Is it discouraged using Java 8 parallel streams inside a Java EE container?

Question

Given that spawning threads in Java EE containers are discouraged. Would using the Java 8 parallel streams, which may spawn threads, inside Java EE be discouraged too?

Answer 1

A heads up, the graceful degradation to single thread is not available. I also thought it was because of Shorn's answer and that mailing list discussion, but I found out it wasn't while researching for this question. The mechanism is not in the Java EE 7 spec and it's not in glassfish 4.1. Even if another container does it, it won't be portable.

You can test this by calling the following method:

@Singleton
public class SomeSingleton {
    public void fireStream() {
        IntStream.range(0, 32)
            .parallel()
            .mapToObj(i -> String.format("Task %d on thread %s", 
                i, Thread.currentThread().getName()))
            .forEach(System.out::println);
    }
}

And you'll get something like:

Info:   Task 20 on thread http-listener-1(4)
Info:   Task 10 on thread ForkJoinPool.commonPool-worker-3
Info:   Task 28 on thread ForkJoinPool.commonPool-worker-0
...

I've also checked glassfish 4.1.1 source code, and there isn't a single use of ForkJoinPool, ForkJoinWorkerThreadFactory or ForkJoinWorkerThread.

The mechanism could be added to EE 8, since many frameworks will leverage jdk8 features, but I don't know if it's part of the spec.

Answer 2

EDIT See alternate answer from andrepnh. The below may have been the plan, but it doesn't appear to have played out that way in practice.

---

The way I read it from the lambda-dev mailing list discussion mentioned in the comments: it's not discouraged the way spawning threads is - but won't do anything much for you in a Java EE context.

From the linked discussion:

the Java EE concurrency folks had been already talked through this, and the current outcome is that FJP will gracefully degrade to single-threaded (even caller-context) execution when running from within the EE container

So you're able to safely use parallel streams in a procedure or library that runs in both contexts. When it's run in a SE environment, it will make with the magical parallel shenanigans - but when it's run in an EE environment it will gracefully degrade to serial execution.

Note: the phrase quoted above is future tense - does anyone have a citation for some definitive documentation?