366 lines
15 KiB
TypeScript
366 lines
15 KiB
TypeScript
/**
|
||
* scripts/regions.test.ts — validation for the committed `public/regions.geo.json`.
|
||
*
|
||
* The dataset is REAL named sub-national administrative regions (Natural
|
||
* Earth admin-1: states, provinces, governorates, oblasts, …) that the
|
||
* coarsening step (`scripts/coarsen-regions.mjs`, v0.14) has merged down to a
|
||
* ~100×100 km (10,000 km²) floor: every small non-island region was dissolved
|
||
* into its nearest same-country neighbor, so the only regions left below the
|
||
* floor are islands (no same-country sibling nearby). One feature per named
|
||
* region, each carrying `properties = { name, parent }` where `parent` is the
|
||
* country's ISO_A3 (the RegionMap key the map resolves ownership through).
|
||
* Tiny or uncovered countries collapse to a single whole-country region
|
||
* (`<ISO_A3>-01`).
|
||
*
|
||
* This test is self-contained (no import of the generator): it re-implements
|
||
* the same planar area model the generator/analysis use, so the area-based
|
||
* assertions measure the same quantity:
|
||
*
|
||
* x = lon · 111.32 · cos(midLat), y = lat · 110.574 (km)
|
||
*
|
||
* with a shoelace per OUTER ring, each ring measured at its own mean latitude.
|
||
* Outer rings only — a region's inner rings (holes) are lakes and are ignored
|
||
* by design, the same way the coverage band already absorbed them.
|
||
*
|
||
* It reads each file exactly once and runs in O(total vertices).
|
||
*/
|
||
import { readFileSync } from 'node:fs';
|
||
import path from 'node:path';
|
||
import { fileURLToPath } from 'node:url';
|
||
import { describe, expect, it } from 'vitest';
|
||
|
||
const here = path.dirname(fileURLToPath(import.meta.url));
|
||
const root = path.resolve(here, '..');
|
||
|
||
// --- area model (planar approximation; matches scripts/generate-regions.mjs) --
|
||
const KM_LON = 111.32;
|
||
const KM_LAT = 110.574;
|
||
const rad = (deg: number) => (deg * Math.PI) / 180;
|
||
|
||
type Ring = [number, number][];
|
||
type Geometry =
|
||
| { type: 'Polygon'; coordinates: Ring[] }
|
||
| { type: 'MultiPolygon'; coordinates: Ring[][] };
|
||
|
||
function ringAreaKm2(ring: Ring, midLat: number): number {
|
||
const kx = KM_LON * Math.cos(rad(midLat));
|
||
let a = 0;
|
||
const n = ring.length;
|
||
for (let i = 0; i < n; i++) {
|
||
const [lon1, lat1] = ring[i];
|
||
const [lon2, lat2] = ring[(i + 1) % n];
|
||
a += lon1 * kx * (lat2 * KM_LAT) - lon2 * kx * (lat1 * KM_LAT);
|
||
}
|
||
return Math.abs(a / 2);
|
||
}
|
||
|
||
function outerRings(geom: Geometry): Ring[] {
|
||
if (geom.type === 'Polygon') return [geom.coordinates[0]];
|
||
if (geom.type === 'MultiPolygon') return geom.coordinates.map((p) => p[0]);
|
||
return [];
|
||
}
|
||
|
||
/** Sum of a feature's outer-ring areas, each ring at its own mean latitude. */
|
||
function featureAreaKm2(geom: Geometry): number {
|
||
let total = 0;
|
||
for (const ring of outerRings(geom)) {
|
||
let lat = 0;
|
||
for (const [, la] of ring) lat += la;
|
||
total += ringAreaKm2(ring, ring.length ? lat / ring.length : 0);
|
||
}
|
||
return total;
|
||
}
|
||
|
||
function median(nums: number[]): number {
|
||
if (!nums.length) return 0;
|
||
const s = [...nums].sort((a, b) => a - b);
|
||
const m = Math.floor(s.length / 2);
|
||
return s.length % 2 ? s[m] : (s[m - 1] + s[m]) / 2;
|
||
}
|
||
// ------------------------------------------------------------------------------
|
||
|
||
interface RegionFeature {
|
||
type: 'Feature';
|
||
id: string;
|
||
properties: { name: string; parent: string };
|
||
geometry: Geometry;
|
||
}
|
||
interface CountryFeature {
|
||
type: 'Feature';
|
||
id: string;
|
||
properties: { name: string };
|
||
geometry: Geometry;
|
||
}
|
||
|
||
// Read each file exactly once.
