Typing in our ecosystem and with docstub

[lagru]

As has already been mentioned, I'd like to take at least some time to talk with people about using type annotations in our ecosystem in general and with docstub specifically.

We've been developing docstub to generate stub files and type annotations for scikit-image. It's still in early development. But while it's shaping up, I'm very curious to explore its usefulness to other projects.

Questions that I'm curious about:

• Do other project want to adopt type annotations? How much effort do you want to spend on this?
• Have you tried docstub already? If so, is there something critical missing? Does it make assumptions that don't work well with your project?
• If you use NumPyDoc-style docstrings, what kind of type description do you use? Are you married to any conventions or would you be willing to standardize more with the ecosystem?
• Current state when it comes to typing NumPy arrays and other challenges with fitting Python's type system to our conventions and patterns.
  • (For example scikit-image is struggling with Type checking only typed input.)

If you have an interest in any of this, please feel free to add here, elsewhere and approach me at the summit! 😊