Github user tdas commented on a diff in the pull request:
https://github.com/apache/spark/pull/19452#discussion_r144435539
--- Diff:
sql/core/src/main/scala/org/apache/spark/sql/execution/streaming/StreamingSymmetricHashJoinExec.scala
---
@@ -221,43 +237,36 @@ case class StreamingSymmetricHashJoinExec(
// matching new left input with new right input, since the new left
input has become stored
// by that point. This tiny asymmetry is necessary to avoid
duplication.
val leftOutputIter =
leftSideJoiner.storeAndJoinWithOtherSide(rightSideJoiner) {
- (input: UnsafeRow, matched: UnsafeRow) =>
joinedRow.withLeft(input).withRight(matched)
+ (input: InternalRow, matched: InternalRow) =>
joinedRow.withLeft(input).withRight(matched)
}
val rightOutputIter =
rightSideJoiner.storeAndJoinWithOtherSide(leftSideJoiner) {
- (input: UnsafeRow, matched: UnsafeRow) =>
joinedRow.withLeft(matched).withRight(input)
+ (input: InternalRow, matched: InternalRow) =>
joinedRow.withLeft(matched).withRight(input)
}
- // Filter the joined rows based on the given condition.
- val outputFilterFunction =
newPredicate(condition.getOrElse(Literal(true)), output).eval _
-
// We need to save the time that the inner join output iterator
completes, since outer join
// output counts as both update and removal time.
var innerOutputCompletionTimeNs: Long = 0
def onInnerOutputCompletion = {
innerOutputCompletionTimeNs = System.nanoTime
}
- val filteredInnerOutputIter = CompletionIterator[InternalRow,
Iterator[InternalRow]](
- (leftOutputIter ++ rightOutputIter).filter(outputFilterFunction),
onInnerOutputCompletion)
+ val innerOutputIter = CompletionIterator[InternalRow,
Iterator[InternalRow]](
--- End diff --
nit: can you add a description of what is "inner output". the code is
getting more complex, so I think better to add more docs.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
