@astryxdesign/core 0.7.5 → 0.7.6-canary.0ef60dd

package/src/utils/timeFormat.ts

@@ -0,0 +1,314 @@
+// Copyright (c) Meta Platforms, Inc. and affiliates.
+
+/**
+ * @file timeFormat.ts
+ * @input An ISO time or typed text, plus the provider's BCP 47 locale
+ * @output Exports resolveHourFormat, formatLocaleTime, and parseLocaleTime
+ * @position Internal shared helper for TimeInput and DateTimeInput; not part
+ *   of the public surface
+ *
+ * Time of day as the locale writes it. The hour cycle, digits, separator, and
+ * day-period words all come from Intl rather than a table, so a field reads
+ * the same way `Timestamp` does beside it.
+ *
+ * The hour cycle is the locale's own: `en-US` and `ko-KR` read twelve-hour,
+ * `en-GB`, `de-DE`, and `vi-VN` read twenty-four-hour. A `-u-hc-` extension
+ * on the provider locale (`en-US-u-hc-h23`) carries an app's per-user
+ * preference through the same path, so every Astryx time honors it.
+ *
+ * The parser reads back what the formatter writes: the locale's digits, its
+ * hour-minute separator, and its day-period words, on either side of the time.
+ * Whatever is left goes through the shared English-tolerant parser, so "2:30
+ * PM" and "14:30" keep working in every locale.
+ *
+ * SYNC: When modified, update these files to stay in sync:
+ * - /packages/core/src/utils/timeFormat.test.ts
+ * - /packages/core/src/TimeInput/TimeInput.tsx
+ * - /packages/core/src/DateTimeInput/DateTimeInput.tsx
+ */
+
+import type {Locale} from '../i18n/types';
+import {parseISOTime, parseTimeInput, type ISOTimeString} from './timeParser';
+import {INVISIBLE_CHARS, toAsciiDigits} from './unicodeDigits';
+
+export type HourFormat = '12h' | '24h';
+
+interface LocaleTimeSymbols {
+  hourFormat: HourFormat;
+  /**
+   * The text between hour and minute and between minute and second, trimmed:
+   * ":" nearly everywhere, "." in fi-FI, "h" and "min" in fr-CA. Read from
+   * both readings, because some locales write them differently: `as` writes
+   * "১৪:৩০" but "অপৰাহ্ন ২.৩০".
+   */
+  separators: string[];
+  /**
+   * Words the locale writes before or after the digits other than the day
+   * period, trimmed: "s" after fr-CA seconds, "kl." before nds-DE times, "ga"
+   * between an ee day period and its hour. Longest first.
+   */
+  edgeWords: string[];
+  /** Every day-period word the locale writes, longest first. */
+  dayPeriods: {word: string; isPM: boolean}[];
+  /**
+   * The locale's twelve-hour reading writes 02:00 and 14:00 the same (fr-CM
+   * writes no day period), so its twelve-hour text is English instead.
+   */
+  twelveHourIsAmbiguous: boolean;
+}
+
+const symbolsByLocale = new Map<string, LocaleTimeSymbols>();
+
+// The space family Intl writes between a time and its day period ("2:30 PM"
+// uses U+202F). A typed field shows a plain space so people can edit it.
+const SPACES = /[\u00A0\u2007\u202F]/g;
+// What may sit on a peeled word's inner side: a digit (digits are already
+// ASCII here), whitespace, or nothing.
+const WORD_BOUNDARY = /^[\d\s]?$/;
+
+function utcDate(hour: number, minute: number, second: number): Date {
+  return new Date(Date.UTC(2000, 0, 1, hour, minute, second));
+}
+
+function timeFormatter(
+  locale: Locale,
+  hourFormat: HourFormat,
+  includeSeconds: boolean,
+): Intl.DateTimeFormat {
+  const options: Intl.DateTimeFormatOptions = {
+    hour: hourFormat === '12h' ? 'numeric' : '2-digit',
+    minute: '2-digit',
+    second: includeSeconds ? '2-digit' : undefined,
+    // `hourCycle`, not `hour12`: `hour12: true` resolves to h11 in ja-JP and
+    // writes noon as "午後0:30".
