*: fix bug that UnionScan can't keep order caused wrong result

[tiancaiamao]

What problem does this PR solve?

Issue Number: close #33175

Problem Summary:

What is changed and how it works?

Actually, there are two bugs that cause the UnionScan fail to keep order,
and since it can't keep order, the following plan return wrong result.
Limit1->UnionScan->XXX doesn't return the max(id)

mysql> explain select max(id) from t;
+------------------------------+---------+-----------+---------------+-------------------------------------+
| id                           | estRows | task      | access object | operator info                       |
+------------------------------+---------+-----------+---------------+-------------------------------------+
| StreamAgg_10                 | 1.00    | root      |               | funcs:max([test.t.id](http://test.t.id/))->Column#3      |
| └─Limit_17                   | 1.00    | root      |               | offset:0, count:1                   |
|   └─UnionScan_21             | 1.00    | root      |               |                                     |
|     └─TableReader_23         | 1.00    | root      |               | data:TableFullScan_22               |
|       └─TableFullScan_22     | 1.00    | cop[tikv] | table:t       | keep order:true, desc, stats:pseudo |
+------------------------------+---------+-----------+---------------+-------------------------------------+
5 rows in set (0.00 sec)

The first bug is the datum comparing, the data type is KindUint64, but the old code treat it as GetInt64(),
10353107668348738101 become a negative number and 10353107668348738101 < 33

The second bug is the handling of kv range, in TiKV, the range is singed int
Here the data type is uint, so correct range handling is:

handling range: 7480000000000000415f720000000000000000 7480000000000000415f728000000000000000
handling range: 7480000000000000415f728000000000000000 7480000000000000415f72ffffffffffffffff00
add Row is nil, get snapshot row... [KindUint64 10353107668348738101]
add Row is nil, get snapshot row... [KindUint64 9734095886065816707]
add Row is nil, get snapshot row... [KindUint64 0]

the wrong range handling is:

handling range: 7480000000000000415f728000000000000000 7480000000000000415f72ffffffffffffffff00
in iterator ... add row == [KindUint64 0]
handling range: 7480000000000000415f720000000000000000 7480000000000000415f728000000000000000
in iterator ... add row == [KindUint64 9734095886065816707]
in iterator ... add row == [KindUint64 10353107668348738101]

i.e. the kv range order should be [-max int64, 0] [0, max int64]

Check List

Tests

• Unit test
• Integration test
• Manual test (add detailed scripts or steps below)
• No code

Side effects

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

Documentation

• Affects user behaviors
• Contains syntax changes
• Contains variable changes
• Contains experimental features
• Changes MySQL compatibility

Release note

Fix bug that 'select max(PK) ...' query return wrong result, the trigger condition is that UnionScan is used and the type of `PK` is uint64

[ti-chi-bot]

[REVIEW NOTIFICATION]

This pull request has been approved by:

• lcwangchao
• winoros

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.

[sre-bot]

Code Coverage Details: https://codecov.io/github/pingcap/tidb/commit/5d5555ed1845632a82528c440f8bebfe50310596

[jackysp]

When was this introduced? @tiancaiamao

[tiancaiamao]

When was this introduced? @tiancaiamao

I believe it's been there for quite a long time...
Maybe v3.0 or even v2.1 ... we already have it @jackysp

[lcwangchao]

It may still cause bugs when the int is signed, try the following:

create global temporary table `tmp2` (id bigint primary key) on commit delete rows;
begin;
insert into tmp2 values(-2),(-1),(0),(1),(2);
-- The following  query will give a wrong result: 
-- expected 2, 1, 0, -1, -2 , actual: -1, -2, 2, 1
select * from tmp2 where id <= -1 or id > 0 order by id desc; 

[tiancaiamao]

Update, PTAL @lcwangchao

[tiancaiamao]

/run-unit-test

[tiancaiamao]

/merge

[ti-chi-bot]

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

Commit hash: 41b3f5b

[ti-srebot]

cherry pick to release-6.0 in PR #33319

[ti-srebot]

cherry pick to release-5.4 in PR #37805