single interface for fetch

[DxCx]

since the fetcher shouldn't know if it is expected to return promise or observable,
it added a small logic to convert observable into promises when promise is a must.

[leebyron]

I'm concerned that this will break functionality for subscriptions which use Observables to provide multiple payloads for a single request over time.

Could you clarify what the problem/bug is that you're trying to solve?

[leebyron]

why both conditions above? Are both needed or is only one sufficient?

[DxCx]

Only the second, the first is more for readability.
If you want ill remove it..

[leebyron]

Yeah - only the second is necessary.

[DxCx]

Actually ive written a fetch function which returns observable (and hence, tested it with it) and the issue is that the schema cannot be observable only promise, so, no, its not breaking its just means that the schema returned can be observable which emits one value.

As you know from the other thread im creating a full flow example of subscriptions using observables and this a small issue ive found on the way. (Fetch function expects to return both promise and observables on different places)

[leebyron]

Ah thanks for the extra info. I see you're only wrapping the schema fetching so this looks good

[leebyron]

I think you actually want the first value, a "takeOne" kind of logic, rather than waiting for the last. If fetcher returned some sort of unclosing socket, this could never complete.

Perhaps:

if (isObservable(promiseOrObservable)) {
  return new Promise((resolve, reject) => promiseOrObservable.subscribe(resolve, reject, reject));
}
return promiseOrObservable;

[DxCx]

I dont mind changing it, the reason it is done that way is to be aligned with toPromise() operator which is waiting for last value.
Because of that it is often used with .take(1) operator before it.

Let me know if you want to change the behavior of the function or keep it aligned

[leebyron]

Yeah I think we want a sort of inlined .take(1).toPromise()

[DxCx]

Ok, Gotcha, so in that case unsubscribe should be called as well.
also, reject won't give any helpful information.
I've added a commit with both CR notes. :)

[asiandrummer]

This lgtm as well - thanks for the contribution!