|
||
const regions = JSON.parse(
|
||
readFileSync(path.join(root, 'public', 'regions.geo.json'), 'utf8'),
|
||
) as { type: 'FeatureCollection'; features: RegionFeature[] };
|
||
const countries = JSON.parse(
|
||
readFileSync(path.join(root, 'public', 'countries.geo.json'), 'utf8'),
|
||
) as { type: 'FeatureCollection'; features: CountryFeature[] };
|
||
|
||
// Precompute once (O(total vertices)).
|
||
const countryIds = new Set(countries.features.map((c) => c.id));
|
||
const countryArea = new Map<string, number>();
|
||
for (const c of countries.features) countryArea.set(c.id, featureAreaKm2(c.geometry));
|
||
const regionArea = new Map<string, number>();
|
||
for (const f of regions.features) regionArea.set(f.id, featureAreaKm2(f.geometry));
|
||
const sumByParent = new Map<string, number>();
|
||
for (const f of regions.features)
|
||
sumByParent.set(f.properties.parent, (sumByParent.get(f.properties.parent) ?? 0) + (regionArea.get(f.id) ?? 0));
|
||
const countByParent = new Map<string, number>();
|
||
for (const f of regions.features)
|
||
countByParent.set(f.properties.parent, (countByParent.get(f.properties.parent) ?? 0) + 1);
|
||
|
||
// --- island/close-neighbor helpers (mirror scripts/coarsen-regions.mjs) ------
|
||
|
||
interface LonInt {
|
||
a: number;
|
||
b: number;
|
||
full: boolean;
|
||
}
|
||
interface RegionBBox {
|
||
latMin: number;
|
||
latMax: number;
|
||
lon: LonInt;
|
||
}
|
||
|
||
const COARSE_FLOOR_KM2 = 10_000; // the coarsener's ~100×100 km floor
|
||
const CLOSE_DEG = 0.2; // "spatially close" bbox gap (both lat and lon ≤ this)
|
||
|
||
/**
|
||
* Unwrap a ring's longitudes to a continuous line (np.unwrap shortest-diff
|
||
* scheme, anchored at the first vertex's [0,360) value). `full` when the
|
||
* unwrapped span is ≥ 300° (a full-360° antimeridian band, e.g. Antarctica).
|
||
* Must stay in sync with the coarsening script.
|
||
*/
|
||
function unwrapRing(ring: Ring) {
|
||
const n = ring.length;
|
||
const norm = (x: number) => ((x % 360) + 360) % 360;
|
||
const u: number[] = new Array(n);
|
||
u[0] = norm(ring[0][0]);
|
||
for (let i = 1; i < n; i++) {
|
||
// the previous UNWRAPPED value may lie outside [0,360); normalize it
|
||
// before measuring the shortest signed difference, or a bogus ~180° jump
|
||
// is inserted and the ring's span balloons.
|
||
const a = norm(u[i - 1]);
|
||
const b = norm(ring[i][0]);
|
||
const d = (((b - a) % 360) + 540) % 360 - 180;
|
||
u[i] = u[i - 1] + d;
|
||
}
|
||
let min = Infinity;
|
||
let max = -Infinity;
|
||
for (const v of u) {
|
||
if (v < min) min = v;
|
||
if (v > max) max = v;
|
||
}
|
||
return { lons: u, full: max - min >= 300 };
|
||
}
|
||
|
||
/**
|
||
* Cyclic 1-D gap (degrees) between two tight longitude intervals in their own
|
||
* unwrapped spaces; a `full` interval is close to everything.
