@sourcey/catalog-verifier 1.1.3 → 1.1.5

package/dist/modules/evidence-grounding/src/figures.js

@@ -1,727 +0,0 @@
-/**
- * The figures a passage writes, read by one strict grammar so that a value a
- * reader reported can be checked against the passage it cited. The grammar
- * reads amounts in US dollars, euros and pounds (`$`, `US$`, `€`, `£`) or an
- * ISO code, percentages, periods, dates with their year and counts, written
- * with comma thousands and point decimals, each on its own line. Everything
- * else that looks like a figure (another number format, currency, scale word,
- * fraction, range or rate, or a marker on the next line) is unread: it keeps
- * its extent, so no quote can cut a figure out of it, but gives no token, and
- * the text is incomplete. The detector of what is unread is deliberately
- * broad: reading too little only leaves a value unchecked, never admits one.
- * This parses written form; it never judges what a passage means.
- */
-/** ISO 4217 minor-unit exponents other than two. */
-const CURRENCY_EXPONENT = {
-    BHD: 3,
-    CLP: 0,
-    ISK: 0,
-    JOD: 3,
-    JPY: 0,
-    KRW: 0,
-    KWD: 3,
-    OMR: 3,
-    TND: 3,
-    VND: 0,
-};
-/** The symbols the grammar reads, each naming one currency. */
-const SYMBOLS = { US$: "USD", $: "USD", "€": "EUR", "£": "GBP" };
-/** ISO 4217 codes; a code that is also an English word never reads as money ("TOP 10"). */
-const SUPPORTED_CURRENCIES = new Set(Intl.supportedValuesOf("currency"));
-const WORD_CODES = new Set(["ALL", "BOB", "CUP", "GEL", "MAD", "MOP", "PEN", "SOS", "TOP", "TRY"]);
-const MULTIPLIERS = {
-    k: 1_000,
-    K: 1_000,
-    M: 1_000_000,
-    B: 1_000_000_000,
-    bn: 1_000_000_000,
-    mn: 1_000_000,
-    thousand: 1_000,
-    million: 1_000_000,
-    billion: 1_000_000_000,
-};
-/** Glued letters read as a multiplier ("10k", "$1.5M"); the words may also stand one space apart. */
-const MULTIPLIER = /^(?:(k|K|M|B|bn|mn)|[ \t]?(thousand|million|billion))(?![\p{L}\p{N}])/u;
-const PERIOD_UNITS = {
-    day: "D",
-    days: "D",
-    week: "W",
-    weeks: "W",
-    month: "M",
-    months: "M",
-    mo: "M",
-    mos: "M",
-    year: "Y",
-    years: "Y",
-    yr: "Y",
-    yrs: "Y",
-};
-const UNIT = Object.keys(PERIOD_UNITS).join("|");
-const PERIOD_AFTER = new RegExp(String.raw `^[ \t-]?(${UNIT})(?![\p{L}\p{N}])`, "iu");
-const PERCENT_AFTER = /^[ \t]?(?:%|percent(?![\p{L}])|per[ \t]?cent(?![\p{L}]))/iu;
-/**
- * Words that can change what a figure beside them means, in any language: a
- * currency, a scale, a letter currency, a nationality or a fraction the
- * grammar does not read. Broad by design: one beside a figure leaves it
- * unread, which can only leave a value unchecked.
