• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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 <climits>
38 #include <cstdarg>
39 
40 // ICU PATCH: Customize header file paths for ICU.
41 
42 #include "double-conversion-bignum.h"
43 #include "double-conversion-cached-powers.h"
44 #include "double-conversion-ieee.h"
45 #include "double-conversion-strtod.h"
46 
47 // ICU PATCH: Wrap in ICU namespace
48 U_NAMESPACE_BEGIN
49 
50 namespace double_conversion {
51 
52 #if defined(DOUBLE_CONVERSION_CORRECT_DOUBLE_OPERATIONS)
53 // 2^53 = 9007199254740992.
54 // Any integer with at most 15 decimal digits will hence fit into a double
55 // (which has a 53bit significand) without loss of precision.
56 static const int kMaxExactDoubleIntegerDecimalDigits = 15;
57 #endif // #if defined(DOUBLE_CONVERSION_CORRECT_DOUBLE_OPERATIONS)
58 // 2^64 = 18446744073709551616 > 10^19
59 static const int kMaxUint64DecimalDigits = 19;
60 
61 // Max double: 1.7976931348623157 x 10^308
62 // Min non-zero double: 4.9406564584124654 x 10^-324
63 // Any x >= 10^309 is interpreted as +infinity.
64 // Any x <= 10^-324 is interpreted as 0.
65 // Note that 2.5e-324 (despite being smaller than the min double) will be read
66 // as non-zero (equal to the min non-zero double).
67 static const int kMaxDecimalPower = 309;
68 static const int kMinDecimalPower = -324;
69 
70 // 2^64 = 18446744073709551616
71 static const uint64_t kMaxUint64 = DOUBLE_CONVERSION_UINT64_2PART_C(0xFFFFFFFF, FFFFFFFF);
72 
73 
74 #if defined(DOUBLE_CONVERSION_CORRECT_DOUBLE_OPERATIONS)
75 static const double exact_powers_of_ten[] = {
76   1.0,  // 10^0
77   10.0,
78   100.0,
79   1000.0,
80   10000.0,
81   100000.0,
82   1000000.0,
83   10000000.0,
84   100000000.0,
85   1000000000.0,
86   10000000000.0,  // 10^10
87   100000000000.0,
88   1000000000000.0,
89   10000000000000.0,
90   100000000000000.0,
91   1000000000000000.0,
92   10000000000000000.0,
93   100000000000000000.0,
94   1000000000000000000.0,
95   10000000000000000000.0,
96   100000000000000000000.0,  // 10^20
97   1000000000000000000000.0,
98   // 10^22 = 0x21e19e0c9bab2400000 = 0x878678326eac9 * 2^22
99   10000000000000000000000.0
100 };
101 static const int kExactPowersOfTenSize = DOUBLE_CONVERSION_ARRAY_SIZE(exact_powers_of_ten);
102 #endif // #if defined(DOUBLE_CONVERSION_CORRECT_DOUBLE_OPERATIONS)
103 
104 // Maximum number of significant digits in the decimal representation.
105 // In fact the value is 772 (see conversions.cc), but to give us some margin
106 // we round up to 780.
107 static const int kMaxSignificantDecimalDigits = 780;
108 
TrimLeadingZeros(Vector<const char> buffer)109 static Vector<const char> TrimLeadingZeros(Vector<const char> buffer) {
110   for (int i = 0; i < buffer.length(); i++) {
111     if (buffer[i] != '0') {
112       return buffer.SubVector(i, buffer.length());
113     }
114   }
115   return Vector<const char>(buffer.start(), 0);
116 }
117 
CutToMaxSignificantDigits(Vector<const char> buffer,int exponent,char * significant_buffer,int * significant_exponent)118 static void CutToMaxSignificantDigits(Vector<const char> buffer,
119                                        int exponent,
120                                        char* significant_buffer,
121                                        int* significant_exponent) {
122   for (int i = 0; i < kMaxSignificantDecimalDigits - 1; ++i) {
123     significant_buffer[i] = buffer[i];
124   }
125   // The input buffer has been trimmed. Therefore the last digit must be
126   // different from '0'.
