@whiteslove/parsing-lexicon 0.9.7 → 0.9.9

package/README.md

@@ -85,7 +85,44 @@ parseHiringContext(
 );
 ```
 
-Subpath exports are available for consumers that need narrower modules, including `./geo`, `./locations`, `./housing-context`, `./housing-structured`, `./hiring-context`, `./hiring-professions` and `./money`.
+Subpath exports are available for consumers that need narrower modules, including `./geo`, `./locations`, `./housing-context`, `./housing-structured`, `./hiring-context`, `./hiring-professions`, `./housing-money`, `./money` and `./currency`.
+
+### Money & currency
+
+Currency and magnitude vocabulary (symbols, ISO codes, spelled-out names, and
+regional stand-ins like "у.е.") lives here once, in `money-lexicon.js`. A
+consumer that needs to know whether some text mentions money at all — a fast
+pre-filter before running the heavier parsers below, e.g. when scanning many
+HTML card candidates scraped from a page — should use `moneyMentionPattern()`
+rather than hand-copying a currency symbol list. A hand-copied list silently
+drifts from this one; that exact bug (a missing "₸"/KZT symbol, and a missing
+"у.е." token) is why this section exists.
+
+```js
+import {
+  moneyCurrencyPattern,   // just the currency alternation, e.g. for building a custom matcher
+  moneyMentionPattern,    // currency OR a bare magnitude word ("15 млн") — boundary-guarded
+  moneyCurrencyFromText,  // -> 'KZT' | 'USD' | ... | null
+} from '@whiteslove/parsing-lexicon/currency';
+import { parseHousingPrice } from '@whiteslove/parsing-lexicon/housing-money';
+import { parseSalary } from '@whiteslove/parsing-lexicon/money';
+
+const looksLikeMoney = new RegExp(moneyMentionPattern(), 'iu');
+looksLikeMoney.test('300 000 ₸ в месяц'); // true
+looksLikeMoney.test('2 до 3 месяцев');    // false — "м" inside "месяцев" does not count
+
+moneyCurrencyFromText('300 000 ₸ в месяц'); // 'KZT'
+
+parseHousingPrice('12 500 000 сум', 'UZS');
+// -> { amount: 12500000, currency: 'UZS' }
+```
+
+Do not write a source-local regex for "does this text contain a price" or
+"what currency is this" — both already exist here and are kept in sync with
+the full currency/magnitude vocabulary used by `parseHousingPrice`/
+`parseSalary`. If a currency form you need is missing, add it to
+`CURRENCY_TERMS` in `money-lexicon.js` (and `CURRENCY_SYMBOL_CANDIDATES` if
+it's a standalone symbol) rather than working around the gap in the consumer.
 
 ## Data-quality rules
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@whiteslove/parsing-lexicon",
-  "version": "0.9.7",
+  "version": "0.9.9",
   "description": "Shared deterministic multilingual parsing lexicon for WhitesLove housing and hiring services",
   "repository": {
     "type": "git",

package/src/alias-prefilter.js

@@ -113,8 +113,13 @@ function indexFor(entries) {
  * Equivalent to `entries.find((entry) => entry.re.test(text))`, including the
  * "first in list order wins" tie-break, but only compiles and runs the regexes
  * of entries the text could plausibly contain.
+ *
+ * `accept(entry, matchedText)` optionally vets each hit before it wins. It
+ * receives the substring the alias regex actually matched, which is what a
+ * caller needs to tell a genuine name from an alias that merely repeats some
+ * other place's name; rejecting a hit continues the scan rather than ending it.
  */
-export function matchFirstEntry(entries, text) {
+export function matchFirstEntry(entries, text, accept) {
   if (!Array.isArray(entries) || !entries.length) return undefined;
   const value = String(text || '');
   if (!value) return undefined;
@@ -130,7 +135,11 @@ export function matchFirstEntry(entries, text) {
 
   for (const i of [...candidates].sort((a, b) => a - b)) {
     const entry = entries[i];
-    if (entry?.re?.test(value)) return entry;
+    // `re` carries no /g flag, so exec() is stateless and safe to reuse here.
