What do you think of introducing to Haskell-2 ad hoc ordering for the instance operlap? For example, ---------------------------------------------------------------------- data C a = C a class Convertible a b where cv :: a -> b -> b instance (order 0) Convertible a a where cv a _ = a -- Ia instance Convertible a b => Convertible a (C b) -- Ic where cv a (C b) = C (cv a b) ---------------------------------------------------------------------- I am suffering from that some compilers regard this particular program as ambiguous. Here the types (a,a) and (a,C b) overlap and none of them is the substitutional instance of another. But let (order 0) point that on the overlapping area the instance Ia has to apply - because its order is smaller than the default order for Ic. Generally, each instance of class <Cs> is assigned the order o >= 0 - explicitly or by default. Then for the type T matching the instances I1..Ik of <Cs> the compiler has to choose I(j) of the minimal order. ? ------------------ Sergey Mechveliani [EMAIL PROTECTED]
