Fix: limit is missing after removing SPM #14569

xudong963 · 2025-02-10T04:35:45Z

Rationale for this change

It's clear that fetch will miss after removing SPM

What changes are included in this PR?

If SPM is with fetch, we won't remove it.

Are these changes tested?

Yes, ut

Are there any user-facing changes?

No

datafusion/core/tests/physical_optimizer/enforce_sorting.rs

zhuqi-lucas · 2025-02-10T07:05:55Z

datafusion/core/tests/physical_optimizer/enforce_sorting.rs

+    ];
+    let expected_optimized = [
+        "SortPreservingMergeExec: [non_nullable_col@1 ASC], fetch=100",
+        "  SortExec: expr=[non_nullable_col@1 ASC], preserve_partitioning=[false]",


I am wandering do we expect to optimize only sortExec with Topk(fetch) for the only one partition case?

datafusion/core/tests/physical_optimizer/enforce_sorting.rs

datafusion/physical-optimizer/src/enforce_sorting/mod.rs

xudong963 · 2025-02-10T12:22:14Z

datafusion/core/tests/physical_optimizer/enforce_sorting.rs

+        "LocalLimitExec: fetch=100",
+        "  SortExec: expr=[non_nullable_col@1 ASC], preserve_partitioning=[false]",


we should optimize such pattern to topk later

alamb

This makes sense to me -- thank you @xudong963

In general it seems like we have a class of bugs related to removing fetch -- maybe we should re-evaluate enforce_sorting for a more holistic fix if we find any more issues

xudong963 · 2025-02-10T14:22:37Z

This makes sense to me -- thank you @xudong963

In general it seems like we have a class of bugs related to removing fetch -- maybe we should re-evaluate enforce_sorting for a more holistic fix if we find any more issues

yes, I also have the thought, enforce_sorting has so many condition branches, I'll do a refactor and add more tests for EnforceSorting when I have time.

zhuqi-lucas

LGTM now, thanks @xudong963 !

xudong963 · 2025-02-10T14:38:03Z

Thanks all, let's go!

