[enhance](orc) Optimize ORC Predicate Pushdown for OR-connected Predicate #43255
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
Thank you for your contribution to Apache Doris. Since 2024-03-18, the Document has been moved to doris-website. |
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
Contributor
Author
|
run buildall |
|
TeamCity be ut coverage result: |
Contributor
Author
|
run buildall |
Contributor
Author
|
run buildall |
suxiaogang223
force-pushed
the
fix_orc_pushdown2
branch
2 times, most recently
from
921f87f to
5056b99
Compare
Contributor
Author
|
run buildall |
|
TeamCity be ut coverage result: |
Contributor
Author
|
run external |
1 similar comment
Contributor
Author
|
run external |
Contributor
Author
|
run external |
Contributor
Author
|
run buildall |
|
TeamCity be ut coverage result: |
morningman
reviewed
Nov 10, 2024
| bool OrcReader::_init_search_argument( | ||
| std::unordered_map<std::string, ColumnValueRangeType>* colname_to_value_range) { | ||
| if ((!_enable_filter_by_min_max) || colname_to_value_range->empty()) { | ||
| bool OrcReader::_init_search_argument(const VExprContextSPtrs& conjuncts) { |
Contributor
There was a problem hiding this comment.
For a certain query, the conjuncts is identical in all orc readers.
So I think we can just init search argument once and use the result for every orc reader?
Contributor
Author
There was a problem hiding this comment.
it is possible to cache searchArgument into member variables of OrcReader
Contributor
Author
|
run external |
16 tasks
morningman
pushed a commit
that referenced
this pull request
Apr 23, 2025
### What problem does this PR solve? ### Release note relate pr: #43255 Improved ACID table column handling - Added support for ACID column prefix in ORC column initialization - Fixed column name handling for ACID tables - Improved type mapping for ACID table columns
koarz
pushed a commit
to koarz/doris
that referenced
this pull request
Jun 4, 2025
### What problem does this PR solve? Related PR: apache#43255 Problem Summary: Example: ```sql CREATE TABLE table_a ( id INT, age INT ) STORED AS ORC; INSERT INTO table_a VALUES (1, null), (2, 18), (3, null), (4, 25); CREATE TABLE table_b ( id INT, age INT ) STORED AS ORC; INSERT INTO table_b VALUES (1, null), (2, null), (3, 1000000), (4, 100); ``` run sql ``` select * from table_a inner join table_b on table_a.age <=> table_b.age and table_b.id in (1,3); ``` When executing this SQL, the backend generates a runtime filter on the table_a side during the join operation, resulting in a condition like WHERE table_a.age IN (NULL, 1000000). It’s important to note that since <=> is a null-aware comparison operator, the IN predicate must also be null-aware. However, the ORC predicate pushdown API does not support null-aware IN predicates. As a result, our current approach ignores null values, leading to an empty result set for this query. To fix this bug, we’ve adjusted the logic so that predicates with null-aware comparisons are not pushed down, ensuring the correct result as follows: ```text +------+------+------+------+ | id | age | id | age | +------+------+------+------+ | 1 | NULL | 1 | NULL | | 3 | NULL | 1 | NULL | +------+------+------+------+ ```
koarz
pushed a commit
to koarz/doris
that referenced
this pull request
Jun 4, 2025
### What problem does this PR solve? ### Release note relate pr: apache#43255 Improved ACID table column handling - Added support for ACID column prefix in ORC column initialization - Fixed column name handling for ACID tables - Improved type mapping for ACID table columns
suxiaogang223
added a commit
to suxiaogang223/doris
that referenced
this pull request
Jun 24, 2025
…cate (apache#43255) Problem Summary: This issue addresses a limitation in Apache Doris where only predicates joined by AND are pushed down to the ORC reader, leaving OR-connected predicates unoptimized. By extending pushdown functionality to handle these OR conditions, the aim is to better leverage ORC’s predicate pushdown capabilities, reducing data reads and improving query performance.
