[Haskell-cafe] Subtypes and Co/Contra-variance

[Ashley Yakeley]

Consider this:

   type Super = forall a. [a]
   type Sub = forall b. b

   cast0 :: Sub -> Super
   cast0 s = s

   cast1a :: forall p. (Super -> p) -> (Sub -> p)
   cast1a sp x = sp x

   cast1b :: forall p. (Super -> p) -> (Sub -> p)
   cast1b sp = sp

This compiles except for cast1b (ghc -c -fglasgow-exts):

     Occurs check: cannot construct the infinite type: a = [a]
     When generalising the type(s) for `cast1b'

I get a similar result when I try a different Super and Sub:

   type Super = (?x :: Bool) => Int
   type Sub = Int

     Couldn't match expected type `Sub' against inferred type `Super'
     When matching `(?x::Bool) => Int'
               and `Int'
       Expected type: Sub -> p
       Inferred type: Super -> p
     In the expression: sp

In either case, if cast1a compiles, shouldn't cast1b compile?

-- 
Ashley Yakeley