Idea: precompute value distributions

[pvdz]

Right now the value distributors are applied during runtime and every step in the search tree has to figure out the next value over and over again. This is simple for min and max, of course, but less so for others.

What if we could somehow precompute the order of resolving?

Sounds good but it's not that trivial;

• You would need to extrapolate all the numbers in the domain to order them properly
  • Expanding the domain to a list of numbers can result in a huge set for domains like [0, SUP] or anything like that.
• Unless you want to copy and reduce this list on every step (which is infeasible for memory reasons) you'd still need to search for the next valid value of the precomputed list that is still part of the domain at the current step. That's not likely to be any faster.