suxiaogang223
added a commit
to suxiaogang223/doris
that referenced
this pull request
Jun 24, 2025
…4615) In the old logic, the `check_expr_can_push_down` function does not check whether the `orc::Literal` are constructed successfully, but only checks during `build_search_argument`. However, if it is found that the `orc::Literal` fails to be constructed after `builder->startNot`, it will fail because the builder cannot end `startNot`. Therefore, we advance the behavior of constructing `orc::Literal` to the `check_expr_can_push_down` function and save the result to the map, so that it will never fail in the `build_search_argument` phase. Related PR: apache#43255
suxiaogang223
added a commit
to suxiaogang223/doris
that referenced
this pull request
Jun 24, 2025
Related PR: apache#43255 Problem Summary: Example: ```sql CREATE TABLE table_a ( id INT, age INT ) STORED AS ORC; INSERT INTO table_a VALUES (1, null), (2, 18), (3, null), (4, 25); CREATE TABLE table_b ( id INT, age INT ) STORED AS ORC; INSERT INTO table_b VALUES (1, null), (2, null), (3, 1000000), (4, 100); ``` run sql ``` select * from table_a inner join table_b on table_a.age <=> table_b.age and table_b.id in (1,3); ``` When executing this SQL, the backend generates a runtime filter on the table_a side during the join operation, resulting in a condition like WHERE table_a.age IN (NULL, 1000000). It’s important to note that since <=> is a null-aware comparison operator, the IN predicate must also be null-aware. However, the ORC predicate pushdown API does not support null-aware IN predicates. As a result, our current approach ignores null values, leading to an empty result set for this query. To fix this bug, we’ve adjusted the logic so that predicates with null-aware comparisons are not pushed down, ensuring the correct result as follows: ```text +------+------+------+------+ | id | age | id | age | +------+------+------+------+ | 1 | NULL | 1 | NULL | | 3 | NULL | 1 | NULL | +------+------+------+------+ ```
suxiaogang223
added a commit
to suxiaogang223/doris
that referenced
this pull request
Jun 24, 2025
…ins (apache#45104) Related PR: apache#43255 Problem Summary: Should ignore null values when the literals of in_predicate contains null value, like `in (1, null)` For example, init table in hive: ```sql CREATE TABLE sample_orc_table ( id INT, name STRING, age INT ) STORED AS ORC; INSERT INTO TABLE sample_orc_table VALUES (1, 'Alice', 25), (2, NULL, NULL); ``` select result in Doris should be: ```sql mysql> select * from sample_orc_table where age in (null,25); +------+-------+------+ | id | name | age | +------+-------+------+ | 1 | Alice | 25 | +------+-------+------+ 1 row in set (0.30 sec) mysql> select * from sample_orc_table where age in (25); +------+-------+------+ | id | name | age | +------+-------+------+ | 1 | Alice | 25 | +------+-------+------+ 1 row in set (0.27 sec) mysql> select * from sample_orc_table where age in (null); Empty set (0.01 sec) mysql> select * from sample_orc_table where age is null; +------+------+------+ | id | name | age | +------+------+------+ | 2 | NULL | NULL | +------+------+------+ 1 row in set (0.11 sec) ```
suxiaogang223
added a commit
to suxiaogang223/doris
that referenced
this pull request
Jun 24, 2025
relate pr: apache#43255 Improved ACID table column handling - Added support for ACID column prefix in ORC column initialization - Fixed column name handling for ACID tables - Improved type mapping for ACID table columns
suxiaogang223
added a commit
to suxiaogang223/doris
that referenced
this pull request
Jun 25, 2025
…cate (apache#43255) Problem Summary: This issue addresses a limitation in Apache Doris where only predicates joined by AND are pushed down to the ORC reader, leaving OR-connected predicates unoptimized. By extending pushdown functionality to handle these OR conditions, the aim is to better leverage ORC’s predicate pushdown capabilities, reducing data reads and improving query performance.