- */
-const MARKER_WORD = /^(?:dollars?|d[oó]lar(?:es)?|dollari?|euros?|pounds?|sterling|yen|yuan|renminbi|rmb|rs|rp|rm|rupees?|rupias?|rupiahs?|pesos?|reais|real|rand|won|ringgits?|baht|dong|liras?|lire|francs?|kron(?:a|e|or|er)|zlot(?:y|ys|ych)|forints?|korun(?:a|y)|shekels?|dirhams?|riyals?|nairas?|cedis?|shillings?|thousands|millions|billions|trillions?|lakhs?|lacs?|crores?|crs?|mil|mill|mln|mio|mrd|bln|tsd|millionen|milliarden|mill[oó]n(?:es)?|milh(?:ão|ões)|milion[ei]|milliards?|miliar|juta|ribu|тыс|млн|млрд|руб\p{L}*|canadian|australian|mexican|jamaican|canadien\p{L}*|australien\p{L}*|am[ée]ricain\p{L}*|estadounidense\p{L}*|canadiense\p{L}*|australian[oa]s?|mexican[oa]s?|americano\p{L}*|half|halves|quarters?|thirds?|[万萬億亿千百円元圆원만억천]|लाख|करोड़)$/iu;
-/** Country prefixes that make a dollar sign another country's: "AU $", "$5,000 CDN". */
-const COUNTRY_PREFIXES = new Set("US USA AU AUS CA CAN CDN NZ HK SG NT MX RD CN UK EU".split(" "));
-const MONTHS = {
-    january: 1,
-    jan: 1,
-    february: 2,
-    feb: 2,
-    march: 3,
-    mar: 3,
-    april: 4,
-    apr: 4,
-    may: 5,
-    june: 6,
-    jun: 6,
-    july: 7,
-    jul: 7,
-    august: 8,
-    aug: 8,
-    september: 9,
-    sept: 9,
-    sep: 9,
-    october: 10,
-    oct: 10,
-    november: 11,
-    nov: 11,
-    december: 12,
-    dec: 12,
-};
-/** Figures written as words. */
-const NUMBER_WORDS = {
-    zero: 0,
-    one: 1,
-    two: 2,
-    three: 3,
-    four: 4,
-    five: 5,
-    six: 6,
-    seven: 7,
-    eight: 8,
-    nine: 9,
-    ten: 10,
-    eleven: 11,
-    twelve: 12,
-    thirteen: 13,
-    fourteen: 14,
-    fifteen: 15,
-    sixteen: 16,
-    seventeen: 17,
-    eighteen: 18,
-    nineteen: 19,
-    twenty: 20,
-    thirty: 30,
-    forty: 40,
-    fifty: 50,
-    sixty: 60,
-    seventy: 70,
-    eighty: 80,
-    ninety: 90,
-};
-const SCALE_WORDS = {
-    hundred: 100,
-    thousand: 1_000,
-    million: 1_000_000,
-    billion: 1_000_000_000,
-};
-/**
- * Words that write a quantity the grammar does not value: a number word that
- * is not one figure ("two three"), a magnitude, multiple, fraction or long
- * period ("thousands", "twice", "half", "a dozen", "a decade"). "one" is left
- * out, since it is prose far more often, and so is "third" ("third-party").
- */
-const UNVALUED_WORDS = new Set([
-    ...Object.keys(NUMBER_WORDS).filter((word) => word !== "one"),
-    ...Object.keys(SCALE_WORDS),
-    ...Object.keys(SCALE_WORDS).map((word) => `${word}s`),
-    ..."tens half halves quarter quarters thirds double twice triple thrice quadruple dozen dozens".split(" "),
-    ..."fortnight fortnights decade decades century centuries semester semesters biennium".split(" "),
-]);
-const longestFirst = (words) => [...words].sort((left, right) => right.length - left.length).join("|");
-const NUMBER = /(?<![\p{N}])(?:\d{1,3}(?:,\d{3})+(?:\.\d+)?|\d+(?:\.\d+)?)(?![\p{N}])/gu;
-const WORD = longestFirst([...Object.keys(NUMBER_WORDS), ...Object.keys(SCALE_WORDS)]);
-const SCALE = longestFirst(Object.keys(SCALE_WORDS));
-// An article counts one only before a scale word ("a million"); "a six-month trial" is six.
-const WORD_FIGURE = new RegExp(String.raw `(?<![\p{L}-])(?:an?[ \t-]+(?=(?:${SCALE})(?![\p{L}])))?(?:${WORD})(?:[ \t-]+(?:and[ \t-]+)?(?:${WORD}))*(?![\p{L}])`, "giu");
-/** "a year", "an extra month" is not one: only an article or "first" right before a singular unit counts. */
-const ARTICLE_PERIOD = /(?<![\p{L}-])(a|an|first)[ \t-]+(day|week|month|mo|year|yr)(?![\p{L}\p{N}])/giu;
-const FRACTION = "half|halves|quarters?|thirds?";
-const FRACTION_BEFORE = new RegExp(String.raw `(?<![\p{L}])(?:${FRACTION})\s+(?:(?:of|an?)\s+)*$`, "iu");
-const FRACTION_AFTER = new RegExp(String.raw `^\s+and\s+(?:an?\s+)?(?:${FRACTION})(?![\p{L}-])`, "iu");
-const UNVALUED_WORD = new RegExp(String.raw `(?<![\p{L}])(?:${longestFirst([...UNVALUED_WORDS])})(?![\p{L}])`, "giu");
-// A period unit counted by no figure states a period the text does not value ("for months"),
-// unless it names a rate ("per month", "/year").
