nx_codewiki_lib.nx

nx_codewiki_lib.nx -- THE CODE WIKI: an always-regenerable wiki OF the source tree. ===== WHY THIS EXISTS ============================================ The ecosystem has a charter wiki (runtime/wiki/, ~100 curated pages) and a corpus search (nishi_search), but NOTHING that answers, for a given module: what does it do, what does it import, WHO IMPORTS IT, what functions does it export, and where do I look next. That is the Google-Code-Wiki class of artifact: generated FROM the code, never hand-maintained, stale-proof because regeneration is one command and the output is deterministic. It is deliberately a STATIC generator, not a daemon: - the charter wiki daemon is sized for ~100 docs (docs_cap) and serves public charter sites; 17k code pages do not belong in its pools - static bytes are gateable (byte-identical re-run), diffable, and servable by any route -- or browsable as local files with zero infra ===== WHAT IS REUSED (the dupfunc law: the tool must not duplicate) ====== gl_walk / gl_read (nx_gatelib) -- the ONE tree walk + file read nis_scan (nx_import_scan) -- the ONE accurate import scanner EcoGraph (nx_eco_graph) -- intern + CSR fan-in/fan-out gl_walk's edge mode is NOT used: it interns import paths verbatim, so a subdir import ("wiki/nx_foo.nx") would become a node DISTINCT from the basename node of the actual file and every such link would dangle. We intern the BASENAME of each import ourselves -- same convention as file nodes -- so links resolve, and what still cannot resolve is reported as an honest unresolved count instead of a dead href. ===== ENVELOPES (declared, per the scale law) ===================== CW_MAXNODE 24000 nodes | CW_EDGECAP 262144 edges | file cap GL_FILECAP CW_PAGECAP 4 MiB/page -- overflow appends a LOUD PAGE-TRUNCATED banner CW_MAXDIRS 256 | desc arena 2 MiB | per-file import dedup window 128 license_tier: ORIGINAL