Sets.difference should accept general Collection as second argument

Current signature of Sets.difference method is

SetView<E> difference(final Set<E> set1, final Set<?> set2)

however, for all usages of set2 argument everything would work as well if argument was of type Collection<?>, without impact on performance for current usages. Why not to make signature less strict?

In my use case, I do following manipulation:

BiMap<Foo,Foo> specialFoos = ...
Set<Foo> nonspecialLeftFoos  = Sets.difference(leftFoos ,specialFoos.keySet());
Set<Foo> nonspecialRightFoos = Sets.difference(rightFoos,specialFoos.values());

where this issue is the only reason why specialFoos is a BiMap. Thanks for response.

[jbduncan]

If you really want that behavior you can also always get it with Sets.filter(set, not(in(Collection))).

Or, if you’re on Java 8 and a fan of lambdas: Sets.filter(set, e -> !collection.contains(e)).

[lowasser]

If you really want that behavior you can also always get it with Sets.filter(set, not(in(Collection))).

On Mon, Feb 13, 2017, 9:03 AM Michał Sobkiewicz [e-mail] wrote:

(…) for all usages of set2 argument everything would work as well if argument was of type Collection<?>, without impact on performance for current usages.

Collection.contains(), which Sets.difference() uses a lot, is O(1) for hash based sets and O(log(n)) for tree based sets. You can’t guarantee anything below O(n) in general (List, Queue and so on). Iterating over Sets.difference() would be O(n^2) in that case - that’s not something you want - it’s about O(n) now.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/google/guava/issues/2738#issuecomment-279454201, or mute the thread https://github.com/notifications/unsubscribe-auth/AAhPObEnpQzq_tLYJbM7coIWoy1r034lks5rcIzbgaJpZM4L_Zdk .