[GitHub] spark pull request #14140: [SPARK-16426][MLlib] Fix bug that caused NaNs in ...

[srowen <gi...@git.apache.org>]

Github user srowen commented on a diff in the pull request:

    https://github.com/apache/spark/pull/14140#discussion_r70429952
  
    --- Diff: mllib/src/main/scala/org/apache/spark/mllib/regression/IsotonicRegression.scala ---
    @@ -408,8 +409,12 @@ class IsotonicRegression private (private var isotonic: Boolean) extends Seriali
        */
       private def parallelPoolAdjacentViolators(
           input: RDD[(Double, Double, Double)]): Array[(Double, Double, Double)] = {
    -    val parallelStepResult = input
    -      .sortBy(x => (x._2, x._1))
    +    val keyedInput = input
    --- End diff --
    
    I think there may be shorter ways to write this with `groupBy`, but, this and other approaches like that have the big drawback of reading lots of data into memory. Here you have to sort the whole partition in memory (!).
    
    How about `repartitionAndSortWithinPartitions`? oddly specific method, but, likely just what you need here, to both partition according to some criteria but then end up with sorted partitions. It's more scalable.


---
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 infrastructure@apache.org or file a JIRA ticket
with INFRA.
---

---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org