[fix](fe) Backport runtime filter outer join fix to 4.1#64162
Open
BiteTheDDDDt wants to merge 1 commit into
Open
[fix](fe) Backport runtime filter outer join fix to 4.1#64162BiteTheDDDDt wants to merge 1 commit into
BiteTheDDDDt wants to merge 1 commit into
Conversation
…pache#64102) related with: apache#57425 Runtime filters from a parent inner join could be pushed through an outer join into the null-generating child even when the probe expression was not null-propagating for that child. The problem can be reproduced with this SQL shape: ```sql create table rf_outer_join_nullable_a (pk int) duplicate key(pk) distributed by hash(pk) buckets 1 properties("replication_num" = "1"); create table rf_outer_join_nullable_b (pk int) duplicate key(pk) distributed by hash(pk) buckets 1 properties("replication_num" = "1"); create table rf_outer_join_nullable_c (pk int) duplicate key(pk) distributed by hash(pk) buckets 1 properties("replication_num" = "1"); insert into rf_outer_join_nullable_a values (1); insert into rf_outer_join_nullable_b values (1); insert into rf_outer_join_nullable_c values (0); set disable_join_reorder = true; select coalesce(b.pk, 0) as k, count(*) as cnt from rf_outer_join_nullable_a a left join rf_outer_join_nullable_b b on a.pk = b.pk inner join rf_outer_join_nullable_c c on coalesce(b.pk, 0) = c.pk group by 1 order by 1; ``` The correct result is empty. `a.pk = 1` matches `b.pk = 1` in the left outer join, then the parent inner join evaluates `coalesce(1, 0) = 0`, which is false. The wrong plan generated a runtime filter from the parent inner join, effectively `c.pk -> coalesce(b.pk, 0)`, and pushed it through the lower `LEFT OUTER JOIN` into the right side scan of `b`. If `b.pk = 1` is pre-filtered before the left outer join, the join emits a NULL-extended row for `b`; then `coalesce(NULL, 0) = 0` becomes true and incorrectly returns `(0, 1)`. Therefore the runtime filter `c.pk -> coalesce(b.pk, 0)` must not be planned on the null-generating side of the lower outer join. This PR blocks runtime filter pushdown through an outer join's null-generating child unless the probe expression preserves NULL semantics for slots from that child. Normal pushdown through preserved sides and null-propagating expressions is kept unchanged. The bug became observable after apache#57425 changed the target lookup for expression runtime filters from `ctx.probeExpr` to `ctx.probeSlot`. Before that change, an expression such as `coalesce(b.pk, 0)` could not resolve the target relation in this path and the unsafe pushdown was not generated. None - Test <!-- At least one of them must be included. --> - [x] Regression test - [x] Unit Test - [ ] Manual test (add detailed scripts or steps below) - [ ] No need to test or manual test. Explain why: - [ ] This is a refactor/code format and no logic has been changed. - [ ] Previous test can cover this change. - [ ] No code files have been changed. - [ ] Other reason <!-- Add your reason? --> Added regression case with `disable_join_reorder`, `qt_shape`, and empty result verification: `regression-test/suites/correctness_p0/test_runtime_filter_outer_join_nullable_side.groovy` Unit test: `./run-fe-ut.sh --run org.apache.doris.nereids.postprocess.RuntimeFilterTest#testDoNotPushDownNonNullPropagatingRuntimeFilterThroughOuterJoin,org.apache.doris.nereids.postprocess.RuntimeFilterTest#testPushDownNullPropagatingRuntimeFilterThroughOuterJoin` The SQL regression case was not run locally against the available 9333 cluster because that cluster was the unpatched repro cluster. - Behavior changed: - [ ] No. - [x] Yes. Runtime filters are no longer pushed through an outer join into its null-generating child when the probe expression can convert NULL to a non-NULL value. - Does this need documentation? - [x] No. - [ ] Yes. <!-- Add document PR link here. eg: apache/doris-website#1214 --> - [ ] Confirm the release note - [ ] Confirm test cases - [ ] Confirm document - [ ] Add branch pick label <!-- Add branch pick label that this PR should merge into --> (cherry picked from commit 691189a)
Contributor
|
Thank you for your contribution to Apache Doris. Please clearly describe your PR:
|
Contributor
Author
|
run buildall |
Contributor
There was a problem hiding this comment.
Pull request overview
Backports the fix from #64102 to branch-4.1 to prevent unsafe runtime filter pushdown through the null-generating side of outer joins, specifically when the probe expression is not NULL-propagating (e.g., coalesce(b.pk, 0)), which can change query semantics.
Changes:
- Add outer-join-aware pushdown gating in
RuntimeFilterPushDownVisitorto block pushing runtime filters into null-generating children unless the probe expression is NULL-propagating. - Add FE unit tests covering both the blocked (
coalesce) and allowed (slot-equality) cases. - Add a regression test reproducer that validates both plan shape and the empty-result correctness.
Reviewed changes
Copilot reviewed 4 out of 4 changed files in this pull request and generated no comments.
| File | Description |
|---|---|
fe/fe-core/src/main/java/org/apache/doris/nereids/processor/post/RuntimeFilterPushDownVisitor.java |
Adds null-generating-side checks for join-child pushdown and NULL-propagation detection via PropagateNullable/Cast/Slot. |
fe/fe-core/src/test/java/org/apache/doris/nereids/postprocess/RuntimeFilterTest.java |
Adds unit tests validating runtime filter presence/absence across the outer-join boundary depending on NULL-propagation. |
regression-test/suites/correctness_p0/test_runtime_filter_outer_join_nullable_side.groovy |
Introduces a regression suite that reproduces the original wrong-results scenario and asserts stable plan shape + correct result. |
regression-test/data/correctness_p0/test_runtime_filter_outer_join_nullable_side.out |
Golden output for the new regression test (shape + empty result). |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Contributor
FE UT Coverage ReportIncrement line coverage |
Contributor
FE Regression Coverage ReportIncrement line coverage |
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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: N/A
Related PR: #64102
Problem Summary:
Backport #64102 to branch-4.1. The fix prevents unsafe runtime filter pushdown through the null-generating side of outer joins. The implementation was adapted to the branch-4.1 RuntimeFilterPushDownVisitor structure.
Release note
None
Check List (For Author)