid.indexing.common.task.Task; +import org.apache.druid.indexing.overlord.config.TaskLockConfig; +import org.apache.druid.java.util.common.ISE; +import org.apache.druid.java.util.common.granularity.Granularities; +import org.apache.druid.java.util.metrics.StubServiceEmitter; +import org.apache.druid.segment.realtime.appenderator.SegmentIdWithShardSpec; +import org.apache.druid.server.coordinator.simulate.BlockingExecutorService; +import org.apache.druid.server.coordinator.simulate.WrappingScheduledExecutorService; +import org.junit.After; +import org.junit.Assert; +import org.junit.Before; +import org.junit.Rule; +import org.junit.Test; + +import java.util.ArrayList; +import java.util.List; +import java.util.concurrent.ExecutionException; +import java.util.concurrent.Future; +import java.util.concurrent.TimeUnit; +import java.util.concurrent.TimeoutException; + +public class SegmentAllocationQueueTest +{ + @Rule + public TaskActionTestKit taskActionTestKit = new TaskActionTestKit(); + + private static final String DS_WIKI = "wiki"; + private static final String DS_KOALA = "koala"; + + private SegmentAllocationQueue allocationQueue; + + private StubServiceEmitter emitter; + private BlockingExecutorService executor; + + @Before + public void setUp() + { + executor = new BlockingExecutorService("alloc-test-exec"); + emitter = new StubServiceEmitter("overlord", "alloc-test"); + + final TaskLockConfig lockConfig = new TaskLockConfig() + { + @Override + public boolean isBatchSegmentAllocation() + { + return true; + } + + @Override + public long getBatchAllocationMaxWaitTime() + { + return 0; + } + }; + + allocationQueue = new SegmentAllocationQueue( + taskActionTestKit.getTaskLockbox(), + lockConfig, + taskActionTestKit.getMetadataStorageCoordinator(), + emitter, + (corePoolSize, nameFormat) + -> new WrappingScheduledExecutorService(nameFormat, executor, false) + ); + allocationQueue.start(); + allocationQueue.becomeLeader(); + } + + @After + public void tearDown() + { + if (allocationQueue != null) { + allocationQueue.stop(); + } + if (executor != null) { + executor.shutdownNow(); + } + emitter.flush(); + } + + @Test + public void testBatchWithMultipleTimestamps() + { + verifyAllocationWithBatching( + allocateRequest().forTask(createTask(DS_WIKI, "group_1")) + .forTimestamp("2022-01-01T01:00:00") + .withSegmentGranularity(Granularities.DAY) + .withQueryGranularity(Granularities.SECOND) + .withLockGranularity(LockGranularity.TIME_CHUNK) + .withSequenceName("seq_1") + .build(), + allocateRequest().forTask(createTask(DS_WIKI, "group_1")) + .forTimestamp("2022-01-01T02:00:00") + .withSegmentGranularity(Granularities.DAY) + .withQueryGranularity(Granularities.SECOND) + .withLockGranularity(LockGranularity.TIME_CHUNK) + .withSequenceName("seq_2") + .build(), + true + ); + } + + @Test + public void testBatchWithExclusiveLocks() + { + verifyAllocationWithBatching( + allocateRequest().forTask(createTask(DS_WIKI, "group_1")) + .withTaskLockType(TaskLockType.EXCLUSIVE).build(), + allocateRequest().forTask(createTask(DS_WIKI, "group_1")) + .withTaskLockType(TaskLockType.EXCLUSIVE).build(), + true + ); + } + + @Test + public void testBatchWithSharedLocks() + { + verifyAllocationWithBatching( + allocateRequest().forTask(createTask(DS_WIKI, "group_1")) + .withTaskLockType(TaskLockType.SHARED).build(), + allocateRequest().forTask(createTask(DS_WIKI, "group_1")) + .withTaskLockType(TaskLockType.SHARED).build(), + true + ); + } + + @Test + public void testBatchWithMultipleQueryGranularities() + { + verifyAllocationWithBatching( + allocateRequest().forTask(createTask(DS_WIKI, "group_1")) + .withQueryGranularity(Granularities.SECOND).build(), + allocateRequest().forTask(createTask(DS_WIKI, "group_1")) + .withQueryGranularity(Granularities.MINUTE).build(), + true + ); + } + + @Test + public void testMultipleDatasourcesCannotBatch() + { + verifyAllocationWithBatching( + allocateRequest().forTask(createTask(DS_WIKI, "group_1")).build(), + allocateRequest().forTask(createTask(DS_KOALA, "group_1")).build(), + false + ); + } + + @Test + public void testMultipleGroupIdsCannotBatch() + { + verifyAllocationWithBatching( + allocateRequest().forTask(createTask(DS_WIKI, "group_2")).build(), + allocateRequest().forTask(createTask(DS_WIKI, "group_3")).build(), + false + ); + } + + @Test + public void testMultipleLockGranularitiesCannotBatch() + { + verifyAllocationWithBatching( + allocateRequest().forTask(createTask(DS_WIKI, "group_1")) + .withLockGranularity(LockGranularity.TIME_CHUNK).build(), + allocateRequest().forTask(createTask(DS_WIKI, "group_1")) + .withLockGranularity(LockGranularity.SEGMENT).build(), + false + ); + } + + @Test + public void testMultipleAllocateIntervalsCannotBatch() + { + verifyAllocationWithBatching( + allocateRequest().forTask(createTask(DS_WIKI, "group_1")) + .forTimestamp("2022-01-01") + .withSegmentGranularity(Granularities.DAY).build(), + allocateRequest().forTask(createTask(DS_WIKI, "group_1")) + .forTimestamp("2022-01-02") + .withSegmentGranularity(Granularities.DAY).build(), + false + ); + } + + @Test + public void testConflictingPendingSegment() + { + SegmentAllocateRequest hourSegmentRequest = + allocateRequest().forTask(createTask(DS_WIKI, "group_1")) + .withSegmentGranularity(Granularities.HOUR) + .build(); + Future hourSegmentFuture = allocationQueue.add(hourSegmentRequest); + + SegmentAllocateRequest halfHourSegmentRequest = + allocateRequest().forTask(createTask(DS_WIKI, "group_1")) + .withSegmentGranularity(Granularities.THIRTY_MINUTE) + .build(); + Future halfHourSegmentFuture = allocationQueue.add(halfHourSegmentRequest); + + executor.finishNextPendingTask(); + + Assert.assertNotNull(getSegmentId(hourSegmentFuture)); + Throwable t = Assert.assertThrows(ISE.class, () -> getSegmentId(halfHourSegmentFuture)); + Assert.assertEquals("Storage coordinator could not allocate segment.", t.getMessage()); + } + + @Test + public void testFullAllocationQueue() + { + for (int i = 0; i < 2000; ++i) { + SegmentAllocateRequest request = + allocateRequest().forTask(createTask(DS_WIKI, "group_" + i)).build(); + allocationQueue.add(request); + } + + SegmentAllocateRequest request = + allocateRequest().forTask(createTask(DS_WIKI, "next_group")).build(); + Future future = allocationQueue.add(request); + + // Verify that the future is already complete and segment allocation has failed + Throwable t = Assert.assertThrows(ISE.class, () -> getSegmentId(future)); + Assert.assertEquals( + "Segment allocation queue is full. Check the metric `task/action/batch/runTime` " + + "to determine if metadata operations are slow.", + t.getMessage() + ); + } + + @Test + public void testMultipleRequestsForSameSegment() + { + final List> segmentFutures = new ArrayList<>(); + for (int i = 0; i < 10; ++i) { + SegmentAllocateRequest request = + allocateRequest().forTask(createTask(DS_WIKI, "group_" + i)) + .withSequenceName("sequence_1") + .withPreviousSegmentId("segment_1") + .build(); + segmentFutures.add(allocationQueue.add(request)); + } + + executor.finishNextPendingTask(); + + SegmentIdWithShardSpec segmentId1 = getSegmentId(segmentFutures.get(0)); + + for (Future future : segmentFutures) { + Assert.assertEquals(getSegmentId(future), segmentId1); + } + } + + @Test + public void testMaxWaitTime() + { + // Verify that the batch is due yet + } + + @Test + public void testRequestsFailOnLeaderChange() + { + final List> segmentFutures = new ArrayList<>(); + for (int i = 0; i < 10; ++i) { + SegmentAllocateRequest request = + allocateRequest().forTask(createTask(DS_WIKI, "group_" + i)).build(); + segmentFutures.add(allocationQueue.add(request)); + } + + allocationQueue.stopBeingLeader(); + executor.finishNextPendingTask(); + + for (Future future : segmentFutures) { + Throwable t = Assert.assertThrows(ISE.class, () -> getSegmentId(future)); + Assert.assertEquals("Cannot allocate segment if not leader", t.getMessage()); + } + } + + private void verifyAllocationWithBatching( + SegmentAllocateRequest a, + SegmentAllocateRequest b, + boolean canBatch + ) + { + Assert.assertEquals(0, allocationQueue.size()); + final Future futureA = allocationQueue.add(a); + final Future futureB = allocationQueue.add(b); + + final int expectedCount = canBatch ? 1 : 2; + Assert.assertEquals(expectedCount, allocationQueue.size()); + + executor.finishNextPendingTask(); + emitter.verifyEmitted("task/action/batch/size", expectedCount); + + Assert.assertNotNull(getSegmentId(futureA)); + Assert.assertNotNull(getSegmentId(futureB)); + } + + private SegmentIdWithShardSpec getSegmentId(Future future) + { + try { + return future.get(5, TimeUnit.SECONDS); + } + catch (ExecutionException e) { + throw new ISE(e.getCause().getMessage()); + } + catch (InterruptedException | TimeoutException e) { + throw new RuntimeException(e); + } + } + + private SegmentAllocateActionBuilder allocateRequest() + { + return new SegmentAllocateActionBuilder() + .forDatasource(DS_WIKI) + .forTimestamp("2022-01-01") + .withLockGranularity(LockGranularity.TIME_CHUNK) + .withTaskLockType(TaskLockType.SHARED) + .withQueryGranularity(Granularities.SECOND) + .withSegmentGranularity(Granularities.HOUR); + } + + private Task createTask(String datasource, String groupId) + { + Task task = new NoopTask(null, groupId, datasource, 0, 0, null, null, null); + taskActionTestKit.getTaskLockbox().add(task); + return task; + } +} diff --git a/indexing-service/src/test/java/org/apache/druid/indexing/common/actions/TaskActionTestKit.java b/indexing-service/src/test/java/org/apache/druid/indexing/common/actions/TaskActionTestKit.java index 2d6b22732a69..189d3fe8779c 100644 --- a/indexing-service/src/test/java/org/apache/druid/indexing/common/actions/TaskActionTestKit.java +++ b/indexing-service/src/test/java/org/apache/druid/indexing/common/actions/TaskActionTestKit.java @@ -27,7 +27,10 @@ import org.apache.druid.indexing.overlord.IndexerMetadataStorageCoordinator; import org.apache.druid.indexing.overlord.TaskLockbox; import org.apache.druid.indexing.overlord.TaskStorage; +import org.apache.druid.indexing.overlord.config.TaskLockConfig; import org.apache.druid.indexing.overlord.supervisor.SupervisorManager; +import org.apache.druid.java.util.common.concurrent.ScheduledExecutors; +import org.apache.druid.java.util.emitter.service.ServiceEmitter; import org.apache.druid.metadata.IndexerSQLMetadataStorageCoordinator; import org.apache.druid.metadata.MetadataStorageConnectorConfig; import org.apache.druid.metadata.MetadataStorageTablesConfig; @@ -99,11 +102,34 @@ public int getSqlMetadataMaxRetry() Suppliers.ofInstance(metadataStorageTablesConfig), testDerbyConnector ); + final ServiceEmitter noopEmitter = new NoopServiceEmitter(); + final TaskLockConfig taskLockConfig = new TaskLockConfig() + { + @Override + public boolean isBatchSegmentAllocation() + { + return true; + } + + @Override + public long getBatchAllocationMaxWaitTime() + { + return 10L; + } + }; + taskActionToolbox = new TaskActionToolbox( taskLockbox, taskStorage, metadataStorageCoordinator, - new NoopServiceEmitter(), + new SegmentAllocationQueue( + taskLockbox, + taskLockConfig, + metadataStorageCoordinator, + noopEmitter, + ScheduledExecutors::fixed + ), + noopEmitter, EasyMock.createMock(SupervisorManager.class), objectMapper ); diff --git a/indexing-service/src/test/java/org/apache/druid/indexing/overlord/http/OverlordTest.java b/indexing-service/src/test/java/org/apache/druid/indexing/overlord/http/OverlordTest.java index 6f9a01c25b0b..3ada645ff88a 100644 --- a/indexing-service/src/test/java/org/apache/druid/indexing/overlord/http/OverlordTest.java +++ b/indexing-service/src/test/java/org/apache/druid/indexing/overlord/http/OverlordTest.java @@ -40,6 +40,7 @@ import org.apache.druid.indexer.TaskStatusPlus; import org.apache.druid.indexing.common.TaskLock; import org.apache.druid.indexing.common.TimeChunkLock; +import org.apache.druid.indexing.common.actions.SegmentAllocationQueue; import org.apache.druid.indexing.common.actions.TaskActionClientFactory; import org.apache.druid.indexing.common.config.TaskStorageConfig; import org.apache.druid.indexing.common.task.NoopTask; @@ -226,7 +227,8 @@ public MockTaskRunner get() serviceEmitter, supervisorManager, EasyMock.createNiceMock(OverlordHelperManager.class), - new TestDruidLeaderSelector() + new TestDruidLeaderSelector(), + EasyMock.createNiceMock(SegmentAllocationQueue.class) ); EmittingLogger.registerEmitter(serviceEmitter); } diff --git a/indexing-service/src/test/java/org/apache/druid/indexing/test/TestIndexerMetadataStorageCoordinator.java b/indexing-service/src/test/java/org/apache/druid/indexing/test/TestIndexerMetadataStorageCoordinator.java index 2dcff7694678..d64bd1d22263 100644 --- a/indexing-service/src/test/java/org/apache/druid/indexing/test/TestIndexerMetadataStorageCoordinator.java +++ b/indexing-service/src/test/java/org/apache/druid/indexing/test/TestIndexerMetadataStorageCoordinator.java @@ -25,6 +25,7 @@ import com.google.common.collect.Sets; import org.apache.druid.indexing.overlord.DataSourceMetadata; import org.apache.druid.indexing.overlord.IndexerMetadataStorageCoordinator; +import org.apache.druid.indexing.overlord.SegmentCreateRequest; import org.apache.druid.indexing.overlord.SegmentPublishResult; import org.apache.druid.indexing.overlord.Segments; import org.apache.druid.jackson.DefaultObjectMapper; @@ -36,8 +37,10 @@ import javax.annotation.Nullable; import java.util.ArrayList; +import java.util.Collections; import java.util.HashSet; import java.util.List; +import java.util.Map; import java.util.Set; public class TestIndexerMetadataStorageCoordinator implements IndexerMetadataStorageCoordinator @@ -124,6 +127,17 @@ public Set announceHistoricalSegments(Set segments) return ImmutableSet.copyOf(added); } + @Override + public Map allocatePendingSegments( + String dataSource, + Interval interval, + boolean skipSegmentLineageCheck, + List requests + ) + { + return Collections.emptyMap(); + } + @Override public SegmentPublishResult announceHistoricalSegments( Set segments, diff --git a/server/src/main/java/org/apache/druid/indexing/overlord/IndexerMetadataStorageCoordinator.java b/server/src/main/java/org/apache/druid/indexing/overlord/IndexerMetadataStorageCoordinator.java index acb617f78a4c..b3c70f0cdbe9 100644 --- a/server/src/main/java/org/apache/druid/indexing/overlord/IndexerMetadataStorageCoordinator.java +++ b/server/src/main/java/org/apache/druid/indexing/overlord/IndexerMetadataStorageCoordinator.java @@ -31,6 +31,7 @@ import java.util.Collection; import java.util.Collections; import java.util.List; +import java.util.Map; import java.util.Set; /** @@ -154,6 +155,27 @@ Collection retrieveUsedSegmentsForIntervals( */ Set announceHistoricalSegments(Set segments) throws IOException; + /** + * Allocates pending segments for the given requests in the pending segments table. + * The segment id allocated for a request will not be given out again unless a + * request is made with the same {@link SegmentCreateRequest}. + * + * @param dataSource dataSource for which to allocate a segment + * @param interval interval for which to allocate a segment + * @param skipSegmentLineageCheck if true, perform lineage validation using previousSegmentId for this sequence. + * Should be set to false if replica tasks would index events in same order + * @param requests Requests for which to allocate segments. All + * the requests must share the same partition space. + * @return Map from request to allocated segment id. The map does not contain + * entries for failed requests. + */ + Map allocatePendingSegments( + String dataSource, + Interval interval, + boolean skipSegmentLineageCheck, + List requests + ); + /** * Allocate a new pending segment in the pending segments table. This segment identifier will never be given out * again, unless another call is made with the same dataSource, sequenceName, and previousSegmentId. diff --git a/server/src/main/java/org/apache/druid/indexing/overlord/SegmentCreateRequest.java b/server/src/main/java/org/apache/druid/indexing/overlord/SegmentCreateRequest.java new file mode 100644 index 000000000000..b43e46d8e7a5 --- /dev/null +++ b/server/src/main/java/org/apache/druid/indexing/overlord/SegmentCreateRequest.java @@ -0,0 +1,78 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.indexing.overlord; + +import org.apache.druid.timeline.partition.PartialShardSpec; + +/** + * Contains information used by {@link IndexerMetadataStorageCoordinator} for + * creating a new segment. + *

    + * The {@code sequenceName} and {@code previousSegmentId} fields are meant to + * make it easy for two independent ingestion tasks to produce the same series + * of segments. + */ +public class SegmentCreateRequest +{ + // DO NOT IMPLEMENT equals or hashCode for this class as each request must be + // treated as unique even if it is for the same parameters + + private final String version; + private final String sequenceName; + private final String previousSegmentId; + private final PartialShardSpec partialShardSpec; + + public SegmentCreateRequest( + String sequenceName, + String previousSegmentId, + String version, + PartialShardSpec partialShardSpec + ) + { + this.sequenceName = sequenceName; + this.previousSegmentId = previousSegmentId == null ? "" : previousSegmentId; + this.version = version; + this.partialShardSpec = partialShardSpec; + } + + public String getSequenceName() + { + return sequenceName; + } + + /** + * Non-null previous segment id. This can be used for persisting to the + * pending segments table in the metadata store. + */ + public String getPreviousSegmentId() + { + return previousSegmentId; + } + + public String getVersion() + { + return version; + } + + public PartialShardSpec getPartialShardSpec() + { + return partialShardSpec; + } +} diff --git a/server/src/main/java/org/apache/druid/metadata/IndexerSQLMetadataStorageCoordinator.java b/server/src/main/java/org/apache/druid/metadata/IndexerSQLMetadataStorageCoordinator.java index cc42d77e1b70..bb7759a8b5d6 100644 --- a/server/src/main/java/org/apache/druid/metadata/IndexerSQLMetadataStorageCoordinator.java +++ b/server/src/main/java/org/apache/druid/metadata/IndexerSQLMetadataStorageCoordinator.java @@ -28,12 +28,14 @@ import com.google.common.collect.ImmutableSet; import com.google.common.collect.Iterables; import com.google.common.collect.Lists; +import com.google.common.hash.Hasher; import com.google.common.hash.Hashing; import com.google.common.io.BaseEncoding; import com.google.inject.Inject; import org.apache.commons.lang.StringEscapeUtils; import org.apache.druid.indexing.overlord.DataSourceMetadata; import org.apache.druid.indexing.overlord.IndexerMetadataStorageCoordinator; +import org.apache.druid.indexing.overlord.SegmentCreateRequest; import org.apache.druid.indexing.overlord.SegmentPublishResult; import org.apache.druid.indexing.overlord.Segments; import org.apache.druid.java.util.common.DateTimes; @@ -75,10 +77,13 @@ import javax.validation.constraints.NotNull; import java.io.IOException; import java.sql.ResultSet; +import java.sql.SQLException; import java.util.ArrayList; import java.util.Arrays; import java.util.Collection; import java.util.Collections; +import java.util.Comparator; +import java.util.HashMap; import java.util.HashSet; import java.util.List; import java.util.Map; @@ -221,6 +226,10 @@ public int markSegmentsAsUnusedWithinInterval(String dataSource, Interval interv return numSegmentsMarkedUnused; } + /** + * Fetches all the pending segments, whose interval overlaps with the given + * search interval from the metadata store. + */ private Set getPendingSegmentsForIntervalWithHandle( final Handle handle, final String dataSource, @@ -481,6 +490,23 @@ public int getSqlMetadataMaxRetry() return SQLMetadataConnector.DEFAULT_MAX_TRIES; } + @Override + public Map allocatePendingSegments( + String dataSource, + Interval allocateInterval, + boolean skipSegmentLineageCheck, + List requests + ) + { + Preconditions.checkNotNull(dataSource, "dataSource"); + Preconditions.checkNotNull(allocateInterval, "interval"); + + final Interval interval = allocateInterval.withChronology(ISOChronology.getInstanceUTC()); + return connector.retryWithHandle( + handle -> allocatePendingSegments(handle, dataSource, interval, skipSegmentLineageCheck, requests) + ); + } + @Override public SegmentIdWithShardSpec allocatePendingSegment( final String dataSource, @@ -599,6 +625,81 @@ private SegmentIdWithShardSpec allocatePendingSegmentWithSegmentLineageCheck( return newIdentifier; } + private Map allocatePendingSegments( + final Handle handle, + final String dataSource, + final Interval interval, + final boolean skipSegmentLineageCheck, + final List requests + ) throws IOException + { + final Map existingSegmentIds; + if (skipSegmentLineageCheck) { + existingSegmentIds = getExistingSegmentIdsSkipLineageCheck(handle, dataSource, interval, requests); + } else { + existingSegmentIds = getExistingSegmentIdsWithLineageCheck(handle, dataSource, interval, requests); + } + + // For every request see if a segment id already exists + final Map allocatedSegmentIds = new HashMap<>(); + final List requestsForNewSegments = new ArrayList<>(); + for (SegmentCreateRequest request : requests) { + CheckExistingSegmentIdResult existingSegmentId = existingSegmentIds.get(request); + if (existingSegmentId == null || !existingSegmentId.found) { + requestsForNewSegments.add(request); + } else if (existingSegmentId.segmentIdentifier != null) { + log.info("Found valid existing segment [%s] for request.", existingSegmentId.segmentIdentifier); + allocatedSegmentIds.put(request, existingSegmentId.segmentIdentifier); + } else { + log.info("Found clashing existing segment [%s] for request.", existingSegmentId); + } + } + + // For each of the remaining requests, create a new segment + final Map createdSegments = + createNewSegments(handle, dataSource, interval, skipSegmentLineageCheck, requestsForNewSegments); + + // SELECT -> INSERT can fail due to races; callers must be prepared to retry. + // Avoiding ON DUPLICATE KEY since it's not portable. + // Avoiding try/catch since it may cause inadvertent transaction-splitting. + + // UNIQUE key for the row, ensuring we don't have more than one segment per sequence per interval. + // Using a single column instead of (sequence_name, sequence_prev_id) as some MySQL storage engines + // have difficulty with large unique keys (see https://github.com/apache/druid/issues/2319) + insertPendingSegmentsIntoMetastore( + handle, + createdSegments, + dataSource, + interval, + skipSegmentLineageCheck + ); + + allocatedSegmentIds.putAll(createdSegments); + return allocatedSegmentIds; + } + + @SuppressWarnings("UnstableApiUsage") + private String getSequenceNameAndPrevIdSha( + SegmentCreateRequest request, + Interval interval, + boolean skipSegmentLineageCheck + ) + { + final Hasher hasher = Hashing.sha1().newHasher() + .putBytes(StringUtils.toUtf8(request.getSequenceName())) + .putByte((byte) 0xff); + if (skipSegmentLineageCheck) { + hasher + .putLong(interval.getStartMillis()) + .putLong(interval.getEndMillis()); + } else { + hasher + .putBytes(StringUtils.toUtf8(request.getPreviousSegmentId())); + } + + return BaseEncoding.base16().encode(hasher.hash().asBytes()); + } + @Nullable private SegmentIdWithShardSpec allocatePendingSegment( final Handle handle, @@ -631,7 +732,6 @@ private SegmentIdWithShardSpec allocatePendingSegment( ); if (result.found) { - // The found existing segment identifier can be null if its interval doesn't match with the given interval return result.segmentIdentifier; } @@ -672,6 +772,95 @@ private SegmentIdWithShardSpec allocatePendingSegment( return newIdentifier; } + /** + * Returns a map from sequenceName to segment id. + */ + private Map getExistingSegmentIdsSkipLineageCheck( + Handle handle, + String dataSource, + Interval interval, + List requests + ) throws IOException + { + final Query> query = handle + .createQuery( + StringUtils.format( + "SELECT sequence_name, payload " + + "FROM %s WHERE " + + "dataSource = :dataSource AND " + + "start = :start AND " + + "%2$send%2$s = :end", + dbTables.getPendingSegmentsTable(), + connector.getQuoteString() + ) + ) + .bind("dataSource", dataSource) + .bind("start", interval.getStart().toString()) + .bind("end", interval.getEnd().toString()); + + final ResultIterator dbSegments = query + .map((index, r, ctx) -> PendingSegmentsRecord.fromResultSet(r)) + .iterator(); + + // Map from sequenceName to segment id + final Map sequenceToSegmentId = new HashMap<>(); + while (dbSegments.hasNext()) { + final PendingSegmentsRecord record = dbSegments.next(); + final SegmentIdWithShardSpec segmentId = + jsonMapper.readValue(record.getPayload(), SegmentIdWithShardSpec.class); + sequenceToSegmentId.put(record.getSequenceName(), segmentId); + } + + final Map requestToResult = new HashMap<>(); + for (SegmentCreateRequest request : requests) { + SegmentIdWithShardSpec segmentId = sequenceToSegmentId.get(request.getSequenceName()); + requestToResult.put(request, new CheckExistingSegmentIdResult(segmentId != null, segmentId)); + } + + return requestToResult; + } + + /** + * Returns a map from sequenceName to segment id. + */ + private Map getExistingSegmentIdsWithLineageCheck( + Handle handle, + String dataSource, + Interval interval, + List requests + ) throws IOException + { + // This cannot be batched because there doesn't seem to be a clean option: + // 1. WHERE must have sequence_name and sequence_prev_id but not start or end. + // (sequence columns are used to find the matching segment whereas start and + // end are used to determine if the found segment is valid or not) + // 2. IN filters on sequence_name and sequence_prev_id might perform worse than individual SELECTs? + // 3. IN filter on sequence_name alone might be a feasible option worth evaluating + final String sql = StringUtils.format( + "SELECT payload FROM %s WHERE " + + "dataSource = :dataSource AND " + + "sequence_name = :sequence_name AND " + + "sequence_prev_id = :sequence_prev_id", + dbTables.getPendingSegmentsTable() + ); + + final Map requestToResult = new HashMap<>(); + for (SegmentCreateRequest request : requests) { + CheckExistingSegmentIdResult result = checkAndGetExistingSegmentId( + handle.createQuery(sql) + .bind("dataSource", dataSource) + .bind("sequence_name", request.getSequenceName()) + .bind("sequence_prev_id", request.getPreviousSegmentId()), + interval, + request.getSequenceName(), + request.getPreviousSegmentId() + ); + requestToResult.put(request, result); + } + + return requestToResult; + } + private CheckExistingSegmentIdResult checkAndGetExistingSegmentId( final Query> query, final Interval interval, @@ -686,50 +875,36 @@ private CheckExistingSegmentIdResult checkAndGetExistingSegmentId( } final List existingBytes = boundQuery.map(ByteArrayMapper.FIRST).list(); - if (!existingBytes.isEmpty()) { + if (existingBytes.isEmpty()) { + return new CheckExistingSegmentIdResult(false, null); + } else { final SegmentIdWithShardSpec existingIdentifier = jsonMapper.readValue( Iterables.getOnlyElement(existingBytes), SegmentIdWithShardSpec.class ); - if (existingIdentifier.getInterval().getStartMillis() == interval.getStartMillis() - && existingIdentifier.getInterval().getEndMillis() == interval.getEndMillis()) { - if (previousSegmentId == null) { - log.info("Found existing pending segment [%s] for sequence[%s] in DB", existingIdentifier, sequenceName); - } else { - log.info( - "Found existing pending segment [%s] for sequence[%s] (previous = [%s]) in DB", - existingIdentifier, - sequenceName, - previousSegmentId - ); - } + if (existingIdentifier.getInterval().isEqual(interval)) { + log.info( + "Found existing pending segment [%s] for sequence[%s] (previous = [%s]) in DB", + existingIdentifier, + sequenceName, + previousSegmentId + ); return new CheckExistingSegmentIdResult(true, existingIdentifier); } else { - if (previousSegmentId == null) { - log.warn( - "Cannot use existing pending segment [%s] for sequence[%s] in DB, " - + "does not match requested interval[%s]", - existingIdentifier, - sequenceName, - interval - ); - } else { - log.warn( - "Cannot use existing pending segment [%s] for sequence[%s] (previous = [%s]) in DB, " - + "does not match requested interval[%s]", - existingIdentifier, - sequenceName, - previousSegmentId, - interval - ); - } + log.warn( + "Cannot use existing pending segment [%s] for sequence[%s] (previous = [%s]) in DB, " + + "does not match requested interval[%s]", + existingIdentifier, + sequenceName, + previousSegmentId, + interval + ); return new CheckExistingSegmentIdResult(true, null); } } - return new CheckExistingSegmentIdResult(false, null); } private static class CheckExistingSegmentIdResult @@ -745,6 +920,48 @@ private static class CheckExistingSegmentIdResult } } + private void insertPendingSegmentsIntoMetastore( + Handle handle, + Map createdSegments, + String dataSource, + Interval interval, + boolean skipSegmentLineageCheck + ) throws JsonProcessingException + { + final PreparedBatch insertBatch = handle.prepareBatch( + StringUtils.format( + "INSERT INTO %1$s (id, dataSource, created_date, start, %2$send%2$s, sequence_name, sequence_prev_id, " + + "sequence_name_prev_id_sha1, payload) " + + "VALUES (:id, :dataSource, :created_date, :start, :end, :sequence_name, :sequence_prev_id, " + + ":sequence_name_prev_id_sha1, :payload)", + dbTables.getPendingSegmentsTable(), + connector.getQuoteString() + )); + + // Deduplicate the segment ids by inverting the map + Map segmentIdToRequest = new HashMap<>(); + createdSegments.forEach((request, segmentId) -> segmentIdToRequest.put(segmentId, request)); + + for (Map.Entry entry : segmentIdToRequest.entrySet()) { + final SegmentCreateRequest request = entry.getValue(); + final SegmentIdWithShardSpec segmentId = entry.getKey(); + insertBatch.add() + .bind("id", segmentId.toString()) + .bind("dataSource", dataSource) + .bind("created_date", DateTimes.nowUtc().toString()) + .bind("start", interval.getStart().toString()) + .bind("end", interval.getEnd().toString()) + .bind("sequence_name", request.getSequenceName()) + .bind("sequence_prev_id", request.getPreviousSegmentId()) + .bind( + "sequence_name_prev_id_sha1", + getSequenceNameAndPrevIdSha(request, interval, skipSegmentLineageCheck) + ) + .bind("payload", jsonMapper.writeValueAsBytes(segmentId)); + } + insertBatch.execute(); + } + private void insertPendingSegmentIntoMetastore( Handle handle, SegmentIdWithShardSpec newIdentifier, @@ -777,6 +994,204 @@ private void insertPendingSegmentIntoMetastore( .execute(); } + private Map createNewSegments( + Handle handle, + String dataSource, + Interval interval, + boolean skipSegmentLineageCheck, + List requests + ) throws IOException + { + if (requests.isEmpty()) { + return Collections.emptyMap(); + } + + // Get the time chunk and associated data segments for the given interval, if any + final List> existingChunks = + getTimelineForIntervalsWithHandle(handle, dataSource, Collections.singletonList(interval)) + .lookup(interval); + + if (existingChunks.size() > 1) { + // Not possible to expand more than one chunk with a single segment. + log.warn( + "Cannot allocate new segments for dataSource[%s], interval[%s]: already have [%,d] chunks.", + dataSource, + interval, + existingChunks.size() + ); + return Collections.emptyMap(); + } + + // Shard spec of any of the requests (as they are all compatible) can be used to + // identify existing shard specs that share partition space with the requested ones. + final PartialShardSpec partialShardSpec = requests.get(0).getPartialShardSpec(); + + // max partitionId of published data segments which share the same partition space. + SegmentIdWithShardSpec committedMaxId = null; + + @Nullable + final String versionOfExistingChunk; + if (existingChunks.isEmpty()) { + versionOfExistingChunk = null; + } else { + TimelineObjectHolder existingHolder = Iterables.getOnlyElement(existingChunks); + versionOfExistingChunk = existingHolder.getVersion(); + + // Don't use the stream API for performance. + for (DataSegment segment : FluentIterable + .from(existingHolder.getObject()) + .transform(PartitionChunk::getObject) + // Here we check only the segments of the shardSpec which shares the same partition space with the given + // partialShardSpec. Note that OverwriteShardSpec doesn't share the partition space with others. + // See PartitionIds. + .filter(segment -> segment.getShardSpec().sharePartitionSpace(partialShardSpec))) { + if (committedMaxId == null + || committedMaxId.getShardSpec().getPartitionNum() < segment.getShardSpec().getPartitionNum()) { + committedMaxId = SegmentIdWithShardSpec.fromDataSegment(segment); + } + } + } + + + // Fetch the pending segments for this interval to determine max partitionId + // across all shard specs (published + pending). + // A pending segment having a higher partitionId must also be considered + // to avoid clashes when inserting the pending segment created here. + final Set pendingSegments = + getPendingSegmentsForIntervalWithHandle(handle, dataSource, interval); + + final Map createdSegments = new HashMap<>(); + final Map sequenceHashToSegment = new HashMap<>(); + + for (SegmentCreateRequest request : requests) { + // Check if the required segment has already been created in this batch + final String sequenceHash = getSequenceNameAndPrevIdSha(request, interval, skipSegmentLineageCheck); + + final SegmentIdWithShardSpec createdSegment; + if (sequenceHashToSegment.containsKey(sequenceHash)) { + createdSegment = sequenceHashToSegment.get(sequenceHash); + } else { + createdSegment = createNewSegment( + request, + dataSource, + interval, + versionOfExistingChunk, + committedMaxId, + pendingSegments + ); + + // Add to pendingSegments to consider for partitionId + if (createdSegment != null) { + pendingSegments.add(createdSegment); + sequenceHashToSegment.put(sequenceHash, createdSegment); + log.info("Created new segment [%s]", createdSegment); + } + } + + if (createdSegment != null) { + createdSegments.put(request, createdSegment); + } + } + + log.info("Created [%d] new segments for [%d] allocate requests.", sequenceHashToSegment.size(), requests.size()); + return createdSegments; + } + + private SegmentIdWithShardSpec createNewSegment( + SegmentCreateRequest request, + String dataSource, + Interval interval, + String versionOfExistingChunk, + SegmentIdWithShardSpec committedMaxId, + Set pendingSegments + ) + { + final PartialShardSpec partialShardSpec = request.getPartialShardSpec(); + final String existingVersion = request.getVersion(); + + // Include the committedMaxId while computing the overallMaxId + if (committedMaxId != null) { + pendingSegments.add(committedMaxId); + } + + // If there is an existing chunk, find the max id with the same version as the existing chunk. + // There may still be a pending segment with a higher version (but no corresponding used segments) + // which may generate a clash with an existing segment once the new id is generated + final SegmentIdWithShardSpec overallMaxId = + pendingSegments.stream() + .filter(id -> id.getShardSpec().sharePartitionSpace(partialShardSpec)) + .filter(id -> versionOfExistingChunk == null + || id.getVersion().equals(versionOfExistingChunk)) + .max(Comparator.comparing(SegmentIdWithShardSpec::getVersion) + .thenComparing(id -> id.getShardSpec().getPartitionNum())) + .orElse(null); + + // Determine the version of the new segment + final String newSegmentVersion; + if (versionOfExistingChunk != null) { + newSegmentVersion = versionOfExistingChunk; + } else if (overallMaxId != null) { + newSegmentVersion = overallMaxId.getVersion(); + } else { + // this is the first segment for this interval + newSegmentVersion = null; + } + + if (overallMaxId == null) { + // When appending segments, null overallMaxId means that we are allocating the very initial + // segment for this time chunk. + // This code is executed when the Overlord coordinates segment allocation, which is either you append segments + // or you use segment lock. Since the core partitions set is not determined for appended segments, we set + // it 0. When you use segment lock, the core partitions set doesn't work with it. We simply set it 0 so that the + // OvershadowableManager handles the atomic segment update. + final int newPartitionId = partialShardSpec.useNonRootGenerationPartitionSpace() + ? PartitionIds.NON_ROOT_GEN_START_PARTITION_ID + : PartitionIds.ROOT_GEN_START_PARTITION_ID; + + String version = newSegmentVersion == null ? existingVersion : newSegmentVersion; + return new SegmentIdWithShardSpec( + dataSource, + interval, + version, + partialShardSpec.complete(jsonMapper, newPartitionId, 0) + ); + } else if (!overallMaxId.getInterval().equals(interval) + || overallMaxId.getVersion().compareTo(existingVersion) > 0) { + log.warn( + "Cannot allocate new segment for dataSource[%s], interval[%s], existingVersion[%s]: conflicting segment[%s].", + dataSource, + interval, + existingVersion, + overallMaxId + ); + return null; + } else if (committedMaxId != null + && committedMaxId.getShardSpec().getNumCorePartitions() + == SingleDimensionShardSpec.UNKNOWN_NUM_CORE_PARTITIONS) { + log.warn( + "Cannot allocate new segment because of unknown core partition size of segment[%s], shardSpec[%s]", + committedMaxId, + committedMaxId.getShardSpec() + ); + return null; + } else { + // The number of core partitions must always be chosen from the set of used segments in the SegmentTimeline. + // When the core partitions have been dropped, using pending segments may lead to an incorrect state + // where the chunk is believed to have core partitions and queries results are incorrect. + + return new SegmentIdWithShardSpec( + dataSource, + interval, + Preconditions.checkNotNull(newSegmentVersion, "newSegmentVersion"), + partialShardSpec.complete( + jsonMapper, + overallMaxId.getShardSpec().getPartitionNum() + 1, + committedMaxId == null ? 0 : committedMaxId.getShardSpec().getNumCorePartitions() + ) + ); + } + } + /** * This function creates a new segment for the given datasource/interval/etc. A critical * aspect of the creation is to make sure that the new version & new partition number will make @@ -818,15 +1233,18 @@ private SegmentIdWithShardSpec createNewSegment( return null; } else { - // max partitionId of the committed shardSpecs which share the same partition space. + // max partitionId of published data segments which share the same partition space. SegmentIdWithShardSpec committedMaxId = null; - // max partitionId of the all shardSpecs including the pending ones which share the same partition space. - SegmentIdWithShardSpec overallMaxId; - if (!existingChunks.isEmpty()) { + @Nullable + final String versionOfExistingChunk; + if (existingChunks.isEmpty()) { + versionOfExistingChunk = null; + } else { TimelineObjectHolder existingHolder = Iterables.getOnlyElement(existingChunks); + versionOfExistingChunk = existingHolder.getVersion(); - //noinspection ConstantConditions + // Don't use the stream API for performance. for (DataSegment segment : FluentIterable .from(existingHolder.getObject()) .transform(PartitionChunk::getObject) @@ -834,8 +1252,6 @@ private SegmentIdWithShardSpec createNewSegment( // partialShardSpec. Note that OverwriteShardSpec doesn't share the partition space with others. // See PartitionIds. .filter(segment -> segment.getShardSpec().sharePartitionSpace(partialShardSpec))) { - // Don't use the stream API for performance. - // Note that this will compute the max id of existing, visible, data segments in the time chunk: if (committedMaxId == null || committedMaxId.getShardSpec().getPartitionNum() < segment.getShardSpec().getPartitionNum()) { committedMaxId = SegmentIdWithShardSpec.fromDataSegment(segment); @@ -843,63 +1259,41 @@ private SegmentIdWithShardSpec createNewSegment( } } - // Get the version of the existing chunk, we might need it in some of the cases below - // to compute the new identifier's version - @Nullable - final String versionOfExistingChunk; - if (!existingChunks.isEmpty()) { - // remember only one chunk possible for given interval so get the first & only one - versionOfExistingChunk = existingChunks.get(0).getVersion(); - } else { - versionOfExistingChunk = null; - } - - // next, we need to enrich the overallMaxId computed with committed segments with the information of the pending segments - // it is possible that a pending segment has a higher id in which case we need that, it will work, - // and it will avoid clashes when inserting the new pending segment later in the caller of this method + // Fetch the pending segments for this interval to determine max partitionId + // across all shard specs (published + pending). + // A pending segment having a higher partitionId must also be considered + // to avoid clashes when inserting the pending segment created here. final Set pendings = getPendingSegmentsForIntervalWithHandle( handle, dataSource, interval ); - // Make sure we add the committed max id we obtained from the segments table: if (committedMaxId != null) { pendings.add(committedMaxId); } - // Now compute the overallMaxId with all the information: pendings + segments: - // The versionOfExistingChunks filter is ensure that we pick the max id with the version of the existing chunk - // in the case that there may be a pending segment with a higher version but no corresponding used segments + + // If there is an existing chunk, find the max id with the same version as the existing chunk. + // There may still be a pending segment with a higher version (but no corresponding used segments) // which may generate a clash with an existing segment once the new id is generated + final SegmentIdWithShardSpec overallMaxId; overallMaxId = pendings.stream() .filter(id -> id.getShardSpec().sharePartitionSpace(partialShardSpec)) - .filter(id -> versionOfExistingChunk == null || id.getVersion() - .equals(versionOfExistingChunk)) - .max((id1, id2) -> { - final int versionCompare = id1.getVersion().compareTo(id2.getVersion()); - if (versionCompare != 0) { - return versionCompare; - } else { - return Integer.compare( - id1.getShardSpec().getPartitionNum(), - id2.getShardSpec().getPartitionNum() - ); - } - }) + .filter(id -> versionOfExistingChunk == null + || id.getVersion().equals(versionOfExistingChunk)) + .max(Comparator.comparing(SegmentIdWithShardSpec::getVersion) + .thenComparing(id -> id.getShardSpec().getPartitionNum())) .orElse(null); - // The following code attempts to compute the new version, if this - // new version is not null at the end of next block then it will be - // used as the new version in the case for initial or appended segment + + // Determine the version of the new segment final String newSegmentVersion; if (versionOfExistingChunk != null) { - // segment version overrides, so pick that now that we know it exists newSegmentVersion = versionOfExistingChunk; - } else if (!pendings.isEmpty() && overallMaxId != null) { - // there is no visible segments in the time chunk, so pick the max id of pendings, as computed above + } else if (overallMaxId != null) { newSegmentVersion = overallMaxId.getVersion(); } else { - // no segments, no pendings, so this must be the very first segment created for this interval + // this is the first segment for this interval newSegmentVersion = null; } @@ -940,7 +1334,7 @@ private SegmentIdWithShardSpec createNewSegment( ); return null; } else { - // The number of core partitions must always be chosen from the set of used segments in the VersionedIntervalTimeline. + // The number of core partitions must always be chosen from the set of used segments in the SegmentTimeline. // When the core partitions have been dropped, using pending segments may lead to an incorrect state // where the chunk is believed to have core partitions and queries results are incorrect. @@ -1472,4 +1866,46 @@ public int removeDataSourceMetadataOlderThan(long timestamp, @NotNull Set + *

  1. sequence_name
    2. + *
  2. payload
    3. + * + */ + static PendingSegmentsRecord fromResultSet(ResultSet resultSet) + { + try { + return new PendingSegmentsRecord( + resultSet.getString(1), + resultSet.getBytes(2) + ); + } + catch (SQLException e) { + throw new RuntimeException(e); + } + } + + PendingSegmentsRecord(String sequenceName, byte[] payload) + { + this.payload = payload; + this.sequenceName = sequenceName; + } + + public byte[] getPayload() + { + return payload; + } + + public String getSequenceName() + { + return sequenceName; + } + } + } diff --git a/server/src/test/java/org/apache/druid/indexing/overlord/SegmentCreateRequestTest.java b/server/src/test/java/org/apache/druid/indexing/overlord/SegmentCreateRequestTest.java new file mode 100644 index 000000000000..33641a8417da --- /dev/null +++ b/server/src/test/java/org/apache/druid/indexing/overlord/SegmentCreateRequestTest.java @@ -0,0 +1,46 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.indexing.overlord; + +import org.apache.druid.timeline.partition.NumberedPartialShardSpec; +import org.apache.druid.timeline.partition.PartialShardSpec; +import org.junit.Assert; +import org.junit.Test; + +public class SegmentCreateRequestTest +{ + + @Test + public void testNullPreviousSegmentId() + { + final PartialShardSpec partialShardSpec = NumberedPartialShardSpec.instance(); + SegmentCreateRequest request = new SegmentCreateRequest( + "sequence", + null, + "version", + partialShardSpec + ); + Assert.assertEquals("sequence", request.getSequenceName()); + Assert.assertEquals("", request.getPreviousSegmentId()); + Assert.assertEquals("version", request.getVersion()); + Assert.assertEquals(partialShardSpec, request.getPartialShardSpec()); + } + +} diff --git a/server/src/test/java/org/apache/druid/metadata/IndexerSQLMetadataStorageCoordinatorTest.java b/server/src/test/java/org/apache/druid/metadata/IndexerSQLMetadataStorageCoordinatorTest.java index d7000f8f4d20..f7af9611e51a 100644 --- a/server/src/test/java/org/apache/druid/metadata/IndexerSQLMetadataStorageCoordinatorTest.java +++ b/server/src/test/java/org/apache/druid/metadata/IndexerSQLMetadataStorageCoordinatorTest.java @@ -27,6 +27,7 @@ import org.apache.druid.data.input.StringTuple; import org.apache.druid.indexing.overlord.DataSourceMetadata; import org.apache.druid.indexing.overlord.ObjectMetadata; +import org.apache.druid.indexing.overlord.SegmentCreateRequest; import org.apache.druid.indexing.overlord.SegmentPublishResult; import org.apache.druid.indexing.overlord.Segments; import org.apache.druid.java.util.common.DateTimes; @@ -59,7 +60,6 @@ import org.junit.rules.ExpectedException; import org.skife.jdbi.v2.Handle; import org.skife.jdbi.v2.PreparedBatch; -import org.skife.jdbi.v2.tweak.HandleCallback; import org.skife.jdbi.v2.util.StringMapper; import java.io.IOException; @@ -371,40 +371,12 @@ private void markAllSegmentsUnused(Set segments) Assert.assertEquals( 1, (int) derbyConnector.getDBI().withHandle( - new HandleCallback() - { - @Override - public Integer withHandle(Handle handle) - { - String request = StringUtils.format( - "UPDATE %s SET used = false WHERE id = :id", - derbyConnectorRule.metadataTablesConfigSupplier().get().getSegmentsTable() - ); - return handle.createStatement(request).bind("id", segment.getId().toString()).execute(); - } - } - ) - ); - } - } - - private void markAllSegmentsUsed(Set segments) - { - for (final DataSegment segment : segments) { - Assert.assertEquals( - 1, - (int) derbyConnector.getDBI().withHandle( - new HandleCallback() - { - @Override - public Integer withHandle(Handle handle) - { - String request = StringUtils.format( - "UPDATE %s SET used = true WHERE id = :id", - derbyConnectorRule.metadataTablesConfigSupplier().get().getSegmentsTable() - ); - return handle.createStatement(request).bind("id", segment.getId().toString()).execute(); - } + handle -> { + String request = StringUtils.format( + "UPDATE %s SET used = false WHERE id = :id", + derbyConnectorRule.metadataTablesConfigSupplier().get().getSegmentsTable() + ); + return handle.createStatement(request).bind("id", segment.getId().toString()).execute(); } ) ); @@ -415,32 +387,19 @@ private List retrievePendingSegmentIds() { final String table = derbyConnectorRule.metadataTablesConfigSupplier().get().getPendingSegmentsTable(); return derbyConnector.retryWithHandle( - new HandleCallback>() - { - @Override - public List withHandle(Handle handle) - { - return handle.createQuery("SELECT id FROM " + table + " ORDER BY id") - .map(StringMapper.FIRST) - .list(); - } - } + handle -> handle.createQuery("SELECT id FROM " + table + " ORDER BY id") + .map(StringMapper.FIRST) + .list() ); } + private List retrieveUsedSegmentIds() { final String table = derbyConnectorRule.metadataTablesConfigSupplier().get().getSegmentsTable(); return derbyConnector.retryWithHandle( - new HandleCallback>() - { - @Override - public List withHandle(Handle handle) - { - return handle.createQuery("SELECT id FROM " + table + " WHERE used = true ORDER BY id") - .map(StringMapper.FIRST) - .list(); - } - } + handle -> handle.createQuery("SELECT id FROM " + table + " WHERE used = true ORDER BY id") + .map(StringMapper.FIRST) + .list() ); } @@ -448,16 +407,9 @@ private List retrieveUnusedSegmentIds() { final String table = derbyConnectorRule.metadataTablesConfigSupplier().get().getSegmentsTable(); return derbyConnector.retryWithHandle( - new HandleCallback>() - { - @Override - public List withHandle(Handle handle) - { - return handle.createQuery("SELECT id FROM " + table + " WHERE used = false ORDER BY id") - .map(StringMapper.FIRST) - .list(); - } - } + handle -> handle.createQuery("SELECT id FROM " + table + " WHERE used = false ORDER BY id") + .map(StringMapper.FIRST) + .list() ); } @@ -466,39 +418,34 @@ private Boolean insertUsedSegments(Set dataSegments) { final String table = derbyConnectorRule.metadataTablesConfigSupplier().get().getSegmentsTable(); return derbyConnector.retryWithHandle( - new HandleCallback() - { - @Override - public Boolean withHandle(Handle handle) throws Exception - { - PreparedBatch preparedBatch = handle.prepareBatch( - StringUtils.format( - "INSERT INTO %1$s (id, dataSource, created_date, start, %2$send%2$s, partitioned, version, used, payload) " - + "VALUES (:id, :dataSource, :created_date, :start, :end, :partitioned, :version, :used, :payload)", - table, - derbyConnector.getQuoteString() - ) - ); - for (DataSegment segment : dataSegments) { - preparedBatch.add() - .bind("id", segment.getId().toString()) - .bind("dataSource", segment.getDataSource()) - .bind("created_date", DateTimes.nowUtc().toString()) - .bind("start", segment.getInterval().getStart().toString()) - .bind("end", segment.getInterval().getEnd().toString()) - .bind("partitioned", (segment.getShardSpec() instanceof NoneShardSpec) ? false : true) - .bind("version", segment.getVersion()) - .bind("used", true) - .bind("payload", mapper.writeValueAsBytes(segment)); - } - - final int[] affectedRows = preparedBatch.execute(); - final boolean succeeded = Arrays.stream(affectedRows).allMatch(eachAffectedRows -> eachAffectedRows == 1); - if (!succeeded) { - throw new ISE("Failed to publish segments to DB"); - } - return true; + handle -> { + PreparedBatch preparedBatch = handle.prepareBatch( + StringUtils.format( + "INSERT INTO %1$s (id, dataSource, created_date, start, %2$send%2$s, partitioned, version, used, payload) " + + "VALUES (:id, :dataSource, :created_date, :start, :end, :partitioned, :version, :used, :payload)", + table, + derbyConnector.getQuoteString() + ) + ); + for (DataSegment segment : dataSegments) { + preparedBatch.add() + .bind("id", segment.getId().toString()) + .bind("dataSource", segment.getDataSource()) + .bind("created_date", DateTimes.nowUtc().toString()) + .bind("start", segment.getInterval().getStart().toString()) + .bind("end", segment.getInterval().getEnd().toString()) + .bind("partitioned", !(segment.getShardSpec() instanceof NoneShardSpec)) + .bind("version", segment.getVersion()) + .bind("used", true) + .bind("payload", mapper.writeValueAsBytes(segment)); } + + final int[] affectedRows = preparedBatch.execute(); + final boolean succeeded = Arrays.stream(affectedRows).allMatch(eachAffectedRows -> eachAffectedRows == 1); + if (!succeeded) { + throw new ISE("Failed to publish segments to DB"); + } + return true; } ); } @@ -561,12 +508,12 @@ public void testAnnounceHistoricalSegments() throws IOException ); } - List segmentIds = segments.stream().map(segment -> segment.getId().toString()).collect(Collectors.toList()); - segmentIds.sort(Comparator.naturalOrder()); - Assert.assertEquals( - segmentIds, - retrieveUsedSegmentIds() - ); + List segmentIds = segments.stream() + .map(segment -> segment.getId().toString()) + .sorted(Comparator.naturalOrder()) + .collect(Collectors.toList()); + + Assert.assertEquals(segmentIds, retrieveUsedSegmentIds()); // Should not update dataSource metadata. Assert.assertEquals(0, metadataUpdateCounter.get()); @@ -823,10 +770,7 @@ public void testTransactionalAnnounceFailSegmentDropFailWithRetry() throws IOExc retrieveUsedSegmentIds() ); - DataSegment nonExistingSegment = defaultSegment4; - - Set dropSegments = ImmutableSet.of(existingSegment1, nonExistingSegment); - + Set dropSegments = ImmutableSet.of(existingSegment1, defaultSegment4); final SegmentPublishResult result1 = coordinator.announceHistoricalSegments( SEGMENTS, dropSegments, @@ -1576,17 +1520,19 @@ public void testAllocatePendingSegment() } /** - * This test simulates an issue detected on the field consisting of the following sequence of events: - * - A kafka stream segment was created on a given interval - * - Later, after the above was published, another segment on same interval was created by the stream - * - Later, after the above was published, another segment on same interval was created by the stream - * - Later a compaction was issued for the three segments above - * - Later, after the above was published, another segment on same interval was created by the stream - * - Later, the compacted segment got dropped due to a drop rule + * This test verifies the behaviour in the following sequence of events: + * - create segment1 for an interval and publish + * - create segment2 for same interval and publish + * - create segment3 for same interval and publish + * - compact all segments above and publish new segments + * - create segment4 for the same interval + * - drop the compacted segment + * - create segment5 for the same interval + * - verify that the id for segment5 is correct * - Later, after the above was dropped, another segment on same interval was created by the stream but this - * time there was an integrity violation in the pending segments table because the - * {@link IndexerSQLMetadataStorageCoordinator#createNewSegment(Handle, String, Interval, PartialShardSpec, String)} - * method returned an segment id that already existed in the pending segments table + * time there was an integrity violation in the pending segments table because the + * {@link IndexerSQLMetadataStorageCoordinator#createNewSegment(Handle, String, Interval, PartialShardSpec, String)} + * method returned an segment id that already existed in the pending segments table */ @Test public void testAllocatePendingSegmentAfterDroppingExistingSegment() @@ -1690,13 +1636,13 @@ public void testAllocatePendingSegmentAfterDroppingExistingSegment() /** * Slightly different that the above test but that involves reverted compaction - 1) used segments of version = A, id = 0, 1, 2 - 2) overwrote segments of version = B, id = 0 <= compaction - 3) marked segments unused for version = A, id = 0, 1, 2 <= overshadowing - 4) pending segment of version = B, id = 1 <= appending new data, aborted - 5) reverted compaction, mark segments used for version = A, id = 0, 1, 2, and mark compacted segments unused - 6) used segments of version = A, id = 0, 1, 2 - 7) pending segment of version = B, id = 1 + * 1) used segments of version = A, id = 0, 1, 2 + * 2) overwrote segments of version = B, id = 0 <= compaction + * 3) marked segments unused for version = A, id = 0, 1, 2 <= overshadowing + * 4) pending segment of version = B, id = 1 <= appending new data, aborted + * 5) reverted compaction, mark segments used for version = A, id = 0, 1, 2, and mark compacted segments unused + * 6) used segments of version = A, id = 0, 1, 2 + * 7) pending segment of version = B, id = 1 */ @Test public void testAnotherAllocatePendingSegmentAfterRevertingCompaction() @@ -1842,13 +1788,13 @@ public void testAnotherAllocatePendingSegmentAfterRevertingCompaction() // used segment: version = A, id = 0,1,2 // unused segment: version = B, id = 0 List pendings = retrievePendingSegmentIds(); - Assert.assertTrue(pendings.size() == 4); + Assert.assertEquals(4, pendings.size()); List used = retrieveUsedSegmentIds(); - Assert.assertTrue(used.size() == 3); + Assert.assertEquals(3, used.size()); List unused = retrieveUnusedSegmentIds(); - Assert.assertTrue(unused.size() == 1); + Assert.assertEquals(1, unused.size()); // Simulate one more append load final SegmentIdWithShardSpec identifier4 = coordinator.allocatePendingSegment( @@ -1886,7 +1832,71 @@ public void testAnotherAllocatePendingSegmentAfterRevertingCompaction() Assert.assertEquals("ds_2017-01-01T00:00:00.000Z_2017-02-01T00:00:00.000Z_A_3", ids.get(3)); } - + + @Test + public void testAllocatePendingSegments() + { + final PartialShardSpec partialShardSpec = NumberedPartialShardSpec.instance(); + final String dataSource = "ds"; + final Interval interval = Intervals.of("2017-01-01/2017-02-01"); + final String sequenceName = "seq"; + + final SegmentCreateRequest request = new SegmentCreateRequest(sequenceName, null, "v1", partialShardSpec); + final SegmentIdWithShardSpec segmentId0 = coordinator.allocatePendingSegments( + dataSource, + interval, + false, + Collections.singletonList(request) + ).get(request); + + Assert.assertEquals("ds_2017-01-01T00:00:00.000Z_2017-02-01T00:00:00.000Z_v1", segmentId0.toString()); + + final SegmentCreateRequest request1 = + new SegmentCreateRequest(sequenceName, segmentId0.toString(), segmentId0.getVersion(), partialShardSpec); + final SegmentIdWithShardSpec segmentId1 = coordinator.allocatePendingSegments( + dataSource, + interval, + false, + Collections.singletonList(request1) + ).get(request1); + + Assert.assertEquals("ds_2017-01-01T00:00:00.000Z_2017-02-01T00:00:00.000Z_v1_1", segmentId1.toString()); + + final SegmentCreateRequest request2 = + new SegmentCreateRequest(sequenceName, segmentId1.toString(), segmentId1.getVersion(), partialShardSpec); + final SegmentIdWithShardSpec segmentId2 = coordinator.allocatePendingSegments( + dataSource, + interval, + false, + Collections.singletonList(request2) + ).get(request2); + + Assert.assertEquals("ds_2017-01-01T00:00:00.000Z_2017-02-01T00:00:00.000Z_v1_2", segmentId2.toString()); + + final SegmentCreateRequest request3 = + new SegmentCreateRequest(sequenceName, segmentId1.toString(), segmentId1.getVersion(), partialShardSpec); + final SegmentIdWithShardSpec segmentId3 = coordinator.allocatePendingSegments( + dataSource, + interval, + false, + Collections.singletonList(request3) + ).get(request3); + + Assert.assertEquals("ds_2017-01-01T00:00:00.000Z_2017-02-01T00:00:00.000Z_v1_2", segmentId3.toString()); + Assert.assertEquals(segmentId2, segmentId3); + + final SegmentCreateRequest request4 = + new SegmentCreateRequest("seq1", null, "v1", partialShardSpec); + final SegmentIdWithShardSpec segmentId4 = coordinator.allocatePendingSegments( + dataSource, + interval, + false, + Collections.singletonList(request4) + ).get(request4); + + Assert.assertEquals("ds_2017-01-01T00:00:00.000Z_2017-02-01T00:00:00.000Z_v1_3", segmentId4.toString()); + } + @Test public void testNoPendingSegmentsAndOneUsedSegment() { @@ -1923,9 +1933,8 @@ public void testNoPendingSegmentsAndOneUsedSegment() true ); Assert.assertEquals("ds_2017-01-01T00:00:00.000Z_2017-02-01T00:00:00.000Z_A_1", identifier.toString()); - - } + } @Test From 10bec54acc92ae109db6fee8e60bdcc641f2b36d Mon Sep 17 00:00:00 2001 From: TSFenwick Date: Mon, 5 Dec 2022 02:34:34 -0800 Subject: [PATCH 111/141] Switching emitter. This will allow for a per feed emitter designation. (#13363) * Switching emitter. This will allow for a per feed emitter designation. This will work by looking at an event's feed and direct it to a specific emitter. If no specific feed is specified for a feed. The emitter can direct the event to a default emitter. * fix checkstyle issues and make docs for switching emitter use basic event feeds * fix broken docs, add test, and guard against misconfigurations * add module test add switching emitter module test * fix broken SwitchingEmitterModuleTest * add apache license to top of test * fix checkstyle issues * address comments by adding javadocs, removing a todo, and making druid docs more clear --- .../util/emitter/core/SwitchingEmitter.java | 167 ++++++++++++++++++ .../emitter/core/SwitchingEmitterTest.java | 153 ++++++++++++++++ docs/configuration/index.md | 9 + .../druid/server/emitter/EmitterModule.java | 1 + .../emitter/SwitchingEmitterConfig.java | 53 ++++++ .../emitter/SwitchingEmitterModule.java | 82 +++++++++ .../SwitchingEmitterModuleTest.java | 114 ++++++++++++ 7 files changed, 579 insertions(+) create mode 100644 core/src/main/java/org/apache/druid/java/util/emitter/core/SwitchingEmitter.java create mode 100644 core/src/test/java/org/apache/druid/java/util/emitter/core/SwitchingEmitterTest.java create mode 100644 server/src/main/java/org/apache/druid/server/emitter/SwitchingEmitterConfig.java create mode 100644 server/src/main/java/org/apache/druid/server/emitter/SwitchingEmitterModule.java create mode 100644 server/src/test/java/org/apache/druid/initialization/SwitchingEmitterModuleTest.java diff --git a/core/src/main/java/org/apache/druid/java/util/emitter/core/SwitchingEmitter.java b/core/src/main/java/org/apache/druid/java/util/emitter/core/SwitchingEmitter.java new file mode 100644 index 000000000000..1dedd548294e --- /dev/null +++ b/core/src/main/java/org/apache/druid/java/util/emitter/core/SwitchingEmitter.java @@ -0,0 +1,167 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.java.util.emitter.core; + +import com.google.common.collect.ImmutableSet; +import org.apache.druid.java.util.common.lifecycle.LifecycleStart; +import org.apache.druid.java.util.common.lifecycle.LifecycleStop; +import org.apache.druid.java.util.common.logger.Logger; + +import java.io.IOException; +import java.util.Arrays; +import java.util.List; +import java.util.Map; +import java.util.Set; + +/** + * An emitter than that offers the ability to direct an event to multiple emitters based on the event's feed. + */ +public class SwitchingEmitter implements Emitter +{ + + private static final Logger log = new Logger(SwitchingEmitter.class); + + private final Emitter[] defaultEmitters; + + private final Map> feedToEmitters; + private final Set knownEmitters; + + /** + * Constructor for the SwitchingEmitter + * + * @param feedToEmitters Map of feed to a list of emitters that correspond to each feed, + * @param defaultEmitter A list of emitters to use if there isn't a match of feed to an emitter + */ + public SwitchingEmitter(Map> feedToEmitters, Emitter[] defaultEmitter) + { + this.feedToEmitters = feedToEmitters; + this.defaultEmitters = defaultEmitter; + ImmutableSet.Builder emittersSetBuilder = new ImmutableSet.Builder<>(); + emittersSetBuilder.addAll(Arrays.stream(defaultEmitter).iterator()); + for (List emitterList : feedToEmitters.values()) { + for (Emitter emitter : emitterList) { + emittersSetBuilder.add(emitter); + } + } + this.knownEmitters = emittersSetBuilder.build(); + } + + /** + * Start the emitter. This will start all the emitters the SwitchingEmitter uses. + */ + @Override + @LifecycleStart + public void start() + { + log.info("Starting Switching Emitter."); + + for (Emitter e : knownEmitters) { + log.info("Starting emitter %s.", e.getClass().getName()); + e.start(); + } + } + + /** + * Emit an event. This method must not throw exceptions or block. The emitters that this uses must also not throw + * exceptions or block. + *

    + * This emitter will direct events based on feed to a list of emitters specified. If there is no match the event will + * use a list of default emitters instead. + *

    + * Emitters that this emitter uses that receive too many events and internal queues fill up, should drop events rather + * than blocking or consuming excessive memory. + *

    + * If an emitter that this emitter uses receives input it considers to be invalid, or has an internal problem, it + * should deal with that by logging a warning rather than throwing an exception. Emitters that log warnings + * should consider throttling warnings to avoid excessive logs, since a busy Druid cluster can emit a high volume of + * events. + * + * @param event The event that will be emitted. + */ + @Override + public void emit(Event event) + { + // linear search is likely faster than hashed lookup + for (Map.Entry> feedToEmitters : feedToEmitters.entrySet()) { + if (feedToEmitters.getKey().equals(event.getFeed())) { + for (Emitter emitter : feedToEmitters.getValue()) { + emitter.emit(event); + } + return; + } + } + for (Emitter emitter : defaultEmitters) { + emitter.emit(event); + } + } + + /** + * Triggers this emitter to tell all emitters that this uses to flush. + * @throws IOException + */ + @Override + public void flush() throws IOException + { + boolean fail = false; + log.info("Flushing Switching Emitter."); + + for (Emitter e : knownEmitters) { + try { + log.info("Flushing emitter %s.", e.getClass().getName()); + e.flush(); + } + catch (IOException ex) { + log.error(ex, "Failed to flush emitter [%s]", e.getClass().getName()); + fail = true; + } + } + + if (fail) { + throw new IOException("failed to flush one or more emitters"); + } + } + + /** + * Closes all emitters that the SwitchingEmitter uses + * @throws IOException + */ + @Override + @LifecycleStop + public void close() throws IOException + { + boolean fail = false; + log.info("Closing Switching Emitter."); + + for (Emitter e : knownEmitters) { + try { + log.info("Closing emitter %s.", e.getClass().getName()); + e.close(); + } + catch (IOException ex) { + log.error(ex, "Failed to close emitter [%s]", e.getClass().getName()); + fail = true; + } + } + + if (fail) { + throw new IOException("failed to close one or more emitters"); + } + } +} diff --git a/core/src/test/java/org/apache/druid/java/util/emitter/core/SwitchingEmitterTest.java b/core/src/test/java/org/apache/druid/java/util/emitter/core/SwitchingEmitterTest.java new file mode 100644 index 000000000000..5183b47947af --- /dev/null +++ b/core/src/test/java/org/apache/druid/java/util/emitter/core/SwitchingEmitterTest.java @@ -0,0 +1,153 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.java.util.emitter.core; + +import com.google.common.collect.ImmutableList; +import com.google.common.collect.ImmutableMap; +import org.easymock.EasyMock; +import org.junit.After; +import org.junit.Before; +import org.junit.Test; + +import java.io.IOException; +import java.util.HashSet; +import java.util.List; +import java.util.Map; +import java.util.Set; + +public class SwitchingEmitterTest +{ + + private static final String FEED_1 = "feed1"; + private static final String FEED_2 = "feed2"; + private static final String FEED_3 = "feed3"; + private SwitchingEmitter switchingEmitter; + + private Map> emitters; + private List defaultEmitters; + + private Emitter feed1Emitter1; + private Emitter feed1Emitter2; + private Emitter feed2Emitter1; + private Emitter feed1AndFeed3Emitter; + + private Set allEmitters; + + @Before + public void setup() + { + this.defaultEmitters = ImmutableList.of( + EasyMock.createMock(Emitter.class), + EasyMock.createMock(Emitter.class) + ); + this.feed1Emitter1 = EasyMock.createMock(Emitter.class); + this.feed1Emitter2 = EasyMock.createMock(Emitter.class); + this.feed2Emitter1 = EasyMock.createMock(Emitter.class); + this.feed1AndFeed3Emitter = EasyMock.createMock(Emitter.class); + this.emitters = ImmutableMap.of(FEED_1, ImmutableList.of(feed1Emitter1, feed1Emitter2, feed1AndFeed3Emitter), + FEED_2, ImmutableList.of(feed2Emitter1), + FEED_3, ImmutableList.of(feed1AndFeed3Emitter)); + + allEmitters = new HashSet<>(); + allEmitters.addAll(defaultEmitters); + for (List feedEmitters : emitters.values()) { + allEmitters.addAll(feedEmitters); + } + this.switchingEmitter = new SwitchingEmitter(emitters, defaultEmitters.toArray(new Emitter[0])); + } + + @Test + public void testStart() + { + for (Emitter emitter : allEmitters) { + emitter.start(); + EasyMock.replay(emitter); + } + + switchingEmitter.start(); + } + + @Test + public void testEmit() + { + // test emitting events to all 3 feeds and default emitter + Event feed1Event = EasyMock.createMock(Event.class); + Event feed2Event = EasyMock.createMock(Event.class); + Event feed3Event = EasyMock.createMock(Event.class); + Event eventWithNoMatchingFeed = EasyMock.createMock(Event.class); + + EasyMock.expect(feed1Event.getFeed()).andReturn(FEED_1).anyTimes(); + EasyMock.expect(feed2Event.getFeed()).andReturn(FEED_2).anyTimes(); + EasyMock.expect(feed3Event.getFeed()).andReturn(FEED_3).anyTimes(); + EasyMock.expect(eventWithNoMatchingFeed.getFeed()).andReturn("no-real-feed").anyTimes(); + EasyMock.replay(feed1Event, feed2Event, feed3Event, eventWithNoMatchingFeed); + + for (Emitter emitter : defaultEmitters) { + emitter.emit(eventWithNoMatchingFeed); + } + for (Emitter emitter : emitters.get("feed1")) { + emitter.emit(feed1Event); + } + for (Emitter emitter : emitters.get("feed2")) { + emitter.emit(feed2Event); + } + for (Emitter emitter : emitters.get("feed3")) { + emitter.emit(feed3Event); + } + for (Emitter emitter : allEmitters) { + EasyMock.replay(emitter); + } + + switchingEmitter.emit(feed1Event); + switchingEmitter.emit(feed2Event); + switchingEmitter.emit(feed3Event); + switchingEmitter.emit(eventWithNoMatchingFeed); + } + + @Test + public void testFlush() throws IOException + { + for (Emitter emitter : allEmitters) { + emitter.flush(); + EasyMock.replay(emitter); + } + + switchingEmitter.flush(); + } + + @Test + public void testClose() throws IOException + { + for (Emitter emitter : allEmitters) { + emitter.close(); + EasyMock.replay(emitter); + } + + switchingEmitter.close(); + } + + @After + public void tearDown() + { + for (Emitter emitter : allEmitters) { + EasyMock.verify(emitter); + } + } +} diff --git a/docs/configuration/index.md b/docs/configuration/index.md index 8ac9382b7e21..021ea4cfb44e 100644 --- a/docs/configuration/index.md +++ b/docs/configuration/index.md @@ -418,6 +418,7 @@ There are several emitters available: - [`parametrized`](#parametrized-http-emitter-module) operates like the `http` emitter but fine-tunes the recipient URL based on the event feed. - [`composing`](#composing-emitter-module) initializes multiple emitter modules. - [`graphite`](#graphite-emitter) emits metrics to a [Graphite](https://graphiteapp.org/) Carbon service. +- [`switching`](#switching-emitter) initializes and emits to multiple emitter modules based on the event feed. ##### Logging Emitter Module @@ -483,6 +484,14 @@ Instead use `recipientBaseUrlPattern` described in the table below. To use graphite as emitter set `druid.emitter=graphite`. For configuration details, see [Graphite emitter](../development/extensions-contrib/graphite.md) for the Graphite emitter Druid extension. +##### Switching Emitter + +To use switching as emitter set `druid.emitter=switching`. + +|Property|Description|Default| +|--------|-----------|-------| +|`druid.emitter.switching.emitters`|JSON map of feed to list of emitter modules that will be used for the mapped feed, e.g., {"metrics":["http"], "alerts":["logging"]}|{}| +|`druid.emitter.switching.defaultEmitters`|JSON list of emitter modules to load that will be used if there is no emitter specifically designated for that event's feed, e.g., ["logging","http"].|[]| ### Metadata storage diff --git a/server/src/main/java/org/apache/druid/server/emitter/EmitterModule.java b/server/src/main/java/org/apache/druid/server/emitter/EmitterModule.java index c9db2e1dd962..7f4821768193 100644 --- a/server/src/main/java/org/apache/druid/server/emitter/EmitterModule.java +++ b/server/src/main/java/org/apache/druid/server/emitter/EmitterModule.java @@ -78,6 +78,7 @@ public void configure(Binder binder) binder.install(new HttpEmitterModule()); binder.install(new ParametrizedUriEmitterModule()); binder.install(new ComposingEmitterModule()); + binder.install(new SwitchingEmitterModule()); binder.bind(Emitter.class).toProvider(new EmitterProvider(emitterType)).in(LazySingleton.class); diff --git a/server/src/main/java/org/apache/druid/server/emitter/SwitchingEmitterConfig.java b/server/src/main/java/org/apache/druid/server/emitter/SwitchingEmitterConfig.java new file mode 100644 index 000000000000..2d3da8f2453a --- /dev/null +++ b/server/src/main/java/org/apache/druid/server/emitter/SwitchingEmitterConfig.java @@ -0,0 +1,53 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.server.emitter; + + +import com.fasterxml.jackson.annotation.JsonProperty; +import com.google.common.collect.ImmutableList; +import com.google.common.collect.ImmutableMap; + +import javax.validation.constraints.NotNull; +import java.util.List; +import java.util.Map; + +/** + */ +public class SwitchingEmitterConfig +{ + + @JsonProperty + @NotNull + private Map> emitters = ImmutableMap.of(); + + @JsonProperty + @NotNull + private List defaultEmitters = ImmutableList.of(); + + public Map> getEmitters() + { + return emitters; + } + + public List getDefaultEmitter() + { + return defaultEmitters; + } +} diff --git a/server/src/main/java/org/apache/druid/server/emitter/SwitchingEmitterModule.java b/server/src/main/java/org/apache/druid/server/emitter/SwitchingEmitterModule.java new file mode 100644 index 000000000000..281debd9960a --- /dev/null +++ b/server/src/main/java/org/apache/druid/server/emitter/SwitchingEmitterModule.java @@ -0,0 +1,82 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.server.emitter; + +import com.google.common.collect.ImmutableList; +import com.google.inject.Binder; +import com.google.inject.Injector; +import com.google.inject.Key; +import com.google.inject.Module; +import com.google.inject.Provides; +import com.google.inject.name.Named; +import com.google.inject.name.Names; +import org.apache.druid.guice.JsonConfigProvider; +import org.apache.druid.guice.ManageLifecycle; +import org.apache.druid.java.util.common.logger.Logger; +import org.apache.druid.java.util.emitter.core.Emitter; +import org.apache.druid.java.util.emitter.core.SwitchingEmitter; + +import java.util.List; +import java.util.Map; +import java.util.stream.Collectors; + +public class SwitchingEmitterModule implements Module +{ + + public static final String EMITTER_TYPE = "switching"; + + private static Logger log = new Logger(SwitchingEmitterModule.class); + + @Override + public void configure(Binder binder) + { + JsonConfigProvider.bind(binder, "druid.emitter.switching", SwitchingEmitterConfig.class); + } + + @Provides + @ManageLifecycle + @Named(EMITTER_TYPE) + public Emitter makeEmitter(SwitchingEmitterConfig config, final Injector injector) + { + log.info( + "Createing Switching emitter with %s, and default emitter %s", + config.getEmitters(), + config.getDefaultEmitter() + ); + Map> switchingEmitters = config + .getEmitters() + .entrySet() + .stream() + .collect( + Collectors.toMap( + Map.Entry::getKey, + entry -> entry.getValue() + .stream() + .map( + emitterName -> injector.getInstance(Key.get(Emitter.class, Names.named(emitterName)))) + .collect(Collectors.toList()))); + + ImmutableList.Builder defaultEmittersBuilder = new ImmutableList.Builder<>(); + for (String emitterName : config.getDefaultEmitter()) { + defaultEmittersBuilder.add(injector.getInstance(Key.get(Emitter.class, Names.named(emitterName)))); + } + return new SwitchingEmitter(switchingEmitters, defaultEmittersBuilder.build().toArray(new Emitter[0])); + } +} diff --git a/server/src/test/java/org/apache/druid/initialization/SwitchingEmitterModuleTest.java b/server/src/test/java/org/apache/druid/initialization/SwitchingEmitterModuleTest.java new file mode 100644 index 000000000000..94280aed1d7f --- /dev/null +++ b/server/src/test/java/org/apache/druid/initialization/SwitchingEmitterModuleTest.java @@ -0,0 +1,114 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.initialization; + +import com.google.common.collect.ImmutableList; +import com.google.common.collect.ImmutableMap; +import com.google.inject.Binder; +import com.google.inject.Guice; +import com.google.inject.Injector; +import com.google.inject.Key; +import com.google.inject.Module; +import com.google.inject.name.Names; +import org.apache.druid.guice.DruidGuiceExtensions; +import org.apache.druid.guice.LifecycleModule; +import org.apache.druid.jackson.JacksonModule; +import org.apache.druid.java.util.emitter.core.Emitter; +import org.apache.druid.server.emitter.SwitchingEmitterConfig; +import org.apache.druid.server.emitter.SwitchingEmitterModule; +import org.easymock.EasyMock; +import org.junit.Before; +import org.junit.Test; + +import javax.validation.Validation; +import javax.validation.Validator; +import java.util.Properties; + +public class SwitchingEmitterModuleTest +{ + + private static final String DEFAULT_EMITTER_TYPE = "http"; + private static final String FEED_1_EMITTER_TYPE = "logging"; + private static final String FEED_1 = "metrics"; + private Emitter defaultEmitter; + private Emitter feed1Emitter; + + @Before + public void setup() + { + defaultEmitter = EasyMock.createMock(Emitter.class); + feed1Emitter = EasyMock.createMock(Emitter.class); + defaultEmitter.start(); + feed1Emitter.start(); + EasyMock.replay(defaultEmitter); + EasyMock.replay(feed1Emitter); + } + + @Test + public void testGetEmitter() + { + SwitchingEmitterConfig config = EasyMock.createMock(SwitchingEmitterConfig.class); + EasyMock.expect(config.getDefaultEmitter()).andReturn(ImmutableList.of(DEFAULT_EMITTER_TYPE)).anyTimes(); + EasyMock.expect(config.getEmitters()).andReturn(ImmutableMap.of(FEED_1, ImmutableList.of(FEED_1_EMITTER_TYPE))).anyTimes(); + + Injector injector = EasyMock.createMock(Injector.class); + EasyMock.expect(injector.getInstance(Key.get(Emitter.class, Names.named(DEFAULT_EMITTER_TYPE)))).andReturn( + defaultEmitter); + EasyMock.expect(injector.getInstance(Key.get(Emitter.class, Names.named(FEED_1_EMITTER_TYPE)))).andReturn( + feed1Emitter); + EasyMock.replay(config, injector); + + Emitter switchingEmitter = new SwitchingEmitterModule().makeEmitter(config, injector); + switchingEmitter.start(); + + EasyMock.verify(config, defaultEmitter, feed1Emitter, injector); + } + + @Test + public void testGetEmitterViaRealGuice() + { + Injector injector = Guice.createInjector( + new DruidGuiceExtensions(), + new LifecycleModule(), + new JacksonModule(), + new Module() + { + @Override + public void configure(Binder binder) + { + Properties props = new Properties(); + String defaultEmittersValue = "[\"" + DEFAULT_EMITTER_TYPE + "\"]"; + String emittersValue = "{\"" + FEED_1 + "\":[\"" + FEED_1_EMITTER_TYPE + "\"]}"; + props.put("druid.emitter.switching.defaultEmitters", defaultEmittersValue); + props.put("druid.emitter.switching.emitters", emittersValue); + binder.bind(Properties.class).toInstance(props); + binder.bind(Validator.class).toInstance(Validation.buildDefaultValidatorFactory().getValidator()); + binder.bind(Emitter.class).annotatedWith(Names.named(DEFAULT_EMITTER_TYPE)).toInstance(defaultEmitter); + binder.bind(Emitter.class).annotatedWith(Names.named(FEED_1_EMITTER_TYPE)).toInstance(feed1Emitter); + } + }, + new SwitchingEmitterModule() + ); + injector.getInstance(Key.get(Emitter.class, Names.named("switching"))).start(); + EasyMock.verify(defaultEmitter); + EasyMock.verify(feed1Emitter); + + } +} From 65945a686fa3ca950c18cf43d4fc5a91e572fa82 Mon Sep 17 00:00:00 2001 From: Kashif Faraz Date: Mon, 5 Dec 2022 16:53:10 +0530 Subject: [PATCH 112/141] Docs: Update docs for coordinator dynamic config (#13494) * Update docs for useBatchedSegmentSampler * Update docs for round robin assigment --- docs/configuration/index.md | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/docs/configuration/index.md b/docs/configuration/index.md index 021ea4cfb44e..9582bbd99b20 100644 --- a/docs/configuration/index.md +++ b/docs/configuration/index.md @@ -948,7 +948,7 @@ Issuing a GET request at the same URL will return the spec that is currently in |`mergeBytesLimit`|The maximum total uncompressed size in bytes of segments to merge.|524288000L| |`mergeSegmentsLimit`|The maximum number of segments that can be in a single [append task](../ingestion/tasks.md).|100| |`maxSegmentsToMove`|The maximum number of segments that can be moved at any given time.|5| -|`useBatchedSegmentSampler`|Boolean flag for whether or not we should use the Reservoir Sampling with a reservoir of size k instead of fixed size 1 to pick segments to move. This option can be enabled to speed up segment balancing process, especially if there are huge number of segments in the cluster or if there are too many segments to move.|false| +|`useBatchedSegmentSampler`|Deprecated. Boolean flag for whether or not we should use the Reservoir Sampling with a reservoir of size k instead of fixed size 1 to pick segments to move. This option can be enabled to speed up the sampling of segments to be balanced, especially if there is a large number of segments in the cluster or if there are too many segments to move.|true| |`percentOfSegmentsToConsiderPerMove`|Deprecated. This will eventually be phased out by the batched segment sampler. You can enable the batched segment sampler now by setting the dynamic Coordinator config, `useBatchedSegmentSampler`, to `true`. Note that if you choose to enable the batched segment sampler, `percentOfSegmentsToConsiderPerMove` will no longer have any effect on balancing. If `useBatchedSegmentSampler == false`, this config defines the percentage of the total number of segments in the cluster that are considered every time a segment needs to be selected for a move. Druid orders servers by available capacity ascending (the least available capacity first) and then iterates over the servers. For each server, Druid iterates over the segments on the server, considering them for moving. The default config of 100% means that every segment on every server is a candidate to be moved. This should make sense for most small to medium-sized clusters. However, an admin may find it preferable to drop this value lower if they don't think that it is worthwhile to consider every single segment in the cluster each time it is looking for a segment to move.|100| |`replicantLifetime`|The maximum number of Coordinator runs for a segment to be replicated before we start alerting.|15| |`replicationThrottleLimit`|The maximum number of segments that can be replicated at one time.|10| @@ -957,6 +957,7 @@ Issuing a GET request at the same URL will return the spec that is currently in |`killDataSourceWhitelist`|List of specific data sources for which kill tasks are sent if property `druid.coordinator.kill.on` is true. This can be a list of comma-separated data source names or a JSON array.|none| |`killPendingSegmentsSkipList`|List of data sources for which pendingSegments are _NOT_ cleaned up if property `druid.coordinator.kill.pendingSegments.on` is true. This can be a list of comma-separated data sources or a JSON array.|none| |`maxSegmentsInNodeLoadingQueue`|The maximum number of segments that could be queued for loading to any given server. This parameter could be used to speed up segments loading process, especially if there are "slow" nodes in the cluster (with low loading speed) or if too much segments scheduled to be replicated to some particular node (faster loading could be preferred to better segments distribution). Desired value depends on segments loading speed, acceptable replication time and number of nodes. Value 1000 could be a start point for a rather big cluster. Default value is 100. |100| +|`useRoundRobinSegmentAssignment`|Boolean flag for whether segments should be assigned to historicals in a round robin fashion. When disabled, segment assignment is done using the chosen balancer strategy. When enabled, this can speed up segment assignments leaving balancing to move the segments to their optimal locations (based on the balancer strategy) lazily. |false| |`decommissioningNodes`| List of historical servers to 'decommission'. Coordinator will not assign new segments to 'decommissioning' servers, and segments will be moved away from them to be placed on non-decommissioning servers at the maximum rate specified by `decommissioningMaxPercentOfMaxSegmentsToMove`.|none| |`decommissioningMaxPercentOfMaxSegmentsToMove`| Upper limit of segments the Coordinator can move from decommissioning servers to active non-decommissioning servers during a single run. This value is relative to the total maximum number of segments that can be moved at any given time based upon the value of `maxSegmentsToMove`.

    If `decommissioningMaxPercentOfMaxSegmentsToMove` is 0, the Coordinator does not move segments to decommissioning servers, effectively putting them in a type of "maintenance" mode. In this case, decommissioning servers do not participate in balancing or assignment by load rules. The Coordinator still considers segments on decommissioning servers as candidates to replicate on active servers.

    Decommissioning can stall if there are no available active servers to move the segments to. You can use the maximum percent of decommissioning segment movements to prioritize balancing or to decrease commissioning time to prevent active servers from being overloaded. The value must be between 0 and 100.|70| |`pauseCoordination`| Boolean flag for whether or not the coordinator should execute its various duties of coordinating the cluster. Setting this to true essentially pauses all coordination work while allowing the API to remain up. Duties that are paused include all classes that implement the `CoordinatorDuty` Interface. Such duties include: Segment balancing, Segment compaction, Emission of metrics controlled by the dynamic coordinator config `emitBalancingStats`, Submitting kill tasks for unused segments (if enabled), Logging of used segments in the cluster, Marking of newly unused or overshadowed segments, Matching and execution of load/drop rules for used segments, Unloading segments that are no longer marked as used from Historical servers. An example of when an admin may want to pause coordination would be if they are doing deep storage maintenance on HDFS Name Nodes with downtime and don't want the coordinator to be directing Historical Nodes to hit the Name Node with API requests until maintenance is done and the deep store is declared healthy for use again. |false| From 658a9c2d350e3ecb334cd7afc6faa29cd087091e Mon Sep 17 00:00:00 2001 From: AmatyaAvadhanula Date: Mon, 5 Dec 2022 21:05:07 +0530 Subject: [PATCH 113/141] Early stop on failed start (Alternative to #13087) (#13258) * Make halt configurable. Don't halt in tests --- server/pom.xml | 3 +- .../apache/druid/curator/CuratorModule.java | 42 +++++++++++++++++++ .../druid/curator/CuratorModuleTest.java | 2 +- 3 files changed, 45 insertions(+), 2 deletions(-) diff --git a/server/pom.xml b/server/pom.xml index c62533763bbb..14e1e5745782 100644 --- a/server/pom.xml +++ b/server/pom.xml @@ -454,8 +454,9 @@ org/apache/druid/metadata/BasicDataSourceExt.class - + org/apache/druid/server/QueryResponse.class + org/apache/druid/curator/CuratorModule.class diff --git a/server/src/main/java/org/apache/druid/curator/CuratorModule.java b/server/src/main/java/org/apache/druid/curator/CuratorModule.java index 665024fd7716..07c0ad84780e 100644 --- a/server/src/main/java/org/apache/druid/curator/CuratorModule.java +++ b/server/src/main/java/org/apache/druid/curator/CuratorModule.java @@ -30,6 +30,7 @@ import org.apache.curator.framework.imps.DefaultACLProvider; import org.apache.curator.retry.BoundedExponentialBackoffRetry; import org.apache.curator.shaded.com.google.common.base.Strings; +import org.apache.druid.concurrent.Threads; import org.apache.druid.guice.JsonConfigProvider; import org.apache.druid.guice.LazySingleton; import org.apache.druid.java.util.common.StringUtils; @@ -40,6 +41,7 @@ import java.nio.charset.StandardCharsets; import java.util.List; +import java.util.concurrent.TimeUnit; public class CuratorModule implements Module { @@ -49,6 +51,23 @@ public class CuratorModule implements Module static final int MAX_SLEEP_TIME_MS = 45000; private static final int MAX_RETRIES = 29; + private final boolean haltOnFailedStart; + + public CuratorModule() + { + this(true); + } + + /** + * + * @param haltOnFailedStart set to true if the JVM needs to be halted within 30 seconds of failed initialization + * due to unhandled curator exceptions. + */ + public CuratorModule(boolean haltOnFailedStart) + { + this.haltOnFailedStart = haltOnFailedStart; + } + @Override public void configure(Binder binder) { @@ -97,6 +116,29 @@ public CuratorFramework makeCurator(ZkEnablementConfig zkEnablementConfig, Curat framework.getUnhandledErrorListenable().addListener((message, e) -> { log.error(e, "Unhandled error in Curator, stopping server."); + + if (haltOnFailedStart) { + final long startTime = System.currentTimeMillis(); + final Thread halter = new Thread( + () -> { + try { + Threads.sleepFor(30, TimeUnit.SECONDS); + } + catch (InterruptedException ignored) { + + } + log.warn( + "Could not stop server within %,d millis after unhandled Curator error. Halting immediately.", + System.currentTimeMillis() - startTime + ); + Runtime.getRuntime().halt(1); + }, + "exiter-thread" + ); + halter.setDaemon(true); + halter.start(); + } + shutdown(lifecycle); }); diff --git a/server/src/test/java/org/apache/druid/curator/CuratorModuleTest.java b/server/src/test/java/org/apache/druid/curator/CuratorModuleTest.java index 20b86dc59660..5b88cb59f434 100644 --- a/server/src/test/java/org/apache/druid/curator/CuratorModuleTest.java +++ b/server/src/test/java/org/apache/druid/curator/CuratorModuleTest.java @@ -120,7 +120,7 @@ private Injector newInjector(final Properties props) return new StartupInjectorBuilder() .add( new LifecycleModule(), - new CuratorModule(), + new CuratorModule(false), binder -> binder.bind(Properties.class).toInstance(props) ) .build(); From fda0a1aadde9d640cd4f9ff448a74ee5d2149d2a Mon Sep 17 00:00:00 2001 From: Gian Merlino Date: Mon, 5 Dec 2022 20:53:59 -0800 Subject: [PATCH 114/141] Set chatAsync default to true. (#13491) This functionality was originally added in #13354. --- .../development/extensions-core/kafka-supervisor-reference.md | 4 ++-- docs/development/extensions-core/kinesis-ingestion.md | 4 ++-- .../supervisor/SeekableStreamSupervisorTuningConfig.java | 2 +- 3 files changed, 5 insertions(+), 5 deletions(-) diff --git a/docs/development/extensions-core/kafka-supervisor-reference.md b/docs/development/extensions-core/kafka-supervisor-reference.md index 9dc80d0c6585..360e75a83de4 100644 --- a/docs/development/extensions-core/kafka-supervisor-reference.md +++ b/docs/development/extensions-core/kafka-supervisor-reference.md @@ -204,8 +204,8 @@ The `tuningConfig` is optional and default parameters will be used if no `tuning | `handoffConditionTimeout` | Long | Milliseconds to wait for segment handoff. It must be >= 0, where 0 means to wait forever. | no (default == 0) | | `resetOffsetAutomatically` | Boolean | Controls behavior when Druid needs to read Kafka messages that are no longer available (i.e. when `OffsetOutOfRangeException` is encountered).

    If false, the exception will bubble up, which will cause your tasks to fail and ingestion to halt. If this occurs, manual intervention is required to correct the situation; potentially using the [Reset Supervisor API](../../operations/api-reference.md#supervisors). This mode is useful for production, since it will make you aware of issues with ingestion.

    If true, Druid will automatically reset to the earlier or latest offset available in Kafka, based on the value of the `useEarliestOffset` property (earliest if true, latest if false). Note that this can lead to data being _DROPPED_ (if `useEarliestOffset` is false) or _DUPLICATED_ (if `useEarliestOffset` is true) without your knowledge. Messages will be logged indicating that a reset has occurred, but ingestion will continue. This mode is useful for non-production situations, since it will make Druid attempt to recover from problems automatically, even if they lead to quiet dropping or duplicating of data.

    This feature behaves similarly to the Kafka `auto.offset.reset` consumer property. | no (default == false) | | `workerThreads` | Integer | The number of threads that the supervisor uses to handle requests/responses for worker tasks, along with any other internal asynchronous operation. | no (default == min(10, taskCount)) | -| `chatAsync` | Boolean | If true, use asynchronous communication with indexing tasks, and ignore the `chatThreads` parameter. If false, use synchronous communication in a thread pool of size `chatThreads`. | no (default == false) | -| `chatThreads` | Integer | The number of threads that will be used for communicating with indexing tasks. Ignored if `chatAsync` is `true`. | no (default == min(10, taskCount * replicas)) | +| `chatAsync` | Boolean | If true, use asynchronous communication with indexing tasks, and ignore the `chatThreads` parameter. If false, use synchronous communication in a thread pool of size `chatThreads`. | no (default == true) | +| `chatThreads` | Integer | The number of threads that will be used for communicating with indexing tasks. Ignored if `chatAsync` is `true` (the default). | no (default == min(10, taskCount * replicas)) | | `chatRetries` | Integer | The number of times HTTP requests to indexing tasks will be retried before considering tasks unresponsive. | no (default == 8) | | `httpTimeout` | ISO8601 Period | How long to wait for a HTTP response from an indexing task. | no (default == PT10S) | | `shutdownTimeout` | ISO8601 Period | How long to wait for the supervisor to attempt a graceful shutdown of tasks before exiting. | no (default == PT80S) | diff --git a/docs/development/extensions-core/kinesis-ingestion.md b/docs/development/extensions-core/kinesis-ingestion.md index 31682b54a291..9b14ec767c26 100644 --- a/docs/development/extensions-core/kinesis-ingestion.md +++ b/docs/development/extensions-core/kinesis-ingestion.md @@ -291,8 +291,8 @@ The `tuningConfig` is optional. If no `tuningConfig` is specified, default param |`resetOffsetAutomatically`|Boolean|Controls behavior when Druid needs to read Kinesis messages that are no longer available.

    If false, the exception will bubble up, which will cause your tasks to fail and ingestion to halt. If this occurs, manual intervention is required to correct the situation; potentially using the [Reset Supervisor API](../../operations/api-reference.md#supervisors). This mode is useful for production, since it will make you aware of issues with ingestion.

    If true, Druid will automatically reset to the earlier or latest sequence number available in Kinesis, based on the value of the `useEarliestSequenceNumber` property (earliest if true, latest if false). Please note that this can lead to data being _DROPPED_ (if `useEarliestSequenceNumber` is false) or _DUPLICATED_ (if `useEarliestSequenceNumber` is true) without your knowledge. Messages will be logged indicating that a reset has occurred, but ingestion will continue. This mode is useful for non-production situations, since it will make Druid attempt to recover from problems automatically, even if they lead to quiet dropping or duplicating of data.|no (default == false)| |`skipSequenceNumberAvailabilityCheck`|Boolean|Whether to enable checking if the current sequence number is still available in a particular Kinesis shard. If set to false, the indexing task will attempt to reset the current sequence number (or not), depending on the value of `resetOffsetAutomatically`.|no (default == false)| |`workerThreads`|Integer|The number of threads that the supervisor uses to handle requests/responses for worker tasks, along with any other internal asynchronous operation.|no (default == min(10, taskCount))| -|`chatAsync`|Boolean| If true, use asynchronous communication with indexing tasks, and ignore the `chatThreads` parameter. If false, use synchronous communication in a thread pool of size `chatThreads`. | no (default == false) | -|`chatThreads`|Integer| The number of threads that will be used for communicating with indexing tasks. Ignored if `chatAsync` is `true`.| no (default == min(10, taskCount * replicas))| +|`chatAsync`|Boolean| If true, use asynchronous communication with indexing tasks, and ignore the `chatThreads` parameter. If false, use synchronous communication in a thread pool of size `chatThreads`. | no (default == true) | +|`chatThreads`|Integer| The number of threads that will be used for communicating with indexing tasks. Ignored if `chatAsync` is `true` (the default).| no (default == min(10, taskCount * replicas))| |`chatRetries`|Integer|The number of times HTTP requests to indexing tasks will be retried before considering tasks unresponsive.| no (default == 8)| |`httpTimeout`|ISO8601 Period|How long to wait for a HTTP response from an indexing task.|no (default == PT10S)| |`shutdownTimeout`|ISO8601 Period|How long to wait for the supervisor to attempt a graceful shutdown of tasks before exiting.|no (default == PT80S)| diff --git a/indexing-service/src/main/java/org/apache/druid/indexing/seekablestream/supervisor/SeekableStreamSupervisorTuningConfig.java b/indexing-service/src/main/java/org/apache/druid/indexing/seekablestream/supervisor/SeekableStreamSupervisorTuningConfig.java index c74c2952de03..c518887cfbdb 100644 --- a/indexing-service/src/main/java/org/apache/druid/indexing/seekablestream/supervisor/SeekableStreamSupervisorTuningConfig.java +++ b/indexing-service/src/main/java/org/apache/druid/indexing/seekablestream/supervisor/SeekableStreamSupervisorTuningConfig.java @@ -26,7 +26,7 @@ public interface SeekableStreamSupervisorTuningConfig { - boolean DEFAULT_ASYNC = false; + boolean DEFAULT_ASYNC = true; String DEFAULT_OFFSET_FETCH_PERIOD = "PT30S"; int DEFAULT_CHAT_RETRIES = 8; String DEFAULT_HTTP_TIMEOUT = "PT10S"; From 136322d13b68f1e14552ea606ac70101ff6af400 Mon Sep 17 00:00:00 2001 From: Tejaswini Bandlamudi <96047043+tejaswini-imply@users.noreply.github.com> Date: Tue, 6 Dec 2022 12:08:03 +0530 Subject: [PATCH 115/141] clean install before license checks (#13502) --- .github/workflows/static-checks.yml | 44 ++++++++++++++--------------- 1 file changed, 22 insertions(+), 22 deletions(-) diff --git a/.github/workflows/static-checks.yml b/.github/workflows/static-checks.yml index a3f2d7c0e3d5..37e7a2225857 100644 --- a/.github/workflows/static-checks.yml +++ b/.github/workflows/static-checks.yml @@ -56,28 +56,6 @@ jobs: java-version: ${{ env.java_version }} cache: 'maven' - - name: license checks - if: ${{ matrix.java == 'jdk8' }} - run: ./.github/scripts/license_checks_script.sh - - - name: license checks for hadoop3 - if: ${{ matrix.java == 'jdk8' }} - env: - HADOOP_PROFILE: -Phadoop3 - run: ./.github/scripts/license_checks_script.sh - - - name: analyze dependencies - if: ${{ matrix.java == 'jdk8' }} - run: | - ./.github/scripts/analyze_dependencies_script.sh - - - name: analyze dependencies for hadoop3 - if: ${{ matrix.java == 'jdk8' }} - env: - HADOOP_PROFILE: -Phadoop3 - run: | - ./.github/scripts/analyze_dependencies_script.sh - - name: packaging check run: | ./.github/scripts/setup_generate_license.sh @@ -102,6 +80,28 @@ jobs: ${MVN} clean install -q -ff -pl '!distribution,!:druid-it-image,!:druid-it-cases' ${MAVEN_SKIP} ${MAVEN_SKIP_TESTS} -T1C && ${MVN} install -q -ff -pl 'distribution' ${MAVEN_SKIP} ${MAVEN_SKIP_TESTS} + - name: license checks + if: ${{ matrix.java == 'jdk8' }} + run: ./.github/scripts/license_checks_script.sh + + - name: license checks for hadoop3 + if: ${{ matrix.java == 'jdk8' }} + env: + HADOOP_PROFILE: -Phadoop3 + run: ./.github/scripts/license_checks_script.sh + + - name: analyze dependencies + if: ${{ matrix.java == 'jdk8' }} + run: | + ./.github/scripts/analyze_dependencies_script.sh + + - name: analyze dependencies for hadoop3 + if: ${{ matrix.java == 'jdk8' }} + env: + HADOOP_PROFILE: -Phadoop3 + run: | + ./.github/scripts/analyze_dependencies_script.sh + - name: animal sniffer checks if: ${{ matrix.java == 'jdk8' }} run: ${MVN} animal-sniffer:check --fail-at-end From 71977d69e6cf57b6b13ab71b8caff0fddfb2e96a Mon Sep 17 00:00:00 2001 From: rishabh singh Date: Tue, 6 Dec 2022 18:21:21 +0530 Subject: [PATCH 116/141] memory comptuation for indexer --- examples/bin/start-druid-main.py | 74 +++++++++++-------- examples/conf/druid/auto/indexer/jvm.config | 9 --- .../druid/auto/indexer/runtime.properties | 12 +-- 3 files changed, 49 insertions(+), 46 deletions(-) delete mode 100644 examples/conf/druid/auto/indexer/jvm.config diff --git a/examples/bin/start-druid-main.py b/examples/bin/start-druid-main.py index 25d7e93afc59..d1f4e6114918 100644 --- a/examples/bin/start-druid-main.py +++ b/examples/bin/start-druid-main.py @@ -74,7 +74,8 @@ COORDINATOR: 30, BROKER: 46, HISTORICAL: 80, - TASKS: 30 + TASKS: 30, + INDEXER: 32 } MINIMUM_MEMORY_MB = { @@ -83,7 +84,8 @@ TASKS: 1024, BROKER: 900, COORDINATOR: 256, - HISTORICAL: 900 + HISTORICAL: 900, + INDEXER: 1124 } HEAP_TO_TOTAL_MEM_RATIO = { @@ -92,7 +94,8 @@ COORDINATOR: 1, BROKER: 0.60, HISTORICAL: 0.40, - TASKS: 0.50 + TASKS: 0.50, + INDEXER: 0.50 } LOGGING_ENABLED = False @@ -225,6 +228,9 @@ def parse_arguments(args): service_list.append(service) + if INDEXER in services and MIDDLE_MANAGER in services: + raise ValueError('one of indexer and middleManager can run') + if len(service_list) == 0: # start all services service_list = DEFAULT_SERVICES @@ -242,12 +248,12 @@ def print_startup_config(service_list, config, zk): print_if_verbose('\n') -def middle_manager_task_memory_params_present(config): +def task_memory_params_present(config, service): java_opts_property_present = False worker_capacity_property_present = False - if is_file('{0}/middleManager/runtime.properties'.format(config)): - with open('{0}/middleManager/runtime.properties'.format(config)) as file: + if is_file('{0}/{1}/runtime.properties'.format(config, service)): + with open('{0}/{1}/runtime.properties'.format(config, service)) as file: for line in file: if line.startswith(TASK_JAVA_OPTS_PROPERTY): java_opts_property_present = True @@ -282,7 +288,7 @@ def verify_service_config(service, config): if is_file('{0}/{1}/runtime.properties'.format(config, service)) is False: raise ValueError('{0}/runtime.properties file is missing in the root config'.format(service)) - mm_task_java_opts_property, mm_task_worker_capacity_prop = middle_manager_task_memory_params_present(config) + mm_task_java_opts_prop, mm_task_worker_capacity_prop = task_memory_params_present(config, MIDDLE_MANAGER) if mm_task_java_opts_property is False: raise ValueError('{0} property missing in {1}/runtime.properties'.format(TASK_JAVA_OPTS_PROPERTY, service)) @@ -307,42 +313,36 @@ def should_compute_memory(config, total_memory, service_list): mm_task_property_present = False if MIDDLE_MANAGER in service_list: - mm_task_java_opts_property, mm_task_worker_capacity_prop = middle_manager_task_memory_params_present(config) - mm_task_property_present = mm_task_java_opts_property or mm_task_worker_capacity_prop + mm_task_java_opts_prop, mm_task_worker_capacity_prop = task_memory_params_present(config, MIDDLE_MANAGER) + mm_task_property_present = mm_task_java_opts_prop or mm_task_worker_capacity_prop - # if indexer has to be run, all the memory related parameters need to be specified + indexer_task_worker_capacity_prop = False if INDEXER in service_list: - if MIDDLE_MANAGER in service_list: - raise ValueError("one of indexer or middleManager can run") - if total_memory != "": - raise ValueError( - "If service list includes indexer, jvm.config should be specified for " - "each service and memory argument shouldn't be specified") - if jvm_config_count != len(service_list): - raise ValueError("If service list includes indexer, jvm.config should be specified for each service") - for service in service_list: - verify_service_config(service, config) - - return False + indexer_task_java_opts_prop, indexer_task_worker_capacity_prop = task_memory_params_present(config, INDEXER) # possible error states # 1. memory argument is specified, also jvm.config or middleManger/runtime.properties having # druid.indexer.runner.javaOptsArray or druid.worker.capacity parameters is present # 2. jvm.config is not present for any service, but middleManger/runtime.properties has # druid.indexer.runner.javaOptsArray or druid.worker.capacity parameters + # or indexer/runtime.properties has druid.worker.capacity # 3. jvm.config present for some but not all services # 4. jvm.config file is present for all services, but it doesn't contain required parameters # 5. lastly, if middleManager is to be started, and it is missing task memory properties - if jvm_config_count > 0 or mm_task_property_present: + if jvm_config_count > 0 or mm_task_property_present or indexer_task_worker_capacity_prop: if total_memory != "": raise ValueError( - "If jvm.config for services and/or middleManager configs " + "If jvm.config for services and/or middleManager/indexer configs " "(druid.worker.capacity, druid.indexer.runner.javaOptsArray) is present, " "memory argument shouldn't be specified") - if jvm_config_count == 0: + if jvm_config_count == 0 and mm_task_property_present: raise ValueError("middleManger configs (druid.indexer.runner.javaOptsArray or druid.worker.capacity) " "is present in middleManager/runtime.properties, " "add jvm.config for all other services") + if jvm_config_count == 0 and indexer_task_worker_capacity_prop: + raise ValueError("indexer configs (druid.worker.capacity) " + "is present in indexer/runtime.properties, " + "add jvm.config for all other services") if jvm_config_count != len(service_list): raise ValueError("jvm.config file should be present for all services or none") for service in service_list: @@ -352,6 +352,7 @@ def should_compute_memory(config, total_memory, service_list): # compute memory only when none of the specified services contains jvm.config, # if middleManager is to be started it shouldn't contain task memory properties + # if indexer is present it shouldn't contain task memory properties return True @@ -467,6 +468,14 @@ def build_memory_config(service, allocated_memory): java_opts_array = build_mm_task_java_opts_array(memory_type) return ['-D{0}={1}'.format(TASK_WORKER_CAPACITY_PROPERTY, task_count), java_opts_array], task_memory * task_count + elif service == INDEXER: + heap_memory = HEAP_TO_TOTAL_MEM_RATIO.get(service) * allocated_memory + direct_memory = int(allocated_memory - heap_memory) + heap_memory = int(heap_memory) + memory_type, task_count, task_memory = compute_tasks_memory(allocated_memory) + return ['-D{0}={1}'.format(TASK_WORKER_CAPACITY_PROPERTY, task_count), + '-Xms{0}m -Xmx{0}m -XX:MaxDirectMemorySize={1}m'.format(heap_memory, direct_memory)], \ + task_memory * task_count else: heap_memory = HEAP_TO_TOTAL_MEM_RATIO.get(service) * allocated_memory direct_memory = int(allocated_memory - heap_memory) @@ -547,13 +556,13 @@ def build_supervise_script_arguments(service_list, service_memory_config, config append_command(commands, "!p10 zk bin/run-zk conf") for service in service_list: - jvm_args = service_memory_config.get(service) + memory_config = service_memory_config.get(service) prefix = '' if service == MIDDLE_MANAGER: prefix = '!p90 ' - if jvm_args is None: + if memory_config is None: append_command(commands, '{0}{1} bin/run-druid {1} {2}'.format(prefix, service, config)) else: if service == MIDDLE_MANAGER: @@ -563,10 +572,17 @@ def build_supervise_script_arguments(service_list, service_memory_config, config append_command( commands, '{0}{1} bin/run-druid {1} {2} \'{3}\' \'{4} {5}\'' - .format(prefix, service, config, jvm_args, task_count, task_memory)) + .format(prefix, service, config, memory_config, task_count, task_memory)) + elif service == INDEXER: + task_count = memory_config[0] + jvm_args = memory_config[1] + append_command( + commands, + '{0}{1} bin/run-druid {1} {2} \'{3}\' \'{4}\'' + .format(prefix, service, config, jvm_args, task_count)) else: append_command(commands, - '{0}{1} bin/run-druid {1} {2} \'{3}\''.format(prefix, service, config, jvm_args)) + '{0}{1} bin/run-druid {1} {2} \'{3}\''.format(prefix, service, config, memory_config)) print_if_verbose('Supervise script args:') for item in commands: diff --git a/examples/conf/druid/auto/indexer/jvm.config b/examples/conf/druid/auto/indexer/jvm.config deleted file mode 100644 index 4611a65196a2..000000000000 --- a/examples/conf/druid/auto/indexer/jvm.config +++ /dev/null @@ -1,9 +0,0 @@ --server --Xms4g --Xmx4g --XX:MaxDirectMemorySize=4g --XX:+ExitOnOutOfMemoryError --Duser.timezone=UTC --Dfile.encoding=UTF-8 --Djava.io.tmpdir=var/tmp --Djava.util.logging.manager=org.apache.logging.log4j.jul.LogManager diff --git a/examples/conf/druid/auto/indexer/runtime.properties b/examples/conf/druid/auto/indexer/runtime.properties index b36c9eb42872..5aef64535b67 100644 --- a/examples/conf/druid/auto/indexer/runtime.properties +++ b/examples/conf/druid/auto/indexer/runtime.properties @@ -20,19 +20,15 @@ druid.service=druid/indexer druid.plaintextPort=8091 -# Number of tasks per indexer -druid.worker.capacity=4 +# Number of tasks (druid.worker.capacity) is automatically +# determined based on available processor. # Task launch parameters druid.indexer.task.baseTaskDir=var/druid/task -# HTTP server threads -druid.server.http.numThreads=60 - # Processing threads and buffers on Indexer -druid.processing.numMergeBuffers=2 -druid.processing.buffer.sizeBytes=100MiB -druid.processing.numThreads=4 +# Determined automatically based on available memory. For details on how to manually set parameters: +# https://druid.apache.org/docs/latest/operations/basic-cluster-tuning.html#guidelines-for-processing-threads-and-buffers # Hadoop indexing druid.indexer.task.hadoopWorkingPath=var/druid/hadoop-tmp From cf472162a6cde2808df90e158cb1acfb7418f8bb Mon Sep 17 00:00:00 2001 From: Clint Wylie Date: Tue, 6 Dec 2022 15:52:44 -0800 Subject: [PATCH 117/141] fix issue with jetty graceful shutdown of data servers when druid.serverview.type=http (#13499) * fix issue with http server inventory view blocking data node http server shutdown with long polling * adjust * fix test inspections --- .../apache/druid/guice/AnnouncerModule.java | 2 +- .../BatchDataSegmentAnnouncer.java | 8 ++++ .../coordination/ChangeRequestHistory.java | 19 +++++++++ .../ChangeRequestHistoryTest.java | 41 +++++++++++++++++++ 4 files changed, 69 insertions(+), 1 deletion(-) diff --git a/server/src/main/java/org/apache/druid/guice/AnnouncerModule.java b/server/src/main/java/org/apache/druid/guice/AnnouncerModule.java index 97a468542237..6b0c96641a65 100644 --- a/server/src/main/java/org/apache/druid/guice/AnnouncerModule.java +++ b/server/src/main/java/org/apache/druid/guice/AnnouncerModule.java @@ -54,7 +54,7 @@ public void configure(Binder binder) JsonConfigProvider.bind(binder, "druid.announcer", BatchDataSegmentAnnouncerConfig.class); JsonConfigProvider.bind(binder, "druid.announcer", DataSegmentAnnouncerProvider.class); binder.bind(DataSegmentAnnouncer.class).toProvider(DataSegmentAnnouncerProvider.class); - binder.bind(BatchDataSegmentAnnouncer.class).in(LazySingleton.class); + binder.bind(BatchDataSegmentAnnouncer.class).in(ManageLifecycleAnnouncements.class); if (isZkEnabled) { binder.bind(DataSegmentServerAnnouncer.class).to(CuratorDataSegmentServerAnnouncer.class).in(LazySingleton.class); diff --git a/server/src/main/java/org/apache/druid/server/coordination/BatchDataSegmentAnnouncer.java b/server/src/main/java/org/apache/druid/server/coordination/BatchDataSegmentAnnouncer.java index aa03ec6f4594..3d766be09832 100644 --- a/server/src/main/java/org/apache/druid/server/coordination/BatchDataSegmentAnnouncer.java +++ b/server/src/main/java/org/apache/druid/server/coordination/BatchDataSegmentAnnouncer.java @@ -36,6 +36,7 @@ import org.apache.druid.java.util.common.DateTimes; import org.apache.druid.java.util.common.ISE; import org.apache.druid.java.util.common.StringUtils; +import org.apache.druid.java.util.common.lifecycle.LifecycleStop; import org.apache.druid.java.util.common.logger.Logger; import org.apache.druid.server.initialization.BatchDataSegmentAnnouncerConfig; import org.apache.druid.server.initialization.ZkPathsConfig; @@ -129,6 +130,13 @@ public BatchDataSegmentAnnouncer( this(server, config, zkPaths, () -> announcer, jsonMapper, ZkEnablementConfig.ENABLED); } + @LifecycleStop + public void stop() + { + changes.stop(); + } + + @Override public void announceSegment(DataSegment segment) throws IOException { diff --git a/server/src/main/java/org/apache/druid/server/coordination/ChangeRequestHistory.java b/server/src/main/java/org/apache/druid/server/coordination/ChangeRequestHistory.java index 65c8dbd62473..a0c45b4a3352 100644 --- a/server/src/main/java/org/apache/druid/server/coordination/ChangeRequestHistory.java +++ b/server/src/main/java/org/apache/druid/server/coordination/ChangeRequestHistory.java @@ -32,6 +32,7 @@ import java.util.ArrayList; import java.util.LinkedHashMap; +import java.util.LinkedHashSet; import java.util.List; import java.util.Map; import java.util.concurrent.ExecutorService; @@ -43,6 +44,7 @@ * * Clients call {@link #getRequestsSince} to get updates since given counter. */ + public class ChangeRequestHistory { private static int MAX_SIZE = 1000; @@ -74,11 +76,24 @@ public ChangeRequestHistory(int maxSize) this.singleThreadedExecutor = Execs.singleThreaded("SegmentChangeRequestHistory"); } + public void stop() + { + singleThreadedExecutor.shutdownNow(); + final LinkedHashSet> futures = new LinkedHashSet<>(waitingFutures.keySet()); + waitingFutures.clear(); + for (CustomSettableFuture theFuture : futures) { + theFuture.setException(new IllegalStateException("Server is shutting down.")); + } + } + /** * Add batch of segment changes update. */ public synchronized void addChangeRequests(List requests) { + if (singleThreadedExecutor.isShutdown()) { + return; + } for (T request : requests) { changes.add(new Holder<>(request, getLastCounter().inc())); } @@ -108,6 +123,10 @@ public synchronized void addChangeRequest(T request) public synchronized ListenableFuture> getRequestsSince(final Counter counter) { final CustomSettableFuture future = new CustomSettableFuture<>(waitingFutures); + if (singleThreadedExecutor.isShutdown()) { + future.setException(new IllegalStateException("Server is shutting down.")); + return future; + } if (counter.counter < 0) { future.setException(new IAE("counter[%s] must be >= 0", counter)); diff --git a/server/src/test/java/org/apache/druid/server/coordination/ChangeRequestHistoryTest.java b/server/src/test/java/org/apache/druid/server/coordination/ChangeRequestHistoryTest.java index 82a67b038ed6..ecba53a8a031 100644 --- a/server/src/test/java/org/apache/druid/server/coordination/ChangeRequestHistoryTest.java +++ b/server/src/test/java/org/apache/druid/server/coordination/ChangeRequestHistoryTest.java @@ -25,6 +25,7 @@ import org.junit.Assert; import org.junit.Test; +import java.util.concurrent.ExecutionException; import java.util.concurrent.Future; import java.util.concurrent.TimeUnit; import java.util.concurrent.atomic.AtomicBoolean; @@ -171,4 +172,44 @@ public void testNonImmediateFuture() throws Exception Assert.assertEquals(1, snapshot.getCounter().getCounter()); Assert.assertEquals(1, snapshot.getRequests().size()); } + + @Test + public void testStop() + { + final ChangeRequestHistory history = new ChangeRequestHistory(); + + ListenableFuture> future = history.getRequestsSince( + ChangeRequestHistory.Counter.ZERO + ); + Assert.assertEquals(1, history.waitingFutures.size()); + + final AtomicBoolean callbackExcecuted = new AtomicBoolean(false); + Futures.addCallback( + future, + new FutureCallback>() + { + @Override + public void onSuccess(ChangeRequestsSnapshot result) + { + callbackExcecuted.set(true); + } + + @Override + public void onFailure(Throwable t) + { + callbackExcecuted.set(true); + } + } + ); + + history.stop(); + // any new change requests should be ignored, there should be no waiting futures, and open futures should be resolved + history.addChangeRequest(new SegmentChangeRequestNoop()); + Assert.assertEquals(0, history.waitingFutures.size()); + Assert.assertTrue(callbackExcecuted.get()); + Assert.assertTrue(future.isDone()); + + Throwable thrown = Assert.assertThrows(ExecutionException.class, future::get); + Assert.assertEquals("java.lang.IllegalStateException: Server is shutting down.", thrown.getMessage()); + } } From 83261f9641a750b121dbbb13b1cf53dd0c40f126 Mon Sep 17 00:00:00 2001 From: imply-cheddar <86940447+imply-cheddar@users.noreply.github.com> Date: Wed, 7 Dec 2022 08:54:05 +0900 Subject: [PATCH 118/141] Starting on Window Functions (#13458) * Processors for Window Processing This is an initial take on how to use Processors for Window Processing. A Processor is an interface that transforms RowsAndColumns objects. RowsAndColumns objects are essentially combinations of rows and columns. The intention is that these Processors are the start of a set of operators that more closely resemble what DB engineers would be accustomed to seeing. * Wire up windowed processors with a query type that can run them end-to-end. This code can be used to actually run a query, so yay! * Wire up windowed processors with a query type that can run them end-to-end. This code can be used to actually run a query, so yay! * Some SQL tests for window functions. Added wikipedia data to the indexes available to the SQL queries and tests validating the windowing functionality as it exists now. Co-authored-by: Gian Merlino --- .../druid/segment/column/TypeStrategies.java | 26 +- .../druid/segment/column/TypeStrategy.java | 9 +- .../segment/column/TypeStrategiesTest.java | 10 +- .../columnar/ComplexFrameColumnReader.java | 14 +- .../columnar/DoubleFrameColumnReader.java | 12 +- .../read/columnar/FloatFrameColumnReader.java | 12 +- .../read/columnar/LongFrameColumnReader.java | 12 +- .../columnar/ComplexFrameColumnWriter.java | 90 +- .../write/columnar/ComplexFrameMaker.java | 147 +++ .../columnar/DoubleFrameColumnWriter.java | 55 +- .../write/columnar/DoubleFrameMaker.java | 122 ++ .../columnar/FloatFrameColumnWriter.java | 55 +- .../frame/write/columnar/FloatFrameMaker.java | 122 ++ .../write/columnar/LongFrameColumnWriter.java | 55 +- .../frame/write/columnar/LongFrameMaker.java | 122 ++ .../apache/druid/query/InlineDataSource.java | 8 +- .../java/org/apache/druid/query/Query.java | 20 +- .../query/operator/LimitedRowsAndColumns.java | 115 ++ .../operator/NaivePartitioningOperator.java | 100 ++ .../NaivePartitioningOperatorFactory.java | 68 ++ .../apache/druid/query/operator/Operator.java | 93 ++ .../druid/query/operator/OperatorFactory.java | 59 + .../query/operator/OperatorSequence.java | 125 ++ .../SegmentToRowsAndColumnsOperator.java | 68 ++ .../query/operator/SequenceOperator.java | 87 ++ .../query/operator/WindowOperatorFactory.java | 68 ++ .../query/operator/WindowOperatorQuery.java | 167 +++ ...WindowOperatorQueryQueryRunnerFactory.java | 55 + .../WindowOperatorQueryQueryToolChest.java | 191 +++ .../operator/WindowProcessorOperator.java | 67 ++ .../operator/window/ComposingProcessor.java | 78 ++ .../query/operator/window/Processor.java | 83 ++ .../window/WindowAggregateProcessor.java | 131 +++ .../ranking/WindowCumeDistProcessor.java | 60 + .../ranking/WindowDenseRankProcessor.java | 59 + .../ranking/WindowPercentileProcessor.java | 106 ++ .../window/ranking/WindowRankProcessor.java | 108 ++ .../ranking/WindowRankingProcessorBase.java | 103 ++ .../ranking/WindowRowNumberProcessor.java | 131 +++ .../value/ShiftedColumnAccessorBase.java | 130 +++ .../window/value/WindowFirstProcessor.java | 50 + .../window/value/WindowLastProcessor.java | 53 + .../window/value/WindowOffsetProcessor.java | 89 ++ .../value/WindowValueProcessorBase.java | 103 ++ .../rowsandcols/AppendableRowsAndColumns.java | 38 + .../rowsandcols/ArrayListRowsAndColumns.java | 125 ++ .../rowsandcols/DefaultGroupPartitioner.java | 75 ++ .../DefaultOnHeapAggregatable.java | 271 +++++ .../DefaultSortedGroupPartitioner.java | 93 ++ .../query/rowsandcols/GroupPartitioner.java | 43 + .../query/rowsandcols/OnHeapAggregatable.java | 47 + .../OnHeapCumulativeAggregatable.java | 48 + .../query/rowsandcols/RowsAndColumns.java | 113 ++ .../rowsandcols/SortedGroupPartitioner.java | 54 + .../query/rowsandcols/column/Column.java | 63 + .../rowsandcols/column/ColumnAccessor.java | 104 ++ .../column/ColumnAccessorBasedColumn.java | 44 + .../column/ConstantObjectColumn.java | 103 ++ .../rowsandcols/column/DoubleArrayColumn.java | 101 ++ .../rowsandcols/column/IntArrayColumn.java | 101 ++ .../column/NullColumnAccessor.java | 96 ++ .../rowsandcols/column/ObjectArrayColumn.java | 81 ++ .../column/ObjectColumnAccessorBase.java | 121 ++ .../frame/AppendableMapOfColumns.java | 93 ++ .../frame/MapOfColumnsRowsAndColumns.java | 109 ++ .../druid/segment/ArrayListSegment.java | 126 ++ .../BaseSingleValueDimensionSelector.java | 1 - .../org/apache/druid/segment/Segment.java | 5 +- .../ObjectStrategyComplexTypeStrategy.java | 4 +- .../nested/NestedDataColumnSerializer.java | 4 +- .../query/operator/InlineScanOperator.java | 73 ++ .../NaivePartitioningOperatorTest.java | 105 ++ .../query/operator/OperatorSequenceTest.java | 56 + .../query/operator/SequenceOperatorTest.java | 56 + .../operator/WindowOperatorQueryTest.java | 124 ++ .../operator/WindowProcessorOperatorTest.java | 66 ++ .../window/ComposingProcessorTest.java | 74 ++ .../operator/window/RowsAndColumnsHelper.java | 269 +++++ .../window/WindowAggregateProcessorTest.java | 115 ++ .../ranking/WindowCumeDistProcessorTest.java | 53 + .../ranking/WindowDenseRankProcessorTest.java | 54 + .../WindowPercentileProcessorTest.java | 84 ++ .../ranking/WindowRankProcessorTest.java | 84 ++ .../ranking/WindowRowNumberProcessorTest.java | 61 + .../value/WindowFirstProcessorTest.java | 80 ++ .../window/value/WindowLagProcessorTest.java | 77 ++ .../window/value/WindowLastProcessorTest.java | 78 ++ .../window/value/WindowLeadProcessorTest.java | 76 ++ .../ArrayListRowsAndColumnsTest.java | 66 ++ .../MapOfColumnsRowsAndColumnsTest.java | 78 ++ .../rowsandcols/RowsAndColumnsTestBase.java | 188 +++ .../column/NullColumnAccessorTest.java | 46 + .../nested/NestedDataColumnSupplierTest.java | 2 +- .../druid/guice/QueryRunnerFactoryModule.java | 13 +- .../druid/guice/QueryToolChestModule.java | 13 +- .../druid/segment/InlineSegmentWrangler.java | 16 +- .../server/ClientQuerySegmentWalker.java | 39 +- .../druid/server/LocalQuerySegmentWalker.java | 10 +- .../server/coordination/ServerManager.java | 6 +- .../server/ClientQuerySegmentWalkerTest.java | 1 - .../apache/druid/server/QueryStackTests.java | 6 +- .../server/TestClusterQuerySegmentWalker.java | 4 - sql/pom.xml | 19 +- .../sql/calcite/expression/Expressions.java | 7 +- .../expression/WindowSqlAggregate.java | 70 ++ .../calcite/planner/CalciteRulesManager.java | 11 +- .../calcite/planner/DruidOperatorTable.java | 11 + .../druid/sql/calcite/rel/DruidQuery.java | 89 +- .../sql/calcite/rel/PartialDruidQuery.java | 80 +- .../druid/sql/calcite/rel/Windowing.java | 298 +++++ .../druid/sql/calcite/rule/DruidRules.java | 121 +- .../druid/sql/calcite/rule/GroupByRules.java | 3 +- .../sql/calcite/run/NativeQueryMaker.java | 3 +- .../sql/avatica/DruidAvaticaHandlerTest.java | 12 + .../sql/calcite/BaseCalciteQueryTest.java | 1 - .../druid/sql/calcite/CalciteQueryTest.java | 2 + .../sql/calcite/CalciteWindowQueryTest.java | 229 ++++ .../druid/sql/calcite/QueryTestBuilder.java | 27 +- .../druid/sql/calcite/QueryTestRunner.java | 63 +- .../druid/sql/calcite/QueryVerification.java | 55 + .../SpecificSegmentsQuerySegmentWalker.java | 1 - .../sql/calcite/util/SqlTestFramework.java | 13 +- .../sql/calcite/util/TestDataBuilder.java | 575 +++++---- .../tests/wikiticker-2015-09-12-sampled.jgz | Bin 0 -> 2366222 bytes .../calcite/tests/window/simpleSum.sqlTest | 25 + .../window/wikipediaCumulativeOrdered.sqlTest | 1028 +++++++++++++++++ .../window/wikipediaSimplePartition.sqlTest | 1019 ++++++++++++++++ 127 files changed, 10654 insertions(+), 682 deletions(-) create mode 100644 processing/src/main/java/org/apache/druid/frame/write/columnar/ComplexFrameMaker.java create mode 100644 processing/src/main/java/org/apache/druid/frame/write/columnar/DoubleFrameMaker.java create mode 100644 processing/src/main/java/org/apache/druid/frame/write/columnar/FloatFrameMaker.java create mode 100644 processing/src/main/java/org/apache/druid/frame/write/columnar/LongFrameMaker.java create mode 100644 processing/src/main/java/org/apache/druid/query/operator/LimitedRowsAndColumns.java create mode 100644 processing/src/main/java/org/apache/druid/query/operator/NaivePartitioningOperator.java create mode 100644 processing/src/main/java/org/apache/druid/query/operator/NaivePartitioningOperatorFactory.java create mode 100644 processing/src/main/java/org/apache/druid/query/operator/Operator.java create mode 100644 processing/src/main/java/org/apache/druid/query/operator/OperatorFactory.java create mode 100644 processing/src/main/java/org/apache/druid/query/operator/OperatorSequence.java create mode 100644 processing/src/main/java/org/apache/druid/query/operator/SegmentToRowsAndColumnsOperator.java create mode 100644 processing/src/main/java/org/apache/druid/query/operator/SequenceOperator.java create mode 100644 processing/src/main/java/org/apache/druid/query/operator/WindowOperatorFactory.java create mode 100644 processing/src/main/java/org/apache/druid/query/operator/WindowOperatorQuery.java create mode 100644 processing/src/main/java/org/apache/druid/query/operator/WindowOperatorQueryQueryRunnerFactory.java create mode 100644 processing/src/main/java/org/apache/druid/query/operator/WindowOperatorQueryQueryToolChest.java create mode 100644 processing/src/main/java/org/apache/druid/query/operator/WindowProcessorOperator.java create mode 100644 processing/src/main/java/org/apache/druid/query/operator/window/ComposingProcessor.java create mode 100644 processing/src/main/java/org/apache/druid/query/operator/window/Processor.java create mode 100644 processing/src/main/java/org/apache/druid/query/operator/window/WindowAggregateProcessor.java create mode 100644 processing/src/main/java/org/apache/druid/query/operator/window/ranking/WindowCumeDistProcessor.java create mode 100644 processing/src/main/java/org/apache/druid/query/operator/window/ranking/WindowDenseRankProcessor.java create mode 100644 processing/src/main/java/org/apache/druid/query/operator/window/ranking/WindowPercentileProcessor.java create mode 100644 processing/src/main/java/org/apache/druid/query/operator/window/ranking/WindowRankProcessor.java create mode 100644 processing/src/main/java/org/apache/druid/query/operator/window/ranking/WindowRankingProcessorBase.java create mode 100644 processing/src/main/java/org/apache/druid/query/operator/window/ranking/WindowRowNumberProcessor.java create mode 100644 processing/src/main/java/org/apache/druid/query/operator/window/value/ShiftedColumnAccessorBase.java create mode 100644 processing/src/main/java/org/apache/druid/query/operator/window/value/WindowFirstProcessor.java create mode 100644 processing/src/main/java/org/apache/druid/query/operator/window/value/WindowLastProcessor.java create mode 100644 processing/src/main/java/org/apache/druid/query/operator/window/value/WindowOffsetProcessor.java create mode 100644 processing/src/main/java/org/apache/druid/query/operator/window/value/WindowValueProcessorBase.java create mode 100644 processing/src/main/java/org/apache/druid/query/rowsandcols/AppendableRowsAndColumns.java create mode 100644 processing/src/main/java/org/apache/druid/query/rowsandcols/ArrayListRowsAndColumns.java create mode 100644 processing/src/main/java/org/apache/druid/query/rowsandcols/DefaultGroupPartitioner.java create mode 100644 processing/src/main/java/org/apache/druid/query/rowsandcols/DefaultOnHeapAggregatable.java create mode 100644 processing/src/main/java/org/apache/druid/query/rowsandcols/DefaultSortedGroupPartitioner.java create mode 100644 processing/src/main/java/org/apache/druid/query/rowsandcols/GroupPartitioner.java create mode 100644 processing/src/main/java/org/apache/druid/query/rowsandcols/OnHeapAggregatable.java create mode 100644 processing/src/main/java/org/apache/druid/query/rowsandcols/OnHeapCumulativeAggregatable.java create mode 100644 processing/src/main/java/org/apache/druid/query/rowsandcols/RowsAndColumns.java create mode 100644 processing/src/main/java/org/apache/druid/query/rowsandcols/SortedGroupPartitioner.java create mode 100644 processing/src/main/java/org/apache/druid/query/rowsandcols/column/Column.java create mode 100644 processing/src/main/java/org/apache/druid/query/rowsandcols/column/ColumnAccessor.java create mode 100644 processing/src/main/java/org/apache/druid/query/rowsandcols/column/ColumnAccessorBasedColumn.java create mode 100644 processing/src/main/java/org/apache/druid/query/rowsandcols/column/ConstantObjectColumn.java create mode 100644 processing/src/main/java/org/apache/druid/query/rowsandcols/column/DoubleArrayColumn.java create mode 100644 processing/src/main/java/org/apache/druid/query/rowsandcols/column/IntArrayColumn.java create mode 100644 processing/src/main/java/org/apache/druid/query/rowsandcols/column/NullColumnAccessor.java create mode 100644 processing/src/main/java/org/apache/druid/query/rowsandcols/column/ObjectArrayColumn.java create mode 100644 processing/src/main/java/org/apache/druid/query/rowsandcols/column/ObjectColumnAccessorBase.java create mode 100644 processing/src/main/java/org/apache/druid/query/rowsandcols/frame/AppendableMapOfColumns.java create mode 100644 processing/src/main/java/org/apache/druid/query/rowsandcols/frame/MapOfColumnsRowsAndColumns.java create mode 100644 processing/src/main/java/org/apache/druid/segment/ArrayListSegment.java create mode 100644 processing/src/test/java/org/apache/druid/query/operator/InlineScanOperator.java create mode 100644 processing/src/test/java/org/apache/druid/query/operator/NaivePartitioningOperatorTest.java create mode 100644 processing/src/test/java/org/apache/druid/query/operator/OperatorSequenceTest.java create mode 100644 processing/src/test/java/org/apache/druid/query/operator/SequenceOperatorTest.java create mode 100644 processing/src/test/java/org/apache/druid/query/operator/WindowOperatorQueryTest.java create mode 100644 processing/src/test/java/org/apache/druid/query/operator/WindowProcessorOperatorTest.java create mode 100644 processing/src/test/java/org/apache/druid/query/operator/window/ComposingProcessorTest.java create mode 100644 processing/src/test/java/org/apache/druid/query/operator/window/RowsAndColumnsHelper.java create mode 100644 processing/src/test/java/org/apache/druid/query/operator/window/WindowAggregateProcessorTest.java create mode 100644 processing/src/test/java/org/apache/druid/query/operator/window/ranking/WindowCumeDistProcessorTest.java create mode 100644 processing/src/test/java/org/apache/druid/query/operator/window/ranking/WindowDenseRankProcessorTest.java create mode 100644 processing/src/test/java/org/apache/druid/query/operator/window/ranking/WindowPercentileProcessorTest.java create mode 100644 processing/src/test/java/org/apache/druid/query/operator/window/ranking/WindowRankProcessorTest.java create mode 100644 processing/src/test/java/org/apache/druid/query/operator/window/ranking/WindowRowNumberProcessorTest.java create mode 100644 processing/src/test/java/org/apache/druid/query/operator/window/value/WindowFirstProcessorTest.java create mode 100644 processing/src/test/java/org/apache/druid/query/operator/window/value/WindowLagProcessorTest.java create mode 100644 processing/src/test/java/org/apache/druid/query/operator/window/value/WindowLastProcessorTest.java create mode 100644 processing/src/test/java/org/apache/druid/query/operator/window/value/WindowLeadProcessorTest.java create mode 100644 processing/src/test/java/org/apache/druid/query/rowsandcols/ArrayListRowsAndColumnsTest.java create mode 100644 processing/src/test/java/org/apache/druid/query/rowsandcols/MapOfColumnsRowsAndColumnsTest.java create mode 100644 processing/src/test/java/org/apache/druid/query/rowsandcols/RowsAndColumnsTestBase.java create mode 100644 processing/src/test/java/org/apache/druid/query/rowsandcols/column/NullColumnAccessorTest.java create mode 100644 sql/src/main/java/org/apache/druid/sql/calcite/expression/WindowSqlAggregate.java create mode 100644 sql/src/main/java/org/apache/druid/sql/calcite/rel/Windowing.java create mode 100644 sql/src/test/java/org/apache/druid/sql/calcite/CalciteWindowQueryTest.java create mode 100644 sql/src/test/java/org/apache/druid/sql/calcite/QueryVerification.java create mode 100644 sql/src/test/resources/calcite/tests/wikiticker-2015-09-12-sampled.jgz create mode 100644 sql/src/test/resources/calcite/tests/window/simpleSum.sqlTest create mode 100644 sql/src/test/resources/calcite/tests/window/wikipediaCumulativeOrdered.sqlTest create mode 100644 sql/src/test/resources/calcite/tests/window/wikipediaSimplePartition.sqlTest diff --git a/core/src/main/java/org/apache/druid/segment/column/TypeStrategies.java b/core/src/main/java/org/apache/druid/segment/column/TypeStrategies.java index 54a15b1dbdf3..20507c597b0f 100644 --- a/core/src/main/java/org/apache/druid/segment/column/TypeStrategies.java +++ b/core/src/main/java/org/apache/druid/segment/column/TypeStrategies.java @@ -236,8 +236,6 @@ public static void checkMaxSize(int available, int maxSizeBytes, TypeSignature { - private static final Comparator COMPARATOR = Longs::compare; - @Override public int estimateSizeBytes(Long value) { @@ -276,9 +274,9 @@ public int write(ByteBuffer buffer, Long value, int maxSizeBytes) } @Override - public int compare(Long o1, Long o2) + public int compare(Object o1, Object o2) { - return COMPARATOR.compare(o1, o2); + return Longs.compare(((Number) o1).longValue(), ((Number) o2).longValue()); } } @@ -289,8 +287,6 @@ public int compare(Long o1, Long o2) */ public static final class FloatTypeStrategy implements TypeStrategy { - private static final Comparator COMPARATOR = Floats::compare; - @Override public int estimateSizeBytes(Float value) { @@ -329,9 +325,9 @@ public int write(ByteBuffer buffer, Float value, int maxSizeBytes) } @Override - public int compare(Float o1, Float o2) + public int compare(Object o1, Object o2) { - return COMPARATOR.compare(o1, o2); + return Floats.compare(((Number) o1).floatValue(), ((Number) o2).floatValue()); } } @@ -342,7 +338,6 @@ public int compare(Float o1, Float o2) */ public static final class DoubleTypeStrategy implements TypeStrategy { - private static final Comparator COMPARATOR = Double::compare; @Override public int estimateSizeBytes(Double value) @@ -382,9 +377,9 @@ public int write(ByteBuffer buffer, Double value, int maxSizeBytes) } @Override - public int compare(Double o1, Double o2) + public int compare(Object o1, Object o2) { - return COMPARATOR.compare(o1, o2); + return Double.compare(((Number) o1).doubleValue(), ((Number) o2).doubleValue()); } } @@ -437,7 +432,7 @@ public int write(ByteBuffer buffer, String value, int maxSizeBytes) } @Override - public int compare(String s, String s2) + public int compare(Object s, Object s2) { // copy of lexicographical string comparator in druid processing // Avoid comparisons for equal references @@ -447,7 +442,7 @@ public int compare(String s, String s2) return 0; } - return ORDERING.compare(s, s2); + return ORDERING.compare((String) s, (String) s2); } } @@ -521,8 +516,11 @@ public int write(ByteBuffer buffer, Object[] value, int maxSizeBytes) } @Override - public int compare(@Nullable Object[] o1, @Nullable Object[] o2) + public int compare(@Nullable Object o1Obj, @Nullable Object o2Obj) { + Object[] o1 = (Object[]) o1Obj; + Object[] o2 = (Object[]) o2Obj; + //noinspection ArrayEquality if (o1 == o2) { return 0; diff --git a/core/src/main/java/org/apache/druid/segment/column/TypeStrategy.java b/core/src/main/java/org/apache/druid/segment/column/TypeStrategy.java index 8a97882d54df..e4856f889714 100644 --- a/core/src/main/java/org/apache/druid/segment/column/TypeStrategy.java +++ b/core/src/main/java/org/apache/druid/segment/column/TypeStrategy.java @@ -55,8 +55,15 @@ * Implementations of this interface should be thread safe, but may not use {@link ByteBuffer} in a thread safe manner, * potentially modifying positions and limits, either temporarily or permanently depending on which set of methods is * called. + * + * This interface extends {@code Comparator} instead of {@code Comparator} because trying to specialize the + * type of the comparison method can run into issues for comparators of objects that can sometimes be of a different + * java class type. For example, {@code Comparator} cannot accept Integer objects in its comparison method + * and there is no easy way for this interface definition to allow {@code TypeStrategy} to actually be a + * {@code Comparator}. So, we fall back to effectively erasing the generic type and having them all be + * {@code Comparator}. */ -public interface TypeStrategy extends Comparator +public interface TypeStrategy extends Comparator { /** * Estimate the size in bytes that writing this value to memory would require. This method is not required to be diff --git a/core/src/test/java/org/apache/druid/segment/column/TypeStrategiesTest.java b/core/src/test/java/org/apache/druid/segment/column/TypeStrategiesTest.java index fa6d86d21f50..66f4adcdd02d 100644 --- a/core/src/test/java/org/apache/druid/segment/column/TypeStrategiesTest.java +++ b/core/src/test/java/org/apache/druid/segment/column/TypeStrategiesTest.java @@ -19,6 +19,7 @@ package org.apache.druid.segment.column; +import com.google.common.collect.Ordering; import com.google.common.primitives.Longs; import org.apache.druid.java.util.common.IAE; import org.apache.druid.java.util.common.Pair; @@ -103,7 +104,7 @@ public int write(ByteBuffer buffer, String value, int maxSizeBytes) } @Override - public int compare(String o1, String o2) + public int compare(Object o1, Object o2) { return 0; } @@ -639,10 +640,13 @@ public int compareTo(NullableLongPair o) public static class NullableLongPairTypeStrategy implements TypeStrategy { + + private Ordering ordering = Comparators.naturalNullsFirst(); + @Override - public int compare(NullableLongPair o1, NullableLongPair o2) + public int compare(Object o1, Object o2) { - return Comparators.naturalNullsFirst().compare(o1, o2); + return ordering.compare((NullableLongPair) o1, (NullableLongPair) o2); } @Override diff --git a/processing/src/main/java/org/apache/druid/frame/read/columnar/ComplexFrameColumnReader.java b/processing/src/main/java/org/apache/druid/frame/read/columnar/ComplexFrameColumnReader.java index f7b662d42bb2..e4a39c16f753 100644 --- a/processing/src/main/java/org/apache/druid/frame/read/columnar/ComplexFrameColumnReader.java +++ b/processing/src/main/java/org/apache/druid/frame/read/columnar/ComplexFrameColumnReader.java @@ -22,7 +22,7 @@ import com.google.common.primitives.Ints; import org.apache.datasketches.memory.Memory; import org.apache.druid.frame.Frame; -import org.apache.druid.frame.write.columnar.ComplexFrameColumnWriter; +import org.apache.druid.frame.write.columnar.ComplexFrameMaker; import org.apache.druid.frame.write.columnar.FrameColumnWriters; import org.apache.druid.java.util.common.ISE; import org.apache.druid.java.util.common.StringUtils; @@ -53,10 +53,10 @@ public ColumnPlus readColumn(final Frame frame) final Memory memory = frame.region(columnNumber); validate(memory, frame.numRows()); - final int typeNameLength = memory.getInt(ComplexFrameColumnWriter.TYPE_NAME_LENGTH_POSITION); + final int typeNameLength = memory.getInt(ComplexFrameMaker.TYPE_NAME_LENGTH_POSITION); final byte[] typeNameBytes = new byte[typeNameLength]; - memory.getByteArray(ComplexFrameColumnWriter.TYPE_NAME_POSITION, typeNameBytes, 0, typeNameLength); + memory.getByteArray(ComplexFrameMaker.TYPE_NAME_POSITION, typeNameBytes, 0, typeNameLength); final String typeName = StringUtils.fromUtf8(typeNameBytes); final ComplexMetricSerde serde = ComplexMetrics.getSerdeForType(typeName); @@ -84,7 +84,7 @@ public ColumnPlus readColumn(final Frame frame) private void validate(final Memory region, final int numRows) { - if (region.getCapacity() < ComplexFrameColumnWriter.TYPE_NAME_POSITION) { + if (region.getCapacity() < ComplexFrameMaker.TYPE_NAME_POSITION) { throw new ISE("Column is not big enough for a header"); } @@ -93,9 +93,9 @@ private void validate(final Memory region, final int numRows) throw new ISE("Column does not have the correct type code"); } - final int typeNameLength = region.getInt(ComplexFrameColumnWriter.TYPE_NAME_LENGTH_POSITION); + final int typeNameLength = region.getInt(ComplexFrameMaker.TYPE_NAME_LENGTH_POSITION); if (region.getCapacity() < - ComplexFrameColumnWriter.TYPE_NAME_POSITION + typeNameLength + (long) numRows * Integer.BYTES) { + ComplexFrameMaker.TYPE_NAME_POSITION + typeNameLength + (long) numRows * Integer.BYTES) { throw new ISE("Column is missing offset section"); } } @@ -198,7 +198,7 @@ private Object getObjectForPhysicalRow(final int physicalRow) startOfDataSection + memory.getInt(startOfOffsetSection + (long) Integer.BYTES * (physicalRow - 1)); } - if (memory.getByte(startOffset) == ComplexFrameColumnWriter.NULL_MARKER) { + if (memory.getByte(startOffset) == ComplexFrameMaker.NULL_MARKER) { return null; } else { final int payloadLength = Ints.checkedCast(endOffset - startOffset - Byte.BYTES); diff --git a/processing/src/main/java/org/apache/druid/frame/read/columnar/DoubleFrameColumnReader.java b/processing/src/main/java/org/apache/druid/frame/read/columnar/DoubleFrameColumnReader.java index a8f08feaa6a4..8ac4b1d15d65 100644 --- a/processing/src/main/java/org/apache/druid/frame/read/columnar/DoubleFrameColumnReader.java +++ b/processing/src/main/java/org/apache/druid/frame/read/columnar/DoubleFrameColumnReader.java @@ -22,7 +22,7 @@ import org.apache.datasketches.memory.Memory; import org.apache.druid.common.config.NullHandling; import org.apache.druid.frame.Frame; -import org.apache.druid.frame.write.columnar.DoubleFrameColumnWriter; +import org.apache.druid.frame.write.columnar.DoubleFrameMaker; import org.apache.druid.frame.write.columnar.FrameColumnWriters; import org.apache.druid.java.util.common.ISE; import org.apache.druid.query.monomorphicprocessing.RuntimeShapeInspector; @@ -69,7 +69,7 @@ private void validate(final Memory region, final int numRows) final long memorySize = region.getCapacity(); // Check if column is big enough for a header - if (memorySize < DoubleFrameColumnWriter.DATA_OFFSET) { + if (memorySize < DoubleFrameMaker.DATA_OFFSET) { throw new ISE("Column is not big enough for a header"); } @@ -79,10 +79,10 @@ private void validate(final Memory region, final int numRows) } final boolean hasNulls = getHasNulls(region); - final int sz = DoubleFrameColumnWriter.valueSize(hasNulls); + final int sz = DoubleFrameMaker.valueSize(hasNulls); // Check column length again, now that we know exactly how long it should be. - if (memorySize != DoubleFrameColumnWriter.DATA_OFFSET + (long) sz * numRows) { + if (memorySize != DoubleFrameMaker.DATA_OFFSET + (long) sz * numRows) { throw new ISE("Column does not have the correct length"); } } @@ -108,9 +108,9 @@ private DoubleFrameColumn( { this.frame = frame; this.hasNulls = hasNulls; - this.sz = DoubleFrameColumnWriter.valueSize(hasNulls); + this.sz = DoubleFrameMaker.valueSize(hasNulls); this.memory = memory; - this.memoryPosition = DoubleFrameColumnWriter.DATA_OFFSET; + this.memoryPosition = DoubleFrameMaker.DATA_OFFSET; } @Override diff --git a/processing/src/main/java/org/apache/druid/frame/read/columnar/FloatFrameColumnReader.java b/processing/src/main/java/org/apache/druid/frame/read/columnar/FloatFrameColumnReader.java index dc5c7fa18f27..0f1d09f57bc5 100644 --- a/processing/src/main/java/org/apache/druid/frame/read/columnar/FloatFrameColumnReader.java +++ b/processing/src/main/java/org/apache/druid/frame/read/columnar/FloatFrameColumnReader.java @@ -22,7 +22,7 @@ import org.apache.datasketches.memory.Memory; import org.apache.druid.common.config.NullHandling; import org.apache.druid.frame.Frame; -import org.apache.druid.frame.write.columnar.FloatFrameColumnWriter; +import org.apache.druid.frame.write.columnar.FloatFrameMaker; import org.apache.druid.frame.write.columnar.FrameColumnWriters; import org.apache.druid.java.util.common.ISE; import org.apache.druid.query.monomorphicprocessing.RuntimeShapeInspector; @@ -69,7 +69,7 @@ private void validate(final Memory region, final int numRows) final long memorySize = region.getCapacity(); // Check if column is big enough for a header - if (memorySize < FloatFrameColumnWriter.DATA_OFFSET) { + if (memorySize < FloatFrameMaker.DATA_OFFSET) { throw new ISE("Column is not big enough for a header"); } @@ -79,10 +79,10 @@ private void validate(final Memory region, final int numRows) } final boolean hasNulls = getHasNulls(region); - final int sz = FloatFrameColumnWriter.valueSize(hasNulls); + final int sz = FloatFrameMaker.valueSize(hasNulls); // Check column length again, now that we know exactly how long it should be. - if (memorySize != FloatFrameColumnWriter.DATA_OFFSET + (long) sz * numRows) { + if (memorySize != FloatFrameMaker.DATA_OFFSET + (long) sz * numRows) { throw new ISE("Column does not have the correct length"); } } @@ -108,9 +108,9 @@ private FloatFrameColumn( { this.frame = frame; this.hasNulls = hasNulls; - this.sz = FloatFrameColumnWriter.valueSize(hasNulls); + this.sz = FloatFrameMaker.valueSize(hasNulls); this.memory = memory; - this.memoryPosition = FloatFrameColumnWriter.DATA_OFFSET; + this.memoryPosition = FloatFrameMaker.DATA_OFFSET; } @Override diff --git a/processing/src/main/java/org/apache/druid/frame/read/columnar/LongFrameColumnReader.java b/processing/src/main/java/org/apache/druid/frame/read/columnar/LongFrameColumnReader.java index 91bd898597bf..1662411b912b 100644 --- a/processing/src/main/java/org/apache/druid/frame/read/columnar/LongFrameColumnReader.java +++ b/processing/src/main/java/org/apache/druid/frame/read/columnar/LongFrameColumnReader.java @@ -23,7 +23,7 @@ import org.apache.druid.common.config.NullHandling; import org.apache.druid.frame.Frame; import org.apache.druid.frame.write.columnar.FrameColumnWriters; -import org.apache.druid.frame.write.columnar.LongFrameColumnWriter; +import org.apache.druid.frame.write.columnar.LongFrameMaker; import org.apache.druid.java.util.common.ISE; import org.apache.druid.query.monomorphicprocessing.RuntimeShapeInspector; import org.apache.druid.segment.ColumnValueSelector; @@ -67,7 +67,7 @@ public ColumnPlus readColumn(final Frame frame) private void validate(final Memory region, final int numRows) { // Check if column is big enough for a header - if (region.getCapacity() < LongFrameColumnWriter.DATA_OFFSET) { + if (region.getCapacity() < LongFrameMaker.DATA_OFFSET) { throw new ISE("Column is not big enough for a header"); } @@ -77,10 +77,10 @@ private void validate(final Memory region, final int numRows) } final boolean hasNulls = getHasNulls(region); - final int sz = LongFrameColumnWriter.valueSize(hasNulls); + final int sz = LongFrameMaker.valueSize(hasNulls); // Check column length again, now that we know exactly how long it should be. - if (region.getCapacity() != LongFrameColumnWriter.DATA_OFFSET + (long) sz * numRows) { + if (region.getCapacity() != LongFrameMaker.DATA_OFFSET + (long) sz * numRows) { throw new ISE("Column does not have the correct length"); } } @@ -106,9 +106,9 @@ private LongFrameColumn( { this.frame = frame; this.hasNulls = hasNulls; - this.sz = LongFrameColumnWriter.valueSize(hasNulls); + this.sz = LongFrameMaker.valueSize(hasNulls); this.memory = memory; - this.memoryPosition = LongFrameColumnWriter.DATA_OFFSET; + this.memoryPosition = LongFrameMaker.DATA_OFFSET; } @Override diff --git a/processing/src/main/java/org/apache/druid/frame/write/columnar/ComplexFrameColumnWriter.java b/processing/src/main/java/org/apache/druid/frame/write/columnar/ComplexFrameColumnWriter.java index 08a8aa0b67b4..6e6b985334df 100644 --- a/processing/src/main/java/org/apache/druid/frame/write/columnar/ComplexFrameColumnWriter.java +++ b/processing/src/main/java/org/apache/druid/frame/write/columnar/ComplexFrameColumnWriter.java @@ -19,14 +19,9 @@ package org.apache.druid.frame.write.columnar; -import com.google.common.primitives.Ints; -import it.unimi.dsi.fastutil.bytes.ByteArrays; import org.apache.datasketches.memory.WritableMemory; -import org.apache.druid.frame.allocation.AppendableMemory; import org.apache.druid.frame.allocation.MemoryAllocator; -import org.apache.druid.frame.allocation.MemoryRange; import org.apache.druid.frame.read.columnar.ComplexFrameColumnReader; -import org.apache.druid.java.util.common.ISE; import org.apache.druid.java.util.common.StringUtils; import org.apache.druid.segment.BaseObjectColumnValueSelector; import org.apache.druid.segment.serde.ComplexMetricSerde; @@ -38,22 +33,9 @@ */ public class ComplexFrameColumnWriter implements FrameColumnWriter { - // Less than half of AppendableMemory.DEFAULT_INITIAL_ALLOCATION_SIZE. - // This guarantees we can fit a WorkerMemoryParmeters.MAX_FRAME_COLUMNS number of columns into a frame. - private static final int INITIAL_ALLOCATION_SIZE = 128; - - public static final byte NOT_NULL_MARKER = 0x00; - public static final byte NULL_MARKER = 0x01; - public static final int TYPE_NAME_LENGTH_POSITION = Byte.BYTES; - public static final int TYPE_NAME_POSITION = Byte.BYTES + Integer.BYTES; - private final ComplexMetricSerde serde; private final BaseObjectColumnValueSelector selector; - private final AppendableMemory offsetMemory; - private final AppendableMemory dataMemory; - private final byte[] typeNameBytes; - - private int lastDataLength = -1; + private final ComplexFrameMaker maker; ComplexFrameColumnWriter( final BaseObjectColumnValueSelector selector, @@ -63,95 +45,37 @@ public class ComplexFrameColumnWriter implements FrameColumnWriter { this.selector = selector; this.serde = serde; - this.offsetMemory = AppendableMemory.create(allocator, INITIAL_ALLOCATION_SIZE); - this.dataMemory = AppendableMemory.create(allocator, INITIAL_ALLOCATION_SIZE); - this.typeNameBytes = StringUtils.toUtf8(serde.getTypeName()); + this.maker = new ComplexFrameMaker(allocator, StringUtils.toUtf8(serde.getTypeName())); } @Override public boolean addSelection() { - if (!offsetMemory.reserveAdditional(Integer.BYTES)) { - return false; - } - final Object complexObject = selector.getObject(); - final byte[] complexBytes = complexObject == null ? ByteArrays.EMPTY_ARRAY : serde.toBytes(complexObject); - - if (complexBytes.length == Integer.MAX_VALUE) { - // Cannot handle objects this large. - return false; - } - - final int dataLength = complexBytes.length + 1; - - if (dataMemory.size() + dataLength > Integer.MAX_VALUE || !(dataMemory.reserveAdditional(dataLength))) { - return false; - } - - // All space is reserved. Start writing. - final MemoryRange offsetCursor = offsetMemory.cursor(); - offsetCursor.memory().putInt(offsetCursor.start(), Ints.checkedCast(dataMemory.size() + dataLength)); - offsetMemory.advanceCursor(Integer.BYTES); - - final MemoryRange dataCursor = dataMemory.cursor(); - dataCursor.memory().putByte(dataCursor.start(), complexObject == null ? NULL_MARKER : NOT_NULL_MARKER); - dataCursor.memory().putByteArray(dataCursor.start() + 1, complexBytes, 0, complexBytes.length); - dataMemory.advanceCursor(dataLength); - - lastDataLength = dataLength; - return true; + return maker.add(complexObject == null ? null : serde.toBytes(complexObject)); } @Override public void undo() { - if (lastDataLength == -1) { - throw new ISE("Nothing to undo"); - } - - offsetMemory.rewindCursor(Integer.BYTES); - dataMemory.rewindCursor(lastDataLength); - lastDataLength = -1; + maker.undo(); } @Override public long size() { - return headerSize() + offsetMemory.size() + dataMemory.size(); + return maker.size(); } @Override public long writeTo(final WritableMemory memory, final long startPosition) { - long currentPosition = startPosition; - - memory.putByte(currentPosition, FrameColumnWriters.TYPE_COMPLEX); - currentPosition += 1; - - memory.putInt(currentPosition, typeNameBytes.length); - currentPosition += Integer.BYTES; - - memory.putByteArray(currentPosition, typeNameBytes, 0, typeNameBytes.length); - currentPosition += typeNameBytes.length; - - currentPosition += offsetMemory.writeTo(memory, currentPosition); - currentPosition += dataMemory.writeTo(memory, currentPosition); - - return currentPosition - startPosition; + return maker.writeTo(memory, startPosition); } @Override public void close() { - offsetMemory.close(); - dataMemory.close(); - } - - private int headerSize() - { - return 1 /* type code */ - + Integer.BYTES /* type name length */ - + typeNameBytes.length; + maker.close(); } } diff --git a/processing/src/main/java/org/apache/druid/frame/write/columnar/ComplexFrameMaker.java b/processing/src/main/java/org/apache/druid/frame/write/columnar/ComplexFrameMaker.java new file mode 100644 index 000000000000..8622a7bce202 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/frame/write/columnar/ComplexFrameMaker.java @@ -0,0 +1,147 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.frame.write.columnar; + +import com.google.common.primitives.Ints; +import it.unimi.dsi.fastutil.bytes.ByteArrays; +import org.apache.datasketches.memory.WritableMemory; +import org.apache.druid.frame.allocation.AppendableMemory; +import org.apache.druid.frame.allocation.MemoryAllocator; +import org.apache.druid.frame.allocation.MemoryRange; +import org.apache.druid.frame.read.columnar.ComplexFrameColumnReader; +import org.apache.druid.java.util.common.ISE; + +/** + * Column writer for complex columns. + * + * Dual to {@link ComplexFrameColumnReader}. + */ +public class ComplexFrameMaker +{ + // Less than half of AppendableMemory.DEFAULT_INITIAL_ALLOCATION_SIZE. + // This guarantees we can fit a WorkerMemoryParmeters.MAX_FRAME_COLUMNS number of columns into a frame. + private static final int INITIAL_ALLOCATION_SIZE = 128; + + public static final byte NOT_NULL_MARKER = 0x00; + public static final byte NULL_MARKER = 0x01; + public static final int TYPE_NAME_LENGTH_POSITION = Byte.BYTES; + public static final int TYPE_NAME_POSITION = Byte.BYTES + Integer.BYTES; + + private final AppendableMemory offsetMemory; + private final AppendableMemory dataMemory; + private final byte[] typeName; + + private int lastDataLength = -1; + + ComplexFrameMaker( + final MemoryAllocator allocator, + final byte[] typeName + ) + { + this.offsetMemory = AppendableMemory.create(allocator, INITIAL_ALLOCATION_SIZE); + this.dataMemory = AppendableMemory.create(allocator, INITIAL_ALLOCATION_SIZE); + this.typeName = typeName; + } + + public boolean add(byte[] bytes) + { + if (!offsetMemory.reserveAdditional(Integer.BYTES)) { + return false; + } + + boolean isNull = false; + if (bytes == null) { + isNull = true; + bytes = ByteArrays.EMPTY_ARRAY; + } + + if (bytes.length == Integer.MAX_VALUE) { + // Cannot handle objects this large. + return false; + } + + final int dataLength = bytes.length + 1; + + if (dataMemory.size() + dataLength > Integer.MAX_VALUE || !(dataMemory.reserveAdditional(dataLength))) { + return false; + } + + // All space is reserved. Start writing. + final MemoryRange offsetCursor = offsetMemory.cursor(); + offsetCursor.memory().putInt(offsetCursor.start(), Ints.checkedCast(dataMemory.size() + dataLength)); + offsetMemory.advanceCursor(Integer.BYTES); + + final MemoryRange dataCursor = dataMemory.cursor(); + dataCursor.memory().putByte(dataCursor.start(), isNull ? NULL_MARKER : NOT_NULL_MARKER); + dataCursor.memory().putByteArray(dataCursor.start() + 1, bytes, 0, bytes.length); + dataMemory.advanceCursor(dataLength); + + lastDataLength = dataLength; + return true; + } + + public void undo() + { + if (lastDataLength == -1) { + throw new ISE("Nothing to undo"); + } + + offsetMemory.rewindCursor(Integer.BYTES); + dataMemory.rewindCursor(lastDataLength); + lastDataLength = -1; + } + + public long size() + { + return headerSize() + offsetMemory.size() + dataMemory.size(); + } + + public long writeTo(final WritableMemory memory, final long startPosition) + { + long currentPosition = startPosition; + + memory.putByte(currentPosition, FrameColumnWriters.TYPE_COMPLEX); + currentPosition += 1; + + memory.putInt(currentPosition, typeName.length); + currentPosition += Integer.BYTES; + + memory.putByteArray(currentPosition, typeName, 0, typeName.length); + currentPosition += typeName.length; + + currentPosition += offsetMemory.writeTo(memory, currentPosition); + currentPosition += dataMemory.writeTo(memory, currentPosition); + + return currentPosition - startPosition; + } + + public void close() + { + offsetMemory.close(); + dataMemory.close(); + } + + private int headerSize() + { + return 1 /* type code */ + + Integer.BYTES /* type name length */ + + typeName.length; + } +} diff --git a/processing/src/main/java/org/apache/druid/frame/write/columnar/DoubleFrameColumnWriter.java b/processing/src/main/java/org/apache/druid/frame/write/columnar/DoubleFrameColumnWriter.java index 82d2054d4029..f83e210ef06d 100644 --- a/processing/src/main/java/org/apache/druid/frame/write/columnar/DoubleFrameColumnWriter.java +++ b/processing/src/main/java/org/apache/druid/frame/write/columnar/DoubleFrameColumnWriter.java @@ -20,19 +20,13 @@ package org.apache.druid.frame.write.columnar; import org.apache.datasketches.memory.WritableMemory; -import org.apache.druid.frame.allocation.AppendableMemory; import org.apache.druid.frame.allocation.MemoryAllocator; -import org.apache.druid.frame.allocation.MemoryRange; import org.apache.druid.segment.BaseDoubleColumnValueSelector; public class DoubleFrameColumnWriter implements FrameColumnWriter { - public static final long DATA_OFFSET = 1 /* type code */ + 1 /* has nulls? */; - private final BaseDoubleColumnValueSelector selector; - private final AppendableMemory appendableMemory; - private final boolean hasNulls; - private final int sz; + private final DoubleFrameMaker maker; DoubleFrameColumnWriter( BaseDoubleColumnValueSelector selector, @@ -41,71 +35,40 @@ public class DoubleFrameColumnWriter implements FrameColumnWriter ) { this.selector = selector; - this.appendableMemory = AppendableMemory.create(allocator); - this.hasNulls = hasNulls; - this.sz = valueSize(hasNulls); - } - - public static int valueSize(final boolean hasNulls) - { - return hasNulls ? Double.BYTES + 1 : Double.BYTES; + this.maker = new DoubleFrameMaker(allocator, hasNulls); } @Override public boolean addSelection() { - if (!(appendableMemory.reserveAdditional(sz))) { - return false; - } - - final MemoryRange cursor = appendableMemory.cursor(); - final WritableMemory memory = cursor.memory(); - final long position = cursor.start(); - - if (hasNulls) { - if (selector.isNull()) { - memory.putByte(position, (byte) 1); - memory.putDouble(position + 1, 0); - } else { - memory.putByte(position, (byte) 0); - memory.putDouble(position + 1, selector.getDouble()); - } + if (selector.isNull()) { + return maker.addNull(); } else { - memory.putDouble(position, selector.getDouble()); + return maker.add(selector.getDouble()); } - - appendableMemory.advanceCursor(sz); - return true; } @Override public void undo() { - appendableMemory.rewindCursor(sz); + maker.undo(); } @Override public long size() { - return DATA_OFFSET + appendableMemory.size(); + return maker.size(); } @Override public long writeTo(final WritableMemory memory, final long startPosition) { - long currentPosition = startPosition; - - memory.putByte(currentPosition, FrameColumnWriters.TYPE_DOUBLE); - memory.putByte(currentPosition + 1, hasNulls ? (byte) 1 : (byte) 0); - currentPosition += 2; - - currentPosition += appendableMemory.writeTo(memory, currentPosition); - return currentPosition - startPosition; + return maker.writeTo(memory, startPosition); } @Override public void close() { - appendableMemory.close(); + maker.close(); } } diff --git a/processing/src/main/java/org/apache/druid/frame/write/columnar/DoubleFrameMaker.java b/processing/src/main/java/org/apache/druid/frame/write/columnar/DoubleFrameMaker.java new file mode 100644 index 000000000000..13ee53d6d5fc --- /dev/null +++ b/processing/src/main/java/org/apache/druid/frame/write/columnar/DoubleFrameMaker.java @@ -0,0 +1,122 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.frame.write.columnar; + +import org.apache.datasketches.memory.WritableMemory; +import org.apache.druid.frame.allocation.AppendableMemory; +import org.apache.druid.frame.allocation.MemoryAllocator; +import org.apache.druid.frame.allocation.MemoryRange; +import org.apache.druid.java.util.common.ISE; + +import java.io.Closeable; + +public class DoubleFrameMaker implements Closeable +{ + public static final long DATA_OFFSET = 1 /* type code */ + 1 /* has nulls? */; + + private final AppendableMemory appendableMemory; + private final boolean hasNulls; + private final int sz; + + public DoubleFrameMaker( + MemoryAllocator allocator, + boolean hasNulls + ) + { + this.appendableMemory = AppendableMemory.create(allocator); + this.hasNulls = hasNulls; + this.sz = valueSize(hasNulls); + } + + public static int valueSize(final boolean hasNulls) + { + return hasNulls ? Double.BYTES + 1 : Double.BYTES; + } + + public boolean add(double value) + { + if (!(appendableMemory.reserveAdditional(sz))) { + return false; + } + + final MemoryRange cursor = appendableMemory.cursor(); + final WritableMemory memory = cursor.memory(); + final long position = cursor.start(); + + if (hasNulls) { + memory.putByte(position, (byte) 0); + memory.putDouble(position + 1, value); + } else { + memory.putDouble(position, value); + } + + appendableMemory.advanceCursor(sz); + return true; + } + + public boolean addNull() + { + if (!hasNulls) { + throw new ISE("Was told that null doesn't exist, cannot add null"); + } + + if (!(appendableMemory.reserveAdditional(sz))) { + return false; + } + + final MemoryRange cursor = appendableMemory.cursor(); + final WritableMemory memory = cursor.memory(); + final long position = cursor.start(); + + memory.putByte(position, (byte) 1); + memory.putDouble(position + 1, 0); + + appendableMemory.advanceCursor(sz); + return true; + } + + public void undo() + { + appendableMemory.rewindCursor(sz); + } + + public long size() + { + return DATA_OFFSET + appendableMemory.size(); + } + + public long writeTo(final WritableMemory memory, final long startPosition) + { + long currentPosition = startPosition; + + memory.putByte(currentPosition, FrameColumnWriters.TYPE_DOUBLE); + memory.putByte(currentPosition + 1, hasNulls ? (byte) 1 : (byte) 0); + currentPosition += 2; + + currentPosition += appendableMemory.writeTo(memory, currentPosition); + return currentPosition - startPosition; + } + + @Override + public void close() + { + appendableMemory.close(); + } +} diff --git a/processing/src/main/java/org/apache/druid/frame/write/columnar/FloatFrameColumnWriter.java b/processing/src/main/java/org/apache/druid/frame/write/columnar/FloatFrameColumnWriter.java index 900b72d2076c..5f7843483d5a 100644 --- a/processing/src/main/java/org/apache/druid/frame/write/columnar/FloatFrameColumnWriter.java +++ b/processing/src/main/java/org/apache/druid/frame/write/columnar/FloatFrameColumnWriter.java @@ -20,19 +20,13 @@ package org.apache.druid.frame.write.columnar; import org.apache.datasketches.memory.WritableMemory; -import org.apache.druid.frame.allocation.AppendableMemory; import org.apache.druid.frame.allocation.MemoryAllocator; -import org.apache.druid.frame.allocation.MemoryRange; import org.apache.druid.segment.BaseFloatColumnValueSelector; public class FloatFrameColumnWriter implements FrameColumnWriter { - public static final long DATA_OFFSET = 1 /* type code */ + 1 /* has nulls? */; - private final BaseFloatColumnValueSelector selector; - private final AppendableMemory appendableMemory; - private final boolean hasNulls; - private final int sz; + private final FloatFrameMaker maker; FloatFrameColumnWriter( BaseFloatColumnValueSelector selector, @@ -41,71 +35,40 @@ public class FloatFrameColumnWriter implements FrameColumnWriter ) { this.selector = selector; - this.appendableMemory = AppendableMemory.create(allocator); - this.hasNulls = hasNulls; - this.sz = valueSize(hasNulls); - } - - public static int valueSize(final boolean hasNulls) - { - return hasNulls ? Float.BYTES + 1 : Float.BYTES; + this.maker = new FloatFrameMaker(allocator, hasNulls); } @Override public boolean addSelection() { - if (!(appendableMemory.reserveAdditional(sz))) { - return false; - } - - final MemoryRange cursor = appendableMemory.cursor(); - final WritableMemory memory = cursor.memory(); - final long position = cursor.start(); - - if (hasNulls) { - if (selector.isNull()) { - memory.putByte(position, (byte) 1); - memory.putFloat(position + 1, 0); - } else { - memory.putByte(position, (byte) 0); - memory.putFloat(position + 1, selector.getFloat()); - } + if (selector.isNull()) { + return maker.addNull(); } else { - memory.putFloat(position, selector.getFloat()); + return maker.add(selector.getFloat()); } - - appendableMemory.advanceCursor(sz); - return true; } @Override public void undo() { - appendableMemory.rewindCursor(sz); + maker.undo(); } @Override public long size() { - return DATA_OFFSET + appendableMemory.size(); + return maker.size(); } @Override public long writeTo(final WritableMemory memory, final long startPosition) { - long currentPosition = startPosition; - - memory.putByte(currentPosition, FrameColumnWriters.TYPE_FLOAT); - memory.putByte(currentPosition + 1, hasNulls ? (byte) 1 : (byte) 0); - currentPosition += 2; - - currentPosition += appendableMemory.writeTo(memory, currentPosition); - return currentPosition - startPosition; + return maker.writeTo(memory, startPosition); } @Override public void close() { - appendableMemory.close(); + maker.close(); } } diff --git a/processing/src/main/java/org/apache/druid/frame/write/columnar/FloatFrameMaker.java b/processing/src/main/java/org/apache/druid/frame/write/columnar/FloatFrameMaker.java new file mode 100644 index 000000000000..3c299d0b9106 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/frame/write/columnar/FloatFrameMaker.java @@ -0,0 +1,122 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.frame.write.columnar; + +import org.apache.datasketches.memory.WritableMemory; +import org.apache.druid.frame.allocation.AppendableMemory; +import org.apache.druid.frame.allocation.MemoryAllocator; +import org.apache.druid.frame.allocation.MemoryRange; +import org.apache.druid.java.util.common.ISE; + +import java.io.Closeable; + +public class FloatFrameMaker implements Closeable +{ + public static final long DATA_OFFSET = 1 /* type code */ + 1 /* has nulls? */; + + private final AppendableMemory appendableMemory; + private final boolean hasNulls; + private final int sz; + + public FloatFrameMaker( + MemoryAllocator allocator, + boolean hasNulls + ) + { + this.appendableMemory = AppendableMemory.create(allocator); + this.hasNulls = hasNulls; + this.sz = valueSize(hasNulls); + } + + public static int valueSize(final boolean hasNulls) + { + return hasNulls ? Float.BYTES + 1 : Float.BYTES; + } + + public boolean add(float value) + { + if (!(appendableMemory.reserveAdditional(sz))) { + return false; + } + + final MemoryRange cursor = appendableMemory.cursor(); + final WritableMemory memory = cursor.memory(); + final long position = cursor.start(); + + if (hasNulls) { + memory.putByte(position, (byte) 0); + memory.putFloat(position + 1, value); + } else { + memory.putFloat(position, value); + } + + appendableMemory.advanceCursor(sz); + return true; + } + + public boolean addNull() + { + if (!hasNulls) { + throw new ISE("Was told that null doesn't exist, cannot add null"); + } + + if (!(appendableMemory.reserveAdditional(sz))) { + return false; + } + + final MemoryRange cursor = appendableMemory.cursor(); + final WritableMemory memory = cursor.memory(); + final long position = cursor.start(); + + memory.putByte(position, (byte) 1); + memory.putFloat(position + 1, 0); + + appendableMemory.advanceCursor(sz); + return true; + } + + public void undo() + { + appendableMemory.rewindCursor(sz); + } + + public long size() + { + return DATA_OFFSET + appendableMemory.size(); + } + + public long writeTo(final WritableMemory memory, final long startPosition) + { + long currentPosition = startPosition; + + memory.putByte(currentPosition, FrameColumnWriters.TYPE_FLOAT); + memory.putByte(currentPosition + 1, hasNulls ? (byte) 1 : (byte) 0); + currentPosition += 2; + + currentPosition += appendableMemory.writeTo(memory, currentPosition); + return currentPosition - startPosition; + } + + @Override + public void close() + { + appendableMemory.close(); + } +} diff --git a/processing/src/main/java/org/apache/druid/frame/write/columnar/LongFrameColumnWriter.java b/processing/src/main/java/org/apache/druid/frame/write/columnar/LongFrameColumnWriter.java index c0a2bf4bb3ab..101d36a83a07 100644 --- a/processing/src/main/java/org/apache/druid/frame/write/columnar/LongFrameColumnWriter.java +++ b/processing/src/main/java/org/apache/druid/frame/write/columnar/LongFrameColumnWriter.java @@ -20,19 +20,13 @@ package org.apache.druid.frame.write.columnar; import org.apache.datasketches.memory.WritableMemory; -import org.apache.druid.frame.allocation.AppendableMemory; import org.apache.druid.frame.allocation.MemoryAllocator; -import org.apache.druid.frame.allocation.MemoryRange; import org.apache.druid.segment.BaseLongColumnValueSelector; public class LongFrameColumnWriter implements FrameColumnWriter { - public static final long DATA_OFFSET = 1 /* type code */ + 1 /* has nulls? */; - private final BaseLongColumnValueSelector selector; - private final AppendableMemory appendableMemory; - private final boolean hasNulls; - private final int sz; + private final LongFrameMaker maker; LongFrameColumnWriter( BaseLongColumnValueSelector selector, @@ -41,71 +35,40 @@ public class LongFrameColumnWriter implements FrameColumnWriter ) { this.selector = selector; - this.appendableMemory = AppendableMemory.create(allocator); - this.hasNulls = hasNulls; - this.sz = valueSize(hasNulls); - } - - public static int valueSize(final boolean hasNulls) - { - return hasNulls ? Long.BYTES + 1 : Long.BYTES; + this.maker = new LongFrameMaker(allocator, hasNulls); } @Override public boolean addSelection() { - if (!(appendableMemory.reserveAdditional(sz))) { - return false; - } - - final MemoryRange cursor = appendableMemory.cursor(); - final WritableMemory memory = cursor.memory(); - final long position = cursor.start(); - - if (hasNulls) { - if (selector.isNull()) { - memory.putByte(position, (byte) 1); - memory.putLong(position + 1, 0); - } else { - memory.putByte(position, (byte) 0); - memory.putLong(position + 1, selector.getLong()); - } + if (selector.isNull()) { + return maker.addNull(); } else { - memory.putLong(position, selector.getLong()); + return maker.add(selector.getLong()); } - - appendableMemory.advanceCursor(sz); - return true; } @Override public void undo() { - appendableMemory.rewindCursor(sz); + maker.undo(); } @Override public long size() { - return DATA_OFFSET + appendableMemory.size(); + return maker.size(); } @Override public long writeTo(final WritableMemory memory, final long startPosition) { - long currentPosition = startPosition; - - memory.putByte(currentPosition, FrameColumnWriters.TYPE_LONG); - memory.putByte(currentPosition + 1, hasNulls ? (byte) 1 : (byte) 0); - currentPosition += 2; - - currentPosition += appendableMemory.writeTo(memory, currentPosition); - return currentPosition - startPosition; + return maker.writeTo(memory, startPosition); } @Override public void close() { - appendableMemory.close(); + maker.close(); } } diff --git a/processing/src/main/java/org/apache/druid/frame/write/columnar/LongFrameMaker.java b/processing/src/main/java/org/apache/druid/frame/write/columnar/LongFrameMaker.java new file mode 100644 index 000000000000..54c4e2bc98de --- /dev/null +++ b/processing/src/main/java/org/apache/druid/frame/write/columnar/LongFrameMaker.java @@ -0,0 +1,122 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.frame.write.columnar; + +import org.apache.datasketches.memory.WritableMemory; +import org.apache.druid.frame.allocation.AppendableMemory; +import org.apache.druid.frame.allocation.MemoryAllocator; +import org.apache.druid.frame.allocation.MemoryRange; +import org.apache.druid.java.util.common.ISE; + +import java.io.Closeable; + +public class LongFrameMaker implements Closeable +{ + public static final long DATA_OFFSET = 1 /* type code */ + 1 /* has nulls? */; + + private final AppendableMemory appendableMemory; + private final boolean hasNulls; + private final int sz; + + public LongFrameMaker( + MemoryAllocator allocator, + boolean hasNulls + ) + { + this.appendableMemory = AppendableMemory.create(allocator); + this.hasNulls = hasNulls; + this.sz = valueSize(hasNulls); + } + + public static int valueSize(final boolean hasNulls) + { + return hasNulls ? Long.BYTES + 1 : Long.BYTES; + } + + public boolean add(long value) + { + if (!(appendableMemory.reserveAdditional(sz))) { + return false; + } + + final MemoryRange cursor = appendableMemory.cursor(); + final WritableMemory memory = cursor.memory(); + final long position = cursor.start(); + + if (hasNulls) { + memory.putByte(position, (byte) 0); + memory.putLong(position + 1, value); + } else { + memory.putLong(position, value); + } + + appendableMemory.advanceCursor(sz); + return true; + } + + public boolean addNull() + { + if (!hasNulls) { + throw new ISE("Was told that null doesn't exist, cannot add null"); + } + + if (!(appendableMemory.reserveAdditional(sz))) { + return false; + } + + final MemoryRange cursor = appendableMemory.cursor(); + final WritableMemory memory = cursor.memory(); + final long position = cursor.start(); + + memory.putByte(position, (byte) 1); + memory.putLong(position + 1, 0); + + appendableMemory.advanceCursor(sz); + return true; + } + + public void undo() + { + appendableMemory.rewindCursor(sz); + } + + public long size() + { + return DATA_OFFSET + appendableMemory.size(); + } + + public long writeTo(final WritableMemory memory, final long startPosition) + { + long currentPosition = startPosition; + + memory.putByte(currentPosition, FrameColumnWriters.TYPE_LONG); + memory.putByte(currentPosition + 1, hasNulls ? (byte) 1 : (byte) 0); + currentPosition += 2; + + currentPosition += appendableMemory.writeTo(memory, currentPosition); + return currentPosition - startPosition; + } + + @Override + public void close() + { + appendableMemory.close(); + } +} diff --git a/processing/src/main/java/org/apache/druid/query/InlineDataSource.java b/processing/src/main/java/org/apache/druid/query/InlineDataSource.java index 040d6885782a..b3c602403913 100644 --- a/processing/src/main/java/org/apache/druid/query/InlineDataSource.java +++ b/processing/src/main/java/org/apache/druid/query/InlineDataSource.java @@ -32,6 +32,7 @@ import org.apache.druid.segment.column.RowSignature; import javax.annotation.Nullable; +import java.util.ArrayList; import java.util.Arrays; import java.util.Collections; import java.util.List; @@ -70,7 +71,7 @@ private InlineDataSource( private static InlineDataSource fromJson( @JsonProperty("columnNames") List columnNames, @JsonProperty("columnTypes") List columnTypes, - @JsonProperty("rows") List rows + @JsonProperty("rows") ArrayList rows ) { Preconditions.checkNotNull(columnNames, "'columnNames' must be nonnull"); @@ -200,6 +201,11 @@ public Iterable getRows() return rows; } + public boolean rowsAreArrayList() + { + return rows instanceof ArrayList; + } + @Override public List getChildren() { diff --git a/processing/src/main/java/org/apache/druid/query/Query.java b/processing/src/main/java/org/apache/druid/query/Query.java index 9d38dbe37169..90b43469dce5 100644 --- a/processing/src/main/java/org/apache/druid/query/Query.java +++ b/processing/src/main/java/org/apache/druid/query/Query.java @@ -31,6 +31,7 @@ import org.apache.druid.query.filter.DimFilter; import org.apache.druid.query.groupby.GroupByQuery; import org.apache.druid.query.metadata.metadata.SegmentMetadataQuery; +import org.apache.druid.query.operator.WindowOperatorQuery; import org.apache.druid.query.scan.ScanQuery; import org.apache.druid.query.search.SearchQuery; import org.apache.druid.query.select.SelectQuery; @@ -45,7 +46,6 @@ import org.joda.time.Interval; import javax.annotation.Nullable; - import java.util.List; import java.util.Map; import java.util.Set; @@ -54,27 +54,29 @@ @ExtensionPoint @JsonTypeInfo(use = JsonTypeInfo.Id.NAME, property = "queryType") @JsonSubTypes(value = { - @JsonSubTypes.Type(name = Query.TIMESERIES, value = TimeseriesQuery.class), - @JsonSubTypes.Type(name = Query.SEARCH, value = SearchQuery.class), - @JsonSubTypes.Type(name = Query.TIME_BOUNDARY, value = TimeBoundaryQuery.class), + @JsonSubTypes.Type(name = Query.DATASOURCE_METADATA, value = DataSourceMetadataQuery.class), @JsonSubTypes.Type(name = Query.GROUP_BY, value = GroupByQuery.class), @JsonSubTypes.Type(name = Query.SCAN, value = ScanQuery.class), + @JsonSubTypes.Type(name = Query.SEARCH, value = SearchQuery.class), @JsonSubTypes.Type(name = Query.SEGMENT_METADATA, value = SegmentMetadataQuery.class), @JsonSubTypes.Type(name = Query.SELECT, value = SelectQuery.class), + @JsonSubTypes.Type(name = Query.TIME_BOUNDARY, value = TimeBoundaryQuery.class), + @JsonSubTypes.Type(name = Query.TIMESERIES, value = TimeseriesQuery.class), @JsonSubTypes.Type(name = Query.TOPN, value = TopNQuery.class), - @JsonSubTypes.Type(name = Query.DATASOURCE_METADATA, value = DataSourceMetadataQuery.class) + @JsonSubTypes.Type(name = Query.WINDOW_OPERATOR, value = WindowOperatorQuery.class), }) public interface Query { - String TIMESERIES = "timeseries"; - String SEARCH = "search"; - String TIME_BOUNDARY = "timeBoundary"; + String DATASOURCE_METADATA = "dataSourceMetadata"; String GROUP_BY = "groupBy"; String SCAN = "scan"; + String SEARCH = "search"; String SEGMENT_METADATA = "segmentMetadata"; String SELECT = "select"; + String TIME_BOUNDARY = "timeBoundary"; + String TIMESERIES = "timeseries"; String TOPN = "topN"; - String DATASOURCE_METADATA = "dataSourceMetadata"; + String WINDOW_OPERATOR = "windowOperator"; DataSource getDataSource(); diff --git a/processing/src/main/java/org/apache/druid/query/operator/LimitedRowsAndColumns.java b/processing/src/main/java/org/apache/druid/query/operator/LimitedRowsAndColumns.java new file mode 100644 index 000000000000..d0521dbad0ba --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/operator/LimitedRowsAndColumns.java @@ -0,0 +1,115 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator; + +import org.apache.druid.java.util.common.ISE; +import org.apache.druid.query.operator.window.value.ShiftedColumnAccessorBase; +import org.apache.druid.query.rowsandcols.RowsAndColumns; +import org.apache.druid.query.rowsandcols.column.Column; +import org.apache.druid.query.rowsandcols.column.ColumnAccessor; + +import javax.annotation.Nullable; +import java.util.Collection; + +public class LimitedRowsAndColumns implements RowsAndColumns +{ + private final RowsAndColumns rac; + private final int start; + private final int end; + + public LimitedRowsAndColumns(RowsAndColumns rac, int start, int end) + { + final int numRows = rac.numRows(); + if (numRows < end) { + throw new ISE("end[%d] is out of bounds, cannot be greater than numRows[%d]", end, numRows); + } + + this.rac = rac; + this.start = start; + this.end = end; + } + + @Override + public Collection getColumnNames() + { + return rac.getColumnNames(); + } + + @Override + public int numRows() + { + return end - start; + } + + @Override + public Column findColumn(String name) + { + final Column column = rac.findColumn(name); + if (column == null) { + return null; + } + + return new Column() + { + @Override + public ColumnAccessor toAccessor() + { + final ColumnAccessor columnAccessor = column.toAccessor(); + return new ShiftedColumnAccessorBase(columnAccessor) + { + @Override + public int numRows() + { + return end - start; + } + + @Override + protected int getActualCell(int cell) + { + int retVal = start + cell; + if (retVal >= end) { + throw new ISE("Index out of bounds[%d] >= [%d], start[%s]", retVal, end, start); + } + return retVal; + } + + @Override + protected boolean outsideBounds(int cell) + { + return false; + } + }; + } + + @Override + public T as(Class clazz) + { + return null; + } + }; + } + + @Nullable + @Override + public T as(Class clazz) + { + return null; + } +} diff --git a/processing/src/main/java/org/apache/druid/query/operator/NaivePartitioningOperator.java b/processing/src/main/java/org/apache/druid/query/operator/NaivePartitioningOperator.java new file mode 100644 index 000000000000..8063a78b5984 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/operator/NaivePartitioningOperator.java @@ -0,0 +1,100 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator; + +import org.apache.druid.java.util.common.ISE; +import org.apache.druid.query.rowsandcols.DefaultSortedGroupPartitioner; +import org.apache.druid.query.rowsandcols.RowsAndColumns; +import org.apache.druid.query.rowsandcols.SortedGroupPartitioner; + +import java.util.Iterator; +import java.util.List; + +/** + * This naive partitioning operator assumes that it's child operator always gives it RowsAndColumns objects that are + * a superset of the partitions that it needs to provide. It will never attempt to make a partition larger than a + * single RowsAndColumns object that it is given from its child Operator. A different operator should be used + * if that is an important bit of functionality to have. + *

    + * Additionally, this assumes that data has been pre-sorted according to the partitioning columns. If it is + * given data that has not been pre-sorted, an exception is expected to be thrown. + */ +public class NaivePartitioningOperator implements Operator +{ + private final List partitionColumns; + private final Operator child; + + private Iterator partitionsIter; + + public NaivePartitioningOperator( + List partitionColumns, + Operator child + ) + { + this.partitionColumns = partitionColumns; + this.child = child; + } + + @Override + public void open() + { + child.open(); + } + + @Override + public RowsAndColumns next() + { + if (partitionsIter != null && partitionsIter.hasNext()) { + return partitionsIter.next(); + } + + if (child.hasNext()) { + final RowsAndColumns rac = child.next(); + + SortedGroupPartitioner groupPartitioner = rac.as(SortedGroupPartitioner.class); + if (groupPartitioner == null) { + groupPartitioner = new DefaultSortedGroupPartitioner(rac); + } + + partitionsIter = groupPartitioner.partitionOnBoundaries(partitionColumns).iterator(); + return partitionsIter.next(); + } + + throw new ISE("Asked for next when already complete"); + } + + @Override + public boolean hasNext() + { + if (partitionsIter != null && partitionsIter.hasNext()) { + return true; + } + + return child.hasNext(); + } + + @Override + public void close(boolean cascade) + { + if (cascade) { + child.close(cascade); + } + } +} diff --git a/processing/src/main/java/org/apache/druid/query/operator/NaivePartitioningOperatorFactory.java b/processing/src/main/java/org/apache/druid/query/operator/NaivePartitioningOperatorFactory.java new file mode 100644 index 000000000000..9bd937ab8443 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/operator/NaivePartitioningOperatorFactory.java @@ -0,0 +1,68 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator; + +import com.fasterxml.jackson.annotation.JsonCreator; +import com.fasterxml.jackson.annotation.JsonProperty; + +import java.util.ArrayList; +import java.util.List; + +public class NaivePartitioningOperatorFactory implements OperatorFactory +{ + private final List partitionColumns; + + @JsonCreator + public NaivePartitioningOperatorFactory( + @JsonProperty("partitionColumns") List partitionColumns + ) + { + this.partitionColumns = partitionColumns == null ? new ArrayList<>() : partitionColumns; + } + + @JsonProperty("partitionColumns") + public List getPartitionColumns() + { + return partitionColumns; + } + + @Override + public Operator wrap(Operator op) + { + return new NaivePartitioningOperator(partitionColumns, op); + } + + @Override + public boolean validateEquivalent(OperatorFactory other) + { + if (other instanceof NaivePartitioningOperatorFactory) { + return partitionColumns.equals(((NaivePartitioningOperatorFactory) other).getPartitionColumns()); + } + return false; + } + + @Override + public String toString() + { + return "NaivePartitioningOperatorFactory{" + + "partitionColumns=" + partitionColumns + + '}'; + } +} diff --git a/processing/src/main/java/org/apache/druid/query/operator/Operator.java b/processing/src/main/java/org/apache/druid/query/operator/Operator.java new file mode 100644 index 000000000000..ad64cd1a5235 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/operator/Operator.java @@ -0,0 +1,93 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator; + +import org.apache.druid.query.rowsandcols.RowsAndColumns; + +/** + * An Operator interface that intends to align closely with the Operators that other databases would also tend + * to be implemented using. + *

    + * The lifecycle of an operator is that, after creation, it should be opened, and then iterated using hasNext() and + * next(). Finally, when the Operator is no longer useful, it should be closed. + *

    + * Operator's methods mimic the methods of an {@code Iterator}, but it does not implement {@code Iterator} + * intentionally. An operator should never be wrapped in an {@code Iterator}. Any code that does that should be + * considered a bug and fixed. This is for two reasons: + *

    + * 1. An Operator should never be passed around as an {@code Iterator}. An Operator must be closed, if an operator + * gets returned as an {@code Iterator}, the code that sees the {@code Iterator} loses the knowledge that it's + * dealing with an Operator and might not close it. Even something like a {@code CloseableIterator} is an + * anti-pattern as it's possible to use it in a functional manner with code that loses track of the fact that it + * must be closed. + * 2. To avoid "fluent" style composition of functions on Operators. It is important that there never be a set of + * functional primitives for things like map/filter/reduce to "simplify" the implementation of Operators. This is + * because such fluency produces really hard to decipher stack traces as the stacktrace ends up being just a bunch + * of calls from the scaffolding (map/filter/reduce) and not from the actual Operator itself. By not implementing + * {@code Iterator} we are actively increasing the burden of trying to add such functional operations to the point + * that hopefully, though code review, we can ensure that we never develop them. It is infinitely better to preserve + * the stacktrace and "duplicate" the map/filter/reduce scaffolding code. + */ +public interface Operator +{ + /** + * Called to initiate the lifecycle of the Operator. If an operator needs to checkout resources or anything to do + * its work, this is probably the place to do it. + * + * Work should *never* be done in this method, this method only exists to acquire resources that are known to be + * needed before doing any work. As a litmus test, if there is ever a call to `op.next()` inside of this method, + * then something has been done wrong as that call to `.next()` is actually doing work. Such code should be moved + * into being lazily evaluated as part of a call to `.next()`. + */ + void open(); + + /** + * Returns the next RowsAndColumns object that the Operator can produce. Behavior is undefined if + * {@link #hasNext} returns false. + * + * @return the next RowsAndColumns object that the operator can produce + */ + RowsAndColumns next(); + + /** + * Used to identify if it is safe to call {@link #next} + * + * @return true if it is safe to call {@link #next} + */ + boolean hasNext(); + + /** + * Closes this Operator. The cascade flag can be used to identify that the intent is to close this operator + * and only this operator without actually closing child operators. Other databases us this sort of functionality + * with a planner that is watching over all of the objects and force-closes even if they were closed during normal + * operations. In Druid, in the data pipeline where this was introduced, we are guaranteed to always have close + * called regardless of errors or exceptions during processing, as such, at time of introduction, there is no + * call that passes false for cascade. + *

    + * That said, given that this is a common thing for these interfaces for other databases, we want to preserve the + * optionality of being able to leverage what they do. As such, we define the method this way with the belief + * that it might be used in the future. Semantically, this means that all implementations of Operators must + * expect to be closed multiple times. I.e. after being closed, it is an error for open, next or hasNext to be + * called, but close can be called any number of times. + * + * @param cascade whether to call close on child operators. + */ + void close(boolean cascade); +} diff --git a/processing/src/main/java/org/apache/druid/query/operator/OperatorFactory.java b/processing/src/main/java/org/apache/druid/query/operator/OperatorFactory.java new file mode 100644 index 000000000000..26235c35d8d8 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/operator/OperatorFactory.java @@ -0,0 +1,59 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator; + +import com.fasterxml.jackson.annotation.JsonSubTypes; +import com.fasterxml.jackson.annotation.JsonTypeInfo; + +/** + * A factory for Operators. This class exists to encapsulate the user-definition of an Operator. I.e. which operator, + * what fields it should operate on, etc. etc. These Factory objects are then used to combine Operators together + * and run against concrete data. + */ +@JsonTypeInfo(use = JsonTypeInfo.Id.NAME, property = "type") +@JsonSubTypes(value = { + @JsonSubTypes.Type(name = "naivePartition", value = NaivePartitioningOperatorFactory.class), + @JsonSubTypes.Type(name = "window", value = WindowOperatorFactory.class), +}) +public interface OperatorFactory +{ + /** + * Builds an operator according to the definition of the OperatorFactory and wraps it around the operator passed + * in to this function. + * + * @param op the Operator to wrap + * @return the wrapped Operator + */ + Operator wrap(Operator op); + + /** + * Validates the equivalence of Operators. This is similar to @{code .equals} but is its own method + * so that it can ignore certain fields that would be important for a true equality check. Namely, two Operators + * defined the same way but with different output names can be considered equivalent even though they are not equal. + *

    + * This primarily exists to simplify tests, where this equivalence can be used to validate that the Operators + * created by the SQL planner are actually equivalent to what we expect without needing to be overly dependent on + * how the planner names output columns + * + * @param other the processor to test equivalence of + * @return boolean identifying if these processors should be considered equivalent to each other. + */ + boolean validateEquivalent(OperatorFactory other); +} diff --git a/processing/src/main/java/org/apache/druid/query/operator/OperatorSequence.java b/processing/src/main/java/org/apache/druid/query/operator/OperatorSequence.java new file mode 100644 index 000000000000..45a3bbd23899 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/operator/OperatorSequence.java @@ -0,0 +1,125 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator; + +import org.apache.druid.java.util.common.guava.Accumulator; +import org.apache.druid.java.util.common.guava.Sequence; +import org.apache.druid.java.util.common.guava.Yielder; +import org.apache.druid.java.util.common.guava.Yielders; +import org.apache.druid.java.util.common.guava.YieldingAccumulator; +import org.apache.druid.query.rowsandcols.RowsAndColumns; + +import java.util.function.Supplier; + +public class OperatorSequence implements Sequence +{ + private final Supplier opSupplier; + + public OperatorSequence( + Supplier opSupplier + ) + { + this.opSupplier = opSupplier; + } + + @Override + public OutType accumulate( + OutType initValue, + Accumulator accumulator + ) + { + Operator op = null; + try { + op = opSupplier.get(); + op.open(); + while (op.hasNext()) { + initValue = accumulator.accumulate(initValue, op.next()); + } + return initValue; + } + finally { + if (op != null) { + op.close(true); + } + } + } + + @Override + public Yielder toYielder( + OutType initValue, + YieldingAccumulator accumulator + ) + { + final Operator op = opSupplier.get(); + try { + op.open(); + + while (!accumulator.yielded() && op.hasNext()) { + initValue = accumulator.accumulate(initValue, op.next()); + } + if (accumulator.yielded()) { + OutType finalInitValue = initValue; + return new Yielder() + { + private OutType retVal = finalInitValue; + private boolean done = false; + + @Override + public OutType get() + { + return retVal; + } + + @Override + public Yielder next(OutType initValue) + { + accumulator.reset(); + retVal = initValue; + while (!accumulator.yielded() && op.hasNext()) { + retVal = accumulator.accumulate(retVal, op.next()); + } + if (!accumulator.yielded()) { + done = true; + } + return this; + } + + @Override + public boolean isDone() + { + return done; + } + + @Override + public void close() + { + op.close(true); + } + }; + } else { + return Yielders.done(initValue, () -> op.close(true)); + } + } + catch (RuntimeException e) { + op.close(true); + throw e; + } + } +} diff --git a/processing/src/main/java/org/apache/druid/query/operator/SegmentToRowsAndColumnsOperator.java b/processing/src/main/java/org/apache/druid/query/operator/SegmentToRowsAndColumnsOperator.java new file mode 100644 index 000000000000..dc912bfdd6f6 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/operator/SegmentToRowsAndColumnsOperator.java @@ -0,0 +1,68 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator; + +import org.apache.druid.java.util.common.ISE; +import org.apache.druid.query.rowsandcols.RowsAndColumns; +import org.apache.druid.segment.Segment; + +public class SegmentToRowsAndColumnsOperator implements Operator +{ + private final Segment segment; + private boolean hasNext = true; + + public SegmentToRowsAndColumnsOperator( + Segment segment + ) + { + this.segment = segment; + } + + @Override + public void open() + { + + } + + @Override + public RowsAndColumns next() + { + hasNext = false; + + RowsAndColumns rac = segment.as(RowsAndColumns.class); + if (rac != null) { + return rac; + } + + throw new ISE("Cannot work with segment of type[%s]", segment.getClass()); + } + + @Override + public boolean hasNext() + { + return hasNext; + } + + @Override + public void close(boolean cascade) + { + + } +} diff --git a/processing/src/main/java/org/apache/druid/query/operator/SequenceOperator.java b/processing/src/main/java/org/apache/druid/query/operator/SequenceOperator.java new file mode 100644 index 000000000000..9dc54f9576c8 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/operator/SequenceOperator.java @@ -0,0 +1,87 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator; + +import org.apache.druid.java.util.common.ISE; +import org.apache.druid.java.util.common.RE; +import org.apache.druid.java.util.common.guava.Sequence; +import org.apache.druid.java.util.common.guava.Yielder; +import org.apache.druid.java.util.common.guava.Yielders; +import org.apache.druid.query.rowsandcols.RowsAndColumns; + +import java.io.IOException; +import java.util.NoSuchElementException; + +public class SequenceOperator implements Operator +{ + private final Sequence child; + private Yielder yielder; + private boolean closed = false; + + public SequenceOperator( + Sequence child + ) + { + this.child = child; + } + + @Override + public void open() + { + if (closed) { + throw new ISE("Operator closed, cannot be re-opened"); + } + yielder = Yielders.each(child); + } + + @Override + public RowsAndColumns next() + { + if (closed) { + throw new NoSuchElementException(); + } + final RowsAndColumns retVal = yielder.get(); + yielder = yielder.next(null); + return retVal; + } + + @Override + public boolean hasNext() + { + return !closed && !yielder.isDone(); + } + + @Override + public void close(boolean cascade) + { + if (closed) { + return; + } + try { + yielder.close(); + } + catch (IOException e) { + throw new RE(e, "Exception when closing yielder from Sequence"); + } + finally { + closed = true; + } + } +} diff --git a/processing/src/main/java/org/apache/druid/query/operator/WindowOperatorFactory.java b/processing/src/main/java/org/apache/druid/query/operator/WindowOperatorFactory.java new file mode 100644 index 000000000000..bc4cd5206c26 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/operator/WindowOperatorFactory.java @@ -0,0 +1,68 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator; + +import com.fasterxml.jackson.annotation.JsonCreator; +import com.fasterxml.jackson.annotation.JsonProperty; +import com.google.common.base.Preconditions; +import org.apache.druid.query.operator.window.Processor; + +public class WindowOperatorFactory implements OperatorFactory +{ + private Processor processor; + + @JsonCreator + public WindowOperatorFactory( + @JsonProperty("processor") Processor processor + ) + { + Preconditions.checkNotNull(processor, "processor cannot be null"); + this.processor = processor; + } + + @JsonProperty("processor") + public Processor getProcessor() + { + return processor; + } + + @Override + public Operator wrap(Operator op) + { + return new WindowProcessorOperator(processor, op); + } + + @Override + public boolean validateEquivalent(OperatorFactory other) + { + if (other instanceof WindowOperatorFactory) { + return processor.validateEquivalent(((WindowOperatorFactory) other).getProcessor()); + } + return false; + } + + @Override + public String toString() + { + return "WindowOperatorFactory{" + + "processor=" + processor + + '}'; + } +} diff --git a/processing/src/main/java/org/apache/druid/query/operator/WindowOperatorQuery.java b/processing/src/main/java/org/apache/druid/query/operator/WindowOperatorQuery.java new file mode 100644 index 000000000000..99faf8068f38 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/operator/WindowOperatorQuery.java @@ -0,0 +1,167 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator; + +import com.fasterxml.jackson.annotation.JsonCreator; +import com.fasterxml.jackson.annotation.JsonProperty; +import org.apache.druid.java.util.common.IAE; +import org.apache.druid.java.util.common.Intervals; +import org.apache.druid.java.util.common.UOE; +import org.apache.druid.query.BaseQuery; +import org.apache.druid.query.DataSource; +import org.apache.druid.query.InlineDataSource; +import org.apache.druid.query.Query; +import org.apache.druid.query.QueryDataSource; +import org.apache.druid.query.filter.DimFilter; +import org.apache.druid.query.rowsandcols.RowsAndColumns; +import org.apache.druid.query.spec.LegacySegmentSpec; +import org.apache.druid.query.spec.QuerySegmentSpec; +import org.apache.druid.segment.column.RowSignature; + +import java.util.List; +import java.util.Map; +import java.util.Objects; + +/** + * A query that can compute window functions on top of a completely in-memory inline datasource or query results. + *

    + * It relies on a set of Operators to work on the data that it is given. As such, it doesn't actually encapsulate + * any window-specific logic in-and-of-itself, but rather delegates everything to the operators. This is because + * this is also intended as the initial addition of more explicit Operators to the Druid code base. + *

    + * The assumptions on the incoming data are defined by the operators. At initial time of writing, there is a baked + * in assumption that data has been sorted "correctly" before this runs. + */ +public class WindowOperatorQuery extends BaseQuery +{ + private final RowSignature rowSignature; + private final List operators; + + @JsonCreator + public WindowOperatorQuery( + @JsonProperty("dataSource") DataSource dataSource, + @JsonProperty("context") Map context, + @JsonProperty("outputSignature") RowSignature rowSignature, + @JsonProperty("operatorDefinition") List operators + ) + { + super(dataSource, new LegacySegmentSpec(Intervals.ETERNITY), false, context); + this.rowSignature = rowSignature; + this.operators = operators; + if (!(dataSource instanceof QueryDataSource || dataSource instanceof InlineDataSource)) { + throw new IAE("WindowOperatorQuery must run on top of a query or inline data source, got [%s]", dataSource); + } + } + + @JsonProperty("operatorDefinition") + public List getOperators() + { + return operators; + } + + @JsonProperty("outputSignature") + public RowSignature getRowSignature() + { + return rowSignature; + } + + @Override + public boolean hasFilters() + { + return false; + } + + @Override + public DimFilter getFilter() + { + return null; + } + + @Override + public String getType() + { + return Query.WINDOW_OPERATOR; + } + + @Override + public Query withOverriddenContext(Map contextOverride) + { + return new WindowOperatorQuery( + getDataSource(), + computeOverriddenContext(getContext(), contextOverride), + rowSignature, + operators + ); + } + + @Override + public Query withQuerySegmentSpec(QuerySegmentSpec spec) + { + throw new UOE("Cannot override querySegmentSpec on window operator query. [%s]", spec); + } + + @Override + public Query withDataSource(DataSource dataSource) + { + return new WindowOperatorQuery( + dataSource, + getContext(), + rowSignature, + operators + ); + } + + @Override + public boolean equals(Object o) + { + if (this == o) { + return true; + } + if (o == null || getClass() != o.getClass()) { + return false; + } + if (!super.equals(o)) { + return false; + } + WindowOperatorQuery that = (WindowOperatorQuery) o; + return Objects.equals(rowSignature, that.rowSignature) && Objects.equals( + operators, + that.operators + ); + } + + @Override + public int hashCode() + { + return Objects.hash(super.hashCode(), rowSignature, operators); + } + + @Override + public String toString() + { + return "WindowOperatorQuery{" + + "dataSource='" + getDataSource() + '\'' + + ", querySegmentSpec=" + getQuerySegmentSpec() + + ", context=" + getContext() + + ", rowSignature=" + rowSignature + + ", operators=" + operators + + '}'; + } +} diff --git a/processing/src/main/java/org/apache/druid/query/operator/WindowOperatorQueryQueryRunnerFactory.java b/processing/src/main/java/org/apache/druid/query/operator/WindowOperatorQueryQueryRunnerFactory.java new file mode 100644 index 000000000000..78f64360761e --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/operator/WindowOperatorQueryQueryRunnerFactory.java @@ -0,0 +1,55 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator; + +import com.google.common.collect.Iterables; +import org.apache.druid.query.QueryProcessingPool; +import org.apache.druid.query.QueryRunner; +import org.apache.druid.query.QueryRunnerFactory; +import org.apache.druid.query.QueryToolChest; +import org.apache.druid.query.rowsandcols.RowsAndColumns; +import org.apache.druid.segment.Segment; + +public class WindowOperatorQueryQueryRunnerFactory implements QueryRunnerFactory +{ + public static final WindowOperatorQueryQueryToolChest TOOLCHEST = new WindowOperatorQueryQueryToolChest(); + + @Override + public QueryRunner createRunner(Segment segment) + { + return (queryPlus, responseContext) -> + new OperatorSequence(() -> new SegmentToRowsAndColumnsOperator(segment)); + } + + @Override + public QueryRunner mergeRunners( + QueryProcessingPool queryProcessingPool, + Iterable> queryRunners + ) + { + return Iterables.getOnlyElement(queryRunners); + } + + @Override + public QueryToolChest getToolchest() + { + return TOOLCHEST; + } +} diff --git a/processing/src/main/java/org/apache/druid/query/operator/WindowOperatorQueryQueryToolChest.java b/processing/src/main/java/org/apache/druid/query/operator/WindowOperatorQueryQueryToolChest.java new file mode 100644 index 000000000000..d02596900df0 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/operator/WindowOperatorQueryQueryToolChest.java @@ -0,0 +1,191 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator; + +import com.fasterxml.jackson.core.type.TypeReference; +import com.google.common.base.Function; +import com.google.common.base.Functions; +import com.google.common.collect.ImmutableMap; +import org.apache.druid.java.util.common.guava.Sequence; +import org.apache.druid.java.util.common.guava.Sequences; +import org.apache.druid.query.DefaultQueryMetrics; +import org.apache.druid.query.QueryMetrics; +import org.apache.druid.query.QueryPlus; +import org.apache.druid.query.QueryRunner; +import org.apache.druid.query.QueryToolChest; +import org.apache.druid.query.aggregation.MetricManipulationFn; +import org.apache.druid.query.context.ResponseContext; +import org.apache.druid.query.rowsandcols.RowsAndColumns; +import org.apache.druid.query.rowsandcols.column.Column; +import org.apache.druid.query.rowsandcols.column.ColumnAccessor; +import org.apache.druid.query.rowsandcols.column.NullColumnAccessor; +import org.apache.druid.segment.column.ColumnType; +import org.apache.druid.segment.column.RowSignature; + +import java.util.ArrayList; +import java.util.List; +import java.util.function.Supplier; + +public class WindowOperatorQueryQueryToolChest extends QueryToolChest +{ + + @Override + @SuppressWarnings("unchecked") + public QueryRunner mergeResults(QueryRunner runner) + { + return new RowsAndColumnsUnravelingQueryRunner( + (queryPlus, responseContext) -> { + final WindowOperatorQuery query = (WindowOperatorQuery) queryPlus.getQuery(); + final List opFactories = query.getOperators(); + + Supplier opSupplier = () -> { + Operator retVal = new SequenceOperator(runner.run(queryPlus, responseContext)); + for (OperatorFactory operatorFactory : opFactories) { + retVal = operatorFactory.wrap(retVal); + } + return retVal; + }; + + return new OperatorSequence(opSupplier); + } + ); + } + + @Override + public QueryMetrics makeMetrics(WindowOperatorQuery query) + { + return new DefaultQueryMetrics<>(); + } + + @Override + public Function makePreComputeManipulatorFn( + WindowOperatorQuery query, + MetricManipulationFn fn + ) + { + return Functions.identity(); + } + + @Override + public TypeReference getResultTypeReference() + { + return new TypeReference() + { + }; + } + + @Override + public RowSignature resultArraySignature(WindowOperatorQuery query) + { + return query.getRowSignature(); + } + + @Override + @SuppressWarnings({"unchecked", "rawtypes"}) + public Sequence resultsAsArrays( + WindowOperatorQuery query, + Sequence resultSequence + ) + { + // Dark magic; see RowsAndColumnsUnravelingQueryRunner. + return (Sequence) resultSequence; + } + + /** + * This class exists to unravel the RowsAndColumns that are used in this query and make it the return Sequence + * actually be a Sequence of rows. This is relatively broken in a number of regards, the most obvious of which + * is that it is going to run counter to the stated class on the Generic of the QueryToolChest. That is, the + * code makes it look like you are getting a Sequence of RowsAndColumns, but, by using this, the query will + * actually ultimately produce a Sequence of Object[]. This works because of type Erasure in Java (it's all Object + * at the end of the day). + *

    + * While it might seem like this will break all sorts of things, the Generic type is actually there more as a type + * "hint" to make the writing of the ToolChest and Factory and stuff a bit more simple. Any caller of this cannot + * truly depend on the type anyway other than to just throw it across the wire, so this should just magically work + * even though it looks like it shouldn't even compile. + *

    + * Not our proudest moment, but we use the tools available to us. + */ + @SuppressWarnings({"unchecked", "rawtypes"}) + private static class RowsAndColumnsUnravelingQueryRunner implements QueryRunner + { + + private final QueryRunner baseQueryRunner; + + private RowsAndColumnsUnravelingQueryRunner( + QueryRunner baseQueryRunner + ) + { + this.baseQueryRunner = baseQueryRunner; + } + + @Override + public Sequence run( + QueryPlus queryPlus, + ResponseContext responseContext + ) + { + // We only want to do this operation once at the very, very top of the execution tree. So we check and set + // a context parameter so that if this merge code runs anywhere else, it will skip this part. + final WindowOperatorQuery query = (WindowOperatorQuery) queryPlus.getQuery(); + if (query.context().getBoolean("unravel", true)) { + final Sequence baseSequence = baseQueryRunner.run( + queryPlus.withQuery(query.withOverriddenContext(ImmutableMap.of("unravel", false))), + responseContext + ); + + final RowSignature rowSignature = query.getRowSignature(); + return baseSequence.flatMap( + rac -> { + List results = new ArrayList<>(rac.numRows()); + + ColumnAccessor[] accessors = new ColumnAccessor[rowSignature.size()]; + int index = 0; + for (String columnName : rowSignature.getColumnNames()) { + final Column column = rac.findColumn(columnName); + if (column == null) { + final ColumnType columnType = rowSignature + .getColumnType(columnName) + .orElse(ColumnType.UNKNOWN_COMPLEX); + + accessors[index] = new NullColumnAccessor(columnType, rac.numRows()); + } else { + accessors[index] = column.toAccessor(); + } + ++index; + } + + for (int i = 0; i < rac.numRows(); ++i) { + Object[] objArr = new Object[accessors.length]; + for (int j = 0; j < accessors.length; j++) { + objArr[j] = accessors[j].getObject(i); + } + results.add(objArr); + } + + return Sequences.simple(results); + } + ); + } + + return baseQueryRunner.run(queryPlus, responseContext); + } + } +} diff --git a/processing/src/main/java/org/apache/druid/query/operator/WindowProcessorOperator.java b/processing/src/main/java/org/apache/druid/query/operator/WindowProcessorOperator.java new file mode 100644 index 000000000000..d95f598d3993 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/operator/WindowProcessorOperator.java @@ -0,0 +1,67 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator; + +import org.apache.druid.query.operator.window.Processor; +import org.apache.druid.query.rowsandcols.RowsAndColumns; + +/** + * An Operator that applies a {@link Processor}, see javadoc on that interface for an explanation. + */ +public class WindowProcessorOperator implements Operator +{ + private final Processor windowProcessor; + private final Operator child; + + public WindowProcessorOperator( + Processor windowProcessor, + Operator child + ) + { + this.windowProcessor = windowProcessor; + this.child = child; + } + + @Override + public void open() + { + child.open(); + } + + @Override + public RowsAndColumns next() + { + return windowProcessor.process(child.next()); + } + + @Override + public boolean hasNext() + { + return child.hasNext(); + } + + @Override + public void close(boolean cascade) + { + if (cascade) { + child.close(cascade); + } + } +} diff --git a/processing/src/main/java/org/apache/druid/query/operator/window/ComposingProcessor.java b/processing/src/main/java/org/apache/druid/query/operator/window/ComposingProcessor.java new file mode 100644 index 000000000000..a4fa74967f61 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/operator/window/ComposingProcessor.java @@ -0,0 +1,78 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator.window; + +import com.fasterxml.jackson.annotation.JsonCreator; +import com.fasterxml.jackson.annotation.JsonProperty; +import org.apache.druid.query.rowsandcols.RowsAndColumns; + +import java.util.Arrays; + +public class ComposingProcessor implements Processor +{ + private final Processor[] processors; + + @JsonCreator + public ComposingProcessor( + @JsonProperty("processors") Processor... processors + ) + { + this.processors = processors; + } + + @JsonProperty("processors") + public Processor[] getProcessors() + { + return processors; + } + + @Override + public RowsAndColumns process(RowsAndColumns incomingPartition) + { + RowsAndColumns retVal = incomingPartition; + for (int i = processors.length - 1; i >= 0; --i) { + retVal = processors[i].process(retVal); + } + return retVal; + } + + @Override + public boolean validateEquivalent(Processor otherProcessor) + { + if (otherProcessor instanceof ComposingProcessor) { + ComposingProcessor other = (ComposingProcessor) otherProcessor; + for (int i = 0; i < processors.length; ++i) { + if (!processors[i].validateEquivalent(other.processors[i])) { + return false; + } + } + return true; + } + return false; + } + + @Override + public String toString() + { + return "ComposingProcessor{" + + "processors=" + Arrays.toString(processors) + + '}'; + } +} diff --git a/processing/src/main/java/org/apache/druid/query/operator/window/Processor.java b/processing/src/main/java/org/apache/druid/query/operator/window/Processor.java new file mode 100644 index 000000000000..94e8c74a6b50 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/operator/window/Processor.java @@ -0,0 +1,83 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator.window; + +import com.fasterxml.jackson.annotation.JsonSubTypes; +import com.fasterxml.jackson.annotation.JsonTypeInfo; +import org.apache.druid.query.operator.window.ranking.WindowCumeDistProcessor; +import org.apache.druid.query.operator.window.ranking.WindowDenseRankProcessor; +import org.apache.druid.query.operator.window.ranking.WindowPercentileProcessor; +import org.apache.druid.query.operator.window.ranking.WindowRankProcessor; +import org.apache.druid.query.operator.window.ranking.WindowRowNumberProcessor; +import org.apache.druid.query.operator.window.value.WindowFirstProcessor; +import org.apache.druid.query.operator.window.value.WindowLastProcessor; +import org.apache.druid.query.operator.window.value.WindowOffsetProcessor; +import org.apache.druid.query.rowsandcols.RowsAndColumns; + +/** + * A Processor is a bit of logic that processes a single RowsAndColumns object to produce a new RowsAndColumns + * object. Generally speaking, it is used to add or alter columns in a batch-oriented fashion. + *

    + * This interface was created to support windowing functions, where the windowing function can be implemented + * assuming that each RowsAndColumns object represents one partition. Thus, the window function implementation + * can only need to worry about how to process a single partition at a time and something external to the window + * function worries about providing data with the correct partitioning. + *

    + * Over time, it's possible that this interface is used for other purposes as well, but the fundamental idea of + * usages of the interface should always be doing a one-to-one transformation of RowsAndColumns objects. That is, + * it's a RowsAndColumns in and a RowsAndColumns out. + */ +@JsonTypeInfo(use = JsonTypeInfo.Id.NAME, property = "type") +@JsonSubTypes(value = { + @JsonSubTypes.Type(name = "composing", value = ComposingProcessor.class), + @JsonSubTypes.Type(name = "cumeDist", value = WindowCumeDistProcessor.class), + @JsonSubTypes.Type(name = "denseRank", value = WindowDenseRankProcessor.class), + @JsonSubTypes.Type(name = "percentile", value = WindowPercentileProcessor.class), + @JsonSubTypes.Type(name = "rank", value = WindowRankProcessor.class), + @JsonSubTypes.Type(name = "rowNumber", value = WindowRowNumberProcessor.class), + @JsonSubTypes.Type(name = "first", value = WindowFirstProcessor.class), + @JsonSubTypes.Type(name = "last", value = WindowLastProcessor.class), + @JsonSubTypes.Type(name = "offset", value = WindowOffsetProcessor.class), + @JsonSubTypes.Type(name = "aggregate", value = WindowAggregateProcessor.class), +}) +public interface Processor +{ + /** + * Applies the logic of the processor to a RowsAndColumns object + * + * @param incomingPartition the incoming RowsAndColumns object + * @return the transformed RowsAndColumns object + */ + RowsAndColumns process(RowsAndColumns incomingPartition); + + /** + * Validates the equivalence of the Processors. This is similar to @{code .equals} but is its own method + * so that it can ignore certain fields that would be important for a true equality check. Namely, two Processors + * defined the same way but with different output names can be considered equivalent even though they are not equal. + *

    + * This primarily exists to simplify tests, where this equivalence can be used to validate that the Processors + * created by the SQL planner are actually equivalent to what we expect without needing to be overly dependent on + * how the planner names the output columns + * + * @param otherProcessor the processor to test equivalence of + * @return boolean identifying if these processors should be considered equivalent to each other. + */ + boolean validateEquivalent(Processor otherProcessor); +} diff --git a/processing/src/main/java/org/apache/druid/query/operator/window/WindowAggregateProcessor.java b/processing/src/main/java/org/apache/druid/query/operator/window/WindowAggregateProcessor.java new file mode 100644 index 000000000000..630d1145f770 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/operator/window/WindowAggregateProcessor.java @@ -0,0 +1,131 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator.window; + +import com.fasterxml.jackson.annotation.JsonCreator; +import com.fasterxml.jackson.annotation.JsonProperty; +import org.apache.druid.query.aggregation.AggregatorFactory; +import org.apache.druid.query.rowsandcols.AppendableRowsAndColumns; +import org.apache.druid.query.rowsandcols.DefaultOnHeapAggregatable; +import org.apache.druid.query.rowsandcols.OnHeapAggregatable; +import org.apache.druid.query.rowsandcols.OnHeapCumulativeAggregatable; +import org.apache.druid.query.rowsandcols.RowsAndColumns; +import org.apache.druid.query.rowsandcols.column.ConstantObjectColumn; +import org.apache.druid.query.rowsandcols.column.ObjectArrayColumn; + +import javax.annotation.Nullable; +import java.util.ArrayList; +import java.util.List; +import java.util.Objects; + +public class WindowAggregateProcessor implements Processor +{ + @Nullable + private static List emptyToNull(List list) + { + if (list == null || list.isEmpty()) { + return null; + } else { + return list; + } + } + + private final List aggregations; + private final List cumulativeAggregations; + + @JsonCreator + public WindowAggregateProcessor( + @JsonProperty("aggregations") List aggregations, + @JsonProperty("cumulativeAggregations") List cumulativeAggregations + ) + { + this.aggregations = emptyToNull(aggregations); + this.cumulativeAggregations = emptyToNull(cumulativeAggregations); + } + + @JsonProperty("aggregations") + public List getAggregations() + { + return aggregations; + } + + @JsonProperty("cumulativeAggregations") + public List getCumulativeAggregations() + { + return cumulativeAggregations; + } + + @Override + public RowsAndColumns process(RowsAndColumns inputPartition) + { + AppendableRowsAndColumns retVal = RowsAndColumns.expectAppendable(inputPartition); + + if (aggregations != null) { + OnHeapAggregatable aggregatable = inputPartition.as(OnHeapAggregatable.class); + if (aggregatable == null) { + aggregatable = new DefaultOnHeapAggregatable(inputPartition); + } + final ArrayList aggregatedVals = aggregatable.aggregateAll(aggregations); + + for (int i = 0; i < aggregations.size(); ++i) { + final AggregatorFactory agg = aggregations.get(i); + retVal.addColumn( + agg.getName(), + new ConstantObjectColumn(aggregatedVals.get(i), inputPartition.numRows(), agg.getResultType()) + ); + } + } + + if (cumulativeAggregations != null) { + OnHeapCumulativeAggregatable cummulativeAgg = inputPartition.as(OnHeapCumulativeAggregatable.class); + if (cummulativeAgg == null) { + cummulativeAgg = new DefaultOnHeapAggregatable(inputPartition); + } + final ArrayList cumulativeVals = cummulativeAgg.aggregateCumulative(cumulativeAggregations); + + for (int i = 0; i < cumulativeAggregations.size(); ++i) { + final AggregatorFactory agg = cumulativeAggregations.get(i); + retVal.addColumn(agg.getName(), new ObjectArrayColumn(cumulativeVals.get(i), agg.getResultType())); + } + } + + return retVal; + } + + @Override + public boolean validateEquivalent(Processor otherProcessor) + { + if (otherProcessor instanceof WindowAggregateProcessor) { + WindowAggregateProcessor other = (WindowAggregateProcessor) otherProcessor; + return Objects.equals(aggregations, other.aggregations) + && Objects.equals(cumulativeAggregations, other.cumulativeAggregations); + } + return false; + } + + @Override + public String toString() + { + return "WindowAggregateProcessor{" + + "aggregations=" + aggregations + + ", cumulativeAggregations=" + cumulativeAggregations + + '}'; + } +} diff --git a/processing/src/main/java/org/apache/druid/query/operator/window/ranking/WindowCumeDistProcessor.java b/processing/src/main/java/org/apache/druid/query/operator/window/ranking/WindowCumeDistProcessor.java new file mode 100644 index 000000000000..c798081339e7 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/operator/window/ranking/WindowCumeDistProcessor.java @@ -0,0 +1,60 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator.window.ranking; + +import com.fasterxml.jackson.annotation.JsonCreator; +import com.fasterxml.jackson.annotation.JsonProperty; +import org.apache.druid.query.rowsandcols.RowsAndColumns; +import org.apache.druid.query.rowsandcols.column.DoubleArrayColumn; + +import java.util.Arrays; +import java.util.List; + +/** + * This Processor assumes that data has already been sorted for it. It does not re-sort the data and if it is given + * data that is not in the correct sorted order, its operation is undefined. + */ +public class WindowCumeDistProcessor extends WindowRankingProcessorBase +{ + @JsonCreator + public WindowCumeDistProcessor( + @JsonProperty("group") List groupingCols, + @JsonProperty("outputColumn") String outputColumn + ) + { + super(groupingCols, outputColumn); + } + + @Override + public RowsAndColumns process(RowsAndColumns incomingPartition) + { + return processInternal(incomingPartition, groupings -> { + final double[] ranks = new double[incomingPartition.numRows()]; + for (int i = 1; i < groupings.length; ++i) { + final int start = groupings[i - 1]; + final int end = groupings[i]; + double relativeRank = end / (double) ranks.length; + Arrays.fill(ranks, start, end, relativeRank); + } + + return new DoubleArrayColumn(ranks); + }); + } +} diff --git a/processing/src/main/java/org/apache/druid/query/operator/window/ranking/WindowDenseRankProcessor.java b/processing/src/main/java/org/apache/druid/query/operator/window/ranking/WindowDenseRankProcessor.java new file mode 100644 index 000000000000..24b05a6f3460 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/operator/window/ranking/WindowDenseRankProcessor.java @@ -0,0 +1,59 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator.window.ranking; + +import com.fasterxml.jackson.annotation.JsonCreator; +import com.fasterxml.jackson.annotation.JsonProperty; +import org.apache.druid.query.rowsandcols.RowsAndColumns; +import org.apache.druid.query.rowsandcols.column.IntArrayColumn; + +import java.util.Arrays; +import java.util.List; + +/** + * This Processor assumes that data has already been sorted for it. It does not re-sort the data and if it is given + * data that is not in the correct sorted order, its operation is undefined. + */ +public class WindowDenseRankProcessor extends WindowRankingProcessorBase +{ + @JsonCreator + public WindowDenseRankProcessor( + @JsonProperty("group") List groupingCols, + @JsonProperty("outputColumn") String outputColumn + ) + { + super(groupingCols, outputColumn); + } + + @Override + public RowsAndColumns process(RowsAndColumns incomingPartition) + { + return processInternal(incomingPartition, groupings -> { + final int[] ranks = new int[incomingPartition.numRows()]; + for (int i = 1; i < groupings.length; ++i) { + final int start = groupings[i - 1]; + final int end = groupings[i]; + Arrays.fill(ranks, start, end, i); + } + + return new IntArrayColumn(ranks); + }); + } +} diff --git a/processing/src/main/java/org/apache/druid/query/operator/window/ranking/WindowPercentileProcessor.java b/processing/src/main/java/org/apache/druid/query/operator/window/ranking/WindowPercentileProcessor.java new file mode 100644 index 000000000000..e8c9aec2ab4c --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/operator/window/ranking/WindowPercentileProcessor.java @@ -0,0 +1,106 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator.window.ranking; + +import com.fasterxml.jackson.annotation.JsonCreator; +import com.fasterxml.jackson.annotation.JsonProperty; +import com.google.common.base.Preconditions; +import org.apache.druid.query.operator.window.Processor; +import org.apache.druid.query.rowsandcols.AppendableRowsAndColumns; +import org.apache.druid.query.rowsandcols.RowsAndColumns; +import org.apache.druid.query.rowsandcols.column.IntArrayColumn; + +import java.util.Arrays; + +public class WindowPercentileProcessor implements Processor +{ + private final int numBuckets; + private final String outputColumn; + + @JsonCreator + public WindowPercentileProcessor( + @JsonProperty("outputColumn") String outputColumn, + @JsonProperty("numBuckets") int numBuckets + ) + { + Preconditions.checkArgument(numBuckets > 0, "numBuckets[%s] must be greater than zero", numBuckets); + + this.outputColumn = outputColumn; + this.numBuckets = numBuckets; + } + + @JsonProperty("numBuckets") + public int getNumBuckets() + { + return numBuckets; + } + + @JsonProperty("outputColumn") + public String getOutputColumn() + { + return outputColumn; + } + + @Override + public RowsAndColumns process(RowsAndColumns incomingPartition) + { + AppendableRowsAndColumns retVal = RowsAndColumns.expectAppendable(incomingPartition); + + int numRows = incomingPartition.numRows(); + int countPerBucket = numRows / numBuckets; + int extraRows = numRows % numBuckets; + + int index = 0; + int[] bucketVals = new int[numRows]; + for (int i = 0; i < numBuckets; ++i) { + int nextIndex = index + countPerBucket; + if (extraRows > 0) { + ++nextIndex; + --extraRows; + } + + // Buckets are 1-indexed, so we fill with i+1 + Arrays.fill(bucketVals, index, nextIndex, i + 1); + + index = nextIndex; + } + + retVal.addColumn(outputColumn, new IntArrayColumn(bucketVals)); + return retVal; + } + + @Override + public boolean validateEquivalent(Processor otherProcessor) + { + if (otherProcessor instanceof WindowPercentileProcessor) { + return numBuckets == ((WindowPercentileProcessor) otherProcessor).numBuckets; + } + return false; + } + + @Override + public String toString() + { + return "WindowPercentileProcessor{" + + "numBuckets=" + numBuckets + + ", outputColumn='" + outputColumn + '\'' + + '}'; + } +} diff --git a/processing/src/main/java/org/apache/druid/query/operator/window/ranking/WindowRankProcessor.java b/processing/src/main/java/org/apache/druid/query/operator/window/ranking/WindowRankProcessor.java new file mode 100644 index 000000000000..2d9b21863de8 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/operator/window/ranking/WindowRankProcessor.java @@ -0,0 +1,108 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator.window.ranking; + +import com.fasterxml.jackson.annotation.JsonCreator; +import com.fasterxml.jackson.annotation.JsonProperty; +import org.apache.druid.query.operator.window.Processor; +import org.apache.druid.query.rowsandcols.RowsAndColumns; +import org.apache.druid.query.rowsandcols.column.DoubleArrayColumn; +import org.apache.druid.query.rowsandcols.column.IntArrayColumn; + +import java.util.Arrays; +import java.util.List; + +/** + * This Processor assumes that data has already been sorted for it. It does not re-sort the data and if it is given + * data that is not in the correct sorted order, its operation is undefined. + */ +public class WindowRankProcessor extends WindowRankingProcessorBase +{ + private final boolean asPercent; + + @JsonCreator + public WindowRankProcessor( + @JsonProperty("group") List groupingCols, + @JsonProperty("outputColumn") String outputColumn, + @JsonProperty("asPercent") boolean asPercent + ) + { + super(groupingCols, outputColumn); + this.asPercent = asPercent; + } + + @JsonProperty("asPercent") + public boolean isAsPercent() + { + return asPercent; + } + + @Override + public RowsAndColumns process(RowsAndColumns incomingPartition) + { + if (asPercent) { + return processInternal(incomingPartition, groupings -> { + final double[] percentages = new double[incomingPartition.numRows()]; + if (percentages.length > 1) { + final double denominator = percentages.length - 1; + + for (int i = 1; i < groupings.length; ++i) { + final int start = groupings[i - 1]; + final int end = groupings[i]; + Arrays.fill(percentages, start, end, start / denominator); + } + } + + return new DoubleArrayColumn(percentages); + }); + } + + return processInternal(incomingPartition, groupings -> { + final int[] ranks = new int[incomingPartition.numRows()]; + + for (int i = 1; i < groupings.length; ++i) { + final int start = groupings[i - 1]; + final int end = groupings[i]; + Arrays.fill(ranks, start, end, start + 1); + } + + return new IntArrayColumn(ranks); + }); + } + + @Override + public boolean validateEquivalent(Processor otherProcessor) + { + if (otherProcessor instanceof WindowRankProcessor) { + WindowRankProcessor other = (WindowRankProcessor) otherProcessor; + return asPercent == other.asPercent && intervalValidation(other); + } + return false; + } + + @Override + public String toString() + { + return "WindowRankProcessor{" + + internalToString() + + ", asPercent=" + asPercent + + '}'; + } +} diff --git a/processing/src/main/java/org/apache/druid/query/operator/window/ranking/WindowRankingProcessorBase.java b/processing/src/main/java/org/apache/druid/query/operator/window/ranking/WindowRankingProcessorBase.java new file mode 100644 index 000000000000..105c586aa195 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/operator/window/ranking/WindowRankingProcessorBase.java @@ -0,0 +1,103 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator.window.ranking; + +import com.fasterxml.jackson.annotation.JsonProperty; +import org.apache.druid.query.operator.window.Processor; +import org.apache.druid.query.rowsandcols.AppendableRowsAndColumns; +import org.apache.druid.query.rowsandcols.DefaultSortedGroupPartitioner; +import org.apache.druid.query.rowsandcols.RowsAndColumns; +import org.apache.druid.query.rowsandcols.SortedGroupPartitioner; +import org.apache.druid.query.rowsandcols.column.Column; + +import java.util.List; +import java.util.function.Function; + +/** + * This Processor assumes that data has already been sorted for it. It does not re-sort the data and if it is given + * data that is not in the correct sorted order, its operation is undefined. + */ +public abstract class WindowRankingProcessorBase implements Processor +{ + private final List groupingCols; + private final String outputColumn; + + public WindowRankingProcessorBase( + List groupingCols, + String outputColumn + ) + { + this.groupingCols = groupingCols; + this.outputColumn = outputColumn; + } + + @JsonProperty("group") + public List getGroupingCols() + { + return groupingCols; + } + + @JsonProperty("outputColumn") + public String getOutputColumn() + { + return outputColumn; + } + + public RowsAndColumns processInternal( + RowsAndColumns incomingPartition, + Function fn + ) + { + final AppendableRowsAndColumns retVal = RowsAndColumns.expectAppendable(incomingPartition); + + SortedGroupPartitioner groupPartitioner = incomingPartition.as(SortedGroupPartitioner.class); + if (groupPartitioner == null) { + groupPartitioner = new DefaultSortedGroupPartitioner(incomingPartition); + } + + retVal.addColumn(outputColumn, fn.apply(groupPartitioner.computeBoundaries(groupingCols))); + return retVal; + } + + @Override + public boolean validateEquivalent(Processor otherProcessor) + { + return getClass() == otherProcessor.getClass() + && intervalValidation((WindowRankingProcessorBase) otherProcessor); + } + + protected boolean intervalValidation(WindowRankingProcessorBase other) + { + // Only input needs to be the same for the processors to produce equivalent results + return groupingCols.equals(other.groupingCols); + } + + @Override + public String toString() + { + return getClass().getSimpleName() + "{" + internalToString() + '}'; + } + + protected String internalToString() + { + return "groupingCols=" + groupingCols + + ", outputColumn='" + outputColumn + '\''; + } +} diff --git a/processing/src/main/java/org/apache/druid/query/operator/window/ranking/WindowRowNumberProcessor.java b/processing/src/main/java/org/apache/druid/query/operator/window/ranking/WindowRowNumberProcessor.java new file mode 100644 index 000000000000..6c998afa6099 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/operator/window/ranking/WindowRowNumberProcessor.java @@ -0,0 +1,131 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator.window.ranking; + +import com.fasterxml.jackson.annotation.JsonCreator; +import com.fasterxml.jackson.annotation.JsonProperty; +import org.apache.druid.query.operator.window.Processor; +import org.apache.druid.query.rowsandcols.AppendableRowsAndColumns; +import org.apache.druid.query.rowsandcols.RowsAndColumns; +import org.apache.druid.query.rowsandcols.column.ColumnAccessor; +import org.apache.druid.query.rowsandcols.column.ColumnAccessorBasedColumn; +import org.apache.druid.segment.column.ColumnType; + +public class WindowRowNumberProcessor implements Processor +{ + private final String outputColumn; + + @JsonCreator + public WindowRowNumberProcessor( + @JsonProperty("outputColumn") String outputColumn + ) + { + this.outputColumn = outputColumn; + } + + @JsonProperty("outputColumn") + public String getOutputColumn() + { + return outputColumn; + } + + @Override + public RowsAndColumns process(RowsAndColumns incomingPartition) + { + final AppendableRowsAndColumns retVal = RowsAndColumns.expectAppendable(incomingPartition); + retVal.addColumn( + outputColumn, + new ColumnAccessorBasedColumn( + new ColumnAccessor() + { + @Override + public ColumnType getType() + { + return ColumnType.LONG; + } + + @Override + public int numRows() + { + return incomingPartition.numRows(); + } + + @Override + public boolean isNull(int rowNum) + { + return false; + } + + @Override + public Object getObject(int rowNum) + { + return getInt(rowNum); + } + + @Override + public double getDouble(int rowNum) + { + return getInt(rowNum); + } + + @Override + public float getFloat(int rowNum) + { + return getInt(rowNum); + } + + @Override + public long getLong(int rowNum) + { + return getInt(rowNum); + } + + @Override + public int getInt(int rowNum) + { + // cell is 0-indexed, rowNumbers are 1-indexed, so add 1. + return rowNum + 1; + } + + @Override + public int compareCells(int lhsRowNum, int rhsRowNum) + { + return Integer.compare(lhsRowNum, rhsRowNum); + } + } + ) + ); + return retVal; + } + + @Override + public boolean validateEquivalent(Processor otherProcessor) + { + return otherProcessor instanceof WindowRowNumberProcessor; + } + + @Override + public String toString() + { + return "WindowRowNumberProcessor{" + + "outputColumn='" + outputColumn + '\'' + + '}'; + } +} diff --git a/processing/src/main/java/org/apache/druid/query/operator/window/value/ShiftedColumnAccessorBase.java b/processing/src/main/java/org/apache/druid/query/operator/window/value/ShiftedColumnAccessorBase.java new file mode 100644 index 000000000000..22288a197f2d --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/operator/window/value/ShiftedColumnAccessorBase.java @@ -0,0 +1,130 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator.window.value; + +import org.apache.druid.query.rowsandcols.column.ColumnAccessor; +import org.apache.druid.segment.column.ColumnType; + +public abstract class ShiftedColumnAccessorBase implements ColumnAccessor +{ + private final ColumnAccessor accessor; + + public ShiftedColumnAccessorBase(ColumnAccessor accessor) + { + this.accessor = accessor; + } + + @Override + public ColumnType getType() + { + return accessor.getType(); + } + + @Override + public int numRows() + { + return accessor.numRows(); + } + + @Override + public boolean isNull(int rowNum) + { + final int actualCell = getActualCell(rowNum); + if (outsideBounds(actualCell)) { + return true; + } + return accessor.isNull(actualCell); + } + + @Override + public Object getObject(int rowNum) + { + final int actualCell = getActualCell(rowNum); + if (outsideBounds(actualCell)) { + return null; + } + return accessor.getObject(actualCell); + } + + @Override + public double getDouble(int rowNum) + { + final int actualCell = getActualCell(rowNum); + if (outsideBounds(actualCell)) { + return 0.0D; + } + return accessor.getDouble(actualCell); + } + + @Override + public float getFloat(int rowNum) + { + final int actualCell = getActualCell(rowNum); + if (outsideBounds(actualCell)) { + return 0.0F; + } + return accessor.getFloat(actualCell); + } + + @Override + public long getLong(int rowNum) + { + final int actualCell = getActualCell(rowNum); + if (outsideBounds(actualCell)) { + return 0L; + } + return accessor.getLong(actualCell); + } + + @Override + public int getInt(int rowNum) + { + final int actualCell = getActualCell(rowNum); + if (outsideBounds(actualCell)) { + return 0; + } + return accessor.getInt(actualCell); + } + + @Override + public int compareCells(int lhsRowNum, int rhsRowNum) + { + int actualLhsCell = getActualCell(lhsRowNum); + int actualRhsCell = getActualCell(rhsRowNum); + if (outsideBounds(actualLhsCell)) { + if (outsideBounds(actualRhsCell)) { + // Both are null + return 0; + } else { + return accessor.isNull(actualRhsCell) ? 0 : -1; + } + } else { + if (outsideBounds(actualRhsCell)) { + return accessor.isNull(actualLhsCell) ? 0 : 1; + } else { + return accessor.compareCells(actualLhsCell, actualRhsCell); + } + } + } + + protected abstract int getActualCell(int cell); + + protected abstract boolean outsideBounds(int cell); +} diff --git a/processing/src/main/java/org/apache/druid/query/operator/window/value/WindowFirstProcessor.java b/processing/src/main/java/org/apache/druid/query/operator/window/value/WindowFirstProcessor.java new file mode 100644 index 000000000000..fe0cca8884c3 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/operator/window/value/WindowFirstProcessor.java @@ -0,0 +1,50 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator.window.value; + +import com.fasterxml.jackson.annotation.JsonCreator; +import com.fasterxml.jackson.annotation.JsonProperty; +import org.apache.druid.query.rowsandcols.RowsAndColumns; +import org.apache.druid.query.rowsandcols.column.ColumnAccessor; +import org.apache.druid.query.rowsandcols.column.ConstantObjectColumn; + +public class WindowFirstProcessor extends WindowValueProcessorBase +{ + @JsonCreator + public WindowFirstProcessor( + @JsonProperty("inputColumn") String inputColumn, + @JsonProperty("outputColumn") String outputColumn + ) + { + super(inputColumn, outputColumn); + } + + @Override + public RowsAndColumns process(RowsAndColumns incomingPartition) + { + return processInternal( + incomingPartition, + column -> { + final ColumnAccessor accessor = column.toAccessor(); + return new ConstantObjectColumn(accessor.getObject(0), accessor.numRows(), accessor.getType()); + } + ); + } +} diff --git a/processing/src/main/java/org/apache/druid/query/operator/window/value/WindowLastProcessor.java b/processing/src/main/java/org/apache/druid/query/operator/window/value/WindowLastProcessor.java new file mode 100644 index 000000000000..2e28962e4346 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/operator/window/value/WindowLastProcessor.java @@ -0,0 +1,53 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator.window.value; + +import com.fasterxml.jackson.annotation.JsonCreator; +import com.fasterxml.jackson.annotation.JsonProperty; +import org.apache.druid.java.util.common.ISE; +import org.apache.druid.query.rowsandcols.RowsAndColumns; +import org.apache.druid.query.rowsandcols.column.ColumnAccessor; +import org.apache.druid.query.rowsandcols.column.ConstantObjectColumn; + +public class WindowLastProcessor extends WindowValueProcessorBase +{ + @JsonCreator + public WindowLastProcessor( + @JsonProperty("inputColumn") String inputColumn, + @JsonProperty("outputColumn") String outputColumn + ) + { + super(inputColumn, outputColumn); + } + + @Override + public RowsAndColumns process(RowsAndColumns input) + { + final int lastIndex = input.numRows() - 1; + if (lastIndex < 0) { + throw new ISE("Called with an input partition of size 0. The call site needs to not do that."); + } + + return processInternal(input, column -> { + final ColumnAccessor accessor = column.toAccessor(); + return new ConstantObjectColumn(accessor.getObject(lastIndex), accessor.numRows(), accessor.getType()); + }); + } +} diff --git a/processing/src/main/java/org/apache/druid/query/operator/window/value/WindowOffsetProcessor.java b/processing/src/main/java/org/apache/druid/query/operator/window/value/WindowOffsetProcessor.java new file mode 100644 index 000000000000..4128731be840 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/operator/window/value/WindowOffsetProcessor.java @@ -0,0 +1,89 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator.window.value; + +import com.fasterxml.jackson.annotation.JsonCreator; +import com.fasterxml.jackson.annotation.JsonProperty; +import org.apache.druid.query.operator.window.Processor; +import org.apache.druid.query.rowsandcols.RowsAndColumns; +import org.apache.druid.query.rowsandcols.column.ColumnAccessorBasedColumn; + +public class WindowOffsetProcessor extends WindowValueProcessorBase +{ + private final int offset; + + @JsonCreator + public WindowOffsetProcessor( + @JsonProperty("inputColumn") String inputColumn, + @JsonProperty("outputColumn") String outputColumn, + @JsonProperty("offset") int offset + ) + { + super(inputColumn, outputColumn); + this.offset = offset; + } + + @JsonProperty("offset") + public int getOffset() + { + return offset; + } + + @Override + public RowsAndColumns process(RowsAndColumns input) + { + final int numRows = input.numRows(); + + return processInternal(input, column -> new ColumnAccessorBasedColumn( + new ShiftedColumnAccessorBase(column.toAccessor()) + { + @Override + protected int getActualCell(int cell) + { + return cell + offset; + } + + @Override + protected boolean outsideBounds(int actualCell) + { + return actualCell < 0 || actualCell >= numRows; + } + })); + } + + @Override + public boolean validateEquivalent(Processor otherProcessor) + { + if (otherProcessor instanceof WindowOffsetProcessor) { + WindowOffsetProcessor other = (WindowOffsetProcessor) otherProcessor; + return offset == other.offset && intervalValidation(other); + } + return false; + } + + @Override + public String toString() + { + return "WindowOffsetProcessor{" + + internalToString() + + ", offset=" + offset + + '}'; + } +} diff --git a/processing/src/main/java/org/apache/druid/query/operator/window/value/WindowValueProcessorBase.java b/processing/src/main/java/org/apache/druid/query/operator/window/value/WindowValueProcessorBase.java new file mode 100644 index 000000000000..5486f69b9d3b --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/operator/window/value/WindowValueProcessorBase.java @@ -0,0 +1,103 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator.window.value; + +import com.fasterxml.jackson.annotation.JsonProperty; +import org.apache.druid.java.util.common.ISE; +import org.apache.druid.query.operator.window.Processor; +import org.apache.druid.query.rowsandcols.AppendableRowsAndColumns; +import org.apache.druid.query.rowsandcols.RowsAndColumns; +import org.apache.druid.query.rowsandcols.column.Column; + +import java.util.function.Function; + +public abstract class WindowValueProcessorBase implements Processor +{ + private final String inputColumn; + private final String outputColumn; + + public WindowValueProcessorBase( + String inputColumn, + String outputColumn + ) + { + this.inputColumn = inputColumn; + this.outputColumn = outputColumn; + } + + @JsonProperty("inputColumn") + public String getInputColumn() + { + return inputColumn; + } + + @JsonProperty("outputColumn") + public String getOutputColumn() + { + return outputColumn; + } + + /** + * This implements the common logic between the various value processors. It looks like it could be static, but if + * it is static then the lambda becomes polymorphic. We keep it as a member method of the base class so taht the + * JVM can inline it and specialize the lambda + * + * @param input incoming RowsAndColumns, as in Processor.process + * @param fn function that converts the input column into the output column + * @return RowsAndColumns, as in Processor.process + */ + public RowsAndColumns processInternal(RowsAndColumns input, Function fn) + { + final AppendableRowsAndColumns retVal = RowsAndColumns.expectAppendable(input); + + final Column column = input.findColumn(inputColumn); + if (column == null) { + throw new ISE("column[%s] doesn't exist, but window function FIRST wants it to", inputColumn); + } + + retVal.addColumn(outputColumn, fn.apply(column)); + return retVal; + } + + @Override + public boolean validateEquivalent(Processor otherProcessor) + { + return getClass() == otherProcessor.getClass() + && intervalValidation((WindowValueProcessorBase) otherProcessor); + } + + protected boolean intervalValidation(WindowValueProcessorBase other) + { + // Only input needs to be the same for the processors to produce equivalent results + return inputColumn.equals(other.inputColumn); + } + + @Override + public String toString() + { + return getClass().getSimpleName() + "{" + internalToString() + '}'; + } + + protected String internalToString() + { + return "inputColumn=" + inputColumn + + ", outputColumn='" + outputColumn + '\''; + } +} diff --git a/processing/src/main/java/org/apache/druid/query/rowsandcols/AppendableRowsAndColumns.java b/processing/src/main/java/org/apache/druid/query/rowsandcols/AppendableRowsAndColumns.java new file mode 100644 index 000000000000..55b197db6122 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/rowsandcols/AppendableRowsAndColumns.java @@ -0,0 +1,38 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.rowsandcols; + +import org.apache.druid.query.rowsandcols.column.Column; + +/** + * A RowsAndColumns that supports appending columns. This interface is particularly useful because even if there is + * some composition of code that works with RowsAndColumns, we would like to add the columns to a singular base object + * instead of build up a complex object graph. + */ +public interface AppendableRowsAndColumns extends RowsAndColumns +{ + /** + * Mutates the RowsAndColumns by appending the requested Column. + * + * @param name the name of the new column + * @param column the Column object representing the new column + */ + void addColumn(String name, Column column); +} diff --git a/processing/src/main/java/org/apache/druid/query/rowsandcols/ArrayListRowsAndColumns.java b/processing/src/main/java/org/apache/druid/query/rowsandcols/ArrayListRowsAndColumns.java new file mode 100644 index 000000000000..3162cdfadbfb --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/rowsandcols/ArrayListRowsAndColumns.java @@ -0,0 +1,125 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.rowsandcols; + +import org.apache.druid.query.rowsandcols.column.Column; +import org.apache.druid.query.rowsandcols.column.ColumnAccessor; +import org.apache.druid.query.rowsandcols.column.ObjectColumnAccessorBase; +import org.apache.druid.segment.RowAdapter; +import org.apache.druid.segment.column.ColumnType; +import org.apache.druid.segment.column.RowSignature; + +import javax.annotation.Nullable; +import java.util.ArrayList; +import java.util.Collection; +import java.util.Comparator; +import java.util.Optional; +import java.util.function.Function; + +public class ArrayListRowsAndColumns implements RowsAndColumns +{ + private final ArrayList rows; + private final RowAdapter rowAdapter; + private final RowSignature rowSignature; + + public ArrayListRowsAndColumns( + ArrayList rows, + RowAdapter rowAdapter, + RowSignature rowSignature + ) + { + this.rows = rows; + this.rowAdapter = rowAdapter; + this.rowSignature = rowSignature; + } + + @Override + public Collection getColumnNames() + { + return rowSignature.getColumnNames(); + } + + @Override + public int numRows() + { + return rows.size(); + } + + @Override + @Nullable + public Column findColumn(String name) + { + if (!rowSignature.contains(name)) { + return null; + } + + final Function adapterForValue = rowAdapter.columnFunction(name); + final Optional maybeColumnType = rowSignature.getColumnType(name); + final ColumnType columnType = maybeColumnType.orElse(ColumnType.UNKNOWN_COMPLEX); + final Comparator comparator = Comparator.nullsFirst(columnType.getStrategy()); + + return new Column() + { + @Override + public ColumnAccessor toAccessor() + { + return new ObjectColumnAccessorBase() + { + @Override + protected Object getVal(int cell) + { + return adapterForValue.apply(rows.get(cell)); + } + + @Override + protected Comparator getComparator() + { + return comparator; + } + + @Override + public ColumnType getType() + { + return columnType; + } + + @Override + public int numRows() + { + return rows.size(); + } + }; + } + + @Override + public T as(Class clazz) + { + return null; + } + }; + } + + @Nullable + @Override + public T as(Class clazz) + { + return null; + } +} diff --git a/processing/src/main/java/org/apache/druid/query/rowsandcols/DefaultGroupPartitioner.java b/processing/src/main/java/org/apache/druid/query/rowsandcols/DefaultGroupPartitioner.java new file mode 100644 index 000000000000..96e559e358b2 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/rowsandcols/DefaultGroupPartitioner.java @@ -0,0 +1,75 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.rowsandcols; + +import org.apache.druid.java.util.common.ISE; +import org.apache.druid.query.rowsandcols.column.Column; +import org.apache.druid.query.rowsandcols.column.ColumnAccessor; + +import java.util.List; + +@SuppressWarnings("unused") +public class DefaultGroupPartitioner implements GroupPartitioner +{ + private final RowsAndColumns rac; + + public DefaultGroupPartitioner( + RowsAndColumns rac + ) + { + this.rac = rac; + } + + @Override + public int[] computeGroupings(List columns) + { + int[] retVal = new int[rac.numRows()]; + + for (String column : columns) { + final Column theCol = rac.findColumn(column); + if (theCol == null) { + // The column doesn't exist. In this case, we assume it's always the same value: null. If it's always + // the same, then it doesn't impact grouping at all and can be entirely skipped. + continue; + } + final ColumnAccessor accessor = theCol.toAccessor(); + + int currGroup = 0; + int prevGroupVal = 0; + for (int i = 1; i < retVal.length; ++i) { + if (retVal[i] == prevGroupVal) { + int comparison = accessor.compareCells(i - 1, i); + if (comparison == 0) { + retVal[i] = currGroup; + continue; + } else if (comparison > 0) { // "greater than" + throw new ISE("Pre-sorted data required, rows[%s] and [%s] were not in order", i - 1, i); + } // the 3rd condition ("less than") means create a new group, so let it fall through + } + + // We have a new group, so walk things forward. + prevGroupVal = retVal[i]; + retVal[i] = ++currGroup; + } + } + + return retVal; + } +} diff --git a/processing/src/main/java/org/apache/druid/query/rowsandcols/DefaultOnHeapAggregatable.java b/processing/src/main/java/org/apache/druid/query/rowsandcols/DefaultOnHeapAggregatable.java new file mode 100644 index 000000000000..83aa50aee8de --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/rowsandcols/DefaultOnHeapAggregatable.java @@ -0,0 +1,271 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.rowsandcols; + +import org.apache.druid.java.util.common.ISE; +import org.apache.druid.query.aggregation.Aggregator; +import org.apache.druid.query.aggregation.AggregatorFactory; +import org.apache.druid.query.dimension.DimensionSpec; +import org.apache.druid.query.monomorphicprocessing.RuntimeShapeInspector; +import org.apache.druid.query.rowsandcols.column.Column; +import org.apache.druid.query.rowsandcols.column.ColumnAccessor; +import org.apache.druid.segment.BaseSingleValueDimensionSelector; +import org.apache.druid.segment.ColumnSelectorFactory; +import org.apache.druid.segment.ColumnValueSelector; +import org.apache.druid.segment.DimensionSelector; +import org.apache.druid.segment.column.ColumnCapabilities; +import org.apache.druid.segment.column.ColumnCapabilitiesImpl; +import org.apache.druid.segment.column.ColumnType; +import org.apache.druid.segment.serde.ComplexMetricSerde; +import org.apache.druid.segment.serde.ComplexMetrics; + +import javax.annotation.Nonnull; +import javax.annotation.Nullable; +import java.util.ArrayList; +import java.util.HashMap; +import java.util.List; +import java.util.Map; +import java.util.concurrent.atomic.AtomicInteger; +import java.util.concurrent.atomic.AtomicReference; +import java.util.function.Function; + +public class DefaultOnHeapAggregatable implements OnHeapAggregatable, OnHeapCumulativeAggregatable +{ + private final RowsAndColumns rac; + + public DefaultOnHeapAggregatable( + RowsAndColumns rac + ) + { + this.rac = rac; + } + + @Override + public ArrayList aggregateAll( + List aggFactories + ) + { + Aggregator[] aggs = new Aggregator[aggFactories.size()]; + + AtomicInteger currRow = new AtomicInteger(0); + int index = 0; + for (AggregatorFactory aggFactory : aggFactories) { + aggs[index++] = aggFactory.factorize(new ColumnAccessorBasedColumnSelectorFactory(currRow)); + } + + int numRows = rac.numRows(); + int rowId = currRow.get(); + while (rowId < numRows) { + for (Aggregator agg : aggs) { + agg.aggregate(); + } + rowId = currRow.incrementAndGet(); + } + + ArrayList retVal = new ArrayList<>(aggs.length); + for (Aggregator agg : aggs) { + retVal.add(agg.get()); + } + return retVal; + } + + @Override + public ArrayList aggregateCumulative(List aggFactories) + { + Aggregator[] aggs = new Aggregator[aggFactories.size()]; + ArrayList retVal = new ArrayList<>(aggFactories.size()); + + int numRows = rac.numRows(); + AtomicInteger currRow = new AtomicInteger(0); + int index = 0; + for (AggregatorFactory aggFactory : aggFactories) { + aggs[index++] = aggFactory.factorize(new ColumnAccessorBasedColumnSelectorFactory(currRow)); + retVal.add(new Object[numRows]); + } + + int rowId = currRow.get(); + while (rowId < numRows) { + for (int i = 0; i < aggs.length; ++i) { + aggs[i].aggregate(); + retVal.get(i)[rowId] = aggs[i].get(); + } + rowId = currRow.incrementAndGet(); + } + + return retVal; + } + + private class ColumnAccessorBasedColumnSelectorFactory implements ColumnSelectorFactory + { + private final Map accessorCache = new HashMap<>(); + + private final AtomicInteger cellIdSupplier; + + public ColumnAccessorBasedColumnSelectorFactory(AtomicInteger cellIdSupplier) + { + this.cellIdSupplier = cellIdSupplier; + } + + @Override + public DimensionSelector makeDimensionSelector(DimensionSpec dimensionSpec) + { + return withColumnAccessor(dimensionSpec.getDimension(), columnAccessor -> { + if (columnAccessor == null) { + return DimensionSelector.constant(null); + } else { + return new BaseSingleValueDimensionSelector() + { + @Nullable + @Override + protected String getValue() + { + return String.valueOf(columnAccessor.getObject(cellIdSupplier.get())); + } + + @Override + public void inspectRuntimeShape(RuntimeShapeInspector inspector) + { + + } + }; + } + }); + } + + @SuppressWarnings("rawtypes") + @Override + public ColumnValueSelector makeColumnValueSelector(@Nonnull String columnName) + { + return withColumnAccessor(columnName, columnAccessor -> { + if (columnAccessor == null) { + return DimensionSelector.constant(null); + } else { + return new ColumnValueSelector() + { + private final AtomicReference myClazz = new AtomicReference<>(null); + + @Nullable + @Override + public Object getObject() + { + return columnAccessor.getObject(cellIdSupplier.get()); + } + + @SuppressWarnings("rawtypes") + @Override + public Class classOfObject() + { + Class retVal = myClazz.get(); + if (retVal == null) { + retVal = findClazz(); + myClazz.set(retVal); + } + return retVal; + } + + private Class findClazz() + { + final ColumnType type = columnAccessor.getType(); + switch (type.getType()) { + case LONG: + return long.class; + case DOUBLE: + return double.class; + case FLOAT: + return float.class; + case STRING: + return String.class; + case ARRAY: + return List.class; + case COMPLEX: + final ComplexMetricSerde serdeForType = ComplexMetrics.getSerdeForType(type.getComplexTypeName()); + if (serdeForType != null && serdeForType.getObjectStrategy() != null) { + return serdeForType.getObjectStrategy().getClazz(); + } + + for (int i = 0; i < columnAccessor.numRows(); ++i) { + Object obj = columnAccessor.getObject(i); + if (obj != null) { + return obj.getClass(); + } + } + return Object.class; + default: + throw new ISE("Unknown type[%s]", type.getType()); + } + } + + @Override + public boolean isNull() + { + return columnAccessor.isNull(cellIdSupplier.get()); + } + + @Override + public long getLong() + { + return columnAccessor.getLong(cellIdSupplier.get()); + } + + @Override + public float getFloat() + { + return columnAccessor.getFloat(cellIdSupplier.get()); + } + + @Override + public double getDouble() + { + return columnAccessor.getDouble(cellIdSupplier.get()); + } + + @Override + public void inspectRuntimeShape(RuntimeShapeInspector inspector) + { + + } + }; + } + }); + } + + @Nullable + @Override + public ColumnCapabilities getColumnCapabilities(String column) + { + return withColumnAccessor(column, columnAccessor -> + new ColumnCapabilitiesImpl() + .setType(columnAccessor.getType()) + .setDictionaryEncoded(false) + .setHasBitmapIndexes(false)); + } + + private T withColumnAccessor(String column, Function fn) + { + ColumnAccessor retVal = accessorCache.get(column); + if (retVal == null) { + Column racColumn = rac.findColumn(column); + retVal = racColumn == null ? null : racColumn.toAccessor(); + accessorCache.put(column, retVal); + } + return fn.apply(retVal); + } + } +} diff --git a/processing/src/main/java/org/apache/druid/query/rowsandcols/DefaultSortedGroupPartitioner.java b/processing/src/main/java/org/apache/druid/query/rowsandcols/DefaultSortedGroupPartitioner.java new file mode 100644 index 000000000000..1470ac2da06d --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/rowsandcols/DefaultSortedGroupPartitioner.java @@ -0,0 +1,93 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.rowsandcols; + +import it.unimi.dsi.fastutil.ints.IntArrayList; +import it.unimi.dsi.fastutil.ints.IntList; +import org.apache.druid.java.util.common.ISE; +import org.apache.druid.query.operator.LimitedRowsAndColumns; +import org.apache.druid.query.rowsandcols.column.Column; +import org.apache.druid.query.rowsandcols.column.ColumnAccessor; + +import java.util.ArrayList; +import java.util.List; + +public class DefaultSortedGroupPartitioner implements SortedGroupPartitioner +{ + private final RowsAndColumns rac; + + public DefaultSortedGroupPartitioner( + RowsAndColumns rac + ) + { + this.rac = rac; + } + + @Override + public int[] computeBoundaries(List columns) + { + // Initialize to a grouping of everything + IntList boundaries = new IntArrayList(new int[]{0, rac.numRows()}); + + for (String column : columns) { + final Column theCol = rac.findColumn(column); + if (theCol == null) { + // The column doesn't exist. In this case, we assume it's always the same value: null. If it's always + // the same, then it doesn't impact grouping at all and can be entirely skipped. + continue; + } + final ColumnAccessor accessor = theCol.toAccessor(); + + IntList newBoundaries = new IntArrayList(); + newBoundaries.add(0); + for (int i = 1; i < boundaries.size(); ++i) { + int start = boundaries.getInt(i - 1); + int end = boundaries.getInt(i); + for (int j = start + 1; j < end; ++j) { + int comparison = accessor.compareCells(j - 1, j); + if (comparison < 0) { + newBoundaries.add(j); + } else if (comparison > 0) { + throw new ISE("Pre-sorted data required, rows[%s] and [%s] were not in order", j - 1, j); + } + } + newBoundaries.add(end); + } + boundaries = newBoundaries; + } + + return boundaries.toIntArray(); + } + + @Override + public ArrayList partitionOnBoundaries(List partitionColumns) + { + final int[] boundaries = computeBoundaries(partitionColumns); + ArrayList retVal = new ArrayList<>(boundaries.length - 1); + + for (int i = 1; i < boundaries.length; ++i) { + int start = boundaries[i - 1]; + int end = boundaries[i]; + retVal.add(new LimitedRowsAndColumns(rac, start, end)); + } + + return retVal; + } +} diff --git a/processing/src/main/java/org/apache/druid/query/rowsandcols/GroupPartitioner.java b/processing/src/main/java/org/apache/druid/query/rowsandcols/GroupPartitioner.java new file mode 100644 index 000000000000..e3bae9d4284e --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/rowsandcols/GroupPartitioner.java @@ -0,0 +1,43 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.rowsandcols; + +import java.util.List; + +/** + * A semantic interface used to partition a data set based on a given set of dimensions. + */ +@SuppressWarnings("unused") +public interface GroupPartitioner +{ + /** + * Computes the groupings of the underlying rows based on the columns passed in for grouping. The grouping is + * returned as an int[], the length of the array will be equal to the number of rows of data and the values of + * the elements of the array will be the same when the rows are part of the same group and different when the + * rows are part of different groups. This is contrasted with the SortedGroupPartitioner in that, the + * groupings returned are not necessarily contiguous. There is also no sort-order implied by the `int` values + * assigned to each grouping. + * + * @param columns the columns to group with + * @return the groupings, rows with the same int value are in the same group. There is no sort-order implied by the + * int values. + */ + int[] computeGroupings(List columns); +} diff --git a/processing/src/main/java/org/apache/druid/query/rowsandcols/OnHeapAggregatable.java b/processing/src/main/java/org/apache/druid/query/rowsandcols/OnHeapAggregatable.java new file mode 100644 index 000000000000..9a707737a183 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/rowsandcols/OnHeapAggregatable.java @@ -0,0 +1,47 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.rowsandcols; + +import org.apache.druid.query.aggregation.AggregatorFactory; + +import java.util.ArrayList; +import java.util.List; + +/** + * A semantic interface used to aggregate a list of AggregatorFactories across a given set of data + *

