| 1 | // © 2018 and later: Unicode, Inc. and others. |
| 2 | // License & terms of use: http://www.unicode.org/copyright.html |
| 3 | |
| 4 | #include "unicode/utypes.h" |
| 5 | |
| 6 | #if !UCONFIG_NO_FORMATTING |
| 7 | |
| 8 | // Allow implicit conversion from char16_t* to UnicodeString for this file: |
| 9 | // Helpful in toString methods and elsewhere. |
| 10 | #define UNISTR_FROM_STRING_EXPLICIT |
| 11 | |
| 12 | #include <typeinfo> |
| 13 | #include <array> |
| 14 | #include "number_types.h" |
| 15 | #include "number_patternstring.h" |
| 16 | #include "numparse_types.h" |
| 17 | #include "numparse_impl.h" |
| 18 | #include "numparse_symbols.h" |
| 19 | #include "numparse_decimal.h" |
| 20 | #include "unicode/numberformatter.h" |
| 21 | #include "cstr.h" |
| 22 | #include "number_mapper.h" |
| 23 | #include "static_unicode_sets.h" |
| 24 | |
| 25 | using namespace icu; |
| 26 | using namespace icu::number; |
| 27 | using namespace icu::number::impl; |
| 28 | using namespace icu::numparse; |
| 29 | using namespace icu::numparse::impl; |
| 30 | |
| 31 | |
| 32 | NumberParseMatcher::~NumberParseMatcher() = default; |
| 33 | |
| 34 | |
| 35 | NumberParserImpl* |
| 36 | NumberParserImpl::createSimpleParser(const Locale& locale, const UnicodeString& patternString, |
| 37 | parse_flags_t parseFlags, UErrorCode& status) { |
| 38 | |
| 39 | LocalPointer<NumberParserImpl> parser(new NumberParserImpl(parseFlags)); |
| 40 | DecimalFormatSymbols symbols(locale, status); |
| 41 | |
| 42 | parser->fLocalMatchers.ignorables = {parseFlags}; |
| 43 | IgnorablesMatcher& ignorables = parser->fLocalMatchers.ignorables; |
| 44 | |
| 45 | DecimalFormatSymbols dfs(locale, status); |
| 46 | dfs.setSymbol(DecimalFormatSymbols::kCurrencySymbol, u"IU$" ); |
| 47 | dfs.setSymbol(DecimalFormatSymbols::kIntlCurrencySymbol, u"ICU" ); |
| 48 | CurrencySymbols currencySymbols({u"ICU" , status}, locale, dfs, status); |
| 49 | |
| 50 | ParsedPatternInfo patternInfo; |
| 51 | PatternParser::parseToPatternInfo(patternString, patternInfo, status); |
| 52 | |
| 53 | // The following statements set up the affix matchers. |
| 54 | AffixTokenMatcherSetupData affixSetupData = { |
| 55 | currencySymbols, symbols, ignorables, locale, parseFlags}; |
| 56 | parser->fLocalMatchers.affixTokenMatcherWarehouse = {&affixSetupData}; |
| 57 | parser->fLocalMatchers.affixMatcherWarehouse = {&parser->fLocalMatchers.affixTokenMatcherWarehouse}; |
| 58 | parser->fLocalMatchers.affixMatcherWarehouse.createAffixMatchers( |
| 59 | patternInfo, *parser, ignorables, parseFlags, status); |
| 60 | |
| 61 | Grouper grouper = Grouper::forStrategy(UNUM_GROUPING_AUTO); |
| 62 | grouper.setLocaleData(patternInfo, locale); |
| 63 | |
| 64 | parser->addMatcher(parser->fLocalMatchers.ignorables); |
| 65 | parser->addMatcher(parser->fLocalMatchers.decimal = {symbols, grouper, parseFlags}); |
| 66 | parser->addMatcher(parser->fLocalMatchers.minusSign = {symbols, false}); |
| 67 | parser->addMatcher(parser->fLocalMatchers.plusSign = {symbols, false}); |
| 68 | parser->addMatcher(parser->fLocalMatchers.percent = {symbols}); |
| 69 | parser->addMatcher(parser->fLocalMatchers.permille = {symbols}); |