|
||
*/
|
||
function lonGap(intA: LonInt, intB: LonInt): number {
|
||
if (intA.full || intB.full) return 0;
|
||
let best = Infinity;
|
||
for (const k of [-1, 0, 1]) {
|
||
const a2 = intB.a + 360 * k;
|
||
const b2 = intB.b + 360 * k;
|
||
const gap = intA.b < a2 ? a2 - intA.b : b2 < intA.a ? intA.a - b2 : 0;
|
||
if (gap < best) best = gap;
|
||
}
|
||
return best;
|
||
}
|
||
|
||
// Per-region bbox in antimeridian-safe unwrapped longitude space.
|
||
const regionBBox = new Map<string, RegionBBox>();
|
||
for (const f of regions.features) {
|
||
const polys = f.geometry.type === 'Polygon' ? [f.geometry.coordinates] : f.geometry.coordinates;
|
||
let latMin = Infinity;
|
||
let latMax = -Infinity;
|
||
let a = Infinity;
|
||
let b = -Infinity;
|
||
let full = false;
|
||
for (const poly of polys) {
|
||
const outer = poly[0];
|
||
for (const [, la] of outer) {
|
||
if (la < latMin) latMin = la;
|
||
if (la > latMax) latMax = la;
|
||
}
|
||
const { lons, full: fl } = unwrapRing(outer);
|
||
if (fl) full = true;
|
||
for (const v of lons) {
|
||
if (v < a) a = v;
|
||
if (v > b) b = v;
|
||
}
|
||
}
|
||
regionBBox.set(f.id, { latMin, latMax, lon: { a, b, full } });
|
||
}
|
||
|
||
/** Bounding-box closeness (same definition the coarsening script uses). */
|
||
function bboxesClose(x: RegionBBox, y: RegionBBox): boolean {
|
||
const latGap = x.latMax < y.latMin ? y.latMin - x.latMax : y.latMax < x.latMin ? x.latMin - y.latMax : 0;
|
||
if (latGap > CLOSE_DEG) return false;
|
||
return lonGap(x.lon, y.lon) <= CLOSE_DEG;
|
||
}
|
||
|
||
describe('regions.geo.json', () => {
|
||
it('is a valid GeoJSON FeatureCollection of well-formed features', () => {
|
||
expect(regions.type).toBe('FeatureCollection');
|
||
expect(Array.isArray(regions.features)).toBe(true);
|
||
expect(regions.features.length).toBeGreaterThan(0);
|
||
for (const f of regions.features) {
|
||
expect(f.type, `${f.id ?? '?'}: not a Feature`).toBe('Feature');
|
||
expect(typeof f.id, `${f.id ?? '?'}: id must be a string`).toBe('string');
|
||
expect(f.id.length, `${f.id ?? '?'}: empty id`).toBeGreaterThan(0);
|
||
expect(typeof f.properties?.name, `${f.id}: missing name`).toBe('string');
|
||
expect(f.properties.name.length, `${f.id}: empty name`).toBeGreaterThan(0);
|
||
expect(typeof f.properties.parent, `${f.id}: missing parent`).toBe('string');
|
||
expect(countryIds.has(f.properties.parent), `${f.id}: unknown parent ${f.properties.parent}`).toBe(true);
|
||
const g = f.geometry;
|
||
expect(
|
||
!!g && (g.type === 'Polygon' || g.type === 'MultiPolygon'),
|
||
`${f.id}: bad geometry type ${g?.type}`,
|
||
).toBe(true);
|
||
if (g) {
|
||
const polys = g.type === 'Polygon' ? [g.coordinates] : g.coordinates;
|
||
for (const poly of polys)
|
||
for (const ring of poly)
|
||
expect(ring.length, `${f.id}: ring with ${ring.length} pts`).toBeGreaterThanOrEqual(3);
|
||
}
|
||
}
|
||
});
|
||
|
||
it('has globally unique feature ids (ISO 3166-2 disambiguated on within-country collision)', () => {
|
||
const seen = new Set<string>();
|
||
const dups = new Set<string>();
|
||
for (const f of regions.features) {
|
||
if (seen.has(f.id)) dups.add(f.id);
|
||
seen.add(f.id);
|
||
}
|
||
expect(
