On 1/4/06, Christian Maeder
Adrian Hey wrote:
So I vote we drop the left/right biasing distinction. This way lies madness (as they say:-)
I would not like to drop biasing distinction, as I don't think this costs too much. With biasing, maps could be implemented via "biased sets" (Set (MapEntry a b)):
data MapEntry a b = a := b
instance Eq/Ord a => Eq/Ord (MapEntry a b) where compare (a1 := _) (a2 := _) = compare a1 a2
I would discourage such a use. This is exaclty why we provide Maps, after all.
If there's some good reason why we should care then the corresponding type should not be an instance of Ord. So what should be done in cases like this? It's tempting to think we should add HOF versions like this..
insertUsing :: (a -> a -> COrdering a) -> a -> Set a -> Set a
It may also be possible to pass an order to "empty" and use it further on. The major problem is if varying orders are used ie. for "union".
By good, bad or earlier design Data.Set/Map relies on Ord instances and I'm quite content with it.
A "good" solution would require dependent types or its type-classes emulation, but I think this is overkill. Cheers, JP.