suxiaogang223
added a commit
to suxiaogang223/doris
that referenced
this pull request
Jun 25, 2025
…4615) In the old logic, the `check_expr_can_push_down` function does not check whether the `orc::Literal` are constructed successfully, but only checks during `build_search_argument`. However, if it is found that the `orc::Literal` fails to be constructed after `builder->startNot`, it will fail because the builder cannot end `startNot`. Therefore, we advance the behavior of constructing `orc::Literal` to the `check_expr_can_push_down` function and save the result to the map, so that it will never fail in the `build_search_argument` phase. Related PR: apache#43255
suxiaogang223
added a commit
to suxiaogang223/doris
that referenced
this pull request
Jun 25, 2025
Related PR: apache#43255 Problem Summary: Example: ```sql CREATE TABLE table_a ( id INT, age INT ) STORED AS ORC; INSERT INTO table_a VALUES (1, null), (2, 18), (3, null), (4, 25); CREATE TABLE table_b ( id INT, age INT ) STORED AS ORC; INSERT INTO table_b VALUES (1, null), (2, null), (3, 1000000), (4, 100); ``` run sql ``` select * from table_a inner join table_b on table_a.age <=> table_b.age and table_b.id in (1,3); ``` When executing this SQL, the backend generates a runtime filter on the table_a side during the join operation, resulting in a condition like WHERE table_a.age IN (NULL, 1000000). It’s important to note that since <=> is a null-aware comparison operator, the IN predicate must also be null-aware. However, the ORC predicate pushdown API does not support null-aware IN predicates. As a result, our current approach ignores null values, leading to an empty result set for this query. To fix this bug, we’ve adjusted the logic so that predicates with null-aware comparisons are not pushed down, ensuring the correct result as follows: ```text +------+------+------+------+ | id | age | id | age | +------+------+------+------+ | 1 | NULL | 1 | NULL | | 3 | NULL | 1 | NULL | +------+------+------+------+ ```
suxiaogang223
added a commit
to suxiaogang223/doris
that referenced
this pull request
Jun 25, 2025
…ins (apache#45104) Related PR: apache#43255 Problem Summary: Should ignore null values when the literals of in_predicate contains null value, like `in (1, null)` For example, init table in hive: ```sql CREATE TABLE sample_orc_table ( id INT, name STRING, age INT ) STORED AS ORC; INSERT INTO TABLE sample_orc_table VALUES (1, 'Alice', 25), (2, NULL, NULL); ``` select result in Doris should be: ```sql mysql> select * from sample_orc_table where age in (null,25); +------+-------+------+ | id | name | age | +------+-------+------+ | 1 | Alice | 25 | +------+-------+------+ 1 row in set (0.30 sec) mysql> select * from sample_orc_table where age in (25); +------+-------+------+ | id | name | age | +------+-------+------+ | 1 | Alice | 25 | +------+-------+------+ 1 row in set (0.27 sec) mysql> select * from sample_orc_table where age in (null); Empty set (0.01 sec) mysql> select * from sample_orc_table where age is null; +------+------+------+ | id | name | age | +------+------+------+ | 2 | NULL | NULL | +------+------+------+ 1 row in set (0.11 sec) ```
suxiaogang223
added a commit
to suxiaogang223/doris
that referenced
this pull request
Jun 25, 2025
relate pr: apache#43255 Improved ACID table column handling - Added support for ACID column prefix in ORC column initialization - Fixed column name handling for ACID tables - Improved type mapping for ACID table columns
morningman
pushed a commit
to suxiaogang223/doris
that referenced
this pull request
Jun 25, 2025
…cate (apache#43255) Problem Summary: This issue addresses a limitation in Apache Doris where only predicates joined by AND are pushed down to the ORC reader, leaving OR-connected predicates unoptimized. By extending pushdown functionality to handle these OR conditions, the aim is to better leverage ORC’s predicate pushdown capabilities, reducing data reads and improving query performance.
