fix types for protocols to add type safety

[rchl]

Fixes types for pyright in cases like below where we've blatantly lied about types.

The way it was done before, apart from causing type issues with Pyright, allowed users to just access session_view.session_buffer from Session, for example, and get away with accessing anything on SessionBuffer because the type was Any.

@jwortmann sorry to cause troubles for you again. I don't mean bad. :)

[rchl]

To avoid changing too much code, decided to use the "underscore" property in __init__ and access through the getter elsewhere. This is maybe not super consistent but makes the diff smaller.

Let me know what you prefer. Maybe just assign to _view and use the getter from there on.

[rchl]

Hmm, looks like I've triggered some import loop... Fixed

[rwols]

The = None syntax is the correct syntax for python < 3.6. Because plugin_host uses python 3.3, we cannot annotate properties of classes so this syntax is used. Mypy understands it.

https://mypy.readthedocs.io/en/stable/class_basics.html#instance-and-class-attributes

Type comments work as well, if you need to support Python versions earlier than 3.6:

class A:
    x = None  # type: List[int]  # Declare attribute 'x' of type List[int]

Do we really need accessor methods? It's starting to look a bit enterprisey...

[rwols]

I do appreciate that listener and session_buffer don't have type Any anymore, but the accessor methods are a bit too much I think.

[rchl]

Do we really need accessor methods? It's starting to look a bit enterprisey...

I don't know of another way to do it.

Changing:

session_buffer = None  # type: Any

to

session_buffer = None  # type: SessionViewProtocol

confuses mypy in places where SessionView is passed when SessionViewProtocol is expected so that can't be done either.

[rchl]

I would think that type safety is more important over how it "feels"? And to me it doesn't even feel bad from the *Protocol perspective.

I guess it is a bit weird how it feels in the actual "implementations" of those protocols but it can be seen as a positive thing since the distinction between the private properties and the public interface is made clear that way.