planner: support BatchPointGet for partition table

[xiaodong-ji]

What problem does this PR solve?

Issue Number: close #21847

Problem Summary: Partition table didn't support BatchPointGet.

What is changed and how it works?

What's Changed:

How it Works:

Check List

Tests

• Manual test

create table t1 (c int primary key)
PARTITION BY RANGE (c) (
    PARTITION p0 VALUES LESS THAN (6),
    PARTITION p1 VALUES LESS THAN (11),
    PARTITION p2 VALUES LESS THAN (16),
    PARTITION p3 VALUES LESS THAN (21)
);

mysql> explain select * from t1 where c in (1,10);
+--------------------------+---------+------+---------------+---------------------------------------------+
| id                       | estRows | task | access object | operator info                               |
+--------------------------+---------+------+---------------+---------------------------------------------+
| PartitionUnion_9         | 4.00    | root |               |                                             |
| ├─Batch_Point_Get_10     | 2.00    | root | table:t1      | handle:[1 10], keep order:false, desc:false |
| └─Batch_Point_Get_11     | 2.00    | root | table:t1      | handle:[1 10], keep order:false, desc:false |
+--------------------------+---------+------+---------------+---------------------------------------------+
3 rows in set (0.01 sec)

Release note

• support BatchPointGet for partition table

[xiongjiwei]

(!ds.isPartition || ds.isPartition) always true

[xiaodong-ji]

fixed

[winoros]

@lysu For partition table, multi-range can also triger Get/BatchGet operation, right?

[lysu]

I guess maybe some reason let #14775 (comment) only support single range? @lzmhhh123 😕

[lzmhhh123]

Sorry, I forgot the reason. Maybe the previous point get can't handle the partition table.

[lysu]

oh...current point get can handle the partition

[Yisaer]

I think only modifying planner is not enough as I got the following test result after I applied the changing:

mysql> create table t1 (c int primary key)
    -> PARTITION BY RANGE (c) (
    ->     PARTITION p0 VALUES LESS THAN (6),
    ->     PARTITION p1 VALUES LESS THAN (11),
    ->     PARTITION p2 VALUES LESS THAN (16),
    ->     PARTITION p3 VALUES LESS THAN (21)
    -> );
Query OK, 0 rows affected (0.00 sec)

mysql> insert into t1 (c) values (1);
Query OK, 1 row affected (0.00 sec)

mysql> insert into t1 (c) values (0);
Query OK, 1 row affected (0.00 sec)

mysql> insert into t1 (c) values (10);
Query OK, 1 row affected (0.00 sec)

mysql> explain select * from t1 where c in (1,10);
+--------------------------+---------+------+---------------+---------------------------------------------+
| id                       | estRows | task | access object | operator info                               |
+--------------------------+---------+------+---------------+---------------------------------------------+
| PartitionUnion_9         | 4.00    | root |               |                                             |
| ├─Batch_Point_Get_10     | 2.00    | root | table:t1      | handle:[1 10], keep order:false, desc:false |
| └─Batch_Point_Get_11     | 2.00    | root | table:t1      | handle:[1 10], keep order:false, desc:false |
+--------------------------+---------+------+---------------+---------------------------------------------+
3 rows in set (0.00 sec)

mysql> select * from t1 where c in (1,10);
Empty set (0.00 sec)

The BatchPointGetExecutor.initialize is also needed to be revised. @xiaodong-ji

[zz-jason]

@xiaodong-ji please add some tests to ensure the correctness and performance improvements, we need these tests to check:

• execution plan changes are correct and as expected
• query result is correct
• performance is improved

[ti-srebot]

@xiaodong-ji, please update your pull request.

[ti-srebot]

@xiaodong-ji, please update your pull request.

[ti-srebot]

@xiaodong-ji PR closed due to no update for a long time. Feel free to reopen it anytime.