]>
Commit | Line | Data |
---|---|---|
0f5d89e8 A |
1 | // © 2018 and later: Unicode, Inc. and others. |
2 | // License & terms of use: http://www.unicode.org/copyright.html | |
3 | // | |
4 | // From the double-conversion library. Original license: | |
5 | // | |
6 | // Copyright 2010 the V8 project authors. All rights reserved. | |
7 | // Redistribution and use in source and binary forms, with or without | |
8 | // modification, are permitted provided that the following conditions are | |
9 | // met: | |
10 | // | |
11 | // * Redistributions of source code must retain the above copyright | |
12 | // notice, this list of conditions and the following disclaimer. | |
13 | // * Redistributions in binary form must reproduce the above | |
14 | // copyright notice, this list of conditions and the following | |
15 | // disclaimer in the documentation and/or other materials provided | |
16 | // with the distribution. | |
17 | // * Neither the name of Google Inc. nor the names of its | |
18 | // contributors may be used to endorse or promote products derived | |
19 | // from this software without specific prior written permission. | |
20 | // | |
21 | // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS | |
22 | // "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT | |
23 | // LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR | |
24 | // A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT | |
25 | // OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, | |
26 | // SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT | |
27 | // LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, | |
28 | // DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY | |
29 | // THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT | |
30 | // (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE | |
31 | // OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. | |
32 | ||
33 | // ICU PATCH: ifdef around UCONFIG_NO_FORMATTING | |
34 | #include "unicode/utypes.h" | |
35 | #if !UCONFIG_NO_FORMATTING | |
36 | ||
37 | #include <limits.h> | |
38 | #include <math.h> | |
39 | ||
40 | // ICU PATCH: Customize header file paths for ICU. | |
41 | // The file fixed-dtoa.h is not needed. | |
42 | ||
43 | #include "double-conversion.h" | |
44 | ||
45 | #include "double-conversion-bignum-dtoa.h" | |
46 | #include "double-conversion-fast-dtoa.h" | |
47 | #include "double-conversion-ieee.h" | |
48 | #include "double-conversion-strtod.h" | |
49 | #include "double-conversion-utils.h" | |
50 | ||
51 | // ICU PATCH: Wrap in ICU namespace | |
52 | U_NAMESPACE_BEGIN | |
53 | ||
54 | namespace double_conversion { | |
55 | ||
56 | #if 0 // not needed for ICU | |
57 | const DoubleToStringConverter& DoubleToStringConverter::EcmaScriptConverter() { | |
58 | int flags = UNIQUE_ZERO | EMIT_POSITIVE_EXPONENT_SIGN; | |
59 | static DoubleToStringConverter converter(flags, | |
60 | "Infinity", | |
61 | "NaN", | |
62 | 'e', | |
63 | -6, 21, | |
64 | 6, 0); | |
65 | return converter; | |
66 | } | |
67 | ||
68 | ||
69 | bool DoubleToStringConverter::HandleSpecialValues( | |
70 | double value, | |
71 | StringBuilder* result_builder) const { | |
72 | Double double_inspect(value); | |
73 | if (double_inspect.IsInfinite()) { | |
74 | if (infinity_symbol_ == NULL) return false; | |
75 | if (value < 0) { | |
76 | result_builder->AddCharacter('-'); | |
77 | } | |
78 | result_builder->AddString(infinity_symbol_); | |
79 | return true; | |
80 | } | |
81 | if (double_inspect.IsNan()) { | |
82 | if (nan_symbol_ == NULL) return false; | |
83 | result_builder->AddString(nan_symbol_); | |
84 | return true; | |
85 | } | |
86 | return false; | |
87 | } | |
88 | ||
89 | ||
90 | void DoubleToStringConverter::CreateExponentialRepresentation( | |
91 | const char* decimal_digits, | |
92 | int length, | |
93 | int exponent, | |
94 | StringBuilder* result_builder) const { | |
95 | ASSERT(length != 0); | |
96 | result_builder->AddCharacter(decimal_digits[0]); | |
97 | if (length != 1) { | |
98 | result_builder->AddCharacter('.'); | |
99 | result_builder->AddSubstring(&decimal_digits[1], length-1); | |
100 | } | |
101 | result_builder->AddCharacter(exponent_character_); | |
102 | if (exponent < 0) { | |
103 | result_builder->AddCharacter('-'); | |
104 | exponent = -exponent; | |
105 | } else { | |
106 | if ((flags_ & EMIT_POSITIVE_EXPONENT_SIGN) != 0) { | |
107 | result_builder->AddCharacter('+'); | |
108 | } | |
109 | } | |
110 | if (exponent == 0) { | |
111 | result_builder->AddCharacter('0'); | |
112 | return; | |
113 | } | |
114 | ASSERT(exponent < 1e4); | |
115 | const int kMaxExponentLength = 5; | |
116 | char buffer[kMaxExponentLength + 1]; | |