-const PERIOD_WORD = new RegExp(String.raw `(?<![\p{L}])((?:per|each|every)[ \t]+|\/[ \t]?)?(?:${UNIT})(?![\p{L}])`, "giu");
-const HALF_PRICE = /(?<![\p{L}-])half[ \t-]+(?:price|priced|off)(?![\p{L}])/giu;
-const MONTH = longestFirst(Object.keys(MONTHS));
-const ORDINAL = "(?:st|nd|rd|th)?";
-const ISO_DATE = /(?<![\p{N}-])(\d{4})-(\d{2})-(\d{2})(?![\p{N}])/gu;
-const MONTH_FIRST_DATE = new RegExp(String.raw `(?<![\p{L}])(${MONTH})\.?[ \t]+(\d{1,2})${ORDINAL},?[ \t]+(\d{4})(?![\p{N}])`, "giu");
-const DAY_FIRST_DATE = new RegExp(String.raw `(?<![\p{N}])(\d{1,2})${ORDINAL}[ \t]+(?:of[ \t]+)?(${MONTH})\.?,?[ \t]+(\d{4})(?![\p{N}])`, "giu");
-/** What joins digits into one written figure the grammar reads whole or not at all. */
-const JOINER = /[.,'’/⁄:-]/u;
-/** What after a space continues the figure before it: a group of three digits, or a fraction. */
-const SPACE_GROUP = /^(?:\d{3}(?!\d)|\d+[/⁄]\d)/u;
-/** Superscript and subscript digits: footnote marks after words, part of a figure beside one. */
-const MARK_DIGIT = /[²³¹⁰-⁹₀-₉]/u;
-const OTHER_FRACTION = /[¼-¾⅐-⅞↉⁄]/u;
-const UNREAD_MARK = "\u0000";
-/** A frequency or fraction before an article makes "a year" a rate or a fraction: "once a year", "half a year". */
-const BEFORE_ARTICLE = new RegExp(String.raw `(?:(?<![\p{L}])(once|twice|thrice|times|half|halves|quarters?|thirds?)|(${UNREAD_MARK}))\s+(?:of\s+)?$`, "iu");
-const LETTER = /\p{L}/u;
-const ALPHANUMERIC = /[\p{L}\p{N}]/u;
-const DIGIT = /\p{N}/u;
-/** Every numeric token written in `text`, in order of appearance. */
-export function numericTokens(text) {
-    return writtenFigures(text).tokens;
-}
-/** Where every figure in `text` is written, in its own indexes, in order. */
-export function figureSpans(text) {
-    return writtenFigures(text).spans;
-}
-/** The minor units of `major` whole units of `currency`. */
-export function minorUnitsOf(currency, major) {
-    return Math.round(major * 10 ** (CURRENCY_EXPONENT[currency] ?? 2));
-}
-/** Periods compare by the months or days they denote, so P12M equals P1Y. */
-export function periodsEqual(left, right) {
-    const a = periodLength(left);
-    const b = periodLength(right);
-    return a !== null && b !== null && a.unit === b.unit && a.value === b.value;
-}
-export function writtenFigures(text) {
-    const { source, origins } = foldedSource(text);
-    const figures = [];
-    const identifiers = [];
-    const read = new Uint8Array(source.length + 1);
-    const figure = (at, end, token) => {
-        figures.push({ at, end, token });
-        read.fill(1, at, end);
-    };
-    const name = (at, from) => {
-        let end = from;
-        while (end < source.length && ALPHANUMERIC.test(source[end]))
-            end += 1;
-        identifiers.push({ at, end, text: source.slice(at, end) });
-        read.fill(1, at, end);
-        return end;
-    };
-    for (const date of writtenDates(source))
-        figure(date.at, date.end, date.token);
-    for (const match of source.matchAll(NUMBER)) {
-        const start = match.index ?? 0;
-        const end = start + match[0].length;
-        if (read[start] === 1)
-            continue;
-        let lettersStart = start;
-        while (lettersStart > 0 && LETTER.test(source[lettersStart - 1]))
-            lettersStart -= 1;
-        const letters = source.slice(lettersStart, start);
-        if (letters && !isCurrencyCode(letters)) {
-            // Digits glued to letters that name no currency are part of a name: "H100", "Q3", "A320".
-            name(lettersStart, end);
-            continue;
-        }
-        const run = figureRun(source, start, end);
-        if (run.at < start || run.end > end) {
-            // The figure goes on past what the grammar reads ("1/2", "5.000", "20,00,000"): unread.
