doxa-maps/public/.gitignore
doxa-maps-88655c70 7d0b0e763e
Some checks are pending
Deploy to GitHub Pages / build (push) Waiting to run
Deploy to GitHub Pages / deploy (push) Blocked by required conditions
PPLR data maps: ten research views of the PPLR people-group data join Experimental
The PPLR-validate-data maps (religion, ROP tree, language families, semantic
trees, PG name locales, ROP sunburst, all pins, cross-system ID rosetta,
translation coverage, index) shipped only as a frozen 339 KB IIFE on the old
dx-life research page, registering <doxa-map> and colliding with doxa-simple-map.
They now build from source as plugin-designs/pplr-data-maps: one monolithic
<pplr-data-maps> with a view switcher. The Research Gallery discovers it into
Experimental.

Data: build-data.mjs derives data/pplr from the source's 78 MB: 7.3 MB of pins
without the 17 locale names, which load from pg-names.json only when a pin's
detail card opens, plus a precomputed coverage table. The tree stages it inside
itself; the web-root mirror is ignored like the other data/ mirrors.

Measured on a served built tree in catalog layout: the leaf page, all ten views
script-embedded on another origin, and the Gallery tab each render with 0 page
errors and 0 failed requests. Coverage shades 241 of 251 countries.
2026-09-25 07:50:10 +00:00

38 lines
1.8 KiB
Text

token.local.js
# The build MUTEX's stamp file (internal/build-lock.sh writes it under flock).
# It is pure runtime state — a pid, a timestamp and `by=$(whoami)` — rewritten on
# every `bun run build`. It was tracked, so each rebuild committed the BUILDING
# MACHINE'S USERNAME into the repo. Nothing reads it but the lock itself, and a
# clone that inherits a stale pid is at best noise.
.build.lock
# ── GENERATED MIRRORS OF data/ — 80.7 MB committed TWICE ─────────────────────
# generate-tree.mjs:1353 copies data/<name> -> public/assets/<name>, rmSync'ing
# the destination first, on EVERY `bun run tree`. So these five directories are
# not a source: they are a byte-identical copy of a tracked source, and both
# copies were committed.
#
# MEASURED 2026-09-24, all five verified identical file-set and hash:
# posters 378 files 40.9 MB
# people-groups 2 files 23.0 MB
# tables 19 files 14.2 MB
# legacy 4 files 2.0 MB
# vendor 3 files 0.6 MB
# Proven regenerable, not assumed: deleting public/assets/posters and running
# `bun run tree` restored all 378 files with laos.webp md5 ea65d2f1 — the same
# hash the committed copy had.
#
# This is the "80mb of images in a public folder" directly: a publish was already
# rejected once at a 74 MB asset tree, and half of that weight was a duplicate of
# something already in the same repo.
#
# ⚠ previews/ IS NOT LISTED AND MUST STAY TRACKED. It has no data/ source — it is
# generated by make-previews.py from the full posters — and it is what the poster
# WALL renders from, so a dropped folder needs it (bundlePosters defaults false,
# which ships previews and not posters). Untracking it would blank every wall.
/assets/posters/
/assets/people-groups/
/assets/tables/
/assets/pplr/
/assets/vendor/