You know what? In fact we don’t need a RTree at all. Just BBoxes and some canditate BBox questions (but that is what a RTree does anyway [pluts the fact that does N*N comparisons]). Rounding the BBox domains (user controlled option: say in 2 decimals etc helps as well).
See : Mesh Inclusion Error and the useBox magic option. Not 100% the same thing but close. In fact I was ready to use a trick like this … but then you did the RTree thingy and the rest are history.
More in V3 (and some // stuff) … provited that Valentino could deliver some decent results in the Texas MotoGP (for inspiration) . V4 in the mean time would answer the following critical question: given 1M plates full with Swiss cheese and 1Z rats circling around how we can cluster the clouds in the sky on a per rat color basis?
