tgravescs commented on a change in pull request #29262:
URL: https://github.com/apache/spark/pull/29262#discussion_r461718596
##########
File path:
sql/core/src/main/scala/org/apache/spark/sql/execution/adaptive/OptimizeLocalShuffleReader.scala
##########
@@ -78,10 +79,11 @@ case class OptimizeLocalShuffleReader(conf: SQLConf)
extends Rule[SparkPlan] {
private def getPartitionSpecs(
shuffleStage: ShuffleQueryStageExec,
advisoryParallelism: Option[Int]): Seq[ShufflePartitionSpec] = {
- val shuffleDep = shuffleStage.shuffle.shuffleDependency
- val numReducers = shuffleDep.partitioner.numPartitions
+ val numReducers = shuffleStage.shuffle.numPartitions
val expectedParallelism = advisoryParallelism.getOrElse(numReducers)
- val numMappers = shuffleDep.rdd.getNumPartitions
+ val mapOutputTracker =
SparkEnv.get.mapOutputTracker.asInstanceOf[MapOutputTrackerMaster]
+ val numMappers =
+
mapOutputTracker.shuffleStatuses(shuffleStage.mapStats.get.shuffleId).mapStatuses.length
Review comment:
personally I'm not fond of us reaching into mapOutputTracker and looking
at its internal datastructures. This I believe was exposed for testing purposes
not for people to use it directly. I would rather see either a utility
function made or we do logic like we were before.
##########
File path:
sql/core/src/main/scala/org/apache/spark/sql/execution/adaptive/AdaptiveSparkPlanExec.scala
##########
@@ -376,10 +382,14 @@ case class AdaptiveSparkPlanExec(
private def newQueryStage(e: Exchange): QueryStageExec = {
val optimizedPlan = applyPhysicalRules(e.child, queryStageOptimizerRules)
val queryStage = e match {
- case s: ShuffleExchangeExec =>
- ShuffleQueryStageExec(currentStageId, s.copy(child = optimizedPlan))
+ case s: ShuffleExchangeLike =>
Review comment:
this should be able to stay ShuffleExchangeExec or be consistent and
change braodcast below to be BroadcastExchangeLike
##########
File path:
sql/core/src/main/scala/org/apache/spark/sql/execution/exchange/ShuffleExchangeExec.scala
##########
@@ -40,13 +41,49 @@ import org.apache.spark.sql.types.StructType
import org.apache.spark.util.MutablePair
import org.apache.spark.util.collection.unsafe.sort.{PrefixComparators,
RecordComparator}
+/**
+ * Common trait for all shuffle exchange implementations to facilitate pattern
matching.
+ */
+trait ShuffleExchangeLike extends Exchange {
+
+ /**
+ * Returns the number of mappers of this shuffle.
+ */
+ def numMappers: Int
Review comment:
need to update the tests fto have this
----------------------------------------------------------------
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]