117 | buffer[kMaxExponentLength] = '\0'; | |
118 | int first_char_pos = kMaxExponentLength; | |
119 | while (exponent > 0) { | |
120 | buffer[--first_char_pos] = '0' + (exponent % 10); | |
121 | exponent /= 10; | |
122 | } | |
123 | result_builder->AddSubstring(&buffer[first_char_pos], | |
124 | kMaxExponentLength - first_char_pos); | |
125 | } | |
126 | ||
127 | ||
128 | void DoubleToStringConverter::CreateDecimalRepresentation( | |
129 | const char* decimal_digits, | |
130 | int length, | |
131 | int decimal_point, | |
132 | int digits_after_point, | |
133 | StringBuilder* result_builder) const { | |
134 | // Create a representation that is padded with zeros if needed. | |
135 | if (decimal_point <= 0) { | |
136 | // "0.00000decimal_rep" or "0.000decimal_rep00". | |
137 | result_builder->AddCharacter('0'); | |
138 | if (digits_after_point > 0) { | |
139 | result_builder->AddCharacter('.'); | |
140 | result_builder->AddPadding('0', -decimal_point); | |
141 | ASSERT(length <= digits_after_point - (-decimal_point)); | |
142 | result_builder->AddSubstring(decimal_digits, length); | |
143 | int remaining_digits = digits_after_point - (-decimal_point) - length; | |
144 | result_builder->AddPadding('0', remaining_digits); | |
145 | } | |
146 | } else if (decimal_point >= length) { | |
147 | // "decimal_rep0000.00000" or "decimal_rep.0000". | |
148 | result_builder->AddSubstring(decimal_digits, length); | |
149 | result_builder->AddPadding('0', decimal_point - length); | |
150 | if (digits_after_point > 0) { | |
151 | result_builder->AddCharacter('.'); | |
152 | result_builder->AddPadding('0', digits_after_point); | |
153 | } | |
154 | } else { | |
155 | // "decima.l_rep000". | |
156 | ASSERT(digits_after_point > 0); | |
157 | result_builder->AddSubstring(decimal_digits, decimal_point); | |
158 | result_builder->AddCharacter('.'); | |
159 | ASSERT(length - decimal_point <= digits_after_point); | |
160 | result_builder->AddSubstring(&decimal_digits[decimal_point], | |
161 | length - decimal_point); | |
162 | int remaining_digits = digits_after_point - (length - decimal_point); | |
163 | result_builder->AddPadding('0', remaining_digits); | |
164 | } | |
165 | if (digits_after_point == 0) { | |
166 | if ((flags_ & EMIT_TRAILING_DECIMAL_POINT) != 0) { | |
167 | result_builder->AddCharacter('.'); | |
168 | } | |
169 | if ((flags_ & EMIT_TRAILING_ZERO_AFTER_POINT) != 0) { | |
170 | result_builder->AddCharacter('0'); | |
171 | } | |
172 | } | |
173 | } | |
174 | ||
175 | ||
176 | bool DoubleToStringConverter::ToShortestIeeeNumber( | |
177 | double value, | |
178 | StringBuilder* result_builder, | |
179 | DoubleToStringConverter::DtoaMode mode) const { | |
180 | ASSERT(mode == SHORTEST || mode == SHORTEST_SINGLE); | |
181 | if (Double(value).IsSpecial()) { | |
182 | return HandleSpecialValues(value, result_builder); | |
183 | } | |
184 | ||
185 | int decimal_point; | |
186 | bool sign; | |
187 | const int kDecimalRepCapacity = kBase10MaximalLength + 1; | |
188 | char decimal_rep[kDecimalRepCapacity]; | |
189 | int decimal_rep_length; | |
190 | ||
191 | DoubleToAscii(value, mode, 0, decimal_rep, kDecimalRepCapacity, | |
192 | &sign, &decimal_rep_length, &decimal_point); | |
193 | ||
194 | bool unique_zero = (flags_ & UNIQUE_ZERO) != 0; | |
195 | if (sign && (value != 0.0 || !unique_zero)) { | |
196 | result_builder->AddCharacter('-'); | |
197 | } | |
198 | ||
199 | int exponent = decimal_point - 1; | |
200 | if ((decimal_in_shortest_low_ <= exponent) && | |
201 | (exponent < decimal_in_shortest_high_)) { | |
202 | CreateDecimalRepresentation(decimal_rep, decimal_rep_length, | |
203 | decimal_point, | |
204 | Max(0, decimal_rep_length - decimal_point), | |
205 | result_builder); | |
206 | } else { | |
207 | CreateExponentialRepresentation(decimal_rep, decimal_rep_length, exponent, | |
208 | result_builder); | |
209 | } | |
210 | return true; | |
211 | } | |
212 | ||
213 | ||
214 | bool DoubleToStringConverter::ToFixed(double value, | |
215 | int requested_digits, | |
216 | StringBuilder* result_builder) const { | |
217 | ASSERT(kMaxFixedDigitsBeforePoint == 60); | |
218 | const double kFirstNonFixed = 1e60; | |
219 | ||
220 | if (Double(value).IsSpecial()) { | |
221 | return HandleSpecialValues(value, result_builder); | |
222 | } | |
223 | ||
224 | if (requested_digits > kMaxFixedDigitsAfterPoint) return false; | |
225 | if (value >= kFirstNonFixed || value <= -kFirstNonFixed) return false; | |
226 | ||
227 | // Find a sufficiently precise decimal representation of n. | |
228 | int decimal_point; | |
229 | bool sign; | |
230 | // Add space for the '\0' byte. | |
231 | const int kDecimalRepCapacity = | |
232 | kMaxFixedDigitsBeforePoint + kMaxFixedDigitsAfterPoint + 1; | |
