zhengruifeng commented on a change in pull request #27947:
[SPARK-31182][CORE][ML] PairRDD support aggregateByKeyWithinPartitions
URL: https://github.com/apache/spark/pull/27947#discussion_r394203694
##########
File path: mllib/src/main/scala/org/apache/spark/ml/feature/RobustScaler.scala
##########
@@ -204,17 +204,18 @@ object RobustScaler extends
DefaultParamsReadable[RobustScaler] {
}.reduceByKey { case (s1, s2) => s1.merge(s2) }
} else {
val scale =
math.max(math.ceil(math.sqrt(vectors.getNumPartitions)).toInt, 2)
- vectors.mapPartitionsWithIndex { case (pid, iter) =>
- val p = pid % scale
- iter.flatMap { vec =>
- Iterator.tabulate(numFeatures)(i => ((p, i), vec(i)))
- }.filter(!_._2.isNaN)
- }.aggregateByKey(
+ vectors.flatMap { vec =>
+ vec.iterator.filterNot(_._2.isNaN)
+ }.aggregateByKeyWithinPartitions(
new QuantileSummaries(QuantileSummaries.defaultCompressThreshold,
relativeError))(
seqOp = (s, v) => s.insert(v),
combOp = (s1, s2) => s1.compress.merge(s2.compress)
- ).map { case ((_, i), s) => (i, s)
- }.reduceByKey { case (s1, s2) => s1.compress.merge(s2.compress) }
+ ).mapPartitionsWithIndex { case (pid, iter) =>
+ val p = pid % scale
+ iter.map { case (col, s) => ((p, col), s.compress) }
Review comment:
here we can trigger compression at the _map_ side
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
