nx_handcheck.nx

nx_handcheck.nx -- THE HAND STRUCTURE JUDGE over an NXA SKEL (operator 2026-09-01: "the hands are bad"). It is the SIBLING OF nx_footcheck and deliberately mirrors its contract -- check / selftest, one measured report, one predicate made of things a hand IS rather than a blended score -- because two judges for two ends of the same limb must not have two different shapes. WHY IT EXISTS: the hands were the ONE named character defect in this estate with NO instrument at all. nx_clipcheck measures limb-vs-torso and limb-vs-floor; nx_footcheck measures the distal foot; nothing measured a hand. Grading therefore depended entirely on the operator's eye. THE STRUCTURAL PROBLEM, AND WHY THIS IS NOT NAME MATCHING: NXA's SKEL carries NO JOINT NAMES, by design (nx_nxa_joints.nx states it and quotes nx_nxa_rig_emit: "SKEL carries no names by design"). So a hand cannot be looked up -- it must be DERIVED FROM TOPOLOGY. A hand is the only place on a humanoid where three or more TERMINAL SIMPLE CHAINS hang off one attachment. That is what this organ finds. Nothing here knows a joint index, so the judge keeps working when joints are added or reordered ahead of it -- the same structural stance nx_footcheck takes with its rotation. THE ANATOMICAL CANON IS CITED, NOT DIALLED. Buryanov and Kotiuk, "Proportions of Hand Segments", Int. J. Morphol. 28(3):755-758, 2010 (open access; mirror knowledge/fetched/cmp_handcheck_buryanov2010.pdf, sha256 ec2a49be..., 66 adult hands by X-ray with a radiopaque ruler and no scaling error). Their Table I lengths in mm, reconstructed and then INDEPENDENTLY CONFIRMED by their own Table II ratio row for every one of the five rays -- two instruments inside one document agreeing: ray distal medial proximal metacarpal Tab.II ratio (pd:pm:pp:m) I 21.67 -- 31.57 46.22 1 : -- : 1.5 : 2.1 II 15.82 22.38 39.78 68.12 1 : 1.4 : 2.5 : 4.3 III 17.40 26.33 44.63 64.60 1 : 1.5 : 2.6 : 3.7 IV 17.30 25.65 41.37 58.00 1 : 1.5 : 2.4 : 3.4 V 15.96 18.11 32.74 53.69 1 : 1.1 : 2.1 : 3.4 TWO LAWS FALL OUT OF THAT TABLE AND BOTH ARE ENFORCED HERE: (1) MONOTONIC DISTAL SHORTENING -- proximal > medial > distal holds in ALL FIVE rays without exception. It is dimensionless, so it survives any stylisation of hand size. It is the STRICT clause here and it needs no tolerance at all. (2) PHALANX COUNT -- four fingers carry THREE phalanges and the thumb carries TWO (the paper's medial column is empty for ray I, and its text names "even different phalanges numbers"). In a rig a chain of L joints spans L-1 bones, so a finger is a 4-joint chain and a thumb a 3-joint chain. DECLARED IMPRECISION -- name it or the next reader trusts it as exact: - v1 measures the BIND SKELETON. It does NOT read SKIN weights, so it cannot yet say whether the hand is a distinct skinned region or fused to the forearm; and it does NOT read POSE or ANIM, so