    + * The aggregation specifically happens on-heap and should be used in places where it is known that the data + * set can be worked with entirely on-heap. + *

    + * Note, as we implement frame-handling for window aggregations, it is expected that this interface will undergo a + * transformation. It might be deleted and replaced with something else, or might just see a change done in place. + * Either way, there is no assumption of enforced compatibility with this interface at this point in time. + */ +public interface OnHeapAggregatable +{ + /** + * Aggregates the data using the {@code List aggregateAll(List aggFactories); +} diff --git a/processing/src/main/java/org/apache/druid/query/rowsandcols/OnHeapCumulativeAggregatable.java b/processing/src/main/java/org/apache/druid/query/rowsandcols/OnHeapCumulativeAggregatable.java new file mode 100644 index 000000000000..a931c3dbcab1 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/rowsandcols/OnHeapCumulativeAggregatable.java @@ -0,0 +1,48 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.rowsandcols; + +import org.apache.druid.query.aggregation.AggregatorFactory; + +import java.util.ArrayList; +import java.util.List; + +/** + * A semantic interface used to cumulatively aggregate a list of AggregatorFactories across a given set of data + *

    + * The aggregation specifically happens on-heap and should be used in places where it is known that the data + * set can be worked with entirely on-heap. + *

    + * Note, as we implement frame-handling for window aggregations, it is expected that this interface will undergo a + * transformation. It might be deleted and replaced with something else, or might just see a change done in place. + * Either way, there is no assumption of enforced compatibility with this interface at this point in time. + */ +public interface OnHeapCumulativeAggregatable +{ + /** + * Cumulatively aggregates the data using the {@code List aggregateCumulative(List aggFactories); +} diff --git a/processing/src/main/java/org/apache/druid/query/rowsandcols/RowsAndColumns.java b/processing/src/main/java/org/apache/druid/query/rowsandcols/RowsAndColumns.java new file mode 100644 index 000000000000..c42f5c0a9267 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/rowsandcols/RowsAndColumns.java @@ -0,0 +1,113 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.rowsandcols; + +import org.apache.druid.query.rowsandcols.column.Column; +import org.apache.druid.query.rowsandcols.frame.AppendableMapOfColumns; + +import javax.annotation.Nonnull; +import javax.annotation.Nullable; +import java.util.Collection; + +/** + * An interface representing a chunk of RowsAndColumns. Essentially a RowsAndColumns is just a batch of rows + * with columns. + *

    + * This interface has very little prescriptively defined about what *must* be implemented. This is intentional + * as there are lots of different possible representations of batch of rows each with their own unique positives + * and negatives when it comes to processing. So, any explicit definition of what a RowsAndColumns is will actually, + * by definition, end up as optimal for one specific configuration and sub-optimal for others. Instead of trying to + * explicitly expand the interface to cover all the different possible ways that someone could want to interace + * with a Rows and columns, we rely on semantic interfaces using the {@link RowsAndColumns#as} method instead. + *

    + * That is, the expectation is that anything that works with a RowsAndColumns will tend to first ask the RowsAndColumns + * object to become some other interface, for example, an {@link OnHeapAggregatable}. If a RowsAndColumns knows how + * to do a good job as the requested interface, it can return its own concrete implementation of the interface and + * run the necessary logic in its own optimized fashion. If the RowsAndColumns instance does not know how to implement + * the semantic interface, it is expected that a default implementation of the interface can be instantiated on top of + * the default column access mechanisms that the RowsAndColumns provides. Such default implementations should be + * functionally correct, but are not believed to be optimal. + *

    + * The "default column access mechanisms" here amount to using {@link #findColumn} to load a Column + * and then using {@link Column#toAccessor} to access the individual cells of the column. There is also a + * {@link Column#as} method which a default implementation might attempt to use to create a more optimal runtime. + *

    + * It is intended that this interface can be used by Frames, Segments and even normal on-heap JVM data structures to + * participate in query operations. + */ +public interface RowsAndColumns +{ + @Nonnull + static AppendableRowsAndColumns expectAppendable(RowsAndColumns input) + { + if (input instanceof AppendableRowsAndColumns) { + return (AppendableRowsAndColumns) input; + } + + AppendableRowsAndColumns retVal = input.as(AppendableRowsAndColumns.class); + if (retVal == null) { + retVal = new AppendableMapOfColumns(input); + } + return retVal; + } + + /** + * The set of column names available from the RowsAndColumns + * + * @return The set of column names available from the RowsAndColumns + */ + @SuppressWarnings("unreachable") + Collection getColumnNames(); + + /** + * The number of rows in the RowsAndColumns object + * + * @return the integer number of rows + */ + int numRows(); + + /** + * Finds a column by name. null is returned if the column is not found. The RowsAndColumns object should not + * attempt to default not-found columns to pretend as if they exist, instead the user of the RowsAndColumns object + * should decide the correct semantic interpretation of a column that does not exist. It is expected that most + * locations will choose to believe that the column does exist and is always null, but there are often optimizations + * that can effect this same assumption without doing a lot of extra work if the calling code knows that it does not + * exist. + * + * @param name the name of the column to find + * @return the Column, if found. null if not found. + */ + Column findColumn(String name); + + /** + * Asks the RowsAndColumns to return itself as a concrete implementation of a specific interface. The interface + * asked for will tend to be a semantically-meaningful interface. This method allows the calling code to interrogate + * the RowsAndColumns object about whether it can offer a meaningful optimization of the semantic interface. If a + * RowsAndColumns cannot do anything specifically optimal for the interface requested, it should return null instead + * of trying to come up with its own default implementation. + * + * @param clazz A class object representing the interface that the calling code wants a concrete implementation of + * @param The interface that the calling code wants a concrete implementation of + * @return A concrete implementation of the interface, or null if there is no meaningful optimization to be had + * through a local implementation of the interface. + */ + @Nullable + T as(Class clazz); +} diff --git a/processing/src/main/java/org/apache/druid/query/rowsandcols/SortedGroupPartitioner.java b/processing/src/main/java/org/apache/druid/query/rowsandcols/SortedGroupPartitioner.java new file mode 100644 index 000000000000..a60657937aef --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/rowsandcols/SortedGroupPartitioner.java @@ -0,0 +1,54 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.rowsandcols; + +import java.util.ArrayList; +import java.util.List; + +/** + * A semantic interface used to partition a data set based on a given set of dimensions. + *

    + * This specifically assumes that it is working with sorted data and, as such, the groups returned + * should be contiguous and unique (that is, all rows for a given combination of values exist in only one grouping) + */ +public interface SortedGroupPartitioner +{ + /** + * Computes and returns a list of contiguous boundaries for independent groups. All rows in a specific grouping + * should have the same values for the identified columns. Additionally, as this is assuming it is dealing with + * sorted data, there should only be a single entry in the return value for a given set of values of the columns. + * + * @param columns the columns to partition on + * @return an int[] representing the start (inclusive) and stop (exclusive) offsets of boundaries. Boundaries are + * contiguous, so the stop of the previous boundary is the start of the subsequent one. + */ + int[] computeBoundaries(List columns); + + /** + * Semantically equivalent to computeBoundaries, but returns a list of RowsAndColumns objects instead of just + * boundary positions. This is useful as it allows the concrete implementation to return RowsAndColumns objects + * that are aware of the internal representation of the data and thus can provide optimized implementations of + * other semantic interfaces as the "child" RowsAndColumns are used + * + * @param partitionColumns the columns to partition on + * @return a list of RowsAndColumns representing the data grouped by the partition columns. + */ + ArrayList partitionOnBoundaries(List partitionColumns); +} diff --git a/processing/src/main/java/org/apache/druid/query/rowsandcols/column/Column.java b/processing/src/main/java/org/apache/druid/query/rowsandcols/column/Column.java new file mode 100644 index 000000000000..97327faf9ce6 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/rowsandcols/column/Column.java @@ -0,0 +1,63 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.rowsandcols.column; + +import javax.annotation.Nonnull; + +/** + * An interface representing a Column of data. + * + * This interface prescribes that a {@link ColumnAccessor} must be defined on the column, but also offers an + * {@link #as} method to allow for optimized specific implementations of semantically meaningful logic. + * + * That is, the expectation is that some things work with Column objects might choose to first ask the Column + * object to become some other interface. If the Column knows how to do a good job as the requested interface, it can + * return its own concrete implementation of the interface and run the necessary logic in its own optimized fashion. + * If the Column instance does not know how to implement the semantic interface, it is expected that the + * {@link ColumnAccessor} will be leveraged to implement whatever logic is required. + */ +public interface Column +{ + /** + * Returns the column as a {@link ColumnAccessor}. Semantically, this would be equivalent to calling + * {@Code Column.as(ColumnAccessor.class)}. However, being able to implement this interface is part of the explicit + * contract of implementing this interface, so instead of relying on {@link #as} which allows for returning null, + * we define a top-level method that should never return null. + * + * @return a {@link ColumnAccessor} representation of the column, this should never return null. + */ + @Nonnull + ColumnAccessor toAccessor(); + + /** + * Asks the Column to return itself as a concrete implementation of a specific interface. The interface + * asked for will tend to be a semantically-meaningful interface. This method allows the calling code to interrogate + * the Column object about whether it can offer a meaningful optimization of the semantic interface. If a + * Column cannot do anything specifically optimal for the interface requested, it should return null instead + * of trying to come up with its own default implementation. + * + * @param clazz A class object representing the interface that the calling code wants a concrete implementation of + * @param The interface that the calling code wants a concrete implementation of + * @return A concrete implementation of the interface, or null if there is no meaningful optimization to be had + * through a local implementation of the interface. + */ + @SuppressWarnings("unused") + T as(Class clazz); +} diff --git a/processing/src/main/java/org/apache/druid/query/rowsandcols/column/ColumnAccessor.java b/processing/src/main/java/org/apache/druid/query/rowsandcols/column/ColumnAccessor.java new file mode 100644 index 000000000000..acc6b5806d2f --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/rowsandcols/column/ColumnAccessor.java @@ -0,0 +1,104 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.rowsandcols.column; + +import org.apache.druid.segment.column.ColumnType; + +import javax.annotation.Nullable; + +/** + * Allows for accessing a column, provides methods to enable cell-by-cell access. + */ +public interface ColumnAccessor +{ + /** + * Get the type of the Column + * + * @return the type of the Column + */ + ColumnType getType(); + + /** + * Get the number of cells + * + * @return the number of cells + */ + int numRows(); + + /** + * Get whether the value of a cell is null + * + * @param rowNum the cell id, 0-indexed + * @return true if the value is null + */ + boolean isNull(int rowNum); + + /** + * Get the {@link Object} representation of the cell. + * + * @param rowNum the cell id, 0-indexed + * @return the {@link Object} representation of the cell. Returns {@code null} If {@link #isNull} is true. + */ + @Nullable + Object getObject(int rowNum); + + /** + * Get the primitive {@code double} representation of the cell. + * + * @param rowNum the cell id, 0-indexed + * @return the primitive {@code double} representation of the cell. Returns {@code 0D} If {@link #isNull} is true. + */ + double getDouble(int rowNum); + + /** + * Get the primitive {@code float} representation of the cell. + * + * @param rowNum the cell id, 0-indexed + * @return the primitive {@code float} representation of the cell. Returns {@code 0F} If {@link #isNull} is true. + */ + float getFloat(int rowNum); + + /** + * Get the primitive {@code long} representation of the cell. + * + * @param rowNum the cell id, 0-indexed + * @return the primitive {@code long} representation of the cell. Returns {@code 0L} If {@link #isNull} is true. + */ + long getLong(int rowNum); + + /** + * Get the primitive {@code int} representation of the cell. + * + * @param rowNum the cell id, 0-indexed + * @return the primitive {@code int} representation of the cell. Returns {@code 0} If {@link #isNull} is true. + */ + int getInt(int rowNum); + + /** + * Compares two cells using a comparison that follows the same semantics as {@link java.util.Comparator#compare} + *

    + * This is not comparing the cell Ids, but the values referred to by the cell ids. + * + * @param lhsRowNum the cell id of the left-hand-side of the comparison + * @param rhsRowNum the cell id of the right-hand-side of the comparison + * @return the result of the comparison of the two cells + */ + int compareCells(int lhsRowNum, int rhsRowNum); +} diff --git a/processing/src/main/java/org/apache/druid/query/rowsandcols/column/ColumnAccessorBasedColumn.java b/processing/src/main/java/org/apache/druid/query/rowsandcols/column/ColumnAccessorBasedColumn.java new file mode 100644 index 000000000000..2e6a4a597cef --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/rowsandcols/column/ColumnAccessorBasedColumn.java @@ -0,0 +1,44 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.rowsandcols.column; + +public class ColumnAccessorBasedColumn implements Column +{ + private final ColumnAccessor base; + + public ColumnAccessorBasedColumn( + ColumnAccessor base + ) + { + this.base = base; + } + + @Override + public ColumnAccessor toAccessor() + { + return base; + } + + @Override + public T as(Class clazz) + { + return null; + } +} diff --git a/processing/src/main/java/org/apache/druid/query/rowsandcols/column/ConstantObjectColumn.java b/processing/src/main/java/org/apache/druid/query/rowsandcols/column/ConstantObjectColumn.java new file mode 100644 index 000000000000..09aad34692dd --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/rowsandcols/column/ConstantObjectColumn.java @@ -0,0 +1,103 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.rowsandcols.column; + +import org.apache.druid.segment.column.ColumnType; + +public class ConstantObjectColumn implements Column +{ + private final Object obj; + private final int numCells; + private final ColumnType type; + + public ConstantObjectColumn(Object obj, int numCells, ColumnType type) + { + this.obj = obj; + this.numCells = numCells; + this.type = type; + } + + @Override + public ColumnAccessor toAccessor() + { + return new ColumnAccessor() + { + @Override + public ColumnType getType() + { + return type; + } + + @Override + public int numRows() + { + return numCells; + } + + @Override + public boolean isNull(int rowNum) + { + return obj == null; + } + + @Override + public Object getObject(int rowNum) + { + return obj; + } + + @Override + public double getDouble(int rowNum) + { + return ((Number) obj).doubleValue(); + } + + @Override + public float getFloat(int rowNum) + { + return ((Number) obj).floatValue(); + } + + @Override + public long getLong(int rowNum) + { + return ((Number) obj).longValue(); + } + + @Override + public int getInt(int rowNum) + { + return ((Number) obj).intValue(); + } + + @Override + public int compareCells(int lhsRowNum, int rhsRowNum) + { + return 0; + } + }; + } + + @Override + public T as(Class clazz) + { + return null; + } +} diff --git a/processing/src/main/java/org/apache/druid/query/rowsandcols/column/DoubleArrayColumn.java b/processing/src/main/java/org/apache/druid/query/rowsandcols/column/DoubleArrayColumn.java new file mode 100644 index 000000000000..c2d7547be129 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/rowsandcols/column/DoubleArrayColumn.java @@ -0,0 +1,101 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.rowsandcols.column; + +import org.apache.druid.segment.column.ColumnType; + +public class DoubleArrayColumn implements Column +{ + private final double[] vals; + + public DoubleArrayColumn( + double[] vals + ) + { + this.vals = vals; + } + + @Override + public ColumnAccessor toAccessor() + { + return new ColumnAccessor() + { + @Override + public ColumnType getType() + { + return ColumnType.DOUBLE; + } + + @Override + public int numRows() + { + return vals.length; + } + + @Override + public boolean isNull(int rowNum) + { + return false; + } + + @Override + public Object getObject(int rowNum) + { + return vals[rowNum]; + } + + @Override + public double getDouble(int rowNum) + { + return vals[rowNum]; + } + + @Override + public float getFloat(int rowNum) + { + return (float) vals[rowNum]; + } + + @Override + public long getLong(int rowNum) + { + return (long) vals[rowNum]; + } + + @Override + public int getInt(int rowNum) + { + return (int) vals[rowNum]; + } + + @Override + public int compareCells(int lhsRowNum, int rhsRowNum) + { + return Double.compare(lhsRowNum, rhsRowNum); + } + }; + } + + @Override + public T as(Class clazz) + { + return null; + } +} diff --git a/processing/src/main/java/org/apache/druid/query/rowsandcols/column/IntArrayColumn.java b/processing/src/main/java/org/apache/druid/query/rowsandcols/column/IntArrayColumn.java new file mode 100644 index 000000000000..10f70351efaa --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/rowsandcols/column/IntArrayColumn.java @@ -0,0 +1,101 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.rowsandcols.column; + +import org.apache.druid.segment.column.ColumnType; + +public class IntArrayColumn implements Column +{ + private final int[] vals; + + public IntArrayColumn( + int[] vals + ) + { + this.vals = vals; + } + + @Override + public ColumnAccessor toAccessor() + { + return new ColumnAccessor() + { + @Override + public ColumnType getType() + { + return ColumnType.LONG; + } + + @Override + public int numRows() + { + return vals.length; + } + + @Override + public boolean isNull(int rowNum) + { + return false; + } + + @Override + public Object getObject(int rowNum) + { + return vals[rowNum]; + } + + @Override + public double getDouble(int rowNum) + { + return vals[rowNum]; + } + + @Override + public float getFloat(int rowNum) + { + return vals[rowNum]; + } + + @Override + public long getLong(int rowNum) + { + return vals[rowNum]; + } + + @Override + public int getInt(int rowNum) + { + return vals[rowNum]; + } + + @Override + public int compareCells(int lhsRowNum, int rhsRowNum) + { + return Integer.compare(vals[lhsRowNum], vals[rhsRowNum]); + } + }; + } + + @Override + public T as(Class clazz) + { + return null; + } +} diff --git a/processing/src/main/java/org/apache/druid/query/rowsandcols/column/NullColumnAccessor.java b/processing/src/main/java/org/apache/druid/query/rowsandcols/column/NullColumnAccessor.java new file mode 100644 index 000000000000..ea876c3e1478 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/rowsandcols/column/NullColumnAccessor.java @@ -0,0 +1,96 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.rowsandcols.column; + +import org.apache.druid.segment.column.ColumnType; + +import javax.annotation.Nullable; + +public class NullColumnAccessor implements ColumnAccessor +{ + private final ColumnType type; + private final int size; + + public NullColumnAccessor(int size) + { + this(ColumnType.UNKNOWN_COMPLEX, size); + } + + public NullColumnAccessor(ColumnType type, int size) + { + this.type = type; + this.size = size; + } + + @Override + public ColumnType getType() + { + return type; + } + + @Override + public int numRows() + { + return size; + } + + @Override + public boolean isNull(int rowNum) + { + return true; + } + + @Nullable + @Override + public Object getObject(int rowNum) + { + return null; + } + + @Override + public double getDouble(int rowNum) + { + return 0; + } + + @Override + public float getFloat(int rowNum) + { + return 0; + } + + @Override + public long getLong(int rowNum) + { + return 0; + } + + @Override + public int getInt(int rowNum) + { + return 0; + } + + @Override + public int compareCells(int lhsRowNum, int rhsRowNum) + { + return 0; + } +} diff --git a/processing/src/main/java/org/apache/druid/query/rowsandcols/column/ObjectArrayColumn.java b/processing/src/main/java/org/apache/druid/query/rowsandcols/column/ObjectArrayColumn.java new file mode 100644 index 000000000000..fd850a0956d0 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/rowsandcols/column/ObjectArrayColumn.java @@ -0,0 +1,81 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.rowsandcols.column; + +import org.apache.druid.segment.column.ColumnType; + +import java.util.Comparator; + +public class ObjectArrayColumn implements Column +{ + private final Object[] objects; + private final ColumnType resultType; + private final Comparator comparator; + + public ObjectArrayColumn(Object[] objects, ColumnType resultType) + { + this(objects, resultType, Comparator.nullsFirst(resultType.getStrategy())); + } + + public ObjectArrayColumn(Object[] objects, ColumnType resultType, Comparator comparator) + { + this.objects = objects; + this.resultType = resultType; + this.comparator = comparator; + } + + @Override + public ColumnAccessor toAccessor() + { + return new ObjectColumnAccessorBase() + { + @Override + protected Object getVal(int cell) + { + return objects[cell]; + } + + @Override + protected Comparator getComparator() + { + return comparator; + } + + @Override + public ColumnType getType() + { + return resultType; + } + + @Override + public int numRows() + { + return objects.length; + } + }; + } + + @Override + public T as(Class clazz) + { + return null; + } + +} diff --git a/processing/src/main/java/org/apache/druid/query/rowsandcols/column/ObjectColumnAccessorBase.java b/processing/src/main/java/org/apache/druid/query/rowsandcols/column/ObjectColumnAccessorBase.java new file mode 100644 index 000000000000..8ef2aeae1196 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/rowsandcols/column/ObjectColumnAccessorBase.java @@ -0,0 +1,121 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.rowsandcols.column; + +import javax.annotation.Nullable; +import java.util.Comparator; + +public abstract class ObjectColumnAccessorBase implements ColumnAccessor +{ + @Override + public boolean isNull(int rowNum) + { + return getVal(rowNum) == null; + } + + @Nullable + @Override + public Object getObject(int rowNum) + { + return getVal(rowNum); + } + + @Override + public double getDouble(int rowNum) + { + final Object val = getVal(rowNum); + if (val instanceof Number) { + return ((Number) val).doubleValue(); + } else if (val instanceof String) { + try { + return Double.parseDouble((String) val); + } + catch (NumberFormatException e) { + return 0d; + } + } else { + return 0d; + } + } + + @Override + public float getFloat(int rowNum) + { + final Object val = getVal(rowNum); + if (val instanceof Number) { + return ((Number) val).floatValue(); + } else if (val instanceof String) { + try { + return Float.parseFloat((String) val); + } + catch (NumberFormatException e) { + return 0f; + } + } else { + return 0f; + } + } + + @Override + public long getLong(int rowNum) + { + final Object val = getVal(rowNum); + if (val instanceof Number) { + return ((Number) val).longValue(); + } else if (val instanceof String) { + try { + return Long.parseLong((String) val); + } + catch (NumberFormatException e) { + return 0L; + } + } else { + return 0L; + } + } + + @Override + public int getInt(int rowNum) + { + final Object val = getVal(rowNum); + if (val instanceof Number) { + return ((Number) val).intValue(); + } else if (val instanceof String) { + try { + return Integer.parseInt((String) val); + } + catch (NumberFormatException e) { + return 0; + } + } else { + return 0; + } + } + + @Override + public int compareCells(int lhsRowNum, int rhsRowNum) + { + return getComparator().compare(getVal(lhsRowNum), getVal(rhsRowNum)); + } + + protected abstract Object getVal(int cell); + + protected abstract Comparator getComparator(); +} diff --git a/processing/src/main/java/org/apache/druid/query/rowsandcols/frame/AppendableMapOfColumns.java b/processing/src/main/java/org/apache/druid/query/rowsandcols/frame/AppendableMapOfColumns.java new file mode 100644 index 000000000000..60a8d9fa5310 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/rowsandcols/frame/AppendableMapOfColumns.java @@ -0,0 +1,93 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.rowsandcols.frame; + +import org.apache.druid.java.util.common.ISE; +import org.apache.druid.query.rowsandcols.AppendableRowsAndColumns; +import org.apache.druid.query.rowsandcols.RowsAndColumns; +import org.apache.druid.query.rowsandcols.column.Column; + +import java.util.Collection; +import java.util.LinkedHashMap; +import java.util.LinkedHashSet; +import java.util.Set; + +public class AppendableMapOfColumns implements AppendableRowsAndColumns +{ + private final RowsAndColumns base; + private final LinkedHashMap appendedColumns; + private Set colNames = null; + + public AppendableMapOfColumns( + RowsAndColumns base + ) + { + this.base = base; + this.appendedColumns = new LinkedHashMap<>(); + } + + @Override + public void addColumn(String name, Column column) + { + final Column prevValue = appendedColumns.put(name, column); + if (prevValue != null) { + throw new ISE("Tried to override column[%s]!? Was[%s], now[%s]", name, prevValue, column); + } + if (colNames != null) { + colNames.add(name); + } + } + + @Override + public Collection getColumnNames() + { + if (colNames == null) { + this.colNames = new LinkedHashSet<>(base.getColumnNames()); + this.colNames.addAll(appendedColumns.keySet()); + } + return colNames; + } + + @Override + public int numRows() + { + return base.numRows(); + } + + @Override + public Column findColumn(String name) + { + Column retVal = base.findColumn(name); + if (retVal == null) { + retVal = appendedColumns.get(name); + } + return retVal; + } + + @Override + @SuppressWarnings("unchecked") + public T as(Class clazz) + { + if (AppendableRowsAndColumns.class.equals(clazz)) { + return (T) this; + } + return null; + } +} diff --git a/processing/src/main/java/org/apache/druid/query/rowsandcols/frame/MapOfColumnsRowsAndColumns.java b/processing/src/main/java/org/apache/druid/query/rowsandcols/frame/MapOfColumnsRowsAndColumns.java new file mode 100644 index 000000000000..aee614b51422 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/query/rowsandcols/frame/MapOfColumnsRowsAndColumns.java @@ -0,0 +1,109 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.rowsandcols.frame; + +import com.google.common.collect.ImmutableMap; +import org.apache.druid.java.util.common.ISE; +import org.apache.druid.query.rowsandcols.AppendableRowsAndColumns; +import org.apache.druid.query.rowsandcols.RowsAndColumns; +import org.apache.druid.query.rowsandcols.column.Column; + +import java.util.Iterator; +import java.util.Map; +import java.util.Set; + +public class MapOfColumnsRowsAndColumns implements RowsAndColumns +{ + public static MapOfColumnsRowsAndColumns of(String name, Column col) + { + return fromMap(ImmutableMap.of(name, col)); + } + + public static MapOfColumnsRowsAndColumns of(String name, Column col, String name2, Column col2) + { + return fromMap(ImmutableMap.of(name, col, name2, col2)); + } + + public static MapOfColumnsRowsAndColumns fromMap(Map map) + { + if (map == null || map.isEmpty()) { + throw new ISE("map[%s] cannot be null or empty.", map); + } + + final Iterator> iter = map.entrySet().iterator(); + Map.Entry entry = iter.next(); + int numCells = entry.getValue().toAccessor().numRows(); + if (iter.hasNext()) { + entry = iter.next(); + final int newCells = entry.getValue().toAccessor().numRows(); + if (numCells != newCells) { + throw new ISE( + "Mismatched numCells, expectedNumCells[%s], actual[%s] from col[%s].", + numCells, + newCells, + entry.getKey() + ); + } + } + + return new MapOfColumnsRowsAndColumns(map, map.values().iterator().next().toAccessor().numRows()); + } + + private final Map mapOfColumns; + private final int numRows; + + public MapOfColumnsRowsAndColumns( + Map mapOfColumns, + int numRows + ) + { + this.mapOfColumns = mapOfColumns; + this.numRows = numRows; + } + + @Override + public Set getColumnNames() + { + return mapOfColumns.keySet(); + } + + @Override + public int numRows() + { + return numRows; + } + + @Override + public Column findColumn(String name) + { + return mapOfColumns.get(name); + } + + @Override + @SuppressWarnings("unchecked") + public T as(Class clazz) + { + if (AppendableRowsAndColumns.class.equals(clazz)) { + return (T) new AppendableMapOfColumns(this); + } + return null; + } + +} diff --git a/processing/src/main/java/org/apache/druid/segment/ArrayListSegment.java b/processing/src/main/java/org/apache/druid/segment/ArrayListSegment.java new file mode 100644 index 000000000000..97fe80341616 --- /dev/null +++ b/processing/src/main/java/org/apache/druid/segment/ArrayListSegment.java @@ -0,0 +1,126 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.segment; + +import com.google.common.base.Preconditions; +import org.apache.druid.java.util.common.Intervals; +import org.apache.druid.java.util.common.guava.Sequences; +import org.apache.druid.query.rowsandcols.ArrayListRowsAndColumns; +import org.apache.druid.query.rowsandcols.RowsAndColumns; +import org.apache.druid.segment.column.RowSignature; +import org.apache.druid.timeline.SegmentId; +import org.joda.time.Interval; + +import javax.annotation.Nonnull; +import javax.annotation.Nullable; +import java.util.ArrayList; + +/** + * A {@link Segment} that is based on a stream of objects. + */ +public class ArrayListSegment implements Segment +{ + private final SegmentId segmentId; + private final ArrayList rows; + private final RowAdapter rowAdapter; + private final RowSignature rowSignature; + + /** + * Create a list-based segment. + *

    + * The provided List must be in time-order according to the provided {@link RowAdapter#timestampFunction()}. + * The cursor returned by {@link RowBasedStorageAdapter#makeCursors} makes no attempt to verify this, and callers + * will expect it. + *

    + * The provided "rowSignature" will be used for reporting available columns and their capabilities to users of + * {@link #asStorageAdapter()}. Note that the {@link ColumnSelectorFactory} implementation returned by this segment's + * storage adapter will allow creation of selectors on any field, using the {@link RowAdapter#columnFunction} for that + * field, even if it doesn't appear in "rowSignature". + * + * @param segmentId segment identifier; will be returned by {@link #getId()} + * @param rows objects that comprise this segment. Must be re-iterable if support for {@link Cursor#reset()} + * is required. Otherwise, does not need to be re-iterable. + * @param rowAdapter adapter used for reading these objects + * @param rowSignature signature of the columns in these objects + */ + public ArrayListSegment( + final SegmentId segmentId, + final ArrayList rows, + final RowAdapter rowAdapter, + final RowSignature rowSignature + ) + { + this.segmentId = Preconditions.checkNotNull(segmentId, "segmentId"); + this.rows = rows; + this.rowAdapter = rowAdapter; + this.rowSignature = rowSignature; + } + + @Override + @Nonnull + public SegmentId getId() + { + return segmentId; + } + + @Override + @Nonnull + public Interval getDataInterval() + { + return Intervals.ETERNITY; + } + + @Nullable + @Override + public QueryableIndex asQueryableIndex() + { + return null; + } + + @Override + @Nonnull + public StorageAdapter asStorageAdapter() + { + return new RowBasedStorageAdapter<>(Sequences.simple(rows), rowAdapter, rowSignature); + } + + @Nullable + @Override + @SuppressWarnings("unchecked") + public T as(Class clazz) + { + if (RowsAndColumns.class.equals(clazz)) { + return (T) asRowsAndColumns(); + } + return null; + } + + @Override + public void close() + { + // Do nothing. + } + + private RowsAndColumns asRowsAndColumns() + { + return new ArrayListRowsAndColumns(rows, rowAdapter, rowSignature); + } + +} diff --git a/processing/src/main/java/org/apache/druid/segment/BaseSingleValueDimensionSelector.java b/processing/src/main/java/org/apache/druid/segment/BaseSingleValueDimensionSelector.java index adb0ecd1e6e7..c864e6abbc25 100644 --- a/processing/src/main/java/org/apache/druid/segment/BaseSingleValueDimensionSelector.java +++ b/processing/src/main/java/org/apache/druid/segment/BaseSingleValueDimensionSelector.java @@ -50,7 +50,6 @@ public int getValueCardinality() @Override public String lookupName(int id) { - assert id == 0; return getValue(); } diff --git a/processing/src/main/java/org/apache/druid/segment/Segment.java b/processing/src/main/java/org/apache/druid/segment/Segment.java index 245c776b0aca..104216f38c90 100644 --- a/processing/src/main/java/org/apache/druid/segment/Segment.java +++ b/processing/src/main/java/org/apache/druid/segment/Segment.java @@ -23,6 +23,7 @@ import org.apache.druid.timeline.SegmentId; import org.joda.time.Interval; +import javax.annotation.Nonnull; import javax.annotation.Nullable; import java.io.Closeable; @@ -57,9 +58,9 @@ public interface Segment extends Closeable * @param desired interface * @return instance of clazz, or null if the interface is not supported by this segment */ - @SuppressWarnings("unused") + @SuppressWarnings({"unused", "unchecked"}) @Nullable - default T as(Class clazz) + default T as(@Nonnull Class clazz) { if (clazz.equals(QueryableIndex.class)) { return (T) asQueryableIndex(); diff --git a/processing/src/main/java/org/apache/druid/segment/column/ObjectStrategyComplexTypeStrategy.java b/processing/src/main/java/org/apache/druid/segment/column/ObjectStrategyComplexTypeStrategy.java index d05ba208585c..93b992a4f428 100644 --- a/processing/src/main/java/org/apache/druid/segment/column/ObjectStrategyComplexTypeStrategy.java +++ b/processing/src/main/java/org/apache/druid/segment/column/ObjectStrategyComplexTypeStrategy.java @@ -82,9 +82,9 @@ public int write(ByteBuffer buffer, T value, int maxSizeBytes) } @Override - public int compare(T o1, T o2) + public int compare(Object o1, Object o2) { - return objectStrategy.compare(o1, o2); + return objectStrategy.compare((T) o1, (T) o2); } @Override diff --git a/processing/src/main/java/org/apache/druid/segment/nested/NestedDataColumnSerializer.java b/processing/src/main/java/org/apache/druid/segment/nested/NestedDataColumnSerializer.java index f1d6c234dcd0..41b84a638e7c 100644 --- a/processing/src/main/java/org/apache/druid/segment/nested/NestedDataColumnSerializer.java +++ b/processing/src/main/java/org/apache/druid/segment/nested/NestedDataColumnSerializer.java @@ -407,9 +407,9 @@ public int write(ByteBuffer buffer, Integer value, int maxSizeBytes) } @Override - public int compare(Integer o1, Integer o2) + public int compare(Object o1, Object o2) { - return Integer.compare(o1, o2); + return Integer.compare(((Number) o1).intValue(), ((Number) o2).intValue()); } } } diff --git a/processing/src/test/java/org/apache/druid/query/operator/InlineScanOperator.java b/processing/src/test/java/org/apache/druid/query/operator/InlineScanOperator.java new file mode 100644 index 000000000000..dbe2adf830d0 --- /dev/null +++ b/processing/src/test/java/org/apache/druid/query/operator/InlineScanOperator.java @@ -0,0 +1,73 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator; + +import com.google.common.base.Preconditions; +import com.google.common.collect.Iterators; +import org.apache.druid.query.rowsandcols.RowsAndColumns; + +import java.util.Iterator; +import java.util.List; + +public class InlineScanOperator implements Operator +{ + public static InlineScanOperator make(RowsAndColumns item) + { + return new InlineScanOperator(Iterators.singletonIterator(item)); + } + + public static InlineScanOperator make(List items) + { + return new InlineScanOperator(items.iterator()); + } + + private Iterator iter; + + public InlineScanOperator( + Iterator iter + ) + { + Preconditions.checkNotNull(iter); + this.iter = iter; + } + + @Override + public void open() + { + } + + @Override + public RowsAndColumns next() + { + return iter.next(); + } + + @Override + public boolean hasNext() + { + return iter.hasNext(); + } + + @Override + public void close(boolean cascade) + { + iter = null; + } +} diff --git a/processing/src/test/java/org/apache/druid/query/operator/NaivePartitioningOperatorTest.java b/processing/src/test/java/org/apache/druid/query/operator/NaivePartitioningOperatorTest.java new file mode 100644 index 000000000000..5813f36630d1 --- /dev/null +++ b/processing/src/test/java/org/apache/druid/query/operator/NaivePartitioningOperatorTest.java @@ -0,0 +1,105 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator; + +import com.google.common.collect.ImmutableList; +import com.google.common.collect.ImmutableMap; +import org.apache.druid.java.util.common.ISE; +import org.apache.druid.query.operator.window.RowsAndColumnsHelper; +import org.apache.druid.query.rowsandcols.RowsAndColumns; +import org.apache.druid.query.rowsandcols.column.IntArrayColumn; +import org.apache.druid.query.rowsandcols.frame.MapOfColumnsRowsAndColumns; +import org.junit.Assert; +import org.junit.Test; + +import java.util.Arrays; +import java.util.List; + +public class NaivePartitioningOperatorTest +{ + @Test + public void testDefaultImplementation() + { + RowsAndColumns rac = MapOfColumnsRowsAndColumns.fromMap( + ImmutableMap.of( + "sorted", new IntArrayColumn(new int[]{0, 0, 0, 1, 1, 2, 4, 4, 4}), + "unsorted", new IntArrayColumn(new int[]{3, 54, 21, 1, 5, 54, 2, 3, 92}) + ) + ); + + NaivePartitioningOperator op = new NaivePartitioningOperator( + ImmutableList.of("sorted"), + InlineScanOperator.make(rac) + ); + + op.open(); + + List expectations = Arrays.asList( + new RowsAndColumnsHelper() + .expectColumn("sorted", new int[]{0, 0, 0}) + .expectColumn("unsorted", new int[]{3, 54, 21}), + new RowsAndColumnsHelper() + .expectColumn("sorted", new int[]{1, 1}) + .expectColumn("unsorted", new int[]{1, 5}), + new RowsAndColumnsHelper() + .expectColumn("sorted", new int[]{2}) + .expectColumn("unsorted", new int[]{54}), + new RowsAndColumnsHelper() + .expectColumn("sorted", new int[]{4, 4, 4}) + .expectColumn("unsorted", new int[]{2, 3, 92}) + ); + + for (RowsAndColumnsHelper expectation : expectations) { + Assert.assertTrue(op.hasNext()); + expectation.validate(op.next()); + } + Assert.assertFalse(op.hasNext()); + + op.close(true); + } + + @Test + public void testFailUnsorted() + { + RowsAndColumns rac = MapOfColumnsRowsAndColumns.fromMap( + ImmutableMap.of( + "sorted", new IntArrayColumn(new int[]{0, 0, 0, 1, 1, 2, 4, 4, 4}), + "unsorted", new IntArrayColumn(new int[]{3, 54, 21, 1, 5, 54, 2, 3, 92}) + ) + ); + + NaivePartitioningOperator op = new NaivePartitioningOperator( + ImmutableList.of("unsorted"), + InlineScanOperator.make(rac) + ); + + op.open(); + + boolean exceptionThrown = false; + try { + op.next(); + } + catch (ISE ex) { + Assert.assertEquals("Pre-sorted data required, rows[1] and [2] were not in order", ex.getMessage()); + exceptionThrown = true; + } + Assert.assertTrue(exceptionThrown); + } +} diff --git a/processing/src/test/java/org/apache/druid/query/operator/OperatorSequenceTest.java b/processing/src/test/java/org/apache/druid/query/operator/OperatorSequenceTest.java new file mode 100644 index 000000000000..5d114b8bb2aa --- /dev/null +++ b/processing/src/test/java/org/apache/druid/query/operator/OperatorSequenceTest.java @@ -0,0 +1,56 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator; + +import org.apache.druid.java.util.common.guava.Yielder; +import org.apache.druid.java.util.common.guava.YieldingAccumulator; +import org.apache.druid.query.rowsandcols.RowsAndColumns; +import org.apache.druid.query.rowsandcols.column.IntArrayColumn; +import org.apache.druid.query.rowsandcols.frame.MapOfColumnsRowsAndColumns; +import org.junit.Assert; +import org.junit.Test; + +public class OperatorSequenceTest +{ + @Test + public void testSanity() + { + OperatorSequence seq = new OperatorSequence( + () -> InlineScanOperator.make(MapOfColumnsRowsAndColumns.of("hi", new IntArrayColumn(new int[]{1}))) + ); + + Assert.assertEquals(1, seq.accumulate(0, (accumulated, in) -> accumulated + 1).intValue()); + + Yielder yielder = seq.toYielder(0, new YieldingAccumulator() + { + @Override + public Integer accumulate(Integer accumulated, RowsAndColumns in) + { + yield(); + return accumulated + 1; + } + }); + Assert.assertFalse(yielder.isDone()); + Assert.assertEquals(1, yielder.get().intValue()); + + yielder = yielder.next(0); + Assert.assertTrue(yielder.isDone()); + } +} diff --git a/processing/src/test/java/org/apache/druid/query/operator/SequenceOperatorTest.java b/processing/src/test/java/org/apache/druid/query/operator/SequenceOperatorTest.java new file mode 100644 index 000000000000..09f52562d4b9 --- /dev/null +++ b/processing/src/test/java/org/apache/druid/query/operator/SequenceOperatorTest.java @@ -0,0 +1,56 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator; + +import org.apache.druid.java.util.common.guava.Sequences; +import org.apache.druid.query.operator.window.RowsAndColumnsHelper; +import org.apache.druid.query.rowsandcols.column.IntArrayColumn; +import org.apache.druid.query.rowsandcols.frame.MapOfColumnsRowsAndColumns; +import org.junit.Assert; +import org.junit.Test; + +import java.util.Arrays; + +public class SequenceOperatorTest +{ + @Test + public void testSanity() + { + SequenceOperator op = new SequenceOperator(Sequences.simple(Arrays.asList( + MapOfColumnsRowsAndColumns.of("hi", new IntArrayColumn(new int[]{1})), + MapOfColumnsRowsAndColumns.of("hi", new IntArrayColumn(new int[]{1})) + ))); + + op.open(); + + RowsAndColumnsHelper expectations = new RowsAndColumnsHelper() + .expectColumn("hi", new int[]{1}) + .allColumnsRegistered(); + + expectations.validate(op.next()); + Assert.assertTrue(op.hasNext()); + + expectations.validate(op.next()); + Assert.assertFalse(op.hasNext()); + + op.close(true); + op.close(false); + } +} diff --git a/processing/src/test/java/org/apache/druid/query/operator/WindowOperatorQueryTest.java b/processing/src/test/java/org/apache/druid/query/operator/WindowOperatorQueryTest.java new file mode 100644 index 000000000000..dc2e6e9732d4 --- /dev/null +++ b/processing/src/test/java/org/apache/druid/query/operator/WindowOperatorQueryTest.java @@ -0,0 +1,124 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator; + +import com.google.common.collect.ImmutableMap; +import org.apache.druid.java.util.common.IAE; +import org.apache.druid.query.InlineDataSource; +import org.apache.druid.query.QueryContext; +import org.apache.druid.query.TableDataSource; +import org.apache.druid.segment.column.RowSignature; +import org.junit.Assert; +import org.junit.Before; +import org.junit.Test; + +import java.util.ArrayList; +import java.util.Set; + +/** + * Tests the WindowOperatorQuery, it would actually be a lot better to run this through some tests that actually + * validate the operation of queries, but all of the efforts to build out test scaffolding and framework have gone + * into building things out for SQL query operations. As such, all of the tests that validating the actual native + * functionality actually run from the `druid-sql` module instead of this module. It would be best to de-couple + * these and have all of the native, query processing tests happen directly here in processing and have the SQL + * tests only concern themselves with how they plan SQL into Native, but that's a bit big of a nugget to bite off + * at this point in time, so instead we continue the building of technical debt by making this "test" run lines + * of code without actually testing much meaningful behavior. + *

    + * For now, view CalciteWindowQueryTest for actual tests that validate behavior. + */ +public class WindowOperatorQueryTest +{ + WindowOperatorQuery query; + + @Before + public void setUp() + { + query = new WindowOperatorQuery( + InlineDataSource.fromIterable(new ArrayList<>(), RowSignature.empty()), + ImmutableMap.of("sally", "sue"), + RowSignature.empty(), + new ArrayList<>() + ); + } + + @Test + public void getOperators() + { + Assert.assertTrue(query.getOperators().isEmpty()); + } + + @Test + public void getRowSignature() + { + Assert.assertEquals(0, query.getRowSignature().size()); + } + + @Test + public void hasFilters() + { + Assert.assertFalse(query.hasFilters()); + } + + @Test + public void getFilter() + { + Assert.assertNull(query.getFilter()); + } + + @Test + public void getType() + { + Assert.assertEquals("windowOperator", query.getType()); + } + + @Test + public void withOverriddenContext() + { + Assert.assertEquals("sue", query.context().get("sally")); + final QueryContext context = query.withOverriddenContext(ImmutableMap.of("sally", "soo")).context(); + Assert.assertEquals("soo", context.get("sally")); + } + + @Test + public void withDataSource() + { + final Set tableNames = query.getDataSource().getTableNames(); + Assert.assertEquals(0, tableNames.size()); + + boolean exceptionThrown = false; + try { + query.withDataSource(new TableDataSource("bob")); + } + catch (IAE e) { + // should fail trying to set a TableDataSource as TableDataSource is not currently allowed. + exceptionThrown = true; + } + Assert.assertTrue(exceptionThrown); + } + + @Test + public void testEquals() + { + Assert.assertEquals(query, query); + Assert.assertEquals(query, query.withDataSource(query.getDataSource())); + Assert.assertNotEquals(query, query.toString()); + } +} diff --git a/processing/src/test/java/org/apache/druid/query/operator/WindowProcessorOperatorTest.java b/processing/src/test/java/org/apache/druid/query/operator/WindowProcessorOperatorTest.java new file mode 100644 index 000000000000..39ba5996d5d4 --- /dev/null +++ b/processing/src/test/java/org/apache/druid/query/operator/WindowProcessorOperatorTest.java @@ -0,0 +1,66 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator; + +import com.google.common.collect.ImmutableMap; +import org.apache.druid.query.operator.window.Processor; +import org.apache.druid.query.rowsandcols.RowsAndColumns; +import org.apache.druid.query.rowsandcols.column.IntArrayColumn; +import org.apache.druid.query.rowsandcols.frame.MapOfColumnsRowsAndColumns; +import org.junit.Assert; +import org.junit.Test; + +public class WindowProcessorOperatorTest +{ + @Test + public void testJustRunsTheProcessor() + { + RowsAndColumns rac = MapOfColumnsRowsAndColumns.fromMap( + ImmutableMap.of( + "colA", new IntArrayColumn(new int[]{1, 2, 3}), + "colB", new IntArrayColumn(new int[]{3, 2, 1}) + ) + ); + + WindowProcessorOperator op = new WindowProcessorOperator( + new Processor() + { + @Override + public RowsAndColumns process(RowsAndColumns incomingPartition) + { + return incomingPartition; + } + + @Override + public boolean validateEquivalent(Processor otherProcessor) + { + return true; + } + }, + InlineScanOperator.make(rac) + ); + + op.open(); + Assert.assertTrue(op.hasNext()); + Assert.assertSame(rac, op.next()); + Assert.assertFalse(op.hasNext()); + op.close(true); + } +} diff --git a/processing/src/test/java/org/apache/druid/query/operator/window/ComposingProcessorTest.java b/processing/src/test/java/org/apache/druid/query/operator/window/ComposingProcessorTest.java new file mode 100644 index 000000000000..570cba65d92c --- /dev/null +++ b/processing/src/test/java/org/apache/druid/query/operator/window/ComposingProcessorTest.java @@ -0,0 +1,74 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator.window; + +import org.apache.druid.query.rowsandcols.RowsAndColumns; +import org.junit.Assert; +import org.junit.Test; + +public class ComposingProcessorTest +{ + @Test + public void testSanity() + { + final ProcessorForTesting firstProcessor = new ProcessorForTesting(); + final ProcessorForTesting secondProcessor = new ProcessorForTesting(); + + ComposingProcessor proc = new ComposingProcessor(firstProcessor, secondProcessor); + + proc.process(null); + Assert.assertEquals(1, firstProcessor.processCounter); + Assert.assertEquals(1, secondProcessor.processCounter); + + proc.process(null); + Assert.assertEquals(2, firstProcessor.processCounter); + Assert.assertEquals(2, secondProcessor.processCounter); + + Assert.assertTrue(proc.validateEquivalent(proc)); + Assert.assertEquals(1, firstProcessor.validateCounter); + Assert.assertEquals(1, secondProcessor.validateCounter); + + firstProcessor.validationResult = false; + Assert.assertFalse(proc.validateEquivalent(proc)); + Assert.assertEquals(2, firstProcessor.validateCounter); + Assert.assertEquals(1, secondProcessor.validateCounter); + } + + private static class ProcessorForTesting implements Processor + { + private int processCounter = 0; + private int validateCounter = 0; + private boolean validationResult = true; + + @Override + public RowsAndColumns process(RowsAndColumns incomingPartition) + { + ++processCounter; + return incomingPartition; + } + + @Override + public boolean validateEquivalent(Processor otherProcessor) + { + ++validateCounter; + return validationResult; + } + } +} diff --git a/processing/src/test/java/org/apache/druid/query/operator/window/RowsAndColumnsHelper.java b/processing/src/test/java/org/apache/druid/query/operator/window/RowsAndColumnsHelper.java new file mode 100644 index 000000000000..451ead6a9396 --- /dev/null +++ b/processing/src/test/java/org/apache/druid/query/operator/window/RowsAndColumnsHelper.java @@ -0,0 +1,269 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator.window; + +import com.google.common.collect.ImmutableSet; +import org.apache.druid.java.util.common.ISE; +import org.apache.druid.java.util.common.StringUtils; +import org.apache.druid.query.rowsandcols.RowsAndColumns; +import org.apache.druid.query.rowsandcols.column.Column; +import org.apache.druid.query.rowsandcols.column.ColumnAccessor; +import org.apache.druid.segment.column.ColumnType; +import org.junit.Assert; + +import java.util.Collection; +import java.util.LinkedHashMap; +import java.util.Map; +import java.util.Set; + +public class RowsAndColumnsHelper +{ + public static void assertEquals(RowsAndColumns rac, String name, int[] expectedResults) + { + final Column column = rac.findColumn(name); + Assert.assertNotNull(column); + final ColumnAccessor accessor = column.toAccessor(); + Assert.assertEquals(expectedResults.length, accessor.numRows()); + for (int i = 0; i < expectedResults.length; ++i) { + Assert.assertEquals(StringUtils.format("%s[%s]", name, i), expectedResults[i], accessor.getInt(i)); + } + } + + public static void assertEquals(RowsAndColumns rac, String name, long[] expectedResults) + { + final Column column = rac.findColumn(name); + Assert.assertNotNull(column); + final ColumnAccessor accessor = column.toAccessor(); + Assert.assertEquals(expectedResults.length, accessor.numRows()); + for (int i = 0; i < expectedResults.length; ++i) { + Assert.assertEquals(StringUtils.format("%s[%s]", name, i), expectedResults[i], accessor.getLong(i)); + } + } + + public static void assertEquals(RowsAndColumns rac, String name, double[] expectedResults) + { + final Column column = rac.findColumn(name); + Assert.assertNotNull(column); + final ColumnAccessor accessor = column.toAccessor(); + Assert.assertEquals(expectedResults.length, accessor.numRows()); + for (int i = 0; i < expectedResults.length; ++i) { + Assert.assertEquals(StringUtils.format("%s[%s]", name, i), expectedResults[i], accessor.getDouble(i), 0.0d); + } + } + + private final Map helpers = new LinkedHashMap<>(); + private Set fullColumnSet; + + public RowsAndColumnsHelper() + { + } + + public RowsAndColumnsHelper expectColumn(String col, int[] expectedVals) + { + final ColumnHelper helper = columnHelper(col, expectedVals.length, ColumnType.LONG); + helper.setExpectation(expectedVals); + return this; + } + + public RowsAndColumnsHelper expectColumn(String col, long[] expectedVals) + { + final ColumnHelper helper = columnHelper(col, expectedVals.length, ColumnType.LONG); + helper.setExpectation(expectedVals); + return this; + } + + public RowsAndColumnsHelper expectColumn(String col, double[] expectedVals) + { + final ColumnHelper helper = columnHelper(col, expectedVals.length, ColumnType.DOUBLE); + helper.setExpectation(expectedVals); + return this; + } + + public ColumnHelper columnHelper(String column, int expectedSize, ColumnType expectedType) + { + ColumnHelper retVal = helpers.get(column); + if (retVal == null) { + retVal = new ColumnHelper(expectedSize, expectedType); + helpers.put(column, retVal); + return retVal; + } else { + throw new ISE( + "column[%s] expectations already defined, size[%s], type[%s]", + column, + retVal.expectedVals.length, + retVal.expectedType + ); + } + } + + public RowsAndColumnsHelper expectFullColumns(Set fullColumnSet) + { + this.fullColumnSet = fullColumnSet; + return this; + } + + public RowsAndColumnsHelper allColumnsRegistered() + { + this.fullColumnSet = ImmutableSet.copyOf(helpers.keySet()); + return this; + } + + public void validate(RowsAndColumns rac) + { + validate("", rac); + } + + public void validate(String name, RowsAndColumns rac) + { + if (fullColumnSet != null) { + final Collection columnNames = rac.getColumnNames(); + Assert.assertEquals(name, fullColumnSet.size(), columnNames.size()); + Assert.assertTrue(name, fullColumnSet.containsAll(columnNames)); + } + + for (Map.Entry entry : helpers.entrySet()) { + entry.getValue().validate(StringUtils.format("%s.%s", name, entry.getKey()), rac.findColumn(entry.getKey())); + } + } + + public static class ColumnHelper + { + private final ColumnType expectedType; + private final Object[] expectedVals; + private final boolean[] expectedNulls; + + public ColumnHelper(int expectedSize, ColumnType expectedType) + { + this.expectedType = expectedType; + this.expectedVals = new Object[expectedSize]; + this.expectedNulls = new boolean[expectedVals.length]; + } + + public ColumnHelper setExpectation(int[] expectedVals) + { + for (int i = 0; i < expectedVals.length; i++) { + this.expectedVals[i] = expectedVals[i]; + } + return this; + } + + public ColumnHelper setExpectation(long[] expectedVals) + { + for (int i = 0; i < expectedVals.length; i++) { + this.expectedVals[i] = expectedVals[i]; + } + return this; + } + + public ColumnHelper setExpectation(double[] expectedVals) + { + for (int i = 0; i < expectedVals.length; i++) { + this.expectedVals[i] = expectedVals[i]; + } + return this; + } + + public ColumnHelper setExpectation(float[] expectedVals) + { + for (int i = 0; i < expectedVals.length; i++) { + this.expectedVals[i] = expectedVals[i]; + } + return this; + } + + public ColumnHelper setExpectation(Object[] expectedVals) + { + System.arraycopy(expectedVals, 0, this.expectedVals, 0, expectedVals.length); + return this; + } + + public ColumnHelper setNulls(int[] nullIndexes) + { + for (int nullIndex : nullIndexes) { + this.expectedNulls[nullIndex] = true; + } + return this; + } + + public void validate(String msgBase, Column col) + { + final ColumnAccessor accessor = col.toAccessor(); + + Assert.assertEquals(msgBase, expectedType, accessor.getType()); + Assert.assertEquals(msgBase, expectedVals.length, accessor.numRows()); + for (int i = 0; i < accessor.numRows(); ++i) { + final String msg = StringUtils.format("%s[%s]", msgBase, i); + Object expectedVal = expectedVals[i]; + if (expectedVal == null) { + Assert.assertTrue(msg, expectedNulls[i]); + Assert.assertTrue(msg, accessor.isNull(i)); + Assert.assertNull(msg, accessor.getObject(i)); + } + if (expectedVal instanceof Float) { + if (expectedNulls[i]) { + Assert.assertTrue(msg, accessor.isNull(i)); + Assert.assertEquals(msg, 0.0f, accessor.getFloat(i), 0.0); + } else { + Assert.assertFalse(msg, accessor.isNull(i)); + Assert.assertEquals(msg, (Float) expectedVal, accessor.getFloat(i), 0.0); + } + } else if (expectedVal instanceof Double) { + if (expectedNulls[i]) { + Assert.assertTrue(msg, accessor.isNull(i)); + Assert.assertEquals(msg, 0.0d, accessor.getDouble(i), 0.0); + } else { + Assert.assertFalse(msg, accessor.isNull(i)); + Assert.assertEquals(msg, (Double) expectedVal, accessor.getDouble(i), 0.0); + } + } else if (expectedVal instanceof Integer) { + if (expectedNulls[i]) { + Assert.assertTrue(msg, accessor.isNull(i)); + Assert.assertEquals(msg, 0, accessor.getInt(i)); + } else { + Assert.assertFalse(msg, accessor.isNull(i)); + Assert.assertEquals(msg, ((Integer) expectedVal).intValue(), accessor.getInt(i)); + } + } else if (expectedVal instanceof Long) { + if (expectedNulls[i]) { + Assert.assertTrue(msg, accessor.isNull(i)); + Assert.assertEquals(msg, 0, accessor.getLong(i)); + } else { + Assert.assertFalse(msg, accessor.isNull(i)); + Assert.assertEquals(msg, ((Long) expectedVal).longValue(), accessor.getLong(i)); + } + } else { + if (expectedNulls[i]) { + Assert.assertTrue(msg, accessor.isNull(i)); + Assert.assertNull(msg, accessor.getObject(i)); + // asserting null on the expected value is here for consistency in the tests. If it fails, it's most + // likely indicative of something wrong with the test setup than the actual logic, we keep it for + // sanity's sake to things consistent. + Assert.assertNull(msg, expectedVals[i]); + } else { + final Object obj = accessor.getObject(i); + Assert.assertFalse(msg, accessor.isNull(i)); + Assert.assertNotNull(msg, obj); + Assert.assertEquals(msg, expectedVals[i], obj); + } + } + } + } + } +} diff --git a/processing/src/test/java/org/apache/druid/query/operator/window/WindowAggregateProcessorTest.java b/processing/src/test/java/org/apache/druid/query/operator/window/WindowAggregateProcessorTest.java new file mode 100644 index 000000000000..856dd0a30421 --- /dev/null +++ b/processing/src/test/java/org/apache/druid/query/operator/window/WindowAggregateProcessorTest.java @@ -0,0 +1,115 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator.window; + +import org.apache.druid.common.config.NullHandling; +import org.apache.druid.query.aggregation.DoubleMaxAggregatorFactory; +import org.apache.druid.query.aggregation.DoubleSumAggregatorFactory; +import org.apache.druid.query.aggregation.LongMaxAggregatorFactory; +import org.apache.druid.query.aggregation.LongSumAggregatorFactory; +import org.apache.druid.query.operator.window.ranking.WindowRowNumberProcessor; +import org.apache.druid.query.rowsandcols.RowsAndColumns; +import org.apache.druid.query.rowsandcols.column.Column; +import org.apache.druid.query.rowsandcols.column.DoubleArrayColumn; +import org.apache.druid.query.rowsandcols.column.IntArrayColumn; +import org.apache.druid.query.rowsandcols.column.ObjectArrayColumn; +import org.apache.druid.query.rowsandcols.frame.MapOfColumnsRowsAndColumns; +import org.apache.druid.segment.column.ColumnType; +import org.junit.Assert; +import org.junit.Test; + +import java.util.ArrayList; +import java.util.Arrays; +import java.util.LinkedHashMap; +import java.util.Map; + +public class WindowAggregateProcessorTest +{ + static { + NullHandling.initializeForTests(); + } + + @Test + public void testAggregation() + { + Map map = new LinkedHashMap<>(); + map.put("intCol", new IntArrayColumn(new int[]{0, 1, 2, 3, 4, 5, 6, 7, 8, 9})); + map.put("doubleCol", new DoubleArrayColumn(new double[]{0, 1, 2, 3, 4, 5, 6, 7, 8, 9})); + map.put("objectCol", new ObjectArrayColumn( + new String[]{"a", "b", "c", "d", "e", "f", "g", "h", "i", "j"}, + ColumnType.STRING + ) + ); + + MapOfColumnsRowsAndColumns rac = MapOfColumnsRowsAndColumns.fromMap(map); + + WindowAggregateProcessor processor = new WindowAggregateProcessor( + Arrays.asList( + new LongSumAggregatorFactory("sumFromLong", "intCol"), + new LongSumAggregatorFactory("sumFromDouble", "doubleCol"), + new DoubleMaxAggregatorFactory("maxFromInt", "intCol"), + new DoubleMaxAggregatorFactory("maxFromDouble", "doubleCol") + ), + Arrays.asList( + new LongMaxAggregatorFactory("cummMax", "intCol"), + new DoubleSumAggregatorFactory("cummSum", "doubleCol") + ) + ); + + RowsAndColumnsHelper expectations = new RowsAndColumnsHelper() + .expectColumn("intCol", new int[]{0, 1, 2, 3, 4, 5, 6, 7, 8, 9}) + .expectColumn("doubleCol", new double[]{0, 1, 2, 3, 4, 5, 6, 7, 8, 9}) + .expectColumn("sumFromLong", new long[]{45, 45, 45, 45, 45, 45, 45, 45, 45, 45}) + .expectColumn("sumFromDouble", new long[]{45, 45, 45, 45, 45, 45, 45, 45, 45, 45}) + .expectColumn("maxFromInt", new double[]{9, 9, 9, 9, 9, 9, 9, 9, 9, 9}) + .expectColumn("maxFromDouble", new double[]{9, 9, 9, 9, 9, 9, 9, 9, 9, 9}) + .expectColumn("cummMax", new long[]{0, 1, 2, 3, 4, 5, 6, 7, 8, 9}) + .expectColumn("cummSum", new double[]{0, 1, 3, 6, 10, 15, 21, 28, 36, 45}); + + final RowsAndColumns results = processor.process(rac); + expectations.validate(results); + } + + @Test + public void testValidateEquality() + { + WindowAggregateProcessor processor = new WindowAggregateProcessor( + Arrays.asList( + new LongSumAggregatorFactory("sumFromLong", "intCol"), + new LongSumAggregatorFactory("sumFromDouble", "doubleCol"), + new DoubleMaxAggregatorFactory("maxFromInt", "intCol"), + new DoubleMaxAggregatorFactory("maxFromDouble", "doubleCol") + ), + Arrays.asList( + new LongMaxAggregatorFactory("cummMax", "intCol"), + new DoubleSumAggregatorFactory("cummSum", "doubleCol") + ) + ); + + Assert.assertTrue(processor.validateEquivalent(processor)); + Assert.assertFalse(processor.validateEquivalent(new WindowRowNumberProcessor("bob"))); + Assert.assertFalse(processor.validateEquivalent(new WindowAggregateProcessor(processor.getAggregations(), null))); + Assert.assertFalse(processor.validateEquivalent( + new WindowAggregateProcessor(new ArrayList<>(), processor.getCumulativeAggregations()) + )); + Assert.assertFalse(processor.validateEquivalent(new WindowAggregateProcessor(new ArrayList<>(), null))); + } + +} diff --git a/processing/src/test/java/org/apache/druid/query/operator/window/ranking/WindowCumeDistProcessorTest.java b/processing/src/test/java/org/apache/druid/query/operator/window/ranking/WindowCumeDistProcessorTest.java new file mode 100644 index 000000000000..7b0bf1448170 --- /dev/null +++ b/processing/src/test/java/org/apache/druid/query/operator/window/ranking/WindowCumeDistProcessorTest.java @@ -0,0 +1,53 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator.window.ranking; + +import org.apache.druid.query.operator.window.Processor; +import org.apache.druid.query.operator.window.RowsAndColumnsHelper; +import org.apache.druid.query.rowsandcols.RowsAndColumns; +import org.apache.druid.query.rowsandcols.column.Column; +import org.apache.druid.query.rowsandcols.column.IntArrayColumn; +import org.apache.druid.query.rowsandcols.frame.MapOfColumnsRowsAndColumns; +import org.junit.Test; + +import java.util.Collections; +import java.util.LinkedHashMap; +import java.util.Map; + +public class WindowCumeDistProcessorTest +{ + @Test + public void testCumeDistProcessing() + { + Map map = new LinkedHashMap<>(); + map.put("vals", new IntArrayColumn(new int[]{7, 18, 18, 30, 120, 121, 122, 122, 8290, 8290})); + + MapOfColumnsRowsAndColumns rac = MapOfColumnsRowsAndColumns.fromMap(map); + + Processor processor = new WindowCumeDistProcessor(Collections.singletonList("vals"), "CumeDist"); + + final RowsAndColumnsHelper expectations = new RowsAndColumnsHelper() + .expectColumn("vals", new int[]{7, 18, 18, 30, 120, 121, 122, 122, 8290, 8290}) + .expectColumn("CumeDist", new double[]{0.1, 0.3, 0.3, 0.4, 0.5, 0.6, 0.8, 0.8, 1.0, 1.0}); + + final RowsAndColumns results = processor.process(rac); + expectations.validate(results); + } +} diff --git a/processing/src/test/java/org/apache/druid/query/operator/window/ranking/WindowDenseRankProcessorTest.java b/processing/src/test/java/org/apache/druid/query/operator/window/ranking/WindowDenseRankProcessorTest.java new file mode 100644 index 000000000000..d61c40ca48e2 --- /dev/null +++ b/processing/src/test/java/org/apache/druid/query/operator/window/ranking/WindowDenseRankProcessorTest.java @@ -0,0 +1,54 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator.window.ranking; + +import org.apache.druid.query.operator.window.Processor; +import org.apache.druid.query.operator.window.RowsAndColumnsHelper; +import org.apache.druid.query.rowsandcols.RowsAndColumns; +import org.apache.druid.query.rowsandcols.column.Column; +import org.apache.druid.query.rowsandcols.column.IntArrayColumn; +import org.apache.druid.query.rowsandcols.frame.MapOfColumnsRowsAndColumns; +import org.junit.Test; + +import java.util.Collections; +import java.util.LinkedHashMap; +import java.util.Map; + +public class WindowDenseRankProcessorTest +{ + @Test + public void testDenseRankProcessing() + { + Map map = new LinkedHashMap<>(); + map.put("vals", new IntArrayColumn(new int[]{7, 18, 18, 30, 120, 121, 122, 122, 8290, 8290})); + + MapOfColumnsRowsAndColumns rac = MapOfColumnsRowsAndColumns.fromMap(map); + + Processor processor = new WindowDenseRankProcessor(Collections.singletonList("vals"), "DenseRank"); + + final RowsAndColumnsHelper expectations = new RowsAndColumnsHelper() + .expectColumn("vals", new int[]{7, 18, 18, 30, 120, 121, 122, 122, 8290, 8290}) + .expectColumn("DenseRank", new int[]{1, 2, 2, 3, 4, 5, 6, 6, 7, 7}); + + final RowsAndColumns results = processor.process(rac); + expectations.validate(results); + + } +} diff --git a/processing/src/test/java/org/apache/druid/query/operator/window/ranking/WindowPercentileProcessorTest.java b/processing/src/test/java/org/apache/druid/query/operator/window/ranking/WindowPercentileProcessorTest.java new file mode 100644 index 000000000000..cc6473f10a97 --- /dev/null +++ b/processing/src/test/java/org/apache/druid/query/operator/window/ranking/WindowPercentileProcessorTest.java @@ -0,0 +1,84 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator.window.ranking; + +import org.apache.druid.query.operator.window.ComposingProcessor; +import org.apache.druid.query.operator.window.Processor; +import org.apache.druid.query.operator.window.RowsAndColumnsHelper; +import org.apache.druid.query.rowsandcols.RowsAndColumns; +import org.apache.druid.query.rowsandcols.column.Column; +import org.apache.druid.query.rowsandcols.column.DoubleArrayColumn; +import org.apache.druid.query.rowsandcols.column.IntArrayColumn; +import org.apache.druid.query.rowsandcols.column.ObjectArrayColumn; +import org.apache.druid.query.rowsandcols.frame.MapOfColumnsRowsAndColumns; +import org.apache.druid.segment.column.ColumnType; +import org.junit.Test; + +import java.util.LinkedHashMap; +import java.util.Map; + +public class WindowPercentileProcessorTest +{ + @Test + public void testPercentileProcessing() + { + Map map = new LinkedHashMap<>(); + map.put("intCol", new IntArrayColumn(new int[]{88, 1, 2, 3, 4, 5, 6, 7, 8, 9})); + map.put("doubleCol", new DoubleArrayColumn(new double[]{0.4728, 1, 2, 3, 4, 5, 6, 7, 8, 9})); + map.put("objectCol", new ObjectArrayColumn( + new String[]{"a", "b", "c", "d", "e", "f", "g", "h", "i", "j"}, + ColumnType.STRING + )); + + MapOfColumnsRowsAndColumns rac = MapOfColumnsRowsAndColumns.fromMap(map); + + Processor processor = new ComposingProcessor( + new WindowPercentileProcessor("1", 1), + new WindowPercentileProcessor("2", 2), + new WindowPercentileProcessor("3", 3), + new WindowPercentileProcessor("4", 4), + new WindowPercentileProcessor("5", 5), + new WindowPercentileProcessor("6", 6), + new WindowPercentileProcessor("7", 7), + new WindowPercentileProcessor("8", 8), + new WindowPercentileProcessor("9", 9), + new WindowPercentileProcessor("10", 10), + new WindowPercentileProcessor("10292", 10292) + ); + + final RowsAndColumnsHelper expectations = new RowsAndColumnsHelper() + .expectColumn("intCol", new int[]{88, 1, 2, 3, 4, 5, 6, 7, 8, 9}) + .expectColumn("doubleCol", new double[]{0.4728, 1, 2, 3, 4, 5, 6, 7, 8, 9}) + .expectColumn("1", new int[]{1, 1, 1, 1, 1, 1, 1, 1, 1, 1}) + .expectColumn("2", new int[]{1, 1, 1, 1, 1, 2, 2, 2, 2, 2}) + .expectColumn("3", new int[]{1, 1, 1, 1, 2, 2, 2, 3, 3, 3}) + .expectColumn("4", new int[]{1, 1, 1, 2, 2, 2, 3, 3, 4, 4}) + .expectColumn("5", new int[]{1, 1, 2, 2, 3, 3, 4, 4, 5, 5}) + .expectColumn("6", new int[]{1, 1, 2, 2, 3, 3, 4, 4, 5, 6}) + .expectColumn("7", new int[]{1, 1, 2, 2, 3, 3, 4, 5, 6, 7}) + .expectColumn("8", new int[]{1, 1, 2, 2, 3, 4, 5, 6, 7, 8}) + .expectColumn("9", new int[]{1, 1, 2, 3, 4, 5, 6, 7, 8, 9}) + .expectColumn("10", new int[]{1, 2, 3, 4, 5, 6, 7, 8, 9, 10}) + .expectColumn("10292", new int[]{1, 2, 3, 4, 5, 6, 7, 8, 9, 10}); + + final RowsAndColumns results = processor.process(rac); + expectations.validate(results); + } +} diff --git a/processing/src/test/java/org/apache/druid/query/operator/window/ranking/WindowRankProcessorTest.java b/processing/src/test/java/org/apache/druid/query/operator/window/ranking/WindowRankProcessorTest.java new file mode 100644 index 000000000000..3f004a6fd26b --- /dev/null +++ b/processing/src/test/java/org/apache/druid/query/operator/window/ranking/WindowRankProcessorTest.java @@ -0,0 +1,84 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator.window.ranking; + +import org.apache.druid.query.operator.window.ComposingProcessor; +import org.apache.druid.query.operator.window.Processor; +import org.apache.druid.query.operator.window.RowsAndColumnsHelper; +import org.apache.druid.query.rowsandcols.RowsAndColumns; +import org.apache.druid.query.rowsandcols.column.Column; +import org.apache.druid.query.rowsandcols.column.IntArrayColumn; +import org.apache.druid.query.rowsandcols.frame.MapOfColumnsRowsAndColumns; +import org.junit.Test; + +import java.util.Collections; +import java.util.LinkedHashMap; +import java.util.Map; + +public class WindowRankProcessorTest +{ + @Test + public void testRankProcessing() + { + Map map = new LinkedHashMap<>(); + map.put("vals", new IntArrayColumn(new int[]{7, 18, 18, 30, 120, 121, 122, 122, 8290, 8290})); + + MapOfColumnsRowsAndColumns rac = MapOfColumnsRowsAndColumns.fromMap(map); + + Processor processor = new ComposingProcessor( + new WindowRankProcessor(Collections.singletonList("vals"), "rank", false), + new WindowRankProcessor(Collections.singletonList("vals"), "rankAsPercent", true) + ); + + final RowsAndColumnsHelper expectations = new RowsAndColumnsHelper() + .expectColumn("vals", new int[]{7, 18, 18, 30, 120, 121, 122, 122, 8290, 8290}) + .expectColumn("rank", new int[]{1, 2, 2, 4, 5, 6, 7, 7, 9, 9}) + .expectColumn( + "rankAsPercent", + new double[]{0.0, 1 / 9d, 1 / 9d, 3 / 9d, 4 / 9d, 5 / 9d, 6 / 9d, 6 / 9d, 8 / 9d, 8 / 9d} + ); + + final RowsAndColumns results = processor.process(rac); + expectations.validate(results); + + } + + @Test + public void testRankSingle() + { + Map map = new LinkedHashMap<>(); + map.put("vals", new IntArrayColumn(new int[]{7})); + + MapOfColumnsRowsAndColumns rac = MapOfColumnsRowsAndColumns.fromMap(map); + + Processor processor = new ComposingProcessor( + new WindowRankProcessor(Collections.singletonList("vals"), "rank", false), + new WindowRankProcessor(Collections.singletonList("vals"), "rankAsPercent", true) + ); + + final RowsAndColumnsHelper expectations = new RowsAndColumnsHelper() + .expectColumn("vals", new int[]{7}) + .expectColumn("rank", new int[]{1}) + .expectColumn("rankAsPercent", new double[]{0.0}); + + final RowsAndColumns results = processor.process(rac); + expectations.validate(results); + } +} diff --git a/processing/src/test/java/org/apache/druid/query/operator/window/ranking/WindowRowNumberProcessorTest.java b/processing/src/test/java/org/apache/druid/query/operator/window/ranking/WindowRowNumberProcessorTest.java new file mode 100644 index 000000000000..bc06de60e880 --- /dev/null +++ b/processing/src/test/java/org/apache/druid/query/operator/window/ranking/WindowRowNumberProcessorTest.java @@ -0,0 +1,61 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator.window.ranking; + +import org.apache.druid.query.operator.window.Processor; +import org.apache.druid.query.operator.window.RowsAndColumnsHelper; +import org.apache.druid.query.rowsandcols.RowsAndColumns; +import org.apache.druid.query.rowsandcols.column.Column; +import org.apache.druid.query.rowsandcols.column.DoubleArrayColumn; +import org.apache.druid.query.rowsandcols.column.IntArrayColumn; +import org.apache.druid.query.rowsandcols.column.ObjectArrayColumn; +import org.apache.druid.query.rowsandcols.frame.MapOfColumnsRowsAndColumns; +import org.apache.druid.segment.column.ColumnType; +import org.junit.Test; + +import java.util.LinkedHashMap; +import java.util.Map; + +public class WindowRowNumberProcessorTest +{ + @Test + public void testRowNumberProcessing() + { + Map map = new LinkedHashMap<>(); + map.put("intCol", new IntArrayColumn(new int[]{88, 1, 2, 3, 4, 5, 6, 7, 8, 9})); + map.put("doubleCol", new DoubleArrayColumn(new double[]{0.4728, 1, 2, 3, 4, 5, 6, 7, 8, 9})); + map.put("objectCol", new ObjectArrayColumn( + new String[]{"a", "b", "c", "d", "e", "f", "g", "h", "i", "j"}, + ColumnType.STRING + )); + + MapOfColumnsRowsAndColumns rac = MapOfColumnsRowsAndColumns.fromMap(map); + + Processor processor = new WindowRowNumberProcessor("rowRow"); + + final RowsAndColumnsHelper expectations = new RowsAndColumnsHelper() + .expectColumn("intCol", new int[]{88, 1, 2, 3, 4, 5, 6, 7, 8, 9}) + .expectColumn("doubleCol", new double[]{0.4728, 1, 2, 3, 4, 5, 6, 7, 8, 9}) + .expectColumn("rowRow", new int[]{1, 2, 3, 4, 5, 6, 7, 8, 9, 10}); + + final RowsAndColumns results = processor.process(rac); + expectations.validate(results); + } +} diff --git a/processing/src/test/java/org/apache/druid/query/operator/window/value/WindowFirstProcessorTest.java b/processing/src/test/java/org/apache/druid/query/operator/window/value/WindowFirstProcessorTest.java new file mode 100644 index 000000000000..425ff9036012 --- /dev/null +++ b/processing/src/test/java/org/apache/druid/query/operator/window/value/WindowFirstProcessorTest.java @@ -0,0 +1,80 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator.window.value; + +import org.apache.druid.query.operator.window.ComposingProcessor; +import org.apache.druid.query.operator.window.RowsAndColumnsHelper; +import org.apache.druid.query.rowsandcols.RowsAndColumns; +import org.apache.druid.query.rowsandcols.column.Column; +import org.apache.druid.query.rowsandcols.column.DoubleArrayColumn; +import org.apache.druid.query.rowsandcols.column.IntArrayColumn; +import org.apache.druid.query.rowsandcols.column.ObjectArrayColumn; +import org.apache.druid.query.rowsandcols.frame.MapOfColumnsRowsAndColumns; +import org.apache.druid.segment.column.ColumnType; +import org.junit.Test; + +import java.util.LinkedHashMap; +import java.util.Map; + +public class WindowFirstProcessorTest +{ + @Test + public void testFirstProcessing() + { + Map map = new LinkedHashMap<>(); + map.put("intCol", new IntArrayColumn(new int[]{88, 1, 2, 3, 4, 5, 6, 7, 8, 9})); + map.put("doubleCol", new DoubleArrayColumn(new double[]{0.4728, 1, 2, 3, 4, 5, 6, 7, 8, 9})); + map.put("objectCol", new ObjectArrayColumn( + new String[]{"a", "b", "c", "d", "e", "f", "g", "h", "i", "j"}, + ColumnType.STRING + )); + map.put("nullFirstCol", new ObjectArrayColumn( + new String[]{null, "b", "c", "d", "e", "f", "g", "h", "i", "j"}, + ColumnType.STRING + )); + + MapOfColumnsRowsAndColumns rac = MapOfColumnsRowsAndColumns.fromMap(map); + + ComposingProcessor processor = new ComposingProcessor( + new WindowFirstProcessor("intCol", "FirstIntCol"), + new WindowFirstProcessor("doubleCol", "FirstDoubleCol"), + new WindowFirstProcessor("objectCol", "FirstObjectCol"), + new WindowFirstProcessor("nullFirstCol", "NullFirstCol") + ); + + final RowsAndColumnsHelper expectations = new RowsAndColumnsHelper() + .expectColumn("intCol", new int[]{88, 1, 2, 3, 4, 5, 6, 7, 8, 9}) + .expectColumn("doubleCol", new double[]{0.4728, 1, 2, 3, 4, 5, 6, 7, 8, 9}) + .expectColumn("FirstIntCol", new int[]{88, 88, 88, 88, 88, 88, 88, 88, 88, 88}) + .expectColumn( + "FirstDoubleCol", + new double[]{0.4728, 0.4728, 0.4728, 0.4728, 0.4728, 0.4728, 0.4728, 0.4728, 0.4728, 0.4728} + ); + + expectations.columnHelper("FirstObjectCol", 10, ColumnType.STRING) + .setExpectation(new String[]{"a", "a", "a", "a", "a", "a", "a", "a", "a", "a"}); + + expectations.columnHelper("NullFirstCol", 10, ColumnType.STRING) + .setNulls(new int[]{0, 1, 2, 3, 4, 5, 6, 7, 8, 9}); + + final RowsAndColumns results = processor.process(rac); + expectations.validate(results); + } +} diff --git a/processing/src/test/java/org/apache/druid/query/operator/window/value/WindowLagProcessorTest.java b/processing/src/test/java/org/apache/druid/query/operator/window/value/WindowLagProcessorTest.java new file mode 100644 index 000000000000..3079a6acea80 --- /dev/null +++ b/processing/src/test/java/org/apache/druid/query/operator/window/value/WindowLagProcessorTest.java @@ -0,0 +1,77 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator.window.value; + +import org.apache.druid.query.operator.window.ComposingProcessor; +import org.apache.druid.query.operator.window.RowsAndColumnsHelper; +import org.apache.druid.query.rowsandcols.RowsAndColumns; +import org.apache.druid.query.rowsandcols.column.Column; +import org.apache.druid.query.rowsandcols.column.DoubleArrayColumn; +import org.apache.druid.query.rowsandcols.column.IntArrayColumn; +import org.apache.druid.query.rowsandcols.column.ObjectArrayColumn; +import org.apache.druid.query.rowsandcols.frame.MapOfColumnsRowsAndColumns; +import org.apache.druid.segment.column.ColumnType; +import org.junit.Test; + +import java.util.LinkedHashMap; +import java.util.Map; + +public class WindowLagProcessorTest +{ + @Test + public void testLagProcessing() + { + Map map = new LinkedHashMap<>(); + map.put("intCol", new IntArrayColumn(new int[]{0, 1, 2, 3, 4, 5, 6, 7, 8, 9})); + map.put("doubleCol", new DoubleArrayColumn(new double[]{0, 1, 2, 3, 4, 5, 6, 7, 8, 9})); + map.put("objectCol", new ObjectArrayColumn( + new String[]{"a", "b", "c", "d", "e", "f", "g", "h", "i", "j"}, + ColumnType.STRING + ) + ); + + MapOfColumnsRowsAndColumns rac = MapOfColumnsRowsAndColumns.fromMap(map); + + ComposingProcessor processor = new ComposingProcessor( + new WindowOffsetProcessor("intCol", "laggardIntCol", -2), + new WindowOffsetProcessor("doubleCol", "laggardDoubleCol", -4), + new WindowOffsetProcessor("objectCol", "laggardObjectCol", -1) + ); + + final RowsAndColumnsHelper expectations = new RowsAndColumnsHelper() + .expectColumn("intCol", new int[]{0, 1, 2, 3, 4, 5, 6, 7, 8, 9}) + .expectColumn("doubleCol", new double[]{0, 1, 2, 3, 4, 5, 6, 7, 8, 9}); + + expectations.columnHelper("laggardIntCol", 10, ColumnType.LONG) + .setExpectation(new int[]{0, 0, 0, 1, 2, 3, 4, 5, 6, 7}) + .setNulls(new int[]{0, 1}); + + expectations.columnHelper("laggardDoubleCol", 10, ColumnType.DOUBLE) + .setExpectation(new double[]{0, 0, 0, 0, 0, 1, 2, 3, 4, 5}) + .setNulls(new int[]{0, 1, 2, 3}); + + expectations.columnHelper("laggardObjectCol", 10, ColumnType.STRING) + .setExpectation(new String[]{null, "a", "b", "c", "d", "e", "f", "g", "h", "i"}) + .setNulls(new int[]{0}); + + final RowsAndColumns results = processor.process(rac); + expectations.validate(results); + } +} diff --git a/processing/src/test/java/org/apache/druid/query/operator/window/value/WindowLastProcessorTest.java b/processing/src/test/java/org/apache/druid/query/operator/window/value/WindowLastProcessorTest.java new file mode 100644 index 000000000000..2e6aabba4971 --- /dev/null +++ b/processing/src/test/java/org/apache/druid/query/operator/window/value/WindowLastProcessorTest.java @@ -0,0 +1,78 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator.window.value; + +import org.apache.druid.query.operator.window.ComposingProcessor; +import org.apache.druid.query.operator.window.RowsAndColumnsHelper; +import org.apache.druid.query.rowsandcols.RowsAndColumns; +import org.apache.druid.query.rowsandcols.column.Column; +import org.apache.druid.query.rowsandcols.column.DoubleArrayColumn; +import org.apache.druid.query.rowsandcols.column.IntArrayColumn; +import org.apache.druid.query.rowsandcols.column.ObjectArrayColumn; +import org.apache.druid.query.rowsandcols.frame.MapOfColumnsRowsAndColumns; +import org.apache.druid.segment.column.ColumnType; +import org.junit.Test; + +import java.util.LinkedHashMap; +import java.util.Map; + +public class WindowLastProcessorTest +{ + @Test + public void testLastProcessing() + { + Map map = new LinkedHashMap<>(); + map.put("intCol", new IntArrayColumn(new int[]{88, 1, 2, 3, 4, 5, 6, 7, 8, 9})); + map.put("doubleCol", new DoubleArrayColumn(new double[]{0.4728, 1, 2, 3, 4, 5, 6, 7, 8, 9.84})); + map.put("objectCol", new ObjectArrayColumn( + new String[]{"a", "b", "c", "d", "e", "f", "g", "h", "i", "j"}, + ColumnType.STRING + )); + map.put("nullLastCol", new ObjectArrayColumn( + new String[]{null, "b", "c", "d", "e", "f", "g", "h", "i", null}, + ColumnType.STRING + )); + + MapOfColumnsRowsAndColumns rac = MapOfColumnsRowsAndColumns.fromMap(map); + + ComposingProcessor processor = new ComposingProcessor( + new WindowLastProcessor("intCol", "LastIntCol"), + new WindowLastProcessor("doubleCol", "LastDoubleCol"), + new WindowLastProcessor("objectCol", "LastObjectCol"), + new WindowLastProcessor("nullLastCol", "NullLastCol") + ); + + + final RowsAndColumnsHelper expectations = new RowsAndColumnsHelper() + .expectColumn("intCol", new int[]{88, 1, 2, 3, 4, 5, 6, 7, 8, 9}) + .expectColumn("doubleCol", new double[]{0.4728, 1, 2, 3, 4, 5, 6, 7, 8, 9.84}) + .expectColumn("LastIntCol", new int[]{9, 9, 9, 9, 9, 9, 9, 9, 9, 9}) + .expectColumn("LastDoubleCol", new double[]{9.84, 9.84, 9.84, 9.84, 9.84, 9.84, 9.84, 9.84, 9.84, 9.84}); + + expectations.columnHelper("LastObjectCol", 10, ColumnType.STRING) + .setExpectation(new String[]{"j", "j", "j", "j", "j", "j", "j", "j", "j", "j"}); + + expectations.columnHelper("NullLastCol", 10, ColumnType.STRING) + .setNulls(new int[]{0, 1, 2, 3, 4, 5, 6, 7, 8, 9}); + + final RowsAndColumns results = processor.process(rac); + expectations.validate(results); + } +} diff --git a/processing/src/test/java/org/apache/druid/query/operator/window/value/WindowLeadProcessorTest.java b/processing/src/test/java/org/apache/druid/query/operator/window/value/WindowLeadProcessorTest.java new file mode 100644 index 000000000000..fac4bf5e081d --- /dev/null +++ b/processing/src/test/java/org/apache/druid/query/operator/window/value/WindowLeadProcessorTest.java @@ -0,0 +1,76 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.operator.window.value; + +import org.apache.druid.query.operator.window.ComposingProcessor; +import org.apache.druid.query.operator.window.RowsAndColumnsHelper; +import org.apache.druid.query.rowsandcols.RowsAndColumns; +import org.apache.druid.query.rowsandcols.column.Column; +import org.apache.druid.query.rowsandcols.column.DoubleArrayColumn; +import org.apache.druid.query.rowsandcols.column.IntArrayColumn; +import org.apache.druid.query.rowsandcols.column.ObjectArrayColumn; +import org.apache.druid.query.rowsandcols.frame.MapOfColumnsRowsAndColumns; +import org.apache.druid.segment.column.ColumnType; +import org.junit.Test; + +import java.util.LinkedHashMap; +import java.util.Map; + +public class WindowLeadProcessorTest +{ + @Test + public void testLeadProcessing() + { + Map map = new LinkedHashMap<>(); + map.put("intCol", new IntArrayColumn(new int[]{0, 1, 2, 3, 4, 5, 6, 7, 8, 9})); + map.put("doubleCol", new DoubleArrayColumn(new double[]{0, 1, 2, 3, 4, 5, 6, 7, 8, 9})); + map.put("objectCol", new ObjectArrayColumn( + new String[]{"a", "b", "c", "d", "e", "f", "g", "h", "i", "j"}, + ColumnType.STRING + ) + ); + + MapOfColumnsRowsAndColumns rac = MapOfColumnsRowsAndColumns.fromMap(map); + + ComposingProcessor processor = new ComposingProcessor( + new WindowOffsetProcessor("intCol", "LeadingIntCol", 2), + new WindowOffsetProcessor("doubleCol", "LeadingDoubleCol", 4), + new WindowOffsetProcessor("objectCol", "LeadingObjectCol", 1) + ); + + final RowsAndColumns results = processor.process(rac); + + final RowsAndColumnsHelper expectations = new RowsAndColumnsHelper() + .expectColumn("intCol", new int[]{88, 1, 2, 3, 4, 5, 6, 7, 8, 9}) + .expectColumn("doubleCol", new double[]{0.4728, 1, 2, 3, 4, 5, 6, 7, 8, 9}); + + expectations.columnHelper("LeadingIntCol", 10, ColumnType.LONG) + .setExpectation(new int[]{2, 3, 4, 5, 6, 7, 8, 9, 0, 0}) + .setNulls(new int[]{8, 9}); + + expectations.columnHelper("LeadingDoubleCol", 10, ColumnType.DOUBLE) + .setExpectation(new double[]{4, 5, 6, 7, 8, 9, 0, 0, 0, 0}) + .setNulls(new int[]{6, 7, 8, 9}); + + expectations.columnHelper("LeadingObjectCol", 10, ColumnType.STRING) + .setExpectation(new String[]{"b", "c", "d", "e", "f", "g", "h", "i", "j", null}) + .setNulls(new int[]{9}); + } +} diff --git a/processing/src/test/java/org/apache/druid/query/rowsandcols/ArrayListRowsAndColumnsTest.java b/processing/src/test/java/org/apache/druid/query/rowsandcols/ArrayListRowsAndColumnsTest.java new file mode 100644 index 000000000000..9d2d18870f5a --- /dev/null +++ b/processing/src/test/java/org/apache/druid/query/rowsandcols/ArrayListRowsAndColumnsTest.java @@ -0,0 +1,66 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.rowsandcols; + +import org.apache.druid.java.util.common.ISE; +import org.apache.druid.query.rowsandcols.column.ColumnAccessor; +import org.apache.druid.query.rowsandcols.frame.MapOfColumnsRowsAndColumns; +import org.apache.druid.segment.column.RowSignature; + +import java.util.ArrayList; + +public class ArrayListRowsAndColumnsTest extends RowsAndColumnsTestBase> +{ + + @Override + public ArrayListRowsAndColumns makeRowsAndColumns(MapOfColumnsRowsAndColumns input) + { + ArrayList rows = new ArrayList<>(input.numRows()); + + ArrayList cols = new ArrayList<>(input.getColumnNames()); + final RowSignature.Builder sigBob = RowSignature.builder(); + + for (int i = 0; i < input.numRows(); ++i) { + rows.add(new Object[cols.size()]); + } + + for (int colIndex = 0; colIndex < cols.size(); ++colIndex) { + String col = cols.get(colIndex); + final ColumnAccessor column = input.findColumn(col).toAccessor(); + sigBob.add(col, column.getType()); + + for (int i = 0; i < column.numRows(); ++i) { + rows.get(i)[colIndex] = column.getObject(i); + } + } + + return new ArrayListRowsAndColumns<>( + rows, + columnName -> { + final int i = cols.indexOf(columnName); + if (i < 0) { + throw new ISE("Couldn't find column[%s]!? i[%s]", columnName, i); + } + return objects -> objects[i]; + }, + sigBob.build() + ); + } +} diff --git a/processing/src/test/java/org/apache/druid/query/rowsandcols/MapOfColumnsRowsAndColumnsTest.java b/processing/src/test/java/org/apache/druid/query/rowsandcols/MapOfColumnsRowsAndColumnsTest.java new file mode 100644 index 000000000000..4e85c152f853 --- /dev/null +++ b/processing/src/test/java/org/apache/druid/query/rowsandcols/MapOfColumnsRowsAndColumnsTest.java @@ -0,0 +1,78 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.rowsandcols; + +import org.apache.druid.java.util.common.ISE; +import org.apache.druid.query.rowsandcols.column.IntArrayColumn; +import org.apache.druid.query.rowsandcols.frame.MapOfColumnsRowsAndColumns; +import org.junit.Assert; +import org.junit.Test; + +import java.util.Collections; + +public class MapOfColumnsRowsAndColumnsTest extends RowsAndColumnsTestBase +{ + @Override + public MapOfColumnsRowsAndColumns makeRowsAndColumns(MapOfColumnsRowsAndColumns input) + { + return input; + } + + @Test + public void testMakeWithEmptyAndNull() + { + boolean exceptionThrown = false; + try { + MapOfColumnsRowsAndColumns.fromMap(null); + } + catch (ISE ex) { + Assert.assertEquals("map[null] cannot be null or empty.", ex.getMessage()); + exceptionThrown = true; + } + Assert.assertTrue(exceptionThrown); + + exceptionThrown = false; + try { + MapOfColumnsRowsAndColumns.fromMap(Collections.emptyMap()); + } + catch (ISE ex) { + Assert.assertEquals("map[{}] cannot be null or empty.", ex.getMessage()); + exceptionThrown = true; + } + Assert.assertTrue(exceptionThrown); + } + + @Test + public void testExceptionOnMismatchedCells() + { + boolean exceptionThrown = false; + try { + MapOfColumnsRowsAndColumns.of( + "1", new IntArrayColumn(new int[]{0}), + "2", new IntArrayColumn(new int[]{0, 1}) + ); + } + catch (ISE ex) { + Assert.assertEquals("Mismatched numCells, expectedNumCells[1], actual[2] from col[2].", ex.getMessage()); + exceptionThrown = true; + } + Assert.assertTrue(exceptionThrown); + } +} diff --git a/processing/src/test/java/org/apache/druid/query/rowsandcols/RowsAndColumnsTestBase.java b/processing/src/test/java/org/apache/druid/query/rowsandcols/RowsAndColumnsTestBase.java new file mode 100644 index 000000000000..d0a4c30f0e6e --- /dev/null +++ b/processing/src/test/java/org/apache/druid/query/rowsandcols/RowsAndColumnsTestBase.java @@ -0,0 +1,188 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.rowsandcols; + +import com.google.common.collect.ImmutableMap; +import org.apache.druid.common.config.NullHandling; +import org.apache.druid.java.util.common.ISE; +import org.apache.druid.query.aggregation.LongMaxAggregatorFactory; +import org.apache.druid.query.aggregation.LongMinAggregatorFactory; +import org.apache.druid.query.aggregation.LongSumAggregatorFactory; +import org.apache.druid.query.operator.window.RowsAndColumnsHelper; +import org.apache.druid.query.rowsandcols.column.IntArrayColumn; +import org.apache.druid.query.rowsandcols.frame.AppendableMapOfColumns; +import org.apache.druid.query.rowsandcols.frame.MapOfColumnsRowsAndColumns; +import org.junit.Assert; +import org.junit.Test; + +import java.util.ArrayList; +import java.util.Arrays; +import java.util.Collections; +import java.util.Iterator; +import java.util.List; + +/** + * This base class is intended to serve as a common set of tests to validate specific RowsAndColumns implementations. + *

    + * Different RowsAndColumns implementations will implement different of the semantic interfaces, this base class should + * test all of the possible semantic interfaces that can be implemented. By doing it this way, we can ensure that + * new RowsAndColumns implementations meet all of the corners cases and other issues that have been previously found. + *

    + * It is expected that this base class is going to grow quite large. As it gets extra large, we could perhaps look + * into whether one of the JUnit test runners could allow us to further sub-divide the test functionality into + * semantic-interface-specific tests. The ultimate goal, however, should be that a new RowsAndColumns implementation + * can very simply take advantage of all of the tests by implementing the abstract + * {@link #makeRowsAndColumns(MapOfColumnsRowsAndColumns)} method and be done. + * + * @param + */ +public abstract class RowsAndColumnsTestBase +{ + static { + NullHandling.initializeForTests(); + } + + public abstract T makeRowsAndColumns(MapOfColumnsRowsAndColumns input); + + @Test + public void testDefaultSortedGroupPartitioner() + { + T rac = makeRowsAndColumns(MapOfColumnsRowsAndColumns.fromMap( + ImmutableMap.of( + "sorted", new IntArrayColumn(new int[]{0, 0, 0, 1, 1, 2, 4, 4, 4}), + "unsorted", new IntArrayColumn(new int[]{3, 54, 21, 1, 5, 54, 2, 3, 92}) + ) + )); + + validateSortedGroupPartitioner("default", new DefaultSortedGroupPartitioner(rac)); + + SortedGroupPartitioner specialized = rac.as(SortedGroupPartitioner.class); + if (specialized != null) { + validateSortedGroupPartitioner("specialized", specialized); + } + } + + private void validateSortedGroupPartitioner(String name, SortedGroupPartitioner parter) + { + + int[] expectedBounds = new int[]{0, 3, 5, 6, 9}; + + List expectations = Arrays.asList( + new RowsAndColumnsHelper() + .expectColumn("sorted", new int[]{0, 0, 0}) + .expectColumn("unsorted", new int[]{3, 54, 21}) + .allColumnsRegistered(), + new RowsAndColumnsHelper() + .expectColumn("sorted", new int[]{1, 1}) + .expectColumn("unsorted", new int[]{1, 5}) + .allColumnsRegistered(), + new RowsAndColumnsHelper() + .expectColumn("sorted", new int[]{2}) + .expectColumn("unsorted", new int[]{54}) + .allColumnsRegistered(), + new RowsAndColumnsHelper() + .expectColumn("sorted", new int[]{4, 4, 4}) + .expectColumn("unsorted", new int[]{2, 3, 92}) + .allColumnsRegistered() + ); + + final List partCols = Collections.singletonList("sorted"); + Assert.assertArrayEquals(name, expectedBounds, parter.computeBoundaries(partCols)); + + final Iterator partedChunks = parter.partitionOnBoundaries(partCols).iterator(); + for (RowsAndColumnsHelper expectation : expectations) { + Assert.assertTrue(name, partedChunks.hasNext()); + expectation.validate(name, partedChunks.next()); + } + Assert.assertFalse(name, partedChunks.hasNext()); + + boolean exceptionThrown = false; + try { + parter.partitionOnBoundaries(Collections.singletonList("unsorted")); + } + catch (ISE ex) { + Assert.assertEquals("Pre-sorted data required, rows[1] and [2] were not in order", ex.getMessage()); + exceptionThrown = true; + } + Assert.assertTrue(exceptionThrown); + } + + @Test + public void testOnHeapAggregatable() + { + T rac = makeRowsAndColumns(MapOfColumnsRowsAndColumns.fromMap( + ImmutableMap.of( + "incremented", new IntArrayColumn(new int[]{1, 2, 3, 4, 5, 6, 7, 8, 9, 10}), + "zeroesOut", new IntArrayColumn(new int[]{4, -4, 3, -3, 4, 82, -90, 4, 0, 0}) + ) + )); + + validateOnHeapAggregatable("default", new DefaultOnHeapAggregatable(rac)); + + OnHeapAggregatable specialized = rac.as(OnHeapAggregatable.class); + if (specialized != null) { + validateOnHeapAggregatable("specialized", specialized); + } + } + + private void validateOnHeapAggregatable(String name, OnHeapAggregatable agger) + { + final ArrayList results = agger.aggregateAll(Arrays.asList( + new LongSumAggregatorFactory("incremented", "incremented"), + new LongMaxAggregatorFactory("zeroesOutMax", "zeroesOut"), + new LongMinAggregatorFactory("zeroesOutMin", "zeroesOut") + )); + + Assert.assertEquals(name, 3, results.size()); + Assert.assertEquals(name, 55L, results.get(0)); + Assert.assertEquals(name, 82L, results.get(1)); + Assert.assertEquals(name, -90L, results.get(2)); + } + + @Test + public void testAppendableRowsAndColumns() + { + T rac = makeRowsAndColumns(MapOfColumnsRowsAndColumns.fromMap( + ImmutableMap.of( + "colA", new IntArrayColumn(new int[]{1, 2, 3, 4, 5, 6, 7, 8, 9, 10}), + "colB", new IntArrayColumn(new int[]{4, -4, 3, -3, 4, 82, -90, 4, 0, 0}) + ) + )); + + validateAppendableRowsAndColumns("default", new AppendableMapOfColumns(rac)); + + AppendableRowsAndColumns specialized = rac.as(AppendableRowsAndColumns.class); + if (specialized != null) { + validateAppendableRowsAndColumns("specialized", specialized); + } + } + + public void validateAppendableRowsAndColumns(String name, AppendableRowsAndColumns appender) + { + appender.addColumn("newCol", new IntArrayColumn(new int[]{1, 2, 3, 4, 5, 6, 7, 8, 9, 10})); + + new RowsAndColumnsHelper() + .expectColumn("colA", new int[]{1, 2, 3, 4, 5, 6, 7, 8, 9, 10}) + .expectColumn("colB", new int[]{4, -4, 3, -3, 4, 82, -90, 4, 0, 0}) + .expectColumn("newCol", new int[]{1, 2, 3, 4, 5, 6, 7, 8, 9, 10}) + .allColumnsRegistered() + .validate(name, appender); + } +} diff --git a/processing/src/test/java/org/apache/druid/query/rowsandcols/column/NullColumnAccessorTest.java b/processing/src/test/java/org/apache/druid/query/rowsandcols/column/NullColumnAccessorTest.java new file mode 100644 index 000000000000..89c286165dc8 --- /dev/null +++ b/processing/src/test/java/org/apache/druid/query/rowsandcols/column/NullColumnAccessorTest.java @@ -0,0 +1,46 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.query.rowsandcols.column; + +import org.junit.Assert; +import org.junit.Test; + +public class NullColumnAccessorTest +{ + + @Test + public void testSanity() + { + NullColumnAccessor accessor = new NullColumnAccessor(10); + Assert.assertEquals(10, accessor.numRows()); + + for (int i = 0; i < 10; ++i) { + Assert.assertTrue(accessor.isNull(i)); + Assert.assertNull(accessor.getObject(i)); + Assert.assertEquals(0, accessor.getInt(i)); + Assert.assertEquals(0, accessor.getLong(i)); + Assert.assertEquals(0.0, accessor.getFloat(i), 0); + Assert.assertEquals(0.0, accessor.getDouble(i), 0); + for (int j = 0; j < i; ++j) { + Assert.assertEquals(0, accessor.compareCells(j, i)); + } + } + } +} diff --git a/processing/src/test/java/org/apache/druid/segment/nested/NestedDataColumnSupplierTest.java b/processing/src/test/java/org/apache/druid/segment/nested/NestedDataColumnSupplierTest.java index 98dc97731f5c..b0396c9293f3 100644 --- a/processing/src/test/java/org/apache/druid/segment/nested/NestedDataColumnSupplierTest.java +++ b/processing/src/test/java/org/apache/druid/segment/nested/NestedDataColumnSupplierTest.java @@ -357,7 +357,7 @@ public int write(ByteBuffer buffer, int offset, T value, int maxSizeBytes) } @Override - public int compare(T o1, T o2) + public int compare(Object o1, Object o2) { return delegate.compare(o1, o2); } diff --git a/server/src/main/java/org/apache/druid/guice/QueryRunnerFactoryModule.java b/server/src/main/java/org/apache/druid/guice/QueryRunnerFactoryModule.java index 102c72d890cc..a2fe8f8d2a58 100644 --- a/server/src/main/java/org/apache/druid/guice/QueryRunnerFactoryModule.java +++ b/server/src/main/java/org/apache/druid/guice/QueryRunnerFactoryModule.java @@ -35,6 +35,8 @@ import org.apache.druid.query.groupby.GroupByQueryRunnerFactory; import org.apache.druid.query.metadata.SegmentMetadataQueryRunnerFactory; import org.apache.druid.query.metadata.metadata.SegmentMetadataQuery; +import org.apache.druid.query.operator.WindowOperatorQuery; +import org.apache.druid.query.operator.WindowOperatorQueryQueryRunnerFactory; import org.apache.druid.query.scan.ScanQuery; import org.apache.druid.query.scan.ScanQueryRunnerFactory; import org.apache.druid.query.search.SearchQuery; @@ -56,14 +58,15 @@ public class QueryRunnerFactoryModule extends QueryToolChestModule { private static final Map>, Class>> MAPPINGS = ImmutableMap.>, Class>>builder() - .put(TimeseriesQuery.class, TimeseriesQueryRunnerFactory.class) - .put(SearchQuery.class, SearchQueryRunnerFactory.class) - .put(TimeBoundaryQuery.class, TimeBoundaryQueryRunnerFactory.class) - .put(SegmentMetadataQuery.class, SegmentMetadataQueryRunnerFactory.class) + .put(DataSourceMetadataQuery.class, DataSourceMetadataQueryRunnerFactory.class) .put(GroupByQuery.class, GroupByQueryRunnerFactory.class) .put(ScanQuery.class, ScanQueryRunnerFactory.class) + .put(SearchQuery.class, SearchQueryRunnerFactory.class) + .put(SegmentMetadataQuery.class, SegmentMetadataQueryRunnerFactory.class) + .put(TimeBoundaryQuery.class, TimeBoundaryQueryRunnerFactory.class) + .put(TimeseriesQuery.class, TimeseriesQueryRunnerFactory.class) .put(TopNQuery.class, TopNQueryRunnerFactory.class) - .put(DataSourceMetadataQuery.class, DataSourceMetadataQueryRunnerFactory.class) + .put(WindowOperatorQuery.class, WindowOperatorQueryQueryRunnerFactory.class) .build(); @Override diff --git a/server/src/main/java/org/apache/druid/guice/QueryToolChestModule.java b/server/src/main/java/org/apache/druid/guice/QueryToolChestModule.java index edb892b03a95..da9ab41e299c 100644 --- a/server/src/main/java/org/apache/druid/guice/QueryToolChestModule.java +++ b/server/src/main/java/org/apache/druid/guice/QueryToolChestModule.java @@ -41,6 +41,8 @@ import org.apache.druid.query.metadata.SegmentMetadataQueryConfig; import org.apache.druid.query.metadata.SegmentMetadataQueryQueryToolChest; import org.apache.druid.query.metadata.metadata.SegmentMetadataQuery; +import org.apache.druid.query.operator.WindowOperatorQuery; +import org.apache.druid.query.operator.WindowOperatorQueryQueryToolChest; import org.apache.druid.query.scan.ScanQuery; import org.apache.druid.query.scan.ScanQueryConfig; import org.apache.druid.query.scan.ScanQueryQueryToolChest; @@ -75,14 +77,15 @@ public class QueryToolChestModule implements Module public final Map, Class> mappings = ImmutableMap., Class>builder() - .put(TimeseriesQuery.class, TimeseriesQueryQueryToolChest.class) - .put(SearchQuery.class, SearchQueryQueryToolChest.class) - .put(TimeBoundaryQuery.class, TimeBoundaryQueryQueryToolChest.class) - .put(SegmentMetadataQuery.class, SegmentMetadataQueryQueryToolChest.class) + .put(DataSourceMetadataQuery.class, DataSourceQueryQueryToolChest.class) .put(GroupByQuery.class, GroupByQueryQueryToolChest.class) .put(ScanQuery.class, ScanQueryQueryToolChest.class) + .put(SearchQuery.class, SearchQueryQueryToolChest.class) + .put(SegmentMetadataQuery.class, SegmentMetadataQueryQueryToolChest.class) + .put(TimeBoundaryQuery.class, TimeBoundaryQueryQueryToolChest.class) + .put(TimeseriesQuery.class, TimeseriesQueryQueryToolChest.class) .put(TopNQuery.class, TopNQueryQueryToolChest.class) - .put(DataSourceMetadataQuery.class, DataSourceQueryQueryToolChest.class) + .put(WindowOperatorQuery.class, WindowOperatorQueryQueryToolChest.class) .build(); @Override diff --git a/server/src/main/java/org/apache/druid/segment/InlineSegmentWrangler.java b/server/src/main/java/org/apache/druid/segment/InlineSegmentWrangler.java index a68859359cde..6128dcf49f26 100644 --- a/server/src/main/java/org/apache/druid/segment/InlineSegmentWrangler.java +++ b/server/src/main/java/org/apache/druid/segment/InlineSegmentWrangler.java @@ -22,14 +22,14 @@ import org.apache.druid.java.util.common.guava.Sequences; import org.apache.druid.query.DataSource; import org.apache.druid.query.InlineDataSource; -import org.apache.druid.segment.join.JoinableFactory; import org.apache.druid.timeline.SegmentId; import org.joda.time.Interval; +import java.util.ArrayList; import java.util.Collections; /** - * A {@link JoinableFactory} for {@link InlineDataSource}. + * A {@link SegmentWrangler} for {@link InlineDataSource}. * * It is not valid to pass any other DataSource type to the "getSegmentsForIntervals" method. */ @@ -38,10 +38,22 @@ public class InlineSegmentWrangler implements SegmentWrangler private static final String SEGMENT_ID = "inline"; @Override + @SuppressWarnings("unchecked") public Iterable getSegmentsForIntervals(final DataSource dataSource, final Iterable intervals) { final InlineDataSource inlineDataSource = (InlineDataSource) dataSource; + if (inlineDataSource.rowsAreArrayList()) { + return Collections.singletonList( + new ArrayListSegment<>( + SegmentId.dummy(SEGMENT_ID), + (ArrayList) inlineDataSource.getRowsAsList(), + inlineDataSource.rowAdapter(), + inlineDataSource.getRowSignature() + ) + ); + } + return Collections.singletonList( new RowBasedSegment<>( SegmentId.dummy(SEGMENT_ID), diff --git a/server/src/main/java/org/apache/druid/server/ClientQuerySegmentWalker.java b/server/src/main/java/org/apache/druid/server/ClientQuerySegmentWalker.java index d6fe03a31d63..638e22919201 100644 --- a/server/src/main/java/org/apache/druid/server/ClientQuerySegmentWalker.java +++ b/server/src/main/java/org/apache/druid/server/ClientQuerySegmentWalker.java @@ -72,9 +72,9 @@ /** * Query handler for the Broker processes (see CliBroker). - * + *

    * This class is responsible for: - * + *

    * 1) Running queries on the cluster using its 'clusterClient' * 2) Running queries locally (when all datasources are global) using its 'localClient' * 3) Inlining subqueries if necessary, in service of the above two goals @@ -133,7 +133,7 @@ public ClientQuerySegmentWalker( { this( emitter, - (QuerySegmentWalker) clusterClient, + clusterClient, (QuerySegmentWalker) localClient, warehouse, joinableFactory, @@ -291,20 +291,20 @@ private DataSource globalizeIfPossible( /** * Replace QueryDataSources with InlineDataSources when necessary and possible. "Necessary" is defined as: - * + *

    * 1) For outermost subqueries: inlining is necessary if the toolchest cannot handle it. * 2) For all other subqueries (e.g. those nested under a join): inlining is always necessary. * - * @param dataSource datasource to process. - * @param toolChestIfOutermost if provided, and if the provided datasource is a {@link QueryDataSource}, this method - * will consider whether the toolchest can handle a subquery on the datasource using - * {@link QueryToolChest#canPerformSubquery}. If the toolchest can handle it, then it will - * not be inlined. See {@link org.apache.druid.query.groupby.GroupByQueryQueryToolChest} - * for an example of a toolchest that can handle subqueries. + * @param dataSource datasource to process. + * @param toolChestIfOutermost if provided, and if the provided datasource is a {@link QueryDataSource}, this method + * will consider whether the toolchest can handle a subquery on the datasource using + * {@link QueryToolChest#canPerformSubquery}. If the toolchest can handle it, then it will + * not be inlined. See {@link org.apache.druid.query.groupby.GroupByQueryQueryToolChest} + * for an example of a toolchest that can handle subqueries. * @param subqueryRowLimitAccumulator an accumulator for tracking the number of accumulated rows in all subqueries * for a particular master query - * @param maxSubqueryRows Max rows that all the subqueries generated by a master query can have, combined - * @param dryRun if true, does not actually execute any subqueries, but will inline empty result sets. + * @param maxSubqueryRows Max rows that all the subqueries generated by a master query can have, combined + * @param dryRun if true, does not actually execute any subqueries, but will inline empty result sets. */ @SuppressWarnings({"rawtypes", "unchecked"}) // Subquery, toolchest, runner handling all use raw types private DataSource inlineIfNecessary( @@ -331,15 +331,18 @@ private DataSource inlineIfNecessary( current = Iterables.getOnlyElement(current.getChildren()); } - assert !(current instanceof QueryDataSource); // lgtm [java/contradictory-type-checks] + if (current instanceof QueryDataSource) { + throw new ISE("Got a QueryDataSource[%s], should've walked it away in the loop above.", current); + } current = inlineIfNecessary(current, null, subqueryRowLimitAccumulator, maxSubqueryRows, dryRun); while (!stack.isEmpty()) { current = stack.pop().withChildren(Collections.singletonList(current)); } - assert current instanceof QueryDataSource; - + if (!(current instanceof QueryDataSource)) { + throw new ISE("Should have a QueryDataSource, but got[%s] instead", current); + } if (toolChest.canPerformSubquery(((QueryDataSource) current).getQuery())) { return current; } else { @@ -431,7 +434,9 @@ private QueryRunner decorateClusterRunner(Query query, QueryRunner .postProcess( objectMapper.convertValue( query.context().getString("postProcessing"), - new TypeReference>() {} + new TypeReference>() + { + } ) ) .map( @@ -572,7 +577,7 @@ private static > InlineDataSource toInlineDataSour final RowSignature signature = toolChest.resultArraySignature(query); - final List resultList = new ArrayList<>(); + final ArrayList resultList = new ArrayList<>(); toolChest.resultsAsArrays(query, results).accumulate( resultList, diff --git a/server/src/main/java/org/apache/druid/server/LocalQuerySegmentWalker.java b/server/src/main/java/org/apache/druid/server/LocalQuerySegmentWalker.java index 730eb3121d47..767e5fbd402b 100644 --- a/server/src/main/java/org/apache/druid/server/LocalQuerySegmentWalker.java +++ b/server/src/main/java/org/apache/druid/server/LocalQuerySegmentWalker.java @@ -93,12 +93,10 @@ public QueryRunner getQueryRunnerForIntervals(final Query query, final final AtomicLong cpuAccumulator = new AtomicLong(0L); - final Function segmentMapFn = analysis - .getDataSource() - .createSegmentMapFunction( - query, - cpuAccumulator - ); + final Function segmentMapFn = + analysis + .getDataSource() + .createSegmentMapFunction(query, cpuAccumulator); final QueryRunnerFactory> queryRunnerFactory = conglomerate.findFactory(query); diff --git a/server/src/main/java/org/apache/druid/server/coordination/ServerManager.java b/server/src/main/java/org/apache/druid/server/coordination/ServerManager.java index 26aabc2ddd68..7990a44b8506 100644 --- a/server/src/main/java/org/apache/druid/server/coordination/ServerManager.java +++ b/server/src/main/java/org/apache/druid/server/coordination/ServerManager.java @@ -194,9 +194,11 @@ public QueryRunner getQueryRunnerForSegments(Query query, Iterable(Lists.newArrayList(specs)); } - final Function segmentMapFn = query.getDataSource().createSegmentMapFunction(query, cpuTimeAccumulator); + final Function segmentMapFn = + query.getDataSource() + .createSegmentMapFunction(query, cpuTimeAccumulator); - // We compute the join cache key here itself so it doesn't need to be re-computed for every segment + // We compute the datasource's cache key here itself so it doesn't need to be re-computed for every segment final Optional cacheKeyPrefix = Optional.ofNullable(query.getDataSource().getCacheKey()); final FunctionalIterable> queryRunners = FunctionalIterable diff --git a/server/src/test/java/org/apache/druid/server/ClientQuerySegmentWalkerTest.java b/server/src/test/java/org/apache/druid/server/ClientQuerySegmentWalkerTest.java index 5e0701680cd0..967d9342d78f 100644 --- a/server/src/test/java/org/apache/druid/server/ClientQuerySegmentWalkerTest.java +++ b/server/src/test/java/org/apache/druid/server/ClientQuerySegmentWalkerTest.java @@ -1391,7 +1391,6 @@ public QueryRunner getQueryRunnerForSegments(Query query, Iterable> timelines, - JoinableFactoryWrapper joinableFactoryWraper, QueryRunnerFactoryConglomerate conglomerate, @Nullable QueryScheduler scheduler ) { - return new TestClusterQuerySegmentWalker(timelines, joinableFactoryWraper, conglomerate, scheduler); + return new TestClusterQuerySegmentWalker(timelines, conglomerate, scheduler); } public static LocalQuerySegmentWalker createLocalQuerySegmentWalker( @@ -348,6 +349,7 @@ public int getMinTopNThreshold() ) .put(GroupByQuery.class, groupByQueryRunnerFactory) .put(TimeBoundaryQuery.class, new TimeBoundaryQueryRunnerFactory(QueryRunnerTestHelper.NOOP_QUERYWATCHER)) + .put(WindowOperatorQuery.class, new WindowOperatorQueryQueryRunnerFactory()) .build() ); diff --git a/server/src/test/java/org/apache/druid/server/TestClusterQuerySegmentWalker.java b/server/src/test/java/org/apache/druid/server/TestClusterQuerySegmentWalker.java index 2245b28169a8..000626f16e2c 100644 --- a/server/src/test/java/org/apache/druid/server/TestClusterQuerySegmentWalker.java +++ b/server/src/test/java/org/apache/druid/server/TestClusterQuerySegmentWalker.java @@ -45,7 +45,6 @@ import org.apache.druid.query.spec.SpecificSegmentSpec; import org.apache.druid.segment.ReferenceCountingSegment; import org.apache.druid.segment.SegmentReference; -import org.apache.druid.segment.join.JoinableFactoryWrapper; import org.apache.druid.timeline.TimelineObjectHolder; import org.apache.druid.timeline.VersionedIntervalTimeline; import org.apache.druid.timeline.partition.PartitionChunk; @@ -71,20 +70,17 @@ public class TestClusterQuerySegmentWalker implements QuerySegmentWalker { private final Map> timelines; - private final JoinableFactoryWrapper joinableFactoryWrapper; private final QueryRunnerFactoryConglomerate conglomerate; @Nullable private final QueryScheduler scheduler; TestClusterQuerySegmentWalker( Map> timelines, - JoinableFactoryWrapper joinableFactoryWrapper, QueryRunnerFactoryConglomerate conglomerate, @Nullable QueryScheduler scheduler ) { this.timelines = timelines; - this.joinableFactoryWrapper = joinableFactoryWrapper; this.conglomerate = conglomerate; this.scheduler = scheduler; } diff --git a/sql/pom.xml b/sql/pom.xml index 9d260eddf4f7..963ae0287522 100644 --- a/sql/pom.xml +++ b/sql/pom.xml @@ -18,7 +18,8 @@ ~ under the License. --> - + 4.0.0 druid-sql @@ -186,6 +187,7 @@ provided + junit @@ -208,6 +210,11 @@ easymock test + + com.fasterxml.jackson.dataformat + jackson-dataformat-yaml + test + commons-io commons-io @@ -249,11 +256,11 @@ mockito-core test - - org.jdbi - jdbi - test - + + org.jdbi + jdbi + test + diff --git a/sql/src/main/java/org/apache/druid/sql/calcite/expression/Expressions.java b/sql/src/main/java/org/apache/druid/sql/calcite/expression/Expressions.java index fa09d99d1f06..b1805519ade9 100644 --- a/sql/src/main/java/org/apache/druid/sql/calcite/expression/Expressions.java +++ b/sql/src/main/java/org/apache/druid/sql/calcite/expression/Expressions.java @@ -93,7 +93,12 @@ public static RexNode fromFieldAccess( ) { if (project == null) { - // I don't think the factory impl matters here. + // Gian doesn't think the factory impl matters here, he's likely correct. But, upon reading what this is doing, + // we are re-building the list of things in the RelDataType for every single call to `fromFieldAccess`. + // `fromFieldAccess` is called pretty regularly in pretty low-level areas of the code, so it would make sense + // that we are perhaps re-creating the exact same object over and over and over and over again and wasting CPU + // cycles. It would likely be good to refactor the code such that we ensure we only ever compute the thing + // once and then reuse it. return RexInputRef.of(fieldNumber, RowSignatures.toRelDataType(rowSignature, new JavaTypeFactoryImpl())); } else { return project.getChildExps().get(fieldNumber); diff --git a/sql/src/main/java/org/apache/druid/sql/calcite/expression/WindowSqlAggregate.java b/sql/src/main/java/org/apache/druid/sql/calcite/expression/WindowSqlAggregate.java new file mode 100644 index 000000000000..7dd158d91f3a --- /dev/null +++ b/sql/src/main/java/org/apache/druid/sql/calcite/expression/WindowSqlAggregate.java @@ -0,0 +1,70 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.sql.calcite.expression; + +import org.apache.calcite.rel.core.AggregateCall; +import org.apache.calcite.rel.core.Project; +import org.apache.calcite.rex.RexBuilder; +import org.apache.calcite.sql.SqlAggFunction; +import org.apache.druid.java.util.common.UOE; +import org.apache.druid.segment.column.RowSignature; +import org.apache.druid.sql.calcite.aggregation.Aggregation; +import org.apache.druid.sql.calcite.aggregation.SqlAggregator; +import org.apache.druid.sql.calcite.planner.PlannerContext; +import org.apache.druid.sql.calcite.rel.VirtualColumnRegistry; + +import javax.annotation.Nullable; +import java.util.List; + +/** + * Conversion for SQL operators that map 1-1 onto native functions. + */ +public class WindowSqlAggregate implements SqlAggregator +{ + private final SqlAggFunction operator; + + public WindowSqlAggregate(final SqlAggFunction operator) + { + this.operator = operator; + } + + @Override + public SqlAggFunction calciteFunction() + { + return operator; + } + + @Nullable + @Override + public Aggregation toDruidAggregation( + PlannerContext plannerContext, + RowSignature rowSignature, + VirtualColumnRegistry virtualColumnRegistry, + RexBuilder rexBuilder, + String name, + AggregateCall aggregateCall, + Project project, + List existingAggregations, + boolean finalizeAggregations + ) + { + throw new UOE("%s can only be used in a window function, this method shouldn't be called...", operator.getName()); + } +} diff --git a/sql/src/main/java/org/apache/druid/sql/calcite/planner/CalciteRulesManager.java b/sql/src/main/java/org/apache/druid/sql/calcite/planner/CalciteRulesManager.java index 088a9cb861b1..e2b91f7d79a8 100644 --- a/sql/src/main/java/org/apache/druid/sql/calcite/planner/CalciteRulesManager.java +++ b/sql/src/main/java/org/apache/druid/sql/calcite/planner/CalciteRulesManager.java @@ -206,6 +206,7 @@ public class CalciteRulesManager /** * Manages the rules for planning of SQL queries via Calcite. Also provides methods for extensions to provide custom * rules for planning. + * * @param extensionCalciteRuleProviderSet the set of custom rules coming from extensions */ @Inject @@ -265,11 +266,11 @@ public List druidConventionRuleSet(final PlannerContext plannerConte public List bindableConventionRuleSet(final PlannerContext plannerContext) { return ImmutableList.builder() - .addAll(baseRuleSet(plannerContext)) - .addAll(Bindables.RULES) - .addAll(DEFAULT_BINDABLE_RULES) - .add(AggregateReduceFunctionsRule.INSTANCE) - .build(); + .addAll(baseRuleSet(plannerContext)) + .addAll(Bindables.RULES) + .addAll(DEFAULT_BINDABLE_RULES) + .add(AggregateReduceFunctionsRule.INSTANCE) + .build(); } public List baseRuleSet(final PlannerContext plannerContext) diff --git a/sql/src/main/java/org/apache/druid/sql/calcite/planner/DruidOperatorTable.java b/sql/src/main/java/org/apache/druid/sql/calcite/planner/DruidOperatorTable.java index 2431fc4accb5..26bd698e17c6 100644 --- a/sql/src/main/java/org/apache/druid/sql/calcite/planner/DruidOperatorTable.java +++ b/sql/src/main/java/org/apache/druid/sql/calcite/planner/DruidOperatorTable.java @@ -54,6 +54,7 @@ import org.apache.druid.sql.calcite.expression.UnaryFunctionOperatorConversion; import org.apache.druid.sql.calcite.expression.UnaryPrefixOperatorConversion; import org.apache.druid.sql.calcite.expression.UnarySuffixOperatorConversion; +import org.apache.druid.sql.calcite.expression.WindowSqlAggregate; import org.apache.druid.sql.calcite.expression.builtin.ArrayAppendOperatorConversion; import org.apache.druid.sql.calcite.expression.builtin.ArrayConcatOperatorConversion; import org.apache.druid.sql.calcite.expression.builtin.ArrayConstructorOperatorConversion; @@ -135,6 +136,16 @@ public class DruidOperatorTable implements SqlOperatorTable // COUNT and APPROX_COUNT_DISTINCT are not here because they are added by SqlAggregationModule. private static final List STANDARD_AGGREGATORS = ImmutableList.builder() + .add(new WindowSqlAggregate(SqlStdOperatorTable.LAG)) + .add(new WindowSqlAggregate(SqlStdOperatorTable.LEAD)) + .add(new WindowSqlAggregate(SqlStdOperatorTable.FIRST_VALUE)) + .add(new WindowSqlAggregate(SqlStdOperatorTable.LAST_VALUE)) + .add(new WindowSqlAggregate(SqlStdOperatorTable.CUME_DIST)) + .add(new WindowSqlAggregate(SqlStdOperatorTable.DENSE_RANK)) + .add(new WindowSqlAggregate(SqlStdOperatorTable.NTILE)) + .add(new WindowSqlAggregate(SqlStdOperatorTable.PERCENT_RANK)) + .add(new WindowSqlAggregate(SqlStdOperatorTable.RANK)) + .add(new WindowSqlAggregate(SqlStdOperatorTable.ROW_NUMBER)) .add(new BuiltinApproxCountDistinctSqlAggregator()) .add(new AvgSqlAggregator()) .add(EarliestLatestAnySqlAggregator.EARLIEST) diff --git a/sql/src/main/java/org/apache/druid/sql/calcite/rel/DruidQuery.java b/sql/src/main/java/org/apache/druid/sql/calcite/rel/DruidQuery.java index 8c5aa588f708..f4601c03b37c 100644 --- a/sql/src/main/java/org/apache/druid/sql/calcite/rel/DruidQuery.java +++ b/sql/src/main/java/org/apache/druid/sql/calcite/rel/DruidQuery.java @@ -49,6 +49,7 @@ import org.apache.druid.query.DataSource; import org.apache.druid.query.JoinDataSource; import org.apache.druid.query.Query; +import org.apache.druid.query.QueryContext; import org.apache.druid.query.QueryDataSource; import org.apache.druid.query.aggregation.AggregatorFactory; import org.apache.druid.query.aggregation.LongMaxAggregatorFactory; @@ -61,6 +62,7 @@ import org.apache.druid.query.groupby.having.DimFilterHavingSpec; import org.apache.druid.query.groupby.orderby.DefaultLimitSpec; import org.apache.druid.query.groupby.orderby.OrderByColumnSpec; +import org.apache.druid.query.operator.WindowOperatorQuery; import org.apache.druid.query.ordering.StringComparator; import org.apache.druid.query.ordering.StringComparators; import org.apache.druid.query.planning.DataSourceAnalysis; @@ -97,7 +99,6 @@ import javax.annotation.Nonnull; import javax.annotation.Nullable; - import java.util.ArrayList; import java.util.Collections; import java.util.Comparator; @@ -124,10 +125,11 @@ public class DruidQuery /** * Maximum number of time-granular buckets that we allow for non-Druid tables. - * + *

    * Used by {@link #canUseQueryGranularity}. */ private static final int MAX_TIME_GRAINS_NON_DRUID_TABLE = 100000; + public static final String CTX_ENABLE_WINDOW_FNS = "windowsAreForClosers"; private final DataSource dataSource; private final PlannerContext plannerContext; @@ -144,6 +146,9 @@ public class DruidQuery @Nullable private final Sorting sorting; + @Nullable + private final Windowing windowing; + private final Query query; private final RowSignature outputRowSignature; private final RelDataType outputRowType; @@ -157,6 +162,7 @@ private DruidQuery( @Nullable final Projection selectProjection, @Nullable final Grouping grouping, @Nullable final Sorting sorting, + @Nullable final Windowing windowing, final RowSignature sourceRowSignature, final RelDataType outputRowType, final VirtualColumnRegistry virtualColumnRegistry @@ -168,9 +174,16 @@ private DruidQuery( this.selectProjection = selectProjection; this.grouping = grouping; this.sorting = sorting; + this.windowing = windowing; this.sourceRowSignature = sourceRowSignature; - this.outputRowSignature = computeOutputRowSignature(sourceRowSignature, selectProjection, grouping, sorting); + this.outputRowSignature = computeOutputRowSignature( + sourceRowSignature, + selectProjection, + grouping, + sorting, + windowing + ); this.outputRowType = Preconditions.checkNotNull(outputRowType, "outputRowType"); this.virtualColumnRegistry = Preconditions.checkNotNull(virtualColumnRegistry, "virtualColumnRegistry"); this.query = computeQuery(); @@ -200,6 +213,7 @@ public static DruidQuery fromPartialQuery( final Projection selectProjection; final Grouping grouping; final Sorting sorting; + final Windowing windowing; if (partialQuery.getWhereFilter() != null) { filter = Preconditions.checkNotNull( @@ -221,7 +235,7 @@ public static DruidQuery fromPartialQuery( computeSelectProjection( partialQuery, plannerContext, - computeOutputRowSignature(sourceRowSignature, null, null, null), + computeOutputRowSignature(sourceRowSignature, null, null, null, null), virtualColumnRegistry ) ); @@ -234,7 +248,7 @@ public static DruidQuery fromPartialQuery( computeGrouping( partialQuery, plannerContext, - computeOutputRowSignature(sourceRowSignature, null, null, null), + computeOutputRowSignature(sourceRowSignature, null, null, null, null), virtualColumnRegistry, rexBuilder, finalizeAggregations @@ -249,7 +263,7 @@ public static DruidQuery fromPartialQuery( computeSorting( partialQuery, plannerContext, - computeOutputRowSignature(sourceRowSignature, selectProjection, grouping, null), + computeOutputRowSignature(sourceRowSignature, selectProjection, grouping, null, null), // When sorting follows grouping, virtual columns cannot be used partialQuery.getAggregate() != null ? null : virtualColumnRegistry ) @@ -258,6 +272,25 @@ public static DruidQuery fromPartialQuery( sorting = null; } + if (partialQuery.getWindow() != null) { + final QueryContext queryContext = plannerContext.queryContext(); + if (queryContext.getBoolean(CTX_ENABLE_WINDOW_FNS, false)) { + windowing = Preconditions.checkNotNull( + Windowing.fromCalciteStuff( + partialQuery, + plannerContext, + sourceRowSignature, // Plans immediately after Scan, so safe to use the row signature from scan + rexBuilder + ) + ); + } else { + plannerContext.setPlanningError("Windowing Not Currently Supported"); + throw new CannotBuildQueryException("Windowing Not Currently Supported"); + } + } else { + windowing = null; + } + return new DruidQuery( dataSource, plannerContext, @@ -265,6 +298,7 @@ public static DruidQuery fromPartialQuery( selectProjection, grouping, sorting, + windowing, sourceRowSignature, outputRowType, virtualColumnRegistry @@ -404,9 +438,7 @@ private static Grouping computeGrouping( * @param plannerContext planner context * @param rowSignature source row signature * @param virtualColumnRegistry re-usable virtual column references - * * @return dimensions - * * @throws CannotBuildQueryException if dimensions cannot be computed */ private static List computeDimensions( @@ -438,7 +470,10 @@ private static List computeDimensions( final ColumnType outputType = Calcites.getColumnTypeForRelDataType(dataType); if (Types.isNullOr(outputType, ValueType.COMPLEX)) { // Can't group on unknown or COMPLEX types. - plannerContext.setPlanningError("SQL requires a group-by on a column of type %s that is unsupported.", outputType); + plannerContext.setPlanningError( + "SQL requires a group-by on a column of type %s that is unsupported.", + outputType + ); throw new CannotBuildQueryException(aggregate, rexNode); } @@ -511,9 +546,7 @@ private static Subtotals computeSubtotals( * @param finalizeAggregations true if this query should include explicit finalization for all of its * aggregators, where required. Useful for subqueries where Druid's native query layer * does not do this automatically. - * * @return aggregations - * * @throws CannotBuildQueryException if dimensions cannot be computed */ private static List computeAggregations( @@ -631,10 +664,13 @@ private static RowSignature computeOutputRowSignature( final RowSignature sourceRowSignature, @Nullable final Projection selectProjection, @Nullable final Grouping grouping, - @Nullable final Sorting sorting + @Nullable final Sorting sorting, + @Nullable final Windowing windowing ) { - if (sorting != null && sorting.getProjection() != null) { + if (windowing != null) { + return windowing.getSignature(); + } else if (sorting != null && sorting.getProjection() != null) { return sorting.getProjection().getOutputRowSignature(); } else if (grouping != null) { // Sanity check: cannot have both "grouping" and "selectProjection". @@ -791,7 +827,7 @@ private static Filtration toFiltration(DimFilter filter, VirtualColumnRegistry v /** * Whether the provided combination of dataSource, filtration, and queryGranularity is safe to use in queries. - * + *

