base/string_number_conversions.cc

   1 // Copyright (c) 2012 The Chromium Authors. All rights reserved.
   2 // Use of this source code is governed by a BSD-style license that can be
   3 // found in the LICENSE file.
   4
   5 #include "base/string_number_conversions.h"
   6
   7 #include <ctype.h>
   8 #include <errno.h>
   9 #include <stdlib.h>
  10 #include <wctype.h>
  11
  12 #include <limits>
  13
  14 #include "base/logging.h"
  15 #include "base/third_party/dmg_fp/dmg_fp.h"
  16 #include "base/utf_string_conversions.h"
  17
  18 namespace base {
  19
  20 namespace {
  21
  22 template <typename STR, typename INT, typename UINT, bool NEG>
  23 struct IntToStringT {
  24   // This is to avoid a compiler warning about unary minus on unsigned type.
  25   // For example, say you had the following code:
  26   //   template <typename INT>
  27   //   INT abs(INT value) { return value < 0 ? -value : value; }
  28   // Even though if INT is unsigned, it's impossible for value < 0, so the
  29   // unary minus will never be taken, the compiler will still generate a
  30   // warning.  We do a little specialization dance...
  31   template <typename INT2, typename UINT2, bool NEG2>
  32   struct ToUnsignedT {};
  33
  34   template <typename INT2, typename UINT2>
  35   struct ToUnsignedT<INT2, UINT2, false> {
  36     static UINT2 ToUnsigned(INT2 value) {
  37       return static_cast<UINT2>(value);
  38     }
  39   };
  40
  41   template <typename INT2, typename UINT2>
  42   struct ToUnsignedT<INT2, UINT2, true> {
  43     static UINT2 ToUnsigned(INT2 value) {
  44       return static_cast<UINT2>(value < 0 ? -value : value);
  45     }
  46   };
  47
  48   // This set of templates is very similar to the above templates, but
  49   // for testing whether an integer is negative.
  50   template <typename INT2, bool NEG2>
  51   struct TestNegT {};
  52   template <typename INT2>
  53   struct TestNegT<INT2, false> {
  54     static bool TestNeg(INT2 value) {
  55       // value is unsigned, and can never be negative.
  56       return false;
  57     }
  58   };
  59   template <typename INT2>
  60   struct TestNegT<INT2, true> {
  61     static bool TestNeg(INT2 value) {
  62       return value < 0;
  63     }
  64   };
  65
  66   static STR IntToString(INT value) {
  67     // log10(2) ~= 0.3 bytes needed per bit or per byte log10(2**8) ~= 2.4.
  68     // So round up to allocate 3 output characters per byte, plus 1 for '-'.
  69     const int kOutputBufSize = 3 * sizeof(INT) + 1;
  70
  71     // Allocate the whole string right away, we will right back to front, and
  72     // then return the substr of what we ended up using.
  73     STR outbuf(kOutputBufSize, 0);
  74
  75     bool is_neg = TestNegT<INT, NEG>::TestNeg(value);
  76     // Even though is_neg will never be true when INT is parameterized as
  77     // unsigned, even the presence of the unary operation causes a warning.
  78     UINT res = ToUnsignedT<INT, UINT, NEG>::ToUnsigned(value);
  79
  80     for (typename STR::iterator it = outbuf.end();;) {
  81       --it;
  82       DCHECK(it != outbuf.begin());
  83       *it = static_cast<typename STR::value_type>((res % 10) + '0');
  84       res /= 10;
  85
  86       // We're done..
  87       if (res == 0) {
  88         if (is_neg) {
  89           --it;
  90           DCHECK(it != outbuf.begin());
  91           *it = static_cast<typename STR::value_type>('-');
  92         }
  93         return STR(it, outbuf.end());
  94       }
  95     }
  96     NOTREACHED();
  97     return STR();
  98   }
  99 };
 100
 101 // Utility to convert a character to a digit in a given base
 102 template<typename CHAR, int BASE, bool BASE_LTE_10> class BaseCharToDigit {
 103 };
 104
 105 // Faster specialization for bases <= 10
 106 template<typename CHAR, int BASE> class BaseCharToDigit<CHAR, BASE, true> {
 107  public:
 108   static bool Convert(CHAR c, uint8* digit) {
 109     if (c >= '0' && c < '0' + BASE) {
 110       *digit = c - '0';
 111       return true;
 112     }
 113     return false;
 114   }
 115 };
 116
 117 // Specialization for bases where 10 < base <= 36
 118 template<typename CHAR, int BASE> class BaseCharToDigit<CHAR, BASE, false> {
 119  public:
 120   static bool Convert(CHAR c, uint8* digit) {
 121     if (c >= '0' && c <= '9') {
 122       *digit = c - '0';
 123     } else if (c >= 'a' && c < 'a' + BASE - 10) {
 124       *digit = c - 'a' + 10;
 125     } else if (c >= 'A' && c < 'A' + BASE - 10) {
 126       *digit = c - 'A' + 10;
 127     } else {
 128       return false;
 129     }
 130     return true;
 131   }
 132 };
 133
 134 template<int BASE, typename CHAR> bool CharToDigit(CHAR c, uint8* digit) {
 135   return BaseCharToDigit<CHAR, BASE, BASE <= 10>::Convert(c, digit);
 136 }
 137
 138 // There is an IsWhitespace for wchars defined in string_util.h, but it is
 139 // locale independent, whereas the functions we are replacing were
 140 // locale-dependent. TBD what is desired, but for the moment let's not introduce
 141 // a change in behaviour.
