Fly-a-Kite opened a new issue, #26488:
URL: https://github.com/apache/datafusion/issues/26488
### Describe the bug
### A subquery that ends in ORDER BY ... OFFSET n returns the wrong
surviving rows when its consumer is an aggregate. The logical plan keeps the
Sort, but the physical plan drops the SortExec below the GlobalLimitExec
because the aggregate does not require ordered input. OFFSET still has to skip
rows in the declared order, so the result changes when the physical input row
order changes.
### version:
```text
DataFusion on 53.0.0, 54.0.0 and 55.0.0
```
### To Reproduce
```python
CREATE TABLE t(id BIGINT, g BIGINT, v BIGINT, s VARCHAR);
INSERT INTO t VALUES
(0, 0, 3, '中文'),
(1, 0, 0, 'a'),
(2, 1, -1, 'a'),
(3, 1, -1, 'a'),
(4, 0, 3, '中文'),
(5, 0, 3, NULL);
SELECT g, SUM(v) AS sv
FROM (
SELECT *
FROM t
ORDER BY s ASC NULLS LAST, id ASC NULLS LAST
OFFSET 1
) q
GROUP BY g
ORDER BY g;
```
### Expected behavior
Expected (ORDER BY s, id is a total order; the two 中文 rows are identical):
```text
g | sv
0 | 9
1 | -2
```
DataFusion 55.0.0:
```text
g | sv
0 | 6
1 | -2
```
The offset skipped a 中文 row instead of the first a row. It is not an
aggregate-order artifact: the aggregate reads a different set of rows.
### Additional context
_No response_
### Willingness to contribute
None
--
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]