+    hourCycle: hourFormat === '12h' ? 'h12' : 'h23',
+    timeZone: 'UTC',
+  };
+  try {
+    return new Intl.DateTimeFormat(locale, options);
+  } catch {
+    return new Intl.DateTimeFormat('en', options);
+  }
+}
+
+function getLocaleTimeSymbols(locale: Locale): LocaleTimeSymbols {
+  let symbols = symbolsByLocale.get(locale);
+  if (symbols !== undefined) {
+    return symbols;
+  }
+
+  let hourFormat: HourFormat;
+  let resolved: Intl.ResolvedDateTimeFormatOptions;
+  try {
+    resolved = new Intl.DateTimeFormat(locale, {
+      hour: 'numeric',
+    }).resolvedOptions();
+  } catch {
+    // A malformed tag formats as English everywhere else in this module.
+    resolved = new Intl.DateTimeFormat('en', {
+      hour: 'numeric',
+    }).resolvedOptions();
+  }
+  if (resolved.hourCycle !== undefined) {
+    hourFormat =
+      resolved.hourCycle === 'h11' || resolved.hourCycle === 'h12'
+        ? '12h'
+        : '24h';
+  } else {
+    // Engines that leave hourCycle unresolved still write a day period for a
+    // twelve-hour locale when no cycle is forced.
+    let ambient: Intl.DateTimeFormat;
+    try {
+      ambient = new Intl.DateTimeFormat(locale, {hour: 'numeric'});
+    } catch {
+      ambient = new Intl.DateTimeFormat('en', {hour: 'numeric'});
+    }
+    hourFormat = ambient
+      .formatToParts(utcDate(13, 0, 0))
+      .some(part => part.type === 'dayPeriod')
+      ? '12h'
+      : '24h';
+  }
+
+  // Literals of every reading this module writes: twelve- and twenty-four-hour,
+  // with and without seconds. A literal between two numbers is a separator;
+  // one outside them is an edge word.
+  const separatorSet = new Set<string>();
+  const edgeSet = new Set<string>();
+  for (const cycle of ['12h', '24h'] as const) {
+    for (const withSeconds of [false, true]) {
+      const parts = timeFormatter(locale, cycle, withSeconds).formatToParts(
+        utcDate(13, 30, 5),
+      );
+      const isNumber = (part: Intl.DateTimeFormatPart) =>
+        part.type === 'hour' ||
+        part.type === 'minute' ||
+        part.type === 'second';
+      const first = parts.findIndex(isNumber);
+      const last = parts.length - 1 - [...parts].reverse().findIndex(isNumber);
+      parts.forEach((part, index) => {
+        if (part.type !== 'literal') {
+          return;
+        }
+        const text = normalizeText(part.value).trim();
+        if (text === '' || text === ':') {
+          return;
+        }
+        (index > first && index < last ? separatorSet : edgeSet).add(text);
+      });
+    }
+  }
+  const separators = [...separatorSet];
+  const edgeWords = [...edgeSet].sort((a, b) => b.length - a.length);
+
+  // Every hour, because some locales name several periods across the day
+  // (flexible day periods). A word that appears on both sides of noon cannot
+  // say which half it means, so it is left out.
+  const twelveHour = timeFormatter(locale, '12h', false);
+  const sides = new Map<string, Set<boolean>>();
+  for (let hour = 0; hour < 24; hour++) {
+    const word = twelveHour
+      .formatToParts(utcDate(hour, 0, 0))
+      .find(part => part.type === 'dayPeriod')?.value;
+    if (word === undefined) {
+      continue;
+    }
+    const key = normalizeText(word).trim();
+    if (key === '') {
+      continue;
+    }
+    const seen = sides.get(key) ?? new Set<boolean>();
+    seen.add(hour >= 12);
+    sides.set(key, seen);
+  }
+  const dayPeriods = [...sides]
+    .filter(([, seen]) => seen.size === 1)
+    .map(([word, seen]) => ({word, isPM: seen.has(true)}))
+    .sort((a, b) => b.word.length - a.word.length);
+