233 | char decimal_rep[kDecimalRepCapacity]; | |
234 | int decimal_rep_length; | |
235 | DoubleToAscii(value, FIXED, requested_digits, | |
236 | decimal_rep, kDecimalRepCapacity, | |
237 | &sign, &decimal_rep_length, &decimal_point); | |
238 | ||
239 | bool unique_zero = ((flags_ & UNIQUE_ZERO) != 0); | |
240 | if (sign && (value != 0.0 || !unique_zero)) { | |
241 | result_builder->AddCharacter('-'); | |
242 | } | |
243 | ||
244 | CreateDecimalRepresentation(decimal_rep, decimal_rep_length, decimal_point, | |
245 | requested_digits, result_builder); | |
246 | return true; | |
247 | } | |
248 | ||
249 | ||
250 | bool DoubleToStringConverter::ToExponential( | |
251 | double value, | |
252 | int requested_digits, | |
253 | StringBuilder* result_builder) const { | |
254 | if (Double(value).IsSpecial()) { | |
255 | return HandleSpecialValues(value, result_builder); | |
256 | } | |
257 | ||
258 | if (requested_digits < -1) return false; | |
259 | if (requested_digits > kMaxExponentialDigits) return false; | |
260 | ||
261 | int decimal_point; | |
262 | bool sign; | |
263 | // Add space for digit before the decimal point and the '\0' character. | |
264 | const int kDecimalRepCapacity = kMaxExponentialDigits + 2; | |
265 | ASSERT(kDecimalRepCapacity > kBase10MaximalLength); | |
266 | char decimal_rep[kDecimalRepCapacity]; | |
267 | int decimal_rep_length; | |
268 | ||
269 | if (requested_digits == -1) { | |
270 | DoubleToAscii(value, SHORTEST, 0, | |
271 | decimal_rep, kDecimalRepCapacity, | |
272 | &sign, &decimal_rep_length, &decimal_point); | |
273 | } else { | |
274 | DoubleToAscii(value, PRECISION, requested_digits + 1, | |
275 | decimal_rep, kDecimalRepCapacity, | |
276 | &sign, &decimal_rep_length, &decimal_point); | |
277 | ASSERT(decimal_rep_length <= requested_digits + 1); | |
278 | ||
279 | for (int i = decimal_rep_length; i < requested_digits + 1; ++i) { | |
280 | decimal_rep[i] = '0'; | |
281 | } | |
282 | decimal_rep_length = requested_digits + 1; | |
283 | } | |
284 | ||
285 | bool unique_zero = ((flags_ & UNIQUE_ZERO) != 0); | |
286 | if (sign && (value != 0.0 || !unique_zero)) { | |
287 | result_builder->AddCharacter('-'); | |
288 | } | |
289 | ||
290 | int exponent = decimal_point - 1; | |
291 | CreateExponentialRepresentation(decimal_rep, | |
292 | decimal_rep_length, | |
293 | exponent, | |
294 | result_builder); | |
295 | return true; | |
296 | } | |
297 | ||
298 | ||
299 | bool DoubleToStringConverter::ToPrecision(double value, | |
300 | int precision, | |
301 | StringBuilder* result_builder) const { | |
302 | if (Double(value).IsSpecial()) { | |
303 | return HandleSpecialValues(value, result_builder); | |
304 | } | |
305 | ||
306 | if (precision < kMinPrecisionDigits || precision > kMaxPrecisionDigits) { | |
307 | return false; | |
308 | } | |
309 | ||
310 | // Find a sufficiently precise decimal representation of n. | |
311 | int decimal_point; | |
312 | bool sign; | |
313 | // Add one for the terminating null character. | |
314 | const int kDecimalRepCapacity = kMaxPrecisionDigits + 1; | |
315 | char decimal_rep[kDecimalRepCapacity]; | |
316 | int decimal_rep_length; | |
317 | ||
318 | DoubleToAscii(value, PRECISION, precision, | |
319 | decimal_rep, kDecimalRepCapacity, | |
320 | &sign, &decimal_rep_length, &decimal_point); | |
321 | ASSERT(decimal_rep_length <= precision); | |
322 | ||
323 | bool unique_zero = ((flags_ & UNIQUE_ZERO) != 0); | |
324 | if (sign && (value != 0.0 || !unique_zero)) { | |
325 | result_builder->AddCharacter('-'); | |
326 | } | |
327 | ||
328 | // The exponent if we print the number as x.xxeyyy. That is with the | |
329 | // decimal point after the first digit. | |
330 | int exponent = decimal_point - 1; | |
331 | ||
332 | int extra_zero = ((flags_ & EMIT_TRAILING_ZERO_AFTER_POINT) != 0) ? 1 : 0; | |
333 | if ((-decimal_point + 1 > max_leading_padding_zeroes_in_precision_mode_) || | |
334 | (decimal_point - precision + extra_zero > | |
335 | max_trailing_padding_zeroes_in_precision_mode_)) { | |
336 | // Fill buffer to contain 'precision' digits. | |
337 | // Usually the buffer is already at the correct length, but 'DoubleToAscii' | |
338 | // is allowed to return less characters. | |
339 | for (int i = decimal_rep_length; i < precision; ++i) { | |
340 | decimal_rep[i] = '0'; | |
341 | } | |
342 | ||
343 | CreateExponentialRepresentation(decimal_rep, | |
344 | precision, | |
345 | exponent, | |
346 | result_builder); | |
347 | } else { | |
348 | CreateDecimalRepresentation(decimal_rep, decimal_rep_length, decimal_point, | |
349 | Max(0, precision - decimal_point), | |
350 | result_builder); | |
351 | } | |
352 | return true; | |
353 | } | |
