nx_voronoi.nx

nx_voronoi.nx -- 2D Voronoi diagram (procgen LAYER 3). Sites are scattered in [0, w) x [0, h); each cell of the canvas is assigned to its NEAREST site by squared Euclidean distance. Output is a per-pixel label array (which site index owns each pixel). USE CASES: - City-block partition for procgen city generation (each block becomes a Voronoi cell around its district seed) - Zombie-escape agent territories (each spawn point owns the region closest to it) - Reaction-diffusion / cellular-automaton substrate (cell shape) - Crystal-grain texture for procgen surfaces - Lloyd relaxation seed for centroidal Voronoi tessellation ALGORITHM: Brute-force nearest-site per pixel. For each pixel (x, y), scan all n_sites and pick the index with minimum (dx^2 + dy^2). Cost: O(w * h * n_sites). Patent-clean (squared-distance is textbook). SCALING NOTE: Fortune's sweepline (1986) and Aurenhammer (1991) give O(n log n) for the EDGES of the diagram, but per-pixel labelling for a raster canvas is still O(w*h*n) without spatial acceleration. For substrate use (canvas <= 256x256, sites <= 256), brute-force is fast under qemu-riscv64 -- millions of i64 ops, no f64, no sqrt. Upgrade path: jump-flooding-algorithm (JFA, Rong+Tan 2006) for GPU-equivalent O(w*h*log(max(w,h))) when canvas grows. QUALITATIVE BAND (dual-reading cardinal): Per-site cell-area sealed enum classifies how uniform the partition is, derived from variance of per-site pixel count. COMPOSITION: sites can be supplied by any source -- in practice nx_poisson_disk_sample gives the most visually-pleasing uniform-but-not-grid distribution. Demo wiring in nx_voronoi_test.nx. genealogy_id: voronoi_1908 + dirichlet_1850 + fortune_1986 +