@whiteslove/parsing-lexicon 0.10.4 → 0.10.5

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@whiteslove/parsing-lexicon",
-  "version": "0.10.4",
+  "version": "0.10.5",
   "description": "Shared deterministic multilingual parsing lexicon for WhitesLove housing and hiring services",
   "repository": {
     "type": "git",

package/src/housing-address.d.ts

@@ -10,6 +10,8 @@ export type HousingAddressParts = Readonly<{
   staircase?: string;
   district?: string;
   metro?: string;
+  /** Every metro station named, primary (`metro`) first. */
+  metros?: readonly string[];
   mahalla?: string;
   geoEntities?: Readonly<Partial<Record<'street' | 'district' | 'metro' | 'mahalla' | 'residentialComplex', Readonly<{
     id: string;

package/src/housing-address.js

@@ -2,6 +2,7 @@ import {
   TASHKENT_NUMBERED_AREA_ALIASES,
   matchTashkentHousingDistrict,
   matchTashkentHousingMetro,
+  matchTashkentHousingMetros,
   matchTashkentNumberedArea,
 } from './tashkent-housing-geography.js';
 import { detectNonAddressSpans } from './semantic-spans.js';
@@ -246,10 +247,14 @@ function tashkentGeoComponents(value) {
   const text = String(value ?? '');
   const district = matchTashkentHousingDistrict(text)?.name || null;
   const metro = matchTashkentHousingMetro(text)?.name || null;
+  const metros = metro
+    ? [...new Set([metro, ...matchTashkentHousingMetros(text).map((station) => station.name)])]
+    : [];
   const mahalla = text.match(/(?:^|[^\p{L}])(\p{L}[\p{L}'’ʼ-]{1,48})\s+(?:mahalla(?:si)?|маҳалла(?:си)?|махалл[ая]|mfy)(?=$|[^\p{L}])/iu)?.[1] || null;
   return Object.freeze({
     district,
     metro,
+    metros: metros.length ? Object.freeze(metros) : null,
     mahalla: DESCRIPTIVE_MAHALLA_WORD_RE.test(mahalla || '') ? null : compactStreet(mahalla),
   });
 }

package/src/housing-listing-enrichment.d.ts

@@ -58,6 +58,8 @@ export interface HousingListingEnrichment {
   district?: string | null;
   quarter?: { number: number; suffix: string } | null;
   metro?: string | null;
+  /** Every metro station named, primary (`metro`) first. */
+  metros?: readonly string[];
   developmentArea?: string | null;
   residenceComplex?: string | null;
   address?: string | null;

package/src/housing-listing-enrichment.js

@@ -1,7 +1,7 @@
 import { deepFreeze } from './lexicon-core.js';
 import { findAllCanonical, normalizeUnicode } from './normalization.js';
 import { GENERIC_LANDMARK_TERMS } from './landmarks.js';
-import { matchTashkentHousingDistrict, matchTashkentHousingMetro, matchTashkentHousingQuarter } from './tashkent-housing-geography.js';
+import { matchTashkentHousingDistrict, matchTashkentHousingMetro, matchTashkentHousingMetros, matchTashkentHousingQuarter } from './tashkent-housing-geography.js';
 import { TASHKENT_LANDMARKS } from './tashkent-pois.js';
 import { TASHKENT_RESIDENTIAL_COMPLEXES, matchTashkentResidentialComplex } from './tashkent-residential-complexes.js';
 import { parseHousingRoomCount, parseHousingFloor, parseHousingAreas } from './housing-structured.js';
@@ -161,6 +161,16 @@ function cityMetro(text, country, city) {
   return contextualCityMatches(text, country, city, 'metro', METRO_PREFIX_RE)[0]?.canonical || null;
 }
 