+    const match = entry?.re?.exec(value);
+    if (!match) continue;
+    if (accept && !accept(entry, match[0])) continue;
+    return entry;
   }
   return undefined;
 }

package/src/currency.d.ts

@@ -9,6 +9,7 @@ export const CURRENCY_SYMBOL_CANDIDATES: Readonly<Record<string, readonly string
 export function moneyCurrencyCandidatesFromText(value: unknown): readonly string[];
 export function moneyCurrencyFromText(value: unknown, fallbackCurrency?: string | null): string | null;
 export function moneyCurrencyPattern(): string;
+export function moneyMentionPattern(): string;
 export function currencyDisplay(code: unknown, locale?: string): CurrencyDisplay | null;
 export function currencySymbol(code: unknown, locale?: string): string | null;
 export function currencyName(code: unknown, locale?: string): string | null;

package/src/currency.js

@@ -9,6 +9,7 @@ export {
   moneyCurrencyCandidatesFromText,
   moneyCurrencyFromText,
   moneyCurrencyPattern,
+  moneyMentionPattern,
 } from './money-core.js';
 
 const LOCALE_ALIASES = Object.freeze({

package/src/housing-structured.js

@@ -19,16 +19,16 @@ import {
 } from './housing-numeric-spans.js';
 
 const NUMBER_WORDS = Object.freeze([
-  [/(?<![\p{L}\p{N}_])(?:однушк\p{L}*|однокомнатн\p{L}*|bir\s+xona(?:li)?|бир\s+хона(?:ли|лик)?|1\s*(?:-\s*)?к(?:омн\p{L}*)?|1\s*(?:-\s*)?xona(?:li)?|1\s*(?:-\s*)?хона(?:лик|ли)?|1\s*бөлмелі|one[- ]bedroom|one[- ]room)(?![\p{L}\p{N}_])/iu, 1],
-  [/(?<![\p{L}\p{N}_])(?:двушк\p{L}*|двухкомнатн\p{L}*|ikki\s+xona(?:li)?|икки\s+хона(?:ли|лик)?|2\s*(?:-\s*)?к(?:омн\p{L}*)?|2\s*(?:-\s*)?xona(?:li)?|2\s*(?:-\s*)?хона(?:лик|ли)?|2\s*бөлмелі|two[- ]bedroom|two[- ]room)(?![\p{L}\p{N}_])/iu, 2],
-  [/(?<![\p{L}\p{N}_])(?:тр[её]шк\p{L}*|трехкомнатн\p{L}*|трёхкомнатн\p{L}*|uch\s+xona(?:li)?|уч\s+хона(?:ли|лик)?|3\s*(?:-\s*)?к(?:омн\p{L}*)?|3\s*(?:-\s*)?xona(?:li)?|3\s*(?:-\s*)?хона(?:лик|ли)?|3\s*бөлмелі|three[- ]bedroom|three[- ]room)(?![\p{L}\p{N}_])/iu, 3],
-  [/(?<![\p{L}\p{N}_])(?:четыр[её]хкомнатн\p{L}*|четыр[её]шк\p{L}*|to['’]?rt\s+xona(?:li)?|tort\s+xona(?:li)?|тўрт\s+хона(?:ли|лик)?|турт\s+хона(?:ли|лик)?|4\s*(?:-\s*)?к(?:омн\p{L}*)?|4\s*(?:-\s*)?xona(?:li)?|4\s*(?:-\s*)?хона(?:лик|ли)?|4\s*бөлмелі|four[- ]bedroom|four[- ]room)(?![\p{L}\p{N}_])/iu, 4],
-  [/(?<![\p{L}\p{N}_])(?:пятикомнатн\p{L}*|besh\s+xona(?:li)?|беш\s+хона(?:ли|лик)?|5\s*(?:-\s*)?к(?:омн\p{L}*)?|5\s*(?:-\s*)?xona(?:li)?|5\s*(?:-\s*)?хона(?:лик|ли)?|five[- ]bedroom|five[- ]room)(?![\p{L}\p{N}_])/iu, 5],
