Do not emit unreachable warnings for lines that return NotImplemented.

[tyralla]

I think no one has complained so far. I just encountered this (in my understanding) lack in TypeChecker.is_noop_for_reachability working on #20068.

[github-actions]

Diff from mypy_primer, showing the effect of this PR on open source code:

yarl (https://github.com/aio-libs/yarl)
+ yarl/_url.py:550: error: Unused "type: ignore" comment  [unused-ignore]

[A5rocks]

I think this makes sense to me, even if it complicates the general idea of what exactly is exempt from unreachability warnings. (I guess now it's "anyway to bail out is OK"?)

I can also see this being unnecessary :^). Given nobody has complained, I'm not sure we really need to add this as a feature (since that's ultimately what this is).

[sterliakov]

I think this is a good idea. See #363 (yes, three-digit issue number!) - this removes an unreachable warning from the "should be fine" snippet there. I guess it has never been explicitly reported only because --warn-unreachable is hidden too well, many users aren't even aware that it exists.

[tyralla]

I think this is a good idea. See #363 (yes, three-digit issue number!) - this removes an unreachable warning from the "should be fine" snippet there.

Wow, I overlooked this dinosaur. That's a nice tool you have!

Interpreting NotImplemented as Any seems not to be the only problem. I just played around a little. Mypy does not predict the following crash:

class A1:
    x = 1
    def __add__(self, o: A1) -> float:
        if isinstance(o, A1):
            return self.x + o.x
        return NotImplemented

class A2:
    x = 2
    def __add__(self, o: A2) -> float:
        return self.x + o.x

class B:
    y = 3
    def __radd__(self, o: A1 | A2) -> float:
        return self.y + o.x

A1() + B()  # works
A2() + B()  # fails  (undetected by Mypy)

I will try to find a fix for this as soon as I have some spare time.

I guess it has never been explicitly reported only because --warn-unreachable is hidden too well, many users aren't even aware that it exists.

Yes, I agree. As soon as --warn-unreachable will be included in --strict, it will likely be discussed more frequently.

I think this makes sense to me, even if it complicates the general idea of what exactly is exempt from unreachability warnings. (I
I can also see this being unnecessary :^). Given nobody has complained, I'm not sure we really need to add this as a feature (since that's ultimately what this is).

It would say it is more of a fix than a feature, because return NotImplemented is reachable, and so the current warning is wrong.

[tyralla]

Thanks for the reviews!