+  const twelveHourIsAmbiguous =
+    normalizeText(twelveHour.format(utcDate(2, 0, 0))) ===
+    normalizeText(twelveHour.format(utcDate(14, 0, 0)));
+
+  symbols = {
+    hourFormat,
+    separators,
+    edgeWords,
+    dayPeriods,
+    twelveHourIsAmbiguous,
+  };
+  symbolsByLocale.set(locale, symbols);
+  return symbols;
+}
+
+function normalizeText(text: string): string {
+  const folded = text
+    .normalize('NFKC')
+    .replace(INVISIBLE_CHARS, '')
+    .replace(SPACES, ' ');
+  return (toAsciiDigits(folded) ?? folded).toLowerCase();
+}
+
+function escapeRegExp(text: string): string {
+  return text.replace(/[.*+?^${}()|[\]\\]/g, '\\$&');
+}
+
+/**
+ * The hour format a field shows: the caller's explicit choice, else the
+ * locale's hour cycle (including a `-u-hc-` extension on the tag).
+ */
+export function resolveHourFormat(
+  locale: Locale,
+  hourFormat?: HourFormat,
+): HourFormat {
+  return hourFormat ?? getLocaleTimeSymbols(locale).hourFormat;
+}
+
+/**
+ * Formats an ISO time as the locale writes it, in the given hour format.
+ * en-US: "14:30" → "2:30 PM"; de-DE 24h: "14:30"; ko-KR 12h: "오후 2:30".
+ * Returns '' for an empty or malformed time.
+ */
+export function formatLocaleTime(
+  time: ISOTimeString | undefined,
+  locale: Locale,
+  hourFormat: HourFormat,
+  includeSeconds: boolean = false,
+): string {
+  if (!time) {
+    return '';
+  }
+  const parsed = parseISOTime(time);
+  if (!parsed) {
+    return '';
+  }
+  const formatLocale =
+    hourFormat === '12h' && getLocaleTimeSymbols(locale).twelveHourIsAmbiguous
+      ? 'en'
+      : locale;
+  return timeFormatter(formatLocale, hourFormat, includeSeconds)
+    .format(utcDate(parsed.hour, parsed.minute, parsed.second))
+    .replace(SPACES, ' ');
+}
+
+/**
+ * Parses typed text into an ISO time, accepting the locale's digits,
+ * separator, and day-period words as well as everything `parseTimeInput`
+ * accepts. Returns null when no reading fits.
+ */
+export function parseLocaleTime(
+  input: string,
+  locale: Locale,
+  includeSeconds: boolean = false,
+): ISOTimeString | null {
+  if (!input) {
+    return null;
+  }
+  const {separators, edgeWords, dayPeriods} = getLocaleTimeSymbols(locale);
+  let text = normalizeText(input).trim();
+
+  // Peel the day period and edge words off both ends, in whatever order the
+  // locale writes them ("ŋdi ga 2:30", "kl. 2.30 pm").
+  let isPM: boolean | null = null;
+  // A word counts only when a digit, whitespace, or nothing sits on its inner
+  // side, so gd's "m" (morning) is not read out of an English "pm" or "p.m".
+  const strip = (word: string) => {
+    if (text.startsWith(word) && WORD_BOUNDARY.test(text.charAt(word.length))) {
+      text = text.slice(word.length).trim();
+      return true;
+    }
+    if (
+      text.endsWith(word) &&
+      WORD_BOUNDARY.test(text.charAt(text.length - word.length - 1))
+    ) {
+      text = text.slice(0, -word.length).trim();
+      return true;
+    }
+    return false;
+  };
+  for (let peeled = true; peeled;) {
+    peeled = false;
+    if (isPM === null) {
+      const period = dayPeriods.find(({word}) => strip(word));
+      if (period !== undefined) {
+        isPM = period.isPM;
+        peeled = true;
+      }
+    }
+    if (edgeWords.some(strip)) {
+      peeled = true;
+    }
+  }
+
+  for (const separator of separators) {
+    text = text.replace(
+      new RegExp(`(\\d)\\s*${escapeRegExp(separator)}\\s*(?=\\d)`, 'g'),
+      '$1:',
+    );
+  }
+
+  if (isPM !== null) {
+    text = `${text} ${isPM ? 'pm' : 'am'}`;
+  }
+  return parseTimeInput(text, includeSeconds);
+}