127   DOUBLE_CONVERSION_ASSERT(buffer[buffer.length() - 1] != '0');
128   // Set the last digit to be non-zero. This is sufficient to guarantee
129   // correct rounding.
130   significant_buffer[kMaxSignificantDecimalDigits - 1] = '1';
131   *significant_exponent =
132       exponent + (buffer.length() - kMaxSignificantDecimalDigits);
133 }
134 
135 
136 // Trims the buffer and cuts it to at most kMaxSignificantDecimalDigits.
137 // If possible the input-buffer is reused, but if the buffer needs to be
138 // modified (due to cutting), then the input needs to be copied into the
139 // buffer_copy_space.
TrimAndCut(Vector<const char> buffer,int exponent,char * buffer_copy_space,int space_size,Vector<const char> * trimmed,int * updated_exponent)140 static void TrimAndCut(Vector<const char> buffer, int exponent,
141                        char* buffer_copy_space, int space_size,
142                        Vector<const char>* trimmed, int* updated_exponent) {
143   Vector<const char> left_trimmed = TrimLeadingZeros(buffer);
144   Vector<const char> right_trimmed = TrimTrailingZeros(left_trimmed);
145   exponent += left_trimmed.length() - right_trimmed.length();
146   if (right_trimmed.length() > kMaxSignificantDecimalDigits) {
147     (void) space_size;  // Mark variable as used.
148     DOUBLE_CONVERSION_ASSERT(space_size >= kMaxSignificantDecimalDigits);
149     CutToMaxSignificantDigits(right_trimmed, exponent,
150                               buffer_copy_space, updated_exponent);
151     *trimmed = Vector<const char>(buffer_copy_space,
152                                  kMaxSignificantDecimalDigits);
153   } else {
154     *trimmed = right_trimmed;
155     *updated_exponent = exponent;
156   }
157 }
158 
159 
160 // Reads digits from the buffer and converts them to a uint64.
161 // Reads in as many digits as fit into a uint64.
162 // When the string starts with "1844674407370955161" no further digit is read.
163 // Since 2^64 = 18446744073709551616 it would still be possible read another
164 // digit if it was less or equal than 6, but this would complicate the code.
ReadUint64(Vector<const char> buffer,int * number_of_read_digits)165 static uint64_t ReadUint64(Vector<const char> buffer,
166                            int* number_of_read_digits) {
167   uint64_t result = 0;
168   int i = 0;
169   while (i < buffer.length() && result <= (kMaxUint64 / 10 - 1)) {
170     int digit = buffer[i++] - '0';
171     DOUBLE_CONVERSION_ASSERT(0 <= digit && digit <= 9);
172     result = 10 * result + digit;
173   }
174   *number_of_read_digits = i;
175   return result;
176 }
177 
178 
179 // Reads a DiyFp from the buffer.
180 // The returned DiyFp is not necessarily normalized.
181 // If remaining_decimals is zero then the returned DiyFp is accurate.
182 // Otherwise it has been rounded and has error of at most 1/2 ulp.
ReadDiyFp(Vector<const char> buffer,DiyFp * result,int * remaining_decimals)183 static void ReadDiyFp(Vector<const char> buffer,
184                       DiyFp* result,
185                       int* remaining_decimals) {
186   int read_digits;
187   uint64_t significand = ReadUint64(buffer, &read_digits);
188   if (buffer.length() == read_digits) {
189     *result = DiyFp(significand, 0);
190     *remaining_decimals = 0;
191   } else {
192     // Round the significand.
193     if (buffer[read_digits] >= '5') {
194       significand++;
195     }
196     // Compute the binary exponent.
