Skip to content

Spy 764 - Pull in SPARK-11009 fix for window function #102

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

Merged
merged 2 commits into from
Oct 12, 2015
Merged

Spy 764 - Pull in SPARK-11009 fix for window function #102

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
f5dff3c
fix wrong result of Window function in cluster mode
Oct 9, 2015
e80266c
add a comment
Oct 10, 2015
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
20 changes: 13 additions & 7 deletions sql/core/src/main/scala/org/apache/spark/sql/execution/Window.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -146,6 +146,8 @@ case class Window(
// to the result of bound value projection. This is done manually because we want to use
// Code Generation (if it is enabled).
val (sortExprs, schema) = exprs.map { case e =>
// This AttributeReference does not need to have unique IDs, it's OK to be called
// in executor.
val ref = AttributeReference("ordExpr", e.dataType, e.nullable)()
(SortOrder(ref, e.direction), ref)
}.unzip
Expand Down Expand Up @@ -201,18 +203,17 @@ case class Window(
* This method uses Code Generation. It can only be used on the executor side.
*
* @param expressions unbound ordered function expressions.
* @param attributes output attributes
* @return the final resulting projection.
*/
private[this] def createResultProjection(
expressions: Seq[Expression]): MutableProjection = {
val unboundToAttr = expressions.map {
e => (e, AttributeReference("windowResult", e.dataType, e.nullable)())
}
val unboundToAttrMap = unboundToAttr.toMap
expressions: Seq[Expression],
attributes: Seq[Attribute]): MutableProjection = {
val unboundToAttrMap = expressions.zip(attributes).toMap
val patchedWindowExpression = windowExpression.map(_.transform(unboundToAttrMap))
newMutableProjection(
projectList ++ patchedWindowExpression,
child.output ++ unboundToAttr.map(_._2))()
child.output ++ attributes)()
}

protected override def doExecute(): RDD[InternalRow] = {
Expand Down Expand Up @@ -247,12 +248,17 @@ case class Window(
factories(index) = () => createFrameProcessor(frame, functions, ordinal)
}

// AttributeReference can only be created in driver, or the id will not be unique
val outputAttributes = unboundExpressions.map {
e => AttributeReference("windowResult", e.dataType, e.nullable)()
}

// Start processing.
child.execute().mapPartitions { stream =>
new Iterator[InternalRow] {

// Get all relevant projections.
val result = createResultProjection(unboundExpressions)
val result = createResultProjection(unboundExpressions, outputAttributes)
val grouping = if (child.outputsUnsafeRows) {
UnsafeProjection.create(partitionSpec, child.output)
} else {
Expand Down