[SPARK-40175][SQL]Speed up conversion of Tuple2 to Scala Map #37609
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
caican00
changed the title
Contributor
Author
|
gently ping @cloud-fan @srowen |
Contributor
|
Nice try! Do we have some perf numbers? |
|
Can one of the admins verify this patch? |
Contributor
Author
@cloud-fan Thanks for your reply. I don't have any perf numbers right now but I'm going to do some performance tests to provide some perf numbers. |
Member
|
breakOut was removed or something in scala 2.13, I think? https://www.scala-lang.org/blog/2017/02/28/collections-rework.html |
Contributor
Yes, Scala 2.13 already build failed seems as follows:
|
Contributor
|
The idea is still valid: we can write a while loop manually to build the map, instead of |
Contributor
|
I think we can test the performance of this api in Scala 2.13 to check if there is targeted optimization in Scala 2.13 |
Contributor
Checked, Scala 2.13 has the same problem. @caican00 Does |
Contributor
31 file has |
Contributor
|
Are you still interested in this pr @caican00 |
Contributor
@cloud-fan Do you mean like follows? private def zipToMapUseMapBuilder[A, B, K, V](keys: Seq[A], values: Seq[B]): Map[K, V] = {
import scala.collection.immutable
val builder = immutable.Map.newBuilder[K, V]
val keyIter = keys.iterator
val valueIter = values.iterator
while (keyIter.hasNext && valueIter.hasNext) {
builder += (keyIter.next(), valueIter.next()).asInstanceOf[(K, V)]
}
builder.result()
}
private def zipToMapUseMap[A, B, K, V](keys: Seq[A], values: Seq[B]): Map[K, V] = {
var elems: Map[K, V] = Map.empty[K, V]
val keyIter = keys.iterator
val valueIter = values.iterator
while (keyIter.hasNext && valueIter.hasNext) {
elems += (keyIter.next().asInstanceOf[K] -> valueIter.next().asInstanceOf[V])
}
elems
}I write a microben to compare From the results, the performance of |
Contributor
|
interesting, we need to understand why |
Contributor
|
I seem to know the reason, |
Contributor
|
To check |
Contributor
Add results of input size 150, 200, 300, 400. @cloud-fan , from bench results:
|
Contributor
|
Great! Seems we can always do |
Contributor
Yes, there should be many similar case and I think this should be a valuable optimization However, before I doing something, I still want to ask @caican00 , will you continue this optimization? |
Contributor
|
start this work : #37876 |
Member
|
OK to continue the work here after a rebase? #37876 is merged |
Contributor
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
What changes were proposed in this pull request?
Traversable.toMapchanged tocollections.breakOut, that eliminates intermediate tuple collection creation.I optimized it with reference to this pr:#18693
An introduction to
Collections. BreakOutcan be found at Stack Overflow article.Why are the changes needed?
When
DeserializeToObjectis executed, converting Tuple2 to Scala Map via. ToMaptakes a lot of cpu time.How was this patch tested?
Unit tests run.
No performance tests performed yet.