Skip to content
Closed
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 5 additions & 2 deletions dev/archery/archery/integration/runner.py
Original file line number Diff line number Diff line change
Expand Up @@ -137,10 +137,13 @@ def _gold_tests(self, gold_dir):
skip.add("Go")
skip.add("JS")
skip.add("Rust")
if name == 'zstd':
Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

A downside to removing this comment completely is that someone might not easily figure out how to skip tests based on name (when working on another implementation).

Do you mind adding a comment to the effect that if an implementer is skipping one of the compression formats, they could use if name == 'format'?

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Added a link back to this PR>

skip.add("Java")

# See https://github.com/apache/arrow/pull/9822 for how to
# disable specific compression type tests.

if prefix == '4.0.0-shareddict':
skip.add("Go")

yield datagen.File(name, None, None, skip=skip, path=out_path)

def _run_test_cases(self, producer, consumer, case_runner,
Expand Down
8 changes: 6 additions & 2 deletions docs/source/status.rst
Original file line number Diff line number Diff line change
Expand Up @@ -126,7 +126,7 @@ IPC Format
+-----------------------------+-------+-------+-------+------------+-------+-------+-------+
| Sparse tensors | ✓ | | | | | | |
+-----------------------------+-------+-------+-------+------------+-------+-------+-------+
| Buffer compression | ✓ | | | | | | ✓ |
| Buffer compression | ✓ | ✓ (3) | | | | | ✓ |
+-----------------------------+-------+-------+-------+------------+-------+-------+-------+
| Endianness conversion | ✓ (2) | | | | | | |
+-----------------------------+-------+-------+-------+------------+-------+-------+-------+
Expand All @@ -139,6 +139,8 @@ Notes:

* \(2) Data with non-native endianness can be byte-swapped automatically when reading.

* \(3) LZ4 Codec currently is quite inefficient. ARROW-11901 tracks improving performance.

.. seealso::
The :ref:`format-ipc` specification.

Expand Down Expand Up @@ -221,7 +223,7 @@ Third-Party Data Formats
+-----------------------------+---------+---------+-------+------------+-------+---------+-------+
| ORC | R | | | | | | |
+-----------------------------+---------+---------+-------+------------+-------+---------+-------+
| Parquet | R/W | | | | | R/W (1) | |
| Parquet | R/W | R (2) | | | | R/W (1) | |
+-----------------------------+---------+---------+-------+------------+-------+---------+-------+

Notes:
Expand All @@ -231,3 +233,5 @@ Notes:
* *W* = Write supported

* \(1) Nested read/write not supported

* \(2) Through JNI bindings to datasets.
7 changes: 4 additions & 3 deletions java/compression/pom.xml
Original file line number Diff line number Diff line change
Expand Up @@ -44,8 +44,9 @@
<version>1.20</version>
</dependency>
<dependency>
<groupId>io.netty</groupId>
<artifactId>netty-common</artifactId>
</dependency>
<groupId>com.github.luben</groupId>
<artifactId>zstd-jni</artifactId>
<version>1.4.9-1</version>
</dependency>
</dependencies>
</project>
Original file line number Diff line number Diff line change
Expand Up @@ -21,7 +21,9 @@
import org.apache.arrow.vector.compression.CompressionUtil;

/**
* A factory implementation based on Apache Commons library.
* Default implementation of factory supported LZ4 and ZSTD compression.
*
* // TODO(ARROW-12115): Rename this class.
*/
public class CommonsCompressionFactory implements CompressionCodec.Factory {

Expand All @@ -32,6 +34,8 @@ public CompressionCodec createCodec(CompressionUtil.CodecType codecType) {
switch (codecType) {
case LZ4_FRAME:
return new Lz4CompressionCodec();
case ZSTD:
return new ZstdCompressionCodec();
default:
throw new IllegalArgumentException("Compression type not supported: " + codecType);
}
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -32,8 +32,6 @@
import org.apache.commons.compress.compressors.lz4.FramedLZ4CompressorOutputStream;
import org.apache.commons.compress.utils.IOUtils;

import io.netty.util.internal.PlatformDependent;

/**
* Compression codec for the LZ4 algorithm.
*/
Expand All @@ -45,7 +43,7 @@ protected ArrowBuf doCompress(BufferAllocator allocator, ArrowBuf uncompressedBu
"The uncompressed buffer size exceeds the integer limit %s.", Integer.MAX_VALUE);

byte[] inBytes = new byte[(int) uncompressedBuffer.writerIndex()];
PlatformDependent.copyMemory(uncompressedBuffer.memoryAddress(), inBytes, 0, uncompressedBuffer.writerIndex());
uncompressedBuffer.getBytes(/*index=*/0, inBytes);
ByteArrayOutputStream baos = new ByteArrayOutputStream();
try (InputStream in = new ByteArrayInputStream(inBytes);
OutputStream out = new FramedLZ4CompressorOutputStream(baos)) {
Expand All @@ -57,8 +55,7 @@ protected ArrowBuf doCompress(BufferAllocator allocator, ArrowBuf uncompressedBu
byte[] outBytes = baos.toByteArray();