354 | #endif // not needed for ICU | |
355 | ||
356 | ||
357 | static BignumDtoaMode DtoaToBignumDtoaMode( | |
358 | DoubleToStringConverter::DtoaMode dtoa_mode) { | |
359 | switch (dtoa_mode) { | |
360 | case DoubleToStringConverter::SHORTEST: return BIGNUM_DTOA_SHORTEST; | |
361 | case DoubleToStringConverter::SHORTEST_SINGLE: | |
362 | return BIGNUM_DTOA_SHORTEST_SINGLE; | |
363 | case DoubleToStringConverter::FIXED: return BIGNUM_DTOA_FIXED; | |
364 | case DoubleToStringConverter::PRECISION: return BIGNUM_DTOA_PRECISION; | |
365 | default: | |
366 | UNREACHABLE(); | |
367 | } | |
368 | } | |
369 | ||
370 | ||
371 | void DoubleToStringConverter::DoubleToAscii(double v, | |
372 | DtoaMode mode, | |
373 | int requested_digits, | |
374 | char* buffer, | |
375 | int buffer_length, | |
376 | bool* sign, | |
377 | int* length, | |
378 | int* point) { | |
379 | Vector<char> vector(buffer, buffer_length); | |
380 | ASSERT(!Double(v).IsSpecial()); | |
381 | ASSERT(mode == SHORTEST || mode == SHORTEST_SINGLE || requested_digits >= 0); | |
382 | ||
383 | if (Double(v).Sign() < 0) { | |
384 | *sign = true; | |
385 | v = -v; | |
386 | } else { | |
387 | *sign = false; | |
388 | } | |
389 | ||
390 | if (mode == PRECISION && requested_digits == 0) { | |
391 | vector[0] = '\0'; | |
392 | *length = 0; | |
393 | return; | |
394 | } | |
395 | ||
396 | if (v == 0) { | |
397 | vector[0] = '0'; | |
398 | vector[1] = '\0'; | |
399 | *length = 1; | |
400 | *point = 1; | |
401 | return; | |
402 | } | |
403 | ||
404 | bool fast_worked; | |
405 | switch (mode) { | |
406 | case SHORTEST: | |
407 | fast_worked = FastDtoa(v, FAST_DTOA_SHORTEST, 0, vector, length, point); | |
408 | break; | |
409 | #if 0 // not needed for ICU | |
410 | case SHORTEST_SINGLE: | |
411 | fast_worked = FastDtoa(v, FAST_DTOA_SHORTEST_SINGLE, 0, | |
412 | vector, length, point); | |
413 | break; | |
414 | case FIXED: | |
415 | fast_worked = FastFixedDtoa(v, requested_digits, vector, length, point); | |
416 | break; | |
417 | case PRECISION: | |
418 | fast_worked = FastDtoa(v, FAST_DTOA_PRECISION, requested_digits, | |
419 | vector, length, point); | |
420 | break; | |
421 | #endif // not needed for ICU | |
422 | default: | |
423 | fast_worked = false; | |
424 | UNREACHABLE(); | |
425 | } | |
426 | if (fast_worked) return; | |
427 | ||
428 | // If the fast dtoa didn't succeed use the slower bignum version. | |
429 | BignumDtoaMode bignum_mode = DtoaToBignumDtoaMode(mode); | |
430 | BignumDtoa(v, bignum_mode, requested_digits, vector, length, point); | |
431 | vector[*length] = '\0'; | |
432 | } | |
433 | ||
434 | ||
435 | // Consumes the given substring from the iterator. | |
436 | // Returns false, if the substring does not match. | |
437 | template <class Iterator> | |
438 | static bool ConsumeSubString(Iterator* current, | |
439 | Iterator end, | |
440 | const char* substring) { | |
441 | ASSERT(**current == *substring); | |
442 | for (substring++; *substring != '\0'; substring++) { | |
443 | ++*current; | |
444 | if (*current == end || **current != *substring) return false; | |
445 | } | |
446 | ++*current; | |
447 | return true; | |
448 | } | |
449 | ||
450 | ||
451 | // Maximum number of significant digits in decimal representation. | |
452 | // The longest possible double in decimal representation is | |
453 | // (2^53 - 1) * 2 ^ -1074 that is (2 ^ 53 - 1) * 5 ^ 1074 / 10 ^ 1074 | |
454 | // (768 digits). If we parse a number whose first digits are equal to a | |
455 | // mean of 2 adjacent doubles (that could have up to 769 digits) the result | |
456 | // must be rounded to the bigger one unless the tail consists of zeros, so | |
457 | // we don't need to preserve all the digits. | |
458 | const int kMaxSignificantDigits = 772; | |
459 | ||
460 | ||
461 | static const char kWhitespaceTable7[] = { 32, 13, 10, 9, 11, 12 }; | |
462 | static const int kWhitespaceTable7Length = ARRAY_SIZE(kWhitespaceTable7); | |
463 | ||
464 | ||
465 | static const uc16 kWhitespaceTable16[] = { | |
466 | 160, 8232, 8233, 5760, 6158, 8192, 8193, 8194, 8195, | |
467 | 8196, 8197, 8198, 8199, 8200, 8201, 8202, 8239, 8287, 12288, 65279 | |
468 | }; | |
469 | static const int kWhitespaceTable16Length = ARRAY_SIZE(kWhitespaceTable16); | |
470 | ||
471 | ||
472 | ||
473 | static bool isWhitespace(int x) { | |
474 | if (x < 128) { | |
475 | for (int i = 0; i < kWhitespaceTable7Length; i++) { | |
476 | if (kWhitespaceTable7[i] == x) return true; | |
477 | } | |
478 | } else { | |
479 | for (int i = 0; i < kWhitespaceTable16Length; i++) { | |
480 | if (kWhitespaceTable16[i] == x) return true; | |
481 | } | |
482 | } | |
483 | return false; | |
484 | } | |
485 | ||
486 | ||
