sunchao commented on a change in pull request #35657:
URL: https://github.com/apache/spark/pull/35657#discussion_r828271471
##########
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 `DataSourcePartitioning` 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) =>
+ checkDataSourceSpec(specs(a)) && checkDataSourceSpec(specs(b)) &&
+ specs(a).isCompatibleWith(specs(b))
+ }.forall(_ == true)
+
children = children.zip(requiredChildDistributions).zipWithIndex.map {
- case ((child, _), idx) if !childrenIndexes.contains(idx) =>
+ case ((child, _), idx) if allCompatible ||
!childrenIndexes.contains(idx) =>
Review comment:
`DataSourceShuffleSpec` (without enabling it to re-shuffle other
children) can't fit nicely into this framework at the moment, I think.
Since its `canCreatePartitioning` is false, we'll treat it the same as
`RangeShuffleSpec`. In case both sides are of `DataSourceShuffleSpec`, the
`bestSpecOpt` will be `None` in `EnsureRequirements` and Spark will create
default partitioning from the required distribution instead.
--
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]
