planner: fix plan explore can not detect the specific index path. by AilinKid · Pull Request #65709 · pingcap/tidb

AilinKid · 2026-01-21T16:45:21Z

What problem does this PR solve?

Issue Number: close #65708

Problem Summary:
Explain Explore can miss alternative index paths for single-table queries when statistics favor one index (e.g., always picking a and never exploring b), which makes it
hard to surface candidate plans for plan binding or verification.

What changed and how does it work?

Plan generation now enumerates candidate single-table indexes from WHERE predicate columns and explores them via USE_INDEX hints.
The exploration is limited to single-table SELECT with predicates and skips subqueries to avoid unsafe hinting.
Added a unit test to verify that EXPLAIN EXPLORE returns both index:a and index:b plans.

Check List

Tests

Unit test
Integration test
Manual test (add detailed scripts or steps below)
No need to test
- I checked and no code files have been changed.

Side effects

Performance regression: Consumes more CPU
Performance regression: Consumes more Memory
Breaking backward compatibility

Documentation

Release note

Please refer to Release Notes Language Style Guide to write a quality release note.

None

tiprow · 2026-01-21T16:45:38Z

Hi @AilinKid. Thanks for your PR.

PRs from untrusted users cannot be marked as trusted with /ok-to-test in this repo meaning untrusted PR authors can never trigger tests themselves. Collaborators can still trigger tests on the PR using /test all.

I understand the commands that are listed here.

Details

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes-sigs/prow repository.

Copilot

Pull request overview

This PR fixes an issue where EXPLAIN EXPLORE could miss alternative index paths for single-table queries when statistics favor one index over another. The fix allows the plan exploration to enumerate and test all candidate indexes that could be used based on the WHERE predicate columns.

Changes:

Added index hint exploration to the breadth-first plan search algorithm
Implemented column extraction from WHERE predicates to identify candidate indexes
Extended the state representation to include index hints alongside existing leading hints and optimizer variables

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File	Description
pkg/bindinfo/binding_plan_generation.go	Adds `indexHint` struct, updates `state` to track index hints, implements `extractSelectIndexHints` to identify candidate indexes from WHERE predicates, and integrates index hint exploration into the BFS plan search
pkg/bindinfo/binding_auto_test.go	Adds test case `TestExplainExploreIndexHints` to verify that EXPLAIN EXPLORE returns plans for both index `a` and index `b`
pkg/bindinfo/BUILD.bazel	Adds dependency on `pkg/meta/model` and increments test shard count from 48 to 49

codecov · 2026-01-21T17:15:51Z

Codecov Report

❌ Patch coverage is 81.90476% with 38 lines in your changes missing coverage. Please review.
✅ Project coverage is 77.6318%. Comparing base (c2202e1) to head (ca41e03).
⚠️ Report is 6 commits behind head on master.

Additional details and impacted files

@@               Coverage Diff                @@
##             master     #65709        +/-   ##
================================================
- Coverage   77.7795%   77.6318%   -0.1478%     
================================================
  Files          2001       1923        -78     
  Lines        545624     533459     -12165     
================================================
- Hits         424384     414134     -10250     
+ Misses       119578     119319       -259     
+ Partials       1662          6      -1656

Flag	Coverage Δ
integration	`41.4752% <0.0000%> (-6.6845%)`	⬇️
unit	`76.7577% <81.9047%> (+0.3459%)`	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Components	Coverage Δ
dumpling	`56.7974% <ø> (ø)`
parser	`∅ <ø> (∅)`
br	`48.8783% <ø> (-12.1125%)`	⬇️

🚀 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.

pkg/bindinfo/binding_plan_generation.go

qw4990 · 2026-01-23T07:42:49Z

pkg/bindinfo/binding_plan_generation.go

 			}
 		}
+		for _, indexHint := range possibleIndexHints {
+			newState := newStateWithIndexHint(currState, indexHint)


It seems like we only consider one index in a state, if there are multiple tables, should we consider multiple index hints?

table levels, and for each table, there are multi index choices? make sense

i will do it in next pr

guo-shaoge · 2026-01-26T05:20:37Z

pkg/bindinfo/binding_plan_generation.go

+
+func extractSelectIndexHints(sctx sessionctx.Context, node ast.StmtNode, tableNames []*tableName) []*indexHint {
+	selStmt, isSel := node.(*ast.SelectStmt)
+	if !isSel || selStmt.Where == nil || len(tableNames) != 1 {


Why we have such restriction that only support one table?
Does it means explain explore will not support explore index for multiple table query?

