sunchao commented on a change in pull request #35657:
URL: https://github.com/apache/spark/pull/35657#discussion_r829346679
##########
File path:
sql/core/src/main/scala/org/apache/spark/sql/execution/exchange/EnsureRequirements.scala
##########
@@ -137,8 +138,16 @@ case class EnsureRequirements(
Some(finalCandidateSpecs.values.maxBy(_.numPartitions))
}
+ // Check if 1) all children are of `KeyGroupedPartitioning` and 2) they
are all compatible
+ // with each other. If both are true, skip shuffle.
+ val allCompatible = childrenIndexes.sliding(2).map {
+ case Seq(a, b) =>
+ checkKeyGroupedSpec(specs(a)) && checkKeyGroupedSpec(specs(b)) &&
Review comment:
> ... It looks like we can always eliminate shuffles if all the children
are compatible.
When children are compatible, it still doesn't mean each child's partition
keys **fully match** the distribution keys. Since in
`KeyGroupedPartitioning.satisfies`, we only check if partition keys are subset
of distribution keys.
That's why we need an additional check here to make sure that when
`spark.sql.requireAllClusterKeysForCoPartition` is true, the child partition
keys fully match the distribution keys.
--
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]