-  [/(?<![\p{L}\p{N}_])(?:шестикомнатн\p{L}*|olti\s+xona(?:li)?|олти\s+хона(?:ли|лик)?|6\s*(?:-\s*)?к(?:омн\p{L}*)?|6\s*(?:-\s*)?xona(?:li)?|6\s*(?:-\s*)?хона(?:лик|ли)?|six[- ]bedroom|six[- ]room)(?![\p{L}\p{N}_])/iu, 6],
-  [/(?<![\p{L}\p{N}_])(?:семикомнатн\p{L}*|yetti\s+xona(?:li)?|етти\s+хона(?:ли|лик)?|7\s*(?:-\s*)?к(?:омн\p{L}*)?|7\s*(?:-\s*)?xona(?:li)?|7\s*(?:-\s*)?хона(?:лик|ли)?|seven[- ]bedroom|seven[- ]room)(?![\p{L}\p{N}_])/iu, 7],
-  [/(?<![\p{L}\p{N}_])(?:восьмикомнатн\p{L}*|sakkiz\s+xona(?:li)?|саккиз\s+хона(?:ли|лик)?|8\s*(?:-\s*)?к(?:омн\p{L}*)?|8\s*(?:-\s*)?xona(?:li)?|8\s*(?:-\s*)?хона(?:лик|ли)?|eight[- ]bedroom|eight[- ]room)(?![\p{L}\p{N}_])/iu, 8],
-  [/(?<![\p{L}\p{N}_])(?:девятикомнатн\p{L}*|to['’]?qqiz\s+xona(?:li)?|toqqiz\s+xona(?:li)?|тўққиз\s+хона(?:ли|лик)?|токкиз\s+хона(?:ли|лик)?|9\s*(?:-\s*)?к(?:омн\p{L}*)?|9\s*(?:-\s*)?xona(?:li)?|9\s*(?:-\s*)?хона(?:лик|ли)?|nine[- ]bedroom|nine[- ]room)(?![\p{L}\p{N}_])/iu, 9],
-  [/(?<![\p{L}\p{N}_])(?:десятикомнатн\p{L}*|o['’]?n\s+xona(?:li)?|on\s+xona(?:li)?|ўн\s+хона(?:ли|лик)?|он\s+хона(?:ли|лик)?|10\s*(?:-\s*)?к(?:омн\p{L}*)?|10\s*(?:-\s*)?xona(?:li)?|10\s*(?:-\s*)?хона(?:лик|ли)?|ten[- ]bedroom|ten[- ]room)(?![\p{L}\p{N}_])/iu, 10],
+  [/(?<![\p{L}\p{N}_])(?:однушк\p{L}*|однокомнатн\p{L}*|однокімнатн\p{L}*|bir\s+xona(?:li)?|бир\s+хона(?:ли|лик)?|1\s*(?:-\s*)?к(?:омн\p{L}*)?|1\s*(?:-\s*)?xona(?:li)?|1\s*(?:-\s*)?хона(?:лик|ли)?|1\s*бөлмелі|one[- ]bedroom|one[- ]room)(?![\p{L}\p{N}_])/iu, 1],
+  [/(?<![\p{L}\p{N}_])(?:двушк\p{L}*|двухкомнатн\p{L}*|двокімнатн\p{L}*|ikki\s+xona(?:li)?|икки\s+хона(?:ли|лик)?|2\s*(?:-\s*)?к(?:омн\p{L}*)?|2\s*(?:-\s*)?xona(?:li)?|2\s*(?:-\s*)?хона(?:лик|ли)?|2\s*бөлмелі|two[- ]bedroom|two[- ]room)(?![\p{L}\p{N}_])/iu, 2],
+  [/(?<![\p{L}\p{N}_])(?:тр[её]шк\p{L}*|трехкомнатн\p{L}*|трёхкомнатн\p{L}*|трикімнатн\p{L}*|uch\s+xona(?:li)?|уч\s+хона(?:ли|лик)?|3\s*(?:-\s*)?к(?:омн\p{L}*)?|3\s*(?:-\s*)?xona(?:li)?|3\s*(?:-\s*)?хона(?:лик|ли)?|3\s*бөлмелі|three[- ]bedroom|three[- ]room)(?![\p{L}\p{N}_])/iu, 3],
+  [/(?<![\p{L}\p{N}_])(?:четыр[её]хкомнатн\p{L}*|четыр[её]шк\p{L}*|чотирикімнатн\p{L}*|to['’]?rt\s+xona(?:li)?|tort\s+xona(?:li)?|тўрт\s+хона(?:ли|лик)?|турт\s+хона(?:ли|лик)?|4\s*(?:-\s*)?к(?:омн\p{L}*)?|4\s*(?:-\s*)?xona(?:li)?|4\s*(?:-\s*)?хона(?:лик|ли)?|4\s*бөлмелі|four[- ]bedroom|four[- ]room)(?![\p{L}\p{N}_])/iu, 4],