* Fix: limit is missing after removing SPM * fix test * use limitexec

* Get working build * Add pool_size method to MemoryPool (#218) (#230) * Add pool_size method to MemoryPool * Fix * Fmt Co-authored-by: Daniël Heres <danielheres@gmail.com> * Respect `IGNORE NULLS` flag in `ARRAY_AGG` (#260) (apache#15544) v48 * Hook for doing distributed `CollectLeft` joins (#269) * Ignore writer shutdown error (#271) * ignore writer shutdown error * cargo check * Fix bug in `swap_hash_join` (#278) * Try and fix swap_hash_join * Only swap projections when join does not have projections * just backport upstream fix * remove println * Support Duration in min/max agg functions (#283) (apache#15310) v47 * Support Duration in min/max agg functions * Attempt to fix build * Attempt to fix build - Fix chrono version * Revert "Attempt to fix build - Fix chrono version" This reverts commit fd76fe6. * Revert "Attempt to fix build" This reverts commit 9114b86. --------- Co-authored-by: svranesevic <svranesevic@users.noreply.github.com> * Fix panics in array_union (#287) (apache#15149) v48 * Drop rust-toolchain * Fix panics in array_union * Fix the chrono * Backport `GroupsAccumulator` for Duration min/max agg (#288) (apache#15322) v47 * Fix array_sort for empty record batch (#290) (apache#15149) v48 * fix: rewrite fetch, skip of the Limit node in correct order (apache#14496) v46 * fix: rewrite fetch, skip of the Limit node in correct order * style: fix clippy * Support aliases in ConstEvaluator (apache#14734) (#281) v46 * Support aliases in ConstEvaluator (apache#14734) Not sure why they are not supported. It seems that if we're not careful, some transformations can introduce aliases nested inside other expressions. * Format Cargo.toml * Preserve the name of grouping sets in SimplifyExpressions (#282) (apache#14888) v46 Whenever we use `recompute_schema` or `with_exprs_and_inputs`, this ensures that we obtain the same schema. * Support Duration in min/max agg functions (#284) (apache#15310) v47 Co-authored-by: svranesevic <svranesevic@users.noreply.github.com> * fix case_column_or_null with nullable when conditions (apache#13886) v45 * fix case_column_or_null with nullable when conditions * improve sqllogictests for case_column_or_null --------- Co-authored-by: zhangli20 <zhangli20@kuaishou.com> * Fix casewhen (apache#14156) v45 * Cherry-pick topk limit pushdown fix (apache#14192) v45 * fix: FULL OUTER JOIN and LIMIT produces wrong results (apache#14338) v45 * fix: FULL OUTER JOIN and LIMIT produces wrong results * Fix minor slt testing * fix test (cherry picked from commit ecc5694) * Cherry-pick global limit fix (apache#14245) v45 * fix: Limits are not applied correctly (apache#14418) v46 * fix: Limits are not applied correctly * Add easy fix * Add fix * Add slt testing * Address comments * Disable grouping set in CSE * Fix spm + limit (apache#14569) v46 * prost 0.13 / fix parquet dep * Delete unreliable checks * Segfault in ByteGroupValueBuilder (#294) (apache#15968) v50 * test to demonstrate segfault in ByteGroupValueBuilder * check for offset overflow * clippy (cherry picked from commit 5bdaeaf) * Update arrow dependency to include rowid (#295) * Update arrow version * Feat: Add fetch to CoalescePartitionsExec (apache#14499) (#298) v46 * add fetch info to CoalescePartitionsExec * use Statistics with_fetch API on CoalescePartitionsExec * check limit_reached only if fetch is assigned Co-authored-by: mertak-synnada <mertak67+synaada@gmail.com> * Fix `CoalescePartitionsExec` proto serialization (apache#15824) (#299) v48 * add fetch to CoalescePartitionsExecNode * gen proto code * Add test * fix * fix build * Fix test build * remove comments Co-authored-by: 张林伟 <lewiszlw520@gmail.com> * Add JoinContext with JoinLeftData to TaskContext in HashJoinExec (#300) * Add JoinContext with JoinLeftData to TaskContext in HashJoinExec * Expose random state as const * re-export ahash::RandomState * JoinContext default impl * Add debug log when setting join left data * Update arrow version for not preserving dict_id (#303) * Use partial aggregation schema for spilling to avoid column mismatch in GroupedHashAggregateStream (apache#13995) (#302) v45 * Refactor spill handling in GroupedHashAggregateStream to use partial aggregate schema * Implement aggregate functions with spill handling in tests * Add tests for aggregate functions with and without spill handling * Move test related imports into mod test * Rename spill pool test functions for clarity and consistency * Refactor aggregate function imports to use fully qualified paths * Remove outdated comments regarding input batch schema for spilling in GroupedHashAggregateStream * Update aggregate test to use AVG instead of MAX * assert spill count * Refactor partial aggregate schema creation to use create_schema function * Refactor partial aggregation schema creation and remove redundant function * Remove unused import of Schema from arrow::datatypes in row_hash.rs * move spill pool testing for aggregate functions to physical-plan/src/aggregates * Use Arc::clone for schema references in aggregate functions (cherry picked from commit 81b50c4) Co-authored-by: kosiew <kosiew@gmail.com> * Update tag * Push limits past windows (#337) (apache#17347) v50 * Restore old method for DQE * feat(optimizer): Enable filter pushdown on window functions (apache#14026) v45 * Avoid Aliased Window Expr Enter Unreachable Code (apache#14109) v45 (cherry picked from commit fda500a) * Use `Expr::qualified_name()` and `Column::new()` to extract partition keys from window and aggregate operators (#355) (apache#17757) v51 * Update PR template to be relevant to our fork * Make limit pushdown work for SortPreservingMergeExec (apache#17893) (#361) * re-publicise functions DQE relies on * Handle columns in with_new_exprs with a Join (apache#15055) (#384) apache#15055 * handle columns in with_new_exprs with Join * test doesn't return result * take join from result * clippy * make test fallible * accept any pair of expression for new_on in with_new_exprs for Join * use with_capacity Co-authored-by: delamarch3 <68732277+delamarch3@users.noreply.github.com> --------- Co-authored-by: Georgi Krastev <georgi.krastev@coralogix.com> Co-authored-by: Daniël Heres <danielheres@gmail.com> Co-authored-by: Dan Harris <1327726+thinkharderdev@users.noreply.github.com> Co-authored-by: Faiaz Sanaulla <105630300+fsdvh@users.noreply.github.com> Co-authored-by: Sava Vranešević <20240220+svranesevic@users.noreply.github.com> Co-authored-by: svranesevic <svranesevic@users.noreply.github.com> Co-authored-by: Yingwen <realevenyag@gmail.com> Co-authored-by: Zhang Li <richselian@gmail.com> Co-authored-by: zhangli20 <zhangli20@kuaishou.com> Co-authored-by: Aleksey Kirilishin <54231417+avkirilishin@users.noreply.github.com> Co-authored-by: xudong.w <wxd963996380@gmail.com> Co-authored-by: Qi Zhu <821684824@qq.com> Co-authored-by: Martins Purins <martins.purins@coralogix.com> Co-authored-by: mertak-synnada <mertak67+synaada@gmail.com> Co-authored-by: 张林伟 <lewiszlw520@gmail.com> Co-authored-by: kosiew <kosiew@gmail.com> Co-authored-by: nuno-faria <nunofpfaria@gmail.com> Co-authored-by: Berkay Şahin <124376117+berkaysynnada@users.noreply.github.com> Co-authored-by: Mason Hall <mason.hall@coralogix.com> Co-authored-by: delamarch3 <68732277+delamarch3@users.noreply.github.com>