487 | // Returns true if a nonspace found and false if the end has reached. | |
488 | template <class Iterator> | |
489 | static inline bool AdvanceToNonspace(Iterator* current, Iterator end) { | |
490 | while (*current != end) { | |
491 | if (!isWhitespace(**current)) return true; | |
492 | ++*current; | |
493 | } | |
494 | return false; | |
495 | } | |
496 | ||
497 | ||
498 | static bool isDigit(int x, int radix) { | |
499 | return (x >= '0' && x <= '9' && x < '0' + radix) | |
500 | || (radix > 10 && x >= 'a' && x < 'a' + radix - 10) | |
501 | || (radix > 10 && x >= 'A' && x < 'A' + radix - 10); | |
502 | } | |
503 | ||
504 | ||
505 | static double SignedZero(bool sign) { | |
506 | return sign ? -0.0 : 0.0; | |
507 | } | |
508 | ||
509 | ||
510 | // Returns true if 'c' is a decimal digit that is valid for the given radix. | |
511 | // | |
512 | // The function is small and could be inlined, but VS2012 emitted a warning | |
513 | // because it constant-propagated the radix and concluded that the last | |
514 | // condition was always true. By moving it into a separate function the | |
515 | // compiler wouldn't warn anymore. | |
516 | #if _MSC_VER | |
517 | #pragma optimize("",off) | |
518 | static bool IsDecimalDigitForRadix(int c, int radix) { | |
519 | return '0' <= c && c <= '9' && (c - '0') < radix; | |
520 | } | |
521 | #pragma optimize("",on) | |
522 | #else | |
523 | static bool inline IsDecimalDigitForRadix(int c, int radix) { | |
524 | return '0' <= c && c <= '9' && (c - '0') < radix; | |
525 | } | |
526 | #endif | |
527 | // Returns true if 'c' is a character digit that is valid for the given radix. | |
528 | // The 'a_character' should be 'a' or 'A'. | |
529 | // | |
530 | // The function is small and could be inlined, but VS2012 emitted a warning | |
531 | // because it constant-propagated the radix and concluded that the first | |
532 | // condition was always false. By moving it into a separate function the | |
533 | // compiler wouldn't warn anymore. | |
534 | static bool IsCharacterDigitForRadix(int c, int radix, char a_character) { | |
535 | return radix > 10 && c >= a_character && c < a_character + radix - 10; | |
536 | } | |
537 | ||
538 | ||
539 | // Parsing integers with radix 2, 4, 8, 16, 32. Assumes current != end. | |
540 | template <int radix_log_2, class Iterator> | |
541 | static double RadixStringToIeee(Iterator* current, | |
542 | Iterator end, | |
543 | bool sign, | |
544 | bool allow_trailing_junk, | |
545 | double junk_string_value, | |
546 | bool read_as_double, | |
547 | bool* result_is_junk) { | |
548 | ASSERT(*current != end); | |
549 | ||
550 | const int kDoubleSize = Double::kSignificandSize; | |
551 | const int kSingleSize = Single::kSignificandSize; | |
552 | const int kSignificandSize = read_as_double? kDoubleSize: kSingleSize; | |
553 | ||
554 | *result_is_junk = true; | |
555 | ||
556 | // Skip leading 0s. | |
557 | while (**current == '0') { | |
558 | ++(*current); | |
559 | if (*current == end) { | |
560 | *result_is_junk = false; | |
561 | return SignedZero(sign); | |
562 | } | |
563 | } | |
564 | ||
565 | int64_t number = 0; | |
566 | int exponent = 0; | |
567 | const int radix = (1 << radix_log_2); | |
568 | ||
569 | do { | |
570 | int digit; | |
571 | if (IsDecimalDigitForRadix(**current, radix)) { | |
572 | digit = static_cast<char>(**current) - '0'; | |
573 | } else if (IsCharacterDigitForRadix(**current, radix, 'a')) { | |
574 | digit = static_cast<char>(**current) - 'a' + 10; | |
575 | } else if (IsCharacterDigitForRadix(**current, radix, 'A')) { | |
576 | digit = static_cast<char>(**current) - 'A' + 10; | |
577 | } else { | |
578 | if (allow_trailing_junk || !AdvanceToNonspace(current, end)) { | |
579 | break; | |
580 | } else { | |
581 | return junk_string_value; | |
582 | } | |
583 | } | |
584 | ||
585 | number = number * radix + digit; | |
586 | int overflow = static_cast<int>(number >> kSignificandSize); | |
587 | if (overflow != 0) { | |
588 | // Overflow occurred. Need to determine which direction to round the | |
589 | // result. | |
590 | int overflow_bits_count = 1; | |
591 | while (overflow > 1) { | |
592 | overflow_bits_count++; | |
593 | overflow >>= 1; | |
594 | } | |
595 | ||
596 | int dropped_bits_mask = ((1 << overflow_bits_count) - 1); | |
597 | int dropped_bits = static_cast<int>(number) & dropped_bits_mask; | |
598 | number >>= overflow_bits_count; | |
599 | exponent = overflow_bits_count; | |
600 | ||
601 | bool zero_tail = true; | |
602 | for (;;) { | |
603 | ++(*current); | |
604 | if (*current == end || !isDigit(**current, radix)) break; | |
605 | zero_tail = zero_tail && **current == '0'; | |
606 | exponent += radix_log_2; | |
607 | } | |
608 | ||
