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/strings/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/numerics/safe_conversions.h" 16 #include "base/numerics/safe_math.h" 17 18 namespace base { 19 20 namespace { 21 22 template <typename STR, typename INT> 23 struct