package/src/utils/touchPresentation.stylex.ts

@@ -0,0 +1,43 @@
+// Copyright (c) Meta Platforms, Inc. and affiliates.
+
+/**
+ * @file touchPresentation.stylex.ts
+ * @input Uses StyleX defineConsts
+ * @output Exports touchKeys: the touch presentation resolver's queries as
+ *   StyleX at-rule keys
+ * @position The reference spellings of the touch, compact and iOS-gate
+ *   conditions. Core styles write these as literal keys wherever they need
+ *   them (`spec:AST-094` FR13, `spec:AST-012` FR9), and a test pins each
+ *   literal to these values; utils/touchPresentation.ts holds the same
+ *   queries for script (matchMedia)
+ *
+ * These are the spellings to match. A style that changes with these
+ * conditions writes the same literal key (`'@media (pointer: coarse)'`, the
+ * iOS gate as that media key outside `'@supports (-webkit-touch-callout:
+ * none)'`), and a test pins every literal in core to these values. Do not use
+ * an imported key as a computed StyleX key: it compiles to a different
+ * priority than the literal (6000 against 3200) and reorders the cascade.
+ * The iOS supports gate is only for a style iOS WebKit itself requires (the
+ * 16px input floor), never a behaviour.
+ *
+ * SYNC: When modified, update touchPresentation.ts and
+ * touchPresentation.test.ts.
+ */
+
+import * as stylex from '@stylexjs/stylex';
+
+export const touchKeys = stylex.defineConsts({
+  /** The primary pointer is coarse (TOUCH_QUERY). */
+  touch: '@media (pointer: coarse)',
+  /** The viewport is compact (COMPACT_QUERY, `spec:AST-012` FR9). */
+  compact: '@media (width < 768px), (height < 500px)',
+  /** Both: a presentation such as a sheet in place of a popover. */
+  compactTouch:
+    '@media (width < 768px) and (pointer: coarse), (height < 500px) and (pointer: coarse)',
+  /**
+   * The iOS WebKit gate, inside `touch` (IOS_SUPPORTS_QUERY): only for a
+   * style iOS WebKit itself requires (the 16px input floor), never a
+   * behaviour; touch behaviour is one model on every touch device.
+   */
+  iosSupports: '@supports (-webkit-touch-callout: none)',
+});

package/src/utils/touchPresentation.test.ts

@@ -22,6 +22,7 @@ import {
   COMPACT_TOUCH_PRESENTATION_QUERY,
   useAdaptivePresentation,
 } from '../hooks/useAdaptivePresentation';
+import {reducedMotionKeys} from '../motion/reducedMotion.stylex';
 import {DEFAULT_COMPACT_DEFINITION} from '../theme/themeAdaptations';
 import {Theme, defineTheme} from '../theme';
 import {
@@ -33,6 +34,7 @@ import {
   readTouchPresentation,
   useTouchPresentation,
 } from './touchPresentation';
+import {touchKeys} from './touchPresentation.stylex';
 
 /**
  * A device: what its primary pointer is, how big its viewport is, and
@@ -90,6 +92,32 @@ afterEach(() => {
   vi.unstubAllGlobals();
 });
 