+  [/(?<![\p{L}\p{N}_])(?:пятикомнатн\p{L}*|п['’]ятикімнатн\p{L}*|besh\s+xona(?:li)?|беш\s+хона(?:ли|лик)?|5\s*(?:-\s*)?к(?:омн\p{L}*)?|5\s*(?:-\s*)?xona(?:li)?|5\s*(?:-\s*)?хона(?:лик|ли)?|five[- ]bedroom|five[- ]room)(?![\p{L}\p{N}_])/iu, 5],
+  [/(?<![\p{L}\p{N}_])(?:шестикомнатн\p{L}*|шестикімнатн\p{L}*|olti\s+xona(?:li)?|олти\s+хона(?:ли|лик)?|6\s*(?:-\s*)?к(?:омн\p{L}*)?|6\s*(?:-\s*)?xona(?:li)?|6\s*(?:-\s*)?хона(?:лик|ли)?|six[- ]bedroom|six[- ]room)(?![\p{L}\p{N}_])/iu, 6],
+  [/(?<![\p{L}\p{N}_])(?:семикомнатн\p{L}*|семикімнатн\p{L}*|yetti\s+xona(?:li)?|етти\s+хона(?:ли|лик)?|7\s*(?:-\s*)?к(?:омн\p{L}*)?|7\s*(?:-\s*)?xona(?:li)?|7\s*(?:-\s*)?хона(?:лик|ли)?|seven[- ]bedroom|seven[- ]room)(?![\p{L}\p{N}_])/iu, 7],
+  [/(?<![\p{L}\p{N}_])(?:восьмикомнатн\p{L}*|восьмикімнатн\p{L}*|sakkiz\s+xona(?:li)?|саккиз\s+хона(?:ли|лик)?|8\s*(?:-\s*)?к(?:омн\p{L}*)?|8\s*(?:-\s*)?xona(?:li)?|8\s*(?:-\s*)?хона(?:лик|ли)?|eight[- ]bedroom|eight[- ]room)(?![\p{L}\p{N}_])/iu, 8],
+  [/(?<![\p{L}\p{N}_])(?:девятикомнатн\p{L}*|дев['’]ятикімнатн\p{L}*|to['’]?qqiz\s+xona(?:li)?|toqqiz\s+xona(?:li)?|тўққиз\s+хона(?:ли|лик)?|токкиз\s+хона(?:ли|лик)?|9\s*(?:-\s*)?к(?:омн\p{L}*)?|9\s*(?:-\s*)?xona(?:li)?|9\s*(?:-\s*)?хона(?:лик|ли)?|nine[- ]bedroom|nine[- ]room)(?![\p{L}\p{N}_])/iu, 9],
+  [/(?<![\p{L}\p{N}_])(?:десятикомнатн\p{L}*|десятикімнатн\p{L}*|o['’]?n\s+xona(?:li)?|on\s+xona(?:li)?|ўн\s+хона(?:ли|лик)?|он\s+хона(?:ли|лик)?|10\s*(?:-\s*)?к(?:омн\p{L}*)?|10\s*(?:-\s*)?xona(?:li)?|10\s*(?:-\s*)?хона(?:лик|ли)?|ten[- ]bedroom|ten[- ]room)(?![\p{L}\p{N}_])/iu, 10],
 ]);
 
 // Common classifieds shorthand: rooms/floor/total floors, optionally followed

package/src/housing.js

@@ -210,7 +210,7 @@ export function looksHousingRoomOnly(value) {
   if (!text) return false;
   const occupancy = resolveHousingOccupancy(text);
   if (occupancy === 'room' || occupancy === 'sharedRoom' || occupancy === 'bedSpace') return true;
-  return /подселени|підселен|комнату\s+в|кімнату\s+в|сда[её]тся\s+комната|сдается\s+комната|сдам\s+комнату|здам\s+кімнат|room\s+in\s+a\s+(?:shared\s+)?flat|room\s+for\s+rent|shared\s+(?:flat|apartment|room)|roommate|flatmate|xona\s+ijaraga|xona\s+beriladi|sherik(?:ka|lik)|шерик(?:ка|лик)|(?:1|бир)\s*та\s*(?:бола|киши|қиз|киз)\s*керак|1\s*хонага[^\r\n]{0,40}(?:киши|одам)\s*турилади|бөлме\s+жалға|închiriez\s+camer[ăa]|ищу[^\r\n]{0,60}сосед|ищем[^\r\n]{0,60}сосед|нужен[^\r\n]{0,60}сосед|нужна[^\r\n]{0,60}сосед|шукаю[^\r\n]{0,60}сусід|шукаємо[^\r\n]{0,60}сусід|потрібен[^\r\n]{0,60}сусід|потрібна[^\r\n]{0,60}сусід|співмешкан|співжител|соседк|сусідк/iu.test(text);
