Yaoxuan Wu created FLINK-40922:
----------------------------------
Summary: DISTINCT over GROUP BY ROLLUP returns duplicate rows
Key: FLINK-40922
URL: https://issues.apache.org/jira/browse/FLINK-40922
Project: Flink
Issue Type: Bug
Components: Table SQL / Planner
Affects Versions: 2.3.0
Reporter: Yaoxuan Wu
When all grouping keys of a ROLLUP group are NULL, that group's row can be
identical to the grand-total row. A DISTINCT on top of the rollup must collapse
the two rows, but Flink returns both.
{code:java}
SELECT DISTINCT * FROM (
SELECT k, COUNT(x) AS c
FROM (VALUES (CAST(NULL AS INT), 1), (CAST(NULL AS INT), 2)) AS v(k, x)
GROUP BY ROLLUP(k)
); {code}
The inner query's result is two rows (which is correct)
{code:java}
+-------------+----------------------+
| k | c |
+-------------+----------------------+
| <NULL> | 2 |
| <NULL> | 2 |
+-------------+----------------------+ {code}
But the DISTINCT over the inner query still returns two rows. (the same as
above)
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
