Question out of curiosity. In the JVM world a coro...

# coroutines

[Edoardo Luppi]

Question out of curiosity.
In the JVM world a coroutine can be converted to a future using CoroutineScope.future(...).
Internally future(...) uses a CompletableFutureCoroutine which overrides onCompleted and onCancelled to be notified and complete or cancel the Future.
In the JS world the CoroutineScope.promise(...) function first calls async(...) to get a Deferred value, and then listen for its completion via invokeOnCompletion.
Why this implementation difference and not create a PromiseCoroutine in the same way as CompletableFutureCoroutine?

[Edoardo Luppi]

Is it only because of

promise.asDynamic().deferred = this

which can then be used in Promise<T>.asDeferred()?
Still have to encounter a place where that is useful, but it's just me obviously.

[Sam]

This is an interesting question! I'm wondering if it's just an accident with no deeper explanation behind it. I can only see one other notable difference, which is that promise() appears to allow lazy start, while future() doesn't. But that just makes me more confused, to be honest.

[Sam]

As far as I can tell, either version could happily be implemented via either approach.

[Edoardo Luppi]

I did experiment a little yesterday, trying to find differences. One notable difference is with my own PromiseCoroutine I shaved off 5 kb from the outputted coroutines JS file.

[Edoardo Luppi]

I'd be curious to see if invokeOnCompletion is less performant than simply implementing the two Coroutine functions. It seems more involved for sure, both in logic and on the amount of instantiated objects.

[Edoardo Luppi]

The problem is debugging K/JS coroutine code is still a problematic experience. Sometimes it just doesn't stop, or you can't step correctly.