points.neighbors (procNode)

What stands around each point, within a radius and optionally a cone: neighbour count, nearest distance, centroid, and the neighbours' mean of an attribute. Use it for crowding, flocking reads and density-driven scale.

core.neighbours in measure mode, which finds each point's neighbours through a spatial hash and writes neighbors, nearest, centroid and, when mean names an attribute, neighbor_<name>. A point with no neighbour has neighbors 0, nearest equal to radius and its own position as centroid. A radius at or below 0, and a direction or mean the points do not carry, are errors.

Built from: core.neighbours.

Inputs

• points (Geometry). The points to look around. Each one's neighbours are the others of the same points; the answers land on it as neighbors, nearest and centroid, and everything it carried stays.
• radius (Float, default 1, at least 0.000001). How far a point looks. Only points closer than this count as its neighbours.
• direction (String, optional). A Vec3 point attribute each point looks along (a velocity, a tangent). Given, a neighbour counts only inside the cone around it.
• cone (Float, default 360). The full angle, in degrees, of the cone around direction a neighbour has to stand in. 360 counts every side.
• mean (String, optional). A numeric point attribute (1 to 4 lanes) whose neighbours' average each point takes, as neighbor_<name>. A point with no neighbours takes its own value.

Outputs

• points (Geometry). The points with neighbors (count), nearest (distance to the closest neighbour, at most radius), centroid (Vec3) and neighbor_<mean> added.

local around = g.points.neighbors({ points = g.points.grid({ countX = 4, countZ = 4 }), radius = 1.1 })