Scala Cookbook

13.12. Using Parallel Collections

Problem

You want to improve the performance of algorithms by using parallel collections.

Solution

When creating a collection, use one of the Scala’s parallel collection classes, or convert an existing collection to a parallel collection. In either case, test your algorithm to make sure you see the benefit you’re expecting.

You can convert an existing collection to a parallel collection. To demonstrate this, first create a sequential collection, such as a Vector:

scala> val v = Vector.range(0, 10)
v: scala.collection.immutable.Vector[Int] = Vector(0, 1, 2, 3, 4, 5, 6, 7, 8, 9)

Next, print the sequence, and you’ll see that it prints as usual:

scala> v.foreach(print)
0123456789

As expected, that example prints the string 0123456789. No matter how many times you print it, you’ll always see that same result; that’s the linear world you’re used to.

Next, call the par method on your collection to turn it into a parallel collection, and repeat the experiment:

scala> v.par.foreach(print)
5678901234

scala> v.par.foreach(print)
0123456789

scala> v.par.foreach{ e => print(e); Thread.sleep(50) }
0516273894

Whoa. Sometimes the collection prints in order, other times it prints in a seemingly random order. That’s because it’s now using an algorithm that runs concurrently. Welcome to the brave, new, parallel world.

That example showed how to convert a “normal” collection to a parallel collection. You can also create a parallel collection directly:

scala> import scala.collection.parallel.immutable.ParVector ...