[Haskell-cafe] Re: Specific to General to Specific

[ezyang]

Hey Oleg,

Thanks for the response!

Excerpts from oleg's message of Thu Aug 19 03:51:05 -0400 2010:
> Granted, we would have to write boilerplate as we have to re-direct
> specialized methods to the general ones.

Interesting.  As far as I can tell, this means that any general method
that I want to use in the specialized typeclass needs to be explicitly
redirected in the specialized typeclass's definition.  What does a user
do if they would like to write their own generic function (using
the other generic functions) and then specialize it?

> Here is the instance of this trick, adopted to use type families
> rather than functional dependencies (no UndecidableInstances is
> required now):

The example (concatenated with the previous full code), doesn't typecheck
for me:

    Prelude> :r
    [1 of 1] Compiling Main             ( Test.hs, interpreted )

    Test.hs:60:18:
        No instance for (Apply (NM mg n))
          arising from a use of `apply' at Test.hs:60:18-38
        Possible fix: add an instance declaration for (Apply (NM mg n))
        In the expression: apply (NM :: NM mg n)
        In the definition of `monadicValue'''':
            monadicValue''' = apply (NM :: NM mg n)

    Test.hs:60:18:
        Couldn't match expected type `m n1 ()'
               against inferred type `Typ (NM mg n)'
          NB: `Typ' is a type function, and may not be injective
        In the expression: apply (NM :: NM mg n)
        In the definition of `monadicValue'''':
            monadicValue''' = apply (NM :: NM mg n)
    Failed, modules loaded: none

Cheers,
Edward