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 "numparse_types.h" |
13 | #include "numparse_decimal.h" |
14 | #include "static_unicode_sets.h" |
15 | #include "numparse_utils.h" |
16 | #include "unicode/uchar.h" |
17 | #include "putilimp.h" |
18 | #include "number_decimalquantity.h" |
19 | #include "string_segment.h" |
20 | |
21 | using namespace icu; |
22 | using namespace icu::numparse; |
23 | using namespace icu::numparse::impl; |
24 | |
25 | |
26 | DecimalMatcher::DecimalMatcher(const DecimalFormatSymbols& symbols, const Grouper& grouper, |
27 | parse_flags_t parseFlags) { |
28 | if (0 != (parseFlags & PARSE_FLAG_MONETARY_SEPARATORS)) { |
29 | groupingSeparator = symbols.getConstSymbol(DecimalFormatSymbols::kMonetaryGroupingSeparatorSymbol); |
30 | decimalSeparator = symbols.getConstSymbol(DecimalFormatSymbols::kMonetarySeparatorSymbol); |
31 | } else { |
32 | groupingSeparator = symbols.getConstSymbol(DecimalFormatSymbols::kGroupingSeparatorSymbol); |
33 | decimalSeparator = symbols.getConstSymbol(DecimalFormatSymbols::kDecimalSeparatorSymbol); |
34 | } |
35 | bool strictSeparators = 0 != (parseFlags & PARSE_FLAG_STRICT_SEPARATORS); |
36 | unisets::Key groupingKey = strictSeparators ? unisets::STRICT_ALL_SEPARATORS |
37 | : unisets::ALL_SEPARATORS; |
38 | |
39 | // Attempt to find separators in the static cache |
40 | |
41 | groupingUniSet = unisets::get(groupingKey); |
42 | unisets::Key decimalKey = unisets::chooseFrom( |
43 | decimalSeparator, |
44 | strictSeparators ? unisets::STRICT_COMMA : unisets::COMMA, |
45 | strictSeparators ? unisets::STRICT_PERIOD : unisets::PERIOD); |
46 | if (decimalKey >= 0) { |
47 | decimalUniSet = unisets::get(decimalKey); |
48 | } else if (!decimalSeparator.isEmpty()) { |
49 | auto* set = new UnicodeSet(); |
50 | set->add(decimalSeparator.char32At(0)); |
51 | set->freeze(); |
52 | decimalUniSet = set; |
53 | fLocalDecimalUniSet.adoptInstead(set); |
54 | } else { |
55 | decimalUniSet = unisets::get(unisets::EMPTY); |
56 | } |
57 | |
58 | if (groupingKey >= 0 && decimalKey >= 0) { |
59 | // Everything is available in the static cache |
60 | separatorSet = groupingUniSet; |
61 | leadSet = unisets::get( |
62 | strictSeparators ? unisets::DIGITS_OR_ALL_SEPARATORS |
63 | : unisets::DIGITS_OR_STRICT_ALL_SEPARATORS); |
64 | } else { |
65 | auto* set = new UnicodeSet(); |
66 | set->addAll(*groupingUniSet); |
67 | set->addAll(*decimalUniSet); |
68 | set->freeze(); |
69 | separatorSet = set; |
70 | fLocalSeparatorSet.adoptInstead(set); |
71 | leadSet = nullptr; |
72 | } |
73 | |
74 | UChar32 cpZero = symbols.getCodePointZero(); |
75 | if (cpZero == -1 || !u_isdigit(cpZero) || u_digit(cpZero, 10) != 0) { |
76 | // Uncommon case: okay to allocate. |
77 | auto digitStrings = new UnicodeString[10]; |
78 | fLocalDigitStrings.adoptInstead(digitStrings); |
79 | for (int32_t i = 0; i <= 9; i++) { |
80 | digitStrings[i] = symbols.getConstDigitSymbol(i); |
81 | } |
82 | } |
83 | |
84 | requireGroupingMatch = 0 != (parseFlags & PARSE_FLAG_STRICT_GROUPING_SIZE); |
85 | groupingDisabled = 0 != (parseFlags & PARSE_FLAG_GROUPING_DISABLED); |
86 | integerOnly = 0 != (parseFlags & PARSE_FLAG_INTEGER_ONLY); |
87 | grouping1 = grouper.getPrimary(); |
88 | grouping2 = grouper.getSecondary(); |
89 | |
90 | // Fraction grouping parsing is disabled for now but could be enabled later. |
