Today's tips is about executing parallel streams in custom thread pools. One of the main drivers for developing the Java 8 streams was their ability to abstract away parallelism. In theory, we could take any stream and make it parallel with the .parallel() method. In reality... not so easy...

One of the caveats with parallel streams is that they, by default, execute on the common Fork Join Pool. So, if there are a lot of things going on in our application, those parallel tasks will compete with all other tasks in the same pool.

This is something that can be fixes easily. A nice thing with all parallel streams is that we can submit the parallel tasks to any Thread Pool, not just the common Thread Pool. In the example below, I am showing how to stream database content in parallel using Speedment, a stream based ORM tool and runtime. This way, complex database operations may execute much faster since the tasks can be spread out on several threads.

However, the same principle would work with any type of stream source such as the ones we get from Java's built-in Collections like List or Set.

    final ForkJoinPool forkJoinPool = new ForkJoinPool(3);
    forkJoinPool.submit(() -> 
        
        users.stream() 
            .parallel()
            .filter(User.EMAIL.isNotEmpty())
            .forEach(slowOperation()); 
            
    );

    try {
        forkJoinPool.shutdown();
        forkJoinPool.awaitTermination(1, TimeUnit.HOURS);
    } catch (InterruptedException ie) {
        ie.printStackTrace();
    } 

This will create a new Fork Join pool with three threads in which the parallel stream will be executed. This way, we limit the number of parallel task to three and they will operate separate from the common Thread Pool. Well, not entirely separate, because tasks in all pools compete for the same limited CPU resources. But that is another story...