Add more tests for Intl.Locale.prototype.getTextInfo

[anba]

Add coverage for:

• Language subtag has more than three letters.
• Script meta data's RTL field is UNKNOWN.
• Don't default to "ltr" when script meta data has no entry for script.
• Missing script is added through add-likely-subtags algorithm.
• Script subtag is present and script's general ordering of characters is known.
• Script subtag doesn't refer to a valid registered script.

[anba]

The are various test failures in JSC and V8, because both implementations don't support returning undefined from TextDirectionOfLocale.

Fixed ICU4C implementation for JSC and V8:

enum class ScriptDirection {
  Unknown,
  LeftToRight,
  RightToLeft,
};

// Input: Canonicalized locale with alias mappings already replaced.
//
// Preferably |locale| contains only language-script-region subtags,
// so ICU4C doesn't reject too long locales.
static ScriptDirection GetScriptDirection(const char* locale) {
  UErrorCode status = U_ZERO_ERROR;

  // Get the script subtag.
  char script[ULOC_SCRIPT_CAPACITY] = {};
  int32_t scriptLength = uloc_getScript(locale, script, std::size(script), &status);
  if (U_FAILURE(status)) {
    return ScriptDirection::Unknown;
  }

  // If no script subtag present, add likely subtags.
  if (scriptLength == 0) {
    char maximal[ULOC_FULLNAME_CAPACITY] = {};
    int32_t maxLength = uloc_addLikelySubtags(locale, maximal, std::size(maximal), &status);
    if (U_FAILURE(status)) {
      return ScriptDirection::Unknown;
    }

    scriptLength = uloc_getScript(maximal, script, std::size(script), &status);

    // If still no script subtag present, return Unknown.
    if (scriptLength == 0) {
      return ScriptDirection::Unknown;
    }
  }

  // Get script code from script.
  UScriptCode scriptCode = (UScriptCode) u_getPropertyValueEnum(UCHAR_SCRIPT, script);
  if (scriptCode == USCRIPT_INVALID_CODE) {
    return ScriptDirection::Unknown;
  }
  if (const char* shortName = uscript_getShortName(scriptCode)) {
    // Ignore Unicode aliases from PropertyValueAliases.txt, because they don't
    // apply here.
    if (std::strcmp(script, shortName)) {
      return ScriptDirection::Unknown;
    }
  }
  switch (scriptCode) {
    // Marked as UNKNOWN in scriptMetadata.txt.
    //
    // ICU4C doesn't allow to query all possible "RTL" field values (YES, NO, UNKNOWN),
    // so the four scripts with UNKNOWN are hard-coded below.
    case USCRIPT_COMMON:     // Zyyy
    case USCRIPT_INHERITED:  // Zinh
    case USCRIPT_UNKNOWN:    // Zzzz
    case USCRIPT_BRAILLE:    // Brai
      return ScriptDirection::Unknown;

    // Not in scriptMetadata.txt
    //
    // Up to the implementations how to handle these cases. Either return
    // UNKNOWN or the correct script direction. But don't return the obviously
    // wrong answer, for example don't return left-to-right for "Aran".
    case USCRIPT_AFAKA:                         // Afak   (LTR)
    case USCRIPT_ARABIC_NASTALIQ:               // Aran   (RTL)
    case USCRIPT_BLISSYMBOLS:                   // Blis   (varies)
    case USCRIPT_CIRTH:                         // Cirt   (varies)
    case USCRIPT_OLD_CHURCH_SLAVONIC_CYRILLIC:  // Cyrs   (LTR)
    case USCRIPT_DEMOTIC_EGYPTIAN:              // Egyd   (mixed)
    case USCRIPT_HIERATIC_EGYPTIAN:             // Egyh   (mixed)
    case USCRIPT_KHUTSURI:                      // Geok   (LTR)
    case USCRIPT_TRADITIONAL_HAN_WITH_LATIN:    // Hntl   (Hant+Latn)
    case USCRIPT_KATAKANA_OR_HIRAGANA:          // Hrkt   (LTR)
    case USCRIPT_HARAPPAN_INDUS:                // Inds   (RTL)
    case USCRIPT_KPELLE:                        // Kpel   (LTR)
    case USCRIPT_LATIN_FRAKTUR:                 // Latf   (LTR)
    case USCRIPT_LATIN_GAELIC:                  // Latg   (LTR)
    case USCRIPT_LOMA:                          // Loma   (LTR)
    case USCRIPT_MAYAN_HIEROGLYPHS:             // Maya   (mixed)
    case USCRIPT_MOON:                          // Moon   (mixed)
    case USCRIPT_NAKHI_GEBA:                    // Nkgb   (LTR)
    case USCRIPT_BOOK_PAHLAVI:                  // Phlv   (mixed)
    case USCRIPT_RONGORONGO:                    // Roro   (mixed)
    case USCRIPT_SARATI:                        // Sara   (mixed)
    case USCRIPT_ESTRANGELO_SYRIAC:             // Syre   (RTL)
    case USCRIPT_WESTERN_SYRIAC:                // Syrj   (RTL)
    case USCRIPT_EASTERN_SYRIAC:                // Syrn   (RTL)
    case USCRIPT_TENGWAR:                       // Teng   (LTR)
    case USCRIPT_VISIBLE_SPEECH:                // Visp   (LTR)
    case USCRIPT_WOLEAI:                        // Wole   (LTR)
    case USCRIPT_MATHEMATICAL_NOTATION:         // Zmth   (UNKNOWN)
    case USCRIPT_SYMBOLS_EMOJI:                 // Zsye   (UNKNOWN)
    case USCRIPT_SYMBOLS:                       // Zsym   (UNKNOWN)
    case USCRIPT_UNWRITTEN_LANGUAGES:           // Zxxx   (UNKNOWN)
      return ScriptDirection::Unknown;

    default:
      break;
  }
  if (uscript_isRightToLeft(scriptCode)) {
    return ScriptDirection::RightToLeft;
  }
  return ScriptDirection::LeftToRight;
}

[github-actions]

7 new or modified tests were run on 7 engines.

View workflow run

Test	Engine262	GraalJS	JavaScriptCore	QuickJS	SpiderMonkey	V8	XS
intl402/Locale/prototype/getTextInfo/language-subtag-with-more-than-three-letters.js	❌	❌	❌	❌	✅	❌	❌
intl402/Locale/prototype/getTextInfo/script-metadata-rtl-is-unknown.js	❌	❌	❌	❌	✅	❌	❌
intl402/Locale/prototype/getTextInfo/script-metadata-unlisted-language-script-is-rtl.js	❌	❌	❌	❌	✅	❌	❌
intl402/Locale/prototype/getTextInfo/script-subtag-missing.js	❌	❌	❌	❌	✅	❌	❌
intl402/Locale/prototype/getTextInfo/script-subtag-present.js	❌	✅	❌	❌	✅	✅	❌
intl402/Locale/prototype/getTextInfo/script-unregistered-or-privateuse.js	❌	❌	❌	❌	✅	❌	❌
intl402/Locale/prototype/maximize/language-subtag-with-more-than-three-letters.js	❌	✅	✅	❌	✅	✅	❌

[ptomato]

@jugglinmike LGTM both on the original tests and your modifications.

[jugglinmike]

Thanks, Philip!