91 | // See http://bugs.icu-project.org/trac/ticket/10794 |
92 | // fractionGrouping = 0 != (parseFlags & PARSE_FLAG_FRACTION_GROUPING_ENABLED); |
93 | } |
94 | |
95 | bool DecimalMatcher::match(StringSegment& segment, ParsedNumber& result, UErrorCode& status) const { |
96 | return match(segment, result, 0, status); |
97 | } |
98 | |
99 | bool DecimalMatcher::match(StringSegment& segment, ParsedNumber& result, int8_t exponentSign, |
100 | UErrorCode&) const { |
101 | if (result.seenNumber() && exponentSign == 0) { |
102 | // A number has already been consumed. |
103 | return false; |
104 | } else if (exponentSign != 0) { |
105 | // scientific notation always comes after the number |
106 | U_ASSERT(!result.quantity.bogus); |
107 | } |
108 | |
109 | // Initial offset before any character consumption. |
110 | int32_t initialOffset = segment.getOffset(); |
111 | |
112 | // Return value: whether to ask for more characters. |
113 | bool maybeMore = false; |
114 | |
115 | // All digits consumed so far. |
116 | number::impl::DecimalQuantity digitsConsumed; |
117 | digitsConsumed.bogus = true; |
118 | |
119 | // The total number of digits after the decimal place, used for scaling the result. |
120 | int32_t digitsAfterDecimalPlace = 0; |
121 | |
122 | // The actual grouping and decimal separators used in the string. |
123 | // If non-null, we have seen that token. |
124 | UnicodeString actualGroupingString; |
125 | UnicodeString actualDecimalString; |
126 | actualGroupingString.setToBogus(); |
127 | actualDecimalString.setToBogus(); |
128 | |
129 | // Information for two groups: the previous group and the current group. |
130 | // |
131 | // Each group has three pieces of information: |
132 | // |
133 | // Offset: the string position of the beginning of the group, including a leading separator |
134 | // if there was a leading separator. This is needed in case we need to rewind the parse to |
135 | // that position. |
136 | // |
137 | // Separator type: |
138 | // 0 => beginning of string |
139 | // 1 => lead separator is a grouping separator |
140 | // 2 => lead separator is a decimal separator |
141 | // |
142 | // Count: the number of digits in the group. If -1, the group has been validated. |
143 | int32_t currGroupOffset = 0; |
144 | int32_t currGroupSepType = 0; |
145 | int32_t currGroupCount = 0; |
146 | int32_t prevGroupOffset = -1; |
147 | int32_t prevGroupSepType = -1; |
148 | int32_t prevGroupCount = -1; |
149 | |
150 | while (segment.length() > 0) { |
151 | maybeMore = false; |
152 | |
153 | // Attempt to match a digit. |
154 | int8_t digit = -1; |
155 | |
156 | // Try by code point digit value. |
157 | UChar32 cp = segment.getCodePoint(); |
158 | if (u_isdigit(cp)) { |
159 | segment.adjustOffset(U16_LENGTH(cp)); |
160 | digit = static_cast<int8_t>(u_digit(cp, 10)); |
161 | } |
162 | |
163 | // Try by digit string. |
164 | if (digit == -1 && !fLocalDigitStrings.isNull()) { |
165 | for (int32_t i = 0; i < 10; i++) { |
166 | const UnicodeString& str = fLocalDigitStrings[i]; |
167 | if (str.isEmpty()) { |
168 | continue; |
169 | } |
170 | int32_t overlap = segment.getCommonPrefixLength(str); |
171 | if (overlap == str.length()) { |
172 | segment.adjustOffset(overlap); |
173 | digit = static_cast<int8_t>(i); |
174 | break; |
175 | } |
176 | maybeMore = maybeMore || (overlap == segment.length()); |
177 | } |
178 | } |
179 | |
