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 #ifndef __NUMPARSE_TYPES_H__ 8 #define __NUMPARSE_TYPES_H__ 9 10 #include "unicode/uobject.h" 11 #include "number_decimalquantity.h" 12 #include "string_segment.h" 13 14 U_NAMESPACE_BEGIN 15 namespace numparse { 16 namespace impl { 17 18 // Forward-declarations 19 class ParsedNumber; 20 21 typedef int32_t result_flags_t; 22 typedef int32_t parse_flags_t; 23 24 /** Flags for the type result_flags_t */ 25 enum ResultFlags { 26 FLAG_NEGATIVE = 0x0001, 27 FLAG_PERCENT = 0x0002, 28 FLAG_PERMILLE = 0x0004, 29 FLAG_HAS_EXPONENT = 0x0008, 30 // FLAG_HAS_DEFAULT_CURRENCY = 0x0010, // no longer used 31 FLAG_HAS_DECIMAL_SEPARATOR = 0x0020, 32 FLAG_NAN = 0x0040, 33 FLAG_INFINITY = 0x0080, 34 FLAG_FAIL = 0x0100, 35 }; 36 37 /** Flags for the type parse_flags_t */ 38 enum ParseFlags { 39 PARSE_FLAG_IGNORE_CASE = 0x0001, 40 PARSE_FLAG_MONETARY_SEPARATORS = 0x0002, 41 PARSE_FLAG_STRICT_SEPARATORS = 0x0004, 42 PARSE_FLAG_STRICT_GROUPING_SIZE = 0x0008, 43 PARSE_FLAG_INTEGER_ONLY = 0x0010, 44 PARSE_FLAG_GROUPING_DISABLED = 0x0020, 45 // PARSE_FLAG_FRACTION_GROUPING_ENABLED = 0x0040, // see #10794 46 PARSE_FLAG_INCLUDE_UNPAIRED_AFFIXES = 0x0080, 47 PARSE_FLAG_USE_FULL_AFFIXES = 0x0100, 48 PARSE_FLAG_EXACT_AFFIX = 0x0200, 49 PARSE_FLAG_PLUS_SIGN_ALLOWED = 0x0400, 50 // PARSE_FLAG_OPTIMIZE = 0x0800, // no longer used 51 // PARSE_FLAG_FORCE_BIG_DECIMAL = 0x1000, // not used in ICU4C 52 PARSE_FLAG_NO_FOREIGN_CURRENCY = 0x2000, 53 PARSE_FLAG_ALLOW_INFINITE_RECURSION = 0x4000, 54 PARSE_FLAG_STRICT_IGNORABLES = 0x8000, 55 }; 56 57 58 // TODO: Is this class worthwhile? 59 template<int32_t stackCapacity> 60 class CompactUnicodeString { 61 public: CompactUnicodeString()62 CompactUnicodeString() { 63 static_assert(stackCapacity > 0, "cannot have zero space on stack"); 64 fBuffer[0] = 0; 65 } 66 CompactUnicodeString(const UnicodeString & text)67 CompactUnicodeString(const UnicodeString& text) 68 : fBuffer(text.length() + 1) { 69 uprv_memcpy(fBuffer.getAlias(), text.getBuffer(), sizeof(UChar) * text.length()); 70 fBuffer[text.length()] = 0; 71 } 72 toAliasedUnicodeString()73 inline UnicodeString toAliasedUnicodeString() const { 74 return UnicodeString(TRUE, fBuffer.getAlias(), -1); 75 } 76 77 bool operator==(const CompactUnicodeString& other) const { 78 // Use the alias-only constructor and then call UnicodeString operator== 79 return toAliasedUnicodeString() == other.toAliasedUnicodeString(); 80 } 81 82 private: 83 MaybeStackArray<UChar, stackCapacity> fBuffer; 84 }; 85 86 87 /** 88 * Struct-like class to hold the results of a parsing routine. 89 * 90 * @author sffc 91 */ 92 // Exported as U_I18N_API for tests 93 class U_I18N_API ParsedNumber { 94 public: 95 96 /** 97 * The numerical value that was parsed. 98 */ 99 ::icu::number::impl::DecimalQuantity quantity; 100 101 /** 102 * The index of the last char consumed during parsing. If parsing started at index 0, this is equal 103 * to the number of chars consumed. This is NOT necessarily the same as the StringSegment offset; 104 * "weak" chars, like whitespace, change the offset, but the charsConsumed is not touched until a 105 * "strong" char is encountered. 106 */ 107 int32_t charEnd; 108 109 /** 110 * Boolean flags (see constants above). 111 */ 112 result_flags_t flags; 113 114 /** 115 * The pattern string corresponding to the prefix that got consumed. 116 */ 117 UnicodeString prefix; 118 119 /** 120 * The pattern string corresponding to the suffix that got consumed. 121 */ 122 UnicodeString suffix; 123 124 /** 125 * The currency that got consumed. 126 */ 127 UChar currencyCode[4]; 128 129 ParsedNumber(); 130 131 ParsedNumber(const ParsedNumber& other) = default; 132 133 ParsedNumber& operator=(const ParsedNumber& other) = default; 134 135 void clear(); 136 137 /** 138 * Call this method to register that a "strong" char was consumed. This should be done after calling 139 * {@link StringSegment#setOffset} or {@link StringSegment#adjustOffset} except when the char is 140 * "weak", like whitespace. 141 * 142 * <p> 143 * <strong>What is a strong versus weak char?</strong> The behavior of number parsing is to "stop" 144 * after reading the number, even if there is other content following the number. For example, after 145 * parsing the string "123 " (123 followed by a space), the cursor should be set to 3, not 4, even 146 * though there are matchers that accept whitespace. In this example, the digits are strong, whereas 147 * the whitespace is weak. Grouping separators are weak, whereas decimal separators are strong. Most 148 * other chars are strong. 149 * 150 * @param segment 151 * The current StringSegment, usually immediately following a call to setOffset. 