union, unionBy are not symmetrical in their arguments

[sharkdp]

This seems weird:

union [0, 0] [1, 1] == [0, 0, 1]

I would have expected the result to be [0, 1], but at least it should be symmetrical in both arguments. I would consider several possible fixes:

1. Remove all duplicates (i.e. union a b == nub (a <> b))
2. Keep the duplicates within both of the arguments (i.e. union [0, 0] [1, 1] == [0, 0, 1, 1]).
3. Remove union and unionBy from Data.Array (given that Set is the better data structure for something like this)
4. Document the behavior if it is intended

I would vote for 1 or 3.

[garyb]

It definitely seems weird, although we seem to have inherited it from Haskell where it is documented as preserving duplicates in the first list. I'd vote for 3 or 4.

[sharkdp]

although we seem to have inherited it from Haskell where it is documented as preserving duplicates in the first list

Ah, good point. The same thing applies for intersect.

[garyb]

I can see how this behaviour makes sense for something like (\\) but I'm not so clear why you'd want it for the "set-like" functions - maybe there is a good reason though? Maybe we should indeed ditch all of the pseudo-set functions if we can't come up with a good reason for this unusual behaviour.