RFC: Hash and Eq requirement for values about left-right HOT 4 CLOSED

Hehe, I have now graduated 🎉 :)

from left-right.

Thanks! Sorry, I didn't realize you were in the middle of your PhD defense. How did it go?

from left-right.

I'm going to close this following the merge of #73 :)

from left-right.

Sorry for the super late reply!

So, this is a tricky one. The hashbag is there very specifically to support efficient removes from value sets, and there it's pretty much impossible to avoid. I think that the workaround here would be to have a complete copy of the implementation that was single-valued (as you observed, this is #45), but that's also a sub-optimal solution.

Now, in your particular case, since you don't need to remove individual values from a value set, you can actually work around this by sticking a newtype around your tuple and just implement Hash + Eq for it with a trivial implementation that always returns true for equality and always hashes to the same value. It feels dirty, but it should work as long as you only ever use that newtype for your thing when it's placed in an evmap value, and as long as you never have more than one value in any given value set. If you do need to have multiple values per value set, but just don't need support for individual removes, you'd have to make Eq return false instead, but that violates the spec of Eq (it's no longer reflexive) and is more likely to cause unexpected issues.

The broader solution here, I think, is to find a way to implement #45 without copying the whole implementation. I'll put a comment over there with some thoughts.

from left-right.