Home | History | Annotate | Download | only in src
      1 // Copyright 2011 the V8 project authors. All rights reserved.
      2 // Redistribution and use in source and binary forms, with or without
      3 // modification, are permitted provided that the following conditions are
      4 // met:
      5 //
      6 //     * Redistributions of source code must retain the above copyright
      7 //       notice, this list of conditions and the following disclaimer.
      8 //     * Redistributions in binary form must reproduce the above
      9 //       copyright notice, this list of conditions and the following
     10 //       disclaimer in the documentation and/or other materials provided
     11 //       with the distribution.
     12 //     * Neither the name of Google Inc. nor the names of its
     13 //       contributors may be used to endorse or promote products derived
     14 //       from this software without specific prior written permission.
     15 //
     16 // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
     17 // "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
     18 // LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
     19 // A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
     20 // OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
     21 // SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
     22 // LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
     23 // DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
     24 // THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
     25 // (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
     26 // OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
     27 
     28 #ifndef V8_CONVERSIONS_INL_H_
     29 #define V8_CONVERSIONS_INL_H_
     30 
     31 #include <limits.h>        // Required for INT_MAX etc.
     32 #include <math.h>
     33 #include <float.h>         // Required for DBL_MAX and on Win32 for finite()
     34 #include <stdarg.h>
     35 #include "globals.h"       // Required for V8_INFINITY
     36 
     37 // ----------------------------------------------------------------------------
     38 // Extra POSIX/ANSI functions for Win32/MSVC.
     39 
     40 #include "conversions.h"
     41 #include "double.h"
     42 #include "platform.h"
     43 #include "scanner.h"
     44 #include "strtod.h"
     45 
     46 namespace v8 {
     47 namespace internal {
     48 
     49 inline double JunkStringValue() {
     50   return BitCast<double, uint64_t>(kQuietNaNMask);
     51 }
     52 
     53 
     54 // The fast double-to-unsigned-int conversion routine does not guarantee
     55 // rounding towards zero, or any reasonable value if the argument is larger
     56 // than what fits in an unsigned 32-bit integer.
     57 inline unsigned int FastD2UI(double x) {
     58   // There is no unsigned version of lrint, so there is no fast path
     59   // in this function as there is in FastD2I. Using lrint doesn't work
     60   // for values of 2^31 and above.
     61 
     62   // Convert "small enough" doubles to uint32_t by fixing the 32
     63   // least significant non-fractional bits in the low 32 bits of the
     64   // double, and reading them from there.
     65   const double k2Pow52 = 4503599627370496.0;
     66   bool negative = x < 0;
     67   if (negative) {
     68     x = -x;
     69   }
     70   if (x < k2Pow52) {
     71     x += k2Pow52;
     72     uint32_t result;
     73     Address mantissa_ptr = reinterpret_cast<Address>(&x);
     74     // Copy least significant 32 bits of mantissa.
     75     memcpy(&result, mantissa_ptr, sizeof(result));
     76     return negative ? ~result + 1 : result;
     77   }
     78   // Large number (outside uint32 range), Infinity or NaN.
     79   return 0x80000000u;  // Return integer indefinite.
     80 }
     81 
     82 
     83 inline double DoubleToInteger(double x) {
     84   if (isnan(x)) return 0;
     85   if (!isfinite(x) || x == 0) return x;
     86   return (x >= 0) ? floor(x) : ceil(x);
     87 }
     88 
     89 
     90 int32_t DoubleToInt32(double x) {
     91   int32_t i = FastD2I(x);
     92   if (FastI2D(i) == x) return i;
     93   Double d(x);
     94   int exponent = d.Exponent();
     95   if (exponent < 0) {
     96     if (exponent <= -Double::kSignificandSize) return 0;
     97     return d.Sign() * static_cast<int32_t>(d.Significand() >> -exponent);
     98   } else {
     99     if (exponent > 31) return 0;
    100     return d.Sign() * static_cast<int32_t>(d.Significand() << exponent);
    101   }
    102 }
    103 
    104 
    105 template <class Iterator, class EndMark>
    106 bool SubStringEquals(Iterator* current,
    107                      EndMark end,
    108                      const char* substring) {
    109   ASSERT(**current == *substring);
    110   for (substring++; *substring != '\0'; substring++) {
    111     ++*current;
    112     if (*current == end || **current != *substring) return false;
    113   }
    114   ++*current;
    115   return true;
    116 }
    117 
    118 
    119 // Returns true if a nonspace character has been found and false if the
    120 // end was been reached before finding a nonspace character.