ArrowBuf compressedBuffer = allocator.buffer(CompressionUtil.SIZE_OF_UNCOMPRESSED_LENGTH + outBytes.length);
PlatformDependent.copyMemory(
outBytes, 0, compressedBuffer.memoryAddress() + CompressionUtil.SIZE_OF_UNCOMPRESSED_LENGTH, outBytes.length);
compressedBuffer.setBytes(CompressionUtil.SIZE_OF_UNCOMPRESSED_LENGTH, outBytes);
compressedBuffer.writerIndex(CompressionUtil.SIZE_OF_UNCOMPRESSED_LENGTH + outBytes.length);
return compressedBuffer;
}
Expand All @@ -71,8 +68,7 @@ protected ArrowBuf doDecompress(BufferAllocator allocator, ArrowBuf compressedBu
long decompressedLength = readUncompressedLength(compressedBuffer);

byte[] inBytes = new byte[(int) (compressedBuffer.writerIndex() - CompressionUtil.SIZE_OF_UNCOMPRESSED_LENGTH)];
PlatformDependent.copyMemory(
compressedBuffer.memoryAddress() + CompressionUtil.SIZE_OF_UNCOMPRESSED_LENGTH, inBytes, 0, inBytes.length);
compressedBuffer.getBytes(CompressionUtil.SIZE_OF_UNCOMPRESSED_LENGTH, inBytes);
ByteArrayOutputStream out = new ByteArrayOutputStream((int) decompressedLength);
try (InputStream in = new FramedLZ4CompressorInputStream(new ByteArrayInputStream(inBytes))) {
IOUtils.copy(in, out);
Expand All @@ -82,8 +78,7 @@ protected ArrowBuf doDecompress(BufferAllocator allocator, ArrowBuf compressedBu

byte[] outBytes = out.toByteArray();
ArrowBuf decompressedBuffer = allocator.buffer(outBytes.length);
PlatformDependent.copyMemory(outBytes, 0, decompressedBuffer.memoryAddress(), outBytes.length);
decompressedBuffer.writerIndex(decompressedLength);
decompressedBuffer.setBytes(/*index=*/0, outBytes);
return decompressedBuffer;
}

Expand Down
Original file line number Diff line number Diff line change
@@ -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.arrow.compression;


import org.apache.arrow.memory.ArrowBuf;
import org.apache.arrow.memory.BufferAllocator;
import org.apache.arrow.vector.compression.AbstractCompressionCodec;
import org.apache.arrow.vector.compression.CompressionUtil;

import com.github.luben.zstd.Zstd;

/**
* Compression codec for the ZSTD algorithm.
*/
public class ZstdCompressionCodec extends AbstractCompressionCodec {

@Override
protected ArrowBuf doCompress(BufferAllocator allocator, ArrowBuf uncompressedBuffer) {
long maxSize = Zstd.compressBound(uncompressedBuffer.writerIndex());
Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is it possible to cause too much waste of memroy in the worst case?

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think this is just a fact of life with compression algorithms. There are potentially ways of mitigating memory waste if changed the APIs around a little bit (trying to pack more data into a single ArrowBuf) or consolidating buffers into a larger buffer if compression level is high. But this is the guidance the documented methods give for sizing buffers.

long dstSize = CompressionUtil.SIZE_OF_UNCOMPRESSED_LENGTH + maxSize;
ArrowBuf compressedBuffer = allocator.buffer(dstSize);
long bytesWritten = Zstd.compressUnsafe(
compressedBuffer.memoryAddress() + CompressionUtil.SIZE_OF_UNCOMPRESSED_LENGTH, dstSize,
/*src*/uncompressedBuffer.memoryAddress(), /*srcSize=*/uncompressedBuffer.writerIndex(),
/*level=*/3);
Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Should we make the level configurable?

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

maybe at some point, the default is 3. LZ4 also has a compression level which I don't think we are setting.

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'll open up a JIRA to track configuration.

if (Zstd.isError(bytesWritten)) {
compressedBuffer.close();
throw new RuntimeException("Error compressing: " + Zstd.getErrorName(bytesWritten));
}
compressedBuffer.writerIndex(CompressionUtil.SIZE_OF_UNCOMPRESSED_LENGTH + bytesWritten);
return compressedBuffer;
}

@Override
protected ArrowBuf doDecompress(BufferAllocator allocator, ArrowBuf compressedBuffer) {
long decompressedLength = readUncompressedLength(compressedBuffer);
ArrowBuf uncompressedBuffer = allocator.buffer(decompressedLength);
long decompressedSize = Zstd.decompressUnsafe(uncompressedBuffer.memoryAddress(), decompressedLength,
/*src=*/compressedBuffer.memoryAddress() + CompressionUtil.SIZE_OF_UNCOMPRESSED_LENGTH,
compressedBuffer.writerIndex() - CompressionUtil.SIZE_OF_UNCOMPRESSED_LENGTH);
if (Zstd.isError(decompressedSize)) {
uncompressedBuffer.close();
throw new RuntimeException("Error decompressing: " + Zstd.getErrorName(decompressedLength));
}
if (decompressedLength != decompressedSize) {
uncompressedBuffer.close();
throw new RuntimeException("Expected != actual decompressed length: " +
decompressedLength + " != " + decompressedSize);
}
uncompressedBuffer.writerIndex(decompressedLength);
return uncompressedBuffer;
}

@Override
public CompressionUtil.CodecType getCodecType() {
return CompressionUtil.CodecType.ZSTD;
}
}
Original file line number Diff line number Diff line change
Expand Up @@ -80,6 +80,10 @@ public static Collection<Object[]> getCodecs() {

CompressionCodec lz4Codec = new Lz4CompressionCodec();
params.add(new Object[]{lz4Codec.getCodecType(), len, lz4Codec});

CompressionCodec zstdCodec = new ZstdCompressionCodec();
params.add(new Object[]{zstdCodec.getCodecType(), len, zstdCodec});

}
return params;
}
Expand Down