    * Necessary because some combinations are unsafe, mainly because they would lead to the creation of too many * time-granular buckets during query processing. */ @@ -870,6 +906,11 @@ public Query getQuery() */ private Query computeQuery() { + if (windowing != null) { + // Windowing can only be handled by window queries. + return toWindowQuery(); + } + if (dataSource instanceof QueryDataSource) { // If there is a subquery, then we prefer the outer query to be a groupBy if possible, since this potentially // enables more efficient execution. (The groupBy query toolchest can handle some subqueries by itself, without @@ -1312,6 +1353,26 @@ private GroupByQuery toGroupByQuery() return query.withOverriddenContext(theContext); } + /** + * Return this query as a {@link WindowOperatorQuery}, or null if this query cannot be run that way. + * + * @return query or null + */ + @Nullable + private WindowOperatorQuery toWindowQuery() + { + if (windowing == null) { + return null; + } + + return new WindowOperatorQuery( + dataSource, + plannerContext.queryContextMap(), + windowing.getSignature(), + windowing.getOperators() + ); + } + /** * Return this query as a Scan query, or null if this query is not compatible with Scan. * diff --git a/sql/src/main/java/org/apache/druid/sql/calcite/rel/PartialDruidQuery.java b/sql/src/main/java/org/apache/druid/sql/calcite/rel/PartialDruidQuery.java index 15bcd5d4bdac..bdd4a4f735fa 100644 --- a/sql/src/main/java/org/apache/druid/sql/calcite/rel/PartialDruidQuery.java +++ b/sql/src/main/java/org/apache/druid/sql/calcite/rel/PartialDruidQuery.java @@ -28,6 +28,7 @@ import org.apache.calcite.rel.core.Project; import org.apache.calcite.rel.core.RelFactories; import org.apache.calcite.rel.core.Sort; +import org.apache.calcite.rel.core.Window; import org.apache.calcite.rel.type.RelDataType; import org.apache.calcite.rex.RexBuilder; import org.apache.calcite.rex.RexNode; @@ -58,24 +59,28 @@ public class PartialDruidQuery private final Project aggregateProject; private final Sort sort; private final Project sortProject; + private final Window window; public enum Stage { // SCAN must be present on all queries. SCAN, - // WHERE_FILTER, SELECT_PROJECT may be present on any query. + // WHERE_FILTER, SELECT_PROJECT may be present on any query, except ones with WINDOW. WHERE_FILTER, SELECT_PROJECT, - // AGGREGATE, HAING_FILTER, AGGREGATE_PROJECT can only be present on aggregating queries. + // AGGREGATE, HAVING_FILTER, AGGREGATE_PROJECT can only be present on non-WINDOW aggregating queries. AGGREGATE, HAVING_FILTER, AGGREGATE_PROJECT, - // SORT, SORT_PROJECT may be present on any query. + // SORT, SORT_PROJECT may be present on any query, except ones with WINDOW. SORT, - SORT_PROJECT + SORT_PROJECT, + + // WINDOW may be present only together with SCAN. + WINDOW } private PartialDruidQuery( @@ -87,7 +92,8 @@ private PartialDruidQuery( final Project aggregateProject, final Filter havingFilter, final Sort sort, - final Project sortProject + final Project sortProject, + final Window window ) { this.builderSupplier = Preconditions.checkNotNull(builderSupplier, "builderSupplier"); @@ -99,6 +105,7 @@ private PartialDruidQuery( this.havingFilter = havingFilter; this.sort = sort; this.sortProject = sortProject; + this.window = window; } public static PartialDruidQuery create(final RelNode scanRel) @@ -107,7 +114,7 @@ public static PartialDruidQuery create(final RelNode scanRel) scanRel.getCluster(), scanRel.getTable() != null ? scanRel.getTable().getRelOptSchema() : null ); - return new PartialDruidQuery(builderSupplier, scanRel, null, null, null, null, null, null, null); + return new PartialDruidQuery(builderSupplier, scanRel, null, null, null, null, null, null, null, null); } public RelNode getScan() @@ -150,6 +157,11 @@ public Project getSortProject() return sortProject; } + public Window getWindow() + { + return window; + } + public PartialDruidQuery withWhereFilter(final Filter newWhereFilter) { validateStage(Stage.WHERE_FILTER); @@ -162,7 +174,8 @@ public PartialDruidQuery withWhereFilter(final Filter newWhereFilter) aggregateProject, havingFilter, sort, - sortProject + sortProject, + window ); } @@ -204,7 +217,8 @@ public PartialDruidQuery withSelectProject(final Project newSelectProject) aggregateProject, havingFilter, sort, - sortProject + sortProject, + window ); } @@ -220,7 +234,8 @@ public PartialDruidQuery withAggregate(final Aggregate newAggregate) aggregateProject, havingFilter, sort, - sortProject + sortProject, + window ); } @@ -236,7 +251,8 @@ public PartialDruidQuery withHavingFilter(final Filter newHavingFilter) aggregateProject, newHavingFilter, sort, - sortProject + sortProject, + window ); } @@ -252,7 +268,8 @@ public PartialDruidQuery withAggregateProject(final Project newAggregateProject) newAggregateProject, havingFilter, sort, - sortProject + sortProject, + window ); } @@ -268,7 +285,8 @@ public PartialDruidQuery withSort(final Sort newSort) aggregateProject, havingFilter, newSort, - sortProject + sortProject, + window ); } @@ -284,7 +302,25 @@ public PartialDruidQuery withSortProject(final Project newSortProject) aggregateProject, havingFilter, sort, - newSortProject + newSortProject, + window + ); + } + + public PartialDruidQuery withWindow(final Window newWindow) + { + validateStage(Stage.WINDOW); + return new PartialDruidQuery( + builderSupplier, + scan, + whereFilter, + selectProject, + aggregate, + aggregateProject, + havingFilter, + sort, + sortProject, + newWindow ); } @@ -341,7 +377,10 @@ public boolean canAccept(final Stage stage) { final Stage currentStage = stage(); - if (currentStage == Stage.SELECT_PROJECT && stage == Stage.SELECT_PROJECT) { + if (stage == Stage.WINDOW) { + // Special case: WINDOW can only be provided along with SCAN. + return currentStage == Stage.SCAN; + } else if (currentStage == Stage.SELECT_PROJECT && stage == Stage.SELECT_PROJECT) { // Special case: allow layering SELECT_PROJECT on top of SELECT_PROJECT. Calcite's builtin rules cannot // always collapse these, so we have to (one example: testSemiJoinWithOuterTimeExtract). See // withSelectProject for the code here that handles this. @@ -352,12 +391,9 @@ public boolean canAccept(final Stage stage) } else if (stage.compareTo(Stage.AGGREGATE) > 0 && stage.compareTo(Stage.SORT) < 0 && aggregate == null) { // Cannot do post-aggregation stages without an aggregation. return false; - } else if (stage.compareTo(Stage.SORT) > 0 && sort == null) { - // Cannot do post-sort stages without a sort. - return false; } else { - // Looks good. - return true; + // If we are after the SORT phase, make sure we have a sort... + return stage.compareTo(Stage.SORT) <= 0 || sort != null; } } @@ -370,7 +406,9 @@ public boolean canAccept(final Stage stage) @SuppressWarnings("VariableNotUsedInsideIf") public Stage stage() { - if (sortProject != null) { + if (window != null) { + return Stage.WINDOW; + } else if (sortProject != null) { return Stage.SORT_PROJECT; } else if (sort != null) { return Stage.SORT; @@ -399,6 +437,8 @@ public RelNode leafRel() final Stage currentStage = stage(); switch (currentStage) { + case WINDOW: + return window; case SORT_PROJECT: return sortProject; case SORT: diff --git a/sql/src/main/java/org/apache/druid/sql/calcite/rel/Windowing.java b/sql/src/main/java/org/apache/druid/sql/calcite/rel/Windowing.java new file mode 100644 index 000000000000..5cbffc6d2a5b --- /dev/null +++ b/sql/src/main/java/org/apache/druid/sql/calcite/rel/Windowing.java @@ -0,0 +1,298 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.sql.calcite.rel; + +import com.google.common.base.Preconditions; +import com.google.common.collect.ImmutableMap; +import com.google.common.collect.Iterables; +import org.apache.calcite.rel.RelFieldCollation; +import org.apache.calcite.rel.core.AggregateCall; +import org.apache.calcite.rel.core.Project; +import org.apache.calcite.rel.core.Window; +import org.apache.calcite.rex.RexBuilder; +import org.apache.calcite.rex.RexLiteral; +import org.apache.calcite.rex.RexNode; +import org.apache.druid.java.util.common.ISE; +import org.apache.druid.query.aggregation.AggregatorFactory; +import org.apache.druid.query.operator.NaivePartitioningOperatorFactory; +import org.apache.druid.query.operator.OperatorFactory; +import org.apache.druid.query.operator.WindowOperatorFactory; +import org.apache.druid.query.operator.window.ComposingProcessor; +import org.apache.druid.query.operator.window.Processor; +import org.apache.druid.query.operator.window.WindowAggregateProcessor; +import org.apache.druid.query.operator.window.ranking.WindowCumeDistProcessor; +import org.apache.druid.query.operator.window.ranking.WindowDenseRankProcessor; +import org.apache.druid.query.operator.window.ranking.WindowPercentileProcessor; +import org.apache.druid.query.operator.window.ranking.WindowRankProcessor; +import org.apache.druid.query.operator.window.ranking.WindowRowNumberProcessor; +import org.apache.druid.query.operator.window.value.WindowFirstProcessor; +import org.apache.druid.query.operator.window.value.WindowLastProcessor; +import org.apache.druid.query.operator.window.value.WindowOffsetProcessor; +import org.apache.druid.segment.column.RowSignature; +import org.apache.druid.sql.calcite.aggregation.Aggregation; +import org.apache.druid.sql.calcite.expression.DruidExpression; +import org.apache.druid.sql.calcite.expression.Expressions; +import org.apache.druid.sql.calcite.planner.Calcites; +import org.apache.druid.sql.calcite.planner.PlannerContext; +import org.apache.druid.sql.calcite.rule.GroupByRules; +import org.apache.druid.sql.calcite.table.RowSignatures; + +import javax.annotation.Nonnull; +import java.util.ArrayList; +import java.util.Arrays; +import java.util.Collections; +import java.util.List; + +/** + * Maps onto a {@link org.apache.druid.query.operator.WindowOperatorQuery}. + */ +public class Windowing +{ + private static final ImmutableMap KNOWN_WINDOW_FNS = ImmutableMap + .builder() + .put("LAG", (agg) -> new WindowOffsetProcessor(agg.getColumn(0), agg.getOutputName(), -agg.getConstantInt(1))) + .put("LEAD", (agg) -> new WindowOffsetProcessor(agg.getColumn(0), agg.getOutputName(), agg.getConstantInt(1))) + .put("FIRST_VALUE", (agg) -> new WindowFirstProcessor(agg.getColumn(0), agg.getOutputName())) + .put("LAST_VALUE", (agg) -> new WindowLastProcessor(agg.getColumn(0), agg.getOutputName())) + .put("CUME_DIST", (agg) -> new WindowCumeDistProcessor(agg.getOrderingColumns(), agg.getOutputName())) + .put("DENSE_RANK", (agg) -> new WindowDenseRankProcessor(agg.getOrderingColumns(), agg.getOutputName())) + .put("NTILE", (agg) -> new WindowPercentileProcessor(agg.getOutputName(), agg.getConstantInt(0))) + .put("PERCENT_RANK", (agg) -> new WindowRankProcessor(agg.getOrderingColumns(), agg.getOutputName(), true)) + .put("RANK", (agg) -> new WindowRankProcessor(agg.getOrderingColumns(), agg.getOutputName(), false)) + .put("ROW_NUMBER", (agg) -> new WindowRowNumberProcessor(agg.getOutputName())) + .build(); + private final List ops; + + @Nonnull + public static Windowing fromCalciteStuff( + final PartialDruidQuery partialQuery, + final PlannerContext plannerContext, + final RowSignature rowSignature, + final RexBuilder rexBuilder + ) + { + final Window window = Preconditions.checkNotNull(partialQuery.getWindow(), "window"); + + // TODO(gianm): insert sorts and split the groups up at the rule stage; by this time, we assume there's one + // window and the dataset is already sorted appropriately. + if (window.groups.size() != 1) { + plannerContext.setPlanningError("Multiple windows are not supported"); + throw new CannotBuildQueryException(window); + } + final Window.Group group = Iterables.getOnlyElement(window.groups); + + // Window. + // TODO(gianm): Validate order-by keys instead of ignoring them. + + final List partitionColumns = new ArrayList<>(); + for (int groupKey : group.keys) { + partitionColumns.add(rowSignature.getColumnName(groupKey)); + } + + // Frame. + // TODO(gianm): Validate ROWS vs RANGE instead of ignoring it. + // TODO(gianm): Support various other kinds of frames. + if (!group.lowerBound.isUnbounded()) { + plannerContext.setPlanningError("Lower bound [%s] is not supported", group.upperBound); + throw new CannotBuildQueryException(window); + } + + final boolean cumulative; + if (group.upperBound.isUnbounded()) { + cumulative = false; + } else if (group.upperBound.isCurrentRow()) { + cumulative = true; + } else { + plannerContext.setPlanningError("Upper bound [%s] is not supported", group.upperBound); + throw new CannotBuildQueryException(window); + } + + // Aggregations. + final String outputNamePrefix = Calcites.findUnusedPrefixForDigits("w", rowSignature.getColumnNames()); + final List aggregateCalls = group.getAggregateCalls(window); + + final List processors = new ArrayList<>(); + final List aggregations = new ArrayList<>(); + final List expectedOutputColumns = new ArrayList<>(rowSignature.getColumnNames()); + + for (int i = 0; i < aggregateCalls.size(); i++) { + final String aggName = outputNamePrefix + i; + expectedOutputColumns.add(aggName); + + final AggregateCall aggCall = aggregateCalls.get(i); + + ProcessorMaker maker = KNOWN_WINDOW_FNS.get(aggCall.getAggregation().getName()); + if (maker == null) { + final Aggregation aggregation = GroupByRules.translateAggregateCall( + plannerContext, + rowSignature, + null, + rexBuilder, + partialQuery.getSelectProject(), + Collections.emptyList(), + aggName, + aggCall, + false // TODO: finalize in a separate operator + ); + + if (aggregation == null + || aggregation.getPostAggregator() != null + || aggregation.getAggregatorFactories().size() != 1) { + if (null == plannerContext.getPlanningError()) { + plannerContext.setPlanningError("Aggregation [%s] is not supported", aggCall); + } + throw new CannotBuildQueryException(window, aggCall); + } + + aggregations.add(Iterables.getOnlyElement(aggregation.getAggregatorFactories())); + } else { + processors.add(maker.make( + new WindowAggregate( + aggName, + aggCall, + rowSignature, + plannerContext, + partialQuery.getSelectProject(), + window.constants, + group + ) + )); + } + } + + if (!aggregations.isEmpty()) { + if (cumulative) { + processors.add(new WindowAggregateProcessor(null, aggregations)); + } else { + processors.add(new WindowAggregateProcessor(aggregations, null)); + } + } + + if (processors.isEmpty()) { + throw new ISE("No processors from Window[%s], why was this code called?", window); + } + + final List ops = Arrays.asList( + new NaivePartitioningOperatorFactory(partitionColumns), + new WindowOperatorFactory( + processors.size() == 1 ? + processors.get(0) : new ComposingProcessor(processors.toArray(new Processor[0])) + ) + ); + + return new Windowing( + RowSignatures.fromRelDataType(expectedOutputColumns, window.getRowType()), + ops + ); + } + + private final RowSignature signature; + + public Windowing( + final RowSignature signature, + List ops + ) + { + this.signature = signature; + this.ops = ops; + } + + public RowSignature getSignature() + { + return signature; + } + + public List getOperators() + { + return ops; + } + + private interface ProcessorMaker + { + Processor make(WindowAggregate agg); + } + + private static class WindowAggregate + { + private final String outputName; + private final AggregateCall call; + private final RowSignature sig; + private final PlannerContext context; + private final Project project; + private final List constants; + private final Window.Group group; + + private WindowAggregate( + String outputName, + AggregateCall call, + RowSignature sig, + PlannerContext context, + Project project, + List constants, + Window.Group group + ) + { + this.outputName = outputName; + this.call = call; + this.sig = sig; + this.context = context; + this.project = project; + this.constants = constants; + this.group = group; + + if (project != null) { + throw new ISE("Suddenly, the project[%s] is no longer null, the code might need to change.", project); + } + } + + public String getOutputName() + { + return outputName; + } + + public ArrayList getOrderingColumns() + { + final List fields = group.orderKeys.getFieldCollations(); + ArrayList retVal = new ArrayList<>(fields.size()); + for (RelFieldCollation field : fields) { + retVal.add(sig.getColumnName(field.getFieldIndex())); + } + return retVal; + } + + public String getColumn(int argPosition) + { + RexNode columnArgument = Expressions.fromFieldAccess(sig, project, call.getArgList().get(argPosition)); + final DruidExpression expression = Expressions.toDruidExpression(context, sig, columnArgument); + return expression.getDirectColumn(); + } + + public RexLiteral getConstantArgument(int argPosition) + { + final Integer constantIndex = call.getArgList().get(argPosition) - sig.size(); + return constants.get(constantIndex); + } + + public int getConstantInt(int argPosition) + { + return ((Number) getConstantArgument(argPosition).getValue()).intValue(); + } + } +} diff --git a/sql/src/main/java/org/apache/druid/sql/calcite/rule/DruidRules.java b/sql/src/main/java/org/apache/druid/sql/calcite/rule/DruidRules.java index 113f37d896fb..8316b1e9868d 100644 --- a/sql/src/main/java/org/apache/druid/sql/calcite/rule/DruidRules.java +++ b/sql/src/main/java/org/apache/druid/sql/calcite/rule/DruidRules.java @@ -28,12 +28,15 @@ import org.apache.calcite.rel.core.Filter; import org.apache.calcite.rel.core.Project; import org.apache.calcite.rel.core.Sort; +import org.apache.calcite.rel.core.Window; import org.apache.druid.java.util.common.StringUtils; import org.apache.druid.sql.calcite.planner.PlannerContext; import org.apache.druid.sql.calcite.rel.DruidOuterQueryRel; +import org.apache.druid.sql.calcite.rel.DruidQuery; import org.apache.druid.sql.calcite.rel.DruidRel; import org.apache.druid.sql.calcite.rel.PartialDruidQuery; +import java.util.ArrayList; import java.util.List; import java.util.function.BiFunction; import java.util.function.Predicate; @@ -50,51 +53,59 @@ private DruidRules() public static List rules(PlannerContext plannerContext) { - return ImmutableList.of( - new DruidQueryRule<>( - Filter.class, - PartialDruidQuery.Stage.WHERE_FILTER, - PartialDruidQuery::withWhereFilter - ), - new DruidQueryRule<>( - Project.class, - PartialDruidQuery.Stage.SELECT_PROJECT, - PartialDruidQuery::withSelectProject - ), - new DruidQueryRule<>( - Aggregate.class, - PartialDruidQuery.Stage.AGGREGATE, - PartialDruidQuery::withAggregate - ), - new DruidQueryRule<>( - Project.class, - PartialDruidQuery.Stage.AGGREGATE_PROJECT, - PartialDruidQuery::withAggregateProject - ), - new DruidQueryRule<>( - Filter.class, - PartialDruidQuery.Stage.HAVING_FILTER, - PartialDruidQuery::withHavingFilter - ), - new DruidQueryRule<>( - Sort.class, - PartialDruidQuery.Stage.SORT, - PartialDruidQuery::withSort - ), - new DruidQueryRule<>( - Project.class, - PartialDruidQuery.Stage.SORT_PROJECT, - PartialDruidQuery::withSortProject - ), - DruidOuterQueryRule.AGGREGATE, - DruidOuterQueryRule.WHERE_FILTER, - DruidOuterQueryRule.SELECT_PROJECT, - DruidOuterQueryRule.SORT, - new DruidUnionRule(plannerContext), - new DruidUnionDataSourceRule(plannerContext), - DruidSortUnionRule.instance(), - DruidJoinRule.instance(plannerContext) + final ArrayList retVal = new ArrayList<>( + ImmutableList.of( + new DruidQueryRule<>( + Filter.class, + PartialDruidQuery.Stage.WHERE_FILTER, + PartialDruidQuery::withWhereFilter + ), + new DruidQueryRule<>( + Project.class, + PartialDruidQuery.Stage.SELECT_PROJECT, + PartialDruidQuery::withSelectProject + ), + new DruidQueryRule<>( + Aggregate.class, + PartialDruidQuery.Stage.AGGREGATE, + PartialDruidQuery::withAggregate + ), + new DruidQueryRule<>( + Project.class, + PartialDruidQuery.Stage.AGGREGATE_PROJECT, + PartialDruidQuery::withAggregateProject + ), + new DruidQueryRule<>( + Filter.class, + PartialDruidQuery.Stage.HAVING_FILTER, + PartialDruidQuery::withHavingFilter + ), + new DruidQueryRule<>( + Sort.class, + PartialDruidQuery.Stage.SORT, + PartialDruidQuery::withSort + ), + new DruidQueryRule<>( + Project.class, + PartialDruidQuery.Stage.SORT_PROJECT, + PartialDruidQuery::withSortProject + ), + DruidOuterQueryRule.AGGREGATE, + DruidOuterQueryRule.WHERE_FILTER, + DruidOuterQueryRule.SELECT_PROJECT, + DruidOuterQueryRule.SORT, + new DruidUnionRule(plannerContext), + new DruidUnionDataSourceRule(plannerContext), + DruidSortUnionRule.instance(), + DruidJoinRule.instance(plannerContext) + ) ); + + if (plannerContext.queryContext().getBoolean(DruidQuery.CTX_ENABLE_WINDOW_FNS, false)) { + retVal.add(new DruidQueryRule<>(Window.class, PartialDruidQuery.Stage.WINDOW, PartialDruidQuery::withWindow)); + retVal.add(DruidOuterQueryRule.WINDOW); + } + return retVal; } public static class DruidQueryRule extends RelOptRule @@ -228,6 +239,28 @@ public void onMatch(final RelOptRuleCall call) } }; + public static final RelOptRule WINDOW = new DruidOuterQueryRule( + operand(Window.class, operandJ(DruidRel.class, null, CAN_BUILD_ON, any())), + "WINDOW" + ) + { + @Override + public void onMatch(final RelOptRuleCall call) + { + final Window window = call.rel(0); + final DruidRel druidRel = call.rel(1); + + final DruidOuterQueryRel outerQueryRel = DruidOuterQueryRel.create( + druidRel, + PartialDruidQuery.create(druidRel.getPartialDruidQuery().leafRel()) + .withWindow(window) + ); + if (outerQueryRel.isValidDruidQuery()) { + call.transformTo(outerQueryRel); + } + } + }; + public DruidOuterQueryRule(final RelOptRuleOperand op, final String description) { super(op, StringUtils.format("%s(%s)", DruidOuterQueryRel.class.getSimpleName(), description)); diff --git a/sql/src/main/java/org/apache/druid/sql/calcite/rule/GroupByRules.java b/sql/src/main/java/org/apache/druid/sql/calcite/rule/GroupByRules.java index cf596ce7f546..549c7899730a 100644 --- a/sql/src/main/java/org/apache/druid/sql/calcite/rule/GroupByRules.java +++ b/sql/src/main/java/org/apache/druid/sql/calcite/rule/GroupByRules.java @@ -34,6 +34,7 @@ import org.apache.druid.sql.calcite.planner.PlannerContext; import org.apache.druid.sql.calcite.rel.VirtualColumnRegistry; +import javax.annotation.Nullable; import java.util.ArrayList; import java.util.List; import java.util.Set; @@ -54,7 +55,7 @@ private GroupByRules() public static Aggregation translateAggregateCall( final PlannerContext plannerContext, final RowSignature rowSignature, - final VirtualColumnRegistry virtualColumnRegistry, + @Nullable final VirtualColumnRegistry virtualColumnRegistry, final RexBuilder rexBuilder, final Project project, final List existingAggregations, diff --git a/sql/src/main/java/org/apache/druid/sql/calcite/run/NativeQueryMaker.java b/sql/src/main/java/org/apache/druid/sql/calcite/run/NativeQueryMaker.java index a9fba5ee3580..258a744827a4 100644 --- a/sql/src/main/java/org/apache/druid/sql/calcite/run/NativeQueryMaker.java +++ b/sql/src/main/java/org/apache/druid/sql/calcite/run/NativeQueryMaker.java @@ -240,9 +240,8 @@ private QueryResponse mapResultSequence( mapping[i] = idx; } - //noinspection unchecked final Sequence sequence = toolChest.resultsAsArrays(query, results.getResults()); - return new QueryResponse( + return new QueryResponse<>( Sequences.map( sequence, array -> { diff --git a/sql/src/test/java/org/apache/druid/sql/avatica/DruidAvaticaHandlerTest.java b/sql/src/test/java/org/apache/druid/sql/avatica/DruidAvaticaHandlerTest.java index 9db98505dc1f..748ed20ce3c5 100644 --- a/sql/src/test/java/org/apache/druid/sql/avatica/DruidAvaticaHandlerTest.java +++ b/sql/src/test/java/org/apache/druid/sql/avatica/DruidAvaticaHandlerTest.java @@ -559,6 +559,12 @@ public void testDatabaseMetaDataTables() throws SQLException Pair.of("TABLE_NAME", CalciteTests.USERVISITDATASOURCE), Pair.of("TABLE_SCHEM", "druid"), Pair.of("TABLE_TYPE", "TABLE") + ), + row( + Pair.of("TABLE_CAT", "druid"), + Pair.of("TABLE_NAME", "wikipedia"), + Pair.of("TABLE_SCHEM", "druid"), + Pair.of("TABLE_TYPE", "TABLE") ) ), getRows( @@ -633,6 +639,12 @@ public void testDatabaseMetaDataTablesAsSuperuser() throws SQLException Pair.of("TABLE_NAME", CalciteTests.USERVISITDATASOURCE), Pair.of("TABLE_SCHEM", "druid"), Pair.of("TABLE_TYPE", "TABLE") + ), + row( + Pair.of("TABLE_CAT", "druid"), + Pair.of("TABLE_NAME", "wikipedia"), + Pair.of("TABLE_SCHEM", "druid"), + Pair.of("TABLE_TYPE", "TABLE") ) ), getRows( diff --git a/sql/src/test/java/org/apache/druid/sql/calcite/BaseCalciteQueryTest.java b/sql/src/test/java/org/apache/druid/sql/calcite/BaseCalciteQueryTest.java index fada41896009..7254e84c51f8 100644 --- a/sql/src/test/java/org/apache/druid/sql/calcite/BaseCalciteQueryTest.java +++ b/sql/src/test/java/org/apache/druid/sql/calcite/BaseCalciteQueryTest.java @@ -114,7 +114,6 @@ import org.junit.rules.TemporaryFolder; import javax.annotation.Nullable; - import java.io.IOException; import java.util.Arrays; import java.util.HashMap; diff --git a/sql/src/test/java/org/apache/druid/sql/calcite/CalciteQueryTest.java b/sql/src/test/java/org/apache/druid/sql/calcite/CalciteQueryTest.java index 516e363f9bb3..c9df143d9ca8 100644 --- a/sql/src/test/java/org/apache/druid/sql/calcite/CalciteQueryTest.java +++ b/sql/src/test/java/org/apache/druid/sql/calcite/CalciteQueryTest.java @@ -218,6 +218,7 @@ public void testInformationSchemaTables() .add(new Object[]{"druid", CalciteTests.SOME_DATASOURCE, "TABLE", "NO", "NO"}) .add(new Object[]{"druid", CalciteTests.SOMEXDATASOURCE, "TABLE", "NO", "NO"}) .add(new Object[]{"druid", CalciteTests.USERVISITDATASOURCE, "TABLE", "NO", "NO"}) + .add(new Object[]{"druid", "wikipedia", "TABLE", "NO", "NO"}) .add(new Object[]{"INFORMATION_SCHEMA", "COLUMNS", "SYSTEM_TABLE", "NO", "NO"}) .add(new Object[]{"INFORMATION_SCHEMA", "SCHEMATA", "SYSTEM_TABLE", "NO", "NO"}) .add(new Object[]{"INFORMATION_SCHEMA", "TABLES", "SYSTEM_TABLE", "NO", "NO"}) @@ -254,6 +255,7 @@ public void testInformationSchemaTables() .add(new Object[]{"druid", CalciteTests.SOME_DATASOURCE, "TABLE", "NO", "NO"}) .add(new Object[]{"druid", CalciteTests.SOMEXDATASOURCE, "TABLE", "NO", "NO"}) .add(new Object[]{"druid", CalciteTests.USERVISITDATASOURCE, "TABLE", "NO", "NO"}) + .add(new Object[]{"druid", "wikipedia", "TABLE", "NO", "NO"}) .add(new Object[]{"INFORMATION_SCHEMA", "COLUMNS", "SYSTEM_TABLE", "NO", "NO"}) .add(new Object[]{"INFORMATION_SCHEMA", "SCHEMATA", "SYSTEM_TABLE", "NO", "NO"}) .add(new Object[]{"INFORMATION_SCHEMA", "TABLES", "SYSTEM_TABLE", "NO", "NO"}) diff --git a/sql/src/test/java/org/apache/druid/sql/calcite/CalciteWindowQueryTest.java b/sql/src/test/java/org/apache/druid/sql/calcite/CalciteWindowQueryTest.java new file mode 100644 index 000000000000..0a12eb8c88ea --- /dev/null +++ b/sql/src/test/java/org/apache/druid/sql/calcite/CalciteWindowQueryTest.java @@ -0,0 +1,229 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.sql.calcite; + +import com.fasterxml.jackson.annotation.JsonProperty; +import com.fasterxml.jackson.core.JsonProcessingException; +import com.fasterxml.jackson.databind.ObjectMapper; +import com.fasterxml.jackson.dataformat.yaml.YAMLFactory; +import com.google.common.collect.ImmutableMap; +import junitparams.JUnitParamsRunner; +import junitparams.Parameters; +import junitparams.naming.TestCaseName; +import org.apache.druid.common.config.NullHandling; +import org.apache.druid.jackson.DefaultObjectMapper; +import org.apache.druid.java.util.common.ISE; +import org.apache.druid.java.util.common.RE; +import org.apache.druid.query.Query; +import org.apache.druid.query.operator.OperatorFactory; +import org.apache.druid.query.operator.WindowOperatorQuery; +import org.apache.druid.segment.column.ColumnType; +import org.apache.druid.segment.column.RowSignature; +import org.junit.AfterClass; +import org.junit.Assert; +import org.junit.Test; +import org.junit.runner.RunWith; + +import java.io.File; +import java.io.IOException; +import java.net.URL; +import java.util.Arrays; +import java.util.List; +import java.util.Locale; +import java.util.Objects; +import java.util.concurrent.atomic.AtomicLong; +import java.util.function.Function; +import java.util.regex.Pattern; + +@RunWith(JUnitParamsRunner.class) +public class CalciteWindowQueryTest extends BaseCalciteQueryTest +{ + + public static final boolean DUMP_EXPECTED_RESULTS = Boolean.parseBoolean( + System.getProperty("druid.tests.sql.dumpExpectedResults") + ); + + static { + NullHandling.initializeForTests(); + } + + private static final ObjectMapper YAML_JACKSON = new DefaultObjectMapper(new YAMLFactory(), "tests"); + + private static final AtomicLong EXPECTED_TESTS = new AtomicLong(); + private static final AtomicLong TEST_COUNTER = new AtomicLong(); + + public Object parametersForWindowQueryTest() throws Exception + { + final URL windowFolderUrl = ClassLoader.getSystemResource("calcite/tests/window"); + File windowFolder = new File(windowFolderUrl.toURI()); + + final File[] listedFiles = windowFolder.listFiles( + pathname -> pathname.getName().toLowerCase(Locale.ROOT).endsWith(".sqltest") + ); + EXPECTED_TESTS.set(listedFiles.length); + + Pattern matcher = Pattern.compile(".*"); + + return Arrays + .stream(Objects.requireNonNull(listedFiles)) + .map(File::getName) + .filter(matcher.asPredicate()) + .toArray(); + } + + @AfterClass + public static void testRanAllTests() + { + // This validation exists to catch issues with the filter Pattern accidentally getting checked in. It validates + // that we ran all of the tests from the directory. If this is failing, most likely, the filter Pattern in + // parametersForWindowQueryTest accidentally got checked in as something other than ".*" + Assert.assertEquals(EXPECTED_TESTS.get(), TEST_COUNTER.get()); + } + + @Test + @Parameters(method = "parametersForWindowQueryTest") + @SuppressWarnings("unchecked") + @TestCaseName("{0}") + public void windowQueryTest(String filename) throws IOException + { + final Function stringManipulator; + if (NullHandling.sqlCompatible()) { + stringManipulator = s -> "".equals(s) ? null : s; + } else { + stringManipulator = Function.identity(); + } + + TEST_COUNTER.incrementAndGet(); + final URL systemResource = ClassLoader.getSystemResource("calcite/tests/window/" + filename); + + final Object objectFromYaml = YAML_JACKSON.readValue(systemResource.openStream(), Object.class); + + final ObjectMapper queryJackson = queryFramework().queryJsonMapper(); + final WindowQueryTestInputClass input = queryJackson.convertValue(objectFromYaml, WindowQueryTestInputClass.class); + + Function jacksonToString = value -> { + try { + return queryJackson.writeValueAsString(value); + } + catch (JsonProcessingException e) { + throw new RE(e); + } + }; + + if ("operatorValidation".equals(input.type)) { + testBuilder() + .skipVectorize(true) + .sql(input.sql) + .queryContext(ImmutableMap.of("windowsAreForClosers", true)) + .addCustomVerification(QueryVerification.ofResults(results -> { + if (results.exception != null) { + throw new RE(results.exception, "Failed to execute because of exception."); + } + + Assert.assertEquals(1, results.recordedQueries.size()); + + final WindowOperatorQuery query = (WindowOperatorQuery) results.recordedQueries.get(0); + for (int i = 0; i < input.expectedOperators.size(); ++i) { + final OperatorFactory expectedOperator = input.expectedOperators.get(i); + final OperatorFactory actualOperator = query.getOperators().get(i); + if (!expectedOperator.validateEquivalent(actualOperator)) { + // This assertion always fails because the validate equivalent failed, but we do it anyway + // so that we get values in the output of the failed test to make it easier to + // debug what happened. Note, we use the Jackson representation when showing the diff. There is + // a chance that this representation is exactly equivalent, but the validation call is still failing + // this is probably indicative of a bug where something that needs to be serialized by Jackson + // currently is not. Check your getters. + + // prepend different values so that we are guaranteed that it is always different + String expected = "e " + jacksonToString.apply(expectedOperator); + String actual = "a " + jacksonToString.apply(actualOperator); + + Assert.assertEquals("Operator Mismatch, index[" + i + "]", expected, actual); + } + } + final RowSignature outputSignature = query.getRowSignature(); + ColumnType[] types = new ColumnType[outputSignature.size()]; + for (int i = 0; i < outputSignature.size(); ++i) { + types[i] = outputSignature.getColumnType(i).get(); + Assert.assertEquals(types[i], results.signature.getColumnType(i).get()); + } + + maybeDumpExpectedResults(jacksonToString, results.results); + for (Object[] result : input.expectedResults) { + for (int i = 0; i < types.length; i++) { + // Jackson deserializes numbers as the minimum size required to store the value. This means that + // Longs can become Integer objects and then they fail equality checks. We read the expected + // results using Jackson, so, we coerce the expected results to the type expected. + if (result[i] != null) { + if (result[i] instanceof Number) { + switch (types[i].getType()) { + case LONG: + result[i] = ((Number) result[i]).longValue(); + break; + case DOUBLE: + result[i] = ((Number) result[i]).doubleValue(); + break; + case FLOAT: + result[i] = ((Number) result[i]).floatValue(); + break; + default: + throw new ISE("result[%s] was type[%s]!? Expected it to be numerical", i, types[i].getType()); + } + } else if (result[i] instanceof String) { + result[i] = stringManipulator.apply((String) result[i]); + } + } + } + } + assertResultsEquals(filename, input.expectedResults, results.results); + })) + .run(); + } + } + + private void maybeDumpExpectedResults( + Function toStrFn, List results + ) + { + if (DUMP_EXPECTED_RESULTS) { + for (Object[] result : results) { + System.out.println(" - " + toStrFn.apply(result)); + } + } + } + + public static class WindowQueryTestInputClass + { + @JsonProperty + public String type; + + @JsonProperty + public String sql; + + @JsonProperty + public Query nativeQuery; + + @JsonProperty + public List expectedOperators; + + @JsonProperty + public List expectedResults; + } +} diff --git a/sql/src/test/java/org/apache/druid/sql/calcite/QueryTestBuilder.java b/sql/src/test/java/org/apache/druid/sql/calcite/QueryTestBuilder.java index f9b843d5d8b3..e0a9320002dd 100644 --- a/sql/src/test/java/org/apache/druid/sql/calcite/QueryTestBuilder.java +++ b/sql/src/test/java/org/apache/druid/sql/calcite/QueryTestBuilder.java @@ -37,7 +37,7 @@ import org.junit.rules.ExpectedException; import javax.annotation.Nullable; - +import java.util.ArrayList; import java.util.Collections; import java.util.List; import java.util.Map; @@ -69,8 +69,11 @@ public class QueryTestBuilder public interface QueryTestConfig { QueryLogHook queryLogHook(); + ExpectedException expectedException(); + ObjectMapper jsonMapper(); + PlannerFixture plannerFixture(PlannerConfig plannerConfig, AuthConfig authConfig); ResultsVerifier defaultResultsVerifier(List expectedResults, RowSignature expectedResultSignature); } @@ -83,10 +86,12 @@ public interface QueryTestConfig protected AuthenticationResult authenticationResult = CalciteTests.REGULAR_USER_AUTH_RESULT; protected List> expectedQueries; protected List expectedResults; + protected List customVerifications = new ArrayList<>(); protected RowSignature expectedResultSignature; protected List expectedResources; protected ResultsVerifier expectedResultsVerifier; - protected @Nullable Consumer expectedExceptionInitializer; + @Nullable + protected Consumer expectedExceptionInitializer; protected boolean skipVectorize; protected boolean queryCannotVectorize; protected AuthConfig authConfig = new AuthConfig(); @@ -151,6 +156,23 @@ public QueryTestBuilder expectedResults( return this; } + public QueryTestBuilder addCustomVerification( + QueryTestRunner.QueryVerifyStepFactory factory + ) + { + this.customVerifications.add(factory); + return this; + } + + public QueryTestBuilder setCustomVerifications( + List factories + ) + { + this.customVerifications = new ArrayList<>(); + this.customVerifications.addAll(factories); + return this; + } + public QueryTestBuilder expectedSignature( final RowSignature expectedResultSignature ) @@ -259,4 +281,5 @@ public QueryResults results() { return build().resultsOnly(); } + } diff --git a/sql/src/test/java/org/apache/druid/sql/calcite/QueryTestRunner.java b/sql/src/test/java/org/apache/druid/sql/calcite/QueryTestRunner.java index f3e1661b09ac..8b64f7a429bc 100644 --- a/sql/src/test/java/org/apache/druid/sql/calcite/QueryTestRunner.java +++ b/sql/src/test/java/org/apache/druid/sql/calcite/QueryTestRunner.java @@ -69,6 +69,11 @@ */ public class QueryTestRunner { + public interface QueryVerifyStepFactory + { + QueryVerifyStep make(ExecuteQuery execStep); + } + /** * Test step that executes or prepares a query. */ @@ -291,26 +296,15 @@ public static Pair> getResults( } /** - * Base class for steps which validate query execution results. + * Verify query results. */ - public abstract static class VerifyExecStep implements QueryVerifyStep + public static class VerifyResults implements QueryVerifyStep { protected final ExecuteQuery execStep; - public VerifyExecStep(ExecuteQuery execStep) - { - this.execStep = execStep; - } - } - - /** - * Verify query results. - */ - public static class VerifyResults extends VerifyExecStep - { public VerifyResults(ExecuteQuery execStep) { - super(execStep); + this.execStep = execStep; } @Override @@ -341,11 +335,13 @@ private void verifyResults(QueryResults queryResults) * Verify the native queries generated by an execution run against a set * provided in the builder. */ - public static class VerifyNativeQueries extends VerifyExecStep + public static class VerifyNativeQueries implements QueryVerifyStep { + protected final ExecuteQuery execStep; + public VerifyNativeQueries(ExecuteQuery execStep) { - super(execStep); + this.execStep = execStep; } @Override @@ -364,6 +360,17 @@ private void verifyQuery(QueryResults queryResults) QueryTestBuilder builder = execStep.builder(); final List> expectedQueries = new ArrayList<>(); for (Query query : builder.expectedQueries) { + // The tests set a lot of various values in the context that are not relevant to how the query actually planned, + // so we effectively ignore these keys in the context during query validation by overwriting whatever + // context had been set in the test with the context produced by the test setup code. This means that any + // context parameter that the tests choose to set will never actually be tested (it will always be overridden) + // while parameters that don't get set by the test can be tested. + // + // This is pretty magical, it would probably be a good thing to move away from this hard-to-predict setting + // of context parameters towards a test setup that is much more explicit and easier to understand. Perhaps + // we could have validations of query objects that are a bit more intelligent. That is, instead of relying on + // equals, perhaps we could have a context validator that only validates that keys set on the expected query + // are set, allowing any other context keys to also be set? expectedQueries.add(BaseCalciteQueryTest.recursivelyOverrideContext(query, queryResults.queryContext)); } @@ -447,11 +454,13 @@ public void verify() /** * Verify resources for a prepared query against the expected list. */ - public static class VerifyExecuteSignature extends VerifyExecStep + public static class VerifyExecuteSignature implements QueryVerifyStep { + private final ExecuteQuery execStep; + public VerifyExecuteSignature(ExecuteQuery execStep) { - super(execStep); + this.execStep = execStep; } @Override @@ -467,11 +476,13 @@ public void verify() } } - public static class VerifyLogicalPlan extends VerifyExecStep + public static class VerifyLogicalPlan implements QueryVerifyStep { + private final ExecuteQuery execStep; + public VerifyLogicalPlan(ExecuteQuery execStep) { - super(execStep); + this.execStep = execStep; } @Override @@ -531,11 +542,13 @@ private String visualizePlan(PlannerCaptureHook hook) * after the first failure. It would be better to check all three * runs, but that's an exercise for later. */ - public static class VerifyExpectedException extends VerifyExecStep + public static class VerifyExpectedException implements QueryVerifyStep { + protected final ExecuteQuery execStep; + public VerifyExpectedException(ExecuteQuery execStep) { - super(execStep); + this.execStep = execStep; } @Override @@ -622,6 +635,12 @@ public QueryTestRunner(QueryTestBuilder builder) verifySteps.add(new QueryTestRunner.VerifyResults(execStep)); } + if (!builder.customVerifications.isEmpty()) { + for (QueryTestRunner.QueryVerifyStepFactory customVerification : builder.customVerifications) { + verifySteps.add(customVerification.make(execStep)); + } + } + // The exception is always verified: either there should be no exception // (the other steps ran), or there should be the defined exception. verifySteps.add(new QueryTestRunner.VerifyExpectedException(execStep)); diff --git a/sql/src/test/java/org/apache/druid/sql/calcite/QueryVerification.java b/sql/src/test/java/org/apache/druid/sql/calcite/QueryVerification.java new file mode 100644 index 000000000000..ba57e603b372 --- /dev/null +++ b/sql/src/test/java/org/apache/druid/sql/calcite/QueryVerification.java @@ -0,0 +1,55 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.sql.calcite; + +public class QueryVerification +{ + public static QueryResultsVerifierFactory ofResults(QueryResultsVerifier verifier) + { + return new QueryResultsVerifierFactory(verifier); + } + + public interface QueryResultsVerifier + { + void verifyResults(QueryTestRunner.QueryResults results); + } + + public static class QueryResultsVerifierFactory implements QueryTestRunner.QueryVerifyStepFactory + { + private final QueryResultsVerifier verifier; + + public QueryResultsVerifierFactory( + QueryResultsVerifier verifier + ) + { + this.verifier = verifier; + } + + @Override + public QueryTestRunner.QueryVerifyStep make(QueryTestRunner.ExecuteQuery execStep) + { + return () -> { + for (QueryTestRunner.QueryResults queryResults : execStep.results()) { + verifier.verifyResults(queryResults); + } + }; + } + } +} diff --git a/sql/src/test/java/org/apache/druid/sql/calcite/util/SpecificSegmentsQuerySegmentWalker.java b/sql/src/test/java/org/apache/druid/sql/calcite/util/SpecificSegmentsQuerySegmentWalker.java index ca312983ffd6..5ca07c83afc1 100644 --- a/sql/src/test/java/org/apache/druid/sql/calcite/util/SpecificSegmentsQuerySegmentWalker.java +++ b/sql/src/test/java/org/apache/druid/sql/calcite/util/SpecificSegmentsQuerySegmentWalker.java @@ -103,7 +103,6 @@ public SpecificSegmentsQuerySegmentWalker( this.walker = QueryStackTests.createClientQuerySegmentWalker( QueryStackTests.createClusterQuerySegmentWalker( timelines, - joinableFactoryWrapper, conglomerate, scheduler ), diff --git a/sql/src/test/java/org/apache/druid/sql/calcite/util/SqlTestFramework.java b/sql/src/test/java/org/apache/druid/sql/calcite/util/SqlTestFramework.java index cace9647e96a..051ad3477327 100644 --- a/sql/src/test/java/org/apache/druid/sql/calcite/util/SqlTestFramework.java +++ b/sql/src/test/java/org/apache/druid/sql/calcite/util/SqlTestFramework.java @@ -62,7 +62,6 @@ import org.apache.druid.timeline.DataSegment; import javax.inject.Singleton; - import java.io.File; import java.io.IOException; import java.util.Properties; @@ -473,13 +472,15 @@ public void configure(Binder binder) binder.bind(DataSegment.PruneSpecsHolder.class).toInstance(DataSegment.PruneSpecsHolder.DEFAULT); } - @Provides @Singleton + @Provides + @Singleton public QueryRunnerFactoryConglomerate conglomerate() { return componentSupplier.createCongolmerate(builder, resourceCloser); } - @Provides @Singleton + @Provides + @Singleton public JoinableFactoryWrapper joinableFactoryWrapper(final Injector injector) { return builder.componentSupplier.createJoinableFactoryWrapper( @@ -487,7 +488,8 @@ public JoinableFactoryWrapper joinableFactoryWrapper(final Injector injector) ); } - @Provides @Singleton + @Provides + @LazySingleton public SpecificSegmentsQuerySegmentWalker segmentsQuerySegmentWalker(final Injector injector) { try { @@ -504,7 +506,8 @@ public SpecificSegmentsQuerySegmentWalker segmentsQuerySegmentWalker(final Injec } } - @Provides @Singleton + @Provides + @Singleton public QueryLifecycleFactory queryLifecycleFactory(final Injector injector) { return QueryFrameworkUtils.createMockQueryLifecycleFactory( diff --git a/sql/src/test/java/org/apache/druid/sql/calcite/util/TestDataBuilder.java b/sql/src/test/java/org/apache/druid/sql/calcite/util/TestDataBuilder.java index d8daa6595594..07a9fbc30abf 100644 --- a/sql/src/test/java/org/apache/druid/sql/calcite/util/TestDataBuilder.java +++ b/sql/src/test/java/org/apache/druid/sql/calcite/util/TestDataBuilder.java @@ -19,6 +19,8 @@ package org.apache.druid.sql.calcite.util; +import com.fasterxml.jackson.core.JsonProcessingException; +import com.fasterxml.jackson.databind.ObjectMapper; import com.google.common.collect.ImmutableList; import com.google.common.collect.ImmutableMap; import com.google.common.collect.ImmutableSet; @@ -35,7 +37,10 @@ import org.apache.druid.data.input.impl.StringDimensionSchema; import org.apache.druid.data.input.impl.TimeAndDimsParseSpec; import org.apache.druid.data.input.impl.TimestampSpec; +import org.apache.druid.jackson.DefaultObjectMapper; import org.apache.druid.java.util.common.DateTimes; +import org.apache.druid.java.util.common.Intervals; +import org.apache.druid.java.util.common.RE; import org.apache.druid.query.DataSource; import org.apache.druid.query.GlobalTableDataSource; import org.apache.druid.query.InlineDataSource; @@ -62,16 +67,24 @@ import org.apache.druid.server.QueryStackTests; import org.apache.druid.timeline.DataSegment; import org.apache.druid.timeline.partition.LinearShardSpec; +import org.apache.druid.timeline.partition.NumberedShardSpec; import org.joda.time.DateTime; import org.joda.time.chrono.ISOChronology; +import java.io.BufferedReader; import java.io.File; +import java.io.IOException; +import java.io.InputStream; +import java.io.InputStreamReader; +import java.nio.charset.StandardCharsets; +import java.util.ArrayList; import java.util.Arrays; import java.util.List; import java.util.Map; import java.util.Optional; import java.util.Set; import java.util.stream.Collectors; +import java.util.zip.GZIPInputStream; /** * Builds a set of test data used by the Calcite query tests. The test data is @@ -119,14 +132,21 @@ public Optional build( new TimestampSpec(TIMESTAMP_COLUMN, "iso", null), new DimensionsSpec( ImmutableList.builder() - .addAll(DimensionsSpec.getDefaultSchemas(ImmutableList.of("dim1", "dim2", "dim3", "dim4", "dim5", "dim6"))) - .add(new DoubleDimensionSchema("d1")) - .add(new DoubleDimensionSchema("d2")) - .add(new FloatDimensionSchema("f1")) - .add(new FloatDimensionSchema("f2")) - .add(new LongDimensionSchema("l1")) - .add(new LongDimensionSchema("l2")) - .build() + .addAll(DimensionsSpec.getDefaultSchemas(ImmutableList.of( + "dim1", + "dim2", + "dim3", + "dim4", + "dim5", + "dim6" + ))) + .add(new DoubleDimensionSchema("d1")) + .add(new DoubleDimensionSchema("d2")) + .add(new FloatDimensionSchema("f1")) + .add(new FloatDimensionSchema("f2")) + .add(new LongDimensionSchema("l1")) + .add(new LongDimensionSchema("l2")) + .build() ) ) ); @@ -137,18 +157,18 @@ public Optional build( new DimensionsSpec( DimensionsSpec.getDefaultSchemas( ImmutableList.builder().add("dimHyperUnique") - .add("dimMultivalEnumerated") - .add("dimMultivalEnumerated2") - .add("dimMultivalSequentialWithNulls") - .add("dimSequential") - .add("dimSequentialHalfNull") - .add("dimUniform") - .add("dimZipf") - .add("metFloatNormal") - .add("metFloatZipf") - .add("metLongSequential") - .add("metLongUniform") - .build() + .add("dimMultivalEnumerated") + .add("dimMultivalEnumerated2") + .add("dimMultivalSequentialWithNulls") + .add("dimSequential") + .add("dimSequentialHalfNull") + .add("dimUniform") + .add("dimZipf") + .add("metFloatNormal") + .add("metFloatZipf") + .add("metLongSequential") + .add("metLongUniform") + .build() ) ) ) @@ -223,111 +243,111 @@ public Optional build( public static final List> RAW_ROWS1 = ImmutableList.of( ImmutableMap.builder() - .put("t", "2000-01-01") - .put("m1", "1.0") - .put("m2", "1.0") - .put("dim1", "") - .put("dim2", ImmutableList.of("a")) - .put("dim3", ImmutableList.of("a", "b")) - .build(), + .put("t", "2000-01-01") + .put("m1", "1.0") + .put("m2", "1.0") + .put("dim1", "") + .put("dim2", ImmutableList.of("a")) + .put("dim3", ImmutableList.of("a", "b")) + .build(), ImmutableMap.builder() - .put("t", "2000-01-02") - .put("m1", "2.0") - .put("m2", "2.0") - .put("dim1", "10.1") - .put("dim2", ImmutableList.of()) - .put("dim3", ImmutableList.of("b", "c")) - .build(), + .put("t", "2000-01-02") + .put("m1", "2.0") + .put("m2", "2.0") + .put("dim1", "10.1") + .put("dim2", ImmutableList.of()) + .put("dim3", ImmutableList.of("b", "c")) + .build(), ImmutableMap.builder() - .put("t", "2000-01-03") - .put("m1", "3.0") - .put("m2", "3.0") - .put("dim1", "2") - .put("dim2", ImmutableList.of("")) - .put("dim3", ImmutableList.of("d")) - .build(), + .put("t", "2000-01-03") + .put("m1", "3.0") + .put("m2", "3.0") + .put("dim1", "2") + .put("dim2", ImmutableList.of("")) + .put("dim3", ImmutableList.of("d")) + .build(), ImmutableMap.builder() - .put("t", "2001-01-01") - .put("m1", "4.0") - .put("m2", "4.0") - .put("dim1", "1") - .put("dim2", ImmutableList.of("a")) - .put("dim3", ImmutableList.of("")) - .build(), + .put("t", "2001-01-01") + .put("m1", "4.0") + .put("m2", "4.0") + .put("dim1", "1") + .put("dim2", ImmutableList.of("a")) + .put("dim3", ImmutableList.of("")) + .build(), ImmutableMap.builder() - .put("t", "2001-01-02") - .put("m1", "5.0") - .put("m2", "5.0") - .put("dim1", "def") - .put("dim2", ImmutableList.of("abc")) - .put("dim3", ImmutableList.of()) - .build(), + .put("t", "2001-01-02") + .put("m1", "5.0") + .put("m2", "5.0") + .put("dim1", "def") + .put("dim2", ImmutableList.of("abc")) + .put("dim3", ImmutableList.of()) + .build(), ImmutableMap.builder() - .put("t", "2001-01-03") - .put("m1", "6.0") - .put("m2", "6.0") - .put("dim1", "abc") - .build() + .put("t", "2001-01-03") + .put("m1", "6.0") + .put("m2", "6.0") + .put("dim1", "abc") + .build() ); public static final List RAW_ROWS1_X = ImmutableList.of( createRow( ImmutableMap.builder() - .put("t", "2000-01-01") - .put("m1_x", "1.0") - .put("m2_x", "1.0") - .put("dim1_x", "") - .put("dim2_x", ImmutableList.of("a")) - .put("dim3_x", ImmutableList.of("a", "b")) - .build() + .put("t", "2000-01-01") + .put("m1_x", "1.0") + .put("m2_x", "1.0") + .put("dim1_x", "") + .put("dim2_x", ImmutableList.of("a")) + .put("dim3_x", ImmutableList.of("a", "b")) + .build() ), createRow( ImmutableMap.builder() - .put("t", "2000-01-02") - .put("m1_x", "2.0") - .put("m2_x", "2.0") - .put("dim1_x", "10.1") - .put("dim2_x", ImmutableList.of()) - .put("dim3_x", ImmutableList.of("b", "c")) - .build() + .put("t", "2000-01-02") + .put("m1_x", "2.0") + .put("m2_x", "2.0") + .put("dim1_x", "10.1") + .put("dim2_x", ImmutableList.of()) + .put("dim3_x", ImmutableList.of("b", "c")) + .build() ), createRow( ImmutableMap.builder() - .put("t", "2000-01-03") - .put("m1_x", "3.0") - .put("m2_x", "3.0") - .put("dim1_x", "2") - .put("dim2_x", ImmutableList.of("")) - .put("dim3_x", ImmutableList.of("d")) - .build() + .put("t", "2000-01-03") + .put("m1_x", "3.0") + .put("m2_x", "3.0") + .put("dim1_x", "2") + .put("dim2_x", ImmutableList.of("")) + .put("dim3_x", ImmutableList.of("d")) + .build() ), createRow( ImmutableMap.builder() - .put("t", "2001-01-01") - .put("m1_x", "4.0") - .put("m2_x", "4.0") - .put("dim1_x", "1") - .put("dim2_x", ImmutableList.of("a")) - .put("dim3_x", ImmutableList.of("")) - .build() + .put("t", "2001-01-01") + .put("m1_x", "4.0") + .put("m2_x", "4.0") + .put("dim1_x", "1") + .put("dim2_x", ImmutableList.of("a")) + .put("dim3_x", ImmutableList.of("")) + .build() ), createRow( ImmutableMap.builder() - .put("t", "2001-01-02") - .put("m1_x", "5.0") - .put("m2_x", "5.0") - .put("dim1_x", "def") - .put("dim2_x", ImmutableList.of("abc")) - .put("dim3_x", ImmutableList.of()) - .build() + .put("t", "2001-01-02") + .put("m1_x", "5.0") + .put("m2_x", "5.0") + .put("dim1_x", "def") + .put("dim2_x", ImmutableList.of("abc")) + .put("dim3_x", ImmutableList.of()) + .build() ), createRow( ImmutableMap.builder() - .put("t", "2001-01-03") - .put("m1_x", "6.0") - .put("m2_x", "6.0") - .put("dim1_x", "abc") - .build() + .put("t", "2001-01-03") + .put("m1_x", "6.0") + .put("m2_x", "6.0") + .put("dim1_x", "abc") + .build() ) ); @@ -336,131 +356,131 @@ public Optional build( public static final List> RAW_ROWS1_WITH_NUMERIC_DIMS = ImmutableList.of( ImmutableMap.builder() - .put("t", "2000-01-01") - .put("m1", "1.0") - .put("m2", "1.0") - .put("d1", 1.0) - .put("f1", 1.0f) - .put("l1", 7L) - .put("dim1", "") - .put("dim2", ImmutableList.of("a")) - .put("dim3", ImmutableList.of("a", "b")) - .put("dim4", "a") - .put("dim5", "aa") - .put("dim6", "1") - .build(), + .put("t", "2000-01-01") + .put("m1", "1.0") + .put("m2", "1.0") + .put("d1", 1.0) + .put("f1", 1.0f) + .put("l1", 7L) + .put("dim1", "") + .put("dim2", ImmutableList.of("a")) + .put("dim3", ImmutableList.of("a", "b")) + .put("dim4", "a") + .put("dim5", "aa") + .put("dim6", "1") + .build(), ImmutableMap.builder() - .put("t", "2000-01-02") - .put("m1", "2.0") - .put("m2", "2.0") - .put("d1", 1.7) - .put("d2", 1.7) - .put("f1", 0.1f) - .put("f2", 0.1f) - .put("l1", 325323L) - .put("l2", 325323L) - .put("dim1", "10.1") - .put("dim2", ImmutableList.of()) - .put("dim3", ImmutableList.of("b", "c")) - .put("dim4", "a") - .put("dim5", "ab") - .put("dim6", "2") - .build(), + .put("t", "2000-01-02") + .put("m1", "2.0") + .put("m2", "2.0") + .put("d1", 1.7) + .put("d2", 1.7) + .put("f1", 0.1f) + .put("f2", 0.1f) + .put("l1", 325323L) + .put("l2", 325323L) + .put("dim1", "10.1") + .put("dim2", ImmutableList.of()) + .put("dim3", ImmutableList.of("b", "c")) + .put("dim4", "a") + .put("dim5", "ab") + .put("dim6", "2") + .build(), ImmutableMap.builder() - .put("t", "2000-01-03") - .put("m1", "3.0") - .put("m2", "3.0") - .put("d1", 0.0) - .put("d2", 0.0) - .put("f1", 0.0) - .put("f2", 0.0) - .put("l1", 0) - .put("l2", 0) - .put("dim1", "2") - .put("dim2", ImmutableList.of("")) - .put("dim3", ImmutableList.of("d")) - .put("dim4", "a") - .put("dim5", "ba") - .put("dim6", "3") - .build(), + .put("t", "2000-01-03") + .put("m1", "3.0") + .put("m2", "3.0") + .put("d1", 0.0) + .put("d2", 0.0) + .put("f1", 0.0) + .put("f2", 0.0) + .put("l1", 0) + .put("l2", 0) + .put("dim1", "2") + .put("dim2", ImmutableList.of("")) + .put("dim3", ImmutableList.of("d")) + .put("dim4", "a") + .put("dim5", "ba") + .put("dim6", "3") + .build(), ImmutableMap.builder() - .put("t", "2001-01-01") - .put("m1", "4.0") - .put("m2", "4.0") - .put("dim1", "1") - .put("dim2", ImmutableList.of("a")) - .put("dim3", ImmutableList.of("")) - .put("dim4", "b") - .put("dim5", "ad") - .put("dim6", "4") - .build(), + .put("t", "2001-01-01") + .put("m1", "4.0") + .put("m2", "4.0") + .put("dim1", "1") + .put("dim2", ImmutableList.of("a")) + .put("dim3", ImmutableList.of("")) + .put("dim4", "b") + .put("dim5", "ad") + .put("dim6", "4") + .build(), ImmutableMap.builder() - .put("t", "2001-01-02") - .put("m1", "5.0") - .put("m2", "5.0") - .put("dim1", "def") - .put("dim2", ImmutableList.of("abc")) - .put("dim3", ImmutableList.of()) - .put("dim4", "b") - .put("dim5", "aa") - .put("dim6", "5") - .build(), + .put("t", "2001-01-02") + .put("m1", "5.0") + .put("m2", "5.0") + .put("dim1", "def") + .put("dim2", ImmutableList.of("abc")) + .put("dim3", ImmutableList.of()) + .put("dim4", "b") + .put("dim5", "aa") + .put("dim6", "5") + .build(), ImmutableMap.builder() - .put("t", "2001-01-03") - .put("m1", "6.0") - .put("m2", "6.0") - .put("dim1", "abc") - .put("dim4", "b") - .put("dim5", "ab") - .put("dim6", "6") - .build() + .put("t", "2001-01-03") + .put("m1", "6.0") + .put("m2", "6.0") + .put("dim1", "abc") + .put("dim4", "b") + .put("dim5", "ab") + .put("dim6", "6") + .build() ); public static final List ROWS1_WITH_NUMERIC_DIMS = RAW_ROWS1_WITH_NUMERIC_DIMS.stream().map(raw -> createRow(raw, PARSER_NUMERIC_DIMS)).collect(Collectors.toList()); public static final List> RAW_ROWS2 = ImmutableList.of( ImmutableMap.builder() - .put("t", "2000-01-01") - .put("dim1", "דרואיד") - .put("dim2", "he") - .put("dim3", 10L) - .put("m1", 1.0) - .build(), + .put("t", "2000-01-01") + .put("dim1", "דרואיד") + .put("dim2", "he") + .put("dim3", 10L) + .put("m1", 1.0) + .build(), ImmutableMap.builder() - .put("t", "2000-01-01") - .put("dim1", "druid") - .put("dim2", "en") - .put("dim3", 11L) - .put("m1", 1.0) - .build(), + .put("t", "2000-01-01") + .put("dim1", "druid") + .put("dim2", "en") + .put("dim3", 11L) + .put("m1", 1.0) + .build(), ImmutableMap.builder() - .put("t", "2000-01-01") - .put("dim1", "друид") - .put("dim2", "ru") - .put("dim3", 12L) - .put("m1", 1.0) - .build() + .put("t", "2000-01-01") + .put("dim1", "друид") + .put("dim2", "ru") + .put("dim3", 12L) + .put("m1", 1.0) + .build() ); public static final List ROWS2 = RAW_ROWS2.stream().map(TestDataBuilder::createRow).collect(Collectors.toList()); public static final List> RAW_ROWS1_WITH_FULL_TIMESTAMP = ImmutableList.of( ImmutableMap.builder() - .put("t", "2000-01-01T10:51:45.695Z") - .put("m1", "1.0") - .put("m2", "1.0") - .put("dim1", "") - .put("dim2", ImmutableList.of("a")) - .put("dim3", ImmutableList.of("a", "b")) - .build(), + .put("t", "2000-01-01T10:51:45.695Z") + .put("m1", "1.0") + .put("m2", "1.0") + .put("dim1", "") + .put("dim2", ImmutableList.of("a")) + .put("dim3", ImmutableList.of("a", "b")) + .build(), ImmutableMap.builder() - .put("t", "2000-01-18T10:51:45.695Z") - .put("m1", "2.0") - .put("m2", "2.0") - .put("dim1", "10.1") - .put("dim2", ImmutableList.of()) - .put("dim3", ImmutableList.of("b", "c")) - .build() + .put("t", "2000-01-18T10:51:45.695Z") + .put("m1", "2.0") + .put("m2", "2.0") + .put("dim1", "10.1") + .put("dim2", ImmutableList.of()) + .put("dim3", ImmutableList.of("b", "c")) + .build() ); public static final List ROWS1_WITH_FULL_TIMESTAMP = RAW_ROWS1_WITH_FULL_TIMESTAMP.stream().map(TestDataBuilder::createRow).collect(Collectors.toList()); @@ -475,36 +495,36 @@ public Optional build( public static final List ROWS_LOTS_OF_COLUMNS = ImmutableList.of( createRow( ImmutableMap.builder() - .put("timestamp", 1576306800000L) - .put("metFloatZipf", 147.0) - .put("dimMultivalSequentialWithNulls", Arrays.asList("1", "2", "3", "4", "5", "6", "7", "8")) - .put("dimMultivalEnumerated2", Arrays.asList(null, "Orange", "Apple")) - .put("metLongUniform", 372) - .put("metFloatNormal", 5000.0) - .put("dimZipf", "27") - .put("dimUniform", "74416") - .put("dimMultivalEnumerated", Arrays.asList("Baz", "World", "Hello", "Baz")) - .put("metLongSequential", 0) - .put("dimHyperUnique", "0") - .put("dimSequential", "0") - .put("dimSequentialHalfNull", "0") - .build(), + .put("timestamp", 1576306800000L) + .put("metFloatZipf", 147.0) + .put("dimMultivalSequentialWithNulls", Arrays.asList("1", "2", "3", "4", "5", "6", "7", "8")) + .put("dimMultivalEnumerated2", Arrays.asList(null, "Orange", "Apple")) + .put("metLongUniform", 372) + .put("metFloatNormal", 5000.0) + .put("dimZipf", "27") + .put("dimUniform", "74416") + .put("dimMultivalEnumerated", Arrays.asList("Baz", "World", "Hello", "Baz")) + .put("metLongSequential", 0) + .put("dimHyperUnique", "0") + .put("dimSequential", "0") + .put("dimSequentialHalfNull", "0") + .build(), PARSER_LOTS_OF_COLUMNS ), createRow( ImmutableMap.builder() - .put("timestamp", 1576306800000L) - .put("metFloatZipf", 25.0) - .put("dimMultivalEnumerated2", Arrays.asList("Xylophone", null, "Corundum")) - .put("metLongUniform", 252) - .put("metFloatNormal", 4999.0) - .put("dimZipf", "9") - .put("dimUniform", "50515") - .put("dimMultivalEnumerated", Arrays.asList("Baz", "World", "ㅑ ㅓ ㅕ ㅗ ㅛ ㅜ ㅠ ㅡ ㅣ")) - .put("metLongSequential", 8) - .put("dimHyperUnique", "8") - .put("dimSequential", "8") - .build(), + .put("timestamp", 1576306800000L) + .put("metFloatZipf", 25.0) + .put("dimMultivalEnumerated2", Arrays.asList("Xylophone", null, "Corundum")) + .put("metLongUniform", 252) + .put("metFloatNormal", 4999.0) + .put("dimZipf", "9") + .put("dimUniform", "50515") + .put("dimMultivalEnumerated", Arrays.asList("Baz", "World", "ㅑ ㅓ ㅕ ㅗ ㅛ ㅜ ㅠ ㅡ ㅣ")) + .put("metLongSequential", 8) + .put("dimHyperUnique", "8") + .put("dimSequential", "8") + .build(), PARSER_LOTS_OF_COLUMNS ) ); @@ -575,6 +595,88 @@ public Optional build( DateTimes.nowUtc().toString() ); + public static QueryableIndex makeWikipediaIndex(File tmpDir) + { + final List dimensions = Arrays.asList( + new StringDimensionSchema("channel"), + new StringDimensionSchema("cityName"), + new StringDimensionSchema("comment"), + new StringDimensionSchema("countryIsoCode"), + new StringDimensionSchema("countryName"), + new StringDimensionSchema("isAnonymous"), + new StringDimensionSchema("isMinor"), + new StringDimensionSchema("isNew"), + new StringDimensionSchema("isRobot"), + new StringDimensionSchema("isUnpatrolled"), + new StringDimensionSchema("metroCode"), + new StringDimensionSchema("namespace"), + new StringDimensionSchema("page"), + new StringDimensionSchema("regionIsoCode"), + new StringDimensionSchema("regionName"), + new StringDimensionSchema("user"), + new LongDimensionSchema("delta"), + new LongDimensionSchema("added"), + new LongDimensionSchema("deleted") + ); + ArrayList dimensionNames = new ArrayList<>(dimensions.size()); + for (DimensionSchema dimension : dimensions) { + dimensionNames.add(dimension.getName()); + } + + return IndexBuilder + .create() + .tmpDir(new File(tmpDir, "wikipedia1")) + .segmentWriteOutMediumFactory(OffHeapMemorySegmentWriteOutMediumFactory.instance()) + .schema(new IncrementalIndexSchema.Builder() + .withRollup(false) + .withDimensionsSpec(new DimensionsSpec(dimensions)) + .build() + ) + .rows( + () -> { + final InputStream is; + try { + is = new GZIPInputStream( + // The extension ".json.gz" appears to not be included in resource bundles, so name it ".jgz"! + ClassLoader.getSystemResourceAsStream("calcite/tests/wikiticker-2015-09-12-sampled.jgz") + ); + } + catch (IOException e) { + throw new RE(e, "problem loading wikipedia dataset for tests"); + } + + ObjectMapper mapper = new DefaultObjectMapper(); + + // This method is returning an iterator over a BufferedReader, attempts are made to try to close the reader if + // exceptions occur, but this is happening in test setup and failures here should generally fail the tests, so + // leaks are not a primary concern. If anything were to actually try to mimic this code in real life, it should + // do a better job of taking care of resources. + BufferedReader lines = new BufferedReader(new InputStreamReader(is, StandardCharsets.UTF_8)); + return lines + .lines() + .map(line -> { + try { + Map map = mapper.readValue(line, Map.class); + final String time = String.valueOf(map.get("time")); + return (InputRow) new MapBasedInputRow(DateTimes.of(time), dimensionNames, map); + } + catch (JsonProcessingException e) { + final RE toThrow = new RE(e, "Problem reading line setting up wikipedia dataset for tests."); + try { + is.close(); + } + catch (IOException logged) { + toThrow.addSuppressed(logged); + } + throw toThrow; + } + }) + .iterator(); + } + ) + .buildMMappedIndex(); + } + public static SpecificSegmentsQuerySegmentWalker createMockWalker( final Injector injector, final QueryRunnerFactoryConglomerate conglomerate, @@ -712,7 +814,6 @@ public static SpecificSegmentsQuerySegmentWalker createMockWalker( .rows(USER_VISIT_ROWS) .buildMMappedIndex(); - return new SpecificSegmentsQuerySegmentWalker( conglomerate, injector.getInstance(LookupExtractorFactoryContainerProvider.class), @@ -808,6 +909,15 @@ public static SpecificSegmentsQuerySegmentWalker createMockWalker( .size(0) .build(), userVisitIndex + ).add( + DataSegment.builder() + .dataSource("wikipedia") + .interval(Intervals.of("2015-09-12/2015-09-13")) + .version("1") + .shardSpec(new NumberedShardSpec(0, 0)) + .size(0) + .build(), + makeWikipediaIndex(tmpDir) ); } @@ -815,6 +925,7 @@ private static MapBasedInputRow toRow(String time, List dimensions, Map< { return new MapBasedInputRow(DateTimes.ISO_DATE_OPTIONAL_TIME.parse(time), dimensions, event); } + public static InputRow createRow(final ImmutableMap map) { return PARSER.parseBatch((Map) map).get(0); diff --git a/sql/src/test/resources/calcite/tests/wikiticker-2015-09-12-sampled.jgz b/sql/src/test/resources/calcite/tests/wikiticker-2015-09-12-sampled.jgz new file mode 100644 index 0000000000000000000000000000000000000000..6ca3233b777cdf165c312267b08b4bac7c786df8 GIT binary patch literal 2366222 zcmV)uK$gEBiwFqWasXBU19xd_X>@60Yh`jRGB7bUEigGPF)}T4VQp}1Wn?aDb8l_{ z?7iD}8&#S&eETS(%Q-GFl=NPf5w7{PVtjefbhK45q`I z{+=1q0-xg*R^Sw0;db>1#m^g#uEV(~mziP1J87Z73Z!C+lPquJJB4G|X7zw2hD5>l ze)p>X7Tn4j!l;&oWA$Zbsctd6HqH%U`p%EA1>f1E$O+qbe$1#)+Z4La3hP(P46gXU z+kh{%Dm0ISpQid8{KyI|HqJ6ftp%zV$TZiBr7|_(Rg)FDFj_|7UGxvYu^9Z)zOj+R z!YEp{$rz5mz;v7TWz@tzZ)&_lxXNrii9PlIJh=%H$xJLJY3W1@5_uXB33qY`H{fU4 zlsx(55WFG`D^KmGo+3{;j4(sj>GJ!O&~lEPIFa{W&%hg4QJB@tB-M4EcXT>4(={M% z)7c!y#*3k09H-o9!nppsLt{6=Og1Qdu*u1F*zhnQY;j3TYC#uJU)ISvrqSnzZr~ru zlud2NB4U>Gmmp#_k0wVk#czCJq{>%+nHsHC(YjPu)6B|e04r$+!spEYcG3|q2 zM5csem&kFdGplbM;6zMPl(>{gNV29hVUDe%919FGO37FY$n!kl+60`);s5*9>zoWt z$r)l4m*1`uyR`g{QS7OumaV``AV6}abNaL}oeuc?$dVf>F=kqS*Pq+4>{88TmkNrA zEK8|`5*3lW&Zt403UMr4IVVm`cw1N|dwg&a{&i6XZoP22g$%6WJW7yRotDeI$V|ry z4bF6l+oxbRn=1(pb8xo+sZu(X+y$r(P?44?DGSSj?U(1nZ&qCfCn?2Z>8QE}Y*+`% zuu*a1?jE89zl)m77cJP7cDxd}5tkuau27>&3Cx>C2C1H>hDG+sX<6%s|D^VY2I?Sn z%CI3bnAj#r{YvLeN7)*sv=mR@jc$%hr+$G51GZF~=C}@@Ak+Z09iB!G+5#Cmdf*u7TqUOCRtmZyU4NfL>!9duB1_@P$ahyRiU1GxImS1_-;!W)xkAp2KMV8{pR8+8r zM5T&oQOjbMou));O!Tik}J-&GK_{x zVB)sAp8`YkK$y2hI?qUv;SY;|hO?g<{TuvAk=%f4Qc9(EH>VAolLbPZnxOD)yWylM zQY^je2&y}R^c;Gd**3a?G}zS=t=fX=1UF=BuhwqVuGhY}`E2c4?efi+NbN27a|SMc z4wu~vZ^NrgwLjlHSNpW~RqZOiulEey+Hc)qU9No%|E}*NwU=QmU(~L_$l$#8`OPzU z=^700YZ&`ywW~Lu3k@rN2j1WPp3h0`{n{6Ir7dV63Zr{nA2?Ci&=v|#PH=0x^@Vs$ zO3AG*q`-PQth`aKmwV7-bcpe)0T0$;VORw)GeW~~xF{#A3Zte^mF}12j>|)qw}e2I zlz1xyCqsDg);_s$`}}*~pZV>bk6*ui{^dJYZ>%m|T)Xl5+J#T=oc*$!*xVztXCpML zXa8gM+#grZ{c82xN2}+~t)Bg4_3W3oU%Pz!`oFI(eZ2PN=WEyATRZ;;Z=~N|J@egr z?}tWjP) zZ^#yz6{c?m}>A^)1b$)^TYjwJottNvhJibp^&2zJwTz6?A4<`!LK)I2iKM zK9U#4Jg8ix;kvMTOn4rfnj&MSSkReC_DoHU^~aOR@Gkaglg(91Tz4OfhOMPGZ7ns} z3QT2UiWF~4xnS02vHNj$isuCxf-4&W_(K@e5ri2;r|{?%Xg<%EuNW9p0HjXYf>Z%H z(`4RdcpqpaHa*2V!PhtbFKt13?~8XtgnpdbC2o;HxIwi;cz*c6Xlrnt+qt6<7!JJx5P0iDXG?A@SNBT$0;Tpy{~02{NdfN|9a&N`QfdrfBop| z&RbvKp~*IACRxSF={%3w7l(w&c19k>4{yHxx3@m};f-IDv481+k}S7^)R>Y6JaS9| z?K_^7WUUEKBGFuA?s2eH2KS0-t&x#U)s((zF;^9>=&ap%@!PMy+60z^6K${@q!l-t zYx|-2{;gNOf9_vzUwW?-4r(h7)Bp9$x$1xY(vJiz6ii@&4aEmVXln{!P^6^Vx=3~% z>s@iIrzBaG+X23Bo^?Ps4zjc@Y_jQAHbo7a)8fW;HZCw1%qYNCrjrUxzkn@_%yU2; z51w#1Y+dYB#bg#-;P6Fod6pa8-Pby;%%Q_=D{}}p?j5l*$An&CIIs~IV8PwL-5-aS zO%^~{g}M0n1@q)=->PAZZDyF^?{ z#25W1Gc#xi`h=GMjd%r2Kuj~xDHx6>gXIqQ!*3b_Qx~nwBV!WY0qn@>_PYneAuPb``1m^U=&2379hi4+m}? zt@cNO1$;Q@{{4CI)lP0i88+{A7%=34P-4I z*-8iDhhB#=g{P?k%j?96@uO3j;q3GXnLaQwweyvm76-F5T=+?7Bz|5!wa|p2=_?87 z21{jB*cqcsNUs9r#EGFs9Ya>nIb{oLQVLEUUIIauH40TS0h((SWMM-S1#z~2XmP)= z;gk0*6wtMGV{&Sc<6O5uRPUiUS-bS| z>e=h7=YG9VxH{F5#@%yE#w}0`%?Ta6*U3mZ7uYVgMehP{m?)oKOQDU)#&kdIL z+>rPZ(t736vD+U>7j?Y*_My7a>8(!Z`Q zy}r8i?&{L@@G1hW8p3ksm>+ zoo5D^)?jJEk9YLtOTDEh9C8dwbY>au4-dKLfS|Bgq0HyVNyp&A=fMY~%V!cRfx!Vg zMr##a)_g#4Ko4lVFOzj_0lFNw=7>ehWopb)+eSy%;$NP9d1mHlxxh^4nK3xWL~#Y>RT{v4#BdW0uLY79syI-6uzCP>Bv_D7E=P4 zppdJ9{To|>Io*hmxpBb)xAKmKhV8&%Q5?YUB~z?ovvLmfcIEFI@6L=dp^IWw@hOTu z=dx_^a(N>Pjb~P;bEkaC!?1epcdKXrw0iDe-OPp$*Dn9tx7T0xg=Cr=S|ad%7T2U$ zTk1|=w?oOcE17|Tg@pylI=R4d5;tbWfXOW3n0aOmSZGgJI(m(DUCPgv|My{P6sJYy zr}95cI5{0oek%8vE`ly3Mslpc9GkfX*$0{Ba5xSPDCp~ICN5{RWKzovDalwyk&~HG zWpp$%Ix;kxNhA`aTUUS#PvJdnRU z`-*p=ab+I?v8fC9g)AQJ10D5L{hd>Iw9jQspx%`u?%_UInne)ACHLA${dkYE_d(HZ z1|8FeIXFyDd)bs>?Q}+d=I~R(=6Nz9K>HPU_teCtxYCvpd+)_=PYGFg;%sU$GU*gy zt5^u{`R^cS`<2dtxh=s;N=Y^H5=TMHOiSPgd(X3S48F@QiGnyq4mxuzCr*V%VH_XQ zDdFhTrel{z-H>;~@N1%>2DIV3H-H%$*E4p1?))%@>7v1US7Up^w@5+#YjR!-A*Ug%m|PJrUU@QDJkBb$F)l% z(Sb(dFkOI0c2<~q<_^Cbj1r+uwlordsB+{a0?#BEneu$hLnfn`?t>HH1G<)pu^wiPG>5LoW-si@<;l=i#c&^%ri<>huofzVF zl7qVObV+;+z>NY7j!`zzl-0gnvO5<8)ish(fV+O-izU1hq+Y#&A zKhn1mvA+Ev&@d`d2Q95iih`DYOp;T+0J!WcAaIWZ0vDHJZ9D63a`fFNN5?&g6eaYs z>|S;>z?*`nB{ID5Z2+C+gDpJu)Q-0Shw-QS)+3%~rfi8E z6c($L$fOGds6rrSO9D$;V08+ihgXD5Kw-awjHSMHsW#=QdmIcnFm1F%t6jKhWlsmY z@Q0`0{Oe0^{P5dN5+Hh~q~~w1{o#Kup9u{{98(-68Pv)|%4eg;ebL!?U@564B&99Z ze*f%`X|r0KXJkL6^Gq*`9!ZXhL$Uq2yZVA(sC8%BAsp`nr_O-bYKyw#iVo_gZt@H2-n%y zudcoN>FqbqZ6YG4kGF})Y04LNi2Llji=Te`)za;^E{BiAVl|B`(m+Y^OUHE6gP9KO z!GxQ{+L&ml84!k*4{tp~R{pT^#jU4T-e0*yR^Gq$^sQ%CF5NocV@qP?%B|;N1fQ;a z5E_qL&%k(}S^2~r{K^+A7vVk^Z#{okik}Aio)S5s=%I3_OjXTJRa5uJi4vEkcpK|k z#SVy5);3wyA>@K%p5$1;Ku=C)NzI>&?^I{({mfsZ{Uwpi;;+W#`K51K* zTIiHLZ_we*Dw7t4F%AYe7;2O54H`;e6E`(DaGIpXBKs?0_c<=iIm}ZaxY&A3*kuia zk_pTju_9yw4I)jnPo49nzXslQ;z=nXM=qW*Sy_Ne+D{r^)??FOb;JKt3}6Vq4mivg z-N4Zu?;J*sZwf#>hFx3I(Vzbw2Vjf|v&hJh&J6@&&pgjFQfZXphObEK7ABgMYO9id zcw?AF40YQj$0B*^z>c41Mi&E_;Ym&rp`;XxZiT|G;W?|iN*{M$GM12d{xR6`ywZ6x z%qrBh%k{eT5o4a4!oX8>!wK}1%)9Adq2W4cCDPVO-{ulBoe42YNoln;o#`PY1kPbi zdlJ$aww`HYYO%#t!ORWjVYGmOV{RgIyy2p-jypJgI<8 z6PS1tnRsAml$aE2uSN*5V#Z85k3DNfsjgdOkHsify(@crA@Op{2x%6!erlA$Z!5kJ zHjqg)w-xs|AU3L`wetxDO<8vJFjQbfFF?_1x?^o%*>Pi07<@X+^CjKYEUnhO;@Qmt%6)^3I(g&;>TA&prdSysMMh)=4bD*tq64*rS(8>O%;lx# z#pMq=*`a--?OuUV6SRwa!QUA={xwB8+5u@C%oNL;3=p%ueBst?(saCUBVK)vQk@hY zIm$Prlw$d)EXgrhO8XRt{!LmlY>8b$iYHpJ3q^adba8d*?CRMU4o_#-KDlw{(rb5K zz4-lGzh8UiqwimP0qLa9zKe_iXaBr<4%w>7>X|e0>bd{C{pR^Q7hhPr@r+BF^WN&x z)2mB2R+rvbUHW)+>HXWUe{lQEA6A$C`0e#yuP(jqGVFZ(-Jf5*bLrL9r5A61^qI@H z10Q)GKC~Y^@)}M(c|Hqv_UHtxc&$dj|!0ne_zWvVm@OvgaYE_?tuc^(w(mf8i zqSKO+j9e*jfGn5-)~*c}zl~LQ+Z(SAzXwjyvgNog;E)xcp!Qgr$*67!oim@KNNXCs zT40+u4qAO1o72PE!YfBDUZ!~`f^WM^O}r2hRZ6#|l|H?xECyh zA~QJEd(0LdDUc~tIcWIQu2E(hoQ#zZ}|S>|kH z(BPIwUG|Ryq#%q(Y!JANFfYxV6^jJ>R9O4!ujOe4I~sC$OBv~OK|Z^4)yWAtHrVEb zthb*z1QNX<%J;w-86kM4Q>!QzIyfH>?LXY=eZ(JbVaBXef&bL!wGMAemE>4Vl9QU0 z@^u*P*E+YM)HZKJi3lsOcScSRKO$@bYBlPhf>VXniJ5Z5aXV+O3~n(+`dXyIOD z>W2=`5Ayy(=v^x~sOTGisChK;nMOOwUg)&kd>njiG(p!iCRe0)!;y7;9joT_2mzGtsq5J4r7d_8xGt*B zyyd;CcbG9ISfxVA1P`BTbe!wD;cI6aZ#v!L0m}wtY`@Z8ejeD3kTvrqK94GT9(nR- zLHpYvsHQ}IuHsZG45`%S85r9|>UDrVm>oGuSvnn)x3~G&UFYrv8RCF9PvNE9U@oWl zc+yqX*FwoV35YT#UPSKHvXS=Jxx!+1TlJafdKJ`n1VmC%k3s`(V^SzJeWF z+G(>ZdO$S#duzKN`tLV#CawOTPj#~Yzu&O_`wc_Pho1I>)Z%O|7T0_`Of&W?2aNqH zNE4v68NGp-n?Q?Oe!tGhgnh)!GKKRmu*QXziOFWIUdwU?^%0C%gOYK_;2o%qf6PxL za@B&9pS!EoAj?uL*;cp~R3o2c^USn)L0E2|nMO;m0xHZ?6_)2z&kPBedY*GUv3yum zR4u;f{Q(Vo5B9NXc#`WcVhcuTW@azGh@_8Wi~Gmmbf>SpdOx0e2TATV23VJSt*_Mw z8-6j)<$PL4f9K8}oA>QN$Tcb57R!BT6>rmlDTqdN(=FW_AW2GU+ny9y_nyL*gJl8w zfJxo38kslk~k$dpq-`YgZqBH7t{k=zO;OoC`D z$}pYa0gc)_HU~CzI&qcJ-LX znv3~r9b*BD9Thj_A>C^61`beV4Ff~bn8<>Wg|cv4-gwprk>yl8A;(kk#$LGmL5IyS z2?|k6mgKZ1#gmI0&RVy|jc+Y(H(Q zR(OYdEI<8(QL8sgrUc5&?)DfGQKeG=p}4;w9^+cm=nwIfIYE??j^%s!+IFa z26Lms>#z#b64G3!B{e+z))dv*Mol;Ei>`Vn!5~#tDb?QnG@{qj76E;prz2HtByVIC?iUqj>aQoaQG@&jgi=CQzF zTZ29d$YDuIweGfo%P{~Uw<6=}%BOhn{K|i>ynpM3-pkfnxwvxW*122fVFVxIK$cdn z!9bpM8w!LpQRv(pPy2Wy<8fc5Lts=iDH(5dlm_O6*cdQaJ@?zyb6>BXy}G*e@tx}* z-+t+L8SE73YB?4)3ORA=si(+4|6PASSzY=Sl03Zg;_BJ+t7o78_Ua3(XJ4taBOu$u zd#g(?xm*yxUR`>c^!NX-dsWKPJ)GChe{kp3i{G7jrU~}?rOfovsr?5=kBuK7N3vtt zqXYHn2(JKdyw6>UOf3yIUSyH^KuE_THWjnVlvp+xeJn!RbXTZ!73?&jDv?Vfyk&TN zaA10Ix>^wgb}vv`W@h#g-zzq}g8l#CMFvKF2F(U+f8;g-L{=;Jyx%+DqwL*BG?i2- zyaiv>)PDCEUg+J%Tjut0 z!K>KaLQ~p^VuNmA@CMuKm~0Byj{@J1f~k%@u(nT%$%+(fV@M1v#m2=v z2(fW&!VLefR5&>(N?1v4y0yxfN4KBo@Zd=#i@(gRKF3!&e`FM}D$q@;0d}xnjg;Tm zo?{q+AUWM}wK$9CIgQfdhto2JmFin*a>6I`5`kBm9G4PuWJ0$GO^f3Op9M8;J9RiY z*&!%?7x6bhS4$9 z%1v01!vYMd4i#@df;MIXC~*m!(WZS#nQF6aS`An>t!5N?Hy+;6#?@)1A`G72=2}re zNR!g)Ze!XYG;tH6;^@>QLym4cW_UfmD_n6Z<}k?dR>xt;sLDCFz4M2}Yxn%&%Aefu z#HUCoMQ$zKdY)i>;@0_<&)~n<|M?TR-gAMMThFZgY31U|$IZdGL+M7~rC;fYo3a&d zYD`Khk@r3?85%T5m7wWtZMEkmRY`@11pc$wJZ#JPYQCwFg<@kBBWxV&J1l%hQufL- zlQ9YQTB)8jJ$-y}`gjjyO}B1~hU;S(?#RymBhSFdDcfXawx3bkU#c2m0jn_c(K8et zKF0~zN_p%uVlWY5yX=bhLV+Ns}mvHLO4rMZ0R35SJmdesRyxwE<= zZ>g_<0hn$psC!u5lr-n)+v%p%(o$00B_D1#q`8qd!TO|8JRRM&q~nSd*IIoYfy=Ch zkO;;G4LpytQlxlU5Q%6~I8LS<7+83M*+F3zU58o&mGC{sR3x5I zQ@c^h&6Vx_Os}|l_V#Gj*wrH)d6#+M1Y3hu9Xmmr<@)%yk5XO3tc!G88;QUb5?O*N z3iBq&F=}w5LjCC)k?mvm7w=p<%*BvaRa+p^ZHXV^(*zvkYFt zw2QR1+r_tCg$5;S9%f*BBT*WXg(RXUr$2ZL<|OiwZ(p!pLn;roEeAGwPvu5|{g6)0 zIkw(cdE;fWr;#_(d%cyXbjB(=vn+Ci7fZH9_KcadOeUylVbwm3)<1sA&KefeeRPOQ zQ;$;u>v2c|5*^HqRX;;mmkeeRj+-oSyVR)p$%#{hEF#|^7)Jeh@>BBn>2%t6$%0EV zp7VnH@&ZT@=AGh8-q9?+lGAB^4$qBQ;oV77bFsMExa8g`);OEpcr!Q|pzC}#><$h$2~LO)Kvd-|WKxbKk)SV^Y2 z*&WH<3WoL_bymzXvHhJSG`HDDTfWu9i%`S5)|rzRd1_X++qzbeN}YY*H91<$3UC6-NfP&27%!1%D7l^U1?>{YG0 z+UJu`o;YE%GF-7)=IHUoqsM1vo_eZxi&nR6h$JVx9A!D4k`yJp(4_OcaF~ICGNOCz z5YhdQ=)u%lcdx*{;bCUYGHz3Z)G1RU98#v-!t}N}^=c=kPip}ErJyzKdpMSqFI-3Er5DbG?MMBv})N1grkI$!%Q3 z2tI@B|2P1GZUIo;tYyZ|(s#C8$Ok}gm=sp{)SL+hcQ4a9W+r@2>SVI%rFM^ltzk`- z;>pPIQ7mVI?bQ-x<~7cA1+q&}#)!>M*}b)<21V7vI;L#SA!8+z@?2G4ba113D z^K}YpDn5br9!$S!TUP}Z*kN-3BZ!8p2*Cbnk}1ZpBYnPPTO~0IauOw=&GJ|fsc8LE zwqq2S(LaiP^LBh*@0JSD7_)GgFvkQ;3`}HZ#z5AXHQ|lnvk=Dkti7?vklbXNbB_b! z31&uHQ^#F11NQqxOyfyrli1UKYSQx^r3mp87g zCTxLG14USioEeNHaNNv09PA_x;;+IC?B--68xF|tznu9Me&OEUD}6aMM209}RR~|_ zFkNDUm~B73zzs{*ToU<)g1AZR=PdyO8=;Nm#t7OFgBy7rBp#b3bk+t-UPs}V%*50C zNWlT4mr|1?Ntqjt%?#YpK=o(lwrU-r6H2BWbKawzoZ{gd-2z<0u(1iFb&Jj4^CUr>Ki9 z=dIOat#9AFRC-70jjb(Gu$LmCwrj2c2njOfz$BXFs9r5scs`0wvSD8CZaF5UxE9$o zTQ3-t-Gad_YvfZ}-G73mrOINv&YlVXyqOpykgnlMAsT}NA zje5&=7XrkQQYPI-G9TEI7z71$o*51!e|LW}79vAY z=l;f(4uS+MIBGM;w`w}So*4(3YcNoDEKAIZ&{!Q7^9+-0d>|$L4a1T_TZYAv9!8CI zuQtGhE7;7JP>=lE%a5tM|8_Ev>#ma)U**Q@t!24&igRmVNR%rMx@iYoAMO;Z*sPp` z-6C<}`o->QH+Z^i7vuA6le+ev~{m ztpnNVEIE8&m{eeYCMDahWCjKn78WE9%DW`Y;y}g8>5aZQ18$^rAP=MMw-_!`iCUFY z9pLOcHr3YxCn(kKa-Xv6ZA|Vx%TFq%ZUyX7`y^kWCO{%cmYWq}b|@D0T8R5W1+7=- zv-k|)9M{23BH53mDqib`IYj>1-VT$p-+Oq;a82r?2z@RyB%HBw{-cl`0X2T;h z>XJ>9f^d-PXMNjqn`IotMoc^2?YR-71C~z9JWt%;PWCiQF7c5IpRt>r=lPBG*PZL; zw1yZ0EDknC1yl+%r{yJZ>F~IV8AIUMpu}|Ckux!^LC2A8e=T19ihRHbwM%PJ! zGxF&2f9i7#WHn|L6xb{W9obDhEQ5J(82iXP_OGNR_%MDDzs~;=zk&xJ8DyuOBB9Q# zdvB*)q(;%QDyD>C`~_wd98@}(_mg|A)17aAtvQoh*hH6%IHssjH}xsP!(`b^hbi~> z=#HSWYjH_Swl3y55nN^e+pmA~!<$c&AD;bX@5M>~=gV_{z4FBm&wuo{r?39cm9ImC z@YnY~`s=SgYRUb)}F#~LG3nn3N<1)g9h8hOg<0noGnzW#^Dw(p` zJTu4yGtKIQra~m*lB^`KEiydHp}`!V80Ge9Ii2wJhDfEFd42A2u(hP9Qe0{E>INqB z`}WTt-@f$j>Y1}U9*yaGp>`9D#+A&sSHE1l@#5O=uHX6Nm)}44uiu}0XLaf0wXa`V zd-j<-7jAre^~`tgy&qoCQC@_H_aHa@gg4C>Hd3RI~wCDTlIiVkK)pv|ulag3YC(H6Ca}ql5|3=P z9TvqRO3^xL0ak@sUgysp?Q+_>T)7NaCNXwLbtK@|MKLr)E%)%D)HJ-bz-E_{mSSyr zet{+CrpfTIu_}rtDX}zmy;*EIZ%c-6dHc6t|Mt5NKE-tS)ulhKo;|;M_62fa7=CU_ z*05C_NVGe@fUB>ZII%x7`RL->ORue-eg5`;o?l&heP(9$%$YlHT)*@A8*4AVcKekJ zYrp&bj+f^fm7}#^zJB|Qr91zACG@j#XpVBzwd5PItW>h7~2@vn!7L1SV<3Vv$*j9|*-}?RRglefbV3Hn*?*`p%7eso3<(oihsLEiwuc zNaUinwfYDX1ODOdH-Gs3zy0w3)4dl0{^9Lk|IgJkKfLuD^4HIP{lk}UgvQ}-zy8f% z&;I7G|Mr_7o_WD1y_z@#B~%0)(o(9;26%X5UJbgw3N^V^;$37SizLX9WvH#LCronl zq=Q5pV%9A;9HB0s&5mC!9V+3nwoT3cERx$=Rg=>JiyDQTI29V7)*IcWk5N;lq!w8r z1c1JK?!)h{{SjsJbJuQv{c113)53{=?t|5H=T^^NT0Q&j?N8sn^XldBm{{~Q_Zd*K z+RUVIj{{;6-=#b0B+9_PpJnFu@{T4jBMVWOfl44=Tx!%Z^l_6Nu5=`r~>PR)}Txh0R)tFS_bv&v@7-^I6c z^79tC&k5`&)gD@W0Iivg^M3o4OLyMA2nNhObI8V;yNr7raK$9&h%M_7O{hRk@$*vn(@qWccqV7`D=2M$eJ8?stlUj5_&}MfQxtJcEYK%<$Da z2aH*mdas&!fwKZkzhjqazT~(|qQL~(6s5_4k~E3E#N^!uqb=#S*doBOh-LGNqoWg| zd-Bxh+Yaj4PGX{2qOI~LFyS(&?GvaHa$9Vsx9{lAd(#3wC&-=&*w~Eacfv0pISmRn z)*i|+_#5*nf=a4$xd2?>3J+P9VzJ0()MniDbwXsf6n{srnJrmZH_Koer(lDi49i;s zKI-T2YGg|SlNxKLUAV`=mXxeY={5q%hgC&kv$KvRcp-{u)ioVO2Ay8p-HFyCh;vpIi zvF5#-sD-XKOd=IaOm1CDrW0J9tNmN;YVAht%FQ#i%Qv5{UA_5I?}aRCzp7oXU9Vli z|L!BTH(?ad!12}ESIv_zYhS@IKZCR9LSuD=&k3DY$su~0&iNNv+PBIA<{Tw4=P1$E zayc-!A3%1T!;itFd@_Z7J7#+iu)|n*?$+Tipat(8xue1!mqnxmXD~D2Rfut>mwX|> zW+`>RO6nwL{=}jSL6;`A!c4EWYW(QbaQ6PG3*ABoZpTcUOyz~$fe!4)GzFCZ1qKSb z-$|b8>?KcaiG(I46Ro0ccRT8|9l0bBsI+3|8)1fLZ11!iMOi#aDXvCNdByYw_!??4 zw?7vQt}I+GkRFh4V@I1{jUCM#9VXfF&?p=dRHw3!EkBYQPE#bct#~(ZKLeMY4YN}n zq+QC4n#{@zOx7JSuv?n72sV*TPq{*y=)xk*yFN2Rsw9u3X98PTz+|)TQAykv`&FM6 zh9$@ZE10m)lPMZLlDj@zKKnInLka*;>yU#I*hM!gFc-$_Qa1AFqUmyY^8!VL=#RBV zYa}oy(-=mUbXrF&3_PtjrE0MaA*NV_QTp?-Lb?% z`vp^&rIAQQ?n{78&8W$k*$bvuON6&-GF>MJnK8$W@M^-N)x0jJ)0)o=PBf?AJq`#! zTuLNbP2s@22ACe3H>v%I<@T)wJrs;CoeY^YKi7Y(vfZ!*h`W!moayj6|DsvHlxgWW zHgHI%g`6(tB4~BpYF(Cc+v3C{cOkA*(mvj2mz;9$@jfzZiZZASvoMdR2ru8i+TCug zq)`)Fr)P82G{RR>|Gaw2M~kHT9OVK1Y34>KIf38ASuf0eOLfu|Ia*N^VzQ3#!2*0@Eil5q3_DGD?-K{z12W!@rtgh}S{R#i8V!)CO+mo%6#yO$e&f8ajO zb)CZ=GGX#Ugoe$-6b$jKX)xQ*9NTo*C;=TdDI~p#wM76QUa$3r9Pk_rvP8?gyS&Sh zwk8#QIgw-Q8+n(9Oh(s9mlwEQB9FlT#%U3TNd~bv1E$S@OJq+@mi1#b0X(g7?l#Hg4J;$bpdMw^K#0^FUy)+9xzx+Yy2kJ3u6dmes zB^M#d2cDyZluSpapgd3pIWZ5i2gN~w*_7)j3x@EwalORKk62g?hKUIk%fa$xE>Jw+ z+c|)OG6o~wJ+|AAF<};Abu`BrK2w2F6I8kX1Ih$gd~H;& z-P=^2eA1xv=uBI5%g)J93DT&-V9BEg4!}j5+d5u-=4|ljw9hE>a~uQ)DJ3NmkwH3MXCU1y zS!@)9prwJl)(xR^W-Yc}0A;M7{Jpy4H`y(@OIrsO-V3ig3`8V3GS);h9dA;+xyQlq zjK`#OEV3Kk-Dn>h$xbff)>s8az#yGOkEkaD%%KbqgK#Q!i%NK<-$*Pn+xWGCFbI?~JbF zaXGmaYQxxeVfxW_H2RJV^=(9>?}z}+0+SKo0X2Nx#uF*mGK(q5oa;A2Iwy-Iwn!FB zVV^K9I+YbO`o{))%lbQrafRb0crM7EM~A{!!4K<7%3?7V8aLZhFHGgJ+px*yjzNCqtsbbTG<819H$Y*qAOvIp%ALEc+S6f>Ma3 zGqISYwxtd^b0JC!-?{l(?d!jsd850vg!g4-)PCz0LM~zP;LR6n*I^{r+`7Tbr1lBC zbOT=cs&?7C@Cv+wRfV6!yT7hoy?F-S{aNjLEj(l55uTspR)x_yAA7YDOZb$%;P|Oh zQr;!B+Tenx=y@m9F`0n%6(xzHn|%+{$ifnj-u}7=cbuO&jDaPm^V2lP?9w9^8HJx& zaCwN_wsXVW!W_y1nQ(~)?|Ms$P&$>y6tGw+skDQ@PO9b z*6Mh3ljb;B&W=rnOvdK=+Lp~JY)4C`s}=aGlcQ$7#4M29n{Z6+2rDNQ+tSJzLJN)z zIi}qKsJ@|WtEDr9{IS9`xX;Q-#w9f&$qBG>G+#nYs%hk;0yQO9gDWs^AAmH* zFmohdvU@NcX4t_I>mq#MbQmsIEn&&&l%Muh?p%|!6#$AXB~z_?{;oKiXQn}`k@EDq zD$W!TXKl^GLz$zFInU=ol|lC?NH5SiEo_1}?&vm0?&K*(En)P8TF@cobsUyHWqKi* zh@9-&yoFCQ2wNYWM$Vg*ayQK54VHmkGfuMx#(KOq&BI{q!+!UeaO_e9b8}ATzQ9j^ zXRp~7PP24IjY(ScJibv(SOYo64I+Ca;}{{a{NQoDm=@+KuMZVugck9O_mOvzy9V;KDqkDH)BF z(I(ZZFjlA%_bbt*h?x_H=Uo)DfLUt-Z8z!P@1qhvR-q=?N_=2wYM4Z6pW3*A4>Z-M zf~$d*a4jLFq9<-T^+~a02vs^EScan}!tE!cKBzX}G zAr0NtccV{?oU+mSZdA&E2GPIc!RW9m250dsvRUeV+!uFCH#ZsgI2ao7m=uqCsLf#s+Lm{2xA)HqFB9oJbf#u>F86J#0Ij%`-Z+;?EQ&0~GQrh4_h zc&dL)qK$FKs*+J=nxJ3%Yh^k(t{-i4T!T6^&y74b4ddidYBGy8J=p2Eq^NFVGfDP4 zP(!Csr7=q<9ldvEou8qfG`?)7FR|9f6=iaW8AgQKJ`R&beTC)yO6UH*$`<{7)0nN3 zXmcjepweyS;;r*5AA%ls>zS1+D<7_Wyz<8Z zKE3rkoWPi#U-=|7&Oe{q$UEsvEtP|&mZsfyTW#Ev!M%EJYaz_(-baopDxsB1uW5G^ z)X~X-)2G8zo{UqprgR?!37<}v2%^==89)I|i7h{0GfZ`os| z>Lm$yP+*zt6xZh|+2an~RR*Ipn^R%{iKfIdN=nJdv6z<8Vsa{z1?@YNjVI-d8Xxf! zeKgl%peTn6x2#C9=-rndIPqu$UIy3mf-s$onee2B&!y|Gu#`b>Z{v*&)uQA-u(8L@ zXE6`=<}YejYu7+t*50apL2jO_U4z&D=H`oV z7_G61BCAp=Daon~QdjYL*?o-FY9PvL59yQ?k6r*gjT8>ETxZA#0V@)t+zo&3zGhR6 zp{e1&AB``f19UWwOyt6>z_)?Wp2biDjBM}DckPaP3_OkS7o24#kiVuROhY4sld zYpzu-@LE-3QaTyAcfc^SN(??obc!4fdA@wbAgOpPp=zm^N4zj%wbT&iVLZs?kd8_2 zKP)0oEp9wTl-FAlV5V>$q&3_-)AA|wTu3uB-J@c-PquxDkx4cm9&^W{WlBI1V}8Qy z2r84Jgwba^i0DSSzU>y<>$;~DW$SE8J>k*pknx0SyE$j@V##wDg|!D;J|2nA%)=)k zX=*DzrJ!5r@^4QPp%%XE0i)PwG5lXYey zVY$6-8vSjyYMqiwph~4-|A9sSNn;ah6qIjxV`Jwc?%vtpELz_~;rm*JeNUw=7OG^B zGt;8^@D$IR>~j5b=Z8Vw`e}$u@#u|VABc3gb$-+wvz>ZT^?m0@^}tBMqu&Zfm?kQ> zzxdPIh1XV>K3+Zd-s-upSI=EqJ@+g3;QZ>@S8l)lhud#{^zF54w=cbT`xh^)F1_eo z`R>Etub%zm>baNkolCE;F8%xJ*|TtyO@!^ii8f(-kXER%BZO@Gz(b>noy@`fOM)Cy z{l06}=erI<3+F+KdU)i19S@u{rU(X$WmAInT9(Z%3#5AYyfKj$ou*20rPW3cT-{xJ zyLO}YY3;_%vl)UuhuWX9TpEtPs(n$rUb}wtB~tsdTM_-)&DUyQ!lfIvE43SV5uFzJ zp35CR^Ot8|0kmtc;Ad_w-F&fLbA6+~_BFb0>X)w8z9cuF#)|8kX9@lixbFMF&DTlo z0^IU*_p2`7eD>yRH(&3us9^1t+MBftIJgU;pOb%iVnc>4-~e{BE>bUMJwS zMShOJDE5_j;*rEPA-q3 zDGc&>-vlpKnDKZY8P6ZWWK?5dEGvgzWRG%_7wW#C1_OjI{URAW{Q}d>qSv+ALd+3t znYRbFL1XVe;uU>7;RxHyMlVxyuGey6hIIJRn0^MsY7lcr2nHu$f>ZEMhJT}}WCGTO zn#`osR5mlD42@(I5N^JO);UXUvm`YSyOg9PTCSnCaXDsTY90 z)eC)CFu$BQNCD>yBDKGk83yHxMc`6VVj%Zo2@rg+-}sEexZ>jqR{|P^2#lm9TT$6G z0Nj_=?Y1--P>3&so{!Pp%dS!J83umQ8n{}!a`W8HGXyOQ0=hpEG+nEGvB_3;X$Cj~ ze^|KF>lsX=Q4|St5t5-1Zfm3<$5Xz%VfQ2$zOpQ-+TF01J<6HwP6xx3M*|JSYOgtA zwj^<$fLSyZH5D49(d7?qMkY#*zl2;(H$zbFaX^?KD)ib~R{_SUOzjpb%ii`RVXPEPxZ$pU)KWK3^NMGq-Ha3K%OJ{9FE=sgy*Iqdq|SYPhp zla$o9q{J|-0KweLmzF;uCCUkGc-E|2xGTIasRdk(!9@Efa|FK+F~H!Y=-t**6)cl9*_j?BqZ zy;`vaFGS$o=w`%G#^gdpiQH?><-A95jtHCQSx!*1Kx|R*`km}(Y_^OC4O6S+4rX9K z(<@*kt&c8(k-&;#gXO3Adp8DZ-smBvNwK!ra9~lH7pCcnJ7geotnOtg-6L<>#}=Bu&1T?X(_aV7=wAIhw~uL>5V& z{FKP>w*Y@r@V5$o(|DcQaD?ympu_%I!nEfYzF}zn8f=0;#U5E;=cBXur$jLd{dqWa z;AOAVi0>XxJw^O%NOCIXe|ctRTbpSGS>TPI?(>v42b1=VCG_L7-zkc;pZ7@iHMVe% z13p7(Nr_Ib1Io{FZX3+9yl!cD*>b~A^Bk-}Y*}80J*M0f3EnM~2PNUade4_w*^R7G z+m4c)SdVGMGASABUxw0ync-AY%Zw@`iOfh!PG^SIq_4uQU+$cpFK>yWsz`G5M&dy! zlG>|Kc1e<&J{iADCa_0F)uePAgLx>9OjdB)*^(Ug?)LFyx-$I4i3~F`9#!Un!-p45 zh9BuUnz$)yz)~D#WU6AqgKF&~8@RATl4o)dPtz`QJsEYgG}lQq{oTp~OHdV4(cApt zLG#B4F}9J-+YZIf2S!1-jUs9HmesvUfT@>SZE$?zh0KtE5e<00UAJ&6G(PSfd;&3* z=qM0Vi7i1%OX+y4NfFqSy>3!q-~oRYHq@HsMaT4{A&yy+B(le`T*y2Y#zokZi#o$3 zeX>Wzl-b12FEnd7tlM4KnQ@V;7M%RtK*=4KD&?*4Xf|%r1T=K@nViuyNo}J=eF&5C zE3ogP-61M0@0bxy_papT%a{w8Y>U|kru(^jo}Z5IAbW28$E`o~9MjLzt>9Uu<+?r(=h9-z+ew+J_pT}9(bY0vB}O| zvZ`$$HHL3cUCL%wp*n^V8<5Y;{1a)TK!$fKe>wB3O+WHr5FxvBk{ul)6DG4jt{e6v z78!CXp+O%$sryvgW`4yE*OqZRZ0qvWOd9- z>^bi>W*xejq#zY^Zh1%Um@F*kS^mJ}%!uNhfn`+kQN=1vYhDRx&MTN^sI_~mH?2|{ zbKqw`;c&gL1tgeJ9H!|JF6d+mRuPDdDdAQL;+2uSgGD==tZyTJeqKeSe8cHP;2U*JDOpm}l8i9~)tB_G zG?Sf_fVoZzQY$8tyDjXRqOlCyHI{J_YGAk2ZEY;!8CWQMeh%o zYn$?yY<2_8ot3dDU6y1yxw!GP=M23mEc3S=Z`9b}Z29ZuOK>Sd(;{z5eUF3MR8rmJ zfDeQwCAG*}Dqv-ya58Ly4LgXJ`zDd$=809O<{Ljcl^M=XkBlCj9M4X}qmZk5W;?ua z4 zqEjaMk}gb9DOGjdiNF)x75}+{M2ac(ZgTN$QRzLsbdcuf(CIrS3U#%{t?@ZvxNT-o z;Jf_jA#?A!Vc;Rmh(D2HlRE4z3b7STMNGHx=mm~y04j2fxX%3dZ#}>AY0vc!_5F18 zNQOpYtNY=qFsnO1ORNL? zMAku0E+VJn>Ch;RF)=%9vg))Fi^ZBiC7T*tGOz}xVj*>81-*M|zVY2O-zaIm@$FCZ zRgl(FPDI{JA0R7$SQQR$jnvh&OPar$lGOBWJz#4^5=@r>X<== zVB1ShsbD5xJQmq%_I5uneRDhalmMki9rTfD$29tf#e#tp5}|Q83>Y9+#P~*@2(`JL zsDXDAEh(j9!Mj^c<*AtyCNh4shH#Eek)su6Y!aXu%C-e)hTs`2?2Mi(E*;CH!1&-O zD=-r|1U#>u#i->nA+&~Ojvi@^%^WrCdLs3D$CZQ>Pf408$-aIova0!fy8$c26wKgD z%B|ae;Cekb!eoO{X1iQ~oe;W@4ah>*>m60dabfD=O{|n6#Z@USBfXyw0$;K~;8cOQ zA|;ZMGnXH@y`FxjWLt*4Uzw(1PYaskmR(F#7}7|W=cTCOm*4bq6hvTBm=Vy&tfD+xhv)+Adv(_ zBiDisHYHrT#Y~=J1DG@DEq2%eeTwvNXUw$wxvTnOi1W7KgSt4Oc!8Thxf&_VbI+E z1MdDRrliE>=+vP`|JYz}#a%}V3&>jI#)`KEq5D$_ zCqS#1$3=ntwTG84z{pIGT=D+PBL(@XBd-Uq1djBR#+UWhT9}6E1LNVKC>fS8$exo9 z`hz1#CeRfP>RT55N@oIhWeWy(6*F|=ZQZ;>P*7E1Z>un~;Mk1(1X;Cc&*8zyqg+Q7 ziR||}-h~uVr|D^?>*62E)GWd#<2BA)x7K}*##BH@ql&~}$z4BJnbc;Wrns4()F*ee}b*3!S_vhuXaEYH*B_mO`7o>?c`%{o8KX4w-CQtG z>@!3~6g#D6+*fTJ`OMl|&W9m}pk=OCd(W?yEtni6HW|aBben~*jXPK_s=gL*5qvpX zLQ2MW$s5`h2W1%03$NgHd1a)*jEbBa`yhppX+6NYT?~IJ<}N919_UR;H2RL(HeG1z z%Qs5%u+B}lMBL;SsG~a|*DrJ`*)Vs@-t&&mkygVY7B7~?@+XB3>m^8%pPbYrSxrfD zIw8pkpFq=+O~-+I91tq(${ktHGRvey?Ru3qsKrOJ&p@R;(@PsZ-o^-E<^c>`Jia#qUYz3)`No2`%YqI#JUW2|V zign3Gp4L-c{Zb5x47TtV+}9Mc(9&znbdb)j-pzWPnMFr0Ax*W89I$#L~SadS)2aI5pBk@JKKI-k^ z7}YXzOo}C2wZ~G>7{@oSeskfQx4(Jgn^(yi^=#(@p7-$)_DkM(GR!p}1z89uCRwhVlZj=NT(o-@o^2;Co-F9zbfjr_~ zJZj5el|NF|Y;hct`3E#Qn7E%gAV=LeM`OjQKCM#?taZ9Mcdcs1!|NL?Qp=`!&@0M3 zZz6rx_GYnYnERy#r>X;k4jv1VHH=nsZokMQV?}3E?$}hDDc7o3-TT~?L@d3fx{kV_ ztM*OV5;-h7G6MCDA8j??#<^t)$4zT(D193Cc#vcb^H_X)R(MJo4U@o9!M#v@k1{>i zNQm&O>R6ymz!YIF@h-8e6#;hKsgB-{zTuHJw8lWAqQ<=_K@Go#4w%^bcktx$f0}c; z(EfH@UTveP0Okxut~3Tj$PwB=2P zVc?$tJp~!VFkK5&{IX*?I&A9}VFsCY3~upFB8fZ75^7qnoh;hSB-1%U;Tu$VSq3Au zxa;0`JVyUU#H5+;ss!x$rDZf#ifOG2X*bjIM=~w3YZEi)!C0TK_gTM-{K|B?8DVsf z10t7@;!0#f$A?OpJ8JU(pS?G2Zre)NMgPj$K2;~~y)B3U2u^43b4sH}+1iq9?Q~in zmdGV?nIOQ8K{6e8mB++!oKDi2$F>qDv6DES6UReR)l6>H{U%klI`ZvL9a|pSPMyDS z-xrijh@jK}L|bB{?#LoQ62e+9)_UjX2}Rl77-J5Z6JxcJy}ddATjR_hp1XbR+MN^Q z_uu$_x%a@}O6vBNGc)hqxPA5NopZ0wocwI&%%}I?IXrXzLp*=x)8Im}OKgPkVXC{6 zj;@@eU$H2npm@q%R;vk-w+qw^-dff;`$3UmH>=e{K{9@QI&5CEd9iMA6EKI?rx(y@ zK$1a1(==0|c2`Tpw1>f?&@AqNqoxr}Mr*cCjY`;qr%93NMKwhJZ^tmx3~pJ@M*67M z-^F}ZIq)ezMUbMQ%PfBtD=7;P6D6aJtE#T?5Lj0YspE@c?NL%(6EiV6i3BR28gmN@!sigqN~B$MjT@nVS@nJ31W+aPJy-hNL_`r_?tUoViM7#Q@#yy1#t4pq$ZhrI7x`uxt3 zqcc~1HPcq7;!#7bEQm=(Kh7QQFC;}s#%c|P)w`d~nQ_#}mXOU$H@&~I zEtC|3(+qPJzOt|Sa%c!FJBwy9Fh?s)$+HyR1B3ep28V_kN80^_8aKc{g7r1V4Smea z2RFyP2#nEc_$TT!{G+5KQHUm%(v#Vvn-)paViD<8R9a$2LsilfLfumNDB{-4Qn5Wk z^(5fR2ItD_&l-$rq7;uW$&!;4L5hb~PfZTLwT;>aCEH6OfS&3gZbj?Pz7$<_`s0G) zsk&Nip&YZMHmDaflx~qDf+8-Z0ixe76sXY@E#uu>W!1UIqg0=6nmy@u+t64^*j)^c z0@&jaXm`-O8_^fv@ zx-neoAXBM1+ZqKg3gvM)f-54(e}$O6U~8EWICai=ZIjIA_Ld;FxDb;Y#Gux$QZ0`K zJGlyx%!*%5HCx%RP5Qw&ff1~SrBxsqZh#x0hk%;xr+ma#N2rnE+LH8pR1_2s!QF1Z zAy7<+zW3g}>!07dc5H!w*_K{Uz-$XGu9SG$hhN`4ef8FXuWo<))~$ob?!IvP_Lcqj zKK(4PJVY@nh%xXI1<6Gi7XknnB!d0qp#hl3mo7XB5j7% z6RI%_Ve6c4l7Sr45iHfM{Kce8{o{d8i6lU(OUZ}sR4uS0H+7PB)ow9pQ&6061fl}? zlQ3x+jdFD!t!2cOc8U1TyC{l~kV4~LfLWd*-Fy!Q#4@^#)lLSh=tGRIJo8My#Z;AI z^DsBbCZ5fa{&Y^))h0`l^*AnMTBSywk=?e&b%W$^G^R6};D*^KPtS8{TH+$o>85$2w8%zh6D0fo3S5&~jvymp`Ak%oh4g zYQUyifor*l7+hRxF-HroFhDe3U2M*4KZ`9^NpjXk$evuLx+7;B=^U2v^iz$rW4^)2 z=o?4}H>lNQE3(VXs^!xdzAjS3(Er=!R7FP$MrKupmMNbh>z$zm-((w+(*|)T>U1@% z)_OSAh7a-(tWJ%ph>Ky8F5;=E4`Km2XsM(hZpdoFWT``ghNkt?x{)V+;J>!RHx}mF z{AaPzc3p$pcco6NiBsPak%gECHz7zgs0)T#K2MuZ82z?sam!Bg@&lq-pa$@UN|Hu#`}3R?2sIrY>Ra&Zj-0pq+*df?pZ^!t>~ay@ILJ(%s3xgs^B8 z-pf5rkKhJjsuMy~5yYemY%uPkNAtbVBv3=1K!iE!f72Vtfgjmx=tVFDOQfwmWfJrV znQCIz8zV5B^p7%wXBUEs?DzJ3_m9Dpt&Diy4wa5p28p<9MHXB)$#Fqaz&!#NNs?Wu z(`dD9BjH%hPg}l* z)mTcF1SJ_73t?w5Q=z&WocnmHn^DU~DuUT69TUVuE!gWunf8=Lh8fd{Ug=N4DIvX2 zE}0Zsc7v|Tm=)lX2OJRbr>$dt=qUuN3kW+VB$V3PQ!M};Qadv(MqsMVJR^CS^AWBV zQp;_~2Sx^MD-y3T_MyJu?3XdsCY__HELw zj)unEAJ>|l8({VEm_I4~ZQb3ZBS#BGgwf6M;EX+;yH!{5i{vU!@>y&V!&c{n2e{$M zlex?@1>Xc~NatG?Fk%*!LIRdetu^VJ*BO>g zCr9Y|VS6v!ql5DBE8%+HW~cKdvI$&GX6AH8)}yNl&YCC&&kX+)>SuS{#Ao0n?lcc7 z{y|yW;qT82;8z--Hn$D5k!_nh!RRo^4r@0bqsYvllQwQ3*eS{}MRJx^OtrR-jg1MM zT2z3|v9&@iT-hI6F{WU)reQ8xOh%1#PM~J79trNsfO$dZ3_3whwD`( zRv47#R!x9y11!LGg1vBy8k+NG&5 za&w=F>>GP{r4>*|gkI--lf{tQN|X%HJT2%}@Lt77ld^=!Zye+J5!rV}W!)ev4I4Z& zn>Kcv*4ZrFDVU+8kD0c*P`T&2jqMBhKwCkBW%R-maf1GtV5CSy`g#9r5t($^ST+Q@ zqa%l#o0esy!>$vvUDx0DbgeJcZyN?HRsEm@GOr3TSx77eoFI!{i`2-MT&}4L%r)v4%{7iq zTFF@GDC|xx!!yKSvyzD=A~8`+CS@{Q!nU9yOLMBGI&{s6I8{Oq8^Q2XGdjU2^4@*( zj)sO1qNQp_&HpH#8XEcyb_KDjRqYBAvpl9CRtyp^(In? zz?WeasFtHkTIz}-L_N%fLHd+$9r)_j!DF`$oNERHv!+Y=gYz@TPRzV~a^{;4?tgY9 zumJA-@yy+iPtCk{1LolV3y1F?cxhc3Dny(=K)J0hjT z1tsR6eh6c$?S`)Hsnwo+Yy^wZ(rh8MxgtV`N<>_A12)p=ZKW+6wi>Lts$@YuJUzeSyAm*_uuJA(Lq)$^ zA0as~KjEWl#$a^|eO83b60yW$GTWll|F+9#G_uj&TPTm$DJSt6<4q<6DUuM95kXRx z#zUD%MWRB)gLkPnqU=zHh=t26JMx_$NP z|Gob1?eD+3bLQijmyh3n{w4Tt=BsySUU~K2^B13a=JvJc?;JRF_t@FHH$HDNCvW!F zZ+sryf~akJc99?Z6^-yTQ9>tcn!LDp6Vr?m*+?}jcmSmi=%Z4< zwp&?r{*yv77JBD1i&%@N+ZKtg_^DsCr{&Qr($BM6Q8%sDxv7dITXz&&Sn$%dUEJUq zZc$f;KXCJT0yAHcNN9*I#7_ac+-YIPZV6Xq;ffbn{Ti8B!e9z+FB@ezOcSGZOY4>x z7~SJCA>vgr#r?}k5h9X*cXyEH*$8K9rZy_k&lrKp(W~IQuC6 z47=dG7zWqN>BurI^LOi-Hw&;s%~aruW=C1aIUd~FUF#jYr*n?+8e5?y?5&kzF7q&W zZ*2l`B}GrfDZ~7)K`{}DVBLKpnsh2WbdwcxV7^;>orksA)*&OK*xStvv;4KDf=Q7F zmxXBr4XQRH!(fWS^vJML7fwg4>J-O(d&SX+5EE-7xA_45eh-hG{N>R8haX%=F+KS5 z%7Y7UJpAA=`Nzc*55GOtWRm*j_y^?S2ge`25j=|9sb&hMr6pr7-tlnuMKmfXa%lU) ziEfT!U`ZsIM|7%SwCe@pPPA;3WCaX$vc7%CK;v-nACI}dXypwKq_r|?ETIZ4xDM6E z)&{b9^JdqcmvB=p`z-iPAb({vvJ~Q$cAKji6doq5V7}6GHK2V3lGDdpkl7$J%F|Bq zcBScP`;LK4I|c`KpfCq$7W-VZ-~YbZ(peqUR`Zocfq*8X&NF*-7YK5Q$ebD22;;XUHR*NGWFTiHS*^#{$10#mR9Wy z``;w>?`DCj>i_2D^n-&<~^d=|))geJUHV_KA^5anHWLoS(W5AG`+F z{~9klE(pS(ucekV4C$i*`!C01%xkqd?Y53E_2CWpy;-_s5bNEGrfhS_Xn-pY29EhR~L@$DU z2)Al6V}s+bm{=n^j5lLAO+2b@R!U*0VXpX1XVF-~E|@7~voG3(B3WMmvjx#*H3oMW z0^^ZuuIkxXHMkw~8Qccp$3P+Y&zBEURkJP90d97_);aa?f34gUY$U#D!3l@?EfyK+ z=jtfKhO*E}*tjCdSYrnMW>j%2j#z!g5nD`FVnm3Fp}TFsp|tc8K6@Evf{I9WP^)}8 zs~eo9uw#nY7Dto}&2m0KjjUY>pk(Z#%KzIgpN5y%8;|iKWH_oIQ55t{`IO4`1^oQb z5TThe>BFDi_~{+^`JBMZ(?6Z~>2grC*WWlE(Zrqrv0P{f%|?R??3qSpsOrx zt8TVb*_A_1w@N+jW}i+U3_CFV;SPxm>PCoQZ9)BK_cifPA|79^`ri-Ny>lXf~j& zUfWFbkr4aU7N%+VMX>GD?q636(tf<3T#~`(fra>hOrRbEYwj1UEzdY_AG# zEQWk%Gt|hH;Sx=qcT>GO7@HZInuV>+(BOMtg)X|A>xP|9b9ahBc4Y{BRxc$n*BeRH zCJk3_WWwU094XQ@H83I(9IZTwUB^j}NxXXHlBOo8paO1pV5T)vIAN@3Fg_IjrJ3un zPa5I9<|S23h9o9wMk)mowo%WXUKuHsf6|Hgw9Yha&MTibVeqU@glJvIfG44Xc-Lu* zr=7<4xkkjekcbJg5R;agPEp`8$iptkYk7;O9#fdW{OdFZAH+QToj1{4$jixd$Q;h z5DWg?EL=dcWVw+?)D=jB)Kp1it&+#5f2@6_9OUcP$k;PE@-M`xZp zLaajZ-uE|Vj-6<-t?d4}tG5n)edoeQ_s)DebM{YnZ(Im0DhKlB^sPgOZXG)6l9Q5f z?O@Js}F_jnK-STJUk9A^{$BQsvgeQZs z!Kd2f-)w-5lxpDkIp64}rqym+madumT(6&$C)>#69)^&~00=QmksLm$={;)2k8y0lwsT;lbKh-OFL z_~D%_n6<-9HS?To)qi|1&GjD-Qqqw#_A)%$4H~|k7Oj$Ll`oEwG1Au?SX}MQ&GucD zlYc2ig=8|crRcX~Z7Z{KhHW;!X>HMdiO18lHanKNzBQ+FzO|nfEmjx~gG^knhfP%G zf#dr`$47C-g;+E+XIiBL8*BS#%}s(^OeMMNK8vMox=~vBjcRq!UfS%eZCXT3#v*}b zwUOx~BL*uC%qNBk&NvN{ii(2hu|yZFeOLz*W!Z-2VF~Zli$zXB8l0fR#;L{^XjE9q zMG=EAo-wdka-@gX$M;^h%@a+BrxCWEG)KFKY{OzySDyk_ODKXQ%Yvl1+In66ZIVAQ z9uyVQO4iog`LA79NkpF{?9!@wjOk;%{dUn6Z~w0s(7ufX>H_-b(9 zS}&wcH!5*J!W@u^5)tJ1QX)-#tU_r-IKsY-B6F|G%|i1@X4l-U2Npxya9R{48C+x6 zP8GDA4rLJ&qMmZPMKDP9bRIN^UcdS6pW?-~?eyrAN4RiLBt!d8gR>R%FVdf*MxNEf ziQToiqgJWx;%QW-E=?jL9tv(fF$FZXE0qB{p=zVWe1|oZ1YD}NqJQ-U2CZDlC{SXt zA~Vx$nUfh}Dk6f>u$0osO>`IxF>8TEv;DW8MKhl-as@0Kwuh%FI&qA(VDTnHxTWBb zZK8zs7pxmW6kU)+QdKIFf2n|%>Zv^mf>D0={o@GEec;rs18?0r@F|9HKRtNw(*w`J#bStf3? zX;7wehh^HzG2#J%4JMjRjAlSO6{674Ap9T|8FW+l#9L3g6T#{;Pnm?y62~C3mxE>p z6HZLs#>%uNQ$0O{gV=FbC{nEy!a#$&&sF|cl}!s6)AdW~Bhf|0BQIsqAgp!;3eq{V z7{MBbgW?jH7zqS2+&BMlXo$3s-{I{Ore$=)bXs?@p?4Q}>8k3GvXIRN!{|X>9OTEa zGDGi_i{gm2RdCi!hxN^o5=>8PonZ!kF8aAPzgb^OxC$B)K2zl)lD|e`wQkjugI;Je zXgXJ+hoU5%GfJkOWm>7(`Y#q5MqsgeFc^a4tDsO5Lc)XCyQDPi=H^YqcIpML@igg0 z<)Nx8DGMv6qLPsG5A0XF{QeER!FC6KapmJJ)~Q_kZpgz_7*1{{bGt}+46}J z*lOi*2UEC-wzeX?xB^)8ui3{!2noiQaU zB$k9tJ6n4N+Qlhlg*$2;*f&rr>SV;BR~{PLKwR%w@XmVkH~1#Y3=O47MfGar=jZ+i zuZ_yVbEPGk!+Y?G?2K)QvF;4sBVrTU*X-=kZo62S+y+yF#D zn{q^l$yEv+mjAVFk}xWG^nzm`HfooYQjv%dOVqk}X5ObfsBP&5P`l}Td7RL^Q)R!m zJdTFYdMQDQ;IsVr9cOuxCz(h%j{rrz& z*}H&5_Y|Lj@>LZ^L~VX&X4>IKu6W8l;Pw!7Fzfk~jsK#e`U_c6)tt@v=d#n!2_D ztU=AB%@!rA+&o!dx%p)PCM>E7_vXOPwe&PA=BR21UnCuyc6PMws0tyGNLHczDhK`qAxnbnAwk@m^C^hi z-+y)I%*Xd%e`n_C7c*zxYO<~B?pNb?UwLlk{QkhxKlAcSGv{8q^U38qUmkN=XF*9g zaQ+oRl-IBnYc?sEI@qTyTV39Gql68iNe`K;ZzbOCgcMOeSUQ1Y1UF}xTLKwm#LVfo znju+c;YnT3gS9pSv$+RLEplLhto;7a8sUwAzAsQKN7fG_z;Tv1P4&U8uTE-MIWWK6 z0mBeEF9{i5l|-%d7ObzX7oo@@b*-Tn!HO=Cw)T`s;6m`nP2{^kmzI-$@RfPisgmm} zpR_Cr9^VLj!lI*C#IT`yHfzuV#=Pnw2Cwi!?Hwu|t<1(o!!8y>F!sVqk`(aykz-kQ zmCVJedMoBzYegj?;Q_<&y(~LfTIEiGiD?fPxYKl>LUvKoLi!(_AqJxvlc+q81lH0X%sN?7Azs)uU3VW%a%y1tu|{+ywY1P@&8 z03j&JSh6ZuS*lhJIj?<+HYy~-E5u)0m|s^*S+xX4;xY#Mqlr3{fWB$w43^Qg3@Gyz zzc^}(IqWO~!t@}6W$^8;K1-XpK=f|W;2rFmaSy1bZxg+;QFZn2`}w(7hK6RYUA_Co zi!;w1nfdD5+gGmsy#M{Do_gxe`}=R-xH$9j*}Jcwy7Pz6W{!S+=h*&x@&~Y{q7Gl z@4q-S^z5@L+b1TIk(qBV-TCI!%<(txp1ykL+{bq=zHw*#=)Mlem~G>RAP7J2|KRSs zhwr?3?*91Y`yYJNT+ZD>^BFvnsH-DsS2d;>k=@YKvhPp`rp-M?=0Rdn1B@0bKqeW3 zgCcZ_nKxpthP7s@#F#pUgyH6i*9XVPk zVy+{{i^0=6C5sN*OGtl9PWn3*+T^Js4M{GfA*^h$98j;hi&RoH1GF$qORyMwb6n-c zBG=d>z#ch)*_!ngwX%{G>ijXS^VZBRJl4SS-KLM>V*(Gp$+HDITnk7PIC?2#dR+=n zvhb_3gNphH)65V6n2oZ7WbC4rJvcTC<3kJ>w&5vsb+V>9ZlI+F3S1o?;rV`G2i0h1 zg@8iPY?d>!QD-Ir7m1Qi&gx{y)@|=!H(w^}w(FzdT%>ht9ykOza!xB1%x%Ku><46& zLE$8XnCQut`3)#%^rEW$@SavaL&{e(boY;EH;^CSp?eveX)-<5&-TKbB=6L>!tJMX z1lCY_++uk$%E&I2rgQq}5ATuoAKzhno#6#rP(Pm7^5fa6tdfap$*c3)pS(m#h{kId zQ?BISBKgbtsq0hYQ`e@xnEIMbor9N`;OCpEGgIS#-R~S>UVcCI37j0C`gH12 zGi21J{xJ2$Uk}6WU!D47>LmOgugcf^UTW$M7~FSL<2Vgh;2gZ1`l|IW-@*sZPJN9B zQ&;|a5MIF~1Qz#S&O0-AiA=o&7ry{Mm*D4XnBDQIud9ziG+te_&TF5Kfg(up(8OG; zsxSfV6W!E{xe_)osZ`6_6|&%^S0_Xx)#$48+Mn6RQg6y|Q`0t*1SKt{h46<5EF;t6 zrrKm{S!z6vCY#_g=%X2)Hc5N=bcpOwA}L5Pg=lw3E;PMpHFFpB+0X`}h=EgG8y9V-bQE`(Oish4d2l@H89P7vdw^P46FPz+{C(KuwsH|h`SAnWtw32xr6Ic{ouN^s65cA zuMMtKzsU`d$Mk1658pg6eg5XLn+KhP>5p$7y?Gepqtl;Ee+_32OrIvxr>4Kb)6<_# ze@<>5nLaap4i4d+CNqGX(LM*GID7Nh+-RSNQ?my!3-I38aJYW@%jvVzr-O^O4Id9T ze~(gKXNvDwjT{q_${I)9BMdK<8QdgUY%MVw+QlbTtysyo5gUzw5nH?HGAszwVn0vk zEU-#VY78?DCnN$#%*4}Z+9Do8o~av%q(sQ5t`S*`5e@e6y>{g@T83$JpL$XpbD_oQ z(2=NBfmV)FU@eU?zPEL5G6IWrBOkG9_!t-3MHk;jqq30nAQ}435yLjx4MSj+j3P(t zw+W&&wcQ@1TUR;4e4AM51J9QoqWXEGq1#1nkbiM4F&9FfKj+NkLB5O2`Nh7M=Lx#CH>yCvL$1&rQBeCN51}o_u-o zAe_23c^poDH+f+4;NR?(aw%RL*9(FcxQowvjGa@GAGD2byGRRexV_oLcj2bOn$TFuJYKCmr1<^rE zAvwz`rdnIa#>Rv(3oJ)X6+JY?-;F6my zCA>H|1}$TF@fOWtp=u@#h?6*=+Xg%l$mto}C*Q1rfO!V-37526ica$Jv3Uaj@gLRF zMqMLMJq0F~mM5Dnva|TqQ)JlCX(mlg>p%VjrXxd3WHV8zjY7^4TDcH{hhnQ57^x}O zw}*xZ%h+kBE|zeOTO3}FGJ&+IDtrf~wm^5oS2cz-ba^-J`Rqs$KUffi74MPM!47oI zLvS_AV9AWpQpRN+x!{f!zj(({L=#ZN+5nsHJjL{>n@6WV2dUmXGJSFS^Xadq&)z(e z>QaljRwr)%`SfM@{oLP<-#k!BXdIdT)(P?-o<6U@H%mtwzZvb;0g|Cez0 zN6Gril)(IabMv_{35(z6=Y^(9fqCeI664)C%(fA!}-ZSw7`vBnYh|A@v0LU*f;T^6C=0?KX}#I z2*}U-KmGHym9D^Rf*(Q09{AzfMb!*9o?cim>#d1{gl zyGZ&$K0Q<;12pUSiZ}yWQ=o5!V1@Xn;|?s;Ei}(|yPRV&?y`2`zO5aEbyx$#@#I;? z!o{%~nPCChAW%_gB%E1pofb0p$5#5Xs)6g`%t$lRZd)n~4&~X&VZtbNrIKALX#zdN@sT!|~$f8t7!1W~G$ zi|=xcZXGd`;Q6wjlqt}tMX^O6CY@}=i3jQ$sSG5rB=8hysCiX0)&0r$_N2h}wFqQ)V zS3xfL$ddF`pRHD-a$QY=hFd3*-)33VXlXNOftX;?(N>dcnPGiT@a0x>A(z2j-7-a= zd#2(58h98Sf(Qr<>=EW5iBOfmfY9u&B-xr&Qti~tGpp0iuxK*0l-w6_MIowbAHQH| ziUZ@HsC99l+_nf4&Q#Sx81QYn%}ualX_IY6sn zJi1S56;@!JjC|$u=2OIE75zP{^Hi?oF*#p};;i@$wI;o$SEhm*8PEv9J9?!UiYXzd zI^>89FFM;(F{KOK;Eq^FD&F3%q!Q7HyhIf!ChO1$uiE^YJ;E(y+zzhu8Wq)55L7~v zLYILy9lthQob;jFNl-i68XE2hvp3*$LZ!j3>cDG<0Q%^mdw2mW#1O)mWvS^jJpG}A_UgfwUxAoVKa;;1u+sAqVOy35=+7r?eH1wP>{P`iiCEI``{DcWLXvF z`sT{VZ=-oe`gEEh>$h$0>m#P+)D(t+h}Q->U8Z>zP&4imQB%Rz|F|dLzIu)Gxoned z=-3n)!eeS=!b!pOZq!4H9AxwAWM0EW? zoZ@v;PVvP=5QU8@Jo^Pk@EA?c&VQ55z61PQfEB)ZbBZ9RRUc2ID?>VUWI+dQ7$Qrq zn>m}3|7yXEyX@`~c6TNEZA(-zB^#GQSB*A#N_MvSf%FKZld2A~+LE)f(fV20=%PYM z3UN=(!djI3u|&D0s+yO4t9eBhl(iDtSr^|-=eU|##v)n;fy|T9wU}tF*<~4^V06$p zA>hXwX?Vp_5lM*1Yj$xR0U_q9N<+jl8!VuhAjd+JCpv*^m8B~L%iC$XlD?}SiRtS1 zc=P?m%KR+9GY?K6XDaD|k70vkV5uY$f*2E!luC4Afr?ctK=c_L1CbS)4@_K`xQsRHPB*~I=)6oG!h-e5=i$udN`3!|HxX(o@jrv196%s~D#ij9 z;XB9*@RkQ@A$Yu#seBgMajw;_+KQ$4Y{iPDL@*aTfPf1wF8HfjM&Ik4oSzxzTACLr z*4`Fr`HYFb4LxJ0E!|+|0;{df8#`+=vY0y3Rz|!N9GAUy;p5izteYPHj}u} zWI;^Cgjl31rNE7}F(+%|VMJQJ8Y+!nH92z4BSpr4+qAIU+0_{4ab&5sfnvAc(gkC5 zw7K#)_uo5w=jbc9ufH33+HZe*xJCENW>;Lw(@V9f1i)FSZ<<}r#PJedu*z~38&&c%I=}!>> zJFqmnDbDnKzLIJjFzPc77{h#0B(|m=UFy-cauk%vFkcY`c`<~APAZ{8>8lC>u7?^3 zT%q!j#xMYdaKfUJkc@{;4z4O`J#DvaFWfi0x^5MBJ|8eEqFUOBR`0&+#oZH{*rI5e8`3a07us9S%cNRo3A;eq+prK;$ ziY!FpYf-^k18+p7qa2+6WjF7XD6AMy0A#_)dOw15(X59;w}-RBu%Ry$L+{w;Suh|6rax*|w&|um(@2D07cm79_J5Oa<#KHxem}MyZkt{1vnzxoWPZpSi z^S4YLG5GF`L-#~&D>$L`(=aS%4(fw0i&K>A8l0lMnA}BCP!hfg;1Mm!LX>+eS8H^3 z?cvrOS6Mx~bEmhGtu{6I`0dXaaycw4m(3cok>TwSIa0^+9QZGf<;nI?vubz7ML`k; zQBed*cDd8>s?{0yX>`W15K4@A_~ZOm8TP}2RX$HMynKn8ZT`FlascPxqt}a+5x{@W*4r)zRDLh(a5FBnq8Y6u?6QZAb z7l&EVRXy(#OG0+{fBLrgkqAPr-SnOutP4Vus|;fh?eN$2bbi=2nhom8C0ExV0t?kk z8(+00Mc=k0?72}-YI$KLwheZ@Hn1R6N9)v}2E#syX39#|Vu{G2TCDnbVg1cUPBqIU z>$~}GUdP&B*U?t1(>kciM|4BuuHsQCg7|)M0mH>fH&UPrD$r*|M;wt1ymWrPvtb*t zDh0MeO}AKzv}H1U4zeY~MzHb{mKauIS(ww*(GKmvo&i{b7TLtJIntlb>AFhRSI7CA zp&&QO2QS&O!(yh z2QkCr;%T#AgYcen`E+AB{XHou#629hMZf+Y2xs#sPgAl(ovWArgZSgmn0%IQc_5ODoo5v^(>I-StQW|QjG~wPg`sdwu$@iUAy<`n={9rzjNmJ zo%b#@pKCC4^8CynuHC+I;_kbL?;rg5_LcoJCr<~KN)Of2rET0rG29|WWA#f!qOnCK zk`!dGv$miW*{35tpT&%MHl*`sNs$>`uUDtDG7LxspV);3gH7-Vy)bJi!5Zs9dh$BP zxZ*Tl9-TNhaUPDpo4CWJVGX}Pn?5Kem8NIOuX%U>C!Bn)Ro}-BI2ch8PbP6?fPg{ zqOR^Hzr9UyMA?XW=nqX^D5`#O_Q!N1!y9O-KcRr)NEXHmmPUwhVxK)c%*`BFC5570 z(Gzpg#OkV&d@n0S5~7|KXFDHYS-tB#g$I#tYvNoZ?S&`O;WzOSERKVd#~r@43uNMt zlW$I3t*k9F@!Tx&+IJzj)VK3=o|{ER9X&8tmzZ!hFeiL=FoU3CU>|w{_d#I2!N+qj z0kFK>VHzfBwt>u+0go|mWQcC?tP>G7n8~P-&b8p9Hq5OiTVPvxfppSQZj$xR4WgIA zjC^qGd^@u(gR0z3`WQlaDM=w2o_Flm5TE$rmA_wmnf&moUB1Lg8g4%aKI7l7o^^&1 zuzu0rM75TA@m<-)&F51n2}DFmN^6y+wf%Xfg`0q?>SiOG^g@J_5(tSB^>FY6i&3-; zgBOp^laqzjm{n<(O)rGju#$JRrzv<&!Mja5sKJLjL8g|w;h}!N!y@et39ROjW3}Qi zNBxUqYxyKRR|YDdF2gLJw2+@SQ~o+b)bdF!PYMn!dY+Fuhs~BI-o`Xax-n4HvJqw@ zxM{bIk!`fdHL{BtuFWk>Z%QI1N`f4(%|@=;N+_$8<;xcsPhiSEmJHoiXBSs~ch&#{ zw~bn?I`?yDQrn;I=qJ7Kazr-@@Q9am+ad!nQzq&EZySje`YVXRfd$l6!7^%6@iCL$>!$dd1xrPUn=GKRtSF)cfTQE?|4uELb@5XM|XqC+4&S%}8guogVF z9B32WhFEIXD(*X)6d2MXlZS}o z%3p-fP41sOaM0AQOk8z^ENb3KRk+UK z=Mwx}4=iTS*XN{MRdlj%MJGqra6s!WM3&}_^{7HELRk@HPatAV$<pEyx>)B#r}U!J%DubdF}O-LTe+c4&>>np? z-5^ypaFRL}Hs}1NTA47208N)6z$%QKxUkXFX{g zv|CINW7waoENSmW_+U>l>6#g1>N#4XdCowsEVCt}$U?%?U%MtDGtZI9BzSpYV7FmJz^}_b5e31o%!YgJZeq} z4J@Ayr(wY+bhmC><Hg>G+UZnr_T$HMFvxTO_r6^)xJHOU|yGr-)b2T~qO*4hSr zx$-5ARqcX=!#*m+5`vqb$<^iMx5+D+65%ppZJ?1ZB%nGQ8)99No6jR~#R2S@pZ;|E zESdi7=3#7~pZAixY5#W6i(6&(Gr|SIO(L&#L!gIt14L1ua8(`VK`Lk#YzGf@g_097V(Ix zX!#V?W(_k_wES85V;%HzS_d6%QG&mDmSQF8};etCNbw3i-)dD`;}RL7Ug4gX;7Y+^(RFIF}fCY@c6WR*e+3R*@F&A z%wfvWTI(<4w^&^ygKQXsb|lNnC$&ta%BPw9+fGI^PC6j49NKm8e+#ZM=oYblJC(Rd zH&wFIU|L&8^8|Au2E*FHqDe-XVc2Qiw9RJAiPRawvLr^5*n%3BYK6HNDkXTkH1A>M zy8rIScmD9ntpjh}I`HYO1BY)N_|v`TKfiO~+U*;!H6vQ(-3u?=xp49B>sRZzx>CH> z)&23fS(%bYeQ}E5z$(OIf*6mx?1SoZi~8R0@ZZR}#K@F#i6- zHxEBJa{b}4lfNAKpxOI{AAb1C!w;@MIKKbkyXRnxfmc#Y6eQ%bjSG^SI=UK~kbH+G zq7fnD*&Ela2a{PmyX=fJNvF?Ls&!F)ucZf{9r)$Imk&O!dIgfZi5JcJ!w-+X z{ovc9!TnJLLQ6;$Xz`?L!MK35eBx*&kv}2fF}H&xDLk3jKBt)~Lo%#rFti#nu#g_r z!B#r3GSs3&Lu8CwxeDRe_O_lbQq+sK>U`mU5(hE{9=O(x-TUTnX~5TFU}`M9I>~ww zd>-|fiYiTMk%fH3^PfN8n8WC z%bN-Y87XyQOuknZ?t2-cGA%m`4%#$1jt+fzST{|^9Mmvu?a2oZUv%r)Y)R8+Q^b!t z9P0NsEGh+XoTb{xMcOZBaxX`wuBu{^lI zWS2_QIeqkp_elGX@36h)Gl6CFg_}0++q4nUs@KZ`*uruXU<%sQ;5x~0GYyIshV0HN*1iXd1xA$w zQAN^w^XqN&=V{R*80JLo6q8vNgIOVTl$bmNChVWTc;=Z@ zhNk9-TKeFMXy3LPv!&3j2}w>UVr<_QX0Squbm+!#NgEpa^R<<3%hps`xoK2 zimjH}+yz28AM{$l)=Ln-~Z*s1wedPd%i%aT5^hSOHNJ- ziU-3Z8-NhyznBiY?nwDe6IKy)Fk=)Bmi*?K+h|XI`zd*s$!u3ulhJ+v` zR(UU01&jLc*FLd`ma~@&i5i5VM1)wh))#zoT@78-(zOlnJi=enXp{9bI5$t$&)$4+ zgm+IXLv=Oz&J=^O5K-2U#9d`*rfpVAS{qq?7MR2nMWcW%=bOZ{jsez5HBNd}on^rQ zU^v?t9<7-FRx?nKEXwJKK0psn0~yFM9TAuU%d74s!l#cAIVR#M#PwTY11mX2Fm~zG zm87~2W~OyG+1jP4FdgPT6KeWx^w*k&vWOvk!V#{DnezI*0{nKPf>IWd0g(CJ$T z&)quo+O0#^ZykE$)}ceU4jsL9=zZt#iAV4>YQG>%t1`01^K-`+ZS(Ye+O zONxbxh}5IR*1Xv{Z#G*VFOfY|3r^Np&!Y_tlZ66Ngs2Cd(f5W|X@KkM-}m!#uM7>% z)sG+pDhw9!lb(5|pK2M?LOzN0oz$?wnnOZ7-E=b3zjK0aoxU5ktOH|yb2|jCm&`A=a z9^SGA6F=+vnQp7|0@@GXKKaX`{n$(VkBcWBetT*IdGO_x2N&LW_`%`XGfkcT`Q`P) z4^A9?@cQcyKX{w`^7{3MAHMRBbNl~s;dq_UwdL{)ZD z)YYON^}R1jvJmr9Xf25ExLqIRq@QML^AQuTPFz8f3kTKu@WhqLw-=(`T-WE#PV_Tl zm_H6#1>X23K7spxd*T|>bBxcjc^q(tbGeSNcI~JY)-HYbsE!!YE`cQyf)tGj(J20^ z3asiv`SzfYu_Rf$(AN6HpWx=Zh2|OAW~XWJ6k^@h@cKaxIKS9qAV=z?N&-024!>#( znfr6iXd)-D8(GIes8E$ZeRlPSBLl2_+iC_6dthF? z9gk8Swqa2xc9Wyr2rQ4SM#;3O%9yJWT#>7lT+VBMdQFZCvZr#$w@lzJ49h7mhACU| za#)y^u#;yYk#tfMg%cgThih?R%MKiHdjFR-Y4JRJN*3J%ZCHc`t5Iz-1+3P%l(Zds*$D2 zfzKI_3Zfi3%kisofxA?erCRVbVPk3~KpZh484q0rysALC2MoQ))39x&OKDZt7%>?& zLtu9{g3k;jr$mH!veuA$a#b!pC9W5k+HCbLbGIE>YF>=as`V}U4ge~$AjLwn;trIG zI=ZGS-cG26xi!-UFJiX<*~K$VC;b)D_`pJtViDBYu~bxq7ww6*cq$=E9jW$cdm8V3#_oMtUC)pa%B|ec5CZvF z=3>2-@u(1w2~j~3qx)R1Yr8DnADeBpOp(4(Vr!;u8)=qlwKb+WsvPe*{n#)>o`_+O zTP$-rM@_QcA^RHPY9V;SYnyIxxQ^lYqK^W73j~_P#Uf4cm?(nmK@%(nXx8YrN-D{# z)GVdK~*m>UR)D}CFWhN%9UG;&LJ}uG#dbUE}SS&L5gIMKVK{moSa-@SBX`6@O1Lp{C zT%fMVZM51<81>mqC^0e94wp1SXlG87HO}+OrF8La#rO1b`eZ!_^pp}X3d@LvrQ9M^6c4emSF~U{J(x3 zO*$;UoU3*SzS6f2xC#fVl9&8TzMksy9rbhBnz?>I!FeNLmwMq4iNrY zkw}OV1*VZf8lru(Ziz_!z&4y;^Z`K4TwCAoyQx{8Yic%|x2u`yMwXN9Jg=+z80m!R zG}Af9DPLlHbBHE$3GsU*c@2Zm6Aq;rWF=qTfT)BOJyK+i=I~Lwtf3E6lbd9erqf*G z%iA(5Iiz2nlAR#+#h}n0o_Jpre|h&f#`YqK-!wa-OjmoqQyKm4r3Xi@JoxZ!^5ES2 zzr6ZB4A3hpKQ+Y>K~@AYfx+I0JGdRRzDmM+kob~1%eso()HBV9iZuX=f|_9JbUPtn zldOm7{hf)<)T&&uO&TD^5 zD=G=eSZJE+PG*iUZqfDTZ-`x>M7K$S!oC4tEPqzM%$yoVPZq*#72`?d0u-X~%tfSq z^AFMJ`%jx82EE1b0t1`-x=8=dwtmuCes|aAPIzHDZ?m-1WJ5Z;`aAl%cD3P$2!7!I z{APpmhGTN}we52fCW!n_5ET|4FR(cq#~SjOd=!M|FP2YNWH9@LRnMgcxWR~N!}FFY zUmfL{5EaQ!gNIrsNLUtTo%EzTV8!H0jqy5y zbC%(AkQ9>f&~B$w8QsnIplmX_jYUw#Y^PnN#jYEk{PJm|LTfNr0ts7$P5MDD7B>r6 znQQ22n4Tsps;LYB20q&-K0pg}@BjO8@FJVGQJ$VBOR4HVO1?8&3K#;3+Q9B=%xqb# zP~wz~sGv5AA#mdwm)24ef)ZZ}ZcGnNx555k8<;)74HQ=q9iv7Dwo$mC(S#P)916is zjiZJ~q69cX>}y0cBO^3zt8k8L8Dd&=WF+|BsM;)JXV^S^vk)a8GGdd6{-OVB<=2LP zj!>{_>>_M=&Wzxuh~v{Z;lnv7ZhgcmF;KyJw~_8(;jvAl<9;0>r? zw*N0;%z6gTK5J{UWz>8x%l!WLk1Tpj8FoBM7dF|Z*UqzO)-teyRbDj74m)@TX^W~B zBCamu7I6V=?M1+~WlZc#(KT@Gxz%OTVO!PB27zmbTUbd!^z{5L7^qp-N1Yb*c?RH? z@>LZf+RCR*YG5lF0shHAMw0<04p z1>0lTX-b(DVgeTf0yhhBSH++hmt3K2pR`~ySn@(NJP6wANxsidx2NaQ?Q}NXzPpl} zheMrf_`%^4*}@A*e&mo$FH`1SMy4FeLrW0EO{>JW9EgSyh84TS8e=r?6d0HoNFCw$ z{U%dwTX}(WV(68ucW$smz$r2E!Q6hYIcF;7UOR-hfmf}K7 zv2)D9((4pR7dM%f)5n%^Ny-r+s;tTRamaa)8V1wIMuBVsQSlHZFAWp2TnGAfZ_fX| zJM)J>-F^K}cmH^9=A%Ok7(<)3dW@k>Iy?|tb!XtqpB27y_Px88FWkR=B6zw{k_9m) zh|#DJcV!h_JjQ-$j6oEN5R+;}K^KCQsw0Z&Q`6^f9-BVf4DraD&)qyY{n_*-GX2@j z!#9siU!4AY`m5=)H;+`MvfxKj12o4BDH4eUmejv=6=4fSHD*i&Mzu3#MGNi=h8g_w+>!*;CElUb@1EZ zxf*kiJE)v=VOcr{K7|7sP^twO4Mbe~%EYCK>kjMs736(&AQ6#;{Q{i%4uJ(PIlzJy z-u27PpyX;W65oY^?T33mUl~|nLG@BQ?^;6D735n+h^81C0B>!lT+E`^^^Bp9kwLD8 zLS|{u7O99rG8#Ht+T;)gXxxD`HtD9xcw;~cg5w;@SOgN2mx4Ch%e8!o^ib9L-30sG zplf9AiaR{@h&8^_>yxQOm_Y!o+w^R!0~pZHiWVyjGlRH-$_)U@7G+n}wCuBNTE>cz zgr^dBRVHzAB}z)vH0|9yy?hL*QLF)pBtu*H137LYCBUM^x>P{OB8 z_A&=!e&sKe6^q9ks}}CWd?G1FvAiHIheqURGtxOeiXO4-GH|iG$Em-sR+T#5B_7WD z&7`B0261OwD_3Adl!k^j1pJB&_6J#Cv9eaOZks=*I72zkfW=*p8U7E33^XAG;nD(o`qK-u;Ut2QyxWX z2*p$0zhIwv9R~r4ovxgF_~9qymv`QO@bWv2^c0>5RMEq?UU~58mk(Y%`QYf`hsRFB zc;Jvcd~5t6oEi^aRxm)tglxGYkuK^!-%^PSF^^P&RCe6^=!^UNKfZh8!tJY9v7qku z_g~#P^YPs;U%z+uY;%bR)BQ8wUb=nd%)OV--Z}T`%=<6iy>jF3<@5IsT)B7jg%EL$ z0^{1!qys)j_)76#+ZL8974@t^3k43A(X#Dh-~=jCXt0@InM{^X3gAD>o)nhfL~rSWD6Yy&zW>k)1u|`dT`Nf9@aIoQ*doq zQrs2Wa1Q(lM-c@jQCouXb`#Fev!nzg6sm45}JJcR|jV{2R2vanidM1ypdpYAf2d?^psVs#z>59TClxZ zIi?qLJOrOmqw?~K1LW_mHdh7-v}c#UD*wn#txlOt^=ryGa{W?lz;7{R0Pj6k{(c%2zY{kehkGBbw=DjVP_ z*6nN#Ou`%;=Ch@Az0}NEwq|Z1+|-^zhDxfcCGxvvnp%x#!goK9S^UEFvumpgue4_420MJb6dZv9u1yz~B1QM~ikC$|o~c~5O7(Bzig)`KEtjil z3u#1Gh2-kGL2f{0>5LfF)3;ssHIC8ldyiS$z~qmMdm`k@%&WMqhoq+(JV zILk_RsxzV}skQ_h#=E-YC8Qt~#4?B|OM$X_423xVKYMTf9mRF#jsGimm~+N*29>(D z-sYQkzEVqKlRz&LFj+WrZg*dGUsPSUbE`_wSUv}VjrZ8e*iIbB2_#O8i5IYOjKRw} zQRa8?lf~pGM+Vzt@;|)KQ!S~h7s*wn21y_%NL|%crMmaI%d>tycgZZc7kN01G=Z=J z&{Rd&VrTYFFv|jctqwJftvLJaonOw9o;#0SzVoX~(Lo#3lnFYQp&-wDa$Q@8t|=)aeD>hD zOZIWA*an5uk1H5;3+2^drg!uoY7_)sGMJf-9uoq-9wdASxB60?vlOlnSIihu&xOq@tP+G*kIfAgT&dq1H4lLSM%~ z)9*?|3ZcZ6WE-7iGzd|fmbN5NDE7tQPyLsgB!{LXeB+bkap|~Zgwc{0okw{jblInYt^8e3+0eD|Jj+gd#;p3OiA?Xgz0k%axflkDecZ)@KWskr8f3A1Z_Wdjo{p zF(5TeKbG89lpJN|jx$D}*{i_3ZRWo3w+6j@3lw|@aweD%pYi?i>1eetu~ zFJ0Mc;yKP7QDG#9HCQ<-hN$-KQwGLH)kJ`IB!*&WqK3Jiy=`O?cIs(nJERBKI%J3; znz8Anv8E3$n4G-J14S4u3yiFip(56fIDG@OFlC7nVHRJmZy-gr1 zT&9yQbMl_7k?}%Vi0FRfF>0}#B|H%6kuvuys@AdfCb4Wb)kb0X0D8$7h{I-}%pj!^ zT;pz>lMIBCR!n2tqW3L&*f2eJ!_s_{7&026Y_fICv7Y1f;Wa1;HoWBi$hObV0qQIckL5Ml zlxdl7u`OzlJ79to>in0>ZRo-t6@Wwx>6iuyCKYvEi1y3YUzRlN>z!n*j8)Ei(9Kfw zsypFR$%ECBaQ7jH;Je^zSz1D`{p*N)=W6qC16AFJ0iOKm`m59$7@Um!*PiK zq+9(Uq%<=f4Cn{is=~5k?o#Q*s1@s;VWn$|-d-&h&%^Jhkiu#>WS1=Nlwl2-q)V|G z7)V==7HPCF!t=}`2#XL|Lc1xKH8SWG1O5mtkVY6@K!&1j7ge+uRZz>~#SJd1NFs!} zs<)*ShRxtdFEKBG*%noElp`)9(9)IiB$Il~V_vzOW#C!Hn zi?eSo&Yr#f>ea>B#};S*xHx-y@$_?WW^wk}#Z$A3v;XD)|J{`U)#B_Mi?gpR&c5S+ zcy00Yv+$7H?|rs7`}@V&-}=va?)J+cEuQ}Q;_NTrfnQ#IW^wiHu7M&{lSM{{q&jJzx*vcH{fvB z14xxY>lx0kIUdqVs*OY?G;_zog_}<KS?-35FB(0F$3K^3%A$O9!Qt65dxS{VGq4j;_Ti03BlCE&BPiX>O2l`3W@lzBC>koF53`-;0RwWH*VkeY?-lMrEc#c^uoY42{~mfMB5BazzC zEIcq`igOnVWo*MTBdwuiQi_JD`bMA>&ompmk^ecI2vt?l6YFt3^^fJ8zCFW-oxX9e zS{2eI)2v*ttnMtP(f&3ujS49W1+gjYv=Q$3JSpra9_0LgNOQeT;XQa^E4ardyY5F4K*f=6?=3(^5)W-_D-#%P96 zC}-m9AcuSsk3g!_rk6mcgf&sN>$SKJ={eXx3{!fN7nqEm z2(^z(4kq7-m;&1_;KDE){#-L;P%MQ+%%<6z4&{V(=OWFBeT?lSHakWgTm1D2x{>Qs z7}3n1o4+*w;r#papU%HG|LKiWWd2if<4HJodHy<_`{c%%hTy*Y+U)#m^OtU%f`cna z)k$tVKK}t8-uUJG)!z9(;K7Zv^B*id0{K2KMF;%fGYy8s;r^fzmk1bfp&)T^ah9lhb1eoy~t@yyx9(^nSHT*Ugo&px^J z^tsLM*lpqZJ@XsCxy0; zkW@korbdO96(|yIt7YhY3>~MVu!W)W*W>WYQmxpEiS0wb1uX1(Idqwk2A@cOaT=cG zNj9^}`MXzEbWcEc#!|Z8H*ustvWW4XEd(jDq z@7`%Vk-eDxXeNy(rEIDD5Y~*Qk9?BSyHMh^>o$r!%ylPEn^! z4pW;G6(hLAqsS>uiKIm?h~w6i#kq}RM{=7NDX6k`kxYm@Es3K?#c{iiaqN~^ui%+`VgN_b&fXS5!sSw-~cSzqAn#!;j+2ldv?FN-?s@_EIT4s~FtS1E^t5 z8@gt$H5OD=$)s90k75LTq5;^@p{@1(+~>0u#f{DFi#$IM3NKIjiRh6lNj+xgsb9RX zvV&&=TX@LS_B6)fO)d6%jry*ng(-s6AGo=8&YIX1;OyTejEORrZjBs>fs>{hnxYyh zC7D(jVyDtRP?h?_pvCYQal)WMW*4I8A?2u56cV(KqGJRX$K)z0 z=*cxw(9;;TH50KBF18oQ5DO%&c+4Vu@hUPTr$O*_D({V1D0N-Y+fk4Rqs$w9l(Z>XO40%X``dvSZKqj%I;0dV=IA=44qI9Fqv&i}x={ z3TfDNMNK3WP0Ix4RRD@UWSl&Waq^7XIv&A&Uzq=7{t8_90sQYhX!R43Cm0Na zpc4?}*IAcOtSK;40~nF>-cjK8Gj}z#wOK|@u{sG=O|2}++J@(1`rvxvC2Ym{+eFA^! z+Kn@$2fq3k9yE*Htq@cH`>7Yr^us#|jGEac5qtuD3r{5TpW$DI=Y0TOjt{~Pf1dp1 zPxuA=NF}hw~rLUm|exX`&iRB8BWGI#JE6pHs4>hUK{>VL*rX zqM&-`UtGy*jF@KHe>fH75Q|~Xn@L4)4<$zdELy4+gJI>=KW=(?tO>S$kp0AC4%oAk zLK0f6aKAjWufHC1jVz)%Lc5rB%WTOLwY;VWcq;wRA?20Xc9ey5lubZ~V$(gF7yL+Q zbi=SKKC=wQffMo3+U^Hy@P|Sm-!3H;Doa!{mqrgPbf-iqgSF1-wzBjn3~_gU^TPjL zy>#a{e<1&O<=5Xn|H>91ux#Oi`^5|2{_(AEpLzY;Cm*}>^y|EUO6ysA1X}Ax*ms#;NsmY_ETFUeQ6gUcx+CPAvjW z;7322@T5%!#c`VqgBEn8Ox?RSeeO=OPkRGM52$aY=+2Xzn4Dzn2zSe)C_T3+wK$W<%bMh-%1yDhNBR zB--+}x3$iSXSw{CSK4r;Q-^i34a{;|KE8Y=HypA#gVpKD_B7^5bq}=Q=c#|MkC5en zT$S5!YGZ5aVnl^_1u6Eh{=k318iDaAr=lyJ`0mbOVUZ3jht5hb4dB5LC=T{ zjdLGmBS>zss_Qu<4KIIES^ZAo?kA7VNUt+6O8F#lna9>*XzNS z`K$9EY_aJx|MvV9j3+@;KJepl@6CU5_KlY@+ynb+k;%NEvKU!DgJUD>cBvbn zwzWBNR|Tx>FoOmeCUMr|0$UM=@-UYrZj*l3!aDXsWFe#zil&;1hIk37_2AEEFmSE4 zvo=If0d|8nLiUN(N+ve_ycj@UE6?p)!dr|EAG7p;pj$AILr^~mDT}0%h>yP(#dei? zO-uOy%>+I_7uAdMs;kt?tWhwN!SHaREg#&O zUM)RqLNHk`6I&huLKY|I{%jos8Isc6pCu_#+nJ&St7(Zd#VyOGo#54Wj1RRfmgC16 zJmxMemeB>aYwiyY%L`d*7tW4>!5z)Q*|A?p*xFpzfB7XLrP4?at-}1$5=Q*3OHW{> zcsqP(l!V7`UjE%Tr+)vyA)7NdV&Y`=p;^RlGd~n|)D=Y88hOCb7sO-Dtqq=)-5F z!y=EI1F}}~0u@ZEt_M<+myW^>8&y@bRBXcLA`g0{wriCn^3A-gE?IA*A`gR< z92$>~rlutmX0{(R?(DFdHnW2&eA119WHy~tO)$x{p{&tAFcjD%8+&AEA(=|jh%KZQ zV**VHr=u<5^g%OH+2Mf<{4p(&0Tp&|Agd2(2AneovqMHs%bKZ-F%$B-Ruqe{kH?D< z&B7JL0;Qh7n>#rB;ic;toLWLlY%LolC%pn#VJEN|Vv_7-P7If4Q{&?n^5aMX?Kyt5 z8v*!8b?!V&8Jb-J^6rCIo1A;Qzy&PUHmG6<;Y}ryisU2~idf5wdWDbx%6HQxJ4ud8 zTA+@ycWl?n_5bP83bo4uuh$ca>IwJ;l_0QTSN0aH#Vt{~+92q-jdJt(+zFT7;#%b+rN;1pD~i z6G?YHO{>Ci?vcVS!my7llBMJxezNZS?3ZjH$n<2U0vOe)&}yJ&D*84dbui2%@=n2% z8y99mp`?^dGPZ7hgjsOFcGg{ia%nf;$c1dZ(6s}?W<^m*8POfIv{D11WCBHbsusZ1 ztRIAvik?t1@o5A5{n8ECsnE56!3e)l#kZsW@5ta5UcMp(~;aX1vTNUF^qF zfP+G9M9S50fpQd!p=fO?%7ay;!J##h!`cPk#^BcMMtuefBCQy0R>uRqm4K;Kh{>?j zz0JhXlw>;g{J0;*GEMD`m58ZmZLYUUSOm#YR>^}JI5kCW;&@1}UV=};V&zq1e*0;e z7sq`@$C<$AR--$3*%*^~7`hZ6jt2G=tD@cH+t(kgQ@+V6q3%=e{KH2 z;4tM@V|d_Eb^=;EML9VJL&YwUgoZB0$m8S(!jXY`(Oxy3S|yh`wdNp6E6KJnN2FAv zpU;2bS9@Q&@jUTsw=ZFBHbyo6b<-;w^S_zDd}H>;nfX6sQc*oC3hUCJz=HN0r|UKB zKC;1b4pO)qvV%X2p7!#?Kz!e5I8pr1;by$4DQdda$_lMI;wK7<{7<^4#NQK7)(E#A zBKvssOxK~Gc<`Dh;wQIYmnRifZ;hyLhn^~|EaTMhb}{y(VkBK{tEF70DG3_L+JVs;geV_zjI46fcH<9pi7nkb0X%Wc=bbrpuil>IVRqdPKG-E;6#;b5D987%bzRVBRFNoIH~ykQYkPbFeIGYNID-bTI{* z5}aLeMf;*cH~dwwYzm@!phahDfjk?M5IPp@Ji7|yj;euPb*UAVk(Vd0I1 zi&?U8;pP*6KlSXw<(p3~yhq^t`wMT}Jo~q2$-?h$J`R^$CJXQW?b-i%7e3s45T92JU;Kw_7(1)FX zA38KSBHbD6Nfo}XDj$@!p_pcyzay;vIUwAu?^Udmg@O9!6a<{?8?q7U+?j!`(4{B1 zPn&}*I;?*F=uuKzVbp_EZ&>oK7c6jg_J z5pOw0)Bt^2ik6HGPkuj(2h+>kjV-80`RIzSu7`x^kZtjbPtY)CQQIN^N=9J?BZJhX zTTg-9fJE6|wKFNFx;iH(ip^BCHXMYLeYHx5Bd7yXSZLva{JuJ zw?2M;@zm_?pZ@%-r=Hwm(!2f3pWc4?-LIbd=+@7l{p!kN(PO0^Jy8@v&Wwg=v~a+J zD4JAH!>fBl&^cZCzX{KfU>Y%*5n7ROSCt{`#h4bQM1+B2-D6|)VS zB0}wmA1~gGMVOw1w>f$=OZE#aJ+0%tt`M~EQuUeK_#plU-L5LoVZ^D0X~Rn;;Pp!f zC7pD$Ip~<8WU%PKE3j#}Zh+dwd*vWm7dGyynQx%kF?fgtdv-FJi0&*|>EnjP(^^9} z^goA_xsgy($<{pH&`wNWrKcr$O=%0nF7_RnMNSIBb{U@E?CQsw;L~HYgzmc{1@pr$ zVA~|?!yF9ShaE{A=7L$3huu6AS%#z3lk3*)4OK~LZMlov2HAJ4zqq7WuxoU~BI`|d z6kBplu^C{Jwd(v^K2vCPSAR;M@qcc9_jQLnp#jRLNY%4zFx|N&rZz53NX1+-vn}Z| z^sOz5M-CAu#=>}RE8Bg0EDe2 z0t#y)w8m96YhXwHqNER=w$Q=IERj-0a5g7 zoy|IJO|#C(bn-xM6D)Q)`F}S(2~q-a=MZ*7(&+KLt`IDOz)f9Ax8pN!YdGIAX_qPn z_j+>fyn_p3&!nXGPdp44MUNJBPs~&%uq$0;?dRoz0~1s zDq%J;0w-2c(6Yo%Q&(g_tzIzUfuoT z+Idd?ulL2Z-@|sfpWD&tn)jEy+85U!x5-~$_~QDTh2@9D&xh2Ie@4q}BTlymVDg}{ zF*`x zVZEog>9*LR~#vK$3?8FDMig7N35czQ(bWprveNqsgM*YC~`wdwh=*vPs);o zGM~n^Q6oj}6g&s>j(ePk>(n-l24k8iyfe4^>bEe*ce3{NdnM0-fwD8y)#I#+E?Fkg zXVp_rc-hP|ea$hj(#Xi-9}&lyEKr9LhnHX}7cSYeXU{r(%`gm-if+V?4mE~>+A<6# zrKofk?R_P}j-$Hka#e$PE^25z7Yz}C)HdTRG$K&l#pQt*bM@qBPO}3PI-pD)SiB`E zc}!yY=)VzK1RGS?VvUOn!k)CaD}1UW=s#R1{GiO-n;4ZDFGyZ>oxa_|g$K;ke{FtW z%SvMdaftTz7vW#Mvm7E^*w$>(JI!o%5|l9-y}EgX3C9YiRYy&r*N`F0OvT2>01IZ+ zhvgiHxo>rs6zSerlzyb|&^S3zWj5&nJ;dfQke6}-jMBqa5ZWik8mt*gBE6nuq&oc= z=usnllC4tubEv)ar^x}jqeUzmQUY|q;#i8dk2z>}MHUB^|Cl-~d8(qP13~l#d}2Bz zcA|y}mMJZ^P`0Z{+J!qnYc7%ncBz)cXqcTwXtPsS6*JSS^+u}ha&cu}WsdEcXq#hu zU8GZ*(befHTrgw(Git9ZGy!fFgqf$fMMsGN~S&F7NO8_FM7IH0V3Uz zJJ8noIbn%P-hb}KuWf_Op_6lO3xfTUC3|2!pgW0$SOLr~P-)Y6Rj6t@slbA#CXi!7 z4=9ZRl`-^IWMHv(>_Yte2Y&_{k6Zzukkct@(dp<+iQPh=0hz$`ORWn-ee?ivFoVQb z)JV>va)~)2pOMQ{XBrwM$veQ5vaQR6A^1V_>y5?!}BpE@! z3{3xB60^f13#Z@qs1#Hk$ZWdBAEOX&#PV`gF# z6Xh_44m1U?O3;rs$))3xDwVV(X|;@7k;`Po!TlG`R4P<%v4w3<2@MF(A)__Nof5~L zsp$IM0hii52w2x^bgrK;)~KJ*F@R{KVrzHX^%K^2)=$7__E(zSv}QZUg><7Qxpt_X z2ngEGV z86_FNIt|Q%12WHDe?Cg#)?FQ7zVS+l*n(IB3klJf!zo*H%e;#U%)JrSetv@VRD_6u zleX_Tt?(&Yf}g>qH`95Og!5|&O4k%kiw%5n-)5Rylx;N3kB)RRsWuhn!3<||;d0Qn zq#ehH->qP#j%XyQ_`t14sI=?Tq;o~QmW02erqkI(wtrwKnKcJGqO3& zP-jA11AA95`DgVm&AhJB39jk*4bjZ#!@%+wn15vduD~cT5)Bi)eh_XX7`l>JmrJ+} z!WXwQ934l5w2C?E`d=XXWZmUO)Qp*MNS=fCw$+g4N2#15qlHlhm&8aM@3(n{T3~&- zwT9%(#{4xy&tC&<_xSU-z{qhF$`9#bu>8gKH(`d@hftec20aJ`ymIG=!pN{S!ftN0qDJNPY z5#Zc40$KJ(*d%)$a|Dv{(F+dxN%lojWL3!uv~G?~unL&rF3Z+4dyBAV!=B`mwF}#- zk^TLnL!>IIp5@>0FZEz2=u3UOhG%vKefU^mQxDnv=tsH!iR^(HAyFNG$GNn$y4Fd0 zRMjxe?%8km1_21UbOqR`!XZCI_F&0rkwwH3I^v$yERglWV$WOeA+`MdB)i z1yc6Ahj1I-%BAFyM{@8D=s7S<9(|M`6acd=i_X1439aNojD`C5+2trt7J8cuda_I_ zI|+W?O)FKW%q83OQSBCEZ>1fPWeHP=HD{kRLN{vY(2bcinF58#NLLpb#$ z=_SX@yif)Wp4mmb*ZKD?Cfh+aNh|dTQ-8t5`X)WP4neuBfr?)YGS@VuriK=prYU+` zwakON8m{bB)Gn1B&Nd$V2Sj9UlbkO{WgI_uLQ==^tPSKj2cckwcI&)37S>U*c=}I^ zXU;C3zOs1c;^OQ(Uw`(=t*6gzE$c^1Br;@dFh_E``$tBG_U{@Z69*0*L|5DX=$N%X zz!g#S1~a#VeuuD*2LmgXjrdF@2iY`}=ps}JTDv4)TQznDLuq=N+cQhQTT!H!zs7^3 zM<33TL191S5_H8P@rfe}Q~sZD6tV~QGK7?_5!e^F#bJ~4U6*bdw>@~GvXm?`ab#xV z2s~_y?RVq6R6)kTVadE=iwV1$cKB-eq@s zLxd|1cE7O>wzaPxUR6vcqp!iWJd-|VPeY86(=81oP)Q_Pcg1Z9Mv>qFhszk={w@g+ zYv<#8Nx)qA>kFX7lu3!tohOB|SDU+Nlgiwc?KoXaoUT!36SBz0z+?lxje_*SwF;oq zs$hqD)o%Jm;%J?QsiZQki(V-#j4|S|Fgf=&C1_{oE%xtQVXA=Ig?t+nMw1U<9*AGI ze3KorxI<-89E%b5;y>}YD}uhOOn?d2KL{x!@CA2qP3Pca9$2^u+*&9B=MC$f$=tOa|-FhtPsx+^p00Sww5m z^rWJyrlOhANut1@XEwzM)e*!VlNNMoL&h1ik&ESj4kuv~l*jP(13QjT%l2FXx+|+^ z@&vI)Sl;38YJtwpfkKu>(C;q#MPzps{L zl`<8SbW&3^BS3c_15L#6VfK@)X5fQWvc_zIeqoqNq{2?{0<~i}OuBjwSgpm^`RuD8>LLRO{JPqCcydrk)+v&7rKGG}Bp49SRWgsDWCN z(2JOs0P`huO~RVyOIx1crq#y2!-(mZwy=t1l!1|GIgTiIC>mLx857w<3W9?DN?(r_ z&;)gJ*E8quv+VdmSVxQGhh-`v&*Dz`@TS>0PGy>F4gsVCDf5VrbeD>9g7w*wS>6VykH{jpv@cyrS z{$fKyt>0v!i5H-WuYCR*`TWh#Uxe?^!-*FOwDUzgck%O|eSQwUzDVH1@8I7{aN~=8 z@Z(s=1qB^u&c(=VS|a-WI)oBJOhlqhp5<5UP~E(vfBaCZp-}H7VK>YKH3Mdtt|Zfnss&V)$dtk$ zFheooQ`2FYv)}oAE1KIYsP-#O$k&MA79w;Ny@Ffs@)$O_mR7Sl{9qbdI-68=Et^c_ zQlQQm{n_FEq2&7J7}9yh&bq!IJXYvyukvbi+p1d;DI`8>gVc_cKZfC3FqDgu9>)Sm zyU2@l)llwco{UqF-!#=7(Lf-VsPImTHTM`M0C-5gXqXu@nK092TG;IgzE0m>;S7Vqd~#;_!_lKe&V5{% zD$Tuamw1VkScMkK?!OVoE0jr|BLdo#upMS^HKdeb3))q7El{(ZDajp%3doQrMYX=u zMfZI$`zK;ui=&wVoiTP|;C<-QKPFC6dy2&h;CIWZpF|S(Oj~)^gDjRQ5Q?zlB+zzG zCI}52+tYDB$OH9=A}xRz5G|L-2zzxR1#hfcVR3Deh*oZ_1L_IxDZwI~C4WEl;{PE( z-Z2c;V6C^tsO%M1@y9#Jk5u9(ZrcTR6t3~Thoc9^|9QWMOwZ2Im6N-FFt&SWKO)&U zUcMLpIrTl9=sQVOa2t$F-A86fZ?U~<1Tz^%sh_HE`_N@QhxU<)!vw3!SqwhXBXk-> zYnOa@+PGN(NyJA_7~3pxftAtnN_^w80+LA6ok+I=S>8Jsp;1p#Ogoj#<&z94pSP4G zTVFr2AiIGflte)FL?$+T+x;-5U~}V`zlrd}JPu)BcZ}s^R?`(d6Pt<@{(jacx36dE z4}*bO5OA_@$VTvVXU2g|i6RO`kwIG>oL_hztgadu)|EULn8Rvgfy|bY-$78yb1Y<5P-HrD{iPGTe(jNbKl32CKDdC5llY%YqJf;+VDz&IHj9NuDj z#DOZeQ6H+p6Htpla)mO32HF(*!6uWe=;t*wIX|LX!iUh`P+98Ql{|pX2C=&K^h<;3 z72Csmmk0}Xk|SdST=@JHwYHlZ-)vd5uMWHhBR=Q~8DYb0TEd87g*b7Vc4h-T$hH}F zQPy=uYXj1Jkk!r#b-*6qK($i`bKGpBfC;-z*1-pOQ8@l4xV-Vfp12wYfgeM{Ad;#i z(=$y+t#|~@mlnBGP>vpbm{4TYp1Wp2izwm|_>nyPlS4)BQ@30E#ActT&EsJoEK?*W z@-znS^DfGd@xZj~)jNX==xbmLBoVVc-3HDVdbRF{)fyMl3ax_TP|vGG7Lkg`bBmGQ zUQ+O+#87zsyP_aASS!bcoFcHNEnQah{2+PYLBd~#3qr^uQhCC35eSu42XWl|WY%j` zwC`df@NG1S|FCxAcp%9k5lC`Kgr4?>s_1Qbnh)|qZ1a*7(wp2+f3;yMdRvv|11%MS z7;+mML@+Xn-sU7-m6dwOIJCtDg9DTaOLa^fc&-m}#PDi2&PD zwy%Gi1-u$i%pB4o(~x9vbO17Y7F0=HdYS6^j;`Q{JHY z*s?lSn8eCxvd8udSx~J8!yhbPgjj$M>}`7CXu?(#V4&$KtOn6^V||k@tr%^j#}95s z^D;_fN!TR^I%LB^8dC_lp4l#{YL?TGv=PxLSEjH7cS^*&#}!o|B|}NJrAkhT@ayJ3 zo`34jYag#7lYIx(j^aKc*+!`$Edic^9QpRCD|eoL{U1-h_3d+)?!0y- z`sxN!Q8R{;PAaK&3MJLSdjz`jFl zC;)J)g`SsMLdhi8M}0SU4UO}+3uTys_k-Dfq=t|_UC9sB8nxV7NPMLfvd*Wp*t~ZS zR4H741^V7^W0lIO=#!{JAzb9TsuVKEYwb>!RSV(9DV9u>M5{SBirwPMxi?n57`sZL}A|qpS*U*#^7#<8Xw$-LPINIivT)Md46;)wzmj)~0(1b`U zYD!mvZju-h(-SboS_?!X!DcZ~L(uWH)~k3Ed`>T zpLn#m^`^`OSV5GmIDMs8%GU@Oj3tbnvl!Hw<1U*-Bvft*hfVWBLq3rqxKkKZ)>5%q zU+!DOcha+X3CYSf-1IdJMa{IT2Peznr_b$QU;OgZvtPaP@vX}*ZN1rZ`{hfwU;FH< zr#`y%+^JhHoxSzzCnL*aAD{9Zk592+%8ilk=p-`IYsw<2WVFzj1s#)HVJVZ8z%-d; zPJKb>tgQ{TdmhHC^`6;yZEcVVo1puy+R@+s9r8$n5|&g<#JCJ15=5sJOJ(69$@qA` zD>OVlLrzB)9;XkqMoejgS(r3uf(NasZ%r32@$sP9WMMgxALoNkgTIkdaY8Gaj+Nn? z-TlzwX{tq!e_@etv1SJd7o67CRkP4dU6DIBRRaRVWNNC;pgYdSD?*xfIv#}l9$Su z5bdAcVjCF35=>Et9w+;kRltL+fCN*x9`x5^b>?yDRuvY2;>4n(@Qg+i{lf zjI#`SHs)-N(BnS1E!ic79cN~CCDxM05Ve_(O$7=Y?c(f9U!Q+=>zROyr$1Ue^Q*f?|C^V8_vL4&zrOIxFRwkbIQ!0*S5JNY$2TI2eV6l2ZNO9@jzG2K3 zvx@?ZSWI+=Klt7EjCy%0=@i*Vt3|%~ok2$ef?E#fMUqi;=l8O#0%K84bS}Jw^J|Vq zus`+n@S=}C@N%vo*-Bd%_fy&)ly`EEIr1hC#3(7 zpz}Kzlt_mSOQeyoKHX-yMs!l(Y_AH(r_*Z?wyol_NF@|ClYxJ_qNQe1UC2I9yt%7Kgthr6)qz$!r|;Pt1Roy(W*-!+r7JVan1`*$5Z zs*oN{H%v8?e3r~y?x3zR*+x9<)94aYAOKu#7(6}q)+#g8K zAhTifpL%$UVF=YG-kJ`zqf+6sT1@L5^XGjLm8E3DvE9 zgEZ1I=z8mf+Ky;XS~Slr3mJAcIHsM(ln_0&foon>6h+m=K&-`GSV2&k-f9b9r#5t9 z9c-kKs3CJm(pe+XLEG7$ag*HQCweDYB^F$0S0|jIDT#DjQt+BNrUV&|n1@-IW7HBA zzYMv_Uoa?X6}0afsd&d=9IaMDxVbYKKqsuXqg_Hf26i+HXvYAki_8*splC3IrMvWG2I^E=kPc?FXQz4UoK}hKiyd~R<-6UrgjyX>44OcHrRnZf% zX)o?ayC1~}&Muz*{o4NnXg{Fe(TKhi?i>1_1O8_FTY)9y6dJ-`<{3szx?EdZ=U$+ zSLa^*^4hcCocU>Fxh0*_u`rGv^&`&BCrQr$mTNj?k`tB(_tI{9Delcawwp>>Bcp{u ze67e@`o4ghDMn>P+UL43V~b4sl&++gEdQ~s)%T-RPLc5`cmoz)YZ!zL&89(URjo6_ zsqgm;2(PtRY4X*T~M@t=y9d<_5a zQ9s3J|9t+p+1~-jbPj0*njsA50J?OUicj2o~;j8P=3;v0p z!#CIcj=c(9a20zJzWAgQ2G(~BlKs3e#hogneu`6k?J|QP6`3}$)qCd|*+nHoi|-)c zYGCq1VDuoWX#MTy&aK}SqBT-Nw2&MPnXH;n(rxTDQ5V{xSG9Pdjx)KH7TVpyE_#wW zjbbMcv8j8t1ep}V%~Q*YSMfB=CqWXtP~Ur&7vf~+thwac0y33EI(B4kJ8kZ})8=wz zDlri#M;&wzjdNULB8$9J)?9Rw<_soZBvP#l>fPRWVVC*L-%KvS_Ord)-izJVY^cR% zNVPbF)E{Q1Er&L2J=fUW`4YD&$y?Nh&6O5uC-U~xbH#qsh>fRkCm9EaAxXc=EQ2P-x$ zI>h2rhyt{9L64u!)IXftJ%c}BvL=+wy~=!8UoK2@8ccSVC18u~9i%Sp9Xe59@G!f; z`W)D2^5R5f`5pi1st`gqoO<W zk`>B3Nx`Cy^L@wl@;eEyU_}uY`G4OM$M4dt>ROpeheIyc=ChXUA+m ze{R{6sRm*TYUtpCX?CFgwylsN+G9_9bt1|tm2d^Jw-DzTtb3x1E8RUYnV2={It)hK zu{?{eEgTdTcwF<8iQ%&{5i)!`&%8}bhcQ-ld0}JS8b%`P)1tJh>D%&Q8(si_vPmkb z_*@L2&P|jV$s_9rjI=4{qTV(xDu@^a+7GjfC^_lbzEc=0(hrYq{>;}7`_3}22MonA z*n(=LDw)HJ41sPfH~^(wa^%Q`9!|zk&+C!8q2h7a+{VkiS`F+`i2(X^BAgn5lFcZo zbZq$dZd5)ts9m;;AHS=Ix>^6I25V(Q254cxsG5>)iw{s4a=& zl1kca!!li!N$emml)=OfpmXBNouG_VWoqq4sDLHZL_54xn<7y2HPcw+hNAG z8U3L3#Kg$V-_O2q^e9qVlrWt5Do60kUqAB5Kj*OoJ3A$2j?rFqhS~7rWVtsnGtGMw zk3LFvgY7K-o$Pt2LiXZnnlv(oLdf^OPnytN@Wl5(^zfkUKc4>R&Xd2){_oXGcYg6! z-^yJxJ_(>rRun=|g}+C#%+7Ykhrt{Fe(H~ZKlR#1I_E-ctP|zAb9R|b^SO&nOmdC0 z*_BR+JeGoYu9V0Gp!$sP<0z#hQ|n2o7#E%__+;8HW=L?Gk(zJjZu;f8Lt@eZrXopE zEeqRBrp?$HNiEN5NGW<^63Qr}B+b}lxA#M+ub%LL9&KpD7uPQr>ZsD)$a~Nm4plvz zM-?QZYl^xJX?j<4v#GSmkU~ZayToxC1G}~xS=lJ*q@uSm^@Nu6Fh4<{7kpS@$tEq7 zA|-ihVT7p&+W;6Fz2rEn=ax;`;y5{8rY_+QaVRhD^jB5?Oqo`z%!Yd;%lgdYc_F4A zf>E%9&%Nna+ZGHvDJD63^tcTwgD2}~=@BjlJV2sK^hng-~|qSfm39H(aI|5E1PSz>Yy9$!v>6a(G9=A zjq92i$OyQNU9@bDvyvs;7(i01n&HtCU_mHmLdm4tswgAyEcM&)!&M<;|0CyE?3#`H$27{%{hkx+i|emO*OWWiM2(~$ zJzUNYiXy3)ZS{-Y$Cw}HR%N3B=KJXsb;{G!=1WZGp&;hRE0VkA`x;yxh0~1)ZEvPC z@m6Ma;w>D38O3Ph_}-?3=|l-Lm?N0fUWifd+yzn?L(L{M#ccPWgxxq{waSwIg6mNW z?0lqpkkp>yU0k|1Vrm_%w*i46;JCn%`GfRuyI^@WH27WT;pj#(t2BwKSMN;Ops&u@ zfYM4Ss@b~DK7c0j5e|1&kk)|g=gh;{s!MwIL>Eg(b2%M|j)$H{VC2O|YEOYzeFOpHDZ-IpldkCz~0-pqv-d$>3Od!VKg9C;_L>_jKlBy}0HZNk4 zg)i8_%sKxT0+rN1ZFZ+_3ktC3V3bc-laZyMGdo~A8x1e65z>pR`pG?d>nsUPOub+FN@5`%C ze|h!P?YGX~e)S4`7u~$$)Gk5C^|tIg(Ot73x0I4jZEFcl@;c;rRC3`ET5$Qah_1o4 zsRY_86w}%o9Md2iEumzL*x5VHlPZT{46xs-D`ct+Lgsluy;2d3$Cb_^Aw<)#T)0+aG2t>;eUYhu*cF!*Y@mlN($Wr26?EdhFSF z|Ki^tp-vf^<3iKMkx*O64F$nRNtx?m{J4Jt1S2#rX;DNc1?~z|3T#1`k0usyP$mtq0- zTA&yQjp6hv#&CK~W7tS4TARiuz0k(6kFPrSi<$XPZ=9b0c>V*#8$~$Q&k(qD{yjuz zB`IA?8)m{JH_qI60^v;I(ogDW>8tXorqYUr#6PN{B{DNjM@w8>EnJ4)gl_&FwBsVS{pMK$9d&u(%FQQlK1M(WPZBu#BOE;rUtf;ySP>HaGO!N9 zP)P)}t3)6i=O2U_U|Kez!iKl5#MULcd9ufzv^>O=C1CZUI=LkhTw7KEVZm!LCP4)( zF;SAVTIL19?MWdk%k>2j_tAVoRW;pA`{m**kgsY-R+xe7l*RE1 zY^C4pBg8$<=+vec(V8UkxiXuc=F!D6EE(Hpi{tPXCFVz6L;&t9pyGs8a6mk!Vi>V4 z#|J9ys^7}T(096$a;0Gs4@}0mNeD2FjxagJNKe1#9H(Uq%t`1aGUS6)7RnFf%-S

