On 6/10/08, Derek Elkins <[EMAIL PROTECTED]> wrote:
> This is the lack of impredicativity.
>
> W :: a -> W a
> To get the result type W (forall a. t a), W must instantiate the a in
> W's type to (forall a. t a).  Further we then pass it to (.) which has
> type (b -> c) -> (a -> b) -> a -> c and thus require instantiating both
> a and b to higher-rank types.  A predicative type system does not allow
> instantiating type variables to quantified types.

Although if I am reading the literature correctly, that definition
will likely typecheck in the next GHC...

http://research.microsoft.com/~simonpj/papers/boxy/

  -- ryan
_______________________________________________
Haskell-Cafe mailing list
[email protected]
http://www.haskell.org/mailman/listinfo/haskell-cafe

Reply via email to