197     int exponent = 0;
198     *result = DiyFp(significand, exponent);
199     *remaining_decimals = buffer.length() - read_digits;
200   }
201 }
202 
203 
DoubleStrtod(Vector<const char> trimmed,int exponent,double * result)204 static bool DoubleStrtod(Vector<const char> trimmed,
205                          int exponent,
206                          double* result) {
207 #if !defined(DOUBLE_CONVERSION_CORRECT_DOUBLE_OPERATIONS)
208   // Avoid "unused parameter" warnings
209   (void) trimmed;
210   (void) exponent;
211   (void) result;
212   // On x86 the floating-point stack can be 64 or 80 bits wide. If it is
213   // 80 bits wide (as is the case on Linux) then double-rounding occurs and the
214   // result is not accurate.
215   // We know that Windows32 uses 64 bits and is therefore accurate.
216   return false;
217 #else
218   if (trimmed.length() <= kMaxExactDoubleIntegerDecimalDigits) {
219     int read_digits;
220     // The trimmed input fits into a double.
221     // If the 10^exponent (resp. 10^-exponent) fits into a double too then we
222     // can compute the result-double simply by multiplying (resp. dividing) the
223     // two numbers.
224     // This is possible because IEEE guarantees that floating-point operations
225     // return the best possible approximation.
226     if (exponent < 0 && -exponent < kExactPowersOfTenSize) {
227       // 10^-exponent fits into a double.
228       *result = static_cast<double>(ReadUint64(trimmed, &read_digits));
229       DOUBLE_CONVERSION_ASSERT(read_digits == trimmed.length());
230       *result /= exact_powers_of_ten[-exponent];
231       return true;
232     }
233     if (0 <= exponent && exponent < kExactPowersOfTenSize) {
234       // 10^exponent fits into a double.
235       *result = static_cast<double>(ReadUint64(trimmed, &read_digits));
236       DOUBLE_CONVERSION_ASSERT(read_digits == trimmed.length());
237       *result *= exact_powers_of_ten[exponent];
238       return true;
239     }
240     int remaining_digits =
241         kMaxExactDoubleIntegerDecimalDigits - trimmed.length();
242     if ((0 <= exponent) &&
243         (exponent - remaining_digits < kExactPowersOfTenSize)) {
244       // The trimmed string was short and we can multiply it with
245       // 10^remaining_digits. As a result the remaining exponent now fits
246       // into a double too.
247       *result = static_cast<double>(ReadUint64(trimmed, &read_digits));
248       DOUBLE_CONVERSION_ASSERT(read_digits == trimmed.length());
249       *result *= exact_powers_of_ten[remaining_digits];
250       *result *= exact_powers_of_ten[exponent - remaining_digits];
251       return true;
252     }
253   }
254   return false;
255 #endif
256 }
257 
258 
259 // Returns 10^exponent as an exact DiyFp.
260 // The given exponent must be in the range [1; kDecimalExponentDistance[.
AdjustmentPowerOfTen(int exponent)261 static DiyFp AdjustmentPowerOfTen(int exponent) {
262   DOUBLE_CONVERSION_ASSERT(0 < exponent);
263   DOUBLE_CONVERSION_ASSERT(exponent < PowersOfTenCache::kDecimalExponentDistance);
264   // Simply hardcode the remaining powers for the given decimal exponent
265   // distance.
266   DOUBLE_CONVERSION_ASSERT(PowersOfTenCache::kDecimalExponentDistance == 8);
267   switch (exponent) {
268     case 1: return DiyFp(DOUBLE_CONVERSION_UINT64_2PART_C(0xa0000000, 00000000), -60);
269     case 2: return DiyFp(DOUBLE_CONVERSION_UINT64_2PART_C(0xc8000000, 00000000), -57);
270     case 3: return DiyFp(DOUBLE_CONVERSION_UINT64_2PART_C(0xfa000000, 00000000), -54);
271     case 4: return DiyFp(DOUBLE_CONVERSION_UINT64_2PART_C(0x9c400000, 00000000), -50);
272     case 5: return DiyFp(DOUBLE_CONVERSION_UINT64_2PART_C(0xc3500000, 00000000), -47);
273     case 6: return DiyFp(DOUBLE_CONVERSION_UINT64_2PART_C(0xf4240000, 00000000), -44);
274     case 7: return DiyFp(DOUBLE_CONVERSION_UINT64_2PART_C(0x98968000, 00000000), -40);
275     default:
276       DOUBLE_CONVERSION_UNREACHABLE();
277   }
278 }
279 
280 
281 // If the function returns true then the result is the correct double.
