Bisection should ignore atomic intervals

[dpsanders]

No description provided.

[dpsanders]

And return an indication of whether the intervalbox is atomic (ie atomic on each direction)

[dpsanders]

Note that a box is atomic only if each interval in the box is.

[dpsanders]

Probably it should return (X, ∅) if X is atomic.

[Kolaru]

Note that a box is atomic only if each interval in the box is.

This is true, but we will have to be careful: the non atomic dimension may not be the biggest one.

julia> X = IntervalBox(0..1, 10.0^100..nextfloat(10.0^100))
[0, 1] × [9.99999e+99, 1.00001e+100]

julia> bisect(X)
([0, 1] × [9.99999e+99, 1.00001e+100], [0, 1] × [1e+100, 1.00001e+100])

Also this doesn't seem to be part of the standard (freedom !), and I like the (X, ∅) proposition.

[dpsanders]

Good point! That will be tricky to handle.

I guess broadcast isatomic over the IntervalBox and bisect the first dimension that is not atomic.