| 70 | parser->addMatcher(parser->fLocalMatchers.nan = {symbols}); |
| 71 | parser->addMatcher(parser->fLocalMatchers.infinity = {symbols}); |
| 72 | parser->addMatcher(parser->fLocalMatchers.padding = {u"@" }); |
| 73 | parser->addMatcher(parser->fLocalMatchers.scientific = {symbols, grouper}); |
| 74 | parser->addMatcher(parser->fLocalMatchers.currency = {currencySymbols, symbols, parseFlags, status}); |
| 75 | parser->addMatcher(parser->fLocalValidators.number = {}); |
| 76 | |
| 77 | parser->freeze(); |
| 78 | return parser.orphan(); |
| 79 | } |
| 80 | |
| 81 | NumberParserImpl* |
| 82 | NumberParserImpl::createParserFromProperties(const number::impl::DecimalFormatProperties& properties, |
| 83 | const DecimalFormatSymbols& symbols, bool parseCurrency, |
| 84 | UErrorCode& status) { |
| 85 | Locale locale = symbols.getLocale(); |
| 86 | PropertiesAffixPatternProvider localPAPP; |
| 87 | CurrencyPluralInfoAffixProvider localCPIAP; |
| 88 | AffixPatternProvider* affixProvider; |
| 89 | if (properties.currencyPluralInfo.fPtr.isNull()) { |
| 90 | localPAPP.setTo(properties, status); |
| 91 | affixProvider = &localPAPP; |
| 92 | } else { |
| 93 | localCPIAP.setTo(*properties.currencyPluralInfo.fPtr, properties, status); |
| 94 | affixProvider = &localCPIAP; |
| 95 | } |
| 96 | if (affixProvider == nullptr || U_FAILURE(status)) { return nullptr; } |
| 97 | CurrencyUnit currency = resolveCurrency(properties, locale, status); |
| 98 | CurrencySymbols currencySymbols(currency, locale, symbols, status); |
| 99 | bool isStrict = properties.parseMode.getOrDefault(PARSE_MODE_STRICT) == PARSE_MODE_STRICT; |
| 100 | Grouper grouper = Grouper::forProperties(properties); |
| 101 | int parseFlags = 0; |
| 102 | if (affixProvider == nullptr || U_FAILURE(status)) { return nullptr; } |
| 103 | if (!properties.parseCaseSensitive) { |
| 104 | parseFlags |= PARSE_FLAG_IGNORE_CASE; |
| 105 | } |
| 106 | if (properties.parseIntegerOnly) { |
| 107 | parseFlags |= PARSE_FLAG_INTEGER_ONLY; |
| 108 | } |
| 109 | if (properties.signAlwaysShown) { |
| 110 | parseFlags |= PARSE_FLAG_PLUS_SIGN_ALLOWED; |
| 111 | } |
| 112 | if (isStrict) { |
| 113 | parseFlags |= PARSE_FLAG_STRICT_GROUPING_SIZE; |
| 114 | parseFlags |= PARSE_FLAG_STRICT_SEPARATORS; |
| 115 | parseFlags |= PARSE_FLAG_USE_FULL_AFFIXES; |
| 116 | parseFlags |= PARSE_FLAG_EXACT_AFFIX; |
| 117 | parseFlags |= PARSE_FLAG_STRICT_IGNORABLES; |
| 118 | } else { |
| 119 | parseFlags |= PARSE_FLAG_INCLUDE_UNPAIRED_AFFIXES; |
| 120 | } |
| 121 | if (grouper.getPrimary() <= 0) { |
| 122 | parseFlags |= PARSE_FLAG_GROUPING_DISABLED; |
| 123 | } |
| 124 | if (parseCurrency || affixProvider->hasCurrencySign()) { |
| 125 | parseFlags |= PARSE_FLAG_MONETARY_SEPARATORS; |
| 126 | } |
| 127 | if (!parseCurrency) { |
| 128 | parseFlags |= PARSE_FLAG_NO_FOREIGN_CURRENCY; |
| 129 | } |
| 130 | |
| 131 | LocalPointer<NumberParserImpl> parser(new NumberParserImpl(parseFlags)); |
| 132 | |
| 133 | parser->fLocalMatchers.ignorables = {parseFlags}; |
| 134 | IgnorablesMatcher& ignorables = parser->fLocalMatchers.ignorables; |
| 135 | |
| 136 | ////////////////////// |
| 137 | /// AFFIX MATCHERS /// |
