If there is time, we could have a discussion about the state of the typing spec and what to prioritize next. I’m happy to prepare some materials and lead this discussion.
We’ve made good progress over the past year in creating a typing spec, filling in missing chapters, and clarifying ambiguities, but there is still much left to do. I’ve been maintaining a long “to do” list. I’d love to get feedback from community members about where to focus next — and perhaps even sign up some volunteers to write these sections.
Oh and I should have mentioned - I’m planning to record this, we can discuss in the meetup the best way to share, assuming everyone is comfortable with that.
I’m wondering whether eventually it would make sense to livestream it (that’s what the PyTorch OSS community does with their meetups).
Very edifying to attend, thank you to the speakers/hosts. I hope in the near future to contribute where I have some relevant experience in the areas @erictraut highlighted.
Also, @adriangb I realized my account on here is too new for messaging. If you’d like to follow up briefly out-of-band about Annotated vis-a-vis weird Pydantic use cases, my public-facing/professional email is dlukes@ucsd.edu.
The resolution is a little low, I’ll have to look more at my settings, but it should be usable plus we can share the slides separately (thanks for posting yours Eric!)
Thanks for the recording! I couldn’t make it live, but listening to it (and occasionally having to look at the screen on a few specifics) was nice to hear more about 746, as well as typing priorities.
As someone who has pushed work around intersections a bit and who wants some level of this to work for python, I agree with @erictraut 's assessment about a low ROI for intersections and HKTs. Right now, I think the most value that’s come from inspecting either of these has been finding places where we have a shaky existing ground to build upon, but even that has not been the most productive process in turning that information into things that can be more immediately addressed.
If there’s interest from other type checkers about overload resolution (as was stated as a condition to that being worth working on next), this is something I’ve already been exploring as part of other work, and can share what notes I already have, but I don’t want to drive the work on this one.