App country dimension: surface the UK end-to-end

[vahid-ahmadi]

Closes #42.

What was done

Data contract (data/lanes.json + committed copy app/public/data/lanes.json, kept identical)

• Every lane now carries an explicit "country" field — "US" for existing lanes, "UK" for dwp-takeup, hbai-poverty, hmrc-personal-tax, obr-welfare, ukmod-stats, ukmod-cases — documented in the _schema block. The diff adds only country lines (plus the writer-canonical trailing newline); stage/note/updated untouched. sync_lane_feed merges by key, so pipeline rewrites preserve the field.

App types (app/src/types.ts)

• New Country type ("US" | "UK"), COUNTRY_LABELS, and a single countryOf() helper: a missing country key always means US, so US-era comparison/populations exports never drop a row. Lane, Row, and PopulationRow gain the optional country field.

Scoreboard views (App.tsx)

• US/UK toggle in the header scopes the coverage spine, headline, scorecard, divergences and gaps views; switching resets row filters (geography codes don't carry across countries).
• Empty states: with UK lanes mid-pipeline, the UK view renders a status panel listing each UK lane and its stage from lanes.json — the country is never hidden. The same panel serves any future country with zero rows.
• Mission control lists every UK lane with its current stage (a "UK pipeline" block that will show stage transitions as lanes move), and running lanes carry a country chip when non-US.
• ComparisonTable/DivergenceBoard/MissionControl generalize the hardcoded "US" national-geography checks to the row's country code, so UK national rows land in the right buckets when they arrive.
• Status taxonomy untouched: bucketOf, SPINE_META, and the status chips are country-blind, so pe_gap and concept_mismatch UK cells will render exactly like US ones and stay on the page.

Reform-validation tab

• States its scope explicitly while single-country: "scope: United States only — … a country filter appears here once claims from another country land". When the populations feed contains a second country, a Country select appears automatically and filters by countryOf.

What remains

• No UK comparison/populations rows exist yet on main — the UK path currently exercises the empty states and lane listing; the pipeline exports (build_comparison/export_populations) still need to emit country on rows when UK claims land (other branches carry "country": "UK" externals).
• The scorecard table's external column header/vintage chips are still Urban-specific; they should key off per-country source metadata once a UK instance ships.
• The page-top provenance stamp describes the Urban instance only (noted in-app).

Verification

• bun install --frozen-lockfile && bun run lint && bun run build in app/: lint clean (no warnings), tsc -b && vite build pass.
• PYTHONPATH=. uv run --with pytest pytest tests/ -q from repo root: 127 passed, 4 skipped.
• data/lanes.json and app/public/data/lanes.json verified byte-identical.

🤖 Generated with Claude Code

[MaxGhenis]

Dual-gate review: HOLD on the UK feed-architecture adjudication (#48 vs #32), by design — PopulationRow.country expects a top-level field no current exporter emits (UK claims would silently default US via countryOf()), and the singleton source_meta/pe_bundle provenance stamp can't faithfully represent mixed-country feeds. Verified safe otherwise: missing country defaults US, all 30,004 legacy comparison rows + 270 population rows retained, both lane mirrors preserve all 30 lanes. Re-gate once the ingest architecture lands.

[DTrim99]

Review — App country dimension (UK end-to-end)

No backward-compat/data-loss bug — the load-bearing contract holds: countryOf() (types.ts:22) defaults a missing country to "US", and the row-scoping filter (App.tsx) plus the Divergence/MissionControl national checks were all migrated from === "US" to === countryOf(r), so no legacy US row/lane/population export is dropped. The two committed lanes.json copies match in the diff, and the toggle handles empty-country states via CountryEmptyState.

Should address

• AttributionPanel.tsx:39,43 still hardcodes geography === "US" / !== "US". This panel is rendered by DivergenceBoard (which now receives UK-scoped data), fetches its own US-only exhibits.json, and ignores the toggle — so under the UK view it renders US attribution unconditionally. It's a separate feed (not a dropped-row bug), but it's the one filter site the PR left unconverted; gate it by country or label it US-only.
• LANE_FEED append path mints country-less entries (ingest_harvest.py:104, entry = {"id": id, **meta}). sync_lane_feed preserves country on rewrite of existing lanes (good), but any new lane added via DB-append silently defaults to US — so UK lanes must be seeded in lanes.json directly, not via LANE_FEED.
• No app tests — countryOf(), the scoped filter, and empty-state paths are untested. At minimum a countryOf() unit test.

Suggestion

• The committed app/public/data/lanes.json copy can drift between builds since nothing verifies equality in CI (the predev/prebuild cp regenerates it, but there's no check). A CI equality check or .gitignore on the generated copy would remove the manual-sync hazard (pre-existing, not introduced here).

Reviewed with Claude Code assistance.

