planner: fix index join on unmatched collation suffix columns paniced

[lysu]

What problem does this PR solve?

Issue Number: close #24706

Problem Summary:

index join on a, b, but b's collation is not matched, b's curIdxOff2KeyOff not be set to -1?

What is changed and how it works?

What's Changed, How it Works:

index join on a, b, but b's collation is not matched, we should only build range (a)

Related changes

• Need to cherry-pick to the release branch 5.0

Check List

Tests

• Unit test

Side effects

• n/a

Release note

• Fix panic caused by index join on unmatch collation columns

[lysu]

@wjhuang2016 PTAL, I'm not very sure about this change - -

[winoros]

Why we did this change? Or should we strengthen the case to STRING type only?

[lysu]

if we didn't change this

for table t1 (a int not null, b varcharbinary not null,index (a, b)) and t2(a int not null, b varchar not null, index(a, b))

the query: 't1 join t2 on t1. a = tb2.a and t2.a = t2.b'

will cause ranges only contain column a's value(caused by https://github.com/pingcap/tidb/pull/21201...not use the column if collation isn't matched), but keyOff2IdxOff include both a and b's offset in

tidb/executor/builder.go

Line 3677 in bfdad7c

for keyOff, idxOff := range keyOff2IdxOff {

so the query will panic

@winoros is any other advise to fix this~? I'm not familiar with those part logic 😂

[lysu]

@winoros PTAL if free thx...

[winoros]

The logic around this function is:

• We want to build an index's range using the join conditions.
• There are two type expression: col_1=col_2, col_1=const, col1 in (const_list).
• So for a join, the expression used to construct ranges may be like idx_col1=col_1 and idx_col2 in (const_list) and idx_col3 = col3.
• The curIdxOff2KeyOff records the map relation between idx_col1 with col1 and idx_col3 with col3, so it might be the case [0, -1, 2].
• The curNotUsedIndexCols records the expression idx_col2 in (const_list). So it would be [idx_col2] for this case.

The removeUselessEqAndInFunc is to maintain the curIdxOff2KeyOff. It will check the index column whose position is occupied by -1 in curIdxOff2KeyOff can be found in curNotUsedIndexCols. Otherwise there's neither idx_col=another_col nor idx_col=const or idx_col in (const_list) expression so we cannot build range for index from its position. So we are going to reset the value to -1 after it.

For this the issue to be fixed in the pr. The collation check for the column is a little later. We'd better add the check earlier at https://github.com/pingcap/tidb/blob/master/planner/core/exhaust_physical_plans.go#L1219.

[winoros]

Also, i think we'd better check the collation after we checked the column type is a STRING type. /cc @wjhuang2016

[lysu]

@winoros @wjhuang2016 PTAL if free thx~

[ti-chi-bot]

[REVIEW NOTIFICATION]

This pull request has been approved by:

• winoros
• wjhuang2016

To complete the pull request process, please ask the reviewers in the list to review by filling /cc @reviewer in the comment.
After your PR has acquired the required number of LGTMs, you can assign this pull request to the committer in the list by filling /assign @committer in the comment to help you merge this pull request.

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

Reviewer can indicate their review by submitting an approval review.
Reviewer can cancel approval by submitting a request changes review.

[lysu]

/merge

[ti-chi-bot]

@lysu: /merge is only allowed for the committers, you can assign this pull request to the committer in list by filling /assign @committer in the comment to help merge this pull request.

In response to this:

/merge

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 ti-community-infra/tichi repository.

[lysu]

/merge

[ti-chi-bot]

This pull request has been accepted and is ready to merge.

Commit hash: 3883bd6

[zhouqiang-cl]

/run-check_dev_2

[lysu]

/run-integration-common-test

[ti-srebot]

cherry pick to release-5.1 in PR #25436