bench: pwmj left semi/anti join by SubhamSinghal · Pull Request #24160 · apache/datafusion

SubhamSinghal · 2026-08-07T08:45:18Z

Which issue does this close?

Benchmark companion to #23870 (LeftSemi / LeftAnti support for
PiecewiseMergeJoinExec), part of EPIC #17427. Closes no issue on its own.

Rationale for this change

#23870 routes existence subqueries with an inequality correlation (WHERE EXISTS (SELECT 1 FROM rhs WHERE lhs.key < rhs.key)) to PiecewiseMergeJoinExec instead of NestedLoopJoinExec. That claim needs a benchmark that can be run on either side of the change, so this PR adds the benchmark separately from the operator work — it is bench code only, no functional change.

What changes are included in this PR?

datafusion/core/benches/pwmj_semi_anti_sql.rs (+ its Cargo.toml entry), a Criterion benchmark with two arms over 20k × 20k Int32 rows, target_partitions=1:

pwmj_enabled — flag on
nlj — flag off, which can only plan NestedLoopJoinExec

Axes: join type (EXISTS → Semi, NOT EXISTS → Anti) × match regime (all_match 100%, no_match 0%, half_match ~50%, so the buffered side is only partially marked and scan depth varies per streamed row).

Are these changes tested?

Are there any user-facing changes?

No. Benchmark code only.

codecov-commenter · 2026-08-07T10:48:31Z

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 81.05%. Comparing base (e64e3f7) to head (cdd33d9).
⚠️ Report is 77 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #24160      +/-   ##
==========================================
+ Coverage   81.02%   81.05%   +0.02%     
==========================================
  Files        1106     1106              
  Lines      380718   382025    +1307     
  Branches   380718   382025    +1307     
==========================================
+ Hits       308489   309651    +1162     
- Misses      54001    54091      +90     
- Partials    18228    18283      +55

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

SubhamSinghal · 2026-08-07T10:53:44Z

@kumarUjjawal build successful. PTAL

kumarUjjawal · 2026-08-07T11:50:53Z

+                BenchmarkId::new(format!("pwmj_{jt}_{regime}"), right_rows),
+                |b| {
+                    b.iter(|| {
+                        let (left, right) = build_inputs();


build_inputs() is timed, including generating 40,000 values and allocating arrays and batches. This can dominate the ~0.5 ms PWMJ result. Could we prebuild the batches and use iter_batched for fresh plans outside the timed section?

resolved in d52523e

kumarUjjawal · 2026-08-07T11:52:01Z

+    // Selectivity is set by how far the right key range sits above the left range.
+    // - "high": right keys mostly above left keys  -> most left rows match (Semi large)
+    // - "low":  right keys mostly below left keys   -> few left rows match  (Anti large)
+    let regimes: [(&str, i32); 2] = [("sel_high", key_span), ("sel_low", -key_span)];


These offsets produce exactly all-match and no-match cases, not “mostly” and “few.” Could we label them as all/none and add an overlapping-range case to measure partial suffix marking?

resolved in d52523e

kumarUjjawal · 2026-08-07T11:53:49Z

+        Arc::new(Column::new("key", 0)),
+    );
+    Arc::new(
+        PiecewiseMergeJoinExec::try_new(


Could we benchmark this through the SQL/physical planner instead of constructing PiecewiseMergeJoinExec directly? The current benchmark will panics on because existence joins are unsupported until #23870.

addressed in d52523e

kumarUjjawal

Hi @SubhamSinghal Thanks for iterating.

This became a little tricky that's why we had to do more follow up. If you have any ideas you can share as-well.

kumarUjjawal · 2026-08-09T05:01:16Z

+            ] {
+                let ctx = create_context(right_offset, pwmj, &s);
+                let name = format!("{arm}_{label}_{regime}");
+                assert_plan_contains(


This still panics on this PR’s current head because enabling PWMJ selects NestedLoopJoinExec until #23870, while this assertion requires PiecewiseMergeJoin. To use this revision as the pre-change benchmark baseline, could we name this arm pwmj_enabled and accept either NLJ or PWMJ, while keeping the disabled arm pinned to NLJ? #23870’s tests can assert that the enabled plan switches to PWMJ.

addressed in cdd33d9

SubhamSinghal · 2026-08-11T16:22:54Z

@comphead this is benchmark PR

kumarUjjawal

Looks good 👍

comphead

Thanks @SubhamSinghal for the PR.
Please update the PR description accordingly

SubhamSinghal · 2026-08-13T16:05:18Z

@comphead updated PR description

SubhamSinghal · 2026-08-15T14:11:42Z

@kumarUjjawal @comphead can we merge this PR?