180 | if (digit >= 0) { |
181 | // Digit was found. |
182 | if (digitsConsumed.bogus) { |
183 | digitsConsumed.bogus = false; |
184 | digitsConsumed.clear(); |
185 | } |
186 | digitsConsumed.appendDigit(digit, 0, true); |
187 | currGroupCount++; |
188 | if (!actualDecimalString.isBogus()) { |
189 | digitsAfterDecimalPlace++; |
190 | } |
191 | continue; |
192 | } |
193 | |
194 | // Attempt to match a literal grouping or decimal separator. |
195 | bool isDecimal = false; |
196 | bool isGrouping = false; |
197 | |
198 | // 1) Attempt the decimal separator string literal. |
199 | // if (we have not seen a decimal separator yet) { ... } |
200 | if (actualDecimalString.isBogus() && !decimalSeparator.isEmpty()) { |
201 | int32_t overlap = segment.getCommonPrefixLength(decimalSeparator); |
202 | maybeMore = maybeMore || (overlap == segment.length()); |
203 | if (overlap == decimalSeparator.length()) { |
204 | isDecimal = true; |
205 | actualDecimalString = decimalSeparator; |
206 | } |
207 | } |
208 | |
209 | // 2) Attempt to match the actual grouping string literal. |
210 | if (!actualGroupingString.isBogus()) { |
211 | int32_t overlap = segment.getCommonPrefixLength(actualGroupingString); |
212 | maybeMore = maybeMore || (overlap == segment.length()); |
213 | if (overlap == actualGroupingString.length()) { |
214 | isGrouping = true; |
215 | } |
216 | } |
217 | |
218 | // 2.5) Attempt to match a new the grouping separator string literal. |
219 | // if (we have not seen a grouping or decimal separator yet) { ... } |
220 | if (!groupingDisabled && actualGroupingString.isBogus() && actualDecimalString.isBogus() && |
221 | !groupingSeparator.isEmpty()) { |
222 | int32_t overlap = segment.getCommonPrefixLength(groupingSeparator); |
223 | maybeMore = maybeMore || (overlap == segment.length()); |
224 | if (overlap == groupingSeparator.length()) { |
225 | isGrouping = true; |
226 | actualGroupingString = groupingSeparator; |
227 | } |
228 | } |
229 | |
230 | // 3) Attempt to match a decimal separator from the equivalence set. |
231 | // if (we have not seen a decimal separator yet) { ... } |
232 | // The !isGrouping is to confirm that we haven't yet matched the current character. |
233 | if (!isGrouping && actualDecimalString.isBogus()) { |
234 | if (decimalUniSet->contains(cp)) { |
235 | isDecimal = true; |
236 | actualDecimalString = UnicodeString(cp); |
237 | } |
238 | } |
239 | |
240 | // 4) Attempt to match a grouping separator from the equivalence set. |
241 | // if (we have not seen a grouping or decimal separator yet) { ... } |
242 | if (!groupingDisabled && actualGroupingString.isBogus() && actualDecimalString.isBogus()) { |
243 | if (groupingUniSet->contains(cp)) { |
244 | isGrouping = true; |
245 | actualGroupingString = UnicodeString(cp); |
246 | } |
247 | } |
248 | |
249 | // Leave if we failed to match this as a separator. |
250 | if (!isDecimal && !isGrouping) { |
251 | break; |
252 | } |
253 | |
254 | // Check for conditions when we don't want to accept the separator. |
255 | if (isDecimal && integerOnly) { |
256 | break; |
257 | } else if (currGroupSepType == 2 && isGrouping) { |
258 | // Fraction grouping |
259 | break; |
260 | } |
261 | |
262 | // Validate intermediate grouping sizes. |
263 | bool prevValidSecondary = validateGroup(prevGroupSepType, prevGroupCount, false); |
