News for hackers. Stuff that matters.
Re:GADTs for Dummies by AnthonBerg on Wednesday May 19, 2021 @05:36AM (#27206341)
Useful take, thanks!

I assume it varies, but I find Idris' dependent type system to be easier to reason about and understand than non-dependent ones I've worked with (say, OCaml and Haskell). Even if Idris can be judged as strictly more powerful in that regard.

There's a sense of a closed loop of consistency and a smaller, tighter set of thought patterns that apply both on this side and the other side – whatever those sides may be, say, values vs. types, compilation vs. runtime, AST vs. raw syntax, parsing vs. printing…

It's an explosion of possibilities!, but as it's a shaped charge of constructive and intuitionistic principles it mostly just blows up shackles and clears the air :)

Re:GADTs for Dummies by siraben on Wednesday May 19, 2021 @06:26AM (#27206694)
Yes! I love the consistency of dependent types. You don't need separate declarations for type aliases or a dozen language extensions to do some type-level evaluation, it's quite uniform.

> It's an explosion of possibilities!, but as it's a shaped charge of constructive and intuitionistic principles it mostly just blows up shackles and clears the air :)

Indeed, I feel like we are just getting started finding the right "design patterns" for working with dependently typed programs. It's sort of a Wild West of styles in the Coq world currently, IMO, which is great when something is well designed but terrible when using a theorem demands you prove some clearly obvious precondition (e.g. Riemann integration in the stdlib is a pain to work with).

Re:GADTs for Dummies by AnthonBerg on Wednesday May 19, 2021 @08:13AM (#27207507)
π–™π–π–Š π–”π–™π–π–Šπ–— π–˜π–Žπ–‰π–Š: Having expressibility of Riemann integration near a stdlib at all is marvelous!