    UdG*7f5Nia8Oq*V9myUCRuU$CYxRf| z7Ex;c-2A2a59dFb|A5S2^@)6zT}2;kDf4swck>^>>!Zu&ED{o3BJ=MFYned~5dCx8MHV;_N$% zv%g(DJ&UBUXWm~t^NYnZZ`IjOzj@{LTR*+N>9x{jf{Z})e3>$1arUvr+1LAuq7Xet zLXu;=QmF-IR5I0ZMkP~gNeT(h+ga@*)ME|Vm>WUZ371KGOT;r;k(nJA^e=%)XKZspQ$T0f9sEOctp_haq^hZu0)4J%MhXkI*QuoH2}MnZM#*ax&jG6=CETI$w!Uvl3A+m- zh`!{Fj*yi9UHf`K&voj_v0RU`de79`9-k`uBjWITWVUV4HglBvMDXr>$CR{wK2p%}1H8KLPr zb{ysfSIGL=pf-{7EZ0M_y+-_{Gvv!=Lu_rOmMYNRoUmv;M}CA=I>{yyN~@NDVehAN zsJhk=S9F$e6qc@}GwVTU-Nt|EfdA5~COTI`!TB|%ky0|rRvj-iRZ+DpVAHAp)lIHU zv=m-gMKa_Hn^W5%!^`~aIm^pO4}Vpt8kiHw1{`BDBpjoLm@j&|Rk;hD>IHLR{+yq{ z^~sHA&~^8|@3uql*XbM2tg?n3Jk+*^9rPTBZ;mx={`~xBz6bKt`D?`I8$UDuXE^aZ zxiJeZeH`h;;rD55@r}oc@7wzbE{BWHA@#P;%k7`QFUb7s^BMB=6fTPy`9CFAc zJLE4#**7sd1FE&J6jVUf;Z+L@Bkg+m6~3pvR42FvcGKqNV$W^{O2I)|9FvVnWhwZ$-t|_yOqP6Kd4{!+92F;r}q$;eMpAbZ43@2AN zBe6PWZKbsJpjD&RY_nRHjB$n;Ta#pTHM^m|eb4Rp99J&+`Zj*s9kXC@caiGPsvQ@i zNDeYw49UNc9EC9$y?#Me6AD-jp!2KgfRdDIn0cv?Rvt*oP!j3bNkQQ^MqEEOI%d90 zG8|0Y_JMg1?CKvP<1plXEPUqNy=2bKu|Az{T%+nj$CRd2{5s5WUh-+z zmd;?s{n9|$g=uj`NlU_>34Dg3)JuNso3&CD6@rZ=$l`%=UvhUXWv_j&@5ehfc@(}= z4YZ`w&|PF03@nHA)Jr#s8r}6e-Ue=@1>haEkTF9Y;f?h8dX%hyreyaX9N%IZ(lGVn zfE7(m`7C=hFspQ~L3%5_res$*U zZ+>xMvrmfE{i%nlu4Hx$J#{Z;HuTb_SVY6|5Yq1eLNQ=>ZcT%JaP`yWLZX7CjsqRg zMJV7HbwoO16=`i4Uk|Q$n6N~lEJcM9YUTN)=WGJ|K}#rlW?M#`WCZt>YYr%N8!jVF z$r#%xu3^dvg46}5f`Y+Oif;CGDldqf-lhADWbswCI`~P@*1*tm-P;vBySonmwfWIe zI%wB>J*!64Oj3(1BA+Wzvf6}}2qp@p8p>TNBo-pbDXpl9?Rw-kH9|NpwW9lsyQ@U# zs-m~$uWk#cAf&?!8(b<5($N*OEra!*27>mpDz&ylKx=ePcY{9xqVTX((mHa6Y2gWr zTZhR(R`D~&imWPG!AHgX(T`S)&@m<*VHc^3M45hc$VYqG-#^>hPLJ z2Ksgl&+K9_0;pAo4TjrRT^&e|sv4%*y&-Zh%16dsG6wha0@9%k(i-W3o~lK6Sq)Iv ztqru~3$Bu;kz6B9Lq{kjEk0l4{k(`5`J}M;hOgs>rld2m!$|uDvEWsh#C%S-%pD)K zR`ChOeZoFkty)_^Lv3T^o90D7kFG~Di-RCMhu~6#&5k~u#>#@({6s=mlYtp!bdJj! zYXoLNT!x}2whPR|ScJotki1BVAGFvYat|G;%8WJ+zXuCa$NS>?&)hng&Jts7`HSoS z<+oxZ=$Z7asVQ1IHp~&I4&;82OMev_FQhE9QiYN-tyE($x#!9Zw6K3_2Dp#45t&KY zh+-$GHw+4m6uDEtj6h-cI1l5}7BDI5h#^7MIaik4aXGc&*?s$k%L^qX8jL41wS1QSLY`ge= zP(KcH$%G9GdJI2vS8%vN2)bgVw?m+9&It?cT!+VzOc#9<24ccwlIy9snkX}}3xVCp z9?SAjBgAU^0^9T}RANIlQ{uYYO&@{Kdlgip!*Z=hM&k{CG4t++hB0j88D;O*p^g3**@8#WgYi%Nl7 zilp4#u8M1_ifimE$_c}B)v z$yf{mN~<|-Fxe<+PKK;#HZTaQ#V1m`kLL)rOL@v2_|0r*qrC@4ry_(X-h+dzz^d+& z)4>1p#Ix(HP(?o(mz}8MuO+xD0t4be7#UN@ zfG45TEka$giz;M-`5^S`@mZU=?8VM%Kggz;#DqMRz_m*irkyptK)%dO;BzC_bBQoI za2cNPOO*Gx4>`-_$4d-uNb^hUBf>9MruCJIDfqY}W7&x#GZRO)9?4g?vy{eZg^W{f zHIZP%n#CB&4_&pgV;{90>X00q+0inc4^)|4f%X{YCy*1iR_zFlx+&l}lWKC!kOy;I z+QIv=WAwm|X2I+jPEHOpK$&HRWy~^><|oxAw$R{nez(CM z1+V2&^2j4O_y(4^0(BpK6wES)UPFt{y@8d8dDtWi_3yKo0ACWamK_new>tzg9MQSB zrDk69%ai>y20F_alg*F{Y5ELm^)O!da!eDc!{!!WHVY~TR?4gHkGtM zoeg(Jfq+aUp{*xCKjd>uP>WdHocVYlj<- z9!4<{lOd_DC;4jy@M#O!)EzX71F}|%0VT6?+wwvdT*<;OWtyg>Q`?GOC`Au7@&YW~ zIX=xTM1}8)suvvw5u}z>RK3mP5rIPIAmtU$E($WrXnqxRz0Edr{MdF-TJ>NzBGEHT z^o{a@6wJwslOi6RJv=0^?F3P1Lrok|75D}T9yi82&sAVsa5yIqXr-psQ>z-ALW+Kqm>1uLSczSuQs6woDY`L zzme51bX@44l`B`Y#v^0d-MN|F!iC$mU=cwcj7&003sY7MhTH*KsllkoWCHu29&kA6 zAw5nehLw#rResf#bjLVMFiuON)gd^FO?TzooAzo92P186z+9R8SQZ+b{@ZN)v<%>xSL8@BV)pKS$CAi$VpHp7Sv~wt#(!v zZa9Q>ntNaR%y3(5#^B;aVYk7T$Kc~b2augk*o-hTitAo<+cz(Ho^dJEhEN*DuNGP; zrjpFGnO_kU`9bE;$+@=$f#!J_$6RgyInciU2um!n(ro+|w9*djqAb~CL)&)}ixSD0 zU7*sY1wQwFbWH*GowT8t1_r6tH-FNKX|^tRlVQsX#+ZvLIjM*up9I~m9y}}vr!x1Z z%RyOnb`r#o5u6k~QiWA@ynVKG#;J6E2-`89`ut|ZzL zccN^1dqs&B{eaK_5`ckLJSNdB?cx@AFXrTO(eFW4g12|bVd|ick02zK=+TrBkDp|A zslp0D33XN>)irfUGpT5)*g5n+_FG9{MO~h&Ki)yOQ=e!D4jw(KkbY3PDpj0+pjX<= z^Fy^KneRlQb;t2G3Q#8$pa7>%Q%SzLra04zo`{_U4AadORdjasTbLPD^P?}U_-|yq zf^|~}5CK~z3}Zip=2f?vhigxNuO1y(bP&`49++v7)y!lzkx(-2YGwqv40Ed@e1@6Y z1}}ZldmY_slG~JwBSa-A{m~6($4i<`+1ZWc?9d|66}>GrV@iYq*-E&+Iu()-K9sHJ5rEpiskt0{*v3<5U4zE^XIV_3~CVHC5hSdDe z;bxnfP;|XjZZxd1w(!Qn`wJHrE-$0FMqnZInt1`t!`v?|YSr)vXnK;quL= z7v3d3H)pZaZ$6Grfkbr0W`dTPY%Z(mRaZd^{eSGeZFk$&nJ)ZSb~S5FoQ?$%0N;C_ za~368ewD3}6ni?UKWvg)uD#7^SWNt|@8 z(V4Fq>XYp`pPndo(#EsqFFg0XL5UDSsQ^SvqOGi0B0!J=aAWU%zg+iqZ9*2B_V_p+ zaad~EMfqa`*3X`xQuTHQMh}^SNh9w@GCnHx3A4=0F#XPPNw|^DCeo^&*0P%+{y;M37>aIiM{y}T)^e#&F0nR3#9o;^D0?hY<>)v^E7ha{b>2Q<{z7%Hs2+B9=71s zjpirI&k(raPhwEVWipB?`Ev%q8yaa>J z5oCrsRfx*K7~pr!Fw64lMG(K(ZmHU8+qTylw~5UdhK%kEdOG{?J0JfWZwzDDh~LQ_ zqo*Bk+w+LP*O8;tDpR}4X$*N=>N0;fNcR%bCKV(wc3Q~Fm~N7tcQi}zv1%|Cg ziB;W7RrXA*0m^*>n|jC=>@*@8lG(k{($)PJ2A|t`=RcM0f?JN_397kV>hlXP!mlh= z^qTL3v-B}Ih2ScE3@<+b&*@`uvMw$^yPv@COUsw=r-JX_Y+eT!>=UHr{geFE;_~<8 z;pkdE+lM~o&yiRd9K}VrtdE*E$YE19&U$W8(9i?pC)#|St;-n~c1fHiCyLneG_Yp$ zBr5btIdw(VeI=Z>QxUQ+D}(TdbaLlt5ejylSru*=!sh+Pim-a_|Mo{b`^Y0VNQmH5 zYYbV<%^({|R*#(%ez@ncd3_>N;_@4X2;I(^+`xy@A**WeWFhowtE^&%;tR+8?M6s78#exuL+CQxGmdtCX603c5ctQRhVO_4r!Bu zZL=D-9qpSwNxTYW^!g-r4~q$52rcEu8YH{NzsXGk;0$t$=jw4O`f1F6+w-CUfo7pnM(9 z7+OL}W;&}a+vnPiR3YjtTeR9RXrFm@{!{OPUl*O|Dud)B89jEa{Lr`zJLV~9DMi~7 zEc^hq3`FVA^x;f430s_5CO70_f0y~5P`Y)LuBVDYfZV^Sv-R19Rk+1!D=!~R}|;SSt?23gk`}XeH>K<5T8cAGvHL+TJbvr zWX2M;wWE0rZ`5wLYR3pJ7OBk}4Q?W9ewoK0nEl7bxwC%1yt_5^O|6+qY(@9gQ)=yL1y&4}RcT2Uqwo)Of-7?7ikd7~@=*fccvaqFK;;Kku znQhoBq~;QehNAQ-C6$<8x7FGCv;NRBw@V6k`cWjjS^CI;E21byA}Me1PaY`4Srxa( z7wy$UFvV+l_M=Nz==sRcz%j(zlQAwVw^Z_Jn#U2K%^)U2-;1c*?vR7rp}b5{{B#!K zemh@1Oy!hpK*Q&e4&Yq)B8GJ8_6|nYS>!FqZB?xd{1-Augom+R)6F4&M-B;7xR$-D zRMuz99Gra_rYW1O;$4VqmYFrEX2{H{*0aNCbYEqR&w!nH$?>NWV-cKYu`Hhsc%V^u zCl1ceWO8V(uiv?G{})ev{c1Ou((W>oM<((0)oXuw?`GRv{=ND83vbCOAUdY=5{ z{R?0J?k#fvr5EqNdu0gk-`AI(BKKc;Sm^=4$--T&mxzg)fW^{a2REtg5R7cKcZ zoaK8yNxkW#VE&6JpEcxh8d}{?TyUxoT{Pq}mzM1DpKBW_*Dko_LKfV>NEwKrVkeazhJ?~dYF03pWX*9)*p3xvUh4^Y zR~8%a)5*>m?aW3<&4Fn#+gLBNvOZ#Z^2wEjk5?A|u(I&(%HqwHg*R3fe{%Zt%EcQi z7vEl4{Qb(}g_VWZRu<^R#m(N0s%@I&W-4$O;38sl$xph)Vyo;|cRfLnHExUUuAI85E0@eD$v{;y3LbrG z>4U1tFnLL)mp&*};hezrwq&W*k=kPv9ByzZif}>|W|CR>GyPSJnq2yzR^>)Hvaxh# z`pQpQdlk|`uREgoE&xio)uRB0+8g9-7=H=-w<=rE`I- zZs@0hohun`S$17#WlDw!Rl*+o)lrm`ARXlnV< zvEf|4FoHiceJnpRmQ3YGGMcufY*Y%C&}{5T>>+U?nPH&D02U{Or#GUdBOY39~23KOYP=w7D`$ zF1M9($0Yn2YWHEsⓈe1`58Jb}X70SZp+im{zYp#x`0<5$ol%QL_vep+^K3BN&-0T>wiAHfMo2p)m24t*eg2_ZhAMEsVfyaP zBBZzdYJ3^e7#T$ml3+y;QY?U5<9H&2E#N$aM13&DZK#VrhCDnd@m>?A!)!G(bc~(l zrac=SksJ7`wy`3!@qnpvN-iLz7@p6O&Pmx=a;z|U@PMZTHR#s8U>7m&Rx9reC%nqn z*l#fsN}`}*O!Tbz2sJp|0+Z8Gi_w}rCi7Ekp!W^bFZ9jiFLHTJS5m=pZ|?vSb=NF8 z{br*cLL{OK6Xmdeo6N>;Lw$JbgEhfwF}g!~*&__b$tmghGo06Xz{=PG;Mkj~lcckX zsw-+vS2Q&@ziz9u<6!-vElH6)8Kc~AEtyY}DZ!D$(+B3^)EImoH3$qYHMzZcwACSY z=-~XJgV?UH555pw)oscL_G;jsWL8la4vU@yJi=?Ls;^@+*1J>u!2Zpi4=f(1Vo#{x z7b6I!;$kirV`o}+4F{ElnE__|{OKRTKDkBq)tRUp?D4+qAeyA5Dx818uEyygnm9V` z8^O_;NpeUKpU8B!YCg*C2o+vI-0h4ZtesQ9$YEo%7GK2P6myi+mwr>>Ce4!rwnIsa z$%j)fpUz2%StWtJvNrjyOSw&OW}IYj)5#MiFAG@)$tKjny2WhxTJ|_$wgXc(S>2%! zk|&QnhJ3WE-$v1G#>|hJNo4kuId+D^x8_Gy|9F1HI}Sp3VLc%BK{@I~{wuT@YF+bfvE*6d99*6e`^ zHp|U{4lj4awJdaTyK4hHF%>`?z!l>%KjmO+ZzR+~+-WzRNo`81JaXt*7ofr6_RiLs zA2tj)EeR%R*&&@)a!EzQZWA8`p4K*o7Ty+VMGp{3y0a0K8h(v*x*iK}^ONQ$8|gLd z<^{0%(((o55XP%meip9br_GO0ouPS^G~a=@K9yv{SG^sm)_}~z@59kIn>Us(z|rr< zl5IH7OS9arGdk<5dS}`RnM~+}Oe&!y^*tG!K!9S@VYvX0`y83!20u4A!)iQ6EufcZ z#vB%<)j+@!X~?|@kKKZl47muu-)(*>nXzA#4<$UM1fIxGmY;&h_JR~3d3yP$%l}N8 zf5Io3Na|=whTU(2^m!O$@541*$DvA^FT+*52ODq`@Mm}<9V3)?DQxim05(K_?Xc{lY9%{`i zLsKQ;psP!!Eo$0`(~oSu_Ls-U8>A9QR;~pIV54NR512FPhtL{oBM2M(P>;p{1H!{* znbpDV!nYX0@ffb;;EoY^bA;CF$Y2Ea9r5U+DEMWg%m&95-kuhg$pk6XndKlEsAkrtJs98OJ zxfUM>1Utme6||^e2FdCl+wBuXI-+-%FZ2l`rYLRVLa2>I3eH+xW-3L>4 zGh-pbA<&nxeWcANDbh4lg#{D8JP4LKOY|9_sQX1kPnB+6paTt>RFd&qs2?dua~yg1 zsMR2Jj?)(D$&UD$YDUQ+HN7Iyq9vD&)!T|8wX~|_)SrApl+_TN< z4rjHZB{7Kj2m2u{=Apn4mNU?S48wi-V~-Uq>ET76`Kk%sKhVDOg?f&8~|LYeo8D%=K>A6W_%G_kQ`W_pZDT_x6Vf0NwfQdGQ-_#*X};|>b;-5^~HzJL{GKK)Euk>@XHU*Qq_zPVIns|8YQWu z)vYwhOf^hV2bZTG-0Lo}>oMVSCdy^bhHPqWAR=znzB(q)xr?1)mn!Co}eYhx5kG{H7YZfN2ik9>(Ano-d ztaY%jeNSYf%?M%fBD1(jVL`D=e1~-k(LPBfkqs>GVKWE?%UQZoat%RB7NceOC1jD% zV-6T8%wtgJ$TC@(5xRPGFZ(iMWP~C!txU6$$>Gr>qr8G-Ojz@S!1ZriLe|V7(;oCUsbyP}8=+afIwR?mHiS5-w|) znP+%)s3t5%YT(kj(LKTT*Z549^UZ>}kkg=ws^iH33&f+mLV+j9zU$IBRHD6FoYcIk zE6Ml@yAPpN&d`!5ax*54at{rY?t-{|bc3hW%;?h3V1{;Vx3<$1uV|v2lF4??Xy>aN z^@NOoX^=*3lq1h($rkPD5@-)$F~?Ql|1fHG?CY#yz5=L z^+V_=v#T*eSHxjkwj)ed=cQSHej$zxaxesA^AVQBb`6ZJxz8QwpLSpTRGpq>WjNII z8ZM%l;-CR;gf&j*n2%;d%_e+=LUJRVDRfoI1O;UxKz=3+?9qiFRib-_)O6p#^E=!D z+ZH8G&%;e;WP}m~G zZh+%xqjEATB!gr9xal~!#c+d0H>m9p_;Zp;5dP9QpJilH*kZ=%FT;-tw_;$%4(6iSIJ_w1 zNGAce*e+pz8^o!;$DsQ|u=yj9Q8cX+d}q0#1@M|bYW_O^`P0vPf)1qlORvFw`8kAF zN+hKOS3cdmnkWDEhvo~iAO4!`A;wPkPnz$MeXzAa!vH6M+Wims#Q-!B$ZGt&S9S9dKS`lWR z$at8P!0EN*qK{e(1_$pMxgT3pSSOC1m^^*@TLkX0*pI~nSUMo$vf&Qdgj#G}U15ep zL`G~zZD9@$@)^<)E-8U8F~h;LqpJ*BPL;w@v|)Q{xMw+-kJLRr-MS^_Ae+x5;yM_kJ-xZ%St1gB2)Bx|8YU8}Jr*8*E_)z~%JhcNAJ&~rOiBUh1# zAec&>30s)G{rc@UZeNo>--VyocKv9${ra7!{_~wXm)oXT-IKrl4!nH&_WO4(!B6@6 z|2|9p^UXWoBe4DYolE5QZ|^+)`GsfieDC(_1P-`H?p%gzzH<8=+ylqL#(QuOzIp>b z^`CFz9U-@`-MM)C_jkS*J(DLaqfAEJI;ISK%spz_&x)K5$%>4o&MMjXt$<*b5-VZj zNSPY_rhIms!%8$u$vCVF(Tf)+>^c(L`%U1(+^0ig?llR5NXHJV9;r0=Isz_ITkXtj ziBs)NR$N0pg>FcD;iEC-Uj6s+rx!#4HvHERaHqR_K?SI zV1tI(lT;>l$#0(E)e=Jf zAM(WX2t;|@kVROmU>r8fE}|c(L8cnE!)m+Vg9izO)(g%?D+q;A1o1HYln=eT4Ekkno?OCLE!VGwFn zF!>n+-FgBm3G#ERv<{I)iP3aPFkPp6w~ZiWOPn>qzddX2$LE<-llB`-dF=|wzSb2y z9=(XwiByi_c+oQ$G>TXd|x)J0iY)Im4UdBNQ6D#a4^D+ix`!OXEy*vrgToD3> zjOt5G$M7+A@G%2~9-#>%BnvT;o65SuyGVnI+!6gS&34ucmM<(XHg7hsHLovUh+fzB*I);{GPJeK18*NojjnS)aRKH$!EHZ+-owswDNTpq8d&NsF+V8L&=Y_=#Oo7spXxV4PZXRyaeF`D%ie_zW)*r%|s8 z><*Wg8QEC=FK$~Zm1uht@^wgC=}I;}6ccPU%}fX}(MSdSTVx9g$MdEcP z40=9#&QD7xGWoILTq3UyXLI?n14#{RFfx))W`^^bLO}<&E#(vGOtwvx3~#Blq6X@V z;RSqzx%-jIo5IJ&7`0GEbJ}v9>NJCIjgFn|{vGihwq#Dxv$2ViCnZrbVlcrQEm03o zu$Bp!kb;oll|roKDGdtKXprND@dKnT>aL`1Cx6?5Aee>0P(G$ox3t=bYOb-`yU zU^c`IVoNB*T@|+bah>b!FfMSa6f-bex=svs?BjzH7-jo$JPBXgHxWATo62+{^?i7W z%BR=UiQ=qyj3O=m@rEUa3RbDg;d<2Q7U>0Fxt>p`N-7&DSUq@AcG7?dzE;HCA`&!M ztalmU?U>d^1o(4PZj99|sLej8cBeE{H(@jc!*yey+P?JQ2)C>%H)zpvZE{9fWoq(1 z%MW(DgstZXV=x)$ST;@v1X#)}G(r~DHb>?UjhsH6CkL#`(re~<239-Tl2coCOvt{Y zoEg1TdEAyn;6Y*4$U%$OqU*IDIfhr5_&%pc%fG!C3kknK9^`u!Yr2 zCEXb53oubiNu{Q`0?~nm zz@wULantQ3`QNr+3_H)i2X10?$doSi;g@Age`sf$ zx5W_h=kMM;%)H$xO5H+vveN* z!MC1wt)-i#*$T7q;QkuZXlZz4>|*pjv%_bZUG$CNOn2^v%qHXwIGi6)7QB*_Z6v{(VEv~(BndTlsV&cCA*$L9BayVV+BWk&sRFeAQKFl}nV zAEDLLr}xA5IENS1>a4IO5e2cli zF*ocqfa)lLh=oRgv)=yW?Q3^_aAyGv!Q3`R8GkTX6Vo(QPT;YkPQ>&jH3{%w&5H- zIW;oM>aNp3N-JT_cWmPG8HCW5NL>q#ugr!t+0vG|{`sXJ zc+AfKee*K>x#WDft zYZ(K-x5P0m>z~uQ&uE+7e(c;1G_ytO+;+izEE!9`u~A{n>v*2vHDq+`3m;*iN z7k%1e2A$(@nkL*|zr)ZTEt$}x1suw3ZY#^r6l0P>a3d=2Lck*!p(ZWw4kGR`OVm7a z=4cj3W|UM)QT4Q>xq!aNW{*O{=d}@}j5>-r@;;h8=bLm&^05&C=*At2w zke}T11Q2Wj&L^xUaM(`!0;tcPiaokg#A0<}IX0$!;A+883OOzuUSjg$Lq_!?EzKID zQt4$*=)qLLH%$o@)+K>BBKPLJJP@H3xO8NghJlA_U(sEQ%`-z$`{lu^X(buHY&$UF zmZ}(p4pYY^qb?=G;MUolq3(fmsE6taz0YpY2RtzT?Z9d_P7T_KN`CyH*0Tb)NF6@y zuJUO3&aT5|18Tc^*SeG`B&4edoRQaJU7X}#l}qGMgWO+}>0XlA)NEO-Z0}8&tJpm1 zR5^vETXL57u#Y~5m~s-u3K9W7aL+hMhZRfWu*cR(DSAL2If|EFIEzh?(szQnP*{?E zr&ML7S+tkMX1+yiRMsK1R02XGwcszn?Qo=| z(XKbgnkUy$Iao;Vcnt|zM8;PpB;W%7RK9c)=p@i^rts|}}G<0W#iQ8OtLG{Oz2(pnwC2vuH> z?lVZ}XHrS06+eAuR7hM3>$9nBY(BAvkYKxRwR$Rcj$#|Ef;IctVB>kPwejzMFCc#M83WQnB26eWS|6;pxC?a~Qm%#sgSqBA4){8iufAM4c%m zmrydO2kK)0i`5(fhR7;fU=;~HU%B}w*9YcdQx`)sHb#zzVG zY()XijupLdG-5Dn4I3_-nr&o#Y0${wP6`XHx!hLPlgU+aR!j!$@qi!T=_8;k>1=Gb{gEmj7FwA+N-ROk1O-)XtnKuqde}BzXpck7D5(IuY-rOx zfgIB9X0UWNkyh1gBDYx&nwrfhY9^(qC?uMmU%%Pe!qRc@Y8#rICGeSXO;7NzUpY9Z zILvI=!{8te)mpkjy$%K#(&r>|w*yv`n6o(C>oRC1N6468=5J8jpwzYL5ZA#o4M@Rk z2Tk4Kgaq zG4MTImS|1Sb|1cac2k?Dm29BbK8lxi(4uqjJ{hBiVUv9xk2BHwBJ$s~m7!ocHELF) zr(_P|Lt7RExk=9ZhGo*X3@1Y`#L0w`);kNIQR2;EUNpF0<-dRK5r>57(!jkkIzts)GQ z&+e>IQc614**m#?jCPY%T8Dt0m)u&PaoPFKJbFMLV6lr%F_1##?@Veda%dBVaQeY* z?*vvca+2GyUV0gl8E({~=R}a_#I+L%KZZ2b)k#QgY=bm5gsa*fOtB5Rw}h}ATBW>F2p8hT-zar5R#=x5V&r;v7glNmdfhu8Q0|S63kcuPeD?Wk&@JiY3;KynGcVu z=@<8WHNv3ECuCLXHACX5U4fj+V z(ep_)k;spwRQNG^U?hJaS%80;1MuE~13sHYh0>)Z#vmXrG82Il?-*yMP0q4nKSZrh zINabib(rf%g|v6($AnvEMG8)gX%Bf-8>hPT2D0Z2jg^NZ8_@rK>&BLUXoa>_t_yL~ z9-3>rl7z+_HNiO)r5N?dt_6_p(lW4rvxh*}6*aXNItP40gO-qmI8K$Lw_c&1QPkvK z1RXFCH`yF_XptMeRBexAp(G;tJCn3_*fp>DYV)J!$IYwDFUfz`$@0bJ#pb)s8_SoQ zpTN&c1U5feUR+*Sz5qvDhGTFi9C5w*$L7^<5%~D0@akss`tkz2^*z|WPMSBoUGNpx z`)XQj{t{05CR`|9$5n!N#(VM6@&&jv3-HbbxXPz-*GuyHZZzLpzSz75mw%!8UgSri z;5wqX^-6LYiTyKyCW1X0#O7@Rk@vu`81pB{sy>Hwx zZ_84li77(x5YNQmp4b54s&}vG=$n>wl367a(D;ht`Hs@F8+4TJE+cfot0U}&iE^2< zA&4T%MR6{&DcWDrHbi9hf^2Twm3{>BUOvr#Hli3*BP9OwKWqbm2@Dwh?Kmk zHsERuQEA9}LJdAsLZ-}b4v}eEK{9d1D<(k_As21~ao9Nr_PTJaoQKs zB`Kta>6~02U8}T;2^=OiYIWl#+s1lA2DvYCLBiU-Cb}+7bd#9sCL?YfqxFP!GlS@o zT>Lt@nC(X|af%vqylnB>HzMElbst@I)4TK3Z^BZa$*bMSP9wFkz#9Zy2Xqz+vJKo~ zwre>3CS2$dz)-sigC5pwa$FMHz}FD3Gbgfz#>*z1^HdO5XEEJJVWEd?WC0VYN;=39 z7lC;DfbCe!6t+kIY0{FsFt!mF=I~@7$uVrv^W4}K8IDO5`fqbl?JW;9kvVG78K?a?$j$up`8f`uN(_3yE+wAi2ITn3 zO<+G6i7Sb(tXu+b1GPr?6KYtDmK&&K#o-Zf5T<34Rh;*Jg47yP#em|$5P^a5NX2jD zhb7cLfV!xo%%Eqf#mKZPZ))Ta4EddXRa~vjZ;LO(&A?hhN$%k*JKcPPvL2(7<9PX6 z5+?F3)z<|Q^=_~h9t^I92Q9{AE@0=^!lS|vwhasWq$mrQ3@al_WNT!Q06MLxDHI>` zPlHhvLNW+TN$Pt!ioh^Is7fsk4O&kJimrimVvD<3Efk7(SReWIM?7$J!7#|MK})kS z1l+jkQd7Xi%RUUFZ3kZsGaK5S*HcP1fDaB${2^VtCytMvn7{X<*Z%zCh0~|MeDT_b z>g1Y|Ls7x36ess>oId@3RP~ zpfc(|a5K63W5VJn#Kz7(FtPgWnJTa$Zh|9WH+5uj<{)_8HgZ~GB{D_Jwf5K z=a3`YY4llRw(rzwbUM9YYj*UqX9h7O*#OdgV@050SdFRS@prB#5aC8mNG_1ZMy#>N z1CbuGobL|#pQA1eB_gj5uA$yd_2+0|V7)a4j#A5n_>J)9+lP*ACgGNwN-bJ4_dDiBB3)2Oe<`rhy zeWz4&^u)2Fh3NvBK78=Nq{^qQXWjhDwIMh|1?zXki_TEb@&c zQ=M~SnMJ63FMZ%R9D{%IgvtGzeSlaUW~S?$_pf2Ad-8{kGZaI5Ham1kl&W^A>KgFF z;xi75ZoBrk`5eKd&yP=roB*>DAd=V$kad!k(9y84#7Yee>q{1QI85|49AYPG+(C>h zvt|qztU=BSR92f27D0R?sap)$=4eZS5&V1jbni$L9?-Hjy>2k#R(@C^1p}s)Q`EqP zJ+e(l8|H!Uj&mQdVNg>yph{3TDHvpt8$%12v>ukCb4*yY=GD3<51gF0WDbm%DOS0u zjft?O*O#s{8>PmB*|64|Ljc@BBfn|}F??x7l`59$`HpS)E~F78lq`K@z;#eB^|kb> z*!*ip3|8dgj5WXd;`HhL_z5+aPbBseS`jv^y1KfbNZ`r7C#FZXsq)^d1$WF6=V7*! znl;f0H;%yIV*h^G7llNXL(1k=P0=!2n-TU}Yq!u^3vIqqFKb?sF~R5 z(T9Nuv86OejUq3>m&DOD?TOl47L{cZ0iIVgEQ1H1C8l6CQe_S~zq7Ovk#&8J@lw@+ zCEzTx>M@i<4p@A4j+(w)XXFSm3yZCiP4D5IL#1T6jbc(IVsj@(VZJAM88BE%QFF27 z{2tcQL7lpBhEOl724I@3k`9ceSrOi~_%?%cvB5z1ufFm1GjEXlFTHsG-7EbN8{U8U zssCDd?dz-8{_@_u2Pd+)aTEr<>@D2xKVv4gYYJO)9 z+`U61^M`~3NB3RQtYA6x9H-#)Hc;ub5q;ruID`8;D!?i}U%nj;>LE#uvJw}Q=>$`U z4h9R@vM5Yyma+TUv@vN*;BBOVw^$i5vlyB|*!GP95qq}Xqd46-vcn8r6+Lq1 zm1W3Om0B?%jNR!?ySuEge^Xeqt&=TEL{okT^!Zhs8z!eMlE%m1gHD*bHyfGo1B0P9~Aw;#t zJkU{fTE9P!o2^NotX=3#>q4WiYiAj^ZPmwkpGviJSMqf@8>TRU*;#~oB!xW8ieTV& zkeX}2%El~u1m|0jfcBmbFI2{pHbGj-l)5rd#`H&7Ffev8E zq>|7=>k$u3fi;c@d+E)Aj&XH(1Ki|DutRPHWiiMkEV2W;0X%t(Lu?^t2t@3bFh#9F z&IlK~uDdKS15-x^)>{RQI>)i`3Cp*w+k-|~Z75Icikg7{Sw;GCU*@Xq5u`&p1YoSR zl8aAH|1dZMSj-Z<4DlSD!351t6|-`Z86)7FZ2~<;EwUM5mBE3wB%f_ZsTm$Z3PCA8 ziNlsCxl3=tV`ywd!fqeV&%@rqljMqh0v z;meBREGrLAVYUQ*)w#v}8Qu157M8dqhm;5@wQ~e*U+YD|5PjBh0cRZ9biL@a08{0Z zY{~{CorCzQE7numWi~Z;PRXP@2VfL)-&^?opRc^Qa`El2-ne=1=l{HN@eeB(Ut78O)XK%* ztz5X!594uV@t;=~Z-QC;mTK+FuPpptzIrpVg^p6kvE9LQgEi)BqtSA?b{B-V5pJeT z=}Ia-Jmw*)BZA-IAUTLIE4(DEIO)S))eyldX(hWQ5F%1N%Ko#$_$UFVswl|FD!0LI z9yIICG9rZWnEBY>yPU=7p*9$`h7FfZ%{H<=4t>o>q!T6|rR5Wed_q^$WS|rhfk^kD z<<aNf zMAnDj&1!y%wD6WnE9n3$^@iQK18_~eR9%;HKgjG-6`L~bnWfjTfz>rDL?OehVlCY? z%eZB8R%K+AGkeyxvDv>{9)9#dAP;|#SrBG6c0~ey%;F5JH786mRjMw1U=}Uc?9!$? zF#PS#Jp3@7YmxM~td~<2J*jAkgrX&VCLHule(U?PHnX6bfXT1ZReZFfVC(<9N#>Sb zi4i~bAPn}5kW$n@IWhDKRIp$K%^_k|BV`X245z!#e%#}WXgBsBuqv9BgcNilQ>^Z=S`i*{b*%|qJ4mcl zQr#}z#@J>#<-5{Oz-Rs>Jhuq`!VbLyWsZG*@I*>YszQmDS&f&-5f1=DsO9hyDj^gb zZrN*Xjf#cr8fjAE&aAZmM)T|1-BPIknZD%2$LJ}Y>`wF0S3WGGdEJHxi&Fb zof`2fWL;Audb>eW57VFN86^jicqhD?hDG__e&f#N+rNhy?9Qdz@7@0W?YD2gbLUe2 zfKjgw-G2SfQ?Tbc{C@iXK6~dP!5whehqqscgPywm=Dh!iJ3r`h1T2H;$NOx~lXv8! zJ5NV8=|9e{8)3hlp&jKxj7>UxX7+Uj?fxsEMI+2dt0 zuszp@ubsfoTZl<4`MJMa_qmfC`}|~FfeNJ%ld*N$qYnx?zCD?oBZ7X{UAkhzeX~k% zxnzQusxcY>3Z}F4o=L_aW-V3yMWA%|8c=$p5|Dg6QVAue?Nt_F^@QiM!h}_U=Pr8r zwI@Xd&mbu!rN?$>9tLChT+7=3_wjSROR=SWR1+aR85KoFwcz~O%V3;5iOK6(q9Jlz z+x4yLv19WY?8*VfW9tln3QF4FtsDdTx&qU zN2e`Xn#H_=;+3dIhQ}u6Cs|!s4jIFH4#B3)9X`kY7kqXaPJY(jzk2BLgYzSXi>Y98 z{NU-+3d|RJGNq~M$H_j}*=JO}BNp6T(mFMQ2!S|{O#6oTR_=77?NNCf;T{_;p=9FI z$gGv7^XoDIMW!=PqEPW$pq;97!$lmdydG2o+C6}(Y4qDBn~h48*$SxuOJZxOl{;8H_8mASr~w^ z8paMus-QBpEI(9K_3@~MX%zCzkW|#%o|0S#-zY2@CZAzeoNg6rKH*>z;Tt~7M5G(_ zER$A}xqvSemirJ+AWvuo&eAdkvi9Q4K*}5`(ZDZ){kz>Vo6dC;w1R@?WlA7uk+Yt~ zhR_|L5@|$i=HL_uer_(Z3CF2xiQz4k?N!u0&Sg7MIT6eDwB$&~{5*u$;QTmXzsJ_x zcK8aE5~^DQyd)~MX?C?aCtzW6kQ$(E&b3|nWAgGwT+5-ynd?|F_2mkdN!@jpOXb=H zkX%^Xp%_}GGk>=?@}iEs%+8EF9UP1Jz6aaaN%Mxc3%=rdU**i3zl4*%2^Wgjag{Xx z1Q+){?D!O}9&g9fxbxLPEfqaDE4a@3;4Fr%QC&I^6NC=U#*tRH!idEb$(L?OS*=mm zY2?YUK})mbMBOHzU-&WkM^P(sS-bZn!VevsKXmZHNrbN&@4+hCsafzIQYl4CYKoeO z;eCoTf*hh|xoyI%mi9HwgifX=lS(!hJHj6NCRtC?LcTdv67>dZ$?qrBuox{j2sb^| zNMuzu$tr4eKfyLNoX}v;eh;s6YM4~~NbrY674)FRm`TR*)L=iEcIBOp9MRX{!#?ug zSZUW}R!Js0=jc5IU2N0B`M-7h-P_mT?BOCV!)A0!mrgL9uhtwoFN%Ph_+UC1$Wu;jaRA;DKdyKORVlm`V?Jl>5p^jhfb8mt3JJ$5}BE!Tg2rK{AMmTP!590LV4VcM@_?!qfL zw$41VSq=)zvbX{6v9Isf2hj;<7X(cy=^j)XNUB>CT-&QJlxs*==~yUFYd?mgikDbX zP^&z2*evlfd|xX&pUq|aq%UsMI3~;$M~AoeylYOwZsJ5THi%&3Ct$`BwaCwa{iv9} z{jm4^8PK!2luu&F+Y0$Cc-U8yWhAybS_wLT!GN_q?dJ9rtbaHFy9H z@lndboq}6$pau>4SSTHAaOr^A!nI1Q-@GiDotmyvqhS*m`VawOi(m}(SA;FO_Bs~K zLDkntv<@Gpr#F3ubS0(5?q2|}t0e_V`_K8BhFdhV^q%SQtVfgyn6YySU)G`R?`A_f zyc*K`X0<(RUL0(mA(JhLJt8&N=A{RU(9?76HgowpoO!i`k__-zM&lYNS&VKh#4%8K zVxYs60|o4gg3I3>$^qlTVh)`CFpVC*sG5r6_DW7uw5;D4K3jw~g{Gv_oeSloh(2$~ zLJ(XQb#@MVu3^DH9e|#1U!VRV2HtlCwFdT+?0TP}i!SyNVONhIICWIb#fZ&1J{*&{+OtAdQ3Dcod$^guj+;T* z4PjAhlFdYq-h#lH!R`cKoA7&S*z~QXI!BP@hW8Gd&o}?nyx#n{`SJ2|Sm$lN+q}B` zbo0jYbN%OuH-Fx|3FqDX1TNu%SJQ<{f{R;z7WREAE7lQ>1QxGR+tZ2~{8{l`R{v&#n!A_QHQ$fQ?0X zMcLbD@Le6y@?2263z3Kuw&9m2%0;n5-sYh7D1g9csTjFLz*;#1RCG%)8uo>9U5d z-Jc=2D?@lMqCWw?TikvIw02@s3!T+WL4csft{Ohlq4=sAtOGZ?l}db&!MLjWIw-u2 zkS5Nd#9W7o>tP{vFqt0_6E!$EbwVT+#U#si4Y zy|0K^eUz*r4#Q#uwTZcQWu_oj>&nWM6sDsKHLF>;;}CggbR_fILIFw#^C^h{X~W*| z3^P2Dv$*zvJCl={?$ju-e^aZblx()s1rAM*zfp-o&FYE@U#(`6N-8_QezUU>(Q)t> zH(&bd&1<->BGvxF<&}kJ$l=j^DxFo+IXwe=-~RIZu=ht#K6xlVd2;^lb3a{K_`$so zF0WktS>F-neD#BCU%q(t?xpWXeh32k2)?A8-~M`TQ}ct*mE8(bXH1+mz>tWTTYXuc z6CMxi+OlLLoacWboX2EmiOot*ccSXuWvS_PfJxlSkHJB3HKHEfPomcBj3p{s(pRw1 zbRSGIB)Uq^s}QxSoz2SjTu7UpbK61v;=141AcqZF{f`iErxzj(6PI1$L!E>O_ zZ8(3D$&)EjV`PGfy1~d1*LKLLFduWsQB$17D>pIh_XjOM7+7!KT`6 zv1AH0(>FHcDY!w%mbcGl)P}V)`FPNf0|yHO>kT<@(4(*#Xp_EY6-}zIs*0bte=yly z9w@nyK9Fb6jFJveIz&Ptdp7Jt0oHMj!%8B8rbSIy7R=}#usb5^4G}eEi&lhR7zToS zpY(SsB=$|m8eT4b&E@3UMBZ{{`Ox!yg=^1dh`A-e~zB^{l&}p-q+8(vFjJ1`wMT~f9cKpzk2?^o_^)) z-@MYcPsbbisWYyb%V^PUb^oU?e0}Np|9bY>ub;W%&rO% zCvf&n`#9TIvr3>8gG%JDf!1@0wngRZkS5iYOnPfDygkXEEf6_jQ`|5!46<(owZ%Dv zG&4dtMQRw)egCc^Kw5X)b5`5*u}CTDK=L_?^K<{T7yfenN;7wY%*&1YIR8;Kt)YT_b^7`xP;tzGd&Yn-qF?^nsswDMJJBQ{s!Q4ML zYNBG%8Mw+iwa$Z|$1wCL3Cm&*m>UxZEtBK06}>6GaGsurSyudQjaro| zv$)Lju8zd^6qybkYoyXjB2X}o!Y^B?(z9T5Xn-lEt!CE$V|!y*J@vTRJ!2{Ic!FL0v9EQ87=$rrt}W0%lY%PJ{tkDP6d zoXrrnElO1b`$P#fio&K9?!;kdP=7q0b z{mq~M`R(ZA!{B6;V*3uGuJ88ySDf%mj+#{TROcv(g6oso?n~DpysryHlZZViU#G=3 zg7T@FFUh8bOvxm%%$-Z^rDS$Y-_>Dc3cS7O(3M`LUy`ypHn$Za?ol?!3{l4&@C>Y3 zZ+d>ZQ5R@AFqu~;Uwg-acVsda%vwHuI!`i^#EE?V^p8kpP`-Rfc-D>!a!`me+$4Cm z_rddDiJoekt}?{ZpYDj=r~$W zD%pUFUFe2eShc1tTIMCb^d33mvN=g;LrZYZwCpvj8x&Tjiff@OXOQXhaP;g>hX#>- zQSY>QEylbid4Ucw1Ll+2hR@4Lbzg8&n_COpH}}92aljH5+yt!DJxHiTD%%bof#*myIv5KN0KerZDCA$Yz$xZKmT1f`9ruXK%F!P-aUuue)q+S(8^kMA>I`^eh zZ3D~K;l@`A5u349bRSZ8-3ITlRJ9s5wJHsZ@1(yCom;eo-a3OP z6dph1nuO)*6f5quCm5@wz3zIV(|{F>(|Y z$PFE#rc14cmr4xvAypeXQE~*VpbUHV(QU2o)P zujf!DcnozVn~1H?JA!zm5{@qi91T)}0cP3K&%!y~>pk54G`5?}9%?R;Ajgj(S*5Vb zWX@J3t*;r2yL`NC(m79pU<(FIb>rO(%V>MR;5BYivn((fgPJ@Nc(Zhc!V_wj;1T6; zAabKI!fc0@!9+;cZDXbATcogbjp7kqK7Nv!?7T3jDddN#{4#jn!F^f(MH0n?Kv{>a zV@K~Z5Yu{3;J_Ru2WICW8jtP*SZBF;WD8IVcqTGRF3{>1e&t`SFuHXVy1>y4K`N8+ z12@6Ywu2x00)2)~xCDhJD6cBm^t%wf;z9|Pn zrGhidO?y8XIdB4x9X>oYVwbw5iw+KNlrB07XHyonZOIq?b!QxIm4%ckq|~}mk!_t! zE1Iq=;F2q9R`aFPeI1B;*cQYTf;ZFv1@>M*OUaCDmw1Vrr5yw}r6f=z*Et}!kEeI4 zVU`=B#4Kv}89lG>%)#dqv(dY%11!BUfS;+nrYfma?0$-I0SntHMh&OhAcOz2^%CO4 zZR(8ZneS0))E%;KQo!og0*;qh3`E3?24CRja!LPE*L|Ve)F^4B3)w?LfSX-`sjdj$ zf~DOVVZ=z@^$H6CX2~gu+}>IgT#M@9wDYoS?Bq(Mt4boi)J*^{w(BGAPSz=IGMf=_ zDrYbNiIp0eu^qSve&Ng*!8Q1P=>T*yG-n^!;fqgR{^G`!m5Xn$EL>Sxyt%UYj{N!k z4GOVGCIX)I5n89agADB6&wnj1Nm*B!otdhmsb{^UAg$$%Hq#j|1bR3 zyU=e$HXFijwUmOPk&iBZP)((}R|r$7P0feyd+eNQ7sV(wnbZ@TDMpSlWEBj=6-SlIVhy~dPHCF=(OMT-3`B~ z*>hd$hM~>=zya3#iM$|8mldY-PaKWipN`K}NSVdE{7NO$G-=HoNWon!N5Bkl4~-QsE@B zCACCCQIY=-dnXeETZ_Y!Fad8#3k(5kC+2X{G{^)k&EoLtJ2!R=D@5b2?Z6UFrcR!S zKDt=NwISf!qBT;Ql1O$26QQ+NnlCgzU0%qO=GErA%L~mVFe$&Kb8n>Us(!1lZF z&L_=JmY->UNMQHP=AYmL7nT>u@+tng;q{H>i}>~U@!cL!HeZID@&586%DA*20{`ATqA78{Em32rFg&7V zqKBBQ#w{_oZXt{j*C5%!mRN!go8;_TZ*kv^Oo?7R#2RtRKUzAW=F`b+J_~m>e;}Ph zoO~gXFX#m|Kb+CV@*^V|zf!5*oh@G9ge*RV@(t+#^nPR5?yqnruq5YeGxA7~r2uY&y~k#Ug8vNkO}? zU=JKW(mAZ)RLBt+hOE&ZOzDi0)s=Jtvp#LxL?-4+V`+X183be|FT15xE!%kAy5rP_ zsOs|zzd-m*#bQi0?$+eq$JX}k(;k0Vx_Bk7f{ZFCF;q69cWZCj!8_Q!!(fvJgEVS| zs{wY*NScMe2H*82>J=_OyTQPm3~CxrB6M?i2pTtk1J3Io;UB6JzPq0^f6{ypoqGar zpqKt0`03Xn0JyRIz37?VOcup86dif6+lJD7M8TmaQdm4Q$w!>poLR`L&uT@=%s8o> z-f+sMu?C+AD1APPwh-qUuI28aAyP^rw-t4bDYtI18Y2dyWoDuKiXh(6iOHjgEAxCs zTeue1DY!W!uyx}*1NOR4eP^K0{6b`w!Atc{-^Q26h2>PqQDMbs#UA6Pi7!1Nqg6y_ zsy)fc*C7p>Rn%nc{`iLuNKrx3F)X|-8_O_&AT)OL;Rqq z#z>o}IaSe;r~w5}v0nztrvnvI2r7-b0DIUKA9lr*z!fC4jJ4i$L#;%o!%SnjUDpC~ zX;#=o^Q+B|njbf>F25xIU6<)%>0Dk0|MSN3B?6nDEH8pLeF2WR49DP3I6@|izeV8V z=&jxa*K`40QrNyuWTF~&!B<@Gt9h>ZOE_uFwc~YM#k}~Z&G%skxZm=2JdHbF9n>P* zyWl!vy`ZIEonsxWPu&FAm&QV>p6N7q6fVHK&n3Gv8|L&TxxntI zBh+Nv5`om-D(d93S2hwK+)j5PjAb_EBv`#3uO+bEraE*?RE8w?f}f6|OB!%8kOmCg zz_og|s-$%MUsHYQg=D&2L6xt=Sv8e_m93LFtQuCV3oLo-^;^HV_2#Xg+8E(WaMz>-w+G;YWN1sc@mWe8fdIJt}_7iGYjFz$Cz09#q+`yM))l&Nj zvDgeZ5lDvzn`Ht6k5v5N$%m!Zafn$0u4dYmcQdM057pqye8BchJBH4LjG;4FeAMD+ zlOERjlA9&T;ZkyUZpzHa@}t)2M9zc_q%n2txlz^zsJVf$pw~ENQM2?QK#h@;OtCaV zGZ{t2Ze1;>_+b>PFSJ#|3QvLoQc5<{=}bfs?3rYFa%2+o(+U{d;*0Za8x4Dyp80Y_d^ag~9#nf5Mlp^)^CUDWxP+ondX165+5VXt{)hAbpp_lj(+4S?YZ-otfN@ z-Xwhz?vpiT(pjdae9KN`NNxtJds#hp{QA(P^6J!@+uXfrrm4I!Mjh8;gBFrG#Rr}pq8_DNm*n1GvFbs1 z=Ljv$PWV@*q;F{oNh=Cgg;_5bde$2lhqJStKyG%zpbZJ4hA**b2|J*6Gx#!#N{Abe zyCZ{eE^q}WdC7rToSdBMl6@U1bY@>ir~!l2GX02VL^CV7q@wA*jxZmpE3}!iUBntf z4?gECqqFv|A7ksSfY3kG4)&OsqGx+7VV5jk+*-p>CZ7cRr8^sj%3-tJ6QhH@oZYr1 zA9KVtxN8$^0z4o2woe&LAI#R^M)_`gdty>U&r8XSl8R3!-mP0bo5<;#b*uaF()0d} z&R+F(2ew+(`)q7MTqOCf#~X;Ch9F18P!65o5N@5H^_6W0yK}v6LiL*QWS?Vqxo$$) zIL>W*wLL}r6@JE%)_P4PJ!8iw9|}gs_QkfJ0^bc7gPUl0;&G@xvhEHiM8pM+(p&Si+y z!WSQ7CO64ZQL{x&7~U<3J{;j*t*d(_+X<3n==4Y`t)w%Z3-TVG?SbOia?x$wiD!$P ze97h$2_+p++1U`vHZTGf#+u}6YiDO{c=yr=Uwr)24Z^`gM+4#DApujvQ6Y0RJKyzl z_tNWMzIgS^mw)}`@1m>SY6)FYv)IfE-mH%~Fq!n_^h28k;ca_0%C!a2qP1Gj6hET( zlw^@?T-RbF0c2gQ;u_SF)qRIClT_h;vYJd3^<4^iwO3>U{%C^?^9ts}VGXlN?O@d~0>27=I= z=#cs>>aAx20N>D|;+(4JX);Kp2Wq>v!yUIFqe0aEt{V{Haj+#~Y?j`Hg^a?nFio!R z&mW^^Ww4A4{X;Nvk@Fos>e@2)gySajBd0J^fPZ$k`=A|DsR5(Y^kfab?N)?s(76~B z>p1?!^`vBx339UtHqZin#(P3$A%x6^nN`HR`-9Xj8|+Rpw+Y2u7NH2aOzZ?t&qi&3 z4M}YJx9qa8W_mCVV}$wHyrwDXbZq+YM|vRgqQQj}1VKHm2RR<~M9!zZAU8rG3( zpt?t^FbjQ8 zrV{uk$q|9Oj^HF+TI=BHp9p69BsplnR}%QEGld_7WX8&0lEYR;!4y0SI&B? z7BYhyb+=l#{pPEZuiOl*ZD2DQDs`kjmKwLZmA=@B@q7g~Nyue(GLvL-Ba^U62FT<> z0we^IOx3#fcljsG&QCKQAQ?#Qzwo{9Y1?W`veMlawgs*PTiv=?r{6y3yqD*Fo_K@r zv*oh3?$WV2_@@xQK+z_TH7*G;F)|Ut!{O7YE~=TS8^*Mg;!5NVVCQi&pcU@VoFVr< zx_1Bc*$i1Y`TVwR+y3$8wR@4xmE`Nvl$@!;B}`+s=)-V0~$z3@J~ zYT@Jy_nvbdy(fe<|PVQa0c>kGma0t)ByYBtw>3>{($ve6D zg15iE_sT2xUw@z6|LpX=(=Q=3pl5{q@Y}QZ-#eY2n@?qhlQ4p;%T-CRlI4O|(L(@gNeMwZVI%_cMlAr;KnGS@gElSPxL z*rsZj4kx3eGl8I&t>!Jwgk7Y>@P=y_3sZ!M4D5jgxCUnRt;yATa+P76dTmM}BfyEA zYZb!7q1!aAP%pcI*EzVJFSBhF4ziel@okY>=+q79u(jul%<`g}hpn+#MDPy;Xrdwu za;gncNjjGFxi0~$ZE2+Bme-_UdeF`3THaGy<6s>4f{1{&7s-c_EN0XDnDxZaOJl)0 zF@mFb4TBt^dIX@@Ahnp|%lgVbIXwP2prO)2LaFa@T|cl`FST*e7< z&`MN=2R16_sb)7bx|!asYdMR*uB_>-yV6~PJp_D#Zp%DIg=6KyFS-QSS1pbtV+L93 zhQN)rc6Z=%mf6MBNv(kSgVwTIMje@f#ad@|vW|weo48uh*4c<})Pi6P-_OYpM zdb_pE(PUNFfv~TJI;LfB!<-{5(m9h_>dxIpUQ@AIEPPhZE>8MOFStVBr)9z$FWB%iK3jE zTKrv)+*o>g8L}}vkrNivpQA^nMvrtJpUa3=D7l781}x?_Q4IYNoA<+oMgdJ@KwA;` z`p7O<>||29rPXC2p04*R0`v6_&t5F(FetlyAgcAjz)*`3b=#p6vlqheRr`BVK8e=a zR0}@Sl7h4*dAt6ld=RETmb1z}@#6t%Eo#4Q_wuyMAn?aJPVRV+PU^5248zU}LGpf{ z6oa83q}F(ZywRjuiN*Y4)jqB2zy)w6EyNlt1VMd_ya{psu} zyG)Ic`yV!KYH5Z~;8sfdaE5_QAjUE=QHYCMghM_L9QZrJ&?JhBUoX?n@)iFr_%-3L}aD>5XhRrP^`McfV24HM_Xn z_!-(5QX&@1$f;N+6;paLvLwecJ$+cxO3Rsqlr$j0^2FI7DjST_-Z=Q z)!bpJoP;&SQ*mED+VQXmn(QsQb}?s~WV@J7DkR#BY^mTL7Geh`7eyf*k32Pb_F{lr zm#C(T2zHBk!ZjY!y~i>q4Kj-4zwm7#6hU}e!2WWug(|f|87E#FZGJ`eo|jga8FyOV^4F5@V7-LnQ^vQK3mwf-J9P%iN{e zoTWPIYMixsUyAMO4quOb#}CblybLztu^u8OMvE|o;U?*!*tcRhi*SOCU8AL#pfn(7 zL%C0-GEQfYWMYv+vRt&N&B!mhP|S&06BM!S9;Yp_-8?J+HW@IF|DvlCJgw0c(|mL> zod<4X(Hsj8HQL3j+z6XoQNR&XX&>H*9|SZWtj!JVRAoJ&@pM`-UD#unJ~o#8+sQww zWZ&p886Mq7>>?uN3KZ6CW@($=Onnu5+_jHUvN-sP?f3{!n?0MkdG_Xiq97>Ys-SP& zeD~%nH(!FsuZOm3t4~_0c&BAw9|}JRXk=tTjjVCM%XBB0MN6b?nK@+LvWY>->~$Uf zoR3t45k5s#+^NUtiP;akb!zV*$H2BG$z!()EYS{9$(*1|j31PgXU`QhgEk9f*Hsud zbOUNH#nl%s#E@%f;Th+}lqg7|EXXK&=0ggJK585>ATthFSQdqt95^)U8bN~{q^{ne ziLu)tI?PuWli6gl$SeqH)ni_35w?+8C93NcG6t8;oP=Z7SWPoyv&kB-G~YqfcFLx$ z@q2c8gpFqG8addt(u`drw07;5teH#;>9mkU$t+(gqBdLOf#<3m7v$)KsSiUB%BTq; z8U!rGjp73UE5u`hlnBfR&^5wLSFfK+@S!ns|Mb~=zkc=p7iY0m{r(?*d+#@I+<)_x z`|qCJ4*!%N-hA<~Ri@A0#_2Q6${wN~yIM_)Z4M@zRY%xlGWit2>#p3=L_#4#B_W71 z{*L?CMc=X**ov}{ZU76dA25D|>JtcoV`K!A(b$)>>9Y!T723uTgr1UZK*pmfld!?n zyD)*tiRM3;{{ne?<}ZSs=0BXjdh3~6&)+(gnZI!B^!!KgaAyAe(W6M`bAJ9ZJpal3 zr^xyPFTQmKpS|@wyz=?^PvPJB%o5Gft!MBfJ=UXBx6aIecI)}67SC<|=(BcVYIlI| zV!^mB!}a-S{!`+=ZlQ08R=h`2ZLyYuFV<;6X>fkR0A_lJcS+e!2qf!R-hcPvcb8te z{hKfG```Y<3wK|+di(jeZa;tFyKnw@_eOX^s$r9=BSmwvW{yNHOP>gwrdMJ@Dtc9T zod}Qn4Oq+=(ZJKyy~SK74m<9coyCIOZ#%Va&wZNV6lr)d@8*DWw>n@#$(MV#ADcJk!N zSo=6*)+Nt6!jM?9k5*8F*%)Igvrx*%3P0Tg%iEQQ9OK z+v1rBy)depN%kFcReM|`6C2dBfab$2Ph?(3?5Q&y*rfJDY=O-S**vcX9I=?~ATtr! zj(_oQCtohZ8I*KiC8FHSE8;$;ZR;bWv>YCEz0cKbf&#Dph~J+J4;W&I90t)A?B3pj zWN1EWxNAgg-CUc-(IU$fCbx%K$(&#+*cet=Ozt4Wm~-sGT}8M0jz?;z9tRvWGWTk z(h&D7JnZIB*Js45I)$ucG)jkyt87dy_azCIwM2b_R0|5nz`iFPd=unbf~!FGnW%|A z2ri0c;7dr>VsxX|y3lY6Zxn^vC@78sM@Y$jt-!1(+HRV~-`~sf{^&4#`AN*k@>*!@ zwB^z_;YyXo9Z`nFI2xm&E~H$KRtJH zZkl*mTljYsp83g7e!^v;=bo!x~OHsH8OU!}*G~okXM5xAs3F3kn z3tU#{8leW+W7@M9>)|uIR{_TKHUeXMO^8~^Fx-Bd0ui$$Doky7i(!^343O>T zs!zGpa!|Lp`5tv>q+yRbgzdp9Qe%3&j1PXBQetB3Tt~S~Ne_da6ZZaQ@;6<*qxE}J zuL)N!4@P#K+Kwbkf)s-dDJH^iA0>$BYnvB?_RWikONccvUk08>(aJliIKj)pi#n-z zH-G$)6mqb-vLE7VjD&|7_EMQz_*vMpMiaqNSJFZ%9=u`U2>P~t3@g1TShHeE9zlhZ z-P=k;Ghj+$@qtSkVKtmud1M&o%TZpZw!MQOq?^CgqAAf*LSNeR9l6h*1|{rqtdlS+tlXf$bHi0-ZF`OCMS zCbynJh0XaZ1aYPaxy_JaTHf@#V(n{t0N_Os@vWz8wp#TinFB6WH3s7>sc>(GPx#lr zeL1J%tGOft&P*_4Qb;zKF)(Pp--e0jXci@~+-N206p-=5esY9a7S(mrC~yvOq=SCS zoxMQxZFHF}R$UW=IDAx*5@zQ(H8~HSmZ`~^PHa|~{TdUCC0N4T3~53!{;%*N)SzCX zA?M8E@bOcNpHn*nf34^^<;>2VlarGI3>_}ZxL~l&@4nT^pPF`_N0~M2gueY&=O9;_ zi`XiwtJvc0hZ{qEJQOB5K7i2~ryF-rn1I2pxPA1znj=<1m+uD zDs!XD6%jC_uAx#ZM~$^pv5k!}c0_ZEL-2Oy_;p)+1lIx4_Y#tAq?Gz7^wH`c)ZjVS zCZ=65%}5Mj?dTzd|C55OM5bnYi137FsX|GfmdfCpdt6-`11FS_sv@pL4lTEUVQY;< zO(YK(1W$v4BeuQBIbbwugy)%R3KCm|_%*dU@hjdEC{Pv|E9e=oe`M-!&%AW>Xs^Y% zQF;1+n|#~PenxuT99S_!vfy)?W$Z1$p7VVWobo_)1a%Tvp+exLtxt<0ab*i)_s5ihYFm&+?4zg3RG7ljRW5?Nlc9$~;&LsBSY z6x2;~9R3CbzuO+r*85* z(osUFN4yb<8cNU!0XUva&;S@XkFdI(UiX?k1^&Ku5pY*pZCqVST9J zf}2AzNBAt$YNHg)daM+KE-lSo$ChKVeFV01-K62yh~>-W+cd(m8%Zbw11s$#?@vL{ zjB&K#Ajn(WIHyXE3rQ(5XEfG%a6Qd?OElu!an;CGh_zV%+015XlKbmHr)cL=!P&IR za8WS*SwjAnohq$#$HMQeb&giTxAb`j0Zo+@(#jV6IvfDiG)^$fDN=}EdYGk0z|Xe@ zdo5++5?ud!>pim3i;CzbBfo*YP!cM-G{I_Y&8L=799l3*7Op|~idVbl5vJ1{Hmnfc zpFaeT@%;8AQQqb>MBOt9SOYrJ6Zfj{@jL_>Z*;(7G$z2vW;hFtuP2*&G-(EaYp9 zWDT1(Wy$aQ_r}uy(sl4r*olca*8BxY@|i_U_#1E&K|63I&LI(V^gzc-t$(Gl$4 zy$XhK6Ibi^XM6i_2*^>*#S?qoa(HLk(j3#Uhc!$e`BEh5=G_YE7TpRGS7~Wz!U~r& zE79>)81XkpmHkJJ;yQVY=Es>u24Fv=cKGx{9J_jictTKo6Qx^Hni@n>S{fuRBw}l! z!RH#i7Bw*c(|MH_ydLC;Y3Ztu1EIG_OJ!_kDr)5j(xN5oV?Lo#-B%F{4Qi1yF-b_K z8`P7xJhUJ6&^XZ9GS~)1J2G-VBpMdEz|ju61I54IxCl3*QT*%mkB!coRehFeFq>?r zdd@A8{9kXJM`l}_g-^S|*De;RMi$#Ym(b~nS`vXj4rrIOkV-_(pL$|^nhMS*W=EGY zqAFXp6%y%qW%NP>amx3&z7f*@GE{9APG=Ii=1|Plr4vACDWYn zl>Y6qTw*Orkb3@yWJ(lMsZ>klVNsG$D_oF7MM$NmRvy;7K6Ni%WPt$tmo&SK)n}DA z1#Wuxw`Cs_K3c@wz)Us^7A+SmJBL^)7lq$y(1Z|`BUXG4P_lV^nQVD)$T2}|r0m?9 z)ju+;HA+St?AY3{%-YB^-ClMC!9Cs&)Cae z{k(d;dbOhr5wL+>3T>r^_VoxHM=}z3B;>V#qrqcR-MoXVAK`x;7meM<1nMFWapgXm zf6CQRbO~HjTskd=R|~PKeO#V3uWR{n@VHb|iH3`2b=WHEjiMJv5!#4**~A!~sM#>? z;}+wO16naDNa^~i7Pt!&ta$6``H$znoIgMRA({UWetmZ9`T0-a-^I?`)D}A}Kf5(E z|2c2Fod00{GW*-t1dnf*cO_{e&!6E-Ke*s>723`*@3vIzx!Y18fb8vb~Q>KMf z^a)u~bB49Hwx2Dui)feQZzhPexQO@15FZI`BDdG~W8G#)}PILlo8~vb}7= zrUE-olzy&*6s}Cop1z=72^bs3Q6@!dL|O*5r&lWmlZw18pjEy!oe(4`El4q#^HOSR z{*o(SG#E)aa3RGVHEP(RdG?&|vJnocu0Q zqM(OamOoRfG9SqSF(@nQeLp(%kJcogyvJQuvV4&`MNCU3Q-- zq?iwI8w(hc#1X_HMJGl>5f3iXi^vpi3~;i+DsEXtl_u^Cq5ETz3Zqv0`rOvXK!cj4 z+uMletMRxP@g$2Chww4vxG4Dqh(klz;u#Sf+XniLEl%By=hRJvPYbn$Q#X!`5=wBE z4{Uhh`X3`@3tXI)R9zO|WSbagbKroLSiB`3O0!g?7czH{gUm4OiatS&X1wHG2OD^| z>fW$uA#@Jn;25MJcvzc_Qi38!USb=3*@Mh1>&)vpf84ajw;a|VaaiS=&E&vFlSDyo zSZl-iD2tkJ-p&CdgJW#HgG~}^6`0^zr^fGgS?GYJV!t=U!YZtMPK1)hZbM~ba6t+* zr>~&>mwY+#fCR~Lj-zSF!L6f&c+oU@i!yb%=9ZXBC&-A44Q=FrUMZEeyxn0j|EOjm z=M&ju=4}|m?8kf%MBN74hUOyt1_#-OoIv;*NF4bZJ7&qt%HbK>ztO_(<*LDm75X%JAWef#aR=P+uGz+J+0SF=mnnk*5Ifm>O>cLk2+z!KHB zdEvblv|T0D&&k}gUQ{;sBB{Pwz0Awkmq_&-#4caLzi+%SZiu~djZ$H%@rp3o*ljRpz)aycoaR`mQeZ+vKJNLh$Af_cNhm1x<4x1wJ-3_~|P zdMU>w+5NkD!U0WS8qIi+H`iQi%|5@(B^#)+}QZ&!A4(qGAq z`+WU)aA!Z>w6}Q+xzuVp9;7&nELR|XuEBTGJQWv z3p;Fba&oHvM18HH;SI~ZXAm7Yf1wWFb z9O8=Py|CpqNlWf;(3)Bi#?`0_l4$`IrWC)JM_B5-I2PFB>9xpk!!Q`EZBaGbEoQr= z*^6A%!&fMjuvO% zHHu_HoBeP_e!q4$Gr~urId=HU*^0MD#~3NdD^g_VRc(%ANJ=SJv!nc2H~Z~NT1sSThPu?8cE)5<)^G6-=$Yi=MEC%4;&AAp z& zh9GAi_gJNOV55JYc)|pZ*2dYiKIY8kbi`nX&skKC``eC1xjBImoNs9!MUbNFExU$_ zR03Wn4@5}j7|HYeLkaPZ8r63oNT7oG;0p3EfO9pbSzaj@4^lGRq^UTu83(|?DT|RD z16wjrEP^8qh=WoHI&l~+2?RYfoza2|pJnQZ6QW}X?`r(K4!aRXUEQYSfTp92>vl1n zi1WdO_c$L1zoi6b@JXMEqAzk51INx%Qb?!Q(t(NjW3K+uajWc5h3wPD7(xsVSQr3= zHp5c8)9w=-4dq)$Ys+X*L&zdb)5y?*0Zc# zNy_6PMDz6aAHZokIJEhTaP=*D;J)i5JI-6wLYqxW#A4vN!$t3jNm3>)ioKZuMHV8R6WHT5PeuOW_8Yj=Fq3!aG3}eR#MAW3C}5d_4vJ5_jMVx57~?4ZR1+F{)=dWz+CE1xbHt2-;{!~>mG)Im18+S_N=Qgs@^Rr*$u+t`9G8hqOcWrNk2GlC2*Rkzm(6w-!mYE;7Be#4U00SxMMB0ZnsfFcN3gL z0dT^!ok>dm`^`yM5hg^sLXT%HA>QrvE0_g zX=Sy;6oNFBh6k#&DOQkVA)SgGfA3=mPmstYwG0mqA)rlMo!a9X$>yobN)>OLYixQ- zZ2tXed2e3n_7=} zu=~>=-GB2U4B>}o-o1D3-J0#{p2%>W8(6ClKM)UXI`4`tUa)^$dg=c8lLW4%f9X_w z@f|-1W=~lZ5)C?|TO@58PtukSUlVG=(dKJd2->fZK#WUcrEdK&F1vFGhdt3@e$e8# z-VjVn0#!uQ4a;(1RiM9L`}Ox{Pk#5uQ+J>H>h_rrZh!sdcOU-w`|0;?pM2r=rQh6n z_1f)^U%T`Ar{BN*hwo2ayZ!5v-+g@P_O*}y_v;t$zH;Ht3m<*=#yhCabK%B!=ij?? z;f=fRU%!3nv)k{#yXn27Ms)V=uddwr{iWN_zkB=bbGP3=55o;@&^xbxaQoL^-g)QK z+rNITX4QBNfgKNsz)oZkO_Pkwe{*p5TtPFa*YTwlG!96k`*48yG8+C}E;;tDT&kw66# zQHgATr^-)l`pdo~>)@Xci$V_w5U!Xm88^&HVjv=)dO6+f2@Q7UVR|e)C}{u01YAiU zJY9;Ve7UZ`Zf86#q~*wXv8)BtdjjKi$wxVa4f~Es{0ZqkW*C}HdMtC&*y+8C5jnv_ zc&|FLnH`Ll!MlHGq0P9@oMcv>+Kkv*p+wPP+r4}Dn$|i+7GjOc%wZ_>mc%h5lQejQ z6+}hS8pf#_yffSE#dACAPw*F&>drkRv<^Ix%Iv&JMjWjiKBL-_&7tPa1a@W=h$>={ zxda}z`fizyYq&dVuy}1?fftc6V*RXrgpFcNxcbHw+>;#nl8W3Kdt@<--Kv-|NQuDCO;?X=89HZ8 zTCpD}iO}78#`W!kbjHv?+qKCHCLNuZGMWwqwJVt(Tg0o9xT(?^Oeacv zY%_(z)G7*vsTL%{NyJISqEo3nR9>T2q9kux8oa;_k8y9G<0v74W&NyNznaZ}3m?S! zRFRN2>%dp0ou(!+0KPKPi7kHr2;3{}sO2Oxa!e2#a%0+R{mJ#FgG8O;D&ZFBvG_e{p3A8#DLxrk`_l+7|C{6wcSxYjeLL|MjQp zTQIcib2D?N{_6%oY!0d3fTL&U&Ts$^4Du2l;uYwgJB_2}*dGM?T=Nt_s@FZJ&q)tL zRQ)Ej(Fd)=HG@ZxP)j0(K_V6zf4$Eva5~ZvR(4n^$1Ea>J9vi(*)9zdONon*ktgy+ z*BFOqFeM_&n?5nnl7;m!G9|!vBMLFnuWAlD!8#dbw6wEU%GqARU58qg#|T|wZOv6+ zDWiNdg1H{r(nk*RKe@tXjz9`rKEi$K0Th{UDLDw$-ET6yr_w&yVK#%#M3MG|#nQp1}3 zsQ1g~$;nBdFUJpp-3bc9oa^s|RXM*_L1N8bh6%9ISggCs(@;WmT7 z8OFV#i?O5#azpiU{pwbO%tU>%-McYww!6eE@V+~&cr~cO5HuWux(!D#d%@AdLy6l_ zyaH>c1f@YME)=E0^q0suFnwuZ=1&`)@rZy7IJBD#3?JA(s`ZlZuD*Zw+i&({5Bi{` zffJETHz(M*nPdJE-%{(9L_^(-c_Igf7E-@I{ehR*nVUoSnf}8VLXoEI|!ZmmB#n#qL zf-||5YgzK}Z4F)%sg0t2{=Fg!u?BYCFa@wZW`U|WK}a56uEKjh%rGeu6nyV>BKvNa&`HZk zm{%|p@l+-y#bOybrNlA=Vl0*E@9$4$`o$h46HoW_gvZ`AQi7*_h#USm;Jtw8=E&?D z4{iPl8&T0m->~dc8rj;Dv3fYcs*I{#Bg`}kFtTng=~gq*{h2e1#Htzc%U}NR!bx)f z^!xWde0q^=_5Q18?oD6$)vw6^`KRt~a{slD{&DTn{nuV1@Iw4`@5Pfpyn1%UTac0U z{vUpO|MX`+y!!h6)7S35`HDcgyZT+=uPi^{wf~ezkghMZZ~?>P*jvI_LTgl~Ipwmr z7IN1WJSZKOj~hm6*Am=kHXWzBIT60JuIXxT5obm$;qzvEWTJr`=(r@r<+T{1zr-xQ zagG$>rieTVdriY;PazS8Ug5LQf8wE;hIozKHnT}an0?6Xn_P?F_`#y4Q}|ouq6VRD z-qP~p%t0Tihv_9`0_p7~<)Z1dPcvklGUbbu(KXUFM0aA@#v5Yzs$hwVlFwGL8f6fF z9L!!yOh`5a9AWG=1mB(MwWF&`K5gX{pfikl^Lgw&(PXjr21frQV zXJl0(x5I)>V1ZlQ!pIoSV;`v2-py=XRM8MbQObO3iYS+K9F|!SY3B zR5q#%h89tTL<7Qo%nGc|3ZsF!s=7qinBjv<4`@2e95gNd{F5qsazAy7uzIm4`&iz@ z`%^epE=>LCD{wyBZnU<%^MPm$ZB$aT!Dn=$*4&GCAoN}hK zDhW&1mn>c1@E9NOV)r2_*SNpgy<2YHTP?5NTaEW#T1ZCU)PrzbY<2@h=ClC&!7~q% z#Ud>92ka?|LMj;8g9!*8{fZk`OXL5EX=&RFnw-e=JHw!X<#Tx z77~s6=V9m)yq%!Oay9UY#I%&gBnp!BlU`U$;ZsXx4BRnYR5C5bCG0!QlntXA&~I&wV?w$??IhIhXfZlIMr{WP@J6+4eYX!QGq4Bw}#WCd7C$ljuX-XJ4#8qh$LMnQTAY$f@3B&l)$x zcw9)wBa_nBa}B|t-}E-x92`SF?r%9D`8QfCA*7-gxt1Pnb_+$Ihume_Wi8LqFg+Aq z>tWL^gr9Rgj9QxC$&`G=zfwTvU$hfMa*4=P{SOSfu`YECxQ5{;&mgtRJPYX0PBu?Y zwz}>LGN;w&Cf6%p(orsF^(=hpD=>;TFza!RGgE!*QK(lhleyEJ;fm9!w;>zib4FK+ z0JXFwtaVELTDN|LF~f}NB%7z|?1hrXyPtbq-El2S4pSXwCM*-Q!w6)83@Nm=+rTFy ziJ4eTh{P62!(Da?tg4{`{?H=0ChO4Zly5 z`SbIi!y0<)X*l=gt?66Ol3Qo~^5Xxz4iDh@>G==mufj_{{mYAF{sO$@)A`TH{D010 zhV$p=Kbt?lZZ)EIO-}!p4{gI;?lH}0zGTv8K7J6)d}2(9C$`G45z4SWIM#zK#imo~ zTI}E*1hi~iP#UP@x188!PV8%3`e(78$8~`|_4vt_@%(#L0D_okq11fsf2-5di-Oi?FNs zD5N5Vf=F{It)*qAhvvueo`O}m$9@&>l2OixXe&zKW;~k zEOa*w*b~`CQ;~t14-s}nr3MzcDaj)?+JafuKJG)=rzjA3rp4oe)Zo3Xy6|=NLj370 zsE7r?GVBVBc7=~~f**HvesmNX?%mqH~bwzAOU;3AllTvM7GhivB@ z0!>!JuKjzuRvM@az8Kj9W?`7HT{Ki8DWqf}5fdcY4^QxQ-us<|c(SFblHeso);{N| zyJCt{9GaHJ9LMC|LWk{D1Dak!dBrha1%PE7)z>K3TMQo-pMbpz4E`@*>;lqvabJDW zMyZ;=oeJ!4D`@C?69vvF=a8z`g232vG?m%}-mpj@s}!FYaPhCfE8?=ODDiXx|0?VvI%_8r_d4Ar zt~&dnk=IPZ$Ov<7k}YWkGyFPR8~yUkZJs!Zw@jQUNHEp_P7B50+^=gT%|K2M@{6t> z3bEyr5Gb2Z+OG9v-qg)fPNPraAI2W0Uvvqi2PQd%S2Ls5z+|qafr(Sn{aE7?<-B1Y z-6G7Y@f{G_>n6p1z#=@CqbezSP1u`0M_addmu4?o$R#}@kgO1X!)|9IlgJCq4dKe8 z?N7OR*UD)QFS2c*Kc%FFbYd;l;1%$hVOBCHz(Z15*iiA=Qp=tW4oodES7Q1UIo?-IyOHgp7*~|`P1E(bBYMmn z85xJ`6Z#J5p^j6jqjyhWVn+~z8QyJ5k zq1J2=me&uizbFZELytvZuLEYyzqxMx!H4_FjeZvgUm_NqZM`#{y>NQ%6r5p zL}FuZc)P&2FBYkW6nDPdfD)^%X#8tm|_#FwjdJf_-u)oca<+j`*DF7v8w%eiPnn(ALHbAz|TLQUcGVt|3gRF}}C1 zm{9}FcC?8W)LmV@*@mokFEywdj&=J%&8yucTOwK4MmYkl8CIn0-+JrQ*#G9T9GB!- zG)6Keq(mX52x2O^Cig)?_bqaZ<|7UJgKKU-P{d!&9}nz}%d()vw@{ps%xQ3akZg(Z zl;Q)0a{6#X>7c+NFm6JSqW4yYko`PREZNv8@GwYzE{$bb?vrzah7?8n5xGw(k&u0W zibUW5K~BLMzQvWalsYcB2Gx#(weq~vxm`?2HCIIp?5o5j@Kx5*bUp$uGh6|4v5lFj z9`VZZ`#l-;Wq4SJ#Ri8pHreeIEz>O&iA5u{qwi%CV|2oArG(E+f&G`bA|#?yOzcPU zKUnQOhS4H*kOq*G7WP_N3BsHsFtN%$8bXvaFs(BDDKIFaeJ3?tCpktIzb)41Y_1ld z(u!Ahk-mQ(P&WUBY`K(Tzn@-_#L%|xsbh5!MW3ztK|o_C!%ziw?WH;`(G#Y* zB7%70iGJAIi%o5@N$@bZfvi6FM&OI+;Eymerc*&${ z{rZgGSAnzNeRAQuPp;!L&)t3V^xfyq+&TM^oD`Fy7>ob!*Kgi_|M%a0`tjed{rdLv z@813Vwe@jk+O^Prcjelh*I)kr}IMdx6E71ZfAz>T5l&-Iz z2F|*`W?22e>;Hy-H?Sd`lmA`yaxQbfAzn5Ho5NQL=!XwAOv_>PUpGfU->=y@3psql z^W^X;n1U&o6htwxMk^%*xj~XVcmw9g^OtU&BJ-E$Kl{sz^IvUxe*_yv|L48$&DO@g0}X2<|7c!#S*pQoigogkuG!2cyO?V+>Kl#Lu&53%0Ro|^%;%%bSyr6KsdX3DT;Y_zT z=3XGxk70B3IAJ?;bF02ry@D;Tuw_rfE`4L}McAodsD4FYRM+NClIqJa{IBr>V3%+c zn?9Z$t~cuA;Fs7sov_DIn&3bVnHskFlG{lkfvK*zAf|mumQ*bt${z=`F{+hD4iJC3 z6vl7*{jgqszRkb{UuhxuvIrBjq4E6D&2B9=p_Z6V4OT8%8`py>BWgs9-dg8Rn`Fr2 zri)pQK8ubynS&I^N8tV&hah!o1-lxTjl@ig?dD1vPd^_$>OqhhW=vRK*D$Z*scdHl zdUs9X4|q*XNUvAnVJ#r87P*syYEd)nTxgRY0NW_(MA~m}AH2>x2x#-9kc>y>#eHa} zBP|$OS*HbUBj=+e3b9zd2l;4m7E{-vdF*$!%cj0zS~y^exDZdS9i^~;U?1tF7K9i) zb=a9uh5kKch#D?y;zk4jDM%uCad2lVF+oiFJoV6VuZ&XDiDZ2S>XBc&bk|{ap0%w| z7HPVq(Nm8qGO(OIQ8~WeTr|vpKmOjdS~I(4SYNePo3dEUxS~&`V|4Q_b&C4FHHZL zf5h2k1&4*{3;gFb_<8i`W|ya48c6MySRUHyd`OZ}!=y zZjSTj=;dOLzm^j?HjqU@O4i>tDhr-cuJxQUvOVkCleDNmw#p*g*Wa+n4q_{uQP~u$ z>>iWpWUpqxDyQU>pP6)9EXyxijtU(uI_orN~VdL9vsM>_4B8yas z+Ibgz`c+zHy83$??9!iiscKrA!Y_%&yC1#lU>>O5=4mr^#OX~cI+9A2&g4gog9mMt@lcs%whPDzO-M^?p<3IeVgDd8*d(zbV9*v>RyOk**)k2a zhuJo*Ff1qKm=m6zV?T8ePoQf0q-pW@FUJl(3ZmfIc_l6+#rg`&>cRG?X+gN%P&4b* zsaE1h-BGP%I>idv$Ea25$dG#ozxY9L8j7YvhFGN;sG#e>{(4X2fZNSIrl0sTp%Pbl(U8lH=tRCjyyYu>?me#xx&XQW}R%wM~ReX$3*E& zPmMJ^#yHMZx1*OPn=wpoQB#CnY!TXvlNvR>vxebfu;!tGO@3~q$jac%s^RaGCgc^8 zZw{{zKogag_eoMv8t6+N-ATNZ{A6z4+~7iL91E$9autF3-#ClZ6Hgp+;E-hoEN16T zIJXllh^Sj6ByMJWp&W_8Zw>m_!PTH#q?mM_y9(p_<6#gC`! zE8xM=lAq0Am_I-NA*xiIx%F)Zjz2|w;j_1%p1*wSng99ppa1gtU%r^~PTzXP_XOFF z&tN?GQ#ke6{D<=w;M7;bU#>ouAwMO5`GRL^p*I(_TR`~~=TncRB%FE7qtxpg{% z+v8Z&uyy1|^@n|FU)&c^^8l=yXoC8AB$F{_9+pU|w+S8@ZH-I|@diQOz`4mG7ni(W zbZP4BMUDQV%j5RJV}^%2!1G2uqY@(~G0X{*(7b7>CV}|WELgMWV0|=&jm)k@5ce@4 zd?nk*G`$po@cDr1s`$|I5odOsDA!r_08&~ z>W#VIRIgS)uU>kx`jSxno$w?IXFQ4UNKYq@D~Oz~UYq-mC#%0hfenNd@~8QG{vg2r zPgc)zI3hlaZ@T&<9DW1uLWt$paQYcI#?M=P(o^j?<;j!43sIQv^Xj)k+2D+f)%W4- zSJfM+VseFl#ot!HA#daAi~DZ@~UcAHT*)c8jWIvU0I3? zvBp$GP*P5OPXR(0blkNPEp(dZa=(SCbMS486;`5pgy3|T=FGj+%F*!^vZLMND)Ybf z0+0~msd^;LqZ@a4$%DMBb8x*GVM}da>Oe{fiALexpz{Q2i7dBlc7hme)?mkdWO)2A zbJ*0GmNRuE($!%Ydt6q4UkCKrk4t5+4QDhQVYAnDBBoQR@I^t}5XPCo@d4564{_8G zk(2dqXc&}3R=3Fryw|nM8vI48GXW0X4$&TDL#k$nRB$^eO+3&xFlRU>=GU;WvKu)x zdf~%eG6GR;Nz-YKaiT+bG@1|`HW^lEGyWKRMg%uOPBn28)=|sXb>N~F*p8ZHYKN(j zERBF{UmijDK%+s1q$Z#cPq@JujI&ncptNi4Bkj| zu3(VYW-l8grNk0(Ig#*m6Z-AC9)E1}WA;uFm-|#&5Jmh3v8lRatTSNuze{U`xcKRr z`uE{FEx@|F=_7!q{>|X*MT0ki!nG>RUbnGq3U5!R)|wZH zCj=$67B+P;os6ODNo><{c(8^IfMkqKA}NkZ97aowP~Gg@Tb2hB38nbao|8+5iSF(H;*OM!ojAAc=HNN-Lf zily8pj{N3jMmXQ{E|3LTj2u%gm9o60*;mWzH*^@a?NzXjYkW<^n2mYo!DtndMjwyK<%u{@uoP41i5iLOd1O9d^sA$A`JO z!~J0nWt9H z#I&1l`>18K<$x(|1#vS;DKgK>!?>lFXx^Np#)fX{Qd~$haG17%D7w#Zm`=LM_E7|G z?6w~xqeVvgEQksaSKdDXdl9yTU=fYjJjqA$=EN4Ev_77}tn>+XEJqC*0bgHFDxLNd z5u}6P7%3svfZ(>RppruGRj(sTZSGC}59NgBW;k2k8LqSQGy==MnVa$OXLD!ZHFy$U zakcv8rWYKl!-$=!UaJ1tL(##guM!T=;}5RSo#YTXMCDzZdxoFlm*o^bJdeORxDH=8 zBMAJl%7yEgE7hJwd6lAt+@fruWaL=0R>Q6I=8t8kvAsao!Y8BJlHF8L(Iy?Q&u%`F zyQxKO+iaVVa-;x2POfDW1)mfA-cE2faf@2HVP$#{EaEylZP=u9=JD_%GK(Luh6JsG zEFX#-nYM8y+0rn)-h!|goA;5rfMc14%?i~?4>Jmw+0+Vf-#joMqg`Z3b;uSzf}A%3 zhmRoqgfYj)7(1dl#UW~qGsj1FB>SX|12Qivh^9y<>htQ47`q?X0~O&ZIC~JWMMz?_ zgWV5I0a5dsdTwDq94ALuF1O7lBZm)k!fIBR>htPte;79C#?u!nfw&N)%;55@UkM8Tw-3i-{13YP947 zmc(k%TmCqpeI+5)K;0OaAP{EU-7hbE|JE~qzxMIMsZSP8J ze4uq-m_;?4CWBiw)kM5!{jkz|i&V49uGL`x>7}n7+Hwt`=m@+m2~-pjLnxi~Qis~q zg*$FVnLf*P2VqM@et~9n`pqGC7K^u*kw<=I$5SaCWR6p@bFS57N_8E~cC9oF7{f5^ ziar58CtjJQ5`H3qgv5f97DQR{ZE!wS)_~J1fv}3Wyd~jTSF0>o;D1q9Z5M_;ysHnE zNeEJEtBTxc8GD0CqZ6nMA~p&pg^0oi8SJ~$~guz~j5%fLdz%yLw zSUb+tqenAjmpO@GK|=oa%0g=LXTKs+MITZ!Ck*(_1ZcFPPKedri6;u#x9&XM3eIE@sUYwPrhr()@3 zTx{Lxb6CrR3w=<7$5Xz?t9JPKU*9p^J$0#5=U0SzfOM8m~ z>z8d2V+r3P%MXH0XoR20ZRmBDoAo;Tm}E=z7=${d%2J!Ny%WF6&gJmm+k$`m4OzF~ z+D?bGNVg`}T^eEd^i8{4q!WmDNu}ybMPbrz-=BQ%?iX*~dH&VA7yi63{kuEY-r{Mu zHx^F)Y2nnTs}K?F9c~CHmn0}L&AQjL*hbOpw#@C^e(A-9Q?Gw_;q|{?e_`R&wS`lk ztTz16ww1u~5z`QhU~g*8r12VBB_0r~lHgbsxjsV|#;6O^R~Kg9_T*C*X3i~4|7jIN z?m5_CNDkiIAmi1c_usho_iN98|N7_OoqOy1H=kcP^)1f}!*FI6re6+iAyJY9NfBW6 zPNw`@Ypl5lGn{W_BUub1QeST)M3L=)wbtChkzr%7f0A*-oHQ__NKxt*F1)4^yg2P+ zHCrFE?Q%;KEPRV=)+pukbZl%5RVJ`XTu>S?{f}4)Healii>Be)#C9zWUg<{30koqI zn?SXhWNa-h$NPIRI?azmm}}dS^8l0}6tl>v$}<7fs2x8f$$2<>s12Qtz*8|x@6$n52=|QTr8`${+&Ovv z_KUywk3GDPf;>wQ!Eo}7gs!^8Lo4} zj}}`xlgZloqGd8?XAjdHix%t?v>0BG+5cwdqx6dq62#*|GCJ4J!<2zz&WQrEG)f>e ziI@y;S7$;3AxUxrKw|w0w4+6fLEH(xB{FA1d6+vRpDijb57Hvw>QUpAqstU+?9 zhaRJz4j}wEF2%~nSMe$iJ>FPiU+z=ec)y}ocXMPv#hg6VF=0QjNKVwGzCLuIzP{ea z$`qCzyeyn_W5xAauC^zI1ZQ#-WxvRw)I7SDS_aojvPp6?Femj-MZ#&JKc9PHt{HE^ zW|~6PcaXTcdbN6eZn|bJ|Je*bxP)xhJF9Ol6IjEO*m(3Gq376s&6=PFT{+2;KCSs= zK)`bn;Va1o2k_A?y?d;Fh6d5M?P$43>tB1A{lGMFAr_0=o%(QXM>&hy*qSSBQi z>OS#!V7>WzgRJc7=tZQ|QbPwPSKIc>Uur1XmnV+q=nQEvfi zkDkoaGVgH3Gca(&Ey2Slp2!;Tw0`s`ILuR|T&(Cz=Pd^!oU*ANJ&ModbyTRa`BS?M z)Jycvb2$b)xsR1K)x_(Kz`2 z5{A4vGF1Cu9%QY2lZCXar?1{|+;3VI{B?(a_r=SfObJPZy(Yj{jIYfK2|;dvmxU{n z-B_5uv@mmdVdmt*%#DSaXBVd5TbOxyVfxF3=@0Szsq+gnuPw}cgr&4ouPvN<=gw;% z+&TNv->+Z){?xTQ&%MxzImU&V4{C7_mvBG&U|tA%M0Iq{psB|Pv7~^x9`67NhfScEdd69zy7=1&tJU#$$#E? z_sq^apH0I-XoKxF^-9SH{*rv`T2fGUtt28t(#gmL>S0b03O*)cy?CEX1BIu(Gm#Kd zNzTO`Pfhs`me74mkeL-P#50Hka!ij9dSbER2L47LHIzhW72ZbS;*OIcM)rYA9^N-> z0EBfb%@Za{i-buM5hF^x!Pr4v!Xav-?zxB9-1HN05j5K6gJyYb9YdO>ySlw|l){IF zH^t)HYoG{11wP|Yc;|2iUIS$RU=&ey;u*l$zDzM zt$Z~Eh!hY4l0*@hMgr7Fvi#-p(ibySSJ%d{Jz&GCUrCg5Ni_gL*01{20~?LT3^I($ zaUE9c8bOUmovIX5?~4Q3kRn$SWn2etn)bJOtDvmU|UHSZLB#MO@4 zBdc$7VPF^?jGvF>&oI0zC@Rbagd52U--PEs!;yx*1sbUe<>u2;ZaFU%1^WfI=H<%1 zQ!!-`LMt(e*+$6BZlm|Ab_dFgss|=2c>11N_7)5SE zg1sdvwobcV53>wvsbq-S<64B6pqZvv^Xwy>Z+ROegj9o)L*Uz>kLAIYM#I={uB7o+ z)C|dKCg~<4%V#k7qdC)c2<1)3aAvY-64iuT$1ojIV$8t@G^cSsrh==f$Y={sdmYHx zf!c#V@}Q5{4mpRX?e28EE;6{9-1G23<45UorJq!NB9r}45g$K z>6-o1wIv`Od<{rHU}TFAPl9U2QyhTuGuv@rSx}_N3#~n7YB!!W_cu1Z0b16#vrEK>Nd3VF1ebp2VQm6g zSCNVc?sJ!&h%iMU?xViBAtn8_pMb?5uuxiv#kZ1ObwM>do_DPYx`ConDFlzC8>Ge` zUCVfBNX#Om>f*-bA;B<;Ac^&TX=6e2kfqsok9mk-feh>($Zq=OZKWM%pL@!MQ4-Z9 zBUGP2oh&XfyJ~mLkC}(l)fC6(6Ap|cq z?FbRz6YRl?yLZQZIXVAY3g=sz zRuNLE$Pz;j={9Gxk_KDdMg~#wl#pm3H`5&`TW!EvZ@e-_&c`{i9Gm#>-Jr-mHE;5xA znZ?K?oTTK?NdNHS+yAM5cV>nVz1We&3!aOX*c`Qt6iVjtj|#G)NeGk7$LG3v0`O#atC_J`7g`rXKEr zj!Clncl$VyH@y*C=jYXuX24AfK8$JE+pr$4?2B-gUz z?Sdu2v?`r9S@s^>KN$MT_Z?Fi9rwW>YUqc+S3DskQd`VMb#WHB<{F3LY*?g~Mzm2P z9XSGi2#X?2H;n>yHn1v^Mp#NLI_2p9XYXCV+o;aI;eTb{?pja6ea92cNeB1sdp|zJ zb`DNb+az}jUR_(GZE44&+3FmeD44ZKLV;3RXiHnl;VJ1Uqy-uZw2%P(;PQSp-{84F zQR5tL=>PCu*OqLJC0U&r%Lm2oDzQeAWzAf(uYDbU*Y9`VEth&%q`*)UX9v6PIZ2Vy z0?gcVTlxeInETy;7F_D=7M|W5VpY;D!gc#KqbM~aH}Gg-iwi>`U7o0_JUNMSNhNL% z--o-?Y*3^Zc5r9X3`NxvhYRWjsk@I9*fBmN+3Y?bgam z|L5LgkF7j)}|><=6F$s^_#!ZuzmxyFON2Z|s$qU;E<2AKw1OU+(?lo!c)z zcJF6@gHv01?xio@yS4JvFYaD_Eq+<9cEv)qZ!KrEjS%7aj<$#GE0-8GdmlQ>bAe;A zZaKDHd$$Z?rj)d~8HNL2R1kCE*0~KHd88cJKCjq}aL)^v z_voXfpE#m^M5K)REN8t23_8iUEpb7yEj1>&OKptRoW`A@$$} zH7eW!nWljoc;uj8;&4=y6r1QQ;!hAwvSQ$sScu_V)xu0E(Wp|;E|3+a_Nh`DeNu1- z%<*jM6u7l>xDvEXTFJK12=1dCBcL2dxMzu?ODm=77SE!qZFe$5?P%9*n@RLdP2~}n zRqX1ncZfSMe+QR{9e2ce_>M&;DN;p<%uo-pw(1dSTfu-j+z81V8-uP@;`122tD)h^ zgWdzRA=-x?TDvAS3=tmIqkJ+rTZF#|SBW~P19z}?H|}OPyZ&U_7{2349?8RzFwIN%VO;GRc@Z1^`pbFL~N4xOlyu*menbhRQ-R0Qt5nVCZ*^okF!$S{~+V1=2P0R;RJc7c`=@$-)- zeiA-D1iR@2+u)41Wj!95hj2);>Og2zrsXoV6O_VBjC?cT7cgxir-e`baJ#9|ZHj7e zQ&WswbHO!^Wcj_Dzqt3Sw^wfc=JrqC=)H2??H7J@=h;{9y>x5k!?$AFVtme?o2xeq z*MgSzF-u1in2KSvG{~d&Vqsq%Ah@&DB%UD?_ACKU9zNpOaK_#OEFPa;kki%>n|W9R z$LC=~tP+btcn;6OX9>59ddk~T%R9p%nsLGR*<=;kJ-T?)&x`nt^|gx@?)im!x9b6R znU2QHDCu-!`Olq+ai-uG2Fny# zurTKOY$wase&dDbJKh7PV(Kk0s2KOaQ9{eGtBN_a4HkqbM0OPn@E}crI|R}bMh{`} z-YPL`PCi>Nn(X>WsbhU;n$Z$`8+Q1c0rTXLZ2Y|C$P{Nq78ZdpGohAqM8K3P!EDhx zXno_<5xg1FF)b=Y(>B0zoH^Nolyk@^s;PxnTl=-v$wq1lk=L7!i&PmIwGE8w zquS}LqCwE3=3r$ngb$jdqjhg=j*gJ9bTvFW7I^*;bKR;^*GZf{yD)tg-rZXf(rDlZ zE+YqBTB$HMepIw}8A3NgrhSZ@=@25TsGFy0%^`A}l=JSJAKiWPYVReRD?fSW?v3YG zp8e~c=igm<`lXepU%UI^@9*4tJU%K!(85$noDVO__H60u#$@Se2}3bjc$)VfR35(l zJH%c5byr5TjS!h)G&V&v1bEnNvv-U$qJ2hswQ_#(dU>bUMeM=sL7#)OEeir?y7)Gv zY$CQDrb@-d4_q=reQ)uDVmN8bHfmzoh($3KQHW*>C(?dzFhXzeh~1D0+sfu(T*gu~ zDsDT?v3}%{t%dskcDv!u2-ik?HOgeQCt8Lc<8L~xAFzpdP&1UwX5zsu&bAK`c|!1( z;lb&FQwzealn{&qPPTZzW#TB{OGen3OV4*D6*t?DOThCTZ-|Vd8=Dy;Ung2HAqsLv zJoGGltm`X17BZq@B$@JElx{z$rBWuD=LIH65q=|n2tVjRx?N!TY$%yCvW=lW`8b+4 z)l>U~w(D>fIB`&Qt9p3QnRVe+=kO>FvymZ%AmqFzrYI_<>5?eZ_@&(`U*stlUMLT1)aSF& zsAbSfnv%_IE=a^R@^IhEm1kFO{CMU4cUFG(`+L9nKliTRP>Ou^;l4wJ`wj7;`j@H! zwCl5DwNm5ipqL!zo~U{xqpALdg|$Zz`sE=Gg|i=kU#wJJUM%^rRSR3r{cKm%(XEJ( z&w-yIQx(RfkNK7e5{OyVFW?A67sW!*r2cICs$sTc)zB~%y+yOpi>kuC^y<=$rN1tH zxb!Yrdhe42yE-iGI#dCYY$3fqw4g~e%^!!|c$b{voHya>K zESt@Syqnn`Kwx&n512^R-poE6d9}D~w#Tmh(IWgY1d35|E`Sh`IH<34!3sPYe^Crm z`0W$){IS59ecRXaP2jfe$MdguGeT z5KTzjsHiE$NQI4*+Q3Yy&W19TteO};ILW>%v2(EDc$jKeYXu05m1{Vo;i-|Yjj8Ga z{3P&mX6At{Lm+LBDTnw)4HA_?f(UvX+zu{TN`?Y!j)iGz!h9%^onTfm^~9yvLD?JR z&a80D-OXdR#}32b!F@923+6fr@VC>AjC*zs?KV|0#@(-izrS31Ah4LZNs4kaQDwK3Dl$AanFTFi?Jy|xa8un#x~z8P?q zAbo}?$cLbz5N5{QOmk`y{tWVP^jtw9N9rOex5BE)qDwt`|gRA0* zGPmGhN?70ts**ZIaz#GeM<9NM9L1)20cQ)LA-t{36{XjJELtvQkO{DBIYAD5j8Uh5 z6!PfBYXU9?c)oD_7rzy+A^a?tOB}iHX@qCsiV9ZCCb;Da&xe?4Yh`9tH5ZETmoG?oj%1q81#@DJ!b;8d2xV8t2375(V%zUuv2n%jdEiX7qHBp8QXve$ zH+d!a&I#D2m2;Kie(QgOd599dzz>mQoH|7>{yap7o`1G|&p+GI_PS!UjA!oMqJF6? zJl-t;n;P<|k=G*ZnnvR2bWiV^tK2Wtt&%)A2T|Nw&eC)UaFvdVxYCdnFRLb|PrHxU z3M$O=>tbW@q46`2F3G8aNeopn())E&Xcg z-KBSy-i5<=Ke>$m!qEr#0)o>%fRlY;=|*g`wIb3IU^X;Fq-9z(@b;vOD+QF$?*`l! zPJ*zadmaukXPLx(fLg^LPmAQKHTQMW%N5}%Um=4wyFi@+#78IDyf;@R!xWre0lN?f z$05OjPk>j>lVfoCSv7vaa+JC*Ee8H<5Tf2NGucMHvUU*Bc)Fq`*T=n&wDa>NKCZQLqVT!C#!y`g3yRLLl6&Q#J1;e%%CjdgFt;))_`CbI7~ zKDIDECJ%K*RcZzK9>RzY3ztzWCy=QierNR2hV@; z;ma#Oy|jGkkIRoevwY=O%UAxea_x^RzrA+*soyU@_QU1NKe_w24_BUkwa1+6?N>g! z^Zbo_zy8zh8}Gz6&6E_^acPBz%#&$UCNmN1lKMbORdTtOEPCsf(8`bf0@qqHQ{n4&Y|yz;BxB!yXwH5Njyr*pWnRN746aZ+T>`8 ztddS|MtgLW+vN=~@_nbr`k0wIN_g{H^T4Xk!LB34_MTwpB(P$A`(tfn^La`JIRtE82zv9D~= zST#PZlc0MIk@y)F*aTv2HF4!)pT(wo_rnM{n<*c0L*3z%H8%i+n;ih*IxJ@;vtJBp zU8C^!ZmIqzS*Dt)7`e@afoEM5%ytlfKIIDsL5|6^E(}uUPT<#sE{y*A!F?idF#$fw zAqPG9ZUx_^)?EBJv!lF4>|zL59vpP4InUNpBi?8+IpzJtzg;(NanmV*Z z;ZbaBj8qFE2rgX6>7nQ)L<1lnM>B+OK>pBdh{#nbFvRWWKf3qWrQ5%}eDCK!UB2|@ z?H|2%=i05_GYS^b{`~gxmG3WKxxRem|HL-Q^5u7zufXp=zkK=En_3w^5mO=NhD~Yy`rVE4I9-CqM6A^ zDJe5DP_=KWf~efIJG6jKxqfL?E%)CKlT#-~2FUTnzg&cfi5xg^aN^X+k;#(- zW3^`YJPmXgPV4#zljN&-Q*rRZTIU}@tOcFHc0K?gPt8&kLGz@6h>8}^#M zGlPBWP1tvayG34Wh1Lz3&cbZ~_dzS{5y^ zGRN)&N0oSs@g7sSZ=AwOdJ(}?HVd6h>}A>w#d{qwAUSg%pR*S8Til`KpjF}tqQ5R7 z6~QR5aMR5okH~S~B%CkhJ}Csm<`DKbR?{fA*}xIF?2FggcM9w~_JJevI(%n@+FUty zdTgQ7iQ=**w{l}l*`f+^dWVf0>rmmz~boDm$0dOlo`daVYG;DJ00GxI+V|i$!2^)RSIJa~ahtZA0q7f@Yzd`s z!pUM=5JSbL7f zxC~qXi$B*0KCV>DH`cEG1{3rhW>gIXlx!lEM-?-Vc%nebdAQGAa!~&GkKt3=1Efyz zRbtfkOV$7Eg?dZJ5sfv?KFAxqxx-TyXEm1_m)|y+q+vOVM&e*ZOdJfP`7v9-<9j-B zsNp6d6k|Kr{x(of&TN5A?sdDADH^6*O`Jq#$tt;)3i6(et@EGX`W+{q-~6pZ{{GzO zw|?1;2-|3+J4QrEf`LGhOmZr@5h}T_%5aW8b1?M9zO*-h#VzO2M|;XNau`0(Xd(Uz z8ASF6HgsybVT@3%IBKQXKzxW1T~ViR2zGQ{7A|Y96U%{23sMnAEVv%DvJt1tZm9-q z|3Vw{952AW(mzgJ+9nz8n;2=%VASx*nM+RaxaxG!l{!nv-q2^H`3Sfi5TF^W86l_2lU%zV9p3T=HETT3WXzb=pq@m}I&(D1^z@Zop2&A{Dq zgp();NO6wqU~%<~<98{?gQCdgat#w}Y6#CA*~DojZ6szOlOX$gDGQ-#mZ(&bSN9M^ zy5Ok`Rj^}$yRaH7#V~I8et{#NsQNY|(}BFWv5kyr zX$Z7$G-~4bFq(~xlwziu3$u~^E7*9tM9XBZ^!d%-pyu2CwE3NclX}A^LvU(c+5Zwi zHw=Vsm@TnL9ox&{CJdFkCxo>;mGkKS!LG0BJ(6NC-Eem*Qy--r+* z1_&m0*7cT2<&A6V!#1DwMT->6G|s{ho7c2Hc5^K~)e%`en9s8+nH03pri^~_SaTV@ z!BNyGq^C1VR#(y~*mqK!a%o^2%#jlZZ^t1CoUo2Me${1?2G$X^x+Q{XByR?$O=V<@+h>qj*^-aQcdWSRlOrKp~mUL+jiT<;=A{qxkPw(uy5(^!jdl*#QK1*8Nd z3xt>Pjo(;(^c{=V=qm6TIBO>;!^4Lj8g$@gduE1Hc)n7q+N;mIe#sRTQJ9&*S1g-) z9(Zhg>8Rtuol<*VxCJ4fJi#iwAS$Q>F*5@%%+fNq6-8+`W7Um2Dbk2sf)Kh|#&>h9 zj+ZHHSK!$aP~)EEHn_nuj)}m-(2!gv6bEFY2Lh_8q+0rXdpz%i7)Ds!b$1R&XG*9; zOR}@)?GjkjOM=(#o?SE**2}(B?s~f*E!GaAPEABBq_pN~m3=DdrCV?yAKd)pv06Pq zRzRM>r=DASTb2htil0d|b4e#C4rr!gCJ#N48f4c8k{yvjgR77q#6K14(xrz_&@y!? z`KI_Tr$K_6Csqw6@bk%Zi3tcc*3t6B71K450Yps>8@}(nonK;3w@}<>$j_OP9i5IG zNs-6$+V*uI+-9{}OMwvx;ORx2+v~m3=a5^4yK96xoY_h8uy*_`YG`m(K8|P=T}i8n zMK~Qn90vK9FNiFJ5fS)U^zx;2JL=g@*BP7oSUVGeTVSrNY8<&xp-!QujrE{1r99k+=6iU3*%_Y1V#vdNnbns) z(8ZcVGnd2ZdxSwupO6kaZlA$VDBl7G- zn{}n8(`i{_%NV`O7Ca_=f7a!&`0FJaa)R-^!{Zlr4IKPV!GBpr(?i4GAu5l^rZJRU zOI{QU20~p1%Oojd-zYeC=$x9tQu+qY?@ zLQhS=4H)WsHf990h%*9Ug0zxLF4cJ7`A95mP)13a zwoIOFox+QjN~w@~WU~X>D13vQW)2VJa>n4mur{a;=-JVcfs8gdv=H%q%?dG5KmPtw zkU2FEo7Ft8P?s0lA)fxl-P9r`lWCNgGZAad>4B7{s7B%}alc8v<4y81JtyKb(r?@% zPc>#P<>QFAk)fEm=D2OoO`R`Zst9+kTeIekfgm;FKu{XdO05Fgdw0*S=QSSu&dy~u z=>b($(%Rkz$1h$Lq`dg5Ace(Olgue}10D~C&>_8SHbljb$|u9TSqA$VmTv-Zy&^zv z8IQm(E?!%Fm$eU}yu-)D;%@{wgVYA(H0AS@s0(eL!$J}Kv0AZca&SS0X&%0f5jxL3 z;myo+FQeR+5NFJlgwY3w8Fd%0#xMJvf;a$$?JanD@v6%Ofnb6bLbad%(D?JZmTk7< zeE{)u^6}^OQO_i+*l_Lzte^Ne^tBrKctPIt2P*ajm(4B|&wWGJvyVQyRl!F0EQn!w z!-BH%aoUCJI7_X=9(4*e&EuFv*!2$x`_W9OjVKSIEn!VlvS#8)vh>07uj7YzwbSdU^l;SoEa9?hKDK#|PmP8+ zyJA)f^va*BSrQ{THy)$vfW~XN+S*;mD&fbHmYhALMhKjs;&wc z3VGspPRT2c00=F!8BV9EAm8uRB0#?i{>u_Yhp$Oh4hiZO$S{`QR_DMHHx5xItO|jZ z>>I*CzdX24+J-g47$k1QS~#@{d|C-Mxnle-es_C^aEoNYtG2Kqi|Dt=kvR{GQ0Iuv zi`*9H5G-?ol85^yMHS9;5YHSr8!b4P5*BAZP6Co)>d*pxxQ`Ts1Dn2$XcJb%8_OnT zt|+}Wsi)<#DLAPza}#tqj)5QSM?U+-YXUCDFc@x=A&dc=rI8q*h6M2Nh!zoZLVV;H z-e3}UcbOy(ne{|tST(l2PEwbaLMGBe-dT(8o()YgT5vRRu$|-FLxw@rS%6gsA%WdJ zM)+-Uj1Qv9w#Qgr2%qQ~J$~tzx|OipURX9PvKdUp*k|HsPQt)%^CES-CK#N^i)9rv zm$+{pp`R1#qD(`2vmo0p-|+k(zrf3=9DzK`->}8`t*do(t+6t_E@i``fvxSXB%tw6 zBPJ%JB@RjVP1fLCcxL9& zM`h}KoBYyXCSuF2z@bONPso`!F#xh8^W9o@9 zO>kQbkRvun_v{pKj%1YC@Ox)Cv-_C?e@(4HKNT`0T$jRInA1>46q`{;RM1)0#Sc#& znI1lM`dgDn$Hu1@+CW0Fs3Qj}2m^ziFF}4qM2*kqJj+W*9XK|hV|Eq2aUNbh?~wDt zog@8(JXm)I^*?rIba>$m_X28<^4@X>HRBMpPW)SLV@bvKtb$qxR>2Gs_2|(BklU4W zELtO0xZ`86}-f%yzezRZ7N~ z>`QE^LP;*<^K#Lho|aNFdg8rpnih+)AMaM+x51HsN&079#_Gu3FfD^Z$kAB{V{qzP zzE~$1tSJFtUg@l*mUaF@hh6ZAoc1LJTJ_{i&Ur|zG3cqE-mV{@LAmZwRDFQuy&%LVHBTHFtAcdz@t30KE5^@3*3mjz%>KnoWxx5_pkSbE9?h0pJiR+ zvP$B9wX~9L;j)UnQXfR&paul3$Er0)ki!}|17(Z>x7mXZN9ZG*ZfuK8G0SB>zM!Vm zO+_mRo=LZ~c{j>O`z98z+O$|}CaqQJ?Be20hZJ3Eap4J4qWQ%)D5(U@lT-)ekQ72j z+QwXcM_TE6rz*9pw)1n`6Q_@i^&8|gICMs&v~x3A@5Er=dLzLn)GtrnP+3U#HB|0a z)KrM6x4%8tY6q&-jD*y*WN^F?rzu#j1P<;{c2IWs1_Sm-M_NpB)F}vud3-B`yJKsr zPfZMxlMH#_iAs(OpCSooagN6yEkYI$uni8_=QhWpoIya1n%>lSeTiGND9DhFB-i8g zJmlDFPB)N1#IaeG48yJjpIxpPgyT>ryd8!2qsSYQ zo>OvWGomvNUED`YZ!LZF$>jmE^bS0JXX(Z#mzO?5>D5~K2GX5{L2&kh9P~=OatLf{ zXb<$5D(jB1$?1(LEj{9pRt8B(j7;L;`jaArS(DUD4?Q#xHR7F_!GvHfk-$<{Q>c$w zK_eN!E?o`$Has=*wNZ2ThnF2Oz!MC9sG7Rjz+Y93kR+l!h-i9U$+WaQVzF5Byv+qf zGL=%596W8KWii2H1bO4@y++9JWOjF2*xV~U(0ji4XwiZDJMhRv)$`|&G$_HWQi)Bn z!fEE5njH~R%V$&NX(OYgTJ(G)TkdE*0S7-Tbpkxm#V*zM*kn?`wtp(9&((Yj?lc5N zLJ)w@O^YBtJVQ&TQV_wY+Q5*ijSP$$T6!Rz9We(o!$YZoK~+N?mEn*Kvl<%GSEEkp zXHc(5OYKWBH~NePUh$~CgJR4Xg40t%K8PIjH%%hJI#1t~ew&lnh zE$2fW#nq0_0C+~>Q0{@R6tiU1UA$gcyzX)fqFhno&USTmx@1~;HVjYZVD0u~i#?9pb@atzLvd@~SE;!(@B zPuN89D~n)^oW2>p!pXp{QAc8>9}9dvYLG$Sr*8kC3oFB`x|}AqOY;YXJYFuu^pK0# z+sR7YLN&`L4dt`wt_U9^;3BOFatyt?zg4y-4>3of-53dpV201HSy8v%5l)2+W&?uX39=tBe{UpcUadu81 zu);wZWKPM@*-^>0mluVq94mr%l~AY_pQ9f0VB^N$%u{Qw=prt9;A^9!!^64U*Wn%b zMipNDce3ttcbc;=LT9!Q4-`zlO2zsxRJq;r0tZT6yYvxJYS)tCU8>GG$gm zPm=50?q%fPIHNW)4Lb!g%YBD=-j;#@)m%77$VJZ*U;{|C6rwi{OR@a*E#b~JhojP{ zy{j%gLgyjnonkI?1vw}x%%ub?cg$_$)%#99mqIm0U(Kw2+nJN)Uj3 zj5dwgy_UsM+ol(|M?5R=cQS|5!5waD9%PL*0l_Gek{)$^S;b|R1nGswFi(ST5s<(7 z$Ve;uTf(uc>GRZ+zHB57xSdqKMLC+iq zv6y5O)y-M_=H3wX{pz~Apg)X&SWZ2&Auu--GrO?< zu(_1h^yV)k)r_uX;Mq6td>@|u$wLo~4@{m} zSb6s6%a@=5xbndZvG0O$H>eiFG?58%9s?1I&AaCI zwF)PwJI8D?6?l}Pf9R=Yy#C9R6V#gv+}b;~zds13Oe?uf$(kLZF8F0ciOXn@6@oSL zF2bbF8N+IN-SPL>N;}Con5xYE_{}B(dyn+E4Fjbc21-XaP)^C|o53edvhO0Q8mxf( zEzv{yNIh&NR(baWC-Z+@Fo!phfgV3n}&MjVd7q6E&1Y0>yp8BVB4E zLXnT7m_Ij=QWbrlWMn-N7Y^IwFzrqfape#}k_bk%f;cS&C z@Wt0$)?=Y$hcoFyG#M(HMl6_uX)|rgX?wL?GaDeibuhbA1)CgVAk^fVfupuS{UvhL zkt`k<)Kfb&fmj1j_t=JY0y`{X7V{KWAmWI_9=$qu-CUTB4W5RpoS zBc0;p9mCXhz|bac@&GP94a%z#$q+t{ZG;o-q9Y1nNfaYQBw0I%Xp5YZP9XeCKuqzSa%zkyP1r;D#?irz9=7YDI)=iO&+zwu7*t%7?$ z`|B6)e{bd4AKZE8vH10QS_RKVvRO4WUmnVUqu%s%{;#zPX zj?N>8=ZGgUnr#zVM(_0E)gqE8S5Uka31KVLrL`9Kk;w&C^f8@;62yq8kc)AHu_h(BodCN1Wu4MPoQ&QGf9&3V_TqQnW>$#-~gC z*P?k^NPA6-D!m3~*OY8aNo!yDB(|~jlx~n=xX&3Z#$x<++*$iX4}kz9TQ=PWPuc~N zmh3uWli?6a3n>f96)@X9TFvb-Udvf^o316?;Evki&LLfknoGQ0N|X`?0uT2M&d$O& zW8aw&0bJ(qF#kK#!h$dQ9r(?_t#K$zs=gA1aDgFTN1ec9))nP7MrE1BCnZ#d&$TFA zrvb8Eyvn;h${TU%YzR#(4+m-{!p>XGMDdG58!{=YA->6VkSQb6=`J)hC8Z`V#eyFh zfsFxz9D+0fJ;>#>T6fJe&+qq`Emfe3+_edU+Y{it$E#xQyl_(Kux;AYh^QTCL~;mb zN~aT#R@e9i$xfX5zH*NnVcLzE}yk`7lakyr`F&z zJKcEfT8G$G$b@j_q@zKnGs0a6UcgbsZ;K2%ehE$|Wbt(SA(3>48Ic@nG^JZccDA=( zt11vOW>najjD^tS7L;S?Z5@iWBW_SA@Z#>5gKg@S`K(fcZ4!ebo`CyxeX+LDuZ-NT zG%^sjC!X&Jg$7_;3vAyK8Yst3kKKL!vn34z7+;Z+zbRM*Ja>tbF=r8uUl3-FV`PuOOaUX%H1!_YEa zv??z3XaXMx*BF@!Hwgzaa}d8Yht_-Dvx$73TP2&h2?!Ls}olndF;5qWKhe5xPL zJe9*zZ!J@g`tAkh$e7}&PtH?sEd>cA=Iy1}*PjHJl-u5+y zy5t=iS9S4?*!LaUs8G+=*7e`})9LorqjbmHF0B~meozxqi?L~$Q{qskOH&0QPT`1@ z!eADyohZp(2F7}<4F1}gwF8zw)vinESTu~L7}_R^Y_+25zb>g}R3)8JG(%TZ)mT_} z)Ld3w|JK#QYAxX#A6pn7lZWss+)~>_641TmofgPJd_6uP(^W9_iZeTjE);Z(uE8^B?A2f@m}GK)qKf=bdT7_F05vMue%*{ImePv83VwNI~o`iBAX=^LNk`0Nt-?4$pF=Ch|hd*ahS zefH$1*Z%t%^4a(P`%$YRJw373(XuGc zl^plkF)9wlwt906-2B=ui+JJJ7FM5KduC?lB;+%vPUB~fx^zD(uIr4ozwOM!Sa=-dZB*a5ox^g{ipB!^0!}n^vb zZSJc?H!@E$K^UkWZPEIhwS!j6%FV~E782~_4WZkI&bl3UE@HU6YaRt?j9ec9&$)XvB_2#L!WdyhW|@YBBo# zM?9YvxByo$>#~9%L(HYfpS?lNdT_F(W$7=^YE3P#0hRVKz&l=y;OWH=Fw`QU?aftLb!1h2BxXeO{^Mr0TK6sySS7$tsED1&we)H)mZEQe zx!W&4rK-1o@w=TaU2S@K-Fx-Yy;pv{^2Q4*w|;&1wIAL6}| z>rclv(nQWknW+%#n-*%=YEc`uDxxS_ImRk>bQ7Je<{>0u2!NmS>73nz>{7>&SN5GU zgXk(g_rnPa+d?s5o9nNt*g4M_i3fQHU8}@pZ&f1&?pc9{&itS!$NIt$mE%Ya><*WB zDWLHG!?j;zMAIZTrMBjGv%l*YXEdE%d}E69%(bWow=*x6DVCCs9zEJHZCqVnv};z$ zWj2!nOtE6%6et-%;%ABa9V}Cy1wA+c)OsofMjmv7VnE|(0LKFASS6gb%L|9Oq5fqf zua2Bn>#CAz>9WL0t0$^{Nk9-(1*=TW7H7TUe7gkgXsd*%jIMS?ayICBQpz+gf(SfU zRxwp2B*_$Vfm-CHqMHz*CzgK!Biy_8+}#iUu=46V%a`7~{j+ygF8{I@btrE?|IwWf z-@N6`(nTBtdQ#Erz(-qUmWR2_s>|7O_R=*+D4Ih0J-fR1LdU4_C zy!iO)9RK)jGAAl__3uCA6epk;K7N(cz=r1+jtPh3TSVg#ap^hXP;d3ODPLG9!}s+T z9(w4%&VXAdRGFE90~C6m#0dnBLfumc8-|3e2VUtjH>(z>V z51m0K36Gy!Cz@`CRnyJLH;@5IUJJY+dSrut5blJKhL_7`dAZ_>dA87aDDCCbA_hG5 z1hv;nDC^BQ&ijWrES*Va!Fq|0>^#d0Y*vISCkfyODzccQGMkxJ9;{Rtg$%*!!Ng$n z=$RpM27kSpr?yQdm~+4*4{B;%n01X_tHk58jtsh-*`8g+B0LY#QDrl!1}$gpAfl;E zMQfS8+%IrNE^yH;Eo>!79Sn%7a*?&|n+^mVJ7d!e^@qLZ z7e|?Mo=eI*JASqfXFWT~Xd%q65hC6)B5!IC?&dO^(Kj1|7x3o7UfXB@=QAWczl%&e z+cbZG{$U!`vd9pvl1WljKpxxn&q#wxqs z4jp~nV7qqH;ghW724t!X+wmL@u@ra^Kiz}VKuO*~%MYj|1ZTzF~+_;1+P?#axxh&MK-3$+ebIvY~;lT&h87Cm%<# zqBfAylyoYwQ1<=SR{!44-zx0*PJWfKQX~ z&NJMx;*)t~bVbb}bR?zdnUI@nIJwBC$SLMtSCO~pIVC0G`dBa8DW)3@Njpq6@mTP* z%i%QIt%uze6f_n3$Pn|%M3tQ2v+;qy--0OD zfeEd^_5+t~EwCor*#G6@XzNPT6isiA&*O|AA!AXEvZy97A*ClMnZ+jSEE2izbXhR= z?lTODDiaR%1sQUw$8E-4iPs(te3lnBK9o0bKbcy*?s7&(;x{)T%<#@rdmRy|mJT62 zA{)?v_w&t8LMA1|nxgO~H>)LBnj=(2Q=w zOdQHk@TpFrUP%fq#FMJ7Z_o}M8`>!5I}X`(L6moe?N+JY9XXtzhwCf_HrLL!*=|Qp zjx`(Zh*Pv7O>9^x3!7I8+s^P`7)_;y2Wg{D(t@FKijlnTw5Rt%mwtCk_rHF#eRN}X z3~qD`DN2M&4w0kc ze4B*0?`TW8qW;FZmCTfqRTWLs6)i;Cmuqw>a}m3g8mj5!qHl7|>&^o&;xfC=saW7H zvmm(y>!8$R#bvdcQ|&`~hp}mj6;SlS0#U%mL&SFe2V z)zj~P_2-xW=e-+WKK*?CDLjDpefh*szIx>y^5tJ1`|@}1;HzI=d#?Yh7k>NY?_MBZ zJ@vv@ul}r^Wp_2sZteWq>|JuCY~9{9{qYaysvrL_6yT>IHbDA#v}#x_$ZW6HU~Vkd zfF{i<*_NvAh9+%Zc3-~!-dB&mEba9E<6ph>SdSI5zIx%(|GfLgm#@B0zWUkYU;gl& z*h}ln%P)NO;(N7!LzYn6hkR2+gAcr-9A zD&(LqgdIPl8gi=^D+R?83)?ItV-vCYr6VbXsci;~HhmtlkPZ03`c99x1aYTJ+~$=E z@zvk!=GRRZj)^ zqc-(Qelkfp_E{Gp)BA-N`Q!!g~j^ixZhan=UNY!z;y-y0az&xKl-1ndV_Mg2F+lTfXdFnb#-J#PbDAX+RG89>=ZDM z5exFn^43o0WgA`RY{ag!E+^xYEBWqi)6Q?LJq@)H;K;|(ENY}6X>XcfMhMC0|6-*6klR;}_rn^4M*h((bhO>+8>x|51W|8-_uVd#qWZLz?wQXt0 z^y?e2J?ov=o^>2*PPGhU?yc6RkG)Sy%G9QL>T)ZI^kg?CT6COY&hBw3?Yc3TTlGt> zs8ovym~6Ucj;xYv3HJ7+Ir7L^!goqU+dz%^Y&vm7P-nsz!7uR5fEM8UEQAkcX2^qs zgAJ{^ye+Ib@Yi?}o}4kRzi^&%iAHqUcLV16-ZqsTw`TvsYa-t)w`p1^a|iY-m-=|* zBT5+u%xkubXZ7&44VRwU7oIxWJ)>mxmJlRrSsz)sX+aiIZR(+F1a;x()QQP$3Ou?M z3Vxe8+{eQGRXl_EmHT1p+QKVvPZZrSy_c2=(R;~*Xv-1h(pw7n(I+oU%pve+4nQEp z5OeK;o!}(}$(5XatQ(72@^Igv-9P9Qs2z9@_aWie!+kXT47*@KpozmF z-wlM%yh9NE+YB~!cP^;ZhyoxHAd+UT*HjG}r=H02a2e+|w+z*^qMM0@K#}=I%!9KL z;7te5BTYpw)O^?5Ae-?^WYD4oR<71hB7QlZy9Lf1*g3MH@vsCOZD0}9KuE0Lci)62 z%(f|{P}Loa*$r%v+FR_!NuEC?03V+j|Bh{$>27TOqWcZ_$Ls@fT} z;|0`~azL=>`X#omdV^6RwdUIRy}mOob6;n;3p;t3q$ZbD)HJv)xbT_GCT#~FlS&+a zz+b~y;E<#awR(+yZO9e0aGqAl5Y6)hLqo_E#fu>l2rX2j)uI+@4QxEN4h_ZATbvgb zBXB)e@x1)v#aiLHcXpvcyTZsYFXt2R1A$9H%Kpu7LAs$BErboRn9S)pNcwC>i$M)k z)2lV}<1Y8ekXxlr{H%2dHWlWE*@4W4Ym$#6S|g*F$#d3829dJQ3&N8?vsI;23p9U>8DrlhJ-s&**#uz6ok)LJ#tNJkKdLEP;Navpx05&Vh-!Vvj3UsSlY z>+z&*#!`mG>?~s?_!jYzXyZXK91_N!Pd2bwoYI=jxu*&e$LSoc#Mg;!Q-{blI=R`1 zI2JGfieonKIdWc~(c3N$soVKASh9KT=(+L%qVo%Z8x#0w` zh^l1L&H2E7p1qxUb_28)5wT{X)V8W8?ihpBj$Aw?_1Geo5qnLUj!a^iNL)y&?>5G978mYWYKPboHMJ4H$j8x^u&#mQP0ZUeO^Xs) zD%=YEHk=k*=KfiiF~ZA8RTWm%QV2OZTk8hYUW5k@WV$u|DHNM^*7Rn%Bdb*oyWqoB znumC(K&##%!hKlw5cElE2YGs1r6}CB)j3)V##YHf;Amh%>aCxLdmyMN=PL0Nn4$GM zuF9mDnlmx3Izt+@Do7o`)9E#7oEtdp}MfXy(( zUP~-e!PD_VKUa&SWyf7|&_<$~QF>TL2dx8cov%x9B zmxC7$k<-j^ysAAxP7q3A;Q_h3UtlH}X zs|Fu9D&UVoSUbUG1q-E}dW_7j%^AeDBxLDq{r1_&Orj?yW)4i5n0qimaeaM#aiQtB zrMTYo_O;?VvVD-nn>L&m#n`Z}g(4Z}6b)iC6}j$Z0kenJ_r{gnxMlX(TWA`Yl(U=Y zpy3>3p_YoS^{DGY*K1uUNXexK_nO{!_=G4PmbrVV0jGWK!6}l;k|TBv{5CVA)O#w^ zXBVc=_JSyO6U2Aj>BNtc=zLh_S~o(j+lc79(Rn(n#i<+Z2j;pPqCxd#wYBxlK;bvf0@LnEr0jlcwc> zdUDB-nl#@dkA!SJeCp@|{5LZL@ieAG>t3V<4A~ZHKjd3bl1vtc>s8^N-ebpk{}87h zEDcK}8FN3&3v5<|wpp{$i@}ExS9VTGw+O9o2ng#N22Yed?*(zx-2dYiMdpNh>(_rD-87f~Fd| zP#4CCY!{SGZ6Tc94}Ky9KWNvj75r8+Mq_lKb`a6186}t5%rMl1oGPXc1P3Us*#i*{ zCdhs3u(AWEZF-T;9rnQC=SBSS!P>8fYw;m=ooqy?LK^#=MypQs1 zlxwR8>@n!D3!&*jc{tD@%9p1B7f{mWlPP#sGI0&N$ng}vhBxRG16X1jAFJV4sq=DZQ)R&rh36hO)>_;b(kQdtRQ3Dp@dQe)Gb=)!#h)GUy8bD4A|sa zThZp|9B^?&jGjGi5{(3x@l%3^(Rk z3Tw_q%_~n?jnJjtJii?=%gJMX>+RGB-<2GL+kqiDaNS<%oP=I$H~?VgHfOW6l1(Mn zz}j?QV}h4O-e!)&1Gp@3jVpjg3yD3muM7OcFUHF>*?(g>5XRK;YS-!_y1X z)r!Dc2yER5a16cT!z*$e3pu-IW(H7%-NR9e01=OL3B&(xO9bb;aFfvbZ1cF~g zHqL|Ppa4G)5s#JS`jRmT-a= z(x$jPy3@=mIoyH(Le&qe!-ET_95!roYpy<4=+4Z*jwBxzLQTvOK0A9jolT`R&CF&N z8o>#~@IHfpM22947V>x^?1Gu8jq2C7Q82ctX;^BA?tl~N^)F%8)TFJCl8)9Tk5U)f0gAh&+awoeN8CZ@9 z0{=f-f?G7UC8k)Odb5F_7j`zA4b@p2>8lxW=rdg-Ip446CCi-Vu^=JJ-14@zu4jbw zXln;ijGI9`Lb`>DYR{4MP0yd2ieDk>4m)24C)=~h)ok|m0`olLLHGcljelT^wP)i_ zm1PhJk~W%4wcD5Mo~rmV3@dZlWERD0&!xHbFAhQ4;-BT&bRolykKxM010?@-7sVX@&X zzYCWnmFGgyT8A*^3j`j|M0ilSD%W%<%xj077g>JHWq8-wVG2hW@2vPVbmx%$6dl! znOr!M7L{S67wvfP`zHRYr9n2qE^te1OK0p~+2}z>#v%!D;F54Cn^Dr~^$jnh!L5W3 zquE5)6*D>g`kqek!zqP6A(aX3R(11^_&f$jHTm2z0nTK9Q~ zR=s~b_Iw>>36D++-(_S3zi#Zv#j-LPU2j+@-9RFyBjOM(p z7CDahqmCp|htCB3AyM4gf^zK6XlB6FluV1_aO8f?u*+Z(2}e>6*|l>Mw%E~qD)#al z<?PR7!i33>yB!}-L0Ic>D@nM)N#2Eu9F|;j?dirngTJQMWkEsS20BG z7~%g_tOB+YHV;!a=nasud>qmKX~k%_iEq<1$*l{OPRBqQYo09pdgM4ZT^`2WX zXQ_=$R%D1Vw;J0LCm}ow9CpAX(=IPCXLCT_tkhFiHzVY{{WPfeS#ke}%N=qmPvh4t z+fGYwAfB4-AfKAWq4`#f{(kBZC3WD1g3ZVglnrwoQszkg?N&(%LKm^c_>^?0IGv7K zod#o@ijhmaW3F$H-~RcH{js ziulx-!#hw*nOdeFa_eKK#}@Qds&ZjwMp2XlTXZ{iYv6n3sq1&1f9>vL-)r2BYflYK zy17Cr@l2}x?NEOsol7-}t@Wd5_mP%T^wu%QI3vP=?ej_rwPPe@B&;da=-c)C`Szfz z&z4}5EFUchE-RO-b)j^lh#((Fv=tVZHUaZbGOsfsed0eJ`z4M#6kP^R5XW}P^63MO zC-0Nlog~B*qvKdz2k8 zRYf(Lo!%ZpL{O_#b-@WEBXTi87}p(n1h9pv7%hxvR#eQTD9q7|#G&NxKeWLJYz{UN zx4d}OcK-enA`Ecu9WP6;_J$tYQl{8Nc#+IOdLOvX;u}^8&IwM?hf|#=!zDW?2P7cV zAODi&1)OwjJAVA?v5)^0+GS#^A7-&_q-}ym-xeP4;_}&%O2w40<|;v@!f@nlo_Rhm z(ptNDlp((ed{elReqbd4^kI*~VU`tgdPYd&$%r_HmBV>eJ(ajyRHssRiUc-OkLPXH zXc}YK*3xhH4rkMfn#n4vo`oAEwXpuM#y~J%F9*X1;LDKGX@;(3G^JJ!86Fxvy)ap8 z>W`w4Aer~L&*#~H!tYPRY0P$eSjQj#AUo zDHZm1ePCF_2cv`ffS%QI1G+Y@3&Sf@(kgN;4HGqMuPA@Fbd3abR8;lH=D>90j zKHKZ~$F`^0R?;PYeS9nwO(js^%pvD@N=@8P#5P6=vRuoD?=Qf?%yV|W_mEuUs5h0D z3lPsFemTNr)jVgR`Onw@aj*nLQR&1Dr3Who>>nMiHzo4mAc*Waa!-C`~ ztg%^cGrTHDx7B!eL0AEPYOV3K^-bOFvGIiKkqK*zJAUY%%Uw zvF&v{Jr|asNs`*=X~oQKMxgk?GQnbuQ>v8%k#lSD5C7H+F7iPySiFjaHzPG6vf4=0 znps+292gm#JbH>u9XWGk@@V{>Z2d2_0+_Yhe-o+?Ai%-o0j5H3-KeGN>N4BX@)9eDqFUSJZ`BoF1$eS&5hVQ0DTlVKZn zKG-CttdbD%MHMr}!;x;8^6eJc!*}&hSURVy9z|)%qG|2-@ z&>bOxOOWvNfCufEy^S)8>1bh1|I4CIfn%)g9bVlf#4MQ%x&3+IPPs>BA+1*knPf%k7LYO(&07-Z3Eb*cVqr~_R7F6FaC;vj+=sLZ zK~z{cc0^!H00{ThaK?FUn?R2;!ye(zoJ|w7$&S&gD;!-5@2*;V$S^$L@&3pvxm@BP_(Ak?jiRe0cTbra zy;Irv-b=5oy!qz6XRogO?aD&9BRv?Q2Ew{D#&wQv)fa6KHOO&WXtW{GU(1Kkz^A-T3MkuOCjU zaHA;MjH0QTh48`5%(}w;y7&F_iC4e8`s&Qg17E%J0#S`@O3mcb|9s{-^3NZ=_tmu@ zfAz}ahYJ!SFgPD z<+HyeU%mMJSATxv|7Y)Ao7*_9{NbY=NXm-6+u{ZLOoIwwgc_ zXmc=)%!L$WrplIlQEX*5$$H~loQ)kjk*vg)Z%KShwrZKYSKg8Ft5}g_ubujTh3A|e zkZ1xRJu@UlLX=A;X9k1e^qlF_=lc7}0Pg2cJxK~*{Pg+6FTx}Kqtib2!YhlrQBas0 z3-KCw@MS;@fzHL{Pv|^efGkS?)>)hXHWdwItV>erw2MjFfc{={El_K zuN8xAKfEUY{Sx~SpLhd?bn}TP6N^OGEe(M#vsMQVVB6G+JH6marn$*RcP=Uj=-n4U zaXliigUINO?y7**D{rT6TI{$^TP)djQCT)#O&JO%t7W_T2a*7p-udY(pZ@BHcfRxf ztv7yh=lQF*Fa7S$@2`IP(e*n&c`{-Z?gw#g=)eB`gCE>__O)AY{NnaYmk!;2>jLZ~ z7TSp;!^#q+P(B+mIiEpPh6(n{WI$ZzgK(-SC+}P62`{lu4M=|d8W4OP@>;G74>yT@a_2X{`TGZ# zZ@>8STR;Ec_KUyy^y;NM&wqIHiC;%h0=M4y`RyP6;qULg4$J=EKX~r;FJE8BhS_Dn zJ!>U)=lQp9eemL~_b$Ld!1;xZPhWoF){Af6e)q%Mk6*cU`G@M*6ARNYz(&+zmV;em zDvqSBw=EmF9dCl0hE_0BBgBJ5r#$LnQM7O|4(NkSD*do%-X$oX#NFN~5?d=6sm;x( zE3!hOzDHRo0HX0DxMF_=ao=Mz4tjbkMIXCfUc(-p36!iB?iMH6IRdBn3`U$hadJ4I zb=ixzx=1QZWbFA<0FBt0HHqzd&r|fES+J*FWnJNhi5LBb=WdyrnZBcYFV~zSOY{+1 zZ}>ItZ8to-p%6C7otVPS!lKmWEW~ZgYxN3MgbyR}Ng10Ag=A@6FK|5qYZxAOz&oW2 zYxX*;m)D+J;eJ`eDrHI>N)|tGU__Ki=zFA5fc5V-utfx9s$TUEd>iIn3vp6b4_O73D+i(In5@Dc4bi zqA-xeR%3Q{Gr4;Z$^=8sYl)!@{+7e4wA_c)iY0Oq+e-;p!I)S*g{UoL|AQf8Fs^jz z)d?Y?!DoH$I7*?mS!mB4xgAqZVBP2Xg+J}og7A=O;@)IVW7viVmfiU+34djKBMrdu z^AK;rXhu|4k zzBxdS;=@TnUDKm=W}Uj1jm2KjGRA#GsfO?a;b1AEgwSz$n)ufi0RVthz`xCOp-%gA zarNTjt5vM93}DHIsWS!!bFmc8V$337fgi1qyMF$Z+_}Y9kwKU#a#}rBJjuPf@DELb zx#leXGKD`7T|mjhD)6!KdJIueSk3OJ`(`Od(>x-4QKQI7=? zOW2x59fWYAcz_coQA7DUiT0xk3To?`M*}K7>m6tFI}w7kU@#s=*t!KWP&zrVRv-vS zJjXG7Pb!?0+_G@-d975=7@Cdsf|{rnYw8ZxEalIIR|XesC!I7V~8Tz@buD1q)9S_CK$tE3h=9ZN@1DFDN^}!GxDxE`1dO_Ll!dyKa z)Ok>@F~`|Tz|}%u0~1+nCm!n z@e0LwGR&}Y^m!4pmQ~?3i$}dbzsxZy!u{Ia3d(JfB*&@eEx!G&5Z%6gavX_;IL_alJRW_U7OfO#`8vR#jG1ep}co4waJ;5quP>AGBBu5=F$6)Yq-8Gr? zgwQ`A0zISU(^n+HG%OMX8{soev`Qow)6j~VYQf`0V4I;kjr}rU7!*Li1&Bkk+lwB0 zf!{SN?9d9jiq~U&$t?xA;u=371-lQ*5&54}1H!I_lo|3EUB#1;#W4FGdklTj8lx5F zef!(Ho#=CFjY~Zxz3B)(CWzo;D%@|eonWEBmcqvG@9x+Z2qf4rYf83nQZytki}@BMe9t!A-y`WAJ{->Xvf>vY<^Pu*2@>ah+;+F|AC% zl`6Z#^=Ys9$>aYq5z#xB7{*V+K*lp4a=G+IQ(z??qvuE_o69LW)aXrSVXV()fs)H= zR(G#g+)jsKHof>728CcAx_(H#H4P68#EAtJ5kx08BorsZ2J_0!;>$iS3q+13Gm9St z)50Ott71CH6$?&Sg$tF1=fU#@I}@4YSRWK%bhvL{KGMWOKZRVJ?q;&^22!~i>4Pyb z7EKPTCNIOP1J)ks;>RAFqUU6h+d`|b?dvOB|5~O|2{9glLAzm{APINLg^!)HXF<0u zkVdWPtQ_}(8nPf)&Ys0j$`19xA(kH;VuE|3{o=gv-Z zcBgognsq!ICQ*&mDVbq)?s9lm$rB#jsWAtE5fS0vX}?A}r2{S(w*}6?)?K5E6?olT za8EmWPLvq|-Ep3h(1&xk4xj6;)C7~`%u8PPRzfU0F4io&3d9<2#uR|9>0Mne@f&qT zqlru$2UfxNxEtXrI=D$zOr-+^yLHi0f+GiHy?^QSAgQXQ;P*=T)M@J1U?2!oJH$bK zCpV~$)Zv0Ev|?gX1QaZGnWb)Dx)<|K4dFE9){Z5f4B7JA1;ezrrP8eL7?&&x#MTO3 z#YnL+2*0>^iA1P+MY3<-KHU7T?;)Dt7cad2#f9JCp=X~bU%c|s=P$oVqEA1+@FVj1 zbFY5>%lF9tzW3%oe)Qhwzx;^&{U;OaBFMjwg`N#KP{Kt=;#LImC=8K;{ z_ZtKv{Ng{K{Ng8Xe*VjMKY#7b{}*)W{}*9&PHAMAx|})w0rE{*ZaEo*6$$5i+T$|i zHQDq7R`e9VbopGRhE^Fq#<)tsbdC|OA>_VQ_G zG*eGqklM-~A8a7Wyu>2L#dIY=hQ6$BC?z{GRr}RL#AWjLpf&ErS9^3CHl?(jwf z7GQselFmJ1ONVKIa6q(5|DVJXL}WZYlSkuG%jXpvD5sDu?ixMTlp`^+^x=uyL*uQkmkTG@+oa|jukulD^ zI&~QWOs4`Wz`6@Po7eQ5q2*N_r<8%3G0Q*=Y#+_$V><`~N2mvG=5**qum4&ANnhXv z))C4>lU+h-0m)5DK<~RE8#|lbaF&QVBcpVVk&@G+`FdcyG_Y1MAWYZwo6a08(n_6f z0So?g_}_-*L(yJ&1(z@%GaqOcGnSUkrOpTL<)Y-%SPbri4Pa?>7iGz$mI? z0XAIFEz=;y%CuE}f?0&wn#=C-Vxk&zMVXgt&Q7&Vn+RiEFjHrVPE&aDLXdh`u3DEc>&49G3wFg3{&H`Db$o+(CjX<$8~`;2d2bqb8Q=sM96;U z=7sCM%z<4Ek9qUr&u?CQ1#kZ1f8D(JQ4dQ%H#;qXsXu>rw)yAp4k&h-A?>t6K~rFE zbGi~ok1Gnp(5$Xf$F0oB;A+~6%gQ!-ZxTyh7xb@_HJ_0|j!b6YOOORe_9T403Crq& z%(Hym55j4(K*K;pM}4eH^bgJpZ}wo-6JaAdjG&Ix|N8LaYd-fDuO1-B1X_(U5v2aD zBa{a`y7W$GWYQH^MB`c+XW~gbN$2L8kJ9q2p;T8JmQpNh#jKSzu_w<=-&l)r3^5bT zBf`X250Mqjn6JvLS^ziT2k=tWBOJP+8ZXnTkhGq|(ajwKVH|WtK(R6#ktx2MVtMU# z`%QhfDxu&BHZXeUV`L7|gIdpq$ZN?qU1a1+gHdu4rdEwc75#c9lTniVrrEnt%G~e; zK@_3XFG$K>dlO_Xf~eEpJSeyCktEU4Q^Uz^s4Z=#H!=5`?@Hsx4t>L%|NQMIzj*dl z^7&KmeevW=eSkT*Rs{d}_wW4U;?KT#;Wz*L=Fbw#;-98zO`*!R72N8W)oZ5Y=SdoWDrc!dmkbb2~@WkgRFh~6YsaD2m;Ag@{GeI%I_G1GAxxO7`Xre?P7 z1Iz7r^V-g6g=}m%y@APw2lHx%GpDj9DlvlQCqZNM|Bz@Kxw-4KR|)a%NLpmEDro%~ zcWO2e@QTi`UJdaBM^AJHDvn|Uy;#%9o5^e0oTg{se?y5?q%b;dTeFN#qi2uBQJGgq z#2WBu@eSmTBf!|%skEz!89p=LiF|kWOq(g@d9(p*9uXeG07A{5r@{U?(Pc;NYrrrS9!r;t?_+Li2u zn1$>XHiYWwXWTsX3HON~&dl&KkJy)Xw-3Ko3MUz>lj9UFz{wz;POkQj&_<(qnj$+r z^HlJ%gX;tmZ2;5BaNy5TMPAq7WLHLSA|J7r@kE&3!(W;(?1t zU<8aJvkE&PIpl*H_xi2Dy#)sI@E~~@uhE>Qjzh+n3!{9<&|B;&8^gnViJ(g!7OOf< zY|-ON1TC>CX<_~`o7s#XKNNyg-GEF8)E@oU5kKfMPL43|yu+Kxr`#a*XIX#}o}sR; z=s#USwU*oQ#9^44)n!H|;Z?kO@tKp zUW9)?mFevdQfR{_!m{V{x|+A?&33r~*QJG7n{JsTnIcCx0_GttN3DtVA9!2e9PsO+ zdT@s0z=5G~h)Ua*XX)vXoEu#kFO5T*OJeeMcxSf9y% zoSr9RVAwCd?xx`3IPAEr>WN2%ze!o!npx;9*q~ngiMjmL(r=co!O!=}(v{^Wm!DdG zX8Gx*>r3w}KfC<2Jhb%trK_-myZx4$TN%sJYa|*4ZoCddxrCdEg*F<95Fc|vEwfSD zRbw0ZRz}NNsR=uV4)>Q7#*h)jK}gXR1ciiZk2C4)(gE3N+``UXNu+G(5bQFiG?B2XnK;SZ4REibnCIeh@C^+ zctnL3p%?k>#E~G0(KbfvxkOXecKzGfp}V=u(W&JTZpb)%$%CJmaP}lE6q<>LUFKC6 zUxy86@wIY*rDDkyN+h3Q=5TjbE)pxRU!iy8rm$h?#u7qBt|Et~ zU+sK@%u0p3D~+ycc-FGUeRZ^*RVar~c7e(tcB{x%)viLx=QTZJYq|mdXB1c7Xjcmv zF^h$4Bw@@JQX3j=$_>4PazjG7s-zpoDl-y>4vevTfnGxy&I*MHu@g_aFXP!T)EtBn zI|*T9n`R+>E3{tdZ+L2l&N6abl#^@0EkUNXu|3mD-<}zx4^6@2!@6;$v220Yv7HR_ zuU~h>{0^h-rYwW+!0Gd42auP%PIg_GKC4|#WW}r|0;%S-Y*+8k-VVSauBgM@3UhZZ z24^E3X0DLhiUd2OBvtu3#5KEF)s2N1l8&fJ9j8eK;qr+EVPxSkni;D&s2?6F4ri@Q zad>zrR~#y2hl{3Z6pGoQ5e1$@SK55^xP3l0;^XSw2rBCZ)FQH>IbLC1oQH$-QE(+? zkkrVWGct{H>p`hJ(rqJpVKt=N!FiY>1?z;<h2BM<<^6ebD+>R$eQ`a(frvPJDHk>)LitK;ud*8kF z@2}o^>$%Uq^WG4Q?qTR zn6d2?GpAX4Z0jFhU$4TuoV3XXlT>ptAlat6DZr*=`PdH4$6jb`4hzAo5@34po*Y2~ zR<3g@u0EHxhh~7bvkRJDPzs97Ogm;Jx8n)Pf(hJ3(w3;8=)&7KFMf3MBF3!<`Sd%# zx(Rdc!sVM!T!EQ~MMj9<@WRcDPh;5;oOt2--#`3e^y!^nd~oZHpWM3ibTljbY<0Zz z^LPIK-m{c@#En4GRx5S!z|2RAQXx_RN3O4hs`(~k?TXJfxzE|;3!64$-4 ze>pS>6U|{b=aBq`mzy$emX^ic$Ye76d>IPE&3guWc*Nn2gENk3kXksswV|-&EIBr& z67yZ|;9YE*r4{fYT{1b_Bu9M=CiPp>Y{@&v8;e&RS|xHYuD=!Z97+RpY#Ycndh|ah zBw&P@t49c@y*L+TMV^0X%+pE{Y=vnos5>hqX)BSaH7DYcvu7VDl3~GocV9s27z#l~ z38uA7s;ScpQ>Xi{UJWPZkZ`8q`H_fLGVvQw!tf+J!K%zv8sQJFQy#EUbpX5vSuLAS zT>>@2oJt*LB=?C316cur^_L!tqQ$uFGwKUB3iwbGCGGv12p#nIxcg}6HTwYJV0|)| zBXA>#4PWx%`#$la+H;rAO^b7hB|GU*zcwPy>6>ZFv@ye*vv!V`@d)<>h=Y?qUk?0m zw%=O+I&XhRa=ckgJn6aAfdBVGgl-ZUkwCRLY}yOmkmB!tzS++9E5!{D;c z;_F=w=HWSvajZuxJ#?zer-XGv)Tg(lF)z6^CWcOuBf~}yN3wKss?(D!xv=^p6`2yl z$?KYKn&3)eYn|GkmozVt#f~t&D=9k@x7>P8*UT(Ai0(mgXmX+53`YE>8_;tclpI{jq*CmUul|8U^IXU12s|9LReHuc?EH;G|x^O+bFs7vCpCd}d?#3&BdFIBOe|h3B z7jFFF#(OtDxbYsj@l)7%`^Fz{{NXQ8z=1#Bc%R7K_u;^E3S7$(<}`|e&o@dJE3u>B$KlN+!8<@-0@z46DtJPzM@ z`^I~Jc_#6D09X9ujo->g`tFT4mDoeOc03o8pCOA)X{Om3=-Xq9BCLs&`1ot5Feb8? z12N+BWJsKK&u2l`invm+i`Ay4uVFR2rPj*o%~z*X;ae8rdfqBQSoko~Kai_7tc~E4#x_81V!$ zc_tSDvLt|q0H%FTQGE)9MqCDsEVh{GcIxS{r<<8Vcb3#aNtd^^naQ%+ArZM9Q?6Mp zV|A7TCTX*kjkNE>&!?83T)u>zwO6Bd+okuHpIN%ze}yw^?p}`WmrL)!PaNU1(ScW% zFD<*`0_TP8e%83%S1DSKdE zRa3O$Zj~8tn_Ax5+vt5Di$;xBDVYP80_#6^iVP)-xfsj1&zH7Md}(H`UF|2gV+tj& z6}pz3B*xrv?g+B~RHMc1LY)WtVU_7oQ6vqZaau+sEevng1V&6{dhbwa=rAc$r<@`p z5nTu_uCQL@ZN=)dW14+AY>TrqsnyxNQ`p=9Z-A5l(m*Z09qP^Q92GA}!`2F$3O~e& z4(+a4)7i8Q17{8yYr+#H`;Sn^5%Y|>$z$nL9&1i>rr@iYN(o*}^Ero2QhFwJvj09( z`sQATQ+}CxWCCFnoJR7No}O6g+1_Ene0HPPK({o*+Vd88WcB$0_`WG{eS6q^X1f>1 zjM<9=vdC*zHnn+int~r8r%3PODIsNJBTHiwxwv&*5vrBtK*mY&-s~RUKxz} z)^HHK6f{t~YcZKhQ3`Vve3q`)cC^ozh zNg~R@)K3vb7^lLcilj11`*SAtYGzrQ-BoI^N4>`ebN!|>N8L0~ds~w>$RYJvE}xn# z8bQHNO%XfO4@(R>a%tVO=e>S;1!S+{^cGml+lc6#2VFUzt$7{Y_0a~z)S zg0dau%dB0$jBS`OQ^vC;S+RoB3tfoY44ucmFWu;zaTONdz&OpOSE@+J=*i)`Gj+DC zkc4`(LDgZrTmv&4ekR9^80J`Wp}<8bsf@Upj67yCGMUt&hA+hfE$+kr&Ya1eJqy24 z;g_7I{BaC|Ta7z`E1yUzpC-AEr{i;O2n(!-!M`-|C>UH3t4>s2O54pprxI>qPg1gJswZ;m zH(4>^oATIVX2lOsbt+=R^eTOfi+w~=1Hi1GsEIVBS-rrq4F#V(_AJ4)RkpL9axaFV zm|NlH&;gId1%>8pQ79@!jcXox%*^!dMfajirYV*V_rw6bdr|>?)t%r?P!(~w0W-eB zg%7qJ4J$ApaK!qsCVZGlUy8udN^*f5=d-X91s*mkZ7P6nTS-_=QXBGxtuW*Z_YL_@ zL%y&(40#)D&93z?_FIJLfYTgoMp@qm%Q+r^^}OuyhRm1E!vwd9D*}@1WjXc%CP74# zNJdTZ(W@-x^|{PfaQIP%Ow zbmWuAFSm~&-iO?~wtNA*k6~x&ZF1&JbZQay-hvO`S-M8#=9BW=>&ur2ZbqM-JxkzP z&-DC~dRCul{-s|my}kU@(v^=dFTMNM>oD$T2zH@=d|4S3!F}1T{S^L89^)3CQz1iP z1M>tb*Sr!nUDGIHZ}`^OTDFjSdQY%#h0G5yaOxwrGi0Kdqn zdG_oBeFyA1?9rJ(S=w>8I3_~43q*jM!g|Hz41p)8R}~G1o>!|Jl}`JZjrIkZ5;Rh? ziEz)u@CtQ<16Wp0#Jszbia9`L!67CblNbd5I+?-x<6SSW?y_O5P&j>jVjKrD#KEL= z-N-9USji2mag=leT=Kd z(yu^Iz7~O3z|w$=$m^i}uSQ3%z|J)o-j$`_w-7FtpI&~Vz4O%4tIHRct|w2$)7e>G zP;$O{R%sW>#?1N_ki^IC!k$UME&>{G^3>4rvuFSBM?d-dd*A)+iPvsjdgkU6*Y5oO z>SsTFy05&@TR;D|zrXj~o##J5euj&WOMZrDZeDy1xkD~ogWbeJI_l1Fj1Fe91qCA6 zESPP~tY{~utV}RqJ7pCcM|4OCuOeJ`@in#rx)2dh_g8jlg3s~As~%`W!QtnAZ)JZ3 zo~}qyXZ$eZj#JfD`6T9aDImmP-cD_^g6Vh_RKzv1)_7fh2K2)_%TNFH`d2=F`{UoW zEZF7el@G{%{Nw?!LcbQ_1T9_p8hrNF(zT_lUjxna6c})C!%6RJAOH4&BKca|P6a)5 z2|j!WEm~>eJ&i{0^`)zD5N`#v6xex7?|gi@&vN&0+hE*F?fL{fflEuT6Bq{=_8(tP z{Bn$7KH2ph8VH45T*2mu3(Zj|nueCor4E9_pX;-=aWZJ0BT>EV{tEYLeVSK8Sz4NW zj!sV38TZ7zvfL%7ktiVXQogJ9+;tMf^}x#HkNP`a@0Yl}@Nfle#^FEzk^R+jfTBj`!#-T`P2@fcA8t8(fn*2D*h8)^{w{UBC=PHrG{blpxlf)Bfig8Kn*_ zf-4U`7Yo)yg50te+^thtC6}EOihOXlI7m5mF9{4wSjhung^Xqxwq|6qnyx0d(}gvx zLNA!D^^5zdTcwWhSZ8YcF!h`!arg`(<6=cv6pHN<($SM$LOSYJb_uHgBeW9EF>KU4 zj&$o-91#SR4Ap3zdh}6BPSXl+(q4$h$4i|#@8cBP(jz?F714vD^1(PfR^ zvdbD92@etL^wij@tDS;o2cAk%qb_(ok*ZG~HSeJuIAAe;&47 z8C8nLnOIH+7CL4u&D0E)=rrG!RX%1@MgduXGMj43kbLj~D+i&+`WF{^q8V13!mW}c zB49m=tiEjV!G%XRG(?9svWnG~uwcY2W-`T$saaNR*!&(M845j_)pWRwtjluNuURB? z;chFW!TYV(`D6*Ok0)+jdg0TTuO^m6$*p*7-qh7L^@I_#ND^7iGCR$;`^@Lva*V-ze&^Tk zfA-Ob2S2;=6g-UESAVQM+DP7%wpsRbDODq*ndyC^lOjMw->5}BsL_C&Ia3_r4l81Y zG-3dQhaF(kozNOc_agZ|y$1X-FgmbjcbkjbMbM^7A{ z><7JiH-JBq)m253$ZPp*r#;VuBw$~D8QFZn5`{^_&m$Ho;SSMtm_Pyc{etTydw%;s zNE>$hj?*f*gai=~sas}_Lp^rrVd_M|O7bwTf*X*$^x_dwtv3C&WUC6SQ8xC{Xqs9f zb01PdG2<&3OP>)A-$_Fr>3}kuf#08$y+?t-A~d9TsFE`BeIh<;Wd+H4*Uzku#xYP5 zh>d~;yCX`vDR4mEunBpmDg-rr~&}ztFX`Ob~i81XiN%UBVK#A44r(4z` zA8esCwBM^&P~?$_15V1qLYUh0Jz+KywD)aK*l-hP#PNCRyPVDYDTohxb%iNq@v*An zBRL?KT<4sK?PS59LnYOyi9l=}eDhr?B)aqScRu^!wAQS6BB*iEs4=(SU|FWPa28)mEWU177RBg}ttsuiW_F?Qrsg9li$)W^6d*u4 zVy!npvK5Zh?hbCQ_L?0a#A7q)jKGlt@)a+g9wb$jeB_>?sdcLiXX{ibA8+DDu{vB$ zO_8+C5+kNvqGR(>HkUf-6-MyqbJPuK5cZi>G8Qc%zC<#>)E2h5NmOrlCGk}XM4gw( zVR-pK-@JCI zxLQuINQ3y&D5$JPYYu^(a7bDOilJa6_mjoU3}dHxP(z3x7ARBAvK1UaJLZC@W3I*M z;*;6aVT>6^AXJxI*|{(Vh{e_lU5y6o*PM+~?ChnHx?4jGz4EfH9~0bls5cuE@>rUY z$$;ak^_V>a4|d_-JQ%WRRvu7MZXRmsS`U#EVxJUlF7t zNWjhoLNarZ$$+70NXFnrba&EzNL=K61W84N@3V8ZPFIxpe*Xb7pF%2{F&SGW8kpd9_Z_VLE#BsG6}XC<;;T#FS`GD|GdwCu7Ayfnxfy z%%%nPFv{ED<-5gswLR|nF!gJ4Fo%vXr-9%HE2B$Zzi|TGoi$5Wwk@e_Y&n?-aUb3`=rU_KnB;T6_0Q za${=lxr}C;$TO6bJeO3AA2*N3$8(opP0L(g7;3aWH1J)bK zRJiLS>I>d(E1Ypwrri6A=!e=jC`t+U0cfKjX0KsZ@J~jHB ze2@g=J^q2HU{mH3ebk~7Y!+kd==S?HRG%27twq$>s932p9o8Q>Kmw1JXNkkyYEZjd zVZp6ZuP14P1?0~!i7>JJj(dK<%P0A)a1^@NY}Qtye=%pLfgq52E|8 z`W`iQ^N0qqmzTJ^Tf4_N44pY-LL=iC4e!N%sOz1Xs@9VrlQo52OvnCiJq!QmH3bbp zR|*aFxRr(=o4l6Ermnu-%cX9U)S)xmTFlm-Y_X>r0(KVmc+)EQGmVXp8SfFC-ZitIbMd(VC4S71ViRn z!aPrScg=xIEyoSO$|t0(WW0WL;ECmmPiC#L*>2X4+pBM28Af+)aFR~s&n{g5?BaW$ z{{H&ymtMYg`GbC_Z|Z>LeD%AxU-;FnSFhZ9{a4BLoRu$R@_CgPvn_=}d?^$(JCoWp zG&6nOsBlcwTyh#qGR3^>u;jIVTftr<#b6jP89M7S@B3Y;{H|Nyec{#{|BZgv?JLiH z_V`OX=XV*lRSm!}s})SK?R67;xm@bXm}sggYIK~5hQp$Yv1AeL5JW^80SO!ZPZL^CAVp z+P6YRX9s0B^opWY)NE`eYnhszPYthjZv;hn=xq&rSE^8mmDRFc$QZHljAbYUV;op( zBZ?y)^UGLSapIA}(shpSH`=O0`P@lX3oY^TF>$(ztu zACHo;?wzpSAYW}D-Qp&qS9{vxWapxib4(h$@dN!0q9-GHH*(*d$ zNt|lP)?^R#BO=r?B9NKGZ8E7lW}Oyc{VrCLgI+kR!r4UoMR>fRBWl8t>+~czC9@P{ zm7=|s2B&FIuB~0qa=n4ybL-*0gQ&1hM9X&N#**m1Ks3-mhxHZIJ;V9|Zv{M{g1!F; zuP$A0>>;Rwt6hdA!4I0?k@sg5Y;qPHNfwa((;&d)?(}KuARP1P;_o~x^H${7n{#13 z%7sXJCivBx!o7}m`e9Yt(BUmr5l|tha33rF&@+r)tc3c{tFO_0@Vu4yf6Qhc_-AQ_ zE~1jX=314RR}R6-3%yyMg86Yvd@T4^&!N3(NK797PuVi?%!s-nHLqDH9NYTsP(tGn_=qDiC1@?Rtx$L!aFpc%XwuZ!_H3 z&J@@u}d_RKze zNz|K1#`xu@lF!n0x#HLHq&IM-O)Ms{(nlO6lv^1zZ-f~|*trHcEv(;ad$y(Uz{)Dt zeY6u#cN(}JrcP#>R)3dSfoorHk|9Y686-DKdR1Aq%Rge4ewYXlZ)Q4Wu&3d;kcZB# zP`8Tc;wd`%wg{5TESee3>WWFwxE*6l*Op#~HRai*Uz1MC`rj@;0iPk^^U_txU;i{& z`XwrBskEP6zPSACZZ{-X?h}^JXXPlbw1wIISrLZ(tX!*KJGAbQ4=ufsSlDCKqidZr z(_V+Q(B9BJGfmAjIz#TURpHkrMuv91`kMkT*omy&mDQ+qq&Nh_@L<}VXB7qgDzP+< z(WWO{Q^i?RFbeg!7=@;U5a|N#OKwQF#)~%1Piz)`l7<3GsL_27RV`}WNPu{a*cab zzL6=?!d@Mz9_s9TnR=T6q%$P}Ko>H{Y}^z+NaKIs{P7n*dzXCv&b$Bn-kbd=lh1$i z-WUJ%y)Sv;jyyp z_dj!H)tns%)gi!z4CvXjWS$2#B3!3QPL{?FlZI%75kSB{wSgl?M;1m^aDSzgz6%?Y z><8s<+q=EpvV-E9GBS#F#>tKPK^g?j&IDNR$S@pV{}r$hC?IR@9NHB^{m^M%;lk-N z8&kRS#NzA5y0R6IEZ*$35Jw>2y`&{6EEEl5*d%IME|M|f2V`7MMUS+r@k%R;c?+rY zG-I=5e<`enaJrennw+-1V@KxB0~ijLc&{sHAYMy3OyWHEqM9AB8_G52?eobnZSHyv zzM)~X=20eoA4Ub}H94o%37s)}Pw0%f(OaRLpwl-)GkTbMpqCv!(-+Kc9h0&KHxIAs zkOnW;d~&rart_I^DG|zcfHLl+)U#S9cfW=^l>;#o%u|nTtM_0iH3~-DA_WUh)yz#n zwp7t@@f(a)no)}wJ_Ic4j#9>`ThE*s3u_d(aJYUqk5 zs8dy0?>9>KShX*>OHE~Yd`8oCT{FzAW-E9}2MawHw8Dnc%4>SIGg`JT+`+)9S$dw%@-^bR{ljPO{Q9N8 zzxUnSm%exB*U$Bm$+7V+nH-}HX_)W4L~dUE{>_W;+`Mr8=EV{l56VLH;ftcI2I;a&ZiS>^!FFzrR z3*fZ(vCa)XdTl{HWBJ)0XMm}neW1^xAxp2o2(H37o?ZF?3kl%&wZ!86#_XD)h@})z z#TKkBO^^HjR~r8-3~Hnl`m*YP5>LxT%( z#u)2Ph>=9n3cMzk#_{Z|qDpNghHS@8nu08H;2rO6u!nLJ#334?Pw31HcT<$eZB0Lg zjV+fMYcuh`sT!-ddMW#1dUWJeM49ZLdYGiGNdZ|8_Crj z!j^HEaRMGoSm)&kSRnoE&}}W(1+KTJE9VgC#5F-)JUDVd0!o%n4IrRR3cPcRkoUEq-Z{ z!*J=5Y@e1jtQ&HNxZhU}06&ue`31(tOux(5S=DQT&{dC|Kqyv|qCh+S?*;D)k z>^`&EtYD#S#TQ~*@m5~T8Jmg7#;o6~P(=L?iI}E7?bB6m;O-Rf8ErYr?RbK;Gg>Y^ zzw;!ox+uujk%-?}j&g{5=yiG%jQYY|Si5`nEICNNi7NRtaskUGw+fcZ5SPub;7opV zV3*I(sWrxnBrZEDD`XPOvqe^|XkAmu;F>XWxVFsUZVDNBP5aTVpImnXbtcyqgxroP zh=OKy6|L_zUkv+Fz4(_WKV+9Xkl&HO8nr55_$spFI78_TGVeB?fgOoaKLhdG=2) z_+;GUehTw*s>aBJAwLJcZ<&mXfCiiChy}UlG`lO#7#rGw2EkSWfZiny7?$D5lVi+v zIos!xBaEi#ohqHD=b7V(f2xC(QDfNX+!jQ(Jq5F|r=XtEOuf@syw9-LEzTGWd!JtZ z-R>K`CH0Hx-NfYdl9Z$j!|r!+_s zZ~p*+6}4PL9zad9B+qVYL@yEB0b+m|7v%nW>8n zT?+zRy0&y>>8+11FTL^Cr{|GmKZTS+6#F0Zi-Hs1+X+7Hi31Whq{1YZ zCVqu_1Qrqa7`%hU*THm=`bmCG{g(U-6)d9Q*dOI;x)48{SI zf6Kw^kYl0C-2?CqQs?Iax^PaHDyki39-B1` zU2jXr&f7{8M@%%CY|+wT+NX{I-ir=Wz}%4g4PM?!4++!J)_p+26uksf)CYHCn;QtS z3N4!4jww|{BTVi0mb`Ke)=;+!j&BZK2(dOphG^4ajFjA<29i4z zeV4&|0RAM|9}Q+V&(9wEa-Fdy&r2e3N11{>;PYyoB9z(Dqet7qOQ>;D_3J&O8~&keUNa0VTQN0Vw-(m!o;{0i#9V8Rzy;uAB+fRm zheg*~l;ndW;|mdR6&XQ+ka6Mj0A7-RhVM_oFlPM&E2o|uU07*3JUJ>8+=gk{hHb*T z1P6D0$a1Z=A8W2)kdYD?im-b`)EPpUCfE3@Perkl%uMfB2##;~64{z&26EisvDmt|g+A7-UrJ^_Bx;-&EI0ZrrE1dbO+S)ez zL|dmI_f-eH^C~=7%ov*4)xwlKCT)-w(zT3D`w?|1#tcCdXY(0T&*w5a>ST;J8=0(@ zx4_Z^+xa2a7ok^X{l?ug>aivnW+%Asu;e8rCm$LeeQ0b_$)PH}=z82(7jdfrnL+B1r;)m06O^I@Ab=oA;3_sz>&d2-!wXn>z$@QqU0QFe_ zDa8_7DuZL}%Xb><{OfcUT!O^H?q-q!)4x5Rt(c*v9A7Lxr@uR?}}>&gWCN^q(jnu;qk!_>0Y{V1>!1)gWHMkc2i->L|34m*&I^n5{k zIs%N9h5ok2e{;6kws@kQcv8g<)_=#c1bJ$Oe#d!-z>GR1jV8s8GQRTAB*Fe=0=FOa+hnXhZ2DX)@E&7IzDUwJeohS& zD0Z(E+wiq=n!PC#X&wJDx5nox_qG3++fM(nU0)p&ha!jhZ!U~)UAENm^G-%Kmv5m1@JYfATG?9^m&=+xL)Y1_c& zw{4afI#vyG=2Rq?8C(a(GyzXv_JHqy$mO^&z&r}l)JGn}KYSuAQzZhRhub;&G)xpuVfF*4d* z{GD5cAHLu0H^PqA00i8>fZIZ*BRwCxKcrh$>i=3FCvSx!VYq-6Q^ zLLrN-0}bv(eI?YH4(lq{EPB;LtI7RcKixV>O6$`C9*RpQM0qxaWmppxIM1zkutqAA zO9^_$4S9m-49(mW2+!UPWP1Y3&>d24mL1`4FGXTD+eSN++cBkUV_j+5%u!^JOr2N& z{dxB6lt*Fi)&KV4PkE7i^7xC7J$B@8AO5(3sib$|gMhoF%K!GkE1XO(z5@sT^!>$` zk(*YK`r`Yvb#Q?Mi`QK8b+T{YzR{z9``{Nx$;d-flcf{n!H53#;rE8f=+P2<1n0Sb zx>$$vj7+Am@~zq0;>%??_AqAwNRZ5gfsfB$0oz_Vzqa@rxAw{7zxw3ypY5bpk4{Pw z+39|Q;4W8*TU&gI%q_kP$Nu)=bLASmV1N7Y`(=3IR1P_Lk%%alX}MMtWo32F*p`BH z812P;(ej#^o^EbWtC7&<{x&y6=((%{>?L>NiPbPQ(@G7{a4%$?C+Rs}77itkGOr@s zZ9frHJgm|5(U{$od6~>ICENEOYxXhJ81I z3zc;%zTaY+TFywF=UsYh>8h-2J9B3F(()gRuz!8|8M6EgB91OU{nzVX`S|URf42}F zUVdKrfb7Rl9sq~tYjEhTrE5!9zqWJ@v#d*R!)NbnAOH4&BKaCzIQnvN`RV0L@F9G8 z>H1%rzT9t^q_4|SU&o9&KFoJNzMS}l7@?974}3+T z@TAyngV~O4FC%MfVKdR$Ks2y)b=m>Xt}o{2x`Da8jhJ_up0Tt-KCw(jk-~;5G`hNC zfW$UGEM3#n_gsy(s>EFpup%i{!0u}4)^MepgMiV=>kaVanCCP}Sq7c!5$K@Pm-=>H zc|OctkNHgV12zw~re-G>YIi?KKHFYC<#tTD!n@Ut!W~=h_BT|FtO0&{>DphfOZ4eW z%TH}xMUB~;7|+~*88450&z70Gq6X80=U~DhV37ls3dX%_w7S+k#FwiOV~H@hhk-bq=01+N{fzjB?! z6^hhgfrGgR79<%vHQ8sz$b_U9uh8TrghO>NlW$8TaVt_mLi(ns4d}`mj#3XREj;kN z?k5c6B@9(sme^9WG^2~9t-|8ZqH%BWRi9U6nU7vD^9SwB0|&@A<{6g$lhbf(vmR$v zaMCK|cmy+gTpob=R+^sfHw?Hxz3Ic*Y!=JQ28YYrFlrQwlL@J}5m~Vrv<plBfmjF@&>fw4kNP9B1pJ1Spwpc`Z$YQdrF|ufHu?H1;^~u*obHy$z|_ACtd+>X z6gwvicw0g#VuimWluJk za4D?TBS!~Iq*Pyg&EsVW_$%WF$rrm2hgD@xMv+Wgt;N=9rmYpax=YG&1H^4wq>K z9+^+7%&UVr*6V7mYe=|X()z>j?dT7Et2jx){e|O>sM9`6(M30n!U@Z&C94vnzM&^dWc^K?)yG*?V&La(K!r`-7 zW7p5K9wTVW@k@u-@~sY~QrMG?HFGjsD73{J?ZlLCPRpCI=Oz4E4`$wsZ2OrK^Ri@f zjh%ySF4J(+u5?vE4+k|QcsUoWAVCp4QFaE_sqd3o)0fHG0|eZd2t~?gtS*uL;Hr!u zc>!F;74tQCjSi5|bwC-coLO#_m-}L!>-LH^><%N#&gdH!k~(rRnxJ*=^&(v!IC-qo z+`@jPV_+?_rebqtHN!GAM6Wdz?&WOT=*q@6y6lWr*beZ{8Bf%S+KoR+>jCxpB;v5K z3YQ~k<3n$G6hb#tZgV3pw|N0mR|Ru3-X~nqz*&lY%@_;(>LKzd_&u;juXKB_z{1I= z(pS0O@e)NK1UW<2Lc&mfCvHn*ALvePw&wG%EG7G=p-Mvw>)xcs>PS&lX)K0q})IS9%pmdOcUj_-3poI-mDz^TxB=<{-7 z`5DD72?yA@)?%!KAS|6Z7{AFX9(oM20RI z6?4pUWy}6{7GE5+a=n*8IKH8nvzy^*iHsDRd;^>NPB2;x88Q1or>>oYattShhuB9h zIXQWnj0h1-Q^x@Zh^C=g9;N}(xRf`Eh5-UHz$NGk>+Cfn1>x#Ea68X&(3M+Zog}Z0 zK~Y#*zL45G`=o zBiwQ7!oy%MjHh{k!0o!s7eVq8r^z}@#%a(pN^M4=ooOn>OEjSjb6U0wB_SbhA>tOl z`^n>%mwvJQ)bgdJH?UAyHk>R!we;@tGjQmw{?magi*7catv;6zP?zIW%nKlYz)z4iU8w_bei(|3N6SPX;H*-RGxsamwR65U7+*d&&| zDk>tXC=yIOpe88`j4spqG_QtHp@r0^6~Q9fcr*c`6S)uPw!ZS!L!|700{Z7&FS{N! zGtn3lOYdQiD%cUG0_H2X<4qzft7W>Nw;DlQs}nw=i&qw35v1f+JT^}zYt(CygDt8) z?(hn&NX}xoXI4kFwK3}XH7Zk^J142ciySUfWR$WXML4v}Ph z(ehnbSxM}Okq{9%Nu5h)sxNRLI}in3Kj1+aNOrDCJSB1_e8ixlp@b|YlV#4*%)OCt zHmnPkm?3F>Aj-(r>@M1sJ=jne;KZRM2&Z|OyE{gO4ieFI&CH~BdC{V#)6>YoiLr}5 zn=hl>k-c+iV(1<1b1CM6j)cUU(O0qMoyL{he{Kq;^88nJesZ-z?;lxG+lCbn{5FlZP!5o<+Xf1HEq|upr|JP z_ThiHRpR~a!ylK)AgwxdhLvlVfcM#p%Lj8g#Q-6)sqq;{cstgY zjZy(S+TKiU1-SI(TK}0_dZAz>k5@D5l5rY3fLqm4Sbe+@m#YQaPc)#7B0R;6+vb&Mg%*%zg4~DvU5V22e?}t_e9mBbz0_&@3@o~Sj(?e#hj=-(B9UqeCXDjmp=XI`Z^kP>U5U|ouYhR zRuAmF2EFxz%eSvyxb@?!`18ddB^Cg<$cABSW=>PJXI57@a`c#SbOWiy@?A?@Y=^a`AU+7r?=M&v}sF=<(|r&F)&h zVpjmUVvdr=;>%TVE{o)`$G~_u78+n(2#T}bFM~$M4N{IOfzF&MIjm0I8eO2UvG@vT zL3#L;@6u^6{8q@$p8fW>+qiW1EnM#>Wc?ud&z7-7JrC6bYNAMYr9?trwDMbGt^cDFCB(nNBG;!DlN6q zPE5HM5V9wB$nJ!2D?Pv_b#t9o=Yi3|F4wZ z!30koTg2p2vH`Gf3roR7h{eH-UM!UgMzN3|F^aiN_Hfa%2L~78ofcu`t-;BSb8$RZ zM(MX1XYa1~%tPEQa~GF1GQmE*Ti15Il;^900|g4RR>7Kl;9&8UIk;K z2f#uDn-l!1KvsRuA0qY833vm93`)ff;0#P4hHw@-S*?TMD0?ET*2wt8P@iQ)r{KZM zu4lxyuEB=SNKN+kHRXdT+)v1peUQ9l@ie zpj>w{q9F?j?$cSP@pO5G2~veCZ7-v>srb3kPCUulNIH|cx^9YAtH}Q|PR2!pNNB6w zU*y%J>|l_=IzCLBKG{F&QA9!>5>5aT^@Uq~Ah}LtGNX@xYf58XzO54Tu@$0if!CT^ zK3HcE?f6l#$cQ(nO^xENHsknct*dT#nDH%a@j}6_>6p{dW1u`^M~= zkikK{cg>u>fuPJ{1`OCP;HPyHa03rh2W<=<-><4w1{JkgJne@E*c}lnd0Gm$tdPCBSVTz4W;qh zh-vNPYNchnP`GB|I$=(;wskh(-GQeTP93WvSU4xZ33fOS;3u5ITyi*I&f=^8^LiZ{ z7(50?f-smxWI~U@?^zg5a>wjgT`dx^dI!2Tln|29ZvkIjXF+kBuMQL^tYLLhXZRO* z*3E%a}7ao3 z{NkxEe)_#He)c0`et!9J^2OUPe(~e~Ouob!8Eg$Q%#3E}1!XN$%*WXFqmk3BT-0uz?Z&eUg7cIIxUkLciQ+cI4slhX#|!4}}Mh%;9PffVMMv?$=X7XLQ z80$3^4vZlKTcx{cf_ffI`*kME*uQj5mN9mYODwmYpCJ>qrcX}OStjXzdiYn5JT%nl zU_BC+X|VXRtGZS&Hwfvk=yYmh-qq>Y^M@iH{OoqyMnI~iXTz}MnMHj zerGugd&%YDlU;Ion6tp8+mRYiyqx^$dyDVF%`ZODLnI1W*h}|&z8yfOvUdv}vFHn0qv2q&U!9nI$;lKchKzF~kW>yNe zO3)a$-ps_dY@>_nZzpgI?urk+QeRX79pVQvPhKJ5W7XkGiRI7($sln@k%>u5-M9<& zM0z2Ee@N4_7ApURj{VLGq0ZA=r9ee8Lb-#0dD00#Jzg9-I(2vfZV-DZD5%{QJGC!D zugrQ^4q1IxIdqKI$!X!tBz{9q`L%gkQ{N9`oj1hT_#Me=g)ZlNCT>;+!IBqWgZ&0p zmdL1IomT02tf<)OIA(HToW_QrLA0=g*)3#L*JSKL%>s^R2u}i0ssf|rzSshLaGJ0~jzm(KrFYsvf|7Y)Ao13`O zwc)?Awx6me%pTlUx70fAcWTOK4mNmfC+ti@O)X2y(n76Xnm*W=IB$Us2{{ePgq)KJ zB$)(qAP^wom{e`gR6QT%c|QQpd>dl}lkA$m@ZR^5Y_%oX>6VSL*klIiZnavy*1Zn* z@w)5`vzb$eIrQfy1{Vca+Ix%f<4T+*>^sb&N0e}^VI+?Lxv?QIU=rwEhTSAKx%{*y z*LC^L4^>WU2gPjXG_$k+IQaSi&0&zo#7$ThZmd&|t6tw^&{s7X^q7(eKr!4&O)w0s z8dplU!I#QV9I05;cd9f&0xQPZx)7m>43x-I?Gk^%AWsW3%R=+t4yighQu z+M{~1yS>v(s1dj#JAxp~>Z2OlMSTR<6149gm^uOiiNj{NCR!or-*hu_7j9fSh=*V@ z=2LTzAa+5b2E(I~5<9h0Vpm;Zm+Xf1y#S+deCC=7_pB$NzdT&s%bR5j{J50)oBB^ps8jU&h*sCIk*Aju0X;%__Q_~cSSF}8N>`!i5Scyj(c1VKU`XtF)idwLf zCW)GeN5D9NpV69OZdDo8jzh@~{5ZI}7YZTDTRa&Jh1F!ZJ*jsmJHp*v$(|myJJ}xX z_DCvl&M8O}v1tCIxf>$uIoxl54+xPe)sC&t$Kcl5z=ut_~?B*`lfuUEd%d$~ujK9U^Oz zZPdhqn&8+E9n^6qs?zyADay<8`Q!Sk13*`F|8by4n4K+a!t4y30U5!fPd)r72;N8# zd=Yme<&dFNRtV$%qUkguTo@mZG91<_+$V&Qq9&5AZZ`48l|;10K=ex%Pzp@&w6LKK zHfXl_5iaZ!0^v#l3Z$h|u);_SV}QbYS{tBPb~EO>D=S7Q!c7*Tm^oxIO7!4mjXS6% zmPMpu(aOFmkNt^W*OYiot-f!>*5zdF3zv@EIelgR$fX=$q9kfmPoteUz$QSrNJ2yebz^pz% zD9KthMa=B|VhUb2{LU)BHR6ZuBYh%IbBz!HNLUP>8`?SKu|qT> z_yo5y+?0J#mPsP2Xu7Vbx~fDpk8K|~zY`>ak7_s!__M4k6*cm=G#!mMN!8=3aGbIs z6RT7@Ct~~KK$23*Z8qadVXj7JQgGnynFh=9IU3ExNb!-{-qLAq6z;AN4qdRAXno$a}CJ`zM0p(P;_-L>U_b$ z5K<@XNU%|x^*q`w{WWV~KecU|%7R23$8PaNF+f{T)EF#<-YUnS@d_O&E7jV{y2hZ2 z3zjGdTl%=mV8s%lxQkgV2O_jA&K@p+V%aiHQ-fR4^=g2$gkDiDN+&*L7=e+nj%nz2 zQ|F;`GmQ{jgTismbnviI-BfgsMz8wU=p!K|T!VWTq{jQd-#GQpt8bGZ&tLfGrL&J} zynm*~J6h2Kqdv1Cbcsz^<6yR2|1v6`EPfCFJ2(46@nZ3M@lx?3!BhBa_9&d4D_(&s z|K3K57vS;_@aHPrRlal6Pp#sgpkaq*r;Fdh>sMa7EYW(OCX4UR9)wnYGkXj!e2=|x zp?J3V4K%y*{b2Hj9=-_AJP!}JOyD>1N$2q=ikD`e_bi{Dqx)^B6|)H!uV*7O%}7YrZWJ zCeRgFj$m5CY@L$R@-ms7f(vJh*Jlsmsx*6Wb_&{vtJzt+4!7Y1YiP1!hn>|luYDe_ zZ=i=)wb1pb5?vFMJ!q%0G$Df(=Q3Rl-T{;w*b<4k?6W2lXuk#K=^k$6nEe>(;wJhb z289UDz1PovO}0Ol$KD{ICm6;A!TNrWFVo9^;&1*%lOfERX&8aKiz4_CLF!zGnbb@% z1Pic{kF0xtftqPzm(uZ`eRh+3Q>v1hsGTKSpq)Hr2M2|1N8T6Kv3uCF;dwASeCni8 zj3C594SShakA5 zk5RK9b#C5pAZR#-BlZ#OQ^JG5-&EKTsjM@eZ?n~|yV;}|F%UKxR*uG>ltcr_vnC;Xt&fsEld4j<@BqwMUyeZ8m-c%I93}f-N>{f$| zBsh;Ur5n&=xK*l!Als-hG2&WIJx;El84Pt3rt6ox!u8b#17^(or{YE;?E+hl3Wp`h zM_-+P{p8*A&o8|BPE(R{-+TL;`RhmL&%QQ)`GQYU?n+lIRf%hg8i^_RPkb#&(jb%s z2&Kl$<#)|AC@EO*WQ--U5!iWQLr0FIJX{yn?}<%EV9tVN?M}djR6W@G&9$z!n{2;m z@nm_yX|fws5)-~)|szbJq=rT)v|$k7lC;ft9FHKTsc@j7JQG0{bYC?p4e<}( zQ^K)obu$=J^YMRP{xkXK`4c~W`eid{oz-!>H-7ry8rkyG;d4K|dp5Y|Y6nQco``9l zcxr37ekoRXl}fA_x)s%sZ?|I00JUshC{S~vTDcgIYNAC_OIalZfm(QqA3#J??iiBW zZukjeXT^9u-D8tvoUEE@2RgOE_9HKWCB`W11=!^Dgbm*>Wa~_(mR&UrE%15-#AmZ* z&^4fpgi1_}9t7dQM2)ff$;>?M*WeaTlJtXBz^oAp-yj8v=E>nvBtyYKAUUi9*(n_4 z2{)*jX}Z-9Uz+4Z1IkGI7*8!*`6~0xXMqNYC#J-r4SVsI7ie@^Oqylbw35;G7;I;@ zJ7POYcf7UqLMm)(|1bRa*v2?e)>gZ*q5`7E;tGOvYFZexa)O6suu7l9TckL`57^$s za@_-~qWetoilMJkukn{Wvd%YX8gFB!H9>Zo@BkzT>>rT%0t3QwY`%As?(A;gwflD+?Yp{^eTB@^En7kB zAz5mZ5oCTTzfCCuQ8qM6SdWWVg)CC*HVkf(4rtDSNzXA%&JkD*py^5Ya$&-fMDb%> zln@WM5}V;}oJ=Llk52aS6atf)naRC-HwrLjou>5e6r-j9QyJ8Z;Nzm)0bR(UoLf$g zu@VUJs^cT7#A@=ZeN(8Jpc<0G%*hNlDMoR(o1>zwlcx+xOEzgD&GoYXenob=mP0yS zC&vsJS!5DMd4^TpQ@)!2okL*`WJT3g1PVe8t$dXhpfA#*>IwMUzr(at=mt@_3Ysj( zmlwaoT}Dyv*VwsF1%=CdFYw^$6^1Y~o}^(l9Pwm*%M*X=3hM8%>f}@_fLS}t3$De% zRDy-m1>@&0)Boo^PT-aikc;U^vkG&)la@%rX5aP1uE`W^5GgQ_2z4Eir_Oq&e{V?D z^x$8n4~z)z8f4~-NmxOoC^2aPHa}raYe&7_Vefr+&z!n@<_g~O!ku>y-+AHa!uwx_V`@y*BZ*C~ z=&fq0d0H9VskQ98OYpt*17^`x;$^XYJW6b9LtQVYt)4nP{4X z8Y6Cgy@=CWSHM(WRVFbd6s}gfedmJYIdk`Kc4<_Xr&+?BInBtHtjOc5aMnPMVg}O# z1?dC@z+$jrcw841$A#ZTQBNpgt>MbFCD8h?4+@rO6sDktU@w{)nOuHaL%_29<|Y1> z$ITQx3xt(&_!!%uD!2PJ96H=|hB;ugNwO}hg!9g&Q5+6DEVhHvKz5tbr`}(3$1iP` zArHS<`OMGgh$ll4B@}HqIb8lCjEBxe{RLLw1c9c~5wLu+c$bas9cJ)oSW%eAErr-W zos-*y(}@aflL>af;%O-eBxD;x|C50k)gACx@I6pu;bl@egJR)}M+fqtoA`w`iKDwY zTvNp8TL%JrWd6wc+gJXuaO&jz2X7^x$Y38dgD-0JXmgKjiF#szgG?Hx5U=m^K zcZm`s_|~}xrsb$NcOUu7pbq;_d4{+&O+~)7$lP({Id8ot&HcWN!M4 zip0HFlfyJa^Sl){1(d0ZPSfDNde?C~J3V5pumOgBaskGZT1ZiAB5@mM+xI}Xrm4|n zW_|gsABaUCN3ia2O-8J5E(7!oZzZ_=z{F?h$`O5n78#%x$DORhwV^?^2J7wws^bIt*7X!oSb`{pDMS2ws5OF5mnNZaRGP^rZKWTQAnR12&N8FPdzE-8_G5 z3i{+ba`PNC=MXgJG`V%?){C-nH@~`h?$(iVJ8wM3G$EU{z$^Dc(h_PySb)y zDZ#vpj&Yl8$-~SxnH5~kL0+xjxO~3=b&3oK*jYWHjKJxYaRld%$2Sl^HjY$vF-6lO zikgT{EI!e2`U=4NT97ri7wvqMAu_Iq?WODJ)iiWyfj8 z3=>%Px|p3pQ?cX(+bUZ$w8$)jzNW!4J0zy+53&7$#kMm`_wjVr<8%$~`Ho}#K_Xh? zb$yf)bl*~f0%Nwt5j96KDVN+^hX8KjKnM{xL8=gN!iH)H!*ql>dPl_XU~9 zx0u}VE7I*ah?+5!5(Y&>xcrID58$VhM0(qn-v(DlxaDLqPXj-D=9vyWCGr$UKYV{u zO@!4+eNqigrbmA@!E7)|7r$08Kr-;Xr=A+N*ceX<)V#LKl#flHAzH&$wA3qs$4s}y zbEJpbPM&&l1hnXKv?JdRB3)IYYXTNnmJHft)A|W2p$drso!tNE!%UMm z$VxfP^NBU7EqX!;C8}lP+ldOlhB{awD;%+Yz(O0M%inqJPj_EDKmYQTM%4ImYUbFo z+_ZSSQWYYP{d9jK844?r8u*|fk)W<8x|*_Pc;V=4^9NtMbMe*r)9>EC@*KH+>CHQD zynN?{qk+YtK_A9c$uI52v8?z|e7aCT7*Qn_YFLG@)T63D2Xq1kH%pr`C$Q;@P#4>> zSXx`+V}*SqxW@Q1C8>!}CFMvS`;#h`goM{(FPAJAmL>zPjF@8vlCW|TWDPm1v!MDz zZ-5#}`gadC+0N8Yt$o7cp!{cqJ@Zv?sdeV9P<#zrP(p&y)D)er!WRHfr954VvU(Bl zq(xZ*WURj;wm)98A}&$bKcGdiPM+8RL+aAqj|~ejDr?2jGE{?`Epr=)w<$~YixEYQ zCg6{*s1dIr7ScSiai8@W2}Gj_*C=wfUq!BXqIkCW9U{V#;w6M)l%R~?AZY9+UjVBh zCXd2v|4_!0yiz<%idW%89$c9{C}AfNSMt*Ab9m`(i5mOeZ18mHcGnTh;6VcyHYO0Q zMvt$BNcT}a+!{v8b~q#3n3*nl7`6}{N5m(>&1=m*{8}T|D zqi@w8#>U+A@ww?gV+O(0wYjM?cszY=Zu*Pc-+z7Q=*781r{<>Tlx~wia$yo`x8|p ztZ323frn&5?K97Kw!Wb(wcy(J0m9Pmt$0pb8B6 zf(vrn^h;z_otFVk+38wo9#}57uUuJp>zKz>@C0pqPcU6mVp_FWM!8S1*qmDy12u=_ z@m`y`7CeMvmG0A(L{*8*EbzKbpKc+o!seSFjBVA-+kX7pocSa-F*Y%Q1ilRvApDl5 zvL1iE8uQc~#C&_g^q7)Zlh5Rl*|3(Hdpkuf(#KedH+LQEgZgSFTwiezXrW|CRbn+s zgh2*D4>v(Z7W8ns8MMqg2`4?@Fa?PaMZf#vd>5l=BtP)i=A5mWRU_0l)? z0malSVG6fW2m}kxZrh|Imiv3RRP#g4G$JR;1!?9qu1G^Eix(Wbtt#(*v#Tvt2@dWM zDYbOwS}NDNjbhSnD0q#@lhn+()Pw4*Rm72Z-&XhBp_k^Sj?EopEEW@3Q6*j#*8YtRo&sa1qq%IH;sRP7=oT|c8TORh%L5F>^L4#y8 zUD0ZY0ewpbbbiZ`P9 zaG6Ld%bicAMi2;63HpZ?t)tfUa*K{&(wi0vMVmpft0S8ysFCNE8i}-_)-xi4r=Tu9 z!6S}}2P5pygMs3RDWS&s1Yn2RJ(TQfM^mS*WFo;yd*|TpT~GEeyEfR}*FN0cC9heq zJB=~RpoNmb_0&^;eVI31efRF*&W#RhBE_<9Cd)i(T+om;2rZ&2@y78&X$odB*(bQk z(wn}|yDU2DR61RrCNYH2$t~D%$QHMNJYQgbk8+~{Ad^m7a4<~PZrjubGydNyf{7$5 zcB?%0DI7$JTO*i;%xV+xKA5&QwxrY51J*wsPi?{CTo$NGREZ8KwdZj`M z8>_DAiC>T{aBY0gM%0RU$R%j~iER zybYmI=@>UBW==F&Nu>AH0}F5+3Xf5EpjsxXz08lQTpJcCy#JT*?&ry^>08f1Lq3H= z+0Y9&&)u3T9exa*cK+sf!Tr+9OI)KBc}IQ9TSbSYTEkzxag(*rivQdiWNZ*#px5>$@yG%E#o7A3mFwH;D|H?jrvMHYAM zAu-A+NefmlAHC)ACtIrI)AG3GW?YIchzFk)U3&RWiV=}o(X{mMZdooGgnc= zc9x6_D{T|zq?#?*+DBo~A(blGGHx;3F&ath>L1wA+O^Z;TMl_@o;6q6e}6CzqdA*M(QzTt<*Xo!Fo-c|D482Q&9kup`LREj!rgsKq9p zZUJq0oEb)QDYj2g6AXu5Y)fzqotmtOpe2(q1~jZGsa{-Hj1++)XBZRXLZp|it)AHBLBPi?15?;rW@51p$-F&5yv`v}aCZo_iV#GI<_0dib=?JZ++Cq9i;H5pcwXxz84+A_#fS%nyd9?*V$epJDL9SpmN}=SJQEsI6(RpvJ*5EyH4?9%kEv(GTLoA)Z(69F9Gri z!>$)c|FahsF5+dziumNlLPsLt=|A*z243EFI)$jFufg@EBKWr&Zdd`4}NgFBq%?jPYqgFC)Y^U1ZG21?!$ zd|-^49UlL4K%O^{iXH*|yWu!A?&0aIYw{e%DsyUU+H#r8nlUUb*wdOSPXP>%j3-3rD}bcjD08!Kr^;JNnEs9m09` z*?(Poq4@&f%bFHER#)}yCBxWb7yA}>WG$(TJ&1QlG8WC}skKS0q2N1KXVwDafSZgA z(}_}MM=*iJbg<(=tB!yM=H0wn_4;a@zPdkYjhmS+DT3g(gVFQB1Dc1AaYaIaSC>_t!;dKTp2BYqQK+Y z3fvr_hC!_qO>cg1?piptfztbv_QeJS1jvYnw4|nM$*xdmvLn$GN_KU1_9Q!1tt;6P z*P}_TJKVJNq8b*`{{&Emkdf1hiPkum~qcxs}2s36i1~>|bqZ zkqSQjCiaheTo#p-NY!UXBnVtn^u|5GpowmS-H z;$Z=a?RG6;GaB5^SXW$F6tu2HYse&4L@{jXM>nYS^)2TKXU5D|$eX>A$G87KYjm7` zwx6fOrciEnP?I^NKQ+wiC0VvS`FO=wRet#=UD#J&~rTcP29se z823p!2yzLBU;;KwbOs{1ed(+DSN?$YcxPXmKXPvVk1x-^cYWdLt6(|Izxs!!tE9hk z{jIyNe7tb-^}sV`{*%w=zj_t^zWu`o^KX0#|M#qo-pmodsUunmOnEJ6t(_L!HNqv@ z47Y7p7P`sOdK(mfxJ0D(IYCBHJ57^)F6iE%wn&hgO|OI0fu$MBiUJ><2rN8U=0|&} z3|1`BeRzl=bl)a>M1z^x@P@WaRJuP&Yy_>ESaM2u=uEI8;i;$kT%O@1!=Vj@0tSp= z^8%Y;gxap|SWnx0#Zr!b>SVc{6CyZ&L?5*>0<@n4o*gV4(sfT(y6$^QMxshI z)UXOU$WU8>i6*lOFu+X9?x&_pC%_OMjKT~PhRDE>HOMHVj*H;JFg_#UFzL2LaJ{2z zSL=wNR@zoQTU)52?8{?+qSA3|tM+oP2M*MK5cvCh`0!fs;_MWB`aE8R+e(-4I~%j< ztrt|(NoW_g1lx3W_8=*Kq>$nVaOo&C?!CY=tbv4T4cFHg!>h|rS3))T7u;_xX@_u< zWC*EG@|Z@38zE91le8!lIBa`QJ+*y+ris>M8FKj2!Bdgj$9QUawY64XBHAhv38Ozc zq*qHM7%Ogj_t4}&o_lrg-g_r6-uvLA`R9(zpSgVZqc0ZT`(fdgk8WSO@~>;}K1S~s zob-K-crLtl`tHl$EWGqW+fakmI!jMjgQ*opH*%sT_PpNn1~d`gl#_?_%Qs(-DO8^q zLtB{_1BH`N5;bKXz6<2b4v(4h7N;v%z#W4%7I23}4{+l?Sir$UpB=*l-p1y^wK(WW zC5%9kiXQXsEV24KOKdeEsDY?TTRM=Db5vMq(jzQ4U+t9WLbTPD91^`9D+OI-nuEx1 z3dcmhFw$ZSfrVUdbF&_xXV>n*YUe~3(nvZL*F-$3s7UY>iz%`2nwBs%rl^S;Eqg_8 zQM*C+A^$_QWV#KG78BA5t&n6^J6SpfhGzyrqumse=@j`9q-WStumr~TXbFqmU}GSt zGD}hAmFr~+)&Q}Btvkle6qk?kbe(6SsfoCvVbzi5@nJ+lo{Ti#qC{7fULC`JRQ}R9 z$=k=!f^&;JZik?IzJ<0u7uj=q6SuPm+_2p#mc9Y6InIEKyA8@S+EXI5Xa&OmsKM0 z`S&(vS!ymOS+}>-0cvGVlO}9^M>l^b@dv3ZjKHxu9}u#Bl5FppjJ<;?69$ z`0rir@$@G`il!zMRShXIk2|D+{S*|ZXpO?-TkAAIH6VqVlNs1clcYpR)Yi#U2Bs>I zNmJw*S@z$rNI4IR^s#YnFOn85zomRN|2qd6kFuibs=+eMOw+}yw7?9W7FAEcpZ*=Z zXaPDwRQ+_9BrJi21K;8PB?;aqhN9&ixFuEY?cgOly}}U7I6x3-&a;vGb$x~rIyMoh zwkr#Mb2(c#{VH8$s?1zVWf}D9K9h7xl)wYfFTt~!cYw9Sav?pkrU@L;lz6CGi+beC ztmDgk1KM?cc3Q#{uhng-*UuPRRZzMT@lO*`M3k+bV6hm=!7srC&2!> z27k+TOZ?syum%sn0-WOyoRl=M&!tc9(x73K7zE*>DcQaHnL`cI%cw<$1njLuOTqzs zCPqmgA7PdgJdCQvLy=@8tR@q&ke-Zp#gckL?MiArU7=)myj@GiL$V%cQKUiXFj1gL zhq;!FAym+iOp@f90Fa4sYCKeAzQ)^P3F*IfNuzNRA^m#sk79C8vDn= z*CP$kz^#^36qE@|izGvuqSnA_ZyeIXpg`8xUBVit6HVqsSbBArWrCLltp_ejMDm!` z6aYUek!!R>PW8Z0seWOoBAEQF)x;wn@&xw{a_rP*xJQF~>6T}>heL7IleQA~ndS## zjw<2CbKKEW2Szq?BqHG&iWk9@^jHnMcaZ$cyQjWdc=LMm31a?}&+Z)mX5q@`ciy>n@4c(@S5Msg^vipH zdS&7Chj%U>4lJ8)i>KmxNcBR;1h%syn2#QAJbiBw@ZS^;OA<*a0B4{?r4aDObheTa zgPJ+i4<+t$C;Uc=5eM|5L0SO){Nc*k>rTN+Z%=@pF(y*9Zdmc>HO3t@l~X;%_%YoB z2&DVOMTlUiA|9=FpgeM31$SNr=hlYa4{M56Q|7U8n!2tdbxz9SB;90+y64RwkeVLW zm1wwHbU^?`=8sIzO<#rOe*U{Z-~IHZJ3st!?$F`p+pp%PKSfTxsY`+7@ULrM{Oj7u ze_gxsuWJ`Qx`L-zG2*vaF@nG%ks5SC-y2k?u&feG@Gv|d(^>L#gLs|SZD|t|?g28~ z-`=W5`-A&$`YY@$g14&0EvfQY16P(CV~BQvy-- zSR!BI9SAlNA~8kRK>$%j^~RPH9;QN{ySt8Q!O_NX3rf(KON2iYNC=xC3{*$SV~%&( zm7zMJ z50EqsTPxiVJQTb`WZ{idIRn%rP^CP3QAMSW_|&L%^u9)GK&dU^o1XG7({6@3lA*G+ z4<#+6W8>!kd5=w24ka^Xv$4`t3AzMnL1F!d-B%Q5&NyN$u=Ivz&g9{>xM%YUSL7D< zFSnQy3RSO{K|BugmoDA?^w{0+zP|I_yUnM^ExdX5&Kob!zjAQ?$Y=A%UcCMNbMxOH zynXqDd*8n@|J{Z8kH082Z2st5cTb)QE~H%vRgIBYf>gbr-{W%VH*;Sha%Cas>SQ#rMkG%*lh_gl3`PTABI1Z%g(AHQ9Q>L9rVNi!}B6enx0z$joO zFVa|~3L91)rNx`dpzcIK-}Z4DJhr{mQJ$Ht)z1*oR!I@j(G?YntR?WZR?G`$To9Xq znqkSwtdEqhth#P5Ly?_aehPE8 zl{}5(rUX>GSaOW1)FvB;kswozpKjUY^LSNPb0-+t+;oH28xL`Fxsyzbdmzo@zO|}w zq7Ffp?>wVX#vQR(YlmA<2`axpEiAa(#S9lzzx!t3+@}=Q(ra^BFcs8Gld+g-djg*g z;OXg@i>JjKUTEcVNze*qTTAu&O{U&-z;rk5RCb*4Ol!)3jU3hzYm7NgaCbqUfv|!( z=Hb9pqZNP*dF)dt@j5_2zI7ML*KFiVwE`=YKyI8&0_Wm&2-3wY)&Q=kFl>=d8=ff< z+*hVc#(TAz5iaM~W{1xI>4Ud_I`Ibi?|0t+@z^^nILvm|7S#>1f;(cst|~^!PGPal z;XC{B>>K~QboQrj-wv+lBR?`2(oqDN^prZbMnd&T%;EZ~azQ5hhQX@uJ9GV3vgD*| zA5QL`6jlLVv!rI?mB;%gT{F!rTqRM(gjF+RKV)?GU0~_hq}P%@%D@Ar4&QyEzU`1$ z#de4$Lu>L0v|FhxA493AB+cyo<-!nX+=@^o>){hEHyMPM`&!-*|`II1cB>WnQ24Dv7W9}NV**xb zu+!8;Ovf;vX|M^>=@tk)Jzpp%sW+KOYuX}7=!2VR$R~O)@waIQ{<}1)aQ$l_LgQhcw%287_5s$2( z1`TxAs6nvGyAMbUYB$+_;Rz7oX|l2Q1|A6##{*+QA;ib>c#Av$a+*&@ttcV^glho1 z?Yc}$1XkCC{=Jo!t%>+CEHEO>AYsdPR-*_ z@NJk%wxYEpW!=8=$-RRsq+_*g+gY;~Z-Ze)b_&U5^5D(V`HQd4pF2mk-23#Mg-c%t z&urAdcvKUrqG4*b2eU4Cj3k=W!(h`kto4@p)Vg3_anR`DW*T(yNl@s{7JmTk?iwnH zaC{y9y+}|g{!T)%pbmUg;;megYB6Yk-=mtm>FZUBZ^JuZk*e=mgpi&+=+TdBK1&vl zV;6wBe-IjanH1lJW_~OCy4=qnjgD!#jI4iD`=WfGfiw!$cj8zR)iWOZl3fpK?V^|H}k9J^KI;;R-Ok3pd4nYtXZOj zD7f&OwnufX=&inG8&zV_YS@A`udv_Dz9fB8xSC#-D=93l*KxxsEjE{0(f0z|4Sw_; zY&y8YezUj~KXmKtr^Ppl9~3_-o+$pQ_#WKw(Nv(Rb+yTM=4~#+Mu5fvYGeapLWj$KTXHCQ+(~{XC<1?n*HUv^iFh_z#DG+@ z_#hq+1i&{D0Q7d`C!^?RRw zx$yq2dJ~`E_F65Hy_|B`(&42f21e^IPI8cikk;EsejvP&%z6_x1l{vq>3Cpte4^;YWF17bVZ;io%wy?IuQjfk;m`w{shQqAx`*2( z5~`rZuvK4Q%Kq~tHH3WTG{ms}A*MGh536k4c`!t}@Z)k)vd-#yCx5 zdix%p%4KQrb+oRlq`pQ7*H;?~ND@?2SLu5wEEbv55#c(dU4ry)?%C*{Q;Vka+_ba2 z(AHiWzrlSbx%9Z#V-kycd?qoUZq2x?KBysaX)twbX|IFUcJegr3^H4-12%{QZ1Y!& zFTH$4dX@GJz$;SHE37mnTS_fj2i~Kw2RYkk018BwP{L~iOM2zt{D&VdynAZ?46oyPoQvZeSo=HJgw&rZQ-R~kv2!T0h^t5C(KB#-?~EG?!aYG^@y zuZ1Oz-(9u}4$hoUQ&a5S`ya64b41Y)b#EYBK)0cmLuFR!FQi67&9p$J1%G_zH0~fn zDT@~zyRBp@tiQ!@pS)PBuB*Ll5jO#Lioxv6g$5A0pQL8SrQWE1a8Thm3=}1>Akp1c9%tU%kQow8xj5;>kfuv#ybA8{X62CYu!8oGU%SBM(oIT^{?C zNEj^eh9%O2QJJO?z(zYn0tZ>*Sm3b+f4i3$r_tEjHZjh|pkrigt$eceVF{~j6Bsz! zMy)~v#dWI&)B-0s#nF==id6%$JTl-n(17o!R*DRAHaZHO|DZA*;s`39sEMrlR$vj} zt%F)4)S|!JCAg#PD`cK-QAqI3twT-36d4|XU5}46u#u4UZy^F|RA`_30MO0?(qMe#dB|CTbC;KKlMPb6i zNH!Vc!jMp#xY!s%z`!+lj*;Of+QxXA31X)##=vG54%w2Y6C+v`QLWDIC|BufT? z9#C|#3p8b?A~JcJ9HPzyB^?&c=IdnMwCrrJ4rA@42r5f+oRot_{t(lNNDtU^fe4-~%gkS| zrvvFe>0PV3Xxm*o=bEFzF_!zS4%afh7O1Bn(q}njBB^SM9$yPjb0341kYtyjc*#%! z`3uto9q0(d!HDb&LNRBFO%T}4Tq8|J*jD5RgO!mxoT6_tt&r_D;SOiGaQxS5o8A2(}=%mb7LKGbvI@S4U;WHFj7asERba^ z!Ciy{1~r*8hf zZTlOkv_k{2Hc)zXOBW7ewm;morFasK&%yWLTfwE-?mA+58HZZGj6+?ej00pDS3))E z0*@@+4ID@9X->9V)C_KP)H8GS70g_H6*D)1Y;uWM!=qq13PGpEqztinB?&+F%rj`Q z%I1?aYV4CGb@o^7rF`N&`~d8)lQ1gSubz2k$Q`krWas09gWbD^`*-ge-iD*&*=I@B zM^ARQ4?g>>=lU+!nEZc!^{cdyJ9$Y{hJAsf8EeA+4tx1Hhwa;Dj;s*G-viCf5qU*X z9($0nS)Y`iRc`98@qMvaINr8HfH4EDC4FcI1`i#T9^=^*zz*`TNydEUy(E$m6_Wus z0Mw`xzRJw77JnE}jg}g-1Z$+-2VA3=C(vFx@hQ`=l1S9#G6ldj>U3$^T7kuFb17VP z(TIkHWio-Gx!<@mr$<%`}YIw2PY)gC@iK1>ed0X(|%+@FS~{xF?Ur z|Ky7)vBoFgAO;g~*~XZiy&9{98b{0|RY5unH9oogw3?5o>gF;bQJFVvTkqtyUU?c; zVC9Hz(7wH18hQ8VC}|)d+=&=#f)(AWb_Bi=660&DI*XqZqv{eU!t|?mM zF~dF>2nMXccNtdwFT;9VNz@Q!=W@O^ zwEA-yLF@DoNGgOr&&lPd)pNP(=GiHf^#6DIZyy&g6fc%a(j^MYn`qmCWmK;7$D9{g z)qgn=PsAsepTci|W%T{uKK|S5voDe2^}n5)J@mKtWa#51S^JMC#mlnx`(W|H>`TqH zL=@lu+t)H{DIoCV2ha}(p*y94c|xLDz63q|t;`zwzW9yT6Iy?X2(LVv!XXLHQA2^` z+Y<_-B9-hNoa`OM5r8KcN$PS721k2ua_I=;NKZZ0!wr^lVic5nI_%&0=Fb9GS(+@Q|j*JYMPy}M5RT3D^U~S zL~BGR+q6(ZD?hM%KTSDASj0|I5MXOsjGH9A_;N}4=E;D+_m6|0HrEh@hm+qjYRH!D zN8!p+IcuWSz`$d(UWCz;c2@C0g#G!Ga5P3ib7>eVW;EYqRN1}4u^K86XBMpBc8Wv0 zx$U+3>pOl{_WMAetMzDN{j*pWr0u*|fjJ^XjyPo(4=Ev0VtGGw3T_^Pj9@A7D9kkM z6?Dl~?oNx8i%*q-=M6|^*`S)B2~vhGBJB;(Ko7T@P8OQrUFYkUAB0z~K0qlal#tr6 zlmi&#JE&z%sG1&n7!2~I##A^-f@^<2Ymlp+63Nnuzts}K6xUSj-)_PTHA+Po9s?E` zMQw+gDM>-t$IV=m&AB0Fjd6H;D%%%)fd$PNL)6}q5>G*izj6|kPh1HnYIZZf4wC_4 z!Z6HJR6H<(MKNQ#US3B3f7aw>^t1guB_2>>M%5nYGRu+)&F!NZ5Qy!}oHWT6H4%%q zs)>Xa+)JXyv3NVCM5D_KjXh;qK4~FAMq#X0uAW1_)7&xY@p~6;v~ibnc#0RO>97aR z1rr!$x0B`z44ubo;H3VHE>`_lbKxMh}z1D?6 z*ohiI8^3dDkx|)psl|ndQn}kSpgZgmk@9{nOSBMaoRzL`dMLT&QrvDMP<%VA7LIGu zG!5>1s>7iiG-(!u1%{j+Yj|S{K!Uc1Wyh)EJPfKvxheNo(eOlbRKL+2lxJAcYfOay zd13z75EmJde2I+UU%(=+Ycgt&=zM3_9=LC0s)1f#gEZrNs&#>_1@kNq?R3B*LyZra zA(42XZ}8b?n@T)`Fl#KE!vel#gqrCF@+dG1mUZkN_QXE}vl@d)bX^InYe}T&N6L>9 zvGIdeBpRwh)yW=V8ELWyHWSCd0oJ)ll9%O$;G5F!JhKdXfKxBKZqT|N6ih;iMb;vi zVJQ+ZDnnynGZidRV4(J`r0}(L_tvQ9-I7R^Vcj1v&e;;}|fZ(RDt=Ob+9J8_#6PP%0}#a3j0YGhWwL z;ACzbF>ytUHe90h$btFu4$L8zvY119sg-68Adc6#Fe~eV?`1&^E790m08MMM6Bi*X z4@hyff!>orfjuNel?GgqWR)RkigxOV1TCy91)1Uo35uWNXOyr3_@oFJOtBEZ)3`YAX zn>g4C(1o5z(m^egIl+r~>K4wfkX9aFCB6uw04o@)x99b69{fki4F66NlupT-)+_S1 z(k1*(Qwgk&f8jG|#y8O7>#}`sH-PwGTO1G^OAv(^RU$Q{hB@I=h(ApqB|m<7>8Hc* zqi5l#Z{PUm#V^TEhfn-;^lkFfi8p>ce}PQ?^udK6KRrx-MSgtc?2j)U{prO|$d505 z^5f;hKOQ^$-^UNb4O3tK^cmi;c=X?gzWnEPk#LTh&(!R_@{3!kfnz`^Ne9o zGc#$qlf~mDPNvyIUhtiZ*uyw>h^*rHX7;7x1P=7>EgF=?Ocfi^p)Jr3tOX-m%0xzGa=hSE?+NGxZX0yW^#ssb!}UcT zUR~mGC0bLh7X%RY?x*MHkGwGd{@Es|ZC);sJ0G6E_vXR*mxE_u&K>$-Zt8`(LqEX( zJpu@vXr*FJOspo<9IVt;6r@X-xYB7M>yRxrYJjqeunvo)48-A18L7N^=1dnI<7up0 z3Z5f1C{nqJ#U5>|QVcrAy5 z9$LnQ0mBtk%ip^UQw$6`^BE;HKjK03Ni8Cz4`EQQb>;CrqfVy}j z*=-cEZJ;F{hAS6QTm1XWw+<1!0JmL6%@IC2J?XvU){AxSAX~KPFPco_z#h4ESZc(# zUId-^Gt|vdZw@Tw-{zLBe9?+=?>Q?7r;-lS@Z?_kvov z7>QQORk*`@CWrT|r~BkS+3KqXx)6phq#*b}aJ>VRElxdPzaDp`UuUI;0gObXT1yO) zBHm8gZCj)`X!|oC3$opmk!Vckrv^!bRxxbR5o%;<8g&c8h#^ro(DX?S9nqS$79!(i zA(O~(4nydg3v}N5)#$aJ3Wv>%LMFZvMn#2i;H=H7{1XWUs z>i70cW`szSQ3|#$#~P=BBUR{o(rj zizn)tEIS5jOqLzAut{u|`Lo9sPQE`kb!u+<@Z8k-z{1eL=u;!o?p9RQW1z$r7W*~W3R*ZS1q((i&?@^y#YYhRzV>Qh01;cMa zh>#3Cxw0bjmv_k1ExXePKwCY)Oip%NY0!4^2!H(7YZHW%e$G+{=+iCa03`*H7pcD- zDJ{mksCD2k@1zfKdfym&HpuaraY1eU$|U{n%sFv@^n&i0X2JK&9W>{Txa7{+{;K)U~JWAu%6V-;N0&>y7cK3IHAgy;+` zrtO60$!=3t=k!x=vbh@f#)jg4qeEfznMUF@wqXFg!)-ICEdhq%1hK`;xw3U%-*-CD zU89H$Fw1>FdPrUP!)^l+(#IF0Beu(6J+za=re?;H(Y!ogJlO&-KhpqrotgkG1JT7I zVMX)ckT#O<5=Lxsy~aEYq_P~p@y8plgLd*c9)1qn`x{bG`Mjjez}An~Ps8@m zPTzPRuY4r$d>s!@!AGazqknHB@PI!`jpz7gxBI1v_8I*71X}-&eBVWdr|Od=wDe8c zC&3k;8*jl^PeHG|QxUYLdtu0Z;zos$T{5IL9LRlC0`xN*D3$=|GVied_ zhfFGpT9ZfaTgE~Z?5(8yOOnVT@20X%CWlU86@-Q3tRTxBvO=VTn}yMknv6zb$%qOE zk*@Y=G8~GTQePRqYU9cgU z5q#6F12Gn<@3o1nD)C52iAL^6p+{4N9*cjHL=R_?%VG9iY|Xp!1Uq*b_;=1MmZ%}IX>&U$UZ+~bjDV{7L)ix##6Mlr+ z*__ZqArEGE@a&8*0(NWRYW2nN8)5t<$q8Yl*RMonL!wDLUF)wW|MHgk*Yl*0y4Hj# z$o4UA+Bq&rhUZOsfbR>Qtn>tK0XAReWog7z$IL*z+6BFFxu3&qAG5egn=I{O@wEes zZ9URCBgs&x9)rzF2761d3*DWk29-GeE!FvE1#W|~D^{}DdxvW*_N51ICbNBrWn42& zr9K3@V_HJ-^;&ccxeI!zVbpPDw)8&UvRqm%T^>qy`2GvYSHCcO6x5)@l9U(Zf*R40**D=7i%O?vr*X6emeCM3Cun$0 zBGNP^qBV@4{mAI@VRSicY+X>C6Lrq(pG_iiXbUTB2=xi`-UiZ>ZzZ;J^kNrAKkmRj8yRB$@j^Rhu{9^H>ZC3_}V`& zo!#tnXSrzLg@1hU&p#bLM@r2KER-j6*>T~GcxsghjOH^LA(D(BZBBIv&36Q*W78w_ z1PPR!xbe9RAH8$q19Ic!jkn={@5q1O!Aos7{&eHC3^ARSr;(7_s!P8@z-*F!Xwj1% z>$kGH`307bVCAFMIH>4&7M<;+6>)V)p4$a1qe_!;VWpdFzwe%q@np;o9p;$^^5ME% zCzS=$KF1~!%e+K^v*IF1<`q+`qnZEn_MDI4wXY%C#09zGTtujxb=Y#$`VvuB7YTxr zS6f7nmNy^ETNIYJ*g7n4G0(OX^S>`45^QuG*ifng%E2h|CJaC6ns(OZ6>PSLf~BI!I`zDWl+_XSPc=B@0{=Id~Mj$diZ{3O%_DKqT^H|=FXfS$FjYG zb%5oRy?e_T*F!t;Fh?C_tMmpsIo@Pq>vMUAlYW{pKtjPCdv&t98!(i0Okj(v4LfcJ z1wqD;qAkH-9zih z9I-K@pwxi_!Lf6@(`m}n!n*+1ckF7bcy@J!*Gx1{m?QNVMxfOzkgKw}3nN4i^~$yz@fS>Q4+!TU9DCC04UAu2>7$GAt%& zp0?UqULDWBedVRwSHGP*bmq=GSMD7DaD|DosS5(zcXmYL3&&0_yztWPYwylaeK&vR ze5qM=LilR5I)q>98&A9*tS1d)N*aQgcz8R4gB$7++%R6P{quBy*#o-~^n|$*xJ*}# zV78lS*s^VYKNhjTm4NMThBMMbP0)kg2Eb%M0oOo{i$)YJ;)wx!imJoD)xk(aNrcvd zwKQOfw40Kl+h*KPcG7ef?29hpAXGtMSwwNkfmsn&)QD$RXdVzj%`b`|2qUiOHCeGi zgyDz+R*#^DVyq_>>&B4A2%G$tISnuo4)Ki1(gzNBK;v4Y^=+f*sshkKwcZdi!d=u( zQ40aG9tJO8r8P_LOMTs4erniL*i}05H>>oRqSb6S!Eqyg9w8!z<Qhgx~kf#pEK=XSr?hbR+f+B3IT z@LR3swu&MmWF%az9R|XYfB((9Cr>q>!hY|Ct9RZ#Jpby;x39l)=fa8L0WeLECgL%@ zB7ln7>a!FbdOr%|y*fp+0Tlo9&)+@uV)Myh{@W{ekDOm86o1uY9?4^WB2dFhDB7^! zc1gQvuBB&vm5TawD66ocK(B*N7pvCkM*W83do83UoH=DN2{`4-odL_IsU#6s2$;}d zL)(kjJ6tPA9zSrvK!jn;uT0IH#ybh=qn0&+pIc&l{V-`aX7&tZg;NT+aalZ`L7w;q z9I%Xs-EN@sHq#C9;n9Lb&j7av7Nsb}Ps$!C$YY3a0dMqNjwpxbEyj1YN6l4Hnkj~X#nqN5J1tRp4nL`0{jVD>Hk_GAZD6y>q|G?~|n3#tqg>zU1^3(1c9tqN|{!R0+! z?y#LE&tyGWRPk8F^Dd8lj;OE_3HkQ%R*$ILmk!>WK6U$tGYemR++2Fhy*Iv_zwn2- z>38R*&(BSrfxj2dURZeR*u77J^IqI}z-4J?7E!UF3FWk)z#`e1b<^CUU?n3SC6#DC zi}VMTd7}HCCVEIo)OZqon_Kq|zI*4xKmY66yZ26By!XLJPdzpN-G%wH*T|MTFTM4P zy?f_>_~PEd59g1bxcAN{3&&oafAy6Ixyr5Cd!>a!YO*^TPlACONya+Vcv4d%v1CV2 zmzL}ahdYz;?ohaaa6kNhA{GjH!HLvJqEfF=I`Jt`tXODRpfJJiy?gG?XNTs`yfA;H z34Iv7LSziK@YbIf-hBS<_pi?%`SZg0Hv)@9@y^@a>anOf^O=CR-$}E<>AiO%DwtIM!u)&~U<`UGC4AYbW7)qZ0MYIKph`o)a$9V?ri#^oN!t))l z2kKqiyBD6DqR=PEB9u%eJ9qad`zAX@VZuT!n2d2@$a2@cdy|AMvdx!(p|=tkp9W7M zV!AvoFe}fchv&Xg%igB~4Cr(+EPFlhKq^b2i(wqooATHitDiI#TP0~KhO`=y8mGVm z)lWADI3+tF%#1 zX08~}23m^6Aq~f>kK9Rfjtd(iPfu13G1nEowb%!t$Z!aWt&*xvwmd%E*<=B{0qPn8 zkw8-1Oa~uGPf|1EQjfPaaNcq(8Agky;TeoPPO#m>mXJc1tLYHiEugyyR{fEQ^2%#( z1&J537L^bLqQJ^HzRjd4579igNH2r4;Le)C_E0?JwN^r5Z+fI(8XORou7n~Dqw>RY zyKv#^{7Y}#zV_O}S063BGc|YU%-r-zbbU;}IyZf->DJ-7sW0ZH4XEzEQPz~XL9>HjUw8q6FOMuw66%(tnmgV4;WEJ_iZQ+!*(l276t1~fNl?q0~wQ6EOH2f1+ZvFwol6JPHEz? z((PbPZhCOk$8C4!q{YP+`U_c$4to={z^lE~`#ibK*j>8~7u6~*0D0_FwqYe+GvPK4 zEwz_fd1_7&Q@Y<^%F2$nNLHkBY=St9=9>&iRl*Xa_{@2dek?1X4|sXV7qHISCV2`Y zz(6px^keM?!`80{=&A>()3@wZRSDHF`1#gehEmy?Z_Uw}Pb^rSpp|Z(3{%IZpp=?~ zS>ud5=))d4Fivv9vUr9w%#hn|W|&R$VDUp2BitZpK0|mf0WEgsTO0HS0vdXnnOi}L zB|K}9c6ZdVxZ_9->FJgMVQ`1%P#1i^rj( zB#$E7IM_C3jv0Z6#Sn+tWx&ed$xaQ&Dr>$x_Bl2{vuzw^E68(Gl0OVGC_8=_ji+mD7c3F!v8K)x9;g^ z>+PBBEhB@B^8L)qK36kQ$re?O==uhE+}CMS_oONhz&4hqJoo_UppG-KY}1Aby8`1t zVinE-kUdP<8dh`PaB_&|{GLRkmOK!Pn8(d$-xfcVE%=@kKboB`UN4@9zrPDE7LkKV zzSAR>`g~=p(3Qn8u~w@=acRJn8?NulT_O!kcBi?68ztIrf9mypX4br4Ex?CBMBvuSc`x5 zkyrk~dgU9kYjq@RZPd_Aykiv%zcb)EJV=$B+*mYuPIj@uYFZd~SOn z5nYL{Ng{)6lvxN=hBA3%lmr_Nd$?h6nkNHUUMSdPOE2P#z^~(?)FJXCbRsyRU_B!x zUOy=tuCF>a;%gy0UxImXL_ZnDx46Sn*~z7&YQTreEoHQaBpDnV zQ#OW>`@SU((;OomhC4dCm28KBVG16++OW4Yq`XnQ&it?5GHrT4${*((cg9lL$$>g~%f1Qy2KAFkj1;qt=!pD%py zXOBkYF;P^%7D@uF8L&{RE5d@f`35X%BH^0h(g%)`7r0oWTt4}@ET8PP=z-v-<4T2- z9?(e5zcivsti~>Bz+aFugbj=2e~xz9wY~y8D(`fpYq6f)#^eSw)7B~S1=o>)M~jPN z5~P&ptYi0Zo1F!1cXYIE@oPh+Z*?NvFS90@WeLmv^~Nod_s7fNuY=lX$>9)q$SifI z*PQXNXNP^J86~iWUJI`UEBy)1#+xr)z8|APL#QBf4F3e*&UV<2JK~9u1ZLe$VCrn7 zhPbN%3^*_rImUZZ7L$sqEwGlPGs{h8Gh#B9X3a*i@;HbETGE4&AF@YSIZ|Splx+`Q zU}Ers@KAd-KqH!vJl1QH){rT(Fl+5B>BK&tAe|CP*#te43|o{NC`-^tp;y*38U4Ox zrqr15L-1y4W_4I@inR^Fwm2g8w{=dWj8by~#7-!HCV(b_bKK4W}j`<$$KsF=ZF=@Zfk zsx}NSW*f3F(qc&c1tVrF$%+CaPL`ts+sZW?WK|!xrLz_kSXi)ZJlq9}UKubuuza>r zGlTlDYmIppzC=Y%3IB3J!nSHT1u@KETbf`wGUD1KCmh-auLl1TFm*T?##&WOtLuPI z2ZV+3YlDu{bkkXI-k0A>I#E|hi|kf1jM+|Lng(Cf+D#bP9CUn(N4=caArGu+$e$P^ zA>_OcH4fbD+iY9(>?|53lXY#uwb^vZWmaT+NpZC$-NYg+THAKeir}JSt#5+nrn3SY z_N3J@{#qmT_sq!3n(?4ou5> z1yIA0cn3@;gSxhBQCZVVwhVxRmvI|$d2gc!=*&5)dho&{m4euazlvBCR8)zD*JOJ< zA_A$b!j(1HZ~sdI_38?|AS9dL7ta;HpPfn)d33dS3Hd3D-xSZzzFWLVwhwjeYAvoG zm+*!#ZoUQ0_g-KJh)5Y^}esJ~>@`w`r^j!6|Z{W7K%O|tb zq<9j(c&>P9_HDR(i<(d)|2TLY{)q>lb2TKMK`Xsb-ntUk8s}Yl6l@G;LShQPO@=3o z21=3YMip0JZH=!g^suHxwY5Z?F+WG-*}pZzZn!#3S~R3WZ^e_LWP3+PcNs zbSGnaXC&FFMbyA@+QqoB%j3ui$2>V#{t#bLMsXz^{r}l})8;m=Gk^4}oN%hHl_goWY|HV!#Y-$(jwHvjwRn*wFIB_; z+?ThgSE7SMhk-2F7On0o+=Ff9WNuY}x!g0!#-lo^_?`Fgy>q;FqLhTa(!SxAZ-;V~Jo zz-#VwC{jvb?wx|!-+uD(k0N>mjAk`Z=)ThZs9{bNM6yA%T@+h>fX!dTZgEqR+hP8j z`du)i`us%>lTHoS$Z1|bGBzq_wXCk04(!VoobB?OFDiw>qWR z<(Onl%b8ol_BZt@k~3|`EroIb{1Z_TNFpXEsvyM#Kbe`2S0N;=0?H&V!~%_Lk(e!B zI=*uH4b1Z_J$d}rnu)4fy5n0Pc41&)s|m+b$l1kpIkoj;ty9| zSvmF6%F`E?j=jEeB05m<(&Ha3oqKWV)1y8+PHr62kekLd)RdqEf@Sv9dh4c#*}MX? zMoc?(7j~j)VzYywML4vV(j}_vInTchI?o4Lu1LmM*#%{j+ae3ea$EyhdkYO>(j7a7 zJL(q_Cw&CloK9;s%PA>{F-efntyX+ho-MQpDk7G71`$@Zu>JpF_hUryA%yv2>3)mZ zITOxx!EWkM+txtoYn-QeE^wyJTEF2>QE-to3mwL70~n#UqMLqeGgK8@RD3p+SyG6r z_xg8l*Gk2@{^400hOaT6O&Zl4$IiM|p|}~zRdF5XxT(lRS%@nw`^J;n0c=kxwwWMy zTRPVEkxYqN^Jn!IBK18b%MY^WvpK@9HRKdh1Jw%=%=(OjX$J&DX#=BvD}f9}?;`8lS%sSslkpT`wgaxGW*6})&QCMS z48xw)r>UV)U*r}E9(y()NBqt#%yCB`qwobPp`~Jxn_U-Gb6j|a_c;Wl|9(&B6 z=N+InOk~;2+rR%kseNmd>Xey=W9nvM_V>TPtBgTmq`vxW*9D4ePNxT9G_a}}DXa}V zvbaZKqjmBY^AiR{=0ZV1r-W317CyZCtuO@~jMP}hY19dQ)D5PXjvKeYBneet>EI) zM31m+nFnS;-G2Y%ORGoV&0EvVYCC1y4NAkvF|rNRE0)K>!x&wk_L@e{@;6^aZ+=E0 zC52e3<+{IL`tov+Ko0C22mH4|(=xd4Hb_g+t((A-QmUSV654f8Vb=@r2;Gxb1zaEt zo&!5jq#PBcRW7g1(0vlpRf|FMB!Z9kFw=E&1`O;hc&dmh(FU&GL$-?<7=QYipGXgs z8@u}f8pO_F-c|(xQ zyqJn2_6Em!AV|E#5aDTC38(!2e?@X^UX2A(=`>7q&)W{GSqD!TEQy(^?Jz#wv}g;& zEW@;5n7T&e311ep(_i&bTPn3aN*u^dk~dg})1$lDDui~M$)nG9SQddB;r{b1y0`m~ zZcf%D&Q!Ir5LC6P5Mc2H2mC8}s9J<%%-E4j~x6Ru}l^(Wm4jz`2pvW6x z?Y#%5vk{C>6BPKcYZU0T79EbaVTtQHDkIrcWyn~q*Rliuvw=|fA3DBSP1ngV%THht zgs$iC4X^uP$}ndP1UU0uiEnB)0ezY&F)UUxNXgY5t%}=Tf;-E>Z-mt`cSr_Zqca5a3YIs_ z%rvJ(^A~)^)R?auO%0j;hleO3NU4@LfBx9u+FtGV^1+gN&k{C+J@QC5!(PgMTGI5{ z%_M5LJz{IRM>KA8^A9kxT>B27_2DqTI7MFpFFyz;P9-KJ0_;-Z!$j?7Oph|d0(%FP zic>W6ktco-6i(JGWqrChnP{krlZmFYj*F<}MDasLUd#Ky%aXBpEEboWf|tdSPbvvN zsvyb9x!S{e7_+*ID_Ca4+xLlUcr1!}u3b#0$gCO7#`9nx%z{0MBgi4ek+ylA^>^m+ z7{(YBUt6<8j78t7xvTcDY5OaHJJrUp&+7WP&+vpGeM_L{FG|nZxrHwmzTy2>R}U?G zhdA5|7qH)|tt_X7mlm$TjeoXqY2j-a$m0Y~V5`|t>^fVx9NnU>?+)u+#{q6C8l)I* zc?Y})0vn)up9kn8=&JXFKWeq1|f?1XZ z7Ae|3s{pb|Nf5ETIIaq+RAbusEt{|+6v42O>uVN^VF9|Yy?E`-Yj0e8_1fFy+Dq3? zUpsN_6(S~-crqzTeDST2 z;Vl`U_hTQT!f!*t{&dFi4++yLB<3@EYLt7iM&>y|2kK6 z5IPRi*<@HNSV&lDkm>m^^eJX>7FZpb5~diYyaz0|TxMh%+#&yG4bI;N&nLVE{x&5B ze8k!EypQ$GAAIo9qXfEU(=0f;YRX?rlJFIigkE71tgt{K`v1Na_SV%S;EY^A--9fi zKo|=|_JC98?#eooWkHJHlgdd{btsdohe)pwz0e@&6*dm9HwmsI#jubJSQXJx>A}3j z1h(BQmQ#alGgos|Z*bOLx&}&*ej^Xb+l;|ajz?lOn;0fEMFZ@6l~K2;w<~DAX_fzV9Xt$c#dpwIqlIdR9n{v? zR7~|%JczO?`I6Xb=%|b;3MqLbXd~lD2ApZXk`oR@*FUtUo9w~wn9Wk26g2L$5gl94 zlODGmSu!K0fs>;-uq%{(24YmaEE$0jx9gPCITU|AsEW+ZHSje&fF%X|GZ4b7^q$xph+gtc>xa@Ii}r$uCd>B zGzgq!M(P7+pslAh155yt*GLBB#a--*0F#oSCIk_-c3;J7nWGV-!gun<{P>m!(x5y)7(IoJ&^Zr8XQoij~+%d5VJ z9E7FAZMp%g?451&Yi^>3WiTf>&j%U`7?$2GbHbl9MyBK zj`5$@n@Ilgd z0CWsue9WITreL&i>aoXizVKvwnR*ZF)0Fg0Wpx@oPv37*V@hNFK1(lpYleaa%1W$0 zjujf71mjqm&eFImX*8o_g`C-|W!!>m!=A8>9VnZmQ&I`9$)diig+~QayXW%9E0zrdnmPZl2Ti!zsS#M1+4ViOYI_fG>)IxL-953 zD`C6#6@+(H)r~aZ^k4}ZB?YR}G)cwoTQA9%2?B0qtO9(EX51~P-G;EpeLBI2Be9F- z4B3YBb$=S3#Da@(7zn?DHR#b}{bCB4!VrrKO%`8FQWdLma*rH>1z@*%5IF-(;RbJ#6@*;z+p~!+L`ta&I zrFFWenzT#MIon(Cej0N^tZ{!6vZe}g`TI!W5Risu8?}rTTD_-bwBlzM- zWn}TJey&*4Sw4zYE}di#0_lMLR=V3_+<0M1CYI!wJ`w9uRzJ47Q>j&Rguf0aC?zJO zS})`GV$2u@ESs<&XLUxV*96lj+-!FxR3Rndf1n`d{0IB?VJsMPYG_Bpht-&%f*whx z1WE8}mU_DTN9V>EtYL@r6hV%KwsATp zV6)F@V9=BU8|D~iVguc5VPK+6^>sGQ!I8kGIcOSraJ}p;v2B8pW3&15ye6B&OXV~M zzZ&eJ{Q+=9kByFcu^rUuU@hgak(>i29`xfSipe;v2%dmR)pUJDtvdu6ri6Hat8vN< zucm)Idg-Sp-Xs4w_x4YJe5D;!ajnFDI`+no&p+{xFOL7?;(I?m`P@$z-h;!)tLmqB zpZMwBkN@%IdmG5nC}~WpC0obBz%DrAmU$H&5=RMMHg3So7hyXuQk!535W8in{WWl{ z$uR-fs;h!u)}Mh50Kpu?^V}HmXSwYEjRTm zJ~}c(F?DDXkgPrXt2TS=q*zQM1MYrIJQ1**o#;uB@q7tCU1S@3<`93S<@86j<5TqY z+-$S&U+wx_ZQ%0^os8=g;c;_jz37CgV*|i8FQmDZgSs`Ns0#6C23J7BXo4oIzanw#$pbl&baIRlRJE~x=I>eC|;P4F9^DO5p`WJoa z<&ar0kW>ns;Eg1gGvL*0JgorFlwoEy9Tg00tzt?{wV$<>aczfJDX$JCvS5A-mat^! zeJiNCfWk^el)-P=$olD~xv7F>Vi{<=g^(WV?;-u1EOi9`O_MP*H&rzCJejry(yiI$ zR_I1{u*u18%QegmAo~0Yya!ua~M$N98K}AoZBvRPD0;Y*9h$-2Z;PkNug|;%Kgb;5%@;k;J zb(zg0iguYXa3t5d`n@X9p7Gv)$t33B7s({RfoSvUP^_BWqZzYcN5UMqMk||QPmx-> z#<=G{sug_8%_qPbz7x5b0A*1~i2(tI&0Y2~yU00Q5oA5DCsK4qq!v5e!$V1g^D{O>#(|Ge>wKyfsqFVITlZpwT0KvZ!{8U8AYK9~k5<6VQ z1W7Vk?`TA7S=00q*4j;arFIX9TYdr6*^@eU$?bBnz@~GI+AY8h?J(_G$zSO0$D|1h zNuwfiF>F6ABMenz-#kW*G4d@hil#Z|!E%}ErO47)`&g?HXBCkyz!bP^ozmk-8L~Gs zKH6r@(QEe|S!Tg@=;)QC@pMcQ)F2t?J+!|&BYke+!_2|s&a6q(ieA0i5b3szp&`SFIY~0MZmo^13e+K5)vbjM=8o||f{I%GLz97~neMCg|Nj&axt zY1y<`O^!VTKbY-iZLn(kdpJjj#&^-KJh+OE!De8+{y)F!?Cc~%m?(9q;cUmWD0bp| zH8@zY^qK=8mn1L5Nbn}xF4i!)Yzg7{(DKeDxPMSh+nFquU) zE+`4XPuQ986uNmY@ASRmFOh5vo| z?3YW=U%1qoFg~xP~3hH+>RPLR_tn3*I@4IRh$o5-qI~Ef}0sxy{Vv-~gicBy0v= zU)+oE21*3B5XIJZ&O7f+Z65d29Wko=N=Q_0k^yRAKQTS%lHvI?nAnXT%MX%o#j7Ex zcu26EIEIj=V1hbIZ&?v6UoZr6It9adQmgQnBAP+FfzJ=a-hfONpoHLYz#-O+Dtd%z z(smG6*yjgzDvi4~l_p)AiF8bcUA7);qIF6qGBH~=+pJO*3*-3yPj(>?O=MXNKRjMb zt`#+TpWuie;T_TjQW)0?27j6=0UgcEuS*DdCjoR?skNE4Of7osJdqL*6;VNqUtf#4 zBvnmM{yL2IN^ndLspVAU&$e7t1J1*iK*k-5=Cx^XVqE)|5JayZ(^@25k~ck+t|-2S zS|wxzD1kv(N!|#P-8iGDF+|eY4sIRFrdCqvo$z?(=~N3V&Hu^ORVMmYnHYMNNiiYW zInQ1tjWa{_7zpG?XVZ_-}aZT!0V9n|DZe1&wIdD=* zVQu8m&3+6p{{~ zo&(q0p$9a3wWXJN=fU@$1>2;Q)!-a_m+1wLx3Xfb=c_hY2~i@zbA8CrcqSlE^CaUs z^A`>9=WxKIF3alC&6c%U^qY{4Ey$*lO$lo2+;Vx%akhSyPOuu7`KJn{&Z(9iXl2zT#Mi_UTd_o&NB zS9fX50B6!c{;esE%%91Dh9LZVh-(lX-)6kehP9a#$2l4wQBbepC&;f|*nr&VMQ~%R z=-7UqylTOP7_!_Vkyb<@83fPRNV5}pK;*A)7qMZR_{e^zEVQ+Vu3aFVelK{q32h99 z%XHFbJG@=Dhh{BiMHg|Sz+p8;91j^IPNY?E3KAPZ<;&P$_1a+Z%_5{&<#ii4X?0eH z3EG=CM`!XZ=Tfgnhi8GejaMM~dNpK%#PfT=P&{bD#z&)DBfDI!gxuUU5O9%+5E2ND z(x5dog5dVdpQR`@B>pwIEe~(IhL)=gzU^fA5cW|QXbaV5m~J^}(UJzgN#@hm<<7EYBKs4|gA?7>J$(Zp8iSG@%(hwd`1=Uc^bwzR z;ja%1TOgK1tPeR5e2)S%o6#Lh1C4E2u<_U$4+F#IhM?i2G>;lCaI>08 z`sXh@h-8VWpXipLW`Id=G9%|fviOX`s);UzG|@p;Ng*lTNo2$k{0JKM7M=6G4&Wk+ zaY;zY;6H3GE~3lpSeS-`Q>z*%uu=vU1 zmy4$sPk@xJA6q<~zW&tWCokW5^V)Qyx2bZdh@U;VhXv}@BOTJOL>WLfOh@@&#P;8Ur-j%6uq$8r+0 zWAPv`dh0CpYD!8ZlFgXS5PFR5GN;+7YfR169Mv!0dnD@()8@;2+Ed8^!~lj9z)Sm+D}5kb(IxEM0RB`pnd*OoK^G+=zkNyG;*h z4mqsEb2nfgF#YpLsa?-mS1mHy?1KgZC8&yy@lfBwvdgPxt@I`x)K z$}N)!uP;lXSAWMY#If*{t=|N;s9|$cCuhQUaG|dU+cVG|s7HCo&@h?HQ_W^+OBAN` z>8xBt{=@$ZDXSbX6&(b%vz>J+xH|eqsv5j9tUjYqNlvU%IVC+%9|_A={2lbs`;C%_ z2_pq~5u((oFP(ht){)bAfJaY)Ubb&Zt2lrGV@ zZy?ch87Q#n#9DRE>oR=`l zZK|yA3HW-`Oiz&qJNlAfHu8jjaDzTpdbmY`MFwvA-|)7L5S#SLCSfubvvgaFQ)CQN=_-|ki zQlLp8)_M>dJKHRf8G0>aEXT@~=EFwutI@$g{A#qQ=~}sr9gJHdcr{pQR^R0sE%0}W zDNc%!5d82*Nnda=8Mb6Om6k*yrM6s(;w5ZhuNEu4_yhfO1=GxvNvb(T*sE}%HA^&z z@b|&{lSD;G#^!2{YF+X>u`q9Eg?gq$58`0P-~z%q9gvKeh?}`GJLSf{x7@_}h9>}= z@7P8t1n%`-`FpQArN-%2sY$*y#W{2X?b)Hh`VA~CD$k!NXa-GJgI*d<_h>mx;!vBI z`uy2aS<9I=r*kdS`Lp(1H`BF)nDX&|28>tC6a+p7W^2RitAAtdiaD}d*SwEy{cUt@ zB!LjxdK0)d&d3kUfT^2C!G|s05{PvwEvtgsdZ!5j4|4XnN@_k$`tt6hE^Uj&yROBP zabdG5xY1F}cCliK8Zke6Z%T>#jGWNBshkqzV6pVZ`;;BCrb{`KEo%E2bI3S_jkast z&EjN8zl3Jx+Mz`k<D#P(#T}p(o89tNwCzs z(sw-zYJyqt6WwY}rUX&MK))dRy~F6$Nhw4iiYXiMC5Aa}V+jHLzzT<7HnCXFwC*}N zRdnMLi3FxJ1{%CTt?1U|aAv@r^527JmcSvFJAEC{czd0SlxVT@gunIuV)rNnyWe5%s7ABrc%{v^g|7o$wPU@!I9b+ z@{GC|>AMJj9X(c3z~Xf1ZOTgqlU zXCe6U04qgL7xb`MGdIP2$vU~Z?a86HJxK!F^6qVW{n%E7^kGPF1=ggAFiT{X^b~1s z{$jiBf@`Cx4y7`XyR;ul2|-ae!lM%C`)@C0*;D-QE1wc${$T-KPTe^4 z(aMR_H=qBuk#(~(=vt4lvWsNih|IUswAkIk+OCyy!wHJ+W=o7CEW|GfGJ{Q956AO0$`%;MnuaVTCvR(+m$)jUmv zHc#cGpe9;Q=j~-W95A#`)`}#5b!Ym@H;bI0X>d`B$Ha6`cUpr=~Ob& zlUBRqsdQhTn2Ibt&g7^Wu8-nPl9ImWS$+^se2O9@e_HwE^vWlf zmyTY%_4?U=e*H&(z3Z*_4&8d^eZ1_-`zxpau=1DJx>ioVaqIXKU3e#5u~b(~ZM*F1 z(sQpaoqK-iwdbR6j=Ib8EH~v3MmE-xN^_|tg+%Mv$|I(;20WZ54`aLBBae8b8eo7? z(law_P-`yZk&W2AoQ0z+^1iJ#A`@^F*LX-^P{>n~uPoh6f6w3?{sY&5SJ zSF^n~v51gGW$>oF=Cw6r)mJ-~2ve^lHu6?s|3=WttL_ ziF(W(%MFb=yztV(pVITM&nn*Rw7v9BNzItTgi-phlq+MG0y!yN}dN;lBCz!g&Vd^d|dPSPU?bh*joL_|F||Pc!@NLqTV}N%bka1PEYUm{KlV_ zjy}D7`tZ`*FL^^+I`;a?iPO>Je_;D!?*s+`t3pz$MIB1GnhumuQc#jDdpEt@2gS8i z^-;}hfr*6!^d={Xp+1Ms~xE~OIQj9FKAZN;ZKo*tHI4<9e=y} zwhh@cd+-auaUX91`!MLw`Z&syq4hLYlm?9JD52i}dG*bIeE!0{m~OL}ZgjE+kc<)& zl5%LWhK^w*D9G2+n2vh2;hdm0PZ0x#{O4aDd#qo>9>qNNl>IcNT? z-w!{c6a3$UWTyt_CT2|1ZP977dRfi?9)zE{`Ljk|%dx;abG)#6kr(3VQx$^0!H~;v=($}xveC0H}xOL+5TkpI-7v`cZA1A35@2#g^Tz>QN(noJa z_oERs1E#YxTVqK@Mb1E3ri4TQT;-nD-dwF66{ij4qN)u#iSI7l&>e9&Hxq40T!-*u zeXwAh_>>R}gnmQElVIW(-dOnN>fwdYuRabxhv4hombJ9zox z)x(h`Ii4$8+SCRb@l^?FFx3DbT-2!G_fNjyagmYAiG@FTV#V>U1qiQr0j{gh&oz5{ z;d2Zs!W&>W!MB}(4^(1^1f~fd0l55Im@>Hh>Qiv~t9|<#+ADV~Ro=q$;2#tW0_Wlz z7~C;9@f}|WUqyZvGCDgzjl5+V+LY-70#~XYiW1U8kq}#2l0tK%cJvgr94(kNc5p9a z2Z`Z2jcCU@9vTnqb`{*iaR+RjJJUHHk4eo-!T6v^H_Wn4xZ7A@WYl5}ilRG)cGX*# zLpu1*re;^I%XnN!CWHi-n5vK0poK`E1d3Tmr8ZKg*KfL(Q>;W{y^2^dl}vp9_{(A{ zrM4kL`!3CP&`>I3QO&fLXC^7Y#pr+8rp}_5%l4W%$Fv5gc?)yvu^kD^o|B1|LHQ8r z#L|qMRtEpDr${ZDbC|^|pELP9dC15!YsNJ5Hjk{2Yb7KqGN&1`0JfP3z21En<`e}v z4z8*-)e4{4c$Vsu@O)S0=Ho(0gtRLqgliC$`i;>Ji~kEd+_$Sc+toMu&h|NhP44rB zFBd+?#_@%B;Z2_MAq(f>*A=+_BH=c{6`oi7{>hi{vOjVo0l(n_rhyhNdDm1v`C@}l zwwc%$-iJB*3^v3!V9Q@P0fYGlM+c+%8V3FbzKR|@H>W^(@7$o4mEuAy)-rXcFQ1~( zHl$(dhau0BKA5T!jUFl;V|heJCZM%)F7`&{eVzE#&|JvKoP>OjL1!bh7WJ^uiBqJT zS5k0iq%h1hyGRC1JrBQI*u;#SMSbOxosD6!*GI+1sZYSvZ)#7#z0_i~hhWsyD}YR8 z&YOaO;;==IQN3IwBQ(#Vr<@yP7p2%ba4NC&JH+AfP!u5*+X#O}wKA{mbYvAK)3lrx zTC(6VXM1$&!?bC^gU4vKTA$txPwjI0`mx22kk;ks;>E>}7C%}1^!m|sA3RxI$c_SM zK8N2={_XhnLmpkq(Z#P8Pr*e`ES{P3Uvd4J23L@6QvCil3pOu)03$dL!#M&!$F4s~ zU|zn0N&3_~j4b8fPSvb4xoV2Yp-mB_hD>e@If}tvY@6-3#^BELWE*UM$Yx$}xwjg< znxe;?oMdb->?XtDRxrmGA&Cymi~>mraW!;=q>(@zq}#Ic8pkTbQuNv(f~wcG+$^s_ z_uTY$5Zzh>L|s{A?fEnwR<>r+3wLBw*NUz_`(0yd*ff0gZECEFKg(Z-5vGg?v_XE| z(1t#&?lf5gicPPv%Xjq|=!M&JV2dE*3kS2(O&g1v@qC-9MX&h=hK9o!$|6%C*+#8{ zW>$mkFZzmHo2LzAf8?|n6Xf{4nzy_3OrS0$gH+dVTg6+VF&!_QZ~9ckg?Q_bwrfzE z2&rkW!-gov;zA-Wz>Q${=UnY!J#$yx#jDg^tKbAa^28C?4|FnK+i+qYhO$jGd#CgSQ`<@dgQ|NQ!xU(~Dbd(F7`_R}~2_{7bp zk1fCbv78VSVp2|RdAm6fR){Zs_1O=X&aOQF>GHW3m)?H%=3ky&zWD0WsaKX>K5^^N zrIlk(M}7t>6SXm+BUx$aWLC#GELB=IE!MHU-W}z2^0r$lVb?_CnyB%CK$h75g?GOe z3L$1~vJK&gVE1BnbdyH}OGQ)!MG+)FQez7)sq7)5ZX`xPClQp9%rJO|rxSJX=>*@7T!`L)$lqgIJyBl6|d6V56=FlzofoX`|eh)eZ3rYKc~ zKfV3lkDp%p@rjeBRYa^@UBV!mBQ(8rsKHR6!0vvxEi-{Qk)ohFK}=!ip(U zotD5w%7V7AHFDo*rYF-+u<4m%9v>G~k6C79LMt#!*k3LXa^HRTjTA^vkufsNi~<}I znuo6`5>FLkuP}Vtfp6W<@JkS6)hJ9+Jmco6MFlbhmoQ^L&BFKFEK3Vus3*xzR)+5( znbnI&u?HK*2QRm+efRxj4wm?Rw`q;vo?3AKd_{h29@42rEUl>dqDl$HS48C= z?n+_EH-^y|LGppo$K|T5`RgzXT|pF#RKP6O!UC7C29O;%WWM+@j~j!B3u4oF?tCtA zrp;o=J!Xz}_LzEfCY}b-;!d%#10yyU6?ib>MyeU^1&lXV<7ei<)&%Qild2ZP-}4QeCFm0N0we*hofL?g&FFg^1l4+)|Edjz5T)R>mRN>esTG&uUEdf z2!r|IyYoMM^}?+q(cuP0S>7_iQ7hojhC)RkoRpA?ht7|2OYmXSnyQcDb<^Aw5}SjG ziDqVwT5P>4uC3AXp~l3f7%o_*XywUXfsDAeouzIAWVMcwhwCj&-~6n?;3q4lB{)!V zKLtxdQ+SciD%WJMSG0YiQU}dba zM5g`N+IKM9W@LmrnvrGIt4pz%FU-{`B^o1%!TDw~QW5cLBenkbf+ri{%*xMb26*XC zE_x@>W}(R`$Um1%1}dwsL$fvha<%X= zvTSy-et2$kd%z_0?}&{z#Pvm^reiT75p2bY(6Bga!8co!6fiT`*pJHu6UiA15q31= zZY7X220LI@rNa-?eL4ZR$KXU!WE)P^{b|^@3*f9$;((Gd!5D3U5k3_YBnh^vSX>ZQ z9}=ACD^ujJ!|Zx0hNCxDv@u}j4{G2tkZqG#pbfSQ!}fjBTLGpQ$hK82$QihY;bN^e6V+{N$+wp92e3z6h(^<3JA{I|EO&_UB zTN10)9{hDUX^WyD1;AE9>&~2Mm*&qn8VqYYKHQ}e)hu|4pVl>IG=kc=V;+8Vu3Oz9 zIk+-<-^VVqGzHJE#T#N*jn25w=!}OnIzb@d4F-+Q|1B-v6FeDXlAY9W%;?#e&F%44 z5pf}=Y(y6#g2cF1Fv_Owq)99lBz+)=%1E@W-v!>Bm%{{;!*&`W(fpjqe4lJJVFQ9v z1u@lf5%)tzUdw|ZrZwL55KqR`cuG{nhK&}L%9bR?QxyTh$i?_v%~5@2OYNnVR*Nk! z@45;WEt^FGnr#ziiHFvAFw2^4p~p%?@P_=vo_=ch0WFnko17A|OHNJ8qL7T$Z(i&2 z(sfLjQ?s>@P-y#Wc2551*T;H#ZydSO_nV!%T5`pKh9 zpM16Q-8W0myw}JE-7y-lL3hyd7QvTBCA4Q~d}#MbX4gZ%X^=k<<@)@=#D6_EHTz%B zb@X6ofC~ea9q+$uUs00F{5zLvsoN1V) z`Lk`&Y+MJDVVTniltn#iSi_;D$8rz0utnx6Jo0{9Kk2jeL#s=y>E96DakqwvyK)iR zBUbq{+Ds+`U~G+38Xd!H#|+bKTlA^ER;s;J1nrf?`YMEcgwhh+LB0GbBik~rJwuB+ z_pcui+f5eCDq?cG7yU+~g|b-Fo|P2evhb~n(2GKq8&rjaUKEB2X5(a=&HE7Sbs9Cu zxQV+a!W52zewmFt|G1Iv#Ch-)6ErS;Q=3%2#W6@IhGzrm=z%|pA$73aV7JdX^%k91 zptJPCb4#y%0*2kqS1#QA{0YyDYh=0&?hKf2gV;~B)8s`hci((ldFDL+$BT8A-#!vq zOt3=;k{Ux+VA)^z;g1;AZ0KSWbS03MA|%CnpEx`U;d=|`7QXbb2QMLG)*4mU=WUhw zSa=x`q7exC;wmyAM-t|cfyWl!j-I@99kUh@K5}7B90}FrAZ$p49X_+4I@&bSMEsKS z%9$+7F|eUnD_cy_rDTU`P^VavQjV*>=w8TLk9b-H7b;mVpgUKpGtNqy>1erjLwdpc zp1**AbCD&u88tc}H|$GlnKlC!*7Dgem!7}Sem=~wX;Up1dDfR-JhO8A!t%3EZy>4{ zN18A>cJ~U~#x-Vn#HRx;wm7=wtYJG88yhFXnW0_XyMOabfZ1p;tW_Wp+v9KF;Tq9}`*R>1DIa=SDlB?g9HDX5 zc)f`@twjOin3&k*@i+XR7ykRos~Rc(zvuq@%3JUR{GS*9`|@`N$b6@!>*T(^63OWL zZG)J;ZQM{g@$NkOdW0Fx7)8+!&fshG58LMtLP~;i-AJM_V-;LHYDWY|SIcRj7I%V9 ziJt9!7&ZqjQNvc#I1toq`W0D-wGP6+zDBHr9hxYzDR|nLRrZQ1*WicbUxHe!AKj)a zz!r=aB_${UR<0&})|!#Gd8)fUEk*Ow7&9)gb@KQfJ()n{qK6t@4bj~XTa3E!3DUQp zHzksZqUG8S$+!isG8&}zJ9Y>9q<0$#@?uuAP?9332|*GSK@@$-!%n4fYC>s}nuw=W zNr(lk%Sg4_m?zr8y+uD82fPoqp%KidkzFRp>S5E3yns?jOo)gkAQitvq;;x!n^X(N zv5->GghWk7Txc&BvRD?E`J?H=OqxXRtYfP-8n~X&@{`;+K!Y+SVL*H znV06>KH0W<`_YrN0Vl0W3lx$SD1osXvPa%?99%(Mzx^oUfi=@yDgG9hhDT_!M?D9iQ0 z5m8c!1Iz|vFjsWi4xLEk>kV3#HL(99$UTUGKlP0_|i|W`6nL``TaJG}IPK_yeSE#hnP3)bU*xL@3 zl}%k*DIOEkJ>A%+D5q0OES*r3nRF_d=t-;H@l?96PfSIY@wFz*=XUeAfbxJ}+H-6b!oX^%%HUjp;o1}hy` z;C;lvd>hM})XH}a85ngk8u9XUjP4A$zleWWhom6cqD08&8(2 z2Bvoqa(N~M*apI4y~48uH&m~3chLL{a{?H^+gq9k`PY_O1uVmu~SJlu8)HA z&E%y@LL!j-Du)HfF~H0r<@pl@&7f(rhvuk_R7573KSNAXoIjhjv^*!gaZ8j;!}|nZ z5=XPBVOFlS*nvjU?G84|c3bqIrbiO(?MaDMM;clGiHVdT;)bZee(1AtR3GzF=+&ks zgjk@uEh2!2j}cija%K_R@wR?{SAc80%uSD8sz=W&4{Lhf3jhsjxt#0k?dn_y=AjwN zLlWbQ$;3ub9x{1XcZh|&+v|0aKNMIT85hks{kEvSRwQD9SqW-^jKQ304fNf3re5ET zn=Y7F{&B3X8Q7d;?932Ee@VOOYr_)5_F*NlWn7i&GbvFZ>iwf_S#4vBNb)xA9h(aof z5{=paO8)b_A1Tbk2-~;s`zK%WofcD0u`>sscdc2{4c8#!^UUhC>tKkr*#_5F0{h8B z?IYXl)lP^O`s7TQ4DN#YgzRgB;IBgpIltP_v$KxHW*&W%zwYASow040p*i2{efu^KD(;SC4SJB?6eAUT zv>a`L6Hd!h+izlauHzvzld=T*K(4nijj)^L1JrdY)gtY))&}N3w3)%O+2S;fUhy^p z4(pUJQ*^DK4a>KpgrW5YZh0BJvrPp>0YzCL{u3kBuNYq|}(2Xol#3Apm5< zL>>ltuI8w|alH1@)z)#$7CrLF@b2++PiAzy?RLbBO@_!WI!k(IaW~kO1@|xdez=$F zXiW{we_`pGbI=!?u7;*VKLfI$NE>Nq>c!D(4jBW}b1&9?3pjgL?g2b74jO zrpZHaZ#f3e7y)yQpC#}YS@gK*x|%q8xq1?#+DWzEYQ1w6elTX( zUBU9oA)T{9EseHWkza!mM3!i<7t0q;2pvyQlY*><&hj4! z{daD@^~uWne_1(l;)lzJZ=Slc^z5t4Z+*7(!gK9VHPbYe|HI{%mYzAY^vV0n?;Yz} zK7SYvB8%+io8K&5I&$lSHC&fw)yQr1BGNY~G|BpM2 z>0(OpwPJ~JwOV`Y9fXrC26_XPgrOxnV`5;Vz<1a!zv$|Ksx-aG^b*;vgRzIL2+Yk! z&p3@~laqe<{%ZY0$a=vPhTqfxscLw*cMK&`omkW0W1uH!o352(w`<|tE@|x1bVGA( z@R$m%fYJ`K{IrH-vi@omDV}It-5gIelZT`TN+6LQ#rqqz5Pho*La9y4X7YKEPKkNW z9-oVU1rK8yI|6xqx@UgC)!RW47BsL_4R~pnH5gaB2^7KqG9?$L998)^Em3RA#|gzgh$$^>SW&@CWqd_WKoDE1LQ$rxr-5nOg9T8zlL8#hPi&f%f?2CJql?p3a;TZ zHCE50`>C;CBcKUkSkWum>p$IkqdrI{(V%Ghb~Nh$mgc4q^(&VAC3x%A(X2%wq)m6- zrC6*gQtu#~P{CCe14Y<((0Q8%Yiq=uX1Z>Y%nY^i?S?-yYO#_={NvaJ->sG9sU!te z4bDhqOe)9}tMyD>LAf5ssF%rfRr7YAF34a<`(&%ijH<#qG6sbk^1m&n(6O+m0ow=a-p9J1J!Aw^#uTlD-PRw*Hn zJGPYjU_E^>t74d#wUUm-(lJ$#Td$=65j_mHu8*FBSgWCfUlK50LKc(+Oq3wXV)VRx zEdhzFRNVpoI-H=y80=^*N3;Xlq8I0pEAiyuo3n|d- zeKy(8@O^MZL1qPl7r}ah?QFz_Z?Oc@XW7&_2+lU#vqcJ)Yvk?ulg>e1%N2um+H({V(Z7Rt}?&lF(sHoS^#%TnqR zHg(F#BrR$1{8}wA69r1vqNa?zM?)48ftXDs8`nxx{jvLQaQ3T z29YTddm^xJnPhU<)CsB(qmxw$rK%E2g_lrLNXA+=NrUB5q8f{dP0OYFuu_`o8m!(D zFQa7#=Ng==Pu(~A*lGb4#wvS;GGxZI^gO{TY?3w2DJ)cT;RJ6J%`#8`Xd>H$dQ2U# zEKIB2vL0hF1^AF4iq9HoS40pW%hy*E6(NB?G_oqh)QuR)HRP#VUIO?K)~@GtH=Q++ zEZCZ%urj-BFu}6sfv#Qf+teMB0m}jDN3 z8={E2M?kzG3K7#PFtS%5yRFH|7M5j$*7?R17I41lV-Oco$@<42G>_)AD9CZriDF6j zjL-F~bypF4)^O+nB-lg=?a~|umC#F;I7M;g0BQr@iHau}wd#K`dcnLIcJ(Tf9E(Xo9?Jz5PUm#->r%~) zQmb4kn(;05>&F&9zWx+h{Alsw^`n0~4lmbf!w)tR3|Y5@OfGVt_Fq4}{^a7P*Pr-X z17N2PF;?%n?ku_o8SHyT)n`p9f|!)RVO0dh=a&Am;sU6wR=YZSnO$V? zK`6_;--Z=cx zjU!**IP&_9BNxfG8%LhJapc(>hreo}fgBMrloV`Sg6u;pkJf_hi6$Y4YRkomeBpqV zWzWGs`}U>D2*NVurA3hldQ!k21oAD`no;UAVUSy%CoFR3(>v$Zj0uo zm{XgiRQ%XMN+|3gE;mF7O|bH5L4!2LM-w7-{033sK(_5M5_h zzMv{pRrMh%dIuqe8W-ZNr#7)7ng{3G%>%IV9%xgvdH? vd(vA>@@?Q`_23Q!8I z0i_sNs~ab4Go5AGirO-^IV6HKQx9h znmbi?vsr6yfEgwkFM*nmERNsOa+&TPrjF*%N=70y^aK(Y5@O5CZc;lyij3y5?i)+; zhj))}CvXcSYwEmRqGEJ_9|M~hhQlx%uQN^$ytVhJdh|4X{*0xu&J4^E8Gtp1J-qq* z`oQ(}jH!rK1tlxyV{>Ss0tZBw5)$#2w<#V7MG^h3WB5PD|082i!zqHjo-a{@W{dPc z#;7UmFBg8-v7KlRDN>tc88b-b-n@F_O39p;#agY87ZvJ#GUHO)D?<%FlYAH}xzoxVFFc~L-+as0Ro7>E+^KCR`>&%@2*K(&Uw@hu! z&@&j?6f^P`ZQ1;(%ppu6m>$g6M9?WWcKPL|&z39;Vu0r~H1d6AH)fU#&(|~vR3@3M zohure1BNXzXs%eZP49t0BRWvYwM$IT!>20Sm&d#oOLd7-n{U&>Ax0MQ1U+CHHEv7f zV2+3x$y9(mX_Ewdn*>x>+I*+v6MO_WL1K!StR}&{gOGAd334zH8^LhuXI8#7bLh6KRO15TONz=|%~1fyq2|)nV0b0eIyf*lIKU6#Z9$aV zEN8U~3*n*%#dlB(i(RU6mzB7$t~LzziYN7?5K}gzkxOX=zEagp6O0yNl3CZU(I(;# zz#7G1Qf-souBN&-IKt@O>866RaQUh+Arl6qysSf~gVXCEmIYuH7|N51s zmrtymINc5#W3335UjAt5g_l;2J+=JdGt0-KV`z-Byk$DziQtcKLyXwEd>5^u7-114kF=wvt*Bls*s36N;B_3CV>?odkZFy2w_>}c z618UAE>BsbG&L~4N?Bh_uo@B!_q<5ySgZl^cr-Oj z3viE?VS?zg7zzcUdN~w4-_~B^C@o<7Mjk%H0xzh{&vzwHQf679cis znj3Q`CF}xAw%*N-Zi$OP&}0`Yv_*w$Bj^Wu5Y4zTR45YJ)>G6B8au)#UPaY<6Gol~Q<=fTsq2Yfi3^~H5Sn7H`m+8J@Lkla z*;8IX0pkPIorK4L5g5-whhgU^Z$k3^pjUmNIWR!D|GBkS@a+d(YyRupRDs#}!S)Ic z^|KP}!v~|Qi*d)yQqA;L+lg^s3dK7J`E106Q~(EZ{jC_0flH>5ui1?o;cUmHq})td z^wMc&BGr>?&-q^0KQI2vRvX0fY*NF$T%YC652y?_O$~=^2akzicm^v`I2+ zx@3|uodX%7E4B7H(<1CZ8QJ&Hy_&_Af1GCz`LI3vi^t)ePMaOT+_t6RIACztK>S-q z^@bhh9eaj)>dlQkS`JLC8pfcQEJ!gyQehFNe3U_*vecLa0nRsj`eac^ww@kFqAhSX zKve&CSC8iKjf3Ri88BRrwyNN7d)W>W_lnojWX`qp2aAqVPIq<9%*+V;(UufUtI%bG zf2Ol8F)7DlN{oE} z*_~432$|F*LM9>dm@L)TaTg;{Y44o>;^6!jHko2zQvKJ9lT4rg8vpWC`+06-tZW|C zocW8E=9VH$B!j7Xy9j^1Dia?eXfkB5OiC*;K@66UMj_3hmDEvt6K3>)7s^a9^*Qn4 zDicVrhPxj2jPPK`wzXTopLCLb-dG|<&;4O(KBmM(q!9J3sc1|yp7`QIY_qC=YoLQy zD}VQn4iD`b92%dveO0}i1{^j~Q>%?f=_)?x47Danp^Oo=m1@ zPz}tOi2#B21SWlKo3W}HlS&9lQQ*|seiPbbB|O`}8bfoh22K??A@z#whp+y8^ZYv00v&@p1Db5m1dDyA z$umfI-#Po?(%Gfw{_w-)%gg6pTsnGw>80aKZ+)}!{$FnX`Owl+mv26O?B>&FmX5u? za^iG!yFo(#O!908pWz_-@#sUr(R*T31U10-5yd^*#R?=vwqY7aa}ZG+{i%r})zAsO zzx_<|+nP*Ksu5o$WY``=loS$T%T0M`oiht;hZ3glkoFYD$s1oQK^Dzihb4Y%NvUZg z-tfTiL@E{&liWlfm&j<}SpV*^;mk-+AN-RU=o{`EnP|DHIT4c*NfG~uPn4;uEx})h z6Q&p!lvp4yww|li(c@CL=&+KRt5J{x7A?)6aI_rtBI*S%WfV}9!@mhA%3*du%b8n3 zOAb;42?(@Ykr`Ql_^Tn)^DNu%pXf6wWU#c+$`FafGo_tc^ zOSRmHve)Y`UpFKwA{atUkQC$#R_1CC_w56J@?XUEUWv6sioDW#bQf`Z$;g1;zmj|v zK>Q%wT9YIp5r{N}w%h&1kGI}CbnBh>m(ISibmje((|=g`%WGXLr{B1B{E13gcCsrb z@7wp!umAYNckkSM_026W!Uj(A%@5yQK7E*MyYOL%o#&g^2Pw25 zZibW)`ff(IWtvm^tWob?APB?dKd+p*xboe(Egde2i6QVLZhm`o>F|sH`Q^%m zD@)&gx%BKyH@~=W^MxZT7r$9L`$2!jtO%)SI51ZDMFMyI1L<9;}Qeb`h%$};z z6INuh)PQd(O8Df5OIKF@dq*-Ojsort`AZl1sR!{v9E zKYhiwX#DZUumMLIJFeob6Jw7d{u(v3g9w<2Iad!KiIk~DLGjI+7Rl)pE0@}^rB=ff z@irt}FU?<^*7WF`->}KD;QP8-mYJnae{fh4eSKD8Wu=HhEXarzAs^#qCB#@etd^h4 z#PsD{%5&5P&&u@AUv{#l?&V@S!F{k@m}tyBjaQh>`y5)19(~N{SrhpttobwGl=uW# z4XY@jqbiClqykMjcMv5o@Xjw0D$prv)FwO4gSt6mPem`W*)Uq5%o&pdWmlo){QcuE zgN%EqGk>N44>(KrYkCP)DpoaEZ0j3})myL1rG7a1`>0;l49zY@Cwm+)T+8%%6rDTX_04A>WTo1S^OD~^Tdf~aH*FO2@*T-(Ya^dFZ zPyF-ir`t9Avvl&*50}oad~p#5w*2;y$cqbWLL|^y@suFT8`z%|P9vS%NMFpjM{au$ zvAV@{?4UpfT_;!6T4*M22wI(Nyk3*ZW{&}^eYtf|vxs$<1a_H%i{(QP5;)@Jm1LWI z|1TvT);Bo`cW)!AWez5jW?P^@YfwumH>UEDn^5^ENM#+9>m%HDSo?F~y@hiNUoL#J z@HttyboJ1}cUK>W6BiaPFMQrsOKst;g_8>}6By373s+zS-zUCZc7Zj3y;TtxeAhI#{L%aD=BDB5l5D2Qx)4!cEKD$zV*V;3+a79G=svpUAnI0ReM z->n$0Zd?3yIDug6fYh?}GUlPT;htlc8OzVEfl^i)&U(gs|0R={gI^?**thTgHg7~5 z$58Abf(l1>=f*OzWDO9YuTC;#oFBx606+EX=bqvVtN13rfBfZctA#FrfcEQb93^QI zBvG&hAeOpEZ*!L{m|k5nPlOy$MV_V|9;Es&9%CY zx+<31^JzS+JYHmYmKr5$twCNz2jLs1Gp4Ukr!&zw@{(v8cu`|QBHl8@8}H!H&F%^{Ovfr z@Q8-U;%AGWlIzFFB7gf6c>7bbcxv$@e6e`(uV={O``4eu%TJN(M;E_>FE|9B`s*1u z!})&T;=dh-7a!vVhw(c6o=>ko^|#~rCjZvU@cR7fg`dMWpI`h4zU}(azn;PGKDBre zkDpq+fFE8w1?R3mdHoQ41TH-d_k9YPkt(-#zW(pt^7Gf{Q480^aT`Yu;Qabep^91tYj{7Z-n0NQt2lni?$rQEJ=FDDd8VG7dJwEEo%>>u~tG@u^w70E`73cMVO9N!Q3ZZAQ~Q0)K0kWFB?lHyPr!6+A2sPDw_y%4Q42k>hTeSsBaGaylDu zXR9(*tHz-kvcgMEi?X1Ko5u2c&xMQjcuW*c*a|V>lR-8)r-h_?BR4M9cuRrvO&^7n zpa#s#&=hMo-@Lr^!G+~F58pieRhwzp8%LhMapdfc!+*YU_=6iqF5WnD<;Ia`ZX7;& zJfhvWlb<8>x!uT?0zD37w%DPcr0m(?$$YFoqDFO5touTU$^~L0@}ds(Brs?J;_- zjkzmRH{{3?9YD~*-TEX&@F}0!(*lDuymp(YR^u$NMBHLgXEkHZbSr9f!CjDR;Rat}%;?D0zjk3GcYThY;_JN$t?QJekO&l6 zhEB?jFc;jrHEtaOB3etqRdF8MVcpzrR-}e_jefYsumV$MK?)GiM+@<>B3_e3xmlJ4771c^R}E-1;uvHCjlZHKc!tHHPAX^G)Ta z3W)$^+2%EngJ&lx2|+@LMp2!sJHEXwq_Iyn~Rd5gh3N zFXg?OhHy0vJtq79@t1pe2gimhx~dgKiI4?Eal|Bw2g;xCwdYiZGf}iO_%e%>?b%$> z1RDh#%c2|e^-VfzW4)<1QJZmIsV>)J7)HccTYl}Eb}&Wingw2X?$&EBFQ5Bt>6y1~ zzWMd7x2`N*c`tfaz|wL_H73eF13q#x3Z#${63LC^1jZd|=xmnsxMeUcKoM|mv9=S6 z6pO`b#?r5W5g*;IKrC%!rRp^egBrg0epNPVXyJ&GkO+_CT`wH4Cv=@;tg@z)2mg&^ z%CPvcEPtyjvEV*~mG`@_;b+LkhBjtJm&*hqKs#o6LyJXJg_N`rF3eHh(+tk-16v>T zr?D7dIAzo37-F0^0wE+ww>i^73F@Gg*9?1V{+kVD!QyF26cPdAnR^ynIE$^9PH%#( zCab=7P&H&fXdHP8g50wI`?PW*8B@Xb?KmY#ie`K`~u^l1mvplRsnhs!T5 zJ#%L1llPb3JJz*){xBRwme|cVzgfC;~pu~h!0LnA;=7gCic94~_Udhb^B7)Y~WG}TG*q>~jYxLa*88twcw_E$l2%ECd z2PKL?e(l1LC}OW|&iy_^EgOjOU>V&*NAS9}D1(Z63D%gQU;0Op7Qn zg*-?mkhjKmv%KWIU^=9v**1JY^C&@K;;bq3r@->UV@}s~R{{B>H<`hJa1?qYY=fV3 zrr5Cdn_X;xjpWDuIbdH3IIL7TBuKrH7-f6A`Ni8SufKEi)5l?BYll6o4wCHrd$*4N zW#!v*k!7$`o1(Vn^JGOIOEP~QQUL1z&)%E0H*sZq<6mVr=Xp-(_uxWRrTM()?Z+dd z!N)eC(@8G2rERHjskTysjfis|uz}EoOr3e40|^O3LSm-GjJfdX|FwJvr*Hbi*i6oM z_^-7kTV+XBsw4vz?e1V#l}fVqUbWX=>4|DS?ED zI9DcEfd+95xUSih#GB3P-%jDHcQ75EznpK;>;O+wrui8SXrNps6auLg#1)q#f-j7c)~wg#7SPRcHq&7;Q_!O&HxA_U?q zboJw*dzU}Gcli+Vw|w(?^BFG>KDjje#__r1{>}S)H{W@1;{4q4w_H0?FjBp}2(H%> zgQ`}kdCqzzVG^N^Op%A_X-mMrm&>d|*lg-6(A1pt;qNXJp$WlkKt2|D?|_q_hX1Pc ztq$T9b7oj>8C7pg2-fIJh#+1^BwC(NG<^_p=HJiX%3qrv&)>*j$={qF%TMI5Pan(Q zgsWFb{#t$#51+R1>WILf#^^f?=#A+k)9>alZXo%;2wbVlFrpjz$?5$h{~nGQ{tp*n z#8>hc;qvRgWhoG=Zs2FSoP*M8Q#|4gM-VVttzbO!6IaC2mSPrYB?b0Ga>CnmjDq&? zv`*H;b_6F`4Q8zldweUma*E57_6>F)5nPC|aTK&^w%n<%^^*!n;g{^am90g8kje>byXN@ZyXAJpp#GG2rN=jsEQm zI`p?MSQ_5eiYru-f=F#br2Y5hu?G{OOhg z{s7rZ=|CL}$~F0bd3N`rq zum-=4X|_3aD#bQNB9bZa|3aE9rT6a*<)tI5koM8yMOq-^xIPCblEJoU?LE$3|(rKi0%}TEz(%F;H z;_XyVVZ>%LhUxut1jDz_Xsm6givm+ANl5W7B2|i2R2KOnc-t$WH*~?ou$IfHC+3=D z7Z4kUjx9=1QOL9boJtiv2E|<0?Ts>42t`RDV zB9mscqN(klCB6U(a{$ScdZ?pkrf%RP_T(cAQOr_AgXQqLp2Kxm7&|>CN@J(T^0%g6 zhYOSWYxxP%0vAc4ZdSyz^0%7I_{e{lzmmT;eZaQ>h(LG0P+rKmW2lkpiPuPtylWI= zs6I594}+s~Zx(45tOD^@vzZ#)un;LUkt2l;{-pw^!K9kTnf<2)Sq;fNhXM z?217Oh6HT?Sb>Lq56iXdDuJ-11fxb&)DaMMBLy;RhGnCVfsr)X2C%izY5G&t}%?vOMcW`vRq$Co8ZYKC-H&G!1Poqr@D|M1n@aEr1n zXTTYRui)lzR$YAxs%OO<3QB4PDB3??v(FdP2SwKOFzI|sy*^RIIEdAcPuHPY2pf`e z3bsr^ciJ43+GeKdT8bH#O_MgNG(nw+e zDrzGUnSPp~CbAFiwYdTQ#Q)X)A9$oXO0ZSe$#&R8P3!+|)c#*HD9@$x>7YfZ*E$LZ z!vQ6(Voz4Itu-EN4+r9H?Ok%bt5pug+d3kVc&FOt5?x7ho&^_{!qJL_`}m=Pno7m? zBPpiNkdkhNaDn`itHdxY9jOphz2TGft&TN9)B5k5vU9vM1uCM=1S+!6sX2J&b2{ z_8eA5!4#7oK{2$x51w6`I>23 z@la&DX_D1f+jkQqSzK*$WaXEeTuS7KV{fulgH?D)JG8?*e|y z5NJV>*5A%@$$AL*w825$g>2jst7>J$Y6>AD!L>#;p2iE!=0^j$nHK^cd-owmh9MNDPs3qZQn0L)>!#*~2UE{QAjR^6STwj}A>f`sTa& zk53V0{@~=xFOx^#P0kBo|G|xiSH2^U#xBFr{E?4;+5gtdFV7!8xc<@CH-2Aw)3&yq1Rm6f zKTVcyefj0}^Pk`Nz4(-mzkhgje^beM{_x59&)$KV@Ga8$4-U?Mb7X$(JD6B^3}YEq zndR4<96&VHP^i2J#Lt?$mA{a`H9Zz5SY!bICa@A=Njd#!4@}NJ0esbDDQ{OKAHRL$ zYfmYaBPiZFSy4pcMnd_!jHH$%CL7j7y=pib2rH@_4%W(W%dvo@D3T&$b;{_HGYt8= z{jjCC;-%&z*#<8fTg!VS(qmHFKjYVO%ois*+%ZWz zHw|W5BqMIxKrHUq=@G1JHVv|*;I8-qr9b>HCZiTSRIV=|SVl}u_$3DU7X`mgKQAVc z3a&mS8AQA?QaZCN0^Um@9R&{w!4Y7hi58ELg09e`ElEw5DC&k)$M!iH3KI-pFSh}H z>l|r%{b$;FGObzg^p4Qd1E`j)Fx8tChR>fOMS|sSzvi{OX6^ZVYR{i<^WQzSPxFBb zN`VTb2G3kSYY$u#DZH8SF9m_$;hC{(Gvj}KFn*G(pBcM^?0(~CXU6|L`_o_R0Nh!* zClR@C%C}zAtDCd-BsP!g1MftGNpp14M-8&op~eQX%cfP}M4#Nl7e+{;dhI?flQl(M zW3i8k$YdnhT*{$GOEGvTv&?^AakrBh;4ORbCo(F!Lb`ry@=*{`?Emu&x7)Anfybu+ zrR%PrF=>t!df%7nU0K~gyW7BHG)Q}zIwm!$HT$=a@im4?z`Rxk?If2_5i8X?6xXUn^t|4mxs=2JGWw4bF6s(z84x(Xoxv}Vm zFHI@N`}t6gr!-TK#5P(WxBnO#czr!YO_#S{-nBm8!rW^QUwxk zO@fJ_A>;_=KCwL59ES0su|QLxA1s>;FYxO5eyl2{)P=j<5{y-^3y15q9xqjq0#VP( zy8Q7f^nz$fZi!;J1+Kg#tI_-skmIr%#V(;}Al}~56^^%tS_AQz+8K(gvDQdj?FfWM zJ>FLyPV@TlSREmunMo1Waq5~5^FuM$zKHhNo`#(yBrDBj19a0w+95q!0)F|A!&zo> zpYDF9sy|URpC{Osl)8%^M7}+(4HU^!{?A5i3K6J*sJ-B!C*3>yR{ge8R?Q6SJx6-M z#4T7kK8gD>rht2|f13}=QYc!Up-5}XnPV^)h&`Fs$_=uAocjJBr^fzq>YIO@`uQKH zCYCYEw(O`d%C;EE#{$3P+iCg7sc-*r>IQuGe|-I>_z9!=P+Yn7k5k`?pWjq5)H=48 z+i5!}KA#lJAF{~7OZu53oZjLR=L87NBKpo;AX59BsEr$^u~!t z=MK$Z^iNc~as^4~{7S+4t6dU|2>J%vWU=qD|^ zUaIYdnQ^tievOG%irqg@a;y+nAM0X zYZ`OKHRcak^-b2xv1s7Yp@aUj*DI+zl`xp9jkL$D(FrxwHiXPNu;#PWf3MIAA*H}W zpvqFP0=>%cy3qekXW%O~*XB9(v1zbD%^Ir1o!CJQI^>G0_($r;DUV2@a3kO}?P-dL z!%dKKEQdDKKncKhw;myYcQZDG@Mhdd`PYMg+&Sp>)MXcPn>h9+iJ&6Mm6TbJq5rho zF&*n!uu;T77Q0W?6*pC4vf^q!^;iufBjtizk=$PkReB6nGSh!{wo0tBB)PC>!G#p? z-#k}kER(M3LNIoM(Ckk+M^i_vl7jlcAHRDHsTc*+TQndkusWjd(u|U-^em>R6sz#^ z*4%-H-GNJ!g(lFaLttG6|R&lA$Ss zA@08@7Ad+95@7|X)84Kv;#h}?BCt6)_QD%28=m}xL=rkRwZU9-S)A~<(2;3@Nhvxy z37*X|vjs*2$5>M4KU3b*>0-bRmhwbFk9i_RKoUh7sfg>snII^*EyghHnPALeX))el ziOapBq8OeGi~R0)&0w$gFEM-YEZ&2UaF@<=Nnzx1!Qbc~&9;s0%kAOZ#cj}{R9&u0 zO_4=lwbtGI03v&E;Jy1l{m=}eSh|E}FP@!!{o3s0nD2@2Gz{1(Eyp$SYIT@rb$OU) zn0h53sTGY*p0_}l`iEEE!QT0g-XV{U?T0;(Ji2=D*JJNs+0Ub~GxNtkCi7n&!kfOn zQGDsw4^EKzq$aQHecCd+bx4Gv2fc0L_M*|n0bEl(wH`&esw7Ye7{5i}QWSj3~> zs0g4Vd2RXCx%`~$3M{->sa-wh4SEse7DN_x?{qKd?{DWul2@=}_Lw~B2x5fQ%=H3W zL``5)c+hrGY&B>zxt+$U-!`^sQ3Ttr)QpS`GkTPN|loJ!_bXAT=vv?3!Xjt)H4$P`#NbR%eVnM z9qD6cO3PA1^rLk-Hbb(`7ScC$7M3At^gD!d*S#>q76i&RYzfhz~GTv9h zh|@$`H1krE33jad$S%7aH?1X}JyX1vf8x45b=cd4}15M_@Z87Ce!fJ#_+Wo(46YTYPZYZqqIOl|gf`aC47g*Yv$=`-n~HJn~6I8}t1 z2`P2QDt|LmD z_G!53;O(=c?mO-rt#SuhuY`Z!WL5IFzq)e>{(ls{{q3)ApTB((26Lt`va#D2;N;4k z!?*q8bpB~*Nr_RgB)D0o zrd^1i)8^XkdcfDokzS4aZ!76^GF;EPc^>^1j;RTrZBlM)-i65EF%qd_xn?L*w>6KE$i@PYI|Kqxnu$s?j6u3DY#$B84iDleDB~hR^{=pg$TZ*a|YZ-R)r+njV6)Gz=0P8S>J) z7hjAsBffy7)C->q`>VZi0gZfPIH@!L(dA?HsD+KT=8n=pX4N&d_Ht^B3@Et3BkeoP;n z9-kh=8}iqu$NgJbVAwREnGx|tl9fg{zg8+chzwgf2HO^{PB>{1C(p>j?52CPB-sjk z)k<&1bu39aB*lDS&URxGJ6VA_Tkurr9%hjqQ!^4;mg>tPnXY#DSxx7pe}p3g^>7tB zgmo|MDw)p3um6GrzWdZkqf{|dK%14NAf_i=u_M=B>RCcjFqi6PwlqZKsE92~Qra-Q zv&mFutC2#&Jf2}m|Hbkhn>ID-?+4>bP?9US5Iyf(LZ{!FejORY;NRp5(te%!FIJ%ghAo+^~M)?!$Yp}_Q5l_N3xcy50 z#`L@Sn^^gPN+=4%cQ-p<6l}MqwtHg zMi%WAyt+x2#bx*{(3scv?`}5Ysg+ZyN~vaE@mMntqJu51qGsOnA}iP%nvrxYTf~Ex zsXOYZFf+Db`7g6p2dFA?XLF01+6K~%!5RN3gR+eLXCVnIp#v_%-(~rGHvNN&6s#ax z^p`or20c<}WWbc6XxPk9Ec-X8k!Z5iE=~~X;%3ItDcK3`h$Y77-(DZlOg)!n?ruG~ zdb^%nqhBwGgtBV1Zr|+17rS@%#XDM`T-U)$31Vl@AN^-~%v^${om85+D;i3gD&$#H zA#iP=d|CMElAIa)c4qAA%-GK}<0k|O(d#qg-!7xEw)RwLtgS^%pQrB^&5T_ULpVD# zc4}tq%*^=s%-E!F$td8>1VfUn!UoWg-x>-@fk-2j4m-Fl2sliZ;AXP!NCo)gCL82t zA;$RZUgy+J_a>H7=ZR>F@)0!YVo3hl*QNzFFIy(C_21cSwT#=9)87bs3fDu9* z<7lLAXrv;$i92=Mo~F#^V9MEAlCqay{&!N5-CexpSxMwB4u~YV&Mn8k`_+R9z@!OZ zvQ%|jD!KJ6OfWZZ8*!d6t0>Wi0fx!hfJS zu$|Cmg?2)?tv#*jFzKa^*FLt_p*rc@o7Fio!HI5IBe`(Bb-=OwLwGDl1gyqN5{`Km zaZC!t>IRi7)Nzt6#vn}nUNXS>5V4CDO+?GUJQC1J@K(&h*_;ee*gFR`U9c32;=;wd zlRV+zrxqsZ@o%V4<{8=n{*hokG(p${%)kHL-)h)t7=+uSkgIzXaxFQ_;5Mq|o9SBP zbA`aFHa>6TL%-~QZyR3;xU)JO_7FBc#44(-2Yq-Zx4*n|;LcxeU-}d) zF_88DLH_3lIC>q=Hr~E;`@)?gw|~IXjdu=VGr*lge?NBn!tcqQBLv>~*6qu;&;7Fh z7~X*I#a4;ir{Uk_;zf8De)8?_;NQZH*jjN!^b-8=`{iOcy_K;W^RYNWzQZ;!4Wc8& zx7;J(z{eyxj8rr(4ebgBs9sSDLkJ-g4V3HXwYHb~+g@NsU=&BmdN?}T0xRRF7~Mhq z?Z4W%yxpZ#Y#j7hZ44rgRAGwnte&5H?oUo)h?xUZH&8=A_Z%5<%p@bj%!G43z?#kU z+D5Y`w^$t!IL554E?M|xj|w5gOSt~73N(kT0y#i(^?Q>`Cw((B%pBP%kv@24MyPJr z0d!nJAmwWG^Xx+kpyP6KhJ9)8yWPEslXK^PdfKo8{w46>lS{L29G^SxAKLNW&37K0 zI6rs%tp?(3p?FZ048m7BMPw#M*$ zi}bg;Xj|k^w7Me@iq;fzSW+unq8lgz>SGDVNKz9yP1lNm%8P*HSSj!#jy*2Lm=vra z(esy!xHvO@V`lsW#yRia{CxJk1NVQvIeYZP%=q=0@pCidmuAL(xcl?>bLYQmGA%j# z)))6Ke>D5{8xLOhZzavANGK zHrd?YyK!#r!yoQmedoc)U-+MU_-$-4hOuU_fL*&Fg;2>a631So5RL~_SaEfmY636V zpbODF*s+k8Ipq|t;CQWnCnMBE99j}w^>V728Hgw_Q>qjK73H6&y#rCgOn1>iG+L6E zIQA$n#3QPRVtDmqMN4ItTZ^p6&6Qi*N)7TnxBQ3lTQgH&~1Hbv;r)h=N3{HTy;vU=M zDT1xJM=p$|j4Ch?TXD}V7Axfd#IZN6tss_L-JF%Jg4~$A`f4`YvIxZc+H1uE*#6z4 z{kuVmO{RRG*VsswYp{P{%!=#2MYV%z`i_!hTz05;GgN`q7ArSve1)ryoWFm5qWQvA zv!7nRcjSuy+JrQ>c}{bg|AA5ojNaW2iufsCMPh(Vc+6}H%W>dkQ7XKgKx%o@^iRt~-x|G(nZ|*D)4lbROqKH6S z$uh>*qf*3`5>{Ny zXC55?a_-1Ccdz}~_fmTSd%k#!Yx@e8L>Tle!cb6BDwxrp97vcjd$W2=aMh>oi&dU( z(uu`OFbs#9nk7c!-wr8936n#vD5PgWM!_my1G!&Wko%8M%^$t`>-cf9zBmW-2T#DL zo+j^gBpwKYy;*nHln<|ZE2SnkVxcC!`G$P{F6;LvMoPz&! zXkuuSV>_Tc_W3kKG+Sm?jWn*U&`CUnZ8_0Kq zWeBV#p`|crZ)*viSsQnAbI3|_|C~OX;FYZ~p*?P2&lNfGTG<#uxQkdtuzGFtYqHm-HW{!R9=eq_k$)x;us2B8RG@ zm%#O!q6Po^K=1<7ecMZijR;<2j4KuBInRdmU7DJGg<9 zquF4%W?k;R-x^|&3m82WB%i)$A`NYjn;AlN^mFZWrtX$yi<1O$5hX>-vc+49XU|B- z^7aBuy#TV;>OpK+QG~S!Mo)vmMtYWaihgAHc4(irbJHv!^!ITGU%)aYD#i+8L@XO! zM=`#%&a#~>GhbSVO15yuO$knSaCA1Yy&2xLfe2kUt}OhP=65^i!G5YgTzE6U36vfPovsTpfF!?;Cm9WE3*u| z&0w%6Thz?qU0L`($PS8*lV)id4ZebtWE+}+1GT(Glw<8io5@G~KYOi7ni(+Bxozx` zi#EcP0*~Aa_-u)49U;1uhV7UQVmfYyVWC?rv`+ElnSn`Yry0^qE&s^qD~pg2iH8DG zM6JIfpmx1}DTdR<3~g^0qo{$(sv}yRlw;Y?LZHe$$Q2J!wFF(U!g-+B4vc9V(0;b;qm zjnBRMs^tt=cH9|hv6+Q=w*P|vJ@=T_k|*%^hIO7ShorE8JC%X~-=gb=DKf##vl1GD zVpy&?L?~FQ=n=;rMHrSM^^34qa5{k0&^EQuQho8o1yX=^4#Lm|9AV>b#BpOHUIBAm zj)X_c&aw8?z5kWpju*w2SAdz>G`e|{I8`NClAlz;>Pl(XE^ab#2@*tzG2wp&SGEWE zSFKS*BZuj2pos3lH=oQ+{xEm>*6iC8bLWrEe)bt29C_#7pxE2J{`Pz@`jn^$Qe{4JqXpoA!qQeT=l_9(fi z6s+j5S(BVs$(jCPSt~^sDQUlm$B1DxuE5?QmnXv31m{_B{(bRE!Np2TF6Mc~DN!j} zfve=5r*;7pEP;fzPF6%=Qtl3!4JEj{*8cd;QEq18?|_8@uEVnuPuXO`*mZ0a({+qQ zWD5aB-y+@2XmWsO21+7TU3fd5w?E`nt!TYogWIWFyzu?MyK@L;Jnjx^eB+>|z>gcxr7S?539DRqoqnKL@1wC`vlnbnz=BBP; zqXZ?bQx|A*0okH?WYvdof@!LeZkpBoduyFZ1a99;hrv7`DQ@heI@?EJDu=0AD6*~J zZvgi=Rsqv&4SSEkl**bG44NTg(M*OKgVeUM5x)?g*_ladrj~+TGKmq_BsGYw!QKHz z&K`Ki46;*CGG;tB<7UO8GiB)K;#+SZg+cfh?Y5Z7WZAX(l@y}tSsw-?QaId50uxKo zSw^(04>PjGvYh3(p;oa)(ngITgCmqzv9v{KNTi=g!e>(d2?F@Go3$ZoIsq3FdR?#= zQ6y7U!{r(QeDabr}_W zrWNbspN`e=3Jl~*k;e)7oG#~ogaO^0J|G75HocsRI zGD79A!R`LLN0;&*y5M`fqW8p*t~6AEr@5?-!{^V|XliNcH)=Y^G|}YCn&^pUGU$s{ zZ*gnizOGD|>kuj$ckicP{@uoTrobk123*1BP*qagn^Ul4`UO4OelaZASL(+tL6N0lVDM$zbeb4S(~%*IvG5QR~lsIi512z9ZMG|P2Io6-mAeRm`Q@ceK$dHsI=Zfp4XeKNReAw@S}Y7Fa{KqGM;VvQShtw>f6U ze{*&p9pNr&jh0|_TFPJzR?3*bfUcj4LE!Y`{?yy`N75@;yEYIR&RVSvBmlqLEX-yy zBfh&Q-qYFIzWIgG9!3-C98shISx@zBnv&@6pBzF+Lq$GIbKUXw+yFDPSfUe-L^t*h zE=+-MNs12=#XYWYe<|AUbuB3&DH3gf%+PC^TL}l`csLl2M?2v_iAAIF_F!u;-X05u z;+<{LP(0Yt)iLVrw(_L7#}6;&%Hst4k?;w`;tA^+Nu9$`9E)`BV|BPBm1*-j*cD@# ze+X5?FFj-G4a?w4!v6WT)%NVFstiC}L!Y~D88Ycsfcr)j6rc+SFB$83USq~c6tP4+0X zQ;QDh)JXW>9(q~Ywp?JCE!9zdYhe1uu!l#fpEepStTqiKu^fEwmK_!D%N8TKN&q@` zCTtE`naKVD0OY75g%J`HIqF=Yfl$dp3VAk=)SwiU>)xP!bcgrSBod(->{$R^TO7<2 zKwVQy`c)izlxs|iRsPy@n>Hmt#bPX*IA8 zz$Syz)*jegMFsxKesZGt$(?&?!p5Ew3zdJeCCNt&l3W;3{FnXTtac^7P!q8v9JGMJp(pJPQQ*lVJhcoZ$V7J^HI& z8dbclK5C>io04rH8dKib4r_*qx)wG#2*xEDVFPuz(5}zaTwz2d5O5_b#fi7|6ADPN z%Dlf1@W4LoIV=DN?x2Z6#=Pmmg^1+KP0am^j#`*HGB|b39^g6{^#zeMUT%~}d7+uk3HYI@l z8{MwgPp?o2RhATK>Lt!IW8mRdF}(JlrSmjBWZfn1PDP&4V~UQk?VP;XgKlvTJ2 zBb{eH!rP~BU%7J-ex4WgZU6Jbog;VtOzs>;GQ!&z?!3P0rE=#MQYGFwjAV%=>A0`P zyLlSkkd%-TRgweJo~-}c;yZ_KpND@x+&PN%fbO~hzc_UuGHkGj)?L^k>Qjl66`ZXv zzSzC9FW%AGd~~ck`?mX*LYtXOu(Xp(FGcLA9>kZI58|T?LQ*8uNUJu&9S9Z$tC0Fy zXvk+|fG2Yc#VSSX5mRKXI_&7+p&@#|Cc5;|5-XfxcF6(PPKUp^>g8CCY|CV<#pF~Z z-qYNK0fE^8HwZ!raJ2;a?m{}-x7@qAfpEM7E(+~%tmLVQV~(8j14EIoc z=Dl`o{iBPaHB=IB$g_A=r4gRWHIADG`4SmSIm|=9Nm;56u#BoaH<9Y*R5ViaIf#OO zUL&=yQ$N>1QB zjlP9Po)&bYL_H0pw!csu>T#~kKJ2k8%)eAVTZh}r5rd{W$0{LT-4QfU$ zy3TdG`YI|*cELLk@X#*SyFUI3lTaX7GsVdEYF4IFKj8gUfCVF8lLS^Pe1?|LTVCCAeZB*@}$Z5@Dt8 z0WIv7>3Nvq*@((Z+YlF zkccdUsZ)*u;Mqj&%W@;d#C=i$R%CT=Uu7y!eET*&fE7R^G?N+P*};vSSfH%*-Gia1kd(N~!yZx5PclB_hdzMjBLmbZ$3YL+W<0&Tw&t8?kl+9h-NGJJl0T8ZmA`-;t=O1aN2bUc z;=`KFU9f8{!Rke1!R3p{P&BbdIzuXIAo?^Tnbt^_@vP3Cl_H3~G);r<+m@!UP=mA% z3{b0%hH#$JkekA_l-z)8UY8O=1w|Uc=e;|5{_f<5Gvn9q|NJf-z-mHLM&vF0VTE}a zpNw!GjGy!^iw9qQ_Tbi=v)}(X_xYLG13#8TQfh?oh9HUpiAd2{-3#b95V5;?5@D)D z;d&pY-z_*E&=+ghiQFVW*sg(;M7P}Ne9QB<3i16`JA+t*iWCafE!KbA-9y^&ZRYAP zsI5}xu*uQ%G&8}w+s>%}ls1lVX+9JCXk>e-k&` zN!W7{6k;NO5iE!+t9`93H6Bh(j~7rbNDqdN&X8szlbjr zxEY}`#>AV%h5Z5mW8A+31;+YPj8LBlO(4XOECnjU3~PB8Z+RE^6Dv`OQYsvad91gT zku6x3%M;#fkc4&17iR=UArJ^*YQ#{E!v7EZ7H}B~q3Vjhd*0ou2H$--WS@d+xe30@ z6$I6p9&1jf^}pnQntmGwF^QYRyPzcxh?DE^Z_>A1zz|f>IF}TWTxmdM#w4|pzR*u< ztq2p;r-8y0S)l*3(F^XY$+E~~n2eKA$JGBLZQI%S#*HH*Bhntr<|c*}H(CYzQ&oeJ zfE@gT#a{W3uxMQv75Dr{@r7rlfnkeb*)WR*d`fL!VG8*m0qa{&!Xy{!bHH+^1IQ_* zF@f5#_>4D-CflM`r&m;$=T}q|M4O5#jT9^Oq;oolhog{GOXw7aGl(sgg=dMp`l?mz zUH;!nO zo+`!aPXqkZTTIqUl&biSN-Nmypw?d@*uyoWPy+$0&0Jh;O}2iv(gapjknJ=<{iBI? zW;6q4qc4%>1;EQv$ECaB(z8HRDHLs_LS?OFvw=sUHOax+C%qDBa}q;vpYO+^YSHF0 zJuuN4O)v^3@G8ks*EFg}jq0of(?O&pxQIs|RDu7baQ(T7rHBK)88B~?*qu4lWJ>jU zEkkv>(trbM^@XI?lt)yGHN^YbXFExagGdIPOcO#~_EJX_zWI~E3(8Q!r2BGv_z*LL zZW&as5vjq?2r^I=DOy4I>luBk9&i|eT*}Nga7+-Nb2HDdO%5DkgXvPR7kr6`*NDaS zBX<`k)*vW3j z{68DjY^z{(6gaqE^MQy;A-QfV+Xo0)J8~~@qe=3eH6h4~Ls?AXMpiuGDMx}XJPdK- zZG8t7Nmc7cK-t0VMb|ZsZ0;LUH0zgkdQV@vNeecp^o0ZdO#*Z-*BgWTS0D zh{irJmJMx)k~RNOABQdhR%R~nFY9t%mUZ_xoFb_^*DCg3IF-1~9NQhnANPbhr^pql( z=Qr1&oR)=hdfG@%AhD1X3RhHOd8JwF0sr2dI%V4fnw}K2X*88w^?Sr zHscC^D$&yHi(`+ouU28=T2^s^IjPeO-N%>GIKB8H!WfEBmPV@+20o45r6N2P3ad5i znJ(QKyL06BPvrJ@w=dlO_RiSfkKH*)ZvSxm0=aYOf6fZkR=&-2Zz=k*FICieDKC_jpB+*(TYU+iW3!d^W@YS9f81fGRVx#)Y&E@64;?7hIDF? zHfAJkqmI>8PA3p7S25_ zCbM!8MIBu{dimugOfRK7em#8R(a96w#m*o9n8?9sKv5&HUyr>*etm0l{`9-^#}AT6 z?|l5|?PHI=xv_yfI=g@V?8#sMdW6jX^b##Izb*?KJ@5~BXF{ENrK1qwxCSv03 z{9*siFn#G9=Y|X6xA!n-<)9)|WZ4cY z<#iZqy#NjO0>JO1U_5`1e!p)4m$d=NYAJmuj=e45Aeh<}MRJDM1^jP1Glo{qp?o6d z?q;gROCnT%{P24Q;l5)AhdyI2lEv3iBpXyS z$%uwjSC&<(gaFNLz>{qi0!lBfTjkbEO@F1|1g{NCqWzZ%_kokMw3pG}AUPML9$iIOSMchdK^3z_B^%$6cKZX`FDcH$ieT%^C z?Tfe16fU2}eA%6YaMz{3AM-7@R(*h*`)JDKE`AC*=&C35EV+=RRG_>1iTZ7Z=Wgnf zxfb&Kt&pGZjwU3;l7dP_{M>V%_5wa^x6j3iJ7at9_O-tsgU9XyI)nHdp3Fi&G+xH$ z=+{BXicIcd^cIPDqc~|cC(k_w7@|1wEzdu`GEi2fxlEB}EkPm1^{r1+f@DTv>1AmC>=62^RAtw}=k^5;QZQaWoQP-SlNy0*tD7`vPd92ms5m zfTT#WqDrB_sQU!-knRBak{dCYE@5Px5}lH%YdQ>vB4BHQN)rAa-j00UGWa*q=8ERD z!(5u2Y;}gQU@1w!(}SIl))Cp+n#njRhk;$=elSAI6vc;X6vYQcQGETxf@`g&{+@w0 z(@8MWB_Z_~^WSz{trsG^OnY3oMt@vTa7-xW!PI57@w$FKlv`RG0t)&5&2R2~`w1d= zo<6cn?cvs*it3Q9*yr~cszUq=WA@Pbd+%Sod+plnPhZbo`tjaRpH?XZk=61-kp92D zIh6a`o25Ds8LUqQgKLr;b>${~Be?~!;5r;Cmx|{ykX|svsbdzxYDs1-tQtFB@d`B( zCj1&OHEZfDvqsC#n+^J82S;BbNi%gYwf1mFcRPnI)z?$a)f9qFNo@qhqz`5_2}3Bv z%A3r<>A}c;m(FuZkqyVfmTcO8^JyEy##49WU&+$(n6C(8kUJPCugzH&Eo$BKT;G?r zhO}I{Zp)t}`ODM$r^oY`@{{=+`AL$W5CFK70(facTxv=nspBvn1lzj_|1NGI`FHZy z3&>dT^D4>z1;%z=fc+kW3m0L$)ym_Z7wtbw)FMn}3Pqo^hG>qGodY_x?B_~aQkAhv zPLfqsk^`>Z;#etI8uQqh62$U(CFYq=e%&U-`T=um#3+|@r;$W%OuVqV>*ul1VR@-; zry~OXgKTH;q@|F<)Y6hLD6q`YT+dQ1Wng!)U>&b0cI+q@W;Zo+n8I>xauHPuhb1|T z7_uel1p}@Mx`4+Lx)9dY$0{~DXUHpF=4E(xFT;a+Gd}^FT>eu8jfAK5QvMb^w;xWw z9nXIzUcEeh1YSA_FW#K~u&{y;!RrMmxG}stKM9s({wE?nvWQtJKIbj`;J@U*cq~}Z zPh?ZffAQS(`_p4E##;r%xB|8&C@=rfZ))VItD`JjTC-t~Yc_)S6-ePhC6cX9Y=x`3+NA(egrBtLn4Qmx0 z=(LvgBBjDCk#5HT7db1!wMCIBK~S|ZDFQ}m*u`Psy9XwODT^u|qSw#?2yN<9^y9fj zQsC@@x@980M36$54TF~A&1dp#GC_JhCd+8tHI6z=9%B!sI!&5O08gqrbTwrZlY-vc z$tvp^HG>v`@CoMa4L$<922KtcO7r3?O(r7upnMBvJ9BFg}wL~094whhuD@f#6 zC``IHE3tqo7eAwOFHP8lo5X?#ge6;&U{sOh!Z(PEuNQCVPc+vqyL6)chj?oupQt4d zNcHFfh2m;Jk}IK)mebekx~I;XT9TIUv<2m6kz5fAApzf?EoUNk|FL4r?Phyd%Wk_( z(>duM;n-n6Tt!K*dtq05xdHrw1CR66NmsXz9F0keB1^%tZXeiQOWEYG=Uo-tAf>Y7 z=Sd{G?i{%DI=ufm(3YgTeNANDaq3Wi=j+ zDxITVZYs|odHP_yQ#4!@*nKvQZr&tLRY{iQCzV22u{;tVVx*1dEYiyohza2P!Rg>g z_YroHUbHCHOV*ScHESV6;E9Clu37W`@I(Sq7-1JAH8$!xE`Jo@JL&twGvkyVVz zG%W5=^(1nrMh-cIXgz90(8ja4?X;5_FsN2c?TM0LA)NlrySm%@w*0BH*-XFQG~8AX zc={Gyr+-3h0e9A+db&uhsl7u1Nr{!`9sJZ(-1}_u!AGCZeR=%eXWwI~=j@k%ojrPD z?##)#<3BfbGkJ=Rep7!q6+6%@ptS4cOrEUcq9 znS`l3%}#@pr0aY@b28v!=?PZQc#l<~Efv5*HO7E<?wzN4NK35|(F!BFOC;tQ-9;3x=!|SKG0(ju)Ud@!=17nT%3YA1Y32Uwzy530 zrs1MUBwHh{kEnu%m}OC9ktCqQ$u`YMC3yy0#PGD;%vqrJNZ*iCEnu{6cLmtrlF-ls z_()Vzz#v1UtDtM!YLS2_x39M9|W*oFL9f7uO5INLj%D9g^S@yS2m@##pbnM~CZ2aG^BV>3NgPnz$ z{+)qsG@-c|L|j#zA&;%mVML^nD`>bhue~YVJc&@apqgMHsgwru&Q?FPNEzsv2CgwD z+qD7HKi~258YrrE++JPKKyt*@r5^Dvil`I`)D3&)E1(vNcLoB1cxy*zAl?;%Be^3K ziMIumu6So#YgarP4a7!0-d828=IP^$ah5pgOp4Z50y{B$@nkp+0C?lY5!XHPt-`e^lgh$JtnEth)IS8_jUd1w$+3w*Ndil7Ht>D zSp6*6<@n(=N2o|rckSL$b#+Xb#u5*&4~jL~xs zM`FQfBoqkN+5>kmP1j5Y`fapySiY0FZdynR?BE+n4yFrBT|gxk;1XMiV-8S*WWkUS zz4TzBQzT>>GO)w}X1MUWK>rYv0GU8$zX?7ZH{DkqY+PA*zeIYMoa5(ZnK{T!j6@)K z$l{-C2I}>t58bXMQ|0@a>xJp!8MJY@DT9$wdhirifttr$dKkmtu?kf4HPW;SsZ}LS z>(~4TlPXEmR9D6W{WXMaiYl})=%Ox8)oT(PH(#4Yv(JDmKE?F z#YoYvAwe(&E3FVXBVF7~F#q{pFy`fWIH1M@prpfcJQD5fh^s0X@@lLTT$Wh7s~*23 zNAFSupQobzCE>M}Ma&}TVfd~#C9l5P4zGZA2czh<*T^dlc%_D})2Y*x(98hVmla-* zqt%q+W-i{#v6i;oF*UmCD@@u8Us_Dbf}62$Dc(MH4dEfgc#^yUVZVscvXO;V*um?F>a=OPo*Uq~Xafr?g1L~eQW2M3 zE4_YPrPrzpYfJUYLuSOpbMMe(u88x*9{(kvB<~J#`R=`*JEmX;ZZab+SD1d4(qR- zbDa**G|d(|Ar{tG$;DSZ_Hadz<~SH?#5(9@36?Da%T-%Oo7$Gk@)onGt71L%v!Nko z5dWB0QCK*UP@zoG%}f-EsxEC8_1H%l!5DF{e%k3GtP3oWEVn?v>5L46jzIr6Q#I~l zZImSyHai8DerVKnjHIQUVEP_K3vb=BwTiW zhQOnsG0Xo}S(+ACpiZrx?^kQ&`y&W{9}AQ(W6wAfm{o8F^dfoYiaOd)L?)XUH&b8< zZ$d6Lfnuan($5`}*|z`qX=`Tctgi7whm+NfC833ClzQ{Y$ECk_?MI%mc%&5VCJGk$4i?ApxOw=?61XT~nijGw#z^Uc|#C+?m9b8`u>edm@P zb7#NsZ}XdUh%uLk?CR!=c(yTvVbHnd*1qSW+Xu&fxPR`{-J3t(d;iSbnUnYae5A?D zk-5v~W)EGycj}h!>7Tp!$=r<-vlH*mUcLO_)5G^}y>;)%Ete{bRbMKxn$id`Qw~OYWLi%@>MEA+fnE^^VEqm&$Sx07~R|AomNmW*s zlo0yIT#mx4QaDg|cxWf8VX&F7#3N322^G70;hkiVn}jAs07<+8r&*I{I6e&$S>M0= zarLCFoY-nMrFCYgrU#RZW$BPf_xKiL$0k>pCKMC6shcGdlq5`QFoIo#G;8^f#O;C=YrzI=kUqL+4^6tXXC1YrQ3TS7u9cP5Yy7a4 zC@YJUN@x%)RIH#3?X0Bh8hY4nT#7{z#ZELwMt zek~TzU{R>4B}lhKx>=6Z5&5ec3Rs3gBwBMf32goB;ncC}}mpjH@Q)_KHS*>>&le zxPTeDP%x!R3dT9YK}m5zD63&NwxdVeCWa6tp$gs(&lRDd2;zbm(xJ%Y)@3p)!QtY@ z;<6kC)oIhcS|KS8tCJP3uX=Z{e0?GZD|oHj8TSukvN$Vlo#_{nH->Q6G4F8K)waz- zSv$4U;HxIUwzkqdE2xx7q3{ihlaZT~I(s!;!}U5)ya!`5+d+J-MhfXXo6H$Re*&eq zIP`^KEwn|HsS2OrHAJAbIcQ#&RTPn3#c&?$eq$oYS^sDd zVn0f&$vDgrY>J81-;h!cnqB(pPbCZ%Ln!l@TJF8CC7ByeGEdx`KX`r>C7ENOX{A^N z+{{{h&0Bo!g?%d!Uh^Fdh{aU}%-BX~JZIk?zrTO{{-xLFK5RxZ&l@)%oc(Tg;^4h& zKl@J!tpv&>h#(A>WqEPNvgM&zJ~U9_htC{%|NiUe?qBrcQNiZt0 zp($;8GlW(3>C{IUp&aa^rGs`BYmz>eE_2a!``Pb+=ltH8%Z* z@OMyKz_V?qBO9=GwSyba*`&v@!2s`d29TN#M*9MIOJp}vs@A61xn`(QJzF30A^n z%&^kjZi24(D(Ljx+|~v`;TcIjLb|B|FZuV|I~{KF^b$KI6m-?U2fg|3AQ}udN9y+P z43GjAt)v;^TiDG<`~hElGc{63i$Qv*xd8C*llY^fV{4y#h4`lXKtY5gwZh->EC?{< z5&a^|xsFzcog=R}++O{`#ws>_o2eme@y6}cBG0oF{Os)L_U{N6KE<^@TxeGzk8oEY zmS8{%hwJY0Dlg46IL0A$%@`701ch!1cm+gP3?WD?UbNE`Or`{;-{1wd7i%}r zmIM-51^4!IKX@`{dqR zXXh?`F#FEIduP7C_rW>*`RT#?<0tR^`PkeyWB1;AeVOde=4}<(oy{C1yN!!j(9^HW zoc-*xyVtJGzIkBw(D~ULR~}ruK0D!`R#lOsk{pvIB`Aem83MoINsw1ak)joex;p9L zc2nE7Hn^+0mLr2(!65QeX}k-ErDy~k?+{2J)^Jog8Y?X{aqLlaVfY5+-nZwaHhbvw z?Dtpa-n%?|=+LScM%fxVb#C_RrMYY8{F{BImWYM|vg&GEjf7nYJ7JIenv7Mh(QtX_ z(3f-c)s6Xs$D7hEdU*NWM}Pg8tbh2^)%m}^`*33aQ*w+z3ZMroVNIVPth1fA5@}6O zl0ll$^xUeirK{O=S{jaTd9FZV9f(s_rAS4l+OtY4q`5l#Mt?c{&dV>)UcYwl+c#zp z9|ARZckMyC(g}%e(m1-S7wi%nLY9mT=6Z)?O;HrU_M)x61V#<0zzp9 zrC6};s6}g1U~|9vYBt-FHNh9d&S6-W1%sl0_h|oaFesW#c_PXp`Bwq9=vy+#60*Z3 zk|k_F9`h<7l!z2+sOBGC{{M6sGmtF=0~edn`8V)}Km6ej&9I+U4JsMn`=oPNi`42^ zwn_CmNVEHfa%LgvyckmoyJAY-io-!hiY&z{Fcth!PGIXAEikf{n0^=Q0QyJ!b6Jix z0PuAStd@(f*zgJxV`70o^5vJ~q?=-3iX>SIK8E~q_%FpP&soK58_F+;n=)nBo)H%t zhRU5JY2zk$GWAg17uJuq7pb&J54Q`ATXj~u;==;INJSN%XQ0$JjTytt09)sGPiWoN zZMt+U^0pGk)^yDRyTx$@zpB3uMvajhrsfdypC8)}nhm^I7gDNR5|Hd&LNV~M*D&c; zHbeH;gO4t~JOAm0rn_+s-A666&%BGc@MyvR$SDINV0AbojXI^e=}UAUTkqG#wE)~sU@VK+z2ybH+%~M z^G$Yb6Mfz7TRMr&_S#r0(n4NZx1AQu7Wl4Qa8_9_t*aI@ZVo*EP3@gzrwJ|*Hqma! zL6P-Kzo!2nOcXhpgHlKauiGWdP^_vfL$R8&3|1i(_RQ}ry1tRvnVSYDHXUHPPDxTD z38c@#>#Kbo?skoH0hoBXICtB02*?!H5jh>cmxuL;^L@5vgtKY@wIC z&E&`yR;7)5-LC$1OC_^zmrc{_O3gMgMS}m7Kv+^jt|jId*cc^*Wme_U{xu`AQ6qBx z$mhR~fBsaB$Pj2@r5sdmO>_FyoWZA~IkCSZQZ7Af*n6zT=GkSE*?pZAsgB;);K`;1oQE%R*5}4cA?Q1tRE6q*>AdjlcQ} z;+D!lECJkd3Y#Zsh4s!5TZP*LRM){ArWxO&*=b@1z|JK&H-at%62Ih-s)9W*iv0A_ zbDgGP*`&+VSklxIX@aoD;1jI+BwabYTGdgkK#H!k*kq53O+m~cwPLsW=L>m!F0d+Z zO&=lok1^0O{aNFIj(W@oMIih{8!51BrzSjGCZU>1MhbMegmhu_pHSFe3rJ}rU;#5? zk8m(821P-hjc-VpbkHU+iZm$}<$?2!M)N{LchQC()J)5kNGp=n7m%T-ctZiA+Yko! z;^#@trG$5A8O`3;lF+GTmF6T-m{tEdB8YkxO&af~ADK>CCKyjTVhxgARICK>)2_BsF8K8 z-JpxKIp~qJ86^~!f|2qdwr2_5|8{cr)Q7X5-JIEf;NG9Ff{9Z%}qT?x@f{CZP;H4 zTRXxr7I*bkFP7h8zc3tztllIe7`mB%ys1tz>#~OVJz24{tP^86ZxyyNIt#azk!O7nl6;!U`e?JJu5f08vLba z{sWZFdoYCS1?fz`-3(P{Wy1Limiz^HvaZ!Zaj+BgpZ1DNamg;dC3{v|YNQIq^04Mn z2!q;Qvi*Orf25K2J~>W|)YOR__ zOzU@Lg|nrond~`i77R+E`nP@$cu>~HHcHb@k4Xp8bkucR?wYx8TY!av{T5tZ+fAh& z4oUEztVAU>(p*ZiS_7+Z$(Qx7`pzOKnB9{bBCzPpY8HIe<;V(JD^m%@tZOR>mvm+{ z^NA&2y@;eWL-%1Qsx>(bLv&0gkhQzHnN9naPAn)XfrzBY@GIn2ngN&6^em~6EXBf& z(2lk-1}mumlHRU!tcu-0V1SNplM!kW%itq$Zc{TQc)(Zpu2wU7*Nc#W;A#}m-jj=! zNE^5FR3UxW$Fi8dgSF?sjjo>#YkOVt3`AT_3gX10L_<}`(!Eptq%m8PBrjo61j1In zv~DqIsu`dwUs~7Ym?FMuGKwt?r5jegbjuc|3E#g^Q;sdveTg*Z{u#8kax3jnf(2-> z?Yh{WTO!q?Sdp4htWZo+D*74x^x-|cc=OSTuba<;M0k?%1M?>*5S}C$STT5#>evf4 zpchnxQ40s^UO{U;{WayhTeIt_U9V{7FJZ@e1CB+x#-u>3?$pw6lc3?gmqqrLWSoo= zE%_ht6I(|{Yd55T5o@tm9QZ3Bdc5C#VZ*Z`_a3h2U@`|;omK(Bv@;y zPOh~)wH7=IvLdN!`J>=D6b)*2?&A})<8RL%`|$r~?_HbQsIvXxud<(ZmU~r z-Oj^*%C}s|&0{+R$f2IvlD4IdTHTuNmK_CC728QLBoJV@6Ci;=5+LLb#<@^+Jo8?D z=Nw)QjxQNP%~$xZwIxSwNjBZG6I;a0#O`jjCGEX>uf49n-`!g$XZ|+Sc=~pJ^v9XY zm+yUb&ihiNdEKybX-7gpc0^5Nw_y{ILRj)udF<{nZm?7gk@|UG9+ya;X=F(I114S5 z$QqWwnuV`|>y`?mTUH!anlRd$0)(L($*lDb4qF2qr#+uwEerHL@ti3#+1bjm*Wvxc z`!4qU9Z5NH>`uhUw&t(gOCP^j@>`|g-dN!_dDgo}Ufo3E8J8Pn5t737`+6PRqL~;^ z=BS{aLj%Xc;K}~|+Zfk4?1Yw06E=oiL@7kP^g>qTLjrS6HgF%n}$PYLtWYj-TZ15DlcPZ1~6mRgRZier}oR3(3)im%)) z^8q%yvrN_ix0ez>X@a-+H^Yy>lS@U}x~rmWttNsgaGkDEXsuOe2MkBB8&9LfU?DdI zhhk40>~iYekC($iNmf)zMrKb(T&=mb7r%@X4|0k~p(>gsk2P#7X>o5wevwiP4jl@| z`5(rgXw?%Y<76u{;E}ZA-n;&mZCrO?QMg^40YwTZb;rE7aD+A(peB}|M-lNr(7K|i z;Nv+Mt5^`-R}2vlW}1d6LJIr$*Tk%Z{jt$X&6X{Wa3pS4n$8F~Dh~s0OAkT3VJQ-E zPx#xakygIAnQX6~;0Z_JVb~W`!m=;CT;bu?9Jg3D%RtC}GcmaMv@*L@adRob)#&+? z)-+G?v_ZBRR*ENBwfHx+KEG+P_&2r2a-<_gV|n&RlE4WmF?kbMDmBL*89V!eI^(e20sg~!Fw6zDsl-w3QxB;-%fhK#8$r+Y+E4T*uQ^*lTDxSC$ zR7g?-b$8*Qd&_l}p=kLe3{;FyQ?8NcU>2!J{)X1Sv$AZR2p*NZYjEV?N=rAU^DMNl z7gE69SDBdyC7R-TA_v=wqs~AMI&v>Tm;Ouu>+yY|4FHB!24MK|{+UDfKYr7*bXO06 zA%HwLY8AA`s;0S@i#-KM3K^U*>=bAghkuZOQ#W!}S`S#cVyw@=7E&;9I&-Fxg~=ja zps976?CI;{@fwb*_sn*E@{dSzXeA|mod*QmrdG0k|FdO-w$t0AQVdFxeG=KlH6#^o zdNG$sGVcXVwdgy;;o8qcIK13L;jdNj@E8`y$9a|`j+zd-ZIx=`>*!<-AfTV$SDDIL zHE8g;?=Q@JHa>G=H&7 z^2-a?UwWx*$iNGC$NqNj^AB2f)Bc!o05?Oowy3-Fg_V$Jq0OB6X!fo1(?j3Pe)I}#X-+TLA9A?Ce%HEF zMZnO?T>YcNblL?2Gc+UF;8ZIM!>EI0U^S>#K#U`cv)zE1jN3e(v&)Ip z5EvUiNyc<)B;t^o!wi<)+QBOJT0{bs&gx_6mzX>uBvnUbaJ3qBRm_Wot8}63$bkp} zJ!lq-Q*gAX#t+g4Yz=EOxy`K}dSo@k6FL6uv1%Vys32Y17kM|S71-h2fXp^9Wu zN)>vJr>cWb3fBwQ3b#dO9#ib&g|R=2@`GE2+l4E|r2H`4b5o?;t2AFnbzv$CFDy^{;cvgq|Wd_6xz+es1>2Z21tHYC_*sc(&TwV z_g?fRM_x?i%sitjiVdnEunQ_}@yFURB8JIVt-_}+|45CjgsT8egCYPMEQbZ4sA4R= z#YTQCJ^4q80kv@xTpvyIo;s`og{lL*HyHK#!8ETsviAU4^kQhi^L|D zcpwrD*V(&R+|SWWE&)?Zb{WZBA}4SN6w=4FIv77IT6>%QwWCV@Wui(!%zF9My1Rk) z3W~@FRc|qbrMF9^v4m$}%mkca_*1Mpv<(tNST`6g!x_RNx$@l!Bc97VC@_BbtNYQ& zyO@mbr!N|$OK`x5puk=ee1}|yz*nh_ggw^N3&%j@mq4PwJlwcuUXVBY=dyV(n0 z+#NeKbLd#x-mU9h=gyuxJ@d)?Ww8{`qn#QPOlT6E|#oGm$NubmY>kh!L)5gxdSh7HpfQ++5U_o^Z6QgUx_* zVv%M{DnBU2?5K5%XwwnzGWeX4Lo!;DVp{h9X*lbmgIRH=!-imse)EM?%J%Je)~P3v zr_8%qi67Ux%aKAZFHgn&`V329Uu89w>wacp$hmIvaityBZ;$pzp5 zrF7k|qHdO;=^~oOgWluf71+$d+5x@UlI6@w8~q%9coxgTo0R3^C{uJ>r5F?>FM%q# zJa*8&ZzD+6XNq^a;G9ACw1+Tu(UE$eXM-yi+Z^YqtZ`SHuO6F>5@a}P={Sm~g<4xk~N zHe!Znh&6HACT*0orzXbZYb^;E?)b8uyZ^j#4mRodKW|(pqMwu4#3>Gk{~Z6xCiXvX zyk)o48g?DITL-B6LJ+hpMb%|ZWGE`jl255jB5df!73;PFdLqaE?_ z3vk0!+bN2`8*xLAbA2t#T0kffDWtB1$Gw~Nv0^B5tyQIWS!~se3^Tcqgw76lPdOk~ zCe^Sy-jTX?R2~Oh;ueIWNZ zCF?d2V8XLpwrd)Fo)>G7>MhR?VZ=WYteXO$ZBK7=bFrLgA;&}D8wiR@Yzq{XMNy_l zYgZS^KELFb6nrj1fx$&*;CtJK19?6a4u`_3KS-KVww<-2EiD5B15)0|*|`{#;2X6J zP&=ObbN^HRTs#>3|C4`B7|H*)k5TODcuL$S!REdHzyCW{`lJ8;Z~U`{t3_(HQrPDG z3{T`?Nv62x=jPd+Af?3sZQo~@&}12Erk{ill3RQ&KKL#2hf%VUXXqh4 zFRq}6R$NxUXpODHqZ3h-@=j1Zai@3uK`Fe^7YT-vq46B{itAuutf`e0@JT+ka(8xJ zT852U7rvUm@YTJMbN4>@tnuumyJHt-UVj6delNW>bL8^O+s9`@tl`LadJiAeKt>JKj;+G=Wn~?aeBljbR8XFF3K8N- z+~hj%GwWgvT{Q)&`fJ2Ry||gXaE;~B)agN!1CMLxCAzj%*J}$UE zQ0-6+ov5bPoP`*aoO;jS!HBT7H?A224j_Mjd7b!Ou9Gk}uY{^#{FhGL?V;&XMO;}v zXyeE7j6|E(ZQ3sE3oY9ZZkQ%V>c`XoxH1Q8r0m*TD^qsuAi6%o)Wr-f6zAf;FH4|% z(=p+P;$x*wt8sZ8V-nx^#E||FdO_fLB9|pmi&zpt7u(!158? z7f+BaxuQL@$6P>%s}$It9hCyx15YmEZ8*dgi6E$gETg7dNhd-WTS4S%xLps(cO*fr zRD)URow~{YnqxL4t!6*d46>8LI_j^ZkUC@foLTPL$+%^5abbi!^9NT3VcjaYG5JBz zZw3KdTemTwhh?$}MlivpBRmwRs|QlOyqM~?%(@o7yJP7rT#9Z-249;d40CsBI4wQg ztOIxTX_i@Jrm5v+qq<%GT zBVSLC9GV_JgLSdP7pF%?*V20M3|9@l?%b8z_YPetj>@x?x|wAgvKn%dN0*x%6_*eh zVN4fP=hD2?8T@+n+WnKK?q53h>u0YvUO)n?Bt9Cu|Jskgz3W{Dy#vOJ*-$`r7}3p2 zZQeO$8Q!_D$PZNI&K^To(=-w$&E&wqG!m1?ZS@>Z%ds}A<@9vV+^Jx>#~-{*6{Tt&*e zc*ONC4*qGlirVXOUK<8dm#$%^Cydxua(DyVYPHMH?tgUY{^3tN%XRh2&%zjjR^_^5 z0;`Gm{09O0*=o`h&5)gRKr_7a^(>3=?73JcEHJYIb;G{2CK4c%786r_D4d0szZSlV zBv_Be^jn~V&YDINw5mm7ra=>e0wtCw+u&6gm*5U5O_cY5_ZHLO35T@GISAQh5+V}y z$&#Y35P*Fl)X-f@M6iVDhc6by>;yR?ISR39maQ!-mUkhxYmD}bp}?2-ET7$sY5agA z#85We-6qW=81i00(!4%>Df>vC(IVZf>}wXse-LenNl4O@Pcw@&@i-$XtZdACUwg~y zP0yk72!i%Rs@T61%%zxP%tjzA0gTmPda~8DQ#q4t!&Yc4;qWYpU9d-(96mSBVKTUr zd#=3RTrW4pCn=FiUtq)VT(RN##m(P7IJ}02X9Nknef3)=C9ZkYfmW8WM4pI`6*+66 zO51xn{-ZTC2){^c=!F-aY;ZimHp(@YCrO;CIzm@q3M7oA<=p+`vTH88bj?9%K}oHG zX7BR=&3@RBqS8E&@j za2Ov62W@x% zdD>Xl&Hle9nSb1(W*QNS;AQPa&yopY?Y}B1{s=|@%MuC%%h3pN>`q1jpA@Mgee)ut zEnF}!QjKhJN8-YofOLqDs5$i?*r0_Go8&TU)d>5bSWs z#w)o*1j*IIfw~hP9^j#OQ!BL#B$qPGNO*A3StUgD;#*4~(HOV3Q^LkWjPDq zcJew1NmiFiGa`_ss^B?1HF9*4?LFkV-o4$=w|4fnHiBKVY}aV1m#akPVYQm9je3Wu z-OaeR+aVYSbc*5Ia6kJq{CO{oBW-P7xOMl|!FDZ&xUppSRzxFdQv3l`iO5fq zCb;;BT3c631$0yTVisKT=db|GyUF!nBAhB9Tw4W%mwRkvpA@R1;CuY6V8JMhOS+F+ z@f0&x$qM#fQHdf!O4zE5fm{6Cep&Jd1=K%gbOzfE6AKz}4O)5FM-O0U4~7kP8HtH= zu*@^qKF-X{#CiCmJtpicNt0$_XJWQ!Didd%HQ4fMjk7$}>LK3iW+~fpxRzckh(Lg# z)csXG6-#fc&lvDQq?;wMGr4kW-AVOiPEa!-BvT%QV&$@^Fky=XrMlHHVCN;hVAxy2 zV5+^%wslu+v)~IH}aRkMsGq_cf@m2?E*?79$6M|Z~_BC9INWC%5}*NmBY zlFZdId$x5~N@*K4^Hg{CW`-3hhy=e8Dd<>aL5Ix~bZzq}VM*~-ZeFfSPP#gp`$T%T zf%Mugc*G0>8Y9aKwp2hXdUve~UI$U}OkeE1jw^sat8puOlq*Y{{ zSwe$V^O0{iSqV{K?R8iyH#M#^LN$$3J?`Bo@XO%Vgrp!o@hdV7{6WM7bG@UqQ9YNW zi~a3lzdu&wW~f<97!3ei@*G$-B*B}z!0F6UowO#6T4ax#{<62Ta!2_~jvW*}_rit( z)|QHK3~)Lf&9Gh*dLWyCPYMTJ%cjN(*$IwoGmSSUS9LG5tkgeVCtF$33iY-?R5j_W zx{o*txg+i>xlsU!?X!{%WxFm~$x7Mo9y^y{$p7`9E%l`aw&YBNW62mM6WIaB`R2_; zAp?x1#S2QaIQ)a;3-*|7fRc9n`{XL;3F^?8fuN*Tae+RD0Seen`PwzKfgn^r7~3nU zL+R8LhG_{}*#@ja*+~ZkIyKn|ijpTu)?^k({lY=j#Azl=c~S?h;FE*i*G$KbZqlCO zahf!U>~n&0)#J9R;T7u9MNr%usD&OKX54%dWLHi%4bp_Q4sB*mw+47T4Kt!~0nVa( zciUB>hkIrqn%lvMf0x? zEiLUZGNtyqn%lMxZrdtO z{gN!n!A8r>V5tDM6(eG1+(izv<~q_rE`1q_SJ1ep8fe zK=K7D{YdJz5Mk=vozbbUK}vUyOkJJ&dg{BWi+7GhJGE@8W$Nsm!*Jy~{66~g@jF8V zFTia-PMw9D4o{sQbl!32wHkMT6A^r}!FaQAMqmu*#bBnsn7TIgEw*u%PGPpNEArf( z*QUHf5k9||F~nc)iuAi{iUe^7U?D4&+H_+fuXG{*cH-6FJ{Vimg$z5bvciB+QnaIB zb4qENiIs~bc(I`f?HxjNHbATanqBv9%Y$@nMJ@K?>(OU8{1P<}n9*(NJdnj8hxd}~ z4!V^eJm_Q)lYI`iUYxjFp{h?(!j&1StnFTodo3`^Ud)*j7ZNxCSB<$6&2mb{&|h?KoTGZkYZ8JI5Q>8ogy%!s;XWsx0X%eRN+$Lhr+GGby66c94h=Yc?d3CDU27cH`L}-I0;|>Na51t z2ZbAj+obSW;risTV5GkSFMLdz3O^Ob@TJM&I=C!ywy`Mb3CVi=1aq$n5{@uZJvQHwrXuO!oSmO+^g2C)y{qZ^YHn2}% zhmoCvX8_Uc61h0mW1%tAXXH#WU~+69CIcxRPZ4$?&X7<8M!`!ROnsbbNS2u4wzTe* zZ$-f%LnCG@;-B>x8c&dCq67uh1ygCmuF_S!+hYvQq;`rt>o7|LwJEj&OEYYtU{%TI zTgi^ugOEnrT3a#Rg|d(4F#?K@ScgfIb##Eqkrxv=GtV4WrK4@gZP$>3tj*yn7vFT7 zqQHFucD=F5!{Ya)$)j+5qQQhwI9(W@9D!|a9Jlew!;{0X@!iHVxO#On5xe9oFu<#@ z$=~w65q2gGNA+g2yhO@&nT)N1$Y1_YWyaQaYp--n+&l=Mo?#?Uvwy#=CH{VS{j0Hm zd`e@QGvivi5bacn za*dRI%+8Z-)QlNsWjdkDFw@3n(v@Ro+z^{ok1?>G+b30j#0`vWM=qWw9Y!LFTP;kJ zN#nIaJ2q|ES(#4Qq8Zp)QBER6f|44MB2vKF;!#$Be%A`nuSjxeC83edzvg&9)kIMf zjPhxkML>9l1-kHfYOJ4WYwl;e_1Tu}O)uvOR+wLi$Yp;`9J`cKPzw3##WaHMW3&Ijz^EGAJC9+G}08DBVddR-d&P@pw+NTYP?>U$|vdO4%8$!Q>1Nd9jF@X7oD3SUXaT zCz!+9aG2!*_i_cVJ5V>EHY~PnTu*QWbjUOUoyW5MJa|;{=ZWaaYbn8~!WO#_8TK9; z!Jashf?3L0Hg$Ai)e$05eUiVbX<;L2vodXa>D+H0zH|S~A<`sA!ogpU9QJPYuU`5F zWUNa5Naccborg{FWUAR@mZ6C{q!Iv2)BgR4I%eorbm8|_&4^L0n`Jeal)*)3q~9k6 zC12yg!%bbh^V*#wcZLP_><^e)MHJm@Q(xnSJ0o@FT2~RSJK#sCoj{;62(qz&S8oAF zd~3Xb{Yb_vt95slweu4E5o|V8&t-*Am~__AZx*eyFcXO=8N}Ap47JHq1pXToq3S`> z6ne7$UZ+1wWP2YjT!EjU<;5ggR9~1|u9G&V*AWZd#%LmDWMYnIB(emC>nY+!Wn4b*DM{8TOzQ(;MO#(HhBSalRS;=lA*21L6#i_WhAS^11{V**5Uq z^%yggY?XP4JX>~ttRDBvu9=TF^?*tFv|&x0UldR3nmC>0Iwf%v9#i-QPx}9S^U_OO zIJ_s8cqRn+66|4=4C?R{!|GrB-&16Z$r6b?*}_bH;#_)jd zO!RL)5q$#x+rK{ua)kqIX@Pyi=CC!|5}d#E(sP(j#>*vgJ1{UaLYb3z@mB(V1>tud ze*5sUut%M51Q#vXdD0ATabAT5Cf*`TO}y|FkuAG94u?7T9OUt`U`u%7Aw6!ZM#9!! znwd?G){5elXwbK4BD@=%xug+^L>$`Rk+M?U=B(f;@=L)g-aOaEgj)=gW{SpScjq&M zW)VL!L9w5gTG$LVadzT5vv41)Zip(q*n*%cvJBo)5cV}$9;`Tr_d=jpLdnFq!2B-u zoKAdUEdRF88(JdI9V z?J%@7wX1+Udtky7wn^jZMnr+^p&F%`X4bsjO8qrTT1l_yYZeT-JB|KYke&P5DxKSX z%(5}cmK5&Nf>M)xlA?qp*!jawY_&2Vq3n)R043#@)T%xg*T}g=04Y%(T7vKr#Ky^D zObz>cQ$`8PzpX7QBwh!)shdmQyOW%cHFt=`m-q}I<_={5A7Gd1uSO7#MX z15(1^qNo7zFyv6E#pjFo)QB7kt3KK9^TGcEr2skHiN3}%_cf_cZbttEeVc`_`X40t zxsM|-Vuo}|1Waii-p4ykWsNwvqSiJg;rh~#Oh9By!j&1Z$F$gS$KNlTY5Moi=)~D| zRPY!;wt;|JH%2*V+E$&$Q>LLK969-OBUEH{a4TMk0#v>h-Dj9u0x_&XM(qr1c)aUd zu4kO7MzwaTQC&_jG8iktO0V8^LHGXsg|`czMSnT^W-&TMN*6?H(4oTh!mYxk2klNb z#QE6diNWPwE?kGV!57H?d?Vg%tZ+^IyHvO_d3M5*cC?(B35!rLL&QiknL0Ohb^(%M>g=5(h|w@QbuM~m=*}o25$(0$ zyIB?)K+do-ag&aChlD=3(pa#3Qx{;2h_-+z5@hPSs07n82;bX;DXWo2$ zdid>`Gq28md~JH@?DWv<(?hqXhd!JhI)=^;yvy{^`_n^*=Ps^ZxzEE|smoSv&WxU) zd*|Hj>Cu^^M?8zPhaRNxNj&f9tqQAUv5I4tt2ZPC{gwTzeQrZ+t?4M)3myiscz-TR zUV5o3Z`paB_VXk~uZRp{cR}7-l!KSUJ~Bvoicu}c9M}6fINw~iC(e*Ti(33n}aZSTz!8Nip16CO{EDOmSE43fdMR59 z3l7enJ~{u^tFt%1ogRAY<(C`5{9bm!Pmi3Q9zG+Y>>pv2eRvF%+1w|GW?y@E?#pwt zuO4c7pxo}AIx_e1cQc1Rod4paXHm|a`Dpg7^Qc@udgbo;xI?J^a$BD`b}7<;6saG; zY!__OXpR;6)H>*ZPLLmjkuHmBV0si$1*;(gR}oP6AuY;fn~{{DFBlE_e7S0Gcl6|E!F+h>G+{jz?xYw~(@tIEMjeoG5H#YW%2v{xoidP^EbYiBHu?WBEfot{#jQM6 zlO@}GLNQcZrlDnJDuYEBs%leTs*mC^4}iDo?114cAAr;7Rxnc%hAE0UHPU6W3^&Mb zBL*i3aY0EJtjHaio(YWpWgJAsJ`^QT zHRug-@+7lx0mpZ;Omj?D@Zu1NSa)YNZ@5D z$`M70k~Y{qNH5nCxDYMUlRJoSMs_mfk7z)CMWk=i9;l5T2G#&+vYD=Ps}68YgN7AZ zarl2;VHH&*Pgbi+w!L8L-2Y(o*Vn!**|1=g{`%>;B}yB2RQKw^Gn$KOJdP8x1NN5n zQRevD2Z!%}`R%XAMmA5$LE+LX#&sB+J7u9IqS|IEDhQr@I@ zkpOx{wgfwydh!XK<;gC?HcShR12EamG;FGv!tKeLypBG!ttL&u)wsiMFPnZr*QQ?} zAO)((2whhaK4$J&Mb-s8WNcT?p)_qyY?Jm=bh@#TxvAU8(_Dr%@71|?17jKQtd7cj zka9!{s4MZ$keQ0Kn?^Rf0vPY0Xdo4lmSD3u{DXyBwn?PNW=vaU+U7dE02RFe*SMOS zo}oHSY(J|)L+v~uAl5a4ak8a&)!JLDjG1 z{v%REsT;Do$XAP2Pu8$39z&L6swXH|jhUM4wN?u_y`hl6Wbpe3ohL86fc>APFggD5 zLJgHvG{f54wsZ{^qmg6_&NNsS7VK#K?9cG0z3}bR*5-v^&I!U@8_Z*V?-~Twm^i!I)I2vLNkotiL%Z zn8NyaCI-IE8kR%_F<&J|Hp=z9Erguyf~c(<3LQN8b76!>>JyLcu2*W>G;%C8vv6n}B8+{+S>m zjz!c;VYueY17R#QN0P}83b!Xm8!vrV7=`aW1cUfV5CRp(3pczQeny(d=?>?H9<>3)UOV;D zOZ)ftMs>r^pbc+web90I!V53IJbUVu`R^{xj9!}kV452IKZi)aR<#ppzL%7Tzg%CfTUEhV(^3qdRMS@sGA?fr(B!X#d281arUzJN6j2 zJ4`5w2+xtqaDwZ}M_c+IAF@Q|?;qRb@2_)RH!PduBEf=>^}b9ha5XifpXpY5;xvhI ztcS%TD{E-%@2}&qJPWE*TLdBIwfGcd6OsZ|k!eq0G#{Lrd*{a8TPJ7!Hq;14Sw++h z!J5CB9vPh;IX^w}seqI?Iz9Zw^zgTyWidT6G(GZ#0H!!KJ?ub>X!h3zM)WU(ix@&I zsGwR|DYubPUs*=|`*;3z<<|XU-bqwfFQX2URiseWX0&?sH(1*xI32c_R8NzqC9;#| ztT?tIAWJma)7Qtry=}{J4Y8xet%O^I5;+ozG{asS$9GhG0n%r{CM>JOvXS;-|9&f- zx8f!i<-G6$Qa9}1pD+e=8~%}Y!F>!a4ZH##w`iR*OdTu&qmKwEU{QX`4YL8tZz{Qc z+{~kQHx6tYux4#1U%VW2G*lH&-0iW7Dg__QTIY5Ht1`$oZZfhxYYp&tY8@0hhcJf{ z^luP7uB!U}?SrxV?|(q<|MG_LV2f0!n$siDQ>bE-Xc=z;La8m*UV6%Gbs=#Qpns6tiVK z+Q!VZl|tRJjoK+XK=ZJ@Wkg1;+}g9vu;bJoEdFY^T5_go7TH40q<5&hj_sY#b~;GS z{AJJF@3L1Q1Of)B)!qKK*czI-F+TUrv6-VsX1@FJ?%4P*hfcoq(r(%xGY*hJf^&dq zhVyT{F?;eHToBdBaAE)c;-$H(H)clP2LTU)!Gk;eQ{<(WUJ{3h-eAOGl4;grFTeaU z+!8Yq`M5?ctTs+z69yI{Cru-lmEp?l*x2mo3Alo<$O3WkKfx-44+8V&mqVY-eRO#C zy-W8_U%&Us=j+@U>Ma4lI(FaEm1uF^ygxUT|23B_+Dr?Nj-$`S@3zqd9-a`G(v?n9 zR@+Du{r;Yd4(5q@@b5I(|Z7*bk&)!;j;*2q<07>N;=33kWf78w7kH+u0T54@ruGw3Qf= zY%FK^IU`4UQ(RAb59Tt{%(rUnz)JGeAnbq?s^WTfz219r$t1IEY$OuAJlEmrz5$Q% zRrq&(a+pjG37Vd3lOvOFOumX_Jp>noh&>%Ij6FQ~{#7$%3tvpW0$&%uG_K8Y$8xWD zCE5$CaDb+Hi(^)-;vv?Nd6sbbZJavY5gRU^xLnk*j{J2u^&x;AW?PJ`s2`~pA7Ps@ zfZBlwgoM(`=06HX0=D6UW+Ho<74M6AIdlZi*jr&_eDPJSYS)3u$7~mG)%z#TU`4wl z;8eZ_%UVB3bxvC*-org-;*4$U@W4{DCEVg)f=N^-C&3Fo%ZMG!WTW&#=}3YrSk zl%30igJ6)XVGvnpJd$D1G%K(PRvh z5g0s;85*|{Zd+UdnObT5_y)kytFT$}$hk)e5-m`ZdtYFCvNXnFb0lyFLOhg(qbQ+r zy>s!zrHH~(sH)<*&qHYF&h0y+f_@OW_g>D|jDXK6_{M&f!K#klz`_(foY;mqW*4tQUVi zK6SP%<{B?X|M%k56*#zvlLMnh>O!2Db5oZeyz)xR(+1VCoT}A~lYRTP6L{)VM$)9& zlp?*D^KZRAWfQvlP$qP{2g%1{2j(LZqR)pX|ac;GKxR@?b*4@R~s@~_{O=p zH;>L-Ix+usjl$7Qa@jIn82>+BPv`&fdO0*N%NR1je;Ay1#0`V~GBL0?b}5gL6qGA# zQ#S0HWu@57;Ik0qep^mYniCf&@>D3kW;qfF)&Z)xI@XSgL?kQ0pxr~$1BQ_xt^G{T zF)WULt)1?WTgNIUh*p}$#1&Q1~2pa~1BQ+o~5&~5K! zE!exI=Sgd>EIG9t=22Y-c~n1GH+4(S%4fi?S^um`6$U0eS;-s~TqteSjO7w=Y>RSW zQLfnqJ6LOmp~}$7R(u=pMh;9SE#_^VOUuEsDZ4g6)u5zQ6(2o@eRVM{Z_-RQm9t37 z&Y1t}a5n<9n7Rqm7rD}1YoCeap-;MQP*w~8g$ zYGpN+C)6TcTVyi85l*wg7M8H;c6YXJ*-0!$>j^3Lrr1sz>!!9N2vJ8Rc?3}wLsg9F zkMlld(SGFoGINjxcyXAt>lVmFmgB~DOk^lxwk-Yu;Lu_UNiX* z`5~>TM?-*KF(mJztbr{n`;{^%Q}M)|G!@vs!W*=xtJ~HF({ROmnc3M%;x1;&UNKf9f zSq5C4bR%TSDw%D}7|bDObu6@A~$5u^_DRkb!emZoHunF$z9 z(&QF1k~t0h&b5q11p+nt!5(M9q*#6}15ts96s*GO>~kv*Wbw%wyb+Qd8W=K;ip0`| z=q?~6`~EM{on74~4_nvquT<-t8v zw+*Y$Vae5%qmLjFml~)Hth**d#CmQ~{K3~fSYUz#_gP*Bso1no~J2|AYo&xX*VK+ z^}{g5yfO=-T!NCnYQtYLUcRZ9rZTMPa5`l3mmG%03E~zQCQVSxMF*FnN`7`zg-AB- zVh4B}Rr=Z&8ttUy8RMWNNo3c=`O>$ltTuXXZxyrPbGbORi_!CidZTbik^>=0@dXys zBzv`%K}K>d2#?ilQ7Z-e!m=c*suY1w8+063_Nh59;Z~M=53=vEQUf&Q5OYAT4N3^tmlZX}AS*@fx=}@V45-g(|d>4sL;m1pXf_=Gn0h*3gV3R*Wt|S-hu8KJD-qC6eE@ z7tz**?_a~5Xx&`}E`n8l#$R)b7>aS%g$WJ?C0|)efk4@u5XbHoi{B6Wt1|rN36=P( z;rXwYz$0GYi)Ci?{rMZO&7J#hX7tNC7K;yr5SGGVkU7?hBb1=J%ci3YkdbXAQ1_xb zx8d9#zt4wxb4KD?F6Q0wp?!VKFFVTj>u`AjvgE6(HK870C8y5bIXw0C)P+04cV4?Q zj7d2-yAJ2COQ*qX!MmdjxS`j+qCU=E#WDZZ zI}8!gSbL-Y-dEq=y?u24%b)I!e>(Ho_}nL7%v^eF=ErOEH(vL=a^R7dvGdH2QBa2y z;Moub8HFXas;75(c+Texvnu-Dh3)w~PPNU%GH_qpEH-kJX&lCovq(Q@1M7NqsXqC4 zyk`^4mNsVU)PygVq60M+*zV^l7nozhO2P|*@ItBtvQxqivqTL!D#cu;v}CM@R_o`4 z5FDmPzkyJsunp~xX~=>BJ1$sTS;J!M-_oo>8L>ML24%)Ou)^~>E!Rb}RCTxqAqQw` z$fa8?V`w&9MgEdYd*BRsXYxG(;)XDWHztn@xU(NgVSI9AQuuns0DIvQL1ux%&BCR1 zFQ~$&h0BGj;#=Pquyu&ec!D4pqkv*W(8duQ5<4Rc7w7D4 zsc>5I{4i_>*Ni+zk;^Nd&O0oWpaWtf=rS@aBWhGwt{V%^WXh>8y3-lVcu!j1n zRb*$vY9xy^!6*5DIrKO96?hVr0e-p9f^aR;y;<<+FhSUw z0O^URIG#t9UIMaDL(r3dfM zP`xM1`}*og3B{NRUz2QKfiIyT?|h(&lF)U9e+h)$jy_HMX;0kbS-S<7YMfau#eA=| zE2r6f(0KxuNR*KOxm-*mPn<`Cg+)6s9FRcspIa8y<8TDa*<`xKcv7D@&q&j=&oW~1 ziF2^oz(dgde-{1y?gxLzbka9*z8@}{M2@I|hD$jko$th11Gd+lxj3h~_o~3BGt|lR zpe~`Wmr>XX;=qKJx&cKNdzsyEe~GnOIGVPBIb`_cQXWxY zTIdy_u%iT5oVb)?1ao8cO0hM8Jl{kHF;eqc-akmq1H7Lw!86@tJ&Al9K6o+4V{sVT zVClDbMP$72#RN}s4BZ(qWY-|Y7-b(Q)4_zn#jO(Q;0eNQf@lH6GGOAt97+BeoSLj* z+9Z}sTAK+oO~bT`nKTEj34$#;i$dCp-vNPK8w`xK$?uCO{>YOefJM6Z$v)4|j$`nS zhsObY;xdMyM3sOf%dS&t8$^W_Vs&Qb?G)&ZR9=U< z@UDwjsT)GU0`JJmQq#=rGt3NHq@qenApeUT7HC8r=0DC74Y6x3-Y}y*0}OTq(rv`k z7D-Vv(E?U?B4!-$9{j0%Cx^&XhYnO-FC>3Z3j1Bh7?zCxY=PMq8!4tEidvHN#8ZYL zsuMfo2F4rW>svLhaeyeWMDN}*-7gAany9|YrZ6wY56m$o0|t!9PJx**015?76cA;j zWf&htgte>|U2DaFUkU|Q!V0&OC7GURgrb8l)$eSw+lBWPfQ+JLcewItRyAjx`l-7e$fFZPoku5(KV#{II9sQfY*)!5wzRIyJeENWL_PSu1 zy(Rc%;oyq6{G;fM$Fju2PlekA>k_d(5exMSwp6=Qlz|s~<9|M{?LfQMrS8={_&=fK89wm>s=1`|%Q8DeHLw z&ELLpZ}_XZcRusHc9ejxEh@{=K&#IO9;Ywb-k}Dga!~1rZV9w+iFUSy9C3)UeIdKF zFZ@wOL2bAa>704*-pIM>k*}vmu1*hM#OTrV@Hf*VN2iCc!PWVnZq2-Qs=)*{bL5-Z zljmly+?>1i&dh~3J&SGbol7%6j+G^&mi;8vrK98*^v6|ITaRI&HNn9Pr$nO}u5;7C zhP73KO!b$=;&(5Wh!ly`9ru9+vx{qKgLL!0KIDgo-I3`=KaDq@S}T9`xS_{IN1^Z{ z;xv=>ETG<0Dh0EYp_UWVTMjsMTEMkZ^9Ln=)zWc|KKK9wp#6mxqNG!|K~Oq#6s)r7 zRu(f%rV}|5R9|@EzeuNL*0A6`y+>c$_7eD}f?Rb$OtMo@vIkN5C4Uu!p%<$H?S~Pt`TjQd?$>kI8-k>G z@sfF_bNAjJ($&4^=}w2m?N$`qs7`H~p=rY*jEFPU?%MngM5J&vq_o?bIXM!J$V7k` zYv50iCPA6Ci^CF{I7MDGa%Ks#OT(TxIJO?#Ow5V%xGPws=}({+pJRQX2zMFqrxO=h zKeyv4vIR?F+4?V$EY=N{g|wC+!SJ`{n2Fr!aKfp^T3Rb-z_a|HpEzGZcIzFW z%`d<+l*%y0V`xDdiBgV!?RXzi8X-xos+Q8-QjPpcW3Z2|9-Y}Eu^YN_Cg=r9k~~Ru zyB-GA-Jl4aO3;z24Y;GFRRdW19rmq+j;GhqSxX7%2ZQ7#6-O+_G8U-5G$WM2ri-1j zLa{Z#c2%RDzn9TWv7Vr4$oDX_pLxfE+Lfj>;T23(g7WSxSI|Z*}>Yb84jan&e@x(%kB7R}RNw8<)X2J$1Sz{*Y>Z6H1 zj{b%P6WC{amg5c%54>&VV!1x}I*vuyyad?sa#+U!upcGWS1Hd8s-=}tEq!?Y{--dY zQ=WypYE(-o*{~FhY(%PbV-4+#KveNdfhvk`*KN|U%fUM;jj(f7z#(5aJ$r6==H$7% zV~6Ii|4@gPOH+$GViR?=*andCFH%*}yrgp5)WeLNR+*VnKtn1B%#7gHp$Pfc*tLI+ zy-Mz1I``|bbL9R<@7%v~0RtBIM@Gr7-;Tk>TA8aI-Bp>ZZkEl(G#*FOWrtDh!dr{t zm-{CV|2BN)*GrfGb@gUhfX=H?P)PSb0-g3I`Pa2WpsP#6Zovuv^~hmz|I|r;=>8Ws zYb3S=XUU?(*1{|=U@BwhL?k5n70C%HEBpNA?W}m>PUilAr1&d0G*3Z};UA|*ZcdMU z3P1mSb9CYKtp^vUhc8c$j7*P={`=;ujkhxI-gs^H{V%77j!h50Iz94saU^13SL;cH zZHWeysN5M2Mdj!g+20urs1Y^V+13`0Dy>0Bb~&IrQu?aPm=%aKEQPAPWpCoVB1AYk z^UC<_&3C3pZrpqC^LvNBxO?m5%-@EJiLo2g!{1JioEPb`o86ty>}kmS*!1wN>5*e7 z(BX^I!#@;9Gd(=+JyG`Mn=?1REQ>DaKnV`Ig$WJ_fP*Rr&-K=~xJYW|#`xSf$7YTm znfdO=yJO?O96I^ZOBiewF`u|ip5X~*kR;B{Ac=#|V3DVYW#?jKkZ>f#u|;3!Q)H(x z@bb&Q9QtJLqrx3{Z_J)Nw|%>3(RNV&*IbqhBW^*u zW#cmFGT(+ZaFQIZ#1C2lvgOrx=SFYOj2&-OaLd_~hvu#i&-{2}cI4`_+qZ9@yLtmb z?$5qG|LPSyIDO8us07BPr~{DXfTQNxb9HYZD&DGcCAufH2|ZwZDHXu3XY;@CnJ7QL zUKpPoF3ZPxHCp%t#xjH@e3NepJUt|}Df~1!0w?hLRk#gf9>OBMrI>G83U5vx znmhuxToHy)v)_9htp*oP#8H+Iaa)(5Vl*vKmD^cWKmPuQV{ZNU2e&r1AAkSHYrlSV zll<$_puh1sjtGoE0^=^+gk}DHsTTJpfgjgIHSTSA zyIbOtz#RPnKIjCN`Fa-kGa60cE=T)y?yQ;c-jS;!HgTw`s>OB1L{~?1pQxN}pxi_| zH8s*hwZ0-g`O0e=Ja;Im2r29V0&01?wPtZY0w!j%oRKDr9e8qWi~7Bj&N{z5p>f-L zNkhBQ*T>jCZl`upGtF$rGEp4pvaSWD_#|Hyi_c>i>9kXZ6*sbL81Gau-YKuB5U-nA zInB0+O=B8E_q^4@2w9{>mW7$sVB411clP$|Bo?EV;hocRt7+uoDWfY7!gYp2y%$NU zMUa5HdCYg&{r+0M*_Q29zS$P8YgA9H3(ss93{EGYWzspDQQKyw_hyN(ij;4Wu%tT6 zumTa;Q7;#9N&Kc_gi`fYVRgDiM?W}mI;}}$Vmu2b1|q&f(C9s~g4F=6L1M9YuCS&B6qi;XB_Jmcsy1Dw2xG`3J z<1T8!lfL-2OvSYYZmm)b4ZqqT{A;u_*j@AhX~QOJYcqM;^p13o4Lg6y@dJoCN4gA~ z+K!}pIqT#y>7+&j0WWksRa%x=78;5slRz|WW-iW137Z|Z4=}nwX3TuD$cyz**`6Pc zF}W;=mJ;;IQKj7%4F`PwXhiLV*$m5DqH>4a9^K+^Q=>uI-{IXas6eZfeGbU$;)zSq zMkHU@b;%M~RXk3UMrE+G%dqU`0cw(Life{tWK(&xb4W93H_Vc=fYmNGw>7~NhbNY( zY9w1JJPKBd4v?5(q$Sd-6GKN>d;*4FUI3~0Q?3cyDaA|%Z<@00Y_z2%jxrK>!IqTC z=~A4w#1okoll3`XYKiMs3s1DD!C)v*YqMV$9eBUT10|78iS$#QJYytl6xg_8fl`Vs zso036I5vInkisf9=qvup{eB}fw<4wguS@U#dg8M z$B_^jeqLJ=T~_3yCB^5Hlt6HB(OD&j?BZLe&P`pI`UWlNsdIN;Ayc1EeK&Ra=i_&d zBA@j5&&Tnze?CtB_dL8N!Wb@!o5U;ErY<(t(}V(mfn4_d9JR5OnLFsTp(3@7sK~O;OXoVo+8*i2!}fvWc0~1$J>(W7vn7Fmh->gI66-H&{I7?%0L-6BqA} zkI!CuYv#!1nYWM6oIG~#&Fk|&T)a1QZ2sn$XYrLLYOz8Nl#Qf-OZO>=$i)8O#xA*d zm+a>zPw-?;TRW)IporRs>+b46zLOL+?ImDa*0E08-bc2R4nu#!CfjH~LuxV7)Jrqq zoVS5+Y(poOp(pFWOia-nax~}~%S!_*h$jURaU$TWyTLlsVESu|_V6wJ-F$EU_FMnHd3^0uwe>1Sej4IOg8D zK6C4G!)eE!-P;0T?~|U*GMgV{X4n^YIu7L;;;_pSVwm}m6jthvrNhi`=elu{Y;G5bHa>URvvtymB8)=k$Ny# zKP%T`hlL5whY3zb9qMai^xh7i)J!Z@1wSjPH%DL?a9E(<4I;B6{y7%FfAoQ z&@Gygu+m1-vL?RNlhkC5w(`1Cl!3RU<+I!(Chvz;SPxiP4@?JqLz_9uaO&ut2nz7v zNZm9`NQ7=NMqnC@2D9y3b66Lvad@`c^KlIWE*RtYE(#b4B5TGqW58JxIhEwSQ$wX2ZhSs} z?v43x-fNjZ_rbm6hehdOiw~?W zcq>uM^M_=BI7p7%avX)CiYx{Em0tD30OIufWj8pSfrf_Rd^#sAnx-9;gPgpFidPAV zGlXatKD9D%sk_C$2MxNXpWy;r>)tT$Eh#Ck!IQ(&#Ct<@m&tW2$qphUlOs&;$Zsib z6I`LFUy*`UmE9?~4SMvTq3fW`8_bsIH8P}Kqk265ghjgHr5JeAJ^TRiN8z*|mUvH! z*VE4R%*6RDcdQ8q1dQw!0|sOQJ76_Ft=oVH&6>sm4to{W*$YHvEobqGt2zQ~f-{rg z@i?AJWDA9FNpd(}ZC?ta9De6ZT#! z2fLLcc2c?!E$Tg44nI!7_2)DW7E|uqIp=0sTQ&(;#-@fDVL2SXmyhGaV_KB1m9}X z6kLt3$gc_u3=U-@6ZU)SrOhy6MC0*v5r*iIM-jsaq!F%PP%+Jjr&(hC^T96LYvm6~ zp^zj8VMFi^T=O)|n}+UpbXrD&W$RoVyOV?rrdgGBzTvah?NWtMTfr2yQ)H{j`}(+@YJS?# znKfARYGrX$?LWnrYEu?8MBEn!)z9(^6yXk-wEE(|k&qPeNpeIIcnS4G(1fvOR|(Y( zhElA!;d-C(U*!L?_pZ%NTgk5 zs6Bvd3-txv#`+>i5ErVx-wYnEG{K%K1wr%_MiX@=l;2rYHLWxOn+=99$E!g+*`v`c z0!FUFp8`2V1YdMD>_LONc;H}GVdqX$lM+owApidU^=}K7 zr}MLlmca%D?_r2PS&R;~Ypb=v{UIR|tvV=&EeBu+fK7nf3rSXGwFv(TE}$wKDYi@& zcs)yH;Th|udo-PNF`Bd(Cl};&{sjIX`^QV(g=XgUr7KaocNkS_A{Df$9T! zI~nXiXokT%O=Bf%F&Y6!1&kR#a?On_IV!gfm%er0Dtzo%8BZk9xRWZG6g&?@81YCl zmFZNeBHdU9hS3rcE9%*ZxT54G@Egig$SJoeWLUjFy$k=p@Y5$ho&D+jPw)M70rf4M zIrq~CKfU$SyLGm3VHhW2An(9HyyN`-^d^k@ugVz@W>L8W2rH z@H@&2pUN&f3#%<)67+=kW};g4)J&;$4QFrHy}+It>kN*YvnX zw#v!WvcU(JQK3*skPy@e=h@ct<1cA7ze}waMZ^ zg#}-z+?IIC2{y)aH**tDm^>pKEqP4*wL9@4NuN|6i%NJPnR+2HOqHa`%MY}qP6yHI zApa)T*)G;2E1E^xjDdS!4SA0!rI=2QKB~fkvMr~utz<9kGQSbUTIP_0xz9^$OH$m&o{fk~c>DB6Mq*l0YxsVI&1?cL$uBo*mE?oS58vW zP4~-GH-@UngzS&+X9N0vX6(Fh-YW?XnGa zUQVTM!9GS)@NHP@!bqzz?e~&yycvSyYPxLLyH#xws(||tRZtSb)dSLOO7KX)_F~n+ z`WDXPEnRP^oOD_t=M=?&Z9uF#e!=n42Q@_dQHPjb4~0mW)ofZ)x*6+%d4fS0Su@pY zP&F~Y^Y9+a2TS)T2_$4^GYPq7GXbeYz0e)N0#_4;wzPndhi9?5$}FN7Il{C&yty4%Ijl=tqP;}}fvaWKb(MhB! z{WAAyYwqaOcslDpXaJY*3KIat%%0uGrnX?;oDNGJRqC+v!X956A8wz5fbw z3EV$&|5z)$I5&Nrz>l-{4{@mev-o=Y0%G|eg3sOHF#d<8&yP%BhO79F3;q411+Gtj z5APYBK2PBCRk(T#Z&ZGp7@59te{A{^5u*_w`M1(h)JK{FL0`cC3~9c9`2G>XRD$Dq zQcAu?56JYD`^U)im&la>KRiDr-Be$izL83~0?AFPp!j@51enUQ=R-~`XY95s#j59} zRo`I$ny@zyQG;i0U?U%(aD>E>232T^zcS@YYyeH$Ivp19HZ_4=N%G@xP&}u#!K@$B zA zv}Bee-g1Cql@7~ETI_iv#RyY9*`u4qP=|NPD2(duip6?|vXNCZO>*G`BtZ(4cjviA zsWfRai7A%@2TXBfn90h^FN1Mlrg$+~iYXb?W)-Mx>S`&?o_%Zf>Sr_GjeC|%OE9xyoD z(o6^6SS4HVbDo8=1169bYfo*#Xx;p9dd1OX4&=08b5pduckW-HLt^I`!Ttf@BkH zVWi-5kmepbBpW8|4c?227iZ1xZdGQ`6PIJq16hg)buO*0H+t18c(bMxwvU3HStPBs z9sg=*YddASpQ+w`gW4`#5sSxzqU)`+X|uL(K+`oVnJN#CbYr`_3UbO&mG?EwH`H)I zOqeC`x3^)&gQ@(aiWkx>J;3w@^727s@zr4~ssx<4KALcXm_k@ys307Zl1>x`=}1** zCv56=wBU48!=UjLh}Sf2%;G_xAcp*c7#2h+Y?E-wEf-z7Kt78P~ zQsj9qOrw~W>e5WhAYzNu@~cZliIM~o@D($;-ubPzS_(YfwnS98y-EUH!Y{@Bh`%iM3TXIVHC>tzFQ?^YO0YiX8RePOUa zJPW6IFIz;pv{6bf-FGo2Av18f=8U^-2jDYBWEH*;D2oFJI^c??8gXiZAO9zdp^>A| zduJ&D0|!=Gym&pffGr!_PW7z%q;|a;89X~>i}3I5Sz_S&uqDP#Rf-ThIDA#tQ+6}Z z)Ub`|OaOuOWff2rz0^n*!6hCma>V25Ygl{OjvQVXp2;5iEh>QBK!#9X6XFEuf_xgLp_OJ)0*3$2#@8+@wnEc+uJNX|Rz*w3I0O#b)CQ3W-XHEl z&7^ndVUq!}TB1h1fthh>`Naz4N{x)1!xxZEVw6U*CK)*qGjdKPT78Y@!_(IbhhLq! zIQii25mGE`+e}rH8d>uDZQHiZp1f7~s$8_aZ z;phj&G2x3(y<_+AdQeG_u&OK|+Ek@vXK;~R;W`$j8WQ}@Vg}Ro5_|zms9CIC#;N;F zbfox#L z_yxa%7}ZmQFx8bTQRKmzwn{w_^R8PU+vQ}+B)zFDO;d$>_qy5?g>0pYqV3&akX<^J z?Y$u0z?6#O2n#+3rPx!LN=-3EgK<%^7jCi_94&o~pvSmWf}#-imz&=W&4rpPm1c?7 z4ywduip@J~p~8YeAsDQ>01gpQv15v}RO?yN7U%O`hfoK9poTS6MZQEcQF5kci^Y2G z!;uhl%oi@--JeFyrE3WtHP-+l`Rf|G32RpgA|b$--_ajcXJy|oC|>9t4ul_Dq4$$` zI_{6d1Rf&OpTnB{2C1Z3h^FQbd-UMHljb57KnuXbdq}_ z@1`vlQ!-0ao%p#VOOTXokQ8hZrnXnzNWj_OOAWGD*D5aQb>5o_TLk zAv&!on+mHl2}N5)k@V@7nIZ^GnJ)IBdsJhxVPvI1$kw`Dl96i>NwCHo)Km?V9Zhxx zyeV__5ul)raK*h?f`aG(3-%Zq_nM;W(V3Bu;SM}Ug-3D%kA5XdsxD?CE{o1cJ0o4N z>PWK~4f>HIM<&nV#|TvhtSadIR?K2~&WnKXqp&8cu6|vbT<&B$X%_b5Avmk4zs>}+ z6oN_f`$Dny4pEAAwnu%jU`K~97VT{BfXVd3;bz~aRcD_?F={gu`GGrmf;|)RRZX7s z)a1ErooBG8uSuFq!ZcUVK{V0W5BYvr1Qg&%nnS5(f?% zr1{_B4^b;=8Row?kwINcBcUJ(MSLN@6!uAV))>9O*z0Ifh$zghc^KAQ-`widOp%OW=^77&H7BdgT_Ojs>o!q2aL@#yrMp5hxvX;#3%$F?4@6%{$|T@a^NE75dLok6-;%xsq=Gcl zli+p!wiT?!b)WTI#I`MSoR^-D;o*E}XOaBT`*!8POOqlcY6Sk;rL4gJ3e|{$q(f(P z(NX4z1{=tM1BNwVm@$rS+NP?MY}R3bhD?WAcQFNS#a@2-Kj42G*1Pk(Vo0Qmsd2hR z*1bz|>73YCo{d?dc;Hf^5g}YX?hdTJoh;5W=3*lt`IDuYOu&Rd27yH6{S29pXkwVE zaaM%fO)&zuKnG}0p+*XRYFYw5vLRh;E9A66_%4I-vVH+Mys!~1W9b>qs9`$c$)w$O zQZt@nCiz#=p-^=wMz*P@#s|2a?X$e=-cHO+>M)?^e&1;isDx}06_@Q7ery`_Iog|^ z+8q(=9H_Z=R;VAdav?_(r)#trSdfJiR~}3rn>+jV?5RT!51yL+_{*8sjueg_DZKdx zp84S0ho5~|4`o3s;=w<-`F7#h`NFrK&3tmSb>`T?nKPHFNepNSO=mJy)0$*kNFf_Z zO~|z-l|&&NEssqOx-C(`p(T(+T)|+Gl}XSdQpft1QKxLgEdyItN#BuwcY>hf!0HrCW-eb-*=ph@8q@ zDBHF-l}$c^F+LU`#_Geu{2Qr6M_ON5jq|i_CD8a??M+KX(zJ`g9om_iq-aSl6zISX z0#hPHqk}tPp`h479BJ#LqepvjR*9JU);z=_6#~|#oRZ+392Or z!xN$|CV$W@(wCCeA^xfZq}a*L+~yaAUIENkYJk-$O5T&!h&|V>X{upgWG#V;iY~CC z@Q}=|e2qPY=AFq@Es?E;a>2Rn$j5XQgO7wS>K}RH(yWyIV0qQ3zGfUx+wkxPy^2)j1(N{4$8n;+meQy9Ui( zFc7npv}r?CG?po%xqnE9txJpsRt2#RA_)jl-)cnCLDLxu8ntSe70?5Wk-dx!aifDD zw(!GWiN(Po>TnN!QD!E*4a=Fp;!+b{4l=Ab*3wmELEBV1lrG}BET~JHwsfg11H9|dQP$`l z=eXEu0cL%gnuOO-%Lbvang$dtKEyLJy!sTEr>dF!QU-2P%+(;@s{vEM7jFCS>Tbzt zak+?C56ezxiky_IAadQNDD)*6e%M?*@5o~To*;EFDL2e}yTHJv2WuXiAqzLIKA1c< zbM5+r$@gd9J6?GG)ZABhX3ne~2t3FyBm_gN>51Ab>xM~82D;z)lSON4tvs2qB+kj- z4!$FX$R@*N$&`ZNKSIYsILotex~=T*(^hsBUUMKOf;A~NiWr+J6fuI2b?>~frr-qu zOd_wMhXR((cG#i@6h^3GrB%897ERj#=cz;-;mubuZBQY^=RlVDVI}XJH&X7+f5jXQ|c# znHB^a7NYJ`o|RJoD-_TjI3Sx8*)Mq8HAwrwfj+tq=5`+$A#xhu`0T=~KgMaXZVS)A zuvIG!FL&>5Q{l4G-!D@*56+#ka9+nSUPc2?9al_{IGA zmHgO*XQAXr@8(B8&yT*IAA2W1b}T=7$Y!NRO4)M$+NE5;5^C(Km3P&(zy=H;U}I3K zn&Ub5Np2!{J$DCt7bda40r{JA-{o#TsqT&%Nh9}mZZdasY7E;mNbVy%fbn3T#vSkO zwq-7rN~qLx7;@Qi7(iOQuzxk-=w2|nkxr-X)K(v_{LPq_T17d+OKd9}99(UjDi9EY zj#7k%n&N88Ou-jw?O=u}!+&d`vI5Zxg(3~<6s1!|2&x7mL0sqHs#_Q8~SClqL>YLX66HIcGP!U)v@lqVagKGc~0g{STV83m^tCWkeUDF1#b|L}KvZ?{V3X(huQAgm(Q>`!9Hamn} zLWj^Mkj-*k+Gh)e**z&)MdC&zvD?h_0m3q}fvtC*WwS$Nz-0CfP2KPSB|-tHv~r zp367jzpxNkRkL(2_#H(y#`(4hG7OJ5az3ZYUaG4ow!peyuPb)fh7p|)X%&S~*t5Vu z&??hIwoxWuiZ&X6osgRk)?sCvsy%*N8_Xa_;3=z0Ykx`4Uk# z*e;v#6bC)%rU}?}XzP|X(R+1`CL3%|#wLQ0y&NFqwjLmeq?tZPXG&vn@WoLDfp;yI zMoR~+6iq!_o61Dd)RU4G zIg^pqC&X<3Plckx{1Et)9BTo!L^oAv){raYZkvAhLJ7BdemuXk*nS?oTIfcH{XG~E z0s%n`*s$g5e3PNK>`AoJJh=nnvOulPz6jiPh|0E>R?UPxwZFe`{kxg#w+hETu^9_9r*F?(yIweZq~5&L!kNzt*RRgJf2nZ$ ztP{y!J*=^67CZWTIDw}4Ep!&P_ zeFAo-A*Ry=BPQEtl0l}>AyZD8+zH+;;VXuRc*iQ8Us?*js6FJ0dP$*+(v2sge6-469u!HJtaJfp@rsv_Q zpav(&U1D`ie=~iPlLC(Z_4xGJFP~<4UDo>EirO;BoR_AC-4vWZ73-7}f4M7bdt}ljGt(Nb!a~%B zvNkMiF<>hL47kkqhp=BY8YtIp3Af>;wjtBP)S=Au^&f&=x?DEvj3K!jfn8mu>VD7j z4=Y5Bg*QcF9lpS(n12)R*AhEp(V#yPYu^;6KYxiY{KeUaB;r0ZjlGAUum~04$ZS*YtgF;LJd`D&}rBn-1g8_swdzJJo&V62s2dy2W3M; zE+`k7CTmdT?m6iJGLeL6xIF3c^feiN24DLs4B^Av1j&7h)KGAEjgw+cz)5)JeUqv| zvmwYQ2v2vpANBOXAnViwlZX<9S#d3sGE-DZuN~P%2zzs*4Rq>7v7~7nyFb=aA^}kd zM#@u88O5bU_GZ&ERVmq@*z=z!<1*by61p~^a%!%u38q?nzseR*>MTKr;@LfFy?M4h zx`iP9GOdD&OEXMOjoLg}8!G=2RCj4bF$D#wdLpi!!^QTF7q)l4u&KYFuyiIx5zIqV zix)a}ZtC2$y>DAvSAV}zXTpiLiX!RLE%5X(z&BqW>M@B#Rl!VY=_1}))gkZ@=K6e& z<>aMbqg~^MX0rNT9k?gKuUBl-fuk0F^egGm`2*EK4)cgxp5Vz@v$hpY4S#8^_}wp-sNf4X!h0); z?BU6yg^3e0AAY%JvAr<11cz(I`S~ku;O>)6GfNaXJ_Jvt_u}#zbZ>_-n=3{ek)aCP zPY0L@Yk)o##sseDz*~2%}d$zp^kNf+d;Y_o-mayW?C^`w9+j@Im=-bpe zGDzcCyFr+)Yr0Xq6&Cq!8QHo8K1rJW0ygG7Q(xXdho-}rbjAifKN-%(@0LJihcB># z#|MvuBm@KH+ofw7&~}-r!%Vdp`IjfN{Z*}ROEjMZ9#Ta%sL6Quj)Ml&etpSv9S(Sw zQWtpiU>w+sI>UZj*&sh~r@+=_+B&xWVWfxAcq&UI!AF{@l1Wh#c?R~Mb#83-v2+HECuTGM z=g-9WU=18=D|kD)X0VWVHP?|HU^<%K+g1^d3S$Ft#K9}m@CWAo0j;%1N50+68elwv zY@$LiQf@k?+*aPU{+9f7;it1dz4Oz_pZ*O0KB|Le#-$SZ>0KDW`JX=HU;d}};N=J z%F(gb{XX?i_}qJBa2FW4_X&Y}SR`SQ=PaGJ@{zNJk9mw!$uL!{hMdv0JrDzn1Irqs1mSicKffZ?Z9{i^4d~d2j;9i=}YL_m1 z5_++V7U{=HbDY8mLEC%pcC5(LSvjdGu8`CWW06KMS~WQWCd-(rX329}wjNp)(Qz*c zf`@MUQf!M3R|hSn7bA(S22~RS+P-S?T1#$ZkJQ^>?R_p`?Za58;R}@;=&r+9AezFP zhi1=xJNw=R{CVoo-1w)nuO6TIYHaq6SC;9u+WLG`nXsC+YG6%1ukj%M+CwE2PM>~o z^Jd}o(Zb=2h1=I2-n>hU_;54gCJYP>)oeDN!Q{V$M33@790oKp|C?Xou%-CPgq%cH z7@`eu)=rt1t_;FfK9bO6{#F4z(FBoAl9mmF7%*jK3jXf@Av+neIg6Pi6>jAB zHLPS3TZ=W-Ak-xGFl2l2I}q3q>g=%*C7(YKCcPB)_9`$NYy-Dhq9t`3a;C2J_F zreyC$R?jn42iqhXT(V%KsLett9=KaHLCjY)AmJ-*e<`|pXvdatWv?u}qFaWX<6_lV z7YJ0x5(JwgN#`bP{^r#kJ1Oj8xm3Mk$BUy0bWB13Bt#^MJlZO;RH* z0jcul1m{}nUXYk52_Ytpb4rdV5^-X z6eS_*NRKp%b(h2g?@SbqzrV)4_z|pY2}R5OAlF+KXlQU&(dstmGulL@!*Y^W3Q~m> zqtSJgXZ-z+9)uQvzZK^d#s;fm&rulF+ZBuT4rL?3)vlzd;0rh`??yDC7){v9kcj}~ zRl9IP$R18`nGcR2FG18%_WtyBbbpe&o4W%4zRyie9fX10%uSNqhq+TNe(x%-2A$jy ztZ4~2T%*QDv;-99M^qF7 zejySNM88y$Qg~5FVb?P93t>mL<0*8&be0Blks5}a#`5(w^u7EcA?m1(ZiFc?OrcOK zQ43yMTdN0u?NTsNA?h!8wOpg8BGrM+!0w<{JXLq~n_cj0fPuxG@@}fPGc_T{wSY}g z*#dZy>sV1h5=5sf)mRv?Ff?7KiPbNRYLZJ4L?;RP8~yh4TgU)4U;~L;Yq^C4kp-mc zPLAr3N49EvvB8mW;F0-*jbtMU_|nnBv+g%W4mHnv$-(ppldgZD1dWeoE+ zr(xcvC$SPuT^|)ky(T=LWCu&IA*K&GrPI{Wv{X1tW`t}OI)x^iHSQBcUl{fg*rEJ3 z++xpVoe?ad2s*$^TrXWSB^w0(LnLVm3o8TH$)7CEWLWlRFjic`?Z2zHCDyB9F>p7{ zwkXwLOK(#Y`jV{hoXRSYxFT3c;S|rZ@u;Rg3TNf=1vGp8)Pu=mbEEIie*exg3#f0G zV*$av_C6N!3GdP9{6Ee>8rS{QS-2qeEw6P51 z-ah#E$%$C;)X2y`PTziXFK7R6|Zm#lc;PLLw(Zw?E8k6XZR(0DWN-hFs* zv~c-gJ+$g9U03rMauQ zySX3W-!058;`!XQjU;z>>Ja=xpygW&XY0$J%e`G3EqwPoQzu}oQ{ym{$*Ci`3DTVV zG&cbkZcdH0@R=FovvYZB9DZK$-m%!Gn|(UXu>Cgc&+jY6%K2-Tl8y>K2fe0;R*i0! zpz1PYW1S9*Z=0Hc=XZqO|mVeutBDHZv*J&JkafoiP4aa z4z3F7^C+h7gORG&%1+MLl^_Y0(OI0++V9@I2)j{K{a6*#-(Sq=YpRkZ+uOP~k&Ko> zd_nlYzm@`4TQ-kuVJbtsD2!NPEBCU&C(Z9{xD{**#e_A9m`UC@L28=W zRrhp95$3|@pjBUXCv0j{($vhByT`fHu%B@vfvML>?k4OD6H}wGSxmyFHwniRuuDwl zZY<+DKmU?Lk3LWL)1lNn@!{Gl(%jpylf9n1HFX?zwh5B^5G#iGjyIW`TqYmf)YjEj z9+4`}9{C+rsF5vs*g-+U=BNNzbU~YcE!s%fMYk;MQKY&EHwaNHo14S(>Km=3kgZ@X zno>N5INIQ;d@`yUi8zBza6*zDIQz1tgLZTJKUY=wv& znZF9mtSG0naNznhmIp2mrdCl#Q3G!*02)hL1xjbb?(hVJtSQ{&Y}d8DRZw&~?T%>0 zRTy1S;z1$c=vJt@fxhtscujoENOFQHZ*&hksGc>^P;?4W+)Iyoo_J_k&$>rBAUS6?9r*!A=GhkpL@!`MIGJM;Hz7s)@~J@oUZXUL=X zZvXw-H{{XDNqF_>=!bth_(p&KqmK?X|NQ0cXG$M>zI_Mz`KxdKetoj^*8cwHM`vz7 z!|(Fpw|~EJu&$&F-}lic@4+xYYeb{s&xa3r4~;!~@6e;Kjy@Xu2Bz4)?!1=)MUi$R z94s#YcMYkx3*LoVl^_a0v5l4`)_wTpg-Te+wv(~8ER7Gfc5AAI<1`4h_K~!vo8Hx9 z7Z1(IWo6m6$c*SRG$Z5>haL9!v4xT}#zQTT;O_0}qgqJzSEL69c+yiVgLLS@ul;Ie~eOgMq| zLnkJz)vRZMbW(jtXLKotSF~~vE4q9T3q=wZq7DP1(Q~xK1wZrJ?ANc&y>fGQ^z+%X zZ_OM#Soq>>;o_wC+)Qoyk5Fn+Arf+go83|w{`l69x8VPmf4uVJ^-2+uEzW|rEjnXy zwk{$fKc4>a+J7DXuR}jx{IAg;um9H=`SI-^e}JLf`0+BF{Qk#p|7*;9lq3=n0)i;{ zk%}kcSqgKnop^BPlfqXw{=+5*pPjbO=aRP$V@W!K0W_i`9yKd?5PVxF27~bvt_|^Q_=lSu`vSEAQnT((<1L`l8;rrc{;e$XW zpAc}AO?it;#lfWEKnA~U?b7x#oy4iZIIxa^)IrQ3MK&nHX9}^?63mF}@&E#4xV8NKn8&R}}eAsjAuqC7ClY~~?H z^1N(3KK8p>ia=3{pt|!?m_HG_cgPrGrg`rRFyDtPyrWzNrL~h;#%q)bC4wA3Rl^UW zTgwAx@WJ}?pwJps3H(X#DakHPO?JUz>ao3PpzTD*@oFp8Y~xz2TvlVTd{d8poythrryeZN172AXzC!GyZTIj|8ELM-hTMW z#r(n1-(dOD)EK`uId!7&>g$Eqk2Q-(j=qujMZ$5Au1t;dH%;*BroSD0* zsV|QM>;F%1&*E%;>?_Z5&5yqUen|1B-Ldf7Dv|ktI~fNgSig;wTFp~xpWl8o@eOHS zn1e%~z^K+Rv4)by@&*SWeM7~9iXEbu6H}i&bSa7@A_PNKXKtPvru1HmbEMlw;8|F! zXG;fhVF^O@&4>+sxxBvo$A^co1R)Yx+Y$tnaYP7&%lE{FrbJDtR7G)9rX=m$%Py&x zMKR`+gn+NSFl1SLn5L~1{FNn6CT6++=hj;$fyv4K&#hO~Wy+Rc-0pDaUX+;$v-B7o zInVavf4+I|D(Jz!d!N9+&$8q{-~69jC*u{II2(iPLh&`sTZ;6gG_sYF?V7Ef9x;(b zzaR#~U=51)_2apH10@AMp;Vp;^3vz~;O5QQSI;jap67ZT;(3l{)=^RUnR7GW-JZSr zZsFpap4W;bf|ig#3J3!dC<($=6XJTkM1-h+6?vczFp83zo{i0Yet7Qo1^7Ap?ss#y zj`jEF$HwMS!bMz}1kavZq(;>!e8qGU>Imky8g|M$d z*I}3`rZR&JP%yzgi=swgS11m@5N$NXSKRc|RJEA$sWxz^%H^QISs6s_M z)_tfHmEy16i5+3oedU42xZACZPY%NOGu@=*z<~~)+BV|UeEDVaCyOBuvqJBkrG%yj z;6W&+zKvLXVZ&>ua6N%KDN8o%_s%Eoo!8|!=}2iAS*^3h3}fL#X)baIlS_g$m|_XzLR2Hh-4b!W1knUdxx&7oVsf9S4QmfFOS%VG6)M{ z2SkfUu%sd$#6<4<+~fjHN$wg!1f|>!xbzxFPC~q-?>KVM9r(Zr0`IxA2%E$)B&JT( znYoucg-B3S$KlK0g|Xcw%`lMHc~pi|5aRp02YKB z@CjkZ0&{IG#I;No-ehhBGgfCbkp)|@m={s-C0VfZHWKg+2EpFNGew5-wq5NcX>YTw zUu@xp@J*AgJ7(U#-AQF-FNLkZ!nx(JOFeW*o<}zE*ua#AN6TJ5Zu9b7yLrfWAFSFs z?A3Hreo|Z7V-ZN74GSx!?KO~Z45iX;=>t?9+DJO6A`fc1T1E8zCG>7HHUZf@GS?1S zASC$3asj#KEZDt!tE>#N*h`i{@!TLa42zSSCdltB22-g126>)Z%HA;2_7XyZcgl?P zGTCHgucar@1F&GRCgF=G%^29DTVRTr4EC{Z!h&r_UIa3NQ;1JRONM4(+n4t<@+_eN z`xR4*<2m@l_Oo$KvC^smS0n-^E+U1tS;@a0{A#`DOWT=hnfvXGstadgFv%e2*7ke1 zah6(PJWNf22|ol9@Q(6SyqkT=hXS^ik>Zid=PD{hd{uXlzOcZO+m=YfYQbVVLMLFeR=xQ{Z~p-C?3Z+3ES2GzBGOI{wsewc>Mk=(`WyBoZLSO zm%p373NQcr$o(VJm*DW~^acJF{sufl?jM4;BW>K-`$xQ6Kup!ML!@2P>0V>VhRkb2 zh7Y(7(gYE2t!lVKeV8cN$Sn~xfQ|KT5-nX$+YE)KF>*Bj?L14OON)cu4m+HkD|vX6OowKjCu$K z5)u5N)j*4E+GjEyxu+Fb9YQDz_zH`j|F5Qpu}uXhQKv?V>Qxj(_A;HcrKxJl8YX+U z%i#M^o9+zSS~*;Ha!3gBAB{AM=PKpVhr#WV_pCaP9xjO`>|QLA5UQRKsIjGa%(^O? zmLT0Sl6ci*Q}bSi2OXg85UG^TB`ltI@X%z)s zI%E=}F9bv7q0gt+G&JH0ZJH3 zz`9t1TM}ETDpy3dJ8Yzo=DR4o32v8}md$vIDYhh6#8(m@f9+C$lHhj`EHv5$SKBW5 zkMT2qpYRUGRl6oZ35&p-2sN&kW`r|Vgu{F;?{@gVr#PXVQIi>*chIU2yh3B%L+$thV9CV$Gtmdk0lBm5CxxO z-*YX!T~t?#;;G!Evjvo-P89~~Gp*@Ks*X6$fN9f0ej~1XX|8C;I7a2&>sBo6-XPnz z*cPK?!?P|PxD=`%jG3wBD%{ zJ~N~9t{JZq0G-++kl$IdYQvKBTqp%%K&N^&G5Dus7n3nAvP)JJnWhcWkpeq-eQabw zpWt%@8|8#c0fMUQX6XTongmRGib0abrA|PL^(BGk0u}HoyBwTd?87 zFXzMS2&!wu*8tv5!g|&~^>*k}V*qus0c`TY9+dT9_xKTQE9hWbSsvCItOff2PA|pV zQencJQfS(_qp3^^O*=LC`uvt_kKnXL)4+Nlo4_hdIYrU5#Q}}{oRwgvrJ%t?O&g|G z7*$6|h$29;jRGQ6N_mG|ro2HKK|zXCy^3I>3D#<9S+@Qo$rbjR9+;s)O*gV;s@0%s z0*q$Qi^O)1k}RSn&pPVTv-Tsem*~JzZD=&79gc;7J=8QcYn`aHsNj#51EN*U@B8Z% znyeuILr0401IS=O*NNyvQSeJui|*GcD#Y?1V6euzsH`VQGquc=_85vuf+a60=rZo* zM>I9bQ8niYBIC`YGe^HF+&x|o{!uv<*sC+|-^h=DUwGr={OGqZjQr>!&tl1sp3RTG z1OM42VOu*s_lqYagdKb~u07V-v$r0e{NnerHOSQDAGfJU3&mqh#ve8Hj~$9UIQYk2 zOCN-pfHQ-C>_(aVvCH~nNZ6BM$zN_CFI@j_?z68;I9q&A@6H+|EQ#j6k;Pa2{e`0! z;BWnM`&5UP$rh92!?IRP2;>ir)!B0}d-h=A%EZiD=L=Wg_nx>ZPF5#FBY|5)3sH+u zRWo7x z94Y=<|Hvl`gDc5)ak$&syFJUVbK9ohZ|brM&n7$fq6-K?$NF<^z;wz6O%KRP3ma)L zFx*2G&Tq^>T7Nt8HWKtvY`{SS$98SKORBspV^D`D534Qrfl-6`5Yo1Rq!2Llh5viK~PICCbkicao+qiV~#6Sjq8X!smsQ_^UN42 zJFYjGcEBEn$y6K~Gp*t;I3fy#t(_4|MfycUHo=DVUSZp#S&C*8Z3)`OC9(|&HY$W1 zp2gGG-od*&h^oa+YwaK`{D^lbMOK5X_u$s&GpB!;J@@VGtDipn>KjCGD?0@idEtx8 zGw*%y;O=Yr(XaF4ALqxe=f}s3NU)5eM9FBTii|w5>3iD%q@ShIjap#}eyAE*1wCNuzXR)ukgJ9)e+ThN-|BGfaVu z147haUaVjZxs-A4)2T7UZRJR{cXAVCfjV>QeK@?Cn?O{=BFJlQa_VUA4xGP1IDYLY zeBu`W)FhtzVQL&dF*Sy@V{nhj+z-V5zh}w(C+^}K@D91_@Ruj-Z^?ZCXK&*jk)AU* z$=`W5cNso~0J~#T$8EREeF-NI3WD&Mw8Ud7=`A0l=Z7bE8m93sm&8>LJzSK=)El{* z_R1H@7F*y4E>%WE2s-!@n6X^4zTIM~W{@_y2pUs*uzK!&nv_*a;s~2daW^Ob_v=4s zMoK0Kg$o`k%((&nD(E{ECZumoFu(NJ4FevV_GQ4k;*dSMDaRF7XRBt5q8SFZ!NTK(K-loO zX;q*>?i83XAlM{QidMJmqyDmRU1{MSXFxOe6dUE373?DO#C2|std4wl}_w) z0(o8_JGH$^5eR?HagzdAZ55HK4k!IOr9|UHGS*`=Fp+3SX(LkYYF8!i zd9F&ut2Sztuxs5Z283Xsd_Q~YSLL59EVE%`m!>6*DzeRwp@c4g+?Ubnk-(kYm%d;q z9A1(8GK!o^f*27*|A_5%Ip?MQZS#DWWV=QT&C=t@rK4h0Kpw=rjBJp&LgjRwJrcc) zl3w^~9QU%hUCS1a475nw(lEF?`9k_Rq zE?h=BTNQP$O5}7#(@hu_d>F&048fJLsEDLs7IlLW-s&BPMO+Elp=ndEBShMpDNg*_ zcV)F(=?quK>3cVg%IdeNRN7ZQ$qSKD(JzPrzYz2ZV#tOX)e@|%go7*V;HVIE;3axl zty^%kq!~o66bqv;Zo_wLm;zY7^O5tI%)|6GvxO^*9bT%2IF>~ zN=aruXHQkOkCFs@%3Sz(LL(iDHNfx%GI&(7`r4g|@phf=mzBq0_IXZ=mw$V2!WF|| zAs7~JgNjZ{F*sH6@@HTzkBQjKyu#> z%}Rd3fsd z?D?BBlb1Y8V&>ZQ2RCh%ostcgTyl#`4uXgX5y$@Z6!yyKay+H!s%#M58e~SCsWobp z@%wD`w(g`7_`5P4li2cnM`xtUr9 z0~!fU-Cz>Vd_|~1 zOeWjX$dQuu7IA651v%AgsN8&O>`!({&7HekxOM&Ei4WFre-cB26kLrW?4UZBH1+2J zVVkT->#9qU2t4y^6Ga9+^0@UjaR`U z^G^+ml8l1xWfT&8(ej$(2Ab7BG38Upk&!V-$U9$J4S=kI$PyZL09iGRizUwQt#2OO z_-OX@)rSX9&D}b-h9-tztPo^vxWL>6)@U3FQC9B-doFqjA_k5>P`*9s)^f#yJ2k`9 zhFdhXrE?w=#byQOK3wSj&rMF9Ai3MvU_W&f{*B}5+;=2*ge+|43{!9?ymKF-V7neP;4)6N~eO-rRYKc7vdL~$f2 zfJFRD)5I;Iik4+)MZpGzaP^og#Xg!{?X7(yec6mg2IYOAEc^R665ACMt~49Z;6;|` z?~f705n}q^4xYrcP7+)KjtE#hyJ6wnMuJa&mQkHwg@YD;(CmNa30OE!WNX}H49S1_ zn8Ef>b{Jr+BC4k+;pbIy|0u!Z|33fM<9|DNeEQ1t+3D~8dc4jo()*+LUx5)_nZ7`# zuiPIk4)6XE!pAuO^3wEWa{tKm<@-nf`$NwX+)C4OB1H{@su5~#fL0xK@PKqb(*M$oxJF=~?P%p}-whz1k*QHewq zUEYtO93;rO6jxQ4njb%%AOAEzexY#v&vjUydXiVDew`n`jfj<_-##3_Qn)t67{3g$AA8Kg8cmI)knw9c@}Ez@Z^G6O(i+I&l>w%g=PyStCF28t7=MOU9=7c zIqFYnHT&Ab&%c{_|78B)=-i2S3K!qREbH#w`LWCSu{#{M_=EiTk^I=_{r&mTujkGk zESxx2XL7@BYV32^xbx%h!iZq7vu_^pEWDjG9@oI8d5&hN$_!Q#n^fx7m0Wh~hGITZ z@P``EKr3x|J(?ne_q+xhUI4sZ*i zWthsdfN`d>|5GWL*3sq&rgc!2Ce}wVty@RTAbjYW`(7k4^D;9kHP-IfULHZ)p_&NJ zSPG$qU;~mOLR1t)pRM>*lx&id+{-Q`1Oip3#$CYC$R<_5>o#RmVK2W-4jkyC`#2YO ziD-NN)gMcY+9TklCsivAFL&>5Q{l4G-_QBFGpVeya9%f4x|Y!r{rz|)u292}4Ss2x zYQW}MJg?~qji2md89BkdWAyn)$RJJ2%370ddw-OAvuS*{Y{XM6O<}SX1%?$*=}b|) zmzeL8l@wFb*w%!c@TKP|>@RJw{I$F132T>p4J@i9YabDB5|nf^vx<(>+83CGkTNb3 zE|1YHYX&v-rsVyWHbl$yb$;|Cbct`h51M-J!WV^$A3XTs+nI}>S5nW<^*GcsHxAZ? zcFvE#l^?sAAAc3x{+B9fWl1X6%C;$)7gR7RL_$I|EcjQ`{}#eprAVOML|r!P+tf$l ztXwXchsO>UZoF5x^V!_xPiIdZn!EGr+_%S;$wKw*a!3Ylx_6z3rtr>}Gw)xTdF@p< zE2lIY3+W~ad~qMBBv4C!yXP0J?D<7kluT5J2FvsDPhX1wq8Mq4We|m?)86gPTCr$^ zu=26$rhaglk6+*5}1t|)y zcX{bpqX=oq7IplC2RFYdOpX-}e_i!faMn_B$x75u2qu5v$W zJE!9W`=f(qFj7SNn6As5TorbPg(+G9l>>`U42Ogu{0iBsOai5v=zz#Uf0t1~lQ zj70rJR!N&>m^uiBsxSvZPy1BovmG4O3u?%S; zR8Ly?VPa~M%m-*3fo*TncrUZEEJ_TT(m=(@_M2MJ!-^iZ!*hOG|Q! zEYiy|CQA>%tcap#$v z=r@~wTfI{AmC|4QwM)rV3I%&B3V#J_%f@cx$FAhZCU}iCR-4ryu0j`0$jsP4_`5xq zT&PMS#bBk_ysj$neZq)7B8pX;fMCoHD2(i-^QJl2dx?$kq?trzafY;jvXTXj2dV)J zuq^ZcL;3?*O|T4uA7BY`?ip}dFoaWQ0_;dxdOU?Ord|M~lqsYcz7F|=rVo)0m{ISPG%xOFY+#RRv-+zD>4%Y2DiCc%ttx3p ztn>1Yw@|JdZM4vCNuk5Y9~B9f7dO~rT*Ml*eut5{Dd;f7y_kSo>u_U|4HLCm znkp1&!;^6A-LUJH6VzBzTxx?I5#8dBgfXcvO6A&W+2;ES63Bg&o5)?xU7I?XyO#St zcW3Ig+%=M$B)K23^&b1u>p?`h&!=AD6!c?AYtL!u7kcc=uEhEbjhKeknmfw}eFsqm z=0D=GrMh^dB}QM^ZuPkrBEpcAH@tX>3BpoN>adN<48uuGO~Tmu)e4!d%}$K`%{o=% z>@lRaJY(d8zmt?n*S#wl=Dm8W)Av>rx0lAr@;c+hQh2*b7(e9APfJ(TRUG}8b@FB>wg2R&FOnuTw#GLeW9a3wd0yH6Yc8Wr}P7o3cIOK|v}NM%JhB(2Pzi#4fTjFEOrAF2>{XzK1O zudnFVRFkTsA;=rrI@(cL)4J z$bkyr+SG2>%oxf28BFb>#d!%FSk$4esZ<& z(TVjw`-?p2B@vXU(Iv^p8D_<35Lkn#dbV9uw501Pl9#l}R9G$GX>c?M$`I{dUUxg)^}>rfYJdwd;AZ zmFfeQo3 zZba}q9JofPInBscS=Wrz8dU5?K!6n}Od3K#p;!YRNN8c&jJTrUdcg=KWXd0hMZD)=E#$ z;yf-+)Z&os<1LQXyDA>iB`FxR+fI_rcyd{+6~SVyfMc~k*wrbax4G_lnbkXPti^Oux`-?g@oeBfv{$|5Uh@NJ1T$lroMqza2{ z`TXrJ4F;@fY51jSOOm)eV@ue+j1HzQ0b5>HiL4H4dV0CC=Vi3{xcbE zrTdvOpljL?f$<;-Co#;0;gsiexA*QUPwJxO->X^RT<@X^&rQz@E&vZ~M8xKJ!B-NW z`iV3$T;V_WFGug%u^S@rz% z@K4)3!X+$Xu4%*~SQF=j(rtK!=oJp%`*iiaa#fg_u zCub%v!QMG@<-M6R;}1tC=Z?ApL4Hh%IEc1qBXdzbMg4h>F;KazBxO2xBhBKcO87>LJGhTcY!DxK`)qv5=d^I+A;TzPVa~Hng)vs6bW0!upeRT14;qcM?=%vEM z33%Txw_mNVU^6f2J<02QKP!B3e)iNW-gVdi~t!=egAJLNPin>-^l3xh9r=e?F0QV~2bISY2EA2b$ zqIEpz6IT+@7N+S*hS6n}>hQ%Y5TuCBE~s4pyh8Oe zh$tik9UaTokjv$h`*dmyyKgx%(4E{wv6mLRb>D}>o4JV?IdCBNHXPl6E60mXx!C_J zkr4uCZ*Q17LCVALxp#|w&+v&iayLotZQk>H7n_B1x8d}`3a5(qoq7ds%fHQJ?h5}3 z6S-Ub@8BJ$-d~9C4_>(9w}cTMDqD;Bt?>Kr*gedVVzQ-0W4a{#b8uz=}@SZr~w|f zcMm)e6+#ZgSJ&EY_Ux^PC%@QE)gh`T_ow7P`h|>|{N?uX!u9XwKKpw2?jj4R5#z-a z-P!-2y?1SH;>z}izsmmmR6US6xUIhFcAoQ`@`XDFk8NOP7}wNRYg_8L)mzgS8xvCn zh7fX3CPQwMOdyklkQ*d~gaApZ#%IoJ`PIOgSBbHKKWJovIP-sfsAlA(=Nr}lSK*}?XT>!sIB7{QE>eK zezlpdf8?6$u&ag*MOsIGvieMZaEosFyh%|^CS_kx);kLEa7i)D#LEHL@0QKv*-ZA- zmFPC@YRx62NxX*eZP5g2qW|NKsZX0ila!EBs*^~Oz}a8J0A1+Q z_OW!v`@hp(V3tfDpFTD{>AiJz`WO*YY9e}QjE-8m!AA-9s<83;>l}beA|kBTUH{$# z>fX$e0!}YvCbsYphvsaIjd$zimP2AP20!WW;dRAE19&z|ELLQWUb}Ph7sDpJptG($ zgFY9%LLVtSfm9(LaM45CGQdV#Ew{)?=lbKJ2X7VB!#^n@xvC)`m;#$ei_@9y1tE3Q z3q#r(FEWfn!Wr8`&zWyfB^&=ud90E2yiXNF6~LdR06I(0r~ND24?4>vV-`+VZ2Z8u zI<%N13sL}qy5+Dpz+wOSjprBYF7Lg4xrsHGNM%(}64hGp(R1GQL%M4<4P_kZkP_rT z?dyHU1nA`!mPCJ<`sEG4v%A(h0#0zPhfk4;DDn)k8ik*EAEB#WoY zONefLGK(!ic@%vR0pSn=kb0^hD|L@)g5N;&LD%$($kr40^ia~zr~wNj19@Jf*6bHH zelZ5T3nu^P6YyJpas{r(V8&=9SsYACrx< z&z`z<>f<^ZI~r4p0EAN{A(g21szXEEf-dUZy?f7kaH$tiY)Ox9kCa>6d~#r#8mQa&3n71{_?;zC@m4ji_aYDw4WUtgR3{ug(nYVd_TuyrIDH(eb4!<|kK#xo99e{w%U{FCFL(o>mKkr+)*o!P>PG?9 ztWK+_bD!SxYOs%$u64lPEwdY?Zz4YvVk{+y@uVOrF(K|N{*uLt`Qp6}`Cx$g5=has zSa^x~a{TPv8&_tJHVqmB$&?aOYO7DKR(6xyY`CMF8k$a<(kv&5SP!7oKimwY>mJSF z2rsa|f7g1jCyMVX%(Jbk+qT|JgqAWkFwTYloGR zE0RHs32ID+O;Is#<%Pj6ub+Jj)*PcC!(YjxHrIau6>Bipmv3W7QpSaNqV5S|;2mHQ+C0x~v6!uKIM-o=T}W+PCt0R@(`zRJ zXVw`rZZ<9Osz@(~h1%4N?gTGxj*nix*y@ssveZfxQ(}VD!nUd{Am`^otWqG>F;W5P zA`5Z=@4ZErdAhhw=-v@A;hRdfc%>Fiz@^~Lp-ECms@3osEt+^Vxp>HC(MS6>fGDac zq~z*ULyIUtl%3RSfKG{yVu?gi1towK8=+>|ixcx$llIQ zFBXoGu6DAY>fS5Fte(HaK#^$<*ByhTANygH%IjnH5 zudZE;5gY*Lm9>kpcJKAWQq->F*x+%}s~d*BiEKAvSIq_&xJNFfmM{}6wsI&n#WREL zaPl49%tq*@CWJ$PI*Opmb@Sym_}SbV_ax*4%S6`CFEylOpG6&gL`|f`jvc)q@*~rr zasywWH|{GuiiebuK#K&Xm{*e;(Q#x0A10XE0-xPkf4PyJx^3&QB({lcF*wNj=mKmj zwZ<)KFg=$qa4ScBV~*pMkyi0=1Fe88q9{m#oP9*psa@_t*BbxBHg3R*#{baAEr%L^ zkYlmpLE-V@BMm}z>cq{dO8O; z5A5D^-=>S-FEZ@rR)O?VEsMa**~o<0jgO&?*g%EV#B?PDSr7J(EDs^QYkWSrwIo4| zK-eTnHsalvX$gOz<5ugj<;qhF(3Zn&eh+r`S!CC|ud|EWq|?m7j@pD3bsO$kBOcNg zPqi(jSzCX8>Pv6PgRgW}5+y-N3S!JZz1pAK51ZVVJaG0vGh>u0v)LDq-9B;*`xkG1 z|Jj|teRun#zs|n+{yO8OTVK93ck$g@Pk)0iUOj*3$m!cBznOjVli3%e*SmMS^kA0z zRaixXg{F!uB&53gxDhNvhhGEI!O|%iw0^kyteGdf^Z#>!?997UA2r*^2$5i3aoE)l zS3d-n(tf!5Dd^Fuk4%!8I<3RYAFh1w5a)-hFFPG8Z$Mii|zLGZYLYjm9J;A`+zNqzUofjII@t_$bckr zOppTYb1hF7Jl$Q?ZA2teu5`}HAsgo;1Ur`lj1p;b*vXe}ef(^*rCf9(+&=p1&1>gp z-+S}c``?4=x%1NH*(Z#DEvz$^TJqbYM=1h*^_8H$7&53YVZlWr(1Y3%a46td0|2q*O4P`E9gZl7AD>)J zr_HYQ2^+uQB9(s~qJ(9}$=04evJoql!HbP<1Z?FQ>S8FCt}5cKsDhXZJ$tW~X~tM~ z9a5x=W?|zUa96406`p&A&YgzQ-Z_jYb!01kG%{7T1EU$D;)}0}e(4one1VfI)o(IC zEX_ANsL5iI+|h>(5MV8Pe1oA6n{0zSMkY_qZku*Rt{s4b^DB`^6@^5gz#>wbka{qI z;H}x*wKS9p3@0>ay{s_K$j}V-U+aOl2A7}%;!*l{eClJ&&6&|n=^nS>9@4WuTiS=L zBZr070Rn*qm#U5>w2=Q+i~I*WaqsOfBU80;Xn#Q2xFD@+4o#oMi`=G0nsfmZ0=E?{ z(5GHA67pWRivwzQQ{Ca~f&Eqt8Fiwf!+r!TW4AZpBzRq88BVQi$97ZGv#=@ACFwT( zaoy6Bz6pPlnDEU-_YOj`X-bI4TE&C4EOp&pecPhsr%E`Q$XEyw55Td9X0SYWm_aON zV=Fl+P-_Igl2BA|m=Z7)qsucuY*Pi<&tX)=9SL zMqxc@t=cZ~E^K)wo%2}Xq*J3=Rv0IpV>p&}49rQ`d;*ezc%Cz>!$;ep7c#udZ-Uge zKvpDOTu|k@Q5%7Dz4URkGdNI!YFdq;G7!po|p&i9k7b=oqor)zv>LU`d)LAPRn^*w=e9^o)xpsof!_ zR6&Ag4%6SVGO_fqTqTVDav6=i>f)?!YXVjWQ!)y?BM?zdsB%n6t@MTP)A|`+&aDaG z>*}xaU$*!?2f%SxoBEuEY2$@oi*RI97b_22>M~CrW7#B5pL+ffnA)lvUuBU=Fr$18u}0N!Q0^4Rfu#v-LaigNA;}>q}vprPB7mW7~LOxoG5R2D6 zIi{?YHV;hvc;cm@p>B(Ld3oG|GKSm+f>Y$Pe2h4ZYI#&gww7lFwqcDEv}%q_%{A0v z(3?3*2J`Ssd;$A9+Erw%rjP>uawIUww@Tm64{Cprl)gi_M?`eIG(B0mwEhgo((4`? znA;&NBHZr5NpSWSBMVFI0zvgNUVgG+k@<&lL0#p z{AI^&pqF*ONV0+u9OR}>dO7TMl@yF_y)E9!SNxKw(8hZmj)!DfP+RC1HDgY$p3^(J z!P~;TWE1*4WW;2bPh>!&;a-Pd#_*%i}jMe>(ep9mOxN ze?RxCKfaz+YRaypvJwRpNl=rmA;%5QoSeJ&%s2P``biTrCo<;yR{?3(iC;%-loPh& zdYJ=^+3a+!Lp5m8*8RJ-Dq6I4A7Wab^s}OyF?6l1iY`s5U*D%2Q|Af`i-<;=R7?<) zh(v}Ju0HJ3L?s@W2ptbpQbHouDp{(nlp2r;sY4a1QjNHIOAxMF5S~7D@68KM6@-{V z4nT50(istZ+!&>fuGIk{QyW6CLKZo6L{&w2{JI0{4@=$CBIs~8-beH560pMxw%aX` zF0Mabz#6E1)FfTZHL;Jt+(e|9lPBG<6RtJhn9`^^fQS{$9jgKOgh-85pS?Es+K21U zaKE2L?7k8;Ii!~hlu!jFF#5DL9b$>I^~S}yC(qt`@tNq6*al69P;kz=qJ%CaZ0kmZ zJ0KD1hI4IxitWPA;w;e?&R3b^dj|sM_+Asi5XgW7g3^8<)_5`|twfIvowja)e~_yQ zrry@w*|wA}4*0S9h0Mh(NCr10T$Df$ZtsRJUypI)Cak4isnW4)35>*FfEp zK@M_1x);2oKo9!t2+>zb9h#JiAg{`z6H`fX_M0znUitIfTYtWF{>a>)ug<;l%*`v$ zE<2`WEz_f+0op_|?h6%#=dPq6i7lk6WfW60P*4@@sa68m1ub)X34+c`7jHlN{_N%F znrLogW3SSxH?Pe)IWQWcgD(bZ&MU=EQpidtpi4p!V|Azf9u4++=ILACy)gUIUz*51 zIJh!^demZ{!aU|mM?W)kAcEb1ChO=bNC8W&#X~DwZ*Ttz$C;md^3JD!p1u5WGdU?G z#NyS#g-0+2!5yAsmTnVAw;DFUsE&>dC9UF6-d4bLMHTezr#G*>F!x6EI=jX(jw%bO zR`u4hDu@hXA6*mj1hbe&4pogcI!dZS9?|oue7dQ2slEGo+s#Vy3O% zw^48REAG?d4bZ|}$THID4%-ejqr;C5IBeL<+ZwPvx{upZl|Yd9IwUD&A)aWN0R&TqdHrMz+v8W(0-n^^c*J}sRr65I(XY=GXrmo zYgDsbAG~hl0tFRERHp*9BP}UbmSRp5d?19h5n=)4$!Oz)3Po<~Ue({PI=o(gd!2Df z7yK%Xk&WAPIgqi3sKF=!>g1xcN;l_5fm)g`xe$GnyXvK|#kzZ45itsk|Dr+1y&>+T zkFGnCEC6h5wGcgHz@Po1)@?=Z+r>2>q(XFP4Hes4 zmB7&BmP9K{v{AhQb62B_G&v^31Jiz5M%j|M=#x+1Ir7f!KfQ71?RRf}e_9foXq;hp zYqGjHg+GJhW2;DH;=3#KD zX+;#`0lT23Z9Hw`=ti~)S!@-XRFhIGF$%20iAIYUEN`adQj)Dm*|I$1Xdc+{*AY0KYPjNo(9koX7&R%x)aqdJqp3(tJxil2^F}F= zcseEul2YxH(6H?w*FU^*^7=>Qz=0c2UcY$#qwAktKY!!N^k$=&@3?;U#&I}v3I3k= z*K;?H5epMf&QpjG@I1XoyC0dmqD~1TZd+Fwtm+yQReRgc4LIZJ3N~c;4 zjWzX^4m-k(rWFjtu@6`bL2pl-jISPk%p6sAm z>Y77C0z$g(?@JHvpBUUfG_>}?)@$G;bnYAwQYlsK=pN`H2G2!ju=yMK=;wp45X0hE z#Ia#GnW%0}4Ub>J3=N+}hm*wt4*%!D(5F%8nTz1SOG#A53*JI4n(w04AL&?h4%%l;siX=@tlAi9DXS&Q>a(wAuY5Xp>W$gUA8*?}*xR|A^l$Fk)4%)w zG*NqrNkItitx{UZe$|vr{gAjoQ^Xe@^Bq9PUDy!mjV_@i}#iiNh6)bA{hWnB)ZSmIChib zf}9Eflh{$LF01m4}yG~ZMc!B?jBi8%kiFAT1_Ql>8{PR63qhRahr+ghvCfPr(8ElyBU^IF~##Y@h zbXxG&$wY@RjN^#o;0*Vcc>YDl-(?wtlY)M*F(O&5Njh~!%B6%vAfpzV@3%6OS=7Ku zek@J)ni@CD8V5(^pez6b=>d85p}fYhVtZM)F`SrLJsT!{u)WE!xV>>~x{e1X#fR>u z+U;6|={cY&nL2J}nN{wpvZk(SBe|*f%$%MhIaZ)r-q}PBx!!sFOZLr6Q6eVOs=5jO z2w*i@f;Y*XLR#@hU(E`3@0uJn27Lg|arRc`>t1yZ_P`gZzw>AUG8 zrE8@N(#D`f~ck(lz|}*Ko&6@Xm$l z$?4|^eC8ya{JeB!`dEzzpsjyzTh#;5){pRE_~V-o#a;R{Eal>K4OFGJ-aY%mhjczeF3K%$!iFPHB zXb(MV&D#~CoA9d~N}hP48{Pq}sZrbo-sk4-TVFOl?R(?MX7qA?+#?uRk z^%!3GtqZ4tl8u`}$TxwynmoQ9Kcdleh9vn1|wd zxHfuUadiSJ;lq^+NtOgQh9?8Fr9y)?cKnL$qF59KhFg{!+~PJxgw0_J>str0gX$=p ze;twO@f$j-Gb@{n`Bjl0#W`fQBH^`o2N0K{1aP+LxjeI-_8ct43w95q&y9U778_n? zR9R?a>7ye$HsI^?i-A+PagoWiw?M(g?b1y$Kp%gcW-HW^;=ak*@sL^E1ddvcwTf4} zTr01eO)R)icr!sT0LcDwq=F0xAf+7FGCN=!<6zHf zhHLAi(c@Tw8~S)H`pgF)VnT?Etuoje=G^Xbo?vyhoKoQ{B{XAUA97&AO%%o&(DE$n z8q1h&Yh#?iFOdUwB}N9g>*N_`9(r}#=B}leeQv`l)#Lus#Lp0t>n0$qBpO6GWY6*J0HJ(=g2#=PkwUy>+?;^kS8#= znyP|qo{`kAwAN{qNWReA?{;5*f{cnIaTE}8=5FV&5*8zui*cb<~Hwp$_ zZ7c8>Xo7a1SHYtQs!!BKu69k7r@XczKs8MUvkAZ@3G9VU1~7eE6hVq`t58RtzY#Q% zg}B<{k{UGGW3cFT-i@FL#&_zj@GjS&YTQJ+ta+!jDy9QYBjQ%-TD~5qMvBM8w31ZP zNih~nt8yZiRunOrR<_8nsd79mZ`qPumG!KoK&1oiAHw!7Tnu$WUF3%kMXy(|(naT4vzyECR)4#2GHrnr7)k68$VzK$$v!Bdf zJA3o$>oZ56zy0CsH?KT9`_l6>M?brH`N-`L&UVaQdw2HJ*P^4zY-K#BvkIBoMY(o~ znY<#M6rfWpu2W0>*c!D~t$(Bn^GM;DCm|%0bytsJ#B^T|N;I!PtaE=?$Kb@^c#$K} z9C)KcLz{^29S7doXg}tm&}D~)(pamMcgmd;Kc4tg`LKOSBZW7)O>^)D6Y$@IB6OqF z+%{%$T_c^aqK(u98bk-JLwD=fu6_^)){#CAx`3}ju_ZABD8@r8wi*=gXGjsyOBVa1 z%WTIc1FYzvy=rQs@zk8?O67RkD@=q6z|8)7nGRDZx{(v#!uUgs3Ew(oDQa?ZU z+}*dY-Fy8#a`&Tich9_CM-T#RkQ5qw7;#PbgG2u4PM~q5FfyL>l~DScTf+9XB(cX%bg_U4J=VY+^AOP$MP80>xmhM6I7xw=aHl``EX)K7D%j*f&kf)h5%jB#4Pt z{iMe?rtaw3rt*!f2+3I8*$%8drl!X_J+KwTE*jXW>=u`+r`#2AVY~CX!PbP^>R(Cp z%k|gUAsR5Y~)q!_qyP*vgl#-Zi7yJ$_sfizzwSV5NSlE5>WSE zha^%;P{bB0UQM^iH9ESP?KEUeTOGN=WLlAglv-Vu5;}N#_sgdU;vC%l>iFIFUr+yh z>aCwHeM)}*^YOc9-Xix-UHj?MC*Cn)|dv9HP%)86$pZ)aBk#$ApdneA^d+*4-<8PCV3wv<!8VOgV`;`*{Q-jz?mOnyVm03)TwF{Ta#S{CA&Vh_ zrxT08!x5RpS;3eyJHvJby(*U^M9+Mn-<>Ag0(ClI1?oXB~w=n z`27@7WHf#1qKRLygIe4*vTHpxT$K=^g(9MeM1B^2%d>jUm#t4JVlv^gTD+q$6i*@K zxtwf`X0D~qVE3%1s2Mh|Mt zPcUfFVO?vahI=5=te?AHj=G$mW<*r#CSp zi78*}Z7QrF!<=rcT9CCcI~s)9p;T0h5?Z%N2p-ka{IxhGuN!)?*o3iiDa^|{^Q$6o@`n&Fy##c4#XTR1Zh9T_L@~p_Cdo;CXAtZg!vEfOX6r^O`$w43`5-kQ~qt(Gx4iVnW=ot-`H8KC!d&et>#ZPASoL4x-GuR$E~ zsX%)4ECE-rpxZXFLHIa%_}>2_{jhnKLo#m8-b9#XaSQ(9eQh_F!;}#=&QXW>!%ki0 z-$1|_TxSAUOeSP>j%`iH$scuy-o=&s1c0x!EVKkxBtZ<{=2-pI^icHkNdX;xW>Q#+ zCk@hI*z_=(C5CQ}96r3RNHbv23@N&Glx1|{bS&1jXRtz&ienrSH8F+rip5mjq1{Io zhLEJxUvxNuY|p=9};&OA4t#-DZo8fQ|a;Fpl|bV)99jVQMfX zRVOkc_P6QgFv8M!-Ci`qU5AY2IoVEV!85|g^5f+Lcze;S6Pn3bA*V-HWvF`G5!U5u zSTmTf1Dh}K@dctuzTW%tK}gc6LQHH8vTocUYuJ$+C%kdiu&nE6BU89)dbLt9;FWL| zJP=ik36fklKd$w$t*h2(V3dgo?6+10NlZniP%p?MYI)-$7Pv3vE2=0Tgd~&<8mKzi z)hY&ET`}mhr|-Ra?%p%eyM7v0{E3uRg{0aF_1CmmrmjQ}Z8X425z&Zjk)#$$SmD8Y zVn7i|0#B?uVBE^}Hz?O1xf=^95(jgzIv5*z_#04UC+rz^V-Kza_?<3hy3QfLiN0_q zJ;<27BQg;dK2_)V0N8(7MBwd~jH)H0%tK8PQnDh*!GXhJ!{MEp<5ELEL>CK?9llWZ zoCSy7CX{U7y%}Mv6oGVe%VLXHax2SB%txA7DvjZBUMPa)sli&A?d7-}eUxGdoiKwG zsHJiI+-BI(sn745cYMDj;AV{jrKdji26}ZCT9 z?42{LZLJUlh{)fwiDF%ZbxsP5?0w8_SdLA4S1O*x8)R(K`c^Re|NoxvFd^j*2#d-;$c?+!A{8prsAbMH`z$oU0F=WBteRW9vjm(h`tDT zB5hR!-)n+yHOOX!k6}Q)+O5vLziF!?OIt5HsB5@{=C?X8KzA%%OAfCOlBunGPXJFI zp8CA28%stACIm@R!F^A{e^P}~311&fB4i&7QY((hiB&M>er_3AvLE-GBdx)>8vaUZ^aT&zT8yw~BVff==UwbSwF!FWx5QXrWhu0C5N zg&(fINDfjnItWpN(7cc+st^~eeFe)6&myHWrE3D2o-AD`T`7H4y0nZe>)#uYW&P+f ztquI?(p#kq(?_R|mcA=}0k2*prK_dOu%=7XN8oUB`WXI=Or?|x(hMJZA>oUps1@c> zL-UR#3UUB0Jha0;;r;IcNw2#3WFH9I@?6%1IVI&2=`zwmI}T7m%%SLWW@_A^eiJ}a z1UV7f+H32x7?u>#Y7?M~nI>8`BedWy!!y4N&JyXN3pz0Qj{!(psX+=_{1T|0^5L)8^k#^_*=fV$H{|1)yAKmdE zt{%kz$p2jUpEFo3|HG9tIvJfhP3#}8{0-l5f4KUF21i%lA^9J!o?nKl56)l+QCwA( zKGQXe+kDs|1CE>3>u7@+C}Jao<@I#>NT<}Mow6eFa6)4ME2CA{vAVjBQz!4edim#5 zZ$zeS~Rv~)of`RAu0QNcn#~%(IT}R6U?CBw_(f>dfHcq3z+TTR8M ziOhVR2DJ)gcCqYKbRC|dd9%n3{Sb>@oVwZ4M@FgHRUuo29p$iE zPZ5Pg3lFho;E7|%zX^dSAWAKSue!ib*2SowS_mlFsdFd?7_JF%P$V%%HukyW?OE2| zGxcRXqnfH8UnM0eG%LFr;O{hkNIjEb*VUffJ!Osm)VA09tQgg4<-o>5&v&tZc zH?y$r!3&QPp|B!BZz`!0VD-1`jGA@^Cb(8@X84-NLiUcyQaTnBk^xkL&{Xl-nGq|8 zTP7zC%`n5DxGrz#F`i%Rp{W&z4E9kk^O2=Q!`^vVo5ea1d6RZ{gA{G%;9NzEk!17; zJaw(5EzObjl!QsEWNVmRO`dWyQ|BK16nCUaTu7zrPKG=xDv5(H&rDm79*Js{Mv^dX zr6|=-R*QQ)?s1~0t{T8QqAVl=qrzJh0g4PzW0Yoj1Ar&%ph}`ucE18u99~DUUjxV@ zA$msW1aDg}2mgAkZ7FF=e=5C`o_cMf^xgDi>00Sh>D$tUp&?TG9@C5HPo} zyTi&ww@W%rCl8AwokiHXaq6X77m4T_Ruc~$qriD7kH_HM-&!~NW!GB z6cC6;*q$CQm`MvnFAw?9olV5nz+z{nO*R}ju!Y*1uDOP5r^z<%Fhj^PVN{3R6X4-= zb&Cni5xTt>F)&wm^lhDR@@^qBA*nG^JUB6bRXcOqy{oWBM{_N(xC*YRYkHA_)inhF zBmS@IrA2Jqg-aOO(vLrmF*uQBB%X@NicmCj^Iszf!iM|w2sI2?r+O~Y(jAHX*lc6Q zD05(|Y*N6ev+4TBUy6xJwMY3N*e0$e1}EFkQh!-okt5r{W!(5bVO&_V_kc=`P_ zT)=YRRir&gSBGl&d=wc6C&6B4BEFAjKseB;^RUq9o-Nq$Y)RHutTghAA^Z4cgann6 z)s|c_tWRX}QA&zaXV9KQDEIB=Fz@KrH5^Yz$%N-yEcxGmk}WI?{HJl!PYY9@TI>*+ z&f0w(tjs7O0u$Rc8&B1E!}7XK9fc2)(lO4*2zd(8{hiU+(L{m|I`ATwn!} zpPa4S;HH1tq^PLAXU97T`ShqlLJHm6-!{lc#`%)CKX`02lkd(TFNW+p$zwaiR@Ff;ky%&})@j$NFY zyx_h5WaikbaEOtg8d%M3qFf!UANuZvZ!xN~YqcMiyy z5Rw6ih6g9fe>-)~poQ{W*-i_L2)fP$d2j(ulIEfa{Md?c)rxRA49rm8(N{Lbn-JUr z>cj;x9y)ssd{tmoNm4MJ_jFGDc;cm@q1~>hdudtqZrHHli6?9~V>{`-&aInwclK^R zd>BrWsSgQgIM96mxtlxeKmU{bzuKo#C(EEfrQ>Q^JIIfO& z07B4FR$JS)Z66GX_U#Te#v_ULMxlXMg--mtQA1ew_f`w*dAcn$=qdt^&PoWP;seu3 zrYdfn_d4VgloXTzrdDW2&a06`NCka(977}?TvGRf6z;Qlk=tNQdq2g4cel}b5~b=~ z#YPoGIyif*fupzXk4&fm?A|?|WgQE+5}IpS=s6av*B;$f$NMf35OE)XLC79#8Ie}w zi5B)wZ2>;Ynwy@GF<&kv>h8vPXy7G&xcUmwrrvdl_QTavWF*fiNE0oOXZrHd3DYQ~ z2^pKQgCU8~^0B^FC=6N=gPA!e?{!1rar^x7L(&k{2^++kj*KKD>TWzx{@)F6qyYJNZJ59AaB{5J-s{`Y^Xl-3O%u(*#I*;J{ATaTu;0zTe2Ygjt8C-pd;fu@MT?MAYjnv(Q+zN$=Eq zmX?Q;qhJN~b7y4gy<*Wjfvt>(n6Bkb27YZGKWOtDwa6GnFyKS(&``5+7gk-U@oPjo zcBbBQ;9-D8TdtYw5GA3?)QimVz05L}xUoseH%Lkft!QOYF{x=kggl1{6Jm|cc8L1Ev zvL1^LI~7`ahv_8?A{$XkMlVn6EW>_p1M2$>%8Ci0gtj15Oii^UkRIJ< zZ=8Ts(R*lh4%agEbf*cBKS~*ZxSDk$W_9QOdV1#QXLFx^Is4)l>ocEw!pqaXGWXWo zbFaOB>*Cefldoav?O)!Sd;Qw%)6e7TJoNcnr#`xM={T&nzDzC->{V1Do)T0)5}{bT zBu8AKA_oYk2y(2I{;S1GJo*gRYEDurhPo2cPzRELbco7p(qy{{RxFNH+072R(dpAe%^cWxuD#@!T=d+fjep7$U^d4I*v2I;H!)^Pske8$>1Zgc8i1Y2gYokTQ*i& zn^;(f$xz)SyI~XZq_e>0F@Pfm0lf|Zni=z(@V=<97&0!bV8ui{0IJneXr4kdqeq{D zA;8G?I?yB~B<1QxfRv9%;_hdIB;Mv?V5|VUD*l$!O+m*!*i#fy8YRO*^?LM-ad2dFQ2WixcjZA z&)s_I#O&vjk?FKcr`$~WMk6Q{8Y$svBnz?NY|;B$S0Kj8&R(ygK#c$W^hkkhlpbp~ z?=^~^y*f4ZO{C8?AE%kdmKFHW!o|>DnhXwjt8TWIow1J^%&G&evjJf!%rw>y2>Eam z#nuh|@um<(7UV#Kew5hYavg)WpZWau+h5&z_rtljPTu*;NANG6`SgQ3$3C1`;k|hr zr^VqQE8XViwKs2{`+R8VF*0-Xpc3}%+Xis3r~8(g+$=QL{gA^S?kEv zTMBl2CzCB9TZ^Cd5vLh|dWzT8y{gbj(OU{Nh@lUAp^^D_y&=6AF|nNuHW@Mbm`Q9b zmGj;#ck;p&zlUuDQ^3sqzHK4Q0=KJG4!br=@Ir-Gw(&zaw0dI!tTm8yX7n_uh+xP< zN)!}dU7Zw9RDzi0gOEg);Z}9$vU_wJ>j-i#Y;qQh5f)1fw^}+_s-B5jb(U=0YBB1N zE=zad$$#v=Wwh^`feQNsPo=}nO{50{WYPO`d$_4tv{0d+T#=b_XrYN%<>G1>w4W8g zqp&@O)A%T`>sl^x{Uy2#Fr!>Z(FuDzP)io17G8O6W5mJsnkG~psMt>)4^Fub zJCAL5xg+ww-{xe*HL)jhLwDZ7em828U5uM~YGui;&3n7I4PtNPMw8kj`hUNTC3M8V z{A!kh#GeqQ328!%E9&9Hgqgs6ot6o*EFmtthbbYzi8{Ob*WtU