kumarUjjawal · 2026-08-15T14:14:58Z

Thank you @SubhamSinghal and @comphead

## Which issue does this close? Benchmark companion to apache#23870 (LeftSemi / LeftAnti support for `PiecewiseMergeJoinExec`), part of EPIC apache#17427. Closes no issue on its own. ## Rationale for this change apache#23870 routes existence subqueries with an inequality correlation (`WHERE EXISTS (SELECT 1 FROM rhs WHERE lhs.key < rhs.key)`) to `PiecewiseMergeJoinExec` instead of `NestedLoopJoinExec`. That claim needs a benchmark that can be run on either side of the change, so this PR adds the benchmark separately from the operator work — it is bench code only, no functional change. ## What changes are included in this PR? `datafusion/core/benches/pwmj_semi_anti_sql.rs` (+ its `Cargo.toml` entry), a Criterion benchmark with two arms over 20k × 20k `Int32` rows, `target_partitions=1`: - **`pwmj_enabled`** — flag on - **`nlj`** — flag off, which can only plan `NestedLoopJoinExec` Axes: join type (`EXISTS` → Semi, `NOT EXISTS` → Anti) × match regime (`all_match` 100%, `no_match` 0%, `half_match` ~50%, so the buffered side is only partially marked and scan depth varies per streamed row). ## Are these changes tested? ## Are there any user-facing changes? No. Benchmark code only.

## Which issue does this PR close? - Part of apache#17427. Companion to apache#24160, which added this benchmark for the `LeftSemi`/`LeftAnti` half. ## Rationale for this change `RightSemi`/`RightAnti` range joins plan as `NestedLoopJoinExec` today, and the existence-join benchmark added in apache#24160 does not cover them, so there is no way to say what routing them `PiecewiseMergeJoinExec` would be worth — before or after such a change lands. This benchmark runs on either side of it: the `pwmj_enabled` arm accepts whichever operator the build plans and prints it, so a run on `main` today is a valid `NestedLoopJoinExec` baseline that a later build compares straight against. ## What changes are included in this PR? Bench code only, all in `datafusion/core/benches/pwmj_semi_anti_sql.rs`: - `RIGHT SEMI` / `RIGHT ANTI` cases across the same three match regimes as the existing ones, written out as joins rather than reached through `EXISTS` (a decorrelated `EXISTS` gives `LeftSemi`, and which side a nested-loop join marks is its own choice, so a subquery cannot pin these join types). - One `bench_case` runner shared by every case, replacing the previous inline per-case plumbing, driven by a `Kind` enum with one variant per `JoinType`. Adding the mark joins later is two more variants rather than a new code path. ## Are these changes tested? The guards above are the tests ## Are there any user-facing changes? No. Benchmark code only.

bench: pwmj left semi/anti join

27b1559

SubhamSinghal mentioned this pull request Aug 7, 2026

feat(pwmj): support LeftSemi/LeftAnti existence joins via classic scan #23870

Merged

github-actions Bot added sqllogictest SQL Logic Tests (.slt) physical-plan Changes to the physical-plan crate labels Aug 7, 2026

remove slt test

6ae681d

github-actions Bot removed the sqllogictest SQL Logic Tests (.slt) label Aug 7, 2026

kumarUjjawal reviewed Aug 7, 2026

View reviewed changes

Modify bench

d52523e

github-actions Bot added core Core DataFusion crate and removed physical-plan Changes to the physical-plan crate labels Aug 7, 2026

kumarUjjawal reviewed Aug 9, 2026

View reviewed changes

bench: let the pwmj_enabled arm accept either operator

cdd33d9

SubhamSinghal requested a review from kumarUjjawal August 10, 2026 05:42

kumarUjjawal approved these changes Aug 13, 2026

View reviewed changes

comphead approved these changes Aug 13, 2026

View reviewed changes

kumarUjjawal added this pull request to the merge queue Aug 15, 2026

Merged via the queue into apache:main with commit 4403c5d Aug 15, 2026
39 checks passed

SubhamSinghal deleted the bench-pwmj branch August 15, 2026 14:26

SubhamSinghal mentioned this pull request Aug 18, 2026

bench: pwmj right semi/anti existence joins #24458

Merged

Uh oh!

Conversation

SubhamSinghal commented Aug 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Which issue does this close?

Rationale for this change

What changes are included in this PR?

Are these changes tested?

Are there any user-facing changes?

Uh oh!

codecov-commenter commented Aug 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

SubhamSinghal commented Aug 7, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

kumarUjjawal left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

SubhamSinghal commented Aug 11, 2026

Uh oh!

kumarUjjawal left a comment

Choose a reason for hiding this comment

Uh oh!

comphead left a comment

Choose a reason for hiding this comment

Uh oh!

SubhamSinghal commented Aug 13, 2026

Uh oh!

SubhamSinghal commented Aug 15, 2026

Uh oh!

kumarUjjawal commented Aug 15, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

SubhamSinghal commented Aug 7, 2026 •

edited

Loading

codecov-commenter commented Aug 7, 2026 •

edited

Loading