-            figure(run.at, run.end, null);
-            continue;
-        }
-        const reading = readFigure(source, start, end, match[0]);
-        if (reading === "name") {
-            name(start, end);
-            continue;
-        }
-        figure(reading.at, reading.end, reading.token);
-    }
-    for (const match of source.matchAll(WORD_FIGURE)) {
-        for (const segment of wordSegments(match[0], match.index ?? 0)) {
-            if (read[segment.start] === 1)
-                continue;
-            const value = wordNumber(segment.words);
-            if (value === null)
-                continue;
-            const scaled = segment.words.some((word) => SCALE_WORDS[word] !== undefined);
-            const unit = unitAfter(source, segment.start, segment.end, value, scaled);
-            if (unit !== "name" && (scaled || unit.token?.kind !== "number" || unit.end > segment.end)) {
-                figure(unit.at, unit.end, unit.token);
-            }
-            else if (!(segment.words.length === 1 && segment.words[0] === "one")) {
-                // A counted quantity ("seven providers") is a number; "one of our partners" is prose.
-                figure(segment.start, segment.end, { kind: "number", value });
-            }
-        }
-    }
-    for (const match of source.matchAll(HALF_PRICE)) {
-        const at = match.index ?? 0;
-        figure(at, at + match[0].length, { kind: "percentage", basisPoints: 5_000 });
-    }
-    const rates = articlePeriods(source, figures, figure);
-    unreadFractions(source, figures);
-    unreadRanges(source, figures);
-    // A digit no figure, date or name read is unread, with the whole figure it is written in.
-    for (let index = 0; index < source.length; index += 1) {
-        if (read[index] === 1 || !DIGIT.test(source[index]))
-            continue;
-        const run = figureRun(source, index, index + 1);
-        figure(run.at, run.end, null);
-    }
-    const tokens = figures
-        .filter((candidate) => candidate.token !== null)
-        .sort((left, right) => left.at - right.at);
-    let complete = tokens.length === figures.length && !source.includes(UNREAD_MARK);
-    const covered = new Uint8Array(source.length + 1);
-    for (const { at, end } of [...tokens, ...rates])
-        covered.fill(1, at, end);
-    for (const match of source.matchAll(UNVALUED_WORD)) {
-        if (covered[match.index ?? 0] !== 1)
-            complete = false;
-    }
-    for (const match of source.matchAll(PERIOD_WORD)) {
-        if (match[1] === undefined && covered[match.index ?? 0] !== 1)
-            complete = false;
-    }
-    return {
-        tokens: tokens.map(({ token }) => token),
-        identifiers: identifiers.map(({ text: identifier }) => identifier),
-        complete,
-        spans: [...figures, ...rates, ...identifiers]
-            .map(({ at, end }) => ({ start: origins[at], end: origins[end] }))
-            .sort((left, right) => left.start - right.start || left.end - right.end),
-    };
-}
-/**
- * The figure written from `[start, end)` with its markers and unit, or
- * "name" when letters glued after it make it part of one ("5G", "10x").
- */
-function readFigure(source, start, end, written) {
-    // A point before exactly three digits groups thousands in some locales ("5.000 €") and ends a
-    // decimal in others ("$1.008"); a figure starting after a point continues one ("Rs.5", ".5").
-    const unreadable = /^\d+\.\d{3}$/u.test(written) || source[start - 1] === ".";
-    const multiplier = MULTIPLIER.exec(source.slice(end, end + 12));
-    const numberEnd = end + (multiplier?.[0].length ?? 0);
-    const factor = multiplier ? (MULTIPLIERS[multiplier[1] ?? multiplier[2] ?? ""] ?? 1) : 1;
-    const value = Number(written.replaceAll(",", "")) * factor;
-    const unit = unitAfter(source, start, numberEnd, value, multiplier !== null);
-    if (unit === "name")
-        return "name";
-    // An unread figure after a point keeps the point in its extent, so no quote takes "5%" from ".5%".
-    const at = source[start - 1] === "." ? Math.min(unit.at, start - 1) : unit.at;
-    return unreadable || !Number.isFinite(value) ? { ...unit, at, token: null } : unit;
-}
-/**
- * The token a figure of `value` at `[start, end)` writes with the markers on
- * its own line: a percent sign, a currency before or after it, or a period
- * unit. Its extent grows over every marker read. The figure is unread when a
- * marker it does not read stands beside it, on its line or at the edge of the
- * next line, or its value is finer than its unit counts.