[vahid-ahmadi]

All three should-address items fixed in c5c17e7:

1. AttributionPanel now takes the selected country (threaded filters.country → DivergenceBoard → panel) and renders nothing outside US — exhibits.json is a US-only feed (SPM poverty, state geographies), so gating beats labeling: the UK view shows no US attribution at all.
2. Country-less lane appends refused: sync_lane_feed raises if a newly appended entry's meta lacks country; every LANE_FEED entry and the populace-reform-validation caller now carry "country": "US" explicitly. Pinned by two tests (all-entries-carry-country + the append-path raise).
3. App tests: bun test src runs a countryOf() suite (missing-key defaults to US, explicit pass-through, and the scoping filter never dropping US-era rows). Test files are excluded from the tsc -b build graph so the CI build stays clean.

Suite: python 129 passed / 4 skipped (was 127); bun run lint + bun run build + bun test src all pass.

The lanes.json drift suggestion is left as-is for now (pre-existing, as you noted — predev/prebuild regenerate the copy); happy to add a CI equality check in a follow-up.

The #57-adjudication HOLD stands — these fixes just clear the re-gate queue for whenever the ingest architecture lands.

[DTrim99]

Re-review — all addressed ✅

• AttributionPanel now takes { country } and returns null unless country === "US" (US-only exhibits.json feed) — no more unconditional US attribution under the UK toggle.
• LANE_FEED append path now raises when an entry lacks country, and every entry is tagged.
• Tests added (types.test.ts: countryOf() default/passthrough + the scoping filter) with a bun test script.

Only cosmetic TS-typing nits left (a string-vs-Country index in ReformValidationView); the lanes.json dual-copy CI-equality check remains a pre-existing nice-to-have. Good to go from my side.

[DTrim99]

Re-reviewed — all findings addressed (AttributionPanel gated by country, LANE_FEED raises on missing country, types.test.ts added). Approving; the remaining TS-typing/lanes.json-dup items are cosmetic/pre-existing.

[MaxGhenis]

Heads-up: the UK foundation landed on main today, so this branch's data surfaces have moved under it.

What changed that touches this PR:

• UK externals → DB ingest path (16,924 claims, five sources) #48 merged — the five UK sources are ingested (15,851 claims + 1,073 Ledger admin-outturn facts), and data/lanes.json now carries five UK lanes at ingested (dwp-takeup, hbai-poverty, hmrc-personal-tax, obr-welfare, ukmod-stats).
• UC deductions: FRR family re-harvested from primary sources + DB ingest #52 merged — the FRR deductions family (7 claims, sources hm_treasury + dwp) and a sixth UK lane, uk-deductions-frr. Also a change worth knowing about when hacking on the app locally: opening the DB no longer rewrites it (the comparisons view used to be dropped/recreated on every open, which dirtied the committed blob for any reader — ScorecardDB now recreates it only on definition drift).
• Campaign-UK staging producer: archive -> resolved claim-id joins + 14 reckoner attaches #72 (in final gate) — 14 UK reckoner claims get PE results (run_id campaign-20260802-reckoner-t2), and data/populations.json goes 270 -> 284 rows, so the Reform validation tab will have UK rows with a constructed effective status from the cross-baseline guard (PE current-law result vs the claims' HMRC indexed baseline).

Suggested sequence: once #72 merges (imminent), rebase onto main — export_populations.py, ingest_harvest.py, and both lanes.json copies will conflict, and main's versions carry gate-hardened behavior (the exporter's per-result annotations/baseline labels, sync_lane_feed's lanes parameter contract), so resolving toward main and re-applying the country dimension on top is the safer direction than keeping this branch's versions. After the rebase we'll run the standard dual gate (fable + sol) on the head.

🤖 Generated with Claude Code

[MaxGhenis]

Rebased onto main in place (your two commits intact, one completion commit on top) — the foundation had moved four merges under this branch since it forked, so I did the mechanical part rather than leaving it to go stale:

• The 3-way merge carried your country tags cleanly onto main's 31-lane feed; the one gap was uk-deductions-frr (merged today, so your branch couldn't have known it) — tagged UK.
• Your sync_lane_feed append guard is exactly right, and it exposed that the UK ingests' own appended lane metas (the five UK externals → DB ingest path (16,924 claims, five sources) #48 lanes + the FRR lane) didn't carry country — they do now, so a fresh-feed append can't file a UK lane under the app's missing-key US default.
• Took your "nice-to-have" too: app/public/data copies refreshed (they had drifted to a 270-row populations.json vs data/'s 284) plus two pinning tests — committed copies must byte-match data/, and every committed lane must carry US|UK.

Suite: 241 python + your 3 bun tests; oxlint + vite build clean; CI green. Dual gate (sol pass) running next — merges on MERGE-SAFE per the standing sequence. Shout if anything in the completion commit doesn't match your intent.

🤖 Generated with Claude Code