You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hive.apache.org by "ASF GitHub Bot (Jira)" <ji...@apache.org> on 2022/11/10 16:24:00 UTC
[jira] [Work logged] (HIVE-26722) HiveFilterSetOpTransposeRule incorrectly prunes UNION ALL operands
[ https://issues.apache.org/jira/browse/HIVE-26722?focusedWorklogId=824984&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-824984 ]
ASF GitHub Bot logged work on HIVE-26722:
-----------------------------------------
Author: ASF GitHub Bot
Created on: 10/Nov/22 16:23
Start Date: 10/Nov/22 16:23
Worklog Time Spent: 10m
Work Description: asolimando opened a new pull request, #3748:
URL: https://github.com/apache/hive/pull/3748
… operands
<!--
Thanks for sending a pull request! Here are some tips for you:
1. If this is your first time, please read our contributor guidelines: https://cwiki.apache.org/confluence/display/Hive/HowToContribute
2. Ensure that you have created an issue on the Hive project JIRA: https://issues.apache.org/jira/projects/HIVE/summary
3. Ensure you have added or run the appropriate tests for your PR:
4. If the PR is unfinished, add '[WIP]' in your PR title, e.g., '[WIP]HIVE-XXXXX: Your PR title ...'.
5. Be sure to keep the PR description updated to reflect all changes.
6. Please write your PR title to summarize what this PR proposes.
7. If possible, provide a concise example to reproduce the issue for a faster review.
-->
### What changes were proposed in this pull request?
<!--
Please clarify what changes you are proposing. The purpose of this section is to outline the changes and how this PR fixes the issue.
If possible, please consider writing useful notes for better and faster reviews in your PR. See the examples below.
1. If you refactor some codes with changing classes, showing the class hierarchy will help reviewers.
2. If you fix some SQL features, you can provide some references of other DBMSes.
3. If there is design documentation, please add the link.
4. If there is a discussion in the mailing list, please add the link.
-->
Change the way UNION ALL operands are pruned to avoid false positives.
### Why are the changes needed?
<!--
Please clarify why the changes are needed. For instance,
1. If you propose a new API, clarify the use case for a new API.
2. If you fix a bug, you can clarify why it is a bug.
-->
Correctness
### Does this PR introduce _any_ user-facing change?
<!--
Note that it means *any* user-facing change including all aspects such as the documentation fix.
If yes, please clarify the previous behavior and the change this PR proposes - provide the console output, description, screenshot and/or a reproducable example to show the behavior difference if possible.
If possible, please also clarify if this is a user-facing change compared to the released Hive versions or within the unreleased branches such as master.
If no, write 'No'.
-->
No.
### How was this patch tested?
<!--
If tests were added, say they were added here. Please make sure to add some test cases that check the changes thoroughly including negative and positive cases if possible.
If it was tested in a way different from regular unit tests, please clarify how you tested step by step, ideally copy and paste-able, so that other reviewers can test and check, and descendants can verify in the future.
If tests were not added, please describe why they were not added and/or why it was difficult to add.
-->
Existing tests plus added the following qtest `mvn test -Dtest=TestMiniLlapLocalCliDriver -Dqfile="union_all_filter_transpose.q" -Dtest.output.overwrite -pl itests/qtest -Pitests`
Issue Time Tracking
-------------------
Worklog Id: (was: 824984)
Remaining Estimate: 0h
Time Spent: 10m
> HiveFilterSetOpTransposeRule incorrectly prunes UNION ALL operands
> ------------------------------------------------------------------
>
> Key: HIVE-26722
> URL: https://issues.apache.org/jira/browse/HIVE-26722
> Project: Hive
> Issue Type: Bug
> Components: CBO
> Affects Versions: 4.0.0-alpha-1
> Reporter: Alessandro Solimando
> Assignee: Alessandro Solimando
> Priority: Major
> Time Spent: 10m
> Remaining Estimate: 0h
>
> h1. Reproducer
> Consider the following query:
> {code:java}
> set hive.cbo.rule.exclusion.regex=ReduceExpressionsRule\(Project\);
> CREATE EXTERNAL TABLE t (a string, b string);
> INSERT INTO t VALUES ('1000', 'b1');
> INSERT INTO t VALUES ('2000', 'b2');