- */
-function unitAfter(source, start, end, value, scaled) {
-    const after = source.slice(end, end + 48);
-    const before = source.slice(Math.max(0, start - 24), start);
-    const unread = (extent) => ({ ...extent, token: null });
-    const percent = PERCENT_AFTER.exec(after);
-    const prefix = currencyBefore(before);
-    const suffix = currencyAfter(after);
-    const period = PERIOD_AFTER.exec(after);
-    // Letters glued after a figure that no unit reads make it part of a name: "5G", "10x".
-    if (!percent && !suffix && !period && LETTER.test(after[0] ?? ""))
-        return "name";
-    if (percent) {
-        const stop = end + percent[0].length;
-        const basisPoints = wholeCount(value * 100);
-        const widened = beside(source, start, stop);
-        return scaled || basisPoints === null || widened
-            ? unread(widened ?? { at: start, end: stop })
-            : { at: start, end: stop, token: { kind: "percentage", basisPoints } };
-    }
-    if (prefix || suffix) {
-        const at = start - (prefix?.length ?? 0);
-        const stop = end + (suffix?.length ?? 0);
-        // Every marker must name the one currency: "USD $5,000" is dollars, "$5,000 CAD" is unread,
-        // and so is an amount that a period unit also claims ("USD 14 DAYS").
-        const named = [prefix, suffix].flatMap((marker) => (marker ? [marker.currency] : []));
-        const currency = named.every((candidate) => candidate === named[0]) ? named[0] : null;
-        const minorUnits = currency
-            ? wholeCount(value * 10 ** (CURRENCY_EXPONENT[currency] ?? 2))
-            : null;
-        const claimed = suffix ? null : period;
-        const widened = beside(source, at, stop + (claimed?.[0].length ?? 0));
-        return !currency || minorUnits === null || claimed || widened
-            ? unread(widened ?? { at, end: stop + (claimed?.[0].length ?? 0) })
-            : { at, end: stop, token: { kind: "money", currency, minorUnits } };
-    }
-    if (period) {
-        const stop = end + period[0].length;
-        const unit = PERIOD_UNITS[(period[1] ?? "").toLowerCase()];
-        const widened = beside(source, start, stop);
-        return scaled || !unit || !Number.isInteger(value) || widened
-            ? unread(widened ?? { at: start, end: stop })
-            : { at: start, end: stop, token: { kind: "period", iso: `P${value}${unit}` } };
-    }
-    const widened = beside(source, start, end);
-    return widened ? unread(widened) : { at: start, end, token: { kind: "number", value } };
-}
-/**
- * The figure at `[at, end)` widened over what the grammar does not read
- * beside it, or null when nothing stands there. It looks on the figure's own
- * line and at the edge of the next line or the line before, where an inline
- * tag may have split a marker off: a code, country prefix, currency, scale or
- * fraction word, another currency sign or an unreadable fraction, and across
- * a line break also a unit, a percent sign, a multiplier or the rest of a
- * decimal ("$29" / ".99"). The widened extent keeps any quote from taking the
- * figure without it.
- */
-function beside(source, at, end) {
-    let widened = null;
-    const gapAfter = /^[ \t]*(?:\n[ \t]*)?/u.exec(source.slice(end, end + 64))?.[0] ?? "";
-    const next = source.slice(end + gapAfter.length, end + gapAfter.length + 32);
-    const opened = next.startsWith("(") ? 1 : 0;
-    const following = (next[opened] === UNREAD_MARK ? 1 : 0) ||
-        markerLength(next.slice(opened)) ||
-        (gapAfter === " " ? (/^[kmbl](?![\p{L}])/iu.exec(next)?.[0].length ?? 0) : 0) ||
-        (gapAfter.includes("\n")
-            ? ((PERIOD_AFTER.exec(next) ??
-                PERCENT_AFTER.exec(next) ??
-                MULTIPLIER.exec(next) ??
-                /^[.,]\p{N}+/u.exec(next))?.[0].length ?? 0)
-            : 0);
-    if (following > 0)
-        widened = { at, end: end + gapAfter.length + opened + following };
-    const head = source.slice(Math.max(0, at - 32), at);
-    const gapBefore = /(?:[ \t]*\n)?[ \t]*$/u.exec(head)?.[0] ?? "";
-    const previous = head.slice(0, head.length - gapBefore.length);
-    const token = previous.endsWith(UNREAD_MARK)
-        ? UNREAD_MARK
-        : /(?:[\p{L}\p{M}]*\p{Sc}|[\p{L}\p{M}]+|[万萬億亿千百円元圆원만억천])\)?:?$/u.exec(previous)?.[0];
-    const bare = token?.replace(/[():]/gu, "");
-    // A sign glued before a figure, with any letters glued to it, is a compound the grammar does
-    // not read ("CA$5,000", "R$5").
