suibianwanwank commented on PR #4382:
URL: https://github.com/apache/calcite/pull/4382#issuecomment-2910233387
Thank you all for your comments and reviews. I've refactored this PR to
address more related issues here. Although they are not directly connected,
they were all exposed and affecting each other within the same test. Given that
the code changes required are minimal, I'm inclined to resolve them together in
this PR.
For this SQL.
```
SELECT e1.*, d.deptno
FROM emp e1
LEFT JOIN dept d
ON e1.deptno = d.deptno
AND d.deptno IN (
SELECT e3.deptno
FROM emp e3
WHERE d.deptno > e3.comm
);
```
The current code has the following issues:
1. This PR addresses the problem described in the Jira ticket, where an
outer join produces output even when there are no matching rows.
2. The previous handling of the right side of the join was incorrect—it
requires not only `shiftFieldAccess` but also shifting
`subQuery.getOperands()`.
3. In `rewriteIn`, if `variablesSet` is not empty, `builder.join`
generates multiple `Correlate` nodes with the same `corId`, which leads to
decorrelation failure (I believe this results in an invalid `RelNode`).
--
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]
