Hi Marc, Marc Weber wrote:
class (Monad m) => GetMV m a where ... instance GetMV m c where (2)
...it would be save to assume that the programmer doesn't want a failure but success. Thus ghc could infer (3) automatically but doesn't
The Monad m on the class declaration does not mean that the programmer _wants_ m to be a monad. It means that the programmer _promises_ to make m a monad - even more, the programmer wants the compiler to force anyone using this class to first make m a monad, and otherwise enforce type safety by failing to compile. This is very good, because then any future use of an instance of GetMV can be safely assumed to have also a Monad instance, no matter what its type. One place you can use that nice assumption is in the method declarations within the original class declaration itself. Another is in function definitions, as you say. But when writing an instance declaration, it is time to pay up. You made a promise. Now you have to show the compiler where to get that Monad instance. There are two ways of doing that. One way is if the type itself already has a separate Monad instance. The other way is if the type is not fully specified - it is a variable - then you can "pass the buck" and say that whoever uses this instance must first make sure that the value of the type variable is a type that already has a Monad instance. Regards, Yitz