morningman
pushed a commit
to suxiaogang223/doris
that referenced
this pull request
Jun 25, 2025
…4615) In the old logic, the `check_expr_can_push_down` function does not check whether the `orc::Literal` are constructed successfully, but only checks during `build_search_argument`. However, if it is found that the `orc::Literal` fails to be constructed after `builder->startNot`, it will fail because the builder cannot end `startNot`. Therefore, we advance the behavior of constructing `orc::Literal` to the `check_expr_can_push_down` function and save the result to the map, so that it will never fail in the `build_search_argument` phase. Related PR: apache#43255
morningman
pushed a commit
to suxiaogang223/doris
that referenced
this pull request
Jun 25, 2025
Related PR: apache#43255 Problem Summary: Example: ```sql CREATE TABLE table_a ( id INT, age INT ) STORED AS ORC; INSERT INTO table_a VALUES (1, null), (2, 18), (3, null), (4, 25); CREATE TABLE table_b ( id INT, age INT ) STORED AS ORC; INSERT INTO table_b VALUES (1, null), (2, null), (3, 1000000), (4, 100); ``` run sql ``` select * from table_a inner join table_b on table_a.age <=> table_b.age and table_b.id in (1,3); ``` When executing this SQL, the backend generates a runtime filter on the table_a side during the join operation, resulting in a condition like WHERE table_a.age IN (NULL, 1000000). It’s important to note that since <=> is a null-aware comparison operator, the IN predicate must also be null-aware. However, the ORC predicate pushdown API does not support null-aware IN predicates. As a result, our current approach ignores null values, leading to an empty result set for this query. To fix this bug, we’ve adjusted the logic so that predicates with null-aware comparisons are not pushed down, ensuring the correct result as follows: ```text +------+------+------+------+ | id | age | id | age | +------+------+------+------+ | 1 | NULL | 1 | NULL | | 3 | NULL | 1 | NULL | +------+------+------+------+ ```
morningman
pushed a commit
to suxiaogang223/doris
that referenced
this pull request
Jun 25, 2025
…ins (apache#45104) Related PR: apache#43255 Problem Summary: Should ignore null values when the literals of in_predicate contains null value, like `in (1, null)` For example, init table in hive: ```sql CREATE TABLE sample_orc_table ( id INT, name STRING, age INT ) STORED AS ORC; INSERT INTO TABLE sample_orc_table VALUES (1, 'Alice', 25), (2, NULL, NULL); ``` select result in Doris should be: ```sql mysql> select * from sample_orc_table where age in (null,25); +------+-------+------+ | id | name | age | +------+-------+------+ | 1 | Alice | 25 | +------+-------+------+ 1 row in set (0.30 sec) mysql> select * from sample_orc_table where age in (25); +------+-------+------+ | id | name | age | +------+-------+------+ | 1 | Alice | 25 | +------+-------+------+ 1 row in set (0.27 sec) mysql> select * from sample_orc_table where age in (null); Empty set (0.01 sec) mysql> select * from sample_orc_table where age is null; +------+------+------+ | id | name | age | +------+------+------+ | 2 | NULL | NULL | +------+------+------+ 1 row in set (0.11 sec) ```
morningman
pushed a commit
to suxiaogang223/doris
that referenced
this pull request
Jun 25, 2025
relate pr: apache#43255 Improved ACID table column handling - Added support for ACID column prefix in ORC column initialization - Fixed column name handling for ACID tables - Improved type mapping for ACID table columns
suxiaogang223
added a commit
to suxiaogang223/doris
that referenced
this pull request
Jun 26, 2025
…cate (apache#43255) Problem Summary: This issue addresses a limitation in Apache Doris where only predicates joined by AND are pushed down to the ORC reader, leaving OR-connected predicates unoptimized. By extending pushdown functionality to handle these OR conditions, the aim is to better leverage ORC’s predicate pushdown capabilities, reducing data reads and improving query performance.
