make list_iterator able to be converted to bool

[ConnorTippets]

Feature or enhancement

Proposal:

Currently, if you cast an empty list iterator (iter([])) to a bool, you get True

> bool(iter([]))
True

This is opposed to casting an empty list ([]) to a bool, where you get the much more reasonable result of False

> bool([])
False

I propose properly implementing boolean conversion for list_iterator, to prevent potentially confusing and unintuitive bugs.

I had a bit of a look under the hood, and I believe implementing tp_as_sequence in the _PyListIterObject class in /Include/internal/pycore_list.h on lines 69-73 would resolve this issue.

Links to previous discussion of this feature:

No response

[sirosen]

Please understand that the requirement for a prior discussion exists in order to manage the project in a sustainable way.

The devguide for CPython is fairly clear about this expectation.

[ConnorTippets]

Whoops, sorry! Haven't done this before, and I mostly just read the page about the issue tracker and not much else.

[terryjreedy]

There is nothing special about listiterators: any iterator can be empty (if finite) and in general the only way to know is to call next(iterator) and see whether it return something or raises StopIteration. bool(any_iterator) is always True by intention. Other examples.

>>> bool(iter(()))
True
>>> bool(iter({}))
True
>>> it = iter([1])   # it is a listiterator
>>> next(it)
1
>>> bool(it)
True

Some and only some iterators have a length_hint. it.__length_hint__() == 0. This is likely to be accurate for the other examples above but this is not guaranteed to be accurate. Hence the 'hint' suffix. It was intentional to not restrict the function to cases where it would be accurate. It would be confusing to have bool(empty_iterator) sometimes be True and sometimes False.

Since the currently behavior is intentional after previous discussion, we might close this.

[sirosen]

Yep, I'm closing! (I'm still learning the norms for triaging. 😅 )

@ConnorTippets, no harm done in opening an issue. But new features and changes in behavior are pretty hard to justify -- it's best to start with the Ideas discussion, and be ready for people to surprise you a bit. That way, the issue tracker stays limited to more active work.