Question

It seems like when I invoke map on a parallel list, the operation runs in parallel, but when I do filter on that list, the operation runs strictly sequentially. So to make filter parallel, I first do map to (A,Boolean), then filter those tuples, and map all back again. It feels not very convenient.

So I am interested – which operations on parallel collections are parallelized and which are not?

Answer 1

There are no parallel lists. Calling par on a List converts the List into the default parallel immutable sequence – a ParVector. This conversion proceeds sequentially. Both the filter and the map should then be parallel.

scala> import scala.collection._
import scala.collection._

scala> List(1, 2, 3).par.filter { x => println(Thread.currentThread); x > 0 }
Thread[ForkJoinPool-1-worker-5,5,main]
Thread[ForkJoinPool-1-worker-3,5,main]
Thread[ForkJoinPool-1-worker-0,5,main]
res0: scala.collection.parallel.immutable.ParSeq[Int] = ParVector(1, 2, 3)

Perhaps you’ve concluded that the filter is not parallel, because you’ve measured both the conversion time and the filter time.

Some operations not parallelized currently: sort* variants, indexOfSlice.