suxiaogang223
added a commit
to suxiaogang223/doris
that referenced
this pull request
Jun 26, 2025
…4615) In the old logic, the `check_expr_can_push_down` function does not check whether the `orc::Literal` are constructed successfully, but only checks during `build_search_argument`. However, if it is found that the `orc::Literal` fails to be constructed after `builder->startNot`, it will fail because the builder cannot end `startNot`. Therefore, we advance the behavior of constructing `orc::Literal` to the `check_expr_can_push_down` function and save the result to the map, so that it will never fail in the `build_search_argument` phase. Related PR: apache#43255
suxiaogang223
added a commit
to suxiaogang223/doris
that referenced
this pull request
Jun 26, 2025
Related PR: apache#43255 Problem Summary: Example: ```sql CREATE TABLE table_a ( id INT, age INT ) STORED AS ORC; INSERT INTO table_a VALUES (1, null), (2, 18), (3, null), (4, 25); CREATE TABLE table_b ( id INT, age INT ) STORED AS ORC; INSERT INTO table_b VALUES (1, null), (2, null), (3, 1000000), (4, 100); ``` run sql ``` select * from table_a inner join table_b on table_a.age <=> table_b.age and table_b.id in (1,3); ``` When executing this SQL, the backend generates a runtime filter on the table_a side during the join operation, resulting in a condition like WHERE table_a.age IN (NULL, 1000000). It’s important to note that since <=> is a null-aware comparison operator, the IN predicate must also be null-aware. However, the ORC predicate pushdown API does not support null-aware IN predicates. As a result, our current approach ignores null values, leading to an empty result set for this query. To fix this bug, we’ve adjusted the logic so that predicates with null-aware comparisons are not pushed down, ensuring the correct result as follows: ```text +------+------+------+------+ | id | age | id | age | +------+------+------+------+ | 1 | NULL | 1 | NULL | | 3 | NULL | 1 | NULL | +------+------+------+------+ ```
suxiaogang223
added a commit
to suxiaogang223/doris
that referenced
this pull request
Jun 26, 2025
…ins (apache#45104) Related PR: apache#43255 Problem Summary: Should ignore null values when the literals of in_predicate contains null value, like `in (1, null)` For example, init table in hive: ```sql CREATE TABLE sample_orc_table ( id INT, name STRING, age INT ) STORED AS ORC; INSERT INTO TABLE sample_orc_table VALUES (1, 'Alice', 25), (2, NULL, NULL); ``` select result in Doris should be: ```sql mysql> select * from sample_orc_table where age in (null,25); +------+-------+------+ | id | name | age | +------+-------+------+ | 1 | Alice | 25 | +------+-------+------+ 1 row in set (0.30 sec) mysql> select * from sample_orc_table where age in (25); +------+-------+------+ | id | name | age | +------+-------+------+ | 1 | Alice | 25 | +------+-------+------+ 1 row in set (0.27 sec) mysql> select * from sample_orc_table where age in (null); Empty set (0.01 sec) mysql> select * from sample_orc_table where age is null; +------+------+------+ | id | name | age | +------+------+------+ | 2 | NULL | NULL | +------+------+------+ 1 row in set (0.11 sec) ```
suxiaogang223
added a commit
to suxiaogang223/doris
that referenced
this pull request
Jun 26, 2025
relate pr: apache#43255 Improved ACID table column handling - Added support for ACID column prefix in ORC column initialization - Fixed column name handling for ACID tables - Improved type mapping for ACID table columns
morrySnow
pushed a commit
that referenced
this pull request
Jun 27, 2025
GoGoWen
pushed a commit
to GoGoWen/incubator-doris
that referenced
this pull request
Sep 26, 2025
) In the old logic, the `check_expr_can_push_down` function does not check whether the `orc::Literal` are constructed successfully, but only checks during `build_search_argument`. However, if it is found that the `orc::Literal` fails to be constructed after `builder->startNot`, it will fail because the builder cannot end `startNot`. Therefore, we advance the behavior of constructing `orc::Literal` to the `check_expr_can_push_down` function and save the result to the map, so that it will never fail in the `build_search_argument` phase. Related PR: apache#43255 Conflicts: be/src/vec/exec/format/orc/vorc_reader.cpp be/test/vec/exec/orc_reader_test.cpp
GoGoWen
pushed a commit
to GoGoWen/incubator-doris
that referenced
this pull request
Sep 26, 2025
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
6 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
What problem does this PR solve?
Issue Number: close #xxx
Related PR: #xxx
Problem Summary:
This issue addresses a limitation in Apache Doris where only predicates joined by AND are pushed down to the ORC reader, leaving OR-connected predicates unoptimized. By extending pushdown functionality to handle these OR conditions, the aim is to better leverage ORC’s predicate pushdown capabilities, reducing data reads and improving query performance.
Check List (For Committer)
Test
Behavior changed:
Does this need documentation?
Release note
None
Check List (For Reviewer who merge this PR)