On Thu, Sep 02, 2010 at 05:10:29PM -0400, Alec Benzer wrote:
Ah, ok, so the reason what I trying didn't work is because I used an actual type instead of a type variable? I got confused because of the emphasis you put on * distinct *.
I think the emphasis is from the error message that GHC spits out, not from Daniel. This is a particularly confusing error message though.
And so, if I want to make Maps applicative functors without dealing with FlexibleInstances, I'd have to do something like this?
import Control.Applicative import qualified Data.Map as M import Data.Monoid
instance (Monoid k, Ord k) => Applicative (M.Map k) where pure x = M.fromList [(mempty,x)] fs <*> xs = M.fromList [(k1 `mappend` k2,v1 v2) | (k1,v1) <- M.assocs fs, (k2,v2) <- M.assocs xs]
Sure. Although there's really no reason to avoid FlexibleInstances.
(sacrificing some functionality, since spaces won't get intercalated between keys if i use strings)
Technically, the version with the intercalated spaces didn't satisfy the Applicative laws anyway. For example, if <*> inserts a space it is not the case that pure f <*> x = f <$> x since there would be an extra space introduced on the left-hand side. I like your more general Monoid-based version much better (and I think it's not too hard to show it satisfies the Applicative laws, although I haven't thought about it too hard). -Brent