+// Every station the listing names, primary first. A flat "between Nizomiy and
+// Chilonzor" must keep both rather than silently dropping the second.
+function listingMetros(text, country, city, primary) {
+  const names = [
+    ...matchTashkentHousingMetros(text).map((station) => station.name),
+    ...contextualCityMatches(text, country, city, 'metro', METRO_PREFIX_RE).map((match) => match.canonical),
+  ];
+  return Object.freeze(unique(primary ? [primary, ...names] : names));
+}
+
 function cityDevelopmentArea(text, country, city) {
   const normalizedCountry = String(country || '').toUpperCase();
   if (!['KZ', 'UZ'].includes(normalizedCountry) || !city) return null;
@@ -310,6 +320,7 @@ export function parseHousingListingEnrichment(value, { country = '', city = '',
   const quarter = matchTashkentHousingQuarter(text);
   const district = matchTashkentHousingDistrict(text)?.name || quarter?.district || null;
   const metro = matchTashkentHousingMetro(text)?.name || cityMetro(text, country, city) || null;
+  const metros = listingMetros(text, country, city, metro);
   const developmentArea = cityDevelopmentArea(text, country, city);
   const primaryResidentialText = withoutNearbyLocationReferences(text);
   const parsedRc = specificResidentialComplex(primaryResidentialText)
@@ -386,6 +397,7 @@ export function parseHousingListingEnrichment(value, { country = '', city = '',
     district: district || null,
     quarter: quarter ? { number: quarter.number, suffix: quarter.suffix } : null,
     metro: metro || null,
+    metros,
     developmentArea,
     residenceComplex: parsedRc || null,
     address: address.address,

package/src/tashkent-housing-geography.d.ts

@@ -46,4 +46,5 @@ export function hasExplicitTashkentDistrict(value: unknown, canonical: string):
 export function matchTashkentHousingDistrict(value: unknown): TashkentDistrictEntry | null;
 export function matchTashkentHousingQuarter(value: unknown): Readonly<{ district: string; number: number; suffix: string }> | null;
 export function matchTashkentHousingMetro(value: unknown): TashkentMetroEntry | null;
+export function matchTashkentHousingMetros(value: unknown): ReadonlyArray<TashkentMetroEntry>;
 export function matchTashkentHousingTransit(value: unknown): TashkentHousingLocationEntry | null;

package/src/tashkent-housing-geography.js

@@ -298,42 +298,75 @@ const METRO_NUMBERED_AREA = Object.freeze({
   Yangihayot: 'Yangihayot',
 });
 
