Albert ten Napel

Interested in dependent types

[Albert ten Napel @Appelflop]

@starsandspirals Alright great. Thanks a lot :)

[Albert ten Napel @Appelflop]

@starsandspirals Hey Dani! I had a question about the "Entente Cordiale" paper. In the copy rule what do you mean with "smuggle out a value which pretends to be truly unique.". I cannot think of an example for this. Related: https://t.co/LlubhAXu5s

[Albert ten Napel @Appelflop]

@starsandspirals Thanks a lot for the reply! :) Additional question: is it then true that `clone t1 as x in t2` could be changed to `clone t`? I guess in the paper this syntax is chosen to make it clear it's some kind of monad, or is there another reason?

[Albert ten Napel @Appelflop]

@GabriellaG439 @sgraf1337 A related idea can be found in the paper: "Spine-local type inference"

[Albert ten Napel @Appelflop]

@therealchreke @zetalyrae How would algebraic effects allow for in-place mutation?

[Albert ten Napel @Appelflop]

@jacksonlango Where can I find the blogposts about your cyberpunk roguelike/TerminalRain? Unfortunately the links do not work anymore... :(

[Albert ten Napel @Appelflop]

@GabriellaG439 Hindley-Milner with only top-level generlization is also very simple. But I feel RankN and forall subtyping is something you'd like to keep, is that true?

[Albert ten Napel @Appelflop]

@GabriellaG439 You might be interested in the paper "Spine-local type inference", it's also about bidirectional typechecking without unification. Also have you considered just sticking to Hindley-Milner? I think everything stays pretty simple in that case. No scope checking needed.

[Albert ten Napel @Appelflop]

@PTOOP @EremondiJoey Do you think we'll ever find a way to get inductive-inductive types from a fixpoint? Or is there something fundamentally different about those.

[Albert ten Napel @Appelflop]

@GabriellaG439 @antonycourtney I'm sorry, I don't quite understand what you mean. Do you mean in the approach where you store mutable references in the type AST itself?

[Albert ten Napel @Appelflop]

@GabriellaG439 I think for System F without subtyping things can stay simple. For each meta associate a set of in-scope type variables, then during occurs check also check for out of scope type variables...just an idea.

[Albert ten Napel @Appelflop]

@GabriellaG439 If you stick to Hindley-Milner unification is very simple. But once you include higher-ranked types you have to keep track of the type variable scoping which is trickier. And then if you also want some subtyping between the foralls...difficult.

[Albert ten Napel @Appelflop]

@GabriellaG439 @antonycourtney I haven't had any problems with mutation+foralls. Could you elaborate where the interaction is poor?

[Albert ten Napel @Appelflop]

@andrejbauer Yes I looked on Vimeo and then I clicked on the links on your blog (redtt specifically), which still goes to Vimeo.

[Albert ten Napel @Appelflop]

@andrejbauer I noticed that the recordings of "Every Proof Assistant" are no longer available. Any chance you could make them available again?

[Albert ten Napel @Appelflop]

@ChShersh But if you don't need let generalization and subsumption. Then you can simplify.

[Albert ten Napel @Appelflop]

@ChShersh The use of the context in "complete and easy" is really good to understand the scoping issues with RankNTypes and also generalization.

[Albert ten Napel @Appelflop]

@kmett I'm curious, what is the status of your Coda project? Is it abandoned?

[Albert ten Napel @Appelflop]

@wouterswierstra There's a mistake at the start of chapter 4: the definition of B and C is swapped.