we will expand to all tableNames in next pr

guo-shaoge · 2026-01-26T05:30:56Z

pkg/bindinfo/binding_plan_generation.go

Guess we should have a sys var to control this behavior.

guo-shaoge · 2026-01-26T05:38:51Z

pkg/bindinfo/binding_auto_test.go

+	tk.MustExec("use test")
+	tk.MustExec(`create table t (a int, b int, c int, key(a), key(b))`)
+
+	rows := tk.MustQuery(`explain explore select * from t where a=1 and b=1`).Rows()


suggestion: add case that has alias table name to cover this situation

looks like we failed to explore index b when adding table alias:

TiDB root@127.0.0.1:test> explain explore select 1 from t t_alias where a=1 and b=1 and c like "%xx%"\G ***************************[ 1. row ]*************************** statement | select 1 from t t_alias where a=1 and b=1 and c like "%xx%" binding_hint | use_index(@`sel_1` `test`.`t_alias` ) plan | Projection 1.00 root 1->Column#5 └─TableReader 1.00 root data:Selection └─Selection 1.00 cop[tikv] eq(test.t.a, 1), eq(test.t.b, 1), like(cast(test.t.c, var_string(20)), "%xx%", 92) └─TableFullScan 2000.00 cop[tikv] table:t_alias keep order:false, stats:partial[a:allEvicted, b:allEvicted, a:allEvicted...(more: 2 allEvicted)] plan_digest | 2d880578187453ef4bec09bd049c955f5e558c549d4e6c0f937b557ec82e8ea8 avg_latency | 0 exec_times | 0 avg_scan_rows | 0 avg_returned_rows | 0 latency_per_returned_row | 0 scan_rows_per_returned_row | 0 recommend | reason | explain_analyze | EXPLAIN ANALYZE '2d880578187453ef4bec09bd049c955f5e558c549d4e6c0f937b557ec82e8ea8' binding | CREATE GLOBAL BINDING FROM HISTORY USING PLAN DIGEST '2d880578187453ef4bec09bd049c955f5e558c549d4e6c0f937b557ec82e8ea8' ***************************[ 2. row ]*************************** statement | select 1 from t t_alias where a=1 and b=1 and c like "%xx%" binding_hint | use_index(@`sel_1` `test`.`t_alias` `a`), no_order_index(@`sel_1` `test`.`t_alias` `a`) plan | Projection 1.00 root 1->Column#5 └─IndexLookUp 1.00 root ├─IndexRangeScan(Build) 2.00 cop[tikv] table:t_alias, index:a(a) range:[1,1], keep order:false, stats:partial[a:allEvicted, b:allEvicted, a:allEvicted...(more: 2 allEvicted)] └─Selection(Probe) 1.00 cop[tikv] eq(test.t.b, 1), like(cast(test.t.c, var_string(20)), "%xx%", 92) └─TableRowIDScan 2.00 cop[tikv] table:t_alias keep order:false, stats:partial[a:allEvicted, b:allEvicted, a:allEvicted...(more: 2 allEvicted)] plan_digest | 53a2f3b7e7fd761ec1e6729d7680306785fae4ccb945f1a6790185ad3c586dd0 avg_latency | 0 exec_times | 0 avg_scan_rows | 0 avg_returned_rows | 0 latency_per_returned_row | 0 scan_rows_per_returned_row | 0 recommend | reason | explain_analyze | EXPLAIN ANALYZE '53a2f3b7e7fd761ec1e6729d7680306785fae4ccb945f1a6790185ad3c586dd0' binding | CREATE GLOBAL BINDING FROM HISTORY USING PLAN DIGEST '53a2f3b7e7fd761ec1e6729d7680306785fae4ccb945f1a6790185ad3c586dd0' 2 rows in set Time: 0.482s

make sense, addressed

guo-shaoge · 2026-01-26T05:39:38Z

pkg/bindinfo/binding_plan_generation.go

+
+func (e *predicateColumnExtractor) Enter(in ast.Node) (node ast.Node, skipChildren bool) {
+	switch n := in.(type) {
+	case *ast.SubqueryExpr:


question: why skip SubqueryExpr?

different select block will mix the column name
select a from t, (select a from t2) x. this case we will fetch column name a, but hard to tell this a is from t or t2, so it will cause misunderstanding here, i will do it in next pr.

Copilot

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.

