[cwhite911]

Added a new lib/octree/ spatial indexing library for 3D points. The octree recursively subdivides space into octants with adaptive subdivision and a depth limit for coincident points. It supports point insertion, axis-aligned range queries, depth-limited traversal, cached subtree point counts, and subtree centroids.

The areas I was thinking about when writing this are, but I'm sure there are many other use cases as well.

• LiDAR ingest (v.in.pdal)
• v.vol.rst
• spatial decimation (v.decimate)
• level-of-detail (LOD) rendering of point clouds in the NVIZ GUI, where loading everything into memory currently bogs the system down.

[petrasovaa]

Are you using it for something specific now? Do you plan to integrate the library into one of the suggested tools?

[metzm]

How about adding some id to OctreePoint3D? With such an id, it would be easy to retrieve the corresponding points together with any attributes from the actual points vector when querying the octree. Or is this covered with any provided user_data?

[cwhite911]

How about adding some id to OctreePoint3D? With such an id, it would be easy to retrieve the corresponding points together with any attributes from the actual points vector when querying the octree. Or is this covered with any provided user_data?

@metzm you're right, user_data doesn't cover it. It's a per-query callback parameter, not a per-point attribute, so there was no way to map a queried point back to its source. I added a caller-defined grass_int64 id to OctreePoint3D that can be set to a vector cat or a source index or leave it OCTREE_NO_ID.

[cwhite911]

Are you using it for something specific now? Do you plan to integrate the library into one of the suggested tools?

Yes. The octree will be used for new 3D tiling capabilities.