    #`Enrw{N%&SaW^JdJ7lQReGQD#b!DYH)>9rf|G z?90;1VY9SA6-kIIE&ccl`q7}yIB3XbrX*BMs3+ync&BZhBNh*ipSKx$bh1{{3ETvK9ldBUgm&_L2HOL3V%h&FSH~hks&0YBsu|!Ky2UkXtEw1ZUY@4gm#n-- z^kvVs>gda!yr*{O>4^hslPN(-36ktfU`W0?Bq?kml8UY4@p!dRTsCl5LnzKnels)q zJp!8?JGCC#pQ?gkpI)6g`q}NLUYUL4!^oWFLAjitId*zg4uTY`F-fr{0BZ<9jT++z zn<)PzSx70X;g#7_CuUzcI&(B4ZeR(x9 zQFgEq9h8}w&5vhUg)Gr0UaHu9pHl0dum@Dx_QTbufNg%b`ZAV1{c!bpoJ#r-HRZoS zyv}@c>PzppH>U|7C+rC+xs|u8;acz4jXX0}Ux_SAK5!^0Bsvs^6NPvnLX{8OW&O{y z|9s=0=l*%>pC7{Cx5+=>`{!wR@mKf{4%b)t!9QQ}e)oq2UYz^qTaoFpH{)hfK2NpX z2cPE`XR5j)Ig)}B3k{xEO^%!2e>V5&-`2b~8Ds@i>CBfRnLm?8u*eKQNOX;nJl96p zIBNfKW_8%0FW{ZXblKUiYlnQ;=DwoSur;SDLeR&gy3MqQMV)k0jXEwR)Tjf@uC^|b zd@?IEF;qdWzuUZ@jWVyAy9_B^3$NJjz==mO0o!om6Ta8gg$4f0@C23-IGh;Nb_ixw zu*e_2U@KTVx|j7?NAj?b=glIE-UL`#>|)!JSi5aWT#?90A-Sq}9XS12-7Sz|2G1;p z4%Qp4z&_X}i<93`C zOK+B5D!p4C=XRxZVS3U(%tnU(m7vJF<*emkfrIrsZq)%9 zvFgNS99yQvWVP7LhZURM`p5{GI%hdk=SG+@bE@pGsqp34? zn%IY?uE7ub$IF|@Fo;cV+f!$9X)-eP)ge}7HoIRLs)nTx_vwxgkP45|okcpVw~v4& zHFeGyCmZ{xE}9e;dTu0o<dDLAgFZ@X>Ad6cbVdw?YjdBjqPT3t;`mkpX0`d$qsl!w`SN!65NWVCa&Cn^a+7)|no$R~xA|IvZ_&U>23qLzL(E=UpJ|+pxBVrtkU{@du^H4s(ND zoEO%evE7}gmci_W1x8N>T5J@I!JNr_UFa2KJEW~0_Xb(V>t-OJ(-TkZ(sLl#rY^y1 zSlJ-3LFTQyOkk(!fL*3@@;i{w?&Hzx19xznaU48Y9!Msjg_+^&U z-p_KTK6TEURT^Ef?AhDX-#O5}ZSTN9g&uXxtB!kc$}Nb7U#K+Uo_7*mu{5a4B#!w% zQ3|x>ghd7g)q)=4sGfBxD#`-2HP?WBT!)$^qk~t#5yEsc)>(6#>@{`#Qx^vHuB@G_ zkfsjs$T)VnN&#OdwASAOhDV{?+uy%=--P95q4X?8R90$ZGHTu&fCUGm;PY>Gh{=Q? zB?K`hNRm9UTi7ITkb~)ciWkm|BL19aJ?kv)mQ;X_`vHF z{`=cE6pUtZnEg zSA3S@f|v+x>9@hYkX_u&1}kcDiA2ME;~H+@Qe}B%Wy#2HIs(6`%i=il*Ct))yKEfX zM_}vcM+ybT5e!zxmf*=Y(Pq$GrxIOCNXQ?e2yBvCO)pnGt%|F```BB|QU`TqXBOEC zci+a;6*JTGlW({EcA}3_Ek8~qAx1V*qnM{8`Pk6VS|6yiLOF-GSZwjNJGPE7lWe7R z7(w0R8D?RmX#48urwFNdwQ3!XP{4WP#Ep|b9(m>Z#TzHCfAr&#=l=Kn_0O-L{nvB< zdM-_FJo&%pu77m>LjvzkUO#*Nv+HN!@~76H-o5b@tnPDft>yKR8z*6P*S~>9U4l1o z!Fc+{iO6*H?)kd+!3E#DqK8wV^{|>0WU0F3Gjui@=veyGLMC|nMN)dJbYc2P>7w`V zrRhmhdb9N1^wH@_yarnnrjNnv@1~zBeOtN&|6TSnt?P-RrFSqvJUuzTK3EC-nSPd( zUV<~2L%urw-1Ko+7%VO_O-rAZu6p~J@Fg%S9qA$4(G%EunTS;@#F4N~es{loieMM| z-LH<{egF0J&!^t{>C&g<=RY66d*&^2@6@%QE`37oy>S^{-8=dEk4Ih@8oKxP@r`%i zzxG(=Lp!_nkh>pz^3&IsE0+!pZM^r^wa2`>y#Cow-yB)vl6WcM@4fZj-HR{({Q70G zabXXRpN3UMk8j>Pb^P83C+|&u0vqkG6{szR#g!?Y6qHnT@GMG@6YI7}k?ZC%HqYAP z#TwHF2OI_$SYYCeci>{cD(HZFP@GDw8_bH{jOV+fk33TCQVwZ_QjP?6aPJJkKN}P;i(=UR6DDigh^+N+7Ld_8*%{q{$OG)9yBDQSb*M(>w1{=>% z!yvuP++ZUJh><1TUax30^6cg&?vCy_Af0^iyr>mG2}~5pf$pT1RkK>F=HBtkKV5vU zsaA}X5Y*HoKlHVmVi%vq>gh~xRBu@Vd&|^73sWShcbefAv0+nkIm0W*s!#9rGHQANykYjiOyD}t)DQd=vhwhF9}0S^JL zXmhrnVfFYGa4+X%*@h- zv$P1Nc$Qb)0l)hX=vTuo@Fm~9%hm10A0P;gt1q}uCDK)G2TS%Z@{_X_B)6IK3tc7Y zgUI#{LOwZi4Xn_X=&cgH)I#29Vqrl7`|0FB_9GJh6)ot-daP9Mr!VN$}rlFpm=my{ZI2uUcN=s=dO$egZquRIgA~g;c85CQc2FTDohO z%c<&^1XjBWvC_38hwR!(V6re77$n-#PfPWaF>WDR-uEEXnl~nlp z=2Bt%=769$k-qNYrbSE*v7DW)1*);o$sbDz67ah!NUAik)Z}4hMUtoB(W~|dmNi4$1{inCgDopf z<3B(1P5P(H7w*0GNr%_JiN&ZAJk^QRMAb1^mww=s>Z^Y~ar*9=(?dg#-Fy3Wg8rux z6Q!S@dx`w~!sUBsUj_@7+_k(M-|3# z0SwN&Z(k!D@4fxpy;DyTFv9Qt?VDO+T@~niKQPC?bpILG%mkAj4;+6Qo;iojE;s zK6qp9^7%U-zkch}r{^x7y7R`nvoE}J`@?tO_3V}CUg`W6EcJlu7&e|nj9)d-tQeZH zd#UAB1`K0Xix!I(A4S~^W*JD-?6_yWdB5Yq;h@(|ur0dbtKPw*^k$&|Tfp-=Tx05* zTM!;EMu*8n7F*JmT}OXO5{m_rf#5+DfZ5w0d4R01y}$nIIFfU3olL?nkG%DpLGO^5 zZy-{tQhb^US$&$sY`z$;yXfs<=MmD+ZJTsvscw-jBby^qET%S-QNVW!gtDVB%B^S4 z-@I~V{c*|cXHU(2adr074{m+`r`a>#&0cshKKuT~*(YC``|Qf>XBTf?dG_YjuObtr zr@NPQ+C^sNyo$wDcoblV%1R5&eC1%MT%~UsSP^Me1gvaF2(J=*Kj_R&3d)V!R zYyrcx{Rr8NMeuNE0^DU#vxzo>q6+b3Yp~ESs3I&wa+)}DFNtLdiNF}rM^_mYsN2Ls zuq5_4#{-=Ntw7HbBmDW-<}RPVef61JAHUJWu#T#N!BBS>?0sx%y-!J$(%<0MUho=o z@%LV-5TyE@bf>8oTzels*irgZ=}PHp=|bttIw&=NlM0XV4dtJDui39f1% z=LXraOg4LPTOjK(n5X)D*0=e|4%m)(pgX7Bm$Ma{6-l3^M<3|m%|45H>7)2+ve+ri zVk>HO(lfNQ!;wz!y?W=^hj(82ZtjJ37<8LE^8VbBv$wx{Y4*bLTUW2mzI5`|YfnWF zAB@;}Q4xKUPwL1&0WA`OlB&Btu|Tto7+_7444dz~Tu+qPrP~hbMmmNdlAdc|@&cBO zVqy#21dZwFGl0tMr)EaC$WB`1b)=97SOF|Po%B)H5B=(!3J?!Fha-ip!csi+M93QK z)9U075~x-uw_J(E$swV6Q#h|v4h*x-h8^q#rhS@pjyNusUg*qXWIweW*UZ^i)z8SU z2lnsj-??l1z#!Qusxh(|h-gi{mtzkqyIQw3w-59WlCT0dUOtij^+5a7TN#|%QqQE~ zUS=6fp?Ou`Y#TK^Q-M}fS7R~kLMdo+huChh$i1X|X56#DGd%0zb>nLwrmh(b0Zs;Z zfgxVLO-C6iFfwfM0$c;$h($9vA|lV>!tLa77laf%nFL791wkMr-QKt6xpr_#0Tv#m zw^)p|TbvfqSIKkBhGkO|n@V64v9WD8LQaxS;KO=w!g{&s!0j?Vj9Il(x~hgOT}`DW z5nSWYHFo97L5rJZ1oK*e^M;NCtk*J5SWH|=ODgT#_O$oveS<-gwC( z@*1+kBkENXM2rK&@1mxL;BzQEeaE%fWlcA{^w4oi20*ywNf)dGY@`4I(s^M2FS)9 zcoTGtt>rm4>~&XT$GDYa4j(fSW!4+r=Zup9Z!u^x^^^;4t>~EZo2?ut1Fml#USz=e z6sBxb$v}NdSX5xomdC;|4+XK9V! zl_$DIBhTVy_*g0`06_|~x3O4t_BN9XNuP}e4ruAE{S#aJaa0{p%1f(t7RG@pHs;WV zhCD!t`8T@3Va8{2{sOWaj8u>*Plh9?SomX@8hPQxGrc?_=h zfy-j&5o~Kd_qv{haP=}G({KkpNMiAXuNtr}$4gO!SOB9UnzZfIyvWHRYG%0yjbIBX zTs3eF1yYzgV-t$4PPCSBxu&8eeV8HE%lJME^Ys=^rTfa55B9>+7y`6cFE1hqSz(EN z6R&V#JUkaB1UOYM7kXa@!nS*1%2X%9y@8g%UK-DhX<`Iyw~vP5b)U>*ue3_^;>S??;F+ef^8;A0n*uk4K)n z{^gAmaQfoEo?8!jfg30P_1yoS``2^u!*6_d{XDLyLaQ2Yrf)on5aBnD-*^hoK81+d z1lIof|DM181zd3cM)V}yHVW22p4v7wbBT`=(KFkFdFtdW&PT_!ysnK@h*42);dC1kLKfmJ zF!}}q^t}D$-51~Z`7bBQMlqF4-F@68)@f%erF zovP?WjUv2gT2~R!35$UOm5!muJ@zx(p{`Z78net^AdQm_yhBY5+*6XVbbIO?vnQ&K z*Bczj!2Fa;aI`hOXaKwE79&{~{cxM<;DK+TI9>q#@X97kopWf$&~4kRS+GeN>cS?& z7B#2NQViv9VEFl*S*~iZ-Qv_aOJ}Y{iU!rD&e_C7$?_mUJXyvBP zc?p5#ZWQ@p>*KD$eMkgR1MT>fDMB(;-PX~nPhTPw+Er=xrRV0Zd_4R4d$Xs{%>M25 z+uy!(Yx45kxpyLytU>kaNIB5AftXxq`B0uoVd+yJT5LVU1gZnuqq<3Y=x7~T-!7)x z8SX3Alv5S)?7a@T|JXEL9j^(UqQa?JXG)h!-(!-cbglFSCSJ;uAWIj8W>X|Y;%Q+B zR_Q|NsxJZMJDV=O3ro8)eQf$^SeloV`rJz@U7mii!@KZk=_*|DC3YDQZ*bNZbq>t} zztQ}ozb8@jRlJDd@)ST)Nfu&(3eT2lwA3^j#<{ETYPcSsvSHSdqzVbS1;Wx+vw^f+ zd_lI)caI7?`nE$WlnkZmhQT#31M8W1)G$txV?r_@ z^jgT1;e(w$xcOa%%j_OE%fwVnY8O+Ehou|j&W6m-3%95#Dy+3x5N6+10DLLE+bQb%VM8cAw}m%P{EsHDV%cmSe2N?y(b zy0q9IT?CD8C-4g}wqt3`_Eg+H-PCmA7Hx)YcJoblSZ3zcj*aDYEswCB8ZVA}q%Bfx z@N(EeXU)^IMsv>qv*58L`@vYzX`zlP>qDISyzOMAmo*u#mlY{03t}LD^eC38a3>G$ z7g)uvR+j_TpmG%?i7JTo_v<_&wgto3FWkknn`+`K2|<O3h9RK zP^(d-ie4)Oq*4*ayA^`^1-7!roxH_0YK%v(*lQeHsfv&cjL8gbwBP*p`&)nhuw3W5 zXRy0-aIw~P_UkX^zP>v9%xPcA?CgsdZh!FV?0c`?zIg4uHHHE z^qqGT~dk%t4m~E9YH>iv`Iy8eK zVxgCuthNtBLoaSv6wo{`u8WE;JL$x!A<5?j+(lSE34Mjz=&@Jly;|p?B%tR@qnJ9YTPKj!og%)_5t=#<@&wAggdK`woo^0%d z4YG60*761e$riolaTnuQIFR+JoJ1wM?!694gOm`DSLe1u+Y$Jg99R-yJQS$o#fdPG z#Ijx}Y0Xoo26j_tj#{}?M}b=i|6+}}#psb*?-oAZkSc1&dmWB4Vy)@R^;>v-%tSa2 z?zD`&Mj3g2h6_zqRS=cByGXI!LbvoB95JuIvYQu~z6^NJzRhh_)N0$uOm@g+#@aw6 z@qT<{vWTgekVqmFiR6bDY?o@5drK=4BQ7N5x>Jh}eMkvz=vpIsGtD{?)KdV@_EC(P zj~5+o<;P7e@0I5|d0J>1A~etiQOII-2V)-;b9-JGFc<2pQ{C5HlUS_D-)wa>eQf~9Uae`d`#_Rh2 zh`>LeFLrzr{O5nq-#89eUHsQ`(amrp!>vO!XDw+isK6|V;c%9aa4A6vOks(Vithxe z*2~FGFeKLjl?LYvH+(7d`7`MO3NH}V!^QTJdC*YAl%qwQrD~E~dH5!Qef6qBHhIc5{K;iFAdPqHGYQr(Cwz*viT@rRb_ z|7eYNZ^USKBvnF4q*f&Z=b?oymf_qX8~^JuEIY!gB&;$jto@WkjjE#Vyspt4Z?{2G zF;e6Pn0s1u6JyJGHfsR=b^lZeN$~}ZLg$c_m=H^~{3SH_c+mk>gq?}Z(sW?=dTJsX zXZ_)@us|9q8`k;*&H?S}7B>O^V5cU_rpd0Z_8jB3<7PY@z+JyxJoJVFSZz_W+>MB5FE-@)%tWJzR^k9Wnf|M`8 zo(i8~SK@*c$goEl5j*$V`?o%Pc0CyJRcy4gOG%QMV;{{-o}ZcgaAxwGnPVqsCf}Vo zc7EpAwVBDQGslkM8N9%I>)0FR7h+6S>oehZ^Nj0IpY>ixnTv>n;KcediGgB!+;uTU zBrqXKAzhoh$sn^h8K5>bDcK@)3t$pbyN;CYe%&Gc+%)wHIpP(07!NHFRY{Q5x&fU@m*gAFqyI<9apPD!!47!yGTy5l1&PLphSZ1 z(rua@rKSVR^W6FJv;Z=A_Z8Kv5aT6wJYFR&CiC|37Ra+3v0r?3alnF%}{B? zl7a-LwWuUkWlyLnK??Z0q4m#!1B-E(naN8iu_r$U>3!$u<+-Py9U4M0t;~M&r6|t5 zby7-zsv;Zb-umz{BElEPemwHR#=V2xkIhVi+<#`x^W6$DsNmW=Kt5(Ob>Lp^Y}d6z zJ{6@Tk_lgfaQP?3r+cpXzg8Wnc}Vkv?eYo$eKp~#AP z-EeJ^bAh!Aw9Y1vMtR8+;;)H`Y6By*myC0+Kor1wnWhixBO{F1BQ(Pdj^h7U9gmI| z-37nfU)xkZPIn7~dDmhc1F*iz&U9>LJZI72=$Xm}c-az`pv6`dTEtp~oaK4Z>&~R? z*AY}JOSVxDoibv1Gq>RP?{H0)^hFLM&Z7kiV5wRCQ<6F@Ew7KFqS6eGNy*;~_t@n& z(gO$fFWe%Hujcg_>6wQmAvU+X@rc{NFy_`cSTP%MBncMc;iya2wP&!y&btmD9`0E9 z&_MY!-N?+$nqZd2E&IQQhSvOy!TrANoF{*@?dhQ2RR{)U{1y0A~ZS==|F_i$;tKN^Dhe|A&bz#(#efL~#w5k2cfH z6q76`a&P11bTXl2-e1-77F%TiAK;Ygdgj4|HP7AB5ybS>X2fxUn+;+W*hqyLLxWrG4XHWlx?L#+jtjb?Vdo_<#OM2nQjMNrK!Cll83K)jQqWRM&3n zkkH873nVBio<>kn7+`b+6qKW40B0?O_v`R#!n$9fCIJa+zQS|u>Lgw1ZemrXrAg8# zq`JDg(tH1EU;8@#E^PnhFX%;w@YaDk9zTr{&ZS_@XvO!F+I(?J2exgUa1V4h=EV(o z#VsQn!w7jr#@qmc?_Q*gs!)u1^mE@hCWJ%e)a*a(mo{k8u{;Om+3``SF!^1_nA!g6C^f zi`0ToC7qXCXD<1_Hc(>80lhHr>;6n-$uFa_AWZKQ#Cu!Ot}O zjKI$V{EYMSpzS}i=uU8wyK;sIzr=f%8pT_IqRUS_gJd(^s==WHzo-DeSrCM$p7a!; z`5II2N}t=lp@* zP$w8?bM01xLDB1^h*@xkJX8&#Y%@g`lnQ!{2DPgfw5j(FPMv;b>eJ8f?|&57j@I1VdJtGqBe}*Kq+3$H20Z7$K;kRYNmt8AOwc;xWdu3QaUzM3&zxL+7{ye zJ7Q|YCG_dz>Gv9^8i6~5;naiScLc}14-83d1P>G0>#Hp3B8qQaL{)@HMG|t}FVz>% z3AgC20~v$fI;z2wJc7X3GAMZgYgy1rh(5%u-8#v!kpi5csJ*{uE^WD=IVS4nJlLj% z#7HE9nlmOwd05ctf}!ok#geum+rc{BjkOo>3VQ_lS#JE)52sKEG1D;kz!-NE`vqsq z`py!-iINJF@U6nAyiB9qiRru1YRSF45j4x>2As|iyR}Gbw7_mOdwdklzI<->qq7f> zT$vp`_A8B9A09mCwpv|i$vXS;`Pq}l9v(T1-`&vB^zg##n zd+Frt$L|Dw6t;SezYwihCC+fFRHTLZHaTFK-`{^`>fkFgZ(f>t^R0V(53B?;unHmB z#%vb!A(K!o+jj5G*JjTAGa?4}qLgmpF$5zeSlnYjt5bGqO2zVRHwZ5GF5jFw|H;hN zi}%hPTnV65)aIyD-(I=*>SZ{J@}oB{PM^GS@5A%=zP>#3&i;E}ADFrF(Y<|N!7u~k zw2m1(*P~DZDH;(ybvi!N(E%h$5ToV3lSas!ltSLr;5l)%YhxV3ts?V8bVV-DOz@o{ z#Bzgr!M2Hw@StVT+C}6Z%igkurV1g#alJWedTHO(s#}Y? zbQI0(KRC1Ri~Hwau0*Cul$#bK)71aNn?r>k-fSvSPDrXC2~t8(J$w`1sv6%lHGmUn z0P=DY!@CJ~i?7)r`g7BI6B|Rs0PUR`e{KTIiou~B9c}%4`U`o62I(%icwnFfdET+% zoo4H4{w7He3?xyzrZF=O7wQE!2ha1mrQQ;EUTN?E+ z8`~c0%)mXKWuBv6N$|X8y=5A1LE8rF=6nG-WI>FD&Z%2X84U&9Wg2m58O;m~-Z1J| zLM3yCS-C=L1|d1P|Nh_a@BQTd$pbSVf4LI3wx7q&-QRz6`tW!6_wEafgUAaUq9Q!p zf+%_Q46(WiyCf2zXH*oktyz(k!naQFy!>@E!|K;`S$4DXBju8-5z6!e_9s!4+C)+Q zY41KbRNzpYK6h&B;0O0^epa{8rp$Bm6S2@_AiZRmxYblB&n%0kbb^Mlk;Xc;avZ3g zJNtOqDAtpm3=p9yt4)kX1u-rPaj>S9=$^$d2L^b?KHQV@jPn~*@YM)KMY6MFeb1g= z!aFXY^&AUOhb+n|fc_8MqaTJcWVOs+x^=^zPTk>|plIucfdK(EOR}QMV(e+u3@2+B zRCp|kWDnC>3)-aX@H@k7iVHL05yVP$XQHz7qazCmwcH=#+ok;A%Bv{(D=y;Mqu1bQ z_TYy!$`3t}YNEYHI>e{R|pkG4h?Fb%1NFbH+{0tMKCIm<3Fb1iN# zY=j^y%N#vzJY^%eO6w37C`KHKTDG3p+_VQAS9F^NuYMA#(fB`c4W0EzhrBnr|k zbe+&zST?YW;0({guVFjtaBC+aD3Q>{XuI25)d$K0)XsI_7QbKw{w$BjidmdaE>HDC#sw4&Iw{Gt1|3-cxbYs>Z7)Da4nA zgZrt=O<`&WtXSjw3-J`JxrJ($MNyhi#4Eb@RGa~OmT<0AI!HkqLOBiuT~TWa^f8Sx zp2A#=IxoL`O|I_CSPDN?6k_2kj8|Fo$4KA0XlBQ66vl51wV?4=$xvMhvj81gh$PC3 zZ%oH0IuqaB9)-VqCvHq!pSU`4bK=^>)s^L$pZEa2^9}dAk4=2nf+qe7Be^hfW#ZcH zeeTgEG;!?qzS}RCPE33Uf3MwsHMsvLjfV@KR>iHJE?kQoy@@3eQ-yeV&(CUHPc5k5 zv2k5jQCyIu@>S$Jmb$n)f9m?!%olG=y>ejc%Wr?aHumG*V^2Nx)X&$>PaS*X=drQr z3-8`L{mJy<@9tfH^ZwCm@Gty7{lN!t&A)FPn*Q*sojaGfc3@!UqtTr^iys*n_-XIU zKkhvar>5V(aR1cR`^W!YPnT`+q96$lzcwj>pNN@#UrZf%ZTiBMsbeSZ|NYGLM<0|P zj6{h~DB+V&Na0E1qSBz;T0-Hsc!O%a&Q#a{6zqf1a;X}SUn6J=r7WWY=e*_Vld|0E z9yiNR*Bjs3RR~04sGGMOu$Ya}2BxQp0YCCtUN<(#n4QA9j%X%A;7N+ z5j8aYZNEo7L}z)%$I638ys^hzq{zkuyBvrtUvOI;B^T0TFlt2xv*BO&Ozo=V9#2j~ zacsf0pl4~$W1*J}bl+IT1u-7FUVJUcRC|QLJcQ^eUpsj!ih>lWAT0@&qeRl~g0UWz zl_a^vJp!k++2egtV$Uvi-&%AbKQl1!G>?qy*s{HeOJW>$YcN+OIVQxVm9-i-=nSMO zZ69!ZqJbe2d_7Hy5nb3%6f^ z*Vho3b?!|r@EhEI3&DpjxM$$13+^3W;lKRq?RQJx;A{9jb9Y~1ZatXOOYQ_*z5PmY zw!Vgud>i=bnD`PVm``pj;`IrsQgyG8DlPe#sIbsNyII#^H-#|=k&)8Kk}v3KB$4i4V3?%5nOVPj>6BuCV^ry?R+%HN9mOy2@XBq5e)_ywwp z!SPd?m13qfemOWw)fL#_48jV}rvMmpgY( z{B`0)a{Plmpbot}=l8z=_QJ$JUB&Ox#26@tFVoU{X^Y*KtA{MpHYw+HMN;|~i z4Qi~m&Mb&(ZD`Nj&p8vMKRJKCs6{@$<~%i19Y2DBe=~9Y_S?7jqUMQHFk_cMjoi!U z4HUtNzun#kFL`w2?cfn&xZMS~{S`Fv72FXt_}d3dkB1uZtlQ!Xe~*Y95!8wngpb_r znzpiHU+qQioc0lHrXRtW*|ceEQyE;aNfz&0|Ej-s;{~i!)QmOTBo^4t%ruagh)1iy z)3m*(X)!!a+ie@0sK>-t2BqV3fH;BLq1H*#)W@H5NQPZf6*< z+gyLB#T--P{BuScW$`dUDMAdCW`(S1S92-yMTIhf-QnH-e0D8 zBg<_%KbMe$;1I4DE|#je4A_`A{}Wc4U|xp}3T8z=%@Fe`3vDGdlZCaOWh{hlo3U&* z|SM z|J+jdz=t@|D@%}zD?B4LBA4&Qje^i+7li)( zz=O9(1A|?=f>4F0u`01;5yh+yW)cH75^bTovxy7XaO%fP0=U3FDK{wOmjZ3(tvBw! z_vO!HUxsjm*M?+4;c2T%)OUCSR9~Zl8x7QpL7v^h46K74#`P9Zh&fHxBMi&!eHDh~ zcHSs4jrw4tcJ;I0-oT8D;*$0y=;5H)_t~K98y2f#tTb(XuH35JN`|p(WY4q9MB;)Z zgGCut1u^0YCWX*IR7HqZc!buroCAX9zb}BW7{B5^bM-O(Ri3ISRkWFD>`ROGr7K5gzPvbn@WS-j6Fcg*H>sS}HnF^t ziq#?V7}Ca1*(eX|=fAmVxb<$RHB>8!rxw!)%asOaD1bG5`1bo?=Us74-Rrjxxp9YY z(Zty8{kLCU2_(@Ie*Bh$nyB*B#8?Gm|Jw9yryMhQ zm%HjkoTZ!x*=Np!niQpK-pBw=)7A^9Q$QWq)Cj0kcqOd8atHA+O~U|Ej$sW2PX%oU zV_EYQZ^riP_u8qzdJ2ds`*uAkYe%kkC zks$l$YkTi~b~-Q=>-xL34cHuZdm=1sPaf{zxKH+Al~ZwwiDgm;aw0HRitI!cMM@|M zIZ`!EcRoD6z+ktC)^Bm1uIHk;EZrpP9iDF~Bs?B=sg;(X+gd&Oa*5PScnT7NT=5h% z472$b=CBE2u%S}~X4-T@6sOe8CX+-#O@z*o4Hr5SBfQkzv56Zu6Cl``*a1}v{G`PVYQOCl4S1999gVdMd2yluCYDJ=YLt4&DOB=pG{VF$Ap6(S_ zLbv!?RW=v4i`_%VUuE^(w`56>;|<1lz-0Gs%-e+86f>PIN0Kd5h{=i|tDy;H!4_M` zV32(0%b3ivKwZc<#h&6w{KE=U#kWw)F1n*LSr}8RJ0kwL_240R3jl-R>AB!YZO>uY zv^uDP`tS=cU=P)Y81ZCEMto*Us+V1gBksqZJ0* zvy7m=Y5~VzGnoi8Gh|ViF<$cS#eF;X-!+Z&4utvu(8>oOl%b{RJ z=NMl`|AClkvy6!c$Iq{@Qt)T7#!{#~pC@_>huC=9b)C1|1#M5qULQx$?nv-|L`XE0 zfjMQM1@$wluDY2b3zE{%1LooAW4XGbp$KAls#XwSeNNu(-T~6;58@NQd+hKGYYvBn zQ^{WPq65OXm>Z7MIciv0ns00GA!gO2mS5XrOiXfwhYM^#!b=qzFGVlhW1mn(1-XL% ze@!l&IR)BsQWzkkHrW-L9CW$MCm~2uLzvAAvuThF{80~Tkqif6a-J5;5Xt`|C^TO3RPtxCWUr|u6h@n$)7tNuC9?A zG~9Yqg;)inPXnXH7=eWyqD5@&Fp3LNrO_PLqc-H>ms-nL7()5n)PxYLAeCE#X?l5+ z<{UlaSP1S@@5ZoEo;@j65a?plCoGd^?i`%_99GDk1Cy60KcD<^@~b-sl3kz(me&#Z?f?KdcH?{=BhD0NV%w&}kI3e0 z+SK1vhF8-j8|(c0*ShW|Fxl}2tn=;!1Xyx%zpgN!^1abU;lLW_=j_2L=Qec`{2~^z|e^zryP$6 zG5Dc)*r+^|@}kc$6Cjrm;uXQcMkY#8lPqSU3O5tO)~*Iyp3i zox#BzX#e%2zm2-7&++;ZDRyJtz|VsHuGqhHediO8=aN*472}yF5AW2SRA88E!KxVL z{H4)ExzJ3L0DLEomHGY(3`!dUu&v*%1SPVA7>>OgU|2`3A;+z4_AJ86H6y+WA|o+vwj?Y;e@MZ zc8a3AV`mN6=63G%o=tAeICrnVXK)1u^q1n5M&|DIul8&#pe^t+KQQp)-s8W~!mM4( zanG1aC_i$fHoZIcPnfLvF`{Oo2d7!=-sSDbTS|S3`AmHR_+>!}K&jZ1r8}wgt9I0`imQ7*`;Jn}MC&;D?W| z_tx2kX#WNHB?MV*#7vhm*)bP3X`TjKztp&BQU?7dya#MG>6?YxIk`$(nCz1nAApbr zHGGoQ+F3DLmKrrIj0?AD)GBNjzC9prXXf@H+rc`uC~TL!YntsC84F(91pTpZ$F|Feqh?V;T@>&b|1?9M_Y_M%%MEAA$k8 zo0$4yb;BEJ^B-!HV}ew{|2!Cgvf<9?{~Vh9YUQ~|ijhzrvPP2^?;M=GeCJ^BkY|pX z?VC#&D6!J2a*zFCphSd7MYz+@{bR5Z>%#;MQUmMc5%$&AEfx0FR>D~pbZn%53+pPd z7Po_Y|I$M^m#4gL4yGcHh8f{*UnC~n22=qfW(h@5R3Qq+mK-aA6gXQk$~lh{O1U+> zMwNwhwpQ9gaF(AauV|2>sX%Zpbga11X5yx z94+5=0@Rx3*f2>07M=6%pbdoOuuU~qJ4ttJ+qR9KvD2|_+v?c1 zZQHhO+Z|hbuD$lfxj(;QjxoP?)caHwrsA)-l~oBQKx9Pr5oR18%)IxWtweb|Aq94n zgQPrTENnY`%kl*(xS}x;QF2YPXiK26t5mC#!&3oFsk;xns9b-k|45s~3LnvUTLoE@ zmacg%5rYBO$MODR-GpLJb~g97x}9d5sL`mr&OLHOAbNa-fP35#^r8$Zexb(c{Mr#s zDw8ZDehLns9045a%q`?xdZEhnGbTf7Eb);b1El`Nq(IFMkE8pM`*-p8Q{O{f4A{)a zLs`nTFAMwUlyVFtjRfOPLp8KKmD8p9QpRwbCWbHPoc|jh4#JY6$55cKNPCHG(C)Ts z0gGYDxA#H8iR9K)g7^U$-ps?AZ27M_T!!A`sLPzQ;xpeJPBo)p85a{OEL@_-slzmK z&*JXakwrh-;%zU_^T!4-uo|$(8L}|hpfFRhVE~Z^YyQH0$&!#&+$b+ z4p5=ep2pEyhT42J$W~BohKvgF@D7;0~>D7pnzq8tl+p}a|8-E-hA@>xYBl1 ziq?&vk|1^E{h*r-r^;L`FZ)s_qkR>xX>YDdFKKT~c=jaakI&5Z;Co+{5|HOEl|(;^ zfXa2vP@S;s5#jV%br+J-M!) z?)+Nnaus-aJ+3=Dcvnp|V7ia#k@-AWDy!-4c0J=Rt4+*QCQUjXYomzvQj6*+g;GcBH>z#PGGswu)3dBiEHumS!OY?jcqurd?_4% zr^h~|{&6VyZ)fMsmHe^$mcq5OMnzOjMz7#flYemo?}ICM=9X_MeaUyMWaKC=#Lhm zV=X<c&r;5IPbBAxYR>if7zoR|t+_vbiSnquDbzh3=BLqsVQKJ7B1d-1C$7dObbEXO($M|5W{OX9qO*G&lozcO2W2tEn8c3&>e|Q+ zKXSLw-d@(7MwOBP&Rv6Ej?Qfr4{yT;%67}hLNJ!*Z3$d{s{}ze=j-CDY(=2jcy7v2 zoizhRTxm(`-*!dLPKSs7TLX>{Xo|#xt3Q|dR}K&SO1@F zexv@ycDvjN?jq`Ia!)r2g;2fjQIy9aN>^k*FPc4inz>er3955~ z_tM;OxN$PINLl&_JOfbKbg=gn*^5;>M7uDUIky0g1|yUka>cY+kW%A(q6mMDxI)xI zumzwlWcuy$W6H7{8+YQ(4QZF)ep`pe;!njio*Oc>wC?W!-Y%k1g! zG@vDBT89$_Gjpw{uQBueC?f%%jRd`-++@%|3iPLDsuumQ*F6{5@6Ur1YcbQ9{I#>v z%fKc5{Vk<$xaQ2>NS};gto8lVT(MeMnC_xMWI^1>@z6=lWcTL8d7pjHG~r)l?UjgE z7DyVh$qspJ2K$1a?PNpl4#rA8lM=$3+7ROC-8sf>>cZP7735K+DEty)YScv`XL#BR zmwB9WlaZ8p4WDjkBN6ZoVL_stnv7m!05My@k3><#11vp~MdH%l%#wM^h!Ch}?kK(K z#9;xRcw=tBfX!#uR8D*(n5%v?+_m026-*Iih}MBNxyowyGeu=yH6Y>F45-zLvn)nM z15)Z=JiV+{^=pkMR+Py1OL9m?DM8+C%GNiwZ^i7>_JTR5EWSVC#HACK3SAnpX*;m` zTz}{^N1ptCNNOR{Z2CtDwu%JV5#V?DzBIh+db#;thn0by{PVt7w%b!JaA@V6=|!b^ zYpf(2uzXm(6kC~PNU`ycWAi*CP5(RvJ$dvzIJ`V9$(+7^34w&IfuPW-2)iOHQ?_Q1 z&_*Vah#fegt%Ae2KGwvTBJ#TUi7%inb_PszV4c@Hq_UP67_2wPgK^&HWne*qsamKa z7KfZvh4J`W7UA>0*tbE>?@*HA^|lZMF6=_p3fzo>hi8J9a#MZ@5n~|%h~?_?<-K z9}D+Ao9P=G&#j^i#wkyv)92Snjuyq05_@%+$=iIs0g4X16k*}sw4rX7>&7Sxz1$qU z!T52@cRwR)PflT4|FZLN#g2E;y47u{{GGl0VSFeesDMwKh>Vr#d5HigflS)AX&?TX zj0cLHkOKq?KP>JZ4l!sLz>&mUb$T3crfEKQdO4j2L=!&^Hv|42Qgy|8E$fhK*{qrk z$Z0T#hQ<+jz(E%Xr3vv`a(aO1b!lI4Sg+NOkbVet%3h6K6SiRjJAo)N zX|cgOZM-=wLagic>_yQ`@|_cNWv%Sr8W8Pm(+528GhF&$iVp#bcQ7TmNJ_@)ptebL zOe_=g(i7*Ls--8)3B~}poWLe%j(pZRT*p@Wu-CuftH}l=2djqNdKN1iE^&tb$hNb^ zbQ;90SF?cP5F6nx=^zR6^8I6xn?X*}42Std3 zmOmFxEYtIVlY8@sJmI&KL=mr(iayswqb2(|(1%mWUq0$~ZoH{v9yC@r?NI**Ij zjd<>;P!&JAbBKb8S^G#D1x&f1mlveo)72XeQhN!Wub7~KsLxWPoVXJrbwkV-{m04H zj}l~9=OFIo*(DQiIPg;M5TQ6!$9HI0LiryZ1*vdOsvR z#2)03dZhO&?}keA2^Qgkp^N6@9*8w`Zy;dgFXss7IZ8J2=*iP*mhqxX{IhRrdJ@fqA81gl}Q+_68&moAH$MSy+ZD+0Q{zQ75j z>{*u2f7ImQxR_Zjo&D~qH6yJeXfen9<7;kaE8BzOTD$9S`n9ju=eFs!T-TSyjjW%1 ztqmC<5;9x^XwQ2i9s&N5g)5NmV2i#pid>P)I!^gvd4QOcmjscS2hs_;9VWy42arVt>6h=f{7mz z!yt7WH1NmlR-9aq^`UyE0K2hhl>3Ds1)`uJ9J@5DG(mWa%81s7U1wzgdp>zALEHJr zv6ke!-((|urPUwQBk9tk$84f$eXrevF`Vqb5J+)yxmstgsMmRHbHU^zrAGE)_{&q- zjXD_Oq#AgbCodtcb0JEe`8Hi`o{lmae0dl>y%^FzUp504Lh$k1n0Y*qr?R(s-oH9{ zky$;gp9%)#ctayFR)ADU7GBrOM3E&Mv_aE&Soyx3EG<+bF)4NYy_X`vNYH3PH<%0_ zFOTEPWLFo2BukGS_3J`+jPLsMmK;OR*FrbP#$oSoH-T>2CO|k;{UW8X#{+qyDV7Og zh8nw8e{h+#A&)Sq@l~Qd6_Ie_oij?HkOR!r?(M>XY8YFNx2|i5l^(rB_v1|fk69yyya$e~rHXD^QegN04&m;lscnRx z=?~6N=6~sg>A^AAms@bC&eybGJmRoGngl()gx?WDywHYN&H_+>GFgOQPt=fD%% zUH$Y}zzRqnL50y5#S5aQYs`nir&1@2sWeQR)Z~GZF_1kx3!#lKaGYUE0Uv51 zxJ+RnnxMqNj+jdlY6V%QCVX80jgql9X_uzZo z|4)ZWYlL;}Ye-SE5A^kq0G`1J0-cci*ZzLGV|)UMWT{tbAY4K~q#Vk|UCBV^?(84c zC}Fvb%8_{*1vMAWyZ)Dssr~Kg<7J!&3@nShk2i55M14}u_uTipx~U1R%5Yebef$<4 z)6x85(E;m+Z^-V|_jOTve{;E2CW#`E6;WIx{fd8y=t`b(per|s?Ps?V*B z<|o_D_3f>r@L&;JzMZA*cZ`>S@E+$73F{V_04`slL}G3ns!xxXtJ>Kd4RsoCMO|N2 zhV!lEwl;oO408GAY=^Zc!nt_9vk3@W02z2K zj=51B?h!+_XgYj8Ua*0fYtA@AHkgQP7`h+B zD2$eXBmrE5g#(wQ))uYRXpNN=e26OBv>4>(ABrRXTRAOt7sfU-ae9tDQ<%hWh$L?g zypO~bK~c)$Yr%7BHmoMn7D4)OWl3F-zeRxHis>ZvT{PVb^&@YU)QLU3s8*=zBKzC6 zA-LnyIIoyN&sA45M!KIL($H8Xu%uhKyl^hCURnQ25I+@wFx( zYDPxgej(1fe!^v~R_l2dT2l%KxP2NlApzgl#)euwp2~LIfWyo)@3RUNS7c}DL}og1 z#t6f=jkEn!3x_li`6F_l!8He?no3xH229`8UpT`PNGgVXWYobk^4ACramSvqRq;5U zm_<-l!Z6Qq5qbWbdcpeQ}TAYziK<(&Z zsQ+xiA@AcpXJ+1nA`!@EFL)Qo7>_kSC+{d7K1;dKO=ZeS>~y2W^6@4UmC4!w%EtkV zi(_`nc{bJaf){(Kgc8&#+Bd3qmDF-1%+IS78sWLa(*piv^*tOtS2@mr}`<4oA=Y9{#OG7u}|zphe|!$ zh(yDsKX({ZuPf6#p6?a!6!daA?yo=uDQAQ~tGnIyd}?4}fqr7XdFab#~jpjnLnE%@5~FD2iPaCgr51w2^JqVFQs#vChV*vGqsM-cG4S<|RzOuL=3jsJ;p_x0JrAs$GB_L$DwE=!H~_-zkTf z<{?#S(oCO*^z)d(LLH}XxTv!Qv(AuCPbERG2q+CNtMQx5KeI!@2EMEU@biW`>%3zi zK}CATr+t-$le4oRRGUmkC=r@#QhCyVIh2EsS&FCsW(%ezFn08qkKxSU%IQje2EV*- zXstPt3A+y1ZSGr2ofZrSCIrnDBM19cTnbOGs;mFrT!rq^Bmxcxd5&F{b#T=3fH)zD zfE94@4!wEHrBH5MaVP*9&SQCxO`E*otaYW5m_GqinWR6iCi>7@`8=hbRDK&bIq>rM zsVNKc@Gl2OI-1{GG+79aaVFUc=aqT6{M%&m)s82a%Bvr7QBkEfn3x?aXYfrnXP!>Cs0r$~~0ROog$g1Upm%tl|T&}JhGrgY#sn<&eb zc0f4vC|d@b`LHh6??&Iui9H0aXqdgG%7B04Bg;Gub}xt^ivMGnv&J0_i_e>Oq?IRtSlGT{+qec*Tr%^EWdZ zdo96XB$MJ-W(yIjeqUxVzL`uP3Xxq@79=e4U>r!&j^ z0jWX{C*)uHlh>ssaYg#xMMXPSf>&+GrVgR|v=##e+d2<&3_qg`mjZ_C$mXJ9>TYCW zRN~@{kZa>aYpVkgp#b6XOOo+Y%B=M>#MogcYohs4F4nc$XwsPz2S7^ccM4_E_$WV% z@ejyZGU~sO#E=^xK{mnciYm5S9%)~f{Jj8*(a+*#-Yic7UL($#-?uD(f-wUp-1?hf z{uYvHf%((frO%YgcD=S)K=CdKxXA-5Z`fG-(R~Kc`6lZZ-t@c$(UE7Cln0HkBr8O~ zJ(QkMXZWjW=E4RY9hTJEAUfj^8cG{h>rhy1b24B5RppEf|I_W1Z^6Y3qc|Quhzjg2 z`lxcpq?u+6XsIdJHIi_6HKT!CT*=0kWlcCL3v!zW3*e*TBy9>;5${y>BAqvLYBO@} zKESx{>2W=dC@6J5zk*Q=;ZJRUa=7u2TZiK1A{JLTUHxTt`AHiy2szG}3Js^Enc&H| z?u=*6d>erp6`MHp`GroOTY9SVG+6Oy91Vd9!Mw@K_-6$Is{v3ki*$ZX4>Pll z7W-pf9w3{@AejEt1G$0((*lRI#l*C^i-e}dhm$e>+x&b?A?x^l(e5*?iT(f@S8Y?; zft+23P!#68SF~?zZVp}f> z^6_po6p`Q~oOXH0(Lo@8|FrPcAaJ$wy)c2{Ft(KbPcBjF3u*0HLkcnu z+Ip_V_wfJOhM1+-8iaY(i-$i;=Uf$PAWj{nK`j@FX0P>&et944PG{M~3!5g#_aO;x zNL!&e&Mb0#r*hv+EoRutb@{mH?e1{sc<$F##;b5X9z!gM^<;B>{(HFa1OsdT?-GRJ zKonC*?$=fkF-FPW%^iJbnd{6Zk3(ZLKD{z};b}d0S|z+d-E;$+6U9N{xjG(&y`Uk` z63LV%wJaX9)9P&S0lqq=y^Oe&lGTg^AD@aCmQ-T>LeBPUN{@tnz8x&LlCvu}Wp$HB zrpXEhn}{6S|F+kxjNv?Bq?$qia zwrt!ktP7hOi32k+bl6Q2X9W~8Pey?(JY2dP2AkG%911V80?l{aPkJ!m!paZ4Z6J42B`cjYtJpZ;C=s985d3L0j9WV5O>waX!c%@2~lC?|Sz# zm>AFM2ps89(TFA6pkr6SoH~cuMGc72KkctQ4T$ki82i2yJ5&^6fJmT>^4tL>h+(97Z&K>_NKP zg;S#VY76BP&>tc}X8D@{^wIf4yknKWRO-HOqc&EIns;m>d6FOlkMr#mt?6)`JcCpHJbQ`W$^(f6$f87ZR#>3~nhKHQV?Pls^K6& zcgi7fiQmykP$i^HTGhL(wwJyUX0Fpcj^C$K?;|^6Q_tVGE1nPtTN|FaK8-xk z6oME7k6tD-ktY7Xd*(mG{hVXkiRD8KvN0?t9_ zYyB82`TygkiBiY}mRb4D%o1nV!572dCW=SB@AZX?Es~)HM?l|9I5ERYZCxEq)g0@h z=tm^Lm|{Wxc;V+#I|v{r8sgun)gS13B^laP!GsIA5hrnsmJi7_iLg5Xfi=H)o1V^2 z+I>|2uzTzaoOb4#^-q|#8Z0!#O5=yFkq#XhXQSSkAP9V$JVSY=zs`BxKb_&&$!vQ) zA`usVb7f5enO%%I3zZtu`evVBxSokKm%-!YaNgbTQWfZNJ)iBmsF{hnx_eBk=wtB} zmPaM6rSCmW;8lJi<3}Fj5Ur0brBndM`-wpMI9j~*qq#v}P$er9sUX>VIO#s&8?3Jt z*Do!el!J+T4>DWq=+s4`t4-yECb90sdgCovA>Q)T!|GGWV2xk2>fM%91ep9rVl7Ro z3M@Al6A5_28>|mAC%k4n!ALEH%twZJB^wPl^-HQv_%Osk*;g1~8Ssoa3>|a$#Yhj3 zKT+)QI$#d0k(LNYB3h?R(~K0~9z;D9lF2mQ`yKa7onDtbDr+=irK?3J$Kz=8Aw=xF z6(uYjv1(Kj@1etGJtW=?h!h&l>4!jgG;>;+m9FYeem43+xAT9QXRlWVN0{G&D>oDy z3^81<*#`G>3#Aks1LLcH~^IMds%Al#M%Q-{k}%fw~BNMH_k&$B{@lN`0sxT|@P~Xjt}Z z{dZ~_2xNK-j5p{$9ErHy^Ma#LbdXFjoO|t!QQ(U-Bna&xH7xB!VV;IG-_m|wk;WFQ zQbrc*x*vo-uUUo!OEpI)Lbpx(&Y;GPSGC2UQqd9ltFBEsq4EokS4-Qa9=j@b^Ttg) z=}N}nJZ7dPSeV&8z8j?s<$M`aW6Azx?+@a83Bk>zQII?uNduSk$@*3yRJId6gjqXC z_|#Xdd$%XBgjLW`Cg6{~0)I>z3|CR*E9XlzgLr9Pn|HfOGsDB)xnRXNwL^O5=T}%E z!NoO;M%|GQH2=Lk?)|7Z-S5vhiMs#w<=0w?b0ns4j>*#Bg+$_#gfqM;n}`;3NZGHwtj^rxKAxuFaxv{$x` zpl3ehlBd(S9Hvt`BC1vlV0Xe_&Csd9Bd75Oe5KnmN~p^@8rEI8_lqo?om zJb*fyg|R%vLYCGzj#P|&FcuNDpTyZ(fTVMae2#Mq&>@mS^OCWAsL)>~6KueYj|Yd$SOGoh!ab1Krw#I?8UC)Nw7Dj#30% zKOgZsx!S-C8Lm4(2h%Qrn=X6W@IF$0V~|%FwNc1z7cauIh0zP%hvPei3B-D1vIj-> zq+xPb75j9gE2pZFcZ$v`)y_TQEx^>Azk>+-k(b?J4AmvZo1fF3NFmO1@4atr{4!{M zgLFKDcE-)%O0cg<)A;Jj@osgz%o6?VM4#A*A-tY8Q_`cO_UF>8IsAp!8Y=H}q;MnZ zDMR;7K>oO6i=l$hwe2~kfN{aUj+LFpi{ri@jGUgDr@|i7+{X0)19cmE?)l%BNV0-x z^lro>e?T&ZTElQHonurbMvb~8=Bw+m4PKYnt@7_^wdKv-B$6Kxbzj}idvOF}qMScX zF)Xq@4n`;A7|vC%L1X)-R|wA_(x4WUcjlYj{M4bxm0Z$_inM#giW>h*s&@< z+ttv2gi{XEtPU>|Q^HLA%oxZcA)U}Bs-045lvRt++a-`iwIg!k^?3#8CY7v6GVQ<* zIna()cm7*uEq41wymN9z%OHpA6v=wL^7kA9CEnP41ri+bj$3S+Et(Cs0HD2<5gru2 ziR!NL8{P}5HjT~-eHM~sfhOv$e&P|Y7IzvmbBREHfc#j?bEcQm^~vec9kv3wH{SOz zioaT%z!i>O!jInd%pI^=rW2l|VofbEiS@bTaT)#GWSJ{smZXcLA9M^UEpvp@;1)4e zKr$k>{Fi70^|Kn3;!f;c;tsfSwAjF`eY(9@z4I+e1@E&P^Y znKpU`TRwM+ROMv7oPtRbi+RKxIF6n9-k$}dM=0a+Od_7VWJ|y9EI1xlU|K!@TG*A! zW;u)JOh3{yhI4~{icpX(ACGIlCZk?MY^U$n&7xn1Kunj*_bS`#*Y--WuBV~i-vu7t z^YpZqX<6^rM_ZNA$Bd#nJYUPVQ%}VdhTP+{0o8~A$nJmK8}Ef}0{iuwlJ%6MbE0k3 zqk^-sPO_76{ou^?t|PMMkEkK_wBin2^n=i96M9^$g_DYtv&MEwbJWR^dft>n-OV+F5F)Sf9(}TDbU5`FwrR}&jyR95lgDXe4T3HTI>&ES0Sta ztrT_wj>0++&`1@ZFqGz@xaMA*>z%gWJ-(7EJ`l1uk&p7)o&wMHJ^G5%e>Z!B(L zoSz!2k`gHZT6b~Ea%{gGk8iLzx|{4ZeN8>Tv!f=#gADcb&NHM~N@{il-pW)#USHli z*KY|R=~KUs-xe<@_%NiqSUh}pE*2mdmb7?Yqd;};cRGJwD6FeQJtX!@^#84=9o)}b zG2FPGdq{gOruS*zw=lc@Jjh|G$xeNJSzt8p;R|#vj8zf|7dFT?R?#r-E+J^9E0Ul4 zQ$JnL9ncZ$Rll=xh1tgPlB9Rq@#rHL4-5T z*0JxH;iXGwPl@LjcgXB?2u#ZqO1OO>%r;A#xbGFTJNi8$n+~hVKtB|*@*`J~0-HY3{7WIKWj zF^BHY0uSgKbdfeJs`s)P6oNH9X}p1C5X~ zVYo&n(HL*HgORLrkj=aAyk{>}CEQedBf4Zc1=vB?v+Gcnl5Ru<#HQ_d?5zeVb-yq? zSyi^i@T2I$RHemBCnO&>_idE2ybT=YY?|_)3S|xYb3>z%tm@$5_x@Z@BD*pBwyetF zSjg>{wT#z$jNcLefxEB)sBTX^LTT1Biiha$F(Go=EP?v2X?V3sLSuMHM7I=K>4b!s zW1B=r753>}?E>>t<+$1qeeRBi$ z^_$2A(s0zBux69|dow1&Y*!@cydY!m_J0I$f%=S#*`SJ%HBz$v!*35i4FXBn)Y0lT) z3E#JFE|xEMwv)oj1E6|LRN^p!sQ)@(ee2>R&inZdsU`jH-=pi^iD>ZLamNZn3n3M9 z7=vbgm|e=egT3yH79$iulGVzG!*&$=+ar|2YVG&r)+6gPTG`y7bse%Eu#!sA^M}!jpw5^wX+gMCo3E zCMZ{1jw>0!%ncwv!fudVMx@=+!aDy}GM(z!UibGIs(96I^ zDl~J4PC9#usrLFF8stGyC9js<%xtyu{J!F_1;&35Kd-3j6c%NqVoRV@GzDOGfPpdJ z-d=s8q)>Ecw6Zd>+gtA!i}8mk7bUckQpO`DP4H>E_gsZ-;`%F9kb*&Sg|zRrgu`Y= zZx=I@rgKYoW;pN%YLx7t+9LgkFilI_vw|xgcJtn5&w@v>Fj$o znU9Mje}w9IgQmFLF2?u?_xyb0{($Bh7WSe4k^C}>H0q@_0&S9Skg-kC7q5vE!#~O+ zOS+sfMd`5zecGZup+5U#Iq~Rc0mcNpirNPN&~(==sR4@@KBfn+%Bh1?By zBOgQRF~Cv}ZlDzp`ge5B|>oKXm@*nGPz6_$2ND20e~SLUTsqBEHhXbCe2 zy*vOF=O0eY`*5!boyWi61XX*0{5W2w`6UV*O;5kd2s$IXd`dR-Ue6>list7kq42YK zv3Qw#u;-=ivPfAtcvtbt-|CgHE~?@CH{m9bVW3rNz06||$(lBrnRk9LaG;%3`v<=L zAM=R>z1p44ga1>P_5Q!qWp4M^>5CiJ?=00n_lF#e@iXKM;0RzJ=4ME3@e1o9s5=|6 zs7Qz5cV&cCcpOKZew5AD2Dz+tvyPEPSVB0P(vDa1U9HQ`%B@@Mh%6ynlF4r=U1Mi` z+J1Yx+idcoWB!G1p1p3HjP<$#(vcLp<9ln=D&DeD(Nm~jdp~E&uzZt(#S^NOX}O0N zn3lqsvxs^?8X;75L^9p}^{o^jeBJ8b?Zp}PXz|7150$lvBufFpWpGCAxE~$_1#`dk5`AQ2!Pkq z{vpK@!$e>c19N4Kq@W=OFVp)KSGtnp<9cMLRgkU!?$O zSAecGmr1n^R9;;nE>b>PAw(3Rpu6o;9m4rq8t1SgAY~8#%EisrVM|b*7nTum_ItOl zAkAP0fI)0ot+)(*k7(`0to}f|9PGd;+y&B8btwM>1!k1g3#`B&drh1xGv-1sp6bv# zyvak<`7kwCCDDi7Xr1_Yhy|l$yMy)qDaAR_TvGTCF)X>!b-lCf#C5<9&hvxRR_(om zq6f=bU~CWB>_fRertT%e{XGcdK$M@S-@n9oYARLiXELXtQ4XsiQ}%I%%FQqv?beY- zdxNjQ&91@?ccO+yyQEU6O6=0x^3Rr&h1Fx^`Yc7{wg*-8Yf9t^-vk-K@*~hl%2r?~E$V3oPsBLLU#OvEjF4-!Zh^GwW!0SHFLuxX; z?9NGZxjjFGNU3i_iOcQ=gsyW#Y_&7^KItn>A9k}42s*ui2ZJNqjjPiwi@(**=;;M+ zk(bBA9%wb`1Crv`J@31{JO~JQ9yi18GZsk|C6^3gE`p(K(0W@B`8BMI&pRcN=#6Q| ze@0%Cj>JX;%~{Hog8ZxV9MPSpBk|FbjyQtC?8`N|ROqq2Ls~YA8Lxzz1Dq}7tL6Un z7XPmeDxmsxq#JL;gTr=haU~|*_?NSM>i4Z2%b`70S{`7159(xZTQ4A&)&|!u{yb4=CJ=A8kU>)pfCcbKUn>J9qlbB zZ}Gv*tKG7VClKV;K+7qs|B&za{MLaz{%4X-m5EUSC#~4eAj=s4fW8(QOHNdnNF#2y z4m~OS^+&iwaM`qiJ|>|hq9Dj`R6sf5t9r!Wn{vGCC|JRT!OEm!4CwSgLe&Q(D785BRJ0^6@^SerR0;V#<%)NUB)fcI`4?en0FRA7 zn5N>>0%vJ}zgAxga$&~OB2R%L!A|De-9=dsemb{{{`u_c!6m6%Idr|&q&C7DO>5G9`tsnd$WyOxxA1qSh=zdT8dcY&Jw#7KzLl`f!50(yEH{Z zeVbmN1>h}bJQgyIpUkKaDg;FJY4o(#0tLDBtM za@6xWqjmd^UAG1xn?diP*s2t)4-?Rj&(m@aDJg%-(B~1qi;A@?c4V&dO(ZvrjErZwUcD86_ zBw@!}1M0mE0#8p>Q1sC;4#5SgNBs3OCvu}nU+uF~fxfWHKiQu{`)5#gY6bm$cBSHi zjs4I2JOYq^UcU-)2-oY3M|yjoDZUBM1HNxRC2rpnc*G@3hG(BCSd2atNo1rwgG9jE zskG7&s)=1A3p2NLQdUXkJ(EZOh22Zvn^kK0yiz0%yIUA${UbLIoXg3`{&rld=8(y{ zLeuo5V7Gf4Gwb5nk@#{iaR#&AO8&&N;AmPtK^#+ru|d>|EJGq1cXSp3NYj>_*kqPO zRNM|fF}KW^X7z7fyu%treeY+-D?X2h%n+E!4>%URPt>n>&ROI!R{}w||RiPZ{W(=AEv|D8Vu*d{-Bw0c# zDT%XEvoo3u|q$xQ(Mw|wIhp<3(cixU~iDDO>>2qGoQb+@dbU+Y6J&{O>#El>a zG>7tTS=57TD1VaBr}}(dneRk#RmQJC_B>XW?E1LjAK2**RxPf-wc$_5`-sDp>c!S~Bafn3B%@-^+n0lTErx-gS$Z;d%MgA5e4zB!)b*dJ zzIaKhc)l$Ufx-)gDRn5*hYON!GZsq?J2m&F@zpn~z>Nu=G3V2YTaf5Z2SAfJj|ig; zx{=(DL`YzxO9E;bsF{f3DWCaXkNF%-!R5ialcMIN7@YfgCTm&$FO8GXp!a52&M?rC z>bERScTOOT9{QvpkVw5}_bheKC^PPRn1dvj<9Bv3y}j^^Si+E(fTt$`S3VgM#v|Q6sd6mMXwh;#F~F)sWixC~ zSY0}FY-0XWMIygz8u+!Z(RDZSFL8(n3(t3?f$Ty&QKa#3UH)J3-|B+d3A3RI0zEp) z)vrI9T-+3W<@)8-4@Bw5ZI+ZgC&aS>xx4gdi{;O4Zm|LLmZB?gtHO~P2whniM`h63 zCXcHvOqtv6-C1}x4d@Whr7kE=MvG26>w}lNys?yQ^X_e zTfto-X7RA>YQul({um&kb9cYcm?hU>Rx*A}dgacWl#WhiXkD8LH5+2H>I&8RR1}{4 zV&5C8>m%;Fd3*Kysfx&#lZoL+26c+VATKT+HHbC+tQ_%gu8HrDsgvx6Ol!W}#wkvQk)H}^=JE>GYKJe*WMX{i)h1o$K`xg}Y8o`SE_{oBQ< zysF5380X4>(|nTub-7h1hdH0sVR#aCPJ1FVV_kJ4$|n|Py}RQNAgdLCX0?rUK};{- z$FmBjKKfVc(@P%VQ4ec9XjT!?yD3B{Sx+d^mIj+Yqv$eAJzS0vJecf-4eVmQL4QNf z#Wn}HkIY`8uO9YR27mW)=V4zWb{<%RIHxWiyW>y|2WPvDsjJi&YLo^bW?;$nfq3~= zYg-302-i!?71&tyI{)MS>6-G#GSEV(lqG_1Ux}{DWa--?8cI^+$7IVuq3MTGus}+> zgk8F)7u>@q21D1+;bQm*#;LGpEmdgB^zV1qs;bfvqE0fZv7<2?E8@tT7ZKPF^7=9Q zRr)QS)kVAiM4j9}6;+$&bSTk&fW9Ia$v@-|xaZIrWvtHcrKE?RdbQo_hQ>uBF|n4P zSn(Nq`rGKbmvwwr$yQ2Nvd%x(J3sCOh#p4fI?ZE4xh~K(lVEWS`tJ5pG zK4?AUj_|>ov{UjRKD1AU(Bye}pSK_R5>*)+AAw!mG=0mCR_GlhvG1AH@`?W+gNJU8 zNz5tX2WiBWjxR4o20S!X+J#N^ZPW|)KL`Ps*6?_Lv+49^my0}pXy_&FoT%`kEVzysY1pDC8Y!cSl`$@{Go#Pl(# zVsFlHk%2H31`xtpYF?+{!#NYaDxi0PH3t{jK;NKKS%Sy8fe(;^f+DM>0skGqGtT$X z90AfWyi2ue9{7+t>HBX4y%;@oFLGk;^O(H8osH154jj3w4zoroKE0oLf+yzm{aLC zyC6jv5N!+vfA?Hw@y3OJ6|L4zCxBpOA|K3{FG((7+{=lvY<2sboEhnu1J5i9D#5?5 z*&~8-xWT7ZH_%T-Mf?!8GUB4=owZ#z3EMD{7Sq!q`RBsmR}4z-7tG4dt*JMLaBo$T z^#R#a{ky@0X1he63;rCiHzeLXW0_Fa%a6FoLv$!6GWVrg*{;EoP&$IPw^&X4PB zPoX87#20+yJEm8zKY9HW&S*tz#<*h08`Z4YHsgIuNWclQmkr2^}Xli9vEh zZDM`WiOqF>qx>HL$v`&0Jg=A0MyA_NWgVJ8IxC_U|NTcT(~k$Bv)ID$agAo+Zj~>c zqRHnfU&5wykDi|YXYM{4Q?b@~AAXV+`A_ihB{1ujh+&m`H;XB_u^@ELR}~L|g1!rtE*}#u|_5z`NS+%niC}WK_G&g<@#D zT{S?`px{I;lp5m8SkKv}MR|Cn1SWQv0Lh{tCL@mW)?rN0OKF*OqkeY%I3CwHVPr^j zku|L1fK^n!*C0&rDVY5Hn47%m|YeF`qxH~r;=`-+>dx3~hWkP^>6fA4(k z;rD+F00Yjw_)QxYJCfqjRYxm#bIZR;`047BMFLmO9z>9Y4t$N{%TWo8tRX~U=t zRD?H7MWe*D=ot3TtkBehg` zXEl^gxzw0?NrvdLA_EXp1SJ_Qlnrxo!PZSnBSt75Y0>E&CkIYn0h50XT7yC99|IV)cag~5K$?EVr7 zg{>MRU}q1~SqmGjGtnkySox6SULr4nO*BWI+i|Tem#)I?JR*nkSb3-1ZLQ1Ug+jVzR4St$cdR zQ7s&*)EXoSA*D2haifNL|8il<@^BgcH1HqNHLevLaM3azvQR2u?@mx(k$k^5?Xi8?cdVX076&4fyXI# z6^GucWIzK82W1)BLW=%1HZhq5!$y!esdmiW%s5@2-<)l4Xv9S!(H<$r5|NHaYYxkX zmt-sQg7I_{S9(;`Kd-3&^xHFUesKTDuDf4seO&#A<9>{k8>~lR4qqUTl7bKHl1Z>1efqb(Mk|fL0SjcN!jmx%+a4mB#~w{R&if zRiH|WYsyr-izuk+P2pG11wnm=nx|uw7@R$S3`QuSzKp5nb!u5eN4ye<1No>_=p`64 zubwU1W|3%F_|(eUEmk@WEMn=EqV&|-<^`7E1v&Vehv9w8M2cYpm#00<_Bwe^?l`v zWpV!`@oz=Fz;tYG#ARZn)2?N%p1>u0&fkisgrwN$;ER+KH(Ga&oWA$g>yBmuo8fCF za!gQCffKlviSN&#uo1eORu+Y1EO55%@T_z>67|X}+qVysVoAdm>D|Qio>7AkWRR?_ zlI%p|fsqlA#1)SA*#F}21pER0vS+S4?{J>5=S;K#^b?IB?!C&wdXUk`N`shmoEQ_c z`%jODl%5z70`wDpFdJXd$l3F$V${>V58?i}ppk}tE zAIdT){!L8+gQvqQF-3|c7fOd06QaOd6Uu@ZmnY`!H73H(zjSeODsLmEh1S7yA_3Up zSO(@Gm&Ez!Q5HEnfYemAwHz&gxd~v7on*KiVKWrc*08JN%WK8LuHw&2E%~hhXUZV7Jq( zLc=?KL%IQ8iFxX1k0d8rN*Ekp5YeQNj;w6A$JybE(r7UglFwOA3sRK7?w<8ZL}w(- zX-|m>aj8*AB1_wi*8LC9&0PC&=DUj_TGxswq=H-5%ODFkgMU7L{Y*F7WSh90q2ZCM ziJTZiZ0fomiIf|&dtaQn|M8Lg$HQaHc3?_-_kwh-#?-_b&Nv-*=cfr^FeX~32W(#7 zlpe4d7pQ4D?0=n35b%$~5XM=GHh|WKhB!A*GaOx;`@81D(Y4vuci>D{Jy9S9)?g|j zNRlXsF3)dI>&|xzG`kJ(Rw807p`Y1t+mvLU7boXsWSXh|kQH4p)KTyZg51tjwM-&{j1QNY;e`VwPQ4Lt zN;8VWfloLbI0V?}xMvN#0OLJd$QUOp>lxtwQ=t?jFX{0S!Fau$d)#9?cZs)i7n|4} zm*c9&6>+BHLO0lXdN_JGpurud5j&1c$H24*);JAO#}H#eG8MQdcrN=JG1a&) z46b@6CQ4Bfq^68)&1)YfP&GF(3!{FbFT#w$bvkvbvwef`0@LB?x*T}^lLRHsA`)v# zT3MDY?~2SU>s*<9nUw-WlRDM!PU2)Cv3OJfys*koa|R8}SJerHUrY#zCQ28>_C93| z!s@FLWARN`8;Jo6C^36rzV%<$QT=FYZ}nXDTJ>!8QgyO=)uq(=trnFycT2iG!owvt zd+nrJo@l%R$-lRvgELrKNHq1H@SF)bn?IN>NVT-nAPxldze%^UcYbgCE?x`gv&KREcx4 zi0yGoERp%lq+rPfFiKm_VS&5_k}QZ#sf>{j%`5G__Rh?q?|+^AcIMm3yI+48rran( zT#CX+UH#ZB5MzvLTCFS@)PV4OI}+zTh@!0XKq+E1Z#a$sh&oR7sOb}!3H2vzwY9VL9W6gHbT6`+ceGRja z#ZW&%1H4OmfSkfXV!|p>hd`>kxg^hSuDB?~mEMA1=OF zy#mHBeB^5N0-D+l|4hOWxY5+^skc$}EdSsn>>Yxm7f|&m+-^_x!qhG}4gbRTU9O&< z+STGE=-RTitKlW++G3lgGq&)veeCP_<}EqBqYQrv%aBy!YDRXio;Ic*(tos|bKaeJLWm_I;?Q||9kvas+FE#G!kdgV8bDDW$-vl7pS$@w z{Mmc+AgXOne+5H2>l@sL4w7Qb)z&BGv9FKBnLOm+Zti6g5hA)xRcvb^68!^GX(646 z=qu~Rc~oO&8%+v3fr%_d+1_qUcETO>a(MNU5@*UtrK19s21T2Mr#{GvkV-`U(*{7Y zATA~irEQ$VsT7h@WBeg1R;?-CyjV4hi@a8KEh*4q9*0+}3Q)CbUJ`|h8p^-#962!a z)kh%)iIkWiOHr~&J^(9WYv)0bX-yG@WYl%Um4vNjSjNDrMM7KrmAH@;qs6~grGUqH zmBJ7nvqRxf1vF$KB{d3&pS;5A$JNtQyPZ(M1;^2TffwObuT{^$9#6ZgA9o@cS9!o< zYJWGH+B>zU7Qm=oF!grF;Co6ThU4(_cTeqs(erU$aJDYNHjMPAsa>-tckwYh(Uh~* zYrZjFjSZ8I;|y~}q{Nuy^6E&Qkxqcx(5T^gZVWso3h_k7DOuRuJ1{gc@TiYAGRsb% zl0Vvn)6NZGKD|`FFo@H*MCFuqmGc_uT;amg5YE-MQNTJ{&x}G5`_DS%!wW`+^O&~ni4_W1(Kg=5#{d|o$Ezv8NN(L<3&${f)hh^&!da(q*pFbeqNNlhR zCbkOLB_VQ436U3sLK0GKrKM~kDk;r1rdL;zMW!xngS7;$SeMIzNj;8XaX=>1bN=bS z2N=_e*ihZ`Rco$AaRI8uI4|{=Q;}&@j~(Q3v5ltm?K`Rgl0*v%{pAb^v9L+4kb9p4 zE-4`uoRxH$1Vk}BineOIJAsrK6Vg&6FNWt{OMM&GY~3`3R&Ckx;t(2Kgq{3It8=|yXZ zPR20GV3TCgFR#8gd!WZT(2YjnOA!CP&N;r<vW5E|KpY+*8IZ{ zi^zZ4^krsI6>Vk~wp{fWzWad0ooZOAF(qr%E4Rqaga0@*eP;R;{CytXJc#)I|9tt6 zL%-}gG<|OR)b!bZ96~qu-#iG{eE*L_i_8LsAN>Q2;r#Ten+JU3v=*$hY#y5+=v1sD zl&I??36JF&abEWtPechpP6+DynM~zMxpIYvp{O<%IJ2S6Wlo9<>EO&vEB_s>Qkl+* zW$qg?ifCt^TC;;*R1?}Zq7@tpM}H8k*URHNc8Nt07q+Aba_}^{Wsm|i+xC)->i($~ zI+gwt=XI{l##E#bHY$WUV!%ofy6Z>+13D47MG?$vLw&=@RGHzlvk#^t)^wwuonRt6 zYG7^O*>2Fqy{jT)?o1~;?{ z8EzeF3K?$IDgV$A$d}XI_Q~^SB*f%=IOe#lU1BrGZVg|o6O&UO5AUS9i411e-mgXq zjcx0?=3~{hje_l>n-0>mkR=j=n2rf@N)V-lJF@7Pc`(ik`{br*@3IV54OA|n;;l{Ne8TT{Ayjxc|2=!sL65^C~hW5&3aoM!SBT=o&G)(VEH+MwYZsC>Th7l(--% zQEL^~3X`>U)z-r7NfqaPKod9xWU^vNLc;T@79# z+YxC>3(4SZIhN8K$Jd-JEp(&J#2CXy`;|li)*KliX{Sxf62b$URD&x`q~?vX$%4?P zcLoqn)#%-Avz}>%nncfYrfnqHv}?iDR=%2bo0@zz>-6k1!g0C*>sgc~IRj@Mzicay zOj9TDZN#*A>#CfSD>s+#&HciH7?%Z6k?Xif_2ycV$L3lxFJh6KnCyL3l>c>V@6;}* zBzzZVsO40yHQLsxcb(euv(-u7%*CN|y;MCt^=e0p$DfM3vwEs}p?anI6O7;*jPx|8 z>pfgO%d5t#KTW-X;LJ7T5d7|ffuE^fKw>(U_8&OVGDtTIuBFI-yEK{im24v?d>JGL zX|x-yGR7(B4B-(2$-MYQj0a`|TrM~POyQ+($-MVqn~u08f(X~^v``%hGA~R@DS-e3?TL6l`z*FI;*ds7VP;U*%O%V4`pZ?@@%VD3VG(JJLVM06)Tx@c z(0_N05yP~IS|Iv=cXh+eL-GWuUk^XTZPSHql?tZUcL*v4pwT%m#1eMuJL zk^T8Ep`u2p)~A#Z(os{qSqCYxRog_S3i^wtVIwRbsgxFC&3cd5&TA|G9c`ed#Ummm z-h^2PWemomF#}b;)6p2k|KCT~_aa!i=qOcThjRCvUCP$d$QDzIXlK?FfDEo27b~#lOb@D7{+Z&W|QAjo!Kg&_P=3?G$_nT}#N(ZRKg+#20 zZ$|UVH~0 z`uDfJAPkms*mSFTkJ0iLhDQ-%$+i+imx$2dpO5F6!N$8$Kf8V$kAqTI4yV*x=FOwh zP3?tLuM7#WD6OVik>w0L3e!Lb8&p1}m4;asK^ur(_ODWJ#$1^oZn5xPRVDNGPRXYH zTwejYW7P|@-6$s=|Hm84IV>l}+7HXsNh~~aSjfC;A>Bmh?5B)m6E9-m495iP$U%b` zJt{7iFwN;a9*%Zu-G~RG;o-&(ob8Z>N9Z zvanu@q@2ZiHyXCbHCqkS)gUW^)LzB*Qc08Fl!JkEtd1mBSFW9Hk7y)?cp}P^UkIvu zpZ#?I!wbJ&eB;i+)4s9lR9>mfo882kii3U8_hdovXUUyg0Pt)lVr^nH>wne~V{w6( zLWHGN-C%?wss<-h$41lAT9 z)?5|6QhmvgJTDoFETo(IJ^Fd|AJtiIAu29$#Vh;F+ceu;gPW`1eH5yi8<2@cnSrv@ zRC!-`rmx(!$6*= z!#@6Xps{V={%BSl%3}kLt=o)9)}ENRH!=b$oi%%*fL1spSi{>UhPRE3Jf(g^!Kcuz z@Z^z&wL%@pM~8YPt{Wr-p0^_V26GM7G>VHZ*LGc{i+fb*6yC2xPDfq97cLdDsG4W+ ze2CSdoYDYM2*oLdh_~a>OQ7m$r)L>YCwhDFIaRBlR%{xqzG~o<=C=xE>w1_NUNOakaHo>GL;mRre!K} zEx3XD_3Gi7qp#h+^Z{CN=i5VfUVFnoYf`5i41iawe+Fgkk_HDWU!pA5 zO=Rl`mr6wAZCHyeTlr$KHDTRA&=y7!(a{cU2PoQlY!}v4%b?|B7U;)@0fRNUYoJs+%MTCav6(O8xzmX9 zd?i)0xgXrx$!g8JIMD}1J8aWJ1}U;J;(XL3DuWYDCjuIx#gbM=CN6W~gtG=gJF$tf z%+6^15PYli8Ti2{Q~NIAS$B~)=jiicWab_ue=2!`vpbnlaI(BM;{wm)_pj>@Ie{+J zr&&cn!=))ok71=(7-*eYfDPyi5kPhy~Tf&%x)dx@E1x0eQoophDmFiV^2v$RDv9)U$!j^65lrP5` z&SffJ`bMIk<_4K7{;kyW>lKd)_6n~ZN+?k$WUiR2dHdGo4=q&u`AYwf}jS z+PX2Pt7RwLJGC?+oHJo^M4*gPRU=O-S?DP#cJLUA*gvUGAEQ~9d6TMFBBwkSA}hVJ z*rLI46ghXpbN9yzxn5nll4nSMpisLcIA2JE=Rxn z=fz!%O8$QP@Xd#>U;O8f$9#+6!^8U?etz)bo^Rm6x{wuH7oUp@3t>V~ns}j>rK(a( zmSh>uaTOk?Iv|n~mVnRxcM8qW`7Ekd| zv8{zjYohRavSj3}bK_z%jhHIYLq$b~=__!2eF>q0e?DG0g)Lm*2@hF({df^)i5^O0 z2!&UXNliUfmoF+tmcz0YOb1d936PW^HTCn1(okv`$98myErYg%~8C<$IH zmKGz~aji3)crN5-k-YtoASr>99hXVzSCLwIiuF*%M;21)#`v9QZ)+_zHjC`iJfh)^ zShump-Ec!zStBbqIGv^^Y7M7HMr!^bo?wECnD?1BJC{cCbeyNdd7Fo|dF@`ozdAR_ z)NR-@nJ&9Y8(hrM9wXujUV_vHG0>8n{OBL;Fy}onCdk2qOG{mI*ugqf{~y3rkl>{( za_|KFrH+k(sp>y0+d*_V4{9SiJA80D-uwKk+q)0lIeX;pS^&G1Hb4K~JzJMqfJqvd^CQ-pYM7Sv^;&cRoX zQDVWBv{MXp!PcOpAUA=fdu0^1sby*~G96irmT1>*UIZriWq8IfX z3&1t+Lb|E;*t43wm1kmcQR&MTsqO?rO`d%~^xkco zdpED@8(8~?iOmG7`7)A(7+Qhhhhvm_{;6@81Qw;a)C-qxJ(rB(4A#j$TRF%d|9{#J zk*3c;LpehX_=#v;9mUp>xCQZ(L_rnG}hV5A@J`R3tYzI^z}J9W4p9qRq*worh_ z7q-l%1vwRkbRVGt>OR(3F@udP4I7)dC`e*NF5tq(qO1f4K$wPDJT6G-NMPKcbgXhc z7$Bf^$c>4|oz*%{?RxOlepqZ_<^5M$hjS^KyLK^WOaAlp;Rom5efY-59n&OQD-sK0 zn-s(*qN2#VJL}!GXc-&b==$lR;q+q+G2|Fe3sSN%xx!yPt!f(v2L^6nV3kJ@otUCo zAeNeU%4`9%Eil(u%;x3}S}{ts;`58G%jtSjM*oI36L_q?@p^&j*n;OfxH;vfxvmSA z=&_6~!TZE0QO@gZ&g&%$hH&fh85L>RKS^?@s@M`vPndKx-7PfTX?(_8Q@LW~c;CrV zhv;y<0NUb}GT zo7a7Vv@tEml$cxxNQx!v*{yur8$@DEh$kCY6wl7h?c4b_)6L$sL@V{K!Py+Kde$;) z09zBae;j`%s%13&cKH0meaGs4%?GDux!=HOAZLQE^lgZ=ybw@iZP@nHWuza2XC;Eiux4jM^m>7kL71+kR`Sq`jVgZK^I zx_kx|=&j2KcwgqF?d`05O3|=!eJ_k~ zh>G-)u35A-i|WJ}>lAlEFjRQ4PGli-rxF~tgSz%Q;Z)+SW||7zBc@O{O7{=EdGDw9 z@4o*1-K$^Sef!$XiSxek>J)2ek{948(aQ6kj*^NdIe^8) zI@bA&P$Y00+l8NTeb?#1lfIiHJ|8V&(z-PW%&Oz z4CQP$s(xC%3`fpY&%pjAR6SZfGqnp&o`X?dsa~s|txirI@D1E*Y?yQ$XPDbr?YF%> zpe6~4h>qIvw4r{S?U+L#h^tKo6`_f-b<`{W+Rbazr>4(MfA`GG&CP@VIP^bX{^QUu zyAHu`{{A0_rcc#zXmPUlf4-kS3mf0xd==f?e{;|Dshj(59^e-om_GH7LumTDn+K=A zSb6in%~z+-z!lRM;Xz!0O@4Oz3|#)IZ%~J@Fop|Mqu3>0@w5xow0Bq&f*hP5d8r#u z8YyE`!nR5c5WjE}^W>C}Ot(`P`w}+bjjfhDCKW*q`=|!YbQfW7Om) zsjF+FtLcvD+1;RkD$zs}8pRykc+FgU=l;>~x|1cH%Aw#?G)sO~)3BW-pq9g|#@ii+ z8D~)9LOSZtGtaNB+9iXkEHE`qyND!&SW~@f)~lkofo3hloYue&1p`GEDVP?~yUogH zypO0!bYc)x`OGkRM1^vKcf@l1tVtGIBjL?dV>aqDGnLQ%H`wkg?!Y*sx-2BAUS8~M zdju>cB$AE&{g$H$ks_zGK;&gkUl^XMmf}rEM9Fh=%oeKJ8pg7q4PXrv1I}Ec+ZIR) z{A`|Bhjwb+Fn7aqMJicj0yCUvMQoyt_UI@H(iV=IJE>Yg19@C>%Z*EHJYysYiqxpa zf!gIgbJaHa=Xje>%o_psY@@^|F9zj$!7TC(l^;wsUx2rc@=07@DT)FbA|;CyGcZ&0 z=ja=^^#;vmHJ45&*ArHqZI6)2V5>G-K%<67$o>yPzXP>y$Xn{%kZi5dV%kgMci#GzS0F{3%|blx|`#@iX0R)^M4V zH0{hYRN~uNGzzmDJgPiWofg51K&s{=Q#oQ(ekd4N&*5yvzbUlEw(>(38LPDeoCPNl ze*_NS^axH$kmSaB&JQ?Xc2mO5P)2nE+8P2IoU5r05jWurl2IJXB7>9&=DnEtC;3VR zcI?JJv=ek%u3T6QuE@b7LcFifs6j}U716&T2R(iPd;%~hU;v^L+sWSs-b&rF_(!OP zh@}c3za?x~{LUt)?9o*MSLLyVFldxa*I@P?1?vXcrGE!ibBsPZU$QNDyWr+|i+|B< zhxV20bR;IzapxzQaGaW)gAir0UCPtNMrYrMlkaz~U$Y5X2lIZ8)!(@uNshO!3z6ds zy>=q^l0^JdkFmf_6U=WF7fTo}4%4j&K#4-SDUAtxT~w#PxOs5;D}dzY{^@hmUrm2E z{r%1Ty{oiRekJ#W!;uTz#r=;%H+Lh)HC}Uoe-Ed=oN!-p^Ys>2pcPW$sj>2P2;J#( zH(%#}oFN@V)0b`@y!q<%S!ZakPoL!>fK$_FroXCPF@4rI^#4`ppI5}H)R$~$+av6< z5KnmaCU^D2MOnMl)vz|eY+;+|B@7SiF`94Jx=mdTKy~pshTp%_`8NSX(O~*GOq#hC zvt`%DwO#XJfDdy*Ni%ObNg%U$h=L>uqLLQGSfUP#?Mf+5XpulMfxzs{V#vTEkdlQVCBdjIgTnZKQ%Irzc7lVA8x zDcYi(e*}1!Ky%AcV6p%^?@rays zK{L%MIvv$mh>5=`rwdL~XM$y^GU!$w7bz%qykCTog{7!KnU)QlUD_@%X+chScCU54 zXpmx~5j9lSSqTrj&swN z@9us3&TB{S@A(|9xO@2Yuaie+K0EO1rAz)<4DNsX{oS`+{4fFRoVX+?anA;A*AUh@ z1umy$uO!wi7n(GlYr7MK+0NLpxtQMvg&z(3#80Wr|mC@lk6 zqobVO)a8FjiXbtPAi_)(Pq~Al?x?8euxMOvhtOwPC1pOgwUG0-N&K@4zC<;xj8;>NO^QY5w4N@a$z2Xb zF#I}HI20&65>E+lum8wnw1Sg3{Dqz>k~Z*U4{Z-a9n&=B~#?&xf}) z;WCWrPN(oM{P6jmBd2E$eK0e*XJ-F5GyBinJMh)LlZWnIe*NyL# ze*NM7%bx@cNr)mCw~8phbCz6T2{B!_ZJljzXu*w?WLtEKONOz>to+qR10-YnhNLx< zy5h<~@eBud^EEC#o-?pF5B2~ePv^;ZN?u;T1(rcWmVxccaoxu1@`5@E6EuC=Bvf)3tgBG!l)-$`n{L4T;%?&bF9cdi| z?-(QmCEZAE;2DDfjbPm_3D5fL-(vahX9hLV5FTYZ^WX2UFN=#=tdX+Y(W$r`o!)1k zv>+;hsdSdOy|^PZk4w5YCW0Z~m>p#oyw6V6PiL7+I8)0-RQ;)XVQLrr**|rpdLFIV zNL7QGY}7(SMXYJ4ADdKt413tKHHB5*gArVUKQN-|`KcpQZ`UzvKG$14Ikm?jeAzQ~ zq^EkRdKwwHd21axc+cu3L$J-uItK2UD4DmWZ3@#nrD%^4hQ~%}Z30O;Mpp zY~}?dKK%afwYTnm{UKU0^Y(0P!ari26Cg6VCLP-3*7tsQ9P zkrY9cqE@<(R5Yz}u|vId1L9`z z7qH6nb~Xb@1vVBYOuX|+Q&o_kmKp;F$r4@xPf(UwmO*ug&;{mbunu~WMi4WyRLA@h zc{q{R%=s0ueN zVV(RAtMHOP-zwwbS?<@dkr_p%Q?tl?gANN%GA4-W7%wxf!`|acp>t~$@x+3-#|1GJ zsTHnylLKRRtP^=KVN?XUDIq_aGhK6L0cWqD=|*cRKNKpbNX7}`hK5#Bh&SZ|dFGOX zsdEECzhC$JS6(4tN13$pg<9jn`W-UJ=pwUsH`-?9OPJs>=k3B8XK97Go%2ZraCZ;j z66fq%3y3ha;8GA1NmsJ4XQMZs7L+E=KhH*QjrwHO(8MpVzB4j%|G=9w-~DCgop)yr zp5DIw-nW1G_451oKRP?}<_CAroV)wuU!GX`N~g;I8%oHX0sK1o+RUFns|~<6Ser=+ zCK&42FYPGuGVjA0n45kzk91{uMyHH2oXXAMe43ozjdRsFPKxc@dx_o)U~XBpdfkQ% z69(ZQ8W}+=S}76FvatPueo{R%wR>uB^@r+tS7)EOl!_K-d#lH* z*QRz?uT?Kk9fFg4VC2_O^`!Iv2~_=QY8TI}I0+ZMRsFGgb?S|&{k5Z~okWK{aHIMs zPfzXl4f-ak7D_JuA<>lz>oF5H?R5P}L}A@WK@no|+jyW^qI}WZm4kU(oYupqkZerr$~iU7IhAUoay9yom&Gx`%{WCm2>dMU4OT7 z66J4Q;>vLb^TAzyS4BCn0pO|1zTsQTR28fFJaf6{Mb{E2dS3$-QHVDs6h`j(dA-PY zzkcW7*}HGO@n5Tc-?t$|zq}F`lqLo?@1&QX-h6fX{Pd>|U*+_Lo4cpKo&FL{fAO5N ze`@;N&HbfKuerHr`U^o2o?BG9ZtI8V{|y3N zOT(Jvp05+G1=QK`_zEe4*c6Rhj!Wnqi63uYvIDI8P`Tj%m!uGj$Q7NCn+D5;LAp^t zJUIVho{nfXN=isZ5gv|)q2@^uCXVuO?W1%L-EdA zlc->@6N9_H;Kv|Zgz^sMylZ0e&BVj3uQpPr5)q%w)u;0I3X;OrEy3vgGBTYku z(^{DU*>Z_9)3iCUIUC3HrN>ujXDI9=*o)n3_(Ij2dDj5N>ZU0z)Yk8 z*sa^nPvA(g?1`L6kB^uU(!Ak_1s(zO- z9wkZ)gM_3Di*gIfcR%DE={v^`iqf4AK3fdS_kvB|?w{Cw|Kq>Te0y}}@?Y16dg*)H89@JA;2gombD2FE@$PtP7f?~8@Oi0Q~_)EUMYd7CX@SXUb zZ%*=B&a0n?6f!X_q?>rDqiiH+@xcXRI0WKjG-tPwjNS^|#GI~KkgDjg1;uBw}6O=3O#`9anJEyi~4=k;E zl>o0aQ*<| z%^V*5uH$T44c1d}4xZ5|2jOf)yxJedX) zFVZL;#jvz8fYMXzz!9d4L>7faQ2Ku!y(C;`v44%rq7 ztpgt5p~Qr=(zptF?jJBrEux`SPGkP;^vU}yY$;rM3_s^(oW-n#Qz z_Uye@Bq7$+hJ9HIguVk8sZN0JMqT-_NfW%6Ovy>VnHQ?MbMWlU z>j&|n)%{NMH z=FF@254?N-oue=$FDTWY(kxQEL2;5-kO?W#g;#x17d`uX6h)Am`~aSPbK9_?*92&& zL{y5k-T?*=o@Q*Go^P&{5;3tCzO#2A7K`;JL@CxgIM|=)RZ_`h@9M#Rsdq3fCMLXH zHZp>G(Ek?0^zC^8d@Eptta*JW9v$lSHd6obC(S0XprK7`T(5)dTGVPg9$$&1AP4U_ zxs+yABSUzvke!5@w4F@acJ++3ASWZudmbQs+6fI3k@pBm#Ujn4u6bZo0=+}0p|p|Y zfbixUQ8ny5K=PO{D{nYW%EEv|5mHT^JtMbJN8Cbx+x6RrM?-T40k%n)|DsU)JaNMW zHdugs+s;T^T8IZnWZ5JRSNMtR(bXtfqguPO|(kS5;Vqi6wi*n z{FvsggR7cs;WdPptUd`cQ~$W|U+x=*WKw|V2iC7NuYIoUYM(26BlqJ-NP^tt!F#Or z+XYVpZD4wq>5U$XA%ac7Lh40QES7o<99xH);F#@kJY-?Z>52a2K~XrN(Bm=T38tGs zUb3j)0Of%gChF&9^ecv6LI40IW{Vg^*T8UBgUe!o!Q)gds@hiFZ;|A=<^Uf+m`>Nh ztMi|DBR}E5#f6j@nM5v=2__#ynrj<(=?Z5>#rjfBQnhbC}c-NH3 zICbNpvJ6k{2?>w1kd8H$qCI|<*-Q*8-?+eVzbfoJW@YKy+O)Ft*`{S+&AEB_OZ^6F zTD;(Kkg1@7p>27Fu$`q0l{8#7dHG{M&7swxNSpZ<>&v#j#`R?z;YA`kau0K^XS*Z| zqMQQbQ;5rLtum4*q?3(;6jgxEW7NBNb>{ftnd9%@{rW@S($%SIPe4i*VoC2|rvayn zfQAO)wKS+|HA>secL!#^`|#JvBX`g5`oC8Wf|fQ0nh<<|f6z9{-Ds;YApGAeuiZU& zdFJ2;zg|5y^XJ`gV$m`bD$1GHWw_#UbA)6VBf}HC;xo^uSZwbc4kL8 zdHLSwyYHQP@7`-??|uL7%sX!eY}hJGLOLZ#k|aoR7ZaZE5^a%}FG)=+U2xd%oI8B? z;I)~_Ln~He?YRL;9@CZiS;VxV57n_XKrBhPoEF#}&eKzJV* z9(gde9VDyouo2bq^*6!joc9a(?Rq!jHF$#mwb)8V;T6g79M68L(d+&hNBNN3nq%YK z+Y(t5ysFCl`CYSe(#VGBJB;(RkH-2OKV{%+VeMl^dGRq?Y#@0Ztqnx>wgne68YpnY zO@ip+K=dvPiXbbEmYlA6eEKgDgV|YP@b)IuNg`cDB6UaSXV3BT^msd{nxAM#n}s`Y)=5R|zfhx}}ud4htnf0VaHrnFw&% zjB`CSLvwDI^F)jDpeh6@;7Un7`X-05n_-mV3cAO}GzO0SMcpC`(~ACDyn_5Uj8K%klZX1;=Y9R-r0UudhqP-sHHB-%NIOqEndC_Gdrek$asI%?y-AHh!YP9S= zve>h@z``=v<+LC+HKogY71SGF-T36j;TtDzyuZjy0QkwL;D`VD#`|c+jrVT61KXe6 zIC|rY8z0_y-+yV`K%VF(%up`TsJpV#o?TgrC`e5e56e+7=0ee9SR?wlO|-!DJZ;9K z6-h|Oqp7N1C&!B6@<5ADSx~}r2(*QX@CQ4`Obw60a2B2CVh9^}2O0CzHeTjkL z<8rpwAHMAU;VUipcL2RWLcgL!+_rIju;(a@=HZ9?d;{CLL|i2)q?+0YMlb_fIeGfw2j8rec$tor z6hui8#AJG+VINGF`F`s8SO50vi3cZ7jEp@0@Z+QCx3?xAo_ZI|qB*xTTq}#bjzx?s z6Yd>uJRL<3P95ZC(y7XKEJZKoYown2mAM-^uv$dM-M-} z`tajJ{>79KT}-E4IcHK`MkVhu8W+;RsV-LjJKAV#7A@tmNsvK`Yz%9hkYo!fY2|0$ zBe7u0=Sx)6IGJwAE|$=WlEF&MgnI^Xsg^-=DY2czNF-BrxKgpYEIZpC0R#0T75VXa zq@RtkZe-oM{Fq(IqEPy+yb)e8i65?HyKnaN$F`e>XYBBVJT{&rx@$T8gVmPvz zWQl>>?}Zp_5ljp5$p2jn)u}oua#M;JMXFPzw!;dPk8eS&NTzsL3ew+n4uz2fi%W?AZM}zt_t*39)pejad_(y$%vW~$soL3c~ z)cO1rk9>X-FXxDhk(9VRK%69zcA_pIkX0)sgk(DM|IEV!t5~=4tdwWr&@D^6MUWh8 zv~i=5&%BV&+oxyV{NUc@*YAG)AzE?&>!bJIx!~WL=~Vbf;=O`or4cf_@VGj)AzZ}H z+cj^^sh+G(Rl>n9` zNKNJ1%W~G6^{^7#ne)PTtsQZo$vo{a>KptRc^Rvm z=#;<&LnkdHq;}k!^&)XZUM_L|zu$o}FiH%ftBXw+*kb20<(__9&k|FkIsZ!80M_7f zkfM8`iHmi(8ZHZRg_iKM!CXXECh%`g11Ioz)xE8k4Ci zPoyBY$aNhx4=WS*L>Y3{4sXe`@dUooK>Q7UoUWd$p5_O+uj;gO;L_9)XZtMQ=bxH-+o>PlzP^7OSb-@kt4mDx4UiOW>aPrVCguRgk* zFPc+_rrw^g3lkf$YOrx!q@cZFHOuL}c@yA?>T&lC_>x+^hN_d*AJNoaSTy+;PQqfj zyLySQtq6YKBy3Ld?}f`JtCM`0y#`-by*9P`mtCL#{qhr1X*<^7tNKax3_P8^)gNZZ zJ6XNpT0vW!?S0I{^S`!guOBPB?3;9`!)IJTAyL_Z(L*N81r8jkyZpT78l@^2j$h(T&qWH>5~%| zfY=DjpxjV8+u?AdL^xpG$x)tpA# zx8L6Lx7&NazrFX|?Y)<8?|t|7-Xpj7etUcG;gONsyC-k&{_ghPuWs+%Hyv*vrIlH}^u|}9~fjOJ2e2Y<*%#M@jJWojW zjoS+bdSOEywY1BUN_)4Y;;|N`&Yp$P`97Ep2e3tQ%qWBOC(?+3=z)T6=J{17&wJ+8 zYO~FA*TejoWyO(^M^hFED!3CymH<+FM*3o5r23sJk<|h zZ(s)w#OE25B=n>^!Q z9jWjGdL8Ow8yl4Mt~Q9NGB~($l@4noF(&{n3%?AnBX4cXxChcXzR2+E%h{$?`Bq1L zDl;vNHsG8cq|@!A=TisGciT*WEG|e=%nhA9eQ5-mRt*-atOG+U>Z zStz3wdArOj;Qbepb&8WX&O0GVVno~JQVocvv7K#5OuNF864LE8*jXZgJPHrwY|9Fi z!q9>EfXr+B(&@;Cbm1||3@aQ46+j~{B$_B8J@;A$H2!|V33P7Tux(`I? z?EM~L zOA-!)j8WcTgodTxi?FIh#eVZJp?sna&*I`mWlf9FvdmU6n9A#B2*O1?&H{KOI+esP z^X^Du6LlrEJc6P|yBz*u!MrZ%L9O=Xu>3XfTkzkX(T>@3)kSH>wMM^7sc~e0Jr_nyHMDi8Y#a zGTzrvLwn9dYq;Ft*~Fk02z71zeOCiQT^q^E)X_`Bulam|m?rPO^g z`F91_>*00JC@Jky`7gQUit3$8ObxVFNl8JH8+{s&ZS~VN07w{En{7j=){Ie?6y4wT z>D^0*W?nz=7$I5PTp_)d!7R9*IT(g7C-=TSaOd1>clQ6_8xskZ{Ukii7_TUDxdr{# zJ;+W~lGrFGo{Mb(hnY(s+&lQeox`UegTo7(n{apmm$3d!U~qfS=ePHKaeMFAxA**b zd(U6*9r-lmYmVXaR_EI{1uv8@6g11(JNM(eQwx^_@@rZx>9 z-iJvHOdznt3uQH&;Q+0=ew>C^EDIPZLb}PL77^^W4e)Qf&%>bLNuURlhaP-7`QY^X zzESQJy)K}f6cS=!N1YR493I|KBZe4;i6w`Mw=RD~d(khuj=l2AaQ@ciy+~uXE}f+4 z)}=4?Km7^dbsz2Bl(TtL1g;c-Z}txG?_vzi?uT9G&Wr^al2+0}3$Db=0kl4lW53kf+&iE&xM#Gi8+Uh)!TFi>#9L`K#iJ}FNrzPu>Z91>sUGyH4J7vXQEZv zoPWlK&CJ&3F@4v0zC8n%7UcG#P)kITW~Vl%sX8m=%bFIFJ6sWkL=#W0XEk7JDGR~^ zfb?aHR7XP=woJ5Q`*x!Dg0{YSRo}qcKTK>USk0G_B*b`{N-2*~>iLn8=bv%FhDou+ zvrk+q_Uw^fXRsz8>q^cXN(1y)EAu6dS!i7yKHqg2GOx=>wAUVb>3GTEz2b>bW{Hl6 z=_XecA(d$K07kBiwn2XQ>PZ;vQRr&gTGfGEJRP=KSr47(d`9k=yD<%g0{xe}(NBn3I@ZdNa)-C=8?7SaMDV3S6k zX%`~QB&B@NHN6vvh%r9}6J47z=py;9hQx*;QnCoAmgEGa%44z{whlFAH_S45a#Pl+ zw}0|k4hm2A_}6t21*&QAo|!09MjGCUh(5{;mGBY6=~1eJ1=B!BS{p=n!?S2cwrkTL zy5=L>1tUU#U8)iK^LKd2Dr;or23%&gHK95GjEp$Z7X}Z*m?d2^7aJ?lYN{2HsWQXAQ+1H3 z+puLaU6x$n`6nF!=#qjMoO5-_>w#HHEYw#F+-|9HVe`+|rSN_-a=J0*n)h00edFkj z!#DnR)nX6oOTmt|B3o0+Qg@L(smi$zj%VYmdp<2(P zb#USyYXcPCk|y0`9y~S^y@l?0=cCDcUmd!$cmK?ZGmB4!oH_mW{f|ER_2L_M4*Jh^ zUTunsB8o{D1a+5F+_TR~N(!-9L>%q}p3D-+b*Fj(1@b9LDM3jGZsiC9@gg@LVd+BK zNX7)qv6Djagw~8f1CLsDD4gBi;kE0wY~9$mX-(gTzKv*j)yB;mR{6%|C88IoRZCgN z&Ko(4O&uD~$i!oUl6aa8NNN`UQt3akUZ=S{j1ery4LB1(b$@)5k@~`;cwTtolAy?q zXxd)eK~_&r?WvxvUYpugy->YcJ&mSz!vWZxIs&_ws;7I=bI(0jeYbkLdU2Iai;D`%Jy@ESB7Yibgds@p87t z-MgV>bSS(@N}|aaFuf)lR!_jY_9KYa1>{H>7u3@rRhOz478NwrgD_R?h6&;)R6Pd& zT&iC7jY*$vv3Yp7Djq$fkrw25;AqW~H_ZyXAico*%Zj=ZC z{ZbE{JJS)Ac%eZ=nEqJ8bXa;2dHMfm@7@TESbyE^2B9 zEHl(VND`b}Q{_4uJJZHah7?($$gl@bii-#5-&q7}QQ!>3Mxev;4{E4Vqj6L6mQ%`%QK8BTa*C7Wz^l;$o8&A~B+Jm6q06fOu%<>q5@)Fev6_UcMG47- z?xR^G%L=U(!V8*?J0&B8nsNkQ9a~iEg1&hPl`Dik)qck5ApCyqrsvz5S8w#)^ILJL zK;2ZL{ESg4{G5F*b7Bo!?4leAY;wZieXe?n!wm}Q3pPE)xn#WB4&WykWu<|$ zx~}mFMVSw=DBE4_n!uwH==<>PQ{x9NkAL#n_^~UGzIbhXV2^KPT-{MYMykZ9NFHNU z;w-O&k7^7}3>Wk1sG1&U#S`TEDse3!N2f*AG)n~lMY2GHNWP8oaf7#FEmtxWtWhNt616zXxe)olk#`Y$iorpv4VY zhVZDwSsjL0lma?6s1!^LTP3DKv{0&*#9M?I$Jg*EdmdL#)MM+a_A*syn?al+Qml=r zhgvzbWarK{3=_F-FooLEi&W_=8U%2Bwq+{TaywNsU6iqXQLT@>q^Nd=G?!a#K3gK@ z)G>GtO3H_v8eOy81G#=OT|y50rM?-j;5cr{ucQbxEYxnYjV_sHwpQbAVqg=sIW1o1 z>IdKN)*rEwXHgTeX$hzm`-QXZ@eLEWloTn`4ilMy3sf(fF!I@;w~Yo6x&#~bG8c=g zM!ETQaFHc_cLfi1s27nqm#WDuSd55FjFJp0;S z!3Q=IS~#0!!)WJDi|^|uRNb|!EXlCUj~r2pEJGqkGlU~A5hh)fRkx#}x zdHvxRAE2e<@17ew^xoK&J^p>R6D^-^5s@l7f#g{!5CtwSRj)>d?R}69|FxX%ap#rc zOHQr_?9IYMvi13PLPz+T5@WUY}2z-E(ZWn2o5#n)HvLhiy&i5(jd#(gcEAut$u-t~1Vi9MFFnp?7 zf%VK@@xlvhsoF(a{%D(+9qYoTX|pk;Wzg?z0{x`@D@)gDuDYPF^^cvMEodbrsGYD( z;U3$_uo|K^!dey{JZrUpn8Hk%Aclc-vTjS?ijL-1O@)rs(%9~MUbBdy5thBWYgd?7 zj`Brwz62|XW?Q`uo5a+MP(IR9)!Iil<@R zc4X?=O=Kz*T1LC*i*|s?bqi&fS{8BF@oN@mW_lLqpn8t8)Tjg>c?4VwU&hcC-72Y# zj35ZK@JTY4h*v`sF|60muzuv&&u<=n^4>+?*fxq`JSz{;)`OWoJh@dv>Ns`D*x9R(&K`Yy{1W4I zTv|bRt)0!Pq;KsS=Q!w*pu+t|aXpQa-^Z`)A3t)`KfXlCmE2wx-_%<1a-p5bHCd6+ zZHMA2RCA%6;F5_*YbuLx%h38;1HHEfRF+2-hVYhP6>|9c(%BY9u(eK+;{3Tk_zueF z^fIkqIHA^n3O~arAzq8++&+H$=I!g>_kVxr`+Y6w_S?6w!q)EFSH6Gs`@OfnMz=4* zf&SZrx4-&+54t^Y`y2Ro?e@U;`)*(Q{?BmL{@dTM>)<4OT!N8y`%^e}87|)kN1Z$T z<@Qzh3cJg|?Xz&dORzuCa{JTUU*EpYcD{cVKG_|=gx|e-`=;}oaNrf^BpmvF@Avz@ zfBp71%WuE={ch**S-1-NL0T?8^&{D1e^g$tU$F!Id%wG>vcmm>HVfj+6^O+--%yt8 z9|gxv5Bo_^M#|OU_^E4sO|T3=DH-819T!FqXoaMVCs;ivxjy2?1p>*G;>WU+>nQ0| zFCIxQQ4>SzCp4}Nn-+6}{l`=91Dg&2AB&+s5Nwp$a`kTiB4-=!#)hG5t~BMOOYBT~ zmpl0wC-YH!$K3H98nzXcgd0YXV1<}uG0MlBQW4Cdi=iq9yu=A|J)ymGR6E$@<-rM=$>U+&e$NaUe9eWTISJtJ;o& zy+&z_d^U1<t^t& zq0(2C*I`=!L38$uTq^h8m(Vikk(zGw{n~7_b2N<#STl9oRYypwObU|sNg>I}$tWOs z)_`P-<(E3wP@(Q z#54_}xJzr5mV)Z4!WEuZ=1+?REtPs7I7w*_=X+OWd>6)I(I* zA^-9#X?zXKxt($@RzRuv8&{GvHf#;Sr*9B8bfiFijJe8(3S!(W4r9D^K0O&Hn5gQ zYl76`U5%Xz+iMI~TD0VLqPXOWXn*U6R24W=4P;|lbj2Bt-nTAv`O7fnKlA_4PYp4_z)4NgfyKiW>Zq0+T zk~QzWhK&r`KtahNTACRCO~czL>oM1A;u5UAi`8P}`JEo^Mh%NC8DUOFi5cQcqXlF7 zf&pmt;_$s(Vv3U!^|17gzju1<>ct0FUN3LV2g;PHNZm_ckDYsa?D)rH`(GP7aQfS? zAH8~G?5zWjZXOuB@%7l1cg9Y<;TxY7ST9i5#O9Kcn8e&*2ltFMJe36*G(V;mpRp-h_?^LAK~ezL$xas-Ck7+u7T$P*5R zh7={GRt*isC^H3xg-D!}f>T&z?hn4E zRYiI?G0=}M{{?N(hmKLCF^yIkx&Z6Y{lUG+8v3%pHm`kFLiYzhSs+lavWu|Pvh+}- zVJ*xungDhZhF_FQjKXp#`l{X@SDjo6*sKH*LV?pSI}=G5_?UP5z|me`pdyNN?9bi=@KDbY|H+(*|;RMZUsQz`GYuvnUwl{st_a3)v6c{7Lmrtfxka%gFr zUMv|jo3l*RO&JXc3o>bj`?M%66E|+zKw0escvt#5I}w~j&riHK?3&FRTehuk>)S?6 z8>_3SO4@RaI}dK&xa|MAjfxF=kDi0~0BxZf4I<-Abk$0tQY}whftJmDOq#hkUQC+v zg^JH$<6YFFn^?H0t1#(Wfh$MXJIIp}pHmeka1X46X53A(+$<>;#c;VH$?GuvhKL9w z8n2$Xyk{4j8a*)bIm-;#Kba~py1!+WTFfnH(f%x69zI|G-$Rq(|NBR-j+}&x_KlqG zb6>$?>T9k*ONIC|^9?+ow| zxmYZ+N4Z8wH)W-EL!Gt+iIXEo)-)N7gc$vjWJyPSvK~g4twbS3$4d9_%Ud%U#x+l& z!f7<=k`1{8mzCGwaaD`_gR3gSnJST{;3QDd@6P^)Daq-|4ju%5J zlFdBSCBxK_|9kZFJAfa7wRD)5Nj@sVM6 zjHvfq;1UwSxRmNGpk}-2}yDmlU$0gHaB}NE~P9XriId!Xz0J;2fKA@ z{C5+|V2zat2c-Z$^`WbXswQHIX_F{EOY zAq99-3EG0YiAt4uwlJ+WSX)cFoz+cN+MZRff=0CyXgF-Hu~`i&mcs%4tNOS0JV&tB z{2W$T0W7qJ=tkBNL?%};*D~gLo&@nR7LOd`?orvL7wV|2S?92n1duDLXGzZrt;09e zx~jeJ$CnRxcD5M=!!IJ%#B${SlASv#8*d7-sHKu!yU@;^9rz{a$(K+cq6K#0vy)qY zQgB%_W=Eja8c4y<0?@4*oK`zKp{x233u2u-Zdf@(FY1}jPIg97vB?tn*{OA!39m-^ zxNc;0=imlXq!~w|#+{vTq6-(O%5mJ1Uun=cjB024o6})i7{Lu_5&AVy+Ua(x=7`Fo zl@M&IEQxiI(o=BNSC>c_U6C<_H4GS@-)msPUdLvMg z`G^{fOmmwUG)prfX@0WE#pFojuLSwQx}Q<&$v>zU5Q$5O)dt;0%KR=#y92BFK1Vcv z*0k(&A1ydRJeoqDRrQ`-yPkUCyJH%~CEIKH6(w*N6{6(XpCZS(gcKN0kiR?N#_&?I zitA-`x35^<(brKbGRpIpD61#947tu&aAv9b3_D3OjA?c?+!J*{pJw4zV3~L9K=b5* zkcV%hME%y1mc<6vn75O0F)>{g`zOI1d?G?z z-=M7~8LYW@vv^m~pvQ5Vaw^~@0@r0-@ID+1aLi-1K|X^3K~O1H4KJ0do=ImVH)(1o zx~(=<4W}nZul)Sw^;rT3)^Du|9N0>VcAA;UKxJO9!*cpm1QGo5;h`t*z6p)F_~h+V zwTLwFeA9G9raEIXULYxO0xxm`h<`lq3K;P1{7kX*JR!I|zZeD%2$n;qwV>IP4H-Id zE>@Gu6Y&Byf*0t?U-v(G@1k#P8^sHhVwrP-RPDR-69F=Ic;DlXULJpE|AT9*rG!k{J`7@FWY}HdF_WZN&YN6A@90@mh6g~{ zP;#`JK_Vkv&7iHTw`>8ZKy1sHXz2?t5UmB&}=FUO9)J!I*Y%_;yqN{gNOzq zfuMaZe}uX_9^LV^?G zk;$VRJB+p%m>QIzHfZrOw2sZadZ_jfA;~3Ef#F&3qKmGSGz+uRkVTa=sBl#voJ0^Y zQ-RYj7d%dy4%bkOY=fpr!{XS2Ati9B8d}Q8??ozpFO{eP@>&{*kjy27DAKSJ+m^Qr zSg$KAw}ME3;Isu)d!M!3JACihy|?b2 zy7&IQxBmo3PTu_XNUI-x-eej|G0;8Mbz5j7i! zKGSf9QIbqF);D}BiBZyA{8FxvLhk{Evcx6DYGYH*_K4ijhdWq26jqlZ3p1AH0&YP+ z_J6!{=k_hDH@B_Zwd=?JqrZ|9q+z*U+oKXzb{{vaon%0?N6=O8CKvqX)to6 zgj!VtFODiI#1~Wyd43i8NflEfpsv8yqYvH3Jafyf# zyc(6d*ryu;3oKfx$WS}0athKXRmd{5f@tKs<9c_4^wUZd8Qb&-Rf0>^N7}PM^wU#1 zO5eXWs3Yb6;L*ltr($LO7d@~4e2hy)HPhWwTA@IV9PJHrLgPg)R+Ajy*<=7p&1NH$ z*7K&4BVELpf2jfO-JtvDF>1}@3@JF=q`K20-6B)1W zrB#&Kpe#xZ8wN^VHyHI0F|4QEL{mLy8Pv2ufkx}Gp`uj(Xj1b zjWILjewcLZ2s8S1ggLQ^!&s@#F{GA9Y+5yT1Y4Mq#1`}n?V<`s37L;2W3dFvP?*@- zSygqcwzM(oT$;97UYEJN9jo2g(xGhzRaIRS`)ZDKcC!38R<{;ej3uY`D;a7kwrRo* zNoVIXh!_T(WxonfSfncMWoPEw;GJwGTH`d+|68FF)Q(NlMFP{z$7W8_hU0U-90^W} zc}~-AY9ovsSw$-;D}1L|)wZIChxd-Xb$smbpU2+)>L1q+Jh;;T_``$$xc=%asS0b? z*BFh~>M#YY*ByB!d`gS46Mr4wd*SiPn`8SgKe~JYnm2xAuW!5rKFJBp@sQ#~v4WZ3 zbSwr}ii>$S>CLe2wG)z=hnzhtaHc>tOV<##91E9||I(_e&8=O`MT^$3ulfdN9ZVi7 zFY&JR&v%eHpdv7mp1?7ei$T3&uD@=qW{BCS+F1TQO$;c3<7*gfBcQ5)vJ9f)9*u?? zv&7I6K^3kmEOX(g!bQ@<7qTNvG_-V(P)!TL4>56({RW_flkxebMXUr-JB>|(;3o@M z1vQ&V)4qXU!xAfTp#r)fR8$9N+av4=P7ovON5%Sq3%W{*;RS>-Oj2AjS?&7ptZZ+f zTE2Nb=-4E0IpvS$(coCkq4gZvj)}s8LhIM2MSf-$e15L(7kMt0ii)7ShD5r)?2Sf@ z2~psZv54#5mBLZLpeOYNm$&N$WG33ANYt(pwM|-}4T~-gBxH+mE-psZ2&N4j#zYlL z%`=8af=kxm;h9B(F;MMiJP;#iM+T6?ns601&yJi%qx(kpF$#r|lcNV=cix$RM$U|! zg(v^Q;kIySsh}CYG1{2My@o67k(Wi6G9-FNDaDDYs6N4z4A5npiXLvJ1fU|txMWQk z=O|O6JiVz3R&Na~LmecaC;6evp%MQwphPYvMs~P=CLCbR!Ws@OSDgVRaf#sc^9yL6 zqxn25i9;-CPnW72AxQ^PoJ-YYMMsR;^)qIF{MnPg?*HYZ1HNHx5MwsOHpwM}GuAJb zNwKKwDhYK?lzC39!O64vEln+ppgduvbGDPTO0~R+3kB6~pj+a%5y)FLi$L?y3S0to zd;_vlSGcv%4OXvv*hn*v8);5JL@tr44nFeins9!O)z`{17lIBAa@seq?4r-HGiV#t z6h@r5QQ4>)hTh{Fl2x|Kstd!}Zs`~bm)tn~@apxkuio=- z%5GSGi=5!(sMOw+h{tkU>WbH|OTyz>x{(~55pMy(G06^Z?S_i21O`trz0=BGIcZQ47!&o{PJoYt?UKYK6Ax40?#QH)p&PYx8G`o z3<8Pc`DitBd0|yjv~&d<#gJArUgi=t#NmD-?c4N1kri>!H5P9V?ZFvjmTMk8^@bi6 zYFPn=?X|SQMMM*YQq}TS)K?&(xhw4 ztpVszvJ9;mx~viZ2D(OBOnFvBC{g2=&3VP?hZE<2IC1lb6MKF*@lL2Crwq-UApCIR z$`2?0#{Pcx=O0d-ML(Rl^uvi8Kb$z{?4N+`3Pk_>+Lm^^fN2T;KRL>}YWWW*K6lQY zgUf&T?6n_GyyqO*`@@N^o#V%RV?6x_>RQ5BOm?IsyO*%MfuWm2iNE8WRF?=2w|=Z^ zI5d)R28hJP;(-GqGxrByQ*z6YL5eCgGt+`Pas-(yg@Tw!!5T^v^yABaK^yd;W3ZyJ z1x*>c0E^oF!MzIv%C)pubYk-`>cW=YApU`7VdfxkVU8eho~6Sif~UhU@h}JT1mjun zVtF2#LYR=a6=$f9{tayocean`nn^S_nQh-NI>=u+27X=0FZZAdXFZ z0d3a}HM5MlNHo-1U}jTzF*@`jl42-BGwcL4klv%A62V3@fp03#b5ITOsU-g|PrYPX z^+vFAs+*ySj3pW^U8Tay4UI+{oc+!4t`HL@V+2BY)_=XI4ujkZ_A{oc>aZLZlETC3D`eD<>qXb6j*0zF7rkuhbiGPg^6eBGE#5? zx5W}yQNX1T3)cWlVvI}h-U}qIWR0XX1Oy#8YdSgNFzLaBI#Yhr(?L}!$K+a_+evSn zu6R!nV{CeU&6?*ow%6j*v~gY2bbOjNQdlFi4w2^sN_-47Fp=Zq0wY^c7NxF%u7o9qf^KMFePJd2cVxkqa7aImat(k6UhHa1*w5kNcnBIlf65Jg? z<9<#T>^!O96BjB2$=UXZj>yI1k?3CuPFHnBp3b=A7G^1ItZ6(h^Ky|YfvpmC4jL&2N=~)(te74lDbRwBa^HPGzdLT-`sk@&PpTO+ zx@iVfeir^>=7+xV?~ux2l@`!?qT#gR=0c2#mGz{&;}HxHj}y`OaIGlNX$DrXX63>f z9}*|lFh9+%UA9B_nNF)?~WW-^0ciOph?X=-0nqhz{o8a}n%Mp#8Og0^4- zXBZQqZWqy3QZ$GOjde(8P#d(KnbLj6^UVU=x$diu`&obq$-&!-}!GvO`RjM$f$ar53Fa2jW(jimxxYSY++0iiGkuCgdG(^%S4%Z$4P=c#Eq zX5H|I&}wXv9BNf5S`XcUmi`X+^netN*n3Be8GLO`>)NXa)l?sWMq8v7R#68r3~K2{ zsWQ~W6zH0aUX>GKF4MJKNp+GvraJLXxw2$L62(%Tuv)hq9&8xRl)%0iA(&E)upW|~ zO-{lZR=x1#N*d1@&DNe|N8IVM0a0z zHm=@1&JN$bjP72%`vq*=aK4NB{?7&2*mL*N-OD_B7VX?w+3)*5*Wt>)vM;fJyLQ3V zcl%-U(%sMCj-Ro6v2!<^zwApF;SP7-fOB8my>|D?-7n$u3v~BuI56n^H2m86yO;lG zzjMW9j^n-^Z26Te9iEB{dWa^R{>seJINvW2pv8c>!n(1)ya`=+8hV#~UWU)|U8mkJ z-u;I09^Wy6Wi!U=N-Y>W_r`-Oua52Ci&{`THg@3U`N|~!;M$=F*RDOjHaK?fm51+q z{_xaSV{aTB`|87oH%>n~?mwr8QOffvo)aW4F1zOB%9QLGBA!du_}wFAu1cB9!UszY z4ow0DphQltDYn|>gBh;P7&<+4n(-SLXbW~KGQ((EW8gvoQzP6e8op=4HVT&xeW9V1 zC9L^Z3|7FC!wrhLq0<@N6=CFCffLGkkJSSx%6qJzljqnwzr4Ti=W|DVgWITtM}Z|g zCTdDkN0VdKHYkAZE|f}d@WaVXFNBM4kV zs1C_6Z14Kd@LPBKhc69Z8@|yD8&`&}-`Nd2=kC0Lh6jdk-08n_2tLoDrEp^J@Z~$N zz$M?@dF>e%iavZEcCHOyxwFstJa^}1*nVSvDLBI?q4C#;Z$b-h41WV{*n?pA3UtLe z1V?YUd*Kq|Q;sjvXRf`PMw*#-`7kTVH?qgn8-b+}*JiVMR&;mnWNE`~7&;2Aunmn^ z7HZGwJ-c>=YW?$%*D7113PDS|s9Nx!*|~w}MdHdEaOK%}O|tkQkghc}sVU;df_-=TSg#xRCC&sIFTF)s5oAmCX z;|N*-2<34L+PQO`)}=E(gJz_gdK&NAg_bN?vPz|BHC1U5ttCp{RC2_CqpTb#EQCm< zrll8grn3{!<&%k2XD2h0PVy)Z z-KY(ncD+3*$>n!;c09g8A}1xHU?7*JO4SR*z*b0BDpBTy8V|E)+0t>Cz`SDi!5`)& z*`!-cM$@}cQMce4id9xzEpHc5+8{Vjf3=baRU;@cOmpVsvmoF&3r!TbRKmXl+Q3Sl zSku@$MM+G!^l{O%SxW>_v?k$b_Tbd#H7C0R4m4G1fm>Jw<*mfDRcp4ymZyR}AsI`| zHJ}_G9MnN@0kzr1T5OuyTAQY0)6|Al+NB$s(^r#0HWDXuyufq3D7dCO*Vy8H46Y$&&#y%YD2-TUC)+vwhrdvD$Q6!5wC zw0lN1wbj;`X-fO;h>fUL%*T?tXuFD@R z@tnj*iTf_+V;4xuuoZal)G0r(l)q%;pl@11Gek3y9lqg1l0c%Xp2Jot?LgU~ z(|Xpx#T+H5x%s~wMT(%z&}j=(RUJA#&q^h2C_g?y2I;nmiA;qu(X@-cXcPZPZK1?4 zh-DGpmEzjWOUW5$GdMmcSTZLCuZp>VhR@V|M(ivBi7aSfS-sO)oQo$SALMKjTMyNi zp>?XtJYS*CaY-%~3+xdArDwgK>%j&^8?ctOp?&j>Ua4c6g39(<*5`C+5>h8x31mYQ=6ql&+llY=a#RdVT8&H8Cf9lbwA}66r9v04xx^57(6f2fK zQ*z`#Lqk>3TCA#i&k_@DH!!twE>v8kSE5(eD}hz9tmTKEH8Tg^DmY53%?0D9KY9G& z!C4Fi9ouUR1#p+1rzC;((HrN-ZoKyB!dv5Ce)?$O@S_i2e{|q;FUVB7Vlo*#A#;2( z#tA&ns^hri0C8reMw}ZMflCG_k69Eh5VR3z^scVPtU0L)NGZ>%rz~-Djc5=k8{E_j zgw1Po-ll?YvP}b9XnIEY3^un0CxBP7IoPI0*y3EQh8fd){VGEg(zPU)$UOQX9- z&WxN!Bd3FEBE0jDg~L3W_}VmSNx$a&{>?DEmw&heSQxg<rCF8N>YsOMRWp50 z2OI4iO``(VOx-qILa0m3;k}DFEVEsXj2}}`al|NDIYZBu8byB-NQaYCur^n(rquE- ze6Gp>$v`&0&|Y{*I{#>jH(+F8BkRBS*EmKEw>?8h%xupP60`Tc$dbQPd{oM|X$8%q zt-6x0TU)M6Q5IPlTe&9FJ1WRMtswXQGY_uxkALvtgX$O@Ns5a}ffMH!x(}Xe`S9W-MV6#3C9Bm` z&t(m_tZar+^V@WJqg2s$-GZrVo33IlfE8!6mM3anC#s7Fr&*X$FA?VKcY#X=XE9hn zQ&mO4GO%e<1>0&kOgN&Gms?|K_LYErgk>GG#OnEHY92l@{2}`Gi{Uq+3H^6=4__J{ zET>U_JBWtQBj?E1&JmV7ok}i;KmHS_hmXN+E)Jg?zIJC1OU52P3j0@wud^Q=K3Bmn z!6z$KCdxY=F^q!|-^4AGV&B)NDjuu1j1J$3;ry+ornYW^=| zx%U4P!6a%C=8m@%rNmB89Z;0gj2`_BwH|%8q-357e}22JQnWGv!$?Zx_ykLempMLO zLCbe!zv$G0ja3_ZJPQ`TYe)Yv&nD0|qg1pA%Y9w-l0`J=EqIAYf!YP;fSiaz^vSxX zE8JQ_HLKS(a9orVZZKqdPBa9LmmJ^jXbeT*QjtMM|cbx1PB$BidYR5mIN1lfgjE?%JQbFm1OMjzDK7A#tyvq==iBw(uvwO)fi3M;LUtGDoT$1?cMSI z&&OYV?eU=*iZL}wQqy!e$~}=O7bluPh1KJdB~Idacb-G4va+W$g%{f3P*yie zEn9SzT2#R*(v2+EXfG}otWKBE1WJ8Vm8hMhlBsWQo_;+x;Wwy;X1RLecT5)dJ9SCl zw0po8%9nE-_q6ArOt#DQ{`|8Zb@z?$SF2SeFJUc147AS7!v=Pxpa`)_z`e8W@g0*u zps%5U@)0(C`@`D<-|xHq_3aC{Z+^e~_SxG5Ew_)|zIl7__Ek81_4~cIuYZ5&`#tkP z-oO2q+gER2gJ5CU4^+UN`FzXmHz(Te-&e5`4wp>Zr4mSqin-O<_6Q?$ zLseUoA~#h1WuFgT`T5uZR`vNcc>cYIe*W{@PY#|4(I+J*xMU(S8&29p3_DE~Tg%vu zQZ#}s)G5&aqI%r3L>%23$b%lJEkm1ZBZF^!a%&(os3capQVw2Bc)`bm=T%Z#hSupl z;py#?Q?)X>WBi3%~2rCB;vNf=?yl&AN&5HXJBHD=cD?JX&QNLl=s;84eE} z-+`MmLzgw{yEpv{@Y_1pqxI`rR}Q`9n)F=kIv&Y&0H`DU%`N%l#8$oU%Kczn4GBkX6c#xHyxwG48=EXbx zGaM>^<<2Xt`iv7{cmU1}`sY#Fhz+V?J{EJC*D4xX%A>i_Yx8g}SOx zHbT(g1&)tEsVGyJVvB^!KPs{MpsAWHCx2GYwzaF;)^A_8eqCoLB8H(GtPC&9L`DCa z5;^1q%o!S15zB@sTSGItomL6TSyr)S`Es;UXT^Lcf3aoFs!glc{eD$@UzTXZfG@JL zzRXlfyAVc2Lx=Ac5P6B3mdT;!Wyl~!6?!_;GI__z8?&0fliRbtJZLG;OLFX)dG!)? zFvgSbj@jA{dg~S6@6U=gd}^A-ym@9m)H&EtC6*wUibnw>ljFGvtx-cXi~t}>kUoQx zXPp{7F!DK|F}i=`!pP?%UyYm{-QTiGE#{Uxq*$MiT!PP+|M$@7ZigicF1tE%5-!>| za=Oob1!IJ&xdJT};?K-8VBz`CIK6Ue39a4ZS6&}` zKg$8x{qfP35%VMxq{)F<1@~L@wuQ71TGfwD7nNa>?Ln zsS62`iIDgV0nJLmnw1MBCLR!xxp;@g1CF30$zOsyEuiY5_ z+DSb7hSfZU3s_EDKkQ$3^zr>H_3axcH;pB}UHNv9sp|j1KlQ4)V`aGE=-PW+D!S`v z%kW>Fx~RSrE93$ zH97jux0G0hU#!Oe<2m1QgTbK{;ZW7Ev?E(k)idNX6J< zwT<8klak5bXE$nDrK+VK+*khit=^2lk(krpjCJt6EXFG8wr!&amNMTj##U^mv1@_^ zv}d6+p7YH}aH3SLV0bov!OAuA>B!a5{f_18mHFp;9o-F2esJXcr0I?Q-8*_PfDg_t zkPHGmO>EGv`j)LE7q6K-{n)4;e>6D$*`=|sF3(2;)la+M`f2y^pZ2`^=*^EFe|lr= z(6tBG`p1vFI=27Iu_J#Od-c!b7jKNca&7$8{-1V#_3(qiv11n>T)X!0)zbm3SuqZT zY@X*UkV%zsBzg9YOR*VBG zq5SCgXesne%)hC911MA-(6y!F?DDmKhn8W%W(Hjpk0>0Mou zZc8}SIPd8HWst7_>gnm>9LK1$H}Nyh!KK183(QNpk(wr21>-n}7?=(4lg?6TVXit0 z601`vrXv5qq^DF@X;$x2^-%2`60Dd~H70i6BPPSg@9Y^q?}T_=Vl--Hi!|)sc^7uC z4WDa4J9iF$bmv`WOJG)n32Oo~AzVf5yKt%>p}ygx!{>&-hFhIO)&F~zS*?%5O(%ZI zx%D@9-eC4__I0@9w}aDfF?`*L#JX|kbyN;N8@}Lt>)i0*oxO0wojX~zOqu6cf7q+GS4Mzl+(qU z)_LpA``1p|NOKsSeFM`dB4+>vl=B+5g1?|#Ar^}@6B7-c?k0J}LUV_%?MM$@P$}vn z>JHl)IzA8Os|}>YOp}`pR$v*eZu5;%A9Nmtg)mh`#7u^_bv^A>GB37x@fa$s-{CbppWv6sqJYxTuQgPENEEwVGw=&BZ zjERA%|5;s4s`U~T*2-KDp(bIo@S064=hvjW(x8?Fa{E?7m7IU^mg+@Y&A3F{%A6{C zpHt(U6giMx4rGNnTCoz2JzaKE|6Jx6j}y>8nIRvtqor8HJm zv|WdWEiuvaLJYPIHG|r6xKLz1@ErATP60)kZ{fvcvPEoJA+Jcah*E52%L;LIYfIZo zUTjINa+tOzWt57D;uSr9rtqZ2XJLWqOvM>H0Kavk2(x6C&nHHH!s!!Rmbt9zF>Tj5 zx$z>G;3G^X#W0Oz4T>98-#pkTk-2!HI!(v(a-*mOYsfAxLuG=w?JJgd^jW2%zSMli z(Tg&joh@i1T-1}JDsj}Fsl8=PHB#V4YN^>dGr1Qw$cj;`vswy)~UCCA3w?8Q(9&R%$6Q5=2pSg+=f) zTlo38UkQPW3z25*1e)op-rGxc)rLjLHj$o7;}QlHtO13*rOGNMc^<7YC)ChJ6VoF! zjQ%#Gvk+2H^fR_X4WwdJ96HWM6#Djy7hY%~TFd09-3s>u$;4VdIlN)yCS%FC$V+`2 zX`Y}pm?$||);l}D9Sk#M!FOczw5RtkljwGu<_p-O?i_{$A9Ll_IXfO9jdOAhH`QVs zZ54G5ZedxC7-B(GH6g{NgeWL%N>GFw&@Nr2VJd4P&!vLL?<{m`(L5|k%g{!w;@&JZ ztT1sBxMVEiteiWUG^^sY(Xdi2!ElnfM67zv@$7_x2WUa9dCSHs$OXh8Xgf$sww5(v z-y-OD?b|nRUB6*ndk0!7GJln244uf5#(8n)mc;|(wcXYF3D7FuZ?@&UfO!k8BpR&$ zYOzZtS+cy{@x=8mU#;hD|LXKcVyM$dXPWuM%;}JDd`?Ug9G?ta`g3N-;(xSnpJxh` zAhn?%W}HW&Yc$D*G;IpYgd_Pl3DC>L@lxkI5}u7M_wY#WZ0@% zG&|3HzGDa8e{}gETKeeng@=dtKDv2$?BJ=fH{bW~>zHW#bPuppF)B!&O$t(+6KX2) zH7r~(#=-@ePL*bM+U&9@M7t)OA(vPB*g46_^f|%aSOhWG-hF)OjX$4xcPDbPW%%zF_2CzQG+n zJi*xwT{`^PokMqCFL&1PH!j^kl4h<$569;MU5;^*T$e?EVMOO>imEHbuyO&cS&a+O zkp)hwDIdIu*NmU@h#3yQ9h9jFK1)EyLA z`XrCddb1Z7;a}h6MrUW^vP^%Im=e$)doYL`xAyl2smrsG0Bl+iU|=9^*xuZ4e7anrX646~p-gK_;f~?O%%n=2seMo3KSt zs{s>V8ms6wzP={cOohU15yhAQ3Mj#9JG5Ug`0 zT29<$9n)Tk>D)Yg8iIB3@D7hDyns1sTIIUzRcJIlx4vmQ8ci#i+b4(C=WNq9(Ymyq z*DZ`HJv0Fn&+}YL1}RzKlA;?g7hImvN^xQ=upqZcf<;(JuNi8pOuPj!Pl(m>MC1gT zR`ceA3*+y##p!g(%<04x1tM0WKt#{z$eb)z8-1feG)-E!n+0Lx&@Paf{MGJg7#FozWg6|pFF(O(bXXihzA8`A^s;(Y8d-c)TqwK0zAAY#+;q`YP-FWjK*FPHn=;HX>SI0j3c(xzi zK<$#Pr4oIU8=al-{qZCHV~1aVaHapjmCql(dSLv>m-FjoS+^wz$|#Dr`;RSAv64?E z<0(GrVghLvXLO^8bG|AfC&jAO1;s~kfwdsX4V^bAW>%E~DWnZzqApv@;EZmV<%%7s z4bzt}+up(~NPy^k6IYkC!quwkw4sgJfDCA+Z`gik(?ZdvSzf4s7NBe?V+!QranE_c zjO5Hh8*m<;y-aO3J(L_9Y6+>d%Q1{8GnF@LdcM8Jl=8er;XzTm1@pG~o0wA)tP0-j zAgzZ7VU_RE8a=DqGZ<2y@2EDUJa6Hg%U%+XbBP!yCOJOkV!3Y?q?ujL(wq>13K5S= zwfz%=3GV6&k-!}OOK?(6V8CJ&PAlR5b!!R1fJkw%cwo{@Mkr>G0tK;SGpwi`6dE&& zRitUtA8V#E=vnkff>o?#o(j8gtF7eO!=SBpHeH$sKt+mbrxDzPxiI{%JyMF3V=%D^ zT-+rx#9b_vo;5l##--#aByLipqo!_JG+V%y4hO=EwurNg&P2_cCx}NB55& z7`ci@PK|s$dSGt@3T%3;oz7`urP#M-z=CAV)#0H`l33eF@W-ep zD@+(mMae#{VPWB{4(NnCFeEO;S5H=cO0JH5ed@urYvUj6c`*3te5|_q_?6ek-uU#< zp*`cLuKJg?oI95`zScbOYeAlgkJV{gk z6lsmqKcys>s)^BzfT{qM#%a6^wVEo{V5o6apGk_xIoZ6>J+ebqG$}#MGT@TghObf zUXCN87V5z!dePS5u{d_??OR$pw)b^xhx^Sl;iMg_q)@btWvCq+*nc^y71^E~wOk=k z0km0}=TcG7#*{W|=q3p%d`1?zWE9a<_D~S!X3mfv(ru8CObe{WVogOL-%RvjW74Rp zigCx6&_-+&0+h|0kuJVDHREKJniG=*7fS{pXedXbyuYTJ7#iN_wj2%2S&xGf(6DJ>keRb^P*B%}C{Nd?q<7fZ; z=+(dZ#&7Jz8$a#->hVXPKmO>vOJLw#gg)gRk4Q>ftiFoo3uHuL!wjAsoC0IhPEJx#Ja^1-q zxSQyVaV5mLj2Ah+roy9VD-JC{ix^MpGK5VwUJQaLt7L^VLNX5TDn=Q*s8WS7ExJZyosGEXepND8_^vFaPH^_IcM+{sL?R9xO`E^dP5;_@})#}8Hh25J!lTejhl zljtiH?!#_gP8{*&mMhuh%1agMOCVT^ahEFBjg@vVx z*6M6~#4FATe6^e%aqR`?+IzBh_mh+ReB;|VuDt}y;pKz#AuV{b9W+DlHuEL4G-sFY z55A)L4qhb#WwCffZ~h$y_Q`9n`IOt)}UuN1%*e7&@-x?O8(d z+G*U(HY##LrQ zt0nBo!lU&_I0=+Bl}RIO&@^(6!(To;+5c$Z%ZIPO7NS8&jB#=*$_$@v)yc9Y$L{Jf zVDV9c3vP{Aazv2cxLZI0!$w^VpuoD+z z=o+!CWoSD!6&=l z+(3#nqZgS+q64e>KG5V$^I6lf(|xp1)D4Rf*gvc4J-c=-`IY<;lc3eDma+y^KxBI@ z+f4q2f1L*G;`~QAn-~3~P zf^qiq8y`IR^t$i<|M}#BCr7XRvir-QKYkyks$ULXK{1Z!IljKNP3Ac{75Q-Ive)X0 zVyhPP1`MV!|y z`q_asvw8E5je&**koF_!|3L!O2z3Snx=PlVIu2QVlnQpGj^r3t2`W z;$&^7VE~mzdH_2|b$;q6v?!k@g}rs6z>|3TIWaTrN;mr2(?jpVM<1+6H1E z(bBNUxeQUR`e~Ba0%{4C`JAdLyWxkMb>!;E$&tT}?i)FcMm`-mHF^MzoMGgwqx)xx zW@&G&iDqfXc7~!>1E-&o3V0(IN4^@lOeyW1l@aa;i+w%33)v;=foI ziyMb2EtTK|$yNTYB9nU*$&)N2I#okXy@*#S3{6wRA@@wOwCIEo#Ujroj%QWdf*%br zb|T4gp7~VdYn=4|63ndc;RT2#SpnTx6wERONJcNv`a)S60!kv{*&=R_JT|f+DL_e< z;hd^bfEFVd*kCCwWp^QSRnrC_B3A+NtQbslcPmsQt-+eSfOe-v3E(9G#g7q zb5+&wFm#0kE@4zJA`&Oo=#+~Qk%=Xp6sXxKp*PTS&V!7V2wZq^Q3Tx)8>#I`{dyKzyg!DcxomQ~}_?`w*BP+^&KGRt#qyX~1md60bc=*J6T;B|Ha5 zH*Gdp5`^AVHDzsd0=-p&Y)xT_BUH0s{`$WWTE%5Kl;Gm%+X2LL@>0{n)-!A;#wC}U zCDVH5p9dN?J>T9`g=Ew78f96u?H0@4Yx*sM)?`L1?*_Gl6pAc)uqx_s+ToUMGf|{A zt=`Zy9oMGSs&15N&nCc4@SMbR0w3de$yF>d?qa!*d!@NgvYc``S#1vT9A`N4`N)~k z1EU972;OI-`$k`dAvp3S3p^Y-&jJsxIVIuGj=sz`_l+JHIfa~h>bplS!&gT4kM1A& z!U<7iv3&5gtE`|pYDOcU!_~V-&cmG#jK18!Q9K>-hNU)-?r}m7PgTI4Jkqj8S3xXX zqv~n00#=Jw#WIL#c5H(OqUC7%)l@AYK9%Hu?0<_-rBc3!JL@{=%gF-u0LejCufFnb zq$=WL%Cq>$a@!*UcbJrK4XC#USg9RWM=T$rKrabgqNWPtB1I%~Yrs%$4QTnH%cW&# zEA0)h)Up?R(fUPeXIWu){CWQ@N!?BB zx7H|6lf<#BDPmaEio*X^&^SSqZ zBP0l{(0l@Ps3gbB?h*!-iW6Q7P5?>*2eNxK&w*4nbTZG1mVfsNt1&1oaE#hUy)&ZUuuTmzq9+!p5YsJ`iHOFd1L-5FvFh= zUmE`A&hFtW!-IDY4WA!AH~bB2Ni%GmcTT#whkSw(S#er*AV--v*%9#oxQ^sofch#!{^cV?W<<6!b-_% zE9_LKnoIH+ZBDD$w3bwCvH}l6fWQe7E2bj29qNAhV^S<(RjiNWVxcIrOp`5TS&CP9 zdTh$6#w%6hLA(}{XjL4KVJ}wb(1jU&7_GHpc&m|hsQR7`@tYBAQd~f-io!hT*ngBH zpX50q#qo&*ELEur9=>zyQ>-j+JUA!u!jJhP%*tRQY>9nj9ds)kwIx& z$!S62@f#(2r&tkKDHfQp*dRqr4UnQ)sKZk88D!}wLrs_!Xx4TDzsu~!OC8eGm0LFS zm2?{^SaZxmot@91X7n8ECA3j#!-kbB{U_tZf*V@oB@30q#~DTGJQE(bQs_-1gP!Bi z8iNr3>@l0Dl1H2DLfWAUnvxdfnWaT}E`(VQY^>TBK{MVZ`He8<@?i_`<$)JGPIp9=3tOD^Jtp-W2fHl{m-`G@sRkU`(nbMSpX_Wvz?fA9lT%+Uf` z!v6Qd3$2!g;l+bt+d(ZDrAyh#XZorRcXm3LBkTU)7x2Sym)BtR{D05gA3RD?Zs@NC z_ADPDC3k=DRb&}MXOtYu+#fv4{(L}D_t2N{+duXnecGM)9l870KGl(G-@kUoHMpl8 zZn<^%{XxD7Sw^7_ za~BIy>BSk0in@Wm{i0**<_#OZ9enEN<2fe-wx#HrVjHHud?hh)*U%ZA<^Li$?a1Y5 z`^MI-%UcZ#US3<({mV=%bvsS-Ecpq$;uL3&rk)!}877MAAjJkyojYZw|1~CZ2yLXM zBUNoBT`Dy#II>b_H0M+OwPw|()$4x0s=W`SNn&8tS))z38|^}Pk|G_Em*C4L2TS^_ zLK<4i^O78UW?l(vZO)KH+o;(?%pP1Gopo}gixs<%=+Uu|VhP8|;Dx^yko4?!pyFbE z8OC*LG<=Fn1+Q$p;0sED)K!jK&FET)rHfK5%Qcw@9DJ}SaMEZtFcC>ZAQHK_5V0gq z83$;VsCsvZx#206_nN55)Qg0V-%|L{ncWxq6B^#SjZdUv^$Q<-+*MZ~?pb9O;+!Bv z*|l?6T9={UIZ7egMhtkK`m2#u+gvMIPNGM;=M>AvjMW&(qu4GtdgGs*eK7dh_&edT z38kE>zCyKdPc5%rB~8&mM}clc3Pq-_>4?2}#OP+s925~`P-abKWFs^AoYr;J_?*_k zlvDZ@{7oJ#c~w&&oH1C16$<}~wuLSKOQOv2@g(!ey5b_5<1;(o<8wOSMV?E=B7Ho> z(Ylglys8s^gQZ#cfcK&sWThf8{M+1WSI5dE%X^+=MUj)mD7G17SF7YyL+YYg+h~MV z9!PE#B`#iLU5^IoZ#hUw+JYqrzCM|&S?916DjZ6JM+&7BD_AVl5W@ILuxGw61u57d z#0c937HFnKo2Zi0RsRO;^)!>w1=n&ZRaQ&Mv&m9Sa$=%7n{+YGi`W1WCj7dNDVD`4 z*KqL7DrZ77SVZauy^zL+lXN_7cyz!|b~_b%SizoFP!*u-^-1ElUN}DQq|L3+4`>6` zGC934LkeNj*4C!!5H-P@*-bTt!m>xvR@_ZA&6S;BGet^2RSwwc$ozio*3FySnpbc1J#e<-Qh^e;i088P^FF+) zEUz;j*aWhmrZb%h;UMZ27@tFzs)xg2onm!r<9rky%M^PMw=fKCnz7Y}1;_^RQ%Xrr zh>G$}8zm~$2p|(mdS{B|4aRDC0z5Aq4$)B?e9E_HQYQL8tCB9OZ`vN{G?qT&msY76m`O4^+T=tJg4VQJH;8HV&31yRM=!ikso&Px zO%=8Po7C*vN{YH+p;b(K>+GCg`qOOf^>3ks4w+U|xS5~1&L%%cq0JO4VjGW2m`yq7 zKq1te`8Gu7oGfx;jX0R|k)pNwyEn~VVvwTx-J2}HvjfCJtio{4>Ak%}$8l!pxQSGVW_Jdw1Ejp3J3&R z*7ncYQ)?%zOr?VNsWizYq`;AN3q3kW&*oU5os(uwyL6R?w(5y7mi#IUk#gZyF0f)3 zs1{BLBtt~t;{QK;@A@6Zajy&iE46h#Y&j97>Dyd%*885N3m`7S1j*pUn-4WTrI})S zx+m4$!lRIPakVjaVjMe(9dLY)!8SG+u({b@>j>}9Vy)wCSo;$J2^5oc{)P8fEs1H( zjFj%yKqINMVy0)NXEa?;SJm(KJU=;Ud*?$WhRCkKLB&#&ff1@h^jxQ8Pd4z@v0E?R zI(qYqq&ak$=Fnja`~_@jtSE^Ze!JB0&%gWNub=(%t?-DpO#xqEn!(v7(nergf8wxN zpxhPRWzm>r0#?ISsxxBrh-yI2@~#L8*xiHn*eEfjY&Ns6e%Kfg@tril?6PZ%Y>i-E z5_c387&u7O8F#2CSrhNf{Xy~ho?VO*ETPcZRnLhh#mzrM6 zZ0=U7uj#>;^qkP7N^19((q%h!Zt9Cyrd~W3J~Gyx+C6|tO5wD$oEBPA%IlqFIua698jfPU?F-cM3(qk8?Ib|v&yn^ zcI|NFBn4-i6sFd8JS04{YXpoLa7`j7{C3caHAV}tn2Xw;yw|IN>G0!`P_-aB%rP7N9L<)Kc=5_EB(exF zFohoU_ZR+v5bDf4^{--;+RBPmZUsY87)%8^>YBQP`eDPfgfuFr1_Ll%vHRH$b6h7v zm#1G^3qjAc1U<7P=vka?(xqC9aTw+>uj`iS#!`O_P{=8qX(Ds1T_}iKh@?H<9xP;> z)|RDY`G?7GC$HW(I(c#O+?-P2rIl9;yg>X==spT{$H^}zzq#@9jUx!&yLe;2_wF~7 zXD7dcYrQNa5-$q*PPkln_;&J|r~N2Y5``}0>yu~U6Yw^ge8-bg^mHUWVaSl`%e^VN zCo83LJ-Jj?m>}7nfow+ZQM6Rg22~#DS+Di3_gOV{envVNKSSIEDj1td=jpi#NrbFc zTuWmbXOh-l50UU}emyQMvN)Y>3_XvA$QbOHI6XRXnp0C>i7g1Xx+X!}NtAwSd#zn? z6Z({Wl#EYGG!r;-qzHZ3BzGptsz`W;w>3$+?k@%GdM0p?aY~CEHCpxDc3P%R7uw?T z@85Av!;LX&q^MZc8n=%%9yQ`rFF9xvMxA}6M!6ZDS=1Ogs(YWo%G(4mX`ISMW_qoW z88l#Bf5**}8X3QSE_`UHjk73&WjV4A@2U@y8*aG^^tEaSbLGJ@q!l`WIeGn>u@rLl%zXT`!1>v)MjU5*4$Dp-B*}d z5NqNl1V2TfWz+w)!o1-jT0}!&K1R^GC6pV!%}!4FB~(&@gLG;dt4&e!q!KP7o)7+f z{HNDX{^RWapFX;P{&D`p2L~_w^zP9I$3BFuIG0s^I&wI?pU}=4Il&Tb^5oWH=!0;R z6ypg1!mcMLtiCx63KWbSwAZo~bxxWi5>}O=+eziO9sq%rr#H)e-n1aTZ@{qDN>V z3cDmXb%#9l*O%|TadGO{iK%0!?tk;Qdsh#~=CVqH_}FYO2hMy@zLzqG^7fH79*yM{@{L2ah;>VAF{NsPsa{ZYNH zPgs`WKPkN+DjinmLI9G+n$+m!ham>vI&$mPo9Azyy7ihzo&98%AbjNJxtm|Lv+Cy& z(EuAr!GiVs$*mXb6yWoM1T0v>p{L6ZQeoIGG1DY_nQvQ3)u)Yj9B|%bvD!q4S&Pe9 z-Iv@$4Z9pqG)5AvK1EC17C#;(i*r_p#j%$JvrkFdpDn<%qKA)H zea(+vq7~76n+pkU-5`-C&azu?3#`&u)e?4)fA;7WuT}Dv9oOs;jd>bpPhf|BlQ|0G z^1NvH``4|+ER;EAf;Y+v)I0q9{*y>B)=0%M;Qpf=iSKz#tjcy7Yul(x%`y`+$wZx7 zC$5<^@A4vCjCY~o6PVlWUt)dGsv{@+Nag`K=8|ao*evCjl+VphFQa<>oJn{W8WL5G z1#E;Rxr47Sli8>BLM98;QUB@9NwIi^AdJXz~p8L9mho<~Xf*RM{b~6kr!-D7D zR{GCl*<0rM+`SwwMX3Gsf+@{(Vn!%{f-kUz`kPO_WZfc>nkAvoaIPsUvo_Z%vivnl zsTXi6!J1qO2r@X^R2d(b5;ug_{R*u^Bb=GcnK-?UI_pNPJSkc$Y=zqDiDH43ED9Up zn0vZdHtHQ6B}_b-IPGu(=F2G_LF=H;Di&bAJ!4Sjn=tkH64(xtb683xS6n z>blFc&J-Cd(S>zayHjI}KK6i6d_^0QXfl~?$~!0Jfxhy<^b=JcC<(S|CKGv1t(5?2 zmAHC4<>(^CN0oF7ErbNBCaB6_4lYNsqAgfuv7D563O|W(w}PrQPAXB8r0v+4)D%u- zlPc(WJivwFj6sA`a!Ey5pttcW&U*>#yo2&|H1DwrZMO{_|NX_6> zCNg6R;M>UZ@Z%i9kxh&-7z!a!%Z0Yn%G0wNBH-Ee!??k~hn`jN%f^{e@CC5tmH}*e zNjP#?PA7eo7L6fPEVz2^+NePG(l*%#kt}E}g>$(?@9JN8JBDQzz2VSBSD(mLR&4=H z2EZZ<#jP2@M4R1saZ+2ff=E&lez<1dKQ(pi#J#t!&78P8b@kmC(-4~I9Uj?wNEDWq ze=q#PrFp8Oei-C>r{h6OFVosw-yOL0Dlxs*M0plIh}$f|6s6!x6Q%i$l$Ip&UXQ%z zo-AU%EZCAtCMo-xcer)JIed-ag2(7_%c57LrX*F!K3EE_Ub3Q>B~1l!`eQPQdw{`)g*84+1Vv)ES}P2 z*=whlURHh4>e-+iVv1&|;){xc2^T@-r+s^s43>k^2+ZOI6ul-XQC4Xzp;%vKwpCcP zbmE!erqYRLOa?y=4LfAE*XE-ypU9?^CA+Y2r{r?fX;WfbCX#Y3*cBc9O*Inz#9*Tq zSmgfV2vx?CoWofS!-lhr>Ihj-`{Z%^-SbYx? zXS=3|-d8#7rt?)WU1*FIsMUq6Dm4-OREP2|M9c!R0XUQwKZXnLrd_2vBL*7y<(4O3 zFMXG9*yeDr=W%wzK#B5^s#~d|&I+@=oUuva7lvorY_>`!+6m)D%{=8r-z=5t9VEQt zfH#7MGfnd}smqx4!eC&*I~Mw=sI%nOZeM*dLIHy2}cMo zwacK0vV$pUaMBv%QhUf^PTDq7v}$CXU0Nv03P8@^pI832$n@I0y=i)FCemnYjM*|Lf5rx#-?TC`fF;Aj%Wt-Nc4lNq%u@G_HBs?*e+F+IHTrz&%e5)sH)xa5B zEsM2O!bpp8E{~L~3b9ciCuRZe{LWEkR8T=^V6wdHSUkL&)kHy){Qa?mw#QZkrAd_n z7h5#m8@mWMXK@;y&AIpVTjuOV!Oyn4B`Jet6|A8Y*7o_I0FLX>|BxQhoSsK**;D!% z7;FTqaJdfU! zmk%9b<5UD6N{$};Tm2Y;S#$lIAte5PX^j_&6DL$@@Trr^KuZ@j-o*&@nJsGpqd}3y zN;c7tU1DVu4#Ijq&P<(zFII2goRMG?R0H>CT&qBBCx+W5kbIUxU^_vQu^haK=IV`t z)bmd?zMiH4hl0~hSA`E=wB`eZ@1CJ} z6sYByW2IkFQx9nNSg(pX3TNXJ5` zzyyTh)Ex#$cRa)_BTQ^8V%}0NSBEZiZ$JP}2qY#_9S`lfQPHq5J{@@gB#X5~J7t04 zpc9itm9tqU8cM1~<#yl@ae1?Z zz5nRQ(*WQ=AHOo;$omck-+@rgauMMaRY(CZ{NI0cwAz?9ZEMnma-~VR4i#e0=ufz3nG-jQ%r#= zmQ;gv^nt?=fvMNBcvY>NaB5&2G@NNtz+Ia-7<)zF&J^SS3l-9*n8S}if&T_lAfb?j~q%OT$fr-{z6h26W`SJd-} z>On>pl!Az7J;U%edfF&}b1@cvgBqfyC2PyX&g+!bIE|J;W7!3exqzYJHSGPx@UEwK zY}vYLcm#DyXgvcQNbyl2)wT(?q^yQG+CEFsL1HG3clUYWZut7}In*J_jXr;nd!3SM zEj0&Hj^7vCM?HK@H z3&H(q5@y>XD24zWn4hLiS<#$0O;P8w&r)QwiBq~EvV-gX-vyuF_V9DmL>0o2luKt; zTn=%BjgAtthDOL3^)$=FhZHMbtbTr`hVEym2ntm$dQIypGA&DRQ%9ghwJYeM3i692 zHwHb{CK66H&iIomC@fM_yO=c!MVT0JBBA0`u5oz^qoOu*=+xBF7w-M;%X`23{r&w% z;!ZmS<$2XH@IsY?rBZ=%F4~WD-~Zi9_fCD6l0u{6`w|~2epTe4vdF6F&LcM|JtUfu z{>d}_&iSb`Z%_a2^Uy%GOEyv!9ed=aRAJ%-&pE+hbx~6=+OMDECMn0}3#(phpQeOI zg3}yMwRYJ8;S0W1!6iy-kqakx&@mzQ1~^nLdp1r1W@bUSwH2|jHF1^gZb=p$1hJIH z@;=`&I8u&<;M-;!h1QS^zf7G~>66nK%`)TrHf-OwgH|o>pbcWE(RRyb4jZHY3qC&r z7f`ml=1<+WabKPIyL&fAowA~8ijsQ@b;65}De}=SXKEKEUaUvdnCS+qkO&>(Tk)XR zsai{~O4XLUYZRPH(kDF~zhb6dre-{CHB`Z(k)sQQ^zN}-2XQygJspt$=qP~0?uz-> zviju-E?(M7ye4)!MWo1?ZW-a{#;*u3SmRlLv5QZb)&6=E2WXZQ}q+18$(o00+RKe1aGLeOHSGgTn$umx>PGr5@cElrO1uI$`t6Nq< zT0;;sdkv#VD^xIk&=%NeV9PHN*G9Wogc_E{M8~?;2H2AJm*8XMl3=Qc;DNWn3#1fh zJXOwMEqZ8s)h}zVO%SP&k}GmraT)@m6HHUj?ZB%W7y<)CP_+@Kw#fa7@JfNAA@ZBw zOuX~k-&+z7K6VZ9ut+@ZeS(n_jchVC5)17DG_ zno4YsB`2v_6pH=C1=~H&d8#N_i|#|MC8%thRboX-!|A40$7|CLDR2T~V30N`{6;cF z6!3I2k!#LigWExMI5ujCIB3V{fN1atKjKM`lrE_pxt$1R-g?8Z_gdU3b?>rxxf+Gv z*6>AjsN_t2>UhTi!OG!mW?6)s)gggH4F+ZbLDiD5ETSl_w5uJiDw@V<(dp7_V&};d zo7tAZ3UTf`Dp&|=!yxo1fhL z{N@+8j`XZIs-^CmAKy9*uUv%37ykQ|TL+MM0ZzMo^J6&a@Xgcv{Ab)c*5VA*DW}^h z%?kiFQAM&fRG}=9@JZ{W9SeM1Jz_)mwqyg< zEY_k&*H;lDj9YnD7>7YS64zPeB}0(Ggag) zm>Baf^^8~(@0v)zedPnQBpyC?(e>B<^X%=bZ^Qj&-@fvNgZd^ugZID@|A65=)$odW zzi_EpVh$(10C$~_7Eo+j^=K|yV`gJl5avV=sx23c43?s+39KS8 z7&lnC3-xkRHm;wG35+bx z5MluCs3BfPqZ8LS8ls{%Uiev5|J=i*PR5ti$ppntk_2U;0lShL9J^k#Xux8y0gf4i>>1cOvIT9~)H^WodiTU%_8^^=%QQR<$@ewBkY#;h1+C52L`xR6 zNxLgy_4yaTsW*<_yYTM4o-O|TlWP3?`^&_WXEFnqf%%)td zsne>?CeE8^#C2R!4Zn~!^~wxi-l_=_MOJQfp2}9hJK(1aFCwA8|I=58fBNXno`3%F zxEibKdx`Sf&#_Rl7`sQ_wQ92^t-J zB2836V@y1vAuqr;aR6Q!o-3<~Cq}LbT87-T9Bo3Cf;uoB_bMHB9zd`P)R}w5i$SArK3N6aqYpE z@5UNEP_jvrTF>xb zHaW?PMZOod3Yi0I6wUrNM(tS*K2l{cSM1pou<+p>OUSxD*-zNlIZ)W*V!@ zVQ{7eCo0#ZZ5IZBnU9Z5fBf>y{&xX`sTYn+eSZ1=ci&7Md-sW7#u^&^oP2v~|4VQ& zsB`-8iF@ySbnoOV;cM=u-}!Fp_wP(!Iez!jOEZUl@0%Qoul7p`Sp1dtTFNw)L%tM8DMMo*N?FKQO0*7B%CrZ_Y;KgAu^k9xAZ5bC>XtoF%Gn2;XSEdC)DwCK7 zOPdCQC}=E;INcVcRGiI3&h4%V0Xze)HCBkySZiuKlF-;5{S7c=@uEoH!VYcF+VZn- z5o|{ZtWUWye9eQ>fG1B(Ub^w(L{+&DCO z6&^26e%Xa4k58VPedWgf$+I_(PF|Y4fF@s?Jq0clP7)U!-qg=(SW&?a#MzWD%`UeF zOkNtXEY2l)(*VUX{kN~4A?W(?i3>4S%cKF#7D~NUBe-|=be>x#LE9~T+^NBXVV4=* zBNVR-us)hpR2tVDi`VpVlddo;f01k3i#8U>_lNDW8ydLH#B}ZPGTGy+XYqxff=8cZ zsfAa)Cc(aJgHppMzVP*3Ur;=I)l+0{U+b#*20?t}ez|hghs?3Cw2pY4Das`H>4PsVGy(mlxs41Ej+{0`x(G>ON z{|Jq2>o0dB%HUjr#R4qi6c$Sh4x-isNaA-r`a58|W$htokQNJVD1BU)ZMjipmqK2& zcv2&K+?$20R1Ni8hUIb_^={v^!mPd_QbU8xFofq@rr~ppvjZ@9{eCmg*KaEqu-{f% zFc+n!R+34C${(SsSe%w95vAsF2C9^VC(;m%H^~^W+g1Kpk48v=<3~ywD~Won-{jUE zYFYpfX+!mJB%5214x_gQT0M}|XGsNCu0pK&DkX>0sYZ6Z8I;`nbAy`Y_JwvH3o)sW z!*y0BJ_1uaypgpH_L;KfOWoFs>*T=2b)rSq^16(yj<#inVd=z(Kq_h3tt`^*uPRuM z-a27Tm{{;;iJ8kO3~t9drD|A;ULCw9^6J*}HLGC7o3T~XSZneh6Rn7-S`nA`|Kr?+ z2Y)$>I;C7L_0y5V;j6~lry7wK?XfaVjepD#8KV)47U>xLQ$VuPoyzv00ZI)dd0kRR zi*zsL<@WKMA63j?Ny+UIit^O5xEN9s7SAFx$-mp({ws%JO=V>zG?W93mv^rB&Cu|Y zZ)w49%|&mfwKjGL;ADiBp`egf0OM7=SUpuH-T_ViPNTl7KQhZyCNoj_U7R z8nAU~TF@uurak~`ad~1^wN!HiKdd5hTCNc}nzPK{nwG<-?VhF0ZmA?kdhmjaek-1h+I`gD7$`x>!G=j!h#S4Flh?$svZQxo&q6 z-RxI{82gQkz_nOlEg6iOh42t)SW0KHB!e%b_{xhTbRGgE(m1Onk)^gqmS7lSj3;G8 zn>IC_Bn0bDZi(@DJlu?6gJz{fli8P?rQC9u9%FVkBYNRw)W{QLO=LZUYA=jMe7Pw= zNx_-u#FsS@Qnf|x32}M!0A5>w(r`MPtZ}}m1$Ymx?X)3g5e<(FR<*`_(u&jT9jGw) z`TJ%cR=8?!pT+kOw1ILH7NG(i4}D)WN%q#mivfd+a9iYXI(o0JwXy3Fotccpx|+&h zoa7>Jdus(I8l%WT#v`67N%I^3jz^ifm+*oU5A!S|m}lvx-n`N2oA-YA`R4!UK=r}do?w7pz(Eu+V|Oo|nf~GrcdxvB z_v%-nA)CAMg(aG5-Evn08}6zEhI=DqfvEzV7K8)_?E-hbDuO759#pVktsSI*oNbw& zlNnge-(>3*|5@>+-~g>)Oe<Hx`L+EE(WXRjecn=#r>S zojZ5;(#e@G58OZW&Gh-R@bvyy=kFaqboaY+cP}5GIdlpJHFf{uiK#Om#JQ}daV8bH z!{sJ(6vpiN9`yJ9f87c*t;{L2VYvk8yu-imKZ(SgBo)VCdWmu*z6Y=g~p!}H~~+rPy6ppW9cPHx?FQeL9#WJXZsRJAFjXb0`pVL3fsdeJ4ESv?y#wRH<# z>|Sq*-O%0#R;EM1S3m_=MlBoB+1Fvmt07(JvcqfeT|%OKMa@FdH57jOPRa|^MxE4l zh|VU?o8fVMIA*>bFKjQ!OplhrN)mq*CP>Y2xoaa!cMamWae4!qyc}7iHIa85g;m^z zbs^3Gn5MQUnNQs7IrqBia)L~^z;w%wHH)vMvD#Gk8Q7lRMGbgDkz=8Up{{=ul{LGz z_l)e?H?nK@?)C+$7Mkz#SS+{Z{b3DI9iMtx(9;=aM+hTtCWKpN&X<;yR9^yY_DMi| zGB}&)SPACh_&BpFHD=`*%p6waF;{(RIV>ejiG9WE1=rByX&#vsbUG!soF?zHJe_K( zCZs5;1e_p;n)XUlE#E>puc2X9@lx;MaFEj}t*1AYO2P63gQ}%_(*2pfp56^gt|zx4 z)zg<+-@lwP3jj&R8F^V0)m1{GcPn{oqZy5=Et@%;*BE z8kVTKiZPR%70frKY1$;gy9L>g6W>mK{?*i%r|*6GR&3F^tmt=`RuXnyi(X6h6T62m|nIHeV0TP6?R*-1*Ly~dzYHiL7%b*&DJ6c`xM!7#dKv)Tg$oS>pX zT-_4JXq?-RB;j0>X}ty)HE=)~hY_G%Xu!Juj+-YnGJgGBe5w>#k9e4ERHd@Vf_pvN zCO!hM2aVD87A!w5suHu&M$U>B-$Tbj!!SfkuqgVPfrm~HWCfv6YAW(eOfx@BcQ3y^ z^X>1Zez$+>?9qE)|7Ggzt5aWJ$mE9LI7UlNO5=2r*YV|b;G=r1d7Z3agJv^{rkKxq zuH+3GK_0t(wzCAbbCb|svc4^FJ&zrmIr4{lfBb6t%quZ2XmScGsm7fsu$SWv30jAS ztT6!?tOOKx)fyUNLUI!h0H~*RN=!yXs^_`qdiv*1>3MhxGv`m)y<3#2q1kfj4Fmh& zKX{2a3+kcgzJO6!dn}F$%+_Z~*WFvKVz;odWl-;`@Qx!_&yJCJoe1X2E zA%aotb!|cCiz0eoE70XN)y;aXBNI}>oGo7@W*6FG=>*`2aodv>ETtRcuz^duyh@?o zvZyg+Hlv)PVZF4M+79Ja0a4j8m|0#?;PjeB3cKGYS>0UZV{IgAM6ws<43?CHQnPqS&vx?$tHemA z0gOaX-F%)w+(Zs%U4(TAd*PFhihnKLO!P9^Lb>Ge%HP>MXf6(PJV!ldSd97p3Aw z0joOQxF5CE_1b1}r-XVdV0-IJI$#SYOLpa{Qj6KVO43Y^axYm-cM}D zps3$+xl<#&uoCp?7Ea|2#B_*VGr)ui4S}3a$vuh~%08*DC#$Sa_4MVkT93RwH_#(( z&<1+cY&PxNN23%}t1?zokyQkf+5OmOcL`j0tHsF1Gs6{}%`B_K;40a%=jlW`OmtR= zr5i?8w42tNl;Q&yK6q2Dcm43bzrS#N_wJb&e>e5{U#5<~Hg)vubI;wqbl~olD^mww zYa=cZh>;Nrxh@@;e(Q_r_YZ_dUcWTK0Hcb&+PEoS;EgEQxwF%j-BD4wp@4p|*z? zrA%60-jb$aNliGJ9Xm=j%PG+cvrz?>uWFp)Kv_?9rm0Y9(YeyG-8E;W80K*qf(Ma@ zFL?2}Q3lo(Dv%n|#rWvrE!=9-bZpwz(EyJpIVS#G{3Ojm$}QCCnKGlysI-VwcMJ`7 zEW`)CqC_~cOODrPW)3K@^k6B%l9a=e=8J3jr0nG2>skh@sm9$VF&m%UxE73PpJlMI zn2O9bxcQo_lhL3n?vOL4w3@jaHZ{%M4a}U22FE{lM_le0aFhLg|LdsYI>dAwxLl*k z%9d$H=%SLFQ47dX%=N%Jprj*9<0R8J&am`?0gGWW=Y(dD)MV5q z46p@>sdbVzyng!2mu9ZMIdl2!)U{WqkG?tmr$0}hyb>C+b`in=I0dWHvV6qVA+}Vh z+Lm2q;}ND$E2*Uj)-GTne2~B`-`3R76IhF>syCp9M7JKfO5h?bk4QQ0LUEXYU>OAbg!u8)s1tYw09D>5sVx zIs|7i7T)HB-cnjkgnUFzVBz|S0ys1gvMjCP$kgG;1cpP#a?-QUSMZTHXthMR1uJnZ zDzX5Hie)LGl=>iHU^<)8L&eL*L&I7sNk=Z0j-+S@8xu{Fb}?JGOq+!l{cUZd)w}5_ z0lVqtJVm8k<8oFCDr3<7GxX!fKfd+j=^y|2<0*K28~ylac=#JUeHR|Z3oA{Z3_vNN zUs^8OB~%#(;h)gx{P;dx|Lf?-<1iSn`{t77D?Zl(ThUSm=b9=E2@itJihSXuiQF-#0zlR<(nVFNr!Kq-seB# z*0C06piVjc)C!w{H_wYpxpnl`3kV*>^_@eV@Ob&=DYzghm&=8>!GBd=(6PqX>`M(; zDVGzXO-giOgw;*LzkTHc6AevVj?oO4Q#jXDS{kO5eBt2mYhd}@RWp11!RlPt*t(-(a?v*zKz? z#@E3jCls18NplhZ2E&PM!$F)IjELk%&bq7~OkjgCBl}d#ofjFLi@tQNA`ED)MrsD> zLc@e#Kc|y2sm1B$Dmkn*^{g_D;KAFCoSODtwr8A)nj~tXpLLJ0Jq5!W2iLMlD}

