FangHao created FLINK-40753:
-------------------------------
Summary: Correlated subqueries over SQL VALUES fail during
decorrelation
Key: FLINK-40753
URL: https://issues.apache.org/jira/browse/FLINK-40753
Project: Flink
Issue Type: Bug
Components: Table SQL / Planner
Affects Versions: 2.4.0
Environment: Flink 2.4-SNAPSHOT, upstream master at 43c220f6dc3.
Java 21 on macOS.
Reproduced in both batch and streaming modes using
TableEnvironment.explainSql().
Reporter: FangHao
Correlated EXISTS / NOT EXISTS subqueries over SQL VALUES fail during planning
in both batch and streaming modes. This is reproducible through the Java
TableEnvironment without the Python DataFrame API, on upstream master at
43c220f6dc3.
Reproducer:
{code:sql}
SELECT l.id
FROM (VALUES (1), (2)) AS l(id)
WHERE NOT EXISTS (
SELECT 1
FROM (VALUES (1), (2)) AS r(right_id)
WHERE l.id = r.right_id
);
{code}
Expected: the query plans successfully and returns no rows.
Actual:
{code}
org.apache.flink.table.api.TableException:
unexpected correlate variable $cor0 in the plan
{code}
SubQueryDecorrelator.decorrelateRel(Values) returns a null Frame, which signals
that decorrelation cannot continue. A Values node does not require
modification, but it should return an identity Frame so its parent
filter/project can still be decorrelated.
Once this path is enabled, project decorrelation must also handle
RelOptUtil.createProject simplifying a projection to Values, rather than
assuming the result is always a LogicalProject.
A fix has been validated locally with batch/streaming plan tests and batch
execution tests covering NULL keys, duplicate rows, empty inputs, projections,
filters, and computed keys.
Discovered while implementing FLINK-40419. Reviewer discussion:
https://github.com/apache/flink/pull/29201#discussion_r4068948306
I would like to work on this fix. Could a committer assign this issue to me?
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
