Skip to content

[SPARK-16189][SQL] Add ExternalRDD logical plan for input with RDD to have a chance to eliminate serialize/deserialize. #13890

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
ueshin wants to merge 4 commits into from
Closed

[SPARK-16189][SQL] Add ExternalRDD logical plan for input with RDD to have a chance to eliminate serialize/deserialize. #13890

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
036d5fd
Add ExistingRDD logical plan for input with RDD to have a chance to e…
ueshin Jun 14, 2016
62aca29
Revert createDataFrame for Java Beans.
ueshin Jun 24, 2016
e218f5f
Rename ExistingRDD to ExternalRDD.
ueshin Jul 8, 2016
61da040
Uncurry `apply` method.
ueshin Jul 9, 2016
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Prev Previous commit
Next Next commit
Rename ExistingRDD to ExternalRDD.
  • Loading branch information
@ueshin
ueshin committed Jul 8, 2016
commit e218f5fbef996ca3e9606cc68bf433c83ebf224e
4 changes: 2 additions & 2 deletions sql/core/src/main/scala/org/apache/spark/sql/SparkSession.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -238,7 +238,7 @@ class SparkSession private(
def createDataFrame[A <: Product : TypeTag](rdd: RDD[A]): DataFrame = {
SparkSession.setActiveSession(this)
val encoder = Encoders.product[A]
Dataset.ofRows(self, ExistingRDD(rdd)(self)(encoder))
Dataset.ofRows(self, ExternalRDD(rdd)(self)(encoder))
}

/**
Expand Down Expand Up @@ -423,7 +423,7 @@ class SparkSession private(
*/
@Experimental
def createDataset[T : Encoder](data: RDD[T]): Dataset[T] = {
Dataset[T](self, ExistingRDD(data)(self))
Dataset[T](self, ExternalRDD(data)(self))
}

/**
Expand Down
16 changes: 8 additions & 8 deletions sql/core/src/main/scala/org/apache/spark/sql/execution/ExistingRDD.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -74,28 +74,28 @@ object RDDConversions {
}
}

private[sql] object ExistingRDD {
private[sql] object ExternalRDD {

def apply[T: Encoder](rdd: RDD[T])(session: SparkSession): LogicalPlan = {
Copy link
Contributor

@cloud-fan cloud-fan Jul 8, 2016

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

why curry here?

Copy link
Member Author

@ueshin ueshin Jul 8, 2016

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Because I wanted to make the signature similar to the case class constructor.
Should I uncurry?

Copy link
Contributor

@cloud-fan cloud-fan Jul 8, 2016

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Actually I'm not sure why we curry the constructor either. Since RDDScan does it, it's ok we follow it. But for this apply method, I don't see the value of doing it. cc @yhuai

Copy link
Contributor

@marmbrus marmbrus Jul 8, 2016 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

There is probably no reason to use multiple parameter lists here. We sometimes use it for case classes so that arguments that should not effect equality are not included in the generated equals method.

val exisitingRdd = ExistingRDD(CatalystSerde.generateObjAttr[T], rdd)(session)
CatalystSerde.serialize[T](exisitingRdd)
val externalRdd = ExternalRDD(CatalystSerde.generateObjAttr[T], rdd)(session)
CatalystSerde.serialize[T](externalRdd)
}
}

/** Logical plan node for scanning data from an RDD. */
private[sql] case class ExistingRDD[T](
private[sql] case class ExternalRDD[T](
outputObjAttr: Attribute,
rdd: RDD[T])(session: SparkSession)
extends LeafNode with ObjectProducer with MultiInstanceRelation {

override protected final def otherCopyArgs: Seq[AnyRef] = session :: Nil

override def newInstance(): ExistingRDD.this.type =
ExistingRDD(outputObjAttr.newInstance(), rdd)(session).asInstanceOf[this.type]
override def newInstance(): ExternalRDD.this.type =
ExternalRDD(outputObjAttr.newInstance(), rdd)(session).asInstanceOf[this.type]

override def sameResult(plan: LogicalPlan): Boolean = {
plan.canonicalized match {
case ExistingRDD(_, otherRDD) => rdd.id == otherRDD.id
case ExternalRDD(_, otherRDD) => rdd.id == otherRDD.id
case _ => false
}
}
Expand All @@ -110,7 +110,7 @@ private[sql] case class ExistingRDD[T](
}

/** Physical plan node for scanning data from an RDD. */
private[sql] case class ExistingRDDScanExec[T](
private[sql] case class ExternalRDDScanExec[T](
outputObjAttr: Attribute,
rdd: RDD[T]) extends LeafExecNode with ObjectProducerExec {

Expand Down
2 changes: 1 addition & 1 deletion sql/core/src/main/scala/org/apache/spark/sql/execution/SparkStrategies.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -409,7 +409,7 @@ private[sql] abstract class SparkStrategies extends QueryPlanner[SparkPlan] {
case logical.RepartitionByExpression(expressions, child, nPartitions) =>
exchange.ShuffleExchange(HashPartitioning(
expressions, nPartitions.getOrElse(numPartitions)), planLater(child)) :: Nil
case ExistingRDD(outputObjAttr, rdd) => ExistingRDDScanExec(outputObjAttr, rdd) :: Nil
case ExternalRDD(outputObjAttr, rdd) => ExternalRDDScanExec(outputObjAttr, rdd) :: Nil
case LogicalRDD(output, rdd) => RDDScanExec(output, rdd, "ExistingRDD") :: Nil
case BroadcastHint(child) => planLater(child) :: Nil
case _ => Nil
Expand Down