| 138 | ////////////////////// |
| 139 | |
| 140 | // The following statements set up the affix matchers. |
| 141 | AffixTokenMatcherSetupData affixSetupData = { |
| 142 | currencySymbols, symbols, ignorables, locale, parseFlags}; |
| 143 | parser->fLocalMatchers.affixTokenMatcherWarehouse = {&affixSetupData}; |
| 144 | parser->fLocalMatchers.affixMatcherWarehouse = {&parser->fLocalMatchers.affixTokenMatcherWarehouse}; |
| 145 | parser->fLocalMatchers.affixMatcherWarehouse.createAffixMatchers( |
| 146 | *affixProvider, *parser, ignorables, parseFlags, status); |
| 147 | |
| 148 | //////////////////////// |
| 149 | /// CURRENCY MATCHER /// |
| 150 | //////////////////////// |
| 151 | |
| 152 | if (parseCurrency || affixProvider->hasCurrencySign()) { |
| 153 | parser->addMatcher(parser->fLocalMatchers.currency = {currencySymbols, symbols, parseFlags, status}); |
| 154 | } |
| 155 | |
| 156 | /////////////// |
| 157 | /// PERCENT /// |
| 158 | /////////////// |
| 159 | |
| 160 | // ICU-TC meeting, April 11, 2018: accept percent/permille only if it is in the pattern, |
| 161 | // and to maintain regressive behavior, divide by 100 even if no percent sign is present. |
| 162 | if (!isStrict && affixProvider->containsSymbolType(AffixPatternType::TYPE_PERCENT, status)) { |
| 163 | parser->addMatcher(parser->fLocalMatchers.percent = {symbols}); |
| 164 | } |
| 165 | if (!isStrict && affixProvider->containsSymbolType(AffixPatternType::TYPE_PERMILLE, status)) { |
| 166 | parser->addMatcher(parser->fLocalMatchers.permille = {symbols}); |
| 167 | } |
| 168 | |
| 169 | /////////////////////////////// |
| 170 | /// OTHER STANDARD MATCHERS /// |
| 171 | /////////////////////////////// |
| 172 | |
| 173 | if (!isStrict) { |
| 174 | parser->addMatcher(parser->fLocalMatchers.plusSign = {symbols, false}); |
| 175 | parser->addMatcher(parser->fLocalMatchers.minusSign = {symbols, false}); |
| 176 | } |
| 177 | parser->addMatcher(parser->fLocalMatchers.nan = {symbols}); |
| 178 | parser->addMatcher(parser->fLocalMatchers.infinity = {symbols}); |
| 179 | UnicodeString padString = properties.padString; |
| 180 | if (!padString.isBogus() && !ignorables.getSet()->contains(padString)) { |
| 181 | parser->addMatcher(parser->fLocalMatchers.padding = {padString}); |
| 182 | } |
| 183 | parser->addMatcher(parser->fLocalMatchers.ignorables); |
| 184 | parser->addMatcher(parser->fLocalMatchers.decimal = {symbols, grouper, parseFlags}); |
| 185 | // NOTE: parseNoExponent doesn't disable scientific parsing if we have a scientific formatter |
| 186 | if (!properties.parseNoExponent || properties.minimumExponentDigits > 0) { |
| 187 | parser->addMatcher(parser->fLocalMatchers.scientific = {symbols, grouper}); |
| 188 | } |
| 189 | |
| 190 | ////////////////// |
| 191 | /// VALIDATORS /// |
| 192 | ////////////////// |
| 193 | |
| 194 | parser->addMatcher(parser->fLocalValidators.number = {}); |
| 195 | if (isStrict) { |
| 196 | parser->addMatcher(parser->fLocalValidators.affix = {}); |
| 197 | } |
| 198 | if (parseCurrency) { |
| 199 | parser->addMatcher(parser->fLocalValidators.currency = {}); |
| 200 | } |
| 201 | if (properties.decimalPatternMatchRequired) { |