    lBUl@qEtayGKm z?LrlYeol#C=ZVXnEg$>$-~7gQ^S0CD?p^1aw&il27Np<)7X5$lzkhEVh>u9zVac}F zP5hydt(83SSNIV+6Fe>3w;E=0bs)XtVDnB@a7s<~B%hs7tGGtnIDC=p1XYC_B(f$I zuZ$s`RPB0rZs_UZ$Aw0_a{RqI(HTJmJWpm4$J6(=s~F{7US^IHLG59E6{ZWVSoX3) zYpRCR*~n(~8t{CEso8~gPP}iT>&I_jeLcXdR> z<)85O9FDprAwGTnQ@Mv|1((sC-7hW7x^ z;p)U`Sd5IGehW}xdDlCuTk=GZMPCLiWwd<@pNgdYbI;ZH{f3hZyLUf@)}b8}ryY_v zm~GRFr%OP&NREhLs14SvO98BvwQdiy$|&VlHG(d|M#~tbHvA=@P2)3Z0x!{83TJak z%+T|);cAg#nMJtpwPHwrs$xA|%#&V`{dkfsWtevrr$3ycUB3yTE zp=raYUW(!vkJAe5-{tzS%aT$TN>VwqGAIfd!P6fYv5s1DXz3uwbS zNy>z`f0@M#U>Wpr*Yw%ITF$3Uk_(QTf|aHiQWB`}3}aZyL~U}X!){+aWA#O7&3zb1+ci9cvgXp;%(Mm3TxV^#VbVLL(%wNcKygWG2sZo zN?ss)VPuh2DD#@mXr%=%9ox5dG~m*)-Fjkln066^RZyqxz~vECwhX$*+DpPun`{ne zQdkm+5WfObuHH^o90r6;G-FAUx%K45D9oNbGvaGiDnQBMbW<;uL{q|#*ZtQ%m^plK z`s}NrVQQ0<1i&PLVW%a6EFw&nH%QfHHmX{pggc(yN0M<`YqYZi&l<3^r!Rav_3GK_ zGp|ga_;UK_@q3rPU3r~oW)7dZ``xjrYi~wzG{L{fWql~rK?sb2f^%}C9k!+$#)AXN z9J|6ykvrwgt$+u|5jd7YalhzLQ?za(@E97frgrj~mNBx-l6YxYN%|oySi(xwfr}(9Y&pVSO4JfyKq%8W?cowU~y|n<@aJ?1x}D5BsxqpvEfSZS){T7#3Q*RC4W6lQpy6PpQ%i(9_R8~7cep0%-Aih-^L;4%@CzzF9VMogsY8&-`D(WV zLy*Br^jg%_Gb_Z_Lit{r&ohFkD0d=lh~h+>6C=XHdWDymFe z8~2Ko!VUW{tud$9HxY7by|`)iNx-33z&A|%5#bvO#$y*6u&%%3=HYPs`njlt?d2{| z8dj4|Y91!6+=iXIt`?^RlZz7=WN=DpoELk@VtvyF8Mo}xtnBy#v-ZxxGj|SsaOdFV zJBN<8Hf}pMZ*R);ZzfewCNKUp|DA)E?i~8)&Y{zH4t{;-(4jjAzqoVg>YYQc-8pm( zjyG-E)C^CDqIN8VXTk8yp`>N8lvS_n*jGu3vtw0Vk0|RflRO_K z>)ej=W`T%;m8OXkT;-3B!6ncFf-@ak#2W=`jrvJzcDI~P$?%KPJ$?O>+%wRZOZ8~| z{i&YZKwp24oKxWGhSd6f!A@)Bw*~xiaju}gITkXAT_?Q3FW%);n4_1dVLYCjsBMMX zbO))SA-zm(+Y0aIK3yewCI}~R2AM6~dYGW|RsJr>(B-5hZzUC{v&p`=z}nwf>?+hp$M^14-1?W z-C<)CiJ$N3W#tWGmTBR)zkO7)SnY#NI16>y#Iv4lWQ>_MwIo>y4c!2Je!${&?81uq z2;?-JO*NWjYrEc;jaw!`{Z`F#A}B^Q7WGEKBm>?{?@VS{6|a3<$*si9^3k$5 z(_}n{k&t*g&&92_-O|UM8axa!B!dW%(M}<{={hTP< z-RANV37;!(Lrzxm`14H?Ie}*i{0jy_q26z07cAVSC<-0kOADUwwvy(~22S@7w1FAn z<7Y#AY<4u={X4h>7va(cVLlqk5mC-??i3?gF|E7%AyVjO153d2LJQX+=F zOGM1kkob)xfdEl>4eBV{OB$z}I%q74QpX@Otj7F#zd>vp4G~+n z97N0l*rJBT7ujz*o*8aRr9WdbkqH@g$gBnX=u_$$(Lr)?u*6lOK~^jY|G;V`jMp(( z&#KQ44cSXPk<&S-GBTFuB?PmoZ!*<$=Rs?L0HzGiCCz9283LVDs+PU3YH`Pkr6x`i zw0T8sWL9xEASjdHPM(|m2LAnW^6KP8IJz`>5lvp0JPU{4+<5uMkrlSWO`a4#{l<%v z7jGQ6aS;C9k0w8dYdJf4Sv;M*5I$FH@x9F3f_p``F<+idk<^xcD!jfFU{VTann(#@ zwxkiOiewGV?m&yKaZeUXErK}1_82wgZebnR;BONLZ9*k7Mv+Y;1gU4g&9@8Xd}WS7 z=aNuaU7m#7s#)M>bwSF!W?0cA+}af~08~_*Qj^lthHVFCKG92&V>!#kM8it-aNFv^ z#Fqmlrd1&^_SX`qWU!(n%$|j)6sZx}HfswgS)5ZFV`pm+ls>|lEp(;n7}Mcf+OTv* z!b6sDDwS*}5Ak4?JQou7u@K!QL8z88k>ObFWgEIzUBkT^!3fPu+P8|y3f2-boX1c! zm^DVaP(Oo#$cv|O!UA?!^hT$vKJ973D8dpqMxSSfL8429wgsv*mYYC?Ryok0jXOY9f&|Ysj~Q z{#sX3Rst$10|qkArXv@Tt_?Y)tB3C&J2!RW%)L*;bNAY%g(L|Ae@0GXL*8uBOD1(I zmdNGstVtP9OP8~)GtqVg7& zi;C#Im!Q%-ZK~V?8$bjMRPpCt#jE)7y)lsTPPM z&=cZaBk)TeQ>(Jbn6*f=KD?}MxT0=x5upL|kIeEcwj53;p;ytKgICG(s7N@eja}jU?d9T z1SN za&S7jZlrhPI+%_t3<3B7Mapd3wUI%Yr4v{hL*vlT$#_l5W>Y?Qwm#*Q;8RY+*-T@M zmIM`;>;4(K`SGn6ZeF^14&D6V*0Gyk-TZhZ6lX021U%+qwR9ws9o1fk~cNgW6-Ze{IXk_C3 zT8!SQmcl71vToK|N&lIw$d}Q{yd6fiCgDsrQ9kg}=aj<`bfIUyefC6#W0adzTypsKF^#0|I^!fT;Y3q$wh5v&s5i%n&P;+F zTB$HoFGnZ@YM0`iBpS?Tl7Of$MQ&1NaLV*jyXvzd<;b-R&Mj}MtPWUaHyE9Ep@|m> zhb^X(kF)4U%i)}qG&Nlcv3u{FzkmAO`@erFw&bWRDD>&3Y{FWcJ-#dtax9xHCpf*V zsUr&}y{0tAV%D~OMtPpFf-S~_!M~KF5n8PRV#_wp>TJ1yWI0(->$P;E*}PQYwrPde z25#5hBMZ8As!8B2>8rVnzM7w1`sdr9JwbgnWzoq#*R)8k!Hw&oHEva_mdzZmM__!U zRavmSla5Xu3rMSQY1|qk@J;RLJ{M_8lw>R=g39w=QWv^@)}(pM2(K&FzFkUEa8?bj z{jw9bRxd2o(U9m_|2&v>CNbvn*{u^k{k=^IpMGKzuMj!@gwF;}%re@@T-)I-)~=4e z{*Hx^bo6mp>g0RECP8DwL7+|m_#kry6aHgq4bCMIC&xuwTK~+)i9;bCIOIXz!g?7xK zuIbOgf-JXB+x4+9rUpsTvA8+PCoYY#vJrvexD5=IvQH^gtH*4#z(E^X#<}#PESnx7 zcG;_)HUvWqrp!9Di!kqK)Z+78xTi-)nV{ee_stUQ<{5c#f9mD#`n&tQ7v?(uh$cl( z_-y-^_19u7LL=Y0{#veA8Oi|(uCnNeo^Ga6F4nHF5&s?30V5mPm5Gl=>7G_xfsF5;jLHcy{H3qamtdCpS&KB~B;iwAwzbpGnp z(Lc{#lQ?+qRA~G}hfht$vILK+uS`*s>o&S~91z4T&Sn!4zxhJNiTBMe)OY*J7y2xv zA)dgSBH>h`Oy&cn4ISme^R;Ec0+teiiv?ln5;cou%Z)E6t*BUS+I7|-CUb^B7aDeL zmm6+;#i4*n8mHvQ$?a8yiN7Gk)~%`=8$(_eTx?lw%&DD%m?8^GL|V??$vxVCF7eanZR>J zm9Q))0)C4`tC?9~R@G)UVsObka;waC#HzIb+==qqn;LH#qloh|%sikqciNurG&U3bN;rcmlTt8RtLK`QpRVJ=|{|*dNoIy4% zi?dpzeGs_4`L?!=|09Yv@A6&7910I#>VdndYK_l3_O5Z1VIJU!Sctd({?)vM9_zE?G)@dk$bFD^iA-6`hAX=@$5XY!2p$HzfnoA^{VHGZ09f>oS*^t^){P0Z^9Z^2K$;JF+hLc|GLQa zghpj6DTBjm`;wudN%D#y$!Bwk{>PGuxi{aLy72qyw_lxk@s-dZwMzm-@tT`2j~cp4 zIBsJ?xPaJ9j}R?rm&}wR+L~t*{U)(yfMPi|fm(z*Zr%B=&Q?kt>k`R2!P(&3w@_xaDbb*#l1 zs8dcqwZeeCd0t$~tz$RO33c6DN0B&yarrnj#=j~r;Du!C89H$o5M2eQlX|!L`rPfM zi7O(c@FnV>xByorR9@qVoGPOCx!Tk*A+aA8*bgs%J9Xl(=>BJKPksK;4(Ke*F4rg08>zpYNII zXWgqivUmH+2XGnB-@fu)6_svZ5m(m73?pw5UO>8g`|4|O-PaFAO#(LVhjPX@H@znT zF->E+iN+duc21wXGWFJ}yH~FCpy_vCy#M*x)ZI&;-23>8so%dmee93-K05V;W42^4 zaOT|uGjG2?b>_{ftMA`G_51su{;B)^sW)d{IovHqu{)LRPK6gV-^?nN+QPZGPHWw4 zL0h8Pf+Bjd%bCa>C7ZS7b`kElX3-X;SBicyyzA*5TefZ*9zmUwhSoCxYvRKqeOv*< zM}w$1+CD3thx%NzK;E%p5Qt}xv{+7LW-dM=a$DXT1EaF zcSo5HUlO5srF|}(EMuh!k;Hac&>Qm1t1~A)nZEXEjE0GS;8> zZ(seWO>U@G?^Ej?2gFLj*<7O|8YwAv^0S%uE>6Gs#oenPO?|un{^zIee}3iOUk*>5 zdGp?hH=l5jg!U+Mh*7R1_3))0Nt6{xIESTFrW-D2`n~W97_+%JvQX1J?W@ZVjF*PB zrb@5G;9jU*a(U+Di}%lc5~H=PW?-*M%JrArt8~t6p^z2lPL;)K(w@`r6uG0L+yx8O zijl{sc%8Zv&uBgnH+1{zWv_XawZXoOMTsS+*T5(bu5;3e(R0u3+TPPYxW0eOu1#Au?cR;xuCq9TT?`{MTr(Z=ym$hW7@+|R zP^(v&ZCt-?!=`7~5AQ2dQv{!8&x5aEqewI^vGNA3z*J_o1D{^EIPRzuyc{jnzM8wo zfj{7&{(b-Jl7;|7d#|_^wsTs6S&Oz(Qgp*p+xs?3xrTzt*2Q(&()U9eXS7BhTi}K* zun6~Z$7_4ig(|Sja%2@znK-m)adB~bXc3hN7u_2R@G5sub>buB|9))oV%+;_fOZpo z+%5ZCfXqJc*uOOSGO6p#LLuy4BxA;5)wI`LrN>lOp+Qj5x zu7nr7#;x(BOH?6Apec!Pu@Ek%T+rHCg+-|$ZkHVbh{!l6Mdm;Gg1Ps<_!Kv>) zo!S4+)X{fm-um$WUOPH#rwkALkEhjr@PM9Hcklkc*Is%|n@nV|`e!%JS{<s`vGu$0SSC9`kg)7`s8 z2T}M=8!lOVfh?zlR1#KlUgJ45YYO&l*uHNE^@t7|#PFl-mdzYCM*o*+Z3@>>w!7v} z-L`RGzu{6?W6-vZyLV%l*ovwtD$MOpc=0g>p4v5~H;Z@B4mUy*tokVUqD)Gr??w8FfeZ$Stw@FCUZX;uIB{iF=uxDTK$1U@#7Pk>6w@iitNo_- zcs*$oX4FF6>jLFgkvLZPcH8aQ1T#1zMdl5{O172esgV0X14fhm`kl9g}H{@;50>RWmj+Um897WZ6C z;U!tZS~hsAjA7VAiJ0+u;#DXtSO&M3YBtmDZoNb+qSOGb@WHeG*HTc&<+cSbF4`mMRND8tBt>muV^u8xo_*R474I3Pwc&Sn$O zxo*x00a+5gD?>Z%I9YTtOH6EkSe8>Wu9iQlkmJZUaX$hFGQ2+rSUk+oD7>()!eCHL z@7k#Ckh}q&n>gPVM62~Kx2kma?uRh$+TJtRzi-fT;5!~!L)!LPa!X{V{^hmznguIL zf!YYG_@-UDE=QjZC`YdfiI3*`_JoG8K&|NVk*P0ZCH<;G|09=lck=C6doAz*$Mtbi zj&;{kQZ0IKu0dR8ZSXDZ_SFNT2WQEec#om}(!~32JXW|OW2wnE3cP!{CbBJ;>#&6l ziMl*S3^f0DpY(mGA!Su^^CeOi5&RM{!h&@GI3qX46zmXP zxcjW{-!u1+>%S{o>q>l#IChyK+bvu9xXUBMgR+A-$b~DiO@|-x>ZJnaU81b7YGw9Gz$v2wM2V8o0t1{0Vb8cMa3zg3C2}#kK>!PfR;mW9U|k{& z=-FdcYQg2rL1lxBK}}9cVV2duk|xn+;_9Vw>Y$C5;|L>Kbapb)Rw8r^yHMY*ok){H z1Gka#0ztjtc@_LykZ(#!A!ZcV*6D;z@YpACfirNmC>o|kH+%-9i8>IWG(!P%y{UXA z38DK#=v;iHZA~U48Hs8NoT_BUn>XIQp!r+`{aH6IeyS=|N2I3W=QRk+)7&Y+*i;kJXW9~Ui$Vge+CbE(J$5360{g*t)~ZU0*Y!p)J`BeXPP~ES)PuLL?cR;?qtZd0zlWA9QLyl)X?@+H@10*- zoO0&BcYNjLvrhuvKN?n~$Z7aBq28zn#uS#oSku&))f)k?ObhN6cZ27%Cf~eq;Kq^3 zZznI@c=g6hlb0qhqZsC6dv9EwI`rxanUkxE5QeZl&;lhM6` z!j@#gB6CShG8(?3pjkLeO|>rgfY+D20~9Q0k~LG56O`aYgMmv6L+lW}#7KCyX}f|b z0E~uH={7cj9o9>Gsqhb~vt3dJ`E+F=!X;aQ){ z_Yfz7jM)}mSQx{vhwuZ#l?F?*F-ggs5E69NDvUW>36|ghvvt=NTy> zw9O-b#!%=1u#1Rvwu!Bu1XKLMl(d9pPXi=_({V>r z05BP6+9G$issT%E$F( zq86~V1redYl<3_0%IOP4XL>xv^Z-W|t4$TTqdbHgRUlQzGc-k53ia->Tn8B<1t;XS zkZ8FwahfxVI;q({8~QQT(px5RU|7(w!--j>h&pJ9qiq)F)G)#WM)@c+{3+G6FNl~9 z?D|SM*kvLwYOB87L&RK|RkfLo@-^Fr)8ebm(S+`WswC^3s5HF(4w?A43k@=3Eb_v#X2)tz z-{$t7ecsZ|7i?o4%h)?{RH-#ROEb&U(`&N2TlNU$EqnnSAY76VAV9$3YQTw&ZSa*; zZNdJNJ*s-yF?IeTLN;Ji`zxGh^+=l5j4Zl)2DfE3*w%DUPixlt^;*yU_qfFJ7-*M* z&MSwUSA7<8QZCz`LKl*GnG_LR>Dag7TFwZsDg4!PlCDoLJ8FL%mR~APiD$uKU?|#- zpOINkrzJT9TfHJ>m9>R7sJbGWO~PAtIJO! zhkteHEh~J!Tu4b7C6|PHq~pOc=X84Tyo@^pe<-$!tWL=V8OFBLi0p6wblD7{61yV2?l?B}L8v zznL>^3T+soZV?yJZi3D5l5c6!e4F9QZBkb0lA`BzSxR*PS*{1S-2Ts*FFyX_)E6iI z>ya-`e{tfAH{j=m|2p!;$uG`(KR;Rv3H!@I`ad6jap8-TVjN$5^u_z=i;v;uWfcHFUz$7e9~c}~d^@U7De+zh z#8Q1vMNi&AGXW+Fl!zJ1WlLZq~FH^%U=~FHg+6hxlw}Lz0*$gw3JInCRn<6 zKp6z<8dqlC`Sgs5`9?JJcwBj0dR9_1Nq|cz@dh;o2hC0jc72>nBq!yP(uZYvATTuS zuzr|n9v54BI-imyEfd)_anJkmYYw$AnsiOajnZhSN6RNIh#}G)FeMzaZApl!`HSbS zpZ&}Hqo;42yfpvV%kwXP(l6`U{Hw2Df92BGm!ACka(J7nSERClp$1DWQQewCX(QUs zCT8BjQ*kb;w4`(xv+FNYlw-pYw7v-Q5)6$*AoyJ|Uo;vf8A($k=Uc23dE{_f6O;}y z8a-)I!!LIv4z9!PY*(#JS&3C^WmrOXjat+YU0CB&w$q}LU+wgkhbW8~Iph#-VG{{f znuJfPi>G63&$fRsQjXgeRv&mL)9ZCq$v5^UKYH%>SH-nbwbq(S5~ z-dG^dzB>CPYFuu7()dWcY@A1pbK-3C;I7yF+L*Fm5FFr-3!3oRCt-9jm@6J@_A_V$ z4C?aiQ-UkJaV30;%or7Il7<*Fm6TswPv1sePp_)4E3%|?5RN8EgU$VMS<+w)7wH5u zX?$seX~7(!MrV>*Uubb$5{Y=kvSZ{sObem_ExP-6ok2E9Crs9>5Lh(0GEz!Uk}O07 znZka@nI#I1z((Rmtm!XlK_`$+V*4G)2rdJwNTYH*^a2?*C8^nj-*Cas9a?DB3qSwj zqS5lqrFd*CRgvV5a*)6_7AF(^dj_`j13GI=^WJ88E-k?hf=yzh&D5+;5Zgh!%p$Q0 zE)Ue-c8hF+af8+(cTw)iGRm$%zrU_~nFv}b zc+iTj;e-dKyc6}7>yF259yYNpAcFn5zVA!pHmTXI3E`Ik+| zu9>|P9(P7Z+W~gJ5QBD^)(&!FRBS}8#UaT#FZ33Ya^wJK%v)89tzue@VgkdWN~sRQ zy7jm`z}Tk_Mpf4~qKCP=T_c*5l@rk}@>mPFfH$HsItfxRd^o6g0$oc>*^a2)`h)~e z*{)Z5R3~Pft{2g8H`5-T3#?;-B-xD1e2Taw!Ay*H+%b)+SAt}C5N#Ve6dHg*S}M3D zMfSzdLRXom^Lk2>J3Nx>z{B!@TgGSyHchuuiNN~X=}d{B%EH{|a*3b5%^Pf!>RsTe z*A*#~Z8tL#|H(4{$rER;UwZGx`{50c-Z@WnRmy6KHQ2YysvG4c?}=*}*ooC|qU(~{ z(XBG@@&TECN$ae_#2{ z+GA1p`qG=%&mV@7g$IdNk#jJoWV@3+87b9azT?}qE9%=#>$q6%YXj-pJ6|5V_TIHq zUmpGPi7y|$P3G-bXXfn~c8k>OphfSF{M&1P`||j;GhaS|uDx;X-D@9A4}y`ksLUoeqEXC? zZh;z6d0e~2BPS_(LSeHSR5$+q`nNBC7^m+kBg(|PBpAF$NKruAFxwzejICbLp#U0+ zl#R|De(!8zeq&2m!A4w+X&F)AE$fjx_O2TZpd@-v(o!`}6+qS{U2XT*hUs2DBwDpO zqCq)~04^)-;oIAW&ij7uiifJ*h5mEb#^O-);&cI& z+pt$`?ew98PB%q3*x|*y!iBz_!%AT*aQ5{5d~3U-d}%5cFv(;jJuB&8$hRQTTMdoH z%V3ZwDJhjode8|=rr??wO%X@5KZA*LB3D;+sRSspq;_b@ffGzBjPtvnFsU;!lpxVY zYrDsWS|nDbI`gIpC#Z#nDT!3S^H7mVgPa&;$7_eW07HQ#a`X69xM`|cp5da(9FqSxo{V>R2z63g<b0)WyhRAS z_AuBNHp7>9u)Bi3kn0GnkJvVYne-_PSxHghFU+-~`4S;C*{8`h4+4UgmQ*bv_g>vp zt>T_BheR?`l?>QIiJ!V}kXpsamQR;PkCKy8>2?fiW^ifrAjKBNXje(tI;g^`cfWeK zDj()3Gp0xdn=q(Nm^<}~``MBDCNXdMR4xyb)*5hV#KtJE!X*1c*Q(+RU8{kGlA4mV zbix#PT9nf;*#NQoAg!_sVKVT+V7_e0&MceWUrUL(^NC1NZ15mOqvXu{ct{(HeB@ zL!5!N7eVo+RUgdhmg0~Ot~P*;l}id(#a3AO?6>_TT&1Xlf710{=;OIe2X@?PDM`*m zj?qoZlA1>SXqP+5o1Cp{s6=v*U|y^;PS8GSP}&>bwS)NtX-&V9X;EQ&uLA-gOIfX5 ziB3GHexSa6>Fo7KUY)=CW{eI~O_kK>l;!t)$kg^2_jr*w(PKhwDui!2g;-6Rlva~I z4XeA{CQLlp%n$8ZDN7_hTuZM|II$iTfdvzF666u--ad_ng=qZMjzxt9#Ny_;$C0Qz zx%tttn{U0C|Ni+`zB_*kegB7JH&48RZashHyYnBQ?~i|Q>-dZReB|lz@msGR+i>%( zEBCiPv3+nCy7~48-+g>x;kxnh4Yyvoa=-Wa7ytU*#Yff@qVRoh9Xs>=%g51%h0)(S z_BxCzeEjIv^T%$zef-vu4`7O05ckU~S7*wrW1N+;N@T@*y=;!6CMr&vfQ+DU$adJl z>u_IO?pKAEZhVX~8A0Blc91u$LsYnkLc+LHkx*+Wfp=FL(G;YrwnD-PzagYW(*-f}dyxUe8N+DbUfqrKQdTd_&P z*D($gUbm=kg=A!3ZK-z<5E4Ox)ou-i$?_QzSw7;krDpE+L$x9XDS3z72ffUl6Mm4b zaECg>6(LK!%yPoRn?FBec9~`~X`eqqYk5Jv*8#!FNs88Mlw{ub4HBW&Asbu80^y}zc~~Q;d4LJ1c}jvw3zLeCdAyNx03{=7 zsmQ$9KMfN!AFwGNAOD{y)Q8q}q`h9_QE2m)1=?jf=S$HGSh|!IBn(+S@?u>tqK;V; zt{rPEB%7C0Ql_Il{CeD07aST6$HJPiYt)Fv#grAb|8f$YS`u^|*Q;*t`l&?{bF$iv zV|-l%;meFw2&dSJkEJOq3jZ~^n>HQ;#UdJ{ti*AxTwe=3tV^l;$iZf%uQR3JJD=Z6 z;RnLpp@vE9O(TbDpwg3KXsC9R3aA)#V+bC*Gw}rlcfYNg74|fI)+~$$F&#M1x6| zvZ;2M0>?JS+#07ca71mAqHHS3{U zy<|o0zY;wRxK6}Uf!d%2#2i^r6)6=xh`uhIrY2kmdihkW&TUndl;{;^_l_{x1*#mw z#5T&Xf{2L)*D#M$uUnD2A&fU0@BsNz^)e`j7#ID;#sK=Z&Fv*MP1nw7jGVU_b#d{%L+mVE!Do_LZz8|l%|4q&@*C(wvG&K-?wf1 zHqohybH)*@N(rH{y5-hkj-Tx5&9=4lT81G^>G)W-SHcw7n=H3@iL%V3>NXxi@3CnFKCGY!y>3x|VTH zOEV|qwdX=JSnd-o;NTe?D!beetfBC%rIl2_oX%#$FYrOrUqKRe+TIx)g~1 z)|+HqXlP}}spU6qI(YD)L@3NWHp)(aS=6#%)0B#26Sa!ukW?$zHaP@WH`z4`-;`Ix zeb$&qeq?ccXe%{~#4ZyW-tOLsoheeq4y|*CM`Ql#mAS+3&K-GX?(l1KhhLaG{MYMOj$A)~`1%WPee>>%p>cVD zl|@B&MYN{#1rA$r!=T9MvXtp)jTzVi!Uve^2!n^uoSc|B=`~rf(NEFB`TIlz=bA_{ z1*?cz4mv5$rT@HMhVjb`UYoCnRbP3M7vA| z-y%ZhWqV+`MO-ciBreYhZjjCvkn0hRCC^cTlL?a?>dL3r?Gn)?rGtkeOe3ey{$#oZ zG=;{nY1R=%2kQs$R1s~Eb6NN`xtwyp@F*2ID}a^DNfB3!`T^_iPuyGxQ`c>$v)_|A zncv|WmHMFRl1)P<<`sgrU=GSQ3!hTEbSK{eOSQ~c?{z>l(^9&_Pq7}CShE7YN=lrV z(fJ5^B7?G$-a*rx7^+K#Y8i7j6=Od^Kqx0=)pjl6w#hJkn_;7Cb{XySBefTcHc8a3 zBG)Ri6=3LlXsoXdhQ5a^DiRAtdr!|V_T%^O&ZLyp5P870DC1)H&S_|Zn$;E5?7n^5 z`j!I&`=V_y%l7WH*qc+9a=MgLrA(@YPyfp4fvaQ(<^<15R!=J4m-aM#h*sG_YD~pS z8IcogD_S;igWGmo1}U*7iRBXXlOIVHSe*85Q5c42+qG?xYvwu>RyY>U=oT(3zn|Z^ zZMU2WuL`h9BKQpaschE$(fmTcsAA zAiXLch=z$^SV`Q~K6f>I*6?i9cpTO_L9voavV&IN%9srkj>EGP^l;(iR!9>LlTLugNr%ecfIf6A(+}WTB)T|c~FcM75-{DN!LTeH4IDA z29wEC{T5^1T1!LlUI)WVNv+U%E3g7L;3Y5o4`hNOXN!j3nA$e>G8>46=~eErf7dA$*0Wqk)+KQ)Yae_1x>|9gF>A}8>;Zg z85CP9d7Os^+C9!gFb5NiTi%o{;FQiv8vHNoN(-jYY1p|uDCctYh|RsTcov*870OFx z+D*DzFIJA=L`k%OWjopwvq}YAkB!xJ*-a%x!U!cIrBn;oO zVl4T{2`Y+IvRXuzgK2vG-!Ph_92B<+lSK}0cDaLTiX>}^PhdfV22BkYyv5Q5(^fA8 zi>69BEpp`{sLeLR2h+_$e^{`TE4XIdef}i&Dd}aEq=yH@YmzNlVFPYP0PSi`9sO&6w0?U6C>! zD~7uPF=_mH_DJJW<8tGZ#?{7o`1=WJ{JrtX>|r={zH!bwc^=8AaiQ^X(9+ZXgYog+kzL*y{jc#S zZ%iM;j9h43M2**A);{s3NX*H_(8v#0Mb93z-K165E?4~NIdbbEkb^%2Dm(=3_H5wP zwufp(?1*(aP^?nRi@(_dUSj3@`JNfO7Fn-W#kf#=2=paAQVLONQ#O3&0md+q&CUbxD;IV$5aRs_0_bP1DhC zcU9P^(n+D>G6mj*k;l@*65s&UT}wjVE$=qmrRLeuLbAJ~FhktCY zZvlx&v~D+d!w{GL+(lTloa?2-`$tL=-RB!qd#CaIj+u{aqg(-=tW7zotr#XW zN}D$ipwWIg?OQpT(ke`79t4C*m$W2sv|JJq3Uk4WPUQYm-7cN9l+_Z7dEahEs%{h$ zvxmD`R>~xh_#v6#)aW7{(GyT|Qcg?KM*ViCi3`Lu!OH1Hhp9-Q$g-p>-DHi9@|lyS zTl9)V1j*b=Y*KxD_Vu-4(YFWFgPtDR>&*AionwYzID+=M=oN}k8*PxY896+sp_-Ny zRg+{{5k*ohT>6!j8;q|Ciz;ayJo4*st@z}84b;mRvvve1qJ>GKeDys@T;3{v6J+TY zwc<|OzChx0E@I}UC(<@51yaV?JEqlEtT8n~-Vxyu=>nq0v1n4T&x^4A7|!I2M{xh*xZO5P$IEe#d-A4NB;cq!#i-H zi22N=(#%;ztqGXDLueW~#5_Q3WX+tMM85z26^#@{+iW!G!oFd_h4$3U75MryCvA#~ zGbg~d_Feq&!(guxr^0wYy!h<7{^k``Y);ELB{Qu}E4g3)y6rBLGbdoC;nVx~4{}`e z5l&v8SEP8`n`E%! zK=jI5R#LU;Wk;gL`u_c`!2j|~zrOOPZ%&>WAHN^X9e#iA$nm)&PoizZc|Dz#b17AZ zvwywu8#w#B{rk7(cki8^f98*KN1nL;(eb&%FZ6J!-kChsZy$dHzWm(LXXlQ*FM1Zf z_f+TyVGHKCQl+l1rb%u&bH&SmkAWtoI-Kn5eE(!M*D{(T)@v|}>@_|=gBs0QdT3!Cu#P26TfyE|aY=<=!zjo3BOKJ2X6)~s)6S36EaOr%P9 z{f`#=0Hr1PlS_$uAD?C0GHJBHCXJQ`yD~}US|nyPuxUnTW$ua0&q}GJzov-nvZ5$; zs0vm&fCcxAD>D)-dP-7KNvT6?oe=JtBsSUwfru>W9o^~!E61EvLEs2FY|<epTIr#S+@`wtq!ud;Z^+nWpweeiz+0X!w?HkzBpYa#2 zsG6^E#XAVN>NF|S(bzF?ebEHt)r}qpy}h>9+d!3~?aVIIDxGq}hg_M}Uvd3}2`x)h z4jhEg(~{bus@-<#`v#fd&TUtx#haGx@R>JDgj2Z7Z3?~mhDSOy=`Dm;u~ok#I`j_$ zVmZuOsgAhs!F}xt^n(ocG_<=6dwu^tVmq}mHvMK#HY>?$N~qLI$`>LC-B?p#l#y~B zSr zv1u2*0J>TsvM3o&B?$RT@pUM)?^Ik zu7%)VT#uj<@1zxib~Bqz_`vCxQ@3%PrB-pB0X_;CD~XrcCs7LV6bZ{-<#tEPXTrSf zq4U+HHyGM6b2T>azbtCF^$xPbz&n}u>i~x#?2r=U^}Nc9{Z%S>kGy>v418Pwi@z+c zw-L3_h|5`xjE^rtxNkJSb!d7kbKo9pAg6Y5r2Y^>(6Mg?g&TVx4T(0b`e07CoCWE? z-U2x-X}Ls&@cYw>*rw)W*}B7cqvdWl<-d&UK|w z+!OV$E;3wH{p#{FXmaLt@5}d~_*HAIA1(0uQ9uLNk`gRmrn*&f?Pk-n7)uCbQ7N7a zo>$loj7U_QIa|UOHp&j#0sG><}EeW#u*AS}#FcC68dbobllH3v330xuv>*t%d z|M<<}b2pxT{5ApU2Sz&r(huO;k3u~bp+fPAC+7~oHh1{z8}gwC7M`v z4K-kzWbZY2`-Wrg5oIubJv`pzrwsq%1-zq-C=TQm~mgsR7qu5jG4muRk6!(T|Nb3>E4dq&Ah?T7=sI3v;k&P4WG)^!$p^qvU(RYx-SEwUw7^34xQRE^rUjIcODCF?&Mr5&rsO_4Kt&0l8z%W`w9dr=e zVg(n-1mgtcK7qyT1; z#!0g2GQy*yr@ndOZ}X3S6dI)7rI~8V3TdY6P@nY0T|rT?q7bL{M4J@F^w}hA1V?5G z+cv9}O;Qk)Ixu$lpyfobFX=^?Rnw)MnvjHa!h`D|l4C_CBc-!R0*T)+rUls1m>rj4 ztmPy*oe+chqIY&p6w}9Q0P86!)1iy6Lr{dg7p3)tV>C&5oPBBl6h%@ynh8)RASecB z7BiVpT_&;VI(Q1cD!uFd48znmT-%24>T7`RJDDFUQ(|LOq-YFVXbZ(KWQH&rGnj2- z15b+nLg5{;*yRlEVN+AoCvnsk^O5&DAa1Ip%kAEz#Ku`xPas{=%skZ5&;f zQhMSuUkbR^>ZK}K`@fLFeH}W%!zF< zIJs%WGGX-gG+^f7L(~@S7fog((mX#tj_Sz3)dGcca4*Km9io2eRqimuWtYPC2;=%W z=D5VB!f%mF)78RoWiHZH5LEK1l$7nL-doQC%jNJV9*G5LC{j+1oLYZxhnc(ru{~w? zt4pV>jc5>~Jv_cbWJQoBsCtqx3`bxTE^sQ=;1%JM@%iQyk3XSuKM)3RspgWu$z-2 zI25#wgvWKbP?mgi1#Du=jvV#r-ZiHP7KN0agnoQd--;&+rwICltlS<<2yCQo`WJNl z+$UhgT|f2I^_M@qe*D=R7d}~gWkKJ3@ap_ue{?cP3!9^?|z*Mb-sWuiGO8m4xo8Ov!w z)$vw-5SPQRp@%FwNw|&8{%sZ*H^df{oz05X1@Kz&oPg0BfG*q3a#}Y&PQc9-8HUk| zwR$0kl$4ZG+D*zNjIg{JIKfr4;l}x6^UofSkIE9Tk|m{sN_-v4sU^dPyYNdRHy`Oz zLMfnEM=ut>x80eTacToVWopt|jmFp$P7&k>n%wTo#J6*r_3fs0TrBssX%pAp`SRGc z_pY7#^5~aOeEH~Y3hu`S`r7d58^dmqq5+Nz-tURO?*7`}zC3>I%$HA~Yj0e8_u9u| zTM8-*Tk-1~`Iw`t(l*eE(glQ!w(O^$C+gt*qMe?XmAM50;X`$Y54H@#@Xf# zW}nHcxm5q`k=Z{Y(GdkTu1ctJUi#6P<`&WIdVO||q=o&8qs%QEVCbiz@iN@+lg3q; z^mC~3TH~~U0+4vRajErmWLrhG*?={*iXddjYLQci*Nx)u5Q{|4knD*>%t<+=qY>2Y z5*zxq&;v7Pt#*kx$OWz5Z6x7HXIH_oT5~?)7lA{VbC@OWWhfYw8?H5pO>EHSDBGo9 zM|;5m+X2H3+bzkL1CghvrK}1P%ZH7xVNk$!L5kpTNGGHweyK;B)haP&dnbOE^3NgBqSR${457P>!%#{w*N}~NR8e-J?C7bx=;`_h!UC_IF z@H?Y7G{?XR+O{n;WD8czFcICa{QmWpju2Qy&{9c5C~qy1&t7xzXRonOUo+i}Xgf7y z^`vXEl<8=Ads|O>-w?pTn;{(7p$(^9zn(vJ;rc5_=bw4vo5LS}ef77uQM9*i?@+Y2 zGQf5_^PKGQE8Ev!J@d^ofBX8vZ@zu$FW-Ff^4FK%n!kAN#)Ydlo_*}=3#VX^p%K#* zNs(j?RBk`xquk2gk^^UN!L2eK^x69iVA*%E+uE*qI{2v*a!YE5tmVbN_x z#H2)gs#XRQmY`q2m6k0sI~_VnyaIU`vTKwPw$WyG=ofu!Ay&Kxo0S%L;wb+Bt)gvK znN5UG{@O}-VT>Fkrs>l%qHwAzqS8N;OL~4aH9sPXWJTi$;!Z|(mim z33nL(l^avTRRBcraYZ&{@9{ultqi-dQRa2Qkx@4z_sZ*u6v|22=n>z0XXUhfbIi8D zgpJiTr^}L}N3JSbKRO(C+*pGSx*};EO>-+OAWr-DfB(eA{C5{l-+JMLO^T9|<+&u9Ly7}yjH=lXx=G#~9%mdLYa}1{M z*0D3Uo_yuzQ;(nxw_d(@>&OSUUOo5y6BoiKHZ4w)Kh&RFIntk7^~unrR4U-UIJtXc0zi?aR5~G|T9s_~!rSw2p1kqo3-iA{F@NFcw@1S_we8^| zQY0mvSmaBKr~uc+yMRodKX)Z?2#*q79!u_aCRfy8MO#TcEg8`KwERhfiL9jr#S#fLi3_%m>JvIfaTW z(IS6jEm9R!Bbn$?g7x5(HnGdSF0Ao+)c__DJ4hI=ngWmmDpSnDJYGn0Qm&&l&N}QM z{nh2?4EX+ET|Q%=Lo+AB#|T#JA+59u{FQ(O{%T&9rL@+rCI;pOUS#rf;E?!;{5%3b z+oq^UC=y%lbjKNxS_D;DkxozpY=Ekm|K2eQ-w7PI!n6~P#wyL98}`631_^e|W(%gS zS^fIe6+Oo4s&J}Os$+!(c8hs%4pbah_~zYMnJ zR$MK(yo4Tbt6Zz>F8c3~j& zLT1TnNmJTWN|OyT2$(a6OU%J&JEyiYb9SO9dt!tznkr~FluY6aY4SXkShZGmZDf<8 zYdCIvdBK27PLh>&M=vph=PrK!{C__3`9=79`SVks|NZl;=<^RhzX*RX!@qz3{Oo@o z{`|~;KKA+FK0mV-@)Vb2_4$jRpNBDA7UO`?T>AX4=<_%J^KtJKynOES3!i`d`PKh? z`t$Hk#_a>!_v{+;*}%RG;cU@j{2MSn9r1!MmY$RIjd2M4IUW z1^~ipJ);qd28t%7Ta@Ek)6OVj*>d5?)&)<0d{cm4>?E3Q;=8wPd&53to z3}7i~DVIrVxmx}^a!52Jiq#lWvS17)3G&Mo=^ExBW{4@e2Un|c!lS07l-ACidLU@w zs;I1c?9utZUR-<1^4#Ggb4PwZclfo?lmD}awr|_D)29r}IX^c=z{*oqkW(lfy@-=s z>-JphDzzw!bqiBfP{)$6P)n>7IGKczTULxE!m6OGP_?dTMDB^uP#s&;a0mynVzgfO zCO)LL;t6U%;wr#JmvX6&PESEIoxNWNOJ!&w)j6;7Vt>^u^YaX5Vmr8CQoAg!w~@%B za5<}y@$scqwr@1Qb!d7kbKo978e>`SqtDoTi!R(%UFtt|m9 zUIqlZt9BjhM}T2RRIA2F+o=kEl5|4KzRbDY6r_j;HG$^Gbc*a z!pPtR7X<^3h5r5hhacWd;btBin@f0&Ap^dO?GW^T|D?YkjaUXQkSSbbsK5XJLDLqi zf;30B`|n@T!w-)RY#DiQVCTrMe~qST73R@eCF9Tiy7W6wVPF4*ake=yvPTyq?8sVymt`r!I34sqj%LhT+hp_ z>H_1mQuY>gtct0oN@^}LIO`x-Z^PY0y)(0V#lIU3Crwhi%PqTCKxw8|TWpf`mpEuD zPB%-aTU64Ll1otWqk>7lf#Eg`6#ZCwkEw#^B%SVvyNBZp5?xj-gDf%z?O+CI#-OyU z0pFoWs&7H#rPyyFY+{0wI@-xf)U*iSWMivXV245j)NvDJ?Csc@Pk*w3O8o$A!-kS;mfoy*vOKC-;{s_)*e&M!b9Yl|vZfW?XuJ)@j@DV$O1X;i^9lF^eWbKj0Sr;DA1O>v{x z6hckVifE~JOadoj{uE?oGp%tAsc+k0Um`{XX2_*v&BYA9R#J=>FegS6lp&d1vIv$f z0^;(zC~Jyy9jKD5wi`y=4ce*fLgRAd^6b;1LD}phjnlJ_H7`5ko#w!L5XZ{Uj6-Yv#?FHxZ1bT!fzshgFPi zhj}NNRdbmsDj$(eEM$Ta3ROouxg@Lt(a~zN~_ux^U$_VHUsK7h&ft+@yl@fy4<=}DN-C+t{E5nD|Vf@Gt&BfHnb zO%Aoalxb`@B+gPCO)!1PxuhHBQu7dz>QIbr(P&aOdXmsR2nH=-%jrS?zG!HqS0MJg zeG}iE48H0lEfTg&U*GgEXo#7N6B~(8AO;&A%y&@7F*4mLv|;y_!TWF1dhDT}btvBt zaku2+`dy)Xd&629)Czj{&W^^u!$dTR!eVNe*VNEIXG=})(~_>i)Wh`Wa$UGj0l|!< zX4}_OVDI%%^Oaao%tE$Xs8ZV&xhWgG5Z=J%p{>*6pYd^&l5%&70x>#z!yHdIE%>~{ z{FuQBJdN9|Nn$%_3ukUE^lBSBD5%)^p5_l`tN#DQ5Foo}b>e`tvk0Kb;VHn8!=-Nf`^2;YevlM{nWay$%M1CU-+kbPw)eml&vG+RQLy?j6OuI%NxWf-T z6vLzhiVmKFuSt+Y1UCw{2o~UDOZbEl7c*!OFAIgBenHk&^nYjrJR^kDN$k+OKOJTH zy52`klj17iY87q9(FOz8I1ZmqJi^fsJPX`cbk?t}m`bdz3T949W!keI60ZMcJMp%m z8-MtC{^h@f#;8~1{#srablHhs;#28OY_M9nZpNk4>xz_(PBU~5A+ivQz%Xr~$j1|k z&^1X@B4^Uw^GW!Om$6Zfkrq-D)EAnZlxOzspao)@Vhfva&B-`TT|FaZ)9q?pQX$q= zh1mBm9~Tv3IXU(UF+B(3DNz{wxj(jW0P^@4uk|LuL(&dNM9fXEmxjaXgb4;;oS3MR zswaBfl4fnXoEBqI<8;B{knM;dC4TaDKRI>?7sW`iDG$0R`d70_0kYpu&P?0sfr~sE z92qIo;gVa2a173zGO1B!likD6qQVd_1HzD%a?$(AuAlv@Ci)b_xqZngDVrqx{xDD$ zHDmM|08WaOQ4%X=nV8{pidhaeW8uQr1&>0yqdD+8gk(uyW)=vXqJShOKr|_rkjnZ! zYuGB0Nc2l9Z(LC{0Xdx{kuJ0JZMRS)hRZqWRnLKFizOpv6P{TwYEuH)z7c))XUcK- zT*aPPEpkp$qA!y5qlK9T3d<%=eOZyEbkfJ%w`{gz9^0}}6e*qUCeM117ECI-uh+_T zk#vKatsZ{SN3CD$5T%q%Ae&VQs@9z{HM~ZPj?0M+*ZlAQeP>|nRnpBsky8Rw zg?8e)$2A)cw$0R>If+VcwcyLP@pT6djGQK=qnn-g&NA|uGj5gI4)3BQyGw;YNlU7l z^q~(RgNSYPOm~a6?`0)9(^cX8JpzjmLNo7dM8ja?gx}11B7@YFAP`Ql*g;%_abBht z*f?dkXful%acXL+EM*cJbf0@h03^IKyqW3*&wy9c&?Wy>2_c3Si&K{8$;Ath%Yd-J->65j@S!8~9WJwZ(qqy$%SFCgs$2 zfUKYZO)%FvAkv+=?GUrrGgR(OKDsEn9AbqCIT*u~SN4vyCV%BLrPU`sEh*VVr2Ddz zT=6Pah9X^*T+2yt-|xI6)R~GL+wTb^IZ4m8i}k=gX!h@K{I2m@e&&Vg#wW8!8dnbwxK015j*`v*WzkTZM>+ik<;|mRz4%&sTNQwehkN8!X_E=M!3i!KFVV_S6QT~5#>m~zcu2RDBRW+%Uk zlc{hL4JT2s0j7ISKhRn{K_S^V{aP5 zHrX_aO(K{aNdGpQ3(|`P)(0D@n?^*}49g&Efe1=$6>YQ*OQ-T<#BoegjSw{+HC=}R zGCpakrdo6!-s@m!>awJFG>eCI5*W=Vy-22ku1RXTJzFVoz3i<&p1b+h zi?@!yc{!8AWM+Bn%e`EDF$8ce5g&IaPAW6ryPF%!w(G zrc=#Oq6{OmMIi4Hn%!*u7Xu z=Tj*u*U_XSaP0A?4?ld+{`IeaTC`C=8Z*H70c9Dn8iFBH3#0q)d%!IU(`XFylGo#n z-gn=RB!Ai%q$BTmyTJT5NePumfxEQgHCkIhu;ZfAS)-#RqCl{s*aHNb?oI8vW$BJkLEGccrR1Rocz`W9?jS%O)- z3<$F(rMn|>xHtN@7@&`rsf|kR0c=Gh2=7$~7XU?<(jA&e7*4`1*lfmnX|*c0P8zpu z8(KBo!^zYXwiv992+rTl%(j^Gd*&?lqN+)cYpPq5~L@ewkY0)V@ttB;qeLF;$s>G_bvTF-&c8cS3 z>!d}Ac{-Jda}ejOW>ZJBlAx?scFNd{z%bK0i7soBmP&jJOTpcWYfX_@7}x;|UCPMq zSOg9mz#`3~eXeC2p8n1_I!JjD6^Mn6GN(q_p(W=o$7GOLt~2E~MHkQH_sv`>bHdSf z6T9J6;4O5*S=P5GbXi$i6g4XyQJwX;RH%vVn%jLzjM_5))xhSzoRySxyS{bX6wN-+ z(j7M0&#eAoX7#mMBgz9m@82?Vo4ldj5B0Upf8TD`w!L2vf6`Fn4A7}IIh?#1s5mY*sZ9wd!ZCV~*bd<}flnL;%Zem=d7EV{G~fu&7cdoxrjj7MQsb zjNR%PJMr7h$TBr>iJ?KbF8ss~3*sgYU;lxB8*a^f%SRvB zMX#i#R4VbbEME>qItlXKz)l1!rH4yGu+AsYEQiKmBAgYuk9SXS$w_*27P))8g9gVI zwOIk%h|?l1xK=UNH6+LvI$F;pf#x42&^(4MB$BQy)@`i0l_Z)Dm?N4NYz}&KHiva% zu@cLr94?I0W>rP@M^{d`dm)an^Q+6xApX_mm*4^W^f_X|KX4|-FoLQ{DK*hI`gUNa z%!xs5LFQyr?8He^lNQ7eaw?&zX;BPRF`yD3!vG*6m{B7ma_=jEb$ozWpFJ1nn#oD3 z+U~v%T!66YU(l`NFWx-y>UU>P+$ONFeRpSIfz(`3hzwCtx1Divcf=ST7}Y}KAS<#Y zXS0$bs5CM@FQ~er7gSvh2|=`w?W*wjUXb-ebqiaLuwP9OC!C2A_&9R_HAPa=?WG}s z&7y~DMeH?`9w=6Y-tS;A05A+vvzb*DrJN;l00StpL)>J4gxQr z%nH1ZfeE4jw#JFs!?RB|E_+l^Yp$-h@w>+P*+;~87tB^0QR6Qn|6ukhkG1JEYCPBY zM36Sk9u^c%;#BBW*7&{0`gFE=Bf)(sH_17vZU+nNpcAnoPXh5fA!3bk3K}BtcD)C@y;Wmf$9|l!u;5PlLwP44S88IM3AlXQ#e+gdj@{y3xt;< zua<5XN>Pq@*M@9a;F_>siAEI4@dt7C4!t6U>aF zp!H1jh|>M&=_Md~22hdOxLTkkx7m#a)==9SG4wSdhL*Zgw6$wkem+0EY<~>gMj&Nm|M!tPQ@{(u4@F#;Gx6rEDh2Y3$odOuCe|VAm+8rWtG6f|`;tS^`YV zvok?W7*rM8=#T)&5=u^tEwlk4Wd-v!v5VU5B7<3?|Ar$nx8np$leCVQtQEAd^1(W) zk|Ltk1Va-XBVzqX8bP4`8WD;qaTQYoBiAYtj`r?E6ox@*krT&3axSNbMrNeQ_(XyF z+C*k@t;mA+Iv_B*l+6ZC&*~ebTvSn&akbs>8O0pez<8lpSf^-cj4!&cfGrNIST*y!#|#duMJ-u z!ONFm6t6cALjy95K}C|dBO_MUwJ4vn^d#R2+$_>KfG?%orq-NWdw{nP5 z^8sXnHp-@ugSo?|X3ly|!i(5}5zd^dRtTCoX(N-&oZ>SlY*cj}?7$0*FtXulgH|fu zg#=*}gORdU;!QIro0H>@HF@Xq+nHfkG294ksD5_Fp28J+fB4)w;f+X-b(N0 zcU5_W(BOk?(g&5A^R;u&1*C_}c#HvBWOaSL?3zVf#Zwgbs-r=qJyTCfO0HeL9SB-( zJOB8l`A09z9scXw;UjZLp8V$B7r*)YX*m4$r8noFzI^@o<+;Nz%sq1Co3{_&ICt#& z2gl|PzXR`n^WF#7UwnG*$dMbzPQV3s{n}d&;N0Qgi$6zxi#8~!lsbR$oEZPn$L5ZH zFn9F#b4PzOcl6}k(LaaZCbwXYD^=<$El4$=O0Nai>2-O9EcNy7#ek`Zu@yo_5Dmzg z_LXMbE>&U$@wJ0hut}$2por9{>G5_@Qo|>#>Gn?+%i36x&$T# zJ-B6jTmreGYPzcB?nfKoeWIXD}VM|i&&_Ntx|+GjY|{JdOrG<+pxy1LK+7T zNM;36(g_|kPj~ILIN6L3;wc1Q3R`F^hHp|t1BD`d2$#`LTrP%}eGgReA>T3$pE#p~ z8PL%#D-`$JpJ$JGyk=J#mmB9BR~r`^=h5up*&_>0+%BTVC#Z35_UP=R9)B8K;=K#6 zHX``<evBHgH=b+UhGh2cg0}JJ#u*sN(Z)xObFG0NoqZezE(R!0 zT=K}`7J24^eIP$(lt~4?`y?7L$stlld#T|t&L*i*4*h`qz|9sPoRs3@%JL3^JU&@L zf{{@o*Q7Le*R3N%+xKnTz74ir!a3teXtyGy`hTTm387KfGMt4R;SwV@DiC;+pxtDG zaAFz6me4R0JN4r4%x@WaaPzi*8yTA}i8?B5iaQb;?oTHW7O^srCRIcZQQLtJ_upRj zqYbjG>8ZP&Fd~L2Vgt= z=VQ-~j}LKzMXre$ZnLoIz5@p;rGofVELH5%fl8$~fhsl}O`rqzfeIqU2@{zU#rsyu zX4%}y{JxoE*q1|Pu>?v`3Xrk90zqci7D>hNGo?Go0if! zNd~i1(S3fO@Fh=K!QhdL&gQjVHbawHCDBJV#x5gSLF}Q+iG$9+9WtvJb8>7}uzP47 z!QR_eGp%5+jA9cuND(AlxpZlub2}U0phr5*w#M=2qGg;z9h3a}*XZGg_uxY?pNG&iqE&I>zb?M|rGaan zQ&^mVp<7_1{42mcV8Pqw_&CLIzE-ZAi|4sh=B&nwP zFb~8^S&6XV?t)_#uIr@?Aj?JvMang;5VWUE&>&-0kr|cnl4mon6AL?FlA7Wcmd`EK z5PInhh~>tr!>Ji5*TDc%3Cg*1^W5Vo4|v@C=-AD-Ud(_0{43v`KZU;k!?Bwu zUO~5>zw+Jr5776=Ke%=L#eY8X^!WI#SC4JD`PP;DTc6kt&*#mzKltwB3k%nck8im3 z%9Z=Q&%gNB?=C*FrjUj2`~A^ZZXJ6SZIE;6%*{uSgjk2F4MT=CM9FU&sPxY{_6 z;KWrgE4i7ve0BDj@o^+lmm6o`)P>o{W}mvtv26Tq_K%HA@b~X9f^)NfoPDN+Z0CFV z#%r^Wz<|%g=_C2ZbMW3LFwjpLXT`gl8n41TfA_j!!xiwx>1L+#>>rmhy5avnhJHN! zXtW*=T}!w0oaRBW^)I+f(g_355-I8gvz+L5L6@7crbt>s$K(UUB!}GS?x=3j04tzF zT?v=uL5k?7XbZ>HD(1*26QcrWn*Aa;2Ylc=-0dd9a%4t1b#V3nXYbvgo4E44;eX{k z-l{ib9%!p?x@}Eu`Nkz+!eWx_3`@OrBppi!wYrtl-8OPa6$V1cWfC$;CXkSsu(<%4 zkU)SyfRL(jW`E4eRZkZ4%uf@?Tn1A6{t55*v~1M6;B>bHjqTnY;#R9A^*NvZp7Xta zzG$NX_L_fGK(l59YpBmv?j)t^3zu{35gbKOrAU`_`<_VIvRPn0)J}yZE~s+ESoTY9 zTa9Kz&k#@&LZYR2sGqExA*-MpdCp{5S$aOFr>O?wVxk)5yl)uRliV215LgCXD7ccJ zJR8GGvcue`XLtu9J--Cqe#6+Z3S=CO6!7S1fTnm-JSL^(n3C?7`UlhTgp`5_RpRMH zBAH0X#X%)45B13|+o^p)?IQKGNlp|JZz&at*0ZMdtU!5LIbma=E@-MKC@s1`5#aM{;0$Kj&=GbeYuueknln=8;NDgO8pgLdY_>j!2& zy?&4n3C*0n{xS^ZW#9Pzw%Fi!OV)LZq{lW=NzUey?5xrCvLK;_J1(H?-!Fe`p)orE zc}ka9Xp@|{Lk!+MZ}}{KuD}!{y!4p{>Sf%$6)@FFiSP2kxYP>|E)0Pc`ByDDF9n#j zI@%1&AgTBUb{Ja|u!LH)=#Jqg)b>H#gt{ocija&&Y$A1I2{T4<&}FbwVT*&I6jfWe z_auYp8Y}@c+buSO0LlUYlK`bUaJY^alUnmxlo6HI19@QPTg9Vwl=+Vxw$Nw!&+ANa z!i&~!6OmK5$*DVPvuv33S*f%b6OvK;qhG|yV0j8hdbc>Wf?*{Ik{WUQxc0+JNjI{X zmBTA8090f_79)>_yzMJzaV!b&O0DPdO3Gdvn1Js?087m{k^ z+_-0F5mw=4{yCl!8IG|fCIu-LSP@u0h}_#;3Yh3c5~`4-gk-Y0qq6U;NGHHPp9JZ& z>}4g#lv-jwqV@2}3K|(F#0rwd5(7;)mPq%CLpRHry@k$hRrypoV_c|OfEStD6yz$VllvOA31yby{~@y`uz~KF*zy3 zqAn-D6bj8z-tZ~Zn6;b|q)0n!$c*HzXR_GNg%mtcL_v}w&d*Q#K;+mqEvlo9QtxkkKN&5_T7gznMHUh*@*8O=Zje$j`PP}q+mmnIcQA(4zSv)uk7I!~4ws>4W03UctM5>He%mL%=NW}G2J>%9wHbyNI z?ww)&$2YKpC72OhN3kwfp2f;x?1OR)p=JuJ2_YU80oG%FIbxe( zX7+TZOfaYBu^(b&`n@OZKRJm!D(>I$t!A{_(OJSL=+$N+U~ zWTDp9)j*LVV!K}sC}j-TV?yk&RD%kXGO^hKI_i}GN{Oiw;v6nnjhK(4Rd7RTG)zr` zranMB4VzOtAdTnHm==L!^WMdXO`QbAlsBdR-+<{_JdP(2{0P?3dJO2~P+ukozlIBF z7#DK%(QivVfUT*Md4o|`W9s_3^ftH^q`7B+*SsL@F`O$=)1(=yQMecp)8dmiHg%HG zQ5t}`^0)9wqk%uFp7H7>-?jEwsDd0-T-GvNNJX2vqOC?7?TVen#W0064r)l?I=|(o zh7_jGQdBPd=js3C`$OjMm%hk`sfr|F*K0PUJiGX;A}VT2%-gz=z%Z=>>J!itMnN|N za2>Cs88kxlQIx|Iwp%`wI_aHr>R31I?V6Ile$w z319O9=dV0iyJF@@tIg}`=qS}gHRr@Q;1uOSyE6oBv*4K751-4ehDcISQ_YstPJACkdNvxPc283* z{^pU(b03`h>C&Z}FP&^3lJ9w9OHUIJJx}1WLoMBa#lrMHR5fvS-+|eE`)2pOgjU@+ z{K3t)F1Kk|*CVES8lmX<4fh6M)9LBH5Sx!lf|yVQPNo}k6|Rf*%|`UR*+?-VE;n0K zp7nK}PDJ%_P4u`Bg7^hR5XL=aeO;&aWfXgpU&cTKou~7FM9v1gV1} z)j^TGgCYr1L@MNpM$uB45z`DSL|6kw5yW`3tGGd7xPtX?L5T~Z6qu3M@znMeEt^+1 zQdq*!NVYIeOpCBG$Eav^H6!xj1PY5h^Nx32M;r7}7y>|G78%xy1w#Fkr)*@o2B(lz zcM3_KokCIqcB;UMc+YG}Z`-`>k>5tzkUDI{8On5Az{OsZQ#TvC{BJ7bJZiwU)&96} zsHfjaUh?m1I3xun*^%pXxh-LOmWI2l$T2~RM==IBBqq)4Ar)-aAQ6RBDsbR-rKHlC zuv6k}XcTERq(CK8qxH;X=$CMo!lBiL4hex=FLGRnMfqCwe61{_W*H8`al6O~A)z#T zlB06JkiFVE&Qbpv4K1a9v9_g3i$4!3K~6>Mr8`e7L%G*F8)h6(5rvo(c+FfHVTCeA z7R}J`7>H^}NrKc_JGte7N`t<^a9|yMOyH2qzF% zso{TQHOeuZM`gk|-8|3Vfm;XxuPz^+zY3sEf37?D63d zg`^a?VcR33C#~i*CB_)ZA=9!m2sVV}ra4gq+j&0e3k@1N)ScES?YC)V@rOkb)Trpe zy#Z*3H32(7Sk1p`sKkX>OG=w(?Wy-yXzqote!BdJ8&}?%d-K@RV>G$()>pSadHvSw z0W_mBn5t;ZwT9MLP|uT}5E3orVJlE1ctUOtmsX=;tl^0~W&YC%x+oDTL5er~GCgl< zw$mQxpXc`!WWyH_-b7;DU^%3bQ47Io*yNZ|f@?5iSkDvwS#D=I)A6VaP@8mwK?8=# z;xXT-xjuL=*U(j&B6+M#5qY6{Yn=Yh!Um{Y;^mlYFWGFXOMp+=ZQ-dkO1$!De3{R$CONby73zR=9eGH}o{Z)iX>rrx0Wwv6L*xsvyZpAr_10g0Q5RH@|!Zt36?$(&1+EzkDbumapQ4m|$i3+wy5t({-Co;4g z)>b16GgyL3IE{YX^T+3&1L6Pc8_z%gEF_)@ooV zLcE14cV&Z-rFsY2S*Lh_4&51V=W=14vFysRGj8F z(}RSqC%Xa=YV|#9Q%-U0E^U=XR}GSP^pdd6M4hr1?T2;j(%jVxbMIaK>FZB!f8w8w zq+97AB5y+9%Jm&NZ2j7+WGB{h%s^|wT8BkHM<&qJq&9{bnFz~mD~UoZ5fM?tD#^2Z z7|}F}5zJg3o-$5FSPF?Hyao;!V8PkkSB%F_!?4bguz2@tNXbH~h2_;Vbtvc}+ZZ0D z*`=UtG{UgeusNIQ)X6efYQTeu^OhDWUKoO=n6y1IM%0JG3hJHTy+1-POf86uhjvaU3-@oe4 z$5$S&e_~_*ljzPTpZ(*)#rf-Y>{xa8=#|Hv&maE$A7AfTQs~0{{(SiDyC3aAtL8_4 zcmD@4D*tQa-M9DO{p8@?y`RApyB)IaLl)xV7^eh9Y~HiH6HlJFesJbf&KLQ@0)gc9 z7t%voxzNiwAUPE!{5|luR~JYjUzquN<~Urmf9B+F_Z6J3vgHc2N{TgsYrkd&zY%WXl`+(99-L;)&EA=W}kbYK3| zHeLq)jdNe0OWinnP!w;x^D$Zjr{M36Kb*ex$?J0;oSA#$<)z?*c*Zs0d8Fz&pEo~0 zcJt7ITYFyk>B{@JPVBky{qx4u_=oB}{voG?L`%Nx3YV}@g-eMEaiv)q^X$SKFcT-DnH1{H9~%U*NOb;HHb_>V-=;uiHJK*#tg~Hm}>ULqMw}Sy5$GuC-BnNJST2 z3yJU|F*Ixk${MTBgUh{_Xz$)jEdGcj1SM)Q+_0RMsU9LxrN~(|q$mQicmyO5X0OHj zhv+#*#?27RtpTr;ph(f~*R#$jl?%3s3Y83_IXle2lo}HfEn4~tgruLCnXW~1gSb&u zNl40pS=V}hh1S~4DiFpU_k`(KL!42|?56kOq3rhXTNchxvKoyTHt+vun-&92D!|rn zjOQw`nCRO0lWq6?tNV6>xX3nFU z6qJZMyuoa-vxZJ=7Jv+@i?Dzy+`_SM7d^sKJvP0!`epU%^q%VZ>NnNXXnGP(z{}}B zz{^Y3(`oeFbJgF&t4r|jTe$2hTzF}EZ@Rj_dbWD8`fc_6^xo&6e?YF)u2uL||5!Z( z_j}bf)^Da?uY=nD{dDz>>P2UQP9taV7g6;nym=Pp;vD?{RsU|;ENtXvvEl+fQ3pNY z4Ri@`7ZNQMHW8;sxV2XWqU+fI2HMICC$`TSTl9gD1+}HEx1TiU^*mPKByuc;Fk|D) zZ!>I$m{Gz`3K2ZH)HFfS&jwhRhf0IQRYK*Pa6sbJDyfzxw2^3rESl6ZN*RNV(}^w! zXppBOE~xS5r^2&*{_k4-)znE|fJM@SQzx@iCrm_iq#I+TlmYC;R8n1PZN0bPjAoz_ z!r9vl->~%Mafw={opUAo1*oBuI7LXT#rTL)syj%je(&JVhcCiPkxHp|Uf3UIGPT4B zN91@EyuMec4~xYZ&2nE%D3$sWz$L_6bXU*0wGRJ(Ejd^>kcBf&8gb4^Ew%_h|B#fI zz_2WuBDDqQDc52!C+p7gu`$xiW4*YSPb4ylWz`mZ#lP)=!Bp+eJsnwRqADa}fsv^tg^(PA zD=pJ=8px{c+c~Q&tdUwVz3?`Fv17;MJV!*srsYl)88-A{P0d|GD&z>vXP(hDn=o3% zbz#jv(!-cHWQJePQo>8i-G%%01yc0BLMDWS8ZpsKUE)6Z4|P=f`=tvO(*J(x7~00e zZ<#iPfWH|1X+upT*BOpwj|9gOCpSsPqbT*g!~r}X6b&zPKuSw7K@HwtZ~0*23deC3 z59@GI7y85o(j_EOh|5u#N!@DV-_Q-x%wQ4%x=2f3_(x5D5M!LMRpsUBLNd0r)QEKk z+Klr7ocD_;4^mxukUT37oUWuLc_$*@YUnj`^vv9;!#7U;$-l?2OSD@O=P`+kQ3`Lv zmbHuGpo#aLZdM5)7FBGyeG*27*KDC(sATIj%<3(P6PBbR${&};opC&{fKQk@w;b!F zjpT|*cqe5pN-#Ng7YiVRfh7rXwI#Ukqq_L|3)c_K9G^Kc^Ys$Q;{CrT;a1PU&7P@4 zQa_jOD^b?Bv3KVD^~sr2*Iz_4pI(1q=FHz-y&iyu)A_18I4QC2QgDtv!Xpcb7N3Hj z;_sWUU!6O8Vs`TLxmVtvJM_lVOTf5&>CA1iOK0iG8RksOG-Q) z?}H;*RpoR%5s#%6ZqU!7_;_BhOFWq9xfZX5l3KN}R?b-qP}P zMWcZ=QX#t8g{W+>Jtis0EkrL-5=1~)@9aJNj|)eCK729M)CVOd$Sn;cJoB(Sk@mi# zWpnp1C5)hvY+;<37GY&Bb482S1uZF9`mtUl%tV{$C_!r}(Q*a27ynl4MwV;vA=JyX zJXUD|HWJvZ(RNfbi@<8dsD!f`&K5MRhgVrDajuJG3%}+H#Kq;=q1?wClAge6E?Ts< zN6quStw7d-D%CyGzlXzyz`rhyM(i>Sb->UJ{@)VKcGCx5uWj&L+MvXRSPNO^N)$kb z^9D}?GvOm+l?o} z!&75|6nx7XDwVB@fLN(@tFj#uvv&JcMl-TS*ygC7Bcru#&S@y#vNG>(IRhV`zj%HB z0P5TDv=b*J3;^ogMDj$>?OxS4(Em7WeZoWXNXc`@SO)L03iB|=rvSEa z9{JVzmcBr>)Vr-;^8Ez0%%AGA-{e&LZV7OH(Pv2#q!vnCUGw-X6)dZq?(H2n43;Ch zMt7Q)V6c4egh7hE|MKjwpR{vpwi+b@ga3cVe?1%jYh*|V+yh(jd9+((Q zD}$nvPOjzt#=+!3IyIn(>11MUI?*Sub(s-$E5h>hO7z(X5Sb. z)OhpK3G#EF{?HqpmcuQhGMLp}E>S=wz0aA?Pl0Y)(QbMk@KGCm)B?m4b(%maot z=qCIzcV{(hn!nynOPv3eFD$=){(01R=@zVE0v4yC8Tm@fy8{?3(G0=*^Upu_fDDGN zk-p5gEUh?>T$l8^Pp8*%qFoIPLT7?_f-Fy#y*EtX56B=b;T+8tT?=hJ4w3R0hk&Pp zcc6((h`L?^DhbVoWJJiEUqx(byHWO9w7ivAyz*)4_|%6Ft30_|D#(ClX9&s}V9S*J zb2R+c&~lXqVenmYEOSPc*1FaeC<^ltfl*tZ2wsWOm7OatNG&>gqy*NKz!65#Caf0& z(2RCT2b4LZi6}48{lwYS!p|+Q@<_V(3UGwo|C;8@&xW*BUyb zy^uL+Ng@_e#xCqZMT0+zA%@vxPDG+8fuo-*qk@6kx;n{gFp}_ku*;m7q=W7X%SU+b z*)nBDvmqHWoQkAHck@%MF!#|PZtpuWyYJK4eP?I)et~|~i|)Oc-TV7n-+nXq@&~hf z_AW6yXLj=S*}eN`C*PT!g#W%VJ9%>MqvJnadS!O<)1NL~@{QGvw=c|{{c3LV zaY>IW3hW$v#4G{oTnC;aC%iw%Z{c5UoCAU0?S8?#{$R6+p6csE8dobo($=77Fm;Z2 zkZb-Mo~9+N>DVd=MOMp_TS!po=lKV{?)P^)r{?qJc=p{RE~tEf^@pKx!g^t5dii|# z_XWA8x&9PLb;^3lBWJzLiA&Td+2@`DQpRkIj%o%A(+(C>LQ;uJTk8o}Apd!6nlu}# zW+=skWJ|k3KMhc}!p~rnAg6yKt1P{GynY5R$FW%;Ikf7T0@Vl_0b6lIBcyC1tYz!~ zjA+A*RFssOaK+n2xqVfNyeK@95brF*e%W}BrcM}Wz|M5bRpe@g;(1?@6hV&4^Z6QP z!Ons;+eD_Vjq)CWfeZ9IG?%2fAf}oH#4}g);7+P@K>C*2)0pHn>uINj6BXYxj|}XZIc%?r02h&cZ7eCIOldhZBS78x|d`0pq9bIJ83#p%_k`p4?!=~t^~ zd9#YcyqSg5$l^<QMZF$4f8q7aVM>%axF2@VV!d&Q*Lac@4WF^9K&k0L%VHUswA}Jvm zP=l)n&yb8sEl>h@#`I$rD%%;2 zngznlCAR->A-IUBK5DR1t+JBUa>smnb4*&}G`SbEXkR2KE>a~ZJ2W&c36&8ocLN9H z+ZX1}oVoSR$G#!z7A-fx1S(o0%6o0}8<;%nH6)*bk`Ur@R9)NP0A)so=mALQy5a&= zQi9mhd@stHY{{C0Pjkc3FvHB7RAYiDN2E&)7H7Uv(j8SgObtzygjf{UX|Qlu)(%c4 z+!ah@K~e*6H7jRj%NqKq9qzKG2&tB|xh@i9NpYvaK|C#|60x+HQetT}IS@<7Qv*Zk zzEo;3Ee;JTyS-iJrq>cBR4{C2J^+TTC((){+4B7aE=sT^C6jIKIyk@RvlthWEw;8f z<~;>t3mDdhPT0~*65KfTl}__$J!NLL;2Vs7Vmiq)3k!i9t7o6pUU*y#l7brXj(3fB z{Dq5m_n!%ucUWgz=co5nPfzci-pBtgJprq}Q@vQd0{_0LUPY^_ z?@sTTex-WFf8cSWVcITH^Ig5Dq+_BGiwBO8EPu~0BUXq7atbH=NJhbWSAkq6W9Z>f zjw_t%Bi^zjc}~BD)yt&>am8oJty=?Fx8S~W1Z_6PU<7Nn7>p~kzTp@KwLFjEHr#H+ zc-{Z*9D4*s6{J)oAq}n&P=a$d!-1Lly0UU51hu8J{YuBhBzDfoR!qw%b6YUHG%$rT zfh1e_`Z9Ts_*BnS&sQ%XP+>0c-0|rbL6x~UeV}?V)D_J)RdDsK>Q|sh?yH`zoH6E0QI#>-Q7gwkF(l&kJYLeT=MmhFaL{;4)2T1mIX_8y`|9|~2FFL%Yj_uI$8cSE?HrY~BnqMwC2lT;*saSi&%HT0cj|rrDt(s{HvuM4 z>4T@HFCXb~C2m?+=M-Goa8frNNK!39JU=?xW}`%PM=R%p0fqWLJvjQ;^)!Le({F2e z%$%PHzX`|F8@8Z+W`M{ult8N%?#+K3FOd?&SWFP(yl-F9twFc%4JIzq8Vv9N0zv)0 zQH6Lc@*g#AIHR^U7G@-mQ+l-c!=wGi4cOc(f1Eq~(v3aG!rX5_1XImki4~Zk@Gkd- zProvK&>>GbA8wX{47ysx`p~2VFg_PWJ48i5mW5bL=+%!R&}X7if;rVx4q?VqB0*A~ zD(9SzHGRC1E^m&ZqcXw8B_=>mU=x%ag0>ZmQrQgPycsb?QR+-;6d^7(`_iHv$BojB z_g}m5%GJ3KzYcN7k>i3SH3u#HWR+~eS_!kM6HCzJ`7~wlkWq{_=dqp-pq_zrl`7@D zD|95v?sQX+*`^973y@ms(V}eA4zf*;@4x%*N#BTeFWXe%bOVW&6o>nEjBF!W+on!g zwq}^9Y%oNgwQ1Stri?9$MAve*j50=>)^6!Xatz_AGlqX(dR1nFU)i z%p@RQB`r5o-FZ#|a+)>u6)l^WIZ?t08p#&Mi3ze_<`5Mvjq2UAbRJUtlkwJ~If9<% zrB?xR8aJ|BqZVCPqP@bR#snqR42x$=w9hn&X`*`T(??#5-qz5zl)p+3*cvyaxmC8# z&~j+Nz^1ht{i{ugZfaQ1`L{p|bkySW>CPaK>*&bd;gJL}+3Zw{EYSv%l*_N(c<-g# zNBrlVcFz)3WkKmEuyZesN;lDX!9W@<64Y-Pdd|oexp|6oGLCpV-C8=!qb+2-GPkKg z@V!&%bqtrVW{mrWYb~x29{kE8jn9kAF)AnP(H`g6BWh}!K=w*j6F5b&)&Y@9H}nJ* zbhqHijUY-R7Pde=;ExK=SIlBfLqnsZut9+sMahp3;sIK#8I>G@u}r}FRrIgV$Rh0dzW zaB=FZ%4)Qk7ER1T^I)jFR8~wj?}MXW5ittg>%aa`{CfQa9%l^=W37}SEFT~$Ytb># zU^>e-QF#w`QE8rb&+Ut0!Fn3TOm(~zf*19MF~q$--8-2cCfca~jbnh6%QRd2(gmeM zblF*=M?*{H{nuk+VEUe;rrrxBx`ZMXz$;%1z)d+uQ+^6G^Yn6^p&rw*4?FE^JwR-3E5S$iz<$-0d#nrB9t zn;O^((KD;^^2}I^3vfjuQ_m411)FHuLg02I9cpBvB&6a|Ov%OZsuN?3rw9jTX6RH% zIKf_&8r&!(MdCE34NFMkLaL?Ozu-0Dbo=&=yxq&xS$L?@Vk)Vu$q*fx6H{0CZ(9}| zlG4;ktwM^s7k_)l4phANRj6Q0OiUfaxiO4ZVYJ?`2C!uj?IAgIx}{JIYxui2u#TR< zx-mw-d(ppUZrV)4F1LDJ=oIflmlTrms3fmZ=t`7@*0D_GjiAMdG|2JZXdvxmSC(kq zVtx&@F-fA3YDv}d>_zC)`5Eph*9a;ZIXX%~>p?ksUnW>NA@EFpRT5-%xBCV6UDrri zYjK{23v+c-f5(o;S0hH9h_QtcCk01iggNh_G9y`%gL|952Z!UePaB0_HqbhdfDGQ< z`0`;rH{DuU*D+XDw=fW6118)uvNWqv&e39^+++7Lw~-W!C4E0tTd1bST>@SYWr`j{ znJTYQQ>19kt)-d)N+v2%o^#f{-h;u`T~&~R_gP&w9p3+b=?e>O24Tqr@R+yn@Lm@Z z*}IUy+SWq1y#j@#K{0`OFijGsZjuCppoMsvmcWL*pI30boih!6jKTyocm}s`UpF7f z9fmcUnQd|Dd1|D^C-{_3c?pgagf<_FZ%~cPi&CQ)wa2~L1vtE+*#$TZTR+WX-Q@t} zKGu}V>&OMsy|egEG*g}a!@(Ww%SdEu`+LE#3mwTyH*`l?Y7_MEoRc}q2+=A?qgkFZ zl3P&Xkgd=EI)WDiaJiwk>yy?qG}IzJL#Acuq50GYL1w8dJjGw8hm8r!Rn%<-J}#D_ zEQe&Ftox4}rnu=O3SulJxca8bbzULaBdw6i>jzaea;Ddv=|;w~!_041c`=$S2j<_3 z@{%4%UJ~Ou$tIXqgO6(rK!@7maoc9>zY&d{driC=a21^v}rX~J!_xpsfvz&*Ew5gN41rBjv zr9I~UW(?~H5rhZT7-!*$=`Q8)rrHqJw4WT`Ptp)CkS`E-!UNH1v8{!w+bWVEw)n9;(~)f7Uj2RbPwA<5c2}=X@2y^`p09pWJq_Dn^;=F$ zQvI@esd^D5WGShtsRWwdH~nJu^zIOyEQ#p-!fJyiYE0}|c`<@l@~t{z(>`+gz@Vaek{o^j=! zu1U4uhf|Ag#EL8=qSlXYrMYNl46KK$HY=*2#-bL3_shG>762YpvGshYWf>~(IxI_p z{m&}|r*WIvI&U8mqFSq@1Vs!S4|M8mdDWo7bOV-MY&o&10@r7C+ame|+&_K9(b+~y zWv$lOexc5D{!Ds=f&o3@gc>zGW-u6I*rI#VCG-WHt1PH!~Jek(Kqdo znpQZ(eR0V*KHaLp zRe1~CWYqk{t=_Rb)9>>)pIbM7x*d1?1mGk`8(BfV0&MwUYvg2 zH((>6J{4hHW2jGqHNJwLrdbPAk1^`>2wt#hm9`3lvM7ovL2W7A^HVDgLY1?#^DPyl z7{r5+{B?wZ>P!O|*{#tBsXcX=voM-&!2f}GyHcq+{V0VWs)3x7wE(0vJ-J@Eu%SHK?@au z9SXWy*MC)y`fr?<6Q?@b4Ax@P+OlQoZR6Vte2LYN$yt{WHcedd?`6hgG>5hzxbd8; zm{=5Dl&GRdO4K+nCQi1b2NrxZJiK=D+L3D~uDyNj(`#>|Yj0fpet0uyOb`aDdx>flB&vngX#SEiBRK-XXd3mm zV+FKoqu539c{mDweVnUGG|LR^P^!bzXfxu!sm_2#FzeTBGE%H0Ro6}!q{C~R*C@A& zn4cyT>tWRc>-bswKF!=7Wn{J)HbcxPVcp>|-eOa*zBJCr3(=RmHS!e1pJX6?2SkEDo!diwIGPAZ;<|}}?sOy^iqO(a| zTG~XvkD4viJSUN2_ zQB$UK6!z*I@n6?EjBQ4<1Px%8F?7sa%!IxR_2Rq$Jl?{ix)Xau_NR-d=05s-?xW*# zr#_k8_vY;0i?jQFH@o+1=YLu0}WC{`ltc)Bkhj-R;{4`bT!`_@65;%}#zXyJvFycK7e?*fBf#>a7!} zZhrIm?BqMMlb_5^?w{TJ%Kuz>b9VBh*~!o6PVc?_hds9@FVrr-edIsjL{u@OwJ{>f9f`y7|FdH&1?b^Y|NG+#()EI=vU9 zd_RhFUtgS^e2>o^Og+r$-o3Y8{B-W%0Y3ZR?gMykoVhf+cM?!>snJR*?$V@duRK1h zsvyOqI&5y=a4u68UR!X$A}%PZXFkSeEPi9ykBt#9UNi&qpc2+mt=7d^?sa3K%#_EE z9+HwA7@I}u@B749BB`qW$vrmk>DX4wCTqwvOs48unRDzBucQ!51U5uDB8J18nPYe~ zTz_hykP#Z74D&C<=VB|_{GqMDOx*pj5-Jz8!gFzy8qzWt_f*kGdm){GJD&3AA zXN{C>--`jxFA9e&q+-qWB3%eXEIWICcA>P9N;kn7Dv-5C* zHXAc)q}MWDXAdWPhjb0bY3>G;EWC@F+%d>Ifs)#JfT@DApsrFl2j68d(vBVE(FJS^ z&N0W^sbRDQz9>M#&%Md&5Ws{WMwxw$dyr*-1wnHpv~4IUNJ_I_5OqTi+70>rD?cB8 z`|k1mzQOEXH{>`k?^jwnTd2sf-u)v}|5)&c@63ri;Bm_({aw05SJsbfm@5Ha0i-2yV%?$ zz^a(w6a|VLIFPYYHm_%GTr$EWIZyCT;mMYk)sdsLfePz|6LW7KyS-=Mt@EdOishmG zw@#i3*<+gE-f+>!XEB&bSMMcYty&Uhg$ z&SKpvgu~8~;GM;lmg43ph8qyve|-5Wj0WBL;NabtKMFC9o8U$3@fKWz_J?``m%;Ei zc^ZhGw_&JDw?)*T!7wScb6a%QB}%HViIV3vQ4~Q-Z%!N=^;&+?`zv()>fc_ye(?H% znJ>`HiR&-IfByFB%o%=W>A9F@KASlX_dRp`|DL?Qf99-j>3POrS`H1^Wd!pabIFg~ zo^4KEgl-O?k%d?yDz0qWu5=SdLU~ScIJ=do2uc(TZ$(8nxPBKruGA?kaY1TH5ckE+ zXqXmSYt*uX&?Y+CRTud~aqw+5a=a+f3Q*Hnu4E_7%eO<~7Pu5kiq-Q%+O!5AA5kJkN^k}RHbBltxg;wS17*9)tEB)6( zwer=puNzpTY9PU>0@Qe5gLtJl0yzVMMoXc30wq~UBs%Ixy*yB;JrtMTFQHpK1asl^D?rtEotA}PJ3S3f z`^GmZ31WhK_9Q`;T+TgLU8r;H5ok$Jqnb5`vBShUFfJ;*XQsnT8(sn|z@iArC=%t) zjWD!|X|ParY&k9@BaVUQohom@OlX~?taK_XNkLH}1){<9(Rc=bNS2Qh6U1axQQ=;@ zSh-MPM!BGoOpwkq-CM#GNr=f2L06-!*trhUxH<(y7SxW&IQuve7os{E8Y4_IjB?aD z>H$478+5RGOPn9kne#4qWZQSmUZiPuf){m)QWRZm(m~)V_-{!1hY}a$mfn3U5|V}e zjuBI%-D_kX43&hCOf@g8IUhZLHjxa`mqLn4KhbS#!iniF_*du8r280I^9;4H+1tk~ zT7-$yt#O-;qCQ>cY&n|$J_z@|((QV1c^itd7@);kt2)x2E^`sK%mgnujYXBfP3nl#V$oQGweYf?R0VXB zNKNE!T|0{z#VA7=F4+J{(6qp=ET?fnjYV$Vo&i|cU8F1w)RhqVCK9!Z(qN6^uLMs=F?uVUuMd1Ek_m`e8{ zFJ2@IG0Cf3}avL$cP`g+|H- zvrM!V=V+UDI6cqw^)zDA^9(i%@H|?EyNWxh2#KU1$+93yaz`qH7&!x5l0NJ#AdQPO?AYuSr)8XVxbJpXw-n`tX!a>bF2C$4jeE_8EkP@)vc6F_&0P7 zo0@BeF@vEG^ljL*sqd*lSNedf&%0+Qp%iP=B4Wi0jgf(s-|SL^E0D}ck{~OcZEbz9 zBmH>b(2gBj;re-MRX$q2mu=s^jv25kiE*^m&VyOCiQv3V{%IS*B{N($PXE{q!}?ai zXrzEy(Z440gkjdKO`VKqvB^niN3pB-ic;4U70;?7Z#LaRDH9OB0R~$B0(OpV{Md8P zaqVMX;!r;*WBT*YKlZ4jHo#k88?4gQL$={`T4uR5ybhe`j)uuSv#{cu1wJ(<&(C#6 z>t8Zy$t99f9aG6WrmBz(uKururBZXO=;dfYJ5=Wu#Q+cqwd5`Dl%fq(P3&@;0mMx%cf%1R;<)p2ZCD*Mi*|U|! zOTJT)W2j*XvqEjOnB>*Pk`iI!Ub{%y(5W8k;*}KCWV55&bG&x@_Cac9ZPSF?{C{nW zYShFQv3U`75$lFY&E8$Lxe|7oyXOxI_lHVW?ZP?K&12IAPZ)~MLMat3cM8W z6<(?=NU3Ir%6(I>d69)ddzuXR8}EF4E$IoPBBT+&ec;{%P)mbFHxS#Fd^#SbDYcv=*CfQigry2*m3m8TT%cDlCJ6ZSm!vBcN$m zw=gv+Mx2=$vm6|D$@~bU>8iZLSL8bB0kJ9J1s(y@80X!kyHo}d0GCuc20Ex11=`$n zD7I!;1x9yu9480{h$tjl=&e2PVPFvAQikSjp2x=Rymgo{c4aw;p<|K!M)Wp)ar_c?h*uvIAi>9tB+N%zSE zE`?Bx)qRo9u}7#B0nSBM_TpAHo(R=H2{5T`_~WsU(uJAh*Z0qSHuL56muEhiIX?4g z?ZhQ`c?MnIJ9A;?1RS0WyO6;@My`9mzVG_W*I$6c<1?S|6TaaZ*4NNd!LUf#rG3|> zD(+pV5`wIDq|Pmms7twE5bkNUER3=RRxu5pJmr61Y9D*4?@Y4yohd2ATPXh{VyK|R zP#+(@^ZJpW4`1|+X!ls3QoQZ7CzITCE9alsxhv9+r*7af<}V(Q8Q@ z@$LLdaoT}a!(l~Qa`Qn~Ml3rxOMW+W8c}XTHMF83DPUrC(BZHvpeRD3rBRqa3ca(O zcy8b1?7qvh`wluJbf3-c`*i84ma}_LIYfGJ_xtBQI)3xW)tkrPn4SFm=8?;DADsK? z(xsa(`PVUj^VG%-PY#c`>^4`(Mf4hSN%10cQ4L%)v105tO$#+Ln&LI*V${s2(SxW{ z&B)ps4ynUKszwCRtPB*qc0=60eIt(pO`T<+)4?NRam{h+q{T?Vn1Gj{F6O3AY88q~ ze?3D=@K!NPi$zl0?dM|9@{0F<2Ekybj*S`G)G2EWqg5EyqG%S;9+DE+twpup(6bYG zO%aNErY;NjXrHCm=R_K(J9 zn#~$w(k&;+_F|G;6pEw}m!qZy-SbV~{o`kM-w)R}?etxbRz6p$>C`)@Xu|-R7MS3) zL+wT8+@?{FBbV}Tz(j69o~~Y*etmlH^s7s6C#oKWyT3gB z>hyu?`RbReQT0&uo9gN5z16GL%hik3uTk}I^-H+q8#wW0_2Tr4{?$1npToK_wkGb% z_^#`7l4qb)FrFj%ahJ~G zp5o2zQ!NPyD^qypHy?Y{Sd+ml)4l5TU?P+fk}a*=?$3C>QN!~B;O^aB%3#yPIfGF! zyJ2$f*8!zx_%}UGxWN2qxMDiL6Mikr7R-vM-&j+DTWr$^rALhQG@{Zo!aED)4W8$? zP`i>)1yNB1QA%)HsX9I>xz4KW9D74Zi3!P;tg8E08EwOGwx!JI$0MK2U3~xcdoSF0 z@5=2%=Wkv9-F*<+_Dl;x@Qve6p;h?!%)NDJ?$ArOPJA_Y^}=5+%zgCkPv5+E86q_h_E_I67QYo5~eyyhwPD}GR3JNVq#$b`f@Kn$>(|HiD{8> z%yJ>N15R;z2sIH!C*9G>F`Uif;QUk_y5nHEOeC6(pn})-Ti4#YcH-JQ*WS4H))F&* zuAPJ%KXUD(Yo8*x`%kYOK`@jf@bB$wZ^GaA`Pnz%#R<6ZE#JsoJLG)(Tj<(AjN z=Y-4riMhziW1h1w6jeyHWX*f_nvOaJ4xn`}F_cC~*3ffk#4f?NU>%WNe5^dk;2V-b zJVtq&7l!SO%g3g)b+IXnLIKJ}B1-J5E7T?~86n39m86i2%H|q91X{(&1rU_3xChjj zAhpmlMDBqGh5DnDKV3dDcj-J@HTUW}KVACl#?kk0A3l5gofH1MadyvJpo)ST^&4yo zAApvIJ4uu6RG4Hzjzu~%7&0VVC>g9^fk#)VjCRTs;m5DwB=5Ct30yEtIQwz5el_kXFglpM5FNv z`Er{vW0=w~4H+@U`9BgZp-w;9Q3Ez(JTc0A*2cM|7nEMR5p6MGZ_oM0V8g%KSaVtS z;(fTD?@#v3r(bLnEF-4(=^xTjmJ`U~CD4feULq^XIlPs0>Z}WicJHP~U>dVoe#l$- zVP;L8%G#P?qKwV!u6uQ(M0K16gaRl4!4-a-K!Lj_cTs2gUZi<#vn zf)h@4vj1~HR6$BcZIBx67pm)aIXtFr4ogB170=Pl73rh7edO#%%qWH?Mo zL2hY`?m0;pwxoZ2`OV$KlXs8rx%=LsJ7+K6`D)L70?-?`wgjNJlCquAXqJPy(O~4! z$Q7ui8vlIct-F8x?9SP9KfiPM=fiLO{O+s1Q4-@S?*bwuV?sRP(jENvVCFETgj6c9 zk+a+b(78Mr#};9$(FS9@ON(A0{aO@-q#C6XHrk>(<=ua=Rt%+4B1c6MQc9$A5g;W$$TU=fH@U@>Vy)pO2*SEfSKSYg9 zObGGL+KVm^x#?#(2ftRXfVq<=+%XWUy(1=tM51|r*r!QPlfktOer}99jUV9t=zUYNM_jES+I(mDcGgyS zMG5*1ZkLh$3z$eTNf6_4K~V(Rm5D9Z6R!B7H_4SGK~e*AFmtCkqZZBb40)0zWk)=h zoM*gGKtlp3AUb;5$ons?AGRlG@d;ZmR@`FZvmGMGLII08GN2liLXs!&O!+5V?uuS0 zse+h@GIrdy6T_rru;5&S9okJ&Tu{YMVt9Q#TOaE=M>*euv%D!-#1i^|RS*kilUlVx zsvI@TwiO_R)vAdo?fFZT`_q^3CnzB#lYv!Lc!GxEW|@wLu*qpz&;Ygyyj3w8DA+7p zxSvMyC~tp5yD;3jZ$Q?<4A4=stB#18s@rtVu}7g!fesRwJ9LFmMU{A6KMAp_AApi$ zf*cF1pe%bwJkL2o7Iyld26ilV1t+i%S~872uW|4j2zENmr^0E~AuO@=Y&~HH-*z_P zjA8p%9iJ>)xJZmK!d!ce)SfU#S{w&ihy`!b#>vbX9mCnm(%VvMRv|S@9EY`VMaD3S z{{4m}!b!FCL7Ypvx}&)IG<%i^MNop*ur5bUe(D{>3_EZ`u1;4dRZ!wlg2ZA_-F)xj z&9_d^{qD`5F1{3^$6bztz8u-WJ^e<7x8Xu#@RMekzIHh&C@rM!p1TLloVb2)=2Jl9 z`U^8>XFi?zeCCVmFQkXGa-r90BJ}CZdH8$aZ?9gTbh-+`Wna%6hl}>loZRic;`+;N zu0X4#csE;72Pvp-f&337{r zBeI2>Y$2LuMh1tgWGk^YTUcZaH5fx?Rs&sJCw727QC&nz1f-(gBn@s{m~+t3L&~>M zWI=9WIExm!rbUjFG7Rf*O{hK}MG=xMnM(db`+Y+ryMH`zXvdBLMtHMU#D`i_*dKfD zxlN=@bG2S_+pt#Loz<{u{@S$c%x+rZRM%L~l3#1a`18*{_JFM51?W09g82aIQ1Vyl zfdVfPSdoX!b*tY zOeso4HPEQYnV@Ym8kQob#swwOyuplmb@hG9NXub6`LH z+9Y5tfpJKjel5>Gj}X7=r3|*S1+rVCd7Z#Yk%!5*;LYcr+n|raT<=1goOgz6?;x~W zH%h4C|9*||7}hORHcX44oDBz@?!ZEtVN(-6K}-Cm8a@GMwW*VNPPrin4=PgZ$v~@J zELlHU6i`ugMu6Wt8$yi7%dmIu`NQx08jY^|sl<-c2bp@3G zD;l683!>P(qD46v4OKd}&fCciV;xZ)tv6u28;t-JJKajh5IF-+tCuk6l~cU;SEyf` zx?&NXJDbp2HgzRi0Nv+Ty=d#h0j66%rF9IOYqn^-Yck6OpcPwhFi`Z?v8j(Tx;xR& zMT_JalDHtXbX_cK9wGVbB{p?7_t(9+x6CFgWolN?s+_@lJE8H)IK??MfrhldUdd2m zefI*|2CIhl-3x0r8(I6W`@2^T@Q^ai=Ldpo4IEM1}`h%I~=e)86bb~!DpHK%-lS41Hl zk7XQ6=_RCN@X)AvSvGEIWg>4j8gK zbkT{RP)D7Y8smtR^V5^ld#hi<@oVtvV)ZP7#7fC4UIv)~fQx(`eCoOB~61St_Ihz$;BTPssNoU#>> z^9m&s&4!idW>|O5y@=8P#+|SB-}(4(y4FSt{ruMcJI9WqyKi6l$NA6D&j&xdd+_j& zdtTeI+WSKaye%H#Dpe>f*1N|K zp*zPvg5lix;Nac&{Y%;1xA))uCg2*pLARbpT>ceIGU3#-V!>KRxs zPVO(@?E|L!jgD}FZHWY^(|NqPglQ$f9Fv3DEv9=tfS}o%d7C> z^7JeIg_5pa`{Zvaq7$ee7j0KANWCUV@>mfhat<3Q71+1D;jt3pHDB-w^cJmx!1}ac zN>G8IacrV_vQMjkEd>ALnTz~JR-^F(%@)wV_rOK}-h*hVtWk%$c`2dI=`tuTp$!{0 z@KnqpJ+EQie-qq6wvqCrCU*gP0C%t=398tvh&(Y=TECUgS4JN>RFs*ArWuR+1`o0qafcndN6t8osR*9xfd_Zy>Ri? zx8KaY{J|1KL`)@Pb6=nP>GFrO`!0*A)WVC&&!5;jyKm3z-j8PYeRSikuWy|C&^J;~ z4XxSoQ{!&tY8 zhDNY9hI0l(PZ*}NKrEet;`@!jZWI*a0#tMOl$zfz`R|mj#0_hEL!H##xOhE0Xium zDMn$EdYHt*l{~RJNV4h1Mx%;?5{u~E-O1LRS}^$T-zgqa8&2vd64De2!Mo_FX0TA> zAgTc*wfYVBrcdvg-dDXeeV}@=yRPzwQrlB~!}0TfQ$1h(*71?oyjvF$ zjEH;4zlL#f@76^$y|;R?`Yrl#&mXHF@e}aoH$Uz<>R+wX$S%xPJ7G{*&nZETu{xHC zc-^n(9D77l5!9BxtC8F_%bgW|{85by@pv;*zFe?x&WEPvJ(TBaPa>~Hfj9O$Z(gZh zhUfVk9^QJ?rLp?Q>KP|K^%abye$NZi+2-A}d)Tn%?45p{i?44?w-D2$C6GTU7IQU~ z2~ha2CWLsbBhCFj61b$~Y>?Yy*5xFVD|p6#^P|fzAVb#SH+YN zYr!RoWCg_c&hOshL%DPOqr0yhz5DVfa-6FlPM8)ym;eBXbMCRHIM*H14c}-D*(F0O8*aChi=rXvag&oc5mZv{sJEc= zwAD(9G88Wf$>J0yb?o@7^uO9TTU=)|yLSlIKlm&FgR-#vF=7R2Gkov+>Il{Sh92F9 z66b`9r6R^Vm)nr&c6c5`iE}(DEiF-3CLjwQLtD!hcw_ETIOGfxCvl3Wnl~KJyj!q@ z&3*KT+xt%3-t*zj^M{tg^3@2%jf01Nx_D{s^y|02IQP>xN9R8OYVOc+-$<;rQNi7G zuAYmjdd$THAW}j+9)+Nqg(JA9w!_6r6oq)QIbW|(@h%wu=i48S{lNd7|KZO+{0I8s zwiy5SBsWBD#%+Ks?cRab>)N-sv6gN1`I6TGC`A9!NJ{ zw_f}6jq?}gPWs=cy2bq{ar&S(9qGK5^1$l&mv8&Dyv-?nEy9S2a|R+hum7VA&T+yQ z>{Ryr%MyrcJ>wemy?oaN>W2Y5I(4$Qk7WyV%;@c_*g4$W$N3{P-|#(=9i_R7YhP39 zE(Yh=BX*p5GH}3ZnXY@VKrk9s0+Pg8hQx@S-6_>BU}EHrQn>)Q@G2)Oa4}Bjxkpl3 ziV0$r<>GcCz&2XpnJxxwqv0jGN}PWwB}V;O8kVtIm4zABDkjN7qJ`;kMM6@yVOfq6 ztcPXAN{XPiEMMI&xgLP4nSnNg#&9@xBfBqYUK2F1IDSio~a zkn?jKQ*4pE!76E)<<^uyMIySRFTsd$Mx$ndUvKhmczvVHLG<2&`plN}`hnf+4GX@b zi$Fi9_~ODyU2uCI`_QvCwhTt$nOJ92odl>*q$Ss~o2Wa}6P|tPaw2jhTG)H(vz{cA8K+_|PqD@#Y zV%|lQuv#Ed$L3DHOAy%oc>}WlpS?GIZsR!jh5wblyj5@H9CL`-u$`-WimP^uN2H}3 zJFHY=&;(i-Od~xrq(ISCT5Z|RiIaGVoj8e|$cy4lw(LaStHR`~4^(xXqVD;0ElaXu zx$-Z(PY+08fC1^50c3E2Y>6|20f^~ecR%~@u}iqJc=!OK2sTQZ7wL$wS^7TZKJN|i z_WFmcck9IJWmH*u3Elbi@bPzTqYcH75N2YmK15@3tT~R&LlAkkjOQ7j%~5|hPuJ1gM^HJBN<7Qe%NRZ}HB)rv?IzMgBRAHH__^=qe}ymspM*G@lv z?evSAAqT1@VesO+H&48N?c|wjC*Qnw>aoa3+JJCnfH8332Lml&3>+ZGVAmdK0c3TIzti38Fg42?aysP`@loq`6~VZZ3>0TX)IRZws_Ue{S5ABN z)$a(pgDWRRVKp3`5=G{hy*lfY-h~r&{^Z(4k3@K-6dgVQ)m{*E!{Uv#_dPo0PXy8M zdBL4orw0`T5ph_CjUp+*u*@MroItARFg3vBW^=I7`duaIRC`}ZI<@W;5S`Wf0<50% zR05%e5v~vi$JXzlDIb-dHn3fpXau`mQO zn%FZkjK*EgDg@mxjMG6SfkunCT!FE+i?k9MjWKK$2#4p|`-8CtCNXLYb{oB`XKmPnTzkK%g zFE9S%z1MI5dJTq%fqjPpUei@7CV}lf#=Z28-aAn1OmyEo_Y^$Z*O%UfNBid4pMUeK zpG5biCUr?x6)ByNl0HwW+E`qnfl5kR0H}T*b8{XujEupIQnwWB7>1B4s8$lG*!h64 zq7D;m6Xf7>jxJrW)>ulT!KF0HYzL!5)YuqnXbjKMJlZQ>kBkNksw8LNudGP2o^Bwb zH)0Ta$RLD=NJi?GU@CZ-I@mOp9-v)`z;FJ6HPg|PG5HS7fm{;tl!n7TxQI1+63(eMu!KQaTyrJ z@VJay-cYUJlM)LjCCMPyQ@(CMa-(sR9I|thhfq4zs?LG>6?!MzrKHdxnkRz&3LO@0 zge^zFgYf)@cT|s2yX4w7jJ9t}yK^nOd!Ht!qDR`8g%|wmCI%lN;377~Ru@j_zGiwx zRaGRMxQ0pWd(aLq1^oa8nX~i)I?7zWl6)IgXo)l zs)3FFYPRP`TA*n*eMnHh@VdZk!3SFdxY$XSKuXJ*7JhKWJhu)UY|ST;f<*Gsoycc7&QmVwS0ek?29&m_o@+ z!(#M~Lrx+v8B?v^mbHzOQOw2TJD#%G6xH0YmPk@j*TZBN=2axrBf`;+;WFVKar!J6 zCpm1IZ9J9((?4jnPo~K+Y&V-GDKIHAlAE+#6XT<3mWpQC#+}zUylT@Lp(!(ii@gm^f=y9H zf@Xo{MQ?Cy`X1|E1cRE4v+H^7wx=XCg&mjs`enG_PiDwEUp1@Bopx#28W}rw znR$uS)G}hxMYiXyi6f;{MN$HZprLogBbZzD&26ee4A5q?qYQ?d=3L?Mu!od{BFSn> zl4VtrGs(r4bJ!dQ(f^b;{8lLX9w7&tr3Qv~E@fMuDLR&>OJLVQ_wDjKA{DEBO$e!1QU0{;F9!JFRw{3W>fnz(@e{)pEH!#n>yyj9!+|9k>}Ulu3#Yz9wt^YZ=t ziO=7Gul}z0ROjoD{r8jT?+<&U^eH^%<V;<9hzBpM>qI}SN*bSbU%=@HjYD=LT~uXu+Q{)09vea z7oM(%Ho2OX)Ig?JXr_$jNW+b1E?ztNOF=ks`d8OZe{}8C<*bM`?3tQe9G)89z31BL z|Gakk^tIFH{`t|l>*rs&_3R(7fAIXu!R1r`{OHleBD0;Hcvg`$IhmI7RTx5QNR>)O z!Zt3_L5nyNT}Ll5;dK2X1DUBs`SyKwWVSHC)O_KOog`|_h-eRbjO2L8dtmxz3Ax`BYCX8FU;O#?moD9S^|>3*zm9g?dil8CupA8$Uh;Gjj&m|$&n>e+(P(5U_T%IVsr+9$Pl1jkA3 zW1pTFy-GmSrA(r)CHl3*1q?&Md#EOADJheQU97fklml!SI7a42_jCh2Atm&FnSieV zmzK-eaLpdWB1@S7RgF!{6*H`7-zYo1GO`2*O< z3zF3QAZNDilPW2nTuFtNEV?S`$v}&t2%58!0A);MkGs6OhO+#?fq=67fGC677zIru zoWKC7KvuugM8-jw911AMX-U?63bL&F{B>%07<5U|`YZwd#HS(%c*=`(A%^nRmHa0? z35H9nfVbyDVp3WuJGDjmA!FQR1$&MWPCc(rG#K_X=P2=IK=`szAVf_hHiMf)gvy%(=hJ>-%DL^MfF$IkBGE*_HtC&VJXi>7FTJVsB)P9ZWxah1SirP>xT3(Zl` zg$u-su5e7vd0a)d<3(6RcHx5$-UqLcJjzpB@W4AWGxwt5mA8ww$f!UZubUy8gS)5& z@3kDWDs6Q8zZv8s*dhhYtb0-YaifPc-7>aRfXgSbl4=9u`bkKTlTtd->f<-UM!Bcl z3a4e7$BhhHinE8{i{}})XJ-Bt4bO^`_OBz1;376%t6I;e1vF7Tk9Lz8OoSIGIzWHy z5))B$SoBSDmfnaA>?C##+?@O87kZ&F75%LGYP!$WX%Rlh?B4k~0h3H8LT?LU;xnCC zndN(>gvfHAWVjU%r^xX!Y%r5ldQ)tu3G$OvAjv;8`N^CtsYxWrOD40UQ;VaFuY_%p zU9}v1Ol;5~Xm%W0x-h$RffLIhh=@jLy4XOw(L+$Y@< zNW&b&$Y74}+|sK&hFfAknV?PVCfXYGCfYjuCYoTt&{F+H<`+a1^m^Nf-&V03vPjmY zOh;)_4y%Y#-r@Wl0qxv_2nL62c#lPku|;#liX(sy9Tk0|edV~RR3p8{I}V9MN>bE* zC86In!E&zMBlor@Xd$%tRk+(ETb>-F1-FO&AvD1bkZ5H zmF)gFF?-1D z0JtQjgce&b4#6$vWAqY9V2d6F*iKM*a*MPq^7AoRq&rPE-{T^lm^p0?rZ0FWGm5;6 zOkbMbNMP4P11hTOVk>l8rYxNr+~uMXj+w_f9;*Xhx4KAZg03T@h2~!CFS4~4nMDMD z8;g^6(%ogICL~pEb@LP>)QJo)AY$5o`?&}ZF1=WQC6;4+=>u9oJLGga^TmlL<&3Ot z!jkg+jN4T-&aJ=wJbIK5!&ul0XfJFy%$MOS`&gwzuR2wh(w(UsTWcq@^0(sIEeGv9PiJSTLCxq)Hr?|vsG5|@#17YO6IW24lM1%#>5o12 z2nm{tK3`~)h3i&`;AZQMfvct@y_`OI}%EP(EJUVBKZ9ZDv2~G&g3o9)$D3etD8d>!jG|@XZ z03wQ{rdpRtpR1sWtKj#iZ$AB#n}7b*mzN*A`KvS6pTBth{Y!BA)hoaF>d|+upZn!k zFTNTXzHU($Xo77b9q7H$&oTDYw?>Y!!En(9MZ2fHku^bUkka}A3V$w^@KT8+gWl+G zS_&jm?1CcyjS}O9A#paen*K|&)0~wn#Br#7w+grKQpwQ>vv5VEvrFb&WMGuxgaQ!A ziL5pTL!6#SB(myQdNi9E)#PkCHIYpXs}nxp8n}gqEiEYllE~1C#Be^3Y*r@7c5_97 z^RS{eo8|!GpK~&}mtYGWz%bVwC;AG4??kvG4GS9N1{}#o4Qn!F2`d1Sk+gJ2@rNz5 zvH9$vf{AloO@fPM3h4yjQnb8!nTl4e1}MGKx;lkuKqY%z z#444s>2<2LhVG*zPx#WG?%XsqLKY}5iAD~D=5qFsXoL_Wx1qKFL1c1EfH*BBI!iR! zGV5>Ef^~{j-m?;OJtL)**jx-=lF8`q@zH%p_Uzj;GlO8lF^&XGpbeq_nUp1jCSA*b zMd?)+6&SHmj=+lq9U`;Dt1Pl5G{!`Gm^+hw*Z6%CdwwuJwOAk);n) zaF#QPmtbu>@aDm_=LBp3vZg0C`pR#f8}Mm4O3*Zo;nyA6C1;w}=%Ctmk*aG?3SH2o zq}5ihM(EnI;d-r$Em*=Fu#qdG*_&t$WVKb*>F>L}-gjZ=LZi(3u}kde6yU+V zH{EvcO|LsKl9HV4XFBOfL+`5Q|h%BYjt;w`&w+|oTjFrr4bMx?_@1y&eXs<{-f+UbULJ;L$ zLqIi0xv=-eD{FA+Zipu1dO}_gu)~d~2;www-+v4=<1_%@1^ox(BH?`ntz=q~Ga{&^ zN?M`;n%=O`^^hjIE|?|M_^OFvikOmHxwqc$6IRdA^rGO0*kRw}sT28`nJgL>lty3{ zxTb@?c=#9o^Lu-}3cKY2-78)}2fqjU1pHF8gS&-X6}cAR7@fs>4A0TL0dO#1s4cSp z1;ACMq}*38TKu0}G_dJ}m$Y;5r!L4QI^bX2CUG{yEyMG4QX||IsFIQz=yMf1ssj@3 zt51J^>(Mheo;&sBhksZ@aP6H82(G=jg01gFV0{}Nw_bbp#&3Rp^QE6f&+d>jqC=9f zenDGHHh{?Z{2H0?ZVfplr3209LQ5-NM#qmI9)9Q{k=`oU6j!}C@W1G7S}&Mmvj{}l z;o+h2t}U9o1yh?kxM7EwG>@bE5=U7Mv@X9rn{02;CfBuS0q~&9zr;c!^!2-spe*cP zU;pXM*DpPr{pRWC|MA`%=$og`eEs6{==Ree{o}pY(KqK_zkTl6FHZb)X6E*9&g}U5 zrH}4uykpHncmPrdeqOQ;dqewatdm99abo@BB zj3VQUqb}#3H9-{$K&0YL{pgOsRc?6GO<-~#l?<&k1=cmMHE@a zE`8lD3Mk`Tu-3-TIPQk0OVK1%iEW~{G$G2OahJ0S5kz$ds8IxM0hAxgH(fc*ZrB|d zuL^YyHwgD4!xSK`19jEcqsTi636m~mQmw{zC2V&10c>KvFmiA@i{@oizITLo+$AIr`T} z(2f$b7pf%~ia5Q$R))|Arq+N`3k@U~8G1i!Wz}m^wS+GgM72Y@>naB` z4#pQW)dEoTEGsh72L1a1sWri03HdoqIgu zWV)&;DH$lJ<`D?lYA;?ndFAoihoWcw8iAZc_Xh$wheVWNV|4Sa{r1YKE05IPL$!}< z7q6U)3`0s#vx88|D3YxB+RzQA+MNuQd-XAth3qeK)-q zdCx7HYzKF-TAQf3Ah6s2eqbEk@SOu;OW^=t z2k?;W3A&(#k+pu}Yd~(=ZjM?GDfHBpq9vttpnqEEEpDvd8nleXgII#<%M(dibo}@M z!`fudefWwFC-6YbyBgK&$jX#7`SO*9&ru2ObJWsOD%D!?xE&XZV7+7@Y%nT=#Bo?m zlHabF>N+Tw0bbJCunBd`Z!BNDdJ-+abM@@;>1su-&%fU`Rz@n55TuSzkF%= z($z=M^6#&ny?So>o#l7Y^84^iXa4&qX!-T4kHAyCg~Y>MeR%oZ|0{kzB3^mZyZ6%a zAH@ri5j?`UnMb2;1pzL;>~UGvePPd#Bx+uYd(3_`xXy z?aij;=(bM#HVqALx*9s+NeA7!U%cF;Tib4-KU*T`c%&_AkbBVs=F%I{eQVuvMJENXO-5-o{x@5caH|FUm_gy0sb#S_ zZ`e8y_eHKQftjJVn;CYf=4J5uwrM!qBHb; zG(}vF-I7g2$s{U)%&}0}En#f^+a~i;-$x6CgYdxAE>V7zkm#&v_w%FJDq`9ICE1W3 z$^0&GI#$Ebq@Rqh)3db>sjTHsh^4B7_^EYCzXYwz>+ zMns;Hja`T3aM40aDhSNL)b#FwW@K8%Wd$N#N+l#Yosg2geo%vHy{)84>+8a#cIueh z8rzsJnm|NRZ~WZ!jyymBh$iXrMa`B1WVMpRWLAOZI5r%DyVV>ZmKq@IQd(}srf)3u z)>v{dpMuXaN@#~ounpsujGm1bOiquXw#z9=?X2o_Yd8xnR;h2KR-qWP^uAl2r~#C; zl#yFOq0U-*$gt_K3bqZ(TGfzIF(L&KXLcwef~()6YiJ~UrUM$u9tWG%yP$|fZuPZ` zV9{5A@}+msBuF~aZbu#1e}AhY0=vZ=wG4_;o}&G(;kkfUYYH-wqDm<#4bO6F;F*~_KpvUQ?jBv- z%^diQE~0q1wDHv^ZD1Mx*#)!Md<>T^E&Bvqj1HDb!2pwrp9iyr*4?Vs;cz@hxJ`RU z{*n~rFKMm68Lp=6;~ZyJj|ugY9{-CPh~uu+#ts(0$lA4A{d9+xF3c`nSgA2LTocT7 z0>iULz4^ezp@6S$f{VJ%jo_?1z{tI*%tXsju${_F7sSA{>p7B^Tm5y@nEN7z>7=BH zPE|7Kb^hjb$p%i|aY&$2QbuoGJ=?ZP^7ThPzIFc9-Ze=EJTj6NzZ1pQ+0Fbm)U3;m zt~64LwUueG&;mR~(ISPz<5P=YJp8knnNd!#$Y$_b8+fV8-#dPMKM~Z$W}qS8$A$4VJrmE21WqNoNGq+P#mFVk(gy_ zVFZU}15d4yCZUC_kp(;zDbW${$s}kz%)%xz=`7iNuJ65~Nz`bOFI{jbiJpYF(y4I| ze8?_Nnz)*MoB(oIQh=nOJJACjZ@2l52cO}@R;oMQ?{rD)*Dcix$%IR_>M_q-voPy> zCW4YuI@PbgSSx4sZ=YbRiuUa3j=d8rZ&gc6i9|ou%kvCS%nKq?WIJG~Rp7Zy)2ohk zyJV2Y%Ei+%lB)LsYLYS)(bTg?4N8-Ot3=E6<&58nO-#`8J87`jDI~I#2_&rM$y#2U z2V72#Vw2NmALE>q9GBnw@2+swSnd2Yqq_rsnkg`b76>=7mA?zln&_bHS;-rGCoRHIpf146CMynO3JO(Lsr->CSa@VqooqzpitVTOMA}K7{Y0G za1LexDWC$$ahJk(Z(tI5DLX=s&R0t@C^YZ9?+%+;qcw?l91;ilZd;W zMC^)ok_RxS4F0W8piT)^qxY>z3fdaEU--8gmLR<96gjK(R0q_PQnKGZ?@iz_ zz2Bi=vtoz1KK;XYF$)80T9^-waOQ%28>5(cZz$Mc=hGXR9Ll-B#ILX&R-2DcMA)R72E30|Uf zd=a#@9j=^IEs7Cmw!?c_!SDX{CWls zIDDM~$CV=4TzJSjI_gpj5ewNO%_YX#0Bd3w&T+uhw_`Ontk6JYq_o=aG3-~sgeTrL zd!>KLvXlr|n|)7Bi>HP=)Nsw7dTZc+MM`E`oob<54lMoZC}L(FoA-i@=MFZ@IE%hG z@!R9a_v2h1^Q8|8OP3I}W?=}Ap+)2n^C+>AwRB+))&K1Da>7J|_T8BCDmq~7@_sHN zdtvD#c!H%1HbwcR7h#y{_doccoKaFUGZGvfUHSm#yN%?8G&9rkp1GwLLCN*rFy*ik z8D!SrG>7l)yYq_^FMo03H+RXO(Y36+mI(yob!V7%n2m9c&b2XiczPe&w`X{4=}F%- z(iNpKk-U?TQ%RFlwXe_DlPGf7fKjOT*DY7NRbdfJotJeflkAKuXZ;M=JuB;*Gn@yV zz%smW*WwB`JIBWFM~mXMA34?m31eIKX}Fr^E7Y!^gq%G|m_4ly%C%WrMlt0K9bm%tV~#_b z(hG*ZkES^;W5Mlb2wGC?Fyw;%2xDu}GHaJ@u#p*D#^*5FF-y&|X#XR(*?$M74!5Qo zG_ki#;W=jc=vy+Xq@`iJrA*3aoHT5c(4eJ3s>e=}?nk!omM38Z=Ww2}=yq$66-#OB zs8-gS6-q)t8w5TWTR(QmLPvo#MU%8ZJBjVMKQ$W=!jI0& zE6?96BSJ|^TB>i{t;h~h?N{ef0K}M0(j_Q0JBDySs zeJXD;2=l>*sz)ZT58&2%Acn5-d82 z_)BJ6B-hMx*P#Zpd#SLgpiJwrlQ=}3Jew!nM#3>WoTDan=vvGz8*33YLox~m871gQ zF^+YkBw;KkMluNy=yC=IKbgvoXGW9R(TUMaHl2i{v5}OoPDp9&bV~T0E-5MfJ_F4a z+pgK0K9e3A>11nGM~M?8P=ErG&2emQ2IFt%UY(t7%C|_cK14>>X~AUNv6#0LcXC;zLWD z{jA;z)UFVcuOjbPinxMEp0ny91~v;|O;D3$cVP~CU6B@wUTD>|rK6RCNC(*A#Vw|f zEKVPpndx3^;I0)Z*#7EBah4q+PVqkM7K+Z$=+qc8Ss{9$;Lb<*^astyEvALWR7*(d zJ|FOE30&RN94v_?il)NC0+iBM!4%7xz`YjkVl& z*tR%q4K5nL_BLTvkZC3RTFI(b!rgM2S^1tupR}~3r8?_e-ly2~C^lh%**0^)mg_Ce zW=arl>6rkg5!@i3{o%jPeD-IAK704G_dojpj^BpApZ?d$&))j%ZR8!EMRF#Y8k93B zZS(mRKYRAG_dfgdvk&3#CGmity(ymf5$_Tl{`|AcpZ)o>kN@kZpM4w|$bG~6rVmav z@XPqTika|kMO~KE_(d37XQTA_NH{i%)k?8y_N+2S7r~R*_M6C`En7XS-w`;gIcP6<50N1E<&25 zD;+h7*c$XD|L?<-E;8VY7g5o*qC29x#a)sTygzDyMQvMeWHv38XgL>)Q>HuHn8-Kk zcT#E1)U(~3O>A56%=9qw3*e+W<5k!~J6MrL*QC3+b!DWa9GYQpU|di=4zx)8QD!o} zR(s=rA7~lCpIf;U4tkL3?CY3Mq%K9Tg0O5X1wMv^xkc^OO1-2yq*SDiv@-5_Ik-06>9v4r1;+L{uZg0CK4UCKh z7QzU&3MS4IySO11!zeE2IK`HjLX4SSCfca@v_gzbE|}JGIxQ)Jw%k`1tE!DX$~z8; zO0tuZ>LbL26M`Iu?!-|4$9I#y(5-@Hk*_5HUzo!GJ2pXlVzrTL%j#~roSihW0q^H!PrGj#i0E^OmT{gMCEo;6 zeLY;%kV%1Q!8)V|dW3}5TtIQW-zo?*-Y9}jPaH!+Q_ijtgxy&*MqCH3|Jxc$(Z*imn*r}Y;?*l*%PCZ23lsqCq-qSrZ0!r^kqd#rTg`sO^&^C z#iRzE1qm3}h18L2FD+PnlJWC~wnHWa^LPU)Rn96!+QY1swBTk*Dt$r+{{k`zsnFPb z4jp-30gsfVW%`vItGOv&7x*4#rT`!rNm2Ttqm`69+aM^HAD94#$6lt%!!2?=~luq}n*}qjKu&6`xz0?O3O-ctETJ?bn096LN z-Y|$2r;xgfXeA|0iS6?6Mh6X|DN!>;%_=GjUbzA?Rt@xcePt7d(6scSFE5s*u!>RY++;Dl7U)ma_b=Qf{C5+JQ?x@oUGzMfu6P(UMN+iZ zyyhs0v)9i*bL;f0w@&=w=6gTed_`q9PknInlb>FH>EoNfI`ie_XKp_Ilbiqf;g^>m zyZO>@Zod4)KR-Gd8I=ha6+4w+nkvbuzMkjPD2iSa-`*Iknl35*x+Omv+|6-1YqA(_ z(N;wC#-62zY#IfKq@+YZrPwar546Z*-lqG3oX~ywmf<%hkqjL>cVzqMayW1GSdGg_ zDLpp3hG7cwV7GEF@_sQuGdet-y?_6rYvl-yj$OCb*IxW%ml8C>*?jau z@-aL|^9BsFS2VFopwY0Ew9vMaF0xTut1zS0(TrQpEa4VGs?d!^<9lKw(ZbML@J_77 zdGBY_wN19vMosM6qQZk%CA5pr&k_FwLie=`P$C}6RbW2|0IJ*4UCNMA83YZV1$@p5wg5p!c~k8 z=1gomA}fBi1)40_kR?S`M6qY5i%^v^0h1_ni)KKtjCr0SIfX1XN6I4P*V5@+wrjoxz#;5m%>?WNN>=Jsa zqYSPs#7*?q(-z|Y?}N82bO7gZ0ow-V(L+aFA}a!HG%2OU?%^=z{f-U4F4nJ$KYGui59FaFynkyHqXuG*QsOA)gsE1w&S(h^DeruqW3Jt^Mz3efQ z5j-ABz)R5&gY@ymF~AIn-&13|Y^jc0G=;+0L+Y1| zpnk~&>aEud{lmj>n+U?1*8Gh-J|Tm_ryN_W2ll|kp+E|UP@fHQ7g9Jv5Og+!GfCV6 z0gFdT=xvfTWHpeZ;Z>Q~gL!xGP;u#9i!|~x(u$;rvgnN9Qte2!R}xYxK%&}TOs$KF zQ$zHIa=X@4Al7nH0QP<}*cLl%78iPNhAFC)38d1j-D6-1lm`R1?s9S_rHW#nOh(J1 zam>vs0#Qpm&Se90LO3WW1qyq*Rak2WyniaN1Kw{@LBu-cV8=^ox%;_ndNQ@HM*|Bw z*zq~sj$UzKYP5ADoMJBAKZ@ZOQm-f5Dxb6_sj?)ilI&w!k4*2K5sW=4Id(4sFTdgt zS#e&CPK+Uklq;qfY){pMJ1gF|K)AcD3GU8HTXm~7;JV{z8>2SJUx!*QJj@*7g4J9| zWbygPuG;Y2(Rasye=2e^Mn;hMsM>a|UudIDe9Mq`t9~IP_;l3xov^>v0o*j<1eZg* z8Q1C#4l_%)gkxCTHMgQIcUtKfxr0ka(DEb$gchNB-9g+fJt}_9=F8)lSh`?UVRPTV z^rF3BE?tT|{JXI!-ir2uvdnyGyh>y1Q$p{3YC=-deK*O5TABZWZMW0~=SU1u2ntb$ z>`YrFX8nkPrra`oN|$<%DGlpd32k1h3c%5CX}(r?#xc4@fEy<^qiT{8;C|eO=Jblg zpHq0bYd-&2X0)nH8MPIWD5XNXL7*_YIVyzicSMg$P+MKzz`}PidzkPXIW{!HnG0%O zj54A62l&-IrG5DUTwfQKl$6f&{Z9VG8;4e?)4^)#&=aU;q(m}6D`SUWD7y$Na6G&8 z242IFG`c_FNE$6t^KN+3qXPlc2rNZ|Vw5Lnm}A#qk#UeS2}w?>qK`^i?#vVtACZc^ zKDNA+w;k6;6#&WOxNv*vLDv*X>vPH8Vf?EQUhY!LMJ#!*X_DGmCBhbg!ynKn;yG$! zE8nfiZ9HU>lCJj`N((YX@O9X%YeDsR(4-|LK#Sm()`~TOG0QDF=h(D0-c(j5G#rdLl{gII#!wY% zzlFzpulBKcwx8E7*50XIgr_@s$iS=_jD_2@?iu6RB^aRo-e0zG#pEZ|`a zUN&cDK-*eGm15Ogz0RE?XBC#8nGvsejTWixy|l*?eYoq_8P7BC;(j7zPQ@WykPR-P zSzM;3BuN{-8)K{6y61)1f-v53xGAKA_8BOP z2;I~G_JKQW|JP3a{@UrsubqBzGnmiKs9gWWJGUPB>6f2A_T`6)DP&fC7;a8znO3iS7`iSg zso3RQ+rVmOW!JVbDibb>V0)->>M1GFZ!z&LNrDxdxOuM|DYKZ<(>y^vBPp5Ie3%jo ziPzWfK7z6U%-4T9^Yu&5X1{s*`G36k2KwfyGhe^>Ji7h#NB?;5b@a`-*KePD_KOog zote4)n=?DUe(9ro8t>RQau9v}^6UTj^W~NMW@dKWe*U9-y!Su*=0Dy)v8f>b`r^~K z&%FK3vzO71m8ZXb<^_1F=nT`hpFVT@<#V@By$(a{_c*q<8|1YyRub8SEGfyprco!L zsT7%k-TFg!D9IX4wAFAuv;`Kbl<2o>UQd#V)M>CRD_%36?gbg6?~Vd3Fx~r3#}?OC z$+c|9B)n&8h5#lhrS(pTB6lra5K%3V!J8rg+)6RWu@QOnC$S^gX-(t{VJSY-HJs%t z20txjlChJZVeqGm1j(6nLS&{RK`Cc%cmVFVi9hrXXW2cu{e%sTGRq*`iLS$WZ|`(} z%CuPqeQDB=Ce&cO2zIND<}opIoDrkc%Y93hrDP!IF7&=7r}Q+!D+O-%pdDU4!2xO$ zwDbZx%3MK$giH)e%65qGrQEq0@udp}e29&93<9#dD4P=ne`u0ZYk|(p{44qwxV)E% zFNH>nT$EU23yla0+L@V+o>I9sf$p5k$3{1|=aSlC6-1_RxX0{q<8%-G1!+?T?@O$Dc2M{o7~1dFrDz!t?urCGm$y#m$MJ zFq=jD7$+N}!NKjvpTGUenaG&PiY&?Lw4|sRNl(R(#S2hmDU*oZjc4n~^_6^ZR<3Yj z+rmZAC1-ZLrJ};_)(_k@yKBlUPTgM@LD3~Gkbk@#L8*J6tENbbVcaP)&Wlych#%|j zPytp_QZjzt-ZlxQbv*k?Zo;4x=sRgmv8_2EZQMwGRd#cp6P99PL$-v zuZs8lKcn!*f>#R<*O_I5f#msA(JW_WuiU+Uy;1f)?qCa}9c5Un2F2Dri>((#gE_YL zTkH3Wg4+il6rNDzV6&7Ro}&h_ENoMI5!6Q5;n29K5Z?Iiqvk2nIvA1XhZu+^zftsZ zP%$$^Ke~R`kme*LQFT}peY$%kGgZ(qGp*sXoo6+<(&ME z_VI3c35KnFkz+K;ah}*kG|A>+40d@W6JGB6DB~cd@CwmV#Kdf;OdX?0{5jc!TKkGw zZS8_tF-7z?yoY^{Tsx1P=`JvfSa!D(=3<$%YD!Y%zB$|MW3bp98!WigA^=Y>oee5% zYf9fQcgHVhvU09xtgZr&jFb)}Zb$1!I7JHO@b}tR_;aKF9yLf znBxSO1Z3U-AZ&QuG=!?kK-(dRS|Y8@Q{!MDruHE2iiDTDzmI=6u`WfEl_c8b=8uA+ zA45|G_16GNjR#X{pGTsxmn)(7a!qT)7Z!RihpqSiio@nDG~rsM&6jgI!d;6YK|4M) z?Ai;28b!;sV9%+`qXRJCN?@*)md?Ag(KVpGCPNc2^W{t?(|{@6Sb*u!A$~0>DVf*? z#nw@QR#J`}QU-C4tvoAYu-Ce1KEzUpw6vrJvW~VPI;#pQ=0+#nNL_&>BV{`4-M2L$ z@q1c175Bxr9|k+N7~rLvGK!3uYjrr8+&9ny2<)#VYCB$+=-%hC z-iPOe?}nN1HG*giOpOgRqcJc=t$cCd4j`HW6Iu4CDHYY%+{w2r!mbKgREa0_*pSI^W!r?2+FN+J2@i<@&(sGeGq|(4dZ{!u~A@d4lQ8F$kTk{J0`NSRM6aVVM z?N`ohAi^IRUX}Dz>`vlaPB;~L)h4}bRv`cZ6|OVgdTRjEq{(r5U3o@#=Sv z4A>MEj$CY(J5ncPT^PCjHRRQii7vxMj0z^s#b{d9g%}Osk(5#aMPfVR;os9>(Yu12 z32aomw-P5MrPRLDpxJ3)x<=Oul;f=#P`>-FoaQ~u%ik)Kt}rsj!9+<&s@AHTMTyFc z;)>&Tw~yF&5~I8iAEi0C2N%%;cpuD zbG^+~0(cZjms@wM@Tr{4azrL7*<@auNs2AvP(+S79NWwUBt%c>SwUR|mtp8k3>N?` zI9LD-h|#e&oV)(&{<~oOwUS7 znM7>bw|Y6i{Zr^*g;;2wiR45iFfz%;Cy$+E+`*<##CYe~QCDPAIEc&%X3iloVrYs~ zKn=(dj^r3+Jhwny5Z@L5i~M1Shs3 zI5mgHEfa>uUIYXke2m&2aVfJAsb8O&K~-ena*o0^_$<&OzM}-dE-k!KW(tsOw3mp~ zMl7BT9ay?>jK*j`dS}==@}H@Sq{^|WMi{TayGVY+ueFJ~OL;Zp49>DY@MhS4R$$W6 zN+G(PgH1e#{XDRZ?JpI2`%^_-RP3b8t*Q?&K_K>FCf9JgL@8Jw>0m);!*eGa)p}Fm5MPuNAm^@Fthwp$|s~k1Ex#q04Yn9 z!nDI6k2xhdv}3YZwQ1>|&5&zw#P@V7cr=P)4+&$=W4Wk5(a^t(%obucxV0^lKL6SHceDcCAtGhFC~Mn0HqtCAwT3Jp2pz=frQ}@ADg8Xtfa1yH*}o(TK@dzKT}PcHFNF?x!RlJ%wtF z63^G+^QJb`2_>{Tp=pxbZ~C+vm2kY}Z`r*RJen?PfePJz=j(1P?YrLjTC};0)8caN z*4G(dNn|Gc97;;*RI62Ih7D#7JAZo-yCvbsv9JrPf*oaRvhfYOc}`s$FRmWcIr#3o z^G29LAaTps&Ngmcy(IYd@|ah}uPPB3I=F_kw46}9cJrHQd*5ebTKVfQ^PnXh{o8B& zZtUa)n+{x7POH~gNp5rQ*vWBr6iNs*EP=J4&QrQR{d z9gZs|qxtyB{v9f+x+W=sc5e&C@Dk?cpPv8fm%q9B)`^?He)FFnJ#q8h6SsbH=H{c1 zYzBv9GcGrN_p9q~{OqgepGP}xo&C>Szq@q($rq!WbbCQvQvIdPvfLjcc}hnCn!I2u#tnI0{h1?G*Q2$Ztbn!A`G?_ zbFXm$R9NJ{gIY&jPKkZz!YmfsyyIQ+E^snSK$aN9yW@YsDY7<$N!O<5?E}~-l>x6r zddM5U$djk>0e&qd;mb|%PC`ziw3LqDBVo(OR7=#$Rm@@T*tDb6pU!i#lx!j)X#oR# zJ1Q)VdXb871=$s9^s>XCA_yily`L?&S}uyqF7F|gfg-pi`f;{4E5#|Y1tR)V_jIu& zrGyrHv22rQb-m=ecOe6SDJ3Q3*8J?yie_VcZ`K_yGv*Tu^ zI3!3Z;L3aO^Wp#g$<>o}mWs2>?=QasH=S9&u;{SJy0Ks%J=J(~>Mn@bd&CZQgc)F=*e`7AhAgF^ShgzA2bJBqtP43hawm!Qpr}4H5(R? z$))pjjyl9yI=^(@`i>kOu_0JD@#-+8B`_CD)U04?xvkW6>B zzon$4)+HD3CiAtAVt6kaEfT9v9UTn=5e%NkzVP}Q(e?F$>x!T@NCf(#^&9`(wM}z^ zEi6!iMhK4Xgl*I6M{N(ex*p_;EXkSJHLzPqel@gk#yPP&kzO4NE=85(MC{l-Y=VG9 z-Zi|C$EMo`r&yWH7d`S$I4ZJvv;&(tw;Ww8{RbfxoBFYkePD!)(JP`ZKdr`Y{~995 z<;xZn6yJ+pdf`snadV5bELw15%OE>VHviB=8$R09hOxYnMOZPh!30lCbf40iiG*51 z%xNLBmK0IVAn2~ita%t0K!${e*OlA|2n zkq7h<69H!nn`RZwh@bQqBrdcTXQ*Q>-Z6#lN-ARyYlS>+jGDxgbGz@=WCiHcm zs8e7UVZ1;w+D!|^l@qDt^r$wesgS?0g2M<0t3X8PBcr~l? z{9xImj`Va3VmlZN6uT(yx5c_2o)t|o0K-*uMbGZh#of$-&)5Ww4p*PlZ^(pw_9-`y zqXRun%RbjYvSDQ=Lz|guLdpaQ5&Cw{U0dr6MA(?OY1>&(XigD>6GzBRVf>&2;!D@x z(&?(#+sgudU11;!`?Xdph`Zs1LG9Vv$5&3)KCQi1`{2r{+9y{|TzP!+Dc`l{;hUeV zeNg+P_Kx@S;*}>*?cFPfMataYP}xIP_?lE(o7So)(r~6>CYbt$c6k(I30mfya#uU!a9>yzMR{ zX@Pbdq5T|($Q-CPB#&eS_c|z7U9(z#YX6A1WpU+ay;fcnn;=KT09IcXCo?nmWYHc6 zQQP{SBYfKVdp+m32-~APBXy6s3D=RsJioW?V8>qZi;L_M1mDFH-wowyUJ}3Q$Ygf< z$l~-7`0Pz47biHj3~GByqo9(AuG{ZolP-td{667iQM9IbaAu8waAv(pqsx*W;F#Qo zTkx=5EmUKtsdufFE0FGGW_~w4*ts+Q zKy(LFnPOj%P;ZDmJ+#=u?gP zq78g-o7{%$6I8e|Ou|YLbJqr2qKr*k#a0Yvo-QbL01GH}0p~_&rq9Wx7e(F>GC0^q zqI->n{_@)K;}c$E2rxGdm+%Th27DD8jP3vRKL!WUxMkoRSpboN1_%E)TC`X>$5*)# zf`4|RnQ#+uESj?R zw6<6kwZ*9bOU?G9M00a6yQ1UJmH{mu4(Sd{Nu~&=-EHyv-MEy-IknI(j;oOeUr8j? zY%(j$dL}zO4F4q4pzNp8x}F`LP{*?A0Omj$zr=(uDw%4e-ljsP-m0R;G^zLFlFb1a zFQ^|sJlUb)AS;p_fW!!!GBb%y6LVstl}jX#Kr#8@z%<8s*D&bPC4^mvm6u-da=>9h z370tNAQqd0cU9oMFot!1IgChQ=_6wQZ~*PtO~fY>Yv~p2Ai+M0IedV)^PuUGJ{=

