How to configure a disconnect policy for Scala ExecutionContexts?

Question

Recently, I came across some odd behavior on a machine where the mapfunction returning Future[T]was executed sequentially. The same problem does not occur on other machines: the work alternates, as one would expect. Later, I discovered that this was probably because Scala was too smart and chose ExecutionContextwhich corresponded to the resources of the machine: one core, one working.

Here is some simple code that reproduces the problem:

import scala.concurrent._
import scala.concurrent.duration._

val ss = List("the", "quick", "brown", "fox",
              "jumped", "over", "the", "lazy", "dog")

def r(s: String) : Future[String] = future {
        for (i <- 1 to 10) yield {
                println(String.format("r(%s) waiting %s more seconds...",
                        s, (10 - i).toString))
                Thread.sleep(1000)

        }
        s.reverse
}

val f_revs = ss.map { r(_) }

println("Look ma, no blocking!")

val rev = f_revs.par.map { Await.result(_, Duration.Inf) }.mkString(" ")

println(rev)

Doing this on a machine with unusual behavior causes consistent output as follows:

Look ma, no blocking!
r(the) waiting 9 more seconds...
r(the) waiting 8 more seconds...
r(the) waiting 7 more seconds...

Custom provision ExecutionContext:

val pool = Executors.newFixedThreadPool(1)
implicit val ec = ExecutionContext.fromExecutor(pool)

. : , . -, FixedThreadPool s, , pool.shutdown() -.

, . , (, ), , ? ExecutionContext, , .

Answer 1

, Java 7 ExecutorService s, , ForkJoinPool, , (. shutdown() ).

, :

val pool = new java.util.concurrent.ForkJoinPool(5)

Java 8, -, .

Answer 2

Scala fork-join, , Java, ExecutionContext , Executors.

, , - ExecutionContext:

• scala.concurrent.context.minThreads, . 1.
• scala.concurrent.context.numThreads, . x1.
• scala.concurrent.context.maxThreads, . x1.

, , x, . , numThreads, maxThreads. x2 .