nx_nxa_joints.nx

nx_nxa_joints.nx -- SOVEREIGN STRUCTURAL JOINT IDENTIFICATION over an NXA SKEL. WHY: NXA's SKEL carries NO JOINT NAMES, by design (nx_nxa_rig_emit: "SKEL carries no names by design"; nx_nxa_anim: "names exist in the FBX even though SKEL stays nameless by design"). So a consumer that needs "where is the head / a breast / a thigh" must derive it from GEOMETRY. That derivation already exists -- but ONLY as JavaScript emitted into the browser viewer by nx_nxa_rig_emit. There is no NishiLang function for it, so nothing server-side can ask the question. This is that function, and it is meant to be the CANONICAL one: the JS emitter can later emit the indices this computes instead of re-deriving them in the browser, which removes the cross-language duplicate rather than adding to it (rule 15). METHOD (same shape as the proven JS): height fraction over the rig's own extent + laterality. h01(j) = (y - ymin) / (ymax - ymin) in permil, so 0 = lowest joint, 1000 = highest lat(j) = x relative to the midline, in permil of height (sign = side) Nothing here is a magic anatomical constant: the bands are derived from THIS rig's extent, so a taller/shorter/differently-scaled rig lands in the same normalised space. OUTPUT is the anchor set for nx_arousal_skin_lib's sk_thresh_field_rig -- (x, y, site) triples -- so skin optics can run over a real imported rig instead of only the hardcoded SDF body. Sites (Masters & Johnson propagation order): 0 submammary 1 breast 2 torso 3 face 4 extremities 5 genital. knowledge/arousal_rig_anchors.conf documents the anatomy->site truth this must AGREE WITH; it is the ORACLE for these rules, not their input (it is name-keyed and NXA has no names). usage: nx_nxa_joints <file.nxa> license_tier: ORIGINAL expect_exit: 0