    wlh8P_wQpZb(-peHk?HD6qj0n%_(yQ=B6A5k>bLoY9&=`z|;V}3P;tQZz46hbk z%VR}1@zN{IuD3FUEBWu$eGv`f=fn~_;(9(p? zLan`1yIlLAcKON~@8`uU53eL=ZXz^xGq4lU;fX_$5f}q$YVT(i>T5)#Yd%{mGze)a z)9)PM`A#h2mO(;#;7+z)1j(-sqP|TWR6ixm+N_R6W0YZC!MeQcCZqW_Vp~U=q%nc#d*+wwLZcB`xV%e;=&NCKFeRoMLp8 zVhe1z0$2nEV%gp51R4*ZjHIirc^u(M1YcVI)AB17`2r`n|5gUt$01>nwJ9D6WqC<4n zBxNsp+s|x^Iwx8OG4i}3CJKNs&3omnLb&{^*W(n{ z%jyMYq4~;YSUq=YYa1Z>%590`faVs1Th0H@JDCi=yppiZ8r7_xkYu&bE4&)2@UqH| z!VOk0ec`$+r2?hBeU18CHHM`(7&HesPKk14z>hYJji ziUF?n4dXb5F+7%_c`6FZQrn{}^>{4z)T#G7HIkL~JWtB{OIpxV_wB4zr#dy5$)eM3 zbkDxhlaiJWo$T7*Vpp?cX^zksHCFBR*Y*{Y>lRGvX-NxI=0)QFe6jXv?edkAwNGjv zpxPz)@vdOZMB@LSUpal{RPAEz;Cvd;7|XD^GY$vm(R0pRf|P{4_W|dROl_q^Q7}y*ckhyCQ6hwy-y+FL!@f-h;>b?7bdqFYZWsj7zeP4`NGJ-=`30h zB;WItSd^dys+hF2bZH*V!!hPgwdDTo6!s{7H!?me+HTc$g3STN#q6Eu9m;WBa*zJ? ztU*wL<5{>TMgVi=!q^2XJ@J7EUx?n`-v}oWRwbmEsYjXfs{o+=wCOyvOYpA*IDp- z;jFH!l`Gl9jGJ(22)n{sETcn3A_%lyQCn-<#EPD5vocuzP*>C(c1=`(XY^Jn721_$ zm;_tMt=wDhlmd%wXnN79R@e^v9?uz-pP9+R)-R%9;;{>y5cI{vzgW5Sy*t->ymY%2 zH`_}%BNl@BjZ2a79K&-oZvY6)@9%GT`@^nUS&#dF(L63J9B^4=sXGm|09NDw#XCf)$(J30e`R zg0~Cy(fVZ*TjH{pohmL?m`&}Y%<4IDSN*0a5j$39RkuXZ6fWWDlg1kXG{Ne~_?EKb z%Oa_tgq%PbDc!GNZd%SZE(>WNzxtuWH7yF51J++?rN!&yDiARm{qHvN@>r)$-(-(#Hntba>y*USQ%a=4Oo(m{)^%;BrJj~CU$y9Gcy#!b(qKI>UHiE zIjb-*3dJj4XJu-8FYOU*`%e8j<9X&?^tECZ@(<18G8F_+8@}&yYom&C#b&nfeb#HF z0cf*}O_x)Nh7|OULr$cWq${mahA3q=nH$faljR$~eCgWB$FH4wRMz2eGelI6_#Uu1 zorj*B+!#x!TX`j#s9@GIvGaVky1iZ|f_9+Z%c@c|Nv?-Y5;f6HR+eLvuuKZA7yx?J zCTP|puoF$=-JD66E|AUTo!sx1u#I-)35Z?GYlIZt>FljWFvId_+7-#kK5eF<%!Ecp zmvp&rhqY{R70l=8tO>HDhaN9AB_#qCprIB2qT^JuLqqfP^Ab!nCQ-{7GAi~Ev4+IC zpbQuxnbnMx4m6AExBj?myXOJMD@71!U8#(2pXwv83%Dk?p==Hf`2Cf9 z-FKBI4yN_kTb|U>e(=F*Q9(2=p@WjJ`fO$mdBS3j$TKq%+A(r)`n!s`#YRh1LJc(4 zMP)}#mw=E$d-m)Z8e4trJyxC?#2#Avks|}_`#rv8?W0PU!*Zo-qFJpRNRbR}FMsjy zFRq<>|Jvzu*G_+Q?erhv-wnUtHm|ZTKYa7nZ+D{>3#yFLp7%gmhxb4eb;7#buLD?3S_V_Rm+gxH4@FV~ zPWou^w{gpvCEOxN79kZ{z>dq&j_GQJc}eo*p2d}mn+&39QWnfQPK>3QQp9AkA16^iF4pU zkpl#0ihP%>UE3$YPcOKVl5q-8E*s%%*TU;)khoaSwqwKG*$?e8u)?tZjk|`GRMf^R zk_|U%=@zW2!(uqE+FN-*M#+P!cni!=sMl-nRYFJPbiE6E-uvO!OrLo1LAPyioL2>1gRjhSNI^CoVOi zWYwhrJvst`9En{5M~qpXo3gwm(G+Zw6fr3=f{UnQgC5#gWQjZ>?2=2#6Yw!k$&c#T zm07rTH?v6ZuI^C~0 znxQ>wn#7G?baa_8i%kFI7hitz-t?Kbr!T&F=lxH@$hk4kGqTpOZ&F85DOxJ!vnh0+ zO`(UkDYA+h@arOhN1JpkPmMCcRf!9;gQ6P!!5E+>wM%5MCB!0{ws6i_TkJW|5{jDW zgj;6Kn5r?Kibw)qYz0&4Jp@$-^E8`e9fk zwNEDx)-KdO@US@pSU|3q(pg`76K=T<|G!zgaln6qs|*oYnf###u0H?9o3)FRhxYD$ zj3Dgi2em7c2PY5JuGX#+_!@84eqQ@^S30w@f~B>ed$aYS2kChOMuTvlA0xo&mD(pT zt2ZYPlG@9)58$qwFbfxIAJ^}!U5Nbr>|v(ag*XTjK@oOV;pu=a&%e4CUfhd5Nl8UZ zc9KG~^&~a}8yKJSx+B>}TUbje=|E$5=%DH#*11&>KVvS@3W{_WeC>jeHfCUZhVf+K z*FmeoMQXa%SUn9(Er-<8uPbYDJ|Sd*V@}zbI7c|y2wPZWkq%JTpz|vqd?C)z0ZvU< za$0O?!$)tSfmQ*BrqE7z6C11(o&2jW=8QdUPF#^JRT?=(S3?J!*X-g*o(j(KJU0`Z zjzPyQa6vpWdO3CmjSi!uPgUqqJ?%3lL;l8jST|79F5SevW|?u zf;8N}Z`n#<+OBwL@}JZV%i6Wpo0ge{!{3WO9~zjJ9jc9boV8jSv(cg?Fc%fdY$~Zy zS_+clXty<%e6v01ucpB5_k`C*goZ>K$* zOe>n!NuSJYxO@qwY)hn(Q8HRb{rQtwu4zzf3EAJJs=FMhyG|V2=kHmpB4i815o>{@ zR7DFQm~}T_F zmAr{1yv6u*P_6K#QfVcbY%XApM2Yu?mm<16c?9!D;M98okQ zS6U|X2URFh`!xn^V3Z%$2R8GiXA^J>yxLv+`TeG7@;RI$3<2TvenjAHS8E^P+afMGG)Q-M2V%fi-dW!&6`V;`O`d58i$4{r`9Ug}avze*OH> zyFWg$@|MU~ub!B`{?VORPDB>R6JnnL`^~S$(>^U8UKnX55kD&WXgf}lwUHx8M}UJ& zd_WM?9aYZE5hhLQ4#V2m#+V#u+$gc--?r3*Gt6kwWe9S$jTKp(KHk9+x8Ru^WXOY| z`WpbFf(+}3OOOi0fyyGJT)w@0f*=R{w;vw;_N`aD|Nio+Z?2ppe}C!dx2I2$doN%A z=E{5I??S-<`a7_&u1?r6q^b2yGDD217%7(|x2u ztE87Y7A4#xe%imXQg%HKmJw2i6kL+2R~1(Bk`XPD$dHLQ>qCnyq20VJT$B0=HyS!B zq=rg+s#}mxL_B_#F)JFj-vo%4ku3k8Kkl+D55N ztpYDlSVZepO(BEz1;ixL>Nq&jWVOZIN69uJS-ByqRH`8>?>MBeu(hVsr|!$@jdHV1 z{KsiT4bTGUdsQ=}GzwDWEk+z>lq_zHTgx9|eZCCTWR?uLpB*lf9_fskOmYqeUAfHJ zzR2R(3h%D?d&aud?7}V2aD7pcWR+~7rf7985N>ixs8w`rhYt(0RU9iZoS!t}17teY zI`!HRQI!I5V6c`99{YAmMN^WU&blUTViuSsM%#mXl16lZbh=q2QCQcepMCT0x#M5n zeEsV;udjp%+5oHqYWl*=-g)=*nX`9ZIWc|ick8Er^3wFPuhj<>Jx8`6<65dwlaLBo zlaNMqfMlScDoUJl`rOs&Q$PFq7pK1b^zc^)kKLy=V0pKRfBoAx?wow%%a5P?`t|7a zcdDu?nx0hDoTlXbq0Y$F{~(RDl3E(J$u=+0**B}hsEw=tubxXA|B2X=`|5@ zKuEHu#GUAD$^D4ZY6_nv?DG`;;z*?zl}AoV1PBPDc=a1u2~2vjVb`u74i5Bx|Azy+ zb`I_A+qvzBaAoJNm7rM`_47AMxMGoU7v$u-FPf}2>^DTcO?OV^`v@|!Aqg_lh;Xok zetIn4d9!w<_R-`KtTCB9j6F<~&k-!u!OD^glP6{iQ?TC%%W<$E2YYp{PQHLqo5>?y zr}&9Qi*nWsYzt_n0f9K2%R)2Neg*?QIC&Dgw!Aj@n+psVF0E5@sZ?}c!>p#LnGEtc zffe9eoRNFp(@15W%Eoq4AN3p^jGq)-6AQK5M+%z8{&!vP1ls0mZ;+CjSd7^`KHv*Bj->e%cA}P4M{|@(@Dn zU|sZKxFd!OoXLU>p(`rl+xqN?RJKvH>m7#^e^ONvol^WpwA-MS>JmDNS`kxCYDy;H z2(P-p#0JBJr^&U6L`aVq#4@dmOkksMy0b=nLGJzB8sp5eqX#Eik~K}Hl~kewOYG%> z#=y2Fd9bgCU2~+3{7Y%<7tU(2!5BKYno*@ch~WBGNsvvFF)R8~41#MW7q%ahi;F#{ zs2w!wxpJ&wDGccr+FFp6N+`Mbbm2YuyksMsBA65o7t3t74|uDv>gG*ukWnGawnsoa zAYf@d95J=Z^BpR2SwNChZZVH1_X(v}*{+byOgiJD7+q{7>X6;Mz`%(si7MITjy1&R zqe#_4Z=@+O00S;rC%iWs-eVOu2D?2mU4yp@elOlK1G_NUF+&<=$Q%(QjcXXT!VbYk z<^WA##}0XLc1VevC0LGFC`OB|utQ>1>?7ZJpe&r(_k&;DBh2BMz88PaQZmD3mcij^ zUN~6QzabE;+QejqT4T#1iM5L&cH{=vF2g}k6l(^D*31{hnnAiBzH(QORiY*eVMZ!B ztg^@|Y9gg*Nnfo+BRv^9Iho2Ssa$LfS|*CAUZm(arED zD3!v>`9ZZ)86SjzcM?)W3n52_7LoS{8?+8%-AP&Ks&JSe%Tv_Mc4*$@b_wsd3Aad( z&FeG|^`!=)+}wAdPWRBlb@reP_}Otb#2J3%8>sD&Azl$O`ikTg-%?CB?D}+Q!(KwR=sn0xcV3`&L}{8rUh46J1LrBG3JR54<>X2Xw}3PS18S#6`&*hZ5-yGuvQR z`%=Up5wP6@^8Fjq;9= ziWVR<3=IeM^wYgt1`ceahE145V7{eI3Tqj4pLwRM>wm-l+8J>OBb3J)wB>DBDL-IRZ(P+nmrL9wV+{_*3?^5A5Oh}>(JC$kjvD&Q&(;szxC{` zL*2KIOyc*FZs=i&Xw;N9?wcgNH_ z@b5hQd%GK6fzQAX#lJzk(?hpT;%6OL@^#y><$Qdz`>8 zY(&9g!a~T%$4TC#RvD!Q*1%Iw^|-bJ_j78Ip1hE}$St_vklZMPE#xp-@#5H4 z5&OXLP|c>Z9^>7z$;pI~^P))^HK8hM8r&@%;ofqQWxt7=FoAxM+C<6+6X2bMl)a{8 z1DzAC+Cf*;Qrsvc<3)?)S+PiEbl;+BTTC`g(qw3hR7)w@faw#u2^Ri-H|Z%mE)rI9 zCYjen%4~o z%C!(qA&+FQW7XL0MN4!MG_d;uP&BJhAZR{-Hf0>lX$!>SxU$&J68H>KXs1%mJ6Pzr z$ohXFJ%tLlIGPApBefFx`R9wIk6I3V?*ao;$Qt#c6cPO+Fq?`L{=7eQ=!927u3mf9KYNM;p10IzK1SWpVF!8J_+!8#R0 z02p*Y)#4;P_}<2JLi@+CxC5wBn1tWbS%6@C&6XW&=F4Wy7ToqcyKc6mFp&YXK}%*7 zHRmIYh+(y*Qc5zhlt!UJ$Hjh4YUbU_eL7B;*S^+Nf{PeUk{e_jFP0p#6IKsHU!D}e zUK2&+R*OLbFOdRd!?vA1MU!kSAMZG%c+yHD)tu=V4hw}pb>-HfsdKkZ{O#cLQ*ZzI zc{26x)Z0@Z+&V#~&P~02>m*$H&)+}&^wfKpJ;kdZ+&VsW{=d)v`FSr}dJHZds2}}s z>hje4Q&+zG-{+?;Vz!ywI{Dw{;iZd>8!&MVpEz~#yPKFSP_l=46{E_&YLOh6diU1J zsSn}RbN(C20r-$Z_@yxEJat}Cl>fX&@SDO{dVA_T`R}uMdAVEu^B<_d6h(e|rathd zU?!V=5{2jwtfW%D&`K(F0yC9SlAQ?TJpBmkzRlVZLqh?Vxs|K?)=ju}aq1F0k5iX% z(M|mZPTnC{RsxIiC_ZJE@Hr!YzWC=$f4=(Xiv%8C`0?(o6L9m}SbE}BnGl8G^LUcn zI&tgRe;@Z&{qdQ5&v_pJx4-LcBY4xXTZg@xl&Rm;m-emW%W!dr=mmf;iSzf4w~S|3 z_}sPb`%G`-N9z8cGxK=qdZVYqI}Z1N=?Nu~h|K{B+I@&ssuS-z3`rD|DxqTHn&T0` ztvpRXM5WcmoUTpSBp3s1Y!|OkzvJjjmxd3q=&F*{nq9}xA@17cIcVBz9!l#4M7O_y z)jpGlCZ8wxsvor#L(sJkFw=bvjz6tkA%oj`zQ1AD0+L&H5VlQaK);zhSi3TLtakHp zQhRCgCv~{q$TG|(luR1;T*cRilWr`nboiB(RFVP4$mP_D^EjF7LhTvqesTx=Jm6Z! z46}&M3VZiHKKtS}c%@Kp*23$=p+)-<>*N`d)UxX-I*(~-iM>32Yd*BP*L5WkXyxe? zOOSra>;kj8wlhBQ23vZ`gxaXgR6|#0VpD#$!vlxCB;|<@#~vq}=~$(0T_k!|Nd?$o zmecn7h;5JaLS)i^@Nzz<{QSOw;+NYo1#;Uv^s<3&|A!?Qn7$&vt@4|MJKfn<`-JzDjzH5rnXP#NR z+yf9p^NKd{th-NA8|Fzk+$Od#1&cu>vxl1{W>&aG2^^XYk;bTe&tk{Fp6HaMHR2DZ zs4}~K0BTueq0g$Z*Vry*8?IL)EvxW$SYMJ2+hpAj)_7pfKUhPCrKrHg5t7a(bUmx- z`h%*0en6sv@90S8z+dFijS~3s70!Bh4*Bw0xkkyecN|hq`T}eUq2=U^RR!v@l~vuc zOB9(zEK7!uSdUd?CO0Tqx5X-gu^7DqX~U>%R((n-GT%=YiOv&&f}_>Afz9RNy@m(Z z6tSD{mXwFjVk^F$YtCIoBK~=?cC~i1b_pv3Cy&lB0j{vHR_%4{i9*^#?qoX{Gp~+8Yo=kjTKYo&2>QLmiveR3qzw~mV5D@pJHAbg48j^_Fa#JJ z>5&y7$H-tD_UV7;#c^KL67x8XGrpLjcM?)kx{~hjJo&s{i#xo~#-w`|(Hc^jPM~+L z%DIAF&m+9hUVxe`LL+3;o$7P1MGhgHC^1TM%ocW4#4dDhga$Hz~IFfz#+8<6Vd#f)%TSi?{{=+W|<6sDk)6wV^f`j8_ zj(WTx$}#+ z?woq>&Z}?Ued*erBgdxSym8jgDPTxNK``d>v-#&cx_K~wNq^}Pi|LUy|B8&I-p$~5# ze(v_6cW)m$>RW^fKZ;A}K!cW3a)Gk;)wngz*x=ao(r)`ZsVd1-Y*+Qs-x}K~8Zhy) z%^TvXOLY9IcA@ss$~=N9gL8`!+2NoupL;oN#i zop1ker>ii@4Z7AQJJ^1Q42de)4sSHcI%CW*8PTKXbX8STv_wWx!9&gYyr_k#7Z(>( zO34Je;8vkN@_eW|KN7KBvs!W`YacjI77+=O0UGF4D3?ag-J}j5mLdjKNJ~=g$&8Xo zcZwhyh8xIZ%v&*SwGVPiGEfN7G2LcNw-L&1jwnA4D>5nTZ9MG~RVtw*vmMq_|BfcC zA~&Sl9yE_ELKGwdXdbImBK0jz6hzy1v{X{j)Xt8!_>Pum5-aey$NAe^lY3%DS;ZV@6(Nj(3JL`@sEP*t?hf=kEs{DxLql_Hg9d(!%}kGuJA@ zJ&iF%PcHT3!C|b35zhQ248A~J=r~e(p`x#hFsF z<6f27$Bz@3-)k8E@XG40P9DSy7rc>Pn|!`@C1%lQ>){%+2tScZcG^FD$uOI_1yP`^ z5(6Z8`fBu_9b z=%Xgt3f4h^uDpov00QMnq+O6s$(4_jo&tvtxk}tZqrJh7hQUwaR~PrLL!qaCpsL$bQdXpI``h$&vC|yxfW(qn4;eHzgbim z+K%1ca7}=m5<`}Ez8fp7V%qTag z$x%=DQIL=WW-2Lgk7Wa&2|=Dz6lFzi61`OBVJM`ZkzOGkVMfoA_c)H2N4ubA{6#;Z ztA2D%$pwJ*R^h(wk62e>=Zi<9+`5Rho>KI5r{LEY->C@6o$)1js7sZIu4HntBW{m+ zD8WS*2HVY7vA?oa#Jch1lu`8fmgs7c6Ss_(#d_k2DZ5l6v~8Z4`UchRd9qqj64_7z zBb$$-w}5CNMvN>e*hx&<-|3oZU3tl< z$A9Jxo=`9iWdY9}0m5QlDn(i|U`k|g3z4qL7D0?!VSYz4tL4CrvSK8P2H;sRN}P6O zvOW_pIvx=SRa4SSE9-v5)>%BCyoIYQrKAF}&<=VR_<_t8X7ob)7NB;4P|}Lp312mm zcQ!YxrI<^lRYAoZ0yAiV0`Bor_!!n#C*S@0+OJn$#A^DDBX@rL{`60N_4TWtPQP~J z>t7zbbLx}nQ_(4F2bfiKxkUy`ydWCV&>G<>Eo6Kuhwux@O!LNNgx$EXdh}P5M?6T0 z3kW${yW!!CV&VbY*e9P~d19))>0uCHp5$X3&L{upAY3@e-~_PMWLyY|)ZUYUOO1h%M5AA5WH{g3bb{POg%V=GKf)7P)udH($L z`3sTd(929Q!U~l-f$i*qN%{DzQX#>Wa|nx&4s3Mc!IfceLu`&(>&gb^PyPdfJ*b`W z(%T4+eDHBHlLo(8yAeI@>$!DRA$(AOdQxwMr|TynWsz3&KpHEwwG2x`)fMstEy>mQbRvstGeQNCO>*K)-eNxH zq{yZ3Ic&R6#4mkc6=AmuHKeFAvsHzoG2(6@r<{^aHHWcFVK=w!tT=H_f(u&a%U-n% z(!44B?e^Yz_ht@{*4X&OB{D25N7%&ROxr@(w^DnDRdNFxvs-mWw<2Nh9M`U_+0XQpnG%Qfz85!Pgnub(%Eq~qL`{mU`cMczm zZl&kbi8TD@+Y}=ssDcFa1B`*JeE%#dxuD$Kya`NXk{8LpIGp2z|JZ8;M& z>KNz0zkr5HjO?Ig_)fi2AUUH=kpG;l8`@1^{`rEC!)yA6*36d}LeN=Y?{dS%EpA41 z8BI|WimGOmWO`|W%YmZn)X>&zHOIitGvik*EHT2JOencPgnKoLVy?``HcEV4IIK!z zHx4b)LrNxLUyIMJbpxZflg~{a>n7^sWb#PuGRA^D6x*9{=^?3{*m9Xcn!Qr{Wb&{t zIOMz9T{{J%LIj0RY8TeeymHYS;`Q1MxO|0p0sVtsod4?N@fi3CTjjjs#(1t%GNtHR z>?Y$?5t4vOK-PtJ1|*WYqNifx0v4pmTZ~9npk-rhpU20yqD(p|h9ou0Op?K&$Wj>C zw4}TPcHzgz{~h)Ux3>?HJp;>r(ift(d2lel4QAbtw7NO3^**o)I+2415jhx>#j$a? z9HU)oUi2W{5OX@ToKp!U9iWG^!tQpSpf(L2*uj}Q%8-c)S(PaEw^S%CgXjp%8}d@+ zW%Ze7$UhNP1IMFUr&>*D-Kiv5*SBGC=rOVr-ZRQ9OfBT$Kzen0qB{GTYuD~(78`e& zN!PAzD|1x{cHNyLOq#4niYzZ(US>WB_P|Ts*`$(9BYmfujQjvRLG2RV=L-=w0!W$A z0i=|sq>~+xAZ%e2?dvwrVYhiIK*SS;)Ob7hu>J2Bu-F(3oo|##2loIX6(JT$6ZY`L zHOZWbvk25F#R#WdMpu2ggrZ{*D5Geptbb!n`gX?9g^{VWq6aYU4LgE7gd7&8C~`(9 zDZ8e_N3iVpU)B?>iCl5qRlNw+<}h>O(l`u9es(^3xO>+CE7Q+@>02m`4NDEbVP%wL zz#(3Z8`fNq!Zlf3VA~~nOR1cq#rJ6+Az6U(FzMs+ClNIm2UC`s=ly7^DinUdHF|GFC zAzF%1k_n^=Po_FT@%b(Ld9}1OP{yT-1aZulD8H6;k2UahAOScAq-! z0U@1<)21j#mce!^i`>zDQ2t%n1Gan+ah=6xZBe@U580P24E4oRZc$ zTiKe4L)fqo9q%YqVX2t0S<#lLQdCVz>9O&Z zVN2W0Z3hb_{$;&~vy5pKRzNftDs5&_bJ=fo4~^-$A)fHkNhyg;r+MTdlz?nwd1}Rh z%bj2U(J;5uq0McN$h(tEYu+AhgXX1Ya>wTMUh3YMQ8HTWSoI?$7>8P8`$Vo#fl!z4y&WS7+|qyLa8aQ`aB!-v8?R z-+X*oC_zU;9^? zbccnv(%RF6f`L~6cmIaCx!T1lxL^T7+FGE#wi z?e>)(pgb~4E>KLk8s*VYG$UM?a0?6{{GvA5bQ;)n0Uq1XtQ7EzR`Rq|rWOchYu(@D z-PmG_YUW8q_$f+|LcuG9w~C-@X#Hlm$AJNK*ZtepySWcY0rfIjrnYal^Uk|_%tA?6 z)>zk2iB*POqueEiyPg1LyN`Fx%tqH1Y_(?K<9C&euBvO5yTIu!zz>@hl&+*+sdQCK z!Vxz7=*o|E^(bAUT;g3@Bps!e(v?c;`?Xw3*;kDo(%nO4#iXOYJvA~?Dm*MnRY@d5 z_r$E(MqSD3A*gx%62037kURpXl~}?dBMdxE&I*q$ApB+0Wv;4=t?BIo?5SQ;h(Zzm z1%VT}k*2EIqKd-pCR+~u@O zds@Y5$h3amJ;)3JyBOKzGAU7yVu-|K>wB7q96d%kaaPV}4Cz|9r}q9WwJ2*{(F44_ zQ6k72s5Hkwmcv+=-{ZwBhpB;Z#vZh!taxbP3(>hOkEQvK^zANcl$c3+;TeTRI>4>y z(PS-=RlLec1b%9WtC0~4oe)c=m8{nB5oU~1hgPM?w~^rzB?6+&#LmEY>a1_`hQ6(P zwrt%3N)3B=Btb#J(jotp$|{8HVihsMyq1?)oPCEoCZg@z9JUtk0^9E>q+iSAL^3LDs2iN>42JEx4&6%B?ck zeJq$i$CWX-e|k5mxq0O{U6}><*UVzi2hutR(E2+R46&uErIHC1w5jxRlP%z2&>Bip z3mqItsY+T8aPfte>L7@$Dx@F<O2`S*D zl2v1;ZoO^22R7cn9slXxy?wA^(Wt>mA*0+WVLSIWR^l!tw*6h(aq|aw z1s)Cu`%1rSiqU7DX=}4~%od`4Zrdp z@ZV>rE@EYb_gr!JGJ_omMO`T42y%UYixU*lvtyvTA&IuPTD#)h-OazI6ZGEmW+WgHwZn+!>2S>xV` zbzpRXi!029QFUF9EURW}oXCXRARt{;v`pxIaLteuV{4jL)gH#mB;8F$z-{QqA#(>T znEL}KEDJ|pPic!9mI*aU`b;+B(cdE6<0y>BW?;80P*=2O=jP^!+|H6`vCLvSs!ArO z=$fMYmXw6L853&nh8zcQRdI4-7 z6WnJv#K_9(DJ2uA;k$2s1BPKykC$gYjs>iX0Vea5WA38NY--XHwH8_`TY~1oT-^Tw zER)DEBUp-03H(JEfW3f8w`Q=Xd134TYqmrNU6YQ&C>l#dMOz^OjubrGKBDcY_@^kO zw31p{%gdwDgoG}IMfa0ola9nBJZ@bdlAcxc_}#1zs|op`)S0xa2e~V^;D(pGzt~jd z4Apa?Aq2sBkf0z*A6yCSdQ8qbyjw6Ry1Mu1)FXBAPD zba>(%T?ySz^ViyzDmUY>du#`tT%=O+qOWTynf0Z)v*9I>S^yJpHKu7C12(;_0)NHK z(fEyLA6H%?WT+Nokx>#$qsx6XWPxRM6$&#MgOR#b=`t02FmnOzvI?cq<4nB$hl?ha zkCR?dlQcGi)Dq>Ps!Ae2`49y%aob!G2W+0qu4D7DgNcoMgWO7k?Xi6^A96fsgXtG=Cte*rI6Id<>$Pk!UCYk%+cgfnd z1I$Y>f_;NbjBD5ae|bsP|IgX$|2%UQ-ix&?VB0m``_xlA7&poHLDoh1n+J#ix5_S2 z?PAtw74Lv+Bj5`6n0dD{`>yR=N<~qYJ2q`oq+;GAMSgU?U$OiTjh0Sied)+V_;s9u z=cZE_;uVHqi=f3^B-Ux8rc=|38XrLrcGf;uN;o0*k6c<wS=N|FwpAl zw9M;vp7;eZC^)qyH>hhyS23+>iBYlFFRL}zx$Y6DSXW)7T35x!zI344FiJlB)=^A-uYdsRLRoKLy1S>sWzW|2 zQ&(;snmTvu1etpG*0HIJe}4Yf@yL>?e^f&D$t%1- z8-Nd-b|V6BK}dV=-v2>*hVf+K*9oZC0XlKcCc7!@8U~+u)5C$FU_R6vW0t$Pn`*Az zVfbrgkeMTNrgC0JcgqH-YtVW6C^}d=3U(n@#MP*gW=w0_K*mZpy7|^R+gi&iS$!$C z^(KY^?rLdldv>4nn8eR*Yt*9Mo z8@9j}u#~`ZeXOdwJ?I^PyCaWl99lk-Ewq!JwfHLPccO&K|^t6lw|xJ?-4SS=LK$rHEgQgA(m7^(bZ%XMs~m>c zACh4mTR3}sEe_)4E`h+Z2`hnv0pC;p4C}Rkylfq1*<7WR3W7+$v^@3 zDoU1aq#ZOG=ix|W%sf%ZfdQyNuOk_K(_4eCcP##%4DNSqa*qS z*uT_7siNPx0zBqVJ1kn+|o648KUwwHx4>#gZF*(7AKjadji zt?1d<1v^D%PJC!1E;5-#=46D`d-In+2~|@E9J}Y!%c330(AX$apPilKd%%_ zw!T*i7qpD9qDQ3bpWsvVK3yMD*=xxAnG2BRbu!rmFOIh(g{@)lX*$`-iF&1jqAHNn zf$eAe+Sit&A*E!H8f7(#BtDzN5+$G}6g_}0zWTsgM71PhAQvx)&k|1#nvw`q5=McE zsB=g1)s7LN`Z2x2LV9V~v8e@6IyyebJfDNQMv2LG`W$Ln(Q?hftz1DWiJu?v9V6=YQ%_<1l{g-vs_(WD?$Et^ zA0zfSlVxU)I}>MZL`hmF%mOcT6E&@CSv94%4zT+{t;_pRh4&EnI|j88<$MPDXJa23 zVdKpC>_t`>|Ln!+4@IBtN^@QZR%+pJbZXHDBC8}5fd?W2s?}f4eE65Q4*liKxxbuw z`7dW~{^iX3e>wAxckser&KyGg1n>MjT!N!xcrtT4-13WMFVMf7x%!tgzr&wz{RBn< zL%8C7^9$bj88~hT`yb!iy`7erg_~L_x~kmFt%3_?j>S!C6e<4NDn?cE!99oWdh%#}!m z$O?P+BF8me>E7%X5L$ENm|Y^&A$#_0LFzv63;uTeXA9lqA#Eb`6VgWzEs4y%xK0~y zw9x?>230p!U5Kk-S%ux_KM}rhMi%`J-VaWkFS_37@7F@M_iG;gXs)BgJQ%?et(w9x zX#rEX@UT@&V^UO^85_i|C00gGNu^`2%{sl(9Jaz(cTyI*Dp)jNd5LO-uFb>#TB>;T zgxn%MU|xA33w5dP8R*{J*LoJx2j!;Cj)HVxifKP6_POI^XnRjrBDp>C z%si@gr0PtZE`N4DvK)2| zurmGZm%gY|)bFDiH0qMo_cHQ!>PA#9hy1| zQki;p>dLL-x1POqsC(+Sw+>IeGxhe=M|<}Y@8A-=`u@~;xcL~|dg~1}-Mwqlgygj_Yx_e~Nt;gLa-Xs%e zN4UczF%j{#u-3Jh2cb4`w%|&n;2^eJtxD$c6#vr`8_15eJzKVKAOk&v4=f0nP>ddKuipj9SM~4=RtT4leHrP)J z2<{+Ap5cB>;k1ef-wq>&sE#fB05WEu8lv)YQT-)aDZoEEM~iKWq9B{Oq-D}iZKY~O z_YDTANk^!^4bq>!37@~wH6@$wl;UF%0yDYi4S<&#n+s}*8dFax=~M@GF_7`zAz~DT zv?R5y>qj+}VM#)oW#5!ih}m$k5E_@NjXPcUkcjo>Ut zXB{4(u0y2{XSopw3>^p5GfE=R?6@xi(*K>8KK}aPo41c#`|8@WU*7ono!2hkdF8^& zE7kw%(-*(~@zHc5k+^&5+Sjk)Xrk*7#`7-CbIVu#8yQDO!}jf5pcZ@;s}K|QM=LT- zF;qyhJE4Np)v03hGEGd5IVCBor7_b4Eh3rW$`LLxp3%l+MiTpjgVYJpvh_V11`qu0 z_)qul?URgR9{>)!ZbYwr`e}3#X2eH&DphdlICyj(9osHeb-*gM4;Uu3?e8Ao6)cNF zgs$&C^9-5))&VdwnXnuI2J2Yh)jEH9PlZXa&!2gwt^9ORYCF8Xkzj#Gy`2Px%!I)# zbUo`6g*6WIh8p?7>E{(%;6~Z^)c7Bp2+vJgNd}09LO(Y>UM7NWWXfsG&?M`!&WFvNkLGnt=p@bP!a*I zw6X{(pl>gqAXsbq?T1Idee2ckzrTFyn=9wY-(NcV?denG-pkj&x$++Q`^oq2oqY9g z2mfR5-g~bfUH9!<*B@)VV{7kD^6l^5`{tvoGxzP?yYAkp>yLTwfA#%uK0dgj#C-4Q zse8|U{P$nJ6j`G8UOsy7cPH;1eh;R!A<`~&VH;CjT&5~4y4W$MJycq-zu(Y>vz=^ zikj9HRf`T~y9r75Jhau>5=9>pemX1^KFAOu>zaMH&@z->{UP$n!ruAhu5HU+Bu!AZ zlI^2pBR8Y(jNA9waO(O=l+|2j;YF0oEGhwAN#~kXS1AHh;h*05)7d}$@=w41)4SwP zuffS{aPm4fXa4EkKfV2@Q-AvTpI%z-;SBilXW`qz&xUXs*Sd#Fj10LFJ8XAwYad+? z*MAA)eECmrkpBJsk%hN$e>Iu#86A-wY>-+;$poY(BUIqjUawu4JT!T@c60It_<5yv zt#)=7vQgLYT(!t5+qlZ9@5-Y$Po21th8vuu1eRv*KZZTuVLZoEaHAnae!#@A$@sR~FK9Y?E zl?=V2G*!`a%~9RZE6PLEy*BwD-g9>lp1M!RpVuNwu=XN+$idnN9?miTjSpu3q%V<` z@=?8~LL2>BQqlF;Wr`cj0+nRfJ~~ceGw3Gy*7bosC~@8gB4RU#Jz{Kp;u0Cg$c+t- ze!0xbq{L+bQ6ij)(|L2^G}t0^oNV0P1;(%w2lHj8Ju-3H>G8R0Nnd$sGGuva5}Q1O zY~!oYoA$84YzN6VO^{lJTJR|H*kDvkuir^xTW2nSNELdQeCjFKV&3SU_{9NG*@s~_ z@vzPJ?j^NParb~t02V&UB(+Q~k}?75I(70xGyq{n#w7etfD1VC8-~n zcQ%6#Gt@o*9@uFvP99nbeq7Tg{VUj-kJK&_+%jQqUWC)| z;d=DQWQ99JlJEC<7Lh4`l1Mw7$TUAEp=mI{J{_P)Qo?NB4bMGyZ9|xXh?)L#IL2yl zQYro0vX@>zw++`TH1ZXJ_%3$$*j45@l(=3oyz9hZZ)QJoMIn>+t?iUAjF}1@#!Mo( zZ?g^5iIOcLO7`~Ad%rr{)+ib5c6w?_%JdQRIfS%m3_aR9ZzGvh5`oxd$Hb{1jD=)l z2xQg1GUX)p|0OfAok0w*ya;Oe3aDuKS+{Pk&X_w_JX_|764aU}%Mzvx zmB&13%?jY*8Xqs#PIYN&WXa@IMbkjJ=$fJ}E%24hD7xAlLJ1u(-XR=rux`?WJX2K% zSCAt}f!kG+juC1RwjZ89Bo^ZpOr8r1yVBw4AqD?ime~G>v41t4Q?!(l%`H9juO>8K z8J%|$QkGdImk1PtM*s-y>+ix{5dIcggJWfz6YyOp&P<$H_GaPyb+w&KO18R|MQzg+ zF-*Yiu1%4#*&j3Ug-pUqB$WWqyLok0&9E)C)KQjusK066sWiAnwn{O|jdJuBvH@l@ z(`-z7sxj%Q@RP18IlVat7P`V0mcbsD$0}ZX<;Cx}Bk1HXB@?fVI}_JsnXKFBPP~z) zrfF9>OrDHh)3JjaVssvM>B3kU>LOqnNhC?l>dpRZX)L5nzBu~D=`XJS*WoWd{NgP6 zzt_Kb>x;`@T>auh0$0z&wbTD~@Qd>+Oj7^%;ukl*`1Fgn;HzK#;`d*ifnnhB@5z51 zf>C|=#d)$WnMkD)Y8U*{6N%`8+f9qK%dc0u)a-pSn%PA&nkjt&62~YcbOS<|r1w4+ z+R)~Jg*H^8h?`ni7ABfUq=|@27TF_UOh|%ex_Po2KXD9~7U?U|N)^LqC0>mzrWDu> zNl+8HoT6s3K07;VGdc(=qomW#8=nXcopz>AOl z(?&@xoBgoty*TlEtF4Q#{#b;mOXR+g(cGo(Ct-YK1mwHLDhQ**V<47lJqg5TRukQ+ zel^vd-Iz#pC;PK~-MPM$+MUg8?9TKgH~LJxhJ6=WxY-4S+v&)~GwjGkP8={j4A&+G z<+e>*1vL<>JWhuN-6@Z2UEuG{d%wEy?ehoky?Usvu|Mp>Tw1-sqq5)vWvO{?SuJBV zj<@DCB@<|cj$)>4u$U~45l{@&92*Dc9(-VKjWFA(c>L~;M3u4p`(qcS4DaW56~Wph zvxVs;`-tTQaq^1*=3WB^p z)GW{RX4sCK_xER*Vsr(xSq#k!MB!QYh6>CQbxlgDFn>&n@i-Jed$?I*W(7DV&$#e$W`(580(s<=?9xi89*L|kHV9NpJO)xgtH|S0*nOMK zbhvmI(kjRmE^T4))^&%)*u7Fnm#hb>>BDytF%wQagqk_K^*hTbUl=0Zv}n4pms&{W zr`VDgN>oO-xha*x%K2eOQ(50i@=ihuDzyNmSO*M+3^225^5_-2&3Guj^i^n)=t?e7 zWW5TvH-vhSgn~yNH>j{wJ(E=u0T9Ygz||LU71Ahio0O^R*yY$_YK6j46N;uaJ2FvH zzS|grCog|M6C}58gb?f00(Ol}t8Q=~( z+^T~nti0^|V0lb}`?$sS`ki-p!=NUP5aUacW%KW@#-q7I&?zeNB%wP#IV2Ta|!7U+d>a*Tc{aDji0ONpc5_a zQCSq7sE^7#c+kU4!dF{<2t_Hc={e8}6#CiO%t(R5(;_U`#~I1Mh%DiF$ExO(mi=;{ zzkBy%l_4YBg#hPs9mT%#=!H@nOuH&XdCZ?3UN||tm{>GbNdy4LRttMdx;5-6iKMUY zIb_)r%)F-P+1R~twjt0w6QCYM6*fuca4b+YyaP%mcK^H1Zyq zka>?3_9tsh{=w&P!6WBzgJabhX@}z zVQ^44Z3@;VHAM~d7p!7QRJny=G#f_Jw#AWDasgQPC^6XqD$AhSIg>6(rotUk7FH4F z$#R(WII*}v0@eCJR1DuJM8>zN4{PEyxDI4cP#L4ZYlF$kB_B4L zA44}}zW_)ttz?wB5@a^(M2=puwl@y+d1rkY=j%WWz&$48eTqWFt=l;sCpSSN{9zN(?Hoxz)lrb z!5bPWQQHeii{W98$Ydr0Nmw8zgB8daYaN%*!x3e#q~qKqv>s!JnM{*G7iL0|o#t4j z%8lsu!7e&hq0-3H!ni+Xt=*@)S6igJH-&wLS|B$ZI(llD$W`Q4A1C#^qb0iXlF>5m zhu(Z31rw^NogT#4J%|_YojFL>)rWrX^{dhIOs(rgOkpQtIzFf1qh8Tr zVMAMM%$FG50wA5Fk_aG>bn2gH>Yulh?^%WDJLC2-f`5YWx_MBbBg}97r+q#3-bqMN z=}IC%bRAk$#`4E7Zt{lrbJL`9tgf6R%V5j6Gb(&WhpPEBlNMSv>1ibwpJb_1;S_%4 zgTi%6WUG*I=vL>|Ii(wQhw0FThk8y)BsvL_dJqbXUxN~`+>1OGT4t+S<#lYK3Ic+M zX87DPIB@&mp*xq}xpVYEA@a9M4^5vud-tVxzB+j9ZwG(+<@H~Eb@pOICc)RbRZ!eo z0+V_vC7q2dpiNZLa)rBELp++QHfmbxCm}_YQWELraw4}J0WtG*?bF&-B=e|!GD|k_ zDg68hE*_tJrS|d4i;C3Vsa>dDtX-RYfvkh0Ym?8z8H_8sDL%lYfN34`Gmu2CtJbks zCmV?3cC7GOE4)mJOLkW2NOT_KR@ve@QVFV^JjkpH0^^w))HPd$1{_cQD=rdfQkx7+oRw@m^5j3^O35j;HqTQne$5hg@>OjCr0mda>v}4gBvgWb zZ8G~Yvd(J(-Og=BCf+3bgeysbAs;(Zo-l0ZnY)?DiL-`hmah9(SWGbU$cplgWZpa0 zPW-sw8eUM4tlK5r3bie%s;4M1N; z3>OVkE?s;1hwCidzG^YLG^j~OsN=UBaz4u;WIdSfG5Q6- zA|jGFTm^sal1J!w6bb!S-g2qEj?7t;N08sMcH?n$v#-E-F4e9er{5cJf_#NHYahVy zPDIb_GTY~os%p~b@pva8C6$D&u~QA}^LbFrg$pR35EVcNo}KfR_Pja zsGK-WdWFesFwWX3x1^E_gasny8P3-}o;(ib&nLBOldnu3uJaia^=}7XO5nehVB*z& zHu(w|b+6Pug>hVlQM}+Y_yVtW*M3!JXojI)M(XC;RWf;mz%f#w9-MrIcoeJf0{s7# z=(4rM|N37(V_i%7taa}sq-1qPPjxg}L!%kAQl(bgRa!>Lsm<|*(3DV?7-eL<_}O7$ zk28xuD1Ww9`PW6E$lb6+5Z)=xaMxWFRPBWTqHt6)8ZB zMgkqVz|y&;kOK5lqg<3Q&nqyrw{wI`PS$OrZqemd^elOwE6p)7)LNX^ZvByS1C#YflT<2Kma6dlt+M968b@>0y+KmJLlNmO181N6h z;>7tk&T^U)tf4@D@Pm_wKx7T~W8o0>!{!D>~Cd_9Z!FYC#_s>^=CdfIT} zY*0gb*SdQO6)#VU%pu0j1Ef zZzdaXiVZso0+k0_MpQzHhDrv+v+T9Zd@*bhWL@PCuh|plj2Jdyslu4Az%?>WrHN^vsU zUz~OK_s>m#^!lAEhse6=^T%Nv3n+^^4q2S+W8{E%nf;I<(M*aLkC<3%=!N_!yH(%4FriQn$^2yK3|_A)rac8OWY zIM)t~IgPwOY9N`qswKiL8W;Iusj|=%Jy75nCDpcBh=LF+&oJ92Ty7(tAldDWAoBEk z-*(UuqUy7a8-&YRNW$eb_V^^yv4g>noC7(-j%f);MAsOt@*>m7=aWlxiWu)_pD_?r zp>-XpG0Fr$L|H`(U}ddlJyN3P@UX*T+s#YVB^vVib>xANoE3)cw3Ll-i^-9;t`bdE zQUT{bbbfx~tWgp!lPiqnrk5iEB{uoN1LcV5oUjAbb*L;f=04S^ph<-+p-CgjO*-8? z_n~1L!Mf^)IkTxN$*wVXoR;~0LXumC0wKvQ4mHOfKu9vOH#p#tThn5ZlL9B-ce^(6 zuAQatM3zV{siY9-1(6}r4Y_nBvkTU&B(sYO2DH1{(e8_AcZ!5V?E~9RV{ea^YBr0b z)D`t|QIxb=;@D?R3?U^51&fRd)LH&5rSpaI1Om41rY0?L7M=UB->ta3;A0hvTnz}~ z$SH|bb2&t4Vu*>ONOJF5MP`u^VcoaluOG`$O>to6ccad)4njEhGPldFZNpoz%#XUm zNRLNg7~>8jyJ5pYG1vbNzPUS#{5vTnmsYf-dSKoe{E2Z_PFy$2MP}m%tSm)0Xp|&_ zeF(9cw{ez_9!8j%Q=jeEF#WAE!$}0@{U7_D5Yq*=1nzpAD7p=^r_djexr5;{@s{-P|$uY?ArR3SI~SFyR9aV zz*oQEb==REJ1_7m){tL=@m=bw_o~6`S8CCdsrUPo842nQov4RaipjK+OgEPrlp=sD z42hA6)A>9n+$+kSIBoEWOXZapP2En5V^j)hk58N~60lRM&cuguWa6wf7F{E4N3cyt zSAF}RPbY;}N?C9iIt(Fm;RtHDl@?-{#xm-YIVF>8)=8l;Oust%t2>w8xqJTLSMQv= zd*l7v2M^ir78)%Q;2>aKe#Xku1TEeQR|GjR8r9cWe8EcU~mfD{e(HBA}qI#Fb&>tapJ5n z!9q2;U2nW82oqkk+q#CL6SjfxI6_SvYg7qs<9Q}cUi7@_Il_)c`p!EJDIy&_>bKCvZ>hAB4XmKhfeS>v?57Un-A_d4q(3*bzFI%msXJ7_Hp97o2X>n57xjNi1&scta4QL-HGf{Q#n1YLH?Dqhq>&1E$*!sNtHD8!f?8OYo>9M#(yA=G{v4W|r?WR<(U4v6>I7J+$GYClq~2 zTy@*sN}k>?AM?E+g6H?Z@RkoC9}}bU1y(%~7&5Jzu*e;5kP7#Lsop-Z#VW+W^h#zF zO;r^wkyO-dt|1&BeSK)rAP-Lml<2-TX;l<(2e27DP@+X8vaxJaU02ljIA4#BgI({C zM7T~1OJatF*``j(G&Un;bFtlpr=RZKGH_sk8(;%)_RKR~UH=>Y_ka{lD72a|!srJB zrk)HUuERZ#hPghH?uB3P&T5L5OeyIc{7Of+3QF)v)hZ)W6q%n}$rndd!v_*|_!gN? ztZH^Cad1}KxQ;U233xcP*c~&gMh975>!sN$*1lMhN-27Pa;GDyW=WM{ywu`Gs|cdw z$tn#7Lvy+-QWxOWBX?eXWBR!hUmZNQ@?4<1uieB(K=0e%Ir-D+pZ?_Q=Z{7f$3~Zw z{IK9rBO)M%tdi22l}%{hryo29SR56Kl)x6X3`QJUw7bdEPk+yP<{9|4r^JoYzj=q> zbCK3_c+5NLtG~QcdcQWiGxzK%F{^&mJ4QxJq5>x?YMpI?IjDMGIe1n5RE~0Mt&Ob7 zb^5<(;)StPIP2Xx3_5h5;Zc1=lfVuT+$e2KewV5* zSW>9!i_X5T=*cCmibtaJ$<4clAuxt-!?c^eYpo@Dp1td@&VHO*AF{x#@(-T#5N-o>stZdj@RVJ%2B-3nH6h~7P zEz=wuh~jVFec|Sp*RG+j_Vd&CN%aiv38Z?4C?8!Wa6n)G_KiCy-vF=k>(@_4mVt^1 z9X+Y2IZg4WbQ<}#Ovrp&2FZ=oY^UdcO|M`^iAhQ-+>oML8Z*uI)#HI!Qc6y1_7Q2= ze8JSaQ|E4-y!Gs@qh#u>TSssG7{oJm@z#l{kI1@P$EME1E61nKKGs4JJ)BijA46oZ zY_?0Z0)LE_q~A1-jHwJ_Nh?|aE^8I8mWC>k!eFvJ5PunLMau*bP9hhN}?cS z+qC*ZWi+9saUYNI(UFj(MG6)dfg7a51uw(^UTU3DoJ=Tcro*{$KPSeuI3>K|0pPb& zN~)3!@JU6fhU($c^a>4;lNO#hhu4W|F)p@?Rm2D@5K1Kc#Aw+HGY7W)|LnbMa}!s( zHvCuC_Wlr(!Hv4rI_{lkmroo52F7;6?8)nzT9THfg<9R3J{SbfTbu%9!X#t@2_z(h zoPZ=G5FjzfRK=O6o)7N_x$4QbnfbJb*pR@~{tNGYw`{8|$xe4$APXzRc6Y1Q(puL# z+{f$YNv0+_O=)DEWtenWBi&;c=q|EFPr>gE7?L?xhZlMxuj$?clDBQo?sO!SwFhfT z4Z%fGlxmbsX)TP^3nNQ%&A^biobNrXF0)P+qE&n;u8q$Fg}l~e6KSo^SxV>04B1L8 zr4%FW-MF@y=X6K9vS%)HBncK`Rk`dd(C(blwH)j}n0e^O4O2g>b1))=LzRGd*?)L< z&*lXTG`+v8GSKw0ojj>63j@sn!zHN~=79q3^9+F$2neCDZJ~)u&fOy{*4{tD;-WZ2 zg;-SyYA^1v9@PXLA!Dg&Bh8Ftrj(3ON8i|EAys%RNd)bVdhYXFX0gQ(k^@4Zit{Y1 zxkTu}l|v*Bd-;QJ4m|keMEsw}PyKTB0{Q2$0}sxeA`g$>{N?HwSZGP%G*5`oP=Hqp1~Dl&nQHjpwyU4qAR~=erHj<qWN*Y5Z8weO}!zxw&LFTESUb;U8sLb4+g-er;; zKpGL`K;;I*pqpEkyLXsmhqK!O$sF9)U5#GabjooBvPS z#5n>@I6iTPteR;GH1fp6DY%4fehYq%)hbEV+cqD_wl3%}ohqao3+eJCN{TbEMrA<^ zpbt<8%4JxkvISOhZGpuMR=F}NrKw9?cPK1tP&K8KbvjRLdVvi$m-GpJIa<6qe#j=Y zyI#D#niNk#%dgw*yGfwON8x1OOjYc!i=#GuoJU{Fua&{BZ9H4lbQTH=p-`psrKyKf zOu{lP$dM{U>sOQe@eHCf*oR~l45aS?X1Ub=9YK@7@Lvno;%0W=)5S9XSRYH+|`)ryh~) zSvLlOu2Oxrf=x1A#(&=Bcu-HZ=Quo6RWazC8756AAn27i)wnvp)!wy5o^z;)QrS0v zhb}P|g-EEy$5D>49m2Q_C52U<>ZHAUkFkwnrAIkZ8O+(#tGXBgxj7brA6Qi6CR^0s zhYq_8&AbIo`4)QdAT05(!}1QI0QTf4EbV)tdq-h;hkNYTu{%(JDDe{8OZ}?iCHomy zuv_qo{ifT+s{}5AR9gJT-X=ziS090B5scKfIKcHq2VzK&S~k7eFjh@TWoWK3$z4&b zO1T8Wmz5K%Wg>r}PSpkmYDJMSAzG!|t-y)XsTU|oSUYWMp2o$iMlqdUY~@D)Ab&kM zOjYk;L-q2CN3nirtVObpQ^VS2QH89^W-Ma<_l-;BkJf)*|48+|`_~a0j4y5N_vphb zn5A4hP~4EG&wnOST~@C!$YHI$-^JdqV%Aqnplvpa_-&rY5r_Q~bR3-I?(lV?3+(g)hAU}U(e(j5-^>vnO03Yg+zB`~9yksqm| zK4KnflWtU~VelcIrKZ8P`f`##7H~RL*jPCCo~zZ%6t2)@8H2J(-nBtq3e|}JSh)*T ziB1=EeP0o!^;xc>f*kWrQL&Qtc>pS@)r+^n7gktsvMTVm6%n_Dh`5Wl{&~VX8EO3@ z;-Z*@9FF=f$zbJEwo?)ob!)p_X<>PXyVb7A(4nWHIj{`cwKWqIo~PT+#eSx9MmoF; z4LoZy)0{`uS33YsTqFVxzyWC?T&V=PCYJBmu309QGaoyj96Q&9A=zR$7$i(3R5uNh z2f@eWf3$pPrO(k^>AAChUa1^vWFNa+{!n;m8Kf&LJHt#@kx;Z=#vNbY-e5!sRmJeE zL}*HKotD>z8|7mQ#)MGI7J#!PMKiQcRSbjWY?7eM5_wU~T8%|pktuB*h13t1)FfR^ zI&#z~_?4iSI4@+W2Jb{m(WzE+v}D%sy?E6_&5?sbw5mL-YmxXlYZ&bt8tH>SAYYmp zMl298iK!7Iuh@s0N|HQHb&wI+qpJStDo)QuwgJW|m65`;c&C=&Dkc3pFr#!02EHe> zot^=k_Y}ucCprWG(X|91hlF5sN$es|$W(%D$JB<6Y=NxGScU(-aR|oRf8RJuNw21- zKrEp3a*mc^@_-WF9a_^l3QwLw3 z`tsVnvp4P^y8Pmc_s)Jg{q{H0CtjOAb(pLw-SqU};Q!n_GX2_L?!9st4&Yh$PTrV0 zb^WL7*YCY@e(LbSsaL+n>#u)0b@-c~Zk>YbQ*WHO|J|XVuD^Ewt55IUI0&!)bn~72 zd-mTy`Nq`Y4}L!RG2Z{m+3Ec+Prv^9{qHVLef-(f#W(O%PQ82o%qLSXzkGMkz9!ST zoL%R6p#0X@83_f$v8ZF1ghGxo^7e`Ap%Rl?MBEGUGcby^rOn0$*~%d6imb`T z&yeD2jN=?XJifmaQ+u^|#SZ4*XGhnLj_(^kLW)O=C+wh5`&aQ>c%c+WdX;?tdGQlF zspK|>f#Ul}MMA*$H?Z)w9UF?#r3dW4GvTMp-EUHSA47r1_ZPpxfY(x!+Z-Xa)vv~X ztvfuAm8}d3+dz(y0)cROw1mt0NZ7SL5({3;ORi{%hK65;Bq0!NafmLm z{w8=*(^acp{5ZZ=;80KqS24W1)>V4cl%^Z?OAvcJB8pw;p4z0Rs9beHM;RuE$9Tq} ze+Nf{*+A)!!$2H%g>E4bClIL&qbWPqM<=~o`kQP9^)izxq}ResIv)&_IX&Qz@N0V=7smvh{=O&5 zijoA7rSk zCg6)vl7iE2s<0gCNLzr~R+@!BJE+N<&$Yaf8J2?DGqjNFEa8LmtB(~_>hLOyWlm(# zB`ITT;f+bS?^dtT-E@Qv_QQX2MO<0+SWunpBnwr$$5xotj_ZJW57)D+doz{G4TlP-f8 zCx|d##E7FTmS^E3hV`(6u~w`APB}_;P(|U%OuhVVK6Ka^CVEN%-x*UYN3yUI)Y5sf z!IEGsE(V0)(lXkg3{GAx#gpEXJZkdr`olpKqE&@_7v7wpGsq@QO;wZQ&G1(cn$I(; zV{!Qff(6^2?_X%I={IS~m;LCsnouyfD7k-zkFmAK#%~Py%TLks|H>m1_pE* zwW#pA1}e9qE)DuLX~Kka(BMU2{|N}9h%_9IZOE7GC!+1H6`zAHYi@8#C@gG^v!%wc zFm_P%&DI3-Xf~dh<+%Z0-L@6=l#WRx+GHBiW{vb)$qX6D;25I@p=2#LGM@3-(BJKt zCxL)trr0O0v57`oS=CO{DpO!pGaBs#j$!d~spTkI&`~sY=7&FP!*vqZyb}_{Fq1kI zX?Z0|`c9TZ4^X|K`i_LvWZd%_|0f-u;c#^F8J|Jw;)!z;Cnkk*_ei%GRzVN>4|aHRW?d=xIy?4FVKwQ!63jzH<*@73{Tg3A%vyq3yRFRg2wkLBY_Mg*kLq zRY>u?qU9)PnIwm`F|l|nSIGu7<%8-g2$Gm63Q`~_h*5`jT3)~0FJECVx=`Cjgrou- zSYeY|TBrF;VL4k9qF51b5mz~d*^;W|9%|dxQMtKnqsSDIHtpMIEd~IHC<(!^5Lt@T zJcjkqLvl+6nzy6$S~@kAA1*a${fY@vxN<{j6{uKF1uEKWmwQ~K+GEVF%iUtvo0Ai` z!qk!{wn1V-P!eKQwjX^V+KILSs;4coj;d+P4h*cEX}1yC^i2d6y!hVL_BapbAA{ z;|e*Xh(LKR*~c!?lZBA9B&qw!@D}AWG;J9qN#P;f53_E)3OE8JVIf>)5prJwo#*pI zETb^}KOPSR^^&=}GP;&SiBv5W#qies&zp?&=n2Qx;<;W;3`;#ns%)jMH6HjhJL1{a z6kvnNGo5QGLQ6@eOiiIiVZgmiO*UIlcU?nL9T{hC=Y%W@fq-uwswZdDM+XD$f3WW- zxWAejbu*O8Xt41;Ti}(spsMv$Hx*VOLo~;gLUS!!@HnQMq?aohnwxEjXH@!mD$kQy zg*h;BP#D_BE>XcU6fHfpJX3Nssh8+*8ai1(@sU~!SUH+NO;Hrh8@qzVgsA^r(_a`%2GHa*7 zc>|zF8RiBV5FGMXz)>zO;F6s+hAF4fs+vvpESClnfx?r-7&vqkvk%A00*Za?au|Vt5@_XA*aFHBQ*blPTRQyj!x&V*s9`MC8mpY# zu2Yg=D}>2x7V_G^kSpX)h-QIfv-G-e6o?Yn0RSHJk;0Ly_WW3srjo^>ami5?oM zc@@;A??3PQ{>E}Q<_^rOZjER08H+CF(vwwrP2nborO!vzx{icqnK^BkJB7fqm)4z+ z7BCe9f*ffTcBofjhgUy%c@MX{Kgn%0y?#kwLKvLIG5 z)GsJnwDrT{n9OEi<1GcEIO112EMphr$&6~LX|_Q9r?&2%whB1f)-gR-DqGt4@Yi0t zlRR%`w4tGDG}<61OQmpU5Q+r_DQFV}L?cUZ&?8Lf z+-l95nIKDr>Z{3mTF_}iQ>+F}oJgx=SA{jtb~q|pqG6sbgayPvx#e!0-ClH9kTsQA zdP9OF0S`$O0^VueqA$QTpx%PQCqbh7CIxBRt!N_9V zpo0oLRvs8JcfsU1GJ-t{2;kr$+-#9id zgeojVs;aQ9z&SJ<&zzu|$(N=?)diM_5DEK^(FCz$twM3iz+RUq*xu#-g3hT*w0vGO zxE*u(%o<0@?=4)T;Y@ugpH}hl`GzT-8fhkLz zrMKGU+BK);g=w^4_Jes+hSC=w8) zs36H9AuN@FDc7=cSIP=u0nuNooFuLbpS}2E@h`>G@#6c%4~y>@cum&2Il3PunYcImt zXz`}!8M_Tu(Dn`Lkf!HgB@}=MRqiUSfW0LI?0sw96UqR_FkO=hKdjt+BBh__NBX;Fo2QZyZzW;~;?p6Oaj zGxH|mXgPK#Kf&lB(Bxt*zy7Qf`G-r{)ufm1N^`vt9J>feq(&H$>kOJqZZ)yg6xl(s z44BH38IxdM09RL&gpuDN)Q@rM@l2ZET0G?$h@KQ0Hn-7?!wuCQiqzIUBa5OD6(TX; z1iQa+_!fhDsy)Y{8&s78G}YcWwB&APSZ>8GOE5V2Z^YC4OCi1;u(b0ero_+A%WYfZ z>$^tQYbHFZj+63H!M}*-@>jc{%hF+6pfzk;zczDu&bgxn%NZc{v74DI#sh(+5SFh9 z@q;B1xB|;j)36%n#uPEJG+5mAVpVN}06PBK2Ez|ISKHpVo8~#r#Nlh4|Z4T|sE@UEz(Xb!| z9QNU`gFo9nM6wX9vJmgexYwKS?>{nq@~e7QR(&ky*q=W=IX(K-^xL2ObnDpk$$itO zPJ5mwzyE#VcfWINeYJPKMK(SWlSV^9-=R53Ghrtg3M8rNdj0%~j|Bm&f7YL&DQ$;= zfiklk?A`@+s_O;PMb(1lta2NRIEv8NCoU&MRESpP7_LN79sPBh#Or znm&B=-stV-t7AX?-fL4YzdH5m$J2-3C>?l4r6ZqL7#Vh&uf4`Xi&$h%3RL=27c7NQ z1C@z06GtchVpqRd{vz*8u{HQzMxuNzjuO~_g*?4vGgr6MHE{jp#JkYA)1-T6w`YXb z?#u@R0VhDU1G~gU0=q=a*o=w31G`p`fcBvFV2EA6Q6sjtgcsAgKs z#F!W&zqOdTiz>;n%hdEvj<-Q0GY$sJL5WfhNN^dCAYuk*FwlxenA!VHmS?L;kHY?X zI@=)Mf{5kfg275inmXtX3xmZ1LO3Xh(sJMPDx&V*fd*VHehm$|4oBC>_&yB3E8c*A zSBux6!TZPedLAaf8h}@O-&=Islfr^nW!G-%QPWHHWM)0p8Jb_3Y}lNJj)NYX88%<9 zCeX7uVlI>7F=+UxJ&f#;1;>Y$#{h~7vVX!eB<#^mo>Z7ISGT1JvL>{!bCz#C0~R;f>XLR%Modm4hNe&*V*&IWRTc{M++5{PoGo|Tcz5FZ zLQ2<=R*fY!Rm*wznut+Jz%&z47NQOxq9YXx4_(fxkPxU!f%G=3y4Lq}ZQQnD;|7p5 zn67I&!Q~J3NAj;?L?C1~I)S9Dyuy=q!6bonic^lCI3yho%pSVL<19wBR!xx^fu|_fW_}Kxz@u*h^QAs;Dd*79D0b z6?7w4uaFRFk>5Ej7?@PoNI)V&%s=Y|>cN?@>nKlO0LA{9HBSygooHr&SwnMF=W42v z;W%LD&uKl0U1QX zR6HV4DISf)B{?F-hr%=wPYy}J_)sDh9&vNi;NWTkU#FDl)7LT5$B@Zs>?4gZLSakL zjKRSsi@)yD^t{1H9xi%!l=myNU}Du`r_U-}JDnJ=FFG=0L9TL_w@gLVrXn+=Q_C+8Jo}ZLcc)zG|OZ}mO`EZfp3#2IK+&y zja78X?IdE-q*xUv=fUc3ZxP-21%7MM6pW1#=gHt88K6d%Jm1kl6mHmFlXf!muHNL9 zoeC?I;sRzD-A+YIX4lNTG-96(4lZ{S`Zh;r&6bp#^Cw_i?VD{!1Wo0*3O%7y6VFT& zIug)bQ(~g>hQG6?9nKfWl$71hNeMF|1F~-h$|qkZP@L#lGjOtc!h?VJs6z=sI`>UDX2_ zT~lGEu7VD@jLlk200Y?v`Hd76BC-$-3u44cQdzqu@B}{(}2U;w?E+iX9b(&|Y z;lut|C(%a2OxSQ${M1TR+e@ov4j$0Js>BlJ#>z3@Zb@7*ciQF46_8)lEQ*`0NQ3%} zkX{b6S7Ebs8hxb~ml|o{Al7lNu)*M>gCr_|dfA$^vj*xURSHSXXdEHBgz370DigKq zdAn|ALaCa`;^cW7oVj^7BZm(&-Q@Nwa0-(xr`bH)@V^~^r^C`^<{qW@*Q&=z~uQzz7QB(#U+Dj9TPIG6OrXX5{%USQ@d^9@B=o3O6^dGP#7cTLY0w zr>du^@+PQpw@Yfy_yHVfXiF>J9^V6J^R74BEP5N%QB1SA0za5raU1lxcV-eXKxZ9) z-A*nkW%3lS+uG|jALYmuI_!v7@SaC3VRpCVKiQ%KKaK3(?WicCwlMR8bno>w3^}mKO=n36|&V`oxV7u7H zL4lz;(gEUv!`2!s>motdx`-GPBCR6jc9x~~!|9Z@nha=J+K~K8DJ}(sXmCkEDNn}k zW@(0RaQ#oOX%~j z$MJM#&fC`7^;$B^yr*mQft0QqJGknQ4Mjb)b`VNZIgAA|TmiheMUY8K&-j9=4DS(5(WoE=0|MyQ zLP&H>=JHH-japcUR7EtYihI<$8B0msVgvK)ehDcZqEWn`c+ z>=LndW$a@$i+PJ6*%GRzbV`i8KK4m=rm7sPRV{0I?V1%c&i?j8G{=?go`G3U)wH2H z^*A|JQ<0Nni((QJ!U5lLfmkR|(@eBhHhdQO$RcGr#x!W8Lm(Z7sp)C&YSzXK$ZUl) z*Of97k+R)5;@WNui(;&*)KT8Ua<(OvkN2@SS*Pi{vSXKbpixBCcEBWI+2nS*?6K)C z7+K5MPRUG*3x=8HWw2_#9N(-Bd&g``PIq#pO3^Z2_OVOAf`TMfN)4`8GXQKotYu|! z8J4p|f|>@Q!z5v))9_-70d^{djh3V+Tu|wMG{YoYG^l9qXb9Fm5RpBDw?S1k+p&>= z5nz{tb|4uk?@RWvOC)6>Tt$np63rX4UbGAwrb1gT$~LNF@Gp;jSUO!r7;>~-A2px#si({3fKM;nxUyCat)n-&&jvswQmf*qIE zM8VR<_n?K}LTfG-Z`s}|WM#8y*uI{1R~4@mFM7A|*PTU6!hG#;l_|t^VIWkhIU_GH z-O^3Q6$=CQ)D%cMLuNBbE}cxZ%w5hzqS-6)UaDucU2C3CX{0}y*&%q3JFp+%kX3?Z zc+yQf8>Hz`iFhW7WVMTVV zsIGwsgFMl4ZI^XFh2^lAaYLoos1UA_URNL{vwK`ZQ*1*=BQXgDgh17TbI|pKmS}sM zOC?-yQvZ9C?**t?ZfFQ^nP%y~!bmsA#2B-fHkNQQ~u z;=pl~pUNXh3iG=fFLfaf33AYP&=Cs$DNvVe|2?ttDbzL|lV0u{h)2ckf#FxMtI~N{ z$iaZ@h-$V^TmlmoqQO>Wu(Bix>ZVRp8ueYOtuAs)v<3lTQ(HnYX4-u&^CAFup4`r|hQu6_mA&ivP&AHQ1m2GRK_{@+VK-um&oA3uSwKKkS5KYk3& zz~j%!f9-`g2WK~d-f&sA|{>XuV_tbiX2lb3a|b9tU8$ItJoX7v7#eesSvCukU~J^8N1)1p|QqT)O|KefQouFm>_upKhI+ zdUNlHtGk*E*r{9R@1MLzR&{iAOuhBx^t)esm&l*vxm;m>u3WoVn+!@^gbgZ)N%>Ns z(y7w|)jOy^U;kw4PaoXB`t8q${ycsD^wgfeOr1IM)Ad)UkH1_WbmavUu0_$61v%0R zOKIyUFx6g3VM!g*6uxx%-bYYo5IGiP{|p!_h1Q4I-C`Cev^l1_dI<20+AkGJVzd$ zIQ;Oy2c9wNr)qjTS2H^tSyG-uM1shEAXX)MEa-7*>u1z7XB?6al(sRJWol}HG^w`d z``apzYI|Pg7z);JQqzv`e)J(bQ8Bg{W^1*bq(86era}5JX}21Xwrw44^8snwMvV-t z8>XFcGgp0279yAkiWQZE4#moA?rzX#LkLuYGvuMb(BBT;IS9Jpor8ZnKl$06L!kOj zeogKi{`--?o&Wohf9yE|n&Z{UZ}04#Jc~9!c=akM#Ahcjz&r3R&?pZr_nGHx&F#sr z;7gx%sL9nY#{c%!E|D2K z!BsfKS0^u_6~Vg=AMna)pbri$ZL&(D5RNU$3-=_jW31&#mclF%Q0(%W!F{ON^&)%% zNPg@j0}N{bVeVW#?Lj$A4lVIk(8?6s{wLo+(f zwllJo>PDrc(GR#R$Rg9Wp_+tZgT72cf@k?~@C#vNR1U^i5)Nw2k+wjjQqtIh`Jm7z z11wW&&}FJ(6< z3Q}u-nS~^;YZkn!8VzKx4q|eAAmlq-*vO-juGqoA8#P^LS<}*)y4>aLiWpC-=g20S zrqny%Rev(0nYb|}nChUXZ5L}VS>mEg7GM#J_%2#g$NZ!Ym17#tNK&DO@$3lCp)N_) zBzv|hDXAlu&+r5;{mAq}`dHrN3Ul?=nJlRCt%4vt85vEz%+PjeJV}g#O7jL!J(7$j zDOfJYV;{Sm@gPuEk>L$_ScM-EM2~8EZ^Z$qur#tRSX_r7{}-;uj}KfSy6z}>x{!S7m< zj}SGtxt zv1?OzrIg=Al@!xc)V8qB+7kl+R0zbtAb=HfRBH50D;KmYmRIeTD=~7g3-%pVOH#cQ zZQO{cLAyk`@62i&3qT>MY1xn*l*DGBAgGb+FrZ}%8ZwMUVzbE(vUkzG{1At&qHL;g z4T&s7Tcyh99KBpkQbpmu3ob8F&62}HP_9&cT~CnTaAn!ghxxt!9HVM`hn2PrleKqi z$lmuyi;)d#+7AAQ99efkZ#vv2a!_V+A=^hagG{Wfk;GgpByvm$ zSFw$*#Fc6`x=+`r*}%O1Am$(ks;t~A5s}i~u^r1uGHdC1q`tO}P*4bjTfwg-2M|v& zdxEGO{SFP<%tNvetD;zNy_s(U9ixQDI;QDqM%L3@9<;1F!tr?a_%5#N87+yDyGcnx zr%97(3iVTsae*2r4l61ix$P7M>AM+nd1jnB>= z{UJ&7HeUc~-DD=vrP&4G!WfwqcYT_DdU71Ey~2 zx~8Wz`{E{+=P4}@^8y(gMu^Q&-3(ULiV`7to`VhnE!-d`OXvLYguxifFa_UCa5$Am zQ6x|fPqB|(j-!YW4*M>DHItAGqo_2JNbx;8ZT8~$>jcXk+xf`IGI+3fF;2e!{NGg# zR20QZvH!Juck%t=6}yhq0Xxxnw0L9u$oRqWS4ZF>qs8mv`>}Le@fHler=EJMc*ai9 zyjvl zw``m5`R23v?z?<<-}SruPv70Y|L(r8?(RQ+cmD<3{(IOnRAF=~2Lw^V!j2A;NuZ|T zcM-s3%taQfDuT9xvGHtvFm_&NC=xeLMUgR2S@Mt76Bj;c(0o3_eI$d`jvg#eo}l~L zBMVa1a@U$YwPp`xil(L;n9Lw!60^9pEY~M%{;(W3HT{p4AIY;`5x9T~x`-?1P>a?1 z5wJS9*G`RGBr{UBGsB~mb7shG;=c|}r`#l1FI3@7HAr5wbTXu~l+KeGlB3Bii4y}Z zCK=#n8_exnlKR+rro*LK4n1PPRmI{2biL;CRrW$7$3_d@=h!Bi;EHCMjv%YD>Bgny zM#h@&p(@*e>lCgBR}K+OYkly|fd`+Qi2w8Wsb8*MApbme;K7+w)2Oqw7VAX?9Za!Up$HvYr7esKAXhwmOGt4ghZ_}-}Z#AVDR`}FX`eP2NLI_9TXTYwiA$5#@9ftIaw##U3^*}Qh#8^a-;DbZr$o$j6K|%@0U}M)6_`a0D=O_8tWp%L!cBrPM zkJ!JP&Tc-ybfnHTHEdw-*!i72X&;KQXvp($=%glmFi$yF&e}X|i?FcCc4~7KgWgJ< zqcbkp6g3KZhVRmH)btaqTR%Nbj0!Qag$(9QBsFx#Qaz#RtdWpNff}(SEkj2QLkCw? z+DOg@8MBO|Re_1FYY#hb6f`}(2G7=XvR$aY?VuD^i*$zZ$HmKBdjs-Zlz59yvy5ZD z4$anx<^2_Gq>g|DtCARq%iW?JkFE^_;=%4{S3K4wi}7e=Z9LKuTu} z(S6N@pQub9u1CVFFG*3VB`LagV%>Z7i~EN@yZ6Ddd*{EoyZ6ic=T6-J?xH??chBCt z`@g_MhJ7F2-S_F;{fB@0?#t;5A2pdEXD3~pxx4q`-F+qM7AyZc|c zyZ`*%{qK8Mhu2|HdX8IC88{#1001~-tEm(mTt_@Q=rd-+1@F8{+p!` zn7Q1@nM?8ed*Ax`?GsZUeGYQQz0)6pEOqbXjj0c={B-^Ly;shAo>=#;9=-R9Bk4cZ~er65H>ti49YO;XWjUa>a$o3C(bPPPMIa0siV?9eY zs@e|42M0+%wX!fCqtPX~wnJed6slz9S=b8QouOG;X@h<~pF~2H^a7^n&h9ExbSG8Q z3Qd)g8XIqh?$789(-hP!7EH*HwOq;7Vh$FAmGXN zPIQslI5KAZMy1xqKdY%2$&8hQerOcYf{Yoou}~`zq@>GhnL-{jATwM<+g*IO)09C*Bp{ou?*_Pn=%D))B%|uu@go zQP;c{_NQSh+toSRCp`>*dBS+AM! zs3wTLo^5a$O%GiFE!Sz<@-fwT&LKcY%5X>A<1Gu3s$9SoxcGHyNvdkYJXt4g#)tGA z!o!84t;5_A{0&~6mH-JV|2v!BfyWWF9gq}DtaS|5f6BEbc6K5tq zoH$2di9R-Q4z9rxkGkf>XRuKnhh=}+Z&mOuPfnbLpO0aSI%$93*=?)hl{hro{E6oR8)L4KGpl!qED$~IR zwe;A=TMY)R94GKh24a~Ge`**Ede>JP40a{4cbE+%DIiEvNDwh-E#$Bo)HWI{ zVlxO~c5A4LhsJfaWQXMs9W0gq9DDRr%1MYYa_o zU~E7L6HQ~$$_eC}Z^qv0&9U=oP3J7W73oMVlPlbys`-Z$Gyag$RCVk;9fDU0h2)gYE!_G=-2_59fRQg_TiS-uwE#7HMM;Qebvi5M(zvRne%hMn!* z2ZTOV-qBUYRW6HhvTHF;k_0(gWyxEJeE~*kN}tIwbzs|TsXWQ^NW4&6nCq!7RNbF1 z4fw_jbDQ96cPL4%W;KAe-fA&!rVD0(0NE{sf)~hd4zG{rD#QSVEQqaC(>a&V4xOhn z{<-IusNR4`9t&5>(q0*Ni|-ddEWT4bTKsUNczb+b@p|#%_&(&{{m!P`{knL&czr=y zk>Uq7;qFbi>DGwz1Rk*J#TT*4%OC6tZ}V@wUA#2DcX04&0*{7IxH`UPd~b2II7<4g z1XoM;u4QdQs-V*Sbo|xgcf}jf#Op{}yqXk07~em>*S>O#6i-77k-!-Fj7P^0c^)j= zn4+|AP_e2qOOY*k7!R;QJM=KArm!h3I(8y3kt;(R9~oQo?%l;RFt#ohuK>(21jq=y zda?K|z6I|qUM0_xid(+_y!a+Oa8L2(_c!4Fo6w2yWO(U2__+Z;-xROTy%#> zU6RMcFoG?mMT9fAAMDhU8IUoMoZqkaWvP2YtOkLH7yHW8#l!b+el>mN_SEYar!O3t ze)sF?!$-D>m>41`bijS~R#&=f{7~^?wajL$>CR zXbPVRAR!Mhmj(4O=1wdAoQGw4bAJezkZV zG$35Cu8i-Qsi0Dtc<_?Hxz?ekbZaN{zXLDNl|Dn5R~+zf(tUzt))_lH!+3dUTqDiw zL5y5=%{EeF4)tN)15=X>&3kBePq86tDW;+6=E8W9RiY?|0`=R~9tUpy!%QU`buDX{ z`q($#*#bM)iLrSxMeSluc~)EuFj*AjDm|m+7peA(Yq)z97xq@z5 z29{7SLc|hna4`*S1UJG}G48 zpM`HfJod4{RL_8{r)t`;vz+E8N$vPNanS(1kSGMJVkWui+A$PoY8+*PO+6i(d-_JU zZS7J({9!sQTf^|^Flooat=uxTcsDn&6hhjfBjrQ5FRAG|GfhrHa0p$QrucMjzq6B} z&OT%*#I7}K>C6}$Ty4Jqca{#j3c59ObraJ{2bH(}veQ0xIe(=ZgJmVIsWYlH=%Cup1`7HiS&+q62%!@X z-AZ$nX-lFJstYt>K?+m`F_~^1K)(S<{`x*mD4H4{h6Rf%?V6IKk1U_CZeHw-aD}Qr zTuT0MiNIC5y$n*-OYv@f7t4Y$qK3m^?|yBjBL9ULi#RC<0%3<2)jn|vS44;fs>FM@ z6@Q@9;-n*KT2$e?04|MSsw6cY6B&o^S+vd3I%jAgUW(%pZp<1614?VUd6L_)fos-V zxdhYU`X)uwp#vBi2*?2L;@ihAdzWN7w+!q8iOwO)_;9#}VnQ%bxixtS2JL1>Qs5Us+$aNASsLVWyY?ktjXldRk2`He41=3Xx>rjQAi#{D4lW5DK|L826xbe ziL}z!W;C(9f9l&M#fpTnCUmGOe$aKSh)L6FL}B-CQydv)TzTmwvV)r$VrCf0 zFeQ({7a5*n0^7+5DE}QGbElgugZKXN3)9EH`1#z$siQ}xj{dn6LE^)4Ww(5OCa+vx z)MS^dS{O+UVpUQ^%j{llcE^Iz++dBYrUYo!;*WGKM8F>O!q|CgFc2O|d4{DcV_SIEGX;igw&2)<$^-?HOM-$h5){Z)l~0o%p!wkl zGqPGe5qT`J%A!Dajh*@H$*~LG=Y^Jocb*|z$Idun;iBb<5p_Q?u%2ZrTY5$UFp^r{ zG6+@jyfG|A;kYOYfp9A(qOGH#{BUVC>F2|=K|+}r6asQ(5TR?olI?fx=;<$o<=ZAw zmLk}=l5F*yg5GL8JHm5$O*g5UWY2=2 zS8;1DpWz8kRRgW?q>q8tq%c=+?WMW-pCSmtQ_J0`G0TW#2Thg-&%Dd=OIrp;4|^ z!CHSyaI{7>7}wOVBbEO>CXwBD_rIX&3U>V*0y7ulrMl(q{U$ZJVe+hJNNPO5uR$-2 znfTFgOJ;2A)Dmj_$ld(R4@nXNRpDeS5Qz@h*GuA}Wmvl!BQ8ooAyQSE$5Zh6SH)X) z&d8qeBh6=DD4r>PhnX*Smh$+1J4^zn$bQlwg|h?%mnM zOE0aS35%J#%gX0rZP#=%_OVVl`f3vJa(3)vMbVlqTBcVsHGO#O<23XICj-!Q1zKB= zz?opMcH~SjxG2gIAt<(LlgzhjXbB@}`DTGz>gpI3qAf%mT8>WBI;iOlRC9vp{A<-z zLNSi&CSam#3Cn0;@=X}?R^=*Nz``_%SRH|)r@w7s%Tp(;#97#d&=;65DHqMfp7@-W zQAr2vjTxFFt7sB-ni`4YZ5un<=0nuBk?JZn=apSRTK<3_ia|k)Ip}wM=s-bPhy+?O z%=32`qo8wIhX`h0J4`T)6*c^tYoE`jsGV3d1QJ<)iEInRJW*ObOSrYHEeF3EE;MW9 zowWcYjWz2vli#Uur?dxg5yJ2A~O@o@W3nru!5U&XuVa&^oRaKl{fs5Om zWMIcIBnu6!As3Fz0U;*0`~(hAxI&Zu{j_z#L=wWGN)g(1DDjIgIs$L!h1`xfPX-5x zNp(8(!zDFERZ^h&A1=YWh!_q;L|K-IK?`u@hf5^IemKqldYWeyo+W9fbAnbecz!|C z>lg6)c*H(YB0U9V?55%w-kRaZ?P7F+g)qA4QYgzpAW|t+v?BAINKK6%pZfO7^sBFY z|LxBwzMFpg`n_9kPoH^X`s!=Vx zs^vMRhT1uzu$VTNWU+~lol&)6n-kZg<7g|((!V+iWrz-57MH{FTL8#M|emu3foz?DX`N8%-CO`SZR@KcD$(`s8)*f+giUD6&sJXBnmg1RwULkSK^% z)j9HRHR7#fXHXL)d7F#OgaZ&DOBL{X`bL-n(^jQpXU9I$h(2~E&r_J}y9(L`cS(V# zSA(ix=Stb{SW6AF(=tt1Mn@dS%~nIV(^AtKS@j%7aC>K-?C$8>u!Z#ZywKCP!Lb?2 z4the_B|Tx-PS%Z7>esHrTG~nH#!c;;Kn`gx05+&L%?9qrR3ROPp}{_FCv~CC4JG2O zpw^g3pl5=5iZAphxy>Zy8LsE3niWL{Q009|bV;cj#zNpRu@a`HmRh>t+re|Y8A`Er zipj&UukbU2T!Lx@H--U5^eVNe6hcT-??7zzss%&!O$uHQt6?Z6h*gAD zE76jxYda`1v=f>{`d2N{xLD%|=0b~A>#ir_UuYG@`pZohySEur0y)DZN8RM!<0m_m z_6}siunec2$RTpMmAN9A2`yDo__`L2kt_h((Qi^K#nc4NsD$Up?{_e6GL^uxYi4!M z(#?mEqeY(4kpclR-Vqi-@(skhV{jrzrAT~jXH<$ubn5Nif~>0GC z%(^v(!d!ZA@%+ORd&z^(&ObQw-U1cYHukku0MgdS@>W9ONrZKyrnBV{{Pp1CrH3EB z{P5kQ58pdMf)9@DArHSk_3-3z&j5*%EPy^Q21P-V9fkqVrnv~#aF13MmspA7o0cHm z1%-Owk!lKUvSZ_Mz4M3xArz~mS#vG4*&jy?&C-+2#vim+XWJ=R%N6ge8ok>MIGrb} zh01D-eeAOHK~|Gn9Cvdu-DyovHMH=m4N0^W(>)uOYG!CoMQ&Yq4!5wlg4GZ%0l+0x zBazh`mq6G0p017CHf-E5I9MtbLl7=Bo&Re{5(w$G-8FMTpX+)yuifywp8kwpD>Q=O&Vrt6;PN-aNVp6(Y1FD-w39XN2_RAUAwk7 z7wiNwt`2FKFhXoD;$_XpP$3~OB@OGMrKXrpUg#wpTETe=DTOg1Fy_B!<1F(T_hRhx~5wVPul_=#|)QY!jd%n7| z@$)Ljw_{wk?dN**GIt~#Sq!wT$Zbk`j3mZcRwLB5?LXV*BhuRqEpX!6T+i-nRxMD)A$~LzGx3x}>`DuLnJLsh(l z9;rBb?a(oIW3G1W_+dD`WEYgZg~e)%C+za7H;SXWh%NIGiokPpe$ItdJG50=B(gPv`RW1xLH;YqVZ@=$y3s)T!Uz_7 zk5n00TQja|#@(`g{Lq$1Hq#dH%r3fnNns(@;^w#geO68zUQGsIdTGOT+QG`4lcGz` z$p~hnhhi;6%9abA4-IiM(@@(*Fi?YUZv}!fH%&H=oq>7cokpX6l3q|0ho%f zr-;smIeNxOI2s5^(O@)!-lBzKkJf1kE3MfYk7oY!lEV~=E6d%Uit#0bWwhPZtO`asZROLeqy8ye@v4e#oHFvO3WYKXT39+T| z#BZ`?#jbuaK2Kn5s)@N)N7YN=Xj05;Fu0Qd6IFfK6uUAEM9P1e?=B3MeNjQz&ZuBW zkg6EKhNydY;hnvIKY~T*|FP%DriJtH#t z_2gN2^H&7kec;aF$*XtvI*d$ZWTNho3Bz{Qs>}N8`c+0#c0sF{kD8u!6h0zCq^d}X zmz0;Om%sVx##_xdaqhi)`=@KKPQ7~Q{+^eo4&vg&JmATY!(5t=Rm?s#rgwG=fFRc>1+uHV;#knS`suSj{_KlF`>}>gXRJ z0~>mK$dXFlER&Mk$1aCVP>^H3Q=e9f!SM}Daw%T6lVBd5ao>;0 zQ4R^AK&9}x;s}-ObvOen3OsSdOW!K;k)Y6dLI)H^!$Ay|sod!W)rn^_wmT!So3oL43kYYo@l7)cmIx?yftlwZ2_fqYhcu-HZ=Wy#)Rcss@qZ0}@ zGI%QnPR%dCGdTEf#M8U`HOu+#xd8!+!uqa}^_mHfdSq3Z>nE|yHn=<{+;_o%&Q+eG zDZAFwfR>?4WlFyAG$rob_m>pJ>R#7P;(Us)&(;kqbI%o^u~Y%bq;{ zW4XsA*#1suY#Q5~k}61pdbHK>{XGA=DBTu^tOd55RAJP=wxD6(XIo)CvQF_v? zyT9K;4-rU=WKPiL6jqERNR>H8DCrx!ZopoOvUtZ3w5IY-#Yz|?Y2R-$gXorFXof9& zBXL(r=58(%VV`BZz!dLd^LnOhDb38Aik4%R!@}g1JpnKc3!z{uXKz+$NHX8#JWKME zMKQ6v+Ue1!E-AsnvU@LltfJ-Bj+!~vEE z-yC@G$%*(skDvPG>IL%8V+S6bIYk~Gzxm76FUUU+fAR3}iGS>QZE*16dk0oM_~ho( z<#%lC+(I6F`o%Bbj?UaSIJoNJshdyR??3V7FW2@o6~O;Iet&nAZ-ZK;)9zXE# z)58zZj`v(pIJi&cs@VG zG73nl_;Qb)9{X6`Wj(F{qiG zXCJ$qia{aRsySMg_jrP)twzgxk}O0;-^@-hmHKEdkEmFAA_zh=B#_JnGuz?ZZPh`} zr7&-2+0hM2M%HV|EGw)TU^HhlPr-j3Nh@!uNz>99oRL*Sx|V|(xCyJx5(PAS3vM5=PX`BA zlc#9XGT=6qv6RA|Cchz1&Agq$RgA`+dDzqcuCt;4wR<;wJ4Uu^mabA|1ViY>XSWtq zHZsI@Jnk0ki_)|G8M{PM>ALjnW`V2|MobuSPd%dX=899Qbdy&f=Wi{V%&xO^YfX;= zqf6IRo+L@0rpPP><4W{Aq~@OJkZohJs8hb3IdwU%qCzm#qDDj>KAtowp$gBU6{|5C zhZGZHRhGsTn3V)0ircPCR!A}LSk%s14wg5~pj$wYg*VdjFPEnWYy)tZ$4I}1FjPh{ z#XszSCwwePVsY~PTv4*Ft-V`b=ouV*WVvCFCy;PsM%zJkSRV5knqb~Vq3VP$Q4ULG zC-lsj%OM~O0jZ^5RL)An;JyogFLhF)c-pEIQ$_>}W2QlvXb>C`A+$6(;>m!+Y1>yc z5GwLMO7zuF1rK69r7B0V+v3jf1<*qIykWtPqs7;K$2KZ60X6sG@U)BO0#uB@i`0B=p@(33sh0O2Xf4Hl*1SqtZ9o3E(V?) z5MotMOxLUryLWeP*gCS6C#jy|?4_65+y7tquaO$98yvd1kt~&qAz1ecO{dIqD7m#? zk4qf^G2Rml#iJc-gYoEEIS`M?Yb98gqTTV36pqC^f|0eJf!vybkI3g4(X}+&~1#7q7Z3u^}@QM6pWl@5NKN-ri#%}zo#4&!u~_so+MB#ljf-oGi!&=(;2gY zS*%ga(hvJ59a<4-;M2#h{rukfsXy}J7~>~ z@gvKKRw#;04$&%2yM4Gii@V!h9W|s3dGhFteArD#vs3-UmHvEHaUU#BR+Imvd7WmJA1?Wvw+^dsc_o!xFQuX=M5;Ig@@{v%l&&RIMs${< z*<@k6y8Kz$Dq!f`FyK8VW1Z7Z6s>J1iZ0sek|0-QI%hnj_)WYqadzVT#PNyG$i({- zZ%>?r(_@X6IoxzVd}aKuCnr9hcoTjfD??HFN_>7B--qiT!u3<|^PXqOCf*e$j>3D7 zPn>qh?tb`GW610f5Pe6q0eI^**scdC2cfr~kUgSpWyYpVN^wto?6NZpv+A{Ysa>7{ z>eCEp7AAG-tl*GBV&#f>_OVNdLP9j=JIrYX*(p*yjbXbDwNZ#+AsDOFyA}*GZ0j65 zZ)U8SoyIY%I(b6jT3V<13}?0HRomLWwhB~AWhn4r^N^YC?8mKvk#1e%H8Ks_!cyCO zFxs|KqdmpiyC~9kw9VlXjR`SH5TR4WaM&rU1!`|paD7p5qJmiE(O7}` z24kQ>+58sETi=}zH>zEZX!w8vkDMt-}@HlbjkMuFCqx(QJ-tSMRa z)Yd6A$Jg!?6oqiOmCW(06Y457@SEc8cDtCx4Qw3DDrn5|nqNT>@+yUN)vYM8C7(xL zAxaFahk$;n??2zY8=Y1OEsG>pW`Zh;k!K3k$q4k8`t0uA{TsS_Hh1)He(5DLlHm9) zpMCMgA>?)5tm!<%b&YK1MusDm^N17(hQ>xyDY&*4{u)d%Qm{P} zElpsrXZna3iv~tQBeA(V4U?we~blPg;3Sr>e`bs@y5i56KCPK?PNG+J06ZtoJRM`n-d@5mGf}r z?TKS}`1ZtG6XzyAn>a}(j+So!7|zi_@d5fI1ibtn{yK?IKu-p~b`D=7aLe(DGw=rc z1icyP$M_T^h&X^*o((5hf7$HgRPz8S9gxl;!QgVaC8~vNC0|KArfh!=LWrH zwZ+RILQAbkZae5h`Qg%PvPChiC5LsCAmtBU7HYPw}KHswT) z3gJNIX5SLhDulE%#qWw&OJdow60t)8AsVQZ4_fF|p*L#BqF-Nz_UaHwc%dc5Av_6@ zA?v7;uoOztrN0e~3kk_Ws49)sOStF&)00f6q)$^*v{b@FNwNBQ;2ujlLYJ1O1UiiL zu>x`s42@kk6PjXEy0nX7&-ip0rln^&b&~6rr9Ic!gaxT;wOg@sMPXJ_v*SSax-Jw7!~+3ATw3<<69q+r{&0FV*~;``syD)1fQU$-2(<#+-6%r2 zJgBbJw3NnaLkO5$Blcs3?l3&%qj-3ovhEQI-0WNYX1}7^{Ke+Rx;no15f-HOte7Ht z5kZbtiqWpK^Amh~n(bhEw%KTT;BibhNhj0!4~M)v+bvz^l&b3;j#EcX7};e}80`M2 z5S3e@)W14;uEK=51TDB_huxrg0GX}Yi|^aKs<$x;r%^LV1Zl3SLZNF2K@sh?3={QF z00&FfUtac83Gf&#UjP5ud)Mzasx)8tU)kgN(4i+Tt4bSG zAguRac%P?aM`g(|RV5{gVm0kpDwSmI{p@G&=lc7>&nJE<6s)OLvTAy5xb|$*0$uve zkeVfkH?VYM>2a){T6!Ek)>JGJ5}3VwDR78o8oiDF<8YdY4cmGEu|Q70vk4d)lfNun z!fav+FK>0OS^JiR5GEu{qRXkp$Rxbr4*CsqCL_D6F$$^o@7%s;Y~S8J0j-KkkbuHS zDyrHXY_r1>@PbZ_Gm~2Pu~`mpmN#fyymd+yxfpE!IN@YQa&BT39#M0-WMoN9+NKUnZGkA#}!IXRn=hn6r^iQCL#-QI|o~XmfJj> zw4|c7q)Lm`kP^If?9$o49U_-rzjWSj0=e{?OMm&>(Mx~1^v0zRF1>Q;txGRmIs=c+ z{_V)6(+@oG*Ym?SNdr68c5&%R__iNhdg*V6F1=g-)(yF}zBc^-z|bz8fdRkj4;QB5 zZ^xqB;>;<{&^)|O1F~#H8}MOBS*DasDt6892y(8L>6pput*+fV!6G|ETe&M?KZ}`j zD092GrR_ks21a+LpCk&ABr4L9gpJ&Gx@u))l&sopWtd@Gn6|?A=ZhEeOjhLLg&D7d zuF=ghEc~w=P#JtWJd!t*ivQ)K$zhi?8}r`?sD%!vX}L8D-XdBJ;AWR@5p71Nf@x#s z9F->7E*YI=4%sw55k0drYVnyF0iQC?gPnYV;ItO{xJar>D%qT0*oY1g_@%CeF6pi~ zWosRAMw3Y;+2Z!}b3H9Jpl9kx!I<9pUp^<)Nfz}^n{6CB{^O8ZmR1sp=FsSdJN9>rf*|{u%p>~GM6v4d61BKv6d9Q+GD$0(DK4LB z&n!?IW|)Y;RMv5|+ zj>Dt%>%*h4Hx-1e4p)T3@%;Ol#b(@vRnMlUi7VE8)0k(8Jd%sUP6qa|%?;04SnFbg zlb{}h8=2(>vu)~dXJH?!2OnwG3fUqB^6um%HA>7S zXu4XgtdOm!qzww*?PGhk?fm)Jd4~}b+HJi)_!g>K_-i$LbiKBy!<$6msDEGgb#ti7l19D?5M|l8UleE z*D9_ZxF=x!EPuj1Hmn`f=gabW*g6fp-YOGa&#W7@r?cDyk!-Ul45w8j15)XNG@x>cn?v-X=dh zdG!02PLOL)UHIFG_^PF%RlfB!S@eD~3zo|5ae=ia#X$Sd&8U-|yMqu;;t*7q;G60 zyakgRu1s-<1r3b5A*D+T@D(MF)&@5YJUe$#qO($_y_Lurr_$v%d^KyFv>q;lJMyO>YA6}lp&G;#uP8?l&T2Zsj zcAi%XpAe`W!^lwW;?kj|r!gw@-qO>x)3tLxsO%AJk-#*8f6G~);{77t(co>Y2h2WR zdVF(`%kdL<^Wh#d;cM^KK3Y141dA}vr~UDM9HX(P_D6qC-XE-giXSFD9ZnrfF> zSH>AdPqtNXeP^hhUj@%l@d}d9GU}OK1FS}=WR;BGoQjKKG zU!(Wtr8EBT4|;(0Z?#k|9reHX$xAO?It~B5bm^5#ZxFcn>fet3?WkW@xl*4PkzRpT zOA3`V4}T^gt;oS7kQUOI_Pc)q`yR_9WlX1Hz0u=JDuPvHlFc%iFWxY&MByqQXwsr@ z4?p_%mma(N@ylO7^ym#({iY7IM3~^S=6r0Rn;%;GF^OC~_4wD%pZey*XTwOMF$pU% zC9H`?{M;-uwzQ`rDMjm}@Pb7?@6#<8g+VuyeSE$v;5r}_P<0I$YT!#<{((= z=4UxaDGZfG4>n}eimK^~n$9Yk8q8n>R7wl4Q))s(U;gvucP_ts<)Ps*t5O`g{Q8xnaODjAKK5UaT{%qf0^Ih|<=5e+qnF=Y z2;On!ku~lho3!*@J+^7W_kSCH-X)j+1cQ0x%0pL<6Bx}S{@9L37V|Glt14eo4dqLP z*S(~YNX4#oyZ7GO?`kg$*Zx?0x%O=BiQ3BxK1=Ai+R3FO)6>ED-8onXlNDxgY7LJH zkPOdzxVD`kpVI9D{CBbT@j~!PqweY7{ko^sH(sorS~@&EeHVd`u2)zu9j=|Noh51( zjNr7Uq;(~k>MaMl_5zHfUQ_qUax4^sqlbM2nvaMN4TRN&hsY)b7p`4cI$AF+hube= zEH!!{{{XYBK_tSG$cnYM0_%i3!H`JwHVqX)Hrj-2YR z5e=dhnE1{|AiP^Ji|ppMSz$7IsALTU!iHvrRxC|b^cM8ND1L-1D!vO*bd($(W{UOjo?>MMVN$7IttzkluTuYU6F!;f8k;hcZ%P;`TPN>jq;mAoV=@zuEnXq|dI4$3Rm)|B}B%J%N$NuZF z%cs!Hz~AVog5h!bb-YZ#3V9f9kYiwffE5DP#xa5)184AYJcEzJr=Esa--X-!H=erk z@Rx@k`+wiXv$@CYiyI9R#&mpTkXN3(K1_I6A7+Ur{LR{oXsn@?%BRDYIzLIGqPv<^t1w2YBHt6 zNbe$>;Noo$skC}M?FI*%h#9h<3(F^^W)<$hJELbb8j}*(-ZKr#TmNy$m6KC)*?txI za$W-yO?mJU?e5i(t3KF8t{rn$)yO2PxU8IK(x=er1w6TR6Bd)gC{?^lg}H>=vzWXo z(0UYc_sHZFnc6u%M)qzaVfc0`ukxUrp?rU-Hc=>fJNJse*U3J#k)`fLsqfn{ky z5C@EM7tGnN@dcqwQvE!-9|T)2G%izVVClErMcxw?(JgZQZ2BsxU|qf2z4~tz(GQaS zl6=47L{j2V3@Eh2Y}=+dQVAu|f_Aw9#Q}zyJ;!Xn!Ot%63b*(Si$OuxIYmjSir!)} zhwd!e$14&X0hmCrH7&a0EHRSN6RC72r4tkO0q_o1{%#^mRHPCZPGpG5Nxh?oj8`3Z z@eRp{D`}a#4CFk*Jy#4*2r!iJMw`xZWG;u#UZ~*34hsI(1S!XS|A^yYCd3@rvirc# zgy|xa%7n62iM#|Ry$X{yj38LKlFJ23S8}2Krc_E%Tkx4Sw0ar{<@&!wMoF?IBN%QbEF$V@w%FXJe3rl*+vobnxB3{nT}t4R5x{Xw|T&<8ULohxy4D2iw`P!3ubE+&f^c=kpL2{wF4TYBstvGH zGKI0d2vhL~n3pGDQeOJ}50NG7157AVQFB_ZA>mkcU3x;vX#El#wBqLr9WdP+NwZ|= zq?xWMT1#DPXkY<2gmzwg>EsQ%S$6FkXqxbWeXQc;!CS%D@-89S&8RqzE`WQ+wvCW$ zCtgH21DK&Sx%SNQYe!#gF$IgVX-697w+GL1z zSEd~NkM5j9dRj@uCw}T#`2QVTQYYS~)Cuf=75l%ldUIj-28~c7r+t}O)E1Q(8MuxG zsBwSGH5s4O?S&Ug=F=1W`4hlT=63z{d@yYEi$9V8VyMKI-Q)8kCOI+9lznI0y zczuzL!suM?ddxPekwvDeX$6bKwWN|x2Q1x&p&MFanxeN9(?;=RUhy52C&}k0KL7pa ze?q6^_rCT1$It%^*H0|J{(G?LpSfv)vzt`E1mFEB7|Rndia&h*LS!NB;$=QdZ5|+P zNHy%OR7iVE!;WITrRhHO&gL4u*A)7cTi1ct6)QVmwR>yxFrFJ!#ksiDO zrSrP;uPeG58$up-v)ycPoZ8Odm@Kfo%^e3BXwk)xN^i;LZUG}evv&(E$at-FW0IF> z&;?NgHlrEx@fp32Q{jRz&%eeKx17wc7MZXBS$gB!_fJF`>W(OHo zUyZkA3|gk4D#@?;6$mS+9YSc|YYvE0R?@P#_RbKXgUS zHam_RR~Ecnu|({;fOch(0VjUr(OK@zGj7IQ|1F0^vf%|LL%V^>a9r=N`MjuT5Rh5{jOQ?e>A&F-~2hSO@jEr(#mXf}i0wOvws+LoY_TMz-x< z83oKWk5TkLRb;zqejjy%&30;_S5*z!I0Pcnlx(tJ>gN})6yWfXE*u~(v%OB%uf$WH zl#*^~W!-?1xo-4wLkdSoQRybGzlN=Ly4t*{MTw&%Orjzp&&PN7fBkD)s$lx}N+XgGSE5PfTPeI#WZ0wHH zlp?oyr4kQ(y@Se?YzPHz4 z2;V(DeO<{8ObqWB?R=NY%`ysBh8l2HLJuA*O7;k3i<36CoFa-T+38r+HeLZ9(9$BPKh^mVVYn-^rDr~yM_Nx_Tv0t8=v4xvJk=q(WkrGxy6wzrUMSz@=~)$qQ-18Ego9#^Qz>-IUqG zi(nF~!O)1FT01nNr`DBcT1jYq*KavA;=m-pc)H{9=^13RX(PA0je;%BvB2Cxs0BnQ z`RNcDyzRAR3$s0Up2sixtUy&L`?iLYy*Zf?xFlG9@QY8y&i z<*F<*Ltv0QMlPs>BFlh{(=S|WRDODv%PXF3xZQ+BP5NDcEr@bayn|hmyS&sJ<(;cV zF;!$_ix4GjE1eXc6~kQ?`FwR>!2Dm^S6;_cP9jTariEB3G@8=)rynN6pcLPKaP<3E zpBet)sT1Fwd7J$3dtFskT{ zw%49IdhNC2*N(geQyg^g*5)2p*HtoAQ<8eKlG$NpY%#mUW_ZcDHInF*dhMsS#S3J1 z@m1-ir13Z;F)dd1*wE(lryEs z+YGygtKjxjjd{D*%Gjo;$vkJC>_WsDYP(f3CCb#9XLC*|dZ=)_oM(=j3&PH%Q;kfc z|2U*vGD<3bBIM4|l~vIwa5H(io3SW0r)1*erEe&g$ME2+F=1otroW@9iq--GeCNk> zCfuq~6lJ^X5gje5Xf2eBcmDkEEyEW-*r9ok-mwnzj%@-hnZ))GkL?Bk zbGO&IzHZkbK3d*i$xefj9l|KF>d+LUWuIBqAxKhAqjweVqqEF*=2rD}XhFEruz00O z>?+jN*ygccIXE@7yPoElk`&IA@)zfyY*0wFaWE4q>e{t>;o49iEoPni$g8a(+ z;{0z7Qs8u92mj*y!#4T!xA-=nb1(2IyMpg5`geYL=(R5oJ-@~p8`wHF(6q(|wlffw zjU|WSAq+$1IzlLVa3H}T#`?oso`~jYioDv?L{7=6N;0cxsYKvx%rxj6{l_5}Tuw2#N^>= z%qcECN?=~#*@v}L%XvtcyV@C?s`|X1!?~$X&(a~7o0Bjzj}FxZnHkhF(T^Bg{QT#Q zg`+1LnR5SexP_Ahvnt(pP6kHi`8>^2cQe`LRXu~2y5qwqH6@*Dj_OCvkk|fLdl?q! z6SbEYV6h#6r3Wh%^W~rTOY=Qn$6H8Ry*fc6DD%nM1wYOE@j~!PqbB*^{hH*}H(rEw ze|UQOE`r;~d$?09?+24bQQ*?+tx#L-4VbixxCxRkk3LPZWK-<~OyVKXB&0#ID{<3%^Fc6X@o20!@J;#!$UgEkEf-13)@t13(h_DAhh6 zcG!Sp$V!W^v*D@53$nzkjKDzPhky10yx-zPShQ+U z3F5-P{1cavl{-Q299o^B&Znnr`Vaqb+iP1Z`degyZDTf*7FTnb0JFhhZtW)g+&Vq@ zN$fFBWZ;ZIO1*+wWf@q`bxD%x&sU(Q1Uu5^=qGGn!RXL^ly&Ekj6S$n+OO+$- zr`DV*H)%+TOqLg4l^ui?y-6mMHEa0KhkbId(yn|l^3(_4c^!F5U-OJf>g!H_ zTFGhs*8E#!kbAS#^4hh?8>U1$bZ{b-QF4hs;`sFvRht(i?`m;sDyJkfeYjXZE&-mx zW-`I#XD=X8Rd>!IJ)x+%PBg9eK!#I1`XSA|Qf9 zcp0clD&4Oa?o_?-vnRj%@Wl7;UWhEyPSywO2yNN7xcz<3VsbN?ys^9KUMf<@(VFvh*4+Pfg>e_E#8S#vy${f2sbB$u%Frt6}Fls{4a{a78<3u-wz09B7O?^?-HE0<7#RROmm_T}Y7Mtn1GZq*&-}mbY&=39e zl*=k?Ga0eyXD_fKj~+tnh%>6D)<|?l>b`lQc9CF%4VIlR9r0U49<5y<@Zgx=d2T1tfbaoLu*Y=v9G>1tRunNf0?=CEv^QZ7EF+#6RO@qbRk&ncK47#!Jj<)JG_ zU{oK%*xn?U&s{lo`Hjo(k}D6wtHw%IhAP+afBugANnbo zq_#`JONJ}lpkzOHUGxq+S=b+APYqEev6VZ9=QuF6oOmvbhczVF`r7b_ZC0fyXQKBw zC&(UQxZEKDw5(jCj3O7hj3Sj#a`BNr?>OtdK&?E&e0I@(#|GI{I~a#!2VF2fsO1y1 z_5{S?YGK?hoW(bhH+S)EN+zgXpf(4gL=ORh#bQJ%i~rCqBLIydzcI#;a!B1G_(sgVWO?g3=VW-6dM`rlC_1 zu;4#?c8-H1y0z+*JhG`G9G4oL5#}1oUA>?Q>w?_wqA)ng>=GRsk#4o*eRd>zm*j3b zBV@q~_@K#b!(j3shm=)P(bT@hv|=&c$D|1+ZD$Ooc*>JfGP!;-0Mp!M&moRyo0J<} z#HUaB1ikTblkOamjLZwk;+HbED}}70#y`$CPP9yLYP-d*hRgJvqPK83-T9W+&N)%= z%A$LwmyW%2x)!^4PL_#h`$X`>0d0UUyP4oSXR3q}iyA)t8#`EGk~7<2cuZ zSy9~@2W9LalLmJf;kGL*CMk?4vgF3A1pflz4zY!c>C%8n7CD`xV^3Tz)d&5(e)_x2 z_O1V7wTl*jjwGsTx-FyZjll#V-VH-xAaPj;u^?8<&v-+_E?=J0g0pek+zu88G4gdshV5C=y79tP#+$cPrHb zhR}zwiVIGVD|HBW@9gi&0&Oun|NO7yDo6oHqD&@c==ohVk>pjpECLL#W+AlE5 zC#b}=dPBUpfywOyO(I$Kx#20AZkGOyQ)8kCqY*a2FJ`gj;gM0~{N%0&9!K;zsj8_; z#wTOa8d6I&6fqTE5mQPc-BMQ{p$uRIJtu0LTw-I_ZeE~^uP>fq4(zSmfNxJtx)y8k zz{uVf$&J_r;fG;5Ew&IYuOG->e6uyU?Oz-APXSjrLlboIZHGBS_3t_~MJ1b|(QOme zH@&k;Qb`3QIH;|i57O322HJH7#N)fh)8y^%T34D!*u-6_69my z{J-11Shto|bUn5PZ74bMa#t+?Evx7)*d#ZOBmAP`aEsbL0PZS*F_XB}idf6}yl1k- z`Mk&W6NHm4b(wQZLI-+6Uw5l3Gsz8VlL_vUEilz&ib=U9=<>k$&VkjUL3^SIzBqo~ zvVhWQC7V%H@D|g0K$oI|FYDO5sAgk_#Cz648}$%Gsbg@8arUsJrl`5*8s{h_x3y<# zAJ)zBS^S)AqlO#fJjvt~H3924ol`V5 zz&xil^gs)*2uUTQHV3hzfW(jO*ot7aCbJ3S_B^v}c(i-*729Oe_QPq+bzyHaZ~G8H zQw@SAMFn##J22${*T_~{jc%mvqE1CiBm!d68)DMKi%C~f@hhzEK-;-UnXs7`)5+}2 zA)icxP@2u=C?PZdp!PEZ`6`)!S)dX>DV0IJ`6puXnVGfX&bOD^tXok%BJ&)Eg(1M9 zee72rb6jU=l*S-%3{Fo#Zkd4KGT|kcRdU(pL+HL6G0_c7iBeTGr+B`3>d)VtIr`sc z5B>eM*KcqF-96EA0^Ln3zCqLdtj@QGPJMmiFJB#b`>O{Z{OZWLua123)sa`edho?H zOfxmvY?@7d_UlshvtKtXGfh(!H3O$o66~|I8jw&_Bpe-E1vC9fX9zck;>pxKj!lr1 zrX*X+uR}Mi(ffqw5?;o`^`8h{7xwMH->nwNR_PVQEHxqvVMKDr6)p>;&QvXcPp*aR z6VZ`YGtttP5V{N6$o1Ha@8d9mvy|A>h2!61q~LLrx|ECy+ht_Llb$^yWDYSG--PL(WfDAj!e(Y*%kmcAWY~2f2@Kl7 zGMFCdw^}W`N&MD$iZ8w?EK%TK5P;`E_6y6)^ElDz2}w=9y4fn8MUcv_L6&vr2LvYY zI5DWK#H>y1OiUn^Q*xO;-H2Z;T1FsO2TntnxqO!jMs-!m#YexngXA&qGJ}G(CBfEm zJu|kg)vi2}?c|g1NO@GMl9ut@?^aQ@t}D9U-?LWQK`IXOOd*khq^s>9dPdPQ%^pdc zRn_BQ&t3~c-|g94NH7NA1Heh%biFP(x?CnpT!L(a?oC!* z)4$28>q;f3WLpg7KHge;cx!(qm;ZF-AsExK$Z~DMQOh-CnhTj-Nc*&onHD6d2uvFH z?3)_fGB!0vwv6o?*|YVYk?nhL5E310&00@Z>J!q9km%UFOY;=$h(O|J<Lr#5_5|8YohX-cjIcsKOM=%c!ij7=_l zdF+Yl=~2lj)(qg_$sBi!Nr1J;AI4k$F|&`ElmCVXH|G* z>gR)js}DTzlbf{Ft}Kj>QFFp+9xVM`br}qgbY=}|&M9^Owmfeb?;9AgTv4Hp!CNvIqn_l#;qKy#H&E7{Q!UE66Q-qlJEYIIf>gN8 z$tD*j&4$z6qH7$A1H12OPF=vlp5>s|l$f0C_6(mirIA|Kl1fI&Bm>8!x^_c@x=uqw z8W{@vlqkN-RxRKM77xi@PcDB->(@?O*T>x$&ln&X>4LlYJKJ( z*n8&reB0YiQg4Vr4=sX}qUHKvA$1YNtT$|52>Kdf>^|7x%>_GD6AdJ9|8YoJfzg%i z+e?iCH_=89RVRu&(%8k8O0|WpcW3la&7$RsJR&jJ!8(Q3o%TE0^!}kfuy~|QKK<<% z=O3lyS6=mtbDy}xT09K{EC274U!4DwOU%U=+vnhkr7#-UUCF+P>S=5)D)Fh?+I057 zhu4$M&PSBO+ymSm+ASQ1VBiPqY{)$o(~pyFlWcIZY7%g+mUBdKg;0m&s|Z&|X1ERW zMdlbQ5$(1&5LU|#=aGzT+f5us<3Q_771?f@-$(sM`L^VzD{3}0Gr_>fvZEWQ0hOYR z0Ek7~o2Al(4Qm{V*7yj0k;{6-QDe*zpIYQTyMZNzD5B^AEwty%95P-)kbokCmeWB_3^fVbH|ES*XVdr~b;+sOw*Dnn>4Jo?0VR;NJ zKeA_gK$1zdL5)y<5>lF3MQiDypD)HrGh97??wh|ndiB^tSKs;Qzt5ig^3aKYu?4pc z%QM-(lKbyBh2h~F{Z4@g9w7hn4}*i`>hWXWzVyedk3aJF58wFagSW4~aPHg3kAMA} zSO5N(N56jZ;??6Phv02j|Kr_nUp)NZXWu4+ga7(3Lyb@Um*oe-wyOVn|NZ~S?O9P` z!(fq>F~#jHNJfqNpZk}g`n`&xe0k{kZ=OB+_0uQ6ed)}%&%b=Tucj$hM&7;v`_Iai z;j54T`s;`P*@fr ztihf3jF4rpoAy><*>ERd>&4*AN&EezYIBe$O2PFccRhdGA&V~(i;@`$D~KC}QWL8P_E8|y%6bQC_w1wW=rc6=^MehT_U$6YNK|?HN#}D?jU-p~EMntX8 zMIP9}T!v6=UYwq)&h1{Ql#*<*%{HJ^mNlg;q~wmUyRIy0CDl(Q1eC?gV?>~Ou$!ua z!Ir)W)=(Aj6`5@cIWxm!@#J<+UG#*KNjBSHk&tnDq39N2N?8w*Hmx$|Hg+*d zV^ECMow72uEHB_9u3c2NZcUl4DLK74trHrn2v)){8MECPCc&ZCKvNDdZWKW#Qxbgq zdH;#MFz5eXb*<-AXz@a!5Z$NWCCyc7>get7X{1Qh`*lx^ZH$5yW^{Z1N-Jr#Pg@x{ z?rm;S~F_CNt85iGV}BE0v2G>{e^#rrn70xNagku4i>K z^^@^C))M`Yc0(pr=?EBLH)Y~?we*t9@%M~dq z)q@?Vz>X)7w35_f$ANmbngjk4H@XO~pTRz=RLjJN4lQh{5=VH_SgSVr_Ogtda3!%a zdWO1WJFRju#VoR$+0i4QThVzhR^YSp0gtrdy<|gHrDcY-gpzFOKyboF5^ASw7vTrV z4Nj7k+P&KQwToE4xAY8Ly!YPW+Ow_Khifmv!%ymEeZ;TwJL|t6USC+}HNTt^{^kYv zsOOd*#8+3UE6*++tDPY`ggZ*zTce5{D|0!7e=x2}iH1Mt=+GkAqFyPJlEmQ>O$+iB<;_^+Qn z`Sr`sfAicsH)%WVSQ+_hDSY$#Z?9hb@asqaBl^V4YPp(cjvJhx>!AWIaG?25LdrmI zr)!By3x^hBfzIIZw z+pr9lxZRs%r-|7rDRMowc9&@pt05{d2Ih@SsYvV^6!m_z! zFdtlh7INQ3rpR{md5a~C5@@Z@S=DQ7_Ukqok5(&1!OgrV=c7wZ!faTu!&$hQH6)cz zH+EhBaY!j;l}!A`qn@=+z_!hEx7evbVGNMC40f@o+2#mUPw%M+uUem%UfMxVz}U{$ zyK-yqZ6>wn{ixhUzd!e*j_;_^J2PRWsEXc##<78YI$O+AHaH9BtiwCN2olTBP9>FO z-_LHyq{5KGh$Fjkmjt4>6V$zP1hQQC%xxyiU2Jm7_(W`(eu+^-Af_kVV9V)h?!Q?hz<$TqZRKC9&713dKX#80@~60QU%p^Io* zS?oq@V>jyA*&p*<5wokdL-U?g)Ku)^{GOfn-C~Z(E^aqj?B>d*+VWxU2%-QcO0e}< zlsHC_BUMNYZCH~pAf8Y%Yp|zoT$`vMsZHJPA}@hZa!NAUhsOvA#o(63%UujhW`|W( zN%k=ouL}k7Z_TcW0J``&h_`aTfbCvn|w+q2(oK2|l4cCJCRwpC!3#fHe7+Bd)lwotN_+r{qeV90yk62!SN z=hg%@;Ru{-Hps>fx=h;CBDi7Q0PS?ZZwOdzgxzCvw9$8g7)?>W3D#J2X~0+VH(a+$ zMqzxCk;$TDb57oqYpBY==-xfeX4t5Z_8h2K!xGD;6%|oTbE^!obi;g3hc=%zUD5mX zO@YDzkD0NnG22fI}qnd49kj+qjV0R~7N zczZ(OuYi@9(8RNdr-Cb=ljMfq(C-qD&w~bcVZVy zYKoTaR|TRi_URLgZ`sAoWXl)l|D#MM|MyEV7*RW6nI%(-mT9g>+kheoIQfK@9HYn( z%(9J)PP=l^m29@zhYXESdhfkExHZFuf9W|Cugy@$LG(_~Hp#zvKKsExj(DyZY`S4# zhVr>>k+RnP6ehVp?QiwJr`bd-&fxDp9AZGp! zRVju9Xhu#~)MN%c+<{v1Xg2LzC2Xe2VYN;;UqQ@m1R)zXZz{bZ+s3xy|4J!vRmmsVrG|3g#^? z$;{#h6640DS#H}FEILw=q9AD*CKc`F8-$o9$66F{a*mnIzAZvaBbMdB=6E^&3d{(_iYF3P$+lP3dq;UV+(GDLi`(7vaB|pO+Db=w=NruXctMzPU6od#%IK~N z%)Ac%emiwlqgYgB)Y(dZI)v)sureoxu<#0eUz1RWfCcHo3rPL~GIF}Tt98m)M&xan zjXWb0RF;UWm--Sgons|Zx6)Z|j#8gnCoZx|4trD8WOI>8^inx$`Ka2`M7fgd#F)es z%&~vs!izIZl6;lSa0$Z2N$N3IXCEwG&R_*wtW2XzYqTsA*}!4cIEz9{DyO8h=7e`> z`1*aQPOt+Lzaww=o~;XhAA?P)1R)6tj@KN&4~AZV&3$^Bz|6p{+lPNSzVJ)tIb~sb z&h+$VxVA%>!d`sahMT@T_Pf9ZvPqOAbI8i`yQ1IU)in`$N8Z2qw)^RG^NvMl$$sWA zYw_dA>tbiA#B6SqOp41ZutDougU?6{yGGQ6l2QBUJFDuG^0ZLwT4@r9L{pOK<`~lr zdn;S4mDySDh?}Gq9V-{w${p~Wupx>Q&xLQ!8lq@@ZFtO<+zUS@`f_D>vMBZDG zt(#gF88KE>19mBO!*+V5*;Y8PwoBRu}n(WR%C9w!J| z`1I1@r6WEc2wv%6`^T-puz$IBwsyg%mE5%S(9(lTM_?F-qBnnFhUWe)XM?){w=()PaE}S>UJ4(c7cE z=0fdR#Q^&Z?Y}d zPt7XX_#j;yB~wi1xLJ^3qjwJnH-|madamDhaJ|FO;d#%Zc6YhXK`fe*NHx1#pXQdmJP~OHa@FcgXO|}S&MyHF@MiLrqLaI7EswdQ8c$yg2vzk7vsbk6E z+?K7G;VoOUTie>>w3L#}HY-k)@K%^E>f70b+X6dR!4_oXS04P?&16z4kp;C2Y*@33 zVE2LWL_-*QcxmX0o@%xRm=l%}UAV&p$%s5ExftFinPUjJI5cKkFmz`Dwm6p_?vcGEdHc#P z=U)!glWolp;@6y>zWVkvSI@oh_jlg-`#a~pe*B56PyG6}KZn*OKHLSSo0m&Qme!v9 zjAX1B5LYVDJy?GdT4cHS#TFaJ`@yL&m<6k`LuDAT4DF<*1v{LcF^69Fm zusWmJCyEli1#y7&Qx&~nYaa*3swSU-2Ua z;5j(A{Ylf@4=W{NXv5dvBBVQLF=ueEEO;?QGEw9fuT&V3qQK@GQpq$j>Y309^^~e; z@sE@{M_+=TsX?7e*Fpt|7SQ6D_B6HlerlQA$!I;6mwgrn@M!G=7^2TE4ud%d!})0Q z3dP`&+F7`UOylnnnCbIKnFkl}_R}!SNTPRk>EXzW=*LLN8+7#{+3Iu1cb#hKwT@H- z>`5?&UBAp1l)VRo>U;c(U22hiFz*hAYsC!N!F?cdgG~5Q=f?05;onurZtw(T6>ERF zQ`}5;3lCFSaVqJ#BPp{~O>f9iYj^-!Xa^vrgLM%?gLwXF|b-Ha175bcd)MH<}+DE zO(<$sRa8CKUi3YcQS=rR;LsM1Uj}8t2?~xUsZdEqq)`=iGJ`cydxSTOux7-m7vtad z>enbD99W~UaCk!UGN>ftXq0_ZcO_jH?i<^-ZKq>*c5K_WI<~!IcWm26$F^--C*L{0 z;M~>BHHysE+Mo1p%N=%x}vXFtH)hnE3 zNOecjR<|50{6g<540lbuNkow6=oP~fyPtfXgx2VT7XBS5%z>b;Ie)8^2vCmReO#Bj zxV4kY78m7y+jPcFhc7k!EukSFL=Pz^K*H}p%Y*uLptPioyy>|{fxTT{py#?Fjo1gr z>`%t5Daq*z={vQ^7|3?RIzhHnKsbE2032s}PdaRx=bD7UmZ`GLE?m+T=ka*+aWG;A z;UeR?wZ#7CR(k}u>3ILxn5ddb$05R%ZjcTn*-_bPglcPlig0ovqJ@&9N98nZrA?DT zEKC%g_E7_bn4fDHU$rx`eao6AgM*H?V}pO^K6*46YHg6D+?ZkJk(xo1sG*6q@;&jthjQP*#>_r$e>AO^_{ z#7+lu?z@`6#^`!?7$$Q3vGWh=?@*0rUmclP|G@E0D^Qa?jy_A!u!=B-0 zX5n)+Fg)vhc+llKWrtfFBkFI^YB5{BULNcF_{n%0tWCiNz}qX%pYF=ot)JAko2n4W z9g#+Pcb*0HAz$s-paG{KZ){Xlk)XHzP~|?XM8;0>kVviQ(suguaxl;|YcKyZ_POKu zAqXa7RY1u3n`e>GYBCDqPjlM zD~s-WuTQJn-k($Hc;v2j_6JEdxqm>uz9$=gpOZzJ=yeS>rL4Q$Iuls7EAJgAiMhP9@sC0boTvr+OEp`(to`W>>wC7>T|*l;ehEjl20 zEG|~Rot+P8w*kug$tbiZ#8Va!|22a@)#`{U{G>J{{-ZvXAu{$=<@Jvc+tF3Z%1VsM z1#fj848X(5bd^%AFFR4~;r@1r{*Z|h!sSfCT^p7vci?W?`75?EMky}8G<|KLlISS$ zURPae8jAJ6VcfF1;Py&l z6ITzhAE|vmF&&vbyAqvkIpx#(?T zLe1DvW{VKbK4udY_~mwbHtX7PF=2`I{sb+z&;~|Phb_Il9hS~i6U3~%6`X%LvLP3@ zY}?vc`ppsdL$BM3WqVl10wfS$L$Aks97!z)5>5z)2O`p7>nV#XRNaP#`cdkq%>+wr zMmXTVoQ((Dyzk&@OKrDxwabi`mQ;{zbpyzgUGnuuy!l^F=4dC~QPdk4ladlG`^*wa z3l&OVPlNOVNd1N$(VRW-TK+|^1dg9tvciEq&foD$V7WnKU7c}EwvA9a!=s4GD(%W- zu)v?vQ11FS8MuMc7IH+6H_i;1mbvs`84T6XGx#cO%1Bf?U3`^uW`C$ijG1Isolvhn zZ{pv6>&oZb511Uc!rR4|H-u%#!85OAFZm>0U?wJmdZ}0M4!^{SN)eZspnZC$=%cBb zpzkN#m<3ITza-SiWzn{m7rxh~R~!UF`793wKGNdFeS6|tH^HVj809~@7OP1wV>E?; zc!9C*j9Lk=L{IoH$7&qW>wQ}N=5{!98lx_IVL0!qKws$XPfDBmT_H6H^Y8qZxb0so zgvWUS&sh_w!=f$UuKgtDsg&JcZc`-37Pz5itCpF{i1N`pSF z*>Jh*M#T)Q1iF4CwECV)Q&bhA)xCmu`@S74GYNiAI$yiYSB5jD+UtE^E|=)XROi)e zRUf7ALsA{{u)6X5hkf{yqvL;p8?uMz@I8oT`^x?PP&f%2MWnI$YvyN-n#qJf#YJF} zZ`)@!@-#_VqkMBSmEc_UPEm|I;i{mkab`t1+QU=xS8;#szX=~I90v`Gq|btZy~6Yc zs`@G)us3Mgznv|oRM9>fPr>JZE92k|bm|rX0_fH+$V1dJ7_wFyCZ4cg!?c~_2Qh`g zV4d2?vf3wr2!AU--)`h+E?5sK8|KSKLDv|}gdyefX7ne)_Yu--QJ^ri2T{|HK%L3x z8E_fe2w5~=&FX@;>Dg@Tw*tFH^t>}At(59jHC{Y;FN`sxRFMrPpya;#{$m=T9bNqK z!AR1jzwA(m3Q~9lkJ~B4|6;B)&I4f5?VNA?xcoY{UFtGXp*qRYn(2WV2_boAXptGp z7%qhM0mX0(%cHc%BG%$=2*KY1kfHX;i!96<>2IZm64~+{yc%%j+dZk4YPjZ{Hp+*4 zbRnQ3@ctMgSE~X%qmwy_Lv}q0PPajtL=fJPkfyhwGyxr#sB*(us4e)*4KzBmsm9o)QQpF*vX`xMPjQ=k!AxNQ`&*k-V+^_p% z>d?K0kDbp}mb%FK|o(-3WyG*z~-fF)3VFD9iN_WM` z0o?Yt?aQB39pA^hj#PQwKL>a3kBjaC1>lk0|y^nnxbSKxGysc z2kl&r49p#ADQ)D`qlAgXv@Gm?HypidOGDWiP!FU*u^f_hs4zD)pRJrZ8{qtM4S!r0*uzDo0nKH!$1%y$#&;W_rFru@LxfV2jc+0 zl7??K3TRvP%Rmb}5mPgJ9X35nreBI6POQ#{x1BJiXRG*dttY z$3O4;V^ghjcdm!J$_=z|xzL@qDbW>At^ZXywkMTaq_lZqOanM7Q$umA1z{V92! zhDFghyUEu?U(!!cI|e!t{@L(;ox9QIYBN3hlJlb)f%_Aye`juD2r00tyC0f9)KCnw z+}vj$risW68VDT>BT6SfY5EXyA@cN=ZSa;R2*y@Fdd#nwZ$PqCqrEVE?miFObZaE; z@9dIWiy{8{;}&dO+{%h=anwR(hEU#dNzKlP8`sp_vf94g(?W7Cj;P zzT5WOQ9neF*VaPU?A}C$gkTM8ORaNDqgxxfS?`d=Mi+Zaqijk1qigbLOXW)+j}L=O zw3lm>hv2igxv&w3oktnLITnFRFXjU<)1SlQ1_?rVvXXe;+ z1javr=q7HwihCg_EA3ex3g1LRA($K2wWC$vCOCVE9@Yj7J7vV^1O%%N~Qj#=VVa9!M>^3(ZPP*XaWq?(@4 zI*Ic)dr&)$eZ;f(fp{q1hh$1#L|g8Ce%ppbC5Jkf@BtCATP0nD7y9 zkg$zbtBa-hoa}cPL)CFN8Txv}NpFdM}jnBvMnH+WTpkb-|Q)kCIJxNl}!4)gC0#E?(;n9Sb4+^j*x=P18+VC&($g_Bk{yCou2b$&dLgA zY}q|cqY-7#Gdk${DHqS&x>GnsHK)zEm`3lsTvAN_4nEAuwHu#r4(rCVkraKvV1*Zu zW=58eqt#ric=(po0pbew4AKL5w!&vOah^Qj)Jt=w!}??93R^6NjYZRU*}?Rr%FMP`{gx|_76krb}uZO*ASU}{qEp8OOz}Pk^0pIRyQDbp;ljg zd?-{M=Z&V%v9CWdjZC~2T$94Um30K8Ij;V*w|y#=ofnQgx0Mg9c}(Tc7&|GM7@);|h!`6)rUMoTI*_%?o$&!=b*+#+EH|@K=2gQ^c^)bC2p$ z34J=?7N0rwhLD8XIJ-{K8_=NwA`{Zt$JnHaWe)&f;x>W}C_?8D_|I%Q#}N&mvI!L{ zIS-B53->9HLTe(5+T;c}jtnm8v3@DzVFF`=f=Trfr|a)W8YPU^=OU}78ur^&Oc_r& zvc9qy$@)ygiMz$r+RS5o>0+t*LhvB^5=*IStCC*9-7=~DkhjQ5O5!}1NS4&GmATYB zqS}GZdXzZxG>H1Uv`SCnWvn_h@PQnf)(ztwR-pGWTHe*8$P0*_lIo7kYyE< zwG6k_l|1RuyqlPg*LD75dcQ@Ja2?wwCw8nyl=fU z9Ety8pY`D6`M{^oFpjo{0y8GFGI_r0;6uMZM4BRn|WQRbX@ST7U zPsi6yv<@g+o-v1(eXiI>ak@d{iM&C-;xGswASmWbViFPEFY|kOvR@~m1B9`ouZbjP z34}e+)nCWTXX{ad6{}cSYvR;sz!UY=d7YXLG#7rB4d(>0k_C%!%Hg!tWLu?Wv*E;O z><~R$Mx#-xIp%kbN1o~Yz&F?6>`lBV&nUoV`(JaTeR98>the8rAuIga?I+al9>2%* zU7{SG+s*WrF(i4y&#xl1gEeKWZr*#$*WDNGWj$Ws%TaCSq5yt*-Sr~1l6Woqw*MS#UzG-qU-myu%0YgbreJ4D2eD^6F z4$?_^d4}%p^k^0HvfIV0%IjaOMnS=ol<;|xr=XXPjqk6C$q zv&}guuO`>_2*3W|x3o!7r)7@5O2(q7o2ARNn(BIDr$*awo9(%@u@^ry0G;kfo?4h< z`fPsMGE?$`4Snq1zW2uD&B-mz4x>8$(4Oj*Hcust?X>fSFQ-8ZuQ2b2Tmp{Xl%IXN zMK5a84uoy4%Ku2uEJ-5IEQ<&eh{d!p-QxcGEy(p&@@u@Wj8MO{y1|;=DO|dYJaS@* zMIv-bv5Sv4q3wxV?Qb=Rf7&ycK~E$E{Ffs_$`iH2DUfbxYQNKl-gPDCH1 z2Q~}dmFDARn%66!_idOmDzK)(U20eZD3(RIF=z3ZXe_MTF#@@zY9hN zJ-C3CIuL#$M-{VPHO;L{J|OZuPH;x3+Q{0Lbk}g_{}vnKje&6bgkDCTZ(CkHK40$Y z^4V`kgGAr&8^-hV`UC>+i@WYWCXH!*evgw1vKsHl9qk_%$8!Z#+%`GNCceT&GD+8e zqLK8uHHi8jXE(XsJRj#Zu^(R*WO3fEFBaNu-{hQBXAG48CSqig0%)u>H>wBDAyrQ7 z&Qip1^gH^7xtKXQXUJm%ST26we)rO*S5CZD6$K+>CU1}^zo6WN!N@37I6x+*2Q2pE ztOy_a=%w#br-&zsnhb!8oYGo}z`Y!nHN8eDog-rhc=ra&aD^3#T-yOB9@|P~j+79U z!8$U=5E<_kIcBh16sH~esnXMP)JBULyh`jauoP|HQG2{ZJzs1*7VfZLd`ym~vPJh7 z91R_`)kC}HIotZO{&IxkYeIneQPnDPnq(qG2&D`n0-8dy%1b6^0eVx)P0P%Fxe(G1 zNrLiCsRoFC%J0Z^&2s%8x*DFHPb_7hFdNcYGR&%}gyP&)CrLAcW@W`r2!rh3jwcjJ z(i@7>(3%IGQ;(Ee85j?sQ`j;~nne6>f7e!vMm{~p=M4Q+i28bt&t7cIQyo{mdz<@) z;RgtgKYx#N+jYYuhpZv3WodPHy1<*mGIk&TcI4>*2h#-3j5C#;-0qfoaSHv~Wu8X= z1BgjWh@LW$O>9q8dPn&=QVJ5dg)6ChVN@?AAVUWW+ zf`3m#-^o6Ffg>w=sQWRnrdzS5`qInM*QXa9dfP3z-uL4U(&(wTa7;O@=8tm ztZP}%y5`t?Qm@0|@~(+(S)j|UI_l$0y#Q?}UD8A%#hjD{ka?PwfWe)ixZ}s5^mQFU z&y1%VBu5Jo`X*PqW+|r8Zi~VR#u861qHi`!+U(7 z%tkqVWb-lDzt84jt^)DubBT2DBX_6H?FR{f;&r|}YjR%hCSKMH6inScmTq;~WePK| z&Q2vSD}t}PYXz-Ing!(w!H}2Wi=s%mp~re2JPrg%&1!UfPCC88uNWXDT=6xDp#SA$ z^pl8@jKF1LxzAKBozHi$xpP)QYg$q=ek&CYO{b~lplg_uGr;|@Hh&&nY47}P!2Ua# z(_%vK-1 zafBuqEw%qHOU)Hgj>gvXxAzqZ>p`04Z+pt$JtF&L2S6bgs z1&|{%nk@k3LcwWSs}4Zqq`RAYW-Q_~CMbp)@IX?hU!c;?xMpTZR?+4IJF3||u7e}2 z8QA}JQJjP?QFm-ij>Gf%D3CvVr%?kP&u31#I%%<_X-c;`F$S>$2!42YPGn(6`&|~2ZymNkhZ3bo2flQ!X9q+v z#3szT?d7IdyvR#Ws#vTh9$bZp1CgjZAB<+u=h2Qtu<)iqChhJ^A- zQJpUP)oT6~MxYmk=EktLOwK5;bDS4M;x`Jsj8es(p!{XG8kiYvf}iMK*#@LOhdvc^ z=bV8b7CG9YNO728XESHb!}%Tu3B#+NzKEa-KlDG6IRSAgOn0UlIhMI#u{{ z067E_^!guPsd7A;h@X;Z$|ejNJ-jjvS)=^x({sV_ zu~^!~@v-2$=OCuBF_ZeJ2W4fZ6j^fg9Qysj%>b}Nb(Kn#Xr4J3Gk$Ovxnt<2@JaZn zghphu&{XsB#8F1ITA09IgY-@ZEUXEV(*O8$A#0FBV}xpHQi_>>2d^XfuTb{51W%qp zkn%_nmL)8Zk;-#P!~O_kD!9Q>d9o`S_;@@C&o$ckcnIBCC)`{>7wUBO=;1EFu`L{F zFn=_q_u(N@(TYzv#wvBaTx$DrxJU}=O%A@hFa0BMB|Fc*0#E5s6_)#vuzr>Vbzh(4L*fKMZ~EyEHWWxP_e+_58UflFX+k{+jywB)ntfZp)B8P0w57V!`@ZN| zZI*ZBX-Tqyn&sBQk44v73U`;LmVIr6# z;z}YpaFHSDeH0D!ir^mN#5#l(@+KSgW$Sj--A9$+>Zs9L`drn?>!}oZ(7PCI`;e3I ze~o|+6cbO_{Ksg({R)k%8v_&&GL`-O(1nl@7FR7Q4gC1kaMzePpgomQ)BgHSXTKrv z2eH8e3(2R|M6Gb-f1yxAi>_mxh2XQ-Ww|_jBfmjzE;1Kg)T31X=(a+fS4} z7t|H^w%f*zcR~^>EH$MM{XaNa5JRl{)&aiqLcbq5n50u1H?3Om?{$k$)#6{s;dm1kQcm%aKA}Jk14H2?M6BVxhrL}UOeDWnZZBWQr&2g4 zM^neAyUFM2ZNAU-CAcL1wGL;;r|h8tqSpOC|2{vp!7qFN<~WW(iuw!_++6HkR0Pm) zogKi%jvy=ZpHi1kd!rL)lMXc-z)iDmiL)7-C^p3kuC!=z*Dg%m(x?Nv(PDX zJbWt_pC&WbJ(v{Qj?X^J9u?9{9Lsyd|1rP9(8ubBVpZ&XVwM%?lRX-27~Vv~%F*b0ncF3k1}#_N z0VB7GV*xHtQx{&+*!wJ3E>R>YtHChvVa*W-GcR@QlCde!dl!x8ATnYbl|eHs94s&~ zVX3fblfo2EV;u;zEf|^YBxcNh91EG4ts0koyp~N5G>n{~`nXb?cTA*3MoB&-H z$e0SKOEOiyQF2&Wz8Xmv;MO2*ZmyCnxJk>Q))GI_r(VJD1Ot1Bydy3!ZmN(puv7j8L4Hmri#K$pus)aRXqdjT zi(xNZYcxpo;4X%^LdOx$)Z;!dbL{5!iC~Iw4gaa2{1K7Gv~9H#c)&Sslj}PfGb^`V zQ_{pD&WKb?+E~dFmpW(I{%0TRU`f*YJ0^}SoU}XWsBNll+aC~$k5`xVJ>oXkb~?5? z06&uRB%;iuf4auu4iN(vl)fI8E@117iLal3hl6oPP6z8U=PnmZxg3ysGucB{=`UO7 z@7^&Y%R+vUB~D1Ns=M?Tc49)AD{~>H3)*-d&9=z<@~}940G9PZ5gCkJYQV{g7Psd5 zrm(n@3aQ7GtIh7%F)iNA=uh85MIx$oJg~HDJt{R=gjnNCBk=43$Q4xN zA(O9Rpp=e%vPrd2xWJULSz_ICbf&Fau!N;dMBt7_Njr;Igja1_6?FxH_KOZjSn^#8b-n zxdF0{bxw2rM`FAdYrcCCDMr|}sll-@EA3OFgPkmFfscH9md<+swY$uU!hk7a@x#X* zm^k`Wo<@TwS9e;dr}6}l_xU^nW{6@DCI$uLK*C!8Nbxubyw;HH&{E}DkCVFtiPmUl zEur@Ney2@w2cMBEz0U59_>K|qTcblsCTzM_^b+z8m5=_Eag$9hiI(uxR;(qNVxIR} z8bl)vzQ`mMD=qORv_x+u){)1va92@af{Ofan?-Zo#864oTU^i(%3O-K)MZTXt%Gv1$Bp5 zEFb}wehD(t;?Ngs)-nSFkGgO%9LktxPv15QaC-om;;?aTGW38c*C)nMEuxLuIr4}9 z%u}cMXLdXrE;4!X6=TSJs?WrWE{2#7AyCrcBz=S^7Nilk&PUaDL}C-iP3!wFlTOw) z5+S+l^dR%ZRGCys$2O?PqY=)72kpV2i)DAoK}@kw(&ZfFE5@K4ZvK(dEZ;i(xnzM^ z3|Ua?d!JDv>xwk5IgiQqKT6HK%s*-A`@J)Lx)U8v6Qk%28w+%eGqmf?M_$Wn>@-@+ zLuiE|GyVW7lx?qMh2z*Yg=0jt%~?F;ZTKzO9NE==;`NzJL3dV<(O-nA2jm}v?0_iG zG*arF{7`D|+YEna-qut!xUL%o)#x=J1jION$fl-C@TDEFY02|dT{n_0H_pdTE$uyn%MAT z9bY9+G;-ihV9ZD@0BhB9gY5b}8iM@`jpc;M63M8R}e(&{_Us@O+V*iZh;ehs=QV~?U9k8iu@Oi5F^ zf*zkC4s*_Vc<>-`Ec9NLhN~^V@D~e%adB5~i8{;{-`6H%8x1cpOjRVuA#{Od)T1WxaJK;#(owymj3|wo_dFukPM2;-E`)9tJ97-H&CCj56 zA9+pql0i?lPhY}TfYHhPmTHS-%s)JgoZlAf%|R+|wU?@$K%P@i1D?~5=(kTVMido_E(fBDI_sv? z7N&TIQ+ZP^bRn zx~+@^*`mf!>FV8A0+stkV#6rk-WP2rGvymTr-g7!j}~E-oe6ovvE^|$ z3?NH?VDv6$?c< z13l?H_Rn!3Pni(%G^y#0+r+^~md7$t2kqN-Uy*b4V^ugn2`_M)06$Pk6k3V`m51$A zU=Eqj6Yr;lnZ0g9H%6M}&$($-OuApk2mY)2v3;r>p6o=g0-92OakCAb)1YqlYHgb= z0k9-fGp_8XXu^bF>xZ>Vi{V*$_41OJlU(4m7?PUXNX3tzafQ<&E-N~#$V(Pb9fTmK6YxM zaZhUZDD+&+NoTqKK{_Q;J|Vc`LbmZkmWp1DV#y8rLyhBoR^4lh?P)i1vmK<>gq5D} zKCcC~c1QfCpz}%*Ccbl}mDY@a!xYm-11DXzqOdS1k-~tsFX}g|}#!NgkQ!BoUla0DjyA(r8>W1|l*95Hg zQKxa=S_rF@r868QhX)tLO!X6zq^L!$FprFhH!R{%wykJk_=1s ztrTavCYO;ce~@Z2MdvQD2x+Z?ydcn{&gT^8t!qq?SjBj#Q0@0tEOyj8Y_5iIX~*rS z8$xu>d9MBz6lH$fG?i~WeG#bzmkHK~|ERYjBny@0$bplw#cwlvP#XVSUV>ZVRyU<} zd8sJV`sUGa(jccUE<0Q@0H;Qzt&y#FLA2y|55BkSZxo^a_r}h1V z*;NLsIhwUR^fCp{j!H?EqBROX|F^=iOEUWn9>2XXHYodC?$3e6I4mT>&ew;cOnJS| z&&9>e>F$nGuIyx!wm%mKlk##~Jf62B`|^;K!v&w&192CMMtVu37M+P3=Y;R%5wM^Y zFl$3gH~*31eicjNK!_}}4XhFO`NCNCViaVg#8ieSFT#Gej|jYSQI2t#PR;-Nh8?-A z@qPY06x{T_OV2D};?$;rCy}^q6NJ9nMt<|)InMT;-hfJbD{Q8@3u=}^?Uw`Ei9!1A z9DwHefHqIg%%L|)+eHs__US)3+Iyl_9kdG+J4;D8+j-M5LrA0p+V9hH>1%bz2y-xb z9t&ID{ynyE{h9ofjgU-`JfUAsL9j!(?euT6Ayug0eT1o6vpwk$vDqDrDS#^x0^Ra( z+WqJ+3yH&;+VH5aAcC&%(y?uF0QnMlyu9Z^WKfUwsdK=$3x?$^^OaQsd5q%{WDlIq`~R%Q3?tUByy( zWC3wCtuWXy$_Mw8B};xb493X9#lloOVZGDWx5u{m`IvHv3hKyudwv$&gJ3uwCYhh#MZfeS+)0bb`dolzybs=LTgt%^+NKfE)zb{o= z9_ID-aG=jzch+c3Oh@}?@R4Mx%2o@(xM}W0$xK%rSHrG07lp*DVY zb^9Ijv2Ix)VIS|k2v;H97aCYFPWW*C17esGOKTxGb2{LNk!f4wRL%j}myk(E_yJTu zpIc0=P^=@#bF3@;@bYGTnEN+a`D{OozN%7x%k*7>X3J3UwVvS_j{5ON+dUE7#;^-? ztVCkYgXZ%f)h!1Nq2++A2fU=25*+p~Ie-DS$^hF159mA+4bQ<3 zv5;)78zeH#s(iNmNMB?&%m5q%r<7hS@nAzVswl)CTq>3mQ8N0Afsr>=3Y;#Ocp+Mc zkqYB$QuG!OOUvGRAg_?x31pi)Pb*$mRfqeQQ@u~xDJ6?K?W~Eb_Lby;+*TY6zHXoie=lZ z$Du!;ZDiIAU8!vVcBUO1ehEWR!ZOPQ?MesA^kyV;@~c*$8njVNz>7e(M$qo-Ohd`< zDH)-(Ndv^%JqsLWDh3O5VSXiENIm`g)b)XpJwH3Yl2$2Ak5=!e4fkH5L)mrnPDcO+JoCX_bopk>qL`Vd#?d^c0_ zLx)9p*+z}&9fHzh&X95c$;}ukEbS3DTYb>@3M2Y< z+iy2;o^M@8&ZoQAOCK@QJ`Ep9L3BoepeF5!!7}7VZ(S04i_xtrhY!kJ9 zBMq|!VSTt3byeVfn;LX;YhLXG8+Q-kJ(XAMRmQLUH4odXOSgAn-MeET3n+a z#5Br3B*3SuHR^g$y)58@746o{m0JmJ+`(?2M%q zgErkN-9y{`2cvCBhJb;&;vwC>e9n$Z})o(9_1TZ`j3$W&Q%W?2&*ql4oDy6evc zVFNoONJFt;^>z$OEp&f9(ZlX%^!514HTp4rU(>hCdq1-C(Dk}J`k{T#)IaUW{l};K zRI^KuX^ngtu3&nYBjvBss3SwHui|BG_5#y0GhFet*q^b6e)DQ;j_o0+R- z`Ww+#4ubNLDOmP?sgSu0>P_JH>8^Y0$b~96GE_oY)>=zKY~tCh>}4t*5;mnP(;3?EX|mZlbLd zE8+L)7}@0D-G65i!c|p_*{dQq4O7cC1J24uFq>&!Ow2U60epDwBpd~cVip4=A;q+O z?^r)X*Qp`BsbryG9|ZP(f5hGCD_#TKu^k^?+J-7L;1+g6bc1R|6yK1;IKl1m@?ZP_ zIGggs1ZP#9Dt;Z#{w*bY^IKad{0eJE!Te7WvVVV4#?U(^so!87eK!hmOU6iQ$ty!> z92sjtW#}+SXB@WEuMtwwlB}bCO%TPzM6LZ7yC9@@lR8DUT1`T7S`dn0|3E5x14%%H zdb?fw$QN~`|6g^Wz2A3O5@giPj^XA(3=Ay|Y-#3YVVDyn8bv!B1ALC19;h;Y-#h6g z!)ONtOSTz-RsT%Tcz4^UD|j@n@)Ybxw3KxV(HXG)S{3r$EwlgZrBu*xTn!7gv6Qp--@)c5`Q2nTh&^&X`D+($tZ61kKadBV(aa5Rol<|J&a3m zix`#RIvL?;R=elx<7igh-sdq)I~U@bgU1Z>eh9t6<{F)Wt_G#5zv<_aH<6UQy-~cc zGP*qPA9aetC?#xv_9mnMvQUzeD0fYB?7{qg`W8OQp>crANhg=Y!r5;g8*K%|wCyJu zGrcx4!-v@zLrACWeX}G3ZOtU;DGwLNpja@*>Z6ExzS8|qCX?^ewX5{EsNq$xiJuG+ zLuDe81O=$Y-B~uPAa&vnOC<_TI&4e+oO65 zPm$WS9BnEjgFVj+HSuc#Ar^_{_m=4H2-(B~O3LKshS;gCEl$|^h^TPy&uqGrIaPs1 zQbC+=RjH1TrReZ?AZ&gNtI(> z53+QFzdqLw=UpNUo-6Yh*g1xs{5l?}gkApcnr=eR^<14m2K2s43X`bVo^Q%R!%7fM zocL+@I4~UYyQ|HJN98!Q5|-Kg_Qv{t1Y^qOguM0~BPMSF~AGB$cawt0&4T({LgEGNWDR_xZfJ3unA|Q=evlW9$3u4pYwR3EaDjKEL6m(CTw4)2 z>nc(CRnnt(;|=gPTh#X-mx-}#ne)2ua#KNkz!kqVW6pc>0S=`k=5{-XiF#7uZ%tQu zZ&X3XGOZJ8t?{J|cx9j*RzudQC)>Fl`(VAL!g{Vhz@zIBbzaY^p^`aD*VrX79ddY1 zfycD1@6;#PT!A8SQ`jCgJu4~y3wVRu&Ya?NvFV1GlT;FL!&&DXO@4vDDmtrCz=PBK zi!#q9u?w{O@h2Pw_Ws#6GX^Ut67LefcF2$7cHN)|ICUZJ|9wx0Wg(amfZ5|*ft*HZ7(NM8F`8>VN7d0#+^;M}$W9>FI2#w8xN;{XkKmL^8(u_>@Z!(UcDJi8KgwtC7 zn}}os$J}nYo~^4+eopG9&+hPr=FUi(N@na$#i!M6`@;X|rbl_{hbxkP8OqEVQfEw> zlV+}ilE#9bVwxb5xm`u(tTToa4T7RMrey*)PbeAyuU09CcDf>LNYy zs9r{ntjbj#>{jfK$DCjpN~4^jOr19I*tkOzi+%#mrNBDz2ry%hm%u~EIo(%lzn~)} zUU0PwIJ@%LNjUG>I*hG@N^%z<0abQFSIJs{W&52>ceQ#KAulytq!E*6z5EekHurPk zmKaboN5|RmH954?g!sS=nk5}_%JjjWlohBZBF`Rvnr<)r700K-(}|_(hfbV z%YVHb9k=-U9W7BEvk>2XsuOMVe3lRic)mWRvw7U@znk0V{_?pGdDfO$Zhb6?v+ue; zo1_wKe_e_)S-BV7e-LAtY1T59O0IC)5fI<`E9E7-j zMW$U2UF`a`5JzTV7sX4^Ogf;q&xsLk-;8lv$Jl(ACku9Z>Ku8nwD60D5yVCpVbfdf zpJ(fWS-)NeGKs!>=$JlkPWy;*1@8vfIYdm; zTi%CAEqY!8%!=p~v~BY=V=VGv^hlDa_W9&R8VC>9OCVvJwjZyugQ1HEwbAq)fScL~ zM)B$b8KOLav082~PsgXK{{z!NEWgFs*~QtTi-$g3JbZNV@N->6xq0xp#X}!29{yr+ z_9OSzXBH2?xNqOhqbGlO`NZvmPyYSjZ@<6t#_f~mx|B#{djEd+jnDpg=r=cia~VJL z@Y$Qs|9+E8yxPH3sJqHQw|hqzJgsag+tXaRPemdL#Z&g?p82G`A|>M09>yhk`4 zYxdnAe)CBaH&-F3kUL)2rEIRG+BO1JAm~a*T|K zBGt(w1s0+e=m0Y_l%PhES&@<%^Q2_@RwWcoipHv`WB9G8-?y(46B=O5AQs$aOi~-8 zYcr2AbBZ6ZY?2Ye+`pd~l$xLk?iQ2FoyuX3x=4>>8mJ?V!np~X-z58|4f8npDgh3x ztlu5caJ>=j#t{Xk@-Om+*MqKAvX?ZU*-HroVyl&hTUmS0a+B(r(B6ke=uv(c;|`k8bXu4hJO>8DR4Z5YH=^AKb&^P6CU%xV;%V46EW$fI)t@4-dzvNj0^XHSJ-c| z?19*K>oLHBbJb;w3+kxIUayl>E2_u&M34MRvP8 zTyKMcT55Lm^zG4BgKe=#j;WM)E zDGcJ=joBMVJK+i*KHIr)3MT8rrbM1q1-tLIRoRLvYtl(3bY?WKNefUVDkZmISNLJ4 z(9{kd3B+nh)W8_t{}bN}3OGiP89cwy)F85oZm7`x-?xGeU!b7ZT9})yH>{CVMLmf6 z)(^CZ6w#{9?faiq23d+4b}n4=l*aIgCQHd=wH?)N9J`Lju}99`dG%^wiZ-rYYyxY$ z)$q14&B7y^I7Y*E+cevqCI_n^Pg~2K)SXJropeFaJN3B}R%fp;#&sG(6&t2mZqR9F z3J+ldgGeW^raPv#)niX(P&xbLv%@7s6t+_@iK zKKsL4@7((6y&q>m;4L1yi0K=%rxs@qE*?IHNgRjIEY2QnsVL*}sQyqMRbWCb9{O_e z(4Q6${d)26ON)n}T|9Ke;~uIIP4Q1O%y??I##@o$tfzQJbb^4J45@Ojla-QKW~bC> z+d<3hO7E2ZSbC%M+tMG)<#v}!XRjaj7TbMUy0mI6rSvY`as~cFFH z!{;$iXRUZeRK92~^M%>l zN1wj+@>jP$x!ih0xb^9&TTdRndFf(c5{$43wvIFtnE3>#RMn0l-);Jag*DC0X<-Or zo9yd1|NL}|&HMl}7#ZdHyf7@Xjp`{UA6yi5o3>-7M2dL~LnT@9(k1;8Ly6RZYHn$N z(8IgrL?qT)+f(U}h^>T9pX)Ec?0s4KLQBQGZ1@=C3`L?Q@Q^{72}EMs#konyo?GGvhPJhk)in9OK7 zBW*Wu7UCy4Hjc+6L*-!@G@T=3IR>K+PN*DwKk1vhGVO8c28T`qN~s|8+ab=CJ`G)< zpEW0Rk!n_KMu|zWnlNWOTRt>>6S5j1gVfA0(+a_Fjmne59m5jQd1cszr3cK7vKAOZxpd=^ak`i0jjuw?+M{E0jE1A+d%@h&q zxp@_`Nv!!()%D8&qWSjLM0rIiT{{3B~+euGg5h_Vh z3&iT*pL+4`rOxuD`^raV(j3h%v!r7HIUD3C3&VghDN1khmmbQR^oWCk5<2p7 zx87E9>us&I3U0@8)XthhOkxB)Juy>tzJEXA8TeE_&f)W@MQmo-+b4k~VZXET-W-C_ zF?hHc_jju?WOoW?Xg6bI2=gP&;GR0m9MQaVub)1kteHN5n`*L}6Mg&abl$)! z=X;~B{KCm|GDK|)Sgja5s59zJbB0s3K88~xQnaQe-d5B=yq20xd0K?*KC*5bO;w~w zjZWy>MuRP>8|DB^E@5)Q2qG(&1zF2&QQej$KyXjPTr<*3^)we`2yvnfO18;KCDAD- z)ktvNVGl1*%hQ3-8{P4(B&wKO-wscE!qXX4$t)3GcV92xpeh%-&7eO`eNarHia}f* zed*+(W>d1U+Bw`*B-^NIbE{DSY!8g0Z(=5-NDb<4E`Y0d{^j{M=TFSPOy*yiKQ;eb zIDCEnZ7lqs|05iI09SuUI?4QtaPZRnTLkWS6|TQT=HG(b-+`;|;u{GZ{?w?Y9;^U!az7h47No#vp0 zHe1IIkr(iG=`7}@mp;eThrs0M5feO3cX@~ruU`=vU!x>4iG5C^HK2sPjem^nvT0!u zwaZPf@04S$M+gk(QX(>GZcNObpCT~79L!NNI-Q<&%p%cgN|cyr>o&4XCw**E@Jtukfk96(?uI)1 z4VEtkCk)l;B=zhjmCeNeUXrZSs&w3Mdd!WZ-#))^Zs9om{&ZDusfIAjk{p*MC6b(3 zaaIjaxboJ86APyoKEy1lg%dZPCJTRD_;}&mx6j{riY%O3xb*Gw_}$+=PyTumUMzn~ zS08&E=r&XS1D}qFO$#5v?3==B z^zcV-9D(WhWZ^hjqR08b%W1DL^$h8C3MdFYxY?b_x>7m0>avm}SjY(*?K;#~cq# zmGY-lNTMYxB@phhPqM_NsNBX>_M}%hnI?kJAw`l>JRaKYbzQT)gZ;g`_U_!Z6I26c znxIc2p?;<$Br--*I>S?m^a-+}IY9rO9XtPb|L9DX8O)@*>(aI06EN+<_5H$@$;XMo z7>r0FovTxxY?EbGbK?%1n@d-&fnB&8kr8^JNXc&cc%GvY8+>Q$$xh|C?J3XbRO+W) zC~K|*S(elq-C}EVriC-nyu>FYiXtUzI@7zO!h7xw!6-HV>iqBF%`ht-O|YVED8AJVS(Rcn zA>v?(@W?zc9b^1@M{VYXy$G5KLCv<%j{O50b1!58n`z5n3yPvEI0Lda=Qh4YDzq=2A5=&52;Wk9;=8l z_t+<8l9C#0S0Ge2m^|esYtZfzJmLWq!Nz!6^_CwZ!}jpZv{d$2v0HG!@|&NC*nF>Q73;v_SJ$D;Z<>_}v&XnKqME6SJVTlJmat+*Y8)n_5kPEX8Bjrw>fVG7{$~tde<0Pb(DycQh z6TH%5^*Re{`uPo15l-$>IGo2D5-Iy?UtnK48@WiqnbfB!;$^{e6T zf&M++L;X)YL1x^3yT%%~FW>SXHqFzybE5t|Ix8|GLnTT2&&wNRT>C;?4!&l#sgbDq zAbE^iIV1S884(!lEBD@5-e*DD-}{K`)*RgJCYz1{!#AJe*F%B2mq|Fyfw-;X4yYO-M=&IDC*m zSr1R?Tx7Ak^C|!}n~OdxnRP@Un1>)^+~C1+6?Wyod?{jJY@H4>HCd zH83(WFoNbep7cI4)U{(|W(U_Xn-GWZR!Q#LhgZ6KDYHO0b&AJ%hIKv;c712GuX8f9 zS1K6U{rlY-$~A8t-LZ6M<-2|c#wV!UmghTBn5Rt+YaYD>Ka*WV726G|3X`MBsuQQ6 z3kur$0~T>p9j(ppjIC;>ICpY-d4dA3pFyVQiv&@Z*QW_-5h+&F*VT8Pd3*UZEkAbQ z{K6;SK9BkH$NzeA;rzGHyD9QbT12vxjD{{&fSusRlQ#~#y_=sVHx4bFyYcjmr~Z28 z##2}yfa(9o%he1wo@(|k;e$%$;s}BT0PxHw7ChLSH7|ATG;zA@z=W&3Z#m^CQNamu zkNs^P8rVTK&^(r3IO{g|H;wyy*)%@@MuM2Inz?zXQdAC|`Rm3Crl@Hk!BFpaj5N9n zOebygN#=$j$}b_V>xP3fZIJW#ZbYPGMRrv=NTMhaNztN`90Q}+ZA8`ptfpk7qDQ*N zK4n3yPVi_8dV?#G*DM&~md$j%iAf_`92Q75iv`NCO6TT2Gg78wARIE-(I6Pf&s)sa zT28hyNe{(xhl%{$F-&gd1{s^CYge3|}hOvb@nt%?uX>1{zjK*Lj0DBnDRT*4q}wXqrw>vv3ed4Vw9+ zB!^dJTXyDqsfDRIAwe!1Gxc&rO4P)Q?*FOpag2uPLE+{dQ~s-kjqlcQa8%IePTF%P zTg)&RHeo(UK{zJSAHZN)i1wmGTxVIwlejD=z3`z)p8%pr1XYUF_(8X}32ji|Xq`9B zf|jvmpOR=3rBXEp-#7^~o(>CAxc;(a8H+4q;gdk_Bcgay4-p>BXih1NjK!6aSm>Iu zmfg5}@v&K+HP=*7B2qjST4CKfIZ-CfuLs3_{RvSr)>f)Dh)olIOU8cs^l5_eu%A9V z^3(gTb^Z4%Z~fPW)8xNjKJwFhZ;?B%T=}mHACdn)`q7=Eul@btZ}#uM^Y)Q#KfQnD z!OA;!^*lm;`rxDg`uyV3ef#%syYtqS2i^C-_VIsxaj>PRd*|&>@4RvH&e@LwQ}oU& zNA7%Z^v>arU`oAFtojxI^4c_2fu=Q3l>y-S`#8vP5PeuP(Srirm^)5a>A5vL)=@=^ zj$vWi4(tLQRTODJ=5AntSjMc(P>U2`H%jvY1vPG*!9|NOVv0QK5Qub4JWm~rbO|ue zH3dv`_>n!Z0I~SPh>4_-e#LbWWWc zWlT|Tjl;VWB7?~OCdFP;d}4y<-8;#Fxnugos*OK8`fD@h`z>1-o21!Yd0a$R$Le(L z>4?TVR??_rkeUaaJjI7!qEs>>$?&77E_YRkrFgPKd?ts;Sc|)+kb= znU+&jmg14nITvm#nUs^q@}Nfx88u|L?T#oC)5Vq4)3pkpP5;ZVf!BGa} zrLIf^J7%&%WXL$>b`J4BGu&RQ$zw7hDbZM2&xyomR-VH9?{T8wMv2Zr__)ACUZr?l@X6Ov6xlH zV%B6bbfc4>6G~FBH~rjX<1NNGqda44EZKoQc~ifTX}<#JHp19$&55}@k!E~zu(zP ziflxyv%nC=GS=J4;ghWHAKlCwJvY-io&x#QEE`tC7SHRch96~lK90p{tMt|B7#^$* z9DzZ|3k7bt$u-UQV2|eL1S>vZk;giBGn#c6krg`!L&F+9(u<#)^KaoFY~%bMwxYN-rw%)c~$!sGqX>VuhSYk-+4SXQOTHL#Ka zb~6vQPmY@zA_@qT;+7y93_D?_TcTU!n$^3#IclP~7@{@^ld0fBbCBEV92w+kS4MTG z(X=9?6h77y<4~iOK7a19AD&aNJ}XkABewID)|IF9-)|lz+m=Ru=g7M-ss{tpym5I; z3U&|F!uvCTOM7v}2}I*vMd#c(QloPs_Qa-8HitNHOSw6wQnsjqQ6dTIpByRwM7)54 ziTeKX#2Wr`1*^RjwL1R0s_xP;20U)qqN9o+0|-SyESqBEp6+ax!_1?aGA_(KorZDn zx>Tx7DccDWs<;7U3Z73UU_jg;143l$xh^`!x;s`9sbdNHkyGrmS(&Ki>@h^Gm7-G8 z6S4R7J@(o57%3K);>jB1tktxVjzP-DX-Ab}vZRR2zfR!SEAwy8|Di5@cGuF!GAY{> z34id|sCK+whctsVq*UlC0y_nD&cC1Kjid#ml!p zIQ0FcOFukwve{5U!R6TxpB}mO*6ht=$G^Y&)~%ywZy$g0_A{T{Jo@VG_f7UK%))P+0|CeO`%tRPI- z#F#-wc(&%{I-fP}7#W&@Yw#91nUvuzo;NrKpF6)A?t4dXt-{hPFr#^-T=3*y-1;}>~mE`rFf0uw`zuN z9T_}C&Ac#qf$G!-rYDx;7zVd(raR;SB_^GKTM?vdwJhDuETUHqj%*ubx`n~@;AX(X zpyJVT?(W?*jrv&{Yf?*-Vm0U|reE++1mSq4^vU%nOJ}biEM2^Q==x#!{Q@a{Lg4&t z=_>rYaQ(^3Yj~{;cu~3lcfku6N>||LYUyn2(X!sq1C#MFYH^((=<%BCagVv~6HHgZ zf=Eqkr`veaZO|vuoM4WE#gIx3mcQ>d3<|nlHa$1E9y(kr6SgP7S-IHNJ6$AGoLSS` z;YvYF7t;rVM{7nwQJU~{eGeR_QQYFl8e#-73vQn{ee3sETMlCA=9%Yjoj?2I;o#c; zUSJ(ioFX|-CtObwpyFF9DKcmS?UsdJsRR!<&BB*}tWz*4pq{lmdsc-kd8Ro5O9fkc z#u9DJOb>tP)JVIH@S4RSaVa5(c2#Vhbj&=14>hc?Z6(*AjR~YmO3CEbYr@k@)56M! zu*2`bqK5=;Q(1$nA99cZ{=Hj6b77rfY zAjWBw+j4RCH;c1Z7H5xED0lZ#SJh|(ld?~k6kCxMbZ;X|IUe`GoVq7I>8VMHHkwqG zR&0o7CBrIM(3w=K6BD)|K|FdfbJC(f3xy`>#ZgeU!sBFbA-$NUb?Ss?oJJ_dsw^dI zQc1Qp_wW=mZJGg#B_!zix*14~wuw@+$JjZ?Y}Ov5!loaqj+pp1Vn%t^?K`t{rCe~1 zshwRz2Q0YPv1XPJ_U|X1%prsN4ddPXyQsd>G3f8PWn;fQdAC&CIGhR9y*DYf6ScZX@>d5d7NZovMKBVMd&zo z&NNvkCOWvK0Kzd_uJCMdXq0q2CaBG18)(My-~s1fOqjIbQMN1L0N>F7H32%rdXmCz zQ9uHXutLtHt~AtvD6K{cVOp$nTn}DD+dGxdaWH)BoHy#P)erk?Yf(ar)M@%@_)YHq zxz3D<#&N(s35M-(^bD+h3kPA2J6SzA2kfB+^@bunGq3nIHQh45Y945()vqJ1{>|Us zdG77ywE8=5Uu}R%i%ao# z!vjxZun8Y)iYaTNiK)R+w|cLQ-L%4iP|VZt>Due!5+H{xC2Q&$eCy3Fh)M5YfA6ln zJ9q6wfHV{A0D@HdnUavm7)|L6PbJbPaL2i8^9}Ux*|GC~_m9qGnZZmjlHev69fcYh z;#NyL1;WRP!5EB40xNe_vV+wjt6Jp27Sq)c%$kUCEZ5Yruqx$n9gRQ>E0Tuk1Y%tW zFO2TVPUXO~^z2$ut-cly*VkMaQAty3Fhh$z7sfY7zIpGPi~slVH=liTlKh`5-@O0L zr{7%s<}(6U&%m|!{_nwW&a@c4{?EB@u72~?H}AuvU;O67Z;rt*@c2XWe-FW^KKtej z*`{c*WJCtDDISSLg6oQp(Jbwxd9TKy)YlW?`kJ&ut)p5|Ts8H&V@%|&iH<6JSR>aD zl`fP%FMU~BB|WiYUm@^aUb@8-_A=YRU(?49EWSE*d@kmYL0w>&+|bze$Ul9MXRJIw#v$-oSg5+FE{Uvh6W$!pi3?xOLRW^BLV^*oFv^M9al<)zT|SnD7_$ z0jdjw!2b`kDQ@Kmd)ziz9<23MaB4OeS;ki2Md4XjpOdGAnwT_AZJVYif=e^qu?5Vv zw@*o>(9|z&u$E4(cGw1~rTgKTFK!-v|A#YAE3H6@DsK>j?VaR!HWzsIM;R~D0=(by zCah`|sipaBs}#isf=aXv-dgdQTj`v^cm_imP3)``#V&wKjfEMAH(ow*>*!}U&wX_3 z;PF=A4or8d(8EoB>ZUZ2Q02f>sJw4^Lyy(n&|_;7AXX=|lv3Q2U0?W9#KRbK^+f zM_f{q)#RW7{5?0GyfM3Q{KiuYr*1q=ZXCMtEP=Bl3t!;x8&8%`-d{KmzmLPM&#nfE z`>!>@hMgw;<)(-)EPR2(x^c)I%Z*t$#IZbe2C8*U#A>?1>ABn8-V$*|C|lBSOpsA=O9{CfxK9vR!& zySu*?HnBS1*R_WicvV$nVJxh6Ew`9fzjv@AakVbWbQ)0U%dAHeW< z##<+sggwr6!80su7`zcmxSFV=gs1(&9E0TS`}3OzFutdAV*(7N*3uY;zCLq+q(Emd z$)8`CdoSfe#s?$C$S|9rU%&2Q84t*Y@1UNVl#(@M2!u5I-u2&Q~A$r>j#}*O_t*R(^Biv>9x}5NSh`6-?P_WAlHwSE|xA|e;x++W$8S=gje8&7SIN2 z$c6c+S4rs&7~dx@bn&N{oEkhJ2zyu`%T}mW*@t{@Lz*fjZT+}>lt&Mo!-%Lp>X?Hox_BZ~+DGz~Ab7QuR| z9z7t5{&(kqV>(`^Lj5Whd5t0#6^qRhiO|jcfIjmt&!3!seg2*KSKPmEllkAxzXU%m z*utg!+`(nnG#xj&w>fhqwMX%p+oNK~w?wo$h7g2N`X6_@Vd}1`+!C2qRS9q`Lv~Ow zS%Q~z?aDDDQ_OPH6JXOS!bD@!dU^*wc#plikr`Fwh@oe`&@i&1~nv{&RLFy_o zHyfrQE7pUtwSI$N#kOp*nzn3Pnk;#thjd;yh$0Boyq-3b-IfdVgMeXJkPCX)ZVG?x zoC&HKd;kc&fA82y_QHo#GQe!&HVt&OvtJ13-#bW-TEt{Dox`@yY>E^Zg$+s=1SyTE zR%k!imQ*iZH_%fPuY1(sFlw?3-kcHnO%b^pg{_x~LQWVw4bmMw^jVV@a=}rCVu}=% zB}G;xWj*cDRktuGUafeyp>Eazb@R?gKYef|FhLs!Z=;qgi>h4?0c-N^y!*<3eGW8u z?yWl?Jm0R!`v(~N&C@XZV`G#^IUz-B3MvA5 zCzf`wN`&GeViyadwLB4HIYvfvG$W?gw|Nax(=3u+Y7|`~ZarFyA_E_WZk5xr;nf}| zJ>e8M(?b^OHzLBSVxOXjR$~SSV2hqrHkR&MJdAD!Gqd$g0*SEfW)IVq{IyLcLv%O1Xv12Bdv#p2bp@txM_!Vz`%kIwx4so(D3zi;2o zPu~0S>EHbD);qU8xx9b>;-QO+hdy4MJ+(M{aPjc5#o3P*51%PF5on!^?d2i;0frQq zREvkcTs-ur#Y4YdJp9t);b#{Q9r3vJf|paOSQQhgiRk$*bL~DL92XzjT;o*E6*@?^4iG_FtZ>?njz8|BmeWCIy*_fk)|m&4SE^r?EGJ3#t`|G$;fK|{R{ckuZFt^ zut=vLD{punrq5X8_T^jt!ve_-G^@GsWTPqm^2SHDrt$(@H%Z{q!}xz4$?*sV>bG&w(H!oVc2f4 zt~qvzs!5TWrs6@Ib%Q*cLwMz^DJ(bkK15Aw=QtVXdfq~B7=Gt!^^|*RJbF8g%qH{Z z)g4e^g2}Qf$-uwys04P!x`MN6td0{eP}?M7vsH4uI{Nx+NHW08JT;1&f+!;*JwYCE zDj5PM8HG!hdu z6~Wz<7(5F*ZG#evLMf4&0z#}B3e9)1O6f}Js$0Qvv2pK(yQCu2ltk}EwZtp<^^m(s`k3k_cj3mWcCMx@Y%ZGCstS~dOqv@LbLXe9BO#}{ zL;9vUn@Ud?4N3}P+Mov*`TET%PJ`DSj~1wzmzC(c6bB6y2RXFjkQTZytrnvUC~rmG zErq8RX1J5loG=*~q*D_Jv1WiFk)qZFTLxGPh|M6(i6R&%@R=Z2$Ehc8LGg6N^w|(o z!+Hw2hLzgI$#}M+IGGlB+R>@qM2ve4Yc90f;upU=*uH(csLir-=S}R;I!)wQJkkkT z`P|7#Hj!cmv8Ly);A4R9WKFaD+)2I2CT4u!3D)++-7g?HF+Dx^9?eWrvW=2KVfSLe zsouDrKKH~8(`(ILn&za}Vmb0KtZiYDZNi)sR`8v$*K$lZOJ+GTQ&ILcpRr*L3qF!+ zXybWhHbt+=C?+Q3P|L>61DIH~i)LUWF^LXlww;5iX9QGgi`$NSH?^1Vc*tPX1TNnt ziW~)A5f%V9+zZn=J~0840#`&C^CK;2*|zOAX$ql4EJme$@_vF}ndT-kmq{!ep?C7= zr!hgAh$&{e+1Rk07(R|9Ne{GCa>w#y4hwGBWcS@E4ZM2Fj^jLR8J-flXrv-V++&}M z2sY)K=Hk9V4(NpKkPOQ=E@$xG?91gIKE~W>Vq!$QTNfeV-umcddfLOpooan3d36me zdCe_x2%jvtg*IKiz>TKvxH9ax(Y6`yn(cIRg!ChpMjiR!mt@O z;#&>fzhh@@HNj5PRkWM$j2z;YMfM6)&xG>{C!?tDmwzIjOUzj&tgOz$7Ae*FDc@7;X<(vPox8khh|EFyQsRIN*u;XqFG?dVb?3AIb@@9XbU zwEliDnR~k{gsIqczD+tgA;~om+N-ii$Ea}Z9L*POHwS(Ng&sAq+cI#B1&iHHFkXda zTBGW%$AR%!)Uid1<_x0L0iN#|9UJUeNxrHHTB4(dl{a!CBE>uol1_QGWSq6KaWw3c zp(&wZbV^@dKjPv7T`9xOG}J@6K~`3)I4A=*J<+h^Mr|cXJ8^CuiF5D1^3w~i2PSOe z(&aQPVo;;iZp=V9S1*;$!dfmPlKtn0r5CS1=PvEL3`_2|ExihNe&r%!eRB8ZFaGng zi(`hJPOrc}xO)9?=`7p@uWkD6v-Bc7<|X*~s@#P2`U@UbuX(MjbPS~#qML0m{b8w- z>tz`DXXVLo-yJkDpFK_n(B0uN2V%(zbC*wi@^?&%)--PMt%2@9Bi`QhB&(m7uzU1T zGRCvav?>kRta4UOnDBy~_arLEqtUzMaF2abBOyg=+U1#krL9{LnNJDbF*$Jzu466` z_RIqjJn(;#Q2$m)Fj&$9$5P_P*_{yZC;aj@vs8K_VxFz%d&bqKZa zr$HL$VT<_;0&o|oIWD+G(hp$zk7HvV*DoCmwdChc<&j1t$DGZs`t)%tja>jinRa~w zc79OLoa*ut|QIXVjflA*^GifpDH!AWhB{Jxb z#`K@Sq!<#DoQ+dMi+gC`@%}6R-TzTlQfly?+Kxl7;}EOcZ4M%x3$v-N!>A_3YC0W{ z2k7UqaO%d98%Mu=e&LgEpTBYR##61wIe$F^Pk-v$=gEyH7ruZ&JoWAKfeEtk$-?pf zpS^eeZlgN)$N!am+_k=h(|D{I-POZ6OTOf4JFe}JQ%n|Xd$cX>Xf&hF#g2^mt=LIO z3IST6K)IAlLm@!O9dd=W>hgW@eI>uDQ{GUFlTh0K;rBdSwl&tsW=3&hi?~gQW;7aU z?`QV2pX=uXcR5obnK(Xs>E@(MFd-L+LdYkKLKHhF!X8pY->R&ijWQEs8;B)jk>L`o zoq-}#In`1aT{QWy)fS6OvLr{rv5!r-Uhm&ewjf#tGl`8@@$Sp;1eQsIn!!v3fo~C8 z0enFZwMBi0rE>TUV2y3`O|CWzrt6qd@WzZql@+KyP~;;cE@w^&Me2HUaDG+! zC#0lOi%VnqHAxySYTx;AtD4gc(|=uiy-)mPOWOk*xnJGjqZB8JROTq~W3@IV?a@>O zOkAZl;_Bo#pQjDqDp3R2{XPa7L{#M>nK?8wHFFsLzCv)~*6-m`>ezgl0EoBSnWg5q znzPv=>`#-K5xs_K=f=?IOLG%`e*=&v$n2eQ^+H9e7G1=*MHdN65f7zN5D%yyM%R2+ z9M;H!6Rshh_y7Y%0z)JNh=w($lV~UuSQvZy|G!;^mE$bw%AGX869E#5+zonX)AnR` zArDhAw~icBF!gaMp={#WipmJXvJH(A@Ht08*Rlo-+_iH%!IzN3wdp6{+XeVGDcCSZ zXwHN&jf_w|OR?ZncT7^?Hu~KWU(KyU-%h}J=+4YK9;nM~uLaj97%Cd_!I`9vZo zlghV;UP^|^-mYzF4C5AxFz*)WW;E+G0FzZR3iN=hj5QKhsuAhRfluiqq*!Q^S?$Nd zq44^05t4Wj_Drv$sEf_Rg`ZwH%`84)oNG2eA0uEY$d}=r9EJ~i-S-+>$z>naaceNcZ>bBB5fs$ zI_6rrGU00SIPA0h6Qq%pq8^_!7<`oli=fnm>c_@_jFu-`kl~&eiwKu(FA%FCJ5tUV zYXEM?DKaxlhlfk>dFE15(PUjM6-}-t9YvO6;aW8nq}Z^X4{N-_C&5L&OwE~C2Wn*D z@aJpGm2rQ>wtf24(s(Qbt5kxAYq4LH0oKtlXNkof&49 zkwIh3)0)($(Gu1n50{8eSLsaZ`IX1e>Y?yiYlA?(iR2oxvd|);goxE3;A!e5Ia@;U zmgQ1v>0>$-Nc)TZWg(d2Pz-%M&{J+_QCR_8Oj!k73?v5ozFg}r`lfHV3;f4n(qSXO2`|}KkRom>C>5J*{?mE}7IJz#2iAOWyvDSt z(@{BDo&03ARw3!LRw0%O%Tic=@L;>Qo3v0c)Et8ycl{uwVfw>T)RPzg=}&(bD4g|l zUc*svUK@^iHtn=`P}QhZTCVFepO9n^)N&A<#3gTNM8wO;!Gi~zO!I@&s&nhBDLO_5 zY;(Ny)%DU>N31*#?x&8c<80U*XV!R0uiRjoZx#uxM=@ zpxv#WhK2^IsZU=Rs~g#D**WNm1?&WG@m|1;2NumBc$khgPI`?onkQQw`8Xy`CKgqTTHS^qN`KEB~()#H1NpR>)=63KF&!u)4 zT9%nom!@SYH8X9yHJH`lF^}bUxvOP^8PHH|P};ZC-GP{@m@ohZ$W2%@qyP`2jd4E} z=_>D2FJQwD@l)tR)WNA*=J}Y%oCH| z219Tds^KQ3y6AGOJcS)(r)g5Vp4-?mFj%|w4A6t{qg&H@j%s#w&q1KMWCB6vBqi($ zowz8^BK{Xp?17}->T-%AbJHMe!dlpBaV*{R(uh8odp;{5Ix%%js>A9~*AE6WE^0JA zuCnb67WPIgYr?2~Tyvipk8D-6g{|YB%}$H*S3Y z`Wv@De|?25epjDoi{C|y8>F!B`?vlufBxOur>@?*e5TIs?qY>onu4#M6LYZ0DHaZ) zBZ>bgiVM#su*nt2GRuU=b@6Q#|GZdP7{q3Uq6T*4JMwY=Ni*Yc@B!yE-? zACt}dsA=*B*2VzU=%^dkE>_iGG2a?Y4oeZwx(LA1QhBw}w~u|U(JY8z=~OCh+FZrz z;}OFoV%A{7+Xg^kk-_QfW94&ivSJ_M=UM;;8uYJlVDn&i;lfIh*hlWy2 zPwneVb#!;7d%OB4_VsmWjOxtXzn_fCQaDCh32Z|wv$X3}=cOaZ)I3=_M*Nz=g%u}A z;s0yc{BOc_1Y_g(iQ8kX3_3kyGD^0?vdhp0nhevnZ7h#OyT}-PZ%J}OQW}p${{+oz zaClU%!SL48yv8y@k5qhDJ=%-u(^PB8SMWSKt5c+WX&Ke&+V26W>oh|1g=GdU0rI{`jfeFMoLF(9v%X{r0=- zuirU!F4b>1Hj`uFHn?j3`D^$!Q)lJ<`{yPvyU)y@cx8i&-aEGHgRsiDgD+82&GAuY z-d%Jkj%ABc56pz`p8BDo9b6k>smC0P;<*uOSq_@`dS;=YUjJFzv5nT63ntQ!VL4hS zXJmkJ(6?ibnPEg{4xxG$et6bSktd(rsgHm-4w5~lO(xp=yLJwCk%j*~LX^0a2t^W+ z1exfbKGM5==`DCK5kns`8IO{Q-gHm(9e4`{qJmWd%AK_R_ctd${PxhFH@baQQ@1Y$ z5Zs*7cp>YQ+&Vp>xU_o0XSYfm>n1~;@@vP4k1TJ=C-qhdYqHIv3D(b}6uYM+`TN zXu7wUL5S7tVo~SF8WYSS&6tmm44U@5d`iDIBSde);auyYuPU`5V8VKXYYh=+3(*=AZrj_wW4n_S^68 z-+%k{BlBN={QaS~=a0Yh{U1)=df~V8zkP1QyES$4E8UzLdNR@0G0+Ld#r&7A1{T%A z9bEC2P_-RHy!Iz4EP9OBF0sZk{oN1oLXoqqL5w29{KTjMd!Ys&*lvQ3WMsR;OjFbW z^qJ$VQ2Oc(t)8yj$W;K!U#^mVmd$b;wn6ezULxH%ILlIV)7R~Brjy58+70;k@s>IP zoVBS8aKS_Sma`%yJpd0uKK-*OwBylI?WJ`ZfD^2j=R_1;+fPA#&v{1BBS6T zKL&Cu=9u2r!%U|@w)7j!1WzZ^NF9Ei+ulWHI*xeKN&i$qUc>fB=p>(f64jBIFp@Fp zh&^E$j;XRo7NRZt_&7B)6R?l6N1l9gz!|peRL9=_{;uA^o;|&TJJ2kB>M2t5)~~wK z{ZBpRI^XMP6aV$dBN@;KhD~rJ@C)o5$8A_=TM0Z*V_Z-bR60BG`v}qi6`FugvW|N1 zG;~#P3u9x{wvIGx$AZwT%>q>cu0Y$;1)_C|Y+K}N=+K0zbFkv}PTjR?tW0zsz$M5f^+ox{g!`ZhQU8R1s9%2J7s$+p z&75#+_=Za}hh`?rm&=z)p9PkQ(dUe>U4Qib!F)a==%)my7V>LtB`~4LyLEM;IO5{W zAyR%xfI?po#8fBX&GY5UGsglyKl{L4X~lS*SpdL5Kfi$GxTgqDMAjg}%hzX)%^WUY1#9VL z80F=esR|u2DgUm*p#uZFSGJPuD%6(!S87KSQZm}`y8mevTY;OjmF(0s_||nqn@*%c zvLuHZMq3>k+QGD9s)O3Cv{f_giKSOqI)lucAg&S9D1gis8|hMaPQCoywUe-3ZNq!| z)+@(uy>YqeB3rkf{mtzguiiR+Z2s)a-(7ij{*8-)W%^4@_>212LDR8wcj1{Nus3MklJRD(1(ZU{Z6*f6! zaG!aBls`u&8_vVNFmnPabK*+PP0TajAS+Hs75h(++dqY_300g5q@BA*WUiS$h5&9a0w%|l@K(5 z5N1!$p1XN`_QMKK@lp8yk(-BSPl?0XE3@ZjKfL+O;^BqaD{$!V%@vFGJ}*)YBc{X{ zwqaa%5k6fT%R>oCp=iV9x7Fl!r3`6bd)wdy!ZUBN9wsB$&4eC+#z(LeNR%e&C`ts6 z#gnbBlUvD{VQLvN#z7n51>z8owh}Cx!2yW#aL_6a*7>nmtEIfl$mx0s_76QD+y(5U zqdcR6;0#v{T5woccn%aP5ZAGnHSI;1v?w(Jg|VfnUSxA zkr!j~?nI~GOee@UB1JsDpl=WV)+=Xkz4zAbH$J^bvb#qODD$WPbo<6YETRHG43Gnq2n=ZkRxCE-9uOU!DLoR`mlwyg7>v9o7 zUezea*y%R8(!9ISEstYky5(3PJ@6QCcgvWf9fvS1{en+$oxnS=yIMBEB$SDTVq-Z2 zK3gc_kN3Qd^4Dlld=j>kqG1X^eq%(Gl#MZ#XL<(HSh``;VU3Yi0&_^sjE4P9a9Jhs z1uso9BD!gGwS?#Kr*VO9zp=G;4Qvn1!Wet&-03{p2xrR|M26zV%qxK<^C(ws z!)zUJ3a%WM9Cnrdg?+o4;Rr15O&YjWv@m5lC8l9JdWNo5A9zD9N;oFPVhwMYKPd;2 zXG9N?rE+6E(?#JV7?VG#1Ta=_M&gp3l%;sQx%@};u7tmJ0VeQJ#p@Cg|FfuczZ7oC zSy5JdcI{KZu6{(H?T86cPam%Zy3zeo+x}c*dVcMIhVF}qy7{ol8_=P>RHu1&Jyx~O zM)3)=oxm#hM0}H<`^R>=0y7%{pjkVm3Mwcf*Di}75tu zhiN*~abY-tC9(&+Mn`j8G(A?9!oe*;Ul)J8FYf-ig}$U=ZTlZJJZy z`h!e(D=SL|eK&pV!|b)bD$n-mRQwH?tPyp@a^+$0pyfSPPyNl%|L z4GdG03?uuvnrp;E!svjb87^bb)xGJrc@yLjmLeXGuFbVmT{~F=EC6D_T(9T8lN5!L z@!GJh@0^TZ8V`sX1|9)cFpgYF6 zg?zP$#I==RcE^p-x(LU*0A0I;&(+p#WZQ5QWEa&PYLcC_3*On=PR9?pEXG)UgE6*>y$Fv8g1Oc{BEH#phPjquYuAW*pH!*I#|@pb zykVNbEBJb00WeK<*P2&7=9o{HIf5OPGX)s zaIt`?&MZv9ucurh;MUf9P(ccg2b^M2;|S`-h(?J@HAa|C`lyAyt`6gbOuD2beJ?;ug9BLorvk zbIfPEa}vv26{Xg6{E25X^h^UP&6CWiZZeAhO(t00+y$;7+IrqpFg0Whi|hvJ%h|ZIb$n3eB zQ?sAr5N1zfv&HPG;1Yf;zibJ*x-$GWWhAi()l2>xWDRz*QKlJ1Jh2iSdV5QaC|dUF z9M$b^uqQJHX7<(}6@3?CC6VnWFIYJ`7FZUEm?S4dk^=jZoN(dMwJP-$qq4eYOn|Lk z>$(O4@a6iLkpt}rXU4di*O1}>Y`=ZXw6Xp?(q#IvhpKs|BT+fof)LDklY#l(#nqfO z!n6j^dk1i~%pAhAxa{|8m_half|11X&P3y=y;W zEx?WYvITUTCX@Q3+@b=|pPLNls}fY(?Tk@wa{`mf#~bq9#tgWRt@CWow%kiKw8B~q zt-4slVJQ}AScO)|pEATLZgk^+u*&VWRGG0DzocdfL5@h5z$VtoKn{TmchIrCMPQ&0 zfe}irJgWd7s|HnbM5m*TU=MW|X1-+SSPtpD>=C-G-6<5SpI8pXRuy1ek{bsg`e{W* zKz!>N3<)P;E!B#`H+bgshs07%ZrcVyU|3UhNif1SVi_88^}w@7ixPRfrTZa_j%2t2 zE+4o?S@4sL$6K}v5;cU=HW0mX1hg(skwlnSoWO0l3AfY@!U)ws9g|1`9S*y4Qbb!! zn$-|PLLwbGMn)ZjZzWVm@*Yl*GUk8k*+(0h0->GSwipx6#~?LzV>GzKkL=S?S;F?; zs9W8xWE!n*8%63t)C93ucQBd}jV@Wl%F|c2TsPL%SOos6y>;>5ul@dq^DhUM<-dM^oO+Y!8EXM@}-HI<3)$Vp$OvKDRSL-_f+lYFbyu%cRyML;dc#c zX0alK^$!Z9Em{}`1D*6zwF1JA`2wwx`f%{XBJeE0a}jYV-qvuken>94+5-^{NeQJk zCg(e|jWwuTxAK(aO`c^o*&>4`Z-xIPA8v+?Fs+YHpVTwdba^7xbE&khBg3~&O~JsT zO~0RMfKGPlCT^WtUsm$!vZ|MMUC~@=8L$mCkHt_eH6lacbW!PLk^e(FkbaO1z*;S? zNJ7{YQMVqnzID~g|%#+7DI3eFbt6m9n`K%<;vk|eo4Ic zr*xwcDdtH|1PQ_I733W(2WK$yHdrdw<6>y(tVb4U7N~wN?EeO;zL=TKRn7!kL6Mjg zg*_0A^{~t6Z@|!u#-vE&xZWbu*Ri9kWB0zDyLZBl1x^|^wu96DGbJH`PlGM+a43;Z zV_7BF_O9M-JO8(9U?RJaWkMJm3+{k%4uN;$6tRR%XAA};k=B)?0H9&zNHo-J;kQBX zTRTfSwd_)g^cyNHQ?iW~xYj@qKznv9$8F};ruDI+>cP#KTK?Pu*sl}yA6%olx#6fwSg39yHhC0 z;k(-kln{dAM&!H6#`GGX(6wPENEPw1TOgXIdX^5u_JG{eI&&CiG<1vF<5>pVp-pmJ zA9u2BWcqsce(~3?ndPk8ytQZgI(*)2-C-cdVqo$0P;E5CE$#~OCS3q7NuLNVD8Yyn z@qk)wRts;Kr)3~V$8<`%G(HXsLZBVrRlRWia^aP$h4B%IOEC}fZc+2U$iKQ)QaK|< zvouC+T~ykAb=91{kojtgwiV6sQjuGwCX>s~G1#ZbzQ6tWPM$o(2HY1=iwDm_%d=CRK{Lh@gW4zoQr;QaDmu zB^ji)xP#i$r?N2rOcdQd;n-k)T2!~6$S~`PjG^n(r|<=P0yj0p2YO=qohQI>&0x-6 zSlB!}eQKi{oj+1<+|5Wkn&Sq-Ia;h}v%)Zpt|1)9RSJ*gWL3oCwND|#&J}7HqTtCV z%kLtBx&RkDPs%6Duaoj?0$$+K%pt@gn0dN<89@swxRy^n)nt2E`AuY$6I7BX#VY|d z5Lj+7OP3L!00Dz8BclA<<@d^estPYCwgdYI&$it;gx@r@2D}K|J$746cViGJa=gBcD92Y78Bze(x`HEq z&5)myy;|NVItDVKgA-wna*Pz*tA^dRB0b;Dth@s|ndVeQRy~*MHHH}sh~&9>05PpC z$AM33c^Fyn#=lM@!%0F3yGkgdao5_4`(IsQDIVE`S6#(oq?p-)8&PcPT|e3rlvP}c zc$$)p^^EifS{Q_4MgCe6mIOy-N5O=pXyb$>*f*`_SjG`O z(amTg?Ck3xX*&l~#Z`;!6Pzg>Ia(;9yOiU_;APy~Oo!V&hMs3GSnXt0gC+gt>`^vR zNmd%}&c$jMt#x;f{^QD@{`&FoUq5c!Ijoaj$D$>(323#JI&fN+RqIxoat3pb_G@bD zv=rQ)>tr3&v0PrP3yt2lWa3iP)9)9gbZq*lX!LOO3>v7*>d2a49j#aGSkSr`lvGlR zdAd@3SAC+SUf1B>hE-{C$iRydDVoz44vLdZR~@+2G@8&wN?V4?X{D02lL>~4Q3C-l z8VERaJKQ2UW@^mU3I-GJLcWRnge3r>VVgtPf3@EZCj*Y%C?az#%X|zC^Q4qrCZ{^-;P45ukD7H&h?c$HirQtBBKIIQ;uT>u7G5sFK9CaCsxqt`r#c)H zdsGP<#s0oZ&bu1Zlx^tRvF(adDtejER4joog}^vD470!z_@6o@K$e zx7}uD_^u!&P*PD)nze-ykh^%4s|G&BU04UEEhyAb;4%2-Ow@&b`W@$U^}-7FOY51*Y%NcmeZM<$7Fj}t860eu!kVx z16!uw{b5ANnz;dmt6h8yhI^mTJ2IP78d zF0M7^s!kTKeXg^(6iRq(cE9UvAJas?GAy>GjOc%0i`Blq)Zo5}!F~4*!C|dZg{%9% zV@A$!Obd$`gCn->qS$DibTt#mkwjHWl>>ibl0(>)(5RRk0T#C5S5s|0u1-NMSm16n z(!%{K3R=Xp?J7<~Tc$jJRT_)rP^OziwX?NL-4^ zo)Cgx2#lejN5yrO^784=rav<|w``he+hN!p)SlSUfq;iU9~#;XN-S$&X=SPl3vEaq z+eI9c=S67;GIsx>_PRm=@m&Z!cA2KDF<<=gCe4k~0zD9133gewffBH6*&Ql!1&aL2 zSP4lHrPdw~#42>ULL?;UfRI1uZ1Z_Zf*I+C7{^tzMR5IH!7~jkJ*4*;S;menCtQ0| z!nZb56e$+1O}!3heF|oQtq2(lFB0T3BuHZqLLFrto5C_g2E}uj0 zB~rdx5!B4nqCE66lG0u-e@VVQ^vCk+f@|zDe(KvpZ#E=FifIL@Q`%u-+4rc+AWz6H zzgHailxRdsD4WT2cm7tTCcM%u{IhNeTp5eLVxCxe&aY%Z34U|(&2LVA_|5yzeRJ|t z^nSlNdFh*zUw(7)vN(0}o0DIBbMkDHDaCKz|KOYV|4eWYI20Jf)Hf$Tsf;6N-K%aK z$mg5|Hp8VBNEcas*k>qDPKA-DW)mW9*(x%0mFxOC)$uzbxpnDx`vdFKhorFgQ|CRjO3+pZZ#0VH9I zcw{~3PC~~<9F5p3V0>(hjI0u0O-dd3!7PWLLsZQfCKjgE>!glacJ|h$Uw7(=Gf!RF z*GN*5VCP{Um&&gGGTHZPi6~MuQM+0;ON<*WG5+I`e}4Yv4_A%{7UtTO7|U3or^GjT?pAJ& zQA=P2Si~@OYCG$Oh>Ma!HqTlsV&zWig+}65h6R(_&Dci04WpBzR6_{b0jeXG z#tN;sKiXKG09A$sL5LLQQjFCslJ68!Bq7BTcf(zRRT>>|!U02|)O^9aJq0l1oZwl! z9-8HO$*Gs}i`ZdpxO3BH5rIx5`( z1L1|T4J9quSZgHGD&GYhFv=rbCu`vwxEtj0C@1(qgN%YC9NSoLSp-&79mu(ug?T1h zP<1v!*|UxIaaHB+&AtBKzN#dmQryFgI-c_>iCfRVIRD${ZoP5w)+@~xI-5WG!JSu6 z-hT7#`7?hGJoWp))Goqb*W5?ydn=Ju-IXYkj@BAhLG)5TGWl(hlkr$vJl+&|tc#U7 zEuLDP)PIn6p3ife{72i)0y>mRV8K*xQ58*a#yCYm6gS0_2nf>o&|9Jch%c{EQy_YB zO`ATM%@R6Fth~vG;mZ3(9@n_ER)t^anSNiV8KzrAFCoiwjyU%8XL&faH1f8Ei482s zPG**Lmu`zyt*)3)vr9pKZY9((+0H&{>J=BIKCzChL?uM*mSrg(npk$Ue?Ms@Ka)3l z7H`%1t7QWUEWZxTVc^e@-tF#eLNxA5Cq#WFz(I~tFc%um$~`##r)ubxjCO%vWU{Sb3yy-$(TOoKQmOGu zJBWv>VO4bj1g#RCjag)j4O`s4Z=9VQr$m3il4Zpe23M-aig@i)vWk@Ov`YlTR!(;f zOnm$FZ};!-Kv3KT4%G@M=0i_DiCfFUawHYug8<=%6RJin>yZgAtOCT*d}KnUcFxd1 z+E(F_60_XLt*+~ydWtN+?{Pl|OO0`N&y!QF%gr$W4PPf`eXA@HOUaOdGCalFCtfvXy zcCs8zwIOR;j!PjAUT2UZ-!3d#oIVW-L!|<$yNOGD)8?-op5fCMRA#}$4%YO5;27%^ zX~p0@t7}vHt00MXJW(4e*tEi2Rv74DI|m}OkpF;p z6ktjrfIKd#O(lG<7k)QmIxKS*yx3v63U=Yv`-QDn5CAU{YB)vl^WuVeWmjoGH?%1= zxXgmWqf`esRZ4d1BOGnc;5IhZK+C#1lH*lpH}2cjRbrCjL38pQ5o~SW-Q8T9T{m)w z5<~|XP;-1_r1Cm1IZ|FCyG_orR)Lya556YS%iE7!cB_B0Rf$VUrPh~KsZVM2!xd(w z`-o7`Pd(G0!M4!4jl);pGv#!TO1#C74!$OO`dkulL%I8wKvGJ0fcDqU=OhjO@1cht zYB#iu;7Ebj`*PIAJ&_V{v(!91!T&=Zd4vq&foKAuIomF#+S=^u8Hwv{a1Hz~O}6(9 zR6dCt8RmE^H_Nb5Ns`vpE%n*OxjcZFat%DV4YcgDORLw_NnJVc_xyxqDHiu#yytN| zN0Hz+Oa7-G74JqiC9r30pBa3?{;y#%X_Sut+F|1^r@!j*`<{1X@vV&&ZI*#GUIzBB zBk%n1!XE+)bnVK(6oEA90k&yS7ccCMOX}FJ(5Yz}rvEtY@HQP?PPa^O~ z{PYG2ERY=x5x88KdSMs1fbY~-I3a}+wd-%wS+C=)7x~(;pETsEW6g5o0Im;|sHmf^ zC5Cci;tESqFVbF+TE4{yMy2paC&MPyGa_pF^NkO?vT8-Zk4JJ`Xl-&}A@&aJ2!&jE zEX4&-uJ}bLSFoHp9-PI}tY9&E83pu>D!yPp&U7L7+5 zE>T;KW+<z=BtEdhiSbcRwju<}53 zBOLg;qZtOSs;&L~B?|hWKrNnw&B&&Bn5V(9Ci|&XBIEE)ZeYCzGHA%|?&UPykXioxWPK6qr;+{I$mpX-$f$jx{d*&*6kJi;59WC?n{tp>4)E4N+s;B zP!(crm_wCzr5xB#MLzxYrfBLs@yDIV8nB{63pO)H*>_LFA}s%=L?*hD9ZLc#X39bu&r**fD@6Len`k?bKrN;Qdnl zM%H1&n!^rISL!IRj~YgV*EJK(-4^*jc*Xne_n>4yQLTJR_^f<_#~>j^?ur^iSh&+C zM_?6BpR{>4&v;XCJ}qC3J%(jbHBYt~y2TnGz1?M~1?I9kqOMYksLx`Gu%P}<)Rv*G z7+7fOK&B?DA@;p(a@B6B4Kqw#xbxv_D`0!<*xgbijFue+?1H;1?aucedFsbRa{Jtc z+sD87{+*Zp^7(f+-k3jr>ekuc1(u5fQ>sK_lA=W6-|$^&Ohu98NUd~yqrh$%8d6i+ z`zN;d@86FFENZG_Pfu!F{{&Xt5tHFWWH7)tG=yhTgC;M+uWE%E%at^yv(|8D>u6@5 zRMfNk_mgj*esR?+2evI-SpBA-nSwCba(yG3J*vrJ|AgryKa=tM)3#uhGEAPuKqzsF znn5$nMAj+xYYd~UU0+=TrKdhzkMi4yvMWG$&y_FFOu@g`s?SxucT?MFh8C>0^84k> z2%S9hLiOb_fp9xBIr9u$d9eayi>L6Km*6F0EngBh6sJBaUnpOLljqC9!8vy@txzJy zD04v;2M2oyV~uyxL+2Mnslz^;HEE$h3ZTr9YDPyDLu~}JP;)l2)H~M5^z|G(MM(Nr zxu&%bw(Bw2s!4{znW1V}M}`L?&WbI-)B7f*iw@}d3vZ@+wQrT+IC9Xe5Synp|9 zSD(kql5QKMBX>cq|-MzPo%H&WYdf zXu7xaXWi-E?Gv}JOy2(V(5;uR%s=zteKY3Q$n#p0t*RT$<1C1$by<3_k$oyDh2jnO zVD^K8#{hj6w0*`=S*7-w!1mB*a9yS5^eKW=pB>IHw+b76CD)jF6t*_xa8oTY2--mD zY=@y`RCIW_;litmB;=~1t{nK2R#=w8o-(7Ak=!0LGz#jOWh%97Rk`&VVUd!WShxZ2JG0JS0pX%ds>!@qy@I9QFtCXMF6;VLo>2&tkHNfwagPTv! zel&Y=_Ji4TD}f^RdzX?aiVfC%CH?)>%_FlH$?SVKj}v?iV>usK8c5O&o=!rRj3jU!-(K*wy6$#nkym{t7dRxzB9OyuoU(% z(gh*}Lh2cGV=pawt>sH7AMk*Wz~4jg&b7t!O(rn#y}u}5n>itvj;;h>OU2UYD9`6z zLG)H7T)z}AysC_%Qpls^d|TUlku?x?oN%B8@VkZFj!aL1chuLF?%46Gi9UuKErB&l zwot8@qa^Y0T9Q8A8yz@U?GRry^cD-^#b7NQYukW98>k>E4Kovd9oczTjuL_>hiXOV z+gh0YaQ0I$6Fvt+;pXJcso67VJCK`4VblF^_6mV5_~z3$kAo@k5uAn#&&*zdD~`>6 z@bw9}3O@T0Y|xW``S|7+U!TCM@IH$lI7Z-~Gx*PkaOMoWQMvuiqvY>rX3yb0zCQ7n zYvN0e-8}mB3Ao^MxbhrdpFD<_{{3nA+|A=~gJbyJaErg6BD0^vFPZ&h z_5yr2T#VNdJbeal^MSZ0{3!e{p8G)ju%q~kj!Dw`LTd#!jL@LF`9)=dP;4K;l!JvA z_`wT@LZP1Y|1Q%R_4&x*{Hl*$5)9Sa4LFcyf2Vv2*|y4G!hbK%yg9-E6K=<1yGXTJ#IGudDpK6z<^|+_C7pB_LG6&pB#Qxd z3yi&u&y8KDxN{fXPTK}@7|GQb#{P#5YG#6`DEqp}vS3ytf{B0^-*7Ms@olK7m59bi z5XW1^k_h4yw^kr*z?EnloIrk-E!M*X9F9r`W`{AxMwv-Ol4apw>w5blYqiXF85S$% zNV~~xyMYq5l7B-{x5mqD>Qlvz*R%31=GTa3s!v{zH^?LAR?%E+TL`2DOxZ#rFpJXJh zNKJ0#*0I?Eq>_}v@@6gSkE%u66=ZQiyNfct3yDaui>T-F4}~)ZXq12f2w6 zI}vS10V4F%G88+(k3p>XL1t?#I9gs^d$ECb7LZ>=ig~*3d>6H(9ZU2Z>yDnaMp#Zm z^f|-CA}0oJ&IHJ7iJu~!r)_nvpV+~)5%^dZ96L3)mdI}RMV_71BHLMx@q*4R;uJ`R zGcv$n?)4(Gg6F|I3otz{=&rzYSrLQ-R>HMN00&kIz=`;whsNyIYwye-eqsK^E8pFC zbN;2t6_R!vf@?D#4X-Ak?~na%{_5pA$?x=IwaM?a$;XXl`EEsm*_D!DsE1v^8E$Gv z-`kp;sAE6|0`{kth-uds#rF1_0d*D+jgZ?tFnptmkoMSwN~JYTTUI> z%pHDp?(pgPqwmj6Jv0BsiKb%^zIFNN?U&EqnL0oJ=?j5HkhTYCX^-Oy=>`t{fd~>( z#6tq`fIG7ms4xSYIC2K~O6#e-!XYW@iS-3hUf9@{!Fthwd`Y8uM!-%q4Qir}24Oko z3WS<1{d+xzK|jKZ(G6h`bQrqj6u|u7tJx-P?KZ~9qXtZKHn<{7L?s#i$ciL~?$$I2 zhoz{e{J^(9Ti(VuG<56D4{yEoyIXG_-oJlt^77o|r*nrtnmc@GZtCRR;ZNqK&deP? zevh`DwE`)wQFdZH-(7z_u#Dy=znGi+)7<29b5k#Yrk|TU;!0z=@>ssz*Kkydc<{A+ zudB+||KyWHLrX;z{S^!#stNuEn~ySLsT?lnBglgVx0(3yOZM-7>M0SlunnyD77eyd zW>^eRZI}x{?jjqf8V{~K7vX5rqs(;Uu~yY**uVOWNy%`-tZ18JUd_ii3@z@BpZ(vi8zFIO;zcJ+CTp1I4*A}jQNw&4aeeF~;uK!Siri}_kARu{h39E)g1Rx}Vy zzf;Ij@{88{RL{ItLhLg+HWE?|Y5TZtx9-9OAlkDKH5-chPJGp~EPiw4`!}ARJNz`-6o>!Vj7G%#*%#)|oxAhO z$Mf$@flc)7p*L?`K6UH&M{mFV{;gNe!pUzBy%ktQ(FEpOBnAG7Cfq`D#VHfMcQO!D zJXUK{1W}K&LaYi9zR8Hd#YgaeqKFc!b}&`RV-C_t!5)T){1zX>frkHsRC_7YedyRg7Sk!PB(!oH8`^6f%b_=HtZD^m72c3pR0eb=q!owNr z!UyvU4;Dp0UnR1+Tz-7HsLWjNqIr;3T#9%QdN!GvH`qe2sgYyU&VjvRku8Hc(-_Mb zV?z5nh2T#5a;|2V)#BG^i6RPd!c$<}SpWCTvGUmpYFYEyE#=n*$QpJx-YB1i5q(iU zKQmdrijCg{UR=UHaQGKN*DlRWkuBx-VX#->(n%a}aM5k=>V34UyKBIu0YdRAx@P6T zp9p1Hl0yyWgEqCmdOUBt3XE*&=~$N~FC*kb?epS$$(P@qnVfmHd<`Fdf{TTeuflpc ziztUPhe-L4@YY445^gkCw5|MR1?>R_g3Zla;590P8!*@_%f=X7XyxC+jV=m2NVwH0 zyh-_-&>x^rTpGZIBkvPNJ}N+#6%QVf-vB~8XZk499Y+4+{#0EwY>ag-=?dc5T;mof z2FKM^T>t&lE=RMn%rG0Mq3H$k}h!jsq z3X(%BF3{a@bq@{u-a(am81lN`9du8zXfoTT#NruB3NYV)`S{5vcj_ZBb9vn8>|v^D zR*`AcBoi<<`Xf(1Ik2;{t2f=#`_xlpVwmHPd}L^71hIB^8+lH8__(7R(t%=jLJ5V6 z2Vv`*7=iD#>wJ7<`f?@%=eEJW7SpT@wGCoq!KlnIb3#rgq7%u9*wU4jO*8Oa13CB# zBTx75mn7*g*Y3~iU!$5U*AxX_)gn4gwiy+vxXmwDv2W$T=fe|06tmig$8R&NULtwA zhH%FXOurxPyStJq z1h-$J@Pvmj-62JiRROiik^0r7kyRxWmy(_yyT()ZU7*f}VYkoW1;SJ5a!jMh$Z&}i z5GGC6xl?G~wEp5itI}O%=m@I5(BXt_9q=4vI4sBIa}hjV#S5-LRje9TjQI{LhC@=! z!*vvBS#p9OG!Xy5a$tYk`0P}*wcRF6Fw8L7QsL?jz6!U|KCY_V<^5G{9mTh;g9cNw z;h3SIVZntYz6_1+st;e`WZ1G7_0R#zT}yrDxB+T5n}(W)}h7;*u=m?irH8vTN}M1*!rOC!}bk*0|YB&Qy~# z@E0sMix&)itx!7}%F7@svAkI&^<#np<5VE}t`U7)W1>PH?S{=!osAK-YN2iDooy9M zvXfhEU5rUN;wqYo_%fxx=MN*aaT0==iU{!5O>c=ES3&T9?tf%FN z@UY2RZ_58JhK#@h-EGJ+nbFejVVhF_Y{va=}AzGE5x0ka18w zE)L6TO5SHiK89GH!HPpw0SHcqX=d5ZQeyR(fYwF=YgNZ5v{a4KO10y7G}2EG6djxF zC{+jdqatY;WE$18)D2WDC&E$!f$u_+8}dboC8}Y9%7IV$#H47%cf~|Y2S~%D8tFv3 zO1^-P32lEfl4Z8{!2?z0v=i7~-}?Wx-g)WcMu zT*FwsjTs@+LVzXFWtU+#z5@dg2~4b98fBVMbmcj@MG92x1lX=rG}$Nz`4IS8Fl`^Q zh~;Fn%)-3WFGwdV7^+Ea&4-u2OUQ6k?i^4rdqv|C^vWyAC!+54Cd) z;m}k|3%oQX z!gMye_{y<45!DB{JtGFBp>s4)EJKIuw-S&Ff%63J%WRvE(t4K2h7KPN<{0TUz{V2h zR=c!Q+Ah`E+qP}5-P=IZjKYI0BKKrzVi$R8l2Qcya!Hn5<+iPIvVOBlay1!86)EA_ z+njNq3-IxlC!gHr*uw^P)@LagnWL6d;AEI`QG_-J|VdMW@DF0CGe#=v1pEuVA2tbt+WiU_gw9L%=29B9Uo?S@DBrm zzulV_K_N;xMwU-(@F)*jN;0-gBqcfKg3_16F&A2cIPj-75dKuIHEM(4^7Rf5-g@iY ztykZ__2wtdpv#xb1%Cb9)LF#!MO498gGZ+wo;Pgf>Vm0OF#0TDjA7|VGIBQr6a(-z zs!h5%IE})YGGUpjO|rSxjlp$g#=t)&{WKfA_M=-3MMH5n(L}Ur?L>WAJMmOVmZHkt zNTtIuO*kMU8e?`-Q-dLOHn(yl&CM}d3hpt)VfPCmZ3ieXR|quX0?_k`pa)_JONof@ z#-ci6NmKxIZp=*0yih(jbEqltf!D;=q{}hmvJ~+Eiu&e= zeatbZuY*{qGa`647V!o!pvtWSFuz&x&D`Z-E^4i zOc#=>m&x5&4@nIkNX>5#2&v_$xmNqGR;wJ5LgCHM%+=y*Jj0y=u}THjT0l}{YaOc| zKwz+2Mr)0l5x-v8?77(wXFr(zC`D#Z&z`+GNp61e^@*FuZ$3SH=H{{4557J@ZXW&m z#Owz*pZ@j`od5d7-%r80zn{8!jLe>yJx3N#!Mkw$0e%WTar5}>C-CM#mO z6EdtZ>^>YC+6zA~)lD6f?f^4sV(HcX{SR4$QB%X<(Za{A6w$z@Lb(>GULqzNVJ7m- zT3bnuXLDMKIQlRZyek>f`rm|F7PG{2@OVqd<1G^DI#|@O>koJQF~sJl#UgwOp|)s* zL}N#ELvH+}avdvenuN}Rv(ty#ee(}s!EJa|Kk zo_SB7MdWxq^6jCQ5~1dW7pg%eHF7%nm0{)?*-|t#5!0f2aHJaWn@Ad#2s{WFq3`9Y zp;1@Yv8yl0Z(C3tIep_Em|Ooqmxi4pJqS%Z{SG-`IHq6>qP64^y%_gk%N~6tVR7|9E+f24{wHV2~MbHsc0Lj$X}?X zR$f4eT(qVh(TuUDp4#wcL>+6eo!hFx5otAeLP^Hf%ET|D+vD3t`QQ#P1;d$xi5AGK^(abc@mKdrx-up-DbQ!6ar&8&iS_3ioJK6m@$&u*W1^VXS{?_9Xi3}NzB zTUful`r`aEr{+I>@76oV+vbmc_T9Bt0*kGuWRbWc+XwGzDMcbuBvz{oH_ZgEnNZ~L z3&&~jqK37z6C$v2B^&PY`bo9wm%#hQ3N;YSh{1HOI$9&}SZh=}7{~Iou*dHW1i5N} zGv9-~ojSJ)24R98gcM+BTB&k+V-=fK838zPY#5izjPSmf+NM%41axW!Hv%j8$T2Jv zauq>@1NDRx4vn@c}=*2W2uIY61~oVxGqP7590BWnc)qTIum#*5iSJdZf9}Jj4qH zo~=M+BH44s(Dmt4%))**n_J>4tONq9z#>Ij!tO|-Jye+*GN9&|o`wAgp?R4zilvlA zYLV%WY;&;@VfvIgeG2Y3eX59UCE{1qF$A~udkn!!c4&Pp!Q~>~S)C-W;9FwVw#UV8 zgA$BX%wy{wlf#Hoc*Ty#I!497S5wYc*M}(? zBZi4>@oJ6?v+*Kl>NxxAh-Hsalz+xFhC#|UQ*+>{jO!bpEw?sWJI2{pQ>2G8YrK>X zZUX8VbDSM$hkJZ=y#(`;cNLTej#(s-ge~N$AXph9U^mr^j_Fy*8VuFKDmT7;&oA2g zhJK;M6S}gk-DW1URw`)T&;T3lG|$HB(SEn=^)#&RwQXvWK6F+Dh~Q!@P)0K^Bqd~Y zi|$TJ2`Sz<*Qp>nMG}OI$a(k$DZeda`7Jw9O*cCBvjOk)aH5u60IdrIP(yCIi zJc*8JR*(m=z(8#VTgXM4D9mbiTS)C&sWb;`0ftLe)d5qa4THx$d-ik$mhYp~rWwP^ zIquz3agkq&1E11`=OkIXrdMpIY8f>0C8JjH-u~nhIW{--`P|`;uq$cmDrsFo+3xD| ztZ2a}w;>kv+~hNJldsK99iE$d_PfiM=O#a$oBD9>@XK>kH$=15Z@#;H=+2)%sH2SK zaIG>P{OY-U>8t0eYFJSeNshpRRA6mJs)CDzs*6&*_9?cQluR}pJsBF>!L<>Vdd#sX zo*SW-g%#6|o*}<sS1^!tzS%ZOGWzm+pL`la26U6R|aH!$jUj@?UL>H@Zh^;NCP& z_A~33pxD&lMwR=M3r~lO%E&L1Q6d1KKwrNVmZfmxV1Ay8t+eJ8bS^-P472E1q$6jd zF9e4$MOn6$bkG8bE|X=ObzH&fxT4jyAN9TT!(c0W;GF}t`BB(>%_W~!T?tGp?KQvB z4ie628^i+JMNPAmAv`!!Cu~TE8@FmCCFHRxdA`QL<(JKTs8N+Eowp+S*CdHZ?QxaDd{m@nYFTcW7@DpF|4zL>99^b(V}iv!gp0{!N9ba z2m$TprkcwcBO@!Uq%JSB^gyvPnd`rbU_>>esAJ)5-&@FX!`u<_x&Og|$$yjZi`#~f6k|T0#k=tvBp_NF#0oyz2Dij-`t*jx2t!9e+l4mDU zQLyzq97I8~r~`(cT6GaM}$cFId(LM+;t@-D?aFm)wYZN>ycv*FJ>- zmRY=EVTcO*F5Mnsrp}0rHZT{C*llK( zx2!4kOX*h?R6>d;?w-T9Xc{)CWN?>87rKz)38WS0r*F?-C%8k+f{4CKp)ii5I@GqM z$&3Wg8kpcGf(PNU43aLbS2^&r)RKsGlJuYe1#yD5IXXj>;U;x`(j^nYM+ZO$1TVLe z!9WAMIEGbME(^PAU;CCqL<)P5w|%GEelBnHRnb{O0DuFOhykGimbV?3g;wwFRL6J; zPB3e{jpePPF&=xJ4f{=vCNW(ej(SKZSMr2vW!1^797EJJ zaE{<&l}}&LNg^7G#gj@1orj^JF005?PPJ4Af(Y|rM@4(WYLk;8DTy`7NEbG->o!Iy*;eV$ zSPa}%<92@6pB#~5;f6b0Fl;w{(Z(8;oudXg8_j2?FGAmgufcd$C%uL_22&31(>89x zTZ^MO*y}n$@0%N5yP;1zTa_w&4wSnT`BxdZx$=Z#?x6z%36UG=kMU0Grl0Cpl zBpgl3iFlY$2ZVv{>p+bmCv;dF*kei*6DDCvNlYv|TD0Rp=?Ma@<(Fm_%;fqO+yru6&=7{AWhP3T~Z=9dLFc$bhST;AVIsjK0fd569l)_5GuvtyJ2)9uV zzu9Sw2w-g%AxJM7Zg>L%6Je)rGqVUAX{Fd03~jKBR7Na5&bEr&#R3pclWkFZM!Ct! zmhIHZI=q3Dc?UBD7{{XhRyE&zizY556L*sz-_1&d9mEf38KGJhd^kH-u*epbnKmEc zn8#=}bglGoGuU-CPN{%z;#MVAIoLfwc%i5<9K3iC8{nV{g~F*wM=I78*_MiiLZMVT zk&dL2+oH)-XCm2^>X7e}A|;uMfG6t-to!c892!bhpd327yVAW~{S*88Iy6RgX2QNT ziU}0bN-9KrOVuk&ecc^7rsm1gG46EzOR!Z4W2I79$#8>8H4Me?1oV0VCyDztTsGjM>5STxr`jmZ{uW)##7;w)rE`I$Q$2*L6R` z-VG79w_p^RQZ+ZcMFi|S20DM1mO~TpFM)r-SdkXq8QeE9xNj}PPxoi3;}M7~%i1b@ z7&bnwaEKE^tP(P#2bN@yx}9Y*Zt{;vRg@9`qD)AMM6C%M1jB#+^H1lGzkloHFK?fI zWp47*ThD*^{h{;CSaX_;CBL71cK+Gl2M^*SLkE^xM5Khfd?_&XUlO!7+`_jw>em1- zEB|dmLDUrgdc|=!MZRI#6|wibNTL`ptgXz}{p!0tMKmtKb+LGck^<~)fBE>yCwJ;2 zFl~7Y3`_wn2&zz{CYgYF(I0v8$$_1nUA^g^-lv`-6T=*Ta6VZwA(v{#wWgZzC>d5h7+(j!V(anRC;xWLJ*aT@E zF!V=ykY434*VZ#Z6xb~pz|v)h{0e-kEVpvOF+RmKkZlXFkJZN~nXD>V@!IDD6qlmP zT}7N1_#kuxL+DKt$y^nuXB(roW5RMMaur_?u-}kGs*!kRS{pXZahf#^%N0$NU3H4S zMU@17QES@-A|Jr!s>^~_;TX0diN2yHkX21f)$8Ri%Gb&l;lJmrjZ~Kfr@$9Dz?l~s z5E476!6e#NKH`eEv_`}A>wLp2$|)SevV*v%_IhPT&95e>FJ!*DTDw=mbK?jSO(#=2 zptDNn(-``ic)s~!WS9p>sLT;g2_lQo)*dWN(%^`K26?DXw6vwWvvxOKm@M+B!AA}3 zVdTZ?&YB2IQLuGFk}NA#Au3g8Qt{_Zq6{%|UF%E+T5IpZN;?M+<00{keo?+e;OU$a zR2bKm=rGE^Ex(1A;{V@i#-%uO7{+uK{^F?4!stFPU)qWzI$ORjh+J;q!6i8OooHJ; z4>NKs@bYXovKguyqpdNQTxECjx6cv>lx$RrH?Bp8hRQFN-%gePSbh)m&~MA{O$e!? z5}E?rr2OSV`zkWntQaGgB?X}j!(ao-k%{Gp)hPKBl_>f0iv@ekB$8CX7r^&i63-Vl zj6Mn`6}RF_b7kSn?~3OKS$=R$oP*bg%h$K!!nlDQucZ8vc;f~>Tr(%&^?7_9_zBvh zvu>epWC>!L<0bLoME#af+IV6%qtY7f8ZQrUULB zWoXdma}*7@R*m!1$2R=b*RB2^aK>@0q>Eb&%v9DuS7Ca$$t*@qHRlT2RBhCVKN}SV z5S1c|2k6Fcb#vOWM6t5&=vixoMY__fW;wp!R?Op4rZzXoOvk}9#dSw@R;vqsf4ANr(rvoM*?Gb79NkQ zU@bPCtI>yn$`VGKf~S_sZ-7yGL3kfmQJyaIFKERq3L+*YJvEv^@)BL?fx+ez!1o`& zHvja~^Phft>#Z|;Tjj*u)TOz@hv%j$9K)5M{#l%?;5wA%X`YviP(0)+W|rOERlXMq zmi&284gA)xi{`4OQuEX{u!&U6Nw#(A0yS#F;<`-_@(QAxVG&k24?bIosI)N5;T+t{ zSZoPymK5$zq(|3D+QRb9=?ew$NHFtBI;T$uFJfDoHn*3xf}6FfpA`;Ek4tEeRuiv{JC>5$l%FA&}6ZQBq!jn>`Fv6;CF>%lH%dT&HJS?0a5*M?$1qJ+1l{G zvX7_gi+K-j)HmJUn(ve^*xU?0wgZ!4T&K3AZK+YKTc@T63=5An{HuPNgBA1E#86LnZIOPnaVJVOGJE zvhZQO1+S;0X2E;f&~WB$2dk&d+`3rpu*JTY77S{okmlAj^}r^HhB?u-cv>8W$ZSUoo0^*mwbuO#?JWMvYb#VzJp z++w~LH)Nks8)lGi+I-TWrKRYzYnV_{6+x;YZFB-qay~=<>jRo=YtK4UH`!86@PoC! z*h0_I+Am4%&Is)A79IZ2OGb(sq(BSs2-RT?WGhHpG^bMwo+7$tJIoFa$AP6vl0$+R ziZ1fj8cLoc7AMrOL~_&4l+8M2CUVRsKM4xdJC4fzA7#u9k;Z09NIz^#BjrVw@mi1 z1j;DcB+X#IhU;s#9eZm>FX=Tg1=HIASi;VjISv@G0Q3C#h|7M{&GBx~k1 z>{vnW-tALg8rTg$1~f);npHkAgQZpDUz%(#VOY`wuVQMbi%};Nd|~s9UB$GjD8@W( z>3HDNyi%P;rN@724LDjpW10E2Ystj1$6&o-fCw)TV6(3_T)=H|*fMiuf%mU-0VK>@ zGi!1&VoV|ofndyD?;T!scpf?nDYj{zo+G*JrEVH<6a1}#aX|1l7V);Jo_w0>H=iaV zgdrNL&Zj;8Gpv_p+y;adS8sMnq7eQSu55$UwrOhZ_#SZN6+y0{aq_Ldv1gu|%vrs< zWdqGLz7TK=iW!yhuRIr@dPQfX7Zn16Q#f_fyaRUK0)xMZUtjgCLB5wD3Hxdq)?N_j zLzw1`V5}VO5eAqB{I7mCV(JYv1=qwfZDRe_MRDogMKnfCh}O{BJ^t%bKTGGRW2PD) zP+h&1NtWs$LNfvK@VDQ(ee(9%+wb0f`}TQw^wwHe8-17ozU)c(y6@h8t744GbMa0$ z!H~okW;NKWEbkZ0K0Q4sBI}54gj&3xi~X%=OCH+;?B{T zkMBJH{ecrRpU-?ebBWwJdgl;4bL7qocitlZ7oLNEm;Q0$&M_>cz!~RfF7cmV;Mwav zZwBA<;>=&k|Gj`;7QcGMtTividm&V8OVDh_(Hhu1BGtbgi;dp~uV%w0lysa^dbYE#%v)bk33;jHX>iS{J@&oeFZx?yVK&t@3s4){@BB z=@B##!TUmOa#h719d9{V@zj%jyAE6u;nqQIYunk`!dnh;z4G=umaT)51wTMfr zwSW;2eN&@*?K@HMG=|3nAv1i|grgxf7LG(kk|-?M76vzv7>)=M+(gQ__n?|MZPAJN z47P(grb$MSbS?o7aDv(_Mc@yQW~f6jPA9FJfmwZN1A%c$HQ6080K=5BM`+q0q9g_1 zEG^1%YzIrHnd3zz1{I{o+d>kg$VzyldRfl23dGKUt&B8ohF$F)>wW*LX1hyR6t1Ba z1sh$&9RuNc*ahX%nqO zmKy|&v)$5c3VU_dGEr;IMpiXgU(@Tq*`!V}reeH$m6}1g&q!P?ZuW2S#RfL#Q!>ANM z9X`QQ$X(8iHOEE!Fu<}jIJ-uerL!~y5~IwpgC{9=Q=`MR*rHwDqb?ou4MtEAWcBL< zM)hh*%S;v$Yd0OFcqjxuK*I&oeg?M?;*{Y`$F3Xw*qcNVoN6&7$SAg{jxRo}4t}am zu7p2CM}6>ljV1U?J58i7d_)FcEpP4_GZoz0l`CBcdF%IEiOcj`fvF`S+_+ryV~}@n z)26M{(Nf7`v&LnXSwYv*l!!4kx=D?#JR1>IKnd-FwLRYyZ7o~elx-R#8VtjtSclpg zHRfp^P+Xh*7G`6+ZJRJ1IYI$T`R#ZX=`WuKv#%c9r{(Edtj63G-_B-{3jAx zI;fG(n%U~$wHsAymXc;3+)`tt!3@>b4I>zi2`SMT4j^AX89?csIvGD0`fBr3#-nVV zuqeb3hOHu9k~eSeNt^q&4FtTQhT*4=`$8C2H-*D%vJi@*AlC@GxY-~S8CN*(;RGyv z#rMj*WtWlfVfvlvSL4M~{M9Sd$Ka(S@Z#6g?-0&TbrfDN@nDoW29fLQE5bkW8~Aq< z{^94mLdpbT4@1`e(GYUIf$WNSvC{`g@pVr3aRZ0pO+H9R;Q#M%28CeeAD~8odOEs- zOD>`PWg!%;E+zZ!GXCkx3k2KyfBN#sPk(tg{`1?XvBUW1w~qXD_B46$_Rask`YHMO z*ryMUz5D%vSB8cjynm$ir@!3XPE~0&5SbW>{B-n4@HFm&w~svd=-7ippTgjJ-Ou_hCLGl*@6gWk_1umWYv|*lu>-rE=X}HB&c$A zqSn8{nU>5Bn#kAOPm`LFZZabNAq}wN4!dH^=N3XYai=F?aB4U>n^3=;V*LPR+e@ zvSNZrrDPxRM`ex%kcEi2BBU9j%+B0?^Y&}E&y_G|Yi)*gi8X7hkEd7$P9dQlomfK< z6J&FrdNFD0j10qWJ@qlwY=+z9{{42_`0uv$92NwT6T)4+jNhY~IK8etb%CM& z4@ofUKHEkuhcis|Qp;xLG)};VC1h){neOiMSS_gL2U;yGgln*yfy4puz`}GdfdGHl zL{MUN?@$QHmSGp829*#z0aU~q#6OPW%4u;5>Ww%h7s;3nvRqzYXzEHWbX9EHVWu~g zvvr1x&>vC#&HTyn-U7D}{Dpa8v_5PlNjmZNJssolZxvG@k7l-WA{!Gfy@7@WI_6n( z12d_?y`c(Htoq&vvQdgrlA{`8{i3)f(tW1ufKB2s3#KlhOd*SRYkH122=tx0KvJj_ ztW(knTc^#Mz*hY3@{nMcj^mvgTuqK{+M?MA&xO^2Z>dRZ>Ox7CW>dSB%xEL<9F9$3 z%l4Uij#?fK7YOA%`U5_}H4-wig~BUA5DqYk%8a5ECj)t_#MS^)vyBX9T+2?;0*MN0 zh~$Nwu$rs-BgsB)cas^0T;>)VHFY<+FZc}XgWEr;G0!ZND#cKI=+8`wEXa|^P$ptC z7hP!^a4e=Jnyos{MOALV^-PfBYj8;irVnFAe!xuT1?w=6JdHBZN!)N(}7w#UqHZ(N*(wW(#f4O_$(7Ior z=N$vjemd_e+&z36xmFJ!o4>+O5xRQ##k+^T<`~w`;nm=!1&2``oE{I-MXtm;vOh^+ zG+$H*)g;p%e~=LjUk?60e0!r2d9&3+8maY=M!w^iemKLz2w>_mL3#QPi;lQv+IVOV z&8l@YFE&R9atz)K-mp2)+vnNwJV}4wS$|cA@vp{cKdG**pt^SIHPTwX^an@Iz$LA1 zflXBfaYbCfn@8XwrV0iNd6Nul&w}&05nQ!AvY@BpMAx!9E0FJ>e*+&fz3>Pk!bjqF zK)@jKI2$7fhjC5Fz^2@0X1JbuKclHqsiN^HQH=^>xM9LzFr`PH9D+nS#ATk>w_j1y z+n&}K215<>2y(r4u)#1V=xnr5DkEZ2u2Bi30fO~AM`teIIl}*aJo5>e`S8y3caGk9 ziOhU*=g6IdGhf_!>CQ2D>ZO?rYn`AAZlRe^XFkSDxy*med;&)Y16vA41jYk<%CW(G z=F**G9xtylab%w}M@DKvrSaSwE`rZGND-%#62oBAnPmX(dz#16x?z)U&1wMGk7}uu z(?)K!v+%&%sIF98Z^ftU4MvFwF{R<8*JH+B(+3}^B{h@mC>Y7$WOb{^VgrVW32F_k z(2`QNwIS`EzWBiQp`nc=YnoQ7U{I&9k)$j$!BadbnoXj%?M9Nf#tM1OE}%D&vNPIn zvq2_z_H+_eiUlW4?bEU}WqKx%>|wa!kNsIn1n<2DJ*7E*>1!>gtj*h8j@cm9IlQ|8 z=HSh4&eZcBKF3PN$7dl&MmE8yRGlRBJw0#((a0DZHU9IYV~rX0sQ=gVJwIq#?tdK# zJn+5LVXi%vhUI_{(8Br@^(;_6vj$*-m>@-deLS^0n45HUGjBJ560DoDkt^GS?0b8V zMA#mx@d1zj;BPVY9MjXLv4)MRGJ-Qga>K#L-n|%mfLH$6%;$HG#y5j4HX&aVJaZM9 zoc?hFX`PV3>FCTg@Vq~VmoAL|;u)m%&Y$o6Wy7Ni@K-5WbLaV)Nr){l%(`9Z^U{;Ds-#0q3ocRRUqwS@ZEw7 zF>Fsrswjs%BOmghIQ@>h3=w^#asA>lueMFo4(YH=Cv8oAIrRbi-(-|#EbOMy3|&iR z2+NRRc+5x#O=d@_g}4nyC=47)Y0PG1Cv*4VTWpF2(T#*AAlfoB22v@GvAjdz!$#q$ zF_$FRC`0AvCi5QVuP|Ze^TD&-+e`0SfK5f}zTfgP7`K z`iMrJfq1w6;X)Pq{SY8Pp*qGM9{3M zAEzinMD$&jS+c!*@7~Rue?K%t1VJG1_}QfH+S&*sENL2MPD_$8xMAx0Wlhn|-A9w! zFkD4H%b5-%Tgy$|y-1@5+a-s1QJwUfhFc%xzOo+gN_6BqI8q{kx~R4>)7a;>R4sWA zUOM;l^Jjkg=+kCzp?YxW(+A(2!7i#l{pG>??~(^^pLuZd2>I`?CVx75ppLZ^wc1)R zK)Oy7Zgx>am4wYyVpxcT1+c)1_9rzIL}9&UH`BE&(u|CxP0LJGTdw$eaT2?N5$QF3 zq&PYK>JldGucc1y$9(Ymuv_u;^x^5F#jDeYidRrO9aYpflHyy%iQ-MTr~{<<20Wac zKFHbeH%C!?pRX7bWcuLr zAr6QA8kUdeixY9Oe}C~T2dy50rSo|*4(ra%;?>`hg=Z&8@q_8Z#Y@E-@RoOrSK$<7 zVYu;IGLEmp=N*QZCh+heo`b;K-wwdoIiwHwM%QkWgnjumffJ75D#ll~iGN!-8~)fh zr*?pa&HL=4Z}9zrkJhr@7}(iea`=MZK^piMp4_n1s<$zW<2vaQptKnd}h5AWS9ztetps3 z00iD^sEwf@?jcme^d5*!8pY2c>~S&-#-RxtD+icKo^g%fPQ?IR_ke3~*QikdmxB0) zSg7GNdrBK2%pkC>fG6uFof530iBf`{rjfR$&hsA1ex~Ju*Aly^VY&$x^*DA)#SV&Z zLx_?Ptzlgq%lMxO_fOrJ{rXZ9PJpdFq)(E%E?t`a{QB%?ul_JuCmHk$ zR;b4xQn5kAzYU^7w1!^H@VguD%I7WSIF#60ijo{lk-zxSPp)9n;>Ok$=9 z)lFpkxB>tZXZb&By}ZZ8NJj=W=ep3{^v2mk{cDOa2QB!ZAbc}V*?KU$! zXunFm-Y5H`f{L0EsXF(!bb7uWmOq15mB@Hk_V>@fHV*%=54)U$ntruT7V2+*Exv|1 zZcHAv!Gj4rr9_MQP4O1lV_JF&Yb<-Hm9*%vvytbZE)}n0qMWmZ-k3f?;Ei8%a?r`? zqonv1{DJG!2d0n1A3F|DVA>r1>Q(p>UzffL{{7;$jqrx6#cP=Mnm(`*lVpeRo!=k$ zsO}QovixOGL)p1C)SWxbfTW(TKm9Yb+MpiO7gE%r*^?p zBv24gAcV2<`GUQF|K7a|&CdNLd2lXcX22vGVZ=^q za5}GJ;!utM@+m_@u$?fI<3h5kw z?9<+1AuLuyyn>OfH`6TcYaW@wfbp0dgAwzB4>C^?yV z9+^)3+_Z)oft{vN?g|ZF<3m}FkSc@~A?$%o1a56rkRL~>aibsqL_{xh%)FVPNqFbl zu`nnoa2Ts0YVYAe&K4xB(5H#l&%nmsimDZ4(p1HQ4pXf8`v2 zzm%lbnx=16MUd3$^sR5|;oEQBe*5;>+vje70u$l($G6|T{l>i5ef;((w@=+Z1BV}w z+kb{<&cnkqw@<>%x&8X>*KYsm_DAsa+ogB@`Sv^T`djfHno2XLtt3Fvq&gIEWq-VV z8V2JN0>g6wuKZMK5Z;DEWM}>a9{mwt^Nb3C6HgNumh&(yZbS0^c0Zm~^Dy zGb1Z-CB-LPNkO=hTtjK(8{e>#xvV-r!oF?l@|ewrNvAz^Hd|==Y}3CY zl+Kwt=fIYc%y2~ACU7LdHO)H^V!_onMAy0*DGrRZokc!9vjmi$pWP7;>@4EnkWnF` zHq07<;K2B%O{1fu0<|2?rbZilQOkum(Q3<6OS1(y`C-}e;pE;2<}U!bb{l9yZBL}! zr0JnwQI%+A(<>eLv`tKi##TaWv;&C{OA2UBFrlhp@m9&5E2Sg%P>XEc8rT>E8XL}P zMwe&36Fsb%qF>rfH6D@$C0d@x_FQ482^e=O7zdn0<;@)eAH^Ca7N|?_QF|cLo zexAofO8s({vP?Zo1p~sgYZw<@Qfb9FERU?$37wGpYqx1ROmNV}HTB`~ z#fRXWI;K~O6+-KOMYeXK6lHdHboY(-Gmfb5MyrrM)7CKk{V(v=LAZjfy>b53ojv27 zy34>vCp&wFh6K_o$%-n+qz$AMp4`w}ZtWm=IT|&bjAiBvf$yElKzMGLNc|foviuDr z)xkOhLEdhmxrA$_VgKI_J`G#dt@9ySqJtT(GX~$dOS7}pEKGG?j1NBDF%Ud;Z9gL< zw|IDE!X9vR*e^C(ML?EVD)AGbMQe=?fvst1~eTk$1^{lb} znt|%g8qZ|egZoEk#tn|A4oiY0DcHUzL_PcsVcDbID*I&4P?3{Hj{f?z;s>biXu2L8 z2en#BFc^77P$DY{#CSSjSYs*TFq&%$FwJ5kJiWQQoose36Y~~f^6&h53XSN}bt64L z-(N!>k-#>oJ}tep3@^RPRh*QDox}DNvego5lweMHrRd!yO~clZ0+O?s8EK)(Of`_9 z7Tjn#mNTv3j@WiRVU<2^AzY4l!eyVBCKYLZ)D;PAJvMBlqzf#O#57Ejv?Ji7kt+0Z z#H4{s(JG=vMD^FlE!SgECu8QRUa;3v5kN&Pq*x96=+ev>-5RsaDpm}bB3m7bw-edf zKDdk6$hrc{S$#0a*1;N|3*Ws~9v9{x2+F%GPD${m;ye4&Zk>gzrMrgaqR%o%G_3QQ z!=zg`=K&t`$PHB#ln??pMAa)B9Ibd?{IO4yVIM(m01d|b6Iu{#>9r#*GqUz1Y&aa_W8Q= z7N+MPi-K7h%X+nV4O^Ov*Ggbh-tZfE3d11c=N#ZlI*rq*sfZ>~Ytjt9ISOaa-#Ipe zmCcn>72bKNE>)ovZZ@LAnalq;G4nZrfq9w)EM_!2vVhz9}P^HT?vsp?8 zdgW&D6ze48+YLu+((Wk2 zbA(bO%|={3wUOj?N*x=^Ap|@VOV{?Lm_h!TFdla?9(F-E%;0?Zk-~OxJWqolc?7eM z4(FF!V{*>m;<_UYv25M0<=72K5cFxx5%y^NvdqhC@4>XmenD+2BCEoY8f2AUhq8T) zOkEfmHm#JaPhHNEVN0{^sf*6k1)C8VG4P}x4eFxd{hxznrY_V1X!vvSsTVz0mjy51 zX0|hR)}ou93T-4$shdJlTPS#uvfbfdyMlg-g~A?4V(G|_0mJc-D2Sm)C?T=#*U`Gd z5e>9R)Cfn&o(U%Vw+{qG7*;n? z_53NDjG9)KP@5!m3V$57Mw>BGMz-ZGb6C?^A|)hrGi~QhhXdNAP$;%3q-?Uy9Bbp5 zR+`XR8<&f-+ccvMPSSLpw&7R$H%SsQPVQl?$E#=kZcX=So8nQtvveI~x0b|qC|Yi+ zVPwG49E}WQOgEJ=k#Q{WoxyQk3Cm4-K#bmr?)%md(Pyppi+pEhkkAq{F<&9YWAWU;5n1Cyp?wX$uf)A zr#+CKz?tVT!ZRYF>Uq30ckBQ&UA<-TzWMV@*W&*@Iq~3+pKg+*kPrkr;PLVkMf-Jf1P@zX~) z>*Z3LhMD%KkFWiFxQ>iXi&*iSs>XSqdsdPKEMAK-!PE2B=0Q#_itXiPEv^c34HaIH zBHer_Ql=w-Z7A(xt!~i=tm}n!0T*7C^IrUX97PtAdb!~!33m}9cQ#7P7LJX3o!C2y8O=Se^?le}2Vqv5MVdKC2huIIMrM@JhrX7s?Qw z{JtDv+-+)#y6m!z8s+QKmTB{Achj!z5GfYehA?ZBW|HRFqE3ylhsGi7m&PHCEIgr_ zopac4%{J?ftp9bgnw1$&b;{r|mI;<0;{+`DjnZt`>T<2PT=zI0~Yn`rS2yZ7mt*)PuCJ$Uu* zp%Zrx9=dz*Na<<#5s~1S# zwQQHAnqdl6v(e^wGo)!?fcUhbCE;y^8So3PFgOrthpXA58j6Lyg;%jY23%YQ1P*qJ z!q!&pRQS%FAb_QKWBQeOiZMj5OuyP>#wmUPU;P#)HcC1m)5qJ2Z(_z{J{1(a#$Zv{ zC_I2~kH_wJ0x6;(Newqx4^EwPv~-qf4%u#D7jiO_gB>iz9EzM%%z_PxPAf{cwa&9l@go|R7Vz0fJL5UpV}@$HpwcXOuWxFpBn3$9H> zRSJb;vbvG%GmR`I4$Eq_U21NLi7Kk1I8ZDqsPeey@z4-il>G9ZmSwn&J4|Az5jYLBP9%v)raoJPnNb87eVhvl2MPtOK);?N?h3^;ci>(kZ zSeol4;S}p9y~JgcUP@SyA`K_Kke&-R2Axx;mI(Ngrtap9X2YVl!zI{QHJ}YBGY;6I z7g;37JjH>SZ$D9q3ej-2K2i`%41zNc3p#g2lFZ;OXZ?(qdQv1|nOV5yT9uHp5o<`c zn`aihz)~BC#fD3%%humSnyi_#boaGo5OSV^xuofhw0i8J1sE>solPF>tzYPEIIc>9 z*f`k;LSFLJwpVYHdn6|4z$RtWNVg4QKSl#v+>WWsIRm4A$$SgfcfVdGD=>AFdspJ^Er`3t;&~bSu!0u2U>f1PBWm^j^@Mk% zQOlm4cyQ#>gZCy7o*Iq)bo5AYA93{|+)a$PS96zR7ZenIg=CTL11cu9wqEVd}d^ff9{>AjA_nMr@W# z=#;ny1z$9mhdB#A(bQ$4!M9F=mp65$B>JoqTY~^2#|5D0|AI5dPZr33{KtQ+dl$B= zTmzcAyeR!yelCs$34%PwG3<^#qflzHpT1q9Zqv%qCL~Ls zdfv6tnw`}wOtH7N>-I3yQ^BY3JW4%7-K@7V=PAtj3@{>yZwM(36OR1n%Mg)Iea$(v zVcrl!OJ*FB(ip+i21{l%UQL)fonV&DoG~mRjF<-51;_c}sdGA(R$yRK|Frn2-`1Y7 z?kC=2YuEe!!K}P<>H_>dmh~*sYW+o8UEacBArf5?!0dzKEd-Q(!_$g~;1S}ozDA|t z=1XQ5&z96fF=tB3dIw*WG#{{q@j4gcEw1Rf3B_TtkgcxbVx@B4fdJJA4jHSXu+zZ)CBvH8cmc1UVKRjikJ@?0QU${Q zDw2k_me@ZBx_;9xhQ{GHf!{+z8-(&JgL}pY_cUK|b~`JO97SvLyWU2Io#46!+qbld zo{C@aF6juW?8stuzSMU(&NsUOEp1T8wG0ZexJA=scABIYADI}BqL(6*4y8HtEuL~v^Gx#|4x)vpUV9o~SX{U+Q zmy32xDdee)gnXM^!DLwVT=nhWV6lBf)7^aDB)B4%rY(4$Im&tNu=+bTRu4S=C|Q*n zpiP9^ElXj+;Z^H23SpVVfdL)SY;D90zNW;M1i_<9NN&MAv{6zg*)OUrO3M-U@XtI-{V zk+(BLA#BFo+_7PkN216)TcY0EU0Vh0w#uYyiCUBB|ZTd&Oh>5umhoewMoUnq-Ge_fJR_GF&@y&+ya1LpXWA7CQBc$U{G zxb`%PW0#)40RxW;)sxdN1mAjTijG)PDCEI;Hxy~AF(DT5t^QrF1o0-nkS`Frj)PIU z=Ppzj%n^D!w2l*26%yoBNXIB!uu>?!K|-0!0Z^@+;WV2~$JpA|rT-n+22 z5<^kst*IrG3^cpX!VP>J=IQm)%tfs=a8ziH3YJUS$n?uNXR+9dbAAGOOp0G`#Jy+! zZ4+er1zvsq3d_!=`@K|-z0Mboo4762ZQE(-f4at3@5d~_`he+`I(K+U8L3+M)dx1V z7*jE26GfHhSjE~(B@?0eOhiPHi6*L6FXg_Ak2cdsm|+yaGpD&Wte<2xk~OV;>wXbl zY?gkOHFav_rp`eil{H=6ZJ-Eg=go)*GAm1!RG{SB9EzaIzH|635X6pXdM(ZN(~hP? zbhNaL-uez{|Id51!w+$4%i`g0*+N~%ZcDMY0cuk8CF`E#Gup*xRu2x{KxVHVdxp&`XiD6aHpWgJIQdPv$jnxD-Vy7+L-G$%U8aA1cuJ(@Rnh9HxRG}?znFu~cNOI0Kv?RZv z;M+~L=t%>b$Wc@Y5^L0Go>0O!P6>bh&Py;NGuMzyDX>kenBW`GosV+jmzpN*fZ~&| zJBb9&b?WL(mOFW+a7c_tL!o$7gdbvOtRvnPjVSSMX>&NfIV#5DU2;Sn_jg)#|F6#v z$8(2tq!7UmFmxqy(XfjP=(+9esWK?r9B@WxoJ9@Y7+|}K^wR3Cx z4!BD=pWdiuBi%59>^34M15aIG$C{@}S(+Sc8#P>N6^N+*X8z=OZ-Lvvqeb%YiU*J8 z4(ia+@8Zr^E24RV7f(Cr*gkK9$de!Pd*jGhTUtpc^OGCrk>5BwG@X+6T+VEumbmzq zkt?Mn*=H>YMdXEWtooLD{C)IJYFRXwFoUC?SIt=f)d$I(X&Ca%>hIs2ewOHwS&M2xoBJ!1yQWdt@#FI zby`yw<`W^0Kmlg;b_*|v%q8G7r8qWx#MH^u83Nk~ayzMktE0=Ab#sEFKiKyS8I#)DxGKDScO4PxH|jZwpP> zD8W5SEhji`dr-@HmCS-Gc7UNGf)uR=YzH9}x_9!%?3pV+T)%$r#S6_hd1k-)>+GS| z?;g5!_wYs1diT(u?;bvW_wd=^NriSZYc8%@hbuk+f9%s9F+r}u|N5RdJCNhdVtJ}r zae~4yWH&Ri82lq!^@3td+y7hrT18EUw85MjjWZ(-6RUqm|M)YeOJE-*8K&p|soA`U zb|aO6i`$%XOZ5mcU?w%@6yW*=GQg5#Aek{uJ+Qg=Fx}3qpk^0CgppK3gB=9u-37C) z*Gy@{ai29@7nloGf5M!4tFj?k0WjB!-1_LMpZ$K3bB|V${xgb zV2vxVjv55--Ll0^*``qoS=hT5DGPgCYQbBKx|65n>46%nJZ)%Ltl+q4`;nv%)Uf;u zkZqN8xRz_+b-;aQF}qGPbxUuJE_kn&%xG!O&oghNP*gyqk0?h3QT2El9+;1BGew0^ zbR`M=eiO;z&0!)-O^4FP(fIzrn_`$ey}7%c>_P6Sr|g9W5S#2^H0?If4F*G*VOpA* zR@&3@UU36`qJCls!V1+8e&+plgY`|FO=|`vdCSb3wgV4vpDJa+`^TBCLD)^FKA!r} zB-Yf~yp}?*uoW%~&XPhKQ2P%XuS)O07XA*`!Qf-u^LYL5XY;Bk9?6u2r{e9F#o$yM zMH)+tVaRg8bd#h^n{0+b!{4x+z^}mfK`({9v6EVw$(02b>Ls4WK-Kq56(u30R4=H$ zZ-w>|%{GZ`CSY@*WRdApa9@;fls~jAgbV@4W@zXSEsx3l^0e(voduhbGEW+eQRU95 z^J09PX;7PF8)&tue}jk-&mB@a@Og`VLsQGBN+Fx4McjcK?a_CTLtYDF6s6O@ae_jfXrHM*HL=|ACRXp%bv62H5eI~(T zs9B}dXpslmXSZlJidNz|%@-u@WVqZonc$iEPGrx`(tUPL(+eK7LM1fx85qX6=2Wyg zqmoGbZHdCBiZ4m=Qn@SYN^t^nUQK3*;+v?Q$(x5JN%14z&c&OGf`{u_mdsMi9PuQ& z>gTzZNpr;zXBH8wcQwOro5(-w5P+{<&-5d4ItS4c3q0IAsX@JH9Zzet2*9v{zz1iWcVkkf}}(%wh4^pk8vncjkJ1@9_87KlXGXz z&VBLC{r8(OfqUz#`={TTd+ETBXWxV?xqtG{foBO~$qH&n@hV(2#04Bfsv@~&Kk)4* z_3=QLw@Q@cXG@5!dAg!9u&XcL*T1W4XsF4NpaaZubC|1!D~1(0cnd{8cw|U9%sh2O z(ZgCJ`em(&q4ZWvtuBH*{yWPc#L9W5HJ}HxdK-%(B8|Sr|AJU5D3lkC8Vs{Ts-aY*B-NTpf9=>t+@T+$ZAHI9|*!*iU(rK4}4lQDuikX3|Ig-L=QvPY*?es=LvMUdYW2|c*z!X zv2x?tN90ODPxf8WQ$vDWGx)wevAI{?n0@W#xl>o>-f6b5{Db5mBd%la8g+q8&7JxJY%_c0li3sR{Ljr3v!A^?_s+E+CNKVY z@$)7#SbKXrX`Ma&?)|^M6&zkPcl!L?yKl~(y8u5u(@2g)JUmCGBcB#h>JZ!n;u(a2 zpb67j=v!V>a|CA>M+wO2fMYTQ@(NM;d2u5*+zu05b>;OVyy0>CdQVvSX2nrln#RU06nf|OkH4Y9dLEj zC8oRh8M7Jop00(4=!)$s&&3yCV&I9jE9uY?CT#KgY#xWr_uPsJN{wucZzg@NSVye{ zgVo|;H6lH znPabIU0@ZqjHy>pL39se!Fg>uY2E=_Z-JpVlO76f52Dg1o)YXKhXh5ep}-8Vy|;>C zS!@%-ir9p@h#eG0Qb$XqT=Ki!TemV|Yg6ZwdT>DeE-gnB9wfXNkvsrbe`t8L2W%lq zb+dwRZIt$OT7z*mN#0^P%_QA0TbP}!r;51V#Zb~NR^$q7?Zr5G-`cO9(8F^9I0^b`;b*{*s@Pr@uTLTnXP;1%{pQr^i zsuu9!&7WR9`QZ5bfz7&VY5{1|un?(P*L|m4FgZpY$2HRSumzWeJ+mo@39PIQql6-a z&F4{XVVYq<1WC5;7B-SWrs+fsY@&7}Wig83k)?ZG>X^r+j`?<}m6#ycWH5us4fa`D z&LkG^0&Y5kpll=Mo^-E?dq7Gfd8+Gl#Dsk!M{Nl9b;h0Q;0q*-kU@uC1Ke}!g{FDx z&HjtcDmH&M3}==xV>t*(+Lk%WWna4}?~1{usSE}o18 z#{;}~umN~`hnvfJ*%X7e45Be(f>=Y2<=dD$O=p2yBwnMPY?!hTV+48_JmbS>Mf9Td%ezzvRlecn$HF=KWMrs#=;(2AQdc9$>s3_KgJZ}AyJ4% z8F_| z7Q@*AaSJU zuujw9RU~Kw^oIk<%qUB-u%~P(dsb}O|B@}rf+{wQuK;WAVTt)Apt+pac8C%a6oUHJ zc0-3vhFQU-1gS z*t@s*M)8CA_s_pJ4*y^iD!$Ka7!%X4Rw*(UC*tI9f8D>o_%2WUPfQ<}K1#+(ak6+5 z&aL|2Z%OgB;@7-X@HJ^gJ>lYw=~wY{H{g42@yfwv6yHQOXZ|6jqDlGP3!hv322Uki zi?^c>437AlO_x4k`LveRC z*wD~MGGbae7#~=&QLp&>1ApqGj*BaJf@R^)Io}^R-DJkw$+8v}7OGe^0-JFg&1sg$ ztrNpuPDtOefgBNJshW@NNk)1RJJq#s<}STBJ9%gg)6sHFki!kj3Oz}3uFWz{S&ktA ztTuJU2~N7O3JFdnB*c{Jt#wIgMauvk(b76>EoRH&3;^5*yV^_P|4tVIN&=Ut88l9w z!vA4M#tto=ajP$(Wa>Q8bOQI@m`SLfaItPcy66ZaJruqY-ge3PCiF2F z2ZqfML<6V9)onPUPS;>)rFoX%kRXa-Ar=)R4-=v2p}-M+a^T=yBMH$Oq({mxrPQFN z4^K_*V<`r6*pmN)px3;Wu0mzCWLFuX83KC=L}$o8-po`TrkbX>M@`hPFpWUsI*}x%Zf&=@C1s6 zB{NR&B%;*YO>e*e^sAHF%;yQ^yr{bf}a)W&5v z_fX3y`Je-N1{cF>FFJy4t&RLfT<_(EWr0$QE+}{* zQFmE~>5!d09!ErqdMJ+gfloWCLQJf7U>|R8c3X3(u?p<$jl>Ho_>baBY6x@#*+(}L zd+J=y(Gh0OG?XDKJ#GSQ_OfM!r*ow?fKA&i=8nN`1Q!+@@8CkboS?StaqUD8QK;xM zZz>@_n_5Gq@4Kwa`Iru3CmL@|n?5*wxKcjFvg#zKY=URL;8GKn=gXv2%|>upd=tL* zO@7f=i`V(}P84rUpTOR`;~f8aVUuog*`!+<9T<A6%Zf2)|#N`8ara zmF6_T&S-`~M?LA%m{=)-^T$5z0(+yl60N!?dALVD@pr>@WlgeG*Vu@?;{IZ#a^EQa zZ<~k^4pn1RAOCu9Uk0u%w}!Mc5?3!&s-w0b6Sgx0fhf+cVJt=AoJ90Wvbh#il>gn% zSf2DUvV|HDi9SVk1&2Otp%$zbozy6Jd4PjzN}^R{AyU(8`}og?Erq<9HgoyR+Q|Un z;HyG3T#e)MO~bf<|K7a|o5kkrILi(VJ@*_0y?Kj{rp&yVG82Zz2+P_qX$tr*!cv7~ zVV_ZGu1WKMjs)K-7Gu91VrVuqs6{ ztF%-`1J#yeQ4|>z!t(f{Bh+KvySI`ZSbQozH8ivV{u+a2as(+2bj{8+n~T6~hc>WH z^>p@ma=l`GQd)5tR$7U3($(a8pLmq(`gVE7`+VHC%fczX=?BZ0aWbm8lE1M&Jj@P<`sGZjMcz zg(o;!^3-KZn>w4#!rG9f`J@eZtz+6`A03&x=)h@awyCNMz0?@C;5Kc~F^lQJXI-c5 zS~^|qKY27Pd`~?YZlhQY4n9D2?%OkeKlt|-zn%R1(Z3(~_8a(l@Y~DZe*W#HW=N1O zHp#cIe|zKGi{IY*_N%`iBHy0-`{8dt|N8;BoJ)TX4wmZSQuSo>NSYZM>@A-BXeB%9 z+eXOrTD_~&w{UG*je1T0$XOA zX^*gw>eY|0)Z`?eH94?}D#$hcd4WjQUN25!>Y4Wz9pTw%q*a^5E~MhE=@)q8&vl-) zZ6?bU-{aj*x8NeK7q6no^YjZFkw|W0`q1===~uA>3Y)GbxlHnn;sihGYo34)2rFUW zuH|6sl0qR*zoi)R6z4)di$xMbM`P+raNC{Wv7+W&H!}C|=9!IzwdJU$gVSo;uxW4v zH}^}yg>|$M4K?A>=m?$_IB?eU!ozA9d{~Cx0?P|J%jIzhZcfV_28)6#zOCP5&6r2_ zv~=Llo=QkiYp7;@D+zAximMQ6#-xyP_uyyuUOaa9;N-ZmB>d70kXRiEk{rbHZF9fd_(Oj-T zb`6sb-SvRI8UUzB2-_2*)qY7p63d$cK}mU7X}&C8T|%|t>GqZ?8?`*0#XgXSq3j&o zPQUxz?+CA1UPp4xTcr3kT+_=!VAF^Z)Qb^R^dvltw33&m!f$sT7)Pl?fByJqJi%u- z(`zU?BOwJ~s0Oz6_z(O*!JsT>=GU_Jq98L%wS(XrO|p}@u}1B!9EZRH#h+qj zbYn8(EG0ePwYg@6>S0*Dc^Kr`epaY1&Cb>#n=d~X--Btf&Wsz}1qe&9Iw^vL#8M%z zf9}~pd`f78!4Tpo*A&a_U`&k6q6Bk)rbSbzKui8NF)fEH(oS7mJ@KYTQUchllR?)q zOyus>De2ej;7eHtGxj032D3cqtl;`(5^AG{-uNKz$(B>7GusZl|xDUO(_ZaG3W^4f~ zdA2+^)cW%{s&rH|8{4M4EED{uu|gd`qord{%VLAHbkGzt+V)J{w473-avJQBF8MHJ zJH;d+0^SEKVv5K6X!FnjENXg*ENdoFh}1yqe8Up>R2$fiJ>yx^v{K;yI)#lSWtj=& zXh8xHgdkv1(;6$}HM@W_FlA@7VW-Iw5j%T2i7JPr;0vRcMJWPLVqLDpviz~1l~G37 z;YjF7C(N=v<6nF=cj<$`7Fw+a7#R_Vk(!3_CpPlhXzs0V=DxTw`|_)^Z$JOT&G&x% z?6cX=KAF4x_Kz>WICu8g8rn{TFf3MGyn150MJ&OGj@*0o%-orq@E=#&5ofPmxqs@$ zj~~4=`_c=+ea2NZo+QXMY&O1)H+TB_?8);#T)!SC_gZ_ddQc`_lOzKZix= z(A>ojXHTF1@iX}E`Z~AqJY=Y1=LHZ^2b|0t&DfNK`@mraVwr-wm0B)WSZw3K z&z(883Dy=kSdB^8CW%Y5$&1VM$;*h7mBTA(SKOgF5R|X7kk6x`{B_f=?T`-2L)Bmw z*nzM`E5W!F_WWK~+3NhRgOVtSk(Km9Ic6U7!&aG*?GF&37$&lUCbKAtmO{GuEWDr@ zsJDdgdswQn4=;aQARR6 z*QjHHLydb&(zMXC#vT_eGT2T5{B7ot)KjMAWUy|M&#WmyL}1&MFWNGLH?p={=|yDn zas~Y?`$Rv3)xt1ks*``206y34k8hv6ec|@ox6i}x_sQ))!mp3u@C^LMCz@@8x&0b^ z?N4t1>GpZTui>NH!SjB%{|p20CV`LoG-cxbvOIAi-r;IG7T0VJvPBO&34svf{bE~Sj0)u!6NH2yNizS7jkq)93mGz0@dr>C%y(F2DZZz2ND$)k81Ih+c}+fSQAB zO(+9ja>7?NvGA~F7WR@>Y=B4cE{oArhU%uZ$%cJ2(SW{aZ<1qrB- z6M#Lt6s-2q*&^^3(?K1o!Fw^r>jft!I4#?tiCO6mvrw@?1;pZ$l1)L{E>&F#acc+D z4G2$LJ2r1vYa^_GhYe5^|Ek`rh2sLa`^04TJJ+y^kzqEx4B2$!Tb%Fd> z;00UMM$MKrWW!PrSrU>P1}FSFkR>0ToPGbL*%Q~-3`3$I!XPZIMACSxK}I2zcW}++ zd}UxD+jOlYBV5ic#R*)TjAoOhPPH6KnannxatIrNB^o~mr5IAAhZ&6QV6ZFcJWG&A zRv~0}M+}BFqOnn%*tEbuAB%Wol-dNxHbvUI2RHWrEx z?+F_d#BhyT{F97pGiF>B1U5u#8U|Jm;MEN`v&HESaKGG&-~a<@CJ|z9<5QjZS-4jJ zKlrL^gI${p2j=?{gKx6c_%Nm7gety$P)QNwnw+|S>g-#^OK_8-2=a}x-0)TGHAflb zucwbsA6@sew0IuA>?Qd2w~AkTW~=Afc<}?ctQ&mlPEH?~J`5LnbNV=5D6)dzn0_5* z@Nt4o_NQT%pABrt?WPW6pA}=?kRmV|Q$Z=!h`JJGv;meh!L`|AT6zjaZ<`_Ru8Q$! zP8%ZwG^J(T;0vdgO0!Ng5BsZ!9$WSZjmv(O#udan?V6 zqAO=^`tOF^`s8%)gx@f?R_yTXK19>7YhBTjmAyN`q?#49gOs!{C;h&|xq` zHKfRV2+>3E8CNDI&4UZV#Wzrg9`3$65rjvnt!6yt2^T~m781meDoAq7GY2dANX2(P z5=A}bhW*T6d1W8TvyzzOAe}6Sc(&k5$aTv49=SHjcUBF8$e6s6bj@zq(@c_vml4dC zH|%=TL|I@FIFZXisDtrtNbpQ5O__Ws`L_h(ZuSsk1ESzxz zqTnxz6B~wx{&wRb8GO~~b*qAV&(|lF>k4f69^D+F$*jlw42zz^f^WlzVP$T3HHAnO zD_33Ap~Nspn{4WBVFoxI+gx2^!Sj@zt_54dplfBBr~26@OR>7q7(Bn68Pt#vswqM@ zNKD;d&YR40Q}gsmznQ~i0(CN`t|h_KGGYH?7#i%*Y^u}Wl+pc7$I#ji3%|p!X}6XH zCla=EysDk1JHp^I+u9A3Lk?`-ZfZN)FnhLQ+E~S~F`rHU3gUO<8mz~Xspy*GsSe=+ zaq}Aq|HClF+7cF{S={LOXp|}H@ffU(0Us|h7(9tYbuzG(SgsB}%;Ul#uS(Gf*}^nC z(?HB_!DwYP+7|JOJ$vE<-{FC(3elPZLl9WpX3ZF5xn!&wnJ2i)ufys|Dt#=bExGS={3~DQ4zEwD%BZ8-&%V!=kFYw z`2-fEJ4a_O&wMiT+05s6j>b2GMK+aW>}IaQ@8|z<;?6-XWe2BSoB0?{Ix=%%+^x|(A4CNTujLKJSNs1Oa0FFLB06I*<0 zSxyXN03G5Mf)Ni^k>Sro!h)#6c`^7Gel!c`szr9AG?mdv($(w|OL{q@D@tWVdC_&CnygK{U`#+w1<%es>8{tfn zIKWU8WJM4o5wG{GdfaE94dNq{sdUFm^ND{&kK?w<>2TIc>N9(sr!1K%5Nfw20jh%dVzryY* zsR}{xLQ-e;2GYuL+`YIuPMsn9OxG%_6z~Qbcx*46OsuI3No1!Zt^W**9paqxaN!1p zXHqW7nMR6trjD38ewLlph!~5it0{r@aPohA*>+kyJ#2U!=LJ^HVw>8l2OxMhaO#3( zE}Em2gp==dldK9+rP{+Z{89^;xh!>7Pdf0CBow+}aidn2xR_fayL*X7?2)O-Tn@Ke zDHIX|8)CDDNY36x{7`KkdE7;$0OE4ifaxsD4a?9KKU`>dnMc-7AA~0ka#=c#G(mg~ zV(%v*rhXoNHeFN9U*JlvaFEc=;>7gf=|iNocoTjf<-I=>m_i92we4`9MF_6p&8@2u zkEh|YIYLFkp=h*vA$`Ks(I`{L)zd${dh+L!?*_K$>M?bwD8(kJ)j@~v&9Q&#LR!bm zX|9PjXwlp#E#TJMuOZzn3JOgHkG^+mSu-37g**}`fg4p+6jzHhPIst^(byC3#E*9} zkkuZFVU5kvX>wYW?>zVM?72_wo%%4a zp;oKIDU57LivP}+#>$*xq$Is53!)qvUwl|o=v;jAGL_CQnxZ=yTRpObM-!8n@a=a~ zczf_pjG78(TYc>DGRR{vxV)0KZQcpW2wb*_%{1`Go6OtUMw4cOv|BnyuLsm*!1rqQ zzprSw0}H4^tTqJ`pwx5cD2I)mpSjiyu#YNu)UAJQB6x|?uP?bj z=vtal$Y<6N{ToKDpim7J#FL4#NGR{~w|>0*_5F8_&whM-EoTWLzp7)Y$A6Y|QkDeQ z%FM4Faa=eqNia*Q^GCk>MA}_@jA_Y?;Tmasm}PYEiL<6vWQg3I@$;r8jML= z2irRaRu$9xNX%Ecw7}NcLJga`cGhIN$Eyp>F$VUD3QEoL_hh45%-#6u54WD5n|OWh zkFValdg+JB19KNXm|N3LOO{~M^3C8^ZM7$Dlhr1Uh2$uKy?`N@9!7JmS|`I;b&g04 z=a(v$Eww!F>DV+l?iBK7tG$5?YtORO&`_L+A(Fw|{`b$n0gv1G<5p?IBe3S9Syovd zU4cQcNj9)0yXc6PO2Wu5QYHx3%^QW6HDpu>*T4r-ezO+;QfJ6py0*7e4S7o!<0bo+ zbpYY%aZ{vq7)IVHq_hS!TB8xT^94x`3$h1=E_)cmd_p(EaRol7W_etio~YN&C1{DM zW)VA)o?BK$pk?dM8hdOli!!b~IJn(#?Fdq9r8>e$p`z4i$LIaFZ)MADV6@40mOcd3Y`5q>Ghxv&bF@Hutg-(*nF^kq z=rGe-az(zL5)ssxZ?zg&M;)8%z%H5=dst!~AzXKj+FUaydYF}?#=~fA3~r7Nh?v|Q z>Cl)C&U4#;-6)J|ELmrQbnK~~ARRQtjJ7>fH!a6uNI9A&{rvUHbcsnqL=mJg%$bmP zvPA07mdJ9(QbU5O`j%E*iel}%$y47=e)!$w#CMY~eK$Gz-Q?NtCJ%l$`4@bA?HE6t zKjGSe?}Le(~MpMf{1U!Qs{KCjb21jrrklQ=>vu5LH;^R>JxWhKHM|M(yKCQe@SWA|}7{)+8wc zH%T0&h)J2P z={mJZtEfs!a3^z6%l*>|ol692)bR2>{t4rLxL7S^=IzxI+-oRftj6lhoX3CgVMZg1 z7lVslrNIwgh>su?LRM=+TE9C}7lXJQJV(9)?|OC^CUX-ykfkq9S_hIDHxqpQ0S{3U z|9|%0wYiBaT^s%@YkOC{A$xFJ-8y@}`+dqM4#9xOHn3;1sa;Fbvb3UBFU{(18#zoB z*dc*H&Y4USl1a!U10*4V014+*Z4LE2pE7biA7W-c?IGqcFtz`}d*3Y^wbinxTQ=AN z&jed)wPf|Wd#!aJuKT(W2@SE(k}^ecq$|;q^_ik4xj#V-2R6rY^(VxUoFw7F;V$g1 z_kc@ccykV2I_0F7$}u{=prp!bPCESDwo(q+tj-!J7Pbv)qMKAafc?vBa;^l#WhQG^&%qa96Z+<=jS)@^Gu65 zwbSQryo5vxNL_(M3+Le1OAful*TKt5dP>M~nw7#uIUUz(tC;iHrGCK;k>*#!KtYY)>mZF24ME~X2gJQr~Xp5kHG zF-LtGgpF8Cj_4)J8sU400Mj6EdmaMC7BfXRg}9IBsi~Jgg(yC_^g-7sN(V8w7;ezG zJ&qDiF;Ba^?_xR3j-M`EzVnCucaFb#=g>C`?C+brMV?IpX3rMkl+9OukH35L^2~ul zcfLLF)1^x@uYG#^i^I1MT?}jw6}(_AiX53qMblzS++4H>cntM9Gv8fM8M=%DtdGPK z9;D;KE{IVx-=Mhjk7h!aTJTko>zc}0){?mu%9XD4y=7p!JH8a1f>~zm*4rP3c|lmY z64^whY$7Z2%vfHG(|l5O2ZC`|BI~JKj3z(>JwoUO zDqVHacC9g3Lt6x`5l28($UkDB-Zb;|1$r=HJtr>UOr)ote;(P7W^pyM`T1Y}det+< z%~cqjR`*~nvY!cG6lx|z*b5v$rv!)5rjW`nB#A2RAXMkC=jg2_MdqT<< zE6|h=WpyME8G@r$`v7NBI#XsNGvEmFq%=WOtrUd`Bbjkv*|Ua8)V7j@Z}&SARpN2a zWHAE?BidO`Gh!j|Ti=i;cQaJ1j?l{IZMsaZ3zD;8A>$DF^WZKTO51C=A#3nki?~*= z+S=vY81V*KgU5eg?<+Dc1;o-T2cm$nPb!f_{R$O?JNf41?`}?BzB%#T&57dnnECYNPZv(z`s&in{a?*od+XL`r)J(d24~)V=aakpeh*LY z>^sm39>RP>-rRrf=ERY}Ms$Yf)Xj;naR?_a-kkXK=H%qf$^A8p2Ihc@BL`H%OT!hc zw03U#D-@PNKJ{x@hx+roH(qMFhPvtRrq4{DBGaEcLm&gTmifL1l-{LP6o zxMoej;@6;1-R6Se%#FhJSMsInue7-eKM6$*B~hJ4(_93-YQ#0SB0gtVqRPkg?bDStZ(7X}|e3*~R|8tEt*f=z7dyKb6d`j91zdAQI}`T?oEW!G%g}k2Z$zMi+tb zCJ@4_c@lSPlB>%lxq9Pe)HNe8PA@{3jCxSxp6z)Rnd=}++0vk*ZygHEkiX00`z zXg(WH5~wYi^q^AL3cE`v?0)d^-_D#{!NP6=@!&BJX300w_-8-2S+$zbN=xiUiABWr zL6&WTW2TQAS!!g>Jn{sK>cuyYmPnvfqZ(fd)VrU7TkNpfj;(q-^l~#_6&b5pqjs#b z^83o`mEX2UqQM1ppApaHjORqj<>QB3b;IftpB70fF^_-fNQ$x4c{(jemVQjKo&k3{EMORF#CLgyV`Dafx6@BQ>a}^Vr`#qNz$U zT<@RyX8v-r$Do>yXMF&!IAm!!u@ymlXbIKfB%J#;i`heNSD*G zaWoZuN04njRJY;IBAKVUG8vD)UzY__H8Y3xZM1A|9`-Don=PuZjztp(=db(TQe&i_ zXKh9rD2TOnZ>+Bw3`jA`jWoxNi*z%Yr3FzdASB{SLV-0^a}!bpZstiK^QD&!X-zS8 zl{s#8r9hF+8{Kfy79O0uWxif&6WFcY;gEn$IB}Lk4jF>Zn>XK|$eQGu*Pzh7>bUFBL z_L|?SGIROT?Jo|`ym(;dtBXHfxb)Y3A3XoOAuKlBMfPq~|Lyj*S7(l1y>;b_+izbc zdx!sRJl76?((JtR>X}>bPTan9cIMOf$-0^EU%9jI!0pd|clX*yPyZ{NV~us*`ta46 z&yL?d_loO`yVpK=`d`Vfe@&i$ezP$GR@#H{Ay1nmjntEAEJf|*1sL~#BeM@oKFjdT&;v0|4u%Tn{VpN( z44Ko8G;_QOHHr<==GIo!1aPI`w3us)i}~K;A`rTHs276(V{e^0H}l#8z^>J0o0)?r zZ@+y4mZF-09vtf2O0y*hE`GZD0itz3J{JjWmFg$ByzStCJN!=x`v=@LFvrcwlQ$=? z+?+gibMkCUqU|PMyE*ZPn-drB{{EwxxAq6Nf~JNPHHw;DF+~fx!FF10>u?_XG)PiO zhU<3@-z5Vl`PgSzBM>z+qGEuH%%c$04ohYj4h<T{qHiFkq$J8Y>y{mt*`+#2A`k?iHQ%k+9<^#0~@JFr{znBmVd@3v_vxt@;qhxNw!jVcw8Qs4 zU-^D2coWhusWmF3;U*nV)Xb3ZZ-$5xi!2F8OeJH>+Rpz-+QNnrwO~@Rl}X91;M{RT zaLru>bkzhpbK7_t_7+znFSz#wwm=r~dM@R)oh5*<3M(>L*kD%~bGQ_;JUA8FGVLpf z8TR0If>b8@XCniOfEva(b| zp6HBdaWVpLmK<(aY#t$<9=QZfbF_RZz(zw_3z`X-WgtSIIm6;&2 z9waTWLfK*#K)y4*WRjsYX9#o{s0z{rvw@?#NBS~Z>ZWL|70ALq>tjj8wR*`AgG>b& zPy}QUQ>63%s}&{$Y8TZ}v^_Y~3C>HFrKK>~7!^8KuY&4c^Y448cHZ3j9(N zF7D94BrHjUBWfPUWsNt~ZK&IauHHIyn5=`bQ)e!oyLr5J(I>5zAe@IV z51)Ou`xo1XcJEfmItKzWyle09uHCzzUP&m$lJ~Y}s#I2^!6W)2l@=X$ThG;*SbgHp zf?>qtKY4BO!#3i%RxqWAT(2H9pj-$^-!@ET)+QUNkwx8d%OkwAeS{e%AEl&dGCL#w z&wCc<#{W6WI0niE#DJw4B^i;1M+TLRRA}1Fa4=-k=cdp6`Q-HHWcvJ#7oF?4Nd630 zIDLY^Wx^}Rr!P&PMjkJ?VmRY#l*YO7VqohI2wRR*BLf~=QAkV4u%f$nHyo}-#nmT1 ztqeDdr)kM|f(*|ZX{}_@&qt}w$`?w(=T{FcTJTYuV3bALmGF0Uc={dnNJxpR_1?Yj zstYWNl}{?yDwo=)4py#BO)d~_4SKuw2C+9`rS!-uA5HCpZ+*UUX=c0@bh%#O6AJbVN$t3rY5Q(n`()DJ{a=ovx!MUjmY%=cp^E`v5({Yp=Z`YD?YC!Juh zF{FkuG91Nvf}=!Xv&zhiyCbz3aKxG|d|M(KQnX0DC49G|q;*COf~4nF zF>OP%rzE3@J|fPA&o=Jib4xgPAUB+DKa8iTPKGuUiIK!((^;n_wP=WJVtT>03PN&g zJh<~HVbF4%w#O4Lza@CDQ4$GllU}CoWmIFq9j#~=i_9YFSuv9ql4Eg|Ugjj=A1s)p zc4-cduD@$&m<(^;+DEo;Abnkf! zC)ah36@ZHW(7Cdr)5*##zwDJs3aUbPI^`nZ^VSV~H z60vgaa~f?Z$)qSsE4w=|uJZ;W@U67T&WdEx6@oO=jE(nER0(+lLcjOXARA>8Zykp> z3$N(eXNQU`4bE1l0|?|Pvm+jXSP9#=6je*?o%d|_Zmj0jG8*$w#Y4MyKaBy5PGNb3 zzlEG3_-8gh(@k}~wQ{OMjFDcNA%k}C2Hq}ANwR*HWrm9hyVeonuOS5O9aG|--svFQ z?SHiO2$;v)M+oVX1rr>j7I0zb8)8SwzJ9EH%3#4!UR!CFr%mc=42`;CfB&-%DV2Rw zfwXOha6w?SUZ08S<{~R;F~=8Zr*7LY=*h)Az<91eMXx>=Q$Cp~pOidJx|zXgrhyaM zmhdPt>~s)zwQ`=Os*0M7!?NL0baGY0eb&OkEDlfaQUbdH37?Fg-g=YbfdeY>mMXwv zE(Oh`aovUWc6etr@Lbt0d6Aajzr3UGJ7<_$1up8Z{)UE$DDha+j1OxOpDW_ig^}e& zzCw^8v}HnxYE9F>z$Cl#{Tn}B{{75>Lo=r?wB9^(zV{yj5B@Hpk2vH>&BD83Gi|M#ET z+eu&V=3&y^*S~!i8Q!#ci1csX)<@dg|39)#2#Ek?SQ4Lob_3U0XYI&1-D~CSLdu|A z-z#}GXMyKFPFFMn^KlTIn3wZd+tsm(KQ(jnvM--*+xDIEs0Yr*k=B zq;qZ}GZzn)Z)Z_UK;*oZgBhkt&$# z0XC{~VfXHTC7z|X@;1WFrY0*_5R)}E;Xtj9z@x*J^9$jy=6~pe1$z_Ii=7*3Cd;h$ z>axB3tM?;GdR5+qE1x>No>bnL+BY>hHSxdxiB#UiL4Z-Z00VKga(e1D0uT4Y2wlUG z!KdHDgY%Vh@R~DhQ%9VazX^PURo;iUUqnRQQg5+{67{sKtP!6#c~{&!{OP@mFWh?z zzWmj_?oRyv*4J+b zPhu}m0~r_#k-?ru<;Opg-IAl3mE*y|aStmAj})};a`3X&f*;cE^$%*nY_|H#WN>Gv zQS~?CYI^k@c|_t$I8yJi2FakQYFLSC*tw^uA#_At53SLg4%es$G5ql*Hfhm|fe z1i=K&)3Ihc7o-$cp1USVIOfu2i1`DEfq|l`644UPwu?Ofyvfo|+hh4-;`EMiCV&0w zRd1R)&UZ$Q_I_b=lWgP69!m*Z21iV6Pg``9DVtoedfe5f<=Y^d0_JPhAPzRsVV!wc z%ofWAIp}?$tWmQ>?O5k#ljcokENf#=tY(mye}gOu!q>7Jcn`#yhA8Gr*2OcD8B(9X6eG&_=0sK(mIeC|HbGNe7v>7( z8dXH?MPv1jfp7nLdm+Om3LB%p!*%lPv#57Lh7A0Jp8rPD+llwgu>XW)wN0;Kc^Vp}EqKf9B(SQ7Rr(O=1YjSE>xs?-2!r@OS zTBLrK1cBjfH^`15!i`a(k5)MdxdA(pNx;KrCUOVX`lqUu)m_~k!+Wh#QLHncCL??g z%k19WNy5o$iS@59z6p=poyQNqU+R(MWw-DG=@nL51gCFun9kB*7`Hw`dt+Vng=Nvf z&T2@}nrGv8IR6aWBpswnT0GBmjAXabbZ$iAHLmt>=^rTi4oJ!h+%AyKg`(5F6g=Uu zQ6OEV?y`~=az#%e0L?%$zn|W^isX!{r&n=JX>gU@P<8P)EfHi1l430IjZ-HySi5J! z{%@J65{s@evn{Htb@n%RP92&#aPaQoH#cwIyn?ZYnots9&xX5TS2$wPn9)`@$2Ulp zwi&9PF4P#_+t*fSiMEYQ7O1g0v|jscGfy*|Y?erh+X#z1G~71d9&JN(4}9tNE+a>E zw89+2AuO25grbHbu*12VG+aFzKEdKD!uGtCDc_yqJpz25SsfQ)J-%+4JHNY$qD2zw}NkqD%eIt2VYU+t~D;8{4Sa$+#n0TwkfXaO>qG4AJPje zR5!KtCGLyOP3D*q^$4ghhjLC#O|%{u`r)hQ#A^NJF@Fu_XrzRa^k^ISHqy^rkf0$1 zVv;dbZD3C0nj680Aflhmhi=B_pqukw87gwa;KoxAB!F=~>g*9fJRc3)n2n^{CUryv zcoPKZjX<4hDC)}n_@1C)6*h)?+xremAK0M>^9Ox0TmIA-XJDSWDl+TbV(XSI5zAU+ zhRiW;!0!5VtC?(rZRD8&H!FPm;AOpyaHv3WrpIC~_o%5zBaF4|`bfj~WbMj0964+c z=ENl%VUkt5<63DM-9&ARweRE%E`n7rdW)q3UXTv|(N?x~LgqN+Qv8kFUSf=r0*{ zco9<6WEm$xI(wwOdgSp*lM1p9oW@M>Vno9{X;5htG-p!LB1@4*8D1bB@A6V#cG)H@ zg!+oHUp96RM|gUcwye*VqKNlr^=tDcTDl@hh|QBqB%;J!{lcnNYo>P|`?F0%RU*w# zHDB}znhK$6B8-d(iW-XTt$SWip-_MNA|wiMH43cmaCb!%H5zHHnb|FE2#PldX_lOw zD{$ih5^s!B!&EgF0cN|lfE3!oT4;C|M3iW#euftHh4N?J?D_ zLPXwfG+hV!J=1kS;F23?u09-o=l}gygLGMcZ+*J#`u_6S%=JrTr{JS-4hp8%l4kL^ zl0Z>8>_c++nvj#QNEg>$|py%=DHcjU@ z+S2am(_t@`!r*C==20G;;qc$)DG2RYIB@t~JKH9Q7>!wms z=pwa(d#Z*Y&qTC-CO+hD#zM15muE&!DC$2Wd_@5kT#_~wr% ze*6?3z1<3-_I!i<_&R*=4}ScJ{P;FJc<;w|fBg8z6HT~-xoY;cdAd9e@tTojeY@L& zh@rM_WW^s9S)GfvI-sdS&?~z|7z(#FtXVu|yF8?sgFcgh;U(GLw|p(*l*!;du?) z!;W~Bqf%oseS`a4K!xb;Q=-`8M8Lej<^_`G%8d$qhW;$nZ%OXs|IPk31S*?(cV}-PvaZ8pz z&9aU}Qc)ujM^s+ZVteO5eN=&O(;L{L@^L%Cv!^;Y3$|Hcdd6--b>iQDHNQh3S~I5e z*ymirl3IPZW#@gGvvLli`EymNNiZQq;8kX?vqEtyk z-x24%-^0Hv&e3aAui_`XS^21S0gsipP@;3{1-y1r`2YrE-_$`zgfqC}<`6YX)D^xp zK`8^)jVs|$eH|!BE8Z5#u?Np+y_FIJ>24wN2s9ZIwjMmEu}!2>rNNp7;)z<&<~;Uk zgQOAxr66m2+B;V+a`IAGI$=?pbfhHYm>^SEGif@tSJd?DQ%N#GNHDs z$z;fP$1<|J6BE1O(VIkA9v*3baO5$nxIJZ(p>$3NoviD!(~e5hN@4UgM83v6^pI>J z1W>)HfA;LNKfM3LnLoYo!*@TN{o(sR?f>EHAI^0C@Wu~U;n2k&F8y#G4qSnsSARG| zwr%X0Z4dDOp3RHB<$ZW?@lQwoG(pzEi<9v3dFROF4_EN$(~G?Itsl<(a2ftxgQw^J zG=W$&{F?;M`yTGHAKrAn$;ChIgLA(A;T(LMi}1F(y2R5hwqbw)NHQEHaAjwHxCmoQ z)Py6o7I5D#aw*Js(xsBECS45TNuLOK4e`D)51CRRo%G*E>F%k_@i;sVKs$Udyh^^A$=*qEWY6uQi<4Zyz3V4?TeZ~v~Z1Tip1@83j z8kjL+8A45>BTCw2MJh2l|0ps^k7SP@glBO7k#&8h$@V-=B<1=TEunW~@g~wr&Q{+; z4HrF7yV^Lfp)Q)(z|=yxKC85ruy2+M`ok;tKRkE;9k`meuiSs<+`ZH1?!EQFil*u# zF(ul(vO~)*n>|8iO1el>SzJ;441$827l?Wys6&3bynp8V$vek?fBTawfi1OMydfH* zhgDD2KS^gpBjuf@ZJ`1M;FcI&R&vj zWIN`ru6p6Y3t&Tbk;xP_U7hZ(&K1pPr@Mwy_Tj`*=m>^bS{NC!OGq7lQrQYds`oPr ze4Gr?49^G0QV;MJ)`WbUI3La8`)wQ$swq>S;^& z9h5sk=}TA4;jtvjvk$9?F+=nrH1h zV2-WJ*i>&%3ke&W8X-pE=>;HEgWi_hK?K0Bu72FjhQ)MzM3kflCLLr_^s)jf zLXRw@F}}K{MK?2SYuw!eJ9nhBPv^=&$1GRu7?XL-Bn3yvI3MFmF=&k$I0L6 z9eAe7f*s@aR*@{DtS8(U5moil7Ru`F7$OKwss?oH;JzY7kr^wWea@gHFH)RY!HalZ zMn;aa=aP}7?H)-b>cK1d-txCFgBq)TcGO{_B^=I~IhqMP(n-c`a0vGnA(}> z4uv|^SWi!<7TU0(b3>o1bw)zbo|-{w*i4_u%y1_-1o7s%HDKi$VA2%T-sF*%SiI`? zbhwrh>1T{QBa9|mZMJgULAMHNxr^Is4auPTL^6aCqn?Z}g{ga}WTbOQJYYJ>`_?nY zkkgwATY63}zz8%qVG=3~XE@Xo>I^61J)J$hu?-MubSFD|<65k6~F3S=e+&%2y=WYMVSi3fq7A zQ+p(M=It3>kPXxzYBK3!u4|7rZkmcN(lrGJODeHYefG`ti>VYTSw1!{^6QCNz9ebJ z;4Md7_e(JjFOs~iTW|?RDV-xlTEHfwJQ;;>j1}Obf;-UtcClK*m}{b{!78FYQ5E4% zO;h05Qnnu{MwrvQ4qH%~O2#ld>eL}p@Xe9XGYgN>6qz!R`xNCjbh1OHAZVcZMp4i5 zPyjtFYi^}7ZJTf%*uNh<>WL^?G@_{Cu%fvMfTA`3I_f*>(V|MoLr}4nPHi@wTFjL& zl(ljzluivJKsn~=v<}qm%MFfz3S?a4DV>pZ8r&h+lv9+kV<_uue{E1tSeCYE1Lz6xVEM@vP@>4 z{)<~`X@03awwPVRA@Et6#>iqPS&F9f^C|Fps6NUK*q@R2Zk+>{9p1Hfc-L}54Ifgl zyr{hC_H8?0I}|dw_xWH%-038jG*;aAJPfN!!lS$#h>Q}pKT8x+PNouWNsA1^2H6r3 zK?7>FT`a3}Bf~mO*o3Rem|&Kf`+%jol4006o1aJiz~zxWuw;^g11pb^@^6XrbUnd! z5Wko60z7HQC+kTJ8N46>cMiweXP-PK=(H_dnn>y1182x)(P;I+^Coic;dle4!!@vs zSGNAn%cpL?@p*OEpSgIhnJnY-z-GdzULQ#=33O5Ah=o{5S_Vw394``h17E07wQcjD zw^|~deC5hV4V#z_OG3VhGD%ob)sPZSC<#}iu&evcw>>l^9;wg#2B}+Z6|#W2t%4wo zq$M{k4k4UX{@984U>POrz{51k-w<73TlhE^gc9(;@PXu6N-}V6Q3~)&3%rq&va0J{ zwaX7|!zhX^C^3{va@Bs@)yA&mqDG81XgBuFK=b2oethr8#Gp)55{rDDq z_hUc)?#Is^3Yt%U{1|?}_2cV5z8lyeTX=zw!p*C@cBPu5sQLC2BgqCNXhpwsYJ-rp zQ1wLy`*!Y?4o^1EP!v3&Ceme>kC)HGa+hIV%4~b~#g45sofAD&a*?Kc-A~|C5G@aL zuzX2}D?sIkQErxhFBbp6z5=z)Zei`6d$4==dVEJn?F>iPV@)eeXbRgP9bI;qS4CZ@jJjwY zHL5Cc4}Nhi&l7C<*v8U%QW-{36_W;_Awl)cYl3LFa4}a zH(Negl_(n!Fo8I0wVccbVM}yViULAm=g|Pw>2tw-l%5=gZzkO(M6K2-sQKi~i6Hha z?$HrkOU~4ZAm9G>o7;!qyY<$~E0&UpAen3^UJo5kQ@=VsI3%!y!MCy~*sJHkx&|iu zmCXndkt&2JZUAzXcy zdLyt|DT%cV=;9gB7f#I_n7sA=Co4F^5Y4AHtzwNJHV1@ETWd?&$i*mJtwSLTj%?uZ z@3RX+FS=DtYcs_6$VZfzCl#|6Q!OE;djI6TSAMgCF;xWFWMh%~GHj5Pb(d*U-NNj% zDRdj@pIe~qw^tV zE0JU?O+#>O3A5E)3GW3ggGb>aRPu@Zi*#xsCE`g!F*8U~5#cB{COJDdNM@!59PH>b zbhvDDuZfD!d$^NP)rEMnF$SXou1Ij?`#5l{F3H%ilpbdnJ*tBe|&`-hA=IxtzN&% zuV3TqiNR7OU6B81GYy&{BK60k7QcSX`s2~Uf3y**Gm?RSh4Lxmk4KI2De&h_7$Tm( z{xw-Qpb!Z6dvcYgW?BuXN79}60@Oy=4XK;=k3U9?W5*xPk z;Jt_a%owNc(rk@iNAt2!*ej&ZNi$yzd`&9&?u~D_2!sQ|B zzWwWLRjk3gD&N2<-_0^yUY$BzIZrBA;OW)Mg?6`o+LNFqceHxlQ?JdDtxrrHUY2>= zRG=6RD+y10u?AWQxRY|3rVOsawiw3mE7X}1LH1oW)^y)^6>r@)wD+$szP@|+Ae=8* z8vZE3%fx~Zuo!(MMyh%U8D1T*BosaNzwaq_clVbTC1Uy;rxk%=zE@=pR6#GL7vmlbwb&d+0 z7!ffY19}vd?(FCo8yiz-Hmi&FQL1Oyh(z&tWkhz+ESDWRd;S;ZpOn=R-br=L zBW`d;D0n<63fywRFq`mtBT+^3#Pf#V>QbIsN^4nT3q(sPYOwact%MVEw0xmZKzCV9 zCS$?Nd)%OsY%NJw|5p=IBN3!5jwjb#%B6CA=zI?;@Dfe26?7;g32-P^4vT^fnww1| z6+q}iuW+)-TLmnhc8e91`k#UghQjg)OQicURtTsPMM4RW8sS2T@V0I$(@Ylztxwnt zvjSFGGL)#HPHZ}ewa>P&FC@&{H{{9U3>B+GwGt_uE`#k6t}-q@?d&UZQ)ICA1>0Fa z{3_CngzsUJLit#hr$qxydRupWcE>$va|xcLqDI5m9jjKe`D0e*_`w zkyt&bZw*e-Je=b1AAY%_Sc+&|NqUe;z61C{r#A@#b~w{A5h3R3A2K`Cgr>ycK1?WD zWUu=PDkiz93+A1ON07q65=5t_Mc@x1n+*Pqt?2?%Gd-d}blCtK-8i}D-2T8ue};`x z*w@^w6Vb+L^XQ^!^Jo&BCEq2e#pKa<@(dI3ZPU3{4 z-q=u!*eQyb_Jl`2XU(Y%2nT?hvt)v`OpA`_%w85X#Ncji6xo5p(Wd#ds$~vS6d2Z^ zxz3Hext`1wO%SYW2t0_)@7eXvVhqPez~?hd2)7~V&oU!jO6vmuS{hbWgDJ}LMW2r* zhcsLQk2c4O!Oggn$x=}Y4*n5}L8e{itj*1^Pgn<-GYY%rTIc05O-^a2cox0g6^c5e z*Z~pI>Z6m#KRXW3WK+TZa%IP(*uW6+^qxHaBfq^UinIwx{4x%_rX-_F5xpLU?`u9s zbCRUENfS*hd*wIxe;-^@Z27|RD7H2vL-qL; z-_-aSDN+U_1#os?JuIci4fcR3e?8Idj3KTc{o_k`%c^Aa=FRKL^-EIeB-@4ts>j;N z#vMb;X0}J;6mF5@EInRA==}AGQef+Cg)K1emYUF#u5Q<)&vw@+HbAJ!l}lW2w*ikXS%gU`BvM}C zr^{KW+Vb@%J*?S1JbPiTB+BfO&du|gVkqX7vSb6xuvvIT-zW}G;5%a5(56txmHlj@ zV;XLSgc55Wr@w}p3XGZx-p&$5(0ACQxe-X?-G=B8CKq-j!Q2x@8p0}s-3k;7FZpVbzn3Zm?o-(G*9cT-|19vhOt2vZig2DvBPjG1Va|=1Lg_= zA?myH45LIg@pL};h7*e`k%*#&LyD@o`q%>#N=K1UI_{A@F-Lt`VB>Q_LI8u#L6%8e zx5A-lIF<+{2px4*#=Ch&(rm}BQEnJ}ouAc8c$YC+{?y1&nc-&McEnRUt0&bCU`bqL z@ei%~i8tTC&w!U&6E#OmD$WPnX_8R~<(wjFc_2gc>BnBzTjftU=@ZKU(l~k)k@; zz#WdIIRj}frvK^_pGFERQBMx!@dq&U>C6UySBf-D6gzm#wtPT_C?<_PVhestrb4d% z9;{g$Ldmr9UUY(Uw6kw4PkRCzVeVZ^03k#Z5O_a%eZ`xYAvuG!Rc~U)U#@#U))@*Z zVNW3Y_>ca8$?S~KS&^sVC!9fpcYiLU5i}(cuP;ch#WDf+lOj)HpNDP2CGbMZW=544 zeg%S=7{axaiKYvY{saZBb~DGN42wi()3C~3oJi%0Pa3kuY*C4PP1g(5{<=*jw ztq|$dL2eGdhe(@yUtYNX_6f4?{=56{A3GhKS?w2jJrP>kd`Tz?51YZ-MiwxIpmSvK z#R3rWGwreAx3OXynHo}}9z;zL!pR;Sh%CheoG=p+BiQYSsRd(1$PC$GaHloXDz(_O z!&d)Y+(_q~t0KEZfd=2Iy2j}^!~o*|DNvb(aCjb}i0aZ!QT+<1#1NSu@rXG-0ZzTn zY$?Y1O1txFM2UM=!5~yXeFn)0JB6gXe9uC(1gm*wY^BoT1|^$d8Ee7^?@~6}rlq#{r0oF@4R()@N5@@aCvl&M zyqLrP9;?qr2ZBw$>qH9}_f!FsWNOlR{t`q$3r9h#%RcoK#>jdK!4Jkd2g$n1F$n0c zRnEY_bCvHLTGRcNOK@QqiF4XTgjojnzkBVJFik35OpUQxAur~$kOw9VE1}q0Z}Af4 z7-WfNFgZ&ostgua%mH`06_!FQ*bC7Ef$K=yx@Ak$_bb{u_U+=_?0Qbv7Rfwi8{|1= zlTn^2pA;*ZD+=uOZ8p+EO1o(Bl-`8$?hZ>51k=rf#k#1CoX0*b6>A`T;=9_qO~?Yu zG#G@*Z241T+-j|qU6(#0;70o;%oztN8_=Z~80Z^ZQnfpV%>1!LeYVawfT1gGA^D~? zP7RBo|C&bC+q7iVLXjE5bdzMLl*qUlCmAH`Ul$7{o_<8dzowV90Ua)gSF2190;SnQ zorByeusB{!aMX!Nf&J%2OMwx~G34NndghGp^n$ZjF`XlF`2n#B;f10UMaln_6hsEu zR3T2s5Z%H@tK3?3bIP!U!3Eh!;q_+Fu9h`l5lD6UguY$Eu-Zt_$*-F}Tq=r=2W5BaC7d#yq+i(Fq!2h_NyB-9)Z~v#QKOqt*nhL&#9a9!%Vig6~3!xS~P4qekLNBI+`T3yS62 z(+-3G5Uuwgp7ZNJfI()CUcLMN(c9m>zkn5<8Sq%)8Cs;P!wR?m@Y>8b$8MjzbnC*` zfz6;HtZFnFK`&*iKh89ebE+oVdBs*sL9Q}uE+JH#qQG<;Vj0}s+ZkDAx}RWv7GM3G%a!q z^+PPuBSdkIAOaaivFUNMF?*%K_p{pQx%M=50U+`sNMyJrk`4U7h4lb;Q8`Y*c^J!6 zG9Yv_@1dICR6-%*qJ>(p3rPeH^m}|l_%Lc@R)tfDV}$G`R9ZQ3p9BhP=?rdg3^y83 zWwmV?Xq#`Wwk`H^bj-|)_5o24(#&(0dZ3y}fa{Lp8%jd0na@@1^)dNaq{P69W1@fkU54r)cVt{mV|hwd?mw8#H^65 zU9B%JeEa}<}Yw}59pYlmDY=TL6_;44Vdq20?I4mDMbD$wN zt0XKZI|j%Gp3Naa3E4`ez{GSnHS^3$QC%m)j28Z7q4~JUJ0EipLKKVKSS3sc4X&=i zObmtUXX3Uwrk;Pg$iW@Ns0whPHz zve8XoUElx$yMv(zqR_7VacCfBWX@-{1HkuvM3ji!tsr&&2Ba9M*DG!A$w@4BpEn z#>yE!iYyP*TA`E`STw0<;U_Qc3m$~YkYQsbw=s0hpFm9iMke%9x(Rs7GLC#iNoe&W zzqU6uzc<~V+<$*!C0X+0*ccFr)~6fRl0{x-d|rg_U#KRE*7XW$q_{MKLz|Z`RTM{_ ze9Z&%d;C|zoh!-|5l3EoRc$)4;54lhY++xvF@H44wvE&QO9tP7HVJBEbzx)+++A%C z%#Z>juR(D1(jaNNi6kn zWN<_AQK8!@t~W6q1|-e!tnK1AZi0L|ju`uRygme3SbTI)6j*(M;HiPZ&0D)j&!#>y zykmHHJLwzRJT%-lG}QOhFf%QI^{%W$npq2;w4N^MP!%17%_~pY)!ynx(xii1>+1uv zgEecgWLSy7!V^){aM-ne#TqYQvBg)ggc6I_rvlP`p^kYnP75LhHhsF9w5kWjL-xJ=)8@ ziw@IY-FR{Oi@zNC>%Jq?XQz)(fAg0k#;6Vu0Uya4B{Bhz2OvE%UlkHg!K!?!y_oQLq$Z@dI=KY-Wr(v1VN zZ-q~U@BcX*!e8*VugTLZ>M8|ZRBr5t?|#O)i0Knmn>j-O7Y4%+_%57pgW7!2u5v7<{^7rI&iiKDM&9HYa*Z6kWwsg!&@AA%e6?oD-?uJr<Ft~B6ipGHG6>b%x7tP~$Ah(M$vPN8ZRX;+;2Z05DLKRuv>Qv*yH0_~Dc-1Ds9dX@MVf}G7v=~mkeGs0Ua!2n zo=i<3TLqcgU%7@qkmlhVJh-oN9)2FFoUfdPSK#=CsTV7!3A_j&cohlo;FNEsj!eDk zd|Kr+JeecGz!xXr=(Wm~sYy70mLKCKhYRD<)MSgL1}Y!J)qUp-%~>3t%IWo_@}a{5 zcp6y%VN5HhDp%mzE;|GE4P0k%Rlu$kFNxATp_TrhW$u($#fy^8K5W zU)-F$a&!Mzca9yo_4fBOM=srd`|{2G7vRXucVFGy|Jz#!FXPX*FV4LF%B|O4zjX{* z1j(J3KA$=GBAj${;_S`IV>c(^=fRs3Z~k;)-`&X*cTP;)I`w;a+s*ygoSz@v+&8i6 z?}eL_ui{Wl9<1FE4-R#@hwAnR!TBJ6ZgZJ0!Ox6g8|h!vjIYIOfv)q|-+B?%8mt$- zy`N{Ft?75qX&#ymWE~N-RZ{tmev$2PP|{s?);2Aoh5q@0K4Q1z7B_RGOHWY)N&J@4 zUwSlieT!sf&ft0Sb7}=o#CGX4KhE7#INW#!hZi+sSV?M2(V6e##yHarn$C?eZnhwu zS~oK-&d4Y#YNGm8U_)%BGB3uLi2iVB^6QgEftVnvl$^vlt93l&b8;k*?^|siBm(BCCi3pTa}Jf1@4xV2E10NG2>P`H_x^B| z{QaFn_m98ux3j@%1-FmutT61BlSnk&E)t6xBBI2i^6_mGTbx=^%F7gw!__v z_hL9UbNaPg?@g@a+)FBIxM>B8mPJMk2)$G^S(;|JU05+}L;~5TRS!_;@gM$e%tlUy zsub@^4u1sJy!s@Aoexl7c?atq@c*|f-(#U;fsJKHmuF+yk(V$e55sY5qQdcpQ`WCln3#c1=+u?rcy{)R0(?!@V?_A3DX=R z92lGB>*Rh!*=X1apEAkz6b$Gng*o1Uf#e?;VH2a790#Y|1*2<^G!AMaiv~4XLWz4| zTno}SwsljPW;(bW9v=hH;=Tix za72lQo8Dap*{^IYS3%5T78%Qwi1Symfq9+NH`g=N`_CI=d*K(^7~8%3X=V2K@UFeX zyWsh%x7=J_ux7o5Glfe|xT%`UkJEfRa^aW~^-pEnHZ1ZbQNJ1-n46N=cU>u* zP$J3t)T!_6!OYR$-QBluA#AXkfVDTARN_cYlh`}|X+1W$?&LdPow)PWC77-=``@2= z>C)}@Umz;E_2seKSKhgE?T!Dra%AS<>DzCg!NG{fl(3RWJUDUY&?`S(*!R%ArXg;_N4Yl5u#2kNy=%hp#qNtp2o~w4 zku6vb<|aDOtygC3k>wKwMqr=6td~y)XLG^}fI&^Wpk;w!ya^=YjYaASYl4*H`&nvv z>+p}{#-C09lAW6s2ad7lJqQ+YfLi%b5O!6 z>72rKa+4Kl6TRlk!ZMRUkZ&~d>~fAJVVCLCkSh zD4*0zEWfw@0C;)%2d9&q7#}YmqnS}k)={!aSUuFTn7(`=*`w((G5+`l(*?QyF|+c) zC({-sgRMSdEa@gHsfnS7eX5}+9Q3hAI{Wco$hK0LMPud)vSq=45hd)^)(N64y9b5J zsZLNl1Q#^ZPMNSoj1c&)gqsd2b~-I&2H_^vlB|=)=C15b* z7{ZQmeII4(NfAP-V3?$5s9{s^)zxSXl(eGmTUA5Eah|^ko!1)_r!3Q7cod4Fj^PE22Q`4W{I7FsD zfiusL8!y6}zM4LX?Yjq^Z}}ZOflqdSnDeC$+<4jfGjI{7@SERwdG7+pTdm`B`Ye8H zxB?i3FJbUshCw+Q_#RteFkJh*x)88&ab#mOp~MnTp0C?$=ct}xE4M==9X@7HhrWMN z!!btlgpw3rK0%I0v&wGNaRg=XyvY_ilBcAfyV^+Fqm6I+=;9+ERzjhAwoW>Jut~`fNW$wRA@e{oK@HQn^4X-(wH0(-wMRw&xZ#)lotH>eS)N zMS@SKCaT@B*r)q?bn(>vy_| zQ7eAl)AQAn0-L?^@&7%Ef9e0eSGfYiJ~auWco;veb{j-%>toda`WRP|u_sR%bQ`nb zy_6X%x;G`2q^HyF@gMsE!O~n8rbW}a6@3udj8(omR+##r_6S4-jKH;xTa4_S6;7=FaR>4?!1VFG z_yaz7+?M4l>3o)%czSE4G~n;sdD2e{1>!sm9x0ebF1f+;F2%%XZQe$G&s$9cLo}_S ziA8YKgA^mAtH60mw5KHQ6dAH@Ll5B=D>$-K*u#{vjf~C++mqI9laInj5|iQUmLMrc zgaMoFoRs8_q4k8C&O`7dj120GGD4ldfQj3*EhT&bc!w$UjH7hL4LDV#^YB4r7u5@} zYb$BmNphA|?Cj_m8yi!`9vtpqMh7o;6kw#K(GCZ>Xs>>HdpKm}+F>9Z&Q&T)?J(Y) znH{kZM0Lq%I3ABh5)E#)w%)$Bx|C_!;@z*^lr2;mg(wc>ea?g?mT#k^85w{_XU>`@emi{O#gC zd~)IR{oft9clhAF!=DB=%-?_W?){7B{yw=c;zEJ8tI5V?g5+Z60jDPtT?%JpmuXTR z_DacN&jZ(Mv;|Y9a1C&&&WePS1`alnqv?SWusAhDEK)N{#HDZ=@hfqf#Lk6;r-Bd! z0@!7n5N_(cHTZaTxh`wgB()kgK62{Md5i24aEau>YowQ%JZp4xNsH&1juPO(iw}KN z>R9l$T-Arbz1T@?ZO}sX>)hj?Sla}y3|~?Cm?U;2sG9FytG1ogfPoxk)p2$MGw@Cz zIv-(<>plXmx4r<1xvj_BX4*pynxU)1tJbMO25dRj2s_)>-P1PTE^TwoB|WHSo8qpX zI1L%)AudKa&bZ4?K|4#64sW@KE_#iuU|wg*i)Usoe2cue@1MK#>Z^BNJv8&~A%`7T zT~T&iFkVs#Cz^&qx9sY+QBqjjRuXTP1h+v6N9%ooAO+JQ9D6g@9(HP@*QWMOy)d7r zc(v!KDsLj0G5q_!a=vnI_8ZStLp`r`R*qKQs=QJ8q@(hYQz_l&yn+q(FICPvmD6u3 z*8&@~H}}W=vcU7y8l^Rh2JR?GIxOuT`TDhW&^_($9DeQYi*J&3w@w|ob?ER)u6vQF zqCJ6XHa)@+MX{oUElH{K@AfZSV@&T*fd4~x8 zd`IP2l{5t}xr&#W-2p0B5ZiId8KToTR0K}_q|$(4v#qDE-fis3v2@NTx!gs>k7A!k zR8eE$B`uTDkP`A}DEJ4?A}saudO&I)%*PG~XZiW`d6ZE3%aI%V=RmL*r;noy&GgB= z?lW$@+~5pw$zqR8aCMp1R2Y$$5lT#^fA3txnJVBEkDi(S+yR^hEC=T2P4P%AW{&s{ znxo*`cu0XB|K)WU!G(ntRsBw43rWNycUaW@`;~(tGtLc4IJthMl;(^SWoe9B#|Ej8wrxNtBnzL>Oz1zY=ppGh_G6}%VY!W zGDCI|Ih#kBk#GQ@0%&gB;X_P!3f)z-}L^ zw}bB~)sLNujUuFBFH82YE%R)3fJ=tUS7aX6f+1ScX>hOt(#s$(O$?$LH@GxeYXnUC zX#}hqAfiM)gmAtKt*~sZc>jqPc^E99tQsKNfXTz~ z3q;sn$Z*NCOd?{C>##-OzMdlkRA%9i%ZJty`;gS5=YM;$E>pgcA!@yd=1yLN#aERwU}_);Y?XC`i6IOWT|E{2MGt**klkw=YU)`$@5DQXNaqSv z2j2rb4^qrDXS*L({iLaD%ojy&1mC}U5izW%V3)96*|d0V)r>&a8mj*!+KN(3*)I^qK^_ah#5_4dy`^J&0@l3Ws4^$KbS z;HPGTW4U(;sb^BsPP@~X(8BGiD+sK$gl67fXl=SIV z*mf4p<@F0}m9?l8yD~hg3tY)yLpJ7YbCaCtc14xhPniqtpyTiYpPleRkeaCuvaanDVm4PB}itimou^f`?@$Bk~CZX z7KuGtZma1=86@T~ko1|&Z1`I-!|0}4Y$B$KL>lN+EZlb2Pi@!?%_?ia{7rv>Y18F} zU(fZmu8ualIgXJG&J$`&l4S;KB!%8K?38TgtNT#1Kn-`VovJ~+so_+CDU(#)Ra93I z)%WTagAaT1kb_-&O0tQ<+VUl3B1t70Z(3IAiRkaS`^lNRpPab;(jV@gyX>?#Mpv%A z(Xm-nZNN4Zh?MjrY^9ibZE9lb2&r5nm7`AU#ns9+_;=oEynB<(^`GG>Z&$um=AU)i z;VrFvSNW!@4SUIH(3^m_>?75eKfuRue9~d{ys(~B-*bNU<)c%R_)J;J(p~U-!94~J zjp?`eDDb^dc^y9DoYU)fkyK7PBXwquBGnZ=HB&7I?1 ztENO0H5OGgHCi)L!c_tHKZT<(g_~Y>!8*6U{4#v&tMmVJ<>1T*AKlq^cILph|8wO~ z>-mPaez*VbzK>=Oy>jRKqd#3ZHFNap?L!BH*NBGUUhQU7E^#-E#*}39ik}!qy)VPY!?~v9snb6O+30cTc ztL5+}!#Po)CUK+y$4s(rD4pXYBPNyb+hTdE34e27pl`5djA(6oMSV^$6=C2Z4>l+W zQ{4t`lwes#mvSqUX-;-h2@4A8Y+v)BWWZ)#(IBx_3cuA2qQ4DlC zN1n5vv(=C*Ia}Mrd`TWxjTS@Q-6i-IunEt6_3f<>Cud$hdh4C5Gp`=L_0}stUHzaH zHnY0W?dW&6&z$|Q@9v)b^3IzRx8C_Wuvu<>_^PXz?#74v^z~GPhsVS9;aZSM)LTB8 z*118sDU22*QQwJW3%E{V3ELttS5Ww?&_vc_r%qj9$96T^I13nE%ou7&i6`qPN06Ls zA2-ICZqRgYjG>qrtZe#Ni&>2q5j$s>xWTQeKZNe^Hqu`+%1~s{>jY6w!|C;aZUyUl^c+)FaH)jU8 z`8Ff}^pNJqazRaNWb?$fxXw**AZ_WIFAZPs9OROZ=(ZWWWwl?7f&omIebji^#aHV* z@oA&D5>9%Gv3?_unx`vgN#$dg=F`!@rs)=Wo~B!B zrfH&&RYlBi+|!%ydjn-t%e4 zF^~yU=U;fAr)8tITK4Xijb(wSCRl2sinB_bY(1#XeT9kHew5joq6!w9-}5 zDcQ!f3IG0k*FhomJC=G`QWzwSxPJy~ zUaT@0Q=-v^cNc79oIQp8J;mJmmi27RX2U1=?5C2E5o%4qA+k*_9qEHP6}o?ww+|DM zS!XGUW8?9@@5I5yfAso%o@P{|*ihQM9v>A;ssk?EkTmr;_x60iMh|yoQu53*-SCOf zjWjiX{cG|oa3q-yhRN(Fl+Z#7{QPv~^CX~9C&Cxi|cZ5UbEmP|OHBoei%(Q{|xeKx!9f!@wNy~ES{ zhq^IafDl+@LRCUh(!SUyw3Ixk*1jNMbiT^pgB*CnOYN%$DNr5e$7FDU5xWbSlLlL67Fl>`H0Cmx6nqs-N;gfK!B3@OAduPL`mmK9ry5d8vKz_5imK@^nRA=9& zW9^ZQW)Oc@b{G|m1mCi#UcwPCbwKkZyh8J+Ncf-zA3g7Bm`-t{sjgEvqD14h!Gwce zT_-FFeev4DyPvLL0`3}enSi@!kv6NTtAHG`wfm+clPk)d3^wz(M%mH5@auCL0Y_ zT~w_t6HH}&B|y)Yza>k&A9K$m?}yzcat3)q=8oEIA4rRGqGg-;x~xi|$uz0iac}jU?bAetP&-;2=#)k+Sp#k_zjn z1fxp8#U`9FpQ=1=?yK=uybJ4#4VgHuLo|0uu@jQ+7uPFl? z1korlwbonpSW(fI^q$k zBb24qEK;n>+$WS+wBf{#9VlRAWN0TzQ-cx2e4OBHl5C5Hi9i@{JZwCUsPVSt)mfmL znPZe}gNc0n7dvVU+h;+Va6^jguGbRwtCu=vm86Dj?{Eoeg z)8apeuxz>OEwdZwA`Q;jNdmjHa)*BayIqE)>^L(T>y1ZXj{dIW}_&b9A{ zsVs^-)zCGF_*Zv&Mj8`hdEbwt><Hu%pyfqUE(*$kueYUs+Ugkpwr}NMk0qqgU5p z?8da65kACiY&}gFSapZT$Hqu$5u@55a4i8}Y$@H=MOpr!e_eKef~rz$R|6~13kFu< z7YrtYsuGIS29&J4k3jh!%Gb)5%h%>+%2(%JZ@svl4SaO2e0lEm^2NECxfhVjo8a3D zbQD5J*?>`Y>>o+Nuh>!WGromZez|<@|9u&;shm>t&j|@K3)SgO-7+iIU{<8Tax^7R zn>I&EQw$44xOa87ZfVf}Hl!hnlnh$P45pDj217tfZ5U+V>?JcLG<>Nh5{Y%uv9bCr zAF)+PK0vThCE9pB(P)lIMkxFjj8^=+%>CA1y#!>Ys|ou z1hJ!