Twan van Laarhoven wrote:
My proposal would be to just call things as they are, and add the functions findGreater, findLess, findGreaterEqual and findLessEqual.
I vote for that proposal. I hate adding functions to our libraries. But if we are adding these, we probably also need to add: deleteGreater, deleteLess, deleteGreaterEqual, deleteLessEqual Those common operations are also much more expensive when done from outside the library. Wait - horrors! I hadn't looked at the library lately. We now have, for each of Min and Max: find, delete, deleteFind, update, updateWithKey, view, viewWithKey For consistency, we would need to add each of those seven for each of the new comparisons - 28 new functions, for a total of 42! We have finally discovered the Answer to the Ultimate Question of Life, the Universe, and Everything. Why did someone have to add all of those functions to begin with? Look what bloat it's causing. Here are some ideas of what to do about it: 1. Define all 42 of these functions in a separate module. (This has the pleasant advantage of giving us a new bikeshed to paint to name this new module. What fun!) 2. Get rid of all of the functions except just one that is general enough to provide all of the others as a special case. Add copious documentation, with many clear examples, so that anyone can get whatever they need without having to think too much. The general function might have a type something like: data Extremality = Min | Max | Greater | Less | GreaterEqual | LessEqual Extremality -> (k -> a -> Maybe a) -> Map k a -> Maybe ((k, a), Map k a) 2'. As a variation on (2), we could, in addition to the general function, leave in the most commonly used functions in their simple form. (Which ones? Woohoo, more bikesheds!) 3. Use type trickery to construct a single interface that provides each of the possibilities. So that you could write something like update greater withKey and get the function you want. This one is the most fun to write, even though we probably won't end up using it. Regards, Yitz