282 // Otherwise it is either the correct double or the double that is just below
283 // the correct double.
DiyFpStrtod(Vector<const char> buffer,int exponent,double * result)284 static bool DiyFpStrtod(Vector<const char> buffer,
285                         int exponent,
286                         double* result) {
287   DiyFp input;
288   int remaining_decimals;
289   ReadDiyFp(buffer, &input, &remaining_decimals);
290   // Since we may have dropped some digits the input is not accurate.
291   // If remaining_decimals is different than 0 than the error is at most
292   // .5 ulp (unit in the last place).
293   // We don't want to deal with fractions and therefore keep a common
294   // denominator.
295   const int kDenominatorLog = 3;
296   const int kDenominator = 1 << kDenominatorLog;
297   // Move the remaining decimals into the exponent.
298   exponent += remaining_decimals;
299   uint64_t error = (remaining_decimals == 0 ? 0 : kDenominator / 2);
300 
301   int old_e = input.e();
302   input.Normalize();
303   error <<= old_e - input.e();
304 
305   DOUBLE_CONVERSION_ASSERT(exponent <= PowersOfTenCache::kMaxDecimalExponent);
306   if (exponent < PowersOfTenCache::kMinDecimalExponent) {
307     *result = 0.0;
308     return true;
309   }
310   DiyFp cached_power;
311   int cached_decimal_exponent;
312   PowersOfTenCache::GetCachedPowerForDecimalExponent(exponent,
313                                                      &cached_power,
314                                                      &cached_decimal_exponent);
315 
316   if (cached_decimal_exponent != exponent) {
317     int adjustment_exponent = exponent - cached_decimal_exponent;
318     DiyFp adjustment_power = AdjustmentPowerOfTen(adjustment_exponent);
319     input.Multiply(adjustment_power);
320     if (kMaxUint64DecimalDigits - buffer.length() >= adjustment_exponent) {
321       // The product of input with the adjustment power fits into a 64 bit
322       // integer.
323       DOUBLE_CONVERSION_ASSERT(DiyFp::kSignificandSize == 64);
324     } else {
325       // The adjustment power is exact. There is hence only an error of 0.5.
326       error += kDenominator / 2;
327     }
328   }
329 
330   input.Multiply(cached_power);
331   // The error introduced by a multiplication of a*b equals
332   //   error_a + error_b + error_a*error_b/2^64 + 0.5
333   // Substituting a with 'input' and b with 'cached_power' we have
334   //   error_b = 0.5  (all cached powers have an error of less than 0.5 ulp),
335   //   error_ab = 0 or 1 / kDenominator > error_a*error_b/ 2^64
336   int error_b = kDenominator / 2;
337   int error_ab = (error == 0 ? 0 : 1);  // We round up to 1.
338   int fixed_error = kDenominator / 2;
339   error += error_b + error_ab + fixed_error;
340 
341   old_e = input.e();
342   input.Normalize();
343   error <<= old_e - input.e();
344 
345   // See if the double's significand changes if we add/subtract the error.
346   int order_of_magnitude = DiyFp::kSignificandSize + input.e();
347   int effective_significand_size =
348       Double::SignificandSizeForOrderOfMagnitude(order_of_magnitude);
349   int precision_digits_count =
350       DiyFp::kSignificandSize - effective_significand_size;
351   if (precision_digits_count + kDenominatorLog >= DiyFp::kSignificandSize) {
352     // This can only happen for very small denormals. In this case the
353     // half-way multiplied by the denominator exceeds the range of an uint64.
354     // Simply shift everything to the right.
355     int shift_amount = (precision_digits_count + kDenominatorLog) -
356         DiyFp::kSignificandSize + 1;
357     input.set_f(input.f() >> shift_amount);
358     input.set_e(input.e() + shift_amount);
359     // We add 1 for the lost precision of error, and kDenominator for
360     // the lost precision of input.f().
