haohuaijin opened a new issue, #24414:
URL: https://github.com/apache/datafusion/issues/24414
### Describe the bug
`SortMergeJoinExec` may return incorrect results after projection pushdown
when the join has a non-equi filter.
### To Reproduce
```sql
set datafusion.optimizer.prefer_hash_join = false;
set datafusion.optimizer.repartition_joins = true;
set datafusion.execution.target_partitions = 2;
select t1.column2 as left_b1, t2.column2 as right_b1
from (values (100, 1, 0)) t1
join (values (10, 1)) t2
on t1.column2 = t2.column2
and t1.column1 > t2.column1;
```
The expected result is `(1, 1)`, but the query returns no rows.
`SortMergeJoinExec::try_swapping_with_projection` projects the join inputs
but reuses the original `JoinFilter`. Its column indices can therefore refer to
the wrong columns in the new schemas.
The projection swap should rewrite the filter with `update_join_filter`, or
be rejected when the filter columns are not projected.
### Expected behavior
_No response_
### Additional context
_No response_
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
