Proposal: syntactic sugar for proxies

Change Proposals
#1

I’m not sure if this has been proposed already, but after writing PS for ~7 months now, I find that the bit of syntactic sugar I find myself wanting most is a shorter way to write.

a = Proxy :: Proxy Int

I’m not sure if this has been proposed already, nor do I know what would be the best solution. Some candidates:

  • @Int
  • !Int
  • ^Int
  • :muscle:Int
  • :arrow_down:Int

With perhaps the same operatoin at the type level.

I think this would make code that makes heavy use of proxies easier to write and read.

z = @Int

foo :: forall z x. (@z -> x) -> x
foo i = i @z

Compared to:

z = Proxy :: Proxy Int

foo :: forall z x. ((Proxy :: Proxy z) -> x) -> x
foo i = i (Proxy :: Proxy z)

Thoughts? If this seems reasonable, I can take a crack at a PR (it’d be my first to purescript/purescript … gulp …).

#2

I think there’s multiple mentions of proxy syntax in purescripts issue tracker, including https://github.com/purescript/purescript/issues/3137

You can make it slightly more terse already by using wildcards: p = (Proxy :: _ Int). Indeed you could also alias Proxy to get it even smaller

#3

I think that purs was once already there (with @Int syntax) but the feature was somewhat problematic and didn’t land in the official compiler release.

#4

It was problematic because the kind checker was broken. With PolyKinds it wouldn’t be an issue, but I think visible type applications is what we really want anyway.

4 Likes
#5

Thanks. I didn’t know about the wildcard syntax. P :: _ Int is indeed more terse than Proxy :: Proxy Int, although I’d have to write it out several times to get a sense if it is more readable. I agree with Nate that visible type applications would be better as that’s a superset of what I’m talking about. Is there a GH issue and/or PR for that yet?

#6

I’m pretty sure the @TypeName wasn’t added in v0.14.0 because we were first deprecating the @ symbol so that it wasn’t allowed anywhere except in pattern matching:

case _ of
  foo@(Tuple a b) -> ...

It was my understanding that it would be added in v0.15.0, but I’m not sure. I also haven’t read through the above linked issue.

1 Like