[Haskell-cafe] ST and Transformers

[Stefan O'Rear]

On Fri, Dec 22, 2006 at 06:44:54PM +0100, Pepe Iborra wrote:
> Rank-2 types seem to interact badly with (.) and ($), but my type  
> theory educated neuron doesn't know why. I think this is folklore  
> knowledge?


You say:
> runErrSt = (.) runST runErrorT

runErrorT has type ErrorT e m a -> m (Either e a)
runST has type (forall s. ST s a) -> a
(.) has type (b -> c) -> (a -> b) -> (a -> c)

since runST is the first argument to (.), we must unify (forall s. ST s a) with b

Quoting the GHC Manual:
> There is one place you cannot put a forall: you cannot instantiate
> a type variable with a forall-type. So you cannot make a forall-type
> the argument of a type constructor. So these types are illegal:

I am not aware of why this restriction exists, but it is a consequence of a documented
restriction.  (I hear rumors about limited support for "left-to-right impredicative
instantiation" in GHC HEAD, but it's way over my head :( )