609 | if (!allow_trailing_junk && AdvanceToNonspace(current, end)) { | |
610 | return junk_string_value; | |
611 | } | |
612 | ||
613 | int middle_value = (1 << (overflow_bits_count - 1)); | |
614 | if (dropped_bits > middle_value) { | |
615 | number++; // Rounding up. | |
616 | } else if (dropped_bits == middle_value) { | |
617 | // Rounding to even to consistency with decimals: half-way case rounds | |
618 | // up if significant part is odd and down otherwise. | |
619 | if ((number & 1) != 0 || !zero_tail) { | |
620 | number++; // Rounding up. | |
621 | } | |
622 | } | |
623 | ||
624 | // Rounding up may cause overflow. | |
625 | if ((number & ((int64_t)1 << kSignificandSize)) != 0) { | |
626 | exponent++; | |
627 | number >>= 1; | |
628 | } | |
629 | break; | |
630 | } | |
631 | ++(*current); | |
632 | } while (*current != end); | |
633 | ||
634 | ASSERT(number < ((int64_t)1 << kSignificandSize)); | |
635 | ASSERT(static_cast<int64_t>(static_cast<double>(number)) == number); | |
636 | ||
637 | *result_is_junk = false; | |
638 | ||
639 | if (exponent == 0) { | |
640 | if (sign) { | |
641 | if (number == 0) return -0.0; | |
642 | number = -number; | |
643 | } | |
644 | return static_cast<double>(number); | |
645 | } | |
646 | ||
647 | ASSERT(number != 0); | |
648 | return Double(DiyFp(number, exponent)).value(); | |
649 | } | |
650 | ||
651 | template <class Iterator> | |
652 | double StringToDoubleConverter::StringToIeee( | |
653 | Iterator input, | |
654 | int length, | |
655 | bool read_as_double, | |
656 | int* processed_characters_count) const { | |
657 | Iterator current = input; | |
658 | Iterator end = input + length; | |
659 | ||
660 | *processed_characters_count = 0; | |
661 | ||
662 | const bool allow_trailing_junk = (flags_ & ALLOW_TRAILING_JUNK) != 0; | |
663 | const bool allow_leading_spaces = (flags_ & ALLOW_LEADING_SPACES) != 0; | |
664 | const bool allow_trailing_spaces = (flags_ & ALLOW_TRAILING_SPACES) != 0; | |
665 | const bool allow_spaces_after_sign = (flags_ & ALLOW_SPACES_AFTER_SIGN) != 0; | |
666 | ||
667 | // To make sure that iterator dereferencing is valid the following | |
668 | // convention is used: | |
669 | // 1. Each '++current' statement is followed by check for equality to 'end'. | |
670 | // 2. If AdvanceToNonspace returned false then current == end. | |
671 | // 3. If 'current' becomes equal to 'end' the function returns or goes to | |
672 | // 'parsing_done'. | |
673 | // 4. 'current' is not dereferenced after the 'parsing_done' label. | |
674 | // 5. Code before 'parsing_done' may rely on 'current != end'. | |
675 | if (current == end) return empty_string_value_; | |
676 | ||
677 | if (allow_leading_spaces || allow_trailing_spaces) { | |
678 | if (!AdvanceToNonspace(¤t, end)) { | |
679 | *processed_characters_count = static_cast<int>(current - input); | |
680 | return empty_string_value_; | |
681 | } | |
682 | if (!allow_leading_spaces && (input != current)) { | |
683 | // No leading spaces allowed, but AdvanceToNonspace moved forward. | |
684 | return junk_string_value_; | |
685 | } | |
686 | } | |
687 | ||
688 | // The longest form of simplified number is: "-<significant digits>.1eXXX\0". | |
689 | const int kBufferSize = kMaxSignificantDigits + 10; | |
690 | char buffer[kBufferSize]; // NOLINT: size is known at compile time. | |
691 | int buffer_pos = 0; | |
692 | ||
693 | // Exponent will be adjusted if insignificant digits of the integer part | |
694 | // or insignificant leading zeros of the fractional part are dropped. | |
695 | int exponent = 0; | |
696 | int significant_digits = 0; | |
697 | int insignificant_digits = 0; | |
698 | bool nonzero_digit_dropped = false; | |
699 | ||
700 | bool sign = false; | |
701 | ||
702 | if (*current == '+' || *current == '-') { | |
703 | sign = (*current == '-'); | |
704 | ++current; | |
705 | Iterator next_non_space = current; | |
706 | // Skip following spaces (if allowed). | |
707 | if (!AdvanceToNonspace(&next_non_space, end)) return junk_string_value_; | |
708 | if (!allow_spaces_after_sign && (current != next_non_space)) { | |
709 | return junk_string_value_; | |
710 | } | |
711 | current = next_non_space; | |
712 | } | |
713 | ||
714 | if (infinity_symbol_ != NULL) { | |
715 | if (*current == infinity_symbol_[0]) { | |
716 | if (!ConsumeSubString(¤t, end, infinity_symbol_)) { | |
717 | return junk_string_value_; | |
718 | } | |
719 | ||
720 | if (!(allow_trailing_spaces || allow_trailing_junk) && (current != end)) { | |
721 | return junk_string_value_; | |
722 | } | |
723 | if (!allow_trailing_junk && AdvanceToNonspace(¤t, end)) { | |
724 | return junk_string_value_; | |
725 | } | |
726 | ||
727 | ASSERT(buffer_pos == 0); | |
