kaafil-react-uikit 0.11.8 → 0.13.0

package/CHANGELOG.md

@@ -6,6 +6,86 @@ The format follows [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 and this package uses [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 While on `0.x` a minor bump may carry a breaking change; the entry will say so.
 
+## [0.13.0] — 2026-09-21
+
+### Fixed
+
+- **The manifest's diet figures were wrong, and quietly so.** `normalizeDiet`
+  existed as three verbatim copies in three families, and all three tested raw
+  EXACT EQUALITY against `V|VEG|VEGETARIAN` / `NV|NON-VEG|NONVEG|NON_VEG` /
+  `J|JAIN` over `Traveller.dietary` — a free-text column the CRM supplies, whose
+  own published example is `"Vegetarian, no nuts"`.
+
+  Case was handled; nothing else was. `"Pure veg"`, `"Vegetarian, no nuts"`,
+  `"Vegan"`, `"Halal"` and `"Non veg"` (space, no hyphen) all normalised to
+  `null`, which the UI rendered identically to "nothing recorded". A 20-pax trip
+  with twelve vegetarians spelled five ways showed a **`Veg 4 / 20`** stat card,
+  and the kitchen ordered four meals. The filter chips did not sum to the roster
+  and offered no bucket the other eight could appear in, so tapping "Veg" to
+  build a meal list silently excluded them.
+
+  There is now one implementation in `core/`, with 34 tests.
+
+- **A traveller's dietary text could vanish from their profile.**
+  `deriveDietaryRows` pushed the Meal row with `value: mealPreference ?? ''`, so
+  when the CRM sent only `dietary`, the raw string was never rendered: a
+  `"Vegetarian, no nuts"` traveller showed a `V` badge and the nut allergy
+  appeared nowhere on the screen. The raw value is now shown when no separate
+  meal preference exists.
+
+### Changed
+
+- **BREAKING — `ManifestDiet` is now `'V' | 'NV' | 'J' | 'OTHER'`.** An
+  exhaustive `switch` or a `Record<ManifestDiet, …>` in consumer code will no
+  longer compile until it handles `OTHER`. This is the fix, not a side effect:
+  the previous three-state set had nowhere to put a recorded-but-unrecognised
+  value, so it used `null` and the value disappeared. `null` now means only that
+  the column is empty, which is what makes the counts account for every
+  traveller.
+
+  `VEGAN` and `HALAL` bucket as `OTHER` rather than `V`. Folding them into the
+  vegetarian count would report them as served by the vegetarian meal and remove
+  the only signal that something further is needed.
+
+- **A new "Other" diet filter chip**, with its `en`/`hi` labels, so the chip
+  counts plus the unrecorded remainder account for the whole roster.
+
+- **`diet` is gone from the agency settings screen's rooming strategy picker.**
+  It was the last place a human could still select a heuristic key the engine had
+  stopped accepting.
+
+- **Peer floor moves to `kaafil-js@^0.9.0`**, which is why this is a minor rather
+  than a patch on that count too. The engine no longer accepts `diet` in a
+  rooming `strategyOrder`, and `kaafil-js@0.8.0`'s `RoomingHeuristicKey` still
+  lists it — so on the old floor a consumer could write code that typechecks and
+  is then refused at runtime. Following the precedent 0.12.0 set: when the fix
+  lives elsewhere, the floor is how a consumer actually gets it.
+
+### Notes
+
+- The bucket rules mirror the engine's own `dietBucketOf`, and were verified to
+  agree with it on 32 inputs including every edge case in the corpus. The engine
+  additionally publishes `facets.dietBuckets` and a `?dietBucket=` filter; this
+  package still buckets client-side and does not require either.
+
+## [0.12.0] — 2026-09-18
+
+### Changed
+
+- **Peer floor moves to `kaafil-js@^0.8.0`**, which is why this is a minor and
+  not a patch: a consumer pinned to `0.7.x` can no longer install this version.
+  The previous floor move was shipped as a patch; that understated it, and this
+  entry says so rather than repeating it.
+
+  0.8.0 makes a drain pull the lanes it applied something on. That closes the
+  gap 0.11.8 documented under "Known": a write enqueued without an
+  `entity`/`optimistic` pairing reaches the server and leaves the local
+  snapshot untouched, so a snapshot-backed list stayed missing its own
+  successful write until the screen was reopened. The Recent expenses list on
+  the manager's Money tab was the case found in the field. Nothing in this
+  package changed for it — the fix is in the engine, and this floor is how a
+  consumer actually gets it.
+
 ## [0.11.8] — 2026-09-18
 
 ### Fixed