Get Your Feet Wet With Java Parallel Streams

Published: Oct 28, 2021 • Updated: Nov 1, 2021

Parallelism in Java can be spooky, but it doesn’t have to be. Let’s see how streams look when they’re not sequential.

Run it #

These samples are also available in src/ You can run them with:

javac && java App

Boilerplate #

To reduce boilerplate, the following functions will be used. You can see their definitions in src/

Sequential Streams #

Let’s start with a list of numbers:

static List<Integer> createNumbers() {
  return Arrays.asList(1, 2, 3);

Then let’s simulate a costly mapper. We’ll log the current thread name, sleep for 1 second, then multiply the arg by 1 just for kicks.

static Integer costlyMapper(Integer i) {
  log("costlyMapper | val: %s | thread: %s", i, Thread.currentThread().getName());
  return i * 1;

Then let’s use a stream to call our costly mapper and get the sum of the numbers. A stream is sequential by default, I just tacked on sequential() here to make it explicit.

static void sumSequential() {
  int sum = createNumbers()
  log("Sum: %s", sum);

What do you think will happen when timing the function?

time("sumSequential", App::sumSequential);

Which thread(s) will be used? How long will it take?

Note that:

Started sumSequential
costlyMapper | val: 1 | thread: main
costlyMapper | val: 2 | thread: main
costlyMapper | val: 3 | thread: main
Sum: 6
Completed sumSequential in 3.01 second(s)

Parallel Streams #

Okay, now onto the main event. This is what you came for, right?

Let’s sum the numbers again, but this time, in parallel:

static void sumParallel() {
  int sum = createNumbers()
    log("Sum: %s", sum);

Then time it:

time("sumParallel", App::sumParallel);

Again: Which thread(s) will be used? How long will it take?

Note that:

Started sumParallel
costlyMapper | val: 2 | thread: main
costlyMapper | val: 3 | thread: ForkJoinPool.commonPool-worker-5
costlyMapper | val: 1 | thread: ForkJoinPool.commonPool-worker-19
Sum: 6
Completed sumParallel in 1.01 second(s)

There are 3 elements in the list, so 3 threads were used during the parallel run.

Let’s compare the available processors on my current machine to the available threads for parallelism.


This outputs:

Available processors: 12

So, this means there must be 12 available threads, right?


The answer may surprise you:

Common pool parallelism: 11

11 threads? What?! Did they short us? Nope. The missing thread is actually the main thread.

I hope this little exploration gave you more confidence with parallel streams. See the below references for deeper dives as well as Do’s and Don’ts. Happy computing.

References #