GHC infers the type g :: (Foo [[a]] [b], Foo [[b]] [a]) => [a] -> [b] -> Int As you imply, if GHC used the instance decl to simplify the first constraint to Foo [a] b, for example, then the functional dependency would kick in, and there would be an infinite loop. But GHC's constraint simplification is (now) lazy, unless forced to be eager by supplying a type signature. I think Hugs is the same, and I think that's probably why it terminates. Interesting example! Simon | -----Original Message----- | From: Dominic Duggan [mailto:dduggan@cs.stevens-tech.edu] | Sent: 28 February 2001 16:00 | To: hugs-bugs@haskell.org | Subject: please confirm that this is a feature | | | I am finishing up revisions on a paper that | has been accepted to JFP, on type-checking | with multi-parameter type classes. | With the following definitions: | | | class Foo a b | a -> b where foo :: a -> b -> Int | | | | instance Foo Int Float where foo x y = 0 | | | | instance Foo a b => Foo [a] [b] where foo [x] [y] = foo x y | | | | g x y = (foo [x] y) + (foo [y] x) | | In our system type-checking fails to terminate when | type-checking "g", but Hugs 98 (Feb 2001) returns | a type. | | I assume that this is because of a depth bound in the | type-checker, can you confirm this before I submit | the final version of the paper? | | Thanks | --dd |