What problem does this PR solve?

When the child of MaxOneRow operator returns a chunk with only one row, we should check whether the child has more result by calling the child.Next again. In the before, this validation was not employed.

Before this PR, when we set tidb_max_chunk_size to 1, the behavior is wrong:

TiDB(localhost:4000) > set @@tidb_max_chunk_size=1;
Query OK, 0 rows affected (0.00 sec)

TiDB(localhost:4000) > select (select t1.a from t1 where t1.a > t2.a) as a from t2;
+------+
| a    |
+------+
|    1 |
+------+
1 row in set (0.00 sec)

TiDB(localhost:4000) > set @@tidb_max_chunk_size=1024;
Query OK, 0 rows affected (0.00 sec)

TiDB(localhost:4000) > select (select t1.a from t1 where t1.a > t2.a) as a from t2;
ERROR 1105 (HY000): subquery returns more than 1 row

What is changed and how it works?

In this PR, when the first Chunk returned from the child, if the number of records is not equal to 1, we return an error about "subquery returns more than 1 row", Otherwise we call e.children[0].Next again to check whether there are more records existed in the child operator.

Check List

Tests

• Integration test

Code changes

• Has exported function/method change

Related changes

• Need to cherry-pick to the release branch