728 | *processed_characters_count = static_cast<int>(current - input); | |
729 | return sign ? -Double::Infinity() : Double::Infinity(); | |
730 | } | |
731 | } | |
732 | ||
733 | if (nan_symbol_ != NULL) { | |
734 | if (*current == nan_symbol_[0]) { | |
735 | if (!ConsumeSubString(¤t, end, nan_symbol_)) { | |
736 | return junk_string_value_; | |
737 | } | |
738 | ||
739 | if (!(allow_trailing_spaces || allow_trailing_junk) && (current != end)) { | |
740 | return junk_string_value_; | |
741 | } | |
742 | if (!allow_trailing_junk && AdvanceToNonspace(¤t, end)) { | |
743 | return junk_string_value_; | |
744 | } | |
745 | ||
746 | ASSERT(buffer_pos == 0); | |
747 | *processed_characters_count = static_cast<int>(current - input); | |
748 | return sign ? -Double::NaN() : Double::NaN(); | |
749 | } | |
750 | } | |
751 | ||
752 | bool leading_zero = false; | |
753 | if (*current == '0') { | |
754 | ++current; | |
755 | if (current == end) { | |
756 | *processed_characters_count = static_cast<int>(current - input); | |
757 | return SignedZero(sign); | |
758 | } | |
759 | ||
760 | leading_zero = true; | |
761 | ||
762 | // It could be hexadecimal value. | |
763 | if ((flags_ & ALLOW_HEX) && (*current == 'x' || *current == 'X')) { | |
764 | ++current; | |
765 | if (current == end || !isDigit(*current, 16)) { | |
766 | return junk_string_value_; // "0x". | |
767 | } | |
768 | ||
769 | bool result_is_junk; | |
770 | double result = RadixStringToIeee<4>(¤t, | |
771 | end, | |
772 | sign, | |
773 | allow_trailing_junk, | |
774 | junk_string_value_, | |
775 | read_as_double, | |
776 | &result_is_junk); | |
777 | if (!result_is_junk) { | |
778 | if (allow_trailing_spaces) AdvanceToNonspace(¤t, end); | |
779 | *processed_characters_count = static_cast<int>(current - input); | |
780 | } | |
781 | return result; | |
782 | } | |
783 | ||
784 | // Ignore leading zeros in the integer part. | |
785 | while (*current == '0') { | |
786 | ++current; | |
787 | if (current == end) { | |
788 | *processed_characters_count = static_cast<int>(current - input); | |
789 | return SignedZero(sign); | |
790 | } | |
791 | } | |
792 | } | |
793 | ||
794 | bool octal = leading_zero && (flags_ & ALLOW_OCTALS) != 0; | |
795 | ||
796 | // Copy significant digits of the integer part (if any) to the buffer. | |
797 | while (*current >= '0' && *current <= '9') { | |
798 | if (significant_digits < kMaxSignificantDigits) { | |
799 | ASSERT(buffer_pos < kBufferSize); | |
800 | buffer[buffer_pos++] = static_cast<char>(*current); | |
801 | significant_digits++; | |
802 | // Will later check if it's an octal in the buffer. | |
803 | } else { | |
804 | insignificant_digits++; // Move the digit into the exponential part. | |
805 | nonzero_digit_dropped = nonzero_digit_dropped || *current != '0'; | |
806 | } | |
807 | octal = octal && *current < '8'; | |
808 | ++current; | |
809 | if (current == end) goto parsing_done; | |
810 | } | |
811 | ||
812 | if (significant_digits == 0) { | |
813 | octal = false; | |
814 | } | |
815 | ||
816 | if (*current == '.') { | |
817 | if (octal && !allow_trailing_junk) return junk_string_value_; | |
818 | if (octal) goto parsing_done; | |
819 | ||
820 | ++current; | |
821 | if (current == end) { | |
822 | if (significant_digits == 0 && !leading_zero) { | |
823 | return junk_string_value_; | |
824 | } else { | |
825 | goto parsing_done; | |
826 | } | |
827 | } | |
828 | ||
829 | if (significant_digits == 0) { | |
830 | // octal = false; | |
831 | // Integer part consists of 0 or is absent. Significant digits start after | |
832 | // leading zeros (if any). | |
833 | while (*current == '0') { | |
834 | ++current; | |
835 | if (current == end) { | |
836 | *processed_characters_count = static_cast<int>(current - input); | |
837 | return SignedZero(sign); | |
838 | } | |
839 | exponent--; // Move this 0 into the exponent. | |
840 | } | |
841 | } | |
842 | ||
843 | // There is a fractional part. | |
844 | // We don't emit a '.', but adjust the exponent instead. | |
845 | while (*current >= '0' && *current <= '9') { | |
846 | if (significant_digits < kMaxSignificantDigits) { | |
847 | ASSERT(buffer_pos < kBufferSize); | |
848 | buffer[buffer_pos++] = static_cast<char>(*current); | |
849 | significant_digits++; | |
850 | exponent--; | |
851 | } else { | |
852 | // Ignore insignificant digits in the fractional part. | |
853 | nonzero_digit_dropped = nonzero_digit_dropped || *current != '0'; | |
854 | } | |
855 | ++current; | |
856 | if (current == end) goto parsing_done; | |
857 | } | |
858 | } | |
859 | ||
860 | if (!leading_zero && exponent == 0 && significant_digits == 0) { | |
861 | // If leading_zeros is true then the string contains zeros. | |