 142 template<typename CHAR> class WhitespaceHelper {
 143 };
 144
 145 template<> class WhitespaceHelper<char> {
 146  public:
 147   static bool Invoke(char c) {
 148     return 0 != isspace(static_cast<unsigned char>(c));
 149   }
 150 };
 151
 152 template<> class WhitespaceHelper<char16> {
 153  public:
 154   static bool Invoke(char16 c) {
 155     return 0 != iswspace(c);
 156   }
 157 };
 158
 159 template<typename CHAR> bool LocalIsWhitespace(CHAR c) {
 160   return WhitespaceHelper<CHAR>::Invoke(c);
 161 }
 162
 163 // IteratorRangeToNumberTraits should provide:
 164 //  - a typedef for iterator_type, the iterator type used as input.
 165 //  - a typedef for value_type, the target numeric type.
 166 //  - static functions min, max (returning the minimum and maximum permitted
 167 //    values)
 168 //  - constant kBase, the base in which to interpret the input
 169 template<typename IteratorRangeToNumberTraits>
 170 class IteratorRangeToNumber {
 171  public:
 172   typedef IteratorRangeToNumberTraits traits;
 173   typedef typename traits::iterator_type const_iterator;
 174   typedef typename traits::value_type value_type;
 175
 176   // Generalized iterator-range-to-number conversion.
 177   //
 178   static bool Invoke(const_iterator begin,
 179                      const_iterator end,
 180                      value_type* output) {
 181     bool valid = true;
 182
 183     while (begin != end && LocalIsWhitespace(*begin)) {
 184       valid = false;
 185       ++begin;
 186     }
 187
 188     if (begin != end && *begin == '-') {
 189       if (!Negative::Invoke(begin + 1, end, output)) {
 190         valid = false;
 191       }
 192     } else {
 193       if (begin != end && *begin == '+') {
 194         ++begin;
 195       }
 196       if (!Positive::Invoke(begin, end, output)) {
 197         valid = false;
 198       }
 199     }
 200
 201     return valid;
 202   }
 203
 204  private:
 205   // Sign provides:
 206   //  - a static function, CheckBounds, that determines whether the next digit
 207   //    causes an overflow/underflow
 208   //  - a static function, Increment, that appends the next digit appropriately
 209   //    according to the sign of the number being parsed.