+// Position of the first mention of `station` that really names the station,
+// or -1. An explicit marker must win even if an earlier bare occurrence shares
+// a name with a district: "Sergeli tumani, metro Sergeli" still refers to the
+// station at the second occurrence.
+function stationMentionIndex(text, station) {
+  const flags = [...new Set(`${station.re.flags.replace(/g/gu, '')}g`)].join('');
+  const matches = [...text.matchAll(new RegExp(station.re.source, flags))];
+  if (!matches.length) return -1;
+  const explicit = matches.find((match) => hasExplicitMetroContext(text, match));
+  if (explicit) return explicit.index ?? 0;
+  if (hasExplicitTashkentDistrict(text, station.name)) return -1;
+  // "Toshkent" is both a metro station and the city's own name, so a bare
+  // mention ("Toshkent shahri") is not evidence of the station the way a bare
+  // mention of any other station name would be. Require an explicit metro
+  // context for this one station specifically.
+  if (station.name === 'Toshkent') return -1;
+  if (station.name === 'Qoyliq' && QOYLIQ_MASSIF_RE.test(text)) return -1;
+  const areaCanonical = METRO_NUMBERED_AREA[station.name];
+  if (areaCanonical && matchTashkentNumberedArea(text, areaCanonical)) return -1;
+  for (const match of matches) {
+    if (hasExplicitDistrictContext(text, match) || hasExplicitAreaContext(text, match) || hasExplicitMahallaContext(text, match) || hasExplicitLandmarkContext(text, match)) continue;
+    return match.index ?? 0;
+  }
+  return -1;
+}
+
+function extraAliasMentionIndex(text, aliases) {
+  const match = text.match(aliasesToRegex(aliases));
+  if (!match) return -1;
+  if (hasExplicitDistrictContext(text, match) || hasExplicitAreaContext(text, match) || hasExplicitMahallaContext(text, match)) return -1;
+  return match.index ?? 0;
+}
+
 /** Resolve listing typos/transliterations while respecting explicit non-metro geography. */
 export function matchTashkentHousingMetro(value) {
   const text = String(value ?? '');
   if (!text) return null;
   for (const station of TASHKENT_METRO) {
-    const flags = [...new Set(`${station.re.flags.replace(/g/gu, '')}g`)].join('');
-    const matches = [...text.matchAll(new RegExp(station.re.source, flags))];
-    if (!matches.length) continue;
-    // An explicit marker must win even if an earlier bare occurrence shares a
-    // name with a district. For example, "Sergeli tumani, metro Sergeli"
-    // still refers to the station at the second occurrence.
-    if (matches.some((match) => hasExplicitMetroContext(text, match))) return station;
-    const sameNamedDistrict = hasExplicitTashkentDistrict(text, station.name);
-    for (const match of matches) {
-      if (sameNamedDistrict) continue;
-    // "Toshkent" is both a metro station and the city's own name, so a bare
-    // mention ("Toshkent shahri") is not evidence of the station the way a
-    // bare mention of any other station name would be. Require an explicit
-    // metro context for this one station specifically.
-      if (station.name === 'Toshkent') continue;
-      if (station.name === 'Qoyliq' && QOYLIQ_MASSIF_RE.test(text)) continue;
-      const areaCanonical = METRO_NUMBERED_AREA[station.name];
-      if (areaCanonical && matchTashkentNumberedArea(text, areaCanonical)) continue;
-      if (hasExplicitDistrictContext(text, match) || hasExplicitAreaContext(text, match) || hasExplicitMahallaContext(text, match) || hasExplicitLandmarkContext(text, match)) continue;
-      return station;
-    }
+    if (stationMentionIndex(text, station) !== -1) return station;
   }
   for (const [canonical, aliases] of Object.entries(EXTRA_METRO_ALIASES)) {
-    const match = text.match(aliasesToRegex(aliases));
-    if (!match) continue;
-    if (hasExplicitDistrictContext(text, match) || hasExplicitAreaContext(text, match) || hasExplicitMahallaContext(text, match)) continue;
+    if (extraAliasMentionIndex(text, aliases) === -1) continue;
     return TASHKENT_METRO.find((station) => station.name === canonical) || null;
   }
   return null;
 }
 
+/**
+ * Every Tashkent metro station a listing names, in text order. Listings often
+ * sit between stations ("Nizomiy & Chilonzor metrosiga yaqin"); the singular
+ * matcher above keeps its historical first-station contract.
+ */
+export function matchTashkentHousingMetros(value) {
+  const text = String(value ?? '');
+  if (!text) return Object.freeze([]);
+  const found = new Map();
+  for (const station of TASHKENT_METRO) {
+    const index = stationMentionIndex(text, station);
+    if (index !== -1) found.set(station.name, { station, index });
+  }
+  for (const [canonical, aliases] of Object.entries(EXTRA_METRO_ALIASES)) {
+    if (found.has(canonical)) continue;
+    const index = extraAliasMentionIndex(text, aliases);
+    const station = TASHKENT_METRO.find((entry) => entry.name === canonical);
+    if (index !== -1 && station) found.set(canonical, { station, index });
+  }
+  return Object.freeze([...found.values()].sort((a, b) => a.index - b.index).map((item) => item.station));
+}
+
 const TASHKENT_HOUSING_TRANSIT = Object.freeze([
   locationEntry('Tashkent North Railway Station', 'railway', [
     'Ташкент Северный вокзал',