These things are weight-balanced search trees augmented with semi-heaps. We have play, which joins together matching pieces. So that means if we can write a split and splitMaybe (like in the Data.Map internals), we should probably be able to get variations on union, intersection, and difference, along with something like the mergeA interface of Data.Map. I doubt we'll be able to come up with solid proofs of performance (those are quite subtle), but I would expect very good practical performance. Of course, this is all assuming that there isn't something major I'm missing; I haven't actually finished reading Hinze's paper.
These things are weight-balanced search trees augmented with semi-heaps. We have
play, which joins together matching pieces. So that means if we can write asplitandsplitMaybe(like in theData.Mapinternals), we should probably be able to get variations onunion,intersection, anddifference, along with something like themergeAinterface ofData.Map. I doubt we'll be able to come up with solid proofs of performance (those are quite subtle), but I would expect very good practical performance. Of course, this is all assuming that there isn't something major I'm missing; I haven't actually finished reading Hinze's paper.