AveryQi115 opened a new pull request, #57614:
URL: https://github.com/apache/spark/pull/57614
### What changes were proposed in this pull request?
In `DecorrelateInnerQuery`, the `Limit` case peels the `Sort` off the
subquery and saves its ordering. When the correlated predicate references only
outer columns, decorrelation produces no partitioning key
(`partitionFields.isEmpty`) and that branch re-emits the limit as `Limit(k,
newChild)` / `Limit(k, Offset(m, newChild))` without re-applying the saved
ordering, turning `ORDER BY ... LIMIT k` into an arbitrary `LIMIT k`.
This PR re-applies the saved ordering as a global `Sort` below the limit in
the `partitionFields.isEmpty` branch (covering both `LIMIT` and `LIMIT ...
OFFSET`), so ordered limits stay order-preserving and deterministic. A new
internal flag
`spark.sql.optimizer.decorrelateLimitOffsetLegacyIncorrectOrderHandling.enabled`
(default false) can restore the legacy behavior.
The standalone `Offset` case (no `LIMIT`) is not affected: it decorrelates
the original input with the `Sort` retained, so its ordering is already
preserved.
### Why are the changes needed?
The dropped `ORDER BY` makes an ordered-limit subquery return a
non-deterministic row. For example:
```
SELECT t1a, t1b FROM t1
WHERE t1c = (SELECT t2c FROM t2 WHERE t1b < t1d ORDER BY t2c LIMIT 1);
```
With distinct `t2c` values `{12, 16, NULL}` and Spark's default `ASC NULLS
FIRST`, the correct ordered `LIMIT 1` returns `NULL`, so the row is correctly
excluded. With the bug, an arbitrary row is picked, which can incorrectly emit
a row and varies by physical execution order.
### Does this PR introduce any user-facing change?
Yes. Correlated scalar/predicate subqueries of the form `ORDER BY ... LIMIT`
whose correlation references only the outer table now return correct,
deterministic results. Results that previously depended on the dropped ordering
may change. The legacy behavior can be restored via the config above.
### How was this patch tested?
New unit tests in `DecorrelateInnerQuerySuite` covering default ordering,
explicit `ASC NULLS LAST`, `LIMIT ... OFFSET`, and the legacy-flag path.
Golden-file results for `scalar-subquery-predicate.sql` are updated accordingly.
### Was this patch authored or co-authored using generative AI tooling?
Yes, drafted with assistance from Claude.
<!--
Thanks for sending a pull request! Here are some tips for you:
1. If this is your first time, please read our contributor guidelines:
https://spark.apache.org/contributing.html
2. Ensure you have added or run the appropriate tests for your PR:
https://spark.apache.org/developer-tools.html
3. If the PR is unfinished, add '[WIP]' in your PR title, e.g.,
'[WIP][SPARK-XXXX] Your PR title ...'.
4. Be sure to keep the PR description updated to reflect all changes.
5. Please write your PR title to summarize what this PR proposes.
6. If possible, provide a concise example to reproduce the issue for a
faster review.
7. If you want to add a new configuration, please read the guideline first
for naming configurations in
'core/src/main/scala/org/apache/spark/internal/config/ConfigEntry.scala'.
8. If you want to add or modify an error type or message, please read the
guideline first in
'common/utils/src/main/resources/error/README.md'.
-->
--
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]