+  return /подселени|підселен|комнату\s+в(?![\p{L}\p{N}_])|кімнату\s+в(?![\p{L}\p{N}_])|сда[её]тся\s+комната|сдается\s+комната|сдам\s+комнату|здам\s+кімнат|room\s+in\s+a\s+(?:shared\s+)?flat|room\s+for\s+rent|shared\s+(?:flat|apartment|room)|roommate|flatmate|xona\s+ijaraga|xona\s+beriladi|sherik(?:ka|lik)|шерик(?:ка|лик)|(?:1|бир)\s*та\s*(?:бола|киши|қиз|киз)\s*керак|1\s*хонага[^\r\n]{0,40}(?:киши|одам)\s*турилади|бөлме\s+жалға|închiriez\s+camer[ăa]|ищу[^\r\n]{0,60}сосед|ищем[^\r\n]{0,60}сосед|нужен[^\r\n]{0,60}сосед|нужна[^\r\n]{0,60}сосед|шукаю[^\r\n]{0,60}сусід|шукаємо[^\r\n]{0,60}сусід|потрібен[^\r\n]{0,60}сусід|потрібна[^\r\n]{0,60}сусід|співмешкан|співжител|соседк|сусідк/iu.test(text);
 }
 
 export function resolveHousingPropertyType(value) {

package/src/money-core.d.ts

@@ -8,3 +8,4 @@ export function parseScaledAmount(raw: unknown, scale?: unknown): number | null;
 export function moneyCurrencyCandidatesFromText(value: unknown): readonly string[];
 export function moneyCurrencyFromText(value: unknown, fallbackCurrency?: string | null): string | null;
 export function moneyCurrencyPattern(): string;
+export function moneyMentionPattern(): string;

package/src/money-core.js

@@ -112,3 +112,19 @@ export function moneyCurrencyPattern() {
     .map(escapeRegex)
     .join('|');
 }
+
+// A cheap "does this text mention money at all" pre-filter — a currency term
+// or a bare magnitude word (e.g. "15 млн", no currency spelled out) both
+// count. Consumers that need to reject non-money text fast before running a
+// full parseSalary/parseHousingPrice pass (e.g. scanning many HTML card
+// candidates) can test against this instead of hand-copying a currency list,
+// which drifts from the lexicon's own (see MONEY_SCALE_PATTERN for the
+// magnitude half already used by MONEY_RANGE_RE/MONEY_SINGLE_RE).
+//
+// Word-boundary guarded for the same reason MONEY_RANGE_RE/MONEY_SINGLE_RE
+// are: MONEY_SCALE_PATTERN includes bare single-letter abbreviations like "м"
+// (million) that would otherwise match inside an unrelated word — e.g. "2 до
+// 3 месяцев" ("months") must never register as a money mention.