264 | bool currValidPrimary = validateGroup(currGroupSepType, currGroupCount, true); |
265 | if (!prevValidSecondary || (isDecimal && !currValidPrimary)) { |
266 | // Invalid grouping sizes. |
267 | if (isGrouping && currGroupCount == 0) { |
268 | // Trailing grouping separators: these are taken care of below |
269 | U_ASSERT(currGroupSepType == 1); |
270 | } else if (requireGroupingMatch) { |
271 | // Strict mode: reject the parse |
272 | digitsConsumed.clear(); |
273 | digitsConsumed.bogus = true; |
274 | } |
275 | break; |
276 | } else if (requireGroupingMatch && currGroupCount == 0 && currGroupSepType == 1) { |
277 | break; |
278 | } else { |
279 | // Grouping sizes OK so far. |
280 | prevGroupOffset = currGroupOffset; |
281 | prevGroupCount = currGroupCount; |
282 | if (isDecimal) { |
283 | // Do not validate this group any more. |
284 | prevGroupSepType = -1; |
285 | } else { |
286 | prevGroupSepType = currGroupSepType; |
287 | } |
288 | } |
289 | |
290 | // OK to accept the separator. |
291 | // Special case: don't update currGroup if it is empty; this allows two grouping |
292 | // separators in a row in lenient mode. |
293 | if (currGroupCount != 0) { |
294 | currGroupOffset = segment.getOffset(); |
295 | } |
296 | currGroupSepType = isGrouping ? 1 : 2; |
297 | currGroupCount = 0; |
298 | if (isGrouping) { |
299 | segment.adjustOffset(actualGroupingString.length()); |
300 | } else { |
301 | segment.adjustOffset(actualDecimalString.length()); |
302 | } |
303 | } |
304 | |
305 | // End of main loop. |
306 | // Back up if there was a trailing grouping separator. |
307 | // Shift prev -> curr so we can check it as a final group. |
308 | if (currGroupSepType != 2 && currGroupCount == 0) { |
309 | maybeMore = true; |
310 | segment.setOffset(currGroupOffset); |
311 | currGroupOffset = prevGroupOffset; |
312 | currGroupSepType = prevGroupSepType; |
313 | currGroupCount = prevGroupCount; |
314 | prevGroupOffset = -1; |
315 | prevGroupSepType = 0; |
316 | prevGroupCount = 1; |
317 | } |
318 | |
319 | // Validate final grouping sizes. |
320 | bool prevValidSecondary = validateGroup(prevGroupSepType, prevGroupCount, false); |
321 | bool currValidPrimary = validateGroup(currGroupSepType, currGroupCount, true); |
322 | if (!requireGroupingMatch) { |
323 | // The cases we need to handle here are lone digits. |
324 | // Examples: "1,1" "1,1," "1,1,1" "1,1,1," ",1" (all parse as 1) |
325 | // See more examples in numberformattestspecification.txt |
326 | int32_t digitsToRemove = 0; |
327 | if (!prevValidSecondary) { |
328 | segment.setOffset(prevGroupOffset); |
329 | digitsToRemove += prevGroupCount; |
330 | digitsToRemove += currGroupCount; |
331 | } else if (!currValidPrimary && (prevGroupSepType != 0 || prevGroupCount != 0)) { |
332 | maybeMore = true; |
333 | segment.setOffset(currGroupOffset); |
334 | digitsToRemove += currGroupCount; |
335 | } |
336 | if (digitsToRemove != 0) { |
337 | digitsConsumed.adjustMagnitude(-digitsToRemove); |
338 | digitsConsumed.truncate(); |
339 | } |
340 | prevValidSecondary = true; |
341 | currValidPrimary = true; |
342 | } |
343 | if (currGroupSepType != 2 && (!prevValidSecondary || !currValidPrimary)) { |
344 | // Grouping failure. |
345 | digitsConsumed.bogus = true; |
346 | } |
347 | |
348 | // Strings that start with a separator but have no digits, |
349 | // or strings that failed a grouping size check. |