|
||
dups.size,
|
||
`duplicate ids: ${[...dups].slice(0, 10).join(', ')}`,
|
||
).toBe(0);
|
||
expect(seen.size).toBe(regions.features.length);
|
||
});
|
||
|
||
it('covers every one of the 177 countries with at least one region', () => {
|
||
const counts = countries.features.map((c) => countByParent.get(c.id) ?? 0);
|
||
for (const c of countries.features) {
|
||
expect(countByParent.get(c.id) ?? 0, `country ${c.id} has no regions`).toBeGreaterThanOrEqual(1);
|
||
}
|
||
const min = Math.min(...counts);
|
||
const max = Math.max(...counts);
|
||
const med = median(counts);
|
||
const maxCountry = [...countByParent.entries()].sort((a, b) => b[1] - a[1])[0];
|
||
const maxCountryName = countries.features.find((c) => c.id === maxCountry[0])?.properties.name ?? '';
|
||
console.log(
|
||
`\n per-country region count: min ${min} / median ${med} / max ${max}` +
|
||
` (max: ${maxCountry[0]} ${maxCountryName} with ${maxCountry[1]})`,
|
||
);
|
||
});
|
||
|
||
it('has sane region areas: median >= 100 km2, and no non-Antarctica region >= 5,000,000 km2', () => {
|
||
const areas = regions.features.map((f) => regionArea.get(f.id) ?? 0);
|
||
const med = median(areas);
|
||
const maxRegion = regions.features.reduce((m, f) => ((regionArea.get(f.id) ?? 0) > (regionArea.get(m.id) ?? 0) ? f : m));
|
||
const minRegion = regions.features.reduce((m, f) => ((regionArea.get(f.id) ?? 0) < (regionArea.get(m.id) ?? 0) ? f : m));
|
||
const tiny = areas.filter((a) => a < 1).length;
|
||
// Antarctica (ATA) is a full-360° wrap ring, so the planar model can't
|
||
// measure it (it reports ~21.8M km2 for a ~14M km2 landmass) — exempt it.
|
||
for (const f of regions.features) {
|
||
if (f.properties.parent === 'ATA') continue;
|
||
const a = regionArea.get(f.id) ?? 0;
|
||
expect(a < 5_000_000, `${f.id}: area ${Math.round(a).toLocaleString('en-US')} km2 is absurdly large`).toBe(true);
|
||
}
|
||
expect(med, `median region area ${med.toFixed(1)} km2 is too small`).toBeGreaterThanOrEqual(100);
|
||
console.log(
|
||
`\n region area: min ${minRegion.id} (${minRegion.properties.name}) ≈ ${(regionArea.get(minRegion.id) ?? 0).toExponential(2)} km2` +
|
||
` / median ${med.toLocaleString('en-US')} km2` +
|
||
` / max ${maxRegion.id} (${maxRegion.properties.name}) ≈ ${Math.round(regionArea.get(maxRegion.id) ?? 0).toLocaleString('en-US')} km2` +
|
||
` — ${tiny} regions < 1 km2 (legit micro-regions: tiny islands, small urban districts, degenerate disputed enclaves)`,
|
||
);
|
||
});
|
||
|
||
it('leaves only islands below the ~10,000 km2 coarsening floor (non-island small regions merged)', () => {
|
||
// v0.14 coarsening: every region smaller than the floor that has a
|
||
// same-country sibling within CLOSE_DEG of its bbox is dissolved into its
|
||
// nearest such neighbor, so a below-floor region may only survive if it
|
||
// is an island. Count the non-island residuals below the floor and allow
|
||
// a small tolerance band (strictly < 1% of all regions) — the coarsener's
|
||
// fixed point leaves none, and this catches a regression in the merge
|
||
// logic (e.g. a broken neighbor test that exempts land regions) without
|
||
// being brittle to one or two borderline cases.