| 202 | bool patternHasDecimalSeparator = |
| 203 | properties.decimalSeparatorAlwaysShown || properties.maximumFractionDigits != 0; |
| 204 | parser->addMatcher(parser->fLocalValidators.decimalSeparator = {patternHasDecimalSeparator}); |
| 205 | } |
| 206 | // The multiplier takes care of scaling percentages. |
| 207 | Scale multiplier = scaleFromProperties(properties); |
| 208 | if (multiplier.isValid()) { |
| 209 | parser->addMatcher(parser->fLocalValidators.multiplier = {multiplier}); |
| 210 | } |
| 211 | |
| 212 | parser->freeze(); |
| 213 | return parser.orphan(); |
| 214 | } |
| 215 | |
| 216 | NumberParserImpl::NumberParserImpl(parse_flags_t parseFlags) |
| 217 | : fParseFlags(parseFlags) { |
| 218 | } |
| 219 | |
| 220 | NumberParserImpl::~NumberParserImpl() { |
| 221 | fNumMatchers = 0; |
| 222 | } |
| 223 | |
| 224 | void NumberParserImpl::addMatcher(NumberParseMatcher& matcher) { |
| 225 | if (fNumMatchers + 1 > fMatchers.getCapacity()) { |
| 226 | fMatchers.resize(fNumMatchers * 2, fNumMatchers); |
| 227 | } |
| 228 | fMatchers[fNumMatchers] = &matcher; |
| 229 | fNumMatchers++; |
| 230 | } |
| 231 | |
| 232 | void NumberParserImpl::freeze() { |
| 233 | fFrozen = true; |
| 234 | } |
| 235 | |
| 236 | parse_flags_t NumberParserImpl::getParseFlags() const { |
| 237 | return fParseFlags; |
| 238 | } |
| 239 | |
| 240 | void NumberParserImpl::parse(const UnicodeString& input, bool greedy, ParsedNumber& result, |
| 241 | UErrorCode& status) const { |
| 242 | return parse(input, 0, greedy, result, status); |
| 243 | } |
| 244 | |
| 245 | void NumberParserImpl::parse(const UnicodeString& input, int32_t start, bool greedy, ParsedNumber& result, |
| 246 | UErrorCode& status) const { |
| 247 | if (U_FAILURE(status)) { |
| 248 | return; |
| 249 | } |
| 250 | U_ASSERT(fFrozen); |
| 251 | // TODO: Check start >= 0 and start < input.length() |
| 252 | StringSegment segment(input, 0 != (fParseFlags & PARSE_FLAG_IGNORE_CASE)); |
| 253 | segment.adjustOffset(start); |
| 254 | if (greedy) { |
| 255 | parseGreedy(segment, result, status); |
| 256 | } else if (0 != (fParseFlags & PARSE_FLAG_ALLOW_INFINITE_RECURSION)) { |
| 257 | // Start at 1 so that recursionLevels never gets to 0 |
| 258 | parseLongestRecursive(segment, result, 1, status); |
| 259 | } else { |
| 260 | // Arbitrary recursion safety limit: 100 levels. |
| 261 | parseLongestRecursive(segment, result, -100, status); |
| 262 | } |
| 263 | for (int32_t i = 0; i < fNumMatchers; i++) { |
| 264 | fMatchers[i]->postProcess(result); |
| 265 | } |
| 266 | result.postProcess(); |
| 267 | } |
| 268 | |
| 269 | void NumberParserImpl::parseGreedy(StringSegment& segment, ParsedNumber& result, |
| 270 | UErrorCode& status) const { |
| 271 | // Note: this method is not recursive in order to avoid stack overflow. |
| 272 | for (int i = 0; i <fNumMatchers;) { |
| 273 | // Base Case |
| 274 | if (segment.length() == 0) { |
| 275 | return; |
| 276 | } |
| 277 | const NumberParseMatcher* matcher = fMatchers[i]; |
| 278 | if (!matcher->smokeTest(segment)) { |
| 279 | // Matcher failed smoke test: try the next one |
| 280 | i++; |
| 281 | continue; |
| 282 | } |
| 283 | int32_t initialOffset = segment.getOffset(); |
