Re: Proposal: more general unionWith for Data.Map

On Tue, Jan 24, 2012 at 11:58 AM, Milan Straka wrote:

...

FYI, the new intersectionWithKey is not released yet -- maybe we could leave intersectionWithKey as it is, and provide mergeWith[Key] instead. Opinions?

I think so. We should avoid the words union and intersection if they don't correspond to the mathematical idea, or we risk confusing people.

I think that with intersectionWithKey, we are still holding to the mathematical idea. Imagine I have a Map Hash List. When performing intersection of two such maps, it can happen that for one hash there are different values in the maps and so in the intersection the has is not present. Generally if the Map k a is representing a set using both k and a, the intersection on the same key can still be empty. But this is not so in union -- if key is any of maps, it should be in the result. So personally I would generalize the combining function of intersectionWith to (a -> b -> Maybe c), but not generalize the combining function of unionWith, and instead provide mergeWith with combining function (a -> a -> Maybe a). (Similarly for *Key variants). Cheers, Milan