Csongor Kiss

@sliminality functorial.com/the-future-is-… I think this might be close to what you're looking for?

I can't find anything written about the relationship between React contexts and the Reader monad. To me, they solve identical problems in structurally similar ways (modulo the React runtime). I am surely not the first person to make this connection. Anyone have pointers?

@evie_fp don't the built in 'gf' and 'gx' commands work for you? (for (1) and (2) respectively)

Anyone got some (neo)vim plugins that: - recognise (relative) paths and allow me to open said file? - recognise (web) URLs and allow me to open them in my browser?

@cattheory @ulysses4ever @EremondiJoey the "cookie" pronunciation sounds like the Hungarian word "kuki" (meaning penis)

@ulysses4ever @EremondiJoey ah, right.

@am_i_tom Whynad

Can the compiler handle linear state for us? Yes, by teaching linear logic to type class constraints. Written up in an #ICFP 2021 submission by Jean-Philippe Bernardy, @RaeHaskell, @Lowert, @aspiwack, and @nicolasgwu (3/5) arxiv.org/abs/2103.06127

@ZevEisenberg @sliminality @modernserf @mitpress but why does @mitpress exist

@sliminality @modernserf Because @mitpress exists!

@mpeytonjones @tweagio The main issue with implicit params is that you have to name them, such as "?x". If you want to use them to represent an "open file" capability for example, then maybe you call this capability "?open". But now you can't have multiple open files, as their tokens' name would clash!

@tweagio Of course, constraints give you implicit evidence so it's not so different in the end, but if it's the evidence that's used linearly, it's perhaps nicer to talk about that directly rather than the type-level thing...

Linear types in Haskell, research phase 2: arxiv.org/abs/2103.06127. This time we're focusing on much improved developer experience. Linear type class constraints are the secret sauce.

@aspiwack @GMorrisett @tweagio You would need something like `Dup (Show a) => ...`

@Lowert @GMorrisett @tweagio I'd add that Linear Constraints builds up on Linear Haskell, which is a somewhat different system than Clamp. But I don't see any reason why we couldn't build the same on top of Clamp. Though I don't see what it would look like out of the top of my head.

@Iceland_jack You must read this paper exactly once

A quantifier for every citizen

@VBragilevsky @int_index Great news! Congrats :)

I'm happy to welcome Vladislav Zavialov @int_index to the GHC Steering Committee!

@acid2 That's a nice trick!

@Lowert I think I erased way too much information from my tweet, but got it working with gist.github.com/ocharles/351ad…. The problem was I originally had: DecodeNotNull :: Hasql.Value a -> HasqlDecoder a Yoneda to the rescue!

I have a 'data Foo a = Foo ... (F a)', and data Foo a where MkFoo :: Bar x -> (Maybe x -> Either String a) -> Foo a Is all hope lost to think that I can coerce a 'Foo a' into a 'Foo b'? I want this because 'Foo' is part of a type class I want to use with GND.

@am_i_tom @hdgarrood @kmett I believe it was the other way around! I was going to write naturality...

@aspiwack I'd like to hear more about this

You know what? I'm starting to think that page limits on conference article don't encourage the healthiest writing habits.

@mattoflambda Great trick! With the next version of GHC (9 I think?) it will also be possible to write: gross :: forall {k} (a :: k) b. C a b => b -> String

Now, we can write: gross :: forall a b k. (KindOf k a, C a b) => b -> String And it works out fine!

You may have a class with a polykinded type index, and you want to use type-applications: class C (a :: k) b where gross_ :: b -> String Unfortunately, the explicit type of `gross_` is: gross_ :: forall k (a :: k) b. b -> String `k` must come before `a`.

The GHC Steering Committee just accepted “Unsaturated Type Families” by @Lowert github.com/ghc-proposals/…

@alpha_convert Just tell them to substitute (x/e)/(e/x)

I have a TA session tonight where I'm talking about unification. Usually would be fine, but this class has been using x/e substitution notation instead of e/x, and so I'm bound to mess it up and confuse everyone

@taylorfausak @googleson78 Ah. That's because you don't have PolyKinds turned on in your ghci session, so the kinds get defaulted to Type. Try `:set -XPolyKinds` first!

@Lowert @googleson78 Wait, hmm. This works without a type application: > class From a b where from :: Ambiguous a -> b > instance From Char Int where from = fromEnum > from ('a' :: Char) :: Int > -- 97

Is it possible to write a polymorphic Haskell function that *requires* a type application, even if it's not ambiguous? For example: > into :: forall b a . From a b => a -> b > yFun (into xVal) -- broke > yFun (into @‌Y xVal) -- woke