361     error = (error >> shift_amount) + 1 + kDenominator;
362     precision_digits_count -= shift_amount;
363   }
364   // We use uint64_ts now. This only works if the DiyFp uses uint64_ts too.
365   DOUBLE_CONVERSION_ASSERT(DiyFp::kSignificandSize == 64);
366   DOUBLE_CONVERSION_ASSERT(precision_digits_count < 64);
367   uint64_t one64 = 1;
368   uint64_t precision_bits_mask = (one64 << precision_digits_count) - 1;
369   uint64_t precision_bits = input.f() & precision_bits_mask;
370   uint64_t half_way = one64 << (precision_digits_count - 1);
371   precision_bits *= kDenominator;
372   half_way *= kDenominator;
373   DiyFp rounded_input(input.f() >> precision_digits_count,
374                       input.e() + precision_digits_count);
375   if (precision_bits >= half_way + error) {
376     rounded_input.set_f(rounded_input.f() + 1);
377   }
378   // If the last_bits are too close to the half-way case than we are too
379   // inaccurate and round down. In this case we return false so that we can
380   // fall back to a more precise algorithm.
381 
382   *result = Double(rounded_input).value();
383   if (half_way - error < precision_bits && precision_bits < half_way + error) {
384     // Too imprecise. The caller will have to fall back to a slower version.
385     // However the returned number is guaranteed to be either the correct
386     // double, or the next-lower double.
387     return false;
388   } else {
389     return true;
390   }
391 }
392 
393 
394 // Returns
395 //   - -1 if buffer*10^exponent < diy_fp.
396 //   -  0 if buffer*10^exponent == diy_fp.
397 //   - +1 if buffer*10^exponent > diy_fp.
398 // Preconditions:
399 //   buffer.length() + exponent <= kMaxDecimalPower + 1
400 //   buffer.length() + exponent > kMinDecimalPower
401 //   buffer.length() <= kMaxDecimalSignificantDigits
CompareBufferWithDiyFp(Vector<const char> buffer,int exponent,DiyFp diy_fp)402 static int CompareBufferWithDiyFp(Vector<const char> buffer,
403                                   int exponent,
404                                   DiyFp diy_fp) {
405   DOUBLE_CONVERSION_ASSERT(buffer.length() + exponent <= kMaxDecimalPower + 1);
406   DOUBLE_CONVERSION_ASSERT(buffer.length() + exponent > kMinDecimalPower);
407   DOUBLE_CONVERSION_ASSERT(buffer.length() <= kMaxSignificantDecimalDigits);
408   // Make sure that the Bignum will be able to hold all our numbers.
409   // Our Bignum implementation has a separate field for exponents. Shifts will
410   // consume at most one bigit (< 64 bits).
411   // ln(10) == 3.3219...
412   DOUBLE_CONVERSION_ASSERT(((kMaxDecimalPower + 1) * 333 / 100) < Bignum::kMaxSignificantBits);
413   Bignum buffer_bignum;
414   Bignum diy_fp_bignum;
415   buffer_bignum.AssignDecimalString(buffer);
416   diy_fp_bignum.AssignUInt64(diy_fp.f());
417   if (exponent >= 0) {
418     buffer_bignum.MultiplyByPowerOfTen(exponent);
419   } else {
420     diy_fp_bignum.MultiplyByPowerOfTen(-exponent);
421   }
422   if (diy_fp.e() > 0) {
423     diy_fp_bignum.ShiftLeft(diy_fp.e());
424   } else {
425     buffer_bignum.ShiftLeft(-diy_fp.e());
426   }
427   return Bignum::Compare(buffer_bignum, diy_fp_bignum);
428 }
429 
430 
431 // Returns true if the guess is the correct double.
432 // Returns false, when guess is either correct or the next-lower double.
