CharacterTypes.cpp 12 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316
  1. /*
  2. * Copyright (c) 2021-2024, Tim Flynn <trflynn89@serenityos.org>
  3. *
  4. * SPDX-License-Identifier: BSD-2-Clause
  5. */
  6. #define AK_DONT_REPLACE_STD
  7. #include <AK/Array.h>
  8. #include <AK/CharacterTypes.h>
  9. #include <AK/Find.h>
  10. #include <AK/Traits.h>
  11. #include <LibLocale/ICU.h>
  12. #include <LibUnicode/CharacterTypes.h>
  13. #include <unicode/uchar.h>
  14. #include <unicode/uscript.h>
  15. namespace Unicode {
  16. template<typename PropertyType>
  17. struct PropertyName {
  18. Optional<StringView> long_name;
  19. Optional<StringView> short_name;
  20. Optional<StringView> additional_name;
  21. };
  22. // From uchar.h:
  23. // Unicode allows for additional names, beyond the long and short name, which would be indicated by U_LONG_PROPERTY_NAME + i
  24. static constexpr auto ADDITIONAL_NAME = static_cast<UPropertyNameChoice>(U_LONG_PROPERTY_NAME + 1);
  25. }
  26. template<typename PropertyType>
  27. struct AK::Traits<Unicode::PropertyName<PropertyType>> {
  28. static constexpr bool equals(Unicode::PropertyName<PropertyType> const& candidate, StringView property)
  29. {
  30. return property == candidate.long_name || property == candidate.short_name || property == candidate.additional_name;
  31. }
  32. };
  33. namespace Unicode {
  34. static constexpr GeneralCategory GENERAL_CATEGORY_CASED_LETTER = U_CHAR_CATEGORY_COUNT + 1;
  35. static constexpr GeneralCategory GENERAL_CATEGORY_LETTER = U_CHAR_CATEGORY_COUNT + 2;
  36. static constexpr GeneralCategory GENERAL_CATEGORY_MARK = U_CHAR_CATEGORY_COUNT + 3;
  37. static constexpr GeneralCategory GENERAL_CATEGORY_NUMBER = U_CHAR_CATEGORY_COUNT + 4;
  38. static constexpr GeneralCategory GENERAL_CATEGORY_PUNCTUATION = U_CHAR_CATEGORY_COUNT + 5;
  39. static constexpr GeneralCategory GENERAL_CATEGORY_SYMBOL = U_CHAR_CATEGORY_COUNT + 6;
  40. static constexpr GeneralCategory GENERAL_CATEGORY_SEPARATOR = U_CHAR_CATEGORY_COUNT + 7;
  41. static constexpr GeneralCategory GENERAL_CATEGORY_OTHER = U_CHAR_CATEGORY_COUNT + 8;
  42. static constexpr GeneralCategory GENERAL_CATEGORY_LIMIT = U_CHAR_CATEGORY_COUNT + 9;
  43. Optional<GeneralCategory> general_category_from_string(StringView general_category)
  44. {
  45. static auto general_category_names = []() {
  46. Array<PropertyName<GeneralCategory>, GENERAL_CATEGORY_LIMIT.value()> names;
  47. auto set_names = [&](auto property, auto index, auto general_category) {
  48. if (char const* name = u_getPropertyValueName(property, general_category, U_LONG_PROPERTY_NAME))
  49. names[index.value()].long_name = StringView { name, strlen(name) };
  50. if (char const* name = u_getPropertyValueName(property, general_category, U_SHORT_PROPERTY_NAME))
  51. names[index.value()].short_name = StringView { name, strlen(name) };
  52. if (char const* name = u_getPropertyValueName(property, general_category, ADDITIONAL_NAME))
  53. names[index.value()].additional_name = StringView { name, strlen(name) };
  54. };
  55. for (GeneralCategory general_category = 0; general_category < U_CHAR_CATEGORY_COUNT; ++general_category)
  56. set_names(UCHAR_GENERAL_CATEGORY, general_category, static_cast<UCharCategory>(general_category.value()));
