David P. Sanders

I agree that we should include the qr test. And we should think about whether we should incorporate the use of three valued logic in the future. We should also...

@gwater How does the 3-valued logic affect performance though? Since that introduces a weak type instability, right?

Now I get ```jl julia> @btime $(Ref(x))[] + $(Ref(y))[] 8.903 ns (0 allocations: 0 bytes) [0.399999, 0.600001] julia> @btime $(Ref(x2))[] + $(Ref(y2))[] 13.377 ns (0 allocations: 0 bytes) [0.399999, 0.600001]...

OK, fair point I guess. In some sense we're punning on the meaning of `in`. But clearly anybody would expect `0.5 ∈ (0..1)` to be true. This is another instantiation...

@JeffBezanson Thanks for the comment -- I totally agree. I would love to remove / improve it, but I don't have any idea how to do so. Do you have...

The thing is it really is a kind of contextual dispatch -- that's even how it's currently implemented, e.g. ```jl julia> +(IntervalRounding{:accurate}(), 0.1, 0.3, RoundDown) 0.39999999999999997 julia> +(IntervalRounding{:none}(), 0.1, 0.3,...

(So the name `IntervalRounding` is actually a misnomer.)

We could do that, but it seems conceptually wrong, for the reason I wrote above (the rounding is a property of the implementation of the operation, not of the intervals...

@mewilhel The problem is that the rounding mode is *not* a property of the intervals, but of the *operations* between intervals. @JeffBezanson Interesting idea. But we worked hard to avoid...

I personally don't like the idea of adding a type parameter to `Interval` to specify the type of directed rounding. But if somebody wants to make a PR or separate...