[Haskell-cafe] Expressing "self-composable" functions at the type level

[Ignas Vyšniauskas]

Hi,

On 11/10/2013 03:18 AM, Dan Doel wrote:
> Doing significantly better is harder, and I don't think you'll ever
> get to the real type you want, which is:
>
> twice : ((a -> b) ∩ (b -> c)) -> a -> c
>
> I.E. look up interstection types if you're into this stuff. They
> tend to be less popular than quantifier-based polymorphism, though.

I think the problem is that having intersection types implies subtyping,
and subtyping is obviously troublesome.

Essentially the same problem [as mine] is discussed in an old thread,
where a partial ("value-level") solution is proposed via typeclass hackery:
http://www.haskell.org/pipermail/haskell-cafe/2006-January/013868.html

--
Ignas