Copilot · 2026-01-27T03:59:25Z

pkg/bindinfo/binding_plan_generation.go

+func extractSelectIndexHints(sctx sessionctx.Context, defaultSchema string, node ast.StmtNode) []*indexHint {
+	selStmt, isSel := node.(*ast.SelectStmt)
+	if !isSel {
+		return nil
+	}
+	tableNames := extractSelectTableNamesWithAlias(defaultSchema, selStmt)
+	if len(tableNames) == 0 {
+		return nil
+	}
+	allowAnyTable := len(tableNames) == 1
+	hints := make([]*indexHint, 0)
+	seen := make(map[string]struct{})
+	for _, target := range tableNames {
+		extractor := &predicateColumnExtractor{
+			table:         target,
+			columns:       make(map[string]struct{}),
+			allowAnyTable: allowAnyTable,
+		}
+		if selStmt.Where != nil {
+			selStmt.Where.Accept(extractor)
+		}
+		if selStmt.From != nil && selStmt.From.TableRefs != nil {
+			collectJoinPredicates(selStmt.From.TableRefs, extractor)
+		}
+		if len(extractor.columns) == 0 {
+			continue
+		}
+		tblInfo, err := sctx.GetLatestInfoSchema().TableInfoByName(ast.NewCIStr(target.schema), ast.NewCIStr(target.name))
+		if err != nil {
+			continue
+		}
+		useInvisible := sctx.GetSessionVars().OptimizerUseInvisibleIndexes
+		for _, index := range tblInfo.Indices {
+			if index.State != model.StatePublic {
+				continue
+			}
+			if !useInvisible && index.Invisible {
+				continue
+			}
+			if index.IsColumnarIndex() || index.InvertedInfo != nil {
+				continue
+			}
+			if tblInfo.IsCommonHandle && index.Primary {
+				continue
+			}
+			if len(index.Columns) == 0 {
+				continue
+			}
+			if _, ok := extractor.columns[index.Columns[0].Name.L]; !ok {
+				continue
+			}
+			hint := &indexHint{
+				table: target,
+				index: index.Name.O,
+			}
+			if _, ok := seen[hint.String()]; ok {
+				continue
+			}
+			seen[hint.String()] = struct{}{}
+			hints = append(hints, hint)
+		}
+	}
+	return hints


extractSelectIndexHints currently generates index hints for any SELECT with table references in the FROM clause, including multi-table joins (it iterates over all tableNames regardless of length), while the PR description states that exploration should be limited to single-table SELECTs with predicates. Please either constrain this helper (for example by returning early when len(tableNames) != 1) or update the PR description to accurately describe that index hints are also applied to joined queries.

AilinKid · 2026-01-27T07:48:11Z

/retest-required

tiprow · 2026-01-27T07:48:43Z

@AilinKid: PRs from untrusted users cannot be marked as trusted with /ok-to-test in this repo meaning untrusted PR authors can never trigger tests themselves. Collaborators can still trigger tests on the PR using /test.

Details

In response to this:

/retest-required

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes-sigs/prow repository.

AilinKid · 2026-01-27T07:51:06Z

/ok-to-test

ti-chi-bot · 2026-01-27T10:02:44Z

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: guo-shaoge, qw4990

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Details

Needs approval from an approver in each of these files:

~~pkg/bindinfo/OWNERS~~ [qw4990]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

qw4990 · 2026-01-27T10:02:47Z

/retest

ti-chi-bot · 2026-01-27T10:02:49Z

[LGTM Timeline notifier]

Timeline:

2026-01-27 04:55:10.493794623 +0000 UTC m=+1074538.107751478: ☑️ agreed by guo-shaoge.
2026-01-27 10:02:47.656586785 +0000 UTC m=+1092995.270543641: ☑️ agreed by qw4990.

AilinKid · 2026-01-27T14:30:09Z

/retest-required

AilinKid · 2026-01-28T02:10:51Z

/retest-required

qw4990 · 2026-01-28T11:53:56Z

/retest

AilinKid · 2026-01-29T09:41:13Z

/retest-required

YangKeao · 2026-01-29T17:00:57Z

/retest

YangKeao · 2026-01-29T17:05:53Z

/retest

YangKeao · 2026-01-29T17:10:53Z

/retest

YangKeao · 2026-01-29T17:15:54Z

/retest

YangKeao · 2026-01-29T17:20:55Z

/retest

YangKeao · 2026-01-29T17:25:53Z

/retest

YangKeao · 2026-01-29T17:30:55Z

/retest

YangKeao · 2026-01-29T17:35:53Z

/retest

YangKeao · 2026-01-29T17:40:53Z

/retest

YangKeao · 2026-01-29T17:45:53Z

/retest

YangKeao · 2026-01-29T17:50:54Z

/retest

YangKeao · 2026-01-29T17:55:52Z

/retest

AilinKid · 2026-02-01T08:35:15Z

/retest-required

Signed-off-by: AilinKid <314806019@qq.com>

hawkingrei · 2026-02-02T15:25:40Z

/retest

Copilot AI review requested due to automatic review settings January 21, 2026 16:45