+export function moneyMentionPattern() {
+  return `(?<![\\p{L}\\p{N}_])(?:${moneyCurrencyPattern()}|${MONEY_SCALE_PATTERN})(?![\\p{L}\\p{N}_])`;
+}

package/src/money-lexicon.js

@@ -6,7 +6,7 @@ export const CURRENCY_TERMS = Object.freeze([
   group('USD', { ru: ['$', 'usd', 'доллар', 'доллара', 'долларов', 'бакс', 'у.е.', 'у е'], en: ['$', 'usd', 'dollar', 'dollars', 'us dollar'], uk: ['$', 'usd', 'долар', 'долари', 'доларів'], ro: ['$', 'usd', 'dolar', 'dolari'], uzLatn: ['$', 'usd', 'dollar'], uzCyrl: ['$', 'доллар'], kk: ['$', 'usd', 'доллар'] }),
   group('EUR', { ru: ['€', 'eur', 'евро'], en: ['€', 'eur', 'euro', 'euros'], uk: ['€', 'eur', 'євро'], ro: ['€', 'eur', 'euro'], uzLatn: ['€', 'eur', 'yevro'], uzCyrl: ['€', 'евро'], kk: ['€', 'eur', 'еуро'] }),
   group('UZS', { ru: ['сум', 'сумов', 'узс'], en: ['uzs', 'uzbek som'], uk: ['uzs', 'сум'], ro: ['uzs'], uzLatn: ["so'm", 'so‘m', 'soʻm', 'sum', 'uzs'], uzCyrl: ['сўм', 'сум'], kk: ['uzs'] }),
-  group('KZT', { ru: ['тенге', 'тг', 'kzt'], en: ['kzt', 'tenge'], uk: ['тенге', 'kzt'], ro: ['kzt', 'tenge'], uzLatn: ['kzt', 'tenge'], uzCyrl: ['kzt', 'тенге'], kk: ['теңге', 'тенге', 'тг', 'kzt'] }),
+  group('KZT', { ru: ['₸', 'тенге', 'тг', 'kzt'], en: ['₸', 'kzt', 'tenge'], uk: ['₸', 'тенге', 'kzt'], ro: ['₸', 'kzt', 'tenge'], uzLatn: ['₸', 'kzt', 'tenge'], uzCyrl: ['₸', 'kzt', 'тенге'], kk: ['₸', 'теңге', 'тенге', 'тг', 'kzt'] }),
   group('UAH', { ru: ['грн', 'гривна', 'гривны', 'гривен', '₴', 'uah'], en: ['uah', 'hryvnia', 'hryvnias', '₴'], uk: ['грн', 'гривня', 'гривні', 'гривень', '₴', 'uah'], ro: ['uah', 'hrivne', 'grivne'], uzLatn: ['uah'], uzCyrl: ['uah'], kk: ['uah'] }),
   group('RUB', { ru: ['₽', 'руб', 'руб.', 'рубль', 'рубля', 'рублей', 'rub', 'rur'], en: ['₽', 'rub', 'rur', 'ruble', 'rubles'], uk: ['₽', 'руб', 'рублів', 'rub'], ro: ['₽', 'rub', 'ruble'], uzLatn: ['₽', 'rub', 'rubl'], uzCyrl: ['₽', 'рубль'], kk: ['₽', 'rub', 'рубль'] }),
   group('GBP', { ru: ['£', 'gbp', 'фунт', 'фунтов'], en: ['£', 'gbp', 'pound', 'pounds', 'sterling'], uk: ['£', 'gbp', 'фунт', 'фунтів'], ro: ['£', 'gbp', 'liră sterlină', 'lira sterlina'], uzLatn: ['£', 'gbp', 'funt'], uzCyrl: ['£', 'фунт'], kk: ['£', 'gbp', 'фунт'] }),
@@ -41,6 +41,7 @@ export const CURRENCY_SYMBOL_CANDIDATES = Object.freeze({
   '€': Object.freeze(['EUR']),
   '₴': Object.freeze(['UAH']),
   '₽': Object.freeze(['RUB']),
+  '₸': Object.freeze(['KZT']),
   '£': Object.freeze(['GBP']),
   '₺': Object.freeze(['TRY']),
   '₾': Object.freeze(['GEL']),

package/src/money.js

@@ -22,6 +22,7 @@ export {
   MONEY_SINGLE_RE,
   moneyCurrencyFromText,
   moneyCurrencyPattern,
+  moneyMentionPattern,
   moneyScaleMultiplier,
   parseNumericAmount,
   parseScaledAmount,