Commit 19a0fca
[branch-52] SortMergeJoin don't wait for all input before emitting (#20699)
## Which issue does this PR close?
Backport of #20482 to branch-52.
## Rationale for this change
Cherry-pick fix and prerequisites so that SortMergeJoin emits output
incrementally instead of waiting for all input to
complete. This resolves OOM issues Comet is seeing with DataFusion 52.
## What changes are included in this PR?
Cherry-picks of the following commits from `main`:
1. #19614 — Extract sort-merge join filter logic into separate module
2. #20463 — Use zero-copy slice instead of take kernel in sort merge
join
3. #20482 — Fix SortMergeJoin to not wait for all input before emitting
## Are these changes tested?
Yes, covered by existing and new tests included in #20482.
## Are there any user-facing changes?
No.
---------
Co-authored-by: Liang-Chi Hsieh <viirya@gmail.com>
Co-authored-by: Claude Sonnet 4.5 <noreply@anthropic.com>
Co-authored-by: Andy Grove <agrove@apache.org>
Co-authored-by: Raz Luvaton <16746759+rluvaton@users.noreply.github.com>1 parent da4014d commit 19a0fca
5 files changed
Lines changed: 1307 additions & 521 deletions
0 commit comments