You'd need GHC extensions to pull that off, Greg, but it could be done. Something like

data Vect (size :: Nat) (a :: *) where
  Cons :: a -> Vect n a -> Vect (Succ n) a
  Nil :: Vect Zero a

data NatUpTo (n :: Nat) where
  Here :: NatUpTo n -- "Here" means 0
  There :: NatUpTo n -> NatUpTo (S n) -- "There x" means 1 + x

vectIndex :: Vect (S n) a -> NatUpTo n -> a
vectIndex (Cons x _) Here = a
vectIndex (Cons _ xs) (There i) = vectIndex xs i

class (Enum a) => EnumSize (size :: Nat) a where
  enumerateAll :: Vect size a
  -- law: (enumerateAll `asTypeOf` [x]) `vectIndex` unsafeIntToNatUpTo (fromEnum x)

It gets tedious, though, dealing with all of that safety. LiquidHaskell is perhaps a more viable option for such a thing. It would be nice to have standard libraries written with LiquidHaskell so that we have a more rigorously proven code base.

-- Dan Burton


On Wed, Sep 11, 2013 at 10:57 AM, Greg Fitzgerald <garious@gmail.com> wrote:
I'd prefer something more like:

    mux :: Enum b => [a] -> b -> a
    mux xs x = xs !! fromEnum x

so then 'bool' could be implemented as:

    bool :: a -> a -> Bool -> a
    bool f t = mux [f, t]

but 'mux' needs a stronger type signature.  The size of the enum is
known at compile-time.  Is there any way to constrain the input list
to be the same size?

Thanks,
Greg




On Wed, Sep 11, 2013 at 2:18 AM, Simon Hengel <sol@typeful.net> wrote:
> On Tue, Sep 10, 2013 at 11:02:19PM +0100, Oliver Charles wrote:
>> I would like to propose that the following is added to Data.Bool in base:
>>
>> bool :: a -> a -> Bool -> a
>> bool f _ False = f
>> bool _ t True  = t
>
> +1
> _______________________________________________
> Libraries mailing list
> Libraries@haskell.org
> http://www.haskell.org/mailman/listinfo/libraries
_______________________________________________
Libraries mailing list
Libraries@haskell.org
http://www.haskell.org/mailman/listinfo/libraries