 210   template<typename Sign>
 211   class Base {
 212    public:
 213     static bool Invoke(const_iterator begin, const_iterator end,
 214                        typename traits::value_type* output) {
 215       *output = 0;
 216
 217       if (begin == end) {
 218         return false;
 219       }
 220
 221       // Note: no performance difference was found when using template
 222       // specialization to remove this check in bases other than 16
 223       if (traits::kBase == 16 && end - begin > 2 && *begin == '0' &&
 224           (*(begin + 1) == 'x' || *(begin + 1) == 'X')) {
 225         begin += 2;
 226       }
 227
 228       for (const_iterator current = begin; current != end; ++current) {
 229         uint8 new_digit = 0;
 230
 231         if (!CharToDigit<traits::kBase>(*current, &new_digit)) {
 232           return false;
 233         }
 234
 235         if (current != begin) {
 236           if (!Sign::CheckBounds(output, new_digit)) {
 237             return false;
 238           }
 239           *output *= traits::kBase;
 240         }
 241
 242         Sign::Increment(new_digit, output);
 243       }
 244       return true;
 245     }
 246   };
 247
 248   class Positive : public Base<Positive> {
 249    public:
 250     static bool CheckBounds(value_type* output, uint8 new_digit) {
 251       if (*output > static_cast<value_type>(traits::max() / traits::kBase) ||
 252           (*output == static_cast<value_type>(traits::max() / traits::kBase) &&
 253            new_digit > traits::max() % traits::kBase)) {
 254         *output = traits::max();
 255         return false;
 256       }
 257       return true;
 258     }
 259     static void Increment(uint8 increment, value_type* output) {
 260       *output += increment;
 261     }
 262   };
 263
 264   class Negative : public Base<Negative> {
 265    public:
 266     static bool CheckBounds(value_type* output, uint8 new_digit) {
 267       if (*output < traits::min() / traits::kBase ||
 268           (*output == traits::min() / traits::kBase &&
 269            new_digit > 0 - traits::min() % traits::kBase)) {
 270         *output = traits::min();
 271         return false;
 272       }
 273       return true;
 274     }
 275     static void Increment(uint8 increment, value_type* output) {
 276       *output -= increment;
 277     }
 278   };
 279 };
 280
 281 template<typename ITERATOR, typename VALUE, int BASE>
 282 class BaseIteratorRangeToNumberTraits {
 283  public:
 284   typedef ITERATOR iterator_type;
 285   typedef VALUE value_type;
 286   static value_type min() {
 287     return std::numeric_limits<value_type>::min();
 288   }
 289   static value_type max() {
 290     return std::numeric_limits<value_type>::max();
 291   }
 292   static const int kBase = BASE;
 293 };
 294
 295 template<typename ITERATOR>
 296 class BaseHexIteratorRangeToIntTraits
 297     : public BaseIteratorRangeToNumberTraits<ITERATOR, int, 16> {
 298  public:
 299   // Allow parsing of 0xFFFFFFFF, which is technically an overflow
 300   static unsigned int max() {
 301     return std::numeric_limits<unsigned int>::max();
 302   }
 303 };
 304
 305 typedef BaseHexIteratorRangeToIntTraits<StringPiece::const_iterator>
 306     HexIteratorRangeToIntTraits;
 307
 308 template<typename STR>
 309 bool HexStringToBytesT(const STR& input, std::vector<uint8>* output) {
 310   DCHECK_EQ(output->size(), 0u);
 311   size_t count = input.size();
 312   if (count == 0 || (count % 2) != 0)
 313     return false;
 314   for (uintptr_t i = 0; i < count / 2; ++i) {
 315     uint8 msb = 0;  // most significant 4 bits
 316     uint8 lsb = 0;  // least significant 4 bits
 317     if (!CharToDigit<16>(input[i * 2], &msb) ||
 318         !CharToDigit<16>(input[i * 2 + 1], &lsb))
 319       return false;
 320     output->push_back((msb << 4) | lsb);
 321   }
 322   return true;
 323 }
 324
 325 template <typename VALUE, int BASE>
 326 class StringPieceToNumberTraits
 327     : public BaseIteratorRangeToNumberTraits<StringPiece::const_iterator,
 328                                              VALUE,
 329                                              BASE> {};
 330
 331 template <typename VALUE>
 332 bool StringToIntImpl(const StringPiece& input, VALUE* output) {
 333   return IteratorRangeToNumber<StringPieceToNumberTraits<VALUE, 10> >::Invoke(
 334       input.begin(), input.end(), output);
 335 }
 336
 337 template <typename VALUE, int BASE>
 338 class StringPiece16ToNumberTraits
 339     : public BaseIteratorRangeToNumberTraits<StringPiece16::const_iterator,
 340                                              VALUE,
 341                                              BASE> {};
 342
 343 template <typename VALUE>
 344 bool String16ToIntImpl(const StringPiece16& input, VALUE* output) {
 345   return IteratorRangeToNumber<StringPiece16ToNumberTraits<VALUE, 10> >::Invoke(
 346       input.begin(), input.end(), output);
 347 }
 348
 349 }  // namespace
 350
 351 std::string IntToString(int value) {
 352   return IntToStringT<std::string, int, unsigned int, true>::
 353       IntToString(value);
 354 }
 355
 356 string16 IntToString16(int value) {
 357   return IntToStringT<string16, int, unsigned int, true>::
 358       IntToString(value);
 359 }
 360
 361 std::string UintToString(unsigned int value) {
 362   return IntToStringT<std::string, unsigned int, unsigned int, false>::
 363       IntToString(value);
 364 }
 365
 366 string16 UintToString16(unsigned int value) {
 367   return IntToStringT<string16, unsigned int, unsigned int, false>::
 368       IntToString(value);
 369 }
 370
 371 std::string Int64ToString(int64 value) {
 372   return IntToStringT<std::string, int64, uint64, true>::
 373       IntToString(value);
 374 }
 375
 376 string16 Int64ToString16(int64 value) {
 377   return IntToStringT<string16, int64, uint64, true>::IntToString(value);
 378 }
 379
 380 std::string Uint64ToString(uint64 value) {
 381   return IntToStringT<std::string, uint64, uint64, false>::
 382       IntToString(value);
 383 }
 384
 385 string16 Uint64ToString16(uint64 value) {
 386   return IntToStringT<string16, uint64, uint64, false>::
 387       IntToString(value);
 388 }
 389
 390 std::string DoubleToString(double value) {
 391   // According to g_fmt.cc, it is sufficient to declare a buffer of size 32.
