Consider something like data Foo a = Foo !Int !a The data constructor Foo needs to be strict in its arguments, so it'd need Seq Int and Seq a. Seq Int would be resolved by the environment, but Seq a would need to come from somewhere. data Seq a => Foo a = Foo !Int !a gives you Foo :: Seq a => Int -> a -> Foo a so it is available when evaluating the constructor. We're not storing the instance as a slot in the constructor, so this isn't. data Foo a where Foo :: Seq a => Int -> a -> Foo a or equivalently data Foo a = Seq a => Foo !Int !a This is where the data type contexts came from. They were originally motivated by the needs of the no-longer existent Seq class, and then subverted to make Complex less useful than it otherwise could be. ;) -Edward On Wed, Apr 2, 2014 at 3:08 AM, Tom Ellis < tom-lists-haskell-cafe-2013@jaguarpaw.co.uk> wrote:
On Tue, Apr 01, 2014 at 05:32:45PM -0400, Edward Kmett wrote:
Unfortunately the old class based solution also carries other baggage, like the old data type contexts being needed in the language for bang patterns. :(
Can you explain why that is so? I don't understand.
Tom
_______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe