# Index of Nearest: Persistent Cache.

*In essence, this task is about storing in the mesh not only the BVH tree used to sample the nearest node, but also the KD tree in order to use a more complex search.*


The current design is very generic.
It can be used for both simulations and modeling.

However, the simulations imply that very similar search trees are often rebuilded.
If we mean the general search tree (without groups, for all domain), then this tree can be stored in a mesh (or in other component). In this case, its reuse, in theory, can be cheaper (A potential update of positions should be cheaper than a full rebalancing).

This implies the addition of a separate node operation mode: Without Groups (the name is in doubt).
Specialized for simulation without splitting into groups, with data stored in the geometry itself.