152 */ 153 void setCharsConsumed(const StringSegment& segment); 154 155 /** Apply certain number-related flags to the DecimalQuantity. */ 156 void postProcess(); 157 158 /** 159 * Returns whether this the parse was successful. To be successful, at least one char must have been 160 * consumed, and the failure flag must not be set. 161 */ 162 bool success() const; 163 164 bool seenNumber() const; 165 166 double getDouble(UErrorCode& status) const; 167 168 void populateFormattable(Formattable& output, parse_flags_t parseFlags) const; 169 170 bool isBetterThan(const ParsedNumber& other); 171 }; 172 173 174 /** 175 * The core interface implemented by all matchers used for number parsing. 176 * 177 * Given a string, there should NOT be more than one way to consume the string with the same matcher 178 * applied multiple times. If there is, the non-greedy parsing algorithm will be unhappy and may enter an 179 * exponential-time loop. For example, consider the "A Matcher" that accepts "any number of As". Given 180 * the string "AAAA", there are 2^N = 8 ways to apply the A Matcher to this string: you could have the A 181 * Matcher apply 4 times to each character; you could have it apply just once to all the characters; you 182 * could have it apply to the first 2 characters and the second 2 characters; and so on. A better version 183 * of the "A Matcher" would be for it to accept exactly one A, and allow the algorithm to run it 184 * repeatedly to consume a string of multiple As. The A Matcher can implement the Flexible interface 185 * below to signal that it can be applied multiple times in a row. 186 * 187 * @author sffc 188 */ 189 // Exported as U_I18N_API for tests 190 class U_I18N_API NumberParseMatcher { 191 public: 192 virtual ~NumberParseMatcher(); 193 194 /** 195 * Matchers can override this method to return true to indicate that they are optional and can be run 196 * repeatedly. Used by SeriesMatcher, primarily in the context of IgnorablesMatcher. 197 */ isFlexible()198 virtual bool isFlexible() const { 199 return false; 200 } 201 202 /** 203 * Runs this matcher starting at the beginning of the given StringSegment. If this matcher finds 204 * something interesting in the StringSegment, it should update the offset of the StringSegment 205 * corresponding to how many chars were matched. 206 * 207 * This method is thread-safe. 208 * 209 * @param segment 210 * The StringSegment to match against. Matches always start at the beginning of the 211 * segment. The segment is guaranteed to contain at least one char. 212 * @param result 213 * The data structure to store results if the match succeeds. 214 * @return Whether this matcher thinks there may be more interesting chars beyond the end of the 215 * string segment. 216 */ 217 virtual bool match(StringSegment& segment, ParsedNumber& result, UErrorCode& status) const = 0; 218 219 /** 220 * Performs a fast "smoke check" for whether or not this matcher could possibly match against the 221 * given string segment. The test should be as fast as possible but also as restrictive as possible. 222 * For example, matchers can maintain a UnicodeSet of all code points that count possibly start a 223 * match. Matchers should use the {@link StringSegment#startsWith} method in order to correctly 224 * handle case folding. 225 * 226 * @param segment 227 * The segment to check against. 228 * @return true if the matcher might be able to match against this segment; false if it definitely 229 * will not be able to match. 230 */ 231 virtual bool smokeTest(const StringSegment& segment) const = 0; 232 233 /** 234 * Method called at the end of a parse, after all matchers have failed to consume any more chars. 235 * Allows a matcher to make final modifications to the result given the knowledge that no more 236 * matches are possible. 237 * 238 * @param result 239 * The data structure to store results. 240 */ postProcess(ParsedNumber &)241 virtual void postProcess(ParsedNumber&) const { 242 // Default implementation: no-op 243 } 244 245 // String for debugging 246 virtual UnicodeString toString() const = 0; 247 248 protected: 249 // No construction except by subclasses! 250 NumberParseMatcher() = default; 251 }; 252 253 254 /** 255 * Interface for use in arguments. 256 */ 257 // Exported as U_I18N_API for tests 258 class U_I18N_API MutableMatcherCollection { 259 public: 260 virtual ~MutableMatcherCollection() = default; 261 262 virtual void addMatcher(NumberParseMatcher& matcher) = 0; 263 }; 264 265 266 } // namespace impl 267 } // namespace numparse 268 U_NAMESPACE_END 269 270 #endif //__NUMPARSE_TYPES_H__ 271 #endif /* #if !UCONFIG_NO_FORMATTING */ 272