ti-chi-bot bot added release-note-none Denotes a PR that doesn't merit a release note. size/L Denotes a PR that changes 100-499 lines, ignoring generated files. labels Jan 21, 2026

ti-chi-bot bot added the sig/planner SIG: Planner label Jan 21, 2026

Copilot started reviewing on behalf of AilinKid January 21, 2026 16:45 View session

Copilot AI reviewed Jan 21, 2026

View reviewed changes

qw4990 reviewed Jan 23, 2026

View reviewed changes

pkg/bindinfo/binding_plan_generation.go Show resolved Hide resolved

qw4990 reviewed Jan 23, 2026

View reviewed changes

pkg/bindinfo/binding_plan_generation.go Outdated Show resolved Hide resolved

qw4990 reviewed Jan 23, 2026

View reviewed changes

guo-shaoge reviewed Jan 26, 2026

View reviewed changes

ti-chi-bot bot added size/XL Denotes a PR that changes 500-999 lines, ignoring generated files. and removed size/L Denotes a PR that changes 100-499 lines, ignoring generated files. labels Jan 27, 2026

Copilot AI review requested due to automatic review settings January 27, 2026 03:49

Copilot started reviewing on behalf of AilinKid January 27, 2026 03:50 View session

Copilot AI reviewed Jan 27, 2026

View reviewed changes

guo-shaoge approved these changes Jan 27, 2026

View reviewed changes

ti-chi-bot bot added the needs-1-more-lgtm Indicates a PR needs 1 more LGTM. label Jan 27, 2026

ti-chi-bot bot added the ok-to-test Indicates a PR is ready to be tested. label Jan 27, 2026

qw4990 approved these changes Jan 27, 2026

View reviewed changes

ti-chi-bot bot added approved lgtm and removed needs-1-more-lgtm Indicates a PR needs 1 more LGTM. labels Jan 27, 2026

AilinKid added 4 commits February 2, 2026 22:14

.

ae4830b

Signed-off-by: AilinKid <314806019@qq.com>

.

1a10d41

Signed-off-by: AilinKid <314806019@qq.com>

.

a89fc25

Signed-off-by: AilinKid <314806019@qq.com>

.

ca41e03

Signed-off-by: AilinKid <314806019@qq.com>

hawkingrei force-pushed the fix-plan-explore-index-hint branch from 83f102c to ca41e03 Compare February 2, 2026 14:14

ti-chi-bot bot merged commit 53c8982 into pingcap:master Feb 2, 2026
30 checks passed

AilinKid deleted the fix-plan-explore-index-hint branch February 3, 2026 09:14

Comments

Conversation

AilinKid commented Jan 21, 2026

What problem does this PR solve?

What changed and how does it work?

Check List

Release note

Uh oh!

tiprow bot commented Jan 21, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

codecov bot commented Jan 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

Uh oh!

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!

Choose a reason for hiding this comment

Uh oh!

guo-shaoge Jan 26, 2026 • 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!

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!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Copilot AI Jan 27, 2026

Choose a reason for hiding this comment

Uh oh!

AilinKid commented Jan 27, 2026

Uh oh!

tiprow bot commented Jan 27, 2026

Uh oh!

AilinKid commented Jan 27, 2026

Uh oh!

ti-chi-bot bot commented Jan 27, 2026

Uh oh!

qw4990 commented Jan 27, 2026

Uh oh!

ti-chi-bot bot commented Jan 27, 2026

[LGTM Timeline notifier]

Uh oh!

AilinKid commented Jan 27, 2026

Uh oh!

AilinKid commented Jan 28, 2026

Uh oh!

qw4990 commented Jan 28, 2026

Uh oh!

AilinKid commented Jan 29, 2026

Uh oh!

YangKeao commented Jan 29, 2026

Uh oh!

YangKeao commented Jan 29, 2026

Uh oh!

YangKeao commented Jan 29, 2026

Uh oh!

YangKeao commented Jan 29, 2026

codecov bot commented Jan 21, 2026 •

edited

Loading

guo-shaoge Jan 26, 2026 •

edited

Loading