Hello, On 4/25/05, Andrew Pimlott <andrew@pimlott.net> wrote:
I appreciated your explanation, but can you also address (to the list) the last case given by the original poster?
On 4/25/05, Andreas Rossberg <rossberg@ps.uni-sb.de> wrote:
in particular, when I compare with the single parameter case:
class C a where fc :: a -> a -> ()
c1 x = let p = fc x in () c2 x = let p y = fc x y in ()
where
c1 :: C a => a -> () c2 :: C a => a -> ()
is inferred, as I would expect.
The inference steps for this case are much the same except, that the inferred type for "p" now will be: "a -> ()", provided that we can solve the constraint "C a". Because we have assumptions about "a" in the environment (namely it is mentioned in the type of the varible "x") we cannot generalize the type of "p". It therefore remains monomorphic, and the constraint "C a" is propagated to the type of "c2". -Iavor