Mypy blows up sometimes with bad reverse method

[sid-kap]

When I run mypy (no special arguments) on the following code (with --show-traceback):

class Foo:
    def __radd__(self, other: Foo, other2: Foo = None) -> Foo:
        return self

it fails with this error:

test_radd.py:2: error: INTERNAL ERROR -- please report a bug at https://github.com/python/mypy/issues version: 0.610
Traceback (most recent call last):
  ...
  ...
  File "/Users/sidharthkapur/.local/share/virtualenvs/mypy-test-VQzn8vfh/lib/python3.7/site-packages/mypy/checker.py", line 957, in check_reverse_op_method
    assert len(reverse_type.arg_types) == 2
AssertionError: 
test_radd.py:2: : note: use --pdb to drop into pdb

The expected behavior should be the same as in the following, similar example:

class Foo:
    def __radd__(self, other: Foo, other2: Foo) -> Foo:
        return self

which gracefully fails with the message:

test_radd.py:2: error: Invalid signature "def (test_radd.Foo, test_radd.Foo, test_radd.Foo) -> test_radd.Foo"

• Python version: 3.7.0
• Mypy version: 0.610

[gvanrossum]

Heh, that sounds like a pretty simple bug. We clearly should use a more sophisticated check than checking for exactly 2 arguments. Maybe you can give it a try yourself?

[sid-kap]

Sure, I can try!

[sid-kap]

After thinking about it some more, I think this is actually a false positive; having a default argument should actually be ok. (Rereading what you wrote above, I think you already realized this.) Running the following code works:

class Foo:
    def __radd__(self, other: object, other2: object = None) -> 'Foo':
        return self

f = Foo()
g = 1 + f

Would

assert is_subtype(reverse_type, method_type)

make sense as a better check here?