planner: add missing CanExprsPushDown checks when generating IndexMerge path for or
#41361
Conversation
|
[REVIEW NOTIFICATION] This pull request has been approved by:
To complete the pull request process, please ask the reviewers in the list to review by filling The full list of commands accepted by this bot can be found here. Reviewer can indicate their review by submitting an approval review. |
ti-chi-bot
added
do-not-merge/needs-triage-completed
release-note-none
ti-chi-bot
added
status/LGT1
ti-chi-bot
added
status/LGT2
|
/merge |
|
This pull request has been accepted and is ready to merge. Commit hash: 06a5ce2
|
ti-chi-bot
added
the
status/can-merge
| expectedRes := []string{"ir1z,f1,e6yd d 13mdezixgcn", "ir1z,f1,e6yd i71o 13mdezixgcn", "nwbk 8ue 13mdezixgcn"} | ||
| tk.MustQuery("select * from t where a between 'e6yd' and 'z' or b <> '8ue';").Sort().Check(testkit.Rows(expectedRes...)) | ||
| tk.MustQuery("select /*+ use_index_merge(t) */ * from t where a between 'e6yd' and 'z' or b <> '8ue';").Sort().Check(testkit.Rows(expectedRes...)) | ||
| // For now tidb doesn't support push set type to TiKV, and column a is a set type, so we shouldn't generate a IndexMerge path. |
There was a problem hiding this comment.
Do we need add some log or warning to indicate why cannot use index merge ?
There was a problem hiding this comment.
It would print a general warning "IndexMerge is inapplicable" like other cases when we failed to generate an IndexMerge path.
There was a problem hiding this comment.
It would be better if it could print the particular reason. But we also don't print the reason in other cases, so I think it's unnecessary to add it in this PR.
|
In response to a cherrypick label: new pull request created to branch |
Signed-off-by: ti-chi-bot <[email protected]>
|
In response to a cherrypick label: new pull request created to branch |
Signed-off-by: ti-chi-bot <[email protected]>
|
In response to a cherrypick label: new pull request created to branch |
Signed-off-by: ti-chi-bot <[email protected]>
…erge path for `or` (pingcap#41361) close pingcap#41273, close pingcap#41293
…erge path for `or` (pingcap#41361) close pingcap#41273, close pingcap#41293
What problem does this PR solve?
Issue Number: close #41273 close #41293
Problem Summary:
This bug is introduced by #30341.
This case is related to the
SETtype mainly. We don't support pushing down theSETtype now.util/rangeralso can't correctly handle expressions onSETcolumn. So it requires these expressions be blocked in advance. After #30341, such expressions can reachutil/rangerwithout theCanExprsPushDowncheck.What is changed and how it works?
Add a
CanExprsPushDowncheck before using an expression to build an IndexMerge partial path.Check List
Tests
Side effects
Documentation
Release note
Please refer to Release Notes Language Style Guide to write a quality release note.