Weekly Roundup 2025-07-08
- Wednesday: Oh no, bidirectional type checking is giving me different error messages when I check malformed syntax trees. Better think about this for over a week!
- Thursday: I came up with a plan/excuse for how emitting error messages is kind of like generating a constraint set.
- Friday: Felt bad, but had some ideas for how to handle these error messages in a more principled way than "it's strings!"
- Saturday: I tried to put together a list of everything that can go wrong for one of these syntax trees.
- Sunday: I took that list and tried to converge the phrasing so I can have half-a-dozen classes/helpers, instead of, like, twenty.
- Monday: "Hm, some of these messages kind of have to glom onto subsequent messages; I need to model that somehow."
Next week, I'm going to take my message drafts and try to put them through a few rounds of prototyping.