Add a Set type matching the std HashSet API

[cuviper]

Just like in std, it can start with the simple definition wrapping a map:

pub struct FlurryHashSet<T, S = RandomState> {
    map: FlurryHashMap<T, (), S>,
}

[jonhoo]

The Java code's KeySetView may also be of interest.

[jonhoo]

We now have a basic HashSet type, but it'd be great to get something along the lines of HashMapRef for HashSet too!

[GarkGarcia]

We now have a basic HashSet type, but it'd be great to get something along the lines of HashMapRef for HashSet too!

Is that a case of simply wrapping HashMapRef<T, ()> in a newtype? I could work on that if that's the case.

[twe4ked]

I think you'd want to wrap HashSet with HashSetRef similar to how HashMap is wrapped by HashMapRef.

[GarkGarcia]

I think you'd want to wrap HashSet with HashSetRef similar to how HashMap is wrapped by HashMapRef.

That's interesting. I'm not quite sure which way would be the most efficient memory wise. @jonhoo @cuviper opinions?

[GarkGarcia]

We'd be wrapping a wrapper either way...
I think wrapping HashMap<T, ()> directly should make the type's layout flatter, but I'm not sure.

[jonhoo]

Ah, no, I was imagining we just do the same trick with HashSetRef as with HashMapRef, not that we wrap HashMapRef to make HashSetRef!

[GarkGarcia]

Ah, no, I was imagining we just do the same trick with HashSetRef as with HashMapRef, not that we wrap HashMapRef to make HashSetRef!

Makes sense.

[GarkGarcia]

@jonhoo What about the wrapping HashMap<T, ()> directly proposal?

[jonhoo]

I think we should wrap HashSet, because HashSet actually has different methods than HashMap, and having the wrapper also do the translation from set-like methods to map-like methods (which HashSet is already doing) seems like unfortunate duplication.

[jonhoo]

Closed by #78!