XiDuo You created SPARK-60904:
---------------------------------
Summary: Preserve the output partitioning of shuffled FULL OUTER
joins
Key: SPARK-60904
URL: https://issues.apache.org/jira/browse/SPARK-60904
Project: Spark
Issue Type: Improvement
Components: SQL
Affects Versions: 5.0.0
Reporter: XiDuo You
A shuffled FULL OUTER join reports UnknownPartitioning, so every consumer keyed
on the join keys shuffles the join output again, e.g. GROUP BY or a window on
the merged key of a FULL OUTER JOIN ... USING, or a chained FULL OUTER JOIN ...
USING. Neither side's partitioning describes the output, since a NULL-padded
row was placed by the other side's key, but the position-wise COALESCE of the
two sides' join keys describes every row. This reports a hash partitioning on
that COALESCE when the children are co-partitioned by hash, every join key is
deterministic and NULL on a NULL-padded row, and keeps it null-aware when
either side spreads NULL keys. Keyed (storage-partitioned) layouts keep
UnknownPartitioning.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
