Classes that implement getitem are reported as incompatible with enumerate

[scuac]

Bug Report

Mypy reports erroneously that a class that implements getitem is not suitable to be used with enumerate.

To Reproduce

Small example:

class A:
    def __init__(self):
        self.data = [1, 2, 3]

    def __len__(self):
        return len(self.data)

    def __getitem__(self, item):
        return self.data[item]


a = A()
for i, item in enumerate(a):
    print(f'item {i} = {item}')

Expected Behavior

Running mypy on the above code should report no errors

Actual Behavior

Mypy reported:

error: Argument 1 to "enumerate" has incompatible type "A"; expected "Iterable[<nothing>]"

Your Environment

• Mypy version used: 0.790
• Mypy command-line flags: none
• Mypy configuration options from mypy.ini (and other config files):

[mypy]
ignore_missing_imports = True

• Python version used: 3.7.6
• Operating system and version: macOS 10.15.7

[JelleZijlstra]

This is because we don't fully understand the "old-style" iteration protocol, where you make an iterable with just __getitem__ and __len__. I believe this has been discussed before but I'm not sure where.

[JelleZijlstra]

An easy fix is to add def __iter__(self): return self.

[scuac]

An easy fix is to add def __iter__(self): return self.

That doesn't work. It does make the mypy error go away but the code fails with
TypeError: iter() returned non-iterator of type 'A'

For now the only work-around I found it to append a # type: ignore to the line with enumerate. But that is cumbersome since it needs to be done every time this class is used in a loop.

[JelleZijlstra]

Maybe you can put the def __iter__ inside an if TYPE_CHECKING: block.

[scuac]

This is because we don't fully understand the "old-style" iteration protocol, where you make an iterable with just __getitem__ and __len__. I believe this has been discussed before but I'm not sure where.

From what I could find, it seems the protocol is like this:
IF __iter__ is not present, fallback to iterating by repeatedly calling __getitem__ with integers starting from 0 (i.e. 0, 1, 2, ...) and stop when receiving an IndexError.

So, the requirement for this to work is that the __getitem__ method has to work with integers as input and raise IndexError, which is the case in the example I gave. That being said, it seems it would be too complicated for mypy to figure that out without assistance, so I guess that it raising the error is correct.

I tried updating the __getitem__ in the example to make all of this explicit, but still get the error:

    def __getitem__(self, index: int):
        if index < 0 or index >= len(self.data):
            raise IndexError
        return self.data[index]

[scuac]

Maybe you can put the def __iter__ inside an if TYPE_CHECKING: block.

That works, and I like it better than the comment since it is in the class and doesn't need to be put in every loop.

[scuac]

Well, it seems that extending a class similar to my example is quite easy after all, simply add:

    def __iter__(self):
        for item in self.data:
            yield item

And everything works.

[hauntsaninja]

You could do return iter(self.data) to save yourself a line :-)