doxa-maps/library
doxa-maps-88655c70 3519ca3934 check:packet-browser asserts its headings, and stops reading them in English
This gate printed nine metrics, folded six into `const bad = a || b || ...` — so a
failure said "FAIL" without naming which — and left three unchecked: page size and
the two headings, which are the very strings its own header claims it covers ("the
gate reports which headings came back translated"). Reports, not asserts.

★ IT LOOKED LIKE IT HAD FOUND A PRODUCT BUG, AND HAD NOT. Under ?lang=es it printed
an EMPTY wheel heading; under hi it blanked BOTH — while the catalog holds "Rueda de
Oración" and "प्रार्थना चक्र". That reads exactly like a broken i18n path, and it
PASSED anyway, because nothing asserted the values.

The cause was the harness. wheelTitle was innerText.match(/^.*Wheel.*$/m) and
pointsTitle /Prayer Points|Puntos/ — neither can match Hindi, and /Wheel/ cannot
match Spanish. The product was correct in every locale all along: verified after the
fix in en/es/hi/ar/zh, RTL and Devanagari included. Had I reported that as a defect
I would have been wrong.

mark5 carried the same fault: innerText.includes('Gadarenes'), an English proper
noun that goes RED the moment point4_passage is translated. Now asserted by
.p4-passage length — the same fix already applied in check-packet-render.

Headings are now compared to the CATALOG entry for the packet's language, with
per-key fallback to English exactly as the renderer resolves it. packet-render.js
gains class="points-title"/"wheel-title" as TEST HOOKS: both were inline-styled with
no selector, which is what forced the English greps in the first place.

MUTATION-TESTED: a renderer forced to ignore the language FAILS with both headings
named and the expected values shown; a renderer emitting an empty heading FAILS
("a blank heading is worse than an English one").

★ ONE MUTATION CORRECTLY PASSED, and it marks a real limit: reverting es
wheel_title to English IN THE CATALOG passes, because rendered and expected then
legitimately agree. This gate cannot tell a translated catalog from an untranslated
one — that is check:packet-strings-parity's job, and it does assert it. Two gates,
two questions.

Partial translation re-verified through the BROWSER path: a 1-of-21-key Swahili
block renders "Mambo 12 ya Maombi" with the wheel heading falling back to "Prayer
Wheel" — translated where present, English where not, neither blank.

That makes four gates with one root cause (tabs-localized, packet-render,
deck-click, and this one twice). Any new gate matching on user-visible text should
resolve it from the catalog.
2026-09-24 10:50:27 +00:00
..
api Add the Gospel Access filter, the ladder the binary flag was flattening 2026-09-23 13:19:22 +00:00
colors A validated WAGF region palette, where the one-palette gate requires it 2026-09-24 02:24:08 +00:00
components Restore the clickable poster preview — the last capability the migration dropped 2026-09-24 10:02:22 +00:00
composables MapLibre drops a layer whose paint it does not recognise 2026-09-24 05:38:29 +00:00
constants doxa-maps — portable bundler for embeddable JAMstack maps 2026-09-10 01:53:07 +00:00
data Regions: expand a region to its country posters, reusing the shared wall 2026-09-24 04:26:29 +00:00
i18n Restore the clickable poster preview — the last capability the migration dropped 2026-09-24 10:02:22 +00:00
layouts Merge main into bundler-2/work 2026-09-23 23:48:10 +00:00
maps Load the RTL text plugin from one place, and in the composite too 2026-09-22 19:55:58 +00:00
packet check:packet-browser asserts its headings, and stops reading them in English 2026-09-24 10:50:27 +00:00
scripts The poster print preview rasterises too, and a deployment can ship sources without rasters 2026-09-24 08:07:22 +00:00
stores fix(spa): make embedded maps survive the sandboxed plugin iframe 2026-09-12 21:07:26 +00:00
styles Fiord: the MapLibre dark that actually looks like the map it replaces 2026-09-24 02:26:14 +00:00
utils Regions: one rollup, ported from theirs, instead of two copies 2026-09-24 04:01:34 +00:00
index.js Brand-neutral shared VerticalTabsLayout with lazy nested-element hosting; fix cap-5 prayer tab leaf path; neutralize library identifiers; add Forgejo CI build-gate workflow 2026-09-19 17:24:30 +00:00
ProfileLoader.vue One page, one map token — and the nested search bar comes back 2026-09-24 01:14:37 +00:00
README.md Complete app-profiles->plugin-designs sweep across comments/docs; build green; layouts kept at root (load-bearing) 2026-09-17 18:18:19 +00:00

library/ (@map) — the shared map template

The other half of the architecture (see ../plugin-designs/README.md): every map in plugin-designs/<your-map>/ composes the reusable, parameterized pieces that live here. Fix something here and every map benefits on the next build; break something here and every map breaks — read components/WARNING.md before touching shared pieces.

Folder What lives there
ProfileLoader.vue The root of every bundle: parses profile-config, loads the profile, provides the stores + config
components/ Shared UI — map, legends, controls, popup (PeopleGroupDetail), tables, charts, hints (README)
composables/ Shared logic — map engine (useMapInstance), provider switch, shadow-DOM styles, theming
stores/ Per-instance Pinia stores (ui / map / data) — provided by ProfileLoader, injected by children
api/ Data-source registry (sources.json) + fetch layer — the JAMstack runtime wiring
colors/ Color sets maps opt into via profile-config.colorSet
i18n/ The shared string catalogs (all maps speak the same languages)
data/ Small shared datasets baked at build time (big served data lives in ../data/)
constants/ Shared constants
scripts/ Maintainer pipelines (posters, research) — not part of any bundle

The rule of thumb (same as WARNING.md): custom code in your app-profile, reusable code here — and things only move from the first to the second once they've earned it.