Internal error: UnionExec chunk column count mismatch

[ChenPeng2013]

Bug Report

Please answer these questions before submitting your issue. Thanks!

1. Minimal reproduce step (Required)

use test;
drop table if exists IDT_HP24172;
CREATE TABLE `IDT_HP24172` (
  `COL1` mediumint(16) DEFAULT NULL,
  `COL2` varchar(20) DEFAULT NULL,
  `COL4` datetime DEFAULT NULL,
  `COL3` bigint(20) DEFAULT NULL,
  `COL5` float DEFAULT NULL,
  KEY `UM_COL` (`COL1`,`COL3`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_bin
PARTITION BY HASH( `COL1`+`COL3` )
PARTITIONS 8;
insert into IDT_HP24172(col1) values(8388607);
select col2 from IDT_HP24172 where col1 = 8388607 and col1 in (select col1 from IDT_HP24172);

2. What did you expect to see? (Required)

mysql> select col2 from IDT_HP24172 where col1 = 8388607 and col1 in (select col1 from IDT_HP24172);
+------+
| col2 |
+------+
| NULL |
+------+
1 row in set (0.01 sec)

3. What did you see instead (Required)

mysql> select col2 from IDT_HP24172 where col1 = 8388607 and col1 in (select col1 from IDT_HP24172);
ERROR 1105 (HY000): Internal error: UnionExec chunk column count mismatch, req: 2, result: 1

4. What is your TiDB version? (Required)

It can be reproduced on branch release-5.0, release-5.1 and master
The release-4.0 branch doesn't have the issue

Release Version: v5.1.0-64-gfb0eaf7b4
Edition: Community
Git Commit Hash: fb0eaf7b44b9ddb1494bf8194d8e361c538146dc
Git Branch: release-5.1
UTC Build Time: 2021-07-26 06:50:38
GoVersion: go1.16.5
Race Enabled: false
TiKV Min Version: v3.0.0-60965b006877ca7234adaced7890d7b029ed1306
Check Table Before Drop: false

[xxchan]

a smaller reproduction:

CREATE TABLE t (
  a int,
  b int
)
PARTITION BY HASH( a )
PARTITIONS 2;
insert into t values(1,2);
select * from t where b = 2 and b in (select b from t);

> explain select * from t where b = 2 and b in (select b from t);
+------------------------------+-----------+-----------+-----------------------+-------------------------------------------------------+
| id                           |   estRows | task      | access object         | operator info                                         |
|------------------------------+-----------+-----------+-----------------------+-------------------------------------------------------|
| HashJoin_21                  |     90.01 | root      |                       | CARTESIAN inner join                                  |
| ├─HashAgg_30(Build)          |      9    | root      |                       | group by:test.t.b, funcs:firstrow(1)->Column#8        |
| │ └─PartitionUnion_31        |      9    | root      |                       |                                                       |
| │   ├─HashAgg_34             |      8    | root      |                       | group by:test.t.b, funcs:firstrow(test.t.b)->test.t.b |
| │   │ └─TableReader_41       |     10    | root      |                       | data:Selection_40                                     |
| │   │   └─Selection_40       |     10    | cop[tikv] |                       | eq(2, test.t.b)                                       |
| │   │     └─TableFullScan_39 |  10000    | cop[tikv] | table:t, partition:p0 | keep order:false, stats:pseudo                        |
| │   └─HashAgg_44             |      1    | root      |                       | group by:test.t.b, funcs:firstrow(test.t.b)->test.t.b |
| │     └─TableReader_51       |      0    | root      |                       | data:Selection_50                                     |
| │       └─Selection_50       |      0    | cop[tikv] |                       | eq(2, test.t.b)                                       |
| │         └─TableFullScan_49 |      1    | cop[tikv] | table:t, partition:p1 | keep order:false, stats:pseudo                        |
| └─PartitionUnion_23(Probe)   |     10    | root      |                       |                                                       |
|   ├─TableReader_26           |     10    | root      |                       | data:Selection_25                                     |
|   │ └─Selection_25           |     10    | cop[tikv] |                       | eq(test.t.b, 2)                                       |
|   │   └─TableFullScan_24     |  10000    | cop[tikv] | table:t, partition:p0 | keep order:false, stats:pseudo                        |
|   └─TableReader_29           |      0    | root      |                       | data:Selection_28                                     |
|     └─Selection_28           |      0    | cop[tikv] |                       | eq(test.t.b, 2)                                       |
|       └─TableFullScan_27     |      1    | cop[tikv] | table:t, partition:p1 | keep order:false, stats:pseudo                        |
+------------------------------+-----------+-----------+-----------------------+-------------------------------------------------------+

It seems that HashAgg children caused the problem? (I don't know how to fix it 🤔)

[XuHuaiyu]

• UnionStore does not have this problem
• Reproduced when using TiKV

[wshwsh12]

Release 4.0 also has this problem, but doesn't throw error.
Because in release 4.0, UnionExec doesn't have the following check:

tidb/executor/executor.go

Lines 1617 to 1620 in 880c9b5

	if result.chk.NumCols() != req.NumCols() {
	return errors.Errorf("Internal error: UnionExec chunk column count mismatch, req: %d, result: %d",
	req.NumCols(), result.chk.NumCols())
	}

[github-actions]

Please check whether the issue should be labeled with 'affects-x.y' or 'backport-x.y.z',
and then remove 'needs-more-info' label.