 392   char buffer[32];
 393   dmg_fp::g_fmt(buffer, value);
 394   return std::string(buffer);
 395 }
 396
 397 bool StringToInt(const StringPiece& input, int* output) {
 398   return StringToIntImpl(input, output);
 399 }
 400
 401 bool StringToInt(const StringPiece16& input, int* output) {
 402   return String16ToIntImpl(input, output);
 403 }
 404
 405 bool StringToUint(const StringPiece& input, unsigned* output) {
 406   return StringToIntImpl(input, output);
 407 }
 408
 409 bool StringToUint(const StringPiece16& input, unsigned* output) {
 410   return String16ToIntImpl(input, output);
 411 }
 412
 413 bool StringToInt64(const StringPiece& input, int64* output) {
 414   return StringToIntImpl(input, output);
 415 }
 416
 417 bool StringToInt64(const StringPiece16& input, int64* output) {
 418   return String16ToIntImpl(input, output);
 419 }
 420
 421 bool StringToUint64(const StringPiece& input, uint64* output) {
 422   return StringToIntImpl(input, output);
 423 }
 424
 425 bool StringToUint64(const StringPiece16& input, uint64* output) {
 426   return String16ToIntImpl(input, output);
 427 }
 428
 429 bool StringToSizeT(const StringPiece& input, size_t* output) {
 430   return StringToIntImpl(input, output);
 431 }
 432
 433 bool StringToSizeT(const StringPiece16& input, size_t* output) {
 434   return String16ToIntImpl(input, output);
 435 }
 436
 437 bool StringToDouble(const std::string& input, double* output) {
 438   errno = 0;  // Thread-safe?  It is on at least Mac, Linux, and Windows.
 439   char* endptr = NULL;
 440   *output = dmg_fp::strtod(input.c_str(), &endptr);
 441
 442   // Cases to return false:
 443   //  - If errno is ERANGE, there was an overflow or underflow.
 444   //  - If the input string is empty, there was nothing to parse.
 445   //  - If endptr does not point to the end of the string, there are either
 446   //    characters remaining in the string after a parsed number, or the string
 447   //    does not begin with a parseable number.  endptr is compared to the
 448   //    expected end given the string's stated length to correctly catch cases
 449   //    where the string contains embedded NUL characters.
 450   //  - If the first character is a space, there was leading whitespace
 451   return errno == 0 &&
 452          !input.empty() &&
 453          input.c_str() + input.length() == endptr &&
 454          !isspace(input[0]);
 455 }
 456
 457 // Note: if you need to add String16ToDouble, first ask yourself if it's
 458 // really necessary. If it is, probably the best implementation here is to
 459 // convert to 8-bit and then use the 8-bit version.
 460
 461 // Note: if you need to add an iterator range version of StringToDouble, first
 462 // ask yourself if it's really necessary. If it is, probably the best
 463 // implementation here is to instantiate a string and use the string version.
 464
 465 std::string HexEncode(const void* bytes, size_t size) {
 466   static const char kHexChars[] = "0123456789ABCDEF";
 467
 468   // Each input byte creates two output hex characters.
 469   std::string ret(size * 2, '\0');
 470
 471   for (size_t i = 0; i < size; ++i) {
 472     char b = reinterpret_cast<const char*>(bytes)[i];
 473     ret[(i * 2)] = kHexChars[(b >> 4) & 0xf];
 474     ret[(i * 2) + 1] = kHexChars[b & 0xf];
 475   }
 476   return ret;
 477 }
 478
 479 bool HexStringToInt(const StringPiece& input, int* output) {
 480   return IteratorRangeToNumber<HexIteratorRangeToIntTraits>::Invoke(
 481     input.begin(), input.end(), output);
 482 }
 483
 484 bool HexStringToBytes(const std::string& input, std::vector<uint8>* output) {
 485   return HexStringToBytesT(input, output);
 486 }
 487
 488 }  // namespace base