> SELECT * FROM (
> SELECT
> a,
> b
> FROM t
> UNION ALL
> SELECT
> a,
> CAST(NULL AS string)
> FROM t) AS t2
> WHERE a = 1000;EXPLAIN CBO
> SELECT * FROM (
> SELECT
> a,
> b
> FROM t
> UNION ALL
> SELECT
> a,
> CAST(NULL AS string)
> FROM t) AS t2
> WHERE a = 1000; {code}
>
> The expected result is:
> {code:java}
> 1000 b1
> 1000 NULL{code}
> An example of correct plan is as follows:
> {noformat}
> CBO PLAN:
> HiveUnion(all=[true])
> HiveProject(a=[$0], b=[$1])
> HiveFilter(condition=[=(CAST($0):DOUBLE, 1000)])
> HiveTableScan(table=[[default, t]], table:alias=[t])
> HiveProject(a=[$0], _o__c1=[null:VARCHAR(2147483647) CHARACTER SET "UTF-16LE"])
> HiveFilter(condition=[=(CAST($0):DOUBLE, 1000)])
> HiveTableScan(table=[[default, t]], table:alias=[t]){noformat}
>
> Consider now a scenario where expression reduction in projections is disabled by setting the following property{_}:{_}
> {noformat}
> set hive.cbo.rule.exclusion.regex=ReduceExpressionsRule\(Project\);
> {noformat}
> In this case, the simplification of _CAST(NULL)_ into _NULL_ does not happen, and we get the following (invalid) result:
> {code:java}
> 1000 b1{code}
> produced by the following invalid plan:
> {code:java}
> CBO PLAN:
> HiveProject(a=[$0], b=[$1])
> HiveFilter(condition=[=(CAST($0):DOUBLE, 1000)])
> HiveTableScan(table=[[default, t]], table:alias=[t]) {code}
> h1. Problem Analysis
> At [HiveFilterSetOpTransposeRule.java#L112|https://github.com/apache/hive/blob/297f510d3b581c9d4079e42caa28aa84f8486012/ql/src/java/org/apache/hadoop/hive/ql/optimizer/calcite/rules/HiveFilterSetOpTransposeRule.java#L112] the _RelMetadataQuery::getPulledUpPredicates_ method infers the following predicate due to the CAST(NULL) in the projection:
> {code:java}
> (=($1, CAST(null:NULL):VARCHAR(2147483647) CHARACTER SET "UTF-16LE")){code}
> When the CAST is simplified to the NULL literal, the IS_NULL($1) predicate is inferred.
> In [HiveFilterSetOpTransposeRule.java#L114-L122|https://github.com/apache/hive/blob/297f510d3b581c9d4079e42caa28aa84f8486012/ql/src/java/org/apache/hadoop/hive/ql/optimizer/calcite/rules/HiveFilterSetOpTransposeRule.java#L114-L122], the rule checks if the conjunction of the predicate coming from the filter (here =(CAST($0):DOUBLE, 1000)) and the inferred predicates is satisfiable or not, under the _UnknownAsFalse_ semantics.
> To summarize, the following expression is simplified under the _UnknownAsFalse_ semantics:
> {code:java}
> AND((=($1, CAST(null:NULL):VARCHAR(2147483647) CHARACTER SET "UTF-16LE")), =(CAST($0):DOUBLE, 1000))
> {code}
> Under In such semantics, (=($1, CAST(null:NULL):...) evaluates to {_}FALSE{_}, because no value is equal to NULL (even NULL itself), AND(FALSE, =(CAST($0):DOUBLE, 1000)) necessarily evaluates to _FALSE_ altogether, and the UNION ALL operand is pruned.
> Only by chance, when _CAST(NULL)_ is simplified to _NULL,_ we avoid the issue, due to the _IS_NULL($1)_ inferred predicate, see [HiveRelMdPredicates.java#L153-L156|https://github.com/apache/hive/blob/297f510d3b581c9d4079e42caa28aa84f8486012/ql/src/java/org/apache/hadoop/hive/ql/optimizer/calcite/stats/HiveRelMdPredicates.java#L153-L156] for understanding how the NULL literal is treated differently during predicate inference.
> The problem lies in the fact that, depending on the input _RelNode_ that we infer predicates from, the semantics is not necessarily {_}UnknownAsFalse{_}, but it might be {_}UnknownAsUnknown{_}, like for {_}Project{_}, as in this case.
> h1. Solution
> In order to correctly simplify a predicate and test if it's always false or not, we should build RexSimplify with _predicates_ as the list of predicates known to hold in the context. In this way, the different semantics are correctly taken into account.
> The code at [HiveFilterSetOpTransposeRule.java#L114-L121|https://github.com/apache/hive/blob/297f510d3b581c9d4079e42caa28aa84f8486012/ql/src/java/org/apache/hadoop/hive/ql/optimizer/calcite/rules/HiveFilterSetOpTransposeRule.java#L114-L121] should be replaced by the following:
> {code:java}
> final RexExecutor executor =
> Util.first(filterRel.getCluster().getPlanner().getExecutor(), RexUtil.EXECUTOR);
> final RexSimplify simplify = new RexSimplify(rexBuilder, predicates, executor);
> final RexNode x = simplify.simplifyUnknownAs(newCondition, RexUnknownAs.FALSE);{code}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)