#rust-lang

JoshTriplett: there *is* one place where we sort of need to add it

JoshTriplett: which is, say, unify(for<'a> fn(<T as Trait<'a>>::X), fn(U)) needs to also output for<'a> <T as Trait<'a>>::X == U

JoshTriplett: but just impl HashSet<MyStruct> {} doesn't do any unification, because unlike Haskell (I guess?), there is no type at the declaration site, but rather a name and generic parameters

JoshTriplett: so you're *instantiating* HashSet with K = MyStruct

JoshTriplett: which results in a type that looks like Adt(HashSet, [MyStruct])

JoshTriplett: and none of that sees the bounds of HashSet

WF is responsible for dealing with things like that, and that runs *later* (because you need to be able to understand all the types ahead of checking they're WF)

JoshTriplett: oh let me rephrase something :)

JoshTriplett: struct HashSet<K: Hash> {} is actually struct HashSet<K> where K: Hash {} *not the other way around*

eddyb: I guess I really don't want there to be a difference between those two.

I always thought they were just syntactically different, not semantically different.

JoshTriplett: the former desugars into the latter, effectively

That's fine.

i.e. forall a.(Hash a) => HashSet a. but Haskell is really bad at this anyway

(I get angry every time I have to deal with its conflation of 2-3 concepts within the same syntax)

JoshTriplett: related: https://twitter.com/eddyb_r/status/836892386835...

frankly I think QWERTY was a mistake

and we should all have ∃, ∀ & Λ on our keyboards

I bet you teens would be using them instead of less precise shorthands

well, not Λ unless they're doing higher-order texting

eddyb: I was thinking more about `data Hash a => HashSet a" (or similar syntax using GADTs or similar).

JoshTriplett: there's an implied list of one element, a, in there

JoshTriplett: not sure there's a way to write it in Haskell. I know forall exists but not its semantics in this case

JoshTriplett: but anyway, even in Haskell, "Hash a" is a bound that happens to be syntactically placed before HashSet a

oh I guess the list is not implicit, it's HashSet a?

depends how complete that example is :P

JoshTriplett: anyway, the bound is not inherent to a, it just *uses* a

eddyb: It's been a while since I wrote actual Haskell code. :)

JoshTriplett: this is even more true in Haskell where there is no asymmetry

struct HashSet(K) where Hash(K)

JoshTriplett: anyway this is all minutia :P