If I have a `java.util.concurrent.ThreadPoolExecutor` with a keepAlive of 0, will it respect that `k...

# coroutines

[simon.vergauwen]

If I have a java.util.concurrent.ThreadPoolExecutor with a keepAlive of 0, will it respect that keepAlive as a CoroutineContext?

[Vsevolod Tolstopyatov [JB]]

do you mean yourExecutor.asCoroutineDispatcher?

[simon.vergauwen]

Yes, below is the full snippet.

class CountingThreadFactory(val name: String) : ThreadFactory {
  private val counter = AtomicInteger()
  override fun newThread(r: Runnable): Thread =
    Thread(r, "$name-${counter.getAndIncrement()}")
}
// Creates a ExecutorService that uses every thread only once, so every task is scheduled on a differently numbered Thread.
fun newTestingScheduler(name: String): ExecutorService =
  ThreadPoolExecutor(0, 2, 0, TimeUnit.MILLISECONDS, SynchronousQueue<Runnable>(), CountingThreadFactory(name))

newTestingScheduler("test").asCoroutineDispatcher()

[simon.vergauwen]

Alternative is a mechanism of locks to test if functions get executed in parallel. Other alternatives (or resources I could check out) would also be very appreciated! 🙂

[Vsevolod Tolstopyatov [JB]]

Yes, asCoroutineDispatcher respects keepAlive.

Alternative is a mechanism of locks to test if functions get executed in parallel

keepAlive is a racy mechanism for that.
E.g. you can have a function that supposed to launch two coroutines (and you want to test they are executed in parallel).
What could happen: the first coroutine is launched and executed, the caller is preempted by the OS. The first coroutine finishes its execution, the same time the caller launches the second coroutine and it is immediately taken by the first thread.
In the end, your counter is still equal to one, though your code is completely correct.

[Vsevolod Tolstopyatov [JB]]

One of the options is to write stress tests (though they are still probabilistic).
Another one is to use locks or a special dispatcher that keeps track of submitted coroutines in an application/test specific way

[simon.vergauwen]

Alright, thanks for the info!