*)KRY@QauvsP!BFIGMF-s>VH)V5aCj7&J zPQZ{XSPKMP?ICJEcMWtmah;haBR$L*lC!6r)~2&%&1=tEFc}Jf+0rn>H5TcyN(zeG zCF%>N<-kJ%^)NuwChQ@WL9$vY4ckJYflzT>0_+BsVJN6X zTm(H2zqdWWj9*zyHkBPIXwx3VYjGSrd)oJ5&j^#**^l%ov&Ghdu(;9QX=N?LWbGA_ z9W*9(gHy??iRITBg}b;=-~s1i!lZc&~wcfpA{YlFsq1J zk?5xH!@_(wvCD#q4M(@g1>=@lTOii9TxxAYOs{kiELGzMLW*ib(P2rh!(9tF{I!5* zhb^oss#@!XH{4jdXf=Rd$ue1>y5X_EbYyg-Js54V=Bvt0`(AqPZqi1!wQxvDxUg~Z zUO_-$ZT)=a)xTf-nEd_R2S5M$^VSP^-FxrrpFex{AMbsB@4Yw4&wo6I|9|g_Z(;o7 zwKwlw{Q4h94~HGyjSj$E&*Fe>ELbZJKk@wCpb)YKd^e-nw7%#PBcmc`D=59qESwRR z&Vjul$!-B7urr!@9G`FT?ADklq@*yN%=L|kXH9U%Eo*`ArbUr$V%tFw+YaJ^TI>7a zpZI=S)P#W$Kc!|?C`E!wJkfBi`BsR%BFj=yV8}U5bx>w$OQZE@WH%S(HfNnWP?L7V>`*ro-$BS9P_52Q8@CB zVk=iBIgPp1447PFV`TONqEVX#!v5%2K{XML23jm?)TJ?t2+NeL&;SPcQ@oGy<>~xj zxUM{d;Z+5iP$DjLYd?Er*ZzGZjC-OQRgt`il-T;EMjn)w!%}i$QY~K<$NfXe9ZY5@ z>_|+{_}|THzPuDdcl5*sy*9;#u;ALw)cvR_K4usuY^tFA) z+6Q>tXpRf5L?#IG!Dbm)QTf@c=^QC)C7r=vSUs4XZ>@}CWl+9lGGOV_I=F@)UI=TZ zL$2!FE$7!MoQ03i-TD1*@4S8N?q{#vdG-CK`M)77r4BxVV6oknN%c$<-QN&G^C!Z! z%VguAVn+(NsCHr$Ak$)|ne*1yp4O|@M0@N%2w_QdC{&v{f5bNaexYX?XpE}2=QluP z6|!;U*VLIaz*2V3jyw&U4AZSbN?Pe$Yi;uVFmi)T2%QXzoPRH+r@&$Is*>>C+Z#d% z_=u}?^%2-VK!xh8m=m&zt?e-T^-6fQOs!?6*4bwlzW8S0+4BqUzS%@OD-=`$@ijBN ztzIzW zUnb>~aQoNF7w4XX;e6#=6y^8J7v_G0?V=yg9i2OdoulP*Saa+!Ega!Z)q8joOHN(Q zy1G}~94CmDZ=k1oX%>nep@CS zUsWtoCFV-xJp6vr0JG9y!^o1+K(F)q`AOk~qPk*jo(m7aejB94A_KDx%jq(fls4}* zmFXOx;M82d)W8E$Q{V(-nQ7nW62!{lk>mhP2{2-W?{;0c!nGYT;N%a z!x_@Q*#=IsDf%SV=NJyl&39*B2mzJ@E_xErhTDYTl2GhoCP}gw2*t|B-0X)YqgHF# z4P8uXTyO8<%*Y5i-hfN1Gtz~Wb(&*QIq0!ful(2{tXqw^xDq^vAMWGyE}VGd?#UH^ z`#0CdQ)4{u>m450JKW=22D@pF@kGSQ)D(1PpgqrjNQthgFS!#G7@I&G1Jl?J-X=^Q z{52&{S=b!ZauR{dA#Q1xLPm0SW0P;u^a~`P4@H6w(Zn63*l~|d;~*SxHeG3H&%){B zI0rW1LV~0akwGp|?_kWw4%6Ky_ZjZ(+c_{WEg60bX?ivaZg0KhHblWYK>9}iHkvWP zbce7);Y2l|Q+eZ2Tu~+JN`$W1sMR*Ygr&9BET?-Py!Q8R-zE2c`|iCzzpz3N@84b^_s+iX^PkV$`{eKnm|tzG+O}Lk_a|qm z6*3r7)I?Z`1(b*bki5OVA-$>@jWxqAriHCe34fB^p)(Uo@GF?)>1H7jg+ZuF zz(u$Dh&VWK71g>n16k8hv;dj9C;A1__H z^P5wxXVcw&)iBT@7ztb39eTl8<%$K>OSpR3<)KW zsBM9I_#gj0Lg%!ZRAFI66b{B}XMU>!YMB9b^x2;dAOG>fZ+^V+`kin8)Ih{|eTtyM zcHQhF2(Y(bXhP36vDiF}u&^PQ0qX~0v~?RdPh2U?s&3yg#40DaOU8lZYU}ej4R>jmRegC zK;pSe3%`3~;lh!H-yQyG<`0!5QWKS1D4?j}TIJ?B$BYsai#_@R*%CRzgk#4hZLzgB z0t-#}HwpSAothHMD*U4Ln*pP%bOVAIkP)voLhKo+^1y-eE9Li-BN6X*(&u zXEXgGeZ!gZMFR7D#vb71xfihTt$bnbIo}UW`8D|BxkYqMxaPYRc+9yIHdpd<_6P8j z&S$EAuSs;3UXw7E|HfU#+Yi6pGHe+|(M+ofm6^E#t zE6V$l3Wt=y7R=w8UAZiipgsgEaI8*DORzU{O+xwunMRa|+AwVW5QTYz9jbjOC)Tlw zbc1pcFo=w#?W6G3B9jdih<0XS8V(93vsEcXtLcZR$D((TLQILefW`cTH}*<1Cokz-P&`9F-Y|=j%w0dd`kU0(Ahu{1g)G6B1kF2rO)7PE}gu z%u0jNrd|+B1q1F9g5+*J#dV}rmmMQT&a_Ehse$%cD0=@YRS^Z_VMjTlcOeCpK(tm- zd0rx5@92+#L`abu#L|g9Q83$#{L^L^QjJeTr-W4y6Gj7E*qzAws|(UiO}a#va=*pO z(P8CTBw;1t0$uiu4gd4k&kzp=Y(FIlYdtDHABw(Z`#S01;&|K=Bezj(N% z3+v4c&9s2-A^-KZ6R1=YS%K=y7O_Obg(6j!fvr{KXok&F5(zsJR z{>d#tJaHxAG8{Z(hneh9tF>ms26d1YxK4&Cx6w}gb}pp3b1zb(j@o$pjThAyNd}P4 zq+wcm@fbdcVEJt#vPBxdHi`B~nije#3QQ9^ncya3r#6W+K7u8q;m}&v)(*B9((=qz z*e@p81e3%@A6a~}{h@*&8Ksg=GU-3gP#1&5H>zR;OHe~`7s};^!Nu+2d7)9uAOlur zl4otqsi(OP&Kh3z4&@#%E92t<2k7Y;|>89Bf)fkcM~mlsQvT(Y8O}P)diS)IAHAN6THRb}b^;I)yVY%S{Z0n+?NwXTKB*V!L zYT=MH`ajz?dn#*PfN459D?mJ=LrwVSd0hdeMvWz^@!rZCk1|at(QvKx;kf{@LTz;5 zK=;x{R{m(QM()boZ%O$pQvL>6vl07_%pIHi9lShJK41Q}e8HxhCUY}$N9-C!#DTlw zSh)R6D<*f)3`}l&zcAtY)0ID)e+if0wjut`SE#Vd7v^59J}&OhYG&-^3Mn^yjWncUCf z?=0t4SF4e2)pJ}sgnRNmEFbFZJGyhrIb#fv@@j8{Kp3UB(;JRs$0@Z(I zQGBI=t;qBYGYE$*4E`{a=-3giabvMYa(o=o*t2Zuj0uKFh7_R;4Lh)muO!1C0prE~ zl7aTLM#=u67Bi#=u-m2rzt*04Y%k1?6#g4;hPWZrm}AL0>}bz|4hNK2yw;5N+}ppm z0InI&+HL(N*I=XVLdHRvAwwl_6n21{GtLl{=>Z97Y00fW%A9);WZU@cg^Wqmr~=QO z!7A;y@o2)o4(>+!6{oe;NJS;c4t&#K!*DmljoLMsjg3x zyQmaXlLtG>$FWNV4xhDc8aT1T#)QVsnIm(rca%S|O_Zzf?-I_nPcWLw4^^cduX;9d z&wfo9qpG!Q&2v<>f_}o8v~0Fk4mW~9Rz`+#lK+jguPeja25n;8up?RsGpy%fD~DY> zotwR4PMJ(6ds7El-b4;s(#P!XS#t`ub{Mb$Q(~iU`oywfa*C57uIJZS_7Xun+>ncU zdEA%-Qn1nBf#Kw8AjAPrOuu^gMYmK#qHC7x^w#Y z-OoO1EqCzt=_~DOP=?RPKvp8XM8I7pW*9RI9S5iHqsdj)<_A!rh#-sKcgQCcQtJu`j;|`AsG_=(f1ab_ z@Ov08`$f357RO+EfFkq2;83_dsQRA#(Mc|~Qjs|-RD6TrM6j09O(p4TZE#&<+(ZlE zUpfeWO@qyjOj)FB_XwfZc$R6i-)#Q2fSSyBI;RQ4$a9&_F|w^oQVYGQ2I~HX!C=C6 zjU2V|PT$FY?RhjA2q=Mgt+4#y3jTERh1>7FG(Yp{{LCBkGw=U+_1%R(%&bt3xyuor zSS-ifMT<7lM)S)t=Z}0ifAqEaBcIP7{q6jbi}OdW&mTQ^=j(TVy!0EgO-;n&3m3m` zR}=9B{B59CQ&p9Kl87l$N2_PJ>TQKRdRri|pc0MLDmKr#q9RN_C&q-(Y&6SFy?&kq z^Vwuo3022&G95o%>P>6A3JNh;G9hjG%kbRvohc2_(Ku8CtAIaZPPMU+s2(Y3%1 z42U$z=-j#Ih9dfM}} zYdv8n?9+@c(n2?XIV~HUZ$YU+nA`;DDkT_jpi?@STD%gqL=m1f?&8nMc}*}IQ?pm# zS7uKcoWMUMN443LnwXkB2?z2R)R{de`RvJjo@(&Em^Oq)O<|CObYk|j38#q-LM~b_ zsAHz5rx@v)J*D|CV(m68snO+u9!I^ucbP>%o7Uoqt+=+@2CYJhR!lpmO9eWRu8i5Y zxf`Vhsm8LjUS6*%@#Rwv$*m*dB`}T0vda_ki*? z$pO3=v~s#y($9lBULO(&QW-nV{wZN! zjpKsdd++6+zr5mGu!8yn>+QeEyWtHtOK)Q53E@U8*MSibODEHrNCia(zFSO#~FTZ z)6Ns3C`!>@=vl%GS#+$j6z0xGL1-c8sjo6N|AC7fHzs*HZ*vkkQ|duSPTh0pA{ety#nXTM{d_tWHyhmxJx@Q7WG<3d&j1d{03 zW$BFUvXCyqf3*WGOi~T(^MrE|Ijr)^8*j3U1eB1gz{{84#H8H^qo*0^Dv^w})zsbV{9>&YG~QK^OLH&_#ku#08G+YikoGO~KuuGdWS&!obZG z6D&jQS+eVSm?*+2lQ}`KSo<*(PO(ds7QdvB3iyKn;mgX`T21bM`y)$nJx?Wt{fqjK z*Y_fFrp|f<2+iSfg%foZZ8)*2OruJ`RY|txarsP6Y77~s4YQ;*F)PkxLH)Pm=50Dv%V-GuZ%{zywc&uA_!CE!jZHYz!L%g2AoCT#!=Gl5%Bc~M<@E}3#1K3?=)yzyWeHYT>CxV+ zV89@qD;vjFcM8f5(+~gN3SRUm)$&#G;Jf6AefPeFbMw-zV>ds!_3X`yw~pOWq z>%`5ox1M*blW?^W+J5a#R4Nkg)^0_{ghx@$elx+OK~3c{L5qnr$y$tIpu6;^N;!(v#Za;4VTDZti$fVsUTbFnOgpMqz)iCMiD#% z@kR&tjS7P~!k({A@zzi{G)1l#*rwp5rs2?IDssFoFyn}r5SYIl@)A!n^E;TywJ^GMLe zEVN+|E4zfjiaD|av;pFHJ#D4a-~sBq#UNK!8isG7t_K!&EE!alaH6&a#B+-;XvN!a ze7x}dbFEF)B0EvT{04>cC6 zI$<%-PFOHnhaGGyqTz4J?f`$93(YcMm)FcjLR9M%(Fn$sK%`b_Zb`FIX%?*%la|(q zVY5E+3Ms8$#IAzSVTUqEo#bjT~;L8?)%r*kYNX{o(Sr16+<{fJYR23tgP+LwL{ z;bU0B8;Ld?rn1*mA3tQ55(osH;F z(`+#4jM>Cm#2BKpt1d^(Pt0()4aWu}u^S69BGp$xW4A%@S}5r8z~SYpv3bvtd906K zQ!ncha*7&2M3{q?VUo$kL-CYl5>1G_K{9;YE-W&)AYzt5#5jO6Lmr_KVs>G5<2{is z+8PYIlWwYm)k<|@!Yxd2#2FxP=bT}=BFNEj@YC!cO7?fJKPAn^h_O2-iZDsZ%830d z#t~3KodqnJ6G-4x;U^VUiAJhrzY&jBzcCx0U9D{+@}nbR8VJ?1tN`ksp_aUl8^I)K zO%1$Cf+d@Irmd$b|A5ryD&V!t8b_ zf{OiLNKpgon({1>fRb<(Wcb<4Y@*lQg4@-hY0!!#Vc||@U}ZoC6z9JaIw3gr zmFFrH+D+zfYeWx9gl^EN-LWw$kng4o7lgi@bnV|q!v13xOlBx(8K!5P6``TJaty61 zN9=lSSdN_)U}$y-BTkYn7Bsh*c>1Yjvi5y;jjewjH(AJWNtu~*#*e~|0Fvj5wMaw> zxk_{WM36?XQ5V-yXWHZw+?+i{s7)aR1LT+PxTs4D2a#?3xL84RmBsGRST!>*B-?aS0b_6)_l+65!U!-Hw9|SE$Ec63Sz>@ zH&Dfyr3hsLcArTltCS5ky4A#!0aXb%UVNtWx(cansNeOtF9#59{&wG88EymngQ^C> z`H3hA*yv-9N8dMoJs*RVh&DrA*X&iud(V|bx_70HatXXaicHof%~yOw8j?^$br8h!;v zdwuRO9GIE&A1oSSGRs1$10gyT2v*lk<&8(-!REHri?!C27H1O-u|GK_2bo@=O@pb6 zW5s90RR{m@tHOSf$s#rLYo@)CN}lP>KC`>C&CwtdOVJT%xH-Qd~HF_v7me$3FY% z&u{sj{eCd|OwbK^N1CqzUS62~m=ccFhAqZ@O#jUa4@nEwPt z4~L(DZ#x4N_2Uz(PI|D8@52l~7TkdYPZ}2DEhPISGtvSMcJ%6q3vEmrTsP?9H?d<%NsyYVI}ei0tokTs%fH;zP9zV%R*6@ z`bRH5&|6^A6gq1J9bz3PKcNJ@^Fq~2o>~Pk{7M)v%W$J;nKqr|uBjcky%&Vs{8{Z) zLF9TC#{_BynN_gnOcy3kvy`Y%2>GnopW5M6R1U7)pvQ!29P1$iu1c(pi;s4O$+iLO z5G$mFC9{vW*j`h;d%XE3MMT3PCF&|3@thL3s}V?H_EU+J5O91e9?^HjTNeRv-SIXoPE} zR1@in;)r7!3O8I$wVz78QuaGiWNENyY;xx%dydN7Q?c^8g z1KOvSM6@nhc498W(Yb#pSu-TxqU_cz)N5pqb9)K}s@3|HH{PTffek8H8$;O01YS-j zLtxi(xI~L>Ps{GTNxKuM(m>jZGS@7kG9tgcl{7;BZ4)X*pT=FgA@l>B3>a=A@8$|2fhXXu};T1Fsna z_M8f?5_3D&(9ts ziF@z7+RWB z$0C8C8j1$PiN}?L#a07*t{t)e7@{yHSRmqP4v1F$w5UffErDg00T=f6ma_F($^AmlFgXsF<>RdN`mKdNfkk_|N!Y|E9#YV^vz`@Ar5HV|6z-IKHYp zqHA%-x3uJ9Ll)ywAngY?V1VGS3(oXL3ZvIcT_%)(3-D@7s%wxchzHD>rU^8tD$zu3 z3}E9z&V93|B%5lnSaap3*CEL4$P#$AK7(N;dZln;NrI93Bvz4CSgb&X2_+V&b-p~edjE9z z?K|&({C}^!{nN?wKfU+ifdjWszqD}i+E35DuyF0$+pk?18(X;km!A&5zi{G%pWgcD z_KPnqy!5+GZ^N&etnQ2i)d=`Y{++B{_!Eg(AP{u09jFl}pk~A?q-Fvuq9egtP3U{= zZLcgmSU(-3*zqm48IL0&1F2=>0RRKrs(;qffeFc|i=?z87y(18QxSIX}ve>ncqH2gOeogwAx*kMSvmA_s7`Qu5N zGEK@K%pIMZfnPdP{+2BL?+LQ}LSa1dk9Ioe-11YOAj6iCBmKgXB>cnSKmYY|orkAw z_{p}Kho@~A<2pDW-b=vNt(K)GPqlogz8{`YOac2p7=qR8&6n&hd24Y&s-$16!(&EpEC&3k93^s2A%5v#fz|n9T4ZH*;VkIW43HFdjf6 zF(nqQm585~u-mU*`|0q^!r8;ECy@CgpUofre17Kr`I(dRGhfXgd2W8@`25W8=8yd0 z_8Twy7R$o#-dOlOTG9 z8IO&PJ)T5*c0wx@^u%WRQ^=FL#tegS^D%?qt-_DNO)?POByu4V1+QF4j!uHBrm^YK z{cxcc(^rjSZ;eV(3q6KkJ!iOm%vKaSd$nNzeo^GDdPtE28s~<6z>kLw55L8bR50B$3_- zY5jQdz1wHsU-;e3kC$Fpc<*`t_F84k?I~w`F#CP6h4hnECC|MJGw8)t7^1T_wSuHLwQ zH4*`6!`-y!LxNo^niis_$N!okKCOH_S!`?beM zCZWXDN3Y_1kV#JUOcNQZ6WFVvhSvh?-fQM`0UjbV7{-~&0E|edxk5`gzpJ3Q4~s*H z1J_|=`8LoWz4-k#`Bsr&9nK$adK6B#WP(hH2Zk&CfE<$^0Z(>Tkf*7Zu^PavL&P=5 zqEH>3fR#5M=RT}Ng0%@s&vOq8qvw9<6eC>cQ=FQVaE+)zVldNWg%nu)c0pAIz&(wP zC#~KtJw&s}O`oBT5mlBD!ZeU>o=)c)Yp{`s#BEB)f(pePaN#lUf{iMHkmsm>Tfdbt zgg(Jxg4&ixF2-?KId0T1cx!vo!9l%vttWMkUg?O5^>x020F8<-g7`42^gM|a**3aw zkw&FX9dX;>&bH-JX&dA)=&T*YxbOTUe^NwIL$T#exK;Uhmrh&>CTd-`EmPkqwA=50 z`Qz1d3!h(amXWUpu;(Tb+d?K><4(uG_!B8^Ards0fmtrGEY5!^!g{#bDAbSwJLHy7 z*Gs6m5=@OK=1Q8@q{?s>cdw z+B-blyJu)%=SasWg=_lnv0i=^N1W1c@ej%-ri5Kc*M6L@ksV}w_KL)9Oz7=MZ$ECs z?r_~QNas_qt~%RyZF-5_Hy`U_RMz`6IZ|X&7Uk@fam(Jl{mHQXU_RB}xMzlviz_i# z*2}Zxu;>gansmZscJ?EEsvdUBCNB8cC2-JW=RSpO8)WIc#`MPEbUKHIdWW6qB1f{w zt0)jIKG-me1LiBT%(f3FS;_$WoWLC*!$DoLpWSYSUF9Nkh$OETPz$eUx$w`m%qhIobpAio`K5@;tilRmfS%zS!T9Z z=yjEY#`+WOZVW!g2lnW$z*rH|VuqAoK~xL-?S;7+-;co4IdRBQx!|8>7oLD%KnaCv zp8!8(hC`C;XACAsVorm3-114Ys!I4v@Vc&rYw_{Qu*p(Ma?fAlIS1{iokX~DoX zY=DJu{q$2P2zFa-6-9BSnkNJ-HR$9r2lwd$w|U!ySxQg+=d)=_igJo&43okV0hMK# zlavez&B${yN%r@Q^bwI_X`M4-QZjDo`Sd{sruyHG=$!P8ju1F|bd(%qhDp*>e>)=d zJZG3^(AhyU&cNEG>4O};`?q88Tl;&u>-p6~11`TBZsEZI2m4=lHi;zZ!Q8UV;7t1L zVbw?PXd8UmmB^%qn1ER)`vjkWlZ+fMI-Kg5UDX>5M-&GUY$H+LP*{n(FtE1p>2PS+ z@-sIB2KOaWg{Yf*cG(cI9{@a(H+(Pe2ZmIKGDX+A;h-aCRa(ywU&4t5Tf&kO-G3N6 zs1Xy%Z@CWI$uJ8P;cNYO;s!ed!DM%=Cm9QNCgZ(8DGV7tTC)=e5u7eEP-0i_b57@%Ei-r|zEo)W7&ttN>9#TPo3rLrhgEc@dsocWLva&1$Oz|8Zs6gehEGwnw-GF&zDTQHf8!({&B=(~$B zneo`q&mL)0X1Y!dZcGm|+g)K&3b~#g zeh+4CbuSLr5s_Vjy;?p2q&sKHbj~pe{v%Z|OUr~FshSWG&o)E^YzSA$)5E{shQJul zYNcY1HIZ^0r$3~q;f5iNMkUn^4ZL_*a|L+NB&Qk~m!e?TX!J>@!z1RbLEqdi)aWt{ z99tFiQ_E!S`#7ir|7ncw$wCh7L}ogODgC47qm&{_#1*nyF^t=`-;z^oc}>O$HK$V4 z+D@1Ws~?0j%8SKzy9sC{os*2G##v4S1(+A$E@nBCrb0%zu8+fKjCAVbvsZG`(zC`C z(sW`qlx~hQS)a+Or`P3D?O=%ARoWB|RWF6d@=0!F>xi^%WKV7EX@njU_MI<9o5C2s zRn(xWB%C!jYJJX8b!E~~6fM{rE;glyU;pi4qR36q5=}KY5=tUkyR7__+m2A(q@)kq z=g7YiuoIA!V|@1X`l-4NiS6y@S_YOr)(kX|IqYKkK{`P}c{up2eMb}2WI$D-u59|F zuo5KPa%R0!A|XegmB+G76`Kj-E>OIOf9lZ)D@ri)*DGJ%*yKl*fZOd^Nei`g7VJl5 zd6u?9B~92Nu)%b(Lg7eTue-}_q*$4e=5NZoXc|0F4Xa0XqqUVOk-gx8)isaXx@-67 zPSP!;HG7t}IpUllB^rR;Rf#$Q2xAWX!I;Myauuup6Rz%FUx{?xG)$JOq|Z0KC0xVz z1@>Z>5P3`}j0j8fud4QlRH`(*w(TLIL=%smj_S2cOX}!ZHO){`k+CA|ntNXvG)rOy z2{C?U(M;g>QNHAO#a4cH2=$E6f9>Ao$F3M=<5-%SUa8x;ZPeWZIBLl}+Z=(&(b+39 zPvOy`rA83e)mT&shQKbyPOyYS;f3l)B15Z0B2_GyPlReqZv0^>r)EF0GDzGYiEh{S z!HkHh3gCYwYCOPlu%=R6%U~>-!V^ff6*>hs^K$AJ=^ScF>_?KkfyT{#re)bM$0Eq8 z+#G;x9`V>vY+!{%x5^RL(F%F%cYnWdrm40^Fs`Vse9yygdkhJ!WYVTqA*jhvK#97N z0#@F;8njWdugtw}lL;K3dwuQ=#G$U#121;^_-j)l_&3o< zb@9tzt|<@{3MzqM?LOtXL>mS4zAXU61pfwva;%~X&T-m?!>&=7K z%QdQmTp?D^-IPBZ{sU}E-PD{tl@+ihCx^haH6<7$7Ww7Aos1ZK_S3Y+#>U8hbLz5l zr^#Q>)<0RFJ;ji4Aw#(_HrD<>{@au4+{rZg*ZLR7nj=ZMiLGxnztmc(#a%QlQe;0O z@)e9J@VT^c|2=VoOmm_ow5#>IwbfPHwN_11P*@C_7gH%p%o$R0`Mz$G3ZvZlBfC)0lWZIGrr~$uBStx^$5^>SL`vDsG z;qae{!O}v{bYz%y2sWVg#e9B1dx>2-ID5*CUuz#r?v)hw(GCL?0kB~gyJrI53dYwq9_qiT@u3ptUy!0w3Fs71a?0vXV2*wOOn~sWXJ3U(*m;~ z&r|-*f?lpMrjMXU=h$sRiRuEWyz#h1qHAfJsZF9Fq9RENk#DihXBVS-dlxNMT;?7r zEM{&e-J$^RGLOO1sRuOh->sV**RpX(@mSDNy&v>izOQ19cgQ87Eu3_x_!f?TzHs_a z3(ubV>4Rted#URLm9An%csN!Y{Pk>TOEy8C_#XsM9)D20LJxBNkP&KSWCM75okqSi znWLK1c2mPEv_phf$wCIP6ge1Ki>otI(velfAgIiAVND>iDmp)lSR_ID7mvx91_v1d&9jhga&E4Z_r`o!!wGPMfgH zRKQSVW+Po7Pl;tVhj0EJt zWQ%L2&`F9KEJJ=c{>uGB?Uh3h2}JM!E5IoKnG8`Tg@1AO&ZTTtJ^PEhvN&P?HGUb$RcZnJ> zl{zH|1(=!Q^lwM%wTfVTHDC71d_X1;MJ;_lZBMqrG3~M0 zQzkXH6Ii-`{erGx9C8{~(>uM*HqMPW(I>l65bqz8qCk+2N@2>d)IF^o;W z;A^+m3>Ho>7=CY|h`li>$xWDdbWD6N-R>NP+mkkf=Y6Zl0kInL?2H5hN+MhvH|80y z#U#gH8iCn{EfF5tm5Aqqoq^?kcqEVv_4US*eVyIC$*w@4I~hpE1A)M_my>Eq z>OEaJ1*%cfv9ZTVhNmcg4{kNj>_?Lmc&unX$0ZRO%aO{(lRL9HF^&H;cUSjN7Pu;H z<#Slw)?IlJm z_K|~8CE#Mj_7i{~6-8pCg~VN@1>EH20>SC{;qV{SC>ey0GiVFxq>oF(Bz;1dDXM8? zfYEFN+rglbiiIKp2S{ZsQuUGT*B&JlQ{tgoAM4?-r|!;aLW;Dcg_QrvU$2VlAXc-7 zU0ptDuL(nNC)#PdhT*|MF@a6QU~Sk%ljvpGRUjp>q+t`Kbq?3(``ygzslSw!;5|(> zz@`oMCS4Yf@Lj=T3oGGLW)J^T8Ds`ru8?Y8=Rq(TQk9r1YT$XD*wX=iHQbd+s|bkT zrTqHW1hWK8*EBl&3AiPN6s(kV<@4kcLSphsJ+F&N9ab4!KqGr)ItNZW+L$dB>|1FV z4OC4?wp^%<7zd$novg@;!Gv0&<~KiRUQHpZ80VC9nkF+aUeQ2j+{jdVilyWfGp0(q zk>`%=dv#F-y)UY$5)RZlf_{v%UK>*fo+K59Mvk88+#%@%t9-V8Dc=pT zqzA!+JVKL4_h2<-W54z&qnHwL!A~9Zs`ys&&s8;{fN8FV z<4Pzzz3g4B$G!a2AFsT9_taT*t>36q(|KY;S z`S~N~=8rx#x9Z5Nf)CQQWI_`s$r6S z+_3NU{$c-{AlR6ePRY(f5y^}sKSTyCJ>#%h8c08dqe?8)a5$+F8tfgJ{^9sbV`JTt zQKVMJNvJ5f$9j5uhn^fDyE=F7dF%nn&{O@b2Nw-DXoq2sG3*}3ik5beHUL|)?5j&_ z$+4Kcucm|&57%lm&o$|J+F}IqU2xwTE0H)E;HLDi#QIAaCo-(>_NQPn5>x__hWGp1 z-(FsL{)1tCLEW~V+5lax2ecqId56d6iv z8fHO%;+bbgcJ}n{=^Wexp4T+!V7SRA4jdSV7s*kYw+?dhl+am0m_j*N%uWXbf#M-V zOBjdWHtSqseD*>n1IPN{ufa4ULrppN7R2m{x<+q5G#ACe6SP_LoFfj46up z*UOvS5US}zT0Td22x>3j_#ix z-T&ZXo$KjQ+9BWNZe=A3+=umkF&uFe!O&I;fuHDXTF#}@|XFeSLToYru6~i2B)$)b%`Ax6sYCwA5DPJJvPv?%7&y+9v zmcbx5O%{wTQ*SG@$HiA^j|*X6N~H02<3}KF6IVB1+$I=NB9YoW*%rosgDdgpGq3&p z@=O1C^BCE-IQo0XK7dg*lW!}2Y(bc!SLqW1VbSu?h06V?xfHuA~Eo^IVs2#C4)P6K23%? z_iTD8)kZI*&i!Zn28)iGw@5rzoqc@marz@lI8eJkZ53TRrE}-ickh0B`0l&!-u>*4 z{v#OcRB;``mKV2)?>T~DTM`EjJZZvPmh6+vNDDaD0ea3vI%bk#-0pXob{%Q4PO%{> zG-2;VzJ+yvS{znius%#ca21$<5VBGvT*bZ*e-yS`6kytIoW>@GVZ#9>mS`9$yk#eU zH_upEtC{ida8QZ3fIS&W(?5oc`{)8&_|9iw@0=OE=EmxQGW|yYbGA_iy~^pQ``c z?5@s@PrjSE@x^yXzdLs0+wYE4<+c2HvV!tSx{-CZXhZEo<^4&4~1LvN-@#T%{4P;5Gzv~R=#_Pv(cI3)!!f!3ki{4jWRT2)3S5qd|CR{nJK!;LcOf4du@ zg-MD?jSdo~1`;9Rm=cN9h9b7i_hm`>s~;>Je);x^m*L+ZFTW1|HqnR=Cy?5AEiHRs zqB2>*H-&XJVp7vGP)vyCih(4nh&3&BB-GaVWLu4d+MJ$$^AQMpgp@oh$cV{kfecfw zP12Gb85;yaSOwp5Qx!FsP!drk7ED}-TiQ07iBUPuD7T7AD z=Kk9{YD@H-JdK{`Kc>V~7k`;o?X{+6~x$4g9R#MlGjlQ^}bH?f-%QJs=};-Pne@q?Khd!;O3`14Qj7 z$#_6fqqu3sl}OOPNonI;+tjQEb3}%N4m6ISS2&IkHju>w4U^v4?F0SY)BWAp!_%vq z<4o#^@V0vq%8ey^n4T|Cjqn25W7{(}VszHFt8Fm%3dca23~3ZujJAPX$Wub*A8)m; zo&lS^COtYRBj z>4LY_#E+gJ`~LH&PSCF-+W$Udk%|Ahd^Syt*{?x}|8?mjo&4uZ|8@B-qRoClR17NN z3BNV#_u{neb5vR^{^~x?u$lh>1eARHz5{UQ>6~2w=NOt>a&6IY+j7aZjlu-y$WTsz zuS53O)3_8WZ`S9MZ-r-RKes-Ez_Oj&#t3WF3T{`84 zVRHiDH#COe3I-m|wgk<1l>m`gCI|6?&@D&0rT*v19!v7^254Wa-@6}1489&ww)$zV`XT{emCqwr&IvA|@WY5ZWG5jsA}rr@r~1hq$*q1jg0oL1#3>R}`= z1!;3wQ6=HxqTA}}@tK&k zN|eAAOxtGV&Fba}V~3Wzu6o?7G<$V}7YefrozqO$?u%T&SybJ%0C+oD8gv%%(x7`O0SuOpGFE1p`khfuTVrDNMAfkiBw((?fVz7bX$9Rw5}1 z-c1Hm=G4BDSt41BO870twv&`)5)h6cSVFI}QaherJ)s3srFH7*2dT^#gn;>Opq$0= z)ukQKK`(B<_W10Cl(?%P!tx3zRK8w5KX^CxqPPlwT)$SjT|_TWIFa*{>p;$ zxPc|W<4mGCgGEtT4re4gI-HNFB4oB>vV-dxc1S7aivMnMQ)5p!uJOO_U*n&Jmf@Kv zz-SJrFrLt6cRFf-b3Q4bnVXqAQoaWNT`iv{<*&*Y;J@>W(|2|5Mc)s^UX2~1WSCFs zrF`{~h({cJAeA>BABspFlDRE>md5c}?tSpuy|>S7B7BxGHqazoF!_GaEAKvk_Qwks z7QXnV6+D#X5}7}8dj9BfyFlgG!bhLndF$GpPhLhEytghdeDL*;moDA;%_-l)dFsjC zJNFKbIQMb~U~gC7Yf@y6-5N3p$(Q20+x5K+Vc`~7SpqRYb$t&$8a!Rz(ocU)a+ zYjgcq3PpEHvw|5jF68|(%8tF7|NOxqHFL{phSh*0KWe}6D7sLc(6r~x6xLMv)$+CS zrMaW!E4UTF{&1#zdF};x^*Wh5Hut*y=1TcOE3CNk+cv=61>EXjKwrV8_nc!Btn+Dd z7nNda@?b~#IBbJB3uoal?vsZ*$|rF$;M^l~uXn%>idF~gopA6{`HXLgFaA(f;<2jj z6!UC4MIuVb)!gB^H43Et%rn-sni!>0GQ-qGAsYGhuL-Q%7%5tM^C<#7n6>*j;y6`O z_>P+2Wrvf4OR39_VQ>@}v3nebD6U^#1OwiGf_C|&ISd{L;fHs9TMwV!PxNj-t@tmb zQ0c?}cTCZ)UmW3qm1-{w)#1D3eyU}J949PerR~JBe|HuP(Uf%7nmBqeRIfc45v;Qh zhH8bfRj0j-`s~!_x8M8t?#I`Ey!7_%GcVtM_4?iGFD*R(>B6&b`xo0f)og>{f=WVd zUccxLiya-+(C zFi1Boa2&QwK zc%|izAiFp_WZHNX7>hHk@p0IZ8PdsOz%89)X*-m+lNgLuf_(*Hn50W`lN&kGNu;0| zgXEAgo-Uf>+sPzekQ$_G@niHw4qEU%!zvW0i=9rLeck=j z_|MoF2`Gup-|^Ig=q8S;CVZRhimLI1{~{Tx=b1E8PIl2i)Y0YOF&_fYOF{_**8=)J z$kS3(Zj`U59l9Z5`9WqZ-a8|M__ZR^J)>UDFfuaxCqag&rTOnj?vr#1AJ(<4Wfe>t z7liwYYq@Qtl+jtKtwug*K9>ki!3qrun;!GRuG*s->0t$tmSAnB7YDWN{(aU7khSd@ z+26KYL~Ym>IwCAE@4$Yw750*wfOQ%%RSCnNhy!lM5ie6c8#ZcCQC+<64}XfWhssF~ z!)?G9Sg$N&1etc+4Apd{nwPeI%ExK#in;m8tz$RO-8ymelbfI3I)X0R&ChN<3;*0a zfAjRM=Wm`SH@~>~?ae=tThGFuPj6nl_5Ak5Q$D_V9=`hto&p!SbsWw=QTgoVXZYzi zaCqkCDRT2{d?j)mxc62V5(2ry;4GJ8b^4#QA_bfBx1C!D+d5;@0zZ zR$+fK+BeBE3GUN04bcavCrAVZ~r!qoc-{` z3pWm*zH#`|8%I98armp5E6?9JcnHog;hg!)jU%U=Gatd3?%L^`K=voj{bSYYjUyj* z;QAE|jSL&yjm+)DXOPp9qIfYF4`6GOo#NZ)%Jnm%wld90TuFqWIfOWvic(@7sC%cZ zT;QMZe9_1kHCnEjMA$h2{*U0)3h$S%2+%2x?M`Ib8yCBG;gwOl%nlH$RT$2L*=5lc z&)u&;Y^4)pMoCC1Uubh(H^Yf?58q$n@R;g&$?A3DjAKiTq>oRa5VU3pyZ4bw{nP>2 z_%a~I#~#~Ft2yHUnQHoWza~Q}XrN{24)sft^t+zB6s1|Qe)Di#bM<^{zR!5(FZ_CdX)enZt%=55b!W{t z!JeY1EG5$&)6&eII6M27w{E`i>1v80ybjNtet+h++T(xEKptWTnhnOTM9$3DAIyCcR??1RmSwKnCPv!Ve870| zoM&yF{FYk1P2#$c@#1(c7H!RXXcpzT1K*Bx{JdivQlr)+X9BDU)2wj#qpOWt=)lfr zwgrzpBmIo#i&gxdT8?U3?=W-7>EIj^nB(MySB&BWcg2g-EnHDLAgU;i5Z$;h`E&nA z9(OVeypRsi(JL2sjA5Tb@64~`U^~Xap1b{*WB+*d+0YVQIovDAG$qA-eV+nr6ZS&} zU#UoP^^+@_7`tnE?&9Ary<-r~{@bNb+zmI}hcg3zyY#+J%nr~7n9n70!Y!2cGFeU} zaoB}?sg)%|ygg$^sb+Y?``JB1LnHrl<@n8~zMB2?>DlMroc;6Vo6mppKUbbwB0;ue z#21F&K`T^W6H)jZhtD|?>#H{o|Fuo{U2fGL&(Fzs&y{N5J?9F*%PC2Yq@`#|QWRJ2 z%SD?Rlr=MsxYlI!!OMLO^O9Ia`j|LE6C+B;wcak)X<8^~MsJmBDjRp&kW1rYGra9p zS#R^eW{u4;Oj~%c%3;pnnAuwq+z{MiW|x5gFt%yy)VAtph`Pa<0F25v`Enki8e##O zE!jg0c$ZP;mdn}f!0?Wzq_l6z3oO~)hFxVE@}6OpUWPpOSpBISwZVp+$~(hG&Sf${ zO$)`?%dFmAtbSg9{BgK)yFPAUglS;@!VN<(fdzPlK_{b3JpcWgY_<7*1HZ6-2p$WH z*TJ+3hUI%@vvnPwPz9|pYA^l*Jdx9=Y3P!4w~7w#QjEeg_Rx7O@tqRZP zn;;KNA`_9NgwHz;Q?4^}`oygxA7GyI$a&aRFwHjm>btXVemwiy$(t`p0eEi0t^UVt+__iB|F5Wow>B4v1IQ-&I-+Vpu^vN}E1$lSZ3rNR% z+U*A<>}FPx361Kic^B-VPBvmj z9x)2->Z^Np_j`?@-G;D=q+ie+buy`Ql1jxShdWA=V@Ve%A#A(l1fnrxzMeS2qqSdN z29?>ncjnk9H(&m2=G?hke|rDcp_j(Gf%|aTimaJq=V#A&9l$HF#Ghop;al% zZIHXmW^$D@%Rx(;69|4xE8d`1;6P%tb{V<&cMbLJ9vYq6Gcuqtsxz@~AF0YxG(mbn z#||)cfy1&j1p|ngB@6H26B|d_T-CBk%3y6ErWn2;%O)LYWz{w;wsIiXd(zYhq0I$a z(0CIH88RSdS&RG3c7QcWG}(i_8WKyM#?^Y&2;g|1>8G|`0Xb zKvgmXpXTGW@GGJ)n7N}Rjbv*-O9tN0l!%l}bxfl)@6!LqrWz+DLx_7-+P_kgsGq<@ zn_wKCDJK=T#jgJub8uyE3NEr)mcz@)*l$?u4s8x4D#hYng}HaoyMD?YikBRPw{TsT zVeh=RA3oKv0?jZ%h_I8zyg-{!s~*XL^=MjDIK&e*!~R7#q=A;T2nU%CHis}e;1*u= z(FM27#i1Duh`j;PB&A5iiyzm68X52r(%bv*q-d`k4f;LVgmxGzXJi+hpv;;K?@m>@ z#`HUhMJEu#7)^G(`8Xu?utZbK7Tur;o+>mu53^qd%X=FYD${dJgeN!eQ}^4kfb7|e9PFnB*RH^gy$J0Is+G$TVNBe60vLo0ih`$ zNBX)nLjvqXtpamx5+kP>dV!A{owUMPDT$_}NHlQywO6A5I>%~6tQLR7noehK0sFvX z_$?gD4CRcbDLVj)CaAEaiXMLZ%TkS2smn4~S|~>)Aj%Qsl9m#_bu$!q+MnmRNH0&H zXj~$VuN!Blk2Eedt|GtY^f5Ah6b>9)cxZ`Ht7sRY(9mFBnxpIYl%oTtKumx|mh zhZe;SGo`rXkZv!B^zw3uO0th4be#}(?GwVv5f}6%5*T_Ck4p(3GHA(+cF%5}Z#obz ziEjfBNjT6gQlTcH`Mk}z<}j!^xi(U!+$4Egt~hTuIae{AU~h#o)m5TV-0W04-|P%& zJD^UL;qd};u{ocI4(X&|ah(~8;?=VFgyRw@vDwU8DNyJ8^U6xF5w5{GpmwgQ2D|{4Rl_i0ESq{#Z zI=Sdc7!k&IZ{%vkAQdsm_QQ`34+T?e`TaZtBAsAlvhdwu!*otydOfq#&`Y*3$p|-0 zrkA)gClh>fKS$#`(Kcb!GvWvTjTi|0TP0(4S%leglhQ`?y9T|Y|L#!zy!zcG@`%B! zaE)2ym6rIEDJhkbl(Zzft16@M*23?+4zdy?k-#?LYn8li%Z5|*G2;Nw!<7ht(P6X> z8w(W`JN$0GA~tOjA-ajUH%f8ulE7okA`gv^BjZJtby)`Zq2_HP38O|}AghuA1_{Qc zWc&eahuv%+tMU$XXQc|7lE{52$2$(bV!gT3G!;Hb(9|jM)n~~MuLw9RKOB5XPLSQK zJPAS}{dB)>l+5}mB-`CM0q&Q+0eN+G6)`a3YZ7)ERg!k?VZF)r&P#F`=Ou|C+oZ4h zd3_5D>3!zGX*Nts@`h%Yb6|9?QamP!ESgatL0{lJe?P+xxE3}v-@DgoVCMAn9?x>v zqNZH0@71<+w`@#UVd`J#8gDW}lwhW;Cj5-)n{@`8^bm(Z)K3?}gWt0y!7CM3a&KWS ztV}#0Rt7`~Zi?bXP76CkT;UDJSH9Z#23>9^qwzdxeAzfVci<96^{kx2$NxZVv4v8kQtB%2Q!1w z{wS)T=G)&Bqq4Zg0FjbNuX&7oVDW^O^7_^9pqvOCp$0 zNp%di?b>y1kcu)BtH+F&L^M^1215w5&-*%ZG-G+FEwGsN!7m$Yi=M?;`NA$u(^juwTkxAwaldJ5$$6^)U;^nZjm_2M~K zU{qP6UEZ=`QEaO2<75xa5|THBVEMUT#%p>F^Hjt?l{a9{N9rF4hFB%oSgm&9&VRN( zRaoBE3|EjXJV9#`VQpz2tSxZb)}kdw4BH|VVT?0}=cLP?{}FDDgB}y+wA2VM1X`eZ z4)wbXyy>~XcGw`}Iv6V3 zoU-J-d(kYLt5L?b&?}i=y<6JT&ve_GWP%KGvt-X}{5%tc|72z>SWy#ZW!aS-Z@vjAHAPCvUT%os%|GBMcmRTAEvl|uU_gAR3bKGB|By6kdey zMzM_wQF8^~S^(;R@aYs{vl2eNx&OmCBQ`G-sNhWNfMIr}VmdJUu(NcI_vzZ%&$6P< z%0^DqRuLMPLe|Pe+`IbR|Fi!aUbFQMFi|Tv@yI3VOCOJiXbADJRx5FgPgs-Z8>OUMB}REP}7a`1H>qkow}`(U9?I?I|$5L7ST?9KMhY1G{VFqu7GiC z{^J45a#KzR=JcH~1wLX-Fs%!Lx><17ij-(nFW9t$;9y_TM%WRrxx>2dxNg9(R0^SC zvX34pj0lG6=L=M)m_OdCF?Bb?_b1Q*2YvPMxTSunA&_Wu} zX#EqTTqIdTql=ngyIF06lG>&a4wmx4!2*lQA(+mQUSyRX$?O@}vu_`nrySY1U55eU z_c#k0nGoEYUDHq_b<3vVWPaM)-Uoh`u3Q0hT%>bhMzfk`%hm3Y|MqnY-g2^2 z)X!^R8to?|;4|#E$ppz-wmQkFUV_b;j9|UB&}&X&E2QNikd1K%D|4nTYE{%hS~-OT zk?F{T=bC0MRxVR(m7_9IMT#lj$q$_Va?s827+@NSm|))+tBxEDU%JRlGEm%Lag$U{ zg71c@Fd5h8?6r{gUJ>>?B^6l&6Lg^j1Ir~YC4J0*f#p&^Q>B{SY(`cwLB$R7yPk2% zVMqR5581eP?+6-wR%XOGrx^v#HrJn8_{`=3Y(Z8(GlebN7}~u=+u;6Ap6C1at^G7} ztMfri+^u7oYz-^Fowz#ESCpwnku0}jm{zYL3e+s~-`nB+vWG;aSU=<90}tH;DCKmk z_Te=GCksln%UXl)FibENVauYnH9ZNw8=LIuzwEOuqt|o6RNtff_GQQ* zJdRn$Ek-760Z*yUY;eX3aO(vt-EW4+r17OKN_XTI#hMXy}J%2)=<}Wo@DL04WhS z@nj?@^<)acta9SP^8vTfie^-bHOmM;|0`$T#A2|JcC5Y_e%JN-7Zxx4@0B!c?$V1P_3n}x&< zM`)8^^)NDZk;bcy&*6VpkWO^1%gj?g?7H(!$T@qlIi%2%YP^6Pw|scBAt;ppwL-ln?F@MoBPO!wo{UoLzQGhe;pa1j*5SsR^nR^SvLt zOjz5IZVL{nOeZS1V@-%v`1J5@CR7VP>ax+?CR%U{EuNI*^n(|H+C>GkRuvo(PfHOW zhc68geY$?``{UQ&zy87XFUj>!;l;V@AAkS!^)ubKfWbhXzW(9$3*R3lI3jrSS!lUj zKMx-~_WjcYejSCW_~iSizd!o@(=8dblp~;Oq_Hu0JQcVxc+YNHFlw~dEH;blT#}mo z48@6#!Kl}?aRhMUDH6|VbrmQ^fP&N1OMIe7eW(!5wL*9e|B zm%M5|I#h8gVab&4iVFRjkSSiq8)&`H(~1+_U3C6Gsb)!%w4uAAKwFt13aod8VkMqN z2AP?of?6HK^KGHk$e1lk%!RX2VlJ<@`6l3Fl$7ERAU)(D&)4!84Kh}-tUQhUkP#mm zqZl;kQ2*0@VrVMWHW7;oi)v+>AwL{^8`M!ytz@Y6^Gu0KD@qfR5RAuI* zRzFoG8}X$^tG{xepCa{BNcXk)L)#2dt)H^U7Pg<^Fw18BOaZPfvx3g5Y_fjFT)1|; zc@3!f6tU-S>Xga<6L02+gYW%t@U8Arb!bo8=XfS}1X({e%wWG+cMN|&F zhRQK1?yK0m|JTr=G8n`i6fdvf@(Ci0c_)6oHqb-9o0|FR!jG5!G<)H*pT2ov=IIZ9 zyz-|`;_NF@D~SJ>5|NIZ9(Ap58IAk!avf&^=khwqT|!mWOiF36^s# zF)*zqwvFzf)$F%**vqMH897?iiPEOadQWz2o7a_IJTG%nic84|c+(Ln;U=LD-AE&i z=v+DNEeH%uDR|?^lMb2S8HY*m_>IF~JFJ4oZyfsc)`^#H9eU-)p`$;3^Xcp-?{rx# z>eiEgn0@Pyv#&h=U&M4uvuV<*O+D<_vNz5Y4v*PH_JOT#S!^?o(87lbSz=5Z_Jv+pu3JPwB>C!%xhfisO#NB8aqJAD4*18E&{XOczdWrd>O0*2npE< zy_Q_NR=yd-C|MvBgIr;4|7-4=g z2ft11f~Hce5FwTC4%aL)L3YrJq4V&?bXj|nGZdXmS<^0UO$S$34m&YvfD=Z(%Nxs8XxjO6G>CsF7{hyQFV?Tx0oCXhA*73zNK{MBG$&N;=UVwUrn8^o88;yzS(U^}1*F)nk{#PEaewD)py4?i4yiTu{ENuKJ>0L0$C znSptkjll5DW`JV;qYK#N=I6V>$o62Ab=Bw{p(T!%5hr`7z>tfaPQ=5@mD-vTS6s+H zCFTMT1rn?hLDo+>`XJiZRD*klF|QtU2xniBVzO7Qte|n(Z_nxluJOq{Bh|SN=v*hX z29Fjud0FG6jJE`;gP{1*x93bsEGF{xE9JX|?!QZroe@K;83$_ciDFG-dib4ez{pX( zMoo(s7WJiyxS;*cn}C9gNl{<*WvHIO;0}S`<4{Z+I?)We6o6;DzuEnEQ&6U@i45#8 z&117zvW@a_O14M?QfSenu#y#fge+XbmP7u3t^g51jq96!md_ z1Xc;y3U%D!_wGgZcTTm;fB_TDb2fs9&CTfE#-vnAlI7Tx`^~<6q?i0sUhB&lwk9F4 z>6BV>TR4<)k0<=frrBuV5vPObLa@!)A!K&pvB&oA9m|041iLMmuC=Mw%YFMEe|+Y} zXPZ~p`ht_BPPyOURkJ6b{OQxPGsn-)zIE#68=tOu(Rx}W9Q$bI?A5XUJI+P8NHPO* zWTFt0QX=ZDSzCDwg=4E*bVZxL+@~OlRv;r-90P=?M5I)*<2>I30igKl?3XiFKV2>8 z1j;BXMSZ-%_rDpRwR5&c#p+c~bgv8NyS*Ui9H@TsZ z`^`0&6$sk8y2Au3X+dcwq-4q~&A@Cpuw(tMr|}|bTxeWvoJRmplw3^pAuLC|L;C1>)jF)VR>wQQnf|-uFVu)_z3mN6Ap>drP z7z5Z`q-+#4-T`0!B%NfMX8cQ;ib4^bS~qiTw6zFF124kZcU@iwc`hfr*-IbY{KL_k zPqtxBxMyYxxc3R)Ua1F7<74rr!3wMj|ENS(Hp7>aLB=$)qJ0fIfL{bKT$V+SjlomJ z3`5{7>c~dGb-Kaa>>4djWh5;t;q8CHRI-aAAAw7H9Yl(w=#cyH@FCjgEDV=u_0vU* zF};2oy*GTPvYcj=O6?kfudVfPSuqadzt|zDR;|^{%`k=G<@G4d7hR0Cikx;OTLSM; z(S#JcBVxEL1LJdLd%4^b+P*9;*@BX6g~P5A-bO%_%WnV=eYC>1-W@jPy*129@VHW$_o@51r& zFI@4dR@GrT7?}XbDHf5^K5lPj23M0|s@j50k}dU5v@&wtw$DcnWqtX`Y^C|!u6{al z2cvrL2xmetGR(9phc61Bkc>*vm?S6TlALleu7~aehC+!+315PL-IiNq%bj@r<_k~E z9y#{orAsrX|FVh}T`bXtTrm{&kzNUVL)Gt*8t=34HO){}^VA*@J=Z{jAl}NuZzPlP) z3XjmdMypaz7%ql-SHdc&6yk~$@v$g}F}PgmU+@76Rz;q+6^kw zCb^j9r6M=OXDeb-M|UmcoeK*uQ?Vf4TT&=qW(b__BI?69CU+zcd7#v zIcyU}xczt<{&U&Ht|n2=>wrQ4)4!a8yb> zUnFzYg?Kn$vy2JaY&d%_#MSY}KIc8wOv~?5;=#xg_PGcLV`(eq#s?M&m#rnBv z*=3TYI_YOR_&Jtk5SU9cCJfy&dPg~aJ=IM^XT8I4h-y}N;;x-SU1Y3SWP!m%yz+XSUnDNLnNAXEeIaY|07;#0CRr6d-v1^rNe zy?5`JQK}jH_RSr{R;+N)p{>P(6~pA_enbDYBuW2uc{LsC(D}`sw$97kGHoyvOT1t= z*=)nosKY#vEk;v3E{8eMiYz&=gFHhr(h#TR2eBr|v~#PNXO@w0I2Bn>aV!gVI{W<1 z3&&GeO~{BqT}OHpj#yGI^8vh-~%59^bWM~a74 zwEX0R6c66OsAn4$bPieM3SkXscFrWQQj@Ha=gi`4iC|XGqkDYGsz-S_M{^T*ga5jN zaIZZ{CpZQU3`6sarDt1T&tkE(W@Jt@No;wClPbwk7dBdP1w@pf5fK?d?sw3T0z}hS zt`vEW!;E5e^EljU_k|gEVQD(7zGjhEOniY$azE7z8W?b9k?f|`@NK2ChCoL%tAW8W!Y=MHWUdK>2^r z`}PsrY!(}D14BErm8u11^=>mP_@=4muX`0`Sh3{&pAcj2jPg?B^TuZgFyDmm`?E+M zI{kdJo$XuB1YC6x`gqaT9pw5)m}&AdY>`SNl4v~xi56xl5fHN^BTH&5c_-$^09}&7 zzHi89`pq%e_X(OKLuHP2q2`(}19wv7qbJQ@p(TP2O;U=bupNpk@({N9E6M{5$*9+* z4O|R>RkxZ3lM2fwDH|&8cVIcGe95Z*bWD;HNdFL(V)3a(??6d6?%mres9t=jx=>lj z_U-%CCL(yn5Eiv4p>PMJ%7k+cvHf|51(kERf!_xnpe)6~c~8Q;#ika$>o&C?nq(!; zG_tXeST-i>1Vdt*T6lW9Y>`)~4$ei5>Mo2z=mrS*6s4tj!n>;9|5oZY!_;hD0K7m$ zzg^YDN=%VbzR+P{1;(NTTX3^%z?y2}8c*?zK{Q%5gh5PDvb;eU8RrK`6%2~53Pv?6 z3Y~>udj}BR5w67(i{nqda{Jl?d3hPdq(tYN-ox;Osi5X_O}`Q#?szPfieZRwsbzMj zQ6=fnS8hhG<))6>Whrd);R%ROhTs)a#!qV)6f+{XeJRi?-RW=9y6!Xg(SgH4m~$wZZ&`q0QaO%zP2- zoUWkF^;n%6w&Ml7%5Z29kn5HI|HPyOI9mkLfkM!F@gqHhOL0OsYVg+!RT^pg&lDraJ z^f4*rYf2MX^tL{(8PwWDaEyEf^J4dCNwl$a+O$og+Ih{e4{+K+bEdv3&of=8CSOES z6q*NAMuJzxN4jx8_bc&-Pa2<3ABDSr0)AeGaU3R%KQ_)zzuZwl#1dUfZc3LoojP7Ee|fHl$4HjY_Ipak~Mv%X@Gg|SkvQkh;EX1kR2*F znQ76S#?2xq7S{>p8usn`H{$*RJhtdk{39k_-H)(+l=MQLXYT1(>D11>QR6KGWO(P! z-a(yJ1Q|67$f^*&cYVJG$3|_gyPttxd8rWAaxp9gA&OC7JBMNjxB1%3*IvJN>e_49 zUb*%Gyn4CooabwAUwif1o7djI_92et?Q3VQorcl8e(fy+FW)5BUVw4E2cvUN|Ir!b zhvDsefj)MvcqdsM5Egi`N$=9C1vOlQ;F7YnxFv76pV8t(bixV9Z`YNL(c470=SP%Rmu5 zKA-h62z$NYd21;o-1gl*yJh=avSRHRD(b=@U={Dv)uNzeL^!CcHnz#wZm(@JM$23y zyQylo3=~;TN=h^$rDRWjG<0T5L3mvH0qmBs%cf>=74upO^0q3zSkaQ!Vb3;jPO5Xh zu1iODu^d~f&BCDHuHVW;nd-GQp&>gQ-vkRxsjw!?v`xsiXZLomKI}3km;f*BfZ+^d zUL%lc9t^oA!%6av!gvZI)qe>GTa{r+#?i zg?;-51f%Ha;`o}i3L74K49mM6X6w8~c4M8Ez?xO_@y9pZ8%v>cY0Wzl9uEIShhd3| z#pa2hT*C{$6fPooa`i$D*a{UznUrEa7WsQuhxZ#;%q@-;U{yvchE_v>q{#@}49pKa z8ct%Q+p6;hMq%K@V&#hIU}wT-Bk#Pw8psyxlW5!d)i3^SMCnSJq-&~li4?cJH@zMTE@UuT|s^$s$U3i5i!lsn;38$;qt zFh4Z9lr6rdDb%ugOwU{7Az|^!&|*kOz^p*1URqM(E@L}1cbejWLw)h~FibZCRKTKC zKO?}a;IPD*gkq~fD_>kfF8zKRu%BrKZkO+ZWI|ivJHPMe{1-iuS~V zhkIhk8+~W&YI#<1!sh>LGudI7MeL{wFN$lwjK~vKK=A&?hp|b5UV_D0%h~1dE7~?j z)gs4EO?R~I$6``6btk-P^{1-!51m~`Wn@5O0?g+nY7_3AP{N0LoCfeJjG}&Kemr0d zP=PrI|M)^n3-)BSOed*uw=QrMeFd%gQV>;{PI*%hB_p7~e!lP|$$*>l^A|^de)qM^ zKVEtB@1K7{{_*nBpHIC>ZohKn@1K86{_*_Bw~xQ}!@)o7+jsk|qZ@yI_sXwYpV-;| z5c&DNkN^JV#kuSD?b~?!%`3lhKL6UMfB))WS2;qr-@5Sg`_IBKLQC}aD@Sj?cl`F@ zk6}vPiKXz+XeASmw5iPxyfzLPLa;nWD9~H4P+@^fr!^f`%Ww|nVS;D{=7xYJf(ORA zU}dTk_<@N~@wx&$?@n9RX5laFe{dsCwYkdS3M_CD7!shQXqXl7h_#jD7`;c}v!V_9 zOtx`NBV*3+LQ8!+Hc+N|5ta)>zza7~vK(t!`_AiN3LTG1%6gh=i8NK8uytyNZzZ`h zDXMrvibTAkSzl{AUu$n4J$L)n3uNQm=x-lA38M;c`>g2No04Kany+;_n?+%&H!sW_ zf9vK;&xdy(R*1ePiYUZL!W$I{<35Zbf9E*%9_QWJa=caZ<8MArOb5glT0Se(gFK4N zhDl#z_;8f6{Y%)(3A4@xIXuOQgC6UX5G`@kCUm%+Y z{Qj&-WwF*R`^MXF^M|I7zu8CrY*& zTCjuQLo7~Grd?Uebig!a)Ik!ayaE4QPzBq#cW*wkb#!X$C|1(pn}LUhGh0Tdwr~w= z>2MlVgn{hci$^kBg$)*@%}r2B{8log6_}9rS73`HKRof`{B@&tg_#%@wdkTtT7Th} z!Z8a>*Iz1bx6#Z5JuBf8pn@1lHLSw^O~*VSCqMl zEa93%M3c{Cn4Z}U&*3jegi&GQKXrI*{wJ*2lGJ65NE$CWgfeFv7f9nJr+evFPVd&Q zoTjav5XbD{*fxu~>P36wZ4IpWvX5X)iYe=?o6EW>U!Ayh`01G^znJ~>((I|@tLUuA zaVh2VDc9o$cN91H$7>hYP~0Gj@a1&!&ODy`vWW&bV@3xhy(`chBZ_d@bkrM1FNSFn zWK{WnkOHCKU9|-lRNucuKfpD-h3(mX1qlMN%TJ7)zoR2ulPGgGCi7-Z_{Y#51Hd z=~K*RO{W~`3QUow2A<|eicC1=7~o21*k~3zl9K@#m&Glc>G|;V&;tTmVNx zO-gcblA9&??z2d4xlVEKrMM5<*))oP$VklVt2_WknO!MbMXHtU@Z13_XPLz!Qry>| zJPaDOAF0*^<2g)<{jbgBQM*9(f>~vvSU;^YJo00}Y-VO877(>=M-KwOCQOy$@NA zw04H(n}8w#(<$ZMs?BO}k!+g-!*!o12ElyY534F8yTE@BEeTkd4L6)OY^6dJ!MrJ7 zv}A>d;MX+s!6+@0F(x?K3R@W^a8e837FCXcWt5f_F#hGV+d^uk(UrhS^ms&4;$A;I zuqb>z8J3~>@o@ae>166YqU_lUs}i@!FfUOUVU>5%@3t*xPom|9Bm(;(ajYf^EY+R? zqk^^T?I0{j#kZ6%GPu*HYJ*HKQ&CzI%9gf@WINLD>CW>=HG%{?+yXBby~Lpd*t)8H zE1RBF3wt|1N7Mm-lP0ft%Uww&r3B*GK_6YFStq$p@OTowHo3up6c-v-9R$VY#@F+} zf-X~|#tYNO8&}}QFKn7`Uh+A%b_uOWJM5CNWs7TNbOFJX;5e{&0<1Q#KtgR(MY^OX zzm^4o41)cJQNr|i*08nkp_EX*t9)k%$!Yt!DSXA3s1f zBhTRGwGK!{R?3lw*;k~5(lMWX_s%~2@yXRoZNa?bcF?h9Hp;5l;aDXx0@uPKvXHGbz^&`#`bhyrwQ;j zd3ap;T(1LF&=TC7i+e1RZoBTm@nxR?FfBTd?#I)K7=itV=i!!qMK(Ivi(zguQa?%d z8@9m8PK^?6p$2sRVj}8i@(#P^#(zcQ-Vhbjxd3j%G?^_-u%`IM)t^FLw%0}C(_A0>4Gae ziV*Z@OiD$h)SXZ*#1QkGRDz3Vo2{CLUd49ti{t~VYxe(X>+SYD>?e(xBc68h{P%_j zBaXfOmRTeRhe#qWw-L&)kDG27J>mj&IZh5IT^h4INpx0S9&gLx4J;32S@0$s)?uP< zVcq=m`I#rr&7Ao%eB^b70HQHO^CmkFHZ261hG{{km?Q<$*wKV;vMhWzA2uN|jzz+G zmn2#~RV-l9Mhsz`DcP$qfsJ~{lraBx`4g}n{&x9QP6o0v`CZS>X2U8=P*ULaQ}E(% zm!Bbep?a_GMj}z1A8S4Za8>|BpAS{y;->_t!6iAuz;sicIQdc-KMmD=$PhG%O^R zQDiC6`ML*I+^}u(d^aXM`btHvk&WYEg@(tiE8zb~X<2e7IXapBm=sHR^&(7{%uznh zutPZPgT%@C<@svI^a6x49=5O%Zc7x+_nEES*0_PGC(sI&Z_Z+`#10Ulp6_w_J1%lL zIViha43WHY%qyjP$9FnnXt1qL(j3Qzy4}WW>fF3FclR_rPm`b4)YaiVAj$zmn35vN zj?=nTS&Aldq_>$2+%vFe-#${AqZ-~gXje3j!Xi`@?1k1J)*J|$Wu27z;rBVBy0}Uya^G_;W-MxdlzPrS0rSBb%`o;%}Fd1m8F>CwJHMV3t?ZA z)2Z|i2Va)c>EvqfXPx*_Xqx7X@Zf8(1oK5+0_R?1WjB%~=IYnyybdUbsHFIqYXfUc zmRXMOb6sTzFs=&gU$YIKUO==&3r1m9V8=3?r)S?jiEQkDXzXr<)mN%K9+L+y3k2R} z1~ePD8nS)+_RVOg=9#(qpN==Yo80;44#2QebKm-PI%YDduNw?`z$j{^)sYj zut}q8beUn|x%(BcOI(aDfgR&$QcC$+um{eg>^AHwmi#vxA!Nvg4I4JfaXIpx1L*A;_v^zf{p(>d2A)=2$o$aR%&tgj4u2x%dr8x3LBj4dUljZ z%X@6T3HJQNB2qNHPC~G3E!*>FZytJQj?(|>Cufeo`qQZoRx#2Xi?*Q_4IB&G#q|=Q z_Bh!_?TV$%A;-JPI_jADdZ$t0+9Dsqwd5|M2}mqE|8Xgj@|NBNj=B76UKl!f3qgFmr`nTn96I3_}I|M&7CLiDho==HOVl_d-`@IVihl48hk4 zAG=wY<|8w2o&e?UK96Dc>_<2Mbadv0XMgR0BP}UD zM#l#LP6hW*Z2{Nns^}oHv2wVoxGW{3-rW6q>MJZZv0R~W5hZ5cs2~DtwX$Au1^lzX zP>eRq*>=5v#m`Pi5lO%WRrwrVyx*`nxMw+vwT9id;4QlS@Rk6!uD47(XOF|5cI`X8 zN_#Y$vMiV_9+#rA2MWYo+UU(!^>a=S3OLxcn#upZmv7v<~n^4hOvgYxnjua zo%B)auluQsY{0i)ojHEyr*j=N=UyS-G>+vAO2q3Mh3Y+XM#U^&(a^EQ*|2RQiWTv> z{fycNdYzgRm62UFxS*HF5p3xqCkb|`E9XR4Y-CQ4QOlL#blzvM%1(6#ws{vkgNB_i zc3?~HU{vo12h(MFxFSQ*pn;({mPW*49oru8_tv}|7HiLM?Hu39!e?BzS9rm&Tzb<% zscIYxA{1YJ-1_piC=7S{%*xZh7*(V@8` znPFNoYh{jIc1ZTS!?rnD;%tKF!}sLG#3)~6FowoV#j*Y%lJc61f#Y|mYKFj(>z1+R z*EpvfM017g<_=^Az;NH(&#`=ByaNY!G@gsOSS97C%WjVbG~DA@Y9PnFS^gy@(Sytg z3_VxPDCo@ekxLEOf*U5c>L-_i6OQ$GMB{XS!bN}8yATXhxOuV@TYV+kG)L+g^^r@> z6@hR{R8B@|mFW&iYpZr!Mz$%e-%h$Eg)fl2I4QEDUx&bT^S}!^oB8U(%(0Jty!`s?2Zv|gd~x=| z)tR%;-TdsSnNy$NJpRPZ!xw+N^vvul$5zo|kP}ig;*F3$&^UIDnoOr!p@Sfv2fQA> zGdIX1CB=N}+Ww#j7+BYSymp$~t0=h)4cFaP-Un=>yRLO!<}2M_-A-ZL{#Ub=DU z)tS>z&AxE~kDU5o=Ig)ATzUS+p^p|0&K@~7bMoAe7Z2V#^1)Bvd_D8@$u3(NXHLIA z`~Kxy@4Y_rZg_tFZ5(S%be&90xN3@(sEaVP`6i%trKMEDYfprUKJ-%|nr&e#nxPjk z_CTvtxBhoG>{zXAY`Y9?CTk10({ZxVguCi3W8d2FA-!P=Hdda+?%bkK6e+Gecm$RM zManFA=2wq>0j56QruKf_npkX2MAB$YJacIFlV@)|`N!EeKASoE;m!9C-a2&g=EooZ zcyD^XY@6@oMAq##c^r zt|s+V7x7;&zV}U}ak_C4270maP2(HVcw_nxaOyL7d!9@mZCq`99a?~m7hu9Jz-+;v zA&oOQVU2T*voMuc8=twiJJ)U(SgNv=2;NJxXS=4AxX2fKw}QVs&UDkFs-M4Zr%}|( z%;_oCOG-vT=T3icq=nNoUSRs<5@C=nqrNc67D1bI601>GxyLmN3k^@^4rTgmvt;Yz zZR);y5=n1Jq@S9-qeZ@`H6ylVOv4YNCS6Pna@sA)bO>2SK|{TKw_(Ut`vkUK@3s~+ zFtuDm21u=bx@58Nb$ESMCa~eWO9ZX?4?OvCDVg>LLhfC~!YD1-g>EZUXx?;ad96f! zLLBQ8Vm>P3zyS}3qNyl2*b-DHgS5(3vT!KNN_r~8ddpOfK?`H7G z&GkJFEuO7*0R*IJF8L^~23tzN4qI`o08o5w2*Y!+DgT^{O^(1i9D|=!T2kcXl;^#d zlg)ei0x#QWmDoCXu_DjZF6Uw!{FkkyzXMv0CE3^#?a5i;__EJ{BBg`75IrM;IrQL* zERwNWrCmO(=g~f&wUIS!@E7~W1<$VyDu;eUr{K~KQ}AJPZCvx7ts`DjV=J{S=6JPp zB8kb89G9{GvlMr?Cu&6{69L;8fIMPSB<)=_xf-JD9TWShMqyM)4w5r;WT)TBRD0EP zFQFqQyI_`S4aBnvMrCr73zB6TwP<*7OrN%&niGriG2u-R?B5dgnjq_F!_Lu$|2XpI z?KfUnLuf;gaSEJ;z`c&z+4$uOGH_$xik#ZA0j?~GZh#GjK zW|8cs)$nks?09b^61lUa4f6W?dUpp7^P>xpA@Df^JYEd-i~5N;{k8mci z*Oo=LBCnx&*X+!dszq;w`3cxs8aKgVv0c^a?ayb@iiEb6l9rc^WxOJR`Zz-1Y& zbzn|7x|5LPWLk>55@H>6I|Ml-B{=lJ=|C22Xv4IO)$Hd6lb@DSzOvEvvbk`g94iWb zz^pIJ$hn_i|$DYnA=DoD}N+-kJmP` z+O{Aif}cS+8dH%fZSUS;4hAUvzfyq}64_<3;SlATFsRxEf!xKy+yiL! zQE#sC0k8@yMV2=vn_MHSV-&*NM?s$+ytyAJYGApkwz+y?z!?SpHtUpsy6HFE9cYp-5=6JEY@?L)Hh+F!1{N#M*&*IwU5uDyNjJvjWzwU_S^ zsqIP$32yu^*UnsfnOythwU;*FI9_mu2LpQ(hj#5P0`Hq|oa1M%orIr1UVDYWvG?KU zm1}W#H$Oo}Z-nUu&V5m@Q&q+RfKlYr48z?mR;4KV~6Wo4$; z9KO{taGOY0W2{03482MIf?#$c$V9P5U@D7j<+i6)0f>#VaTtCxmT#6eXY-sjLyI9T zHaMdhc_UwP?OjP%|LtVJ_S+!Ss1!|hjHqx3`UI;`E;5Al!whI(2w8L!8L^!nb)!6Q zklugCVukkS9G>xs4AV1vMlz$jMyGyj*d$MN2c7+kdxc(FxvjCp$RG`$Y8o>tIn02& z8r+9Y_a+c5AM>Sq0|S^a6__J(DiZ0w{9m7DA@CF#>g!iblWD3U#=}P@cGwzKQtseH z)TLI@5Nef_(*EE?&>Gaa4ome`8POUlVs z5F;5lh!Ksnu|V@dWlT$!4ZA>j?bqddjF6m0O-rwTLIhQDABVwfBx?v(3N58UTFp;V zw^`C25f)np5SBo|dMe^A5xW1K-(9R?8F}!4SJU}*IQPBlT&c7a7_Uz&Q>)hFp8v*) z#@YXVr*W1{A8VY&B8KUMjmz-rO5<#XH2#Dw)Ta-_(Ie!)-+;q->hkn6|NRE}zppng zO`m}4F2KL%r%$wVq<22*bEJ3DNm?qJO*?uuoax3pPDgbt2Kl^kp>c8g`Ib!m=Q54c zjjPiKr;jwQZ2s>*HLGJ_60kyM`tbA-ESSOh@XX%y5&tvM)3?h<%g~eu`3(h~MzPF2 zA5JI8h=hC$?w(IESM_8pX!RsKBQYuEYu_HGKStvP7}b^O=ix~>YkufX+@oj|0Mo`*RrGd8Z2M=-iVLQnRT5pEpa3RMgzq!F0=-Ob0e6qX{XI z_C^)gYcJiB{PEJuvu_@sIeljK%F8n^pPPO2$SPB=gVfSeG~rdZFf~6}%g~4%>pr!@ ztR*G&FRDzCY?bL|iITn&1tSaGY?HvZ(sKQaW?4^2hubj9IqC{)#ar={c;F04Ol}j{ z3QXqYWWAWiVdFLy5|)IQE5fJ<{%@pgs2JA*#$lW?Uz`>R*`nc}vhj7r-(4nQmK?pfiIUNzfLPGPF6F$yyAH ztR=yUxM^TvzIxX-?GZuhzIXUALq%;91<8-CX+a-EIfwr;qH zGvs6}=E}Ro0+!8zxMEV;SIT<-N3BMwVRzD{V&y_sh+bE`%Tiz~Jg;FBQoFK=IDeSv zMsH3qT0&3P8S&z-tLW=u{|2Gcxn368?L^obGl^~DHGB7tG<)l2`u6ls!B5x03!||} zjBLZUnW6*FYDB<%6*=uqrd)D$^@j`W_r}vw!q+f2Fu^g{?wHKvWRgwj-@a~%N!>vU zl&IBh;N2Zmx6Mh?2i|Uel7^4U?=mJ-W5Q*hCt4d)BH(tE%tT}M#C|By;b ziV~4z_evJNGg%Ug5YoN?-u-VG3~1Cg4YR`Yb_WUHm9h_#SXU7B5q;}H_mPKSv(Fb{ zVOrgu5}{}R(FrOnMv$I2Jiu(2BEJjS#uj1BQf(jo)UtDrq^LeLQc&A$@7_n8>l~=n zSpAFo7lLC*SlztMum-3#wQXSEz6^0nift84ubmDG0og$^L&yP$TrT!X7I@qf(9tex zn9RBL@UEw6891+|QCD;3y1EjkW&%XFQaSdb)lL$>0fDiazjkOIpIo0r@z*?p%ngItsz z84B;E?l)lVQJt@GDPuI{f-$86!kChXZI7nBHg%`yz@n|H)Byv+3oMblzp#Hhsa>V2 z5uP2Z4fd}ktl*+*$JT!TTMhrgbfZ$NY1%3lWhar$Tk*mD)-&a9LEU`qnVavvee>w4 zTW`F+YT5B*Mv*1iS82T@BT`S6HTiWH`9AoiKOZ}~gy*xex`GAgcJC-AnT zfLNp6l`k;b1lx)^c#Uze)rs5;l{(ml`f0tik&0h!BBoI$x=rfu!l`kt>n5qx-z6po zA%gj!W`3hY(uqVOnb^0_IfKWhu;fCe7VR?n)2zjqF1XV`Eu{y|Q*RV_KW ztUMD^+E?QfrYrr>tykWhIep=Ot{m?h$@UKX^#0M=lUM%d%2OD{`{~pN@Mh-IXJ?PU z&|O-0^TJ%c#f~mAI>YeKeAGx4ldU-^@}{Jn8Fn71TU; zxJdhkvQs}i@xs1+1F$V2H!2ReX?<*X>@ifMx%>nfvIY2xk3YWQF8Nu3$_;`@ZmzbV z<}kDNcuNkNhrtdot=>FLgbhc)0S5PrFnpdX+94VJ!N?Fbt(GO9N`M6~IUcWT3v1!aqXF^dNyL|@5*>3Xu0y_>B<*G`UokVy6Kb?9tv}kS|`OA&NXKx(-=Ejlpi){P$elr|j zRzhhhmAsR*?{-~9-Y-ryqOvN}j0&FXhTyxWicy0_8}EG7py2_`gF+(xnlT4{zN1h^ zOpy{k_U6Ed^Q|lgJC_MoM0c5TkATesu0GYnOGrsy&2*Rr2rL-2$Vr766|IR3QXQ<_d_l}cv&)Ejngen?PFP$!Kt}E` z3Y@5hyHP+V%~g^}B%gc$X?@`@QL&1M)MO?uOEF)^>@ZbFBf=oNSk35$eR%l>YAVcV ziGm)e)UpNBH#@`HxQ8>9i3wWNoa9S*VvlWDvBcdB?Wj3v1=Feer~p@a1LiT?@7?h)>m3gA7w))juI)$ZWEb z9UAV1K-&%CWOyIb-=fX6Nd?}c;wdQ_?U;h>VO&^xmpM8b-uZ5Jwu_{1yh+wM3Q5H8?yB5 z!lo+Tc;N`%2grJyYh~=bQki@2aAd7a#h2z<(LTf2oA*4v?JnP8@x5;% zjd!Py!0W4PudOf$F(xG+#Kt*-nlL;=6losZyjAluKmZ9T7W3M!_s$j?-~xgDuqyI= zQ5f7TcbOeD$n(x^Gm@Vh$_m&(t>sV-2(?Z*vV&;*)?38%ScL2hL&B{V&f$ijmBM#{!)<^Y{PjSOfh9yQhafo zWirx+s4&N9TAYl=N|Y*0<0g{FwtvDh3HV=xPEesP(cWjxrl$W18!~JhCf+1?mK?M5 zo3SV|-^SqF-M_1?#p(2?(O`O7re``LDRNwjrKFU*sKwoRKgmT=hVpoe!DFjuB&gx~CT z7-p44;)`fo!}2$!uv{hKTPxQisf$1Z-@NwYr6*>eIyn30Q>zw3O*!ot5?;(e7^7qt z>_dWDw!H+d>YnLJc(M{KqRspd zP{EeOwDk=wWjOQ+nj=GH4x)1qW2=vB%ThQ;vPEWD;cFe@k{n5*RiG%Y`je;|cpX$D z6a~JkV~IGpwYjYIf*(!h{p6`pqYTd*_S-eEVo2pio|(+a4-eO^MLB%B7H4UZixzKrXz{?l zw-S?N#T$xTl02x1bD5+*6W@RlEPz{xB#P?5se<*YPo$rrbKL;<5z|ZmT zpCI2KgMY5S0~Z|s{wTgUcm1z;?e#NYEMEWM`uo?ZL{&fo}QAD5=pP8w+d0UWF&bXMAfjZlQ1!!MmCn52wQjgp5Ayx_@N{IuOjBG9 zd3u%h-lnpwyNPATaWCoLdZ|G;iH^_$oTISd7IZ!i;tg+wtw^Oo3R_AjzCPXee=>0! zOpP#1i2nHbI2f!LuVWk7#g$t>-_y2?c zRyAP}L^&QdZ7MjK4~^E(w*=)om6=QZZLZi@URve9x4A%qLnm?3Q6iC4gjR7CGs;m{ zGBFz1qlqS^sLwRM|5a(AR?Zo)HR-FWNLWPo;B~Hoxi?hs=T~#;W3m)WJcv`T8>`qY z8|FPG$-e68!0gKy&BURj9im&@>2xA=-UaFHhpoBL$ovlByw7aqw#E@!i$w@3-<$>G zKWBum(D8ha%e-m{DH>cz@iyHniy=C84YNZ1l&00s7nxRLt1h4Lrmh@`_EMANkVtuB zzpc^=JZj-x*qw;`td*%F`T6KXD>xBY9tkP!D|KJ@Uc!1W8i)H5;J_4_S~JYD!PR!P z@b|`r!D_2W$ECRKk?c`rIVAfIHa&9R^y8l?d#)#^X{V>6L-KRPWg~GHAR#eW>v@V%r;E zPoE%-Z=AA#C*Wn**(;5c(&BPi*FGwY$%*|gd!kUK7-fR_-tloJkb8q784)gi zf>pHIn%B4m1IZodTd#0obkGUDpo+k6vHVu}<^ z`k3^CN~}*+9Y%#p6Y_YML=%YdFMw}72)n4|kxc7YW(Tcdg}iBYUKGL)GdqrcgFal#3II5H|(NPf{(t6RUp5CS#Q>)^3M z`qe^R1Z-@X4*{)*G?qgpqTVFN(qR2TrWJXQXI0C>w5qTjW zbp2GDptKDDxM$E0wx92mGe9dGw2F)|Au#w^L!|BeZ+>fmD_ty z%StIFDZYjvVVYS#_88TxQ@fqgk$my-$8o#)S9#6vHOtmb4Aa0R8a%&oyb`v~p zpc;x4%V-Sjh6YWZGgDYEY5d|B9i`?rOCcs1+3J)oVtM)&MrCl92}V)mtarzlOEAIF zOOwdc67y6O`!0srF6>DLGf_neMtRJfN>dyB@o75Ar7)&Tuv-cPGzGV2P7zw1IVE`p zCWH}Kdrv>T)tG?U-A6{TOqEX%G7Z+b!#vdJzb8qLb5%SI#^3N9=d?j;<#omk(#xb( z5O_W>!YwMw59L(146V4HL}b}O0zW1LOcx*yGAJzO!b#8~$y(UC`o`zu0$SVy2hL32 zX|kMc2$N8(3l-)Q#5S-`m}WW*dAN^P8$>5KJn51KxA2E(mg;G2oxYDe07xUMB-Ay~ z`5TX_4?&IMF0Svz3DB*n))px?wsXR(8agT%I?;Ofpe>>eKLn!P&TJzOf@w`Jx4Ht> z>MFpVNo?CmK7t^Vo9HB?cKa#}l1L5{@&Np-cNW$|gfCpa@Xt^GaO=Xk+n=0Y_|+@l zynd(&o3u4OhDl_R(ZcoGF9=!i5S{tRl58VIpeXe5`h=97mQa~_Hm7rwkx}$}HV$({ z1w*rAGx$K+MkcP<pD!(*H$J}M?uo09JKsP4(RY_05OPzC7U12tUcGzh!1u>51(xUCR}bHP z@7Udgr(sInTU29_s;kQsY>s%-V0y6Xobc)}H!0b=jN`tCOQ#L71JY=#2K9ICDeMd0 zX*$9tFc{0^kSpyHjkwwkBfg`QT2hI((p_KjzlW;i<75K{j7*C?-p~<$l1w6EbVKj^ zNo>yKjP%)fg(Zi3l~a^Gr5v25&z#C36V0;o;J9hax*x&A9p}KiM;zRHm#i=ErU3&I$bQG^`@lz^x~Oiaf8D!DC|FN z>PSjrPmJ1pzI8cMavwDVSp=Q2?I%!YX=vPJzVU`Pl-GRJ*iz3hV=wdO^f`5 zo6fD?I~?4p^fD_qlY|NLk=HYkce@07I(H`qHijG9 z606RCOi3i0-hL30gzH5tM=j;c?Y8rR%c95FgYD(E6?}^rwbL+RwgP`!7DX4R>=K1q zGX+%Nra&YKTIiSjS8z>wrg5R9oU%6DdTDe%c=KQ%-xkzUB z&mO9LQpLnRSGf+yu2s&!IhP3hb*6G1ztVYYrgCET(Co3wr&i?Y4CgJCtUxwRar*f|VNR7lKWsL!)oioVJmYE=54#X@!hRugl1{TiI?|ef#Fi3n!ji_~p@WUw@{` z`%X5%JgN(qFSL!`C>+AF{8;lqgbtAj!c)NF>L#0+M8rK)=vkyP$xzFzmgbwKHNUgn zD~+s+auC`em!bim9T}Z)n*Bm4k<}d;>OQyyV#cKEy^_3 zU<>qZWbWR*YdAGJI6nRaDIiHf86M8^6C^7|+F+nTv$T*Y?=jdDWXTA90*{(`K4CpU zw(v}$io?CO7inFE)#g4D4TplaZVdAR%?cMy6Qe-^>2baw*RZ zyKPZfn0Ym%Bs>r<>k2TS0+igsXcHPId93V7OQbyw?%TCU1FB&q>{(8G{A0R-2JGn{ zo&NfnUyqOP-hJ!cU*0@;<>sODw@$n{K7RASrJDynx_Rh>n}_z_JowJdL#J;Z{P^af zW9`iB2VQtEq<>!f6vh)+LbuLb`1a@v-@JD1*1IpetdEGRAm6u?B1$q*--jEdPHsRL zS#YWyz7<;sL*Fs-e+16>gg_=k1k!C`b^ZftWOJ|-Kuf%jB`x4f^9(Ohy;?gbnIY0m z6wZNzP$;`_Mt zZUHC^%saMSB%5wL#YhTAR0`Y{cb2jgez4;NQO*VL8yq!N)wS7a$(jx6y!JWy+SnswYhq#{BL$&RaFz)AUf|I2EO4AD%6ra9R{ z`2-~cN}obZMw=lw9O60VGW}}H%m3<1Dq&B@{CesA4_11wxjH2L^+$k^!XYK@fhP$P zM?qx&YPlR|D`5W&jC9orr)xkWmPuEgfZ9RiwTY$EWinE31po6dB#}rWg`NUhQHjP~ zq2!Q@ejw!AUJq+Z!m}B*!(`O7hrndqMx{V-RExElb_pdLRiYlNalNc$8N=%{Z{5D~ ztAAd(vhdt%Z7d@R#~buF_~vtQrUX`=iSQ;hM#~JfKx7z@WrnWPnUj>zVj5Ogruunm ziSaV^AGVNo5q4|?&BzX0al=ezSSiD2&SSLs=l0Aw*+d$9ksZdROiW@7RRjyR7&mO3Jihs`@oPC~PK_|uutM9APHF$;L#lO6@!tS@T z$FuwGEJy@``T@4TCW3}|s@KNykgL{{}Z__0bQDWhzV*4!m{Pe z9!BiXua>QX*lX|tU6ExY5mv&g5=~%1=d}AxJvJQ!BPt z&YaH(gY6-EnKbusVhK{cMru%`B~8O;=5+1l_;~BG6%QnNHwX`h31?5hBQGqQEOF0s zse`Ft5;Y#29B&Lfgkgb!I+qL6m^5@{&QN%GqF_!X9zBY%O)P_baIA?{+qxRFsH(&x z^=jC+-E;HMv73jkVv*5-kG}bMKm2{`w`Xp@_u@aVow#-E*D11aygY<0qGuOxj?bmEBGQDq(#`Drk4AY!Lfo- zYEdPutpT{vMC74J=W>>1k{Xk(=0pz+H;ST-D_Zz|I)`>KR?y2;%&5g0!qF0}IWtfC zoMWA3hioOId3atixHy_;l21$oPaRRq%5!Rh69dMF%Ww_V1mwK-x%d)F)YE3-yCvvH z+l4G*`AVnt73|`=&S{Yha0{NRrVbXpnwHoJhGy9F?fb+)$|oPd48oy(>vs-tU<=d3!djB5SP)uh(bLd|KtWFi^~MIHjBw~HH@A3*0N zRzQnPtPB^4l1(ZhPm5Nwl5N%>OGFaU=tCgcAtudC9>huwO8}`rR=*f#-k!9xV7p{n zBjCsiK3QZHr%X+$suD>&dP;McN}4a1sFZnTv{iQ^+l{Ab!?)G%B!9MX)8Z3lwDldn z%Ox{8)rak!=#iOZLP%nk7|WK4b6;3}xj~JF3ByDJGg}AP@WZU6omD197hHC#ubgGw zICd<{$*elm?om7Kd^MAKnhOQ1_{+Dp?^9SIc^KB@ud34l7p_mAt%p!n3!%i;S_nl`L559-!A6kH*mS5IJ=>jw7|=Jm z>ECunr{N#6Gm7vj_l|B>Rp&7L>!if1=a2209@_=)*Zk6KutV;UX(Uda3@pasJlIDf zj~gq7r+4uwDzh$%=q3u_Rfi-=txw23+HAyNgpy%KgEy|$lMx4o7)N8MX37n8)@RgbuS&{c`f%-vb*klHwpi zPV3T`KiF1daZtiwN!ackr*<&atBXtlex!SJ`LKcsX6Y&}H5YBD=Ax_0C8;E$O*1L1 zFSNxbA0sP3<8w-2m1QNTaBezy&wq>7TDVb~El4lj76*aTp`&V{*Y!Zk1|A6VP*Ao$ zlu1~LCz~$s*bZ*}1kX5a!6}dFjzr{DrZ+T6ckkXL>4cTq%=Ik6%F7{POJjVzBTe=d ziQsT_V0;&GY=PyuT}TVeGloRyYA6(P87o?(=9N|7_>^xnBvFpC|D!#R*t~K8} z90wWJ>rLqw8El87|L-+MWnQU6I@U*VXI7N5FM9a<;`J7$1@chf`>s?;Hs$5lQgime0k>CfE2k^=rnmHskEck#VNEb;FnAsY zt5y(asR7BYDV@sk@gI|ZnVWgXm|A{9FSYkFSwNepc5Y`$69$u~k{iq@aG3$0e#6p8 zf>%e4FwnVFX7ki8xVbl6)EoYVt0_qj-OW0p%rcbn!>`{ubmZF)U%7ScH-Tl=!~{$I z@rNxYKO!;(2W2XtLBccnyf89(7tE&`PSlEMs&D+rRKqsGh^K7hVMci$=E9QX^821J*Msc$Q%45;|)V zzA0Iz!`KkE@LZk@3vO8GelTQv8HHzE=)uG6hLv;CbZyicyXM;JqKPQsaMOt@5X}e! zAB-6p2a&(S7Ry$!9jHOJ)5+lUl0$F_Fvl}wh#Gu?IVCB=aFzW94wPuv8W#0roYq03 zTMZgrEC2J^ZVr-(D+v!}d~54$JG`Q{&ZI+{MXW*q6Nr|;8lstMfh>a;I1i%&<9DF? zIIAKq2i%va%!!gtr`%$7r|NwRHrXJ_n=yR~R%8ows&$Oz4@Nd)yeO=EnQW%IzLy2p zd|TlD-I?83&DeHc`Acu<ki4> zC4_TEAmqwrQu(!mf4mM^AqB{8!JDf1H{>O7QjE>A910 zAH#b%{bT1GY>gxvzB>2SnXfKA-T}3`aO)@G=Bs^>i_)(hOZ61F>>2D{ng%%jLGYA` za}8Ivqy~@Z7adVU+O&9Nz5V2V8@N~|&tSCR3#C~G$pqR2A`JqKD&c6oK-UwMZgJiI z=+f5oHQNleb8bTO`XQ*#^UR zUa&kXf;4i>Ae&}BkrQypHZV_SkO7^unKyH6@60=tTRZ z!T^yqhM);DM23?`50Gkzm*Kw40>;t4N*)cHykd#_No#HvBBwJ#Ge5Vv$x)3yflpWl z;~R$Ep$VK(F=1t>4Bj@R#X=!)wL|@>URCXfRiMRFA?JTJ6uz&#S~#IZJ^AN{f0=Ay zy2zEAFfv-VL?R(DQ|e>Q>%0QHs>3gv4u{7*EX#WbQ$yVRxeX6p#x}dt(#{&9)P%R( zkZB+=Wv+%DTuD#2@`T^@1AFDnJC=oI`>AcVC=_UFfa}5(6P%8ZBc#yMWm3bO8G@T4 zzlZ5X+Py_sum?|9|Ji7jXFe%V9j^Ysuq<1p6Sp&ww~L6rUknbp@t-uCw_S!p&D7U? z=dY0pt@+xvN7C3^b2%SmKJ&=%;orqTSLAHQB{ka!Jqoj*RN@}Ofro$gM=d6q$mDg} zURyh=fgw@v=@k4nF(hZDNtE=J~@2n-Q%e{2an%*-h_XVXjHXs0gpA?oUA zYQp}PqjQE8JZLXznXoc*700O|*FzBU{}e=&NT|Md!uK}T2Sx{OT3MC`W&Z?`%(Qij z?4@S49&7FM3G8x>bGg+B##4{9ct&5u^sRyU=DA@8H#$bCku@#GU8TQjOm$VxId6PQ zEvCdh9WTDu?wt!qiEGt<|LTeFE`CV9f93FL1%JO1_l7ski$ zzHxZNop-N3Ui-#i?+$Y3z0=?Q`O@NfJA+@_Ndj>WeP5Jyxy_%o#&qWVD3{)X3w3SKQeb7|0Q$p&mRSu z&AmT&ZtmF}?abv;s~Z;%!O7uY zARAyZ-k&=Qug?Xq&#zB!$7-SAn9oRX7@H)b9)PeQajd5lvW?LkCN=nC@3hu=fvY8S zbSLA-Oa^*XSZS*3n4>q$2_{Wq+AwV7-B5$mp6<$Xvjo;uR&}>f*Hmi4r$P;5KSjt> z{bl%7s6Wz~QD6_q3T>5e`b`JktJF-71s%iMp|-A$z!E_X z4G%Pv+`(=gqB6q_+JrHn^@P+^e>j@b`g;dbiQc|YDiKmQfw`Q}Qt?#FiPSYhXJE=c7)7}>dMQwmJP?0%f3z;fP-#EwM}LnfbI26Evtox{Fu zXH8X-9uqBy?Yy1KN$`*n<_=|i&FO_XGy5U|`CSBa`plY_*ALcWRCxm)X_PoV&CkQf zmz)-%VB-D>&j~Z{BAkyWYl3!O`xJCU(W3Pg)xKR`5cD>ko8azd5eVM|D^NGV&m~|D zQhnyFcJFtA)$AC9FW(zVa32Y2ML8J|#xiB=>Sa#&td9v}pMvIz4L$rD-7u#Tv0mHi zBx!Lap*D@ou(RYiMk5pdiFF@(nQVy0;a!jU_!3}fs;5YGE6*s|08{t)zis`=^wys| zI6#J0a)2(4q#K6YJ%+<=BH!SijGE{I7iET#Y9{2_w|on(C!57ou3SQ&S@U~TT?I3h zTrV4BNElKyQB4F-#TG`TH455~xr$>|H{q&ZNrW|6LhGbI_Zb*3JA^-JTTctKO&E3{ z%4h?q<#ka86(u!5#bxbD(2-ooCZGmdCC<2_dLn@9;%rKsrd?t(-M*EXP!b-b&=t44 zuBYf;s?$QA>0Nc(9wI7bGPZ*oiJc;BpQR8q9^nI16u}}!E5E?ZM$m;_+r1sG+gZu6 zpWB~Y+*x8q79Pu=(kuO1;z|<+lZP>#Ky<(6<05!ojdjK9oO?X&_}kjB1I#|w1S(O zy*eBlvAN;yb&S;v&6saXGZI$f@pV}8)nv8s&L3~T_gphhbW2+OK}Jy}>1h-)Cj*pM zaFLUAf|Gymme_7WC4r?iIl=T4**NpAG|3)2&2?rM$=A>Prh4e!k)pF+J*^C6gQ`ZN zp$_AE7~vMGP?*Mz15^roDzI3e66PceMO`3o8zYTtG$N}OX++{m)YD=92*gE@_9-U} zPGO!(?G2A$+d{H_T?MMF_Qy3*)p^Rz73&AjAo8}t9U#PrP zxjuWSa_I?Dd1dxjm2)tZ3&AZ7C8;UVq@t-V*L$PZaD`~ANhcaol4{d!Ss+z|?rpVfS)>^XGXpBXDe$4~qQeMRJ3ZE4Cs0K^ytaTUA`V%G|5Qi43=PJzNbi$o4&SSzTh_LMtTQuc z8T|3SJl(5Pk|H|YTXr0!p!P;qvb{19S*9RY(2uJLLk%ej&;As|#U0@LNRAl{F*M;x zBd~U`PSCK$Bqb)Bg8N_l)fcc|WvcrXC$l4de_PB+FAfh9N0^=Pan zwS=qKFX6Y?FM_=vAx~uD;SZf|7DleD3p2xMTUjz9*u&v1Px9S-1M6n1F(FK+nQ#vy z1|ufrCJ{52NlAy1cSL_dZDllCEvcCo5WLXdM|RjUcoeCRmMoC{7Lbg{xLOTjVHd?* z*!QNR#+0N7(mO~fb%5)|6*$bgwr=;tv>ooSH4)Mt5_t})Yr_~Va-GnOC7Edk@B7f= ziWUXmG!a*luIAim%_xuhY#WSV=ZEGgLux_)!MV!0HNR-u(>&l>H2&9&7E)9Xj=;(_ zDO>EUDU3-DbJ69N7ueP`M;n7OuZ^zuD8{5$DJ9P{(vyQ*KsI3~=GC78d)h3M4HKXh z8my(Rp{-rZh14|!65%-qE8gAjo{TCgmT|dlg20&y5LH--s7+@oGCHFMu}e>oYHf-k zHm0$`tI9|RQ-UQ$4r5ymO>dk5j?bCl)T|&++g2iN#1?wc-qa7ZE>$%YkA%qcej=)u8MZa$?)6wYlIj7?&5A|Z-3taG)6EQSY;a`J{SCT|su!z? zrq&pr^edWhNYTQNUPtgIn`bF&Uqf)jA$#!e9d(TeVX{g1y?VvfKQRHjzKNSkMi>)( zRhr?h;Vm8oHk{`=FO|5lX7G$^qX;Y?6_Y;^B^FY`F6%B@6HwH@fWk_=d4T3!Rokt@ zY_L&FQY(+o0_lTs$lyu|+o;3@HDd1kdEh#%2)4~=o<0lTvyYwANKA?5G+5XY?7~U3 z$--cSft8&#pMaNG+FtEgbGV>Qcu|DeaCn}Xk=iw!8t(7u+w#-tVWdbd6HN(`4HPvV zNj%;G{f4VTla4G68lg$?Q&N(`LV%hkR$X5?%_QP?ZLJACaQ|Thduzgx`W5H9;@DGT zylwa~!>KIZHk>jZu7cuGi@_QrBjqeieDKl1fE0$lmp>~lz%`_Q9ExiM3qpSD z%2_9CVS5n9JlZu?OQ!U8gw`vo?vwCS z!jD%`QV6V8yBSEy7S&co65wNcxXA-#<*Ge)r98)06C!6$37JDuqiPI*ffppCtXlOX z`Q~npLe<tgfW%qhv?o=J}Cy4WZM26PM+IEmzM%f(OeZ_SEpnf zXmcDy6ZTCUcsI5eJ_7pY3H? zW)M*#d2j=svJEbz(}93=K+F_IZi1N>pKJ!GbL1!eBll4d5XBa)goh=4y~4&VKbSLZ z;qrxrcV1lh@V^$`J$dU7zgu|z^1_klzCG~V!gH_QI&o;>%;8%v|8n8o^9vvTdf}tr zFZ}wYZ~yD{Z{PUj*2k~hK7Xx=a)Q=WZ*k9tN$nupM6~?;i#MHGFrUUgwt7~nVFwYEw<`MGNY6wx$u%=pHR#5GP zOcWKl*tGjnWvT`jS~m$+ExNsY<<_C||Gf03qrKY>1D!mKV^dMV)R-?7|$TSuxpY5$)Vmzles>3O6!NcQ;v@mPPTz zlZqNwv~U=k6kI!0a1ysDc3^oSNPNfBdJF>@x17`gGf5G2+!9ljNmMlvjU|)OI7X&+ z?;b79oPzP_sXovplBaFwtnN5spoEkNG7QG2m%YQ%n<}1rb$3{PYPOa2RS(1r+)-jO zcgzQU;duE~*+vYxQDMWr+B1P5FF$||4lcP~2})Qe_d26$B*Ba}=vVIg+4g#U zJh+$s%+~FV+mU$1W!!*Oqb4Gtk5MOU_o<9EO^HVyJqo%l48AQ)OX$K7rMzfDNw&^q z(;^Ml%_AWLtL-FBP-xJMOMLciemd##8a2;TXJAotbH_+v1M@! ze3$BWrXQ|;L7)z*bR(YSO_fcfziF*jff|zwfaKAc%LnREBIc-5? zgcZdkJZo9Iq@z90(6nc|xzvh`HXK;-+^~_@0PMTWqM?uxNMxCTlONP-Yn6$x6`Zr+ zjq{jv?xPNR18SD!zqd_rAl5WW7aAQ>T@G7Y2QS8MFCz_7rK7w0QkDaq& ztgYLa)W*vSj;|`Wh@yFx)x&SjjUh&K8w<5y;=z_v>utFe8@j>&#BS%TkQ4Pj=w@bS z@;sBbu);}_oesnW!_0u0lXj`Jf>!I=)bH_2H#1qJ#t#t|Y+}DDTFgPxtoJ>)@2dAb zw+jjG^hll?Y~wCwTJTKj&ipMop$Y85Pb%Psx^XR37yYy96W%+Dlt-#3@BZ*Fn87kQ zo*Aq~Y2%uSgdRy{wVjpnl4hCFR)rPOlvvDvxBVz==d>i`@^poS^Cfy4f2^bNxy?bk-OvmUQX{UR*kox_4_!K$8Tf*=Zwq0(X5iuqf4d z6cLv&VWg-@S7IYtOWQ_$r)?vs4eux1J;JEd`#otYWNXRXi{cxa0bjr)LrPlQo50t5)MCl8m|wGv$jUsA_*lNZQ9#>1PfD=YHY0) zrXcR4hb784x}51SPU0IubUG=dASP(@j3A6m@O=a!`rEjqnkeo@ghkAjk5bE)&DatS z2=MZvhmBg}3VUsp3VV=DSc$1k2McnN7K^lX`s@KEM`_WdWrF^RFtEzLR0^)(M!|G4 zY>M$JxOI`aue}9KiG-)Lp&)!p;m(DlBnA5H&L@ZOyn8(L{i`RwyZ9mb{*}Xb-abL@ zzIye$i>Jx=FP^@8?D*IFUlEVdZ7s3hV|=SHjLJ7BtC zZH`?uuoQK)#B{y$0LX^$uE>aR2pjM!IZm5^MjXYtk@vI1#E``%sn#!tbt`%mD_RS! zEh{>P{Sj)UzMKHYf!0YQ3ra|vr#kF^ZYE=r(YkyKvVfSi2&a;k6Q!OR8Ey+PD zY3`I+1eR$_5NTpgO}__K&hWd!c9cpIYx;@Fv($lH{L7!7etN4@-CsbgNvAN=1lxp4 zG7XbqJo)s~qgyxiZ|@n}{)=CbX=H!5*pp;>0yGwap`TF9jgG~YT}je;#(U$aF0HXY2pKYwiQ z1K3yRkIbE)`(W;)xsT_Mr24_R=$<<{e;AHjgnys;>iPKt1P{Pzm*-BxNr&f7O}o#S z|7C+S$ObL;cnj^V2ca0|F2LQNoj-&_f>9hJ;S`zsulZy1N6GwA80BFCpB(#Nfd&7w z!ZN=lQd9jA|C&FpMDMSz2`mHH#fp66BK$1NI#gzCFC1M0qg~tk^t~3Ai5309(^W7fWfX5{oq*P#$4$ z_qc3Shnx{*fbCeUK{pnbk*Hc!!L~3pF}?h4eLq&+$#`sh{P7goX_&&68J1NR;d{n* zO^@w*aB$eIGN`vL7V9BwOfS!oUJ8#nyg`2!TbL;5q9!=hdXR~YZDToRWZm&lZH0iS zw)&$HQ6iqbbUg;KS_~q()?yG^tU+tudZwb;OvUex-Fflw_pcpqW2Pd8b@}0B{ieEd z{o|-*TUL(B<>VbBER(k!Og7NaNc&}yDVxl^*+3p^z232xf-mNkNLR0?+_4u+!8Y!i zxhgG-6%oCjj5wcn)k=jTsuB-5)I$*$?O|{hi5S-1hht60)vN6gyLT_?m{gBx3fNbx zJ)%oph8f*Xo8+Fxfp37i(L$jJ?5IoH7PiSPebZZn1()hLVzz-kGAwjqnN8%TbT~F* zb7NU@s@*Bm$qEWrWgs^>cAJHPA;0-N(Ba9~88&#}vmt|~~Cd(L>=EoB2q?4;j=$lVY z-2UAg-@NeL?f0I8zuV}tgM||I2rHn=j^ZjIXe{RKvD1)(`A}@=znk5f0wyQih5{bW zwEGPh8I~Y$J>>t!FRN-Cj^pQGJ|&c@@zK{iFg!iNP@#|koOjsKO>Xf?_J804V=#t- z`NYzx+c!`5={ACClkJuLUQxXW#G6mxQG1*$AIm{qAT;ZLhdAId# z!k$TtM>LW6>h0}=3%C*&;c%_`srts}qK+!bX#JvIH!*|Bau|{1O{)yL3}i~d{|Ub4 zT4Z4&S&FAvU|pWg%d==(Rm`9Eh%>1~nups2hhdOP$(if~_h8kW(&n)(JAwo$NO;<=gYaXFQU?S?je$)E z7k|7H)2WNb_XZCiemh}|N@h5{B<(X2VDe*1_`Z5cdTFMRlQ6w2fr9QF^&}&A8XUtBqZYNC zMw<4gB>RNWS($9%xdu2a|3VmQ2r*X>i8!oiuE=f^G`ex52?%+%k0A0Gw$nNXZ<5JL zo?#uB-q@jM$slw3dYXV?+awjzs|(alHAw{S=ZRzQKq6FMOVde}&EaV^G##$5Y#tMt zJkR~h!e*oMDx0@cQKn>27+Decp;@V_IT5WrInHZ;&yK2LtbN@CTt!;9KltF*@fU9W z_GlaJKQ*GnlTD{(9>sFGaB^*x%f*qiK=qJHtVeC{BMOgSGusrl1$icdkb!meatZsH zk+Z~dyn4wTT7spn0P_h-50xElKWN=yqWF%MAviklsHE>ahnLRX#D zdh6Jkg%{7QWe*||(^iuRlIdOl)$P~bynXD0Z%%!B>*HVDe)hcvqUr0YmrNhk`GiJbbv4$%w(`L_{EHIAj&A$7go|2{2rw9oS_lV_C3eA&q4 znzK|SNI~RN7#Ne(FtO;AY!db)3yvT%#B|M-JrCF1jIeJ%1Hs5+O#>`p zH3LJyD^qk{MEqxV8f=$gVKssGw)LqVmjlt5H(?dl1ekXum08!{t<&EkUEkWA5xOM2nI&m$z*&OMmsIIcNcbf^?hnHw{T8T643p>{qRiMy}Xu-Eg<+ zDOff(*Kh_#-DoDi7skh=({WI+kj9~-cAj3zj49-tHf^132Hg~Vg$>H=vSgd;Mtbrq zn5fzypB@bR$F8zTBjS%N5^M)|I<6gN4TWrHEQ_?$AT%)IsbQBo%*xxWqau^Rm(3uZ zC{VJEPq_@enA*_Ni>a&1D6Awyj~?I8HuR-SBG-gD*1F9n8imcLz7DIT2dKm1po2wD z(g}{3)}#Q=W{0hSF;S%E6zTozk>GyLHnBG=_8K)Q#A@3|%=h*YMf8<*XQy?}PFe^n z&uP2by5debcBnp5m7#unZ*gM4qJdkoD0%wnMM8Jjh<@=4@^hOJX6QPdIY|jEroqt5 zR6kE;mM}nuz)%9Kw!=UK7wem!d+a?X28^G zw_ECG;#b90-dUWGaLni)l+2zoEy;Lrh=+b1>{clzxan{VE?L*H+Aih1_9+`xNk;2y z=GPTu^9qut(~?bcYPCs2Cb0ZI684nb_-%vu(LKPUnGb1di(RA!k_nk4%SaDAyff#G z@$ny%<)@8F3f4Tb<1IfAD`jD`uo!|1+*<`zAf(pgAw3vC0h_~Di)1?jHB&pVkOz83 zT*0K^*enS|wGSm9J)(|5W*}xo`vOV{tmTiWP5Zo^$^A4$Ed-Y4pv!+M%se+}F1QF< z$o-T^sUaoeL3Q(8jSc3+?e{-i_~`eoG$tO1=2^IQ>h^1w7tXx+?R%&H=j!hlUVQG> zp(6`t&fYry@wXqoa_ju7!F7??S5`GO>54Nn4l}G0Wk_Ikd?-=BDuS4IJ?P%{bO2dR zFEeaw%1ytybR;!GvtW+tGw?B2y>0rpJGB3fXXxG&^#ervw3?`0BIwB{l6I(D1b z#`JPK_XstnBt3A>`2efo%a^`9`Q@oEU;Xj}^5q}C{N0zY!P{4Gg?#w|ygKpa%U`~- z*7X+LJl^$dU%mr(`Va}WiEaARvkxn=BmcK+_wK>P zw=}g`mPtVrOE$K^>}6zc8J>i!T@ZVz0q+pX*37W+@&DUJrr_ZaMUnaIk*c(}Q#tk5 z*RoTbc6YRgx&aHBdIx!SIM|r zFS^vCZ`s7bT5lS*p^sy%H4hGUnlXb&*yl3LFj*Gcl_Qv1=5NjO^+f@Ikw(?tot zNcO@P5C|H6fNa2)WMNyRRK**Bn*hmWm~&zpOoJ>loiY)ufT3wTlUqa>Wa~&@r;>k6 z!k}P&;5@>Lwobuf3vT&r7l;AR+{t=v1+4m2074Vk(x7>|mwmg@KOzJ6%$tUAO5!?6 z7DH6#sDue3lkhMN>hO4y4PWnni>MJ*4X!Qws7&*$lN)jS&*5+l@y>beZ&ijhCGH^* z4dl$71c&4``2T(I;lRB)`Q>}y>!6GCLFeshUmk~hf9uQF$d|tXKkY5Y%?b$Su}`|T zv9OEMD(siiD&cSecwq1XpfBs(v=(_8YJ{iK;u&y&$OPEO&TYT|x<}|9N_upi=Cd?Q z2J;B1vF2sAI*O${DVuCE2Q!|4Dh-$WSoR^@KFR4C%(gd6A91iy>U&xJ zRtSE5>^HDf=FZN2JbyGbciN#2KKk|k=jV^ko&4%~@{d#VN8rQv=bxGT_^ao~$LF4R zIzfG&Pv$iPL+g70fZE_()DPlpS|+_y`Iv;c{E_3FD*ZujmKGg8Qt zvaP>od;iGvuHimxPGAy*If-<0WFr~i`&hO{{&vs!?#r`Gf!uqK;$Ku{oDm5JE!$!b zNZ2i^eCyMxX)@B+8(f%{C6=xAW_Xhl*dG-4l-b`$f6y)}E!)VMy3_T&YsbjowymRM z$bl11kbZ0o$ISeL3V1awrB$84d-g%WB0Uq57ijm)i8PZ`rWtoezjx-U%<>#o@>#sl zh}-dBNOy=yeVKHd$iqf8CF<#33DQBajT;3{rr2KNudi8h%3vMZM$uzr$D5D=@aIWFkt~(}=td&b*ADDyqhImnRG!rC4Jh6No-L`M%Eyg;={Hrf;p_iRNwfNOhi#V_`VPSx)`y|3}OSowoYN{m+qogWeVs7d;Pd4G(63PPIJ~8+PBdM^ zd9L$Pi5qK#8EL8bEp3|w23w*AXC@v%Yhd}!da=%jyOs;EYjK7cUZ9Vb9L$<~^^sW7 zro^I3Y*~j{Ge*9eP{N)SW%!K+rum*6+skCZ>M%{$BSEW;*|Ld(oTqFlIi<#y+SJ#Y zjeYhSCsU3Ho9=ks?1U$nPX(E8Zus{7ysw#v0KMpy#4Wog%{7V?&5(| zAFQG7pTJj}SxY0)bi~bH6KUM&5Lu^)tU7AMcZlB5x25 zvKqI2?CGc7^>3TFNsDQovmLYj7r%IHjaw39%ZvzomjXu@;mXK{?6R>m(MblrzAj>6 zce7HswiktcZ$)8E(bQJ>S0fZnzz%k%2B}_VqfC|?lo>YXE~%jin2S-D2keSL_||Y? zxJbC^B6^rIG6YW@E~*VCcUq%hkx|_Hwy=@d^k^Mv< z`A|{7IQSw!x@k-Dmp>6lQ(6vr5#>5z7I8rAt?oW*ut{om55s(PYXlc4{IW^@a)nR= zkNV6hZ0EN1nN!ZH71-eo$FfL~lrvfw75JfyF0kS=wrS#A*cMoTCg@t*1;f?&^VDW! zH*!x-2pOrdTAZ#A53#6hU^*HO5K}GE69HylIye$qw9ufyJ05 zFmEeVs&w`Ad;Fe7dfXmK)1<%1h1cP+~8S5ncCqIx>^A99G!Z9GkL!gzv=f)fCuFvvDmwt_E|7NT?}kslc*H z(BP$M$$exa`QoL&9sb+lFD`%a=@%cZU{Y-!@t9PbC1V-3CXA>r-udFsUtIa(;@_V6 z+p}L>a|ZI+7ax-^K83dz?!AQ}{`qeQ$rrD}saL)@4WFF-+rhxH3&)k1qD3O$l(-T( zTB6}f(h@7mFA_rTh*#mH5x7_N2ChW%l2q6EWA zwE5i%B)FX$7DK$$&2UDaoo1vP+*RHQCHM~%5-zBAMA<}@ka}N0((Bk`%c8Z*u*bBT z>7e4Q7KtG z6TGaWk4{+NGr6n@*M6k=v%OI~nBS=1k4(RvV8fExW^oUb;QWH2N_Owwp=ZgcEmBW`(MLvjsk97AF>q8IM5C&T>gg0&4}z{31ceD) zTU5L#!?2R{q=0=_O=Ttt+6tgAboOI1K`l$HdCj*1$M^`FWYX#tR`&=)HBM*=d?7|q zXn_TTrXkNzBJBwmwt7&<7R3`)l9Bqv)59P2+0I0s8L)Y@FXV=kpHQNn1VIoEoPKUh zG2Ng@PYuf+9iSf-sMTrK0d@u#m|@alQO9I*ztrhshgne3m>FAlx`>u^mMi_H`Aolo zWHeRLJZAKIhCN`GCy#ZOoFZ_DB@ha{E$MX>XG95m0Z{xJ5>#znPQ4spHZ;+}J2wv@Li&R0Kq$M3*rxP%u56G9Vz@T3H@^u*4n*{!P z1qKEVI=uQFe)dPleP8r^PkniUeEDm5|6Y|Z>6I^Ea%SU$;KdC{AM3K|6Rv(De7b0(>dd14 z3e8YhNhKQc#M=D=NA=8{rCII-`gHV+UDU6`LZA5%yngY}Grx|)a5&8gve~A(P9+%@ zQ^9S{W&E#A1#QD%(L^7;?(!j68^xIq+i2v6BT6VRy&s)q88U!SHq0U>E;xLX7W>A>AAf>qM2DqllG4m62{UraBt^ayHRPo&}uct<>>?%pnXCX z0*fpWQq%;P5uuo(CSBfd6Pb}nQi*ujID7+%|19_UL19iYgEOp~vrXIJe|@c*Zdlt| zt}ZkWikS>Y!_nZ{i)Yy+b%l!}wUra`y>g-J3 zKR;OR#@|iBFREJV-iSMZA!^kfso_ZPUyRY3O&;}alSeg0i`FZ^b-3MSrGIMVn-d4V zJ@7Kwu<-ip3x9s28K{PqAnc;NqDm;fj$Cu!`7fFSe=L)Sec56XOv3gxnW9kR4Iosg z-$>Nl$+`1$AIzO~*orO_czJp5cD-ICF?weHAe?o9z!wfXU;F^B^6~tUl>2K-XN`|PPUb$G`(XZAGWSO~?<|>r z2EOvq+$l2m{`@iL{{9Sa;JV8%#lxD8J!aFjmz@?xo;yE(5Qg|?;?N@Xuy-rdpVuT{`Cmyo1$4eus}QnIxdD=p9HrkkBAbB z*Biq5vVRxyzt8@C|BcUXy#4pXH~w_v6LRCsjdM3H{r%vL3pYNw@$TREcV2P#*5SM` z7zyr~Y-VDIE7-3!4)$xS1p8I28rHP>y);NxXB#CG)V3HVbkJtcQj2zi3>^Pr=9Jgx zaE_(6G1ALT_IgQ0Ve(r-AmrLXl_><1K==_8|H z%I2ErJ?hor*oe&ycRhHl=9R>JdL=5>gU1r}l6?4sn437wNCzNYwlA%tVkNh?LEQH` za_l?om*Nu+^4NqFMRIRZ^&tXJ7V#;Z__QK8!?X>gNz(sfP8_Ri-!pUUAjSY^kD(0vX_JDW$F4+H(PFomvc|f|2 z+eJotrlxF!okNf&ZP=x|%eHM-7v3_vY}>YN+qP|V*|u%lHmAP%V`5elv&hI*?lLp( z=bUq2hf-)KtVV!U$wCULe3}dr6}ndd@8~^5=ij45++ltq-MRaFD!%SpSWT){9r2CYm8xoZYE>5} zF0h@5f#BaSC~*Zv`0-EoP(s~Mzl_v9QRplg`!zCqK@{V{-|!1wvGg(e=#gd&YUOn( z=NlA`iMsAZ2%w#omKQ~6Fa?rBmL}@{V%zN&Nrm80E1599b3J6W zlxP}|SOFH1?#QzR9%@UcZ)QGt?$~rFWtd#Qvm*8P)a3TV({1f+EES2P+VOW@%JoB8 z38>(K8fQdEHN-$HEJVspE{eb~3yg3j4`9`zUxfL=rbDevfY9ax4I6gq5Y2B{7hOJR zapaM*BUPzyk8x%6GGf&>3~6tYqpT$o*^)&*_s|0)^^nYNq1&YAA|+RlwD|2~@WHE7 zonFEqC9$%8q4I|t-K`kJL`7L{vNLrQnY?3Qp5m{w*WpGJ21rw^Rt0Bmu;!sb2Fd3k ztg4^IF$Jx0-qR7uEi-xaZmd|c7Rzb+5Tur z4&G*4nUBZxjI(QQ@jQ;r-RNdJ`Hs+8iF39cBN>6$MtX~9wYY7?*IM77%KF@Do?vNp_HTER*u1@g?- zgrz)%>SsBo05hBv8{)<$O1Mo>>PSW&w4g@<#^Q)EVQhHYqb?)<%-G*lyrJ%VDr$}Z zjRZ^oYG~`R__^r8V$t$%7E4=+3z8J^laSSnHCq%_pSjpuelxByN7wN8^}~~O!MC8z)i)887B@0q_DoZTnND8v_e`@?!(f? zM2P}#^DD|Feev;H6nOF)Jx%}NZ)WuSkZU$6L5289h1Kd|vvLelMdR|k^t5fliM-f& zEawb*6vYdIUn$O7#m~#In)!?(u-51dxBW#Uwes04$e-3piU#6y#Ebqq-Xomo(eWsy zQ~w{jpUd~l$0qBaEpBJe{}q8!;@;+tsmUsPO4AT4Zp!i7Q8?;}^tDjrE!2x98AFeQ%lmNenu#H&GNx{sG* z4DTjmX;7I^$yIxPD3y6TBDo3yd+m3P9=QXcau*Q%%qXTFl9kH_sQuVt55bT$@$C0* zpNO)uyDn2z$dC+SigNvNzk2_N6*A|)7hBxw`l5U}X9BUmvRT@ZO{K_DQCou-PK~b1 z4_BBsat=%O2x+>S5Qh63LIsuyVLs9>Z~C#Nic7FvGO8$#8>`uLIBqDFG*ih($l7;w z%N|Tx{a$8Yd!~EI*dekz`Cyx?tt-n_yb$um2Na-oIQwjC#wu+*xrE+~tdDRFpk@`5 zp)}OaD0elNNi_aCb2Lv%R|Me)xiJY!&&g*Syf`2=G+4eKjXUJ=LMhq#5<}JK^s?OUc0esHAx`C*f z{1(V`_OhxTBLvK=h3(m8c3IZ0NgP_Vs8E!7uJgG#8K)$br{D8>^Sn@3iAg01D}vB( zZDDx++F?6y!YtQoi{=^W-HW+;(0pvSCk$he>Y8+z{3b;#Pni}g^to>Qi^$$4-AwmW z6Pze8oBY37AZdzrs($9DnYD+cHN6lV(|>6$P=}-Ts~6RA(C0r^7wyna!Ol7usMjUq z?tY_Y^bZ!Va8G=0+!Tpo6uvleBh>e=DTr~b7Hou=O74iZ2ERpY<0pKEF;s??nalGz z4e&OzYp|hfwEj?i*jc^Dazz*wJ%Mkq7f76HXX+9yz(w8qwrez)&=mxFgmp!5Qdq*0 zVbTmuHlrH#VrA2WM2BU)>$wz>6tNvYI!CXU;T;4Hv#(pikm355lClZ-N&Gsk!)TJ=dt-5U9+fWICzBtp<2`!K;Z+@6`5!xT$!hE;dUqzQ5HKiZLcgAPu zCf-_7LJ4EoKPHkx5}=+g{y)T<$+OCoVzttrAEFz30+G+_yg>|iu^BdXiXLkk_kScs z_18*!$SynBF1y&Cb%_^PAd#AcF}0T~P%h7B~+`e9?8a!tY{ z>ZFLNVU^u{=^)iZu8wwko@jIq5vm$R9zSR=sP0<)DNcYE=gs5S^P0vQvS*pHGz){p zBs25ViDeGlw675w5#%yjO6X#C{Ta583>ML%hvFqiUARWuwl=%t`>I28wAa(6jn3E6 zb4WM4o9FJni9d9;&<+(rT`dx1bM1{6`iogp9ALK-^x8QR#(JHDLzqRd|CU5vP}Tk@ z3&%ohd*f`~WS(DR(W70(RiSpIRvOlr&DOMY!C3ft+LFk}hs#pfOv^)C8LSoDE#i~|zIKV! z)jASABj-;C&GP>Z>cT6QLjdgM+Q)cv`IWQYp(wPKo%*^U4xJ_0V>m~HW3bYM6V;bp zKc}*@x88$F+#ipM-tOTrZ=`Is-Ai!}U`0WTRwcVr7_ZPl0p! z7r!iHxMv!X3n4r`khi?JboQr$(_SfO*c01cT{FxrpZ(nJ?R{+C$LE6{CcsiQA#5Ov zIAlF7ZyCXy7t>0oYF(i(lU?YmKNwVtOrSPz!qc?23{K*xk3hAI140{n;M2r$QC(H$ z5qA-~;s-LPkN`}K7a|-r;q+6xb|o6UkaaS^O;1c(Ed5MZ<(2R-kvlJAkr~^tsc~78c(L7$wpU80<0B!6I4Z>kDh9zo<0XYO& zQH9C$Y?L$H4YmJzU@|1_ak);JHVs1kA!|D9&(1bi8#Se$kHREF#i&5bGm0LMa!-a) zq=Z~=OQ~TF{MUZw0u)`pZooCb7GxVA5A+L{e8 zTM+~Jkc@5N-C}zM6C;#Z`ip8sh?eLt;zFyaLP*T#jrF8#xE%K9wx1KQ$!#W#2p=`w7@4h<6XyA)`^xHk$3Xn60`|b0z_1GRowT}L&vh?)zcZKs$g*X+r=bw@gd@;+ryC$i* zkjzNVcwZ^1C;>kL{s`NbUP+@`?J3Xw?1d`8+VEE24=01k_BSKLo~(yG zi5Q-^;$kND_Z12!OR8J{3uFS^fZVbF!)Z1Gj9&da1^caT+$wu9327DAK zHdH%?N~tz+q!z5y49m776ka*Y$)G`v=bhmM0CiOfR#SOe;V>sT3fmhG@YRo^DyIwa z^vKg0=7*na%!;mhAdr)MFmomW$u4~_O>?kID=|*iwQyVhXA(1SV0ee%W=PZkLM9I8 z!W4tusA47cR#?sVh9dML$#12c%}<&^JVn_r@t6*$b|h; z3Nc<)|BX&9R6f{vl_bh)lE4+uazAf&RDhNU(ZdDT6i(=35YG&SHHzAtCd?=eg|69B zV?4Q>vJJir$8`>7B%fj`1nb4T_j z#66)l>${*P7dR!jzKqbegp}i}ldrY~hW#q6wmb6(Nt$ly>>orpis=_d*RqHj$u`ME z!D8ByOE1xqM0v~drGqXSNW}u_pTz(x6@YsCkd1kpyC*Y{-2V3wRT8?E=g>BZTT_JcY~}OLKT@@ z-Pvll7O7-FG9N3R0a>-FVq(QH|F~6-Ms1lwuJ*FNMbcjl1;_OA6P_$x?Cd?%<;x1X zP4Dw8&TbDn&l95n0!q2Sp9-|}$Q}v&BfMvcHKf9Yvig7|q2g-Hm1?1A95;;O*+4^+ zR-2;fH8hz3jK~%^RyIkHBX=CRD&A`uUvo48REz~CS~P9QlA$pmX>-Q-n+yH8p7A2I z0QO}|1YGEJWqnJ{*~s_$joyh?r=ytWqd%MJ&^oWqyp2=~^tSzjZ86Bid>K8vT8r%F6W-f?r|Yu)(F#M|6+^^mOc+4JiexTtUh!I_vwXliEdx-Q#^T<+|(bj%?dA zMPP!O7&?Mx#az&@H4;=a-)kn5^Vl0*JrMq0|KFU1O7&s366sAoQjkelWeB>lznW3D zUem)Le%wvb<46Xh`Bo-25edlLFI+yG@NVgZIXxpmf zw<6YDM?Zf?_j&B6M*re@b~nwl#;IlwCuNs6Uyn9U)%HlOwXc;=9L0fg>jj-+GV ztm&rQYC9k4b74)Tws@{hy57S``5yJ_1_DvXmgOSzHDp9$do==pXr2_7Jz(mynNsyJ zgKh8}W;@a|3597er!t&U-6WQ!n8p`f#BqjkviOUbi{l$F8251!V%tj?0it4DRa#gH z2eUSQOJ#dihDjQ+(LAz(kx9AoFFb%HUzN-5hje@p_RU)$e_gGh`yUyS9oUd&@N%1V z=`^Fbk#B_!%Nn0RU z=3@MQrXga5Z8cj(06tz$5{=UT7vQzQG!nUU_iGlIeEURmq7%G__?*@L2GZE z;$Q{JFe$nKw9=05&gY7sEA6iNIs0YZ1O5AH%ZRN<;rg5Zi~FWH;bvdyK2@eo0ii*_utWkHG>9M+j}8DVIxuN)o2X1D zmHj(4u|PicO+yA_Fy)Y_Izu(JvuSmoLSR2@P%5|D=_zHTXJ>D-A^2DAB?&4dXoF_3 z)8^)kNa*Y83rh9xq#`+B6ZWMQ7B& z($!}o(tR(6_lLlD?x3I;hV?OprM^?HjDw)1MbWGfYpSe_R?%KnoAT&7hQ6EpPgiya z+H2zZV-ocyd61qm8K)_ju`RY|o*V0U?*>T#iX-w_75eIqEu`o!l7M7rP7N1VNxHE3 z#6;vkR=hm^KPrVBrG;uynV#*RYf@*{U76los3MdT97dxJ%=1cxc~MnH7M$W}boW4&{kN zoDnr4J432B0@f~VHYqG{gt{-{? zgsMI8mrB(;5aw4ICFVffi?4uBl zK|0cWI3+KxxdxUTB7V8&xNk; z8)zzI<1uMSAF1S(1{?LiPc2qVb2{B<5+o!hRbc0_WZKMRL)c}p%gjQu9>cMUkqv_T zufCFm%Yq`w>@-lBDIILkNd3_8(05Vm=CKC9q+vVF@r#RzStZi7t_1Np?767(kM{%I zF+dz@vPA-4(1i*;Yz-{=Y~WyBv4mA!OmBU}L|dga_m2u;rSA7Q>QordEJUeew?ULA zoUR7f%&An(7`zkY`W|eOLT8h*O^Hywg@HZ%+AtEv!w z6c8T>X$>nE!2sQg3u}$m76tTO@&}VX#=*Ra!srFQ@P>l<9#3J|;6xC(FV|_}OQtv- zc?g*NH{W$dGwz!i#xQ?V?8#w$=sQJxm$g*H%{utFAv9!$1seKe_WRkgPUq_)v`$utLbu;@&3?znS+niO`48OiY%uGf0)6-HUgtCKaqz}p$N8(` z_%OtFPsMiK!S;se^$bWeJV&e4=Z>p`D}=nwlAo2Ci>n`T<4xn@*G1HEx4SoP6E*~I znV{$BxSsN^T!N&=gfoof-AP-P9UU{~s+`R)3*u8@>`>Fs_F&@AwZE1s-?eCAzRVre ztKFO|4edUv__FUV@G$PR?AePfFZvb2iGej;TDq9Dfcc9O_JriX!-KNkHYpbzJ>_&X z30oy>CRoR@JhW#v_V}oc2E)I``nVO+f{n%uslu{7gyYsYuu|B5`7=#71e1i&Qe@K- zc^;lE+KoTU_SF_{@(j4OZf;PSHU2dDePetfjzY?J>1Gt~wfhCe_-kNs>T1FG$@GCHWsD_6RV1^s1KMU$8kcYr zt%9hTSFqDys#5SWyORi zb~m)G&6k~{ghzN5xy{0vio?nJ87NBvCiwNY-6;64?PmcAT1+{8(W$^t#%aq1?s_?R z9&80#Ds|9k>K{DJd1Fj$fR=qEO{Q@CO^`z4c;mR}#g|0nKwJ=c>~Y2d4@tb{f&FDc zP|5E0MxTQ1KDi)LqE95#b%yuj;~??*1Z7E{luB-^1D^P!3|Hb=;ggB!>F`L$+|Zju zmA-nY|8PS7pa!`@&ue2~RT(XkruPTj`%e`D=@R=yqBb4R+=UVi=pq!pU2za|zvNUX zWig)&xtqC<_mMX0C1m0i&R>WKbFo3^k2a|6i=0BWqcwnG-|q;6bQ9yZx#Wo7)y`B( ztUf6!GPA^ZLPGGBD&{|u6ZIGhM!khyz#67BE}#21CuOUdTL0qe#|3yor7Gi@n1ICY z!64uhIdw|2P%tqy2-HBxlpS|3niH_^msN__G}oi;P79~3ysrI5`+KqTb1CS$n&QKoY4( zlVd8>iJ~mOl7n2QFDgJJ#jN0jzLHow{~M>)@pm-Z{!ue5 z4O6aKn;@7%xsUU6C}?tm(%A4@WoT&YzS33C!u}+5MC|3P5afx95VJy;T{)OnEx+9B zrpPdrAwaNwYo7rUnh9BzYfzWWFQ!!tzxJK$a&QXdp2@R!zH`AYRjpR@%FewMQbz4T zP!~XrrNpH3Sa~Blx_BGsoD;}5w2R+mT2Boh=X)`N2=8<>7IBL+XDs24xK4Jsz~fgyA6)wAsO}#Q|;kfs6c!> z5(6q8a!ul;-K%DO&8S;#6Dzf^=({9CAB#X~UpXpCx;EuuWLowf>!jZQSp(|)EwV38 zvcxsroRx8D!oG$h3GdvYP(ArW3o)H^t-Qo(cS0d{`(T$$=6ts{4YSsa&Wa}0C+gBi zU!QOIG*d6cgxnSQ`BN#(9dl&&mKjKXgsp;ySES*wvp2? zb{dV8U016ty|VMymz)>4j0>$WCrTEt97Lo-nnIo%UVYB_{( zUEl>;VRB293rJs;Vt{$_^AFATH#65W%B!OvJGsSbJegLH42$awm<-Sllty0xX2DF{ zcVC^&?e`U>5N$$=@`%hg72y0iH%YGq=9^imd~33VE7NxkIC2+>B)p}QFq(s?$)hr- z1^V2s78~SLEjYgJjmX}V_ZxgBgqV=SH{(%ej7MARBO)aaI`~VnMWT(e9fHy5)kU6p z$2wzIynI3R>Jb0*_u7gkCwJD)D*-863|Hu%&~CV&w2pE)MJym17YBHnZ~z?>g*ADh zB|FYozlkxl<~Kp$J;M8nvG>}$NwkFi2E9kBl%Q@Phy~q7BQq(UgL(|lb*idkf1!`IP zJ@`_MAS2>v$j#3F!nY__RVjssD-*UNSOP>(dzk8lf#!zF&Y3x;2BT|RvWOH`w$da_#+ zBwLK98ev=v26El*e1U=o_Uq2XI;@T|taS8n$#>L+$f$kg>B}QdNsC&5qx;zU(iHJ$ zeNwrFr3Q(94$XWrw0@PNUlZOfl6Qtq^VxtMW2lpQqT~z-l~4v#6`!cAu=n@+GrDQ#S0SN?Rjr`+FB!g9`xalB?5jf^@q+lKuT?n(oZVaGZ5nPTVy;GE{=gD7p&~gZeXRR}J7?{}D zH!Bsp^AY>18Dv)i{=s@;Tz2_S8+~PKq?PdSMUkM8HzsHw&zHxebIpp0O_GVR(2)F*yg!Q>6CRBj} z4QV(7Fl5X#=ej>~IK4&;IRQkG?Bm0x$$v}ifMUR~G@h%h3Je${$VtivDFVz@=q_7v zGA%l9EiA4+LXjhw>G=KZ^nuB;dsi}hMdb`%QKk=!CQKSXf)r5%aOFrABB$Kp(*fwp znyH7*LyZg){$ zNq>Q7$Ye4HgKo&%n0eMtmeHz{;1FQA+J?AmLSDZzrnQg|g+`AF3j&3vTc?)m2k4B! zd3~SeJ^ikEjo|$l&Up)-pFv%-{5(9FMU}YQGx@PIC$FbFA_{^yB{{H%aubEE3_3^nFyx#V_dzYE| zx;a`MYv`ms0w+{5-&uN{reQz|i)72sXt!ej^{7Xn;;! z0oChi8yWpQ#sG$^uH6!EuaL(c;YQUf#LeC|o6Fv22hvmM>DlsZlY>p~0Q2-_$C|P~ zfuz;c%gp@=c3}xi4%r1u!7mwZ-kym+S1{`zWa(=NUtezH#4EAZO_)39m)tV|tS}BF z)glvBnL%L%!nsVQK0#a)NDh|{fddgGl&ILTnsfo?M7lghn_1@7GSU}VE$(=`yW1yW z>ssXe58rsNlpWN|GZRO<_|UNSi_H-2y3`BiS%>7O%u@6c^;Iipsk#r{6GG<+#v8KC zOEx2g*D>zG7fld9(1TaInCq?n+AzxkNRzrnN)a>U2?I%Y_OA=oFURf3CFo!6ScUu_Z1}X1|+6)4TSn2YA~eT3D*7 zkw^fHOq@hGE$z^$iz#a-P8=Wm$}t64%V5J8bYJstLjwGAULD;P$3?tj6Ygbm#hQ)* zX=>)ZJ0=4^NpD@N3nj9r9bX%LcqnYhF@^-wKsCUOnvwNZgc{zKfQ=rzcEp7N6Zl*4 z1K(a9CA zd6h*V_a}R0>tyN+33PO6?%Erot5JffweIC3QqX;&mBi{X+BXFt4aPc*ss1o(Xa|rd zx{Z<=T61WO^7jU3$eMQy>vsPVBNUmdtE!=4;;-9(Em2nCM%_;w52L3+DQAcKhFmC4 zaFw{vSeDB%qi=SP2u9;$cHFa~_w=h8yN9I7<%9oX!>Phb{ek%umYKW%d8%=-mAl1w zz<^oavteX#2_nqVNG^D)z}`MYUOt9dN22NGY>7N@-x5V_?x8{ zg)P`D8++h1xQhzotx8-hLnhAbg`1>j-fC4lEsD|zf6$cleutSvYb?q7S$6}p{zjmc zG6R2Fq{+6+kYXVRC`i}H=T6ryt$$gPgbTJb7~U!~Q9qjv1OFaCokVZqLR;YNwjXZ4 zpBvOD`QBeTt#3ZkZe@14TuOWO-nh5UYW1|RzFJD-?Q6Z3p5AslGvgn*xIFf@doR=B zay*}~x$SQqw#jnyYF(ww~;jlF4hfdAf>Fv6<$6N9gLj zU(4CKdRwor+44Hh_l)4RJp#M}c)y=EUTvnkJ6zwI?=-xf&L>(m9J-TvpAPnpfujp3 z#6_OBX9UPaLxS*Rnvb?xm|gh12Y)JGCb&H13=ByNqVh<$!ErPoNp~x@(b#Nl4@MwZ zi&ei1s-P#Uc?ZMEogz9)@Bli!YiFC%tJp-;eZxj&flVjfAL_{>))84Vcz;nNlg zILv%2f#JePUg4-dOGJKFk*wX3C zdqS&4R z!boj_IcU!HUm@I0-Mzn=)?J~OjgnK1HIcUjQjRFRJ7b%#0-iW;Zbn)d2Q53gVlIw` z)tV63)ZohD57mnQCdV+Ky3O|XFciw_M?+)F(p{VaX3FS21I6qo^3kpcMSp%dBb|om zj9WV;LdypkxEw9^x?<`0d{3>pLxFhmxnto;5iY99FCfaw^PZ4HdPv+)`swVXz`+W+ z(%Kju2s)%~?X75M%=O%0b+rtCk_H@Tji0J4x1;-U3b^K(Y({QmUkE<+f-c6egoe^m z499@|mi_g2kVADPUXx$wb)(zrqtgvXXnP$und>rk!d3brTZ<@E(L)^%7uFPh)OuA^ zunijWb~5<*PJz6j29-oHscjDcKAG5kHo8#0BJ zo~drSmpnwKXekBfPqqo*0$8$<&0HX(-lp+mOjhooQ7ml1olCy?B+&+ToAcprpr5tw zn=H7?!fvGq>b5l+U)iJEaf#E^#@8T|b{p#I&vu&EurNZE5x6w*JIHLv8WvcL3%0a! zfH57V)IZ?~tuA3qM(dI~gA z(`2+p{zQJC?mCD&<=GJ_1P_F+oi0jqx}dB{dEDpc*Apv%71 zXe`=LxI=D6h~^Rgo+uuH_(v<^BQVrnDtdG}x3bXXnWYp#k~%E&@^cCjE&UOHSotyZ z4?^(wLhzBoL_DD6j0|^~JVWJ1A{xgJe4a)s&M{gopk>CPcc=R>TRfz{2CqvU3<;~D z#p=II;MG!zp=D^e;G4B-91Oe~ZXWu|f=aGYwJ-6=dWj2`4}-|jvs2POWCTh?4SziZ z2!iRB{7vyvf3MvVd5rW~JqfCphr68hgH;U$IC5v(GarUI6#`KUt6q%{H$U&Y1Vv&_ z1cgB)%wCjovY-Xv)?dyQ=bpmoD@#9CQHC7jkk$J?^;oI4KI*(d_0_Rhi*FLsI!0|| z@EQAUTBWG|7FO973jWF2a3%E{>u2m$#?G1N1&)pWtV9KaE}z}R#|2}3K*fp;}KhH2V zn`jz@hp5OrDY`OT-%2x6`~gDJBtg(A8qA<=+7Ta33Tjw3sE*pPNt_fs>3+5yuO+4YR!kWMdE+My zLEM!23bD=dBsR@CCJpPuw7nB$&w}hejj=nT2zmYI&&r#pVr4~EVX-#aRxSA0WYL#r zz@1>ChRz`tZ-zg?rXOY#@DQ(;Y<6Q)2_%QZJu^?zpDhPyAEf{#OWQrt^f&UfKX% zpsQj}koz~9$`+etb99uLIY*`|Bv)p{_7*i^OGSoIHcb$Lo2nU_1^Jt2SIGizUltjU zTE%6r5XVtbi^}S1#R_dDv&qI1;hhHWiJNNygz|^e1-K?4x;)L)&)8J`b-2El>wUQ2 z2g(228tzmoY%tj%FXBcsNSo|sd>GWw@%$KEr2F&zc-zhMIOwz*JKg1)<7o#9$E(g0 zNap>q3^<6;=`B*}enIxU>(!YK%r?U4P-b!*sj*2@U?xpob3eTthx#=xFzo33uHux+ zYi08d3{gx)%t&njkzq6{QxbgXe#ks}%JYOH5fU-h~KLKAwVaMiT zTQxb>5|Na8q}^nc>o!1iWTCc5TIXeVciflNu#j170-!EUq<#!%&66`&9NzRBwo-Ek zkQ@<)Mu9$HlOUcJy%U}K#RZT;gNaO`hQ3a;K@6WaRd&eLOP()^>taujgx~yZA%peb zX&=(>H<3%!s4S}Np3N_ZR+A!Pq0FW)PxTm$y~WH zsW=&oyQ1p`eY9$oBO|f4v?0D}Qx{oi;#Ex#YKj9^9!b;Rfv`l7#5=9Y;iTt zdx|rXoxpuUTfjd7^>dLw;kfktJKf-;{k3>Uy#yvK@iO!$@XLXZ=nGS4pc0DP(?qOP zWK;?yibk61pCs?T$3Pw|^^RSJc7T3~g;N2AA-;$q7gN zAsQ_-uWd`3)O=<7PZb^JqG`u2-Ot>GBan_HW5vC!xA}+0PJk`c{H%0eHiEM3sk3=K z5yHtl!Yo)zXfD~JDM`73Fc3lvL}k$b80ArNXD!uWlEwRxv~afx14ar4ROU|JEc_7G z7*UdXPc=Lb72=daV-USq(4QNJoNonk7>PIw~Qq5JUsUwNjmbX>Ow_& zl>P8>cbdVOv`n!|Z_wbkp&QHS{zo*^bqU3cXy2$a-vfstuvL-yKBE1JSaM57iqa%f zBgVc`(Dt>|3>MeouT^PAbK@(_?icR01-J`rd_k+4*=+DLSV(U0S=&)Y8HRpA;wnZ& z`U=J{`W(?h%q{Ge!Uh2cb|zk#OxuN3#{Hqu1gA4a@@dg*Oj1SI`}DNKpXV`?gLQQ6@24{OZtnMm%VA#2ewswDnf81v zSM0->1l4V-BLf4N%<+dB)lXOLrJEM@jGB1^i8iuIm=DX&@R7kF0Q58sm&cuw3aj{uCmK|< z42w34sQW`x;b?L5W|o#b)+vsdGVzULL+FG_dXk-LNv+54lrBo|%$3T}2V1agJBA@X zD}QQkPv6m70$Pu>EbG~r?f*&kIB8s(uVl^9e@N=fSxe4C{mW7s)BFJTgPUiZhLfA$ zKZeul<_HVT37zKH7Dpc2Q&7;IaC8MpP4fCNl2Jw*! z6|x6iVY0z$naHx~alQzHHJqVt!%N$-BJ)r+jMG9DkYMGT*Uo_)6pHC<{n=E3AfxAz z<{$%yf>SPUV5iE&xQl%}6mKa-tu;2ttB9=STOQM+E+xH?jPQ5=o_QOqrjP9}_()Kd zlem997g(68w>#VEai~=+4w9x^@x}RowV;gc_F$CSH(2g=DGYBaIKa7i^U&)%#?F zRSa{80`tHu0pPMm#tE04W>kxlMr!6wW5HK32_8n__vkoL1!zH%BPrji8HR|IWKA&u zOb~V+&=;YVISB45lCdYV)Sg^+mSf{t9tAIn2#@otf-b}TKy@&98!R^J{fE>|I0lv^ z9cP?`=9};?ym2feJYCg~qZB?K1S1l2dOVi#j2R=a22kM6$@f)3(V_JLGdUVkdM$_@gXj`+!cErYyXcGwA~I<6ICAmkL--Pj5@3ZNrs{v4KJa&oBt-n z7SpeqWQH9tLp*AzqvPvePgsh({H#~$SWhbjn=D`O!UR2|@a{8?@J=hE1iQ~7bUfI) z69^uhx<1z0T+~x-x}*?v zh+sQVX_O5lBmkUCK!9@*%98_jHSM}|>)W4IB{14L{xK4%OCX1z7f^Ke^J9K`+Xnug zgMSk9MEyE5+_hNo@3Oo2-YAiVO-gzt%KV0T0n{Uyl>+4eoscKtNK{47>x_U1*?sNH zn8HoET1R1MaGI7rCeI*Xfu0IUJ6i4-37Hflao&yZu9j-dA=sDIl@F4IdzeQ&+7>(- zFoVr<<)g;K49dcaA#_KbHekD4vl)gm;1-Wc02E?k$Z+zrlb_O64t*B4tK4F5aIr1m z0Hgpj3L>;UO5ha9RarZ4l?rd+IqyzIT!;_|3f)KWd@(U(VR?CWIeP^mnZCrO83 zEG|a2W~{}Bb-7JMx86rdBh)4N@deZ?yl`!#o>B+dIZLYWiQcp!IOTaAvY`{%DI0pd3LKXqVWr-(L-%I7F z4eXpJ3Qq=dA}ua$3ppb02tWowvC2y^4k%9fIFBRtK$l>7ybg$}QuV@iZl+Ylc2U`G zD!-p#T%9fWj&iNRMpl*W^2Ianu-#LA%>3P^_t-XgNhyE=RXsR)l<_p1ndlQ|1Pj(} z{puZ?GZu%=#|5vBp==&CLx<1&2kF zU}l1R<;nf)NV?QThKP~Xh5u*N7qvF9H`@x*O$Xnz(+5=5`&acC#zgX0{>5IWS382p z!HSo`0`mv~1>{E;WdW{lC*7m_OK1W9?N-G(ZHldyj)UtU>gF&EmU&nA_v)ny-PY$z z(pAQ_i}v@!|IUanR%~-Sp+UD~zLGTHkd+dPqss6rXmn}OmJN6jj`ZkJA@o7u%xDgu z#{@f9TnC)`q@qb!DeCxp%~P3Ww-t!RaHPRhOXusCTXC1(LNGhHsS0pl&V~9{=QT<- z{qFb9)jVoCc_@XiyMb|-S+O!)+8G$^__c%(^j-g@XMUK9$kUDV8=fnGQe1)B6GJWr zp7>>A26eNhDb?;4i-O^^?W=Dx$3+ZFdigQgA` zg2<6ofY7m~Akmw)2BCt6AJ~1XW}9fWPX{N}{k6(?EdO5s>_8L0snC+Ce}_jX9tgA= z6a$Ufk4nN<-ZZe^h!;ziC^HG}-t@wlT#=B9NBt9dQO`J{#Gz>;yo2T6TxdEsAKB{_ zL>rWXDFk1{qs==)U@&_1m19I0S!(QrhFA_YinO16ed;$SP8=TFeQ;>>F4 z`ST}FetqiodnBt3<7hdw0*8)A{S%H04I}$AL)0)r%W3$~--b?1VtOsNtHG@Ty8HS`e)o%S=!SYa>p#!{=S&Z*T`XB4ugM11TO z+ld1q69-VeaBWz9(%iJs_%_|nqz#}%!T4?;MAvmWZR0ePX9gJ(q7Z%>)l+WCkR@!7 znHQEU5<5Mv#C=81!h{VcsuwKI2u9&><>&?dZMEFt%_WUT9L%u?Wy|#fE*dBMhf35a z^$l9W<%ZD)v5UOlqi*PpNg>|{6i`Z0W9td7-+B4XyDwhtx-_+K=6`hugx)pM1X ztBBU9&DSn>*PN=ohTNKnx*@e!p$VTZhEt!x%EI;d>T~dNNrobySvb>Mdr<;-PQz_; z@b?OGjfNK9K7(a4!aEgPD}3sj`koqBl4|oh7-plIw5Vy9gyl#^&Q(HJn^<^b7Dd`5 z*k^YjGvNgW65K&yA4Bw6gj zI~XlDQz$dbuJgk!`GOms;Ghc_qLI|C7oS>Yc`LdW&rmnp$*Qy{22F{@n;i_HvRdB2 zWw>@tCbC?q&3A*o+_aVaN9_a2Lj6hY`oc>x2?vR(N$vH8mk`Xj@WjFsa!(0gv3|_v znYBv0!bXeZ!>YF=lqMd>{w`vxI%tM_~U}o@Pn!eU`^vxU?d1SW`NVa8=A;V3Fmc}k>R|sZD zcwHv3cx(41oL^Hcn9toTmM}Qy#rlO9fH%8OAVbu4(07ly1A8nix)ELpbF3LJ#956b zS4qH5t|^F=67}IAXwdm_<99b7zwyzH-;o>d+&p{Z^3BsX-o0^w+&KThi4$=2Z;hmdjZa#72omu$gPd6WXD5^(p zoWD7D@+A4cOu;+NycJ~Z+&ok7t;YCSI@90iWjMb1^P5lJe60TVG3cE?Ku^GDeGQ?9 z9%ayp?|{S}^G1t2JY0HUlPjhwYO*k@r!Y*|BS%I}r~Z}GENeR!KbD>4OQ$g3Da)~LnS=FRABEqy zS*J%?c)H-_=8$>|VD~7CI#r`Fh3Q^uDi!I4KjCwLhdE{rf~M+4mTGae;r3lT3OKT` zmgjo#r2}S$Gq}+vB?}bjpJDCW>;u&c1}kp@*l8O%Yha$6n5AZBDR0BD9L5qb%q)BJ zof>`TM#}bxswDS_eFN!5W&ggg%nrNu zV<{!!<8#_d5-8I)k~cT5|8*7me+zpm;nREt>pSKv9i0byOi3i0`jCJ4pYGLatb)uI*mjQvagW6t3tc>*^`au#w&p9v6$X5}j~3xN*QR8LoIMZ9qV8+d4jqDH zEX}r`WtE+zv;=jP$t=<{iZ|J?^td)&Zei`VZUds-%VMJ%n7xfSmOKb3YA}{kz9}3u za$P5$6(r9MvHA0*9KMb`ugya9LU$Q-0)Xs!AW>?(WZYGu5uc=Bb-b`~5 zJevIuhE6xVR8~$B8T@`99{jPJKfigF+&nGAzjN^S$?y{;m(7x?oJZ|z4aHakiJ@Y8 zvlee&3-1NUU5!M3EFkw~9_0e^D4T^MHA+(k%dx&;<~nvJkHNgY#hwVCD$)#XZ)avD zl!R}GD+P_;!w&Z-%V-v=8hrbZQc#u8v)WC{e!p<~6$@EU`#99UCu zep1+vJ>*uN(oz8{OQA&LO2U_|8)kyl3yWn|OM&0-cUJ7DhGcE){5G>68p#YKtAHVR zh}t$ajPRv2QyQ3}suGVXs<&XV+S+}$rZAF9EVYHhLN7t#v3qAL9|FB5?VTYn2~1IAhp&RHj2fR z_!g|!?OCBL$q>#3>SCQ-qY}t$xlKp2K2V6r3R9G9ZW9A3iZSP$w^2)bBPeN9%r}hq z5O1N<5#B!Y?c1k2z3YS*!>!Y2Z=HVQ*6BBIo&LqGxo2;kdF$4h`CD_pyES*|*4%%D z7RT+kUb*x91^livFWj1Y^48oRZk_og zHyd9^Zvy^M(8oGnNDt&);Pl5_Ja4(HLl3G})$?M}mgZk9y5^wLl|-~{T7bGEKPxk7 zPn;Nn^TM=AW-)XG$Aaa~+ zr0eD0?CeV+de17d6trJ#t4Y#}T@YTv&@C%;Hjj8?!HT({coRx0+BR-#kb-eWT~fV} zrpV>8lW=niI_)?q87?BJsuy4rV1&W;!ZAe*rf>%9`DSpKZ~oBd$}OP!RL;QG91xDh zdWYeixD&AGNiNz#+d7}GtH)q{&RrRd4^)SiQ7 zDQ7WRwrv#v+@{X^*#$Gqg`Oe<>2{bsQc-lD#W3RKSwq9tkk&P%g4UzWuA9I?v%{5) zn4(}~vuq`Ej1l+qYZa#;rcGWzm20BJGMpOyLjpv=)2^IXeHd>--(NJQXxi+GqsF#e z_u#L*G+VI?u6>M3H}?SS)fPEGTvX3-ujD z4+lSH7w|lis)V0+)S$;b>vOyzs(5fw!A2Qvo_T?FAZ)b5Oo{|VV-$YRVT6`hX32o5 zUXT_HtfhE3Xo+cf^bH1y-IrBTO+_Mvy2nZNDTg|C?;~a#Vffv{BJPbyX{{@IYpYlE z27#%nqWT>2VG?NTfN_~lx!Glao9xQSd;fmdq6Q~>EpF#wVhvg}UqqOHctGY_-_&lF zkcRoqnlCvr@O(NT+%gggBO?bR{Yam&LyJaI3H?E^5W!TMVp*iSh4Yx_N?a3W8nr?D zwnX*)EJNlEsg)l&GQ?nNx)!ME4|p!HXERIZOBOdXyaclr?^juOT&OpUN@nG}ZrP~NLurDW*)+|nsif)&BWSwrNs!kArT}PlI_Cb8^5hssQFvQZN$g{naThJH-^CH_wci&@CR%H8F zE(iORXBpO8yN2{NTtgZrL94O0>&kCytPQ`iD2W_aax~HcR3b1}J5_{0ljPC9M1y;f zxU0ouKbE}RMsceBa6H4!zUpgfW>Hz93fj)hz5y5h<{38%u9@A4vQ%1p6~!sOrsjlc z+3TvOvzr@5d&%EdyIS@l3YG z6aC}Dh9^NT4vSU#hIZ`_Et?hJur4)MODL&Sb8LKL;mWUE3Bo{`VjK@bqgx{OW9I@opS&83deC5_&TmNnpBPtfTN|ogSCeuGQviSFj+#!Y?HG z?=>UqN)=2mH7?4V-J{cMj~UuD7Hya$h-`B!X~Pk z)Dw}Y9cZ1Xda91@m+a%hxEub=2YK+8N= zAD#x27Q;cy!CI~cBV`!G`y@rDL$l$Nfk)^xXPLamIT1G1R6}HfmT1<-X7T}rb?2pv zcYgc&-B;i1F8|P-59jWD^y+t~U%dO|r{Dhmv)hk7O?KRU^_{yfzI^AU_u=UF$G-_* z?Q7Di6f0>ErMK%A1(WiiMp8jxBN`$VR9{`?ZFfXScLYWH2-q`7Nf?!kL9^k}jvM5O zh^CZSvN@g2E{A|Ta3hqmPT2XsGB)V*LZDeBw8TFKuyj?{0M_QWlVVNDa$k-7R-afvt2pWA)FC%REHyGg_b zy%Zx4gLPUixfXfEl;{HJ1X8ku!)&L8%zE|?^sJOw&t8h0?zDhyTCNzVDQZH6E>~2| zlUdk?ey*pKxDO8-1~_e^~MOeo!P4^Nnac+uxRTW-KKXvtpONR6M6*6d?$qKFMjaDnTs7Xrlatwl4u_bD*1~+mmV+-ascZqk{x>m zHRywg{*>>?4*XtQn3C7nT*~vd>uAtcD(LQUp9+fQ3Nq{-=%4}okPSRX(~4Qgc#Odr zn^fN{igPr z%%R`xvtWWwsxw{Z5{{2xp`F3`=MA^Crm2b= zjVRHm67^8SXsHHpPreT*+_;iRZN1EqAy}SSv9r@Esc1f3CLdInDYD!#maEKEFW9WG zlbCGtyR@2fcXAbq9B(t#b2$Q=P06YL*(z5rn3XnAan#CjNB7{3W1h^>SkTPTC>B(R zM4L^3FyzYngeavc?YO!hzNP{acv<9OYPM7OcQ13NsRQPyC&@6`h(0F+hMz$Inxd!f zJ#*nC%j8WY`Kff8x^OS-hqT}*T)=N_GaY(G(}uC>r#eiv);3dXkHVM|_0ft1&hCe0 z;Ja6^e*5-Iw}1Y`|GD<$-LvQJ{`rsfGymt>Q=MKJ+DQ9<{g=1jx^VZi7sCq#Yw3}4 zLmZ7@Uf@VxG_L5reKIh8Qc097I!$HNyn1Q6%X)r4W=_rs+kqC03BxXMGEE+Kv&gd^ zNre{2J~5pY(_SZ8_&o3^qAUDaC|dyfa`)56zWw8;-@Wm4c%^X-h)cp#I5ZKoa%dE* zKqP&YL$_?s@&3^RqdyHTmJLZQj$-nCJl-~L&(D)Fk;zY^1KBFoE5PB>)m73MtX{J$ z9v%v@*UF|>P+O?0kSDIeo*0$c_*%0HzjbE@&5(l=9h({nCAz&zt&{~xc1#+uai;>O zk2XugE-}NX)8#GEHs7D#->5D8VPQ)0Nk(>1Fg0u^d^BT7ASIJ`iE7Q3wFX&Mv|m@1 zln;s$hK+fEL-Sh;a8Dj%xTA1opG;i&&%3YIt-XW4>D zOQrB({X>&`$==a{;pz(>Q8i3?EpWD=t}9xsSsnQCJGl3A2d0t0VvJ@d(UDQCD-eq| zlOz531%RMOp2>?+zG4_16hN#$5>X-^Sw=3Y(jV?A#ZUz5>G`v(TIj26AS1Hk6{YC zpW8T8vde3d>gnPV2NY&|d|bP@@WjIN3unl}FEQI?;mO5Zkj1>07i8wg^@!o zpbuVx(NJggLb=yvevwF)uxm1HXLHNB_8SZLwZ8C zW@$i#xvhTylu?tK;UmXL?*`4o4c;d;*)0?86`usObyQJ-U36A8Tf=|M2kd=-J{Kk7GoAdA~Q1O zWH%xv>xtz{K)^^!(XhBUs50IPFlwr`dn&c20Fz2Q(RM8!hZRc(br{)A7|GJQEsj;A zI&1!gFIk$-B0PZR$>=Bn3GU&-%3>L;@T({{40F06m1tvHMFY<&RaGK-+b-Nif+1$Y zINa-+wzH8TzWd@DnW$dNIvnnABlCjyPS8xg$h;F{6)e#jP$G~)RMC7gXe;Kr!y9pZ zunXg@Pk!|6sZ)1e{^k1Pt)by{F(llN_>hxZ)@OTgqEFiZf31fQGO!$yN<>>v;e%}= z2Cqz3A!2C9pvLrbU(RG;$&9lysAFJ2s}iz<{9&sVZ_^Nwf;{aS=pWJ9m50p)tW6i= zIhlg=h-nMg%D@k1mw1?j)rZs%nCYGcWkSI(0xN&=@a*K_jpT2=Kf`5EISBvjCxfnq z6)0ig9CY%`niiIkH7ev_eccgl5UUJP%jCk4=&_}}ZI$fxwUXCkn7kg3we17jCMT*V zPvg7?rJfl!r=&UW(Y00Sy-R5{j-PQ3Ro?X_&Vdl zv|~0jiOSFZO-n=*u_SzAyZ5V(*al6i>QfoGB7Calt|E2qm4!&u)^$kKHOi30kXc{% zfnx&~udi3NU}hN>=@P}060!6E%{Z&nsol&Q*&$fg_8Hu5qsaFt*q$XYK&$5wu(mG! zKMJ<6+ANVUw(w}Zqy{``?=H)x&P*8@A@i0DS%Z;p5|(Brf$@`LY3g@aCD} z*d+^#Ad-Dcn0WhmS@bCqD z=hCGI_RKDk#T?ipIdrsWOpj@a2gwdN*<}R=CTWF?h`gAZ3a@}J_kw4gO=>MCby8b% zpu`m2$EUchfrgXh5Vvhm7)($Xgxw)FbP#1iiTg?qY;*3cgLCJTQ{TUT?T53klO0PP zaQ01TRVN8;W0+>2YCmj0&VJ@HTdwVPiLq>D8A7yZ><>*0|MMO-It#A^US*m0E^`C4 zD2Tfce{px84ZB!c91ktO1F$`TO(q$&)JP<C~90y#8X-~8Yx40 z3hD`SH^fRww4RHJH3b<}G@qd##ur;BoTzl4O)_faW%c>cT@OPMT1@td67#sv5)qFl zJQCO!9@CYG&wsXMV-gbv%@mMXe**#&Z6uO|IO0mu$F+VegwcL$?oYSQymxEv_qXOQ z+?u;~=fzLIJ9YZbug`pQ^>?cTEgl~81uY(?C2DSpfW__C&wMlg^V^^O;@0W+?mYS2 zcfb1h*4%Gy&0YGZe+n%jO@lo{Q}mePNgrz@nFr)`01<&@rmx3}Uo0S2FI|h4`{D`M zkX+J>ASDBAMzWV$&WzanxXLy0$b0q;>?fcZO$yf{ytv%BA=p`+eBZDrxY#l6r2`J2 z+j<%lG0!rndAK?Pi&WRZ%HAvNx6F6h7KuINSP5t z$+`vy-LmU^RlAu1Z72Jw-9~Wc&@r0HdLrj~Be`D>$W^2xZ+^@N#c302Hv}ukGR|(A zJ9+ZJi4&AtrjwmDr78N*ZUidiXPrukJyh>G0vBW~I_1nx;>oe5lWIzv9h@D1=){Q= zjCxOyrcNNx-pJRdUVxYIWJ!|~Q0$)7pOd}Md1`NgMh%Sg~TThXEloxwtyjaYZRWalBW{tJVC~miwA2#6jC%I&>scwIp1sVX zhPP6kswyc>QB`#Mu0W`@Ho4Z6R6a(U#);#SOQ{{&W)9LQR{`X+uf$n5HeNTf(PirAF0}A?_0n>S9u_dc2WbhphT0*|w-_vRMrp>g_Ik>O2BZqq5w4Q*0g3XJ z;9AK4bLq=Je|hf9i}3HWFX#XB%$FZ^gKhBTFTecbf1dsFk6&K-^7mi91+V8CqVj*% z|K*EcehLTY{`1tA@5@GB`0}!R{WJXU_y2h+wERYBrJzuYrr{^Op51t)TMIR`ap~AYdXi1N~VqM zy%IHo7bvSq%4q(JM53djy@Y^sk&(%Nr z3>?7K;mr*x7PyZ&W>AZFj$Fa{ZAG0{(#cjYX)ktlbFr(JpZ@ldcY7MIif5IK+L}fQ zQ*M0Y`Om(7``sJAedz1op6$!};*T#}KljV8e)r*xCmy{1+QrbqNP+>A$tW89O8V`W zcvBkQ>tLH@LQ_(0OpIYI7=oWI?0nGb7DI;HMN!F1UMPS$qf>*G!&@$6lJdf`P&L)B z@j5qkYTGn*f+RYHAa-qA#N_PY(H@b2B$4Sr-D92D5C!EWm+WsFMEJZ%7ofC~NPLiK zYlhe*VF*TvPA#WMrKl8nghsB8p@>K|OUam$ZtG~>MPfii@C8%Ip2|Q|(rvlFz|`5E zr;DItE|pwh)s3wSI5{+7nJ{$wj1A5$Jyp-@nc(uX`uP(lNEIH#a-PGvGCJ(6d$sMi zL!nJ!C{OCt)6WL8S$Ob|jmh%w`kgoZZ3UPBBkrcMd6)nC;a|V_{~rA7rN5s0>+i{5 z-}~$3zrOL;&*9(?duhFYEXFO;dfDc-lnuNuC6gU$a>F<_=^L*+t|bSx^o5XSUz5e9uU!Hy9%ZDES^6Z7L-Z}f#YkiTdz5e+5ubzGT>jxk8IX&yAhT)#oST-FS z`upW4Z@hf@%dK|;1Pe0wYMMnbm?BTz@aS9fyfEq8 zk9?$A$)K!RDWvj9WN!vGf8rFVa0H!4-euH(k`3yOvSqtI@CxYoQK(jdmjyL=)&LF%qJd?b!KdW^V^C!rNlCOj411j zC>3Vq1b8A5${M?n%}HU0QYzjWg$cas4saYe_FHKr5AGDG+y&+le(ksKs_W$A$A_6U zd~9}jZsNe9si`xP;fIiKW9MUHwzwp`ESzGMj8oHz#AM0z5~!vHddlaL_)^v0>wr?r zDrv3NDG6*weHV7%oo+yFp8EY7Xq9yGE;1(yOfE1392&L258Po0E(V5xoMmjH z!$6g8!V{!p-7sbp2{UbCQ1CURP2w?ltl-`KiBv|tOWEP#nl;f+GEJY0wQ*{Ajyq)* zj~P+et;WFP;?^d?g{FQ;c`Z%ZR2T+JU4oY`T+unE6FxbZ5Ifjp}ooic6MF# zu;wm>-7ndA0ezy8#?Gk}c;z!=E zpTZ-!06z0b?Q-qY)t}U^)ZRmn5&r<3`@D7m{?L2)?&?$h75b~a45NPt?)_)0KZdb= zOln`iP(H7H+Bp3no(%mU)ZT`fzT(aHIhevLIE_9NxoIE=ejt*&Qbn$BR27ptRh_zB zgJ+1gPo+UwwwlOW=9UJ=D3(Mx&dToxT2^OH6bj67IH4G6t|R<(5X-=!`GFR(3>yM#u1gK0+>2g15sd*$;CS?@&TDFmiBUvuFlprC8*a?uBwk3X;QC;b2m&A+2m& zQIQWNky122enXcA=#@fZm3Vxw=WfIQw+yjl6HQUW*L!rO^)5A>RFzn^wK^(r zgFh^twdb+3A3OcAsr|LJXB)lk*!KSNdUyZZYfrDe0xw@&Lm*bN_Qu+|wYTun7uJ3O zmv^t_|5h0$8%PFkzzx6T-SB$*a9l}b11rf65*{&e$*tEJc;IKWQ0QK^;Px1k6DC`RdFG^o$m0^MWK<%vm?3cl z8E0joqYBZ5cOGUjMfT^gaFHCfIQ%sBH4szZa&DTEe9R!qS61Ke7 za*E)99B;uAd)+Dmtak%YK zJNF;wC;sFYQQL6GWc?{3reMpBZBfyX&bcPzsC>l6STKZ=4d1+)- zNkvEh0;#CBgDmSQr%15A!If1qItELQIztq0bAG>q`dQK0#|l`BglLr{BtQQ)MEmqB@N^`X_DVEUz= zWVlrO1W(?gmC!XVr`oHv_u=**u6`D%uAN)`8K#>E99>*}c=aKI z9x5dKWxe{-BhFLz-m-YD2cW%pshprM-6VVP0{;t<~X!qa!_vz zFt4XN#ie$1+&Fc1^JY)wl4vcCs}$N6UPf0-jqP-tbQ8V zeo}4Ou)r3@@#D3h*M2<=`|st__XJfoq`vB~WPhd^uDh!mj zIXot?5jF36VZtE|^tOtzvbd5+o!NARJ#+{k$K?Ox1xz*I>B#DX zt7l=ezf9&U)x||#DphaK%IsPp$g^OXTqLzuyrP-UYafMvEDlqHE=p}vNO&;WIq;a= zHu9)}OW$T`?)87&QT<<&E}03xw|A@UD~3;2vM$vuK?5q7(8 zq&)_TToPnt0cE=*VsAOOLP-u(Vu^OBEgHCQrqvKWyhQJFO?PmHf?90az-lb1sF?&9 z!e|Wp!f~q4&kyY8Ybhn2XbtxS?!fh?2y8F8a=wCX(=M9(!p)o5L2YnZ#|t7@K@IEZ zpyasl$sAo4ZdokaCEjOi=^zZWAUN`)vGrjr*`@Gs-#@s269_@*w!AbpA;sGY^>%F- zbqt1eB-g8$X*{OHW3A@Z?v;h4RdS?$>#%?3syLV~h$b{DL>haIVr?W5fwkK#lB|1r zj2BkUvvoB#Sv6pCsy4f2q2;ax)LJZo#*MF>kCGYXeud?$KL}oZ5XU4^bX35=gi2~W zv5N?m=9pe|?4I%|foM}od>8gNjnN7>O<}T=3{0P1M$qDEC7FqA$BkHUritHH7RE*e zmrYbPlT62wsZ5>IIagUZ-@(Bju}mg~d+N;ki>6y&uY=mS!02^JR{${|b-0NREOY%J zRn3zdqZt~7pZE+!K-Oq{#Z8%kaZK0l#)T&h*pjRWG{0S2a*3Fdh#ovCGYQs(;Z{ss zXddEJ_t+0^k4Jl$mkE}g+GM|XgH@u#6I}`)X`ZH*Azj;7cQvRb;vE+QoZnU^ni5M# z2H)EYX!V1CF-sjX=}Oz_VNfdprfgevgdUVJWkxKnoFg_X=?vjNrajzka*T0%i+b9f zr3LC-(u~+-eYmzG+@^p!zdt;VofhF0_M9kFYnd)D3R&`#vujCzT}v`>#4C|hGMTQ% zy!$Kv;30u{&2;rLd>Dvp8B7kUX<`dqM|+J$pa(x%rm|2pT+81NGMMhD#namg2u#`7 z?lWn;#5^}Bp)hQdS#}Q@!qeE%7|TS42!T~{3@iZ**TWv=I<{V9#PeKj<}7=D>o6%g zPQ@M`8Zte6d}b{O%di!j>8@l7?4&wV;2Vb)nZ({t{ER^znx{6~`6_a=XzvlicF288 zaD+V&p~ZTTnQp1X!CqVvLK@7<3sz!QnU?&8e(_}1SF0Wn5GjoWZ0&f(yEb@v4_&0a z3nV8sC6jIKg4;S*HcG8RL&^D8-VX++w?f4jI5^stD8n{DW{|_M!VFQ~5vj6~iqNT9 zt$+;yo<*7KcBhu_nen!inb$H!#>JA=%$H@rLr%n%w312sq!SjxCyg9ysm#p~4*&h~ zyEk5V^TtaL{r%J5^n;nVSteh<_t94$pS}L*!+-xGd~@QL4?X$i+24G5_DP>*pe8%l z5hb@3LQ2WD;g#q?x24)Iv3#WVDVB@ASG(~4K3MyFnEdH?$B)As@2x&myHvYSM=}|% z{d)Cm?UUN)wGR+mgLnuf@UePrEuQ!c4jv_lTeA8ne%%#>@j&Pj_=@*dA3t&8PoMTO zIS|15XdQD?)a|qJ3>)b#(Io1d5?V~1wOC*~%m zrbfuYLvvF(GBR^$dSw34kr^^OcjTUlvH1=PSBmc?yy7>t){{dK?P@(u~w9D?^Ud`|DAn+Ffp z=Tpa-5Bb`NJWt=Kbn{`@!Hp#?DF^q><1^Z5EESwP>9I9i& z+UJm*BPwJ+4ljJRW{&H{qOkO$FV3emyrv^BW}x zEr%SeZI2=P@g^@*2D*Q&?{pT1bbbv8lo^w8NW0(zH|?WM^QN=84Dt&rjc_ z#Qe`sU*Kdcr;>jgIL!a^(}&B%k(3np$~k!PpPxQStisA`RpJ!E7AySQ0QoM_SI+61 zKCKs#5VQnW==~+qDs_qiA-4(jYLTr7FsuhBM+T3LkK}xoT6kE&3^J3Yv%Aky7{fTE zFnVgFsmQ(>xmtIFdunWIVq|7w?#!{-F_Te?N!XhfRVAJxgX91|1v}WrPSs3LBVQBQ zIL5M^F=S^2J;|}fQpYkLm|KB>U^+|-eAh-wn@ISS<2sv9dQf8+@HO|jrh~6-R`2y3 zBs0RH>z|@KorJE1#LQlzrU6qWonrWah2x0Lg>-6qi%z-p_K!+hiN_)n%jx)B5VvwS zsarIb7|csXk+6$+qb5|>g*5Qk5gxb5uM#`GmIg)6cvFS=NmwV9(v&xKGmIuVR)ET9V$t#C>aPL zJnK~_mr;$UlD>wuRKR$C#%o!NM`rNZ^GoXCGn-2ANs2JvQaJcUteGQ#oEhj&#nD>$FH&$(ATHyz|v2{y`T3mrC*DHrwku@eeG8S=HR)tXV#u+ ziXwD5c?LV#(rw7EpNd?cAT&)$Cien0H`&^2 z3PhZvX7v;?VGrlJYgZzq6uP7l;|Y*ym zZ4JN$Hb~_nqG@}TxjTE4o8M$b;?XdMpQacsg>Tf} zjcJrw1#m}7RmTaRI5+@aFL=ISuQP5fH)lQ3+?f)A!=6aE))sLNO!*Av#8Hck9vUO} zyNr~nrF-nbbyv)xWOQz7$FGHZ`TZ`h)MG4+j5m6NRxUK9wbd2F)53CbCc>|odnJcM zQSs&HlD-68GO!1oP?bzuZAV~*HR44e|7~Euq9}KHp~mCK4|7?J7Y6!HcXO5nQ-w>Vh;F8KW{UEGm62 zdfW$o9$42V;!3uyPBw7K7OZ)XyF_89k!!m+zx~uOJ z%{j`kKfs$#k7o_WZ|+<9@gf>;^F;yo2Lk3M7&Znuuuf?(X1e*%;+`-`1eiiW@(QSa8$PjQmq6Y<3o#|Zx=8Kf=IOkoHb*Vf%X086 z60OD6z>&T^(9`Ikr;((iDamwX6Qe97Ls~-}%LB+}aF`vxWv$n@`)W1TE8gde418~F zXK@_liau#X@R6i0bXs9>L%xlvX@dkM1kECaS}YscKKE7++(dzvJ=$xpEZp)qUJvk7 zJ6=jH;h=pCcT$rg>>48)86<~H#2G3uQ$z{yS_Qw|FqrM^#%8y1W3$-sl4t{Syz@7j zFO)v7c*z6q|7`V{)w4t;wJWR7*FLOWseQis zDx7?B^-;2V7Oujs=O?lI83Fh&)_bSl-?z>w1m?Rj{QXoS6<1R+WCQrjJLeTFZ`2^; zD+KR2v;LybAg5|i*FIf+3I==@hW#{hX7wE2%K zsWy9o1w~|QP#X&`XaT2P3oK}z0L(x$zpq^A$6SMu!g9hJ24K9Ege?~QFwdImizx)` z9Lr+2OESJk@wriI$*JYRyxw*M^elFhB(&B7^uWwNwVyW-Pd-4lA5y3MX##`PE))2F z^>H)^z(Dxi+xx+m@Y)d2j`;EFPgZ}{_kfeQ`+MfjS+B3)qDR~Ffd>$C8LR=%i24{u zPxjY-?HN^9Hr|9!d2fervz=#UF2&*|&6sUgOlJeuOJ|W>O-lz(;%^a)n4krdvFctd zVw+gRV7xO2Q+yi^rYv%EW!rcN!{AY3$%rUOS}FD+Kia5USgw+3YDUR=%y4A0=HAn~ zQRw2Ua|yaaw78;W19Kkl!LGF~l)$iSgW1lJhHZtDP~vS4??O<+dfHixcaq2sCu*N$ zvgCDP?q77_14xmbO%rB{lDkO5?2FYiAvKZ7l=V`U=g^}qX zjOfWby}0rT=I7qki@P+DH%k|hcA3~Z&C7_axLrx)Wx;4rVr_6XyVi#qXLiXEmC@XI zXnAzSjf#VXkZR56?ILN6S3UD=T^evSoQKt+_edJBoy8)v3449HBNMqZe|zld6DP(b zqiBfX(CVZN`%a(6Fm=OcB?p8oGfC=l;LJWVIX5xrP`l(oRDZ9^><1p$cZ;mh#PaKTsmss@!e?)uon>rSBB3%MGKexjXY3(kz( zcjCk_8RyuP!a?&bxzb?7VlJUp0dBkiit_Yn+s)g~@FX=&@xTMUbj1QJvQcKa&N4WZ zb(S={UNqY5+-`kyI7e-max!WPVb~5LD~4Z>v$R50EuQf0YO%C$U-J%vJqSrv$+ks8 z!qA|6`OxcMKK$gD51;$;>{|!LKf(Y1=gQ;%^`mEQy!^_I7vB6oSAN`oNvvrXoGCJK zsv?;kUO0!9c~KRHDEmrgn|YFeJWg5Ur-`)%g9B&!K`^y*wJ&OySI^cy$3nkPyc9kr z@;~xOKQNiycY>LswnJukq3B>&_ymmt>^UdOdXbl$B1oYu{HaEtWvB%=4U?%3`7Fx} z6TCW@J2S(XyTmYR#VtxGYQr+6CFjHm0`r40tR?!u1B86n0u^13M?;OUgt$yhjndzYdDiov`jsR zpAElUP3GDq-&$2;ejvM8P#`;O*HO}Hm@>EUfXz3-!A>#V%tU=Xs%Bs^_H83jzd z3fSg(Fn@3|L(6RvXW1s2-^$7Oa6>ZzT96x}i!1T$9wp+2bu-1gQ@T)S<}uE~PyGQU&n$zNZgRU(JAtW&0l%0;wtLADvaauDA7gM*Al=7ELM8hTND4bi z_;SO81?mUR4?{iF5CZJ^@m3#s*BW56W1a!l3tH-iIMRyR1_-m)L=vuv1{|otHjzgVPJ~__AoV; zjSrPAN$sI@Qb{OUGS+AMaK0$YVD!K*iPR1{M=$|Iec+Z2&jk(@aY@|;synV^G6I?2 z$t7-rb`0hglnXT%UFxg4wTCArED5Ip>8V2tC=^v>$RlQd=D)pI<$Kb)= zeKS96GuIF{?di4~*A%U-zr7gLHCg%PNg_;xns*b2N{5)!dpWXLS zLn?{~U$iMD+qP)$e1$nCs}%=Kg-Y+(h#T{tQDSX$C}Dh+1IVgl6SyzVw(wbYvRjW6 zLT2WsU;xyPB8hv{Y}V(oL~_~VStZ-%Fy8r#U>2TzSOPslXRS{?(wbGwWVfVYt!gdoW8Jzr$AVNB#AXm;U)L~Kj zY1teyi!{$7RMgzdb$Ck2l&IIPuGcumnsvzD>wv;bD5-Sdu%rjpQ4*zr776CCC~&KA z%lu?pxMjXb`Hn${I85zgfp!SE45+czdVuEm*P||J6_FrdUg4h~fcil0$QK*LBc8ab9 zog<{>-}E~pa;aNDay5jRRMhAU?0viGZ)6)6OsebfqK7-b$J-JakOp0NYBNF!GdWV^ zF2nsY0(+8IKi_u_{Q=j>3v-`UJxLCmZk~)+;aWisS6BYHh{fcI!AKS5BPKg#utn}G zW(tWf*br7$iS1@m7VwFYe}VUGy{JgY(hiQkrOh@l!_X{ibsAS5Bqk-!%6W6;{36Hh z%q<+q19NQ-WWAQ|7Pyb0#WD>atdeopvX?~1hPZ)brqzp_;VW-1dLTgSb>y0&VsCaD zJjsNwj@-AOr~zx2)y{k@ft491z$-!uVt_>kKgb|K0#$4S;-tc<#gG8xlmphNKvGcz z|5)%kg!oQ3t=HLr#U}Kv=E>pNZHI?utBKJkSbgxKq2$4o9fO&pt_+V}IN_FsS@DsF zs6OCAH3+*)L*fY~dvn$5V9Q|j707vE0WHAq@-j|Of9+dS?kQHz3s)A11Jdn9WYO)VrNG@+o~~d0fdPfi)nWvgvmUtNp8Y6LOm3d zb-07W;EEX7@a6|Mu{KVJ7I-_%Nv1Q(NoEU>lZLc2O1zD=dF!a^G}Vh7N$(tsmRq;U z5#1s2eyn7hmI@hV)FS&gUSXyJKHIVPYPNNk9I=d|q+}LWAG3?TkZUZis4+#un%InQ zvkcuDqGD@^)-c%^YT=!X!CtO|@2L8EyH<7FYt`jqc$p-7#NSPr*yhUsz zH3yRfZq@5uL)HY?tZW_C{NEn?`IaB4Fi|nRez#tCBXucE0}q~q!T>XdM$iZ&vYeN$ z?QGHtiDm_OB47o$>T%k%1v&#iY;&0Wvc0w{kP$Pg#QUpWKeGp8FqRwt=Ywke4E!eW z`^1U6l*T*r$Ii?jgX0~)ogcORNCS8@JT0&Rvv9;3vZ-a{Mfk>94^ExdA9C0oV?Kq^ zRPFJA-lvMpIBG0%sKDFEX#s9kT7o#305q}Q8qG{tj|JYn5NI!U>{$fRVCXgI=$|1u zFB=+zq0zcxvzdcE>0rO8YWTewZR}0UFcceeLwr~-)!GXcnAm6<9!7?lX_YVwb?<)r zE&|3DSPk%yu3TAp7gne);OyOm7Qt@BAS$R;FQaYXdC=(6BYE?Yx50OzCtO)M=fRJ7 z6mV)rRnxKLj@Pv#7CR+;K?*z^9>H9m%cN6hfEp>UoX6GYIZHWLdUN2^v-T03rgnn2 zROhg3YMZdDid}K3^d4SF_cq!wFk^(pDS<_Afn<9H76PJ8DcLqqpx{N3SONG+?egj) z-tYGiuF0$K+u56E?d@inW>YBsv%|GtA^jv)Ev`OO`wUqry(YQUhvDFp)rVG}B-@ZP z)e9e+^h{VU>gQ)_6h?XsO%ci=YcJLdEmwa+YOljBT!BgV{`gHs!_%bUYwdFgtHYJ;)P6n5CMp09(rBTHYsE=#U?pu3j?fly7 z|2DAptXH&mrP*M7V9`r0pGIBRdLJr9S^t-ZGP(%P@^#OrI%ti8SV623eS zXWxOB`y0cCLBCGm+Na^#SJr+R9#12kMjHaj^%O1X<7U;8z6f~zAfQ0AN-P~Y{D$mO zbsBMlH3D#$t7qVw^_gNlDDoU{8$*B0gh6nV3mnxw!d^K~4sz*XK82Ik7+M07+Pc5a zH+Cw^ebCbA-{)yH20hZ;jE411Cr$4LVH5 z_#!v4zM?~tI^C>#Y?+gFK2pt7ma3Xkl0K$h)dy)EIMWo5D~UEBgfPkJIk%9nk|{S>Af`{~1#=9!spJx01{cr2UgDkCOMF|2bgI;5K~2viQgUj7 zj?XiEHB9#H+XpIJB7ZNAFE6=KU}Rdz0=3{h%-)VlQ8rl>)!^~thv9qh)#=liH#(zc zvWYY58BKfO0rIy8e*qUv!B{2g(&jbUWaf-^CiTDrn5ROH4;%Iy=G{Son~V~e7>g{! zV@2Q#=0p|r^9)_Ao;l>$6ifWe)Sx8+>bb-XrYN^c?RE>fzGZQ5HN6p{nX+V3sl{Sx z|FZU_C+Y_Qmo}=0-A5QVTa~gX^0L|^w!X&XrvErI<=U^k zbkGMd%7?tH5{~)<^c=nH&^gbeya4Zf3W>Z)ZNGWZRH@!|Vpr-1h1sEC#dK4}Nop z8d^|9uB~RX;cGbjpIndQH@ARf@Ess~Gi-Xbi(n^X0_IaRxi^5kceWfhF9kK5NP0+e zwTq6jXFl$pGD^mkSX)KL zoj>nQJE&Ta!b1S*T0K+K2UG*+Hj)V?+2#W7S`?G%l@Ifvw#BT0Wz9^HQx;rSJi{=rMMsFYK+hs7$spS*}V%7GpW=}(m3oRD8 zJ^>cO6klK-1k{qK6NS_}cyEs=y0sSW5f*h|{A7~KMOG#ID<-WjNd)5zzd=kRoR`HG zsb#C056QAAp5W3+z%J9;>AGtztlZJu1o1e>$ON_uR$%EcGS2LhBPzYHu|Pm6B^i$# z4cyCuw8?nFq?Mv5n8LQ*GO;Vt)me{HAR3}i#A2-t8^K$in6=0(*AYYXpKedGKe7y( z*gc%M4%~i6T%Yy#h7Zs@7i1bSUCi)(r|kx=Z(XP4Z@FX zXyJfST85{aTRxPgY^G_rc&`HrCk`G~WYeXN(^M#M*u?~KL6J%?XS~DrvIE@oX-+&E zK7e>c6@f7b2~$gzlNA;oJf(xqY&L-=19pLG2-*;fw{B>GA)+&?np(%IweKSPJt0k_ z4Yu+cIVr#iHW=bB)S`2!hzxK%0Kay)_uRTGo{pjj@SKlJ=-Xm8p z!omCG>Tj;Td+kwt|6%wA=j(6&;p&I@Ccb?8>PPr}-T_?l?$zI4d;IF#_4huy`qtI= zuU_mF${lSdm>8u+S)c3PyJ0-hEU zhGG|e%*1J50i*Xim@G9FoV&=8FSk-aI6d$Z7IwVsT3A|mnIPpJ?8V*9+8vJ@rG zWf`D!eU^Nx*`1LJ*qs4#(3DKN)rZ@2vWE|H7gmcN0obXNgPrj^*qPvq1hl5Q3k$B` zFZDVyuceiETdiAQ{RG;nT&6pJUunq=PUlEYCHvK^nh39ua*`RAOKmd6;GTuv6hLUe z91R<6<88?7TV@Ccrl?yfGQJsf_W1EhZZ5Fl`&^sixdn`TGyr$M@50u|tbZN>-8I;9 z!I_c!PMjDf=kseSJlLa#8 zy+6>jB;yGsuEaB_5q)dhM*yS-WdPDVa^`Gn>?-i;uE*c7%Qx*aC2a)D@FV1I7s_jd z@U(Q|#0dp_`6Kf?-5HM>wG5xyHeCk#n;bfH=%!RiT)mm(P~UEzV^Y|*C{(Rp7I<{* z`P~I`Pt9t@!9ZhA+axq>%uFO(BaLA)GowP3`l3#}rAD(@rUp3-V-BCS{J|14R^Blv zIkH^kW|{h+%Ldb#j`}B^*;WXdUAYM6xX{52I@Qw(vqVfuW+I0db^}QHxTFh?&*#F! z;n|V7d2UmSlxcx3b1QtAc~@b|L5`Y|Y)dik+7g_jdAKJRyb`$xHaDtYYdhlcOa`%{;|`)6l(4=9vxh!5LTrKgi$(Mli8eV&M^# z+UQ>*n9qrnqG`do(eC9Y61dmHYMVJasRYr;t!TA(G^)WWIrtW5aGY72Ky|<$#@cmF$@(%OgQ?CCE~#xGTr}i(PPO$u1x_V{RB&W!nunWBOx}^8){~UZ)h8yIB^|rB5w_hrBjgC|Nij!eO_~XbBG*!jAb|sTk z{LEINS=BY9>Q1&{F4&798lPR8mx$1d#PJ{&`_HGHTzRvEweg^gZLonp3=e@ZAu;po zv&0iVgy;G}KxxE5O$9Ds8pu(Tp;@pQZ111vCdLoVO^nSSIx^!yD@ri;@+NZ`*WDN|)7P^+E9br3hwBJD9F<}T?*DeeXm_u)OH&DTPmz@$gZ*L;9-9wy4$kYCR#=iS|s-%DeIz7BEDM6!19PInKm4)T^m%boYtGbD##`YhF@#xQQ*Mp9EWX6+C>UCWs&`()I53Hf~H+7k~HD&$)XpW zl4P9eB^=7MS7N|q_|l{aGY=p1{=$0KlMewfApBDt+fB1=_5EQwp6->D7EZEW`=9}A zx$Vh^7%52+p^pZ!7)NpDq;HA%c<9xjMD%fN4$ZbThlbfu@Ltx5q!*+pdp(6Z7EknA zJVDk=2Y9q&tYCs3W04t#*T2AL!zDY%rjy&mrsGHut+oLZ2NumV@*)(lv|9l?dz!+6 z=hTGD$$si$#>*<)b>c)L+d6sh%;donC+;Hq;|l!M_|8{!C@KP;erdDxyi=fxU=UhN z(cqsNe*ZaQ7nf+UDZaEX8l4UrjE-YtXJ!|Y6^^@3k@pbgG47E%XIdwIuNHqecfZS| z8{SavP>G+>e3kg#K|s-@l=v=e?VF&svm}Jk!ypdKeOk#TBNK=mKkn_0<9Gw#cc}sQ zbH;xFd&xYtOXTRth}SxcML&b2@u^&q@ltWnH3z4u?Y$<`qKMfxgN)=2*Cf;QBx7(; zEW^)n%9p4~az!fnLVc6URhSO@*YKz*^wQAHeBT-GD|UX@?zW~Q&Z>7YSrC#?f7vjh z4lZ|inuhAnOvvVWpGMSwG0Lc3j1cQ8QQMUz z;iIPwoimPml%;Km^Si*{)#1{Lfa%w{sMtMhL5*V%q}rCeyz@nHqylC*45EjCB5`aU zQ-im~4;(e>werpBAIGxMg2@TD7GQQxbt0VG(csNFhk6jr2pBOs#RQPAghGuBh?!&5BD?v~=sPA9xe-OuE|BwGVI` zsa8yJg#ka>Yo8vGPX^>f26&*Nr|A8cOZiPvkFC%fM4?Jg3Q?ngc3Y_0oCTS(E zwe_zD~-GYZ}W1G}G| zI6=zzw?Lko`rhl5*Bi{&{Z*Zjb78Srl+4q`C@s7>aQaKdO}2Ma6$Vei#Is6b7h2yr zR_1vs^RC&;cymHkw6=mxZUo@hH=T>1c3$ou3prs=qz|j0jYJq|2CUWNs1uMp+==CTwuCb4SME$^u#9PLWuGuu~P2TJ>ft zZXx%HLu?m2cB*~XHQdr8f~ccMmMn6NR>_eF8M@YtwVL%Y%x433iNukZFq_;1>R<0e zfK@CC%$gSA!B@Ie!j{M=saUH&7FZzn3Y`;O$T}`~pRI{GZU`{CIpULp=}pQ4Z1C`X z^drl38POa61=K{L$mFIz!=$g1D)=JFD(SY!cqM4{njuBr!m1pDmh|dLi``8cr^-xL z!Op1a%U1Vfbz5w`DF%KOsP$PIAzv|K8FHDO%=${+cLOj1))(evBbVGR$Q z*K%0L$KSHmdg^dnxG4u(Z;D!_ozZisW@0fL%nznhBLX*=swEUK8?nrqUoWA5KW3X)Y!L zTZ_q*lFUSgR>w6YSI%Q-qHd=3TTj85No|{vDMscc7iDn&BJ>qZQ&)1GC49ly2%QDV zjH0Dl!+3$W8IZ(LDU4-qae14`JxETMUe`)C=wpcUaWOv3EW6<*ks%)foa@l2H0WH#9{&4_Dnc|zKMy@phJg+ zgel3r~f1bh9Xq_HIH{!$z!laFRo-X@N5H*y!y^~e$^6LS23A>g@_+<(Rb>*9*~IR&7{NjKHjm4%3lW0Q z#C-|u-if!lJS7oO7~fp{5gCRh^Ua5keDmgu!~gT#EB}4zUGhKAKJv}ESID=|UHR`z zzbD`R^1W|=b?Mv3Up#T*+h0Ah|C=|j+|~TV;n5@Ho8SKazdyR{zi$7xuUxsy`-&I; z@ZTRl*jMuW_E#5SB;Wk;N@#h0``jbn{`T>2ANoDasxL96cFy3c+lw`;#M-)imy5w| zzw1vu{Plwme|_OcH=ak+k9_^k7dM`}c;mGz*RMSF<%4H${Ospn{rK@dvq!F9e)#(1 zFWh+kk6-=x$DW>M<-AKBQluu*qLMr^0&%kM^ywM8m=~wW z8Jqz^zDK6veG~M=8HZ!2`FrGuG?0oT&n@auuqoh6tFG;G@S>~UF|~P-6uIG6@I6<$ z4qVQy1xZWHq&8f429X@`yH1}z?Y%rh!J*edCbVL+2OfBUoXHELs$<|4t}0^dAUQ+x zpl|S&zN0A0zFvCj0fIC03~7~p7~qW)!;Se2{pd_mi!YE7lT5lcJdnORH!YiOK120` zfKQR8#1gHJVqm3v_3dkqUwsEeyGcb&n3dv?r++=&d+o9R_{p`i>tu}|d&>BctLM-7 zuekO^hbzc_E!7#PaOc{pxcgi?3v;p`*)4HA;nQsQl{UrSeE1zf;X4x8rj^b_E)clW zqA}CC9}JEk$#o$X6QNoIgpq-ncGdbLI=e zj1qe8k~1EU-Hf8UbBNf-sI7xJMjR~bLLf~?GRwjO^hF`Lai)vwWrPNKas)o)0V?425tHpStcvz7Z-#yTyy0@i7bG3 zVXwUFU>OL3IdeSUX(vZjni#acbXyabcIC&u(UrKR+st-W&Ph77FLpQCmmG>|gIah- z?-56A>hx0M89#zs{V1RSI~DEhS|jjCZZNzwx?}+&t>!L-H`*vjsBp6YTWb&Pz<5?k zw^d9B-i5qB*X-n~4o_GGliNkY%DjMlgdRC5%p|#wQLK~N*q`b!dSl)1*1ApA)*{cO zLnaw@U`LuOiYU+{vs6f5s4>yB^AdsWyo9Qx+mbGOcKl70G_ndt{xJSukwTBsjMpU- zJ`q<^iPlKwo@j5P?!M|2vC;`NlV0}}dL!4QqQzTPt`ju2_(NuKV$6mAJMWabr(*M{f45Ezr5<^sYBFkO8r;z=l7P&+_{U8P3Qmi;7bN;ya%trIV!tF zxXH{QE={nHCy{qe`wfBmx? z=N`ZQ(~q{OKA%4}&>|N6(j||`toPODU;Xam>*qd(F$57wb2c`5Xf#EI$2x;>%Dx_& z(A-@KY&dJr00yqH%{kJ;eYe5G^I))cbd+RU(91ct$l20R{TD)%ka@Vo9vbI3JQ_ao zb&NAm`o5xu&~PgWq`k|;Tb=JctL7#G&n|4SR)L~fV;OwG6umQ9lW|3B%jg7NdY}$x znCmr0N~BG|G&@$J&xj?VFEiVsvs3+mZZL{)*vrTu!R=DLzNjg#4sQ-0=B#3uxEfk) zz7MYJOJ8G1GI(c55)2A0vK!gA2E~M1K?vX;L+-)!gDH{PgNE&%Qk-?cL0P2s{VrSP z;p1!FQis_N&|-UHy-Be;MS+D>)aubTmqb>vOdCvhU>Gb5jIk1{dPwBH(CNK`FA4+H z5rSRKiZX-koKtc{*SX=5bYcq^h!g(56X36!;9mo+mDC6bwlM*rWqec`XbwqQZAv+G zjdT*3%G#OE11E{3o0kH7)e0T_zZPI>cf8Hg;%LJZ+5#WH#?6Ps0=9M6q!m9JQ?}_U zWCQvNN$j0Vwm|^i`Bur4aMAwlC534cTjjJ&tBwm}%1?2NGcr-lV?*iAuQ6|}<|#01 zftd!`KLYb*gWIeZO{N!}2=HWjvCPb6f6qp|iN1(014<;Tq}t-fR?t-4kxB&%U?@RA zMchHipMH1x^dS!=xP-JAPM(^kU127bUWv2yz0;?2hsGynMy6-HvSW@n`8{%G0lwMk zaKXaLWy64T2jI62bLQ=3qClt#PTUzao5`F}wKK_$OKpc5-#dPMUVs}^%AY>%tEln3?UtxbNv_C+L#^;;s5b-TfT+vHr%Zh-IvadR ztLk3egHJkqnOVIld`MykmX?Z~b_eY`Cg4*Zn`{3u(T+2#Ol65oa?>P|a&w7RiCu@l z2`!Luy2y<&YWcNmQ@I9~M_NgxcAuqz!S0G9Dxwf!=DTNWs97bJY4wx$+VXF%E&r`Y zzJ2j*XrXqmEkB8^SjlXwk_*$8uf<}C>mOh2I|cXEBj>*Q@OS_9@~2<__WA2?Jpc8D zD>q(v^TsQ`zy9QNUp@D`>yN+i^||ofiF2ZAQm)41ezA1sNmaMuNln&!oAyvkTR6Bg zy)p%~gpz8j)W7q$o{e0-WV$OCqmVyzZv)a&N;iDgEWcz+ms`z=xC4IFH-0NegJ!a5Bfu^5@*4kJ0oD@^S>Ozdqbou^qL zp=#N5Xpznc>A=&>Buu9oi}?UDRW0t5talJhw%Ci*)@v4~(}gTaVHJqUtrFG~Bb0{M z%|Q;DLVDn3-z8;lvUj@7TH!PKf+MWEi8twPInQJROfuv$T;e)Iq*LmNpo!iXmVYgY zj7Dm7)|K!2@);wxEp%RAb_w2!gfHT0 zC6?TMHsq{HVOsOD8jbq3d)7-l3kzj8+mE~ezj`QugsNn<-3KLbfWbudbl&4&F6_Q} z_dWO(Y!&LI1QJTNEvXY6spg$)j}i}U=G|*gTz!{Z{lnFduKtExdlY`Xef8sOPuxxF zm%MrP(zQpfzJ-_6;l$p(_IUlhtMA~uAA0yM=gHOguRRWzy?yl~c=uzt@(F^Ug3JEc zLBIotzCX|+;DJMu-apVH;+_<=U;WM1-(7nGhWb`$@y3y0G^?aCN~$T|(8bb8?DI$` zTQM^O2gbnw`=94t{PvYc$hVLEv>ys*H_!h!7cPGDt1I8W`m_JO5FTFko8SKan_v9C zIr#}+1lxNZPyiVvk&TRL87?g0!8J@V3r@}no7(I2n|E4h!Z*b58S&{!TK1I{sjB9S zxq1h|1e;WqWLs%wz7SBbwO@O5@1ND)TRmI*eDz1Q4?IfVOQiP6>JzmqAWZLy_u%~H z)rXK6-s6h@p!PYz50Tn4wa;rGBgwD#(xZ?+i!{G|=dRX%>2dj9#?h}n^bbZCTHN=l zsygdi#9?VclGu@uink^^0+%t>ad1kGOHFQbCK<`;#U+M34iza*vXX-_ss3_ZaVp8Z zyc3i5Kaj>@X>4*-BZQZdNkz@>Vtdi1Ra#+IFS`wr9%)hZY{y<%dYgD>uwxJg>oCiW zAlWWer6C#a|tt7Zg#;I0Tag`7&6*q(5^FI?L-^(e{bZpl@HMH?Be z7)hVuk4S8I=qh_$6yq`Ut~gE6kQytp1#-wTgvD%5$+TOWp+(^mAdXAfL zx>MM7(B8FW(BLA;F(L}~%A0*xx0I`{yeHkl%H;^_VS<~CSvhc7>CMf4MoDM)6kQkl z!?9JY=-LF+sIY-c|64G zlLH{vKF5&vfBfXyV;JcExPjdM=zshqyn{z!A7{=@X-{N!X(Kq>wY8lGb3?sCH%}o6 zNG!WoX4??OB>ZJY_O}fpe8oh!#FJGrTB~>hNA@@Bl%{FHq7|=Z&h$zU`bl@qR>8A% z%}5!{F}mQu2lb{H(#=>@I$$g+HLR%}Na**>y_?W^)ru`DMQ$SIxmR*!_sqG~5{lN= zLx1~{EWfUO3h!RQ?;y2T@yk{phc9~&zV4IS; z(a^mGTzeIU^%*9JKwv2F%P8cFp0qz(eRB2D+a##SE3o!+xMi6B+81#97vX7q$|tq9 z_l9fd;Gmuo`-HT0_V4tOjc3b_)|kw(R0h-lIae%dvFhXMj_+EqNYJ|@Rf`>$XTd^ zZMO5_V_E`N-tEt?-+pK@%^^{Go_nSD>qcdB@dPawNnxi$A`!W*`}Aoe|G)z~eplKS zbMlxFRbho!T{*fWdh9T81!O;ac`k2&S zgcqNCQ4@@b{HR@M12gwF1LExyhO`bx@+P$%BoS^%G6^=VqNXxRLiH;;zorYU>Jplg z-J_knHQ~Oz;5HVi_6`{}<#nU9t<^ifzVo0kM74KaKq>4~NJZ~cn6FkkS!)9$?aFY( zwP9_UJ47Fj3n_U)ketJ4nanky#HeLp84-)XzN#v!mW4-H(bVis#nF;6C6?L6%yzaT z^pYI`db@jTBB?3Kc&jebgVrK=(?YlfcjZuW`t%sQBP?5|&I1pS`(1FREz_hcZ&5nM5dN7tO#*XCfR0* z`k0RbSC=H?N_sC@cx$YOva3sBd92kvvmTO3C7p>(i9BBa8)SMG1!!U7!MU3(QJd%# zCkTUKjV;3V_J~O;^TL(=mbE_smbyd6Xr5cli}2l}j4IO#cw5q07U4s$^F`{Ad%3}2 zA1#UsIq3S3(+A@|-Rs}|;QY3tOejeIwki|DVc|*98M6~qyqnI zT3k`HKFd&z`8cR!0h8uwB#=+Kf3s!1)~On=u8f(b^&sKZ{(@%p;S!#%lVdSrg03#q*(Wk*29jx-W3 zq*ARZ)=JP^cgZW?q;h$OuYvI@5*eQKMGn@_43AbFu?Qw3{9d`R6j}(!xNUKoi1{F| z)tGOyi3Q$m)VLCFlf*9C^y{7JtcMW4AcIr~-t#bwLgUPJkcJ0+_r5E~8I!{H@JXk@ zZ4YmFjpz=ws#>NggG9C|1n+e~DWsKbTU$zCyK9c!?=pC#U>r6p^TCP~PMwCN0j>NG z0`h%`W%Lgymc99~Ju*Lhmo#T2w2Y^#1phGMHnD_547%vI3496HWYC0bdf2lTBQLrI ze(F!JdEz=Y1;YB%Yu{DPg@66iYw16|c8ui_`^Ns$YX!FQ26z+F+3})mAplyrB1;i! zdPkS??y@f+jBR>vZN2U1E_tw9L(N6dJQbI4$$EP=IY?lhm?_ zbknHD*8%00P|}ILB9iVoN=INRIz^!~P%5TtlZsekrr9gr&wh3O!|SiTfBl&!5CY)uU;N?5yTAGB-N(Lu=T?OlZq;^m{iiSd*TZjr z_0}`j-@SbOqc?B7`R?^set!L@4~EYv6qyCOspvBEXDlWZGc%cD-oR5IzvwRgL>Y5_!-E4xkv z51{)63u$b^h^6-Svxc;OFgV2vMF%-(!qX9TNn0jjirSXUh(tv+?OI&t6>1V)NLXOB z&YR0E#GT``m3wGPz|tcH0lT<`--Z{?(UYKViXI{L9lJ?r4h3>a}9B->M#r8+CXRYFffvcE19;+*)V;yN5P+j zt?MwX6GN2yt{*K2?+n`P6e&|FIi3r>9-I-&W@gzZh_J1)X=Hy;&_4k%D-eXB*tT; zB&1{|$CbEw_-$-#R(gG5^^l~RpCX`GYt?pi^{3dV$s=^iyw+Vb{ zQF38}dsmJ|KiH*Uc`^mYTI5Q5o%Pw4G|I5ifsq-p1e)8-yXElP(ijum4>qG}ot2xm z?Yw7{WLvB2!dApqjR zphZ9f2PCyTY4*I>R!8~;k*FdTXSiIglP?am^E%<~z zctmongR4(nLxSlv*5NBz__e7z%x7f>HnWrPJ!)hf*|*!&EH{P6Y{3m4p;NY7dsGG~ z5ZXD`7J|-ysn}n>eQdrA$5wxkf@>zIgDq+%!oDQaa z)y@&%I5XIP5o?3vyz?!)W&na?5H%7WQ1{F`HLj!*fvYSBCWV=2R4;CFlV*fNs>~tq z2X*|59Xz6gqir71<}Ac5@QlJQk9)v%Emtwsmalf$*9N8z&9r(#Q{2*pY1u_CSs4S zlOwsUATo-k?rN1K3%v}PX1tCGtyN#`s-?gdu*k?US+!TIKvR`W8*B5Hv9%6%H*6SS zRKboRSI#f2oR`ef;kfVeZ=)2RATBA`jG;wwgqLaFSL5sJ9}3z&l<`_5)YioinBAfF z0a^X&>O;r|PFBxuU~gU`4T_H1M;^P!j~b*8mx+gLdC4QVco6v@8blsnc*h^W$30$% z5AfU-_~1kEQF!^m>RGtp|7Y)QyW6PFz45QIPtIC5<`j=LBaNQU>3uKTv6B!xacz^_ zZEnBV(rimR8qF4aX6(qM{}qz(&_WM|J{;P@;o;DRP-p{53M2v6LhzQ#uj-t>I;21; z>38^F*T{~>8rhv0E3p*^l6W-IjJ5YQ``Xvz@Anvi+r6{=Ec|>33U9-k$T)$NB5?RY^NL5PaT?!!0DtX=t6%@x@)_b?c5U^mUhT67 zX+7n{MUNQeoff%9mr0K?x6vb6M~7zzGijfuNM^hH=x~11M-)?Bv3JaSDXw5(7(6$} zpY*J0Us3r$lnA4aHWquOMkQg(^+KUd!-54-W(SS|~zSoRe=>HfCu&_dM|Ezsc@TBz{rh(WzxR7W_}^qSKP0-Y<{ zbFKPG{q1)p=q{x98JEl;12|XFC9}V7VynooE1BOU&b6~N-f>9jq?Amc_3fU;5nsmN zW9|tK5?6(%Z#0SEI9a5)v?G<-6EP0i-uVAk(XuK)ef1^)ed?}Ndz>|J-1{Ffly&{!;8HM zFyyUyW*XZU3e|Nb9Y~4Y{d%osYn2#$Zxlp{7T{;GM|v)vATmlW02DAE7W^1o#Fe*K zeslfFm5b#1V=M2joV$K@<^7d&D{l}ueC)r^!@rM{m5bLOUwQM-Pj7&=e;`=>e?WS5 zlG|SU-yX#0r=DILMr3KJS+J%OStSiW{Fs-j-jy)XP9~B@GUU0pRrXe+pg=I=Er(|*`{V!Vh z{B4lIbssdiT}0ANVKGGRH}9j8QDQJo)IX zFuufz0}@=|HecX4lQWe%IpEMZkIMuTWSq{=*ZpfZGR+n|EU7LK9HA}wny5*-NWEt2 z7U=iR)}5bLK<5l@wAkEj|vyZJ&;x;ctH%4+vMMp#wMN74X=VLMs zExeqP&F((P-x)5xW$RO*T-UbOmk~7)1)tEnhY7M%VBFkCfnJ0dpC!ShAl%b zvE$^R`{}PN&LoS@s~(IMwYQN>=zT;@cL4|uUF59wf2-kyDe4B8l#>+ZHL?%_WzOx# zKm5a=RdRt;^SyB07RXH9C3Jc{XnDzO@Itn!j39^g>&Y$kjSiCn1J?)(ck06Zr+%nh z-xNcBR`S|W9%H}pT^JWgE$3CS=Q{J$R@i+8fz(<}xU9m6ErbM*lQ;s~@T+XQMHAT^ zn_d0lkwu{mCp6Gns+QLiYCfAtB=YIhsG3h{T0S>C2B$~#F`rG)wg^JYNmp_K#Lp;l zQgb=2?*wuHJA#qa$R^)gO%?=m3e;rgJl5~Quo8ptPcE{|_F|mGa$q$6m_W`S zMa_lRY8ge>V%uDmz}Ob#!kEiy{Urwnr_nIvGHA_eF_FxzUq; zli09tu?1?`s^-J6$+V#gGa;Gga)`#S>7A}+44?vp?eGk{JG1lmu1It{Q3P^Wgso;e zYj8p*7-0I3G+qOXrSXRiA~NF#0uh;UZlTWD5fPagUX5OVlhu_>8q6e^Ie&)7*97Dp zhZI0kN#r^OPz;-t^QMEe05vFi$H zyP>N6Bg?wX!B47ZQd`q}8AVI&zAWEx*6G{$4_Jm6oT3%%%&|B*N+s?hYA&sBRQqu#Up{N1E!zk3)xAHLv8+=4avqf6bM~3c9v1|(4CI_jrWC%GQT_=I}4>pXh z`;=0uJMBYi6V`_u7P4=l#-e$t`62vm)gBx>*8C-6X*4f4FE&5KiiPE8^81+-%!7WM z^f;D0c5GVW#ko1se3y7P_{b|@_|UuR-|;rjHDBuAM(fr%|5t02WOe*U4jhOqe`#uH`19Te6Qt2rAOX~VMtBW>8ljMR3SX?wX{X;sv< zrP4l^$HhmBAS1U*49>_w$)h7eYid$a6Plu`@HguVsO8!j>RjjyHP*Xlwvw|M=Vn!u zB%S9y&C$eIFp>AANw~01rnyyGq-J#fweMCf59K7jSFdN3lo}g|$Qf1QSa-u*#C8x>T6}pKiXnjb;rnKVFr)iG6Vxg> zaOFDe`=})vD4HaAps`_$yg(X^*l^b>GrQ@o!;=Bsb+~$4+NNT}`LS(6n1@(&vaYPI zs9?{g6ixT(si?a?${?j=QoFCka6hbTZisTdW^Er23Qm7oNvC#sFswkU1=rr*b}6rC zunP+Fh5_O|lMLJ_i*{O4V;-CX-zbnD^bxV|=XMxUk!3Q?s+O zaY>6v_h?1$`0NtQk4}h&e?YbYAezwC{IFhTU1@Q4oROom!x+dku-~e};U^eamE!T< z*;RLDPak1eEXEOH#&K9>5Z$flCCRHj136!wUTf<#EwoNk!9WO5LGG$-@d+4%bXsjZ zF;>aekBfNX zJ5XD_EPlj8ODb9*=No$Rhv9@+^sI|TChg;#q>_3LHuwagF6bn>Atb{=%Bn&=+4|85 z!?60MQODHKaf;{*!mtE(D^^HN!f!N|JcA6YM3HjNt+=0_cKgf}7@l{KSkN@{m)*=SmghM|L(g@3vf-; zTa7OKgY@&{(09q0$zwF)jvu1qWM+}!M2s-H;7<~0K0JLbB%nS%68;lRT@mil5;yp$x9MrPi4e;!y;>s%7bY~u+5H>K{LhkmFj4%CfBn_M7LVU8PW4DNkIxeh`Z?cH}EHw zP*gpD2pYl28=Id;6sUPZi;m08jrw+jK&|7FTE*J|q!w9K9+XImLuSMRbHoCf0ev3> zqbr+M)Lc%{w6vn7e6`o;2e8HC!!l zQ1O%4blWB6dT^pM<$4f&LDQFIRaI5VX^N^P6*c3l3DMhCQ+nv4sTABKkPF{fCX*EG zACtobzyLc1GYNXG*pDKalnVNn-NZ3G)hUi)%P50$K7~%YUG|A3fd$f7P6(!%ujhTR zd+c`8&@D(Mo9>i~T@Fk8Mv`b;c%9zI*}syH-#&Ki;M!}NKPV(Sj+s|UK{N2Vxz8Ve zZrum=w|v3l2z?J9Ww3+QL{)t9giA`^=7r%$UwFi%4j*c5YkcT-?Q6H|NobGTr#rh( zR%GotW|XKT!={OmXyHdWQhgPUYWT(KY2n+3bcAo4^TKsVi50PIs>GOA(%DCFJ9i5I zC`JKZ3Mn=_i-ap8r1ZYI^aRPny8Gr2XTSON^ZCF3?1jHwevAD5r)R%;eDUnSH^07ifBTAqBZtX1zj^a-@4n~%+`!EjuHEn5;`!hG z?fp}Im0|RZh+@zXB#U7qy zwlnt@0VYk#7P$S#FB|ORm**w58s{jtKYsaJAHPhijq@g2blwxPnrxg$`WSo(9y%$) z%44L;OBlRqAt(2foTlJnUaXK|>BR9F%Amwp%@^|2+kMpDaYzwm6;1Edx}i&hzSY}) z-8(b!6OHr2DkHD=cU)SE(|3(=KvEG_rkeTe@9}lMY#l;|U=bz*oCx&TP?~ zslq!k9FQKFGio=M*j?m`Pq7NON`r@My=g(oba%run~ET*WCO@5wXhcDxZt?OP(u(w zBzznU2Kc(vV_@6y!q^n?b{LEw_^1tIldU4RDwW{U2?DMi83Q8^e?DY!_z_undxxhF z940ekN5-ZP_W&N%GL>0WxfZa-J2g?7z^~AzrJA!6V$_WZYvLh+fPyWj3x=eIy0RTs!~e9fuQ) zrop<1O{&wk>*7GI4AUxVJ*)eSwJJ2&SOUF`(daw0+;%_xk>`F~obU=x$f%IDRp&Ue z5IP{cR38X_XwwSWQxctJhj;(8zpm`INLw*jYg?;tRcsET3+^gMg7p zXxY3#mLG)^AGR1Tn^&3_FwNV1t$CsOe)DRc%!n$6od#RuvIj951@?=k)U47;r|eb> zCq0gmlWI0mJ8`o45?10h-|^@(A7B3Y@_&%#E31T=SC@Z2H-|T0f5N-q-R32%XY{_( zW#WA=HA@0_d*)>GquUqN3)PVlqg-+GIf?uv9Rb;r!v`mQkCEnM-D$lD_b4S3N+wWE zyz7>=3lDg#)JpHxbW&3?fzWDbh^k}9j!xx=$7Wg`VWX148Z`(GmMn6oOzwN+5!)@; zPJU+U0Qgm~lQ`sLvCN7U9LTpVLH_NZ1_udPz|=%DbEv z!~(p8MZK+kW?@!!Oz_4@ZWu7JRVI0n9>@Ddf6hduE=pXQnVkPQOKH9*Efo?j5T3OZ zEc)0jH1|SZ0d+Dess+g$w@M0<79A$+7yt-8ih(85N-BVQA0tQ8XI|oHS%QyZ!XjZ_ z10f|7FqBnV@rd3c><5ZnEQ6k$_qTk;sn1fe0VK$yVue-ghif%>upH`j_j=8}AOH4| zM-Eu?Fnbl7%nKy)Gwm8PsU#<1daQ3h^2p4Ay<>-lCl7u9`{ZPSPv`#hi$9(J)6f3&I{bW*{OLd7$8X@}%kUFVY%hRy^tJw{7rZb2 z!k>N$W4Z57KZc{9L>AMZegZS_69V({3!jX%_KxZuhm%lBQ*z19p!aUdI~*5|BuA-j zbMrVuno%$X_PPab^`agyDtQtLBneAqOKtX)w0fxshU{B=N(~`kxr77LwZS?j!)B>* zUU~@d#HrVIeuAP?(b--Oe|5hsx+cOf62m1$(0kDZH;I7X3*0n)od(JFmYED$sjgz5 zK`Id62<;24t|UvcKI0jkE`z64MPy5Kj;3jhwX+fJYpjTCDyOK4T@D$G|J1e@@<7EB ze=Sf|yCIm{6WP^K`I5U=bF@VEAE9(XF>9u+h#@M}MZfasrhj{i(i5)`< zdz8aA)3_)rlz_Ea6DIE#z1cGp=w_S&d)P*3t|Njk2yCI_SkSYE%I`H(C@Quk`uk1W zhFfT>O;0MRUATGPU33B`WJ=N*E@!`KM>?N_5pY zAr=j>XtifP-Jbb$NLyP2^Fc`k+89EIX&$N>2(66?=^`eb#R8KH;MI|VbczV@xxLf& z{cC+;`Qa)xoid|j0Os@le?D;Fz)AQ!H+TOX5@76Q6Ekf*io+i|ygDtBWxRg6{8_B- zNvusJom7%qr&lH2@J4#`+2(slIE&C_%V(SKEkA=XQ?Kag?E3TlWuY|BH7~<={{V}5 zmY;&7OEAt49w5!<;OCWASrzdLt}Zt}@@RWMY`z1-d?K`~tty`0IWQe;1D;uX3W-Hn*SE>>_x1~WeBrWi=I z8W!(o0>nidBWjIs3ARV69E8)_&r9V&yM`=_-i$L@bfrWvUT~NYka`USZD9Y*=;-ME z8^jU!jRfL|4_O?`lMce~HAHns#1lt18)28g;>dyN&dB1rnOut}rW7@oQB+M;RMjWh z$niw2)fLdW+lJ0D9%3$&WVm1#%P{#4K`K^+1ZSoV^IdWJqcPwg5lt^r*gq@LSKmQV z;jwvcW1?lto)bLK9hGbP+;Sz(rYDv^uQAf<5kcz-Jsa)FmOK6pR zgRBm=ZhQ0;Ym2GNyZK4*P^np{FGXKhRbJ#pUnW1@?z{1hL&__yTB;<&X^Se;(mxJm6YY+AfZHz6zEsZQA}H7jlD;Hx|?L#&A^gFDJ>XAo!F3mLS= z2w!=go7FA}ggqlOvjHFN)}*b9F9Xh50;@Nbb&j{{qg+!%lCEi9>3yJ1CyI*QCqbFo zs7g&QD=;M+ugx>?NSFFSvu+yae5g7v;hZ&eP%?K3oKD80XS|2G#W?lnyc0f*q(n%J zBn@lw_1splBBz-rmLhzaevnLn7_unHgS^@}Z?PqEh?c9=h`yJx6G+!&*=&MQUzxkF zCn;=4l7@^4nLvO!wA&0@Q!&49`sBW8q!h##qlYK+Dx?Z+eAo$&Pd6H2ZG=i#c(NIJt|YQwQRUyf!Smc zUwV;L1b`w2w`(TtWi#xKVgI6Jv~kH{gT{S_B^j6;{9bf8wWAzq3z)AIIY!JaJ-wG0Z|NY%Bg%rso96j^B4_9t6*y}ig8xTYo7sZy^D$QQd#wT)NrIGh}H zRmmk{fWVH*g-iS@yv=@3VONrN$qjE`c6 z6^yFJ9F7=pw@ibF#>R)E3#ZHmsa0p@mbMxl)RDUTfYCbE2nu9dmq%7fq;@-I30!S$ zBVb7qp~Dk_eDvKfg$Y`sc5xf=0b%AfB^Qidgsr>t(wCSprxqC^`!JGi5nGrBI82xd z^RmqbH^4oa_sY(8_V#hh!18e8yBJuM;Isdg;NI}RyLT~U*}U7NXJam{aQ|3wlji4y??rsH9gDX^?^U%v1pE}pNSy!7P@ zuYd8{xi4P0@a2VPzIgJ*zg+#V8>ipDapuJvXRh5ibM}iXkAD5Lcfa`E^Iu$e`!83o z-Z=f-UtfRzuh)LC;}_d{5#kJf@&2V7kN(a-xEO&%HfU4dLdqt59sh~YYEexn*#IZ? z(nh3(duAz|<(=u1FW>p->le>&kTsb-8pxW!T^7CXzWpU^U%dP4FMs{kmp?r9*H?b> z)yqHq>ii$R{>7_dq|u&=H7ZF>Q`A&i$*Q25d>wv10%R?`STsf3g?{A(tHB;4%PdCG zMej@nHL2(Udesf%>R1IS3v5B~K)?7X;zHRqVHvI1WgjvU@1rfKi`lYk4Vr8aoMdWr zYkS?3(>n8p`>E^Lg9aO%VJ$?dHGX1vxRWY8;P5l0wuLq5!;x7Bysd32EnUe5>e51I zLH4qF>Y9#?*nX=pUVXNqOfWJdxFiD;$QF%27SZGW`(4{v2ileEDu~>sa>yu|TqmVs z==#7(YErp0a(FgRYN96TB7!FqmB5JV75CHAjf=grs&7x9*U?+Wf5Av5p3L7G;*jCi zOkSZ*eF&L)Z6`YKz^!}v_8HYze5;16y-nt|grcUm5;of>-I^e&kFWF{BG$AwyVL~N zu9cZ~)=+&$Mb)*qK@D<*nqI~_)`ldxRm%BGmfD+9Xz^%DI?-vLg?3w~TY#di{t^by zA@P1IW?2V3^wZIqz5hI%ISD_=aAt1qer5Ii?9r36NBgfbDYXyYl8VKmS4kYU%Zt?4 z_Pnl`CA743upDAzg!Zx7w!mTiY-dUh=02n7T4!i23MR$~v;NO^$Noc1J@z)5tVBwz zAYESZ*JK9X${)qipNK4#TbFHVG30h7md^xTKofa8mjW2qi9mKa!lnU_5tazTCmov< z>KhVLCNp5jgV&qLfy=9(T>r9@oU-G?cn&t0JbB1sCu$~Q&Kg9rs#t&xSy!qP3i{W*+nDZ%r7=EA4$?{)c3v{eOu7)g`8=>aX})sV){BtO1~^>- z#^;Tl+Yz?3Ml9*CYX3o%+bw>>GAkION)5_oe|oqxl-pnr<n>_}N3d13hgntfZm^(nX>7v?JJgz(^{U&8j&)Ll&sD;ZoKT z2{oU~Ci5fXdM>Z#$8*{7{K(!hHJ=_EN#ysc$;8PJpXprmVQwp0yB4l)VU?Ic;cJ#y zmD0*K!%RfyodK_|!YndLsg0DG<>+Q8_$16)6lUePuLy52*}Zotxk>L3JUqIR4fGC$ z-UXaRMk|DSnAs-?wG3i1ui&+x{DGs~G&!x3NqE*v0%4xW0Q^GoJw_3fk$4b@9NoL< zF_S6U%Dmj_@9N&(wW)-%N;=&cGAV}jy)D?MNtGJ~+B^H`&WPn8wI!kE;0%u*`PfSr zic7wE*L-zbp-t0dPDuu`9-)`U0RoQ{uhyhkK+n2bCk5)1Nk8z*qGoQ(!9$3d+)ED8 zh5F#ZGK~S9yZ++%J6#GaL%lN$6}l;FKUnG3IcyVb!3KU$|(UWJccBL97UvotN=p>WgmH`1D3wG~*Cqy7 zIii3OJTl+7X3Z0ugK}STCH4Ycn|HmwrxL?nn%+hCZvvOUPpb0<1zWO0f4T#rkmsUTg*asfYA-$Kw6H+S63Bx4y`i zj#F%WU4jWLRwxvQ8=YuIEbWZhuWm-CN|!Is2A|3rlHj43Ml`T*AGExImZ+=oqzuJm*pJuX}aPBqS( zjq|M@yDGIFb}1Zni$v_~xYmP+H8;&HDh)C~T?a-vh@An#=6Kv?SK5;a#JAQaZ!NTy z@hUgKZ`E{viBqQS{CdYB1*t3PWa!$nE*kKQ7k>WNS6=Tw<@WVc7ry%C*)M+l^j|*s z(N~vVj=qpIJ)O&Dd{hgA8&uA0I{g_%4=|?pGD~0&&6b%^bGg*XwS&(1;H?j=rUS5# z*QB;X7BF@R3l#4AJr_f1iG;n#?H!NfwO-ILJlm$GuwY8bfpt)h!yKupV+05%XLMgl zfu3vcX5MkgrI%IGdZ+ObTJ|1}rCaN>RR?Mx3E9g6Tb$$c-Oht0;RohYeHoaL`Ii)S zfaHRf_`N{KZ3uMWptdp40kTekiQm~|9eR}&kV!?ha0~iusU2muQxLASkP${!h~pL* zK@M~t1I=rKRtrp)!ASF?o~xOzJD!@^R3K^C@&eFL8=8)jRN8FS*;{9L%(Q?gu-VdZ zH|iMO+2-NlDwSQOu}4jIT7e_NEy7MVXt;yZ<;JCTwY)hcsi?^_6Ob z?|bvb_x}F;dt{(B0q`n%Eb8XZ&ffgZlQ&Pl2{Y^$aCBDR6k^}!lAU{B=)kYX8!=pQ zTm%HbdNVRpEH5&wT&`i0TE$t7hW3w$M$5vmD~%6GErvnT3J3elk*L+BDeQF60%qXd zU!t|Hjc%TsFjmWOdsV#{qmM3y!14(#wj}P+aLf#jdd(>QQv4nL&E!Q{Sfe#w@GW%oYmyT8v}$OnyeV z4m^$&ZpST5gCSGR2;9LCC7TqeU11)%J1GhZ&#grl3L<)GGI;Tox>IK0##_vp_m>?E zYTetGv`zY!Qh7~P(gDpI1sGsdz*Sy286AOneuCJn3M&gX7z>#?K(5p#CL(g$tVSh; z-SSQ_c-)Z)*9bGK!g7eD++3vfH7Ab@%(#vtVw6?wr3>5`rIw!#yS*O6E=Mh?q_j@Y zJ4$Hzt3SN(*Z+92FI$?we&X4$F1`H4Z+;Yc?o~CVq`*Z^D4OaE@9S*==;0-hQj)pO z9WM$nz?aXz{MApd_FYrimv6uH#WVl)<)x?oa`mam0@x2ed6lO8o^<50!W4GGWHOyy zNTEZFF!k5ZVl1%#@+vED{Q2q6Pd$D8N$&?nKmYsu%0&-n`_FJTdba=Y{^3J=hkerM zUWdC$8Ey*OVDtdFauj^yox{}|r{xnNxa!fve-8?~uKBc4x*hmShb*>8 zVMmOvcjk;k3t`xuFHjeOe@qWR5KZPm#9wUT_m{zuEU_g~!f+tRnp}s>aK~-!gxn(g znXI_hyko=e2u6`bEUXQ;$&yPP*CJ)=v?ivX*4nt)xCVp3q6WJH{xnjRkui^%l;HG7 zmh%Cy6TxJVm{;^yrp2gtN@3$fQtRx1y2of9WHe8yJ!dr6y4Tlen-tij5CuLFD22Bd zLMNk5wm`w!wrRzpx2rI1O(a_u!mzc_c+E1XG`di8b$K)YpeS3g(p)>8i5{b3B|m_p zn)0z`q(b|**s8E4gZhiBSHAlF55Bta3fM*cXIcE^y|=!2;i)fw{=^qg{~R%@-g@ur z-~RNkr{DSdwbvsHWJHK+0sa=t%;&tSK9(mnEXz|0G2qjIjcsH8i%GXypaJ}*eZttl z(>ZV`;OQK&T6f)kN9VhuL<#tr65MMs51QJc)^TcfX|CKexxX`~HA&0hirZz!YvNgR zfE4Tu(O^n_g7LTDhV8*kVM9YE(Yavm<@w|1`MWnTkoIV|CB=_L@R`n(mJ`;{*vBNR zG|sU+`4)C(BGmO#mC!0}oHIXpn;7hqm-weI@rucNm)E*8YjeyfF)~nUT$Ak6mm24I z^gz%wnq8(3n%C5)rH+lH*C>eOzvViGumW@_55L3qG0W|aN#jJ4R3=7Gfj_E+ksy(H}|irEY1J;q6dXh zuDNnT;dYt4OV3uQuOlez<&U`HAM`S#~^THD$y>K1DnxGMbhdk*GrOr2(@HiV z#b}I_Hfy}S5gp~8v6+A}oT;^*=&i(>CM__DY5d;U=rK$<9Z~&*O-4~_nr0q$so|kN$-bmS z4stGhe(EP`n(Skipk!1TS4I_Ldqkv>#gt1Z-l9onm8|C9h`OsVeNz>tYDy-EUVbl3 zyrnSltAX~f?F|!ym{UqBupokIKOpr>lBWVw!*U> zfcHKZbTDm^#)`qlTUH!E)+w;;*2hbHyxF0hIW?`Q0hpZ7y7-WA_&hIqIyfHy@wNo) z7cKn^qCIwO3|ovH^wuPyHm7rQcL>nCXQ}a9t>JAUx4XqBslmZR@0#NKN3z6hpD3@b zfTMhgO2fyAp3G+3wOy?jA;p_jGMU&=a6cm}`hrP{ad?Cw_WKNM52=iEbwwtBxhw$p9Ox4BPBHN-avk&Eng- zNko=J>uX1zb!|AI<+bs|SU!_TtNH9m0;VuMItp5Sd^|rshIHx)ebncPsmZRMn3~*7 zB*}!LC3YWo$NJ>~_Oj7aPDvlTXK0HVQP1ON82eTd6QWk;!iZrk^=_+9>WZ#)TC;^x z$f^Gq_>XvigzsXSrg;fmlKtM0z=;tO2HtnI5Y1mL{}8ii&5yja*_9Z{rbCssP*UwC z{Zz;qrcq{M=SR%iXj**lIvpq4?ty8uc(y1l;_ zoPTEi%U<8cnCya=ST$F|=&^Q=L z7Pk#Q{<{y3Qw`Z2m&P*lOg6L)wr4_Et>9#kNx>?4S@kii1ZuAgEhrFwFXB$GonDv!9%NX`ZX$}^I+|jZn3f& zKWR%Pa+@g#MOQMrJ#|#((%C*c4rGv3lF8UTGsliKpKHF9|NQYEpM<}2bENrV^TXyl zEmEaNmrt*1@H$PlH3m>}HVqjW+eEF%9 z&6nWB)#hcm>I$5G|NGx3_uY42^Le~6QX6%=orgF7ZK!=`@-KmpJV4r4uO8s{{QT5! zKE8UJ)^*QHyE+mp8;&fx;fS57V7D+em+|%T=*f2E&pQtHAYtwFZZBdli(+LPYgN)% z7M%>(kK76Cl`(4pNfX5o{ii##Dq?$Dd%0a~f&6gdBH_IcSvB}x9`?IMR)`)w1&KKB z(vs_29hqdil&qjvj(Bc}L7cx-pI-qJP z!Z6Fi>^qdyh%MMIOuM6XgAL)>F{4-9bf`l!Zef)gE+RZG{El^~@PmH9g>}_1sZG*m zAZD;>c43!&hZ;16?bOX}jFhJlIy|=v%l0N{wT@tdwc0iYYSW18t)@CRqI;|)@$J5w zFTZf}#Y^OyA6!V5N#dtDFiOCN^HKwWD*@)d@VA%2UYWsBH(s2{9#F*=I6uQ5khk>JAZyJ#WIdS0dp|Ocfx-G)A51`jO z4k^T(qGe)3Dr2J?TGrU-)_2WqBg|Z&=AaE8VLmy>4 zrv-vxYGhy*snrJ$G0@P1%o&`Z1xXFjEsORKgfm55hf2dIm&h~>AeWSq3Di18QD2Wd zvX@~hmtiIs3M9$)j^NO zMu`P^&U3R$81g#_up+_owro;tHR9W(M(h2@T27M5$kI4$))%NB+cVwWA5U*CjI5Fk zB)Baz{5qYYW|c~|0jAT=S|sb|{h&#yO(sP2O!l;E2(wleu5U5q+S#vM=;T)_(S;y+ z zD2{*vm~UC=Yd}G|0=rHwsc5PX067jnMjE?V)c6FFeH*4)7mf!}NNk4K&N73PuB3B; za7Kh4up`{ymO($cEROe|hMHMo6?>`iij3}uj!I$Mt*38ow{BN=Y9Wg|Vg55+cu4LH zUe=q}A6+(XnP90XYl9yKodGtV7w<)_&al@6JujdRgUJVvQBnLUZo zQhxv=nQe=_^&*_mb+9)AEGe}p)}CBq1`D)70Yef=_ns%#9X0 zbE9iYI$(}OLa2VZc^PZhNb|Mkt0=ryAzQ29?lio2t$CrF`8fShz=X$%EqnrQP ze0%xP$iG z86}qqP*jKQrttElYa5-&kArVq6mW8A%rarf_DP#t4n4tJ*onf%?qobaH%IEENUH@7 z=ipufF6=8KXVQgDfY-PnlX0AeTNAaFz+(oxEVA*bM}rM!GsW^E*diOHGeK`{^v z?>5fYz(8W&$NS5{8-~dk1nUkAjZsmBQE)4|jW}kK8h5B<+{fOf`F5R9?0k$-qJK>$If4L7Ge7 zB+-?@rkCWFFvh1D#tCut9G&Pd%Y4e@3(@zJeG=bEM>nrcC{GP15g^^f{Ii6kl-SxBA zf4K5y>&0&|00}S8lY#56Tt5&0zj8lW`89mx2dxjRUig0d%=;d!|7mjl(d$2it6shS zGW`Ev_xGQOG2p!NHVpoE-oUP(zWyZ4gEz)EuRrNcihsa(BKl)ct};W_bTyY6 z*oKYRX0S6Mliihi5TQ$=8(%$Cg9xv%729ig9y4iZd-WY(-waHCg;W`JA`9Sc_GaptKTXU?pR z$Ad#UeI>GFCZ%Gq1x|hSRMCrpK{Vj52Y5$r1GrdLy&J*B#)=gtt@*|uq``VeHL3X% zo?$bd8L!x0?KHV|+lC{oQUybgJ+ggcZ8%9y(F0&REJDe^XFvPw!$1Duk7qx-{KrQ> zd;ha5pMCV%@5vv}{P9Pheel^OIC?hF>-*Dir?dufX)U>$sA}$Wccpn-*}Wx~&kZb`AH$=Uv@GLg_Z7=vuc7MvTFPA%z#OxP^@5 z14atHn$t=y7rHdOQ^2c8fgsc(EpUW3_e9`z>)K`T%b5l5O-&MqJ7&LyBxcK88f0{J zU}By;sEjC+3fT_|Y)M47S!T-=TZKSyWTq@?WS?92EzTG`vl(owNG3XUaXCW4men}F zK#j&l@@+yZFwP30#9|AL^Pn}k%}oA(`eDPISz@xn>?P;ZXBW8Rvp_rE$R80W=4tfW zgj<)z$K}i*-%C2zDXY*$k@gZTn8JF1z@jf$*j{DOO0j<2Vz*9JnXH2yS}!j7PC5*s z16rqQq9*AgC0JY1xVZQL`SglxTr@sCO^0f7sb1rDeaDNs(~foOpZ==I$r7>X@kRd0 zFDmxZCy&7hi%SLgX~Vc+qO5wcOlqH=t~+FrJVf~yq{T;0FFR$ zzmplt7g+SH`b-TxRVgfAfFTY8Z^mmXQag>UC{+zJ#WFKovk&WfJr#IhYLpB+Wki52 z+T^H^ra?x*g0I26+htyh9{(AaE_Wt{RbjpWf|^x*mT>4aqn=bU0T8cyrf`1B1KNvG zCf~&j&L(_ju=gUQY%)qh>r`BkC~to7{+FkIedEkqH_rU0mjZtO#_8YQICE_S5Qhh+ z0tPBL%bvr#{TL2kTzKZIpI(6BeDT^ZzkK%mSMUBZvS>6FEUa8g0gaUPAdZ=7tcbOli0JCCyLL@Z2_K$X zBBX}PVy9Lr)fwTwr{VJBEvj3K>}8dThiblcMVC(d6kR%GKrD-mTj}IIUVwmzM(dV? z8T!3+kwrHXx&>r1{(wv-_%q*YT(c2AkPM7V z?kt%!2GZTAuswG8q zEiC8^*rB$y78V4NQ&NH5DT?w_a|&@SgJ=nbU~y;}Uchqnz#k8B2FE4Z2f>9A++>&h z{3mN?^84v1n`C@wgt}nRA{gcj7;v?D#^<9a5>{{4mRimSIp`gSlN(t3!H95_8s-*C z$2(rS;DhFe4-hX+@KN(aFX;d7@|op-Up@_=+WEE3&DUTc?>Ddd0{p9I^3&r}6U3W4 zHBC%W5=t#bJ?2pm`rrX;UdSr!GKXs*6IjOn(2?FfatG%(m0_|AQAfXQdlT*n!cLG` zX0d|geYX91pz~yZ^yS+TuU!Wy`yV*i7vq{#Q&;NS%O8`sxgeuosvcVoMfVtPv9O$(iR|e zcoB9=O0c+(xiOkpdqRwq#m31@{H%cY>qwi+2zG~gL1;m^4%rWLQztVu2|Bai*uJNz zYm!M?7H0IazwhzV5-({9U(=qR^q0DataZ;~-(E6MRu;Mf*K1~&IXE}hr3Sa}!~q`q zq=-RwGi*|e%+FW-t3>xLlhDzvWI{=^94 zEU%{rH9e~(5`$LBt$XagR@o`@LD#a~g4YgW3|eJsQ%gBsE1h)8ZnbbS!c}u~2P`mD za@hwwk~iBKDuUXF$AQdBct$5&0s9!%Ahq5lhrAnt5>k{M-$Aq&%ofMz3gXNBp)w^i zFtSm+PZ3*)B0oeEFq;(+dbvn_D_G0rd?lBwuR=acNhOnsZN~KND*nF=E@hQCiS6Dk zV_ne$k+IzkKGZdy>6j^2&My*E=c?OC=J(fbo_#B_WP8Uzoy8_EH4ub}M65{*+@TvWF74pV z0T?K}AUSEgOiD=bN($gK6wA@gm2oiS8s`MJ$;?Az4^8_d;F{ijv16dGDw`^7Ye&_Uux+k_#XUgvNwjIk$3g<&BlMSKfr9_mPY1{gvNe|Di|C z^#-zT{SKMB^2p4!a_;&Q@Dt7tshP9_&!9(dqMcm#s`X%8fBgDs_~fM)lImmL@Aw_u z__6%@OV>V!kIKqhD{mmE?5ps}w~5Dw^*h8|ef|2A-nV=gUcjCGHyfUzRd^yKXt$HT z<9=frT6y#O4?R%X$6E-q7d?>P0Yucr*^91cVDc^_jT%n=+3Qb`l`9^u?_++wvq8Fl z(|ms_qhxZO7Vw4}^PW*^Azz@=cymw37BFnv0`$$2x4SViFDzsAMfZk>1G9TNWVh!4 zsF>S9S{d0z85eSi68t)#4+db)lHCX^hxhGSFR(pRQY>)T&k?X>xFmTAW(V<(*4Qht zpeE<>2Wg3{PjcD@z|or^M`yj(xmUv2~Sm)lfI8aO6B2tyR=hikj{8_@lJz{pHG)uYPd8|Jq!?_|Y%Fcn$uX`{Ktxh&=sA%djqOFy4Iz zt1t7ehEKigx{?kgBkpZR`4N*AD+N(^XjSxNGxAYgw_b#lPew_k`xQ8HuWME6lhlFj z#Bq9;@4q9Cz&;l%@F=G4!5U#~Otno5?>LO4 za`{A3(Q~nDHo&x)Q1$Pn2_HN%J2tw3AwRo6V93vwDMwPe9bwDwM{k}?5N4IQ6+M}N z*1DDeuQ!!Zv}AH?>75+%}YE%+dtpC^aWAvS2eC&pk>^LLPhL=n$n zQ)SRux|j9dLe~4Kmv8>?{hRMZCz0!2);otim#IKw(7kn+AE7ok8KIT~Cwte%_2e_l1JyGZ$iz`Q)&THLZ0ni+k-J zTL|njZi6SwDfTECad_Z+((eJ1?UpBv0uHd&Lr*5Dtcaz4i2iLFnmvL{GA6gHu3crO z5ho&hh!=(F@32a?;|tz#NYQ|ypYAMYxcdVP>i{&WR8sgh%!*S+&?^Xy95#_A)!kPo zeuxzvDn$+JbG8xiJck{jiR`^RZ4mI>;Vv&o%41ZGcLjR%rD?qvA;ptaQi<3kEEwJq zE^H*!!$=+WdI+oBFyN&(H#>eqS-)x~uw6&YlQC*LV~CVvgP8()tmt@k>_~DPy>4I}LvqN1?Hu;*X0^`PLTG6X8^>J_^s&Z^jw@LUCz%#{NRM-8$`+Bo=+4pO zXx&o`+9oLmkW@xVBswKk3#<0N#XUS5+Bny^%L4#{h%c?4Cma-sv?zSQ zYnNX=TSCGvhZiZg2_r%}q9hxy%)^O(TaL43QKdEr8ukE#$KXRf0uQ>7prdNc*TWkN z+Wn|9i(yWtF2{u-dfh@XQMO+(6<-DN#{?U{Bg#%aLHeuhcbvW&ogzbXE z9wVk+sN3rxC?j(z!4R(8si!B#4i2=2%`YdHDSB?dj;8rJO8LQ;>zm8|0H`IkSjU7PJd~|DJSe&w5Fr>1mb8 zR9TSylpONrYgKNlnpV_AMp1K`qNdxTYp>Ms%QUB`TWWBcl472D%7>a@fX??JA;zRA zQv`@SL`!ut&_ZU2?#u0!3K=nBGEBptuS)9O6IMxmGqI!+N;=h<$dAG`)_eu|!aS(Z zE6qPNFZZ8RnlCj!Tz&#Z@L}_v<)>f-&1=mI%cqdl3rC0lN6#uwa+ptMl47mnU*rgY60HZAa!w7TE2t2uD%$1_ncx`p9!@~Bs zZ~AO4kDHBa#WDirlgUmjWlN$_%+&N8*Te|fC0h2ll9&B}Gt+%`5L zltfa==$&5Iy;9IEfEmNgP$|pYa;wA^RoGIQ6$5|27X_gxUtLl4&~d6g`}q8z?aIOL zF5K!XUC*6RLXU~mWaE5|ws-~*nZG8iB3KXJOS*tiL6XgrL4srVLR|!ef*6qww~%^a zU_o%%O2|wcn49a?{l8~?IuPGS1hyiJ6y6V~VIT2N=Ai?A96`mSnz@FZrGk&n~Q#RpkuZzEU&9<4dNlIXi501>57)}18hYY7Xt7MX$E?lHU^$W{S!E=La57*&kOgDHw_pS8H z2FL&p4F?Rvhboe`y70FH7$E8@`X%_{aD3rb9_rtRG$ix<_vQznHe$u<5pYrkTVGeQ zKHOE!x01B*D@juoEj}^LDjSW%ie-xzmwL!Xb4P+Flfq*4eUe7+k2we)y+1C}-swVb zbO_-Vs1RQYTTTo*tb~Pbk38b2Cl?quzyJLfU_T-`l$lw>8q_ie>yn*d9uc7>sO!s`q7_6mecB}2iL`TeK64>L84JyDJ7?M+9gpy)n}+> zlcL!;kM5%*Hvk6RQLDwnx%08Xv|i_joiYr=0lS?Ip|Bigt@`Q_ql?_Gdz5&$5>wg+ zbyV7>F0PhVQh~TuXlr1mEF_G88DwgbcsxHlT`d2w8-`kBrs_H|%=jKR!k}=nN;1GP zvD^4cb4y+S%d*z@hdI0M_^iI+(LcV7-2%t3s*1!^zw+TZl zvJ|iulUrmOu5W##XtSRJlNXDa;YUjM?d0ux&Q!Q^PMm+eOSC8S{KKS>Jb~-PS|qq} zenAvl@o?M4NcZ}&-l^8E9(0`$*TJ&m5mKD6 z`v)M5qZGIA=a%7uF#`tKfq4Qh0WpQBpsglsX4bLGQg~*yWG08G?;R5)lbtP=j6G9r~AgHynMTt$Zg{BNwgOSfpIM^w<`m107owi?HbY-G1h(X^WJgD#}1zT zTx>{Lc`9 zbVXOvfbUG~?aY6D$7LYa)m>pN#F0=HHLWSCuBh46mfY-ALP-RYqW6;CTWAKQwK6sW z72%=9+J0UXQp5n0?NQP@g#Z@2Ja=Q!1zvPbD!0|HO{EllH=-&#ZUHu3A?N3}kChdt zKdqz#aA2Xkq>;kL#HdtFiV}02e&nsuQ@P0Vk65}nC8_8>GL5L9uQ>Twv(Z^)5v`SP z`QNdHZ}~LV*fgJCezc=SBv^fMnX4=*^g5dlZCW&~h%5&TKKftAm-4+ERGeq1aN&A9h(!c{A^j87IJV1!s1a zN^6iggZrplElP?ek|~OaG|VKv%v5xwBb0}#THqCbyO9L^Y>)4bxgeArI2BEPIN;L3@TwaH~T0 zNrmj^rLw7z5n8IjU3Yq>ZLQ$4Ey1-j`Tca1O)@?-LJ?CBZV76$SgFl3J|BJCT)mmE znxid2Eu;ih?6gS4Pqp0(bp$rXDamX^O{b*Dz&kam1>55G1Uf*vDmW>faqf+)`^Sa* zu#20YJ!rz;31+pF<+`qDj(jGFOmUpK7PB{^u-@7A`THB^&Bp66j)7|3DGO7SIO`UB zOlzH?$0=$L8f^Rt1(-m2?J_p_21kv28O(4H+a>vExPy$Sibq6&NPHi_qUD z4Y_H3_xGwuGes=Av{)zZr`PIEMJ!riV-79a&L_X94RJ4?Unw4szN+9EkXvX}N8#v7 zHqn_6kFw$IV|KC3i)9Cw4Vh&poE>kDbcC*oWmDKTIY<#VbuqFGhS7FeL=Xbs&eb*# zL$)-dDAJ0qZzY-HG5&0nLm3MwuT@6YVMz;QtjZ+<^z4Xq%9OASc%qn;^)~2CbhaGb zLk`2Bl;%SYYiqFZ3M{Lr0f4_aEi3sPDAly0A>Cmn-46kh+ugY`!`PxQ=lv_kXY?gP zTYYLmNoQkI=~;rC*YpgcNGY1yS#5Q%%sLPN zo_+AkAHDnaqfdSD*qeX7`orj^K#vNoRqXA_B|34PL)*+=C>Hd<(kH(#mN*?UxINBK zFnO{jYLYI3N6M{A^aKgYma%gX$F~ z>}dOI{{QT~>vI!Vwm1B*?BlBEg`9)i>Wi-Dne&uvbH~ABn{W~i_0*QMEp625*7U{3 z#CZ!aB!rO5B;-0|2njP_5)#ZM0Rp6IoK&4pGx8_flTT+H8}d$0oqyq7t7Thlwe0B@ z0$bqBVE09mwfE|^_qzUm3$Qecz~Z{f5szObWqk1=Tr_Y2>9nnMj`XmiohqYj1eQ@+ zPtzh>0o9(5Uk;Be@{Ds4Bdt&Cum$R>){Q1N4Ki&KI z=TGi5gXuRj^2W@_9r*5UJ7fM*!r|Ecd}id+nUU9mhi!GO6}O9uxWb`IvfnxODK1G6 zmFi*(-}%Y`QII2~6_&g1#J1gC!`skY8Y+M{w|_rLkPc*%a7el_Cn%Po9q`qNMa`+% zY|GF?XV53qlRyq3Atme6>M|SAGkTnDk4PUb@PCLta?~~ znTe$d>o&S`T;)a5P7hl3Ao|k5JR4%J+L=~KTqa#Fu8^=7k?trSt5%rxRV~rZa))RJ zrb}+bau##3QI=19sz_XjE7d;mvMSQj$5fIW-~-k9qHZuX;jdu@40uymZ%Wd(>M)5d zQ!Kd46M5Zk8q_I5SOc-v@^EToiu#h(nIhMeEF_7r_XYk%B1>uq#dt+T&apoU$s*{_ zI%&K|e|FhvPGMU#47(u;6mzpA#(Xf$OY_9UTJ5Y$S(W6`7DjQ?%n)RuNrO|GrdozO zEFo=G18$>6I>!f?7+8!QOy@3@0#R1>rKs;jh8z>*ryabth|RvH&NFQFx@$!2H;mWc zZ@6z?8AA%cz^`qZf$uZyKH0yY^ieZUHWw&Yv6S6zTxsQ_Nng}L$XXas+tOhK4AHr! zr6RMyMXw@uriEhFR`=7Hn0=13LDFyN1p?oN7C5d7t?dTSfIVa()(B70RwbRoM%c8W zn^;x9df8hgEXbaY@+W`M+iftqYDn%l?NLFg9~T3Rg=Z-@O(%j#m`0w}2UsBmzNr|G zNzGW-xMTo<1N4C|aO55ej3@%3LmIP%L! zV`WnLzaKvq+)!^Ay9)?791>#6!!Rr*VTw6Rl_}FUYv9^~^HF3@E!#56*XebswlbqY zwMWrR8eBv@9j#sPLKXen;M#|XnczQx5{6uOH7?nRzU_#xEW~9GmAv1A2&}>5DV6J4 zgZ6_pyY^A%<{hOW)AD*xcrnqNqpDtLOBr@X(~EWVyq7<~9kg>b%*1_H-i0GIc&V8g zNNHG_WiZm!(A>l0Nvi5<85Dmr$RwlV(E%HtGqTY^DB(2tiq@kKQf-iHnL#X+P~BaG zVkjDMmo@nIn~)zS?Axkn>3~N=rQTkNUfUSp19bT;(b~2pl3G^PTjB2&kUM)Lj4*50 zY?6poN`>YJ@f_1K^bS(f;`!jY@13v?s!J?zR}qlUFgFN962j8chv2a^+Jp0FtzN~d zgiS;7aCPfD_3H;seK!5_)b**)$kc`DV^cS#k4#;j8YNQ~H@x`b^hr1#n;LuRCGz4h zNvqT}fAt?NTyp#6lR5gB^EnqwNjO5wFFUFzxA)AU)QWw zlOJLU5)J*Ye!^$xnnvZz724VqzQuCk9tL;=%^5kU;QTm)B~-fx&*LXw|Yv|1*< zi`-{Vyg&QO?b)|Z{B-xs+?~_2@0`1T?A+|h(Yg2T{B--}2NynjFnaa=>9Y?GA16P3 z_vPH>&+cD-`N5_0Yksp`D!cn{ymtS0zni`HO<-->G34?U0tbR6?9d5`)!VA?r=nuA(GRnohi7`6&_>jGFwcNs$= zk=)xp4F8%AuOt?VM#(mPNUZ~(#(>F!ZN4??BG#(IE0>%WAMxpH@#)D8!A7mO=7x$| zt>y1AMT;fm2~s>|fBr*~fBfQ?yQlwqJ3n$3muU-G; z>(O66IYWN=`t<*IrE#If$D<#=ocZLJtCz{fUp_wd%WEG8cf+^Q%wQ}gx(YL6m7=DY z&zdHXP!u$z?u=y(2&dozblq@g<%utQEUiM_*Gi}b0cCgh8)U2n-vx^(p zth$rghAW;@5vYH8#sxW6H*PjqecQ{P2JA$8GYC4-#|lLie60+^CMU1}I;DP>DP(dS_pdyCA_3yA2WYY1A+ z%_3|Xfu-9JOdpVMSdcxq7whtEjC_rZt=6-WHcGY;Arh_j`hwIxb}{C#2B`Dw5*Xmd zUgiwx8!GB+eub@yv6|p?b<;HO=V`JTln;cICLOvC@}l$Gc)$WjYN)yiUtmabx|GQ%)7Fci zf?CGzcR`%AN^-PzdyyPnNkp;;&fa6d1C55F&5sA#o6ZelQh5y`fr2CzL5x1-bYMkz z1`P-8akWxRa#)CalHyOD6kz*U%{-qR6C{s;;M<=bKTK3gOq*MdbBv5DH01zIG|(k} zEQKVPRLLTlWbi$P+7>l36*&avhmo+ZJ|@)-P*RRUfjNWiYE8DmchEvnx4?u7PNcN= zK>nbM8B!4|yq(Umzo}P5A?hjk)%=EV`?#8-d351aPywc83^bPM0NJ;%-LUfv_ByPq zIG}5ZSOhkmuplB+*|7U$|9-M@-@e`=1M5^xbm&;bn6e#XyKTvvQaoeO?Ed}F5cqS% z>DH-0+)$vy_g5^J1I{&JS)Q&xN) zWFZ!Nm?G2#utCyAjl6!S847|PHOgRjj=Q%hS4=SC+d)F6KMx|;lW#3_(<0JaWw>5! z7;$S|sV*fZh)*GLYj;kimZe)cqha+}bxnUbB#552;wOLlyQm2RT-Aw}YM5s0pa_ZQ zbprNBI#;A>9Y%94m?Sp_S=28G83-sW$l}xP%k6=Id@Os@uxp+tbXU-~#BR2N$Pj!D14J^Sb66P z^8K|7y}ZY?$IlsD-{`a)E+T>~UQ9g4(`Y!SscF)-K4w~VAi20d!=*tC6F9r{T3-OW@Ip1or15CXBsz6*!fm< zp$uLwPBxnJ#FRyv&oe8yDYsLFMJew>%$DRzaM?Nbr*onx#Omi5+JgvP#9#y5sn(3d zLXmJh9*H66gOfo>dou_K&I%(e=;m>oC;vpD#@~v~Q38@8NUE{0Rg&|`XMp%COe50iiAX_^?Q%q}E zlAw+-kc{0Ha{b_!Ld__vk}ZC7sBHog?iOX3cu(rNwou?5PZ( zfuvEGh2b7n)D4R;!_bXJE@+P?`mh3XM~!g@7BNoD67edcD_QPaya*^$z0CFdt;?Sp zo4SZi#|Z*IZ(wWj)MYYtY3lZmr+z#&H8y>G>KgtgQ(sM=fXf%~GWH<90(;wEN1Ycq z&B*Zc6?hT8@(Vb>ICY)Ct?TfD6I0*dtqb@u_}a&1a5gCBG(A#tke^Q7hN-&h9%Mc3 z*q%qA4w6I}`)^|5>OimW){LKzyfOR6w=+k+y#MNnnIjjOP z2XzQzwiT zh-nj;7kIYG*6!A@(~W5B18|)k;>`d{?M^0J^Yc+V)x2;wvE4me!FN8G|BSBExQiko z;!3kQ$37o`m>_#l6nq=Fy*#U-aXWvd?6y0o!I4lA?AD>-`QPB5w}PI-0|W9e@N#n| zJR!)^1Gat+o|gaP2%ZftbfEr6%Nked#b|>&Jh~8o=%g8D=)vvUUaqmFZcC^lSLfKD zT;-4;dN|V8t2`2&x@ASpLWOeTQ%RMF#stsXMl^zMFb>)1Bpr0Y{SWRoh8r zS%`ZGqt~Tczf`vy%!I{|X2OYyYW-z2nR%=4rI(s& zk$R3B*Z@n41|L8#NMqKZZsdbo^w_!0+IROoWHd{Wg-E0yCcnVA3f2&Z*Jf-3gwN1AqTt`atmKXH^ZgdKp*!vA z@KFZuxBjVFI09Bw;iPTmZLPnCFWs_jbIVd0w`^mk=|a_xivrkgk|+vd+>K2rxq7zB zCqCsJ6BJL0(Yjlm=%Sj2Q5-6+<@f=%_jyEu<)e>cA*-X;3s!NR^cgT3 zUIU7!rTvBK<*Qe>tsLnag1=c=#9~naqj_;b6qm>(D%X}uR9;aoVId^d-FUZdj(EiU zEK^PJkAGv+jDZXgs~4sSM;@}SzunGca$qbsgV|r@CFdv@Fu|9*P)JO~V{$?YCzS*o ziV3ANxh2t_+#E}UqcNAwAW4-i$vO5Z6Hr^JZXB#8zV?Wb8OMdGYYy$h7jSe7 zE3R%$eT7^M4gOglu3=O+a%C*DOFFQ;V*Bx>P z=dG9EbH18Bo>=n6g_j{{$<*bkOVh7nJ=2#{V`Tbec+Zy(D~Ll;ac$}wIDrp#e-@nF zfNy(oc;TSgS_oj27jUx692kF?y5`L8tuhnFJTb=fNx1a2LyQsZ*@6$hfdyhHfn!)M zb{lKQrjJ*iHW!1r?`KV}!Q$Y10p7*6Jchn0Hxh*O!Q_*q3V%w6HLmsF7BA=pENURF zAn(Bh(-M2T6Fs|nhM(7Mq@gA`rWXIH*+jiPMS69008ARP4Z&=JgGu)3#T3=3=4$N; zTzv!r3=44&i((KgDbmZ*x|SixpO$a74YdQm6*CsUd%)taK3#GfF=EAr5q+94iqj1j zsuqChceO5^oj6>&F)=c6q;v;<Do%bOp3!*8w|Y<|+C9b+DV8O+!i-CPqr% zmaYUAOP5MByq~N3A-(@04V}0U5`CMOElK!P$}_Cm&ECS*0zncx3;ybSH%@JbJGsZ> za68O|R9z!%LfbVH21s_>wE~8{$=V*bdkFI~60tH)5M^1Al(<_+wc!k6E5I2j*v=+- z;3S^>++`QbQZo;`r&)(@)~NKQf~+!<#{_utXTLi~&GafZGQq?vg5n9s`YzSZ7vDo7rTvt=2~TG|a= z&#OaP_3kit^6-Om@64WfdG7abEwe4`>Gf<2do&JTk@Q-$+!_Apw}x;?Trom^_x-t# zul(to2iI@S-uq_uwYO_*5YIkey+Ob?BF~p|`U^>=Seal>i3?G8K(=-`c7VDEf$*{+ixPQE*Ndh)Hwv)Hfy!Q{nq-#^*t1d7fEkJ>%2(*g1?1|CYuP!|J!*{c1ZwI%_jG@9%3XBT(EX_+HCRgk;=h&wlqC!Ne z4#ck4GbFklGJ!NAQ`A8YSF51GMGRa%@hPCV5cYH^n*%-tl-CV|XBh&a(8jqgkH;|B zU3hN%yrE{8MmAbJW9G4nK3MTuGg2|#F$<_4?hV4J4RQhb< zaA~Y`rF3@_DP1UCfwONXPL;mFzNgaN(!J7Exb`NQI9j?H*F-NL1}OE*im;oo=g zg|C-xPQ2-gjCt-QN*@u7mQ9RIyfSeDuHJ>KZ?uULnK*)H6GtcBER9VZh8e)A4=m?o ziVx{SF0^zhB)bY|eJdQsdp!J_rr({Z+hDd?IKT|1=|rO#;G{p?QxEU~o^E7#^Ret# z_|-K&kPjZZ+1){Uc(LfV0;LK?5gz&!Mz{u6@sl68>9x38)b#9-)*uMnx`H=yR0w&3 z2g|||w)9e!aTu_5(X~{Ood#ph8lW6!5*9h;EP+w&F0;T=S+!4NTWwc95IF{n;%BLmp^5eS>Jnn9qF-KJd!Oi8!0Q|xo$8OUy; zW#1A|gm|cKIEyFOo$aCqrzvJwDzBpfs+K$JN=KAWdZ9)JOsX1JX~DJO zxEgL{za6)VL#Gpe`f3Y$(>YL}T%K#I*qE+hGfOyL9U$;uNQ{-bm#S-YH-k*toJ!V9 zQ#Y^x^~Y1wM;6M0Z#m_!$EHSy-LIHFS>qLCqZE0j$q*YIoW|n_YZ+LmFXWeqGhRsx z#eJuR!V$2bt6d!5^&+s`^{@iZn?uVKDRl4h1Qxn=71okR5LgI4_kDVS^bYEEkhY7_ zm;la)7!CY{8eJq%$)5qb~m*s2IFB8zF6Vvymt~QwpO&^~= zIrRnBz`MpkV2wC2{qoe+;QLfH%4ut)X6zucyiCK`4x_9?PBl=-eV9tuvnp7A+$1Un zCt)Em>eve#l28w0M~!#o^l#Ing!k>+Q?`S*7|cxTaIMYcnnl4#2DhFQ@E!BRo-0kJ zok}=DHfFd<3n`wpX>bp0adO<@yk#13S1A1oBMFliq0i0-$S1yBx;Akb^M(`e5v;2( zeOLG|3B`$#iKC@^6R(u+I;pn1PI7K-Tl`g}!#ta_$p1sSXqKn3r%-1md2ajm zzyd41J@MYes}m!oyANfYgq=Q=cy*oFy4#$1p}=e;lkH?(G|lZgAoyxI(2@&W(K+!c zq?jOj06Lz0cWen#o5|DlG@&;NsV$66DUwvZw59z9!n=R?kY~VTAjKa(WI31*a)9M@ z=SB@NlxG_sd*GdPZ3C%7rx#n%&6DhlEp`xA{A5+bqnf-i5u zmu{MNA-Gs-mU_9DF29^RVdUJA#pk) zH(!15=GgtS*X|#^{PU}0bAPxzd;Y@g`FCgU-1zy`V?Q5x=l(Yz1s2k7eaKFMbsmQ~ zcX`Lb8AHO@(GvI2MFfdZbPkz$+dM!M$*jhic_>fGR^!l6isc>rYkd%L&-3X6Yd^`c zFcLC9_94@9vAxu~n@DD8Avh*>HG_fVf;yM{LY;%?B7ze400x@G1c~q_?k_B!gceYL4m_9Of8QbhHO&^~c!V4-Jqi>6dYicBx6Sm8jN zZ$S*!kF}C%M*4Mw?Ay1E8?-=raxgxV^soV@nvbk%@=Dw*XAbDg=`1# z|6(yXO63bYdx&SHkc-Ywbk(Maerr?02`MBf9-fC~tCw54nNet|jyHTTal{EK9L9Ry zWkT#d-uCbwe5lt(Z+PjfgFtX+;)qk6e1jk)LFszwYPm|;E1#C`%{?tk1?0>P*=bM| z=e(R0>;{B?GFS)h>b;0dRuTWQ5`~Bd9mlt}?c2AvJCWSnJN&no-`c;w2VQTmG&W-5 z93bZgXxR|evYKvEt=USFyDWjsfBROG2G5lAfEkiwS$oK3O}KbZeS*q>h-5+XBuSpO zFavy2&Q__lnxO`4B=eN(Hv*YFnHgY))nstwecYnj7On;_8*dB%c^DhY!k!>$5KCe6 zLF}2Qs*_wm)K7!8&NF6mvs}7bWN0c>szK(7CBd6=tcc8_9o4o=&$i)LBs*-_l9m1Y zNf~T@-@a}`$9mXAa&P-E{A)Jk85nX7+qxLpq8|({Y@?7eY}2KCv5Kr?zGaOZL&|zU z++(%xUbyTKX~PKrjNZkyd_?@SnJhU7?f5Y~6U8kl*<98>R3>DzL%R_R8Q+2M|{ z_o`#?-E}(aoc!3;(!J7mV89-kI2C-`Dh>?{@O<87J+#Wvnz=tYx}t0pL5bE)7W9Oe z51=f1ni|E`G#_L%AD%v@?QT7TJdAAhO83IpUKaLLEch;U70F(fLKA^x+4y-a1GZX| zEl<6StIoUY*d)vxSZl6DC|(*vfAy7RLIc}&dO`!+bTGzt>Q4FY>PH8D{_xuD>+j5c7#x{FLZrATEG2>% zarLkShs_To)lxLPZa_b3S?^#}V@R^Hi!rjXsOu^zs&t4M!7G9qWKAoQ5SFXealJ12p@B&Pj z26y0C)8S#pSM@R@%Q)Fd4Qi|Uz~U|VRz(=G4lmiTVZ;2Zp5=L6rEGwI|N7!P;O_rU z8UF^qO6LSY_|vz|6=|#*f8YnFA{19dC8AUwp7M#$$0sZ(p3;EjQC@fHT1H=r;`cOMK4aeLCW z{O*>`&u*;_fr<`OV zELHcWu6rhp*2|=-aucSahq26R`66M&QK%NR@6donvso0yi!WN@@Brf~n5Bc{n^()e{i9W(y8*GTT|{xno^OL z;$K=3A?{%(Utd5Iy;L(A)6O4tqc=wVqBn+-{44B%{SSf&v(>PRI#K>~yZP*k9%p`G zweOxU^0Lq)B`PHydcoElw3z3p%Ct16b>PMH*;ro{cNyuD3-8aTgd*5C7Yo%LZZlXx z9{@LK-@f^cYAEr%Zui+p!MlI|zs?uAx$Y(l%s?Cl;(FQQ#)<9eA%L=2GGI}Gv@_ZeP}`{D zE+_T7VMVa@G3GHZ9xfWTsAJ=o!D`g|4Vr-?6IYyGR)7VDz;0kUVX@6HgX3eA6v3dR znE^db(|r7zHoS{w4PKz2k*gc4@{!G}jCDy#NbT{f@M{;$nKL-Tw1o)>`~-ATjhmTh zxbg%q_eKFQi(qpFCAhXjFGRk@{O+c5rs zMLQ(K8)vLOswH}l>A>18;IzS+(}F<^J)P&Q&4eX4RY(0$57qLpb){TryRIS?za=UW zY}Sl;Hs7Zy1zbdpN`qI_TQn&CrglF;#OtkwK7JN{qmuxi9y~bX$q~EY;r)t7L<+2@ zkF=`WWOhc!pqQ31R%}|fb7*^YL(LGFFvNbbH4rOi)P&{=H14)7hbggFw~ci0Dx}@L zq}*DfyBxMirI{k)vza1-y-BhMoyfmT=6!2)ws{j0U^%c~1q_dFH0*3@h}fDF33sYd zNJ=*}BF3X)9d#)%b*8J|*~Jj!mmmgYBMFM9O7F?vgO*gPS~;F#2%p)Q9jUH$O9~63 zSRLc_Z7MrEQZjGZ2%5>60ZO0F4LbCMMa-7u5g0h%R69zCu9;w@;kuh4-R;WrfrYcv z$SjGyyHK1&-|<5!BE;ot8+kq4l-O$Xt+q+FGzd*sab+w;{~JqOkfZBF2#r&^zu!6q z10w(Z;o-nyZ5$Iq#GzvHrd9kJ49|nT;|d-TYAs1QCkGqvW>!)0p>k?p$zQXvB*MnWq8>;`1V03Mu#CSSjN0jCfIlAqY5)MDj12wC)N^1 zhQ&+XX=y~TNsL5YUY&E|Z$Xe{A?l$We8|uv9Amiwp~WJjhKU{4DMly(F|=xnum=XF z^dGN$A88nI*@wCj@~wmWA%d+uArHJu+Aq9IM>&~*^(Ww@X6!WP1m^RV#qm#^gac+7 z5==|<78w$!Bo=z>+mI&EDM|^4qoJ5AN2TE|3O#IjlN4j8u zXNuJ29g0fCrMJ$BPuWHUB~tAnu3K4XWvME)3#^0Y8`NzYcc=8?zEgTmGg0-H?At$) z7DD-E6UXwg9>MXaV8svKXy6IhixJlr=T{7hD1_?|Ts|T0xy_sgg0h$OdBE{`2B`R zk{j446g=8Ez|&0AsL5SmRZ_TO$vMY9btfvsJ=?u+kj0{2H0a>k$4{*0{G=K5Mqfz2 zKnBSR6ojDvaL2A=AIZA05?gD-6Rdzv5J`kXHDX#pltc=JqHUoV`LB_;|7+y4{~9^- zUnB4Q*T@IU^aXF-+)^c!maR-TvTFo0Y)c2(N%J{oani?BuxSH}M2-rQED26`ksNj} zVxibF(T~`Q@{xqNQr$4&TRywW3ssSFBS8o@O@SXnw&@l)bFj3LZid7F&OkB0y(Ve> zw`FdSReU6ADm5jl ztSC=Kh=_H|6Wfr``Z1Pay;;kP-WJ_pJgbpi&J2Sb(uU0)NUp7h?(`xC%t`Lp-R0J= z+OP=X3a|(uji?azV3GQUVcxfIAv)k-ck+~tfkLxQ6v074l)#uEE5l39_V35QP$gux z^rb)D`RKvu*#7;`keMT&&m285bM!T`y)zMx#H3Ix5+;cC{3=|1d*8loiJs?%XWx8p z=IF_}uTRVzd4KGqwkQp_pZ#0 zd<%n~IdX0G{3SS;d;bq}XTErF>D~LM&jx;AdTF6BM0WR+cGY%)t4kH*Qu1AS9FEs0 zC4cgLuRh&`8G(m=6)Se4iI5~jLhE@rZ?|n0<^neFG%aALs39RDS#dYR_{_#du`=J= zrsNxt3;}4*O&pnct@JHow>IBiS2|m|SGqfK1b>#sN;fx=(pwXUCyq{xOa#xSl|Cw6 zDcvbunK%r;uanXR2c2y}fugQVcGPDoJBs`^;fF$=*`t|GDN$w~YoIBbqH9ozcZqS* zN5M+&!$^_?n-RQSBB^DDNGF&wDs#cb$*usn>^J@mqKVdkYVwcecM%?vG0+=$^2o$U z={lT#JMmkG5NT}-r%Hp~0{uDa=+C{<)zXzs2uXdobQ9tIK$#q%!rP!=SKz|+GBXz5 z3M{&14OMccrBGWXRznfc34M*kT4VeQPc$mWB( z#nS{S%d@)bIG`|hfepLd5G|NCW)j@aS#3ZrWd%g8D3-y8JlH+{Ic6_7)Usn!U%=5V zq{6&4_0{xAhZz&8XOZx7>PEt0puI5t3jBuaZDJH00B{2kVpW!gm!6g#hUu55kHTxO zFYwM*2yCm~yzoNwA*L=*U7CKCOnnM39V62(!@Itm8YK=9E`HH(;G~KUx1Hwmx_xMP z{&%za_fvnEy5>ykEu8e}lQ`Njn6i_|OZo*Kkd4zvoO{QE*I2>~-as}}6u~j1(zxx= z0Uvj*9+kDj@5&)!Ci`JF;?5zO0TtS8nslfsHsM-*N*XtF6qW%sr(1QjVR+xah~-n8 zKvi0SK^0^Y6GGAI9rz*YFZWWzvbC%U=N6pp?%BC(dv707M$2R~2ce9g&$7qmtk>;~ zbc1a~Rac6~IZyPmEK{|%EqdM#o|NulhFV2E~V z%>c9F1}S}4x<|lhx>p)QmQ8__&cN~4r5h6`@G`vSnqzvvAzryRag;c3{|3AihcV*{ z{6oeJxB@Q*%LHt&YjEYL^WlyOf3XsdtC#rkRw=SIKZO&(v> z&(pM_?FWx#d)z|pgQRzSG(}B|*U@L_1kf^ciMi~umVHluM2LB^<-WJPJ;h9!$RWcl z%8@aWgW-2@o>Qn2S>443xoPG(`~jmaX{2*}fc3H>!LPG_zd$y&@9uj{)s}THxiz-R z4%KF`Vv_CKx3_gJ%)oXHOaNxK`5S#;xx4;mMf8d*-K+Rb_ePO3BBJ<4H}vQpQ&mJ3 z(uDGVYu?zYf=RJcfJrphZ{vTRRPAX52yx4l)@2Ts&42~HGCk;9SdWQ1@2<7f4O@pBFdC?g=D zUfnJlbO0{iB?BtWH-l_CIg)=eVlJ?MVuIRUlBlS0KM@NHQYb25_(GKGNdE4qNG|50 z^@{oB^@<`Xfh<{;3 zW97kFyp-s3JY<6)QDiSOXfek%2Hysj)Zjknw(OytYatHHHCN)WyyBx%tJ%!X`;F0G z2d8ihyrnnL`vaHgE;@|g&1ZP-fYWpokJh~AyK@aLpU`zR0b5Q7Et0MHuE3Jo!c5cM za2}juA&N!PvL`bC4ZEn7&N<6oaCP$*8rG$3;gBH59tOfW3BEQx#0$1X zklmmR+PwA&(U%HBG%d1FdDuY*sA>!mL@RZM8Cc_NSXEscQbpZ->snX>%b{-kUPOsy zde)%*9N`UBV#3DM3{FDT?qr%UR2>XjkcnX-BT4XWsROq-PBlHExYj_|3g}-Z*uGO8 zlX~)Pgq|EtW%LYI^VLC?)3|cJa9D`SbwdJSTtn5Z=2HUP6SwK96t@!2{@6|}OJxX` zj?n2__y8kGqrmI1o4eQ{o^_YFx{y143k9)G>h7>krk(Vsb%-vsRu^>5#+xSLT3Xdj z-6ZgGLP>ka_FYXTxm~Kmv$SMXaMD3xS*wc)_SZHq6m6arT?p4PS4_kIN){KwO5MC- zO}mmU@pcATiWweMs~Khtj^EJ`EVuy`HDaW8t_iNMD7u&s^0<$_ z4FG43V5^5O1v}-7FLuBk9Y#d0mtG>ju^C~Ss?zZbl+Z#79RGCrej>e4@!F+>2q;(0 z$L0CgEzSknA}>5Ur^%@1LVX??+^!2u5i_alukV+cGmE-l^M9ze~ekib*Vo^a1W5PxfB+&)HDT)=N z+d1|pFHsbt9%z+E9g2IhLT@{}c0adDLAMyfct>OFM3COrG?g8U))5S?)shwoyIRsB zzL!EVDkzbMNj@Yk7)ypCnQrq%8`cN|B#0p{$nkY}FRzAxXq+;>M(9Y)Rm)O7@h2P! zwjVE9k6#_>VFO_9A?>hE7GUy7BoL!-GH2S6V$+>VG;;v&b`622L&-)B zzVI``JZ59x-r>Hz&DT9YKlbuu1TUz`8q72wO0y*y9-eAf+kagzJp7p@k4;3DWAciR zjR?A zEh^XQ;0y*g^EsEz8*>@GG5^m7npV|GP2Uu!W-1_(JYuq)$jQg-0FXHb>@!< zako55C$2I-$~x>mTg#Tg?-s!z6j#*2Iq|2!a!3e!XxS~l&krbY7gyEsFUc0$O6e-; zqxtX03g3@4pF(dz(sC^`{+UiTilCV!qkn(aE~qx?cZQa>gZB+@qj{1l8-l{h5*;nLlw>~umV1-8&-}a%Y60CXS8)K z(6ekW6m>X67M8yqeuwnvSsMno(5W9xLOXPm6>}<84NyghShF>HR ziYG#m#1;eFa(O!ce>-#)!y{|mZR#wK)0JIcxKbGlamEr@#9P2@F~PzwV2(TO_M=2s zQ?=-`tvcplX?w)$A3O?inJ!f$GHr$(u?-=I;K~ie8~X9o^pOP^s9RGP;6=x#Mu*+6 zm_AwK6^J(T%$l#IsDg9)3ZW!GY%=E)+R~<>19QW6;jNo4`t#x zj>#jCPu}eTL|ybyQU97K3yKG2A`tv20^1Fv@@L{UY2AsEoji8m_}_1;UAK3as0glh=b^G2IUCXV3A#L|3wTzIlFj?6sd?J@)gFcMxU!^tZDk zZ|>hebM);8-`$-(`N@20u*zQj>F#&;PhV_Y8tD<4turI<&Kx~AbM*Yo$mbQ=yRRl@ zPrX0;<*T!2-}w2$aSXQx52e_a-O=IBBm^cP28qT5Nva!|3DkBwOt+q~(@v|}#*h$t z1{d3Z5!bwub|W*dRS{14Jds4}oLPjX(>CL38?L}zyLWfAVIGue!5Krg(Tu^VX1j8M zV#sCeg#0hDaY2bz2fTc1AIR##xtAYYzVh>jpTf`Ccg{7#Qo7VUnfu_4xvSsKp1&}A z`H!$>{B-;F{Z~hSx_S7)A1?%!%iMc!%)b6zMJP%|eH5SS6GvS1uvEPu28rU2pD~d8 znv(%WdW%T9Y+$!_(@CbC{cx7&E&9V*vauYuC)RvQ;^D*hQmE3I;~e`Gjv^>>-D}&9 z4U~hYqG{u!MBShpJj*pkGBzpb7U|}P4pH(On-c>(8OmY3Fim!d^sBm!s|Pa-9p0EA zWEXSZP;}lfz%daApRtu1c7b60ifhFAobgfTV{G^vCbiQH(|6YRCCfP|>fmp~`dXmr z@lgZ5P=;zI0iWB%hE$#7=bKn0XJWaHNf<&xS@6+C8{93p2iK`J|1L`46C6Sij6?7= z6Yxteyl4}Alk{q1NCgfCI`RnooNd>$gG|$1tblSP>WY8+Oyv0&QeS;n6smm9OkgEtQ8sHE`Ba z+@Qf@3P_E%wEv-e3(dtr{s1UBdBL*ad)qM%F$Kg%j8hTO&#mYYe&x%DuFM>w|7vQrjV%miQ zHIVVs0an{e;lax~aIuf0G`L|w^ToQ?ic~A5urU$gj|DL%3z8f>h|>=HV}`k+dD34C5H-E;8lzJ@c5mfqS#N~cSAFbeE& z(%f=0(%f*~aC71nukc#BcJFOjD!t0Q%}dZ>4#jF_DGF-pUOj8GO^PefO~$b%U-np% z{eFFc=@UPmLIMr=d41U;{Hfc>Z-Tidq}(9WFHfJC8k;^jyv+6L!u)05esSvZ)aMS5 z$wegFa9B)!JT?6?(yHA4@f3cuA5W3Lj>63a1WXE2G(ZK07mwhtVM@#)|2R&jzH-uG zuar4Zz9LA3^4G6mgkL!?a_(H8`f9DmbC*7&e;tMSdWCG9eg%gxb#3~U;P!1#fttA- z&A6P;RykJNbAjtCK5=0o8n523pM3MSBgeR^tLy-?HMW{It`rd>9>B}GOTjm6*wB$P zhD=kpRZ4bJ%cKKfF17Y@9SkS9sN3)ovXP;+lr>ly6z;_rorWtm%+%pOU--+Bza0O| z$e+La%j@UaQ!q}g${^iJ@Z~gg;)<1vt=Uaby6;8hU%i;OgefO8+ zFTM24CZdgxV)%hX3Dw8Hwcu}v@i5}%j(>uGhBIv#|JJYIFZCP0N^0=sXw#K~|Dy>G z{4ni0V5hj2cbRB$rLuWA$3CB>Xbq-@Ao-YXgB3XPNEfLCk(6H8Ihbvbja)O+1~1|^ zvWaBIN7GJ&fD>fIZa#|{{V-qVCQ>xUM@{(9G6Z49c95jgdO0_O0jdq-qp)J&hYoPl zX6iHG>kCpj`LpI@*Ugsuf#)JF#Kr38!gs%qMYCCS#$J5U5{Cz1XfM440=05ZV0vMlY~~4BP*bn#Hs5hKd}qaHQF=Tf zi$W}1H`zwxFvB0wGP;n!ukv)4Tu$sZvQ*=yvvimwMNNk>hLm|01Iwl+)nFZE>TzUq z@S+|*`F2v$;Ff^`A*!xrd79dp)wGsm1)SWv8{U)n=jM}JIAmP7Y>hijuGUa{B{nus^DGBw)NYWu|i)w=fF|GsCl1 zV!>V9j76Us*bx@hhutU0zA9DT$q;s&A>#ni^mGoUrymXv*~Em4V7P!MJAPiZ7{-9% z>LR$%BBIb*-mz9J^LvhJ84Pw8Y_q`BjI9N)3;OrpEZxSWN=nl?_W2-0g|H_U<@-Sp z$zGOH5gRGX#?Ry0{iwFKN2Q#5bwkaN?WRf#JWV>Ns*+wk&6%^v1dR}yIlbU=^efAz z?^RO?2_dQO#slZC5m)^k(w9*U!BnY{&Y4dN%W-Fp8k}sSg%rodKn4G+L^^ufk4TRyIeFR~3Yvd`*+WFu8mb|JWQI+)H? zR{(Sw+{~3=GeJaQLGf@0uV=LoUJmPX$a=NZoueXsy9vWw<}cM@;22f=Yjfn}U^ zng6khn4DvuQi=;v4|a3i<>1Z6TmlXmeCwia8(1Zo86Pd^X(VA`q#qpU@e353tia;g zuJQl;oy+(WU5$*Q-$uqbcDlus$AIfBhLk6ul4cu26)YR@**B{!4{sVo!kNGV=rrtfejdNfr9eU1cjMT% zBziWV;&LVe+5}mfskFMZB8i*KR)ZGC&tdF>q;&(!!ZWm1nMHE9r_3U$Wp(vYlvxCh z3GP~;(}2(5hZxoulXioejx|bZ`VGhLQVGiVw z)7Iiaj9u0YXz9?^0yV0&r3bc4OaRobLfX>fQI9PKBK@uf_2`7He|%JfliH+ZEsBP~}6>Ky84z3-Zc(Zi9^tID}>TsiCZPm4j_mEPwbOX6jaqXKJ!6w{^SMXZ- zg-y0Cl|CumEZvznGI6S0RkmcDffcUwworNp+je1IPQVA=b46VP^3{_>G*Yx zwKA)9m(C65S|nvbL>6vXCtXxrF1&fa(h)~mYxUTTI&k3LeETh&8_7h$p#+8 zjBQzNQW?=zDLI(UdW5j+o6f%-u7Ym z*L11zs2B=`$TmBh!(3w>=V^vj|kMXN)lnqPp+a?}9+r#daW75|%? zt`m(jixe0$1VLy9IjvY>*U3oU&{^=hy50jiGe9~i!*)klMHhjG=##($;Nb zf3*iF2&@Y64WxxS$TYJ$^a(Dv#`}4?+)-vADI#a^MVm1Mx{DSOg>@c7tCw1}T&0#+ z$P)#{38_jDMPP0PF&eMPC}^1hNJtR{S+3iru0aGBlEb315)oYQu6Bg)7@q%K-DAFR zyE$p!_H5pgB%5tRFH*9L(X1VO8SJ5E4#~K=I~E~HaaU7@Kim$Kt}IC5>g1pA(DuLV zdTf#FhShO1V~{%?9myiXq6KGa>Ox`;rpi*(jp!h|ppE?%kKpWwg{TL1)3(2Iq*u0b_#qqvcS5IteTi@0DD{0ty~!~r_8j8<(p}h5t`C%tLpttrcb?=6Q9)k z$reQVL4LE~5~W63Ptzjy%De14*M{a^EyMDPB8>?lPg3G3vF^SfQ>n2D+&7>*Y`~V3>H8yg*>?AD8Yq7(O>h={n-L zm+p~`rPHM=rE8@d6GyP}q;&V0W^i7EW4YtFVTkjCbmpOg>n=qQ$^MPfxAg+L?}hcR z70Dot0_-Zgiy%U7y5al7E(FuCzji;AxFUo+y13p%k?4Vi-l7A+vxK!;CnB!+s&nE~ zHcO@FfZPfvOPyjI@68DYERxJQe^JVEHSQx%3-JSTYQ$zO}E|OQtzdyHq z`*vJBZoraq!JN=VF zV(Kdl+^+eJuZ#wV1Gq3XhVj^nAU(I1gOm&IHG5Iflu%{WbdG&WP7%VMv{WDy<{#iF zt}SU*Y1#)9M?kHvlx`DTQ{DnKyM|U+@LdeKzB0Put|(V(SClI&RFs1_B7&mSjry_f zL@#lo_gDr?=jsTG)OMkz+J$4%3WZ}L7Q@P+rw??nx2V!|PSaDykX=|clSIU7S5d6i zr*-1H`ibu@PM@6m3ihlM1b;Y(C#P=Lk%6otzLQ-bvd*zj`HF(1)E)8_VZB=b;W=a* z<9E{eEX(QY+HZ|sF?IiItj@N^&s*T5rD4UlwQO6wngy?I0d**_b+qVo9XE$#Bzj`liS>3ZK*Z99IW)-N*zFHA=~l z1ZyHptJon$6YZf;C=rdxk%S!IoJ@ojxYn^HuDG-PwR8O|ic1k<(dv~nNR~M|cjmwGn4%_SW^_C1smMS&4+BvFO98PgDrS zp3c~OK1F{xi=`Kxe8Aw}pAFu!(Le(u25Pa6s^SR&!^!=x8pk=b zqTuS@dS*|I-amcu!Qtb7JN(v9cRqSB>WWZ!?}0@X6A;52aUHG_r@YbepUBjwWgcyUQ9rD7B*^ga*U=wO&6hi_aW4sdyT5)leZu7_ z&0k5t7?H+d%*76YlL7AF)R%bDhLeB?HHI8a-diZx_k2k&Vx<+)XD6bFeFxFdW1ymn zM5QMHK*uiVM%(-lg9n0Q}Oh} z@4@0gdqZ8lT(XFb1(A4lQTx*(2J~%{LjyDoC#xxgs1OfVtHQc(o7l>A115hL#a7th z!G{_?Zq&tsQ$F!2p|~J=UtD_pV-hQ(s8{3he2Tx@|9n?Er(Shna{or=_vUgY@%*n3;ul(WXErsA3qW21kk{ zQU-&;w*SvuJReKJtbbLIB|}BN+m4BLr))+N@Zv7qlRe*M;2(@zOFhCv6o(2Y4v5gi>4UzLu9x^k;m}*T94JNm>ln>S%j9$P*@LZ;QX4}8uCFMmr<0f2F$l-M&Tk^^`XRlJ)|ti`gxERb zL6CUzCD3abtZ3}1wGh?GsuJzj(vKqG7L5uyW1Ic>lE4(d(a)-#+ z;U@l#4S#?XsF}7^SkDF0?XY?*-svA7g(1S7O=Ljl2HCf7{MGWE%!YDbsPj?%y6#{O zFTQ^NegW+M7hl8BEWE3N=<9k{6Yyg1vABvp^y*@KYiNVMKijUW9ijXWC{DAVh@E>+N~FDwTzeo zx&=y7(6a_D=6D+I4`K}!8)r5?mdeQ;dQJ}>ZtBSCS~kz!im4 zHVo4;in2-vbVDa57*lDTFpSZX0qVs5*K)w3%HYV*f=-;~mO8Q)-CPYvmNI`=ASHZ4 zBuUr~6Y|t^E!%F^@;u8^%&}-G+sL+5=}PHsf;rSud*JR_wvC>WK_$=LVZ1qTuLgSX+=Qg)2m03&g z05dYGJ^<$V5GA`)Dm5)E2w12U7iA$96(l8&q#F<6VM$JJP5dd%l#hQsv*j$g2yFt1(p$| zuo#cIfnwt|SGD+xGK&aGq(XX&L zdgRVBTh)$k&sMdA79A64^>?b7kux}!k;Bf<_vQz9^sB&vk|a@x;{KxuqT&MclPVe} z`PVNjP^tEtgUD*`k4GQ;@ksNj)b_rflOZ8$Kew4$r0IWYYa%sChivAPL1k?DCp zz>k(_UO&_flfPml{5cGRs@yEtW`U`hy6S+PnW%^*CRGBd&apqyfNF(3_$$7>&$UmT zS)*!nff+lXd%2WPTX7+0_+EP?S%`aLroPQC(1TsnsiZC=hqJ88!Q%((MbihUX3^DF zi{NxuTw?`FxCOdEHN6g*sH&~%JmXrPL>Gy+=#xVmMH3ccay2ORqu>CR#pdr|=wtBI znQjVT9O~fs9V5fDj3f<<=UH$S^Rza!O3`rb>`*kkqA+8EypFi$b#JzCqoB7&g9n;x z`8QJ3<=;p?J5MC+Xb5@Wlp78Bwf*0pbhNJ8sD{Gk8(Y>Z`})Bw@XMI9|PP`EC%hu-L?@$+N`a0mjvrULu3s$~iq-IZzjb zM9RVbDzJkLAS-Pz+~Ki$4Cd;FvOUwkq9`jOe=pU-`9 zZua;^xb@SwpFKEtarXR$pYD7VJd#kJm8FqxSNF2t-en1S8DbuU2H)m*kAvk!3|JzO zQd7W6@dfNIV%q84LPqydvA^zHS;ci__o*GPaT zxjRWV{(SbG*)y-zN;LHS`_}42gKH`la|@ynl?5>t727_4UuM zoC@v@bEbwuqANkyDoM*!*GVf>*GX9HAFn?s1WVf<&D3op4f_y&K$bN{A!OeYwp6F@ zpi7LCKC0!3(>aEH2X*wpB(=;C>2$afhg`C$-44D&H@JkT{vi)>YT9q?+6hu1@UubF zeEgbI&IOVIYk?aqo#xn*k`|(ly`Wv@0!winN z=U$uJ%CNV1KAx2UF=5j>-@L)WMU zq99uhef(@ix|M{i&u%6O%fqEewS`{vixb(YXQ)PrgJhhi=>xQuY2LldW14sC^XxmT zZ<>RHbcn$io*l~bIwG-A#PbqFQ4w4=Ihg7dH4}ZCcHyWHtskBm^sL9uriSLZ!idF? zmxfyKXf}mWQB7{AWl1+Rf+IBS)v<$OkRPH(rXrF^CD#%0nd*?S);Hv-9Dh_+#It&q z8sk^(CNm$LuN^!g(jZTXQm&1}pVrF=6G%>91$+zB_mM)ZFi`HlJ8#&%gQ8 zt=DF61!p>#yLof&#G7-YAI@HW%k7*7h0c(X>L3<<`q~t*d{qDbtZHN5*APwBLE6}; zj04CZUL#av=ssoO^niXpLNwj}egylDOgm*-3Eb90!#xgjEhXb$YNP{9unc|aCDPjZ z|H6Mw5s>@O-*HBq4xdA`{D`?ja#eTBLAh8ciij|=P(tbqg%X`x6mU%99f@Q#Dkl_4 z4Bk+_mm|awTav4F=k$Cg^<=Eoj)m7@iao*9F2?GJ0@Xq(j#pfp@`+Eu$b#ev&pj#w zuA2^kMwWvpO=iU5R}T*G@V z@z0Y#_1!A$tgFnOBRkdA9vj)W^RhIL(_t_Qf3e)6>{KOe*lrrsAgLVXnAE0ZkfmV# zGI>^G#wM~C&N}DMNR|)43O1zMPAx-GA4K|eour))B`HR2i}%|qVOo|tH4xkSx^{)j zNE@0(ghvI4+9-(j8ilRxI4@hp;NTgqH!cW2R52NxW1r%W3$jwZ?D>wVjvvDqhYhk2 zm2f-~3dJI1K-XXs>MgP~r)mPhLJXcV+hRyTCq%(jjt#qx_wOfVMxOEWT9#)?mKA6^ zXZ2^BSb=OV7{NEX&H@s- zKb%8`<>MY_z_-39ooc>LcdU|N($ zOb{b60ZSh%vXCpgrtEvyl){4K$?HD(Cd+1H$jTWyJXIE7RjWmc2vX$fn|}-|S@egq zn@F<%hqHY3%2+`Hu(El&@`5B|yBH=fjphrhZW$lVjE@=!Zxg&OzYBM8rl#SqbInu~ z7hjuHi7iOJBMY)vqwvS{8{*C!V5(l=-@mTuq-Zev5CenTq{J+DFR%3wUUjPOhkNP) zKEUe$huPghdU&z81o$=@aGVHOO|Fsg z_YPSsU&qR{!!`^~spf*}7G+msnCv@SDu)Hdla70O_67tW-kdQQxK69MJ;>PoT^-X2 zk`r|!i#OjM(aY5S)}o<1?9v(VCoJAdxsf(#zeNTaE!NY50uD*{_U_HD9i!s%`)%iv zkxfAMP@fe2CJFvDM&tz;XiiOHo8T+y;P5-G(y%jfn&?d%yl9ziP2I7T)If)ZNHUeN z2dK)2*g#-`ZJ{Y_k|llAGSfMi+a4IOLdHV*hzEve9p%1JDEAe2+_y81yaS%h_9M&b z=-T)Nsx@0W2S!0P(yOPqQAZ(2L-INbG8%gjf=&O5(=mQ6!?QZpv~Ub*;vGgP2CZpE zcL4i(7&1C*jwG?*l2A2kYjCef$1h_!Oi?$8^AfnzPgCQg7EeDeWavgMh}(O26IqNz zTE*Zy*Fg@}hzs|!RjFOgqpYlPQ6VN)cTNVPVSx`u0MuoIU%fj$!LMFbAE0>tNlDg#{68V=*oYF;~lp=r*!_FMcs0=3&GMLP=DE*r)ldJjYjnq3}Aa zxhhs7xu!nIE6E@n65>+z2DE%MxZU7pifYUo4^DwfSj*@{(hIGav`x!E{Es!E4ZSR_ zYZ2R_`Ca zI}#B<`S%k?5n2a+meF=5j+DMD-7G`Up!b3ui~m1+-~Qahm96`)?CVo?egH5 z=A812M{I+~cHm4H?$lOmTiU4At?7r19nUSm1ju9Z8j_F<36O+!@kARIqdU8K;o%a8!J{* zdb@O{+_p$cN8s=rwq8!{wTB49DqZz0z0xW8n9B&&Q#xL{07n-}R~&ISzbu6zY(z4jBIN! zH*DS^m1%Xb>2<@W4{ExjWJZcr_9gq+qxix?IOHm!f1LV)Un`x5uYL)Z#3US@A+RoB z&A}3b@wC9IyHNVZ9@rIFZx`keuFlws#GlA|6tHVGr_PiXL6no0K$HWCVX&Y&KkeC| z!)D>)u^cfo|G4-qQH#ek^8M@oxHL|Te_T3EZSOTr{s2jUOLXU}eRobA zPWK1mS#{nN|;^ToB3*Z2SQGLi7Y>1&^o=IeW} zeeu)F*Ps9C<)3%IeC?BKX9>LFOY+m%pU#tOXRjZ;_T`gs@AU&9=<5gJ6DI5%uV2M? zT|1e$_67Wn&*8R{*Unx01OdzUU;6|;^W^n?tyO>Xr?WeDEFUP>gLwpM46*V=_#A&7 z2Jo)reL3YJ?8yiQ>-IvtL%d4Cz#{Xz-!QfDX<_^ zDT+&q61L+|-P>^k$5(c0Dkb0xTR8-Nrs^^gR|P9uYS8`WMCoX2 z>2T@zzmyJvQ~4^mm1q`DlBxZ0x*M#%FKxGS)mua`mPs3h^;;Og`8lU|-r9A&mMDE> z`^n>_%dMrOHXzQIHXzO!-}2mmIVCykpskf64xEmN=OB$F2~wciK=Mo|D8|HSd>&rX zc#<^R`V#ytMvBMCPGktgXfq|s9LS7u?`EPGCuH~(gEape?4%nkrh|5sc>z%XOii(P zYW4Gc+iH_n>rt*GdJ(t-jC=Uw+-Pm9941 zoD3Xk=7u;H|37lPXB{+$oQ$NCkfA*;ip3hwspCbPXZ&}Z? zIx|TdCduG4n#FinY=IxrllIFS=?J`Ra+P#b1*w8zhSVK`ApD`}0c`-wRy_Q}`!uO& zjwfjDLSlZ(7S6Ko$7z7-1xE}X=FHLZ6ZgnK5yCqASm5#NOK#J%G)1#Uc7!x%tdW0Q zI-vTG!PibqIa;qgTTVd{w%Yn}0Vk*S*!+J7>=4;S8$)k(TSryVzrWgefLHBs-$fV< zqSzyAA!Zc5A@JhaGDaXNeFX#j8i(n>A|gZeTvF4r4lB2U^&IrbT@=Fl`*`^AYbWo@ z^SsIQk@QM-X2I+u;7{Ew*LFP$!hmWzXNf)-*&nSk|FUZS5lqlOrZiP8PT&FiBL*o` z)TDnb9{Z!I)0D;aWz9A|pxjlZQo1S@Tdtn&VPL2k@cvd&A_fDQ=8`ZjoM``I&2TWy zM_^8MM@U_aI#giPyAF&AVX--m>vUwUEotuJQI zT%Ni3>8%(12T%v~k=#g<8K!fh=0p(_E4f7b*rO=Gkq_4G3Ou?%KRxhW1Xar_LwwY? zoND3yM4Wg(5szJYp+ra$#7OlPpEIC|YVEyahpY+o>lD8|C!STGVKsV6Lz zkxhLGaMtF?=~#o=6$5Z)wqWfSX@-rL87(Q+vXO!`Tl#dZfN9`5tUfT~TVOwM1Xx<6 zKyB6txW1$uLp69K{ABGrV7p>+S#tl{ijK9xa-7JTf>+=RrwBFd2n(uusw2#pagY88 z=3pnhJJGhSeGL9JS_{oao(5COu;Eh)$U_;XBg8Q0AtQXI6KH{_I?DB{_q_AkRkPv} zWwRhttn{l6f_m1Snr_YoQ5s7YJxDXtLzBMez01mJO5SNb_^&1l#cT9nWW45arVmXY zpFTDH2`q@|lhcQ%-6 zhC`18iP-c<_Gk~|Fpt5Ak4?YfTb#ewXrbR3iL9*`gG;K#uqeb_WaMiUR@L&FJ-+wm zB1Zt%LzWbE)!nnwHmsWd9e2J)kiY-I1e`8yZ zbAzf~@F^}?@fi*aG0{~d;lqP2$NBT%`(bHd#x(PZ)>fS=Ls{%bR+(-HaDVHtrl;~c zGmI7Wn0?1N`e7iXdwwMT!?IbHqcop)Y-o=1B++|$64t7_3e5bpK(*!Bu9gg?y2_30 zicw*v7_p#mCpG#Of32-IqSQ`|g>$AMN%n z$>pmO1Cb)TTy@#cTB374(a8*|Gt6@iP!(LuUv4~;zfDy+>E7aH73r2*@ls3gGabHVGlvB>AZIkD@;StddthJ=Y?ru4a2F6$ zObEO9G}k;cmuGh3#ha7w-+Jxbig{*XEN_=0)ho(#W3W9Rgsbqa!2EjdxejmY zAhSPNj4)MEXz?T^G?xTrD3@PP$n!MpeF_}``xzzLrN{__x_;HS40p`q4mrzy<}M}H zTV_%#7qr-gYfp43yI@fMcSokPg=tVn;fBYu4Un!Rgk940969b%H7lL5DWMI-z&117 zD5Gj4K3_%~CaPxABv;@QR=pOj@){V-2B;Jygw9aKe@XK15b~L;vDdxq>72-Hss6w*}gF z_mlqwYjvlAlcmEmNk2&uz%t3kZl5Y6~#HRvfGH%67>InoF7&=AqF->B;uEPkac z#R&s^_novr4NZY(lIFE+>%yV>7Fxf;3UiBizHCmcC_wwz+cJ|RA>fK=dPauqD;@{u zfs!26WdxXqS2WCa+op=K&Vdx!msn2?v)wXHut>)o$BpKlW6immDSoK6m{bL$+qdXo z?3y)7hjhkajvNB?-`x{Nb^ytR$+ZT#y1bC9zO&!9T;hSc=|N?M>U?N~D2%3TUWy!M z^5OoudLXVMqFsZ@s@u|4n$-$6Ik0c3Y>nyMQ6%giRV8fq(!}aUulLbe8jUo5XK5SN z)6{BhrxpyZ{IK0;7pd=F;IQ3P6RX=gx5?j=I&XmT;40Kj6-)(~Rq*^R*jGu4lMbG* zgXp+h*v7cwFcKsu5Q=9h#0ANP3$=0{^Z$GSSN2~!S^oW)c+Lw4?ed^6{Nz8XS?xhb z9OaVsvA3rqC<^iauX+5+RWK#qo1i|yFu5Wz2grE(tT$g_y z908=D_xMo|VI@6V;7tSEnyU$LR_hE?J!RMyxFPF|sqnNJi-6{C+T2tn8ra8Es{}B2 zFhl7GCj&IYk~Nk|)3)}eg(7L%21Yt*N7_xa4d-Mdp}Q#|%ST)gBVmVK5vb%#>|>8I ziVI;^sr_2?HBsiF;3O7YjtFcl=!n4DC*GtJl7xT@ImAzFp=+1Pz~@uua#{@+U04sg zSgVI7xnBc$iUHBNKpgACvjDp8W-iuYihStL(XXM{8yW3 zJ(veE6iq|y?=Gei0%J8fhE5CvPBY9T@GH&iAER*A2|tNpB7p$Y3C@nfcV>yo%Xh-H z@{90Jd{-@7Z*z~!*4xa?4AYTSz<2-_pMRctgf$^uhO8xPv4%7~Ggxi#VechZWth2V zOr6OYvtstef47??hUi6}BgQ%cO_z z1e>}5dc2@9Gf81TFy|f_EbGPU82C1%g9ZkD>Mv8Kz{dQ zwRFS7nCB_5HrQgB_|?x7)274sUoUhBqU7xZ0f)67@TjsP*q#-1Ee+56$unw-r-;r99K&2B z5(`8kfp|QITmSa$19^rVi-`^m=AI`lyT+k)zASBQ{s55(w{L$2Mnz1jWE1ypz_T4Y zo+NsK;O2vH$joT8SjPh_=W|x`g}nBu#$$E=y;~Tziabg@0QJj8#lN0c7e%X_0sRV+WP!VYwi| zkzlg}wmo2Z(>lUU_&$n44Dig+bHKkxsAXz7K@gsJI3fSKeLtMR%F}{pi9x34xQcK; zajCUSO>RhYHr}~vdJ5_i3&ysG|bRBl_ zIp^3Rt%#%6FygV;Fp`iE7u{7^W%#M1r#gIDMZ=!$d$iUdx-zY`&1&6dfQ_PYq%kt# zl~v!SUKMzDeET|_#Q*wJ6(faT?2ZUFpoCosoX4ReG)KQbO4W6wO;52Oj@B{x%T<*d z!G@DiYuktK!9SnrMR!D2Exr zl%|)lLvrBvGTf3V0gJNup+*L@0xwR0GfWYZil(>>aH1t0g9OwJn$~jpjD>s! zIi9YgxncR#S2!R9q`FI{%1&#_nR@-bW(`_NJI#vjXlmiB_foQs+wNs!K^1Q7SbpE~ z;b_A_I-+R$e05ZTYhc%aN~@zZitCuJ=w_DgP=}RDQIDWm5$s+GxG)b4A8iZ+Y!t>z zRGQZ5PGr0)G@chTpdeWnvKZvGJS)I&F&>ZAq4h?YbEyR!<%yo<5flQCX|P+RMCI@l z;o`3&i?@vG4lD#$3*DN9;Oc2Td|>v4zux-l*b15l;V3MZ>YeG)UoL|>l`|voK*=k~ zmJ8D#7b1bW4FlLc8W5-tLy)+H^GwStj11{oj`U)MU>y~-8Oq59(b0(LKR+79ZUWIo zPv@E38-!3gR=QC7vUH|&W$GYlp4tm%mr56qEPHBqX%eZj*TIGui$B>2!YaK{rYJ9c z3&S{Hnw)yIB0BeLiPF0ShJJNwk4&YJD8W+nA!X zY_h>{z{=f{2Ptynez4>~;F1t@<&ZqDx_@(6+BU3Zsf4=O(`tyIX?=fF zl?a;FgEJvx{C>roy*sJOnL;|*kmf(15Q_*Bwl+qhLM(EBD-1p%>cR>1A%&;6c93qS z4r^M9^keSjyYUJlW)HtS^UBfLLsxI_-ZQ)V(#%=^ylKm2 z$ww2S2x9Zh^x4!y$)L_vZWbq63`=k6Wn}~2u?G8#6NcTnu^7(Ts@8-9nv!8WM>?#$ zTPjU9c*O$Hi(XAn@^b6VJj9tG3DKzF#JLcI&K#1*BvRC_-WS&(6n;x&H@|s#=Jb)7 zgP&sI*xRqJzVoW4^2DiADSM{ zx?C>n1*u-)u~V5<&7s#pj(RZg2Y=$8)CNV8D%X5~YGx(i_;ncxt z!R3uqPql>O6=^xJYdlLTE`%a$wVfVJnVD~o&3yfS*_2sXJ4Ol!QCIuN<6Q}($|Iz? z%hFj3Gh2Ezt}jC^Or=Cj^jISnMS_@UxH?+7(`#SU%=4#a-}&O!p-VI4uin`E#xLJq znR)S8gSl}x_U^y2ch8Nz-`?2!@s06wGY2l-{@a(nh0$Z>IN_SzPC8$_UWs+YJjXhu zkPwg5T@};M3QWgtf4CD&{3Y5Dr9e;!M+Hg5ZlAFQXVp!J3vZo=<7h62wl<7y++d#u z1yK+q4@pTbUx!>uWg}!@n5zCe6StW;ZenYr;*3=XrhB-W;%UBe^&3%an~BBi1{;5D7Rjww zUYhxKeCEK*E4fI*U~W~%B1|8NmTo$t(kv2%H(XAoThlFq*h&5i2L`BKsLZ@n0fU!3 zQX;_2#{@A{t)Beg7thDOSi+DqI^PM__)5#bgRM{&h#usc8cyGff4Bj6NFV+AIL z?*hfVJR3_L)JRcmy8%G!q{As(w+JPU|PZICI|$|XUm!FJDbNeV6jE6YyL zkrXp2SIWg-$d9Qt*xu95bfu1cr_D+^Kv!DAwb6;fOP~`)vHvCZQ^JGBrN6yL7em71qs7?VoxH^;+rb z)b6Q$aP=I#b_R~hwRjDr=t}R|DbEAY&07F+GdRus7XM9vy>&< zs@%7HHdFBK2Iebz@2#}BW2t#;uZd#kRLDhMuoh8EtY;|}jr~6ys-_w^8_N>+3_0Lp z?Gukuk_0hOooDr9*ZrPRy@i2k0Poiqt2bKB>*g7QkpbA9l3L0?&*d|!&ZrW-FWb%T zd#byW;9ga7ho)<)md|K9Ram0|r*~v1Gc^MY98kaYR3LZIA(QG{b_8WA213ZQfq?x_ zu2Q^5-$Lu;nBW@8t)y)hjtap*wQc5$CY};m=LRQ?q!kYC3g}!=lvFiWJVNGMARSy~ zIZD7KQ?UNY|8jsuWGRSU=#YwEg10Dg+(cXu8f#TJ)zZ3s>Gj)J-xIvVc;~}YcMl%kvE%N0`Lswl)P_X*oYu_Izto1LCA@1~F&~HOK~?=Zp*;_AP;p zqnZ2I#bEX$LO4=A`@W(+ZR99j@`t!a4n;5v3H@N>Y&t;taPz}Vp zHK)ndGB)|b?RSpOzI9;c<+oPcT@s^0C|12-{1|9k!KgN|35qmx)yOj$mJUG$iFm@7rvEj+-Nn>#5&T13dOmQg?t zBE2i&InqrPro!<$Vu0JOk{nfRlGi*n%3zJF+q>PW=?YCd60ma@5}Rp(r|mt+P?#O? z>SwBIm^zcQb2$3F8fY7xrdqc3`hEZgXp-qII_w_0=#2A4YRCB^OFjotL3H(o`9WaW zqz)+-7BQ@NHr9XM!vZ@>WeCdz8b$>r0}r!*K$RXVr4_hLLa{)=QCKHca^#Z740+6E znUSik4nGXa+cZlVVz8VTS+7w0_rJE1oMjj!t>Ahkie{`MDVkwec&FQI&RG3rb=w>h zFekoymlB|DXx1(#?8vYjuS0}v#6gQ7#vO6LxaY`UFi@l8?$I|xI%rbIiqw@PMu=fA zis(w{de&nKA}rZ;!muf)iW7N+mB80X$|C)l;(4{)`74ZdQj+Tg+XM-J? zUjo&6CC?QrvcjlJto*t;6nIr@0j*85Xr0?LL8vTidJ4}?&5ly=g-r*WYe4fO0-<-9;gH@THvNZp=GEZBEX~D-S&L+KhMG?k5>%mZ{2E+YYT-IGBHh%l) zrCYD~N1a)|xNImYgj_X5kG^W_<+84UI^g}j!V&qWJAyDbWsqXxF|>%R0;$IDHtAGcv=kuRCRM|1PLg2QUw1|IT;2IV!>Odf2AuLs!5uO9B&-8C?Nm2`yyh?O9AT0#1Z3SHkU!19A*g|XqJEmkZIco~5 zUTRgzMU%?eoSnSTng6D-K|^!W2a2J1?c78tzNC<%HE2)LKCm;^E4WoL5W_A81LDPV z%UTifklK})UO>=P|N3i_!4)2GBnm^7-6P~NbQH(x{!pk|clpUlv{^&2SQ@SrWdk+9 zdCM}zzqcchZh>qR$auc8n`M{ag1!8oo4ngFAy_n*K{?bDULv^@7nt~*51bg)itQz z+a{0?#W3w1rQswr%qiJGQ=0$fvVjkoqu^$wNz4&VYms7$*t)SLj9m~xSO~h>$UI}R zmq|3Og>dRRTxEt)6L!3v!-1|BFB%4?WtvSc!#Fj|3!=OG1Bhk7!Pa#ml3`%j?MRiD$uP(xs{KsRMI-rmrr-gx3E|Q0Ru%o!V16 zGqrbWKPjDn!JW4oj?YXz@0L+h@0M*%3x!mfs5wE2V6{mI`ve=87z{aNR#{S#_a!CO zXvXy%34tX|tb7WV6IkM%zwah0*2qz$w#aB&mf%WzSZ;US_W6=^u$??bMqqEws@iZX zj?f;L{mp$ZF?jn%7#PbkmWK(uKTv{rZ4MG&ZSlu#D8$`U&m$<})OhKljT4BVh;VRr z>P1+37vZ^`heX_mt>mT4^Ay9S3k|p)YYrI}( z{Qg865dx95PPyNb=FKZ_&R+Rw_Pw(!=tD~}A?P9=@w^JRZ?~~2e%}Au82kfkzVx2s zfrQ6!w+*&AIkmrZCPBXY^ttCsN8qXb8W!nAWH+CB*+wz_hRmIug?C>jWDGy&j16^q zro2?ig8!Z({j9(XRIzXU01+dI`@g5asVE(XFENf_5~Xiy{HYy3@Ba9^OAkoHFB|<~ zbzHSP^k(Nk{MTQ#sqiPLk;!V(f+iI)5{NheH_K<nBWF1$ z-I{7=b9`dTFjeQBw7?v!PEuw4N}ks*XzhqAz2GMm)TJ;P#;kIg3{z}Ot8J%tcDeS& z4K`f=$g0>FH&}ecBF&~YtomOHu_$&v!p;{H#E65lw?(X-qFz!SVAV>GALK|cW0@5W zgaspl7^@C*`0=j$SzZU<8P4bY>xOfpw{i_;%A}-+7yP%pJAEH>@S!VLU=}a85mZnA8U$l~E9>s@Gk5o$z5DJNa5Cbt z*q!J1`QK)i4~#L6?Fmw}dYkcF%Zq5W++;d6EgkCz2{`rgPDZlY5Bp?oxY63`Ix<}0 zI}Le+Sn?<{Ot57}$bgpD=3}DNe+9xI_?ms1!VO38y`ls?GJeOgH$i#~k+8aQAdwaA z|AncK2qCH38_9SryqP{UeSG@V^e3RwrcX{Eo_=HcO*nmL`cPu}b$I0+{70tWhJzyn z?tKkjeiKekk?9G%@*3RyFsggR?i`tsni?2Y(DFuMzxFZCKi9Dv6PsLhCiU+_&=h+oh!eL1Wg{q%>f7rKEFA)rSm(EYU zj;ObIa@7vzjg!*1bC9rCOBa0v`B#@MQ+mI2c52Vmc0>ihIV9 z1n~BTL_1Sc{#V|{E$v(Swsdq<1Rtr`JrU1_PdHp7F7+5Ky-t?nx?D$_#PTh@Xbsru zM_&S9zGw=oXU|!tsSgFPKOz{e-glB|k5fONL#dHL`FZ#I+qakAg@wKw7WyQ-fD64_ zQ*snR#AKFB>fA65VP`%)W@fBhax54K1#=iNtYg^=4fxI)P7 zmXfBr&*bWI;e1r3R;vbN4~EzpcHz8D;x6Pr#vHP%9*|5{az0&g2Oa9%na5{L6vDCU z#GN0$8f=?jGS5@h~bDfc{T^XL_HqO_euH`fS8MN9)WLDq6i96 z7nQcpN`|ZVlA#wCWGLrORWRTuN_8}ro+^sQ^jZ{-CaKM_V9?J5OGy~cn)n9 z-`zFJ4x# za+;1hKTWp3L-BZ$mhZ=RZZdM9QJML}>u~n-?vH=o{a!6|w||Sv-0j!qyG54S+O4B{ z#a|>A5kR-%|_d^5QPg7j8$zU5IZQrTW zA+sV3`!umx!+IL>w3bmxue_DZ**f4b2DBkXTwp6zqBoMa=ncvvB*a|d>NWD@L|Z}U z3hAJ#|L8&u(&T8skx{fyJW3}j1YJlaMq#m5g^iS@ZO{?5$yM=E&=lBIi6&>rK-v^Y zAIs;N{)9n#H2B;RvXR-W${mW8B<--1+sWd*UObV_TQy7Ln+Dv4Vgti0W&iP&uM@+x zZabrnB{(3@dFbQFvEg>e8mN&z8MKP_KB19QqS z%XpoPrACPTSJ4-4wxO{)20H)cneZ6=BF}_(?08a`yFR#WY;YT#uX-(ss9+V*;aNfq zfVC0Cel|gr!mYovxTj=(l$XcJS6*$fL zClIblF<;35mQOqi2}>}lSJ_%Kf{V82y;o+w9iP2&sE#PjvNR*0sG>r^#n-pI3b9|Z z6=J{ChzfC`#Blm>Q%+~KEO|!d*tWJ_=n#m(sG<4y2?sP;W_klJ3(=3bG$50@dDxrib610?Xt^Wu?L5Ek_vFT zPEHO9^5{APPL1rk0sB!!%YhXLJm)@eCBPj1(9v@@x$`G zmuerN`uAtEH23}4AN&`0Enm_%7zQ!a4afiJ8iODnN>rE;E_2oYLaF69MadQHopFI_EtTl%te#m;73v6-tIX>aP~>5NJKH|eEmE`w*UYs?^9H*NAQ zlF}PfZ%w^0HD0>nSTf*^E0RA>?M|zEPcGK^AI;(>ezq1%aic>Y!-2sZ$TF}5%%VB~S#H1L-bKf2N?pUFCTrss#6O+7#i2)&-!~(8zBQY?_^ei(*G1Z$b zel@DI+z)5rAcbwhyfqdO;&DL?3qf(ralB*4lj}&q{*&-AO7XN=prl=R*3c9aBu9+- zPlZn@PUtK_RF)%5m3datYy`ZlWx!p