HIVE-29121: Modify HiveSubQueryRemoveRule to use InnerJoin instead of SemiJoin for uncorrelated IN/EXISTS subqueries #6007
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
…related IN/EXiSTS subqueries
zabetak
requested changes
Aug 5, 2025
There was a problem hiding this comment.
I am unsure if the solution should be in HiveSubQueryRemoveRule. Let's continue the discussion under HIVE-29121 and once we converge/clarify what the problem is I will do a full review of the PR.
|
This pull request has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 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Outdated suggestions cannot be applied.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 changes were proposed in this pull request?
This patch modifies
HiveSubQueryRemoveRuleto use InnerJoin instead of SemiJoin for uncorrelated IN/EXISTS subqueries withlogic == RelOptUtil.Logic.TRUE. This restores the logic inHiveSubQRemoveRelBuilder$join, which is replaced by Calcite's RelBuilder.Why are the changes needed?
To restore the previous logic and increase the chances of applying optimizations to the query plan.
We started using SemiJoin in
HiveSubQueryRemoveRulewith HIVE-17767, which introduced SemiJoin for correlated subqueries in conjunction with removing the input Aggregate. Since HIVE-24685, we have been using SemiJoin for both uncorrelated and correlated subqueries, but we currently do not remove the input Aggregate for uncorrelated subqueries.In addition, some optimization rules are not applicable to HiveSemiJoin, which reduces the chances of finding an optimal plan. Therefore, I suggest restoring the previous logic by using InnerJoin for uncorrelated cases.
Does this PR introduce any user-facing change?
No
How was this patch tested?
I added a new
EXPLAIN CBOquery with uncorrelated IN subquery insubquery_in.qto check whether join reordering is applied. One can verify that CalcitePlanner produces a query plan with a different join order when this patch is applied.