-    if (bare &&
-        (bare === UNREAD_MARK || /\p{Sc}/u.test(bare) || markerLength(bare) === bare.length)) {
-        widened = {
-            at: at - gapBefore.length - (token?.length ?? 0),
-            end: widened?.end ?? end,
-        };
-    }
-    return widened;
-}
-/** The length of the word or sign starting `text` that can change a figure beside it, or 0. */
-function markerLength(text) {
-    if (/^\p{Sc}(?![ \t]?\p{N})/u.test(text))
-        return 1;
-    const code = /^([A-Z]{2,3})(?![\p{L}])/u.exec(text)?.[1];
-    if (code)
-        return COUNTRY_PREFIXES.has(code) || isCurrencyCode(code) ? code.length : 0;
-    const word = /^(?:[\p{L}\p{M}]+|[万萬億亿千百円元圆원만억천])/u.exec(text)?.[0];
-    return word !== undefined && MARKER_WORD.test(word) ? word.length : 0;
-}
-function isCurrencyCode(letters) {
-    return (/^[A-Z]{3}$/u.test(letters) && SUPPORTED_CURRENCIES.has(letters) && !WORD_CODES.has(letters));
-}
-/**
- * The currency written directly before a figure on its line: a read symbol,
- * a code, or a code and a symbol. A word before a symbol that is no code is
- * prose ("AWS $5,000").
- */
-function currencyBefore(before) {
-    const symbol = /(?<![\p{L}\p{N}\p{Sc}])(US\$|[$€£])[ \t]?$/u.exec(before);
-    const rest = symbol ? before.slice(0, before.length - symbol[0].length) : before;
-    const code = /(?<![\p{L}\p{N}])([A-Z]{3})[ \t]?$/u.exec(rest);
-    const coded = code && isCurrencyCode(code[1]) ? code : null;
-    if (!symbol && !coded)
-        return null;
-    const named = symbol ? (SYMBOLS[symbol[1]] ?? null) : null;
-    const currency = coded?.[1] ?? named;
-    return {
-        currency: coded && named && coded[1] !== named ? null : currency,
-        length: (symbol?.[0].length ?? 0) + (coded?.[0].length ?? 0),
-    };
-}
-/** The currency written directly after a figure on its line: a code, "(code)", or a euro or pound sign. */
-function currencyAfter(after) {
-    const marker = /^[ \t]?(?:(\(?)([A-Z]{3})(\)?)(?![\p{L}])|([€£]))/u.exec(after);
-    if (!marker)
-        return null;
-    const [written, open, code, close, symbol] = marker;
-    if (code) {
-        return isCurrencyCode(code) && (open === "") === (close === "")
-            ? { currency: code, length: written.length }
-            : null;
-    }
-    return { currency: SYMBOLS[symbol] ?? null, length: written.length };
-}
-/**
- * `scaled` as the whole count it writes, or null when the figure is written
- * finer than its unit counts ("$0.0035", "0.125%").
- */
-function wholeCount(scaled) {
-    const count = Math.round(scaled);
-    return Math.abs(scaled - count) < 1e-6 ? count : null;
-}
-/**
- * The whole written figure around the digits at `[start, end)`: every digit
- * joined to them by a separator, slash, colon or hyphen, or by a space
- * before a group of exactly three digits ("10 000") or a fraction ("1 1/2").
- */
-function figureRun(source, start, end) {
-    const joins = (index) => DIGIT.test(source[index - 1] ?? "") &&
-        (JOINER.test(source[index] ?? "")
-            ? DIGIT.test(source[index + 1] ?? "")
-            : source[index] === " " && SPACE_GROUP.test(source.slice(index + 1, index + 8)));
-    let at = start;
-    while (DIGIT.test(source[at - 1] ?? ""))
-        at -= 1;
-    while (joins(at - 1)) {
-        at -= 1;
-        while (DIGIT.test(source[at - 1] ?? ""))
-            at -= 1;
-    }
-    let stop = end;
-    while (DIGIT.test(source[stop] ?? ""))
-        stop += 1;
-    while (joins(stop)) {
-        stop += 1;
-        while (DIGIT.test(source[stop] ?? ""))
-            stop += 1;
-    }
-    return { at, end: stop };
-}
-/**
- * "a", "an" and "first" count one period ("a year", "your first month"). A
- * figure, frequency or fraction right before them, even across a line break,
- * makes "a month" a rate ("$100 a month", "once a year") or a fraction ("½ a
- * year"): an amount keeps its token and takes the rate into its extent, a
- * bare figure is unread with it, and a frequency's rate is returned so its
- * unit counts as read.