  57. set_names(UCHAR_GENERAL_CATEGORY_MASK, GENERAL_CATEGORY_CASED_LETTER, U_GC_LC_MASK);
  58. set_names(UCHAR_GENERAL_CATEGORY_MASK, GENERAL_CATEGORY_LETTER, U_GC_L_MASK);
  59. set_names(UCHAR_GENERAL_CATEGORY_MASK, GENERAL_CATEGORY_MARK, U_GC_M_MASK);
  60. set_names(UCHAR_GENERAL_CATEGORY_MASK, GENERAL_CATEGORY_NUMBER, U_GC_N_MASK);
  61. set_names(UCHAR_GENERAL_CATEGORY_MASK, GENERAL_CATEGORY_PUNCTUATION, U_GC_P_MASK);
  62. set_names(UCHAR_GENERAL_CATEGORY_MASK, GENERAL_CATEGORY_SYMBOL, U_GC_S_MASK);
  63. set_names(UCHAR_GENERAL_CATEGORY_MASK, GENERAL_CATEGORY_SEPARATOR, U_GC_Z_MASK);
  64. set_names(UCHAR_GENERAL_CATEGORY_MASK, GENERAL_CATEGORY_OTHER, U_GC_C_MASK);
  65. return names;
  66. }();
  67. if (auto index = find_index(general_category_names.begin(), general_category_names.end(), general_category); index != general_category_names.size())
  68. return static_cast<GeneralCategory>(index);
  69. return {};
  70. }
  71. bool code_point_has_general_category(u32 code_point, GeneralCategory general_category)
  72. {
  73. auto icu_code_point = static_cast<UChar32>(code_point);
  74. auto icu_general_category = static_cast<UCharCategory>(general_category.value());
  75. if (general_category == GENERAL_CATEGORY_CASED_LETTER)
  76. return (U_GET_GC_MASK(icu_code_point) & U_GC_LC_MASK) != 0;
  77. if (general_category == GENERAL_CATEGORY_LETTER)
  78. return (U_GET_GC_MASK(icu_code_point) & U_GC_L_MASK) != 0;
  79. if (general_category == GENERAL_CATEGORY_MARK)
  80. return (U_GET_GC_MASK(icu_code_point) & U_GC_M_MASK) != 0;
  81. if (general_category == GENERAL_CATEGORY_NUMBER)
  82. return (U_GET_GC_MASK(icu_code_point) & U_GC_N_MASK) != 0;
  83. if (general_category == GENERAL_CATEGORY_PUNCTUATION)
  84. return (U_GET_GC_MASK(icu_code_point) & U_GC_P_MASK) != 0;
  85. if (general_category == GENERAL_CATEGORY_SYMBOL)
  86. return (U_GET_GC_MASK(icu_code_point) & U_GC_S_MASK) != 0;
  87. if (general_category == GENERAL_CATEGORY_SEPARATOR)
  88. return (U_GET_GC_MASK(icu_code_point) & U_GC_Z_MASK) != 0;
  89. if (general_category == GENERAL_CATEGORY_OTHER)
  90. return (U_GET_GC_MASK(icu_code_point) & U_GC_C_MASK) != 0;
  91. return u_charType(icu_code_point) == icu_general_category;
  92. }
  93. bool code_point_has_control_general_category(u32 code_point)
  94. {
  95. return code_point_has_general_category(code_point, U_CONTROL_CHAR);
  96. }
  97. bool code_point_has_space_separator_general_category(u32 code_point)
  98. {
  99. return code_point_has_general_category(code_point, U_SPACE_SEPARATOR);
  100. }
  101. static constexpr Property PROPERTY_ANY = UCHAR_BINARY_LIMIT + 1;
  102. static constexpr Property PROPERTY_ASCII = UCHAR_BINARY_LIMIT + 2;
  103. static constexpr Property PROPERTY_ASSIGNED = UCHAR_BINARY_LIMIT + 3;
  104. static constexpr Property PROPERTY_LIMIT = UCHAR_BINARY_LIMIT + 4;
  105. Optional<Property> property_from_string(StringView property)
  106. {
  107. static auto property_names = []() {
  108. Array<PropertyName<Property>, PROPERTY_LIMIT.value()> names;
  109. for (Property property = 0; property < UCHAR_BINARY_LIMIT; ++property) {
  110. auto icu_property = static_cast<UProperty>(property.value());
  111. if (char const* name = u_getPropertyName(icu_property, U_LONG_PROPERTY_NAME))