+/**
+ * The Levenshtein distance between two strings, or more than `limit` as soon
+ * as it must exceed it: how many characters a spelling is off another.
+ */
+function editDistance(a: string, b: string, limit: number): number {
+  if (Math.abs(a.length - b.length) > limit) {
+    return limit + 1;
+  }
+  let previous = Array.from({length: b.length + 1}, (_, j) => j);
+  for (let i = 1; i <= a.length; i++) {
+    const current = [i];
+    for (let j = 1; j <= b.length; j++) {
+      current[j] = Math.min(
+        previous[j] + 1,
+        current[j - 1] + 1,
+        previous[j - 1] + (a[i - 1] === b[j - 1] ? 0 : 1),
+      );
+    }
+    if (Math.min(...current) > limit) {
+      return limit + 1;
+    }
+    previous = current;
+  }
+  return previous[b.length];
+}
+
 describe('touch presentation', () => {
   it('keys touch on the primary pointer and compact on the width', () => {
     const {result} = renderHook(() => useTouchPresentation());
@@ -181,6 +209,116 @@ describe('touch presentation', () => {
     });
   });
 
+  it('holds the same keys the resolver defines', () => {
+    expect(touchKeys.touch).toBe(`@media ${TOUCH_QUERY}`);
+    expect(touchKeys.compact).toBe(`@media ${COMPACT_QUERY}`);
+    expect(touchKeys.compactTouch).toBe(`@media ${COMPACT_TOUCH_QUERY}`);
+    expect(touchKeys.iosSupports).toBe(`@supports ${IOS_SUPPORTS_QUERY}`);
+    expect(IOS_TOUCH_STYLE_KEYS).toEqual([
+      touchKeys.touch,
+      touchKeys.iosSupports,
+    ]);
+  });
+
+  it('keeps every touch and Reduce Motion key literal in core: no computed StyleX key', () => {
+    // An imported defineConsts at-rule key compiles to a different StyleX
+    // priority than the same literal (6000 vs 3200), which reorders the
+    // cascade; core spells the keys and matches them to the constants.
+    const root = path.resolve(__dirname, '..');
+    const computed: string[] = [];
+    const walk = (dir: string): void => {
+      for (const entry of fs.readdirSync(dir, {withFileTypes: true})) {
+        const full = path.join(dir, entry.name);
+        if (entry.isDirectory()) {
+          if (entry.name !== '__tests__' && entry.name !== 'node_modules') {
+            walk(full);
+          }
+        } else if (
+          /\.(ts|tsx)$/.test(entry.name) &&
+          !/\.test\.|\.spec\./.test(entry.name)
+        ) {
+          fs.readFileSync(full, 'utf8')
+            .split('\n')
+            .forEach((line, index) => {
+              if (
+                !/^\s*(\*|\/\/)/.test(line) &&
+                /\[(touchKeys|reducedMotionKeys)\.[a-zA-Z]+\]\s*:/.test(line)
+              ) {
+                computed.push(`${path.relative(root, full)}:${index + 1}`);
+              }
+            });
+        }
+      }
+    };
+    walk(root);
+    expect(computed).toEqual([]);
+  });
+
+  it('keeps every quoted at-rule key in core that names the pointer, the touch callout or a compact size, or that is a few characters off a reference spelling, one of the reference spellings', () => {
+    // Every quoted string that starts with `@` is read, whatever its at-rule
+    // is spelled. Two rules, so no drift depends on which character moved:
+    // - one that names the pointer or the touch callout (any word of either
+    //   query, any case), or a compact size with its breakpoint, is one of
+    //   the resolver's own spellings or a query the presentation does not
+    //   decide (a second pointer, a fine pointer, Reduce Motion split by
+    //   pointer);
+    // - one within three characters of any of those spellings, case aside,
+    //   is that spelling exactly: a capital, a digit or a typo anywhere, in
+    //   the at-rule, a feature or a value, fails.
+    const pointerKeys = new Set<string>([
+      touchKeys.touch,
+      touchKeys.compactTouch,
+      touchKeys.iosSupports,
+      '@media (any-pointer: coarse)',
+      '@media (pointer: fine)',
+      reducedMotionKeys.fade,
+      reducedMotionKeys.instant,
+    ]);
+    const references = new Set<string>([...pointerKeys, touchKeys.compact]);
+    const lowered = [...references].map(reference => reference.toLowerCase());
+    const root = path.resolve(__dirname, '..');
+    const drifted: string[] = [];
+    let pointerLiterals = 0;
+    const walk = (dir: string): void => {
+      for (const entry of fs.readdirSync(dir, {withFileTypes: true})) {
+        const full = path.join(dir, entry.name);
+        if (entry.isDirectory()) {
+          if (entry.name !== '__tests__' && entry.name !== 'node_modules') {
+            walk(full);
+          }
+        } else if (
+          /\.(ts|tsx)$/.test(entry.name) &&
+          !/\.(test|spec|stories)\./.test(entry.name)
+        ) {
+          const source = fs.readFileSync(full, 'utf8');
+          for (const match of source.matchAll(/'(@[^'\n]*)'/g)) {
+            const key = match[1];
+            if (references.has(key)) {
+              if (pointerKeys.has(key)) {
+                pointerLiterals++;
+              }
+              continue;
+            }
+            const names =
+              /pointer|coarse|fine|touch|callout/i.test(key) ||
+              (/\b(width|height)\b/i.test(key) && /\b(768|500)px\b/.test(key));
+            const lower = key.toLowerCase();
+            const isNear = lowered.some(
+              reference => editDistance(lower, reference, 3) <= 3,
+            );
+            if (names || isNear) {
+              drifted.push(`${path.relative(root, full)}: ${key}`);
+            }
+          }
+        }
+      }
+    };
+    walk(root);
+    expect(drifted).toEqual([]);
+    // Core writes the touch key as a literal in many places.
+    expect(pointerLiterals).toBeGreaterThan(50);
+  });
+
   it('keeps every iOS gate in core spelled and nested as the resolver', () => {
     // The gate is the media query outside and the supports query inside; a
     // single supports query that names the pointer never matches anywhere.
@@ -201,7 +339,8 @@ describe('touch presentation', () => {
         } else if (
           /\.(ts|tsx)$/.test(entry.name) &&
           !/\.test\.|\.spec\./.test(entry.name) &&
-          entry.name !== 'touchPresentation.ts'
+          entry.name !== 'touchPresentation.ts' &&
+          entry.name !== 'touchPresentation.stylex.ts'
         ) {
           const lines = fs.readFileSync(full, 'utf8').split('\n');
           lines.forEach((line, index) => {