- */
-function articlePeriods(source, figures, figure) {
-    const rates = [];
-    const endingAt = new Map(figures.map((candidate) => [candidate.end, candidate]));
-    for (const match of source.matchAll(ARTICLE_PERIOD)) {
-        const at = match.index ?? 0;
-        const end = at + match[0].length;
-        const head = source.slice(Math.max(0, at - 16), at);
-        const gap = /\s+(?:of\s+)?$/iu.exec(head)?.[0].length ?? 0;
-        const figureBefore = gap > 0 ? endingAt.get(at - gap) : undefined;
-        if (figureBefore) {
-            // An amount's rate joins its extent; a bare figure before "a year" is a fraction of it.
-            if (figureBefore.token === null || figureBefore.token.kind === "number") {
-                figureBefore.token = null;
-            }
-            figureBefore.end = end;
-            continue;
-        }
-        const lead = BEFORE_ARTICLE.exec(head);
-        if (lead) {
-            if (/^(?:once|twice|thrice|times)/iu.test(lead[1] ?? ""))
-                rates.push({ at, end });
-            else
-                figure(at - lead[0].length, end, null);
-            continue;
-        }
-        const unit = PERIOD_UNITS[(match[2] ?? "").toLowerCase()];
-        if (unit)
-            figure(at, end, { kind: "period", iso: `P1${unit}` });
-    }
-    return rates;
-}
-/**
- * A fraction word beside a figure changes its value ("half a year", "a year
- * and a half", "5 and a half years", "a quarter of a million"), so the
- * figure is unread, with the words in its extent.
- */
-function unreadFractions(source, figures) {
-    for (const candidate of figures) {
-        const before = FRACTION_BEFORE.exec(source.slice(Math.max(0, candidate.at - 24), candidate.at));
-        const after = FRACTION_AFTER.exec(source.slice(candidate.end, candidate.end + 24));
-        if (!before && !after)
-            continue;
-        candidate.token = null;
-        candidate.at -= before?.[0].length ?? 0;
-        candidate.end += after?.[0].length ?? 0;
-    }
-}
-/**
- * Two figures joined by a range ("5-10k", "$1 to $5 million", "between $1
- * and $5 million") share what only one of them writes, so both are unread,
- * as one extent.
- */
-function unreadRanges(source, figures) {
-    figures.sort((left, right) => left.at - right.at || left.end - right.end);
-    for (let index = 1; index < figures.length; index += 1) {
-        const low = figures[index - 1];
-        const high = figures[index];
-        if (low.end > high.at)
-            continue;
-        const joiner = source.slice(low.end, high.at);
-        const between = /(?<![\p{L}])between\s+$/iu.test(source.slice(Math.max(0, low.at - 12), low.at));
-        if (!/^\s*(?:\p{Pd}|to|through|and)\s*$/iu.test(joiner))
-            continue;
-        if (/and/iu.test(joiner) && !between)
-            continue;
-        low.token = null;
-        high.token = null;
-        low.end = high.end;
-    }
-}
-/**
- * The text as the tokenizer reads it, one code point at a time under
- * compatibility folding (so full-width digits and no-break spaces read as
- * their plain forms), and the index in the original text of every character
- * read. A superscript or subscript digit after a word is a footnote mark and
- * reads as a space; beside a figure or a fraction slash it is part of that
- * figure ("$4⁹⁹", "¹⁄₂") and, like every fraction character, unreadable.