350 | if (digitsConsumed.bogus) { |
351 | maybeMore = maybeMore || (segment.length() == 0); |
352 | segment.setOffset(initialOffset); |
353 | return maybeMore; |
354 | } |
355 | |
356 | // We passed all inspections. Start post-processing. |
357 | |
358 | // Adjust for fraction part. |
359 | digitsConsumed.adjustMagnitude(-digitsAfterDecimalPlace); |
360 | |
361 | // Set the digits, either normal or exponent. |
362 | if (exponentSign != 0 && segment.getOffset() != initialOffset) { |
363 | bool overflow = false; |
364 | if (digitsConsumed.fitsInLong()) { |
365 | int64_t exponentLong = digitsConsumed.toLong(false); |
366 | U_ASSERT(exponentLong >= 0); |
367 | if (exponentLong <= INT32_MAX) { |
368 | auto exponentInt = static_cast<int32_t>(exponentLong); |
369 | if (result.quantity.adjustMagnitude(exponentSign * exponentInt)) { |
370 | overflow = true; |
371 | } |
372 | } else { |
373 | overflow = true; |
374 | } |
375 | } else { |
376 | overflow = true; |
377 | } |
378 | if (overflow) { |
379 | if (exponentSign == -1) { |
380 | // Set to zero |
381 | result.quantity.clear(); |
382 | } else { |
383 | // Set to infinity |
384 | result.quantity.bogus = true; |
385 | result.flags |= FLAG_INFINITY; |
386 | } |
387 | } |
388 | } else { |
389 | result.quantity = digitsConsumed; |
390 | } |
391 | |
392 | // Set other information into the result and return. |
393 | if (!actualDecimalString.isBogus()) { |
394 | result.flags |= FLAG_HAS_DECIMAL_SEPARATOR; |
395 | } |
396 | result.setCharsConsumed(segment); |
397 | return segment.length() == 0 || maybeMore; |
398 | } |
399 | |
400 | bool DecimalMatcher::validateGroup(int32_t sepType, int32_t count, bool isPrimary) const { |
401 | if (requireGroupingMatch) { |
402 | if (sepType == -1) { |
403 | // No such group (prevGroup before first shift). |
404 | return true; |
405 | } else if (sepType == 0) { |
406 | // First group. |
407 | if (isPrimary) { |
408 | // No grouping separators is OK. |
409 | return true; |
410 | } else { |
411 | return count != 0 && count <= grouping2; |
412 | } |
413 | } else if (sepType == 1) { |
414 | // Middle group. |
415 | if (isPrimary) { |
416 | return count == grouping1; |
417 | } else { |
418 | return count == grouping2; |
419 | } |
420 | } else { |
421 | U_ASSERT(sepType == 2); |
422 | // After the decimal separator. |
423 | return true; |
424 | } |
425 | } else { |
426 | if (sepType == 1) { |
427 | // #11230: don't accept middle groups with only 1 digit. |
428 | return count != 1; |
429 | } else { |
430 | return true; |
431 | } |
432 | } |
433 | } |
434 | |
435 | bool DecimalMatcher::smokeTest(const StringSegment& segment) const { |
436 | // The common case uses a static leadSet for efficiency. |
437 | if (fLocalDigitStrings.isNull() && leadSet != nullptr) { |
438 | return segment.startsWith(*leadSet); |
439 | } |
440 | if (segment.startsWith(*separatorSet) || u_isdigit(segment.getCodePoint())) { |
441 | return true; |
442 | } |
443 | if (fLocalDigitStrings.isNull()) { |
444 | return false; |
445 | } |
446 | for (int32_t i = 0; i < 10; i++) { |
447 | if (segment.startsWith(fLocalDigitStrings[i])) { |
448 | return true; |
449 | } |
450 | } |
451 | return false; |
452 | } |
453 | |
454 | UnicodeString DecimalMatcher::toString() const { |
455 | return u"<Decimal>" ; |
456 | } |
457 | |
458 | |
459 | #endif /* #if !UCONFIG_NO_FORMATTING */ |
460 | |