  112. names[property.value()].long_name = StringView { name, strlen(name) };
  113. if (char const* name = u_getPropertyName(icu_property, U_SHORT_PROPERTY_NAME))
  114. names[property.value()].short_name = StringView { name, strlen(name) };
  115. if (char const* name = u_getPropertyName(icu_property, ADDITIONAL_NAME))
  116. names[property.value()].additional_name = StringView { name, strlen(name) };
  117. }
  118. names[PROPERTY_ANY.value()] = { "Any"sv, {}, {} };
  119. names[PROPERTY_ASCII.value()] = { "ASCII"sv, {}, {} };
  120. names[PROPERTY_ASSIGNED.value()] = { "Assigned"sv, {}, {} };
  121. return names;
  122. }();
  123. if (auto index = find_index(property_names.begin(), property_names.end(), property); index != property_names.size())
  124. return static_cast<Property>(index);
  125. return {};
  126. }
  127. bool code_point_has_property(u32 code_point, Property property)
  128. {
  129. auto icu_code_point = static_cast<UChar32>(code_point);
  130. auto icu_property = static_cast<UProperty>(property.value());
  131. if (property == PROPERTY_ANY)
  132. return is_unicode(code_point);
  133. if (property == PROPERTY_ASCII)
  134. return is_ascii(code_point);
  135. if (property == PROPERTY_ASSIGNED)
  136. return u_isdefined(icu_code_point);
  137. return static_cast<bool>(u_hasBinaryProperty(icu_code_point, icu_property));
  138. }
  139. bool code_point_has_emoji_property(u32 code_point)
  140. {
  141. return code_point_has_property(code_point, UCHAR_EMOJI);
  142. }
  143. bool code_point_has_emoji_modifier_base_property(u32 code_point)
  144. {
  145. return code_point_has_property(code_point, UCHAR_EMOJI_MODIFIER_BASE);
  146. }
  147. bool code_point_has_emoji_presentation_property(u32 code_point)
  148. {
  149. return code_point_has_property(code_point, UCHAR_EMOJI_PRESENTATION);
  150. }
  151. bool code_point_has_identifier_start_property(u32 code_point)
  152. {
  153. return u_isIDStart(static_cast<UChar32>(code_point));
  154. }
  155. bool code_point_has_identifier_continue_property(u32 code_point)
  156. {
  157. return u_isIDPart(static_cast<UChar32>(code_point));
  158. }
  159. bool code_point_has_regional_indicator_property(u32 code_point)
  160. {
  161. return code_point_has_property(code_point, UCHAR_REGIONAL_INDICATOR);
  162. }
  163. bool code_point_has_variation_selector_property(u32 code_point)
  164. {
  165. return code_point_has_property(code_point, UCHAR_VARIATION_SELECTOR);
  166. }
  167. // https://tc39.es/ecma262/#table-binary-unicode-properties
  168. bool is_ecma262_property(Property property)
  169. {
  170. if (property == PROPERTY_ANY || property == PROPERTY_ASCII || property == PROPERTY_ASSIGNED)
  171. return true;
  172. switch (property.value()) {
  173. case UCHAR_ASCII_HEX_DIGIT:
  174. case UCHAR_ALPHABETIC:
  175. case UCHAR_BIDI_CONTROL:
  176. case UCHAR_BIDI_MIRRORED:
  177. case UCHAR_CASE_IGNORABLE:
  178. case UCHAR_CASED:
  179. case UCHAR_CHANGES_WHEN_CASEFOLDED:
  180. case UCHAR_CHANGES_WHEN_CASEMAPPED:
  181. case UCHAR_CHANGES_WHEN_LOWERCASED:
  182. case UCHAR_CHANGES_WHEN_NFKC_CASEFOLDED:
  183. case UCHAR_CHANGES_WHEN_TITLECASED:
  184. case UCHAR_CHANGES_WHEN_UPPERCASED:
  185. case UCHAR_DASH:
  186. case UCHAR_DEFAULT_IGNORABLE_CODE_POINT:
  187. case UCHAR_DEPRECATED:
  188. case UCHAR_DIACRITIC:
  189. case UCHAR_EMOJI:
  190. case UCHAR_EMOJI_COMPONENT:
  191. case UCHAR_EMOJI_MODIFIER:
  192. case UCHAR_EMOJI_MODIFIER_BASE:
  193. case UCHAR_EMOJI_PRESENTATION:
  194. case UCHAR_EXTENDED_PICTOGRAPHIC:
  195. case UCHAR_EXTENDER:
  196. case UCHAR_GRAPHEME_BASE:
  197. case UCHAR_GRAPHEME_EXTEND:
  198. case UCHAR_HEX_DIGIT:
  199. case UCHAR_IDS_BINARY_OPERATOR:
  200. case UCHAR_IDS_TRINARY_OPERATOR:
  201. case UCHAR_ID_CONTINUE:
  202. case UCHAR_ID_START:
  203. case UCHAR_IDEOGRAPHIC:
  204. case UCHAR_JOIN_CONTROL:
  205. case UCHAR_LOGICAL_ORDER_EXCEPTION:
  206. case UCHAR_LOWERCASE:
  207. case UCHAR_MATH:
  208. case UCHAR_NONCHARACTER_CODE_POINT:
  209. case UCHAR_PATTERN_SYNTAX:
  210. case UCHAR_PATTERN_WHITE_SPACE:
  211. case UCHAR_QUOTATION_MARK:
  212. case UCHAR_RADICAL:
  213. case UCHAR_REGIONAL_INDICATOR:
  214. case UCHAR_S_TERM:
  215. case UCHAR_SOFT_DOTTED:
  216. case UCHAR_TERMINAL_PUNCTUATION:
  217. case UCHAR_UNIFIED_IDEOGRAPH:
  218. case UCHAR_UPPERCASE:
  219. case UCHAR_VARIATION_SELECTOR:
  220. case UCHAR_WHITE_SPACE:
  221. case UCHAR_XID_CONTINUE:
  222. case UCHAR_XID_START:
  223. return true;
  224. default:
  225. return false;
  226. }
  227. }
  228. Optional<Script> script_from_string(StringView script)
  229. {
  230. static auto script_names = []() {
  231. Array<PropertyName<Script>, static_cast<size_t>(USCRIPT_CODE_LIMIT)> names;
  232. for (Script script = 0; script < USCRIPT_CODE_LIMIT; ++script) {
  233. auto icu_script = static_cast<UScriptCode>(script.value());
  234. if (char const* name = uscript_getName(icu_script))
  235. names[script.value()].long_name = StringView { name, strlen(name) };
  236. if (char const* name = uscript_getShortName(icu_script))
  237. names[script.value()].short_name = StringView { name, strlen(name) };
  238. if (char const* name = u_getPropertyValueName(UCHAR_SCRIPT, icu_script, ADDITIONAL_NAME))
  239. names[script.value()].additional_name = StringView { name, strlen(name) };
  240. }
  241. return names;
  242. }();
  243. if (auto index = find_index(script_names.begin(), script_names.end(), script); index != script_names.size())
  244. return static_cast<Script>(index);
  245. return {};
  246. }
  247. bool code_point_has_script(u32 code_point, Script script)
  248. {
  249. UErrorCode status = U_ZERO_ERROR;
  250. auto icu_code_point = static_cast<UChar32>(code_point);
  251. auto icu_script = static_cast<UScriptCode>(script.value());
  252. if (auto result = uscript_getScript(icu_code_point, &status); Locale::icu_success(status))
  253. return result == icu_script;
  254. return false;
  255. }
  256. bool code_point_has_script_extension(u32 code_point, Script script)
  257. {
  258. auto icu_code_point = static_cast<UChar32>(code_point);
  259. auto icu_script = static_cast<UScriptCode>(script.value());
  260. return static_cast<bool>(uscript_hasScript(icu_code_point, icu_script));
  261. }
  262. Optional<BidirectionalClass> __attribute__((weak)) bidirectional_class_from_string(StringView) { return {}; }
  263. Optional<BidirectionalClass> __attribute__((weak)) bidirectional_class(u32) { return {}; }
  264. }