|
||
const smallNonIslands: string[] = [];
|
||
let smallIslands = 0;
|
||
let smallTotal = 0;
|
||
for (const f of regions.features) {
|
||
const a = regionArea.get(f.id) ?? 0;
|
||
if (a >= COARSE_FLOOR_KM2) continue;
|
||
smallTotal++;
|
||
const bb = regionBBox.get(f.id)!;
|
||
let hasNeighbor = false;
|
||
for (const g of regions.features) {
|
||
if (g.id === f.id || g.properties.parent !== f.properties.parent) continue;
|
||
if (bboxesClose(bb, regionBBox.get(g.id)!)) {
|
||
hasNeighbor = true;
|
||
break;
|
||
}
|
||
}
|
||
if (hasNeighbor) {
|
||
smallNonIslands.push(`${f.id} (${Math.round(a)} km2)`);
|
||
} else smallIslands++;
|
||
}
|
||
const frac = regions.features.length ? smallNonIslands.length / regions.features.length : 0;
|
||
expect(
|
||
frac < 0.01,
|
||
`${smallNonIslands.length} non-island region(s) below the ${COARSE_FLOOR_KM2} km2 floor ` +
|
||
`(${(100 * frac).toFixed(2)}% of all regions): ${smallNonIslands.slice(0, 10).join(', ')}`,
|
||
).toBe(true);
|
||
console.log(
|
||
`\n below-floor regions: ${smallTotal} — ${smallIslands} islands (exempt, no same-country neighbor within ${CLOSE_DEG}°)` +
|
||
` / ${smallNonIslands.length} non-island residuals (must stay < 1% of all regions)`,
|
||
);
|
||
});
|
||
|
||
it('sums to within [40%, 175%] of each country area (median in [70%, 130%])', () => {
|
||
// The band is a planar-approximation sanity check. It is wider than a naive
|
||
// [40%, 110%] because Natural Earth's admin-1 polygons are digitized
|
||
// independently of the admin-0 country polygons (they are NOT clipped to
|
||
// them), so for a handful of small/island nations the region union
|
||
// over/under-fills the country: measured range is ~[52%, 151%]
|
||
// (Brunei under-fills, Vanuatu over-fills). The band brackets that with
|
||
// margin while still catching a genuinely missing or doubled-up region.
|
||
const ratios: number[] = [];
|
||
let minRatio = Infinity;
|
||
let maxRatio = -Infinity;
|
||
let minId = '';
|
||
let maxId = '';
|
||
for (const c of countries.features) {
|
||
const ca = countryArea.get(c.id) ?? 0;
|
||
if (ca <= 0) continue;
|
||
const ratio = (sumByParent.get(c.id) ?? 0) / ca;
|
||
ratios.push(ratio);
|
||
if (ratio < minRatio) { minRatio = ratio; minId = c.id; }
|
||
if (ratio > maxRatio) { maxRatio = ratio; maxId = c.id; }
|
||
expect(
|
||
ratio >= 0.4 && ratio <= 1.75,
|
||
`${c.id}: coverage ${(100 * ratio).toFixed(1)}% is outside [40%, 175%]`,
|
||
).toBe(true);
|
||
}
|
||
const medRatio = median(ratios);
|
||
expect(
|
||
medRatio >= 0.7 && medRatio <= 1.3,
|
||
`median coverage ${(100 * medRatio).toFixed(1)}% is outside [70%, 130%]`,
|
||
).toBe(true);
|
||
console.log(
|
||
`\n coverage range: min ${minId} ${(100 * minRatio).toFixed(1)}%` +
|
||
` / median ${(100 * medRatio).toFixed(1)}%` +
|
||
` / max ${maxId} ${(100 * maxRatio).toFixed(1)}% (band 40–175%)`,
|
||
);
|
||
});
|
||
});
|