I have a class whose instances can extract a field of type a. Is there a way of referring to that type 'a' in the signatures of functions that use the class? Sigh! That wasn't very clear. I'll try explaining via example: I am trying to define a `Pairable' class, which is an abstraction of all kinds of pairs: (a,b), Pair a b, etc.
class Ord p => Pairable p where is_pair :: p -> Bool pair_fst :: Ord a => p -> a -- Precondition: is_pair pair_snd :: Ord b => p -> b -- Precondition: is_pair make_pair :: (Ord a, Ord b) => a -> b -> p
instance (Ord a, Ord b) => Pairable (a,b) where is_pair (a,b) = True pair_fst (a,b) = a pair_snd (a,b) = b make_pair a b = (a,b)
I then want to define a `Relation' to be a set of Pairable values.
newtype Pairable p => Reln p = Reln [p]
This works okay for many of my relation operations. But I run into an interesting problem with `inverse' (which swaps the components of each pair). inverse :: (Pairable p, Pairable p2) => Reln p -> Reln p2 inverse rs = Reln[make_pair (pair_snd p) (pair_fst p) | p <- reln2list rs] Hugs gives the error: ERROR /home/utting/jaza/tmp.hs:19 - Cannot justify constraints in explicitly typed binding *** Expression : inverse *** Type : (Pairable a, Pairable b) => Reln a -> Reln b *** Given context : (Pairable a, Pairable b) *** Constraints : (Ord c, Ord d) which I can understand. BUT, I cannot see any way of naming the `component types', c and d, in the signature of inverse! It seems like I want to pass the component types into the class definition somehow: class Pairable (p a b) How does one normally get around this kind of problem? Thanks. Mark Utting Professeur Invité Laboratoire d'Informatique de l'Université de Franch-Comté 16, route de Gray - 25000 Besançon, cedex, FRANCE Tel: (33) 3 81 66 20 69 Fax: (33) 3 81 66 64 50 Email: marku@cs.waikato.ac.nz