    121 template <class Iterator, class EndMark>
    122 inline bool AdvanceToNonspace(UnicodeCache* unicode_cache,
    123                               Iterator* current,
    124                               EndMark end) {
    125   while (*current != end) {
    126     if (!unicode_cache->IsWhiteSpace(**current)) return true;
    127     ++*current;
    128   }
    129   return false;
    130 }
    131 
    132 
    133 // Parsing integers with radix 2, 4, 8, 16, 32. Assumes current != end.
    134 template <int radix_log_2, class Iterator, class EndMark>
    135 double InternalStringToIntDouble(UnicodeCache* unicode_cache,
    136                                  Iterator current,
    137                                  EndMark end,
    138                                  bool negative,
    139                                  bool allow_trailing_junk) {
    140   ASSERT(current != end);
    141 
    142   // Skip leading 0s.
    143   while (*current == '0') {
    144     ++current;
    145     if (current == end) return SignedZero(negative);
    146   }
    147 
    148   int64_t number = 0;
    149   int exponent = 0;
    150   const int radix = (1 << radix_log_2);
    151 
    152   do {
    153     int digit;
    154     if (*current >= '0' && *current <= '9' && *current < '0' + radix) {
    155       digit = static_cast<char>(*current) - '0';
    156     } else if (radix > 10 && *current >= 'a' && *current < 'a' + radix - 10) {
    157       digit = static_cast<char>(*current) - 'a' + 10;
    158     } else if (radix > 10 && *current >= 'A' && *current < 'A' + radix - 10) {
    159       digit = static_cast<char>(*current) - 'A' + 10;
    160     } else {
    161       if (allow_trailing_junk ||
    162           !AdvanceToNonspace(unicode_cache, &current, end)) {
    163         break;
    164       } else {
    165         return JunkStringValue();
    166       }
    167     }
    168 
    169     number = number * radix + digit;
    170     int overflow = static_cast<int>(number >> 53);
    171     if (overflow != 0) {
    172       // Overflow occurred. Need to determine which direction to round the
    173       // result.
    174       int overflow_bits_count = 1;
    175       while (overflow > 1) {
    176         overflow_bits_count++;
    177         overflow >>= 1;
    178       }
    179 
    180       int dropped_bits_mask = ((1 << overflow_bits_count) - 1);
    181       int dropped_bits = static_cast<int>(number) & dropped_bits_mask;
    182       number >>= overflow_bits_count;
    183       exponent = overflow_bits_count;
    184 
    185       bool zero_tail = true;
    186       while (true) {
    187         ++current;
    188         if (current == end || !isDigit(*current, radix)) break;
    189         zero_tail = zero_tail && *current == '0';
    190         exponent += radix_log_2;
    191       }
    192 
    193       if (!allow_trailing_junk &&
    194           AdvanceToNonspace(unicode_cache, &current, end)) {
    195         return JunkStringValue();
    196       }
    197 
    198       int middle_value = (1 << (overflow_bits_count - 1));
    199       if (dropped_bits > middle_value) {
    200         number++;  // Rounding up.
    201       } else if (dropped_bits == middle_value) {
    202         // Rounding to even to consistency with decimals: half-way case rounds
    203         // up if significant part is odd and down otherwise.
    204         if ((number & 1) != 0 || !zero_tail) {
    205           number++;  // Rounding up.
    206         }
    207       }
    208 
    209       // Rounding up may cause overflow.
