dbtsai commented on a change in pull request #24049: [SPARK-27123][SQL] Improve
CollapseProject to handle projects cross limit/repartition/sample
URL: https://github.com/apache/spark/pull/24049#discussion_r264537308
##########
File path:
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/Optimizer.scala
##########
@@ -699,6 +699,24 @@ object CollapseProject extends Rule[LogicalPlan] {
agg.copy(aggregateExpressions = buildCleanedProjectList(
p.projectList, agg.aggregateExpressions))
}
+ case p1 @ Project(_, g @ GlobalLimit(_, l @ LocalLimit(_, p2: Project))) =>
+ if (haveCommonNonDeterministicOutput(p1.projectList, p2.projectList)) {
+ p1
+ } else {
+ val newProjectList = buildCleanedProjectList(p1.projectList,
p2.projectList)
+ g.copy(child = l.copy(child = p2.copy(projectList = newProjectList)))
+ }
+ case p1 @ Project(_, l @ LocalLimit(_, p2: Project)) =>
+ if (haveCommonNonDeterministicOutput(p1.projectList, p2.projectList)) {
+ p1
+ } else {
+ val newProjectList = buildCleanedProjectList(p1.projectList,
p2.projectList)
+ l.copy(child = p2.copy(projectList = newProjectList))
+ }
+ case Project(list, r @ Repartition(_, _, p: Project)) if
list.forall(_.deterministic) =>
+ r.copy(child = p.copy(projectList = buildCleanedProjectList(list,
p.projectList)))
Review comment:
Is it possible that this can increase the shuffled data?
For example, if we apply a deterministic UDF taking two input in the project
which produces like 10 columns as output after repartition, with this rule,
this UDF will be moved before the repartition which can result more shuffle,
right?
Thanks!
----------------------------------------------------------------
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]
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
