PEP 484: Allow annotating first argument of instance and class methods - #89

@gvanrossum I implemented your comments in the new commit.

This looks good to me, and I believe that it won't be very hard to get this working in mypy. But I haven't actually done more than thought experiments.

Let's wait until we've prototyped this successfully in mypy.

In python/mypy#2193 (comment), @elazarg expresses some doubts about the feasibility of this approach. I haven't thoroughly ingested his comments yet but I thought I'd call it out here.

I didn't mean to question the feasibility of anything. I only say we need to distinguish between

• Mixins that magically copy / create instances of their dynamic types, and
• Interfaces that require overriding methods to return an instance of their static type.

These are two different things, and need different syntax.

I also think the syntax suggested here is more appropriate for the interface thing, while the examples in this PEP change seems to be about mixins. It is of course somewhat a matter of taste, but I think it is not entirely subjective.

@gvanrossum @elazarg To be honest, I still could not clearly see the difference between the two use cases (probably it is to much blurred for me). But what I could say is in my opinion both examples in the PR (and two examples Guido gave in mypy tracker) if what we are interested in. Also I think the syntax should be self: T, at least for me it is clear what it means (it more-or-less simply allows generic function semantics for methods).

@gvanrossum Since support for this feature (at least the functionality described in this PR) is now in mypy python/mypy#2193 , I think this PR could be merged.