    210       if ((number & ((int64_t)1 << 53)) != 0) {
    211         exponent++;
    212         number >>= 1;
    213       }
    214       break;
    215     }
    216     ++current;
    217   } while (current != end);
    218 
    219   ASSERT(number < ((int64_t)1 << 53));
    220   ASSERT(static_cast<int64_t>(static_cast<double>(number)) == number);
    221 
    222   if (exponent == 0) {
    223     if (negative) {
    224       if (number == 0) return -0.0;
    225       number = -number;
    226     }
    227     return static_cast<double>(number);
    228   }
    229 
    230   ASSERT(number != 0);
    231   // The double could be constructed faster from number (mantissa), exponent
    232   // and sign. Assuming it's a rare case more simple code is used.
    233   return static_cast<double>(negative ? -number : number) * pow(2.0, exponent);
    234 }
    235 
    236 
    237 template <class Iterator, class EndMark>
    238 double InternalStringToInt(UnicodeCache* unicode_cache,
    239                            Iterator current,
    240                            EndMark end,
    241                            int radix) {
    242   const bool allow_trailing_junk = true;
    243   const double empty_string_val = JunkStringValue();
    244 
    245   if (!AdvanceToNonspace(unicode_cache, &current, end)) {
    246     return empty_string_val;
    247   }
    248 
    249   bool negative = false;
    250   bool leading_zero = false;
    251 
    252   if (*current == '+') {
    253     // Ignore leading sign; skip following spaces.
    254     ++current;
    255     if (current == end) {
    256       return JunkStringValue();
    257     }
    258   } else if (*current == '-') {
    259     ++current;
    260     if (current == end) {
    261       return JunkStringValue();
    262     }
    263     negative = true;
    264   }
    265 
    266   if (radix == 0) {
    267     // Radix detection.
    268     if (*current == '0') {
    269       ++current;
    270       if (current == end) return SignedZero(negative);
    271       if (*current == 'x' || *current == 'X') {
    272         radix = 16;
    273         ++current;
    274         if (current == end) return JunkStringValue();
    275       } else {
    276         radix = 8;
    277         leading_zero = true;
    278       }
    279     } else {
    280       radix = 10;
    281     }
    282   } else if (radix == 16) {
    283     if (*current == '0') {
    284       // Allow "0x" prefix.
    285       ++current;
    286       if (current == end) return SignedZero(negative);
    287       if (*current == 'x' || *current == 'X') {
    288         ++current;
    289         if (current == end) return JunkStringValue();
    290       } else {
    291         leading_zero = true;
    292       }
    293     }
    294   }
    295 
    296   if (radix < 2 || radix > 36) return JunkStringValue();
    297 
    298   // Skip leading zeros.
    299   while (*current == '0') {
    300     leading_zero = true;
    301     ++current;
    302     if (current == end) return SignedZero(negative);
    303   }
    304 
    305   if (!leading_zero && !isDigit(*current, radix)) {
    306     return JunkStringValue();
    307   }
    308 
    309   if (IsPowerOf2(radix)) {
    310     switch (radix) {
    311       case 2:
    312         return InternalStringToIntDouble<1>(
    313             unicode_cache, current, end, negative, allow_trailing_junk);
    314       case 4:
    315         return InternalStringToIntDouble<2>(
    316             unicode_cache, current, end, negative, allow_trailing_junk);
    317       case 8:
    318         return InternalStringToIntDouble<3>(
    319             unicode_cache, current, end, negative, allow_trailing_junk);
    320 
    321       case 16:
    322         return InternalStringToIntDouble<4>(
    323             unicode_cache, current, end, negative, allow_trailing_junk);
    324 
    325       case 32:
    326         return InternalStringToIntDouble<5>(
    327             unicode_cache, current, end, negative, allow_trailing_junk);
    328       default:
    329         UNREACHABLE();
    330     }
    331   }
    332 
    333   if (radix == 10) {
    334     // Parsing with strtod.
    335     const int kMaxSignificantDigits = 309;  // Doubles are less than 1.8e308.
    336     // The buffer may contain up to kMaxSignificantDigits + 1 digits and a zero
    337     // end.
