Removes splinepy coordinate type based clouds and trees, as a preparation step for more dynamic proximity query.
Removes dim as tparam, which makes the tree even more flexible and reduces explicit instantiation dramatically. This also has a benefit that the tree dimension is not limited by instantiated dimension. More importantly, the tree's dim is determined during init, so shouldn't have much effect.
For L2_metric, we only use nanoflann::L2_Simple_Adaptor. Unlike L2_Adaptor, it computes metric with simple for loop. It has been told that kdtree is not so efficient in high dim, and if anything L2_Adaptor is useful for dim > 4. Thus, the decision to use only L2_Simple_Adaptor
nanoflann::L2_Simple_Adaptor
. UnlikeL2_Adaptor
, it computes metric with simple for loop. It has been told that kdtree is not so efficient in high dim, and if anythingL2_Adaptor
is useful for dim > 4. Thus, the decision to use onlyL2_Simple_Adaptor