- */
-function foldedSource(text) {
-    const parts = [];
-    const origins = [];
-    let last = "";
-    for (let index = 0; index < text.length;) {
-        const character = String.fromCodePoint(text.codePointAt(index));
-        const following = String.fromCodePoint(text.codePointAt(index + character.length) ?? 32);
-        const folded = MARK_DIGIT.test(character)
-            ? DIGIT.test(last) ||
-                OTHER_FRACTION.test(last) ||
-                last === UNREAD_MARK ||
-                OTHER_FRACTION.test(following)
-                ? UNREAD_MARK
-                : " "
-            : OTHER_FRACTION.test(character)
-                ? UNREAD_MARK
-                : character.normalize("NFKC");
-        for (let unit = 0; unit < folded.length; unit += 1)
-            origins.push(index);
-        parts.push(folded);
-        last = folded.at(-1) ?? last;
-        index += character.length;
-    }
-    origins.push(text.length);
-    return { source: parts.join(""), origins: Uint32Array.from(origins) };
-}
-/**
- * Calendar dates written with their year, in ISO, month-first and day-first
- * forms, as tokens. A numeric date in any other form ("05/06/2026",
- * "31-12-2026") is not read: its digits are unread as one figure.
- */
-function writtenDates(source) {
-    const dates = [];
-    const add = (match, year, month, day) => {
-        const iso = month === undefined ? null : calendarDate(year, month, day);
-        const at = match.index ?? 0;
-        if (iso)
-            dates.push({ at, end: at + match[0].length, token: { kind: "date", iso } });
-    };
-    for (const match of source.matchAll(ISO_DATE)) {
-        add(match, Number(match[1]), Number(match[2]), Number(match[3]));
-    }
-    for (const match of source.matchAll(MONTH_FIRST_DATE)) {
-        add(match, Number(match[3]), MONTHS[(match[1] ?? "").toLowerCase()], Number(match[2]));
-    }
-    for (const match of source.matchAll(DAY_FIRST_DATE)) {
-        add(match, Number(match[3]), MONTHS[(match[2] ?? "").toLowerCase()], Number(match[1]));
-    }
-    return dates;
-}
-function calendarDate(year, month, day) {
-    const date = new Date(Date.UTC(year, month - 1, day));
-    if (date.getUTCFullYear() !== year ||
-        date.getUTCMonth() !== month - 1 ||
-        date.getUTCDate() !== day) {
-        return null;
-    }
-    return date.toISOString().slice(0, 10);
-}
-/** A written figure's words split into figures: "one and three years" is two, "one hundred and five" one. */
-function wordSegments(written, offset) {
-    const segments = [];
-    let current;
-    for (const match of written.matchAll(/[a-z]+/giu)) {
-        const word = (match[0] ?? "").toLowerCase();
-        const start = offset + (match.index ?? 0);
-        const end = start + word.length;
-        const previous = current?.words.at(-1);
-        if (word === "and" && (previous === undefined || SCALE_WORDS[previous] === undefined)) {
-            if (current)
-                segments.push(current);
-            current = undefined;
-            continue;
-        }
-        current ??= { words: [], start, end };
-        current.words.push(word);
-        current.end = end;
-    }
-    if (current)
-        segments.push(current);
-    return segments;
-}
-/** The value of a figure written in words, or null when the words are not one figure. */
-function wordNumber(words) {
-    let total = 0;
-    let current = 0;
-    let last = null;
-    for (const word of words) {
-        if (word === "and") {
-            if (last !== "hundred" && last !== "scale")
-                return null;
-            continue;
-        }
-        if (word === "a" || word === "an") {
-            if (last !== null)
-                return null;
-            current = 1;
-            last = "article";
-            continue;
-        }
-        const scale = SCALE_WORDS[word];
-        if (scale !== undefined) {
-            if (current === 0)
-                return null;
-            if (scale === 100) {
-                if (current >= 100)
-                    return null;
-                current *= 100;
-                last = "hundred";
-            }
-            else {
-                total += current * scale;
-                current = 0;
-                last = "scale";
-            }
-            continue;
-        }
-        const value = NUMBER_WORDS[word];
-        if (value === undefined || last === "article" || last === "unit")
-            return null;
-        if (last === "tens" && (value >= 10 || value === 0))
-            return null;
-        current += value;
-        last = value >= 20 ? "tens" : "unit";
-    }
-    return last === "article" ? null : total + current;
-}
-function periodLength(iso) {
-    const match = /^P(\d+(?:\.\d+)?)([DWMY])$/u.exec(iso);
-    if (!match)
-        return null;
-    const count = Number(match[1]);
-    switch (match[2]) {
-        case "Y":
-            return { unit: "months", value: count * 12 };
-        case "M":
-            return { unit: "months", value: count };
-        case "W":
-            return { unit: "days", value: count * 7 };
-        default:
-            return { unit: "days", value: count };
-    }
-}
-//# sourceMappingURL=figures.js.map