862 | // If exponent < 0 then string was [+-]\.0*... | |
863 | // If significant_digits != 0 the string is not equal to 0. | |
864 | // Otherwise there are no digits in the string. | |
865 | return junk_string_value_; | |
866 | } | |
867 | ||
868 | // Parse exponential part. | |
869 | if (*current == 'e' || *current == 'E') { | |
870 | if (octal && !allow_trailing_junk) return junk_string_value_; | |
871 | if (octal) goto parsing_done; | |
872 | ++current; | |
873 | if (current == end) { | |
874 | if (allow_trailing_junk) { | |
875 | goto parsing_done; | |
876 | } else { | |
877 | return junk_string_value_; | |
878 | } | |
879 | } | |
880 | char exponen_sign = '+'; | |
881 | if (*current == '+' || *current == '-') { | |
882 | exponen_sign = static_cast<char>(*current); | |
883 | ++current; | |
884 | if (current == end) { | |
885 | if (allow_trailing_junk) { | |
886 | goto parsing_done; | |
887 | } else { | |
888 | return junk_string_value_; | |
889 | } | |
890 | } | |
891 | } | |
892 | ||
893 | if (current == end || *current < '0' || *current > '9') { | |
894 | if (allow_trailing_junk) { | |
895 | goto parsing_done; | |
896 | } else { | |
897 | return junk_string_value_; | |
898 | } | |
899 | } | |
900 | ||
901 | const int max_exponent = INT_MAX / 2; | |
902 | ASSERT(-max_exponent / 2 <= exponent && exponent <= max_exponent / 2); | |
903 | int num = 0; | |
904 | do { | |
905 | // Check overflow. | |
906 | int digit = *current - '0'; | |
907 | if (num >= max_exponent / 10 | |
908 | && !(num == max_exponent / 10 && digit <= max_exponent % 10)) { | |
909 | num = max_exponent; | |
910 | } else { | |
911 | num = num * 10 + digit; | |
912 | } | |
913 | ++current; | |
914 | } while (current != end && *current >= '0' && *current <= '9'); | |
915 | ||
916 | exponent += (exponen_sign == '-' ? -num : num); | |
917 | } | |
918 | ||
919 | if (!(allow_trailing_spaces || allow_trailing_junk) && (current != end)) { | |
920 | return junk_string_value_; | |
921 | } | |
922 | if (!allow_trailing_junk && AdvanceToNonspace(¤t, end)) { | |
923 | return junk_string_value_; | |
924 | } | |
925 | if (allow_trailing_spaces) { | |
926 | AdvanceToNonspace(¤t, end); | |
927 | } | |
928 | ||
929 | parsing_done: | |
930 | exponent += insignificant_digits; | |
931 | ||
932 | if (octal) { | |
933 | double result; | |
934 | bool result_is_junk; | |
935 | char* start = buffer; | |
936 | result = RadixStringToIeee<3>(&start, | |
937 | buffer + buffer_pos, | |
938 | sign, | |
939 | allow_trailing_junk, | |
940 | junk_string_value_, | |
941 | read_as_double, | |
942 | &result_is_junk); | |
943 | ASSERT(!result_is_junk); | |
944 | *processed_characters_count = static_cast<int>(current - input); | |
945 | return result; | |
946 | } | |
947 | ||
948 | if (nonzero_digit_dropped) { | |
949 | buffer[buffer_pos++] = '1'; | |
950 | exponent--; | |
951 | } | |
952 | ||
953 | ASSERT(buffer_pos < kBufferSize); | |
954 | buffer[buffer_pos] = '\0'; | |
955 | ||
956 | double converted; | |
957 | if (read_as_double) { | |
958 | converted = Strtod(Vector<const char>(buffer, buffer_pos), exponent); | |
959 | } else { | |
960 | converted = Strtof(Vector<const char>(buffer, buffer_pos), exponent); | |
961 | } | |
962 | *processed_characters_count = static_cast<int>(current - input); | |
963 | return sign? -converted: converted; | |
964 | } | |
965 | ||
966 | ||
967 | double StringToDoubleConverter::StringToDouble( | |
968 | const char* buffer, | |
969 | int length, | |
970 | int* processed_characters_count) const { | |
971 | return StringToIeee(buffer, length, true, processed_characters_count); | |
972 | } | |
973 | ||
974 | ||
975 | double StringToDoubleConverter::StringToDouble( | |
976 | const uc16* buffer, | |
977 | int length, | |
978 | int* processed_characters_count) const { | |
979 | return StringToIeee(buffer, length, true, processed_characters_count); | |
980 | } | |
981 | ||
982 | ||
983 | float StringToDoubleConverter::StringToFloat( | |
984 | const char* buffer, | |
985 | int length, | |
986 | int* processed_characters_count) const { | |
987 | return static_cast<float>(StringToIeee(buffer, length, false, | |
988 | processed_characters_count)); | |
989 | } | |
990 | ||
991 | ||
992 | float StringToDoubleConverter::StringToFloat( | |
993 | const uc16* buffer, | |
994 | int length, | |
995 | int* processed_characters_count) const { | |
996 | return static_cast<float>(StringToIeee(buffer, length, false, | |
997 | processed_characters_count)); | |
998 | } | |
999 | ||
1000 | } // namespace double_conversion | |
1001 | ||
1002 | // ICU PATCH: Close ICU namespace | |
1003 | U_NAMESPACE_END | |
1004 | #endif // ICU PATCH: close #if !UCONFIG_NO_FORMATTING |