[Haskell-cafe] difference between class context and deriving

Ryan Ingram ryani.spam at gmail.com
Tue Aug 2 19:16:15 CEST 2011

On Tue, Aug 2, 2011 at 5:57 AM, Patrick Browne <patrick.browne at dit.ie>wrote:

> data Eq a => Set1 a = NilSet1 | ConsSet1 a (Set1 a)
> data         Set2 a = NilSet2 | ConsSet2 a (Set2 a) deriving Eq

The former declaration is a language wart, IMO.  All it does is attach a
restriction to the constructors of Set1; try

> :t NilSet1
NilSet1 :: Eq a => Set1 a
> :t NilSet2
NilSet2 :: Set2 a

But it doesn't give you that restriction back when you use it:

> let f (ConsSet1 v _) = v == v
> :t f
f :: Eq a => Set1 a -> Bool

You'd think that since you had to provide the evidence (Eq a) when you
constructed ConsSet1, that it'd be available, but it's not.

-- Seems to have same type
> :t ConsSet2 1 NilSet2
> ConsSet2 1 NilSet2 :: forall t. (Num t) => Set2 t
> :t ConsSet1 1 NilSet1
> ConsSet1 1 NilSet1 :: forall t. (Num t) => Set1 t

Remember that Eq is a superclass of Num.

> let f1 x = ConsSet1 x NilSet1
> let f2 x = ConsSet2 x NilSet2
> :t f1
f1 :: Eq a => a -> Set1 a
f2 :: a -> Set2 a

'deriving Eq' on the definition of Set2 makes the instance  'instance Eq a
=> Eq (Set2 a)'.  So you can construct Set2 at any type, and when the
underlying type 'a' has equality, then 'Set2 a' does as well.

  -- ryan
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.haskell.org/pipermail/haskell-cafe/attachments/20110802/4e20b6aa/attachment.htm>

More information about the Haskell-Cafe mailing list