Think of it this way:
-- Here is some data representing the typeclass 'Show'
data ShowDict a = ShowD (a -> String)
show :: ShowDict a -> a -> String
show (ShowD f) a = f a
-- Here's a sample implementation for Strings
showString :: ShowDict String
showString = ShowD (\s -> "\"" ++ escape s ++ "\"") where
escape = concatMap escapeChar
escapeChar '\\' = "\\\\"
escapeChar '"' = "\\\""
escapeChar c = [c]
-- Here's an implementation for pairs that uses the implementation for each piece of the pair
showPair :: ShowDict a -> ShowDict b -> ShowDict (a,b)
showPair (ShowD sa) (ShowD sb) = ShowD (\(a,b) -> "(" ++ sa a ++ ", " ++ sb b ++ ")")
-- Here is what you are asking for
implementMe :: ShowDict (a,b) -> ShowDict a
implementMe = ????
Cannot deduce (Show x) from context (Show (x, y)).
Cannot deduce (Show y) from context (Show (x, y)).
Um... seriously?
>From Prelude, we have
Show x, Show y => Show (x, y)
So clearly it works in the forward direction. But apparently not in the reverse direction.
Is this a bug or a feature? (I.e., is there some obscure possibility I haven't thought of which means that doing the reverse inference would be incorrect?)
_______________________________________________
Haskell-Cafe mailing list
Haskell-Cafe@haskell.org
http://www.haskell.org/mailman/listinfo/haskell-cafe