futures_of returns empty list for dict values input

[bnaul]

Definitely debatable whether this is a bug or not but to me it seems undesirable:

[ins] In [2]: c = distributed.Client()

[nav] In [3]: f = c.submit(sum, [1, 2])

[ins] In [4]: distributed.futures_of([f])
Out[4]: [<Future: finished, type: builtins.int, key: sum-503d2c5687479ce1a33513b097c1af5c>]

[ins] In [5]: distributed.futures_of({"f": f})
Out[5]: [<Future: finished, type: builtins.int, key: sum-503d2c5687479ce1a33513b097c1af5c>]

[ins] In [6]: distributed.futures_of({"f": f}.values())
Out[6]: []

I was going to just make a PR but I'm actually not sure what the best way to test that something is a dict view......? If there's not a super clean way of doing it then maybe it's not worth the complexity. But in that case it would be nice if we could raise an exception here instead of incorrectly returning [].

[mrocklin]

Hrm, yeah, that feels like a borderline situation. futures_of({...}.values()) seems like a reasonable thing to expect to work. On the other hand, this could get out of hand if we include all possible builtin sequence types.

I don't mind having this added though if it would help. You're probably not the only person to run across this.

[jakirkham]

Maybe something like this would be reasonable?

In [1]: from collections.abc import Iterable                                    

In [2]: d = {"b": 2}                                                            

In [3]: isinstance(d.values(), Iterable)                                        
Out[3]: True

[bnaul]

That was my first thought, but I think that includes str which you'd want to skip... gets messy pretty quickly.

…

On Mon, Jul 6, 2020, 4:59 PM jakirkham ***@***.***> wrote: Maybe something like this would be reasonable? In [1]: from collections.abc import Iterable In [2]: d = {"b": 2} In [3]: isinstance(d.values(), Iterable) Out[3]: True — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#3909 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAG4TZZNVELJKVSZOA3CJGTR2JQOHANCNFSM4N77VQBA> .

[jakirkham]

I suppose that wouldn't have any Futures though 😉

Yeah I can see why this starts to get tricky to do well. Is there an ABC that captures this better? 😉