consider exposing the Executor from a CoroutineDispatcher.

[hsyed]

In various contexts (--e.g., dagger2 producers, various guava async utils) the Executor could be the same as the coroutine dispatcher. Could we get access to the executor so that we can use it directly at least on CommonPool and user created ThreadPoolDispatcher. Having to inject both or provide both is annoying.

[qwwdfsad]

Why executor.asCoroutineDispatcher() is not applicable for this purpose?

[hsyed]

Sorry for the delay. That is going from executor to dispatcher. I think going from dispatcher (CommonPool, ThreadPoolDispatcher) back should be possible as well.

Its quite common to be in a coroutine and have to interact with something expecting an executor and that the executor should be the same pool. If the graph is wired up through di you would have to bind an underlying executor as well as a dispatcher.

If you only bind and executor and go from to executor to dispatcher thats extra allocation each time no ?

[qwwdfsad]

Hi @hsyed
In general, going back is not always possible.
E.g. in CommonPool underlying pool will be soon replaced by our own hand-rolled thread pool which may be not an instance of Executor.

One of the solutions is to provide access to the underlying executor for executor-based dispatchers (asCoroutineDispatcher, newFixedThreadPoolContext, newSingleThreadContext).
But going from executor to dispatcher will always cost you an allocation if you have only an executor. Will it be enough?

[hsyed]

@qwwdfsad Yes that is what I was thinking of.