Implement PartialOrd for HashSet

[nham]

cc #15294

Does it make sense to just implement PartialOrd for every type that implements Set? There is a well-defined partial order on sets based on set inclusion. I'm thinking something like this:

impl<T, S: Set<T>> PartialOrd for S {
    fn partial_cmp(&self, other: &S) -> Option<Ordering> {
        match (self.is_subset(other), other.is_subset(self)) {
            (true, true) => Some(Equal),
            (true, false) => Some(Less),
            (false, true) => Some(Greater),
            (false, false) => None
        }
    }
}

[thestinger]

I don't think so. We've been consistent in using PartialOrd and Ord as lexicographic orderings, including for types like TreeSet. You can't have it both ways, and it would be incredibly confusing to have it use one definition for some types (HashSet) and another definition for others (TreeSet).

[nham]

Hmm, you're right. I don't know why I thought this would work. I will nix the Hash implementations for now, I need to study the code more closely.

[nham]

@thestinger Okay, I see what you mean. So to keep it in alignment with the way TreeSet works, for example, we'd need some way to iterate through the HashSet in order, and then we could use the same comparison method that TreeSet uses.

[nham]

Although this seems counterintuitive to me. If I'm comparing two sets, I would expect the usual partial order from math.