`asyncio.wait` should not check for emptiness of iterable by calling its `__bool__` method

[jonathandung]

Bug report

Bug description:

asyncio.wait is documented to accept an iterable of non-coroutine awaitable objects in its docstring:

The fs iterable must not be empty.

and the python docs page:

The aws iterable must not be empty.

We notice a discrepancy: The parameter is called fs in asyncio.tasks:

async def wait(fs, *, timeout=None, return_when=ALL_COMPLETED)

but aws in the docs. It should be aws for 'awaitables'.

Besides, the following line exists in its source:

if not fs:
    raise ValueError('Set of Tasks/Futures is empty.')

This may be error-prone for iterables implementing a __bool__ method with different semantics. This check should either be documented or abolished.

See the source and the docs.

CPython versions tested on:

CPython main branch

Operating systems tested on:

No response

[sergey-miryanov]

cc @kumaraditya303

[picnixz]

This may be error-prone for iterables implementing a bool method with different semantics. This check should either be documented or abolished.

This is not our concern honestly. Iterables falsey are meant to be considered empty. And we do not document those kind of checks as they are impl. details.

We can fortify the check by using next(iter(fs), None) checks instead though.

cc @kumaraditya303

[kumaraditya303]

We can fortify the check by using next(iter(fs), None) checks instead though.

I don't think that would work for generators which is accepted by this API.

FWIW, I don't think there is any real issue here, there is a general assumption that empty sequences are falsy and if someone does something different there are probably lot of cases which wouldn't work.

[picnixz]

I don't think that would work for generators which is accepted by this API.

Do you mean it's because it would consume the first item of the generator and that would not necessarily be desirable to consume at that specific point? Because afterwards, there is the call fs = set(fs) so you anyway consume what remains.

More generally, if we really want something more robust, we can just move the emptiness check after fs = set(fs).

[kumaraditya303]

Do you mean it's because it would consume the first item of the generator and that would not necessarily be desirable to consume at that specific point? Because afterwards, there is the call fs = set(fs) so you anyway consume what remains.

Yes, we would have to add that consumed item back.

[jonathandung]

Even if that is so, the other problem I mentioned about the parameter name mismatch persists and is easily fixable. Should I open a PR?