nx_canoncat.nx

nx_canoncat -- concatenate N canon files into one, renumbering feature part references. ★THE ENTIRE "PERFECT ONE UNIT, THEN AGGREGATE" MECHANISM IS THIS ORGAN (plan rung S4). A skull, a hand, a torso are each authored and gated as their OWN canon; composition into a body is a cat. That is only true because of one discipline enforced here: P and R rows pass through BYTE-VERBATIM. This organ does not parse them beyond the leading letter, so when the canon grows fields (P grew mat, then rotZ, then rotY) concatenation keeps working without an edit here -- an organ that re-emitted parsed fields would silently STRIP any field added after it was written, which is exactly how format evolution corrupts a pipeline. F rows are the one exception, because they carry the ONLY cross-row reference in the format: a part INDEX, where parts are numbered by order of P appearance. File k's F rows are shifted by the P count of files 0..k-1. Everything after the index is passed through verbatim for the same reason as above. Fail-closed on purpose: - an F row referencing a part its OWN file never declared is a corrupt canon -> refuse (exit 5), because shifting a dangling index would turn one file's corruption into another file's feature. - an unknown row type -> refuse (exit 6). A row this organ cannot classify is a row whose renumbering rule is UNKNOWN; passing it through silently would be a guess with a file format. usage: nx_canoncat <out.canon> <in1.canon> [in2.canon ...] nx_canoncat selftest