ComputeGuess(Vector<const char> trimmed,int exponent,double * guess)433 static bool ComputeGuess(Vector<const char> trimmed, int exponent,
434                          double* guess) {
435   if (trimmed.length() == 0) {
436     *guess = 0.0;
437     return true;
438   }
439   if (exponent + trimmed.length() - 1 >= kMaxDecimalPower) {
440     *guess = Double::Infinity();
441     return true;
442   }
443   if (exponent + trimmed.length() <= kMinDecimalPower) {
444     *guess = 0.0;
445     return true;
446   }
447 
448   if (DoubleStrtod(trimmed, exponent, guess) ||
449       DiyFpStrtod(trimmed, exponent, guess)) {
450     return true;
451   }
452   if (*guess == Double::Infinity()) {
453     return true;
454   }
455   return false;
456 }
457 
458 #if U_DEBUG // needed for ICU only in debug mode
IsDigit(const char d)459 static bool IsDigit(const char d) {
460   return ('0' <= d) && (d <= '9');
461 }
462 
IsNonZeroDigit(const char d)463 static bool IsNonZeroDigit(const char d) {
464   return ('1' <= d) && (d <= '9');
465 }
466 
467 #ifdef __has_cpp_attribute
468 #if __has_cpp_attribute(maybe_unused)
469 [[maybe_unused]]
470 #endif
471 #endif
AssertTrimmedDigits(const Vector<const char> & buffer)472 static bool AssertTrimmedDigits(const Vector<const char>& buffer) {
473   for(int i = 0; i < buffer.length(); ++i) {
474     if(!IsDigit(buffer[i])) {
475       return false;
476     }
477   }
478   return (buffer.length() == 0) || (IsNonZeroDigit(buffer[0]) && IsNonZeroDigit(buffer[buffer.length()-1]));
479 }
480 #endif // needed for ICU only in debug mode
481 
StrtodTrimmed(Vector<const char> trimmed,int exponent)482 double StrtodTrimmed(Vector<const char> trimmed, int exponent) {
483   DOUBLE_CONVERSION_ASSERT(trimmed.length() <= kMaxSignificantDecimalDigits);
484   DOUBLE_CONVERSION_ASSERT(AssertTrimmedDigits(trimmed));
485   double guess;
486   const bool is_correct = ComputeGuess(trimmed, exponent, &guess);
487   if (is_correct) {
488     return guess;
489   }
490   DiyFp upper_boundary = Double(guess).UpperBoundary();
491   int comparison = CompareBufferWithDiyFp(trimmed, exponent, upper_boundary);
492   if (comparison < 0) {
493     return guess;
494   } else if (comparison > 0) {
495     return Double(guess).NextDouble();
496   } else if ((Double(guess).Significand() & 1) == 0) {
497     // Round towards even.
498     return guess;
499   } else {
500     return Double(guess).NextDouble();
501   }
502 }
503 
Strtod(Vector<const char> buffer,int exponent)504 double Strtod(Vector<const char> buffer, int exponent) {
505   char copy_buffer[kMaxSignificantDecimalDigits];
506   Vector<const char> trimmed;
507   int updated_exponent;
508   TrimAndCut(buffer, exponent, copy_buffer, kMaxSignificantDecimalDigits,
509              &trimmed, &updated_exponent);
510   return StrtodTrimmed(trimmed, updated_exponent);
511 }
512 
SanitizedDoubletof(double d)513 static float SanitizedDoubletof(double d) {
514   DOUBLE_CONVERSION_ASSERT(d >= 0.0);
515   // ASAN has a sanitize check that disallows casting doubles to floats if
516   // they are too big.
517   // https://clang.llvm.org/docs/UndefinedBehaviorSanitizer.html#available-checks
518   // The behavior should be covered by IEEE 754, but some projects use this
519   // flag, so work around it.
520   float max_finite = 3.4028234663852885981170418348451692544e+38;
521   // The half-way point between the max-finite and infinity value.
522   // Since infinity has an even significand everything equal or greater than
523   // this value should become infinity.