| 284 | matcher->match(segment, result, status); |
| 285 | if (U_FAILURE(status)) { |
| 286 | return; |
| 287 | } |
| 288 | if (segment.getOffset() != initialOffset) { |
| 289 | // Greedy heuristic: accept the match and loop back |
| 290 | i = 0; |
| 291 | continue; |
| 292 | } else { |
| 293 | // Matcher did not match: try the next one |
| 294 | i++; |
| 295 | continue; |
| 296 | } |
| 297 | UPRV_UNREACHABLE; |
| 298 | } |
| 299 | |
| 300 | // NOTE: If we get here, the greedy parse completed without consuming the entire string. |
| 301 | } |
| 302 | |
| 303 | void NumberParserImpl::parseLongestRecursive(StringSegment& segment, ParsedNumber& result, |
| 304 | int32_t recursionLevels, |
| 305 | UErrorCode& status) const { |
| 306 | // Base Case |
| 307 | if (segment.length() == 0) { |
| 308 | return; |
| 309 | } |
| 310 | |
| 311 | // Safety against stack overflow |
| 312 | if (recursionLevels == 0) { |
| 313 | return; |
| 314 | } |
| 315 | |
| 316 | // TODO: Give a nice way for the matcher to reset the ParsedNumber? |
| 317 | ParsedNumber initial(result); |
| 318 | ParsedNumber candidate; |
| 319 | |
| 320 | int initialOffset = segment.getOffset(); |
| 321 | for (int32_t i = 0; i < fNumMatchers; i++) { |
| 322 | const NumberParseMatcher* matcher = fMatchers[i]; |
| 323 | if (!matcher->smokeTest(segment)) { |
| 324 | continue; |
| 325 | } |
| 326 | |
| 327 | // In a non-greedy parse, we attempt all possible matches and pick the best. |
| 328 | for (int32_t charsToConsume = 0; charsToConsume < segment.length();) { |
| 329 | charsToConsume += U16_LENGTH(segment.codePointAt(charsToConsume)); |
| 330 | |
| 331 | // Run the matcher on a segment of the current length. |
| 332 | candidate = initial; |
| 333 | segment.setLength(charsToConsume); |
| 334 | bool maybeMore = matcher->match(segment, candidate, status); |
| 335 | segment.resetLength(); |
| 336 | if (U_FAILURE(status)) { |
| 337 | return; |
| 338 | } |
| 339 | |
| 340 | // If the entire segment was consumed, recurse. |
| 341 | if (segment.getOffset() - initialOffset == charsToConsume) { |
| 342 | parseLongestRecursive(segment, candidate, recursionLevels + 1, status); |
| 343 | if (U_FAILURE(status)) { |
| 344 | return; |
| 345 | } |
| 346 | if (candidate.isBetterThan(result)) { |
| 347 | result = candidate; |
| 348 | } |
| 349 | } |
| 350 | |
| 351 | // Since the segment can be re-used, reset the offset. |
| 352 | // This does not have an effect if the matcher did not consume any chars. |
| 353 | segment.setOffset(initialOffset); |
| 354 | |
| 355 | // Unless the matcher wants to see the next char, continue to the next matcher. |
| 356 | if (!maybeMore) { |
| 357 | break; |
| 358 | } |
| 359 | } |
| 360 | } |
| 361 | } |
| 362 | |
| 363 | UnicodeString NumberParserImpl::toString() const { |
| 364 | UnicodeString result(u"<NumberParserImpl matchers:[" ); |
| 365 | for (int32_t i = 0; i < fNumMatchers; i++) { |
| 366 | result.append(u' '); |
| 367 | result.append(fMatchers[i]->toString()); |
| 368 | } |
| 369 | result.append(u" ]>" , -1); |
| 370 | return result; |
| 371 | } |
| 372 | |
| 373 | |
| 374 | #endif /* #if !UCONFIG_NO_FORMATTING */ |
| 375 | |