matching bytes regular expression against unicode

[matthiaskramm]

The following is valid Python 2:

re.compile(b"foo").match(u"foo")

Should typeshed allow this? Right now, https://github.com/python/typeshed/blob/master/stdlib/2/re.pyi defines compile as:

def compile(pattern: AnyStr, flags: int = ...) -> Pattern[AnyStr]: ...

And match(), in Pattern, is defined in https://github.com/python/typeshed/blob/master/stdlib/2/typing.pyi#L350 as:

class Pattern(Generic[AnyStr]):
  def match(self, string: AnyStr) -> Match[AnyStr]: ...

This disallows mixing bytes/unicode. Is this what we want?

Matching a bytes regexp against unicode is a TypeError in Python 3, so there's a case for not supporting this; however we do have some cases on our side were this is flagged as a type error in valid Python 2 code.

[gvanrossum]

I'm okay with supporting this for PY2.

[JelleZijlstra]

This is a duplicate of #273.