    338     const int kBufferSize = kMaxSignificantDigits + 2;
    339     char buffer[kBufferSize];
    340     int buffer_pos = 0;
    341     while (*current >= '0' && *current <= '9') {
    342       if (buffer_pos <= kMaxSignificantDigits) {
    343         // If the number has more than kMaxSignificantDigits it will be parsed
    344         // as infinity.
    345         ASSERT(buffer_pos < kBufferSize);
    346         buffer[buffer_pos++] = static_cast<char>(*current);
    347       }
    348       ++current;
    349       if (current == end) break;
    350     }
    351 
    352     if (!allow_trailing_junk &&
    353         AdvanceToNonspace(unicode_cache, &current, end)) {
    354       return JunkStringValue();
    355     }
    356 
    357     ASSERT(buffer_pos < kBufferSize);
    358     buffer[buffer_pos] = '\0';
    359     Vector<const char> buffer_vector(buffer, buffer_pos);
    360     return negative ? -Strtod(buffer_vector, 0) : Strtod(buffer_vector, 0);
    361   }
    362 
    363   // The following code causes accumulating rounding error for numbers greater
    364   // than ~2^56. It's explicitly allowed in the spec: "if R is not 2, 4, 8, 10,
    365   // 16, or 32, then mathInt may be an implementation-dependent approximation to
    366   // the mathematical integer value" (15.1.2.2).
    367 
    368   int lim_0 = '0' + (radix < 10 ? radix : 10);
    369   int lim_a = 'a' + (radix - 10);
    370   int lim_A = 'A' + (radix - 10);
    371 
    372   // NOTE: The code for computing the value may seem a bit complex at
    373   // first glance. It is structured to use 32-bit multiply-and-add
    374   // loops as long as possible to avoid loosing precision.
    375 
    376   double v = 0.0;
    377   bool done = false;
    378   do {
    379     // Parse the longest part of the string starting at index j
    380     // possible while keeping the multiplier, and thus the part
    381     // itself, within 32 bits.
    382     unsigned int part = 0, multiplier = 1;
    383     while (true) {
    384       int d;
    385       if (*current >= '0' && *current < lim_0) {
    386         d = *current - '0';
    387       } else if (*current >= 'a' && *current < lim_a) {
    388         d = *current - 'a' + 10;
    389       } else if (*current >= 'A' && *current < lim_A) {
    390         d = *current - 'A' + 10;
    391       } else {
    392         done = true;
    393         break;
    394       }
    395 
    396       // Update the value of the part as long as the multiplier fits
    397       // in 32 bits. When we can't guarantee that the next iteration
    398       // will not overflow the multiplier, we stop parsing the part
    399       // by leaving the loop.
    400       const unsigned int kMaximumMultiplier = 0xffffffffU / 36;
    401       uint32_t m = multiplier * radix;
    402       if (m > kMaximumMultiplier) break;
    403       part = part * radix + d;
    404       multiplier = m;
    405       ASSERT(multiplier > part);
    406 
    407       ++current;
    408       if (current == end) {
    409         done = true;
    410         break;
    411       }
    412     }
    413 
    414     // Update the value and skip the part in the string.
    415     v = v * multiplier + part;
    416   } while (!done);
    417 
    418   if (!allow_trailing_junk &&
    419       AdvanceToNonspace(unicode_cache, &current, end)) {
    420     return JunkStringValue();
    421   }
    422 
    423   return negative ? -v : v;
    424 }
    425 
    426 
    427 // Converts a string to a double value. Assumes the Iterator supports
    428 // the following operations:
    429 // 1. current == end (other ops are not allowed), current != end.
    430 // 2. *current - gets the current character in the sequence.
    431 // 3. ++current (advances the position).
    432 template <class Iterator, class EndMark>
    433 double InternalStringToDouble(UnicodeCache* unicode_cache,
    434                               Iterator current,
    435                               EndMark end,
    436                               int flags,
    437                               double empty_string_val) {
    438   // To make sure that iterator dereferencing is valid the following
    439   // convention is used:
    440   // 1. Each '++current' statement is followed by check for equality to 'end'.
    441   // 2. If AdvanceToNonspace returned false then current == end.
