[jorenham]

This increase type coverage by filling in the missing annotations for the public-facing symbols in the submodules of zmq.backend.cffi.

According to pyrefly coverage check --public-only (and excluding zmq/tests), this increases type coverage by 7.65%, for a grand total of 93.17% ( 🎉 ).
The remanining untyped typables live in devices, eventloop, green, log, and sugar. My plan is to have a go at those next, with that oh-so-sweet 100% type coverage as (intermediate) life goal (see https://jorenham.github.io/typestats/dashboard/report/#pyzmq for details and some fancy pie charts).

No behavior changes (modulo some trivial asserts for type narroing purposes to appease mypy).

Oh and because Python 3.9 is still supported I had to use __future__.annotatiions imports. So we can get rid of those once py39 support is dropped (it'd require stringifying some type-check-only annotations though, such as Buffer).

---

No AI touched the code. I only used AI to review the annotations that I wrote myself _{and I'm pretty sure I'm human}.

[jorenham]

CI could use a rerun

[jorenham]

Any | None isn't valid or necessary

That's not quite true, as explained in the typing spec (https://typing.python.org/en/latest/spec/concepts.html#union-types):

The union T | Any is not reducible to a simpler form. It represents an unknown static type with lower bound T. That is, it represents an unknown set of objects which may be as large as object, or as small as T, but no smaller.

[minrk]

In this particular case, I think it's fine, because it doesn't reduce it at all or narrow anything. It's also invalid syntax on 3.9, causing the cffi backend to fail to import. I realize adding from __future__ import annotation would also have done it.

[minrk]

Thanks so much for your contributions and patience!

[jorenham]

Thanks so much for your contributions and patience!

Happy to help :)