[Haskell-cafe] Question about type families

[Gábor Lehel]

On Tue, Sep 13, 2011 at 4:58 PM, Grigory Sarnitskiy <sargrigory at ya.ru> wrote:
> Is there a way to make the following code working?
>
> {-# LANGUAGE TypeFamilies #-}
>
> data family Foo a
>
> data instance (Num a)        => Foo a = A a deriving Show
>
> data instance (Fractional a) => Foo a = B a deriving Show
>
>
> I want to have different constructors for 'Foo a' depending on a class of 'a'. Note also, that in the example above I also meant constructor A to be available for (Fractional a) => Foo, since in that case 'a' has Num too. How can I achieve it, maybe not with TypeFamilies? Current error is
>
>    Conflicting family instance declarations:
>      data instance Foo a -- Defined at 1.hs:7:33-35
>      data instance Foo a -- Defined at 1.hs:5:33-35

Directly, with current GHC? Doubly, maybe triply impossible. Type and
data families aren't allowed to overlap, and there's no way to
dispatch over whether a type is or is not a member of a class. (You
can require that it be a member, but you can't say "if not, do this
other thing"). Also, you can't give type and data families superclass
contexts the way you can classes. I haven't actually encountered this
before, but I think that what you've written here is datatype contexts
for the various instances of the data family, which means you can only
construct a 'Foo a' if the 'a' is a member of the class - but it in no
way affects which instance is chosen. Datatype contexts are considered
a misfeature, besides.[1]

Anyway. What's your wider goal?

[1] http://hackage.haskell.org/trac/haskell-prime/wiki/NoDatatypeContexts

>
> _______________________________________________
> Haskell-Cafe mailing list
> Haskell-Cafe at haskell.org
> http://www.haskell.org/mailman/listinfo/haskell-cafe
>



-- 
Work is punishment for failing to procrastinate effectively.