[Haskell-cafe] More experiments with ATs

[Andrew Coppin]

Sjoerd Visscher wrote:
> On Jul 3, 2010, at 4:39 PM, Andrew Coppin wrote:
>
>   
>> class Container c => Functor c where
>>   fmap :: (Functor cx, Functor cy, Element cx ~ x, Element cy ~ y) => (x -> y) -> (cx -> cy)
>>
>> However, this fails horribly: The type signature fails to mention c.
>>     
>
> You have to mention c, this means an extra argument to fmap. But if you do that you also get the opportunity to restrict what x and y can be. 
>   

Well, you can say

  class Container cy => Functor cy where
    fmap :: (Functor cx, Element cx ~ x, Element cy ~ y) => (x -> y) -> 
(cx -> cy)

But that's still wrong.

> As you'll have to pass around this extra argument, it will usually be easier to just pass around the map function though.
>
>   
>> type family F f a :: *
>> class RFunctor f where
>>   (%) :: f a b -> (a -> b) -> F f a -> F f b
>>     

I have literally no idea what a type family is. I understand ATs (I 
think!), but TFs make no sense to me.

(For this reason, most if not all of the rest of this post doesn't make 
sense.)