cloud-fan commented on a change in pull request #29614:
URL: https://github.com/apache/spark/pull/29614#discussion_r481688764
##########
File path:
sql/core/src/main/scala/org/apache/spark/sql/execution/adaptive/EliminateJoinToEmptyRelation.scala
##########
@@ -36,22 +36,31 @@ import
org.apache.spark.sql.execution.joins.{EmptyHashedRelation, HashedRelation
*/
object EliminateJoinToEmptyRelation extends Rule[LogicalPlan] {
- private def canEliminate(plan: LogicalPlan, relation: HashedRelation):
Boolean = plan match {
- case LogicalQueryStage(_, stage: BroadcastQueryStageExec) if
stage.resultOption.get().isDefined
- && stage.broadcast.relationFuture.get().value == relation => true
- case _ => false
+ private def canEliminate(streamedPlan: LogicalPlan,
+ buildPlan: LogicalPlan, relation: HashedRelation): Boolean = {
+ // If streamedSide of the Join is
ShuffleQueryStageExec(canChangeNumPartitions== false)
+ // it can't be rewritten to EmptyRelation because the conversion will lost
user specified
+ // number partition information.
+ (streamedPlan, buildPlan) match {
+ case (LogicalQueryStage(_, stage: ShuffleQueryStageExec), _)
Review comment:
what if the stream side has some more operators above the shuffle stage?
e.g. `df1.repartition(...).filter(...).join(df2)`
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