524   double half_max_finite_infinity =
525       3.40282356779733661637539395458142568448e+38;
526   if (d >= max_finite) {
527     if (d >= half_max_finite_infinity) {
528       return Single::Infinity();
529     } else {
530       return max_finite;
531     }
532   } else {
533     return static_cast<float>(d);
534   }
535 }
536 
Strtof(Vector<const char> buffer,int exponent)537 float Strtof(Vector<const char> buffer, int exponent) {
538   char copy_buffer[kMaxSignificantDecimalDigits];
539   Vector<const char> trimmed;
540   int updated_exponent;
541   TrimAndCut(buffer, exponent, copy_buffer, kMaxSignificantDecimalDigits,
542              &trimmed, &updated_exponent);
543   exponent = updated_exponent;
544   return StrtofTrimmed(trimmed, exponent);
545 }
546 
StrtofTrimmed(Vector<const char> trimmed,int exponent)547 float StrtofTrimmed(Vector<const char> trimmed, int exponent) {
548   DOUBLE_CONVERSION_ASSERT(trimmed.length() <= kMaxSignificantDecimalDigits);
549   DOUBLE_CONVERSION_ASSERT(AssertTrimmedDigits(trimmed));
550 
551   double double_guess;
552   bool is_correct = ComputeGuess(trimmed, exponent, &double_guess);
553 
554   float float_guess = SanitizedDoubletof(double_guess);
555   if (float_guess == double_guess) {
556     // This shortcut triggers for integer values.
557     return float_guess;
558   }
559 
560   // We must catch double-rounding. Say the double has been rounded up, and is
561   // now a boundary of a float, and rounds up again. This is why we have to
562   // look at previous too.
563   // Example (in decimal numbers):
564   //    input: 12349
565   //    high-precision (4 digits): 1235
566   //    low-precision (3 digits):
567   //       when read from input: 123
568   //       when rounded from high precision: 124.
569   // To do this we simply look at the neigbors of the correct result and see
570   // if they would round to the same float. If the guess is not correct we have
571   // to look at four values (since two different doubles could be the correct
572   // double).
573 
574   double double_next = Double(double_guess).NextDouble();
575   double double_previous = Double(double_guess).PreviousDouble();
576 
577   float f1 = SanitizedDoubletof(double_previous);
578   float f2 = float_guess;
579   float f3 = SanitizedDoubletof(double_next);
580   float f4;
581   if (is_correct) {
582     f4 = f3;
583   } else {
584     double double_next2 = Double(double_next).NextDouble();
585     f4 = SanitizedDoubletof(double_next2);
586   }
587   (void) f2;  // Mark variable as used.
588   DOUBLE_CONVERSION_ASSERT(f1 <= f2 && f2 <= f3 && f3 <= f4);
589 
590   // If the guess doesn't lie near a single-precision boundary we can simply
591   // return its float-value.
592   if (f1 == f4) {
593     return float_guess;
594   }
595 
596   DOUBLE_CONVERSION_ASSERT((f1 != f2 && f2 == f3 && f3 == f4) ||
597          (f1 == f2 && f2 != f3 && f3 == f4) ||
598          (f1 == f2 && f2 == f3 && f3 != f4));
599 
600   // guess and next are the two possible candidates (in the same way that
601   // double_guess was the lower candidate for a double-precision guess).
602   float guess = f1;
603   float next = f4;
604   DiyFp upper_boundary;
605   if (guess == 0.0f) {
606     float min_float = 1e-45f;
607     upper_boundary = Double(static_cast<double>(min_float) / 2).AsDiyFp();
608   } else {
609     upper_boundary = Single(guess).UpperBoundary();
610   }
611   int comparison = CompareBufferWithDiyFp(trimmed, exponent, upper_boundary);
612   if (comparison < 0) {
613     return guess;
614   } else if (comparison > 0) {
615     return next;
616   } else if ((Single(guess).Significand() & 1) == 0) {
617     // Round towards even.
618     return guess;
619   } else {
620     return next;
621   }
622 }
623 
624 }  // namespace double_conversion
625 
626 // ICU PATCH: Close ICU namespace
627 U_NAMESPACE_END
628 #endif // ICU PATCH: close #if !UCONFIG_NO_FORMATTING
629