Read primitive values directly in branchless IN LIST filters - #24747
Open
geoffreyclaude wants to merge 1 commit into
Open
Read primitive values directly in branchless IN LIST filters#24747geoffreyclaude wants to merge 1 commit into
geoffreyclaude wants to merge 1 commit into
Conversation
The branchless filter only needs the array's primitive values. Read them from the value buffer instead of rebuilding ArrayData for each batch. The value buffer already accounts for array slices, so behavior is unchanged.
Contributor
Author
|
run benchmark in_list_strategy |
|
🤖 Benchmark running (GKE) | trigger CPU Details (lscpu)Comparing codex/in-list-branchless-values (c8c8c9b) to c56a7ab (merge-base) diff Run configurationrun benchmark in_list_strategyResults will be posted here when complete File an issue against this benchmark runner |
|
🤖 Benchmark completed (GKE) | trigger Instance: Comparing codex/in-list-branchless-values (c8c8c9b) to c56a7ab (merge-base) diff Run configurationrun benchmark in_list_strategyCPU Details (lscpu)Details
Resource Usagein_list_strategy — base (merge-base)
in_list_strategy — branch
File an issue against this benchmark runner |
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #24747 +/- ##
==========================================
- Coverage 81.47% 81.47% -0.01%
==========================================
Files 1122 1122
Lines 404140 404137 -3
Branches 404140 404137 -3
==========================================
- Hits 329284 329272 -12
- Misses 55546 55551 +5
- Partials 19310 19314 +4 ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
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
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.
Which issue does this PR close?
Rationale for this change
The branchless IN LIST filter only needs the primitive values in an array. It currently rebuilds
ArrayDatafor every batch before recovering the same value buffer, which adds fixed work to each filter evaluation.What changes are included in this PR?
Read the primitive value buffer directly. The value buffer already accounts for array slices, so behavior is unchanged.
Why is the new offset zero?
By the time DataFusion receives a
PrimitiveArray, its values buffer already starts at the first logical value. Arrow applies offsets while constructing or slicing the array:The last conversion is what
to_data()uses, and the newArrayDataBuilderdefaults to offset zero. See the upstream import, slice, export, and builder default.Thus the old
data.offset()was zero, just like the new explicit0; no offset is discarded.Are these changes tested?
Yes. Existing branchless filter tests cover slices, nulls, and all supported physical widths.
The benchmark bot independently confirms the fixed-cost reduction: plain branchless probes had 35–37% lower latency for batches of 1–64 rows and roughly 3–4% lower latency at 8,192 rows (full results).
Are there any user-facing changes?
No. Results and public APIs are unchanged.