    442   // 3. If 'current' becomes be equal to 'end' the function returns or goes to
    443   // 'parsing_done'.
    444   // 4. 'current' is not dereferenced after the 'parsing_done' label.
    445   // 5. Code before 'parsing_done' may rely on 'current != end'.
    446   if (!AdvanceToNonspace(unicode_cache, &current, end)) {
    447     return empty_string_val;
    448   }
    449 
    450   const bool allow_trailing_junk = (flags & ALLOW_TRAILING_JUNK) != 0;
    451 
    452   // The longest form of simplified number is: "-<significant digits>'.1eXXX\0".
    453   const int kBufferSize = kMaxSignificantDigits + 10;
    454   char buffer[kBufferSize];  // NOLINT: size is known at compile time.
    455   int buffer_pos = 0;
    456 
    457   // Exponent will be adjusted if insignificant digits of the integer part
    458   // or insignificant leading zeros of the fractional part are dropped.
    459   int exponent = 0;
    460   int significant_digits = 0;
    461   int insignificant_digits = 0;
    462   bool nonzero_digit_dropped = false;
    463 
    464   bool negative = false;
    465 
    466   if (*current == '+') {
    467     // Ignore leading sign.
    468     ++current;
    469     if (current == end) return JunkStringValue();
    470   } else if (*current == '-') {
    471     ++current;
    472     if (current == end) return JunkStringValue();
    473     negative = true;
    474   }
    475 
    476   static const char kInfinitySymbol[] = "Infinity";
    477   if (*current == kInfinitySymbol[0]) {
    478     if (!SubStringEquals(&current, end, kInfinitySymbol)) {
    479       return JunkStringValue();
    480     }
    481 
    482     if (!allow_trailing_junk &&
    483         AdvanceToNonspace(unicode_cache, &current, end)) {
    484       return JunkStringValue();
    485     }
    486 
    487     ASSERT(buffer_pos == 0);
    488     return negative ? -V8_INFINITY : V8_INFINITY;
    489   }
    490 
    491   bool leading_zero = false;
    492   if (*current == '0') {
    493     ++current;
    494     if (current == end) return SignedZero(negative);
    495 
    496     leading_zero = true;
    497 
    498     // It could be hexadecimal value.
    499     if ((flags & ALLOW_HEX) && (*current == 'x' || *current == 'X')) {
    500       ++current;
    501       if (current == end || !isDigit(*current, 16)) {
    502         return JunkStringValue();  // "0x".
    503       }
    504 
    505       return InternalStringToIntDouble<4>(unicode_cache,
    506                                           current,
    507                                           end,
    508                                           negative,
    509                                           allow_trailing_junk);
    510     }
    511 
    512     // Ignore leading zeros in the integer part.
    513     while (*current == '0') {
    514       ++current;
    515       if (current == end) return SignedZero(negative);
    516     }
    517   }
    518 
    519   bool octal = leading_zero && (flags & ALLOW_OCTALS) != 0;
    520 
    521   // Copy significant digits of the integer part (if any) to the buffer.
    522   while (*current >= '0' && *current <= '9') {
    523     if (significant_digits < kMaxSignificantDigits) {
    524       ASSERT(buffer_pos < kBufferSize);
    525       buffer[buffer_pos++] = static_cast<char>(*current);
    526       significant_digits++;
    527       // Will later check if it's an octal in the buffer.
    528     } else {
    529       insignificant_digits++;  // Move the digit into the exponential part.
    530       nonzero_digit_dropped = nonzero_digit_dropped || *current != '0';
    531     }
    532     octal = octal && *current < '8';
    533     ++current;
    534     if (current == end) goto parsing_done;
    535   }
    536 
    537   if (significant_digits == 0) {
    538     octal = false;
    539   }
    540 
    541   if (*current == '.') {
    542     if (octal && !allow_trailing_junk) return JunkStringValue();
    543     if (octal) goto parsing_done;
    544 
    545     ++current;
    546     if (current == end) {
    547       if (significant_digits == 0 && !leading_zero) {
    548         return JunkStringValue();
    549       } else {
    550         goto parsing_done;
    551       }
    552     }
    553 
    554     if (significant_digits == 0) {
    555       // octal = false;
    556       // Integer part consists of 0 or is absent. Significant digits start after
    557       // leading zeros (if any).
