Github user advancedxy commented on a diff in the pull request:
https://github.com/apache/spark/pull/4002#discussion_r22838579
--- Diff:
core/src/main/scala/org/apache/spark/rdd/ParallelCollectionRDD.scala ---
@@ -127,18 +127,15 @@ private object ParallelCollectionRDD {
})
}
seq match {
- case r: Range.Inclusive => {
- val sign = if (r.step < 0) {
- -1
- } else {
- 1
- }
- slice(new Range(
- r.start, r.end + sign, r.step).asInstanceOf[Seq[T]], numSlices)
- }
case r: Range => {
- positions(r.length, numSlices).map({
- case (start, end) =>
+ // 1 to Int.MaxValue and (-2 to Int.MinValue by -1) can trigger
exclusive range int overflow
+ val needsInclusiveRange = r.isInclusive && (r.end == Int.MaxValue
|| r.end == Int.MinValue)
+ positions(r.length, numSlices).zipWithIndex.map({
+ // we need an inclusive range to avoid int overflow and setting
the last range to be
+ // inclusive is suffice
+ case ((start, end), index) if needsInclusiveRange && index ==
numSlices - 1 =>
+ new Range.Inclusive(r.start + start * r.step, r.end, r.step)
+ case ((start, end), _) =>
new Range(r.start + start * r.step, r.start + end * r.step,
r.step)
--- End diff --
No doubt that your version is more straightforward than mine. When I wrote
my code, I didn't consider splitting normal inclusive range using inclusive
range. However the benefit of my implementation is that the splitting result
will be same as in the master for normal inclusive ranges. I wonder there may
be some spark code rely on the exclusive range output. And of course, I think
we should update the corresponding document for this kind of change.
I will covert the pattern matching to one case and update the
implementation when we decided which one fits better.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