package/src/utils/touchPresentation.ts

@@ -46,9 +46,11 @@
  * test keeps core's compact-touch keys identical to COMPACT_TOUCH_QUERY and
  * every touch-callout key spelled and nested as above. The iOS style gate
  * is only for a style iOS WebKit itself requires (core's 16px input floor);
- * it never selects a behaviour. The keys are literal because the Astryx
- * StyleX build does not resolve a key shared across files through
- * `defineConsts`. The hook is for behaviour. It answers the server default
+ * it never selects a behaviour. Core spells the keys literally: an imported
+ * `defineConsts` key compiles to a different StyleX priority than the same
+ * literal, which would reorder the cascade. `touchKeys`
+ * (touchPresentation.stylex.ts, internal to core) holds the same keys, and a
+ * test keeps every literal equal to them. The hook is for behaviour. It answers the server default
  * (not touch, not compact) until hydrated unless the caller passes a server
  * hint, and follows the pointer and size when they change (a tablet docking
  * a mouse).

package/src/utils/unicodeDigits.test.ts

@@ -0,0 +1,44 @@
+// Copyright (c) Meta Platforms, Inc. and affiliates.
+
+/**
+ * @file unicodeDigits.test.ts
+ * @input Uses vitest and the unicodeDigits helpers
+ * @output Unit tests for digit values across scripts
+ * @position Testing; validates utils/unicodeDigits.ts
+ *
+ * SYNC: When unicodeDigits.ts changes, update tests to match new behavior
+ */
+
+import {describe, expect, it} from 'vitest';
+import {INVISIBLE_CHARS, digitValue, toAsciiDigits} from './unicodeDigits';
+
+describe('digitValue', () => {
+  it.each([
+    ['7', 7],
+    ['٧', 7], // Arabic-Indic
+    ['۷', 7], // Extended Arabic-Indic
+    ['७', 7], // Devanagari
+    ['৭', 7], // Bengali
+    ['７', 7], // full-width
+  ])('reads %s as %i', (char, value) => {
+    expect(digitValue(char)).toBe(value);
+  });
+
+  it('returns null for a non-digit', () => {
+    expect(digitValue('a')).toBeNull();
+    expect(digitValue('')).toBeNull();
+  });
+});
+
+describe('toAsciiDigits', () => {
+  it('rewrites digits and leaves everything else', () => {
+    expect(toAsciiDigits('٢:٣٠ م')).toBe('2:30 م');
+    expect(toAsciiDigits('१२,३४५.६')).toBe('12,345.6');
+  });
+});
+
+describe('INVISIBLE_CHARS', () => {
+  it('strips bidi marks and zero-width characters', () => {
+    expect('\u200F١٢\u061C\uFEFF'.replace(INVISIBLE_CHARS, '')).toBe('١٢');
+  });
+});

package/src/utils/unicodeDigits.ts

@@ -0,0 +1,71 @@
+// Copyright (c) Meta Platforms, Inc. and affiliates.
+
+/**
+ * @file unicodeDigits.ts
+ * @input Text in any digit script
+ * @output Exports INVISIBLE_CHARS, digitValue, and toAsciiDigits
+ * @position Internal shared utility for the locale-aware parsers
+ *   (NumberInput/numberParser.ts, utils/timeFormat.ts); not part of the
+ *   public surface
+ *
+ * Locale-independent: a digit's value is a Unicode property, not a locale
+ * preference, so this needs no Intl and no locale.
+ *
+ * SYNC: When modified, update these files to stay in sync:
+ * - /packages/core/src/utils/unicodeDigits.test.ts
+ */
+
+// Built at runtime rather than written as a literal: Babel's
+// transform-unicode-property-regex rewrites a `\p{...}` regex LITERAL into an
+// enumerated character class, and the copy bundled with Next rejects the
+// General_Category short names, failing the sandbox build. Property escapes are
+// ES2018 and need no transform in any browser this package supports.
+const DECIMAL_DIGIT = new RegExp('\\p{Nd}', 'u');
+
+/**
+ * Zero-width joiners, bidi marks and isolates (which Intl writes around RTL
+ * numbers and day periods), and the BOM Excel prepends.
+ */
+export const INVISIBLE_CHARS =
+  /[\u200B-\u200D\u200E\u200F\u061C\u2066-\u2069\uFEFF]/g;
+
+/**
+ * Value of any Unicode decimal digit. Each digit script lays its ten digits
+ * out consecutively, so walking back to the script's zero derives the value
+ * for Arabic-Indic, Devanagari, full-width and the rest without shipping a
+ * table of them. Returns null for a non-digit, and in the one shape the walk
+ * cannot resolve: two digit blocks adjacent in code space (the mathematical
+ * alphanumerics).
+ */
+export function digitValue(char: string): number | null {
+  const codePoint = char.codePointAt(0);
+  if (codePoint == null || !DECIMAL_DIGIT.test(char)) {
+    return null;
+  }
+  for (let zero = codePoint; codePoint - zero < 10; zero--) {
+    if (!DECIMAL_DIGIT.test(String.fromCodePoint(zero - 1))) {
+      return codePoint - zero;
+    }
+  }
+  return null;
+}
+
+/**
+ * Rewrites every decimal digit as its ASCII digit, leaving other characters
+ * alone. Returns null when a digit's value cannot be resolved.
+ */
+export function toAsciiDigits(text: string): string | null {
+  let out = '';
+  for (const char of text) {
+    if (!DECIMAL_DIGIT.test(char)) {
+      out += char;
+      continue;
+    }
+    const value = digitValue(char);
+    if (value == null) {
+      return null;
+    }
+    out += String(value);
+  }
+  return out;
+}