    558       while (*current == '0') {
    559         ++current;
    560         if (current == end) return SignedZero(negative);
    561         exponent--;  // Move this 0 into the exponent.
    562       }
    563     }
    564 
    565     // There is a fractional part.  We don't emit a '.', but adjust the exponent
    566     // instead.
    567     while (*current >= '0' && *current <= '9') {
    568       if (significant_digits < kMaxSignificantDigits) {
    569         ASSERT(buffer_pos < kBufferSize);
    570         buffer[buffer_pos++] = static_cast<char>(*current);
    571         significant_digits++;
    572         exponent--;
    573       } else {
    574         // Ignore insignificant digits in the fractional part.
    575         nonzero_digit_dropped = nonzero_digit_dropped || *current != '0';
    576       }
    577       ++current;
    578       if (current == end) goto parsing_done;
    579     }
    580   }
    581 
    582   if (!leading_zero && exponent == 0 && significant_digits == 0) {
    583     // If leading_zeros is true then the string contains zeros.
    584     // If exponent < 0 then string was [+-]\.0*...
    585     // If significant_digits != 0 the string is not equal to 0.
    586     // Otherwise there are no digits in the string.
    587     return JunkStringValue();
    588   }
    589 
    590   // Parse exponential part.
    591   if (*current == 'e' || *current == 'E') {
    592     if (octal) return JunkStringValue();
    593     ++current;
    594     if (current == end) {
    595       if (allow_trailing_junk) {
    596         goto parsing_done;
    597       } else {
    598         return JunkStringValue();
    599       }
    600     }
    601     char sign = '+';
    602     if (*current == '+' || *current == '-') {
    603       sign = static_cast<char>(*current);
    604       ++current;
    605       if (current == end) {
    606         if (allow_trailing_junk) {
    607           goto parsing_done;
    608         } else {
    609           return JunkStringValue();
    610         }
    611       }
    612     }
    613 
    614     if (current == end || *current < '0' || *current > '9') {
    615       if (allow_trailing_junk) {
    616         goto parsing_done;
    617       } else {
    618         return JunkStringValue();
    619       }
    620     }
    621 
    622     const int max_exponent = INT_MAX / 2;
    623     ASSERT(-max_exponent / 2 <= exponent && exponent <= max_exponent / 2);
    624     int num = 0;
    625     do {
    626       // Check overflow.
    627       int digit = *current - '0';
    628       if (num >= max_exponent / 10
    629           && !(num == max_exponent / 10 && digit <= max_exponent % 10)) {
    630         num = max_exponent;
    631       } else {
    632         num = num * 10 + digit;
    633       }
    634       ++current;
    635     } while (current != end && *current >= '0' && *current <= '9');
    636 
    637     exponent += (sign == '-' ? -num : num);
    638   }
    639 
    640   if (!allow_trailing_junk &&
    641       AdvanceToNonspace(unicode_cache, &current, end)) {
    642     return JunkStringValue();
    643   }
    644 
    645   parsing_done:
    646   exponent += insignificant_digits;
    647 
    648   if (octal) {
    649     return InternalStringToIntDouble<3>(unicode_cache,
    650                                         buffer,
    651                                         buffer + buffer_pos,
    652                                         negative,
    653                                         allow_trailing_junk);
    654   }
    655 
    656   if (nonzero_digit_dropped) {
    657     buffer[buffer_pos++] = '1';
    658     exponent--;
    659   }
    660 
    661   ASSERT(buffer_pos < kBufferSize);
    662   buffer[buffer_pos] = '\0';
    663 
    664   double converted = Strtod(Vector<const char>(buffer, buffer_pos), exponent);
    665   return negative ? -converted : converted;
    666 }
    667 
    668 } }  // namespace v8::internal
    669 
    670 #endif  // V8_CONVERSIONS_INL_H_
    671