Fix: limit is missing after removing SPM

acce616

github-actions bot added the core Core DataFusion crate label Feb 10, 2025

fix test

2eb672f

xudong963 requested a review from akurmustafa February 10, 2025 04:57

github-actions bot added the optimizer Optimizer rules label Feb 10, 2025

zhuqi-lucas reviewed Feb 10, 2025

View reviewed changes

datafusion/core/tests/physical_optimizer/enforce_sorting.rs Show resolved Hide resolved

xudong963 commented Feb 10, 2025

View reviewed changes

datafusion/physical-optimizer/src/enforce_sorting/mod.rs Outdated Show resolved Hide resolved

use limitexec

12cd613

xudong963 force-pushed the fix_limit_smp branch from dc76129 to 12cd613 Compare February 10, 2025 12:06

xudong963 commented Feb 10, 2025

View reviewed changes

xudong963 mentioned this pull request Feb 10, 2025

Release DataFusion 46.0.0 #14123

Closed

26 tasks

alamb approved these changes Feb 10, 2025

View reviewed changes

xudong963 requested a review from zhuqi-lucas February 10, 2025 14:23

zhuqi-lucas approved these changes Feb 10, 2025

View reviewed changes

xudong963 merged commit dee0dc7 into apache:main Feb 10, 2025
24 checks passed

xudong963 deleted the fix_limit_smp branch February 10, 2025 14:38

alamb mentioned this pull request Feb 10, 2025

fix: fetch is missed during EnforceDistribution #14207

Closed

xudong963 added a commit to massive-com/arrow-datafusion that referenced this pull request Feb 12, 2025

Fix: limit is missing after removing SPM (apache#14569)

20b89c2

* Fix: limit is missing after removing SPM * fix test * use limitexec

alamb mentioned this pull request Mar 4, 2025

March 4, 2025: This week(s) in DataFusion #15005

Closed

avantgardnerio pushed a commit to coralogix/arrow-datafusion that referenced this pull request Oct 1, 2025

Fix spm + limit (apache#14569) v46

23f00cf

avantgardnerio pushed a commit to coralogix/arrow-datafusion that referenced this pull request Oct 15, 2025

Fix spm + limit (apache#14569) v46

b8d9852

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Fix: limit is missing after removing SPM #14569

Fix: limit is missing after removing SPM #14569

Uh oh!

xudong963 commented Feb 10, 2025

Uh oh!

Uh oh!

zhuqi-lucas Feb 10, 2025

Uh oh!

Uh oh!

Uh oh!

xudong963 Feb 10, 2025 •

edited

Loading

Uh oh!

alamb left a comment

Uh oh!

xudong963 commented Feb 10, 2025 •

edited

Loading

Uh oh!

zhuqi-lucas left a comment

Uh oh!

xudong963 commented Feb 10, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

		"LocalLimitExec: fetch=100",
		" SortExec: expr=[non_nullable_col@1 ASC], preserve_partitioning=[false]",

Fix: limit is missing after removing SPM #14569

Fix: limit is missing after removing SPM #14569

Uh oh!

Conversation

xudong963 commented Feb 10, 2025

Rationale for this change

What changes are included in this PR?

Are these changes tested?

Are there any user-facing changes?

Uh oh!

Uh oh!

zhuqi-lucas Feb 10, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

xudong963 Feb 10, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

alamb left a comment

Choose a reason for hiding this comment

Uh oh!

xudong963 commented Feb 10, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

zhuqi-lucas left a comment

Choose a reason for hiding this comment

Uh oh!

xudong963 commented Feb 10, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

xudong963 Feb 10, 2025 •

edited

Loading

xudong963 commented Feb 10, 2025 •

edited

Loading