Re: [Haskell-beginners] Guard in class def

OK - thanks (both)! On 8 Jan 2012, at 11:44, Chaddaï Fouché wrote:

On Sun, Jan 8, 2012 at 10:29 AM, Henry Lockyer wrote:

...

Hi, I was just looking through the 'Monad Transformers' chapter in Real World Haskell. They are using the "reader" monad to illustrate the transformer structure but I fell off at the first bend when I saw the following:

class (Monad m) => MonadReader r m | m -> r where ask :: m r local :: (r -> r) -> m a -> m a

Could someone explain the use of a guard here / how to read this with the "| m -> r" ? I haven't come across this usage before (as far as I have noticed :-) and the meaning hasn't jumped out at me yet...

It's not a "guard", it's a functional dependency : "m -> r" tells the compiler that for any m, r is uniquely determined. That is if you have the monad "ReaderT Int Identity" for m, it is clear that "Int" is the only possibility for r. This is necessary for type inference to process in more case (often the type of the monad is fixed by the context, but not the type of r, thanks to the FD, the compiler can confidently use the instance it found for m).

Note that this solution never really satisfied the Haskell users/developers since it is more from the logic paradigm (like Prolog) than the functional paradigm, so since two or three years now, this tends to be replaced by a new solution, more functional in style, called type family, which would read like that :

...

class (Monad m) => MonadReader m where type RC :: * -> * -- Reader content ask :: m (RC m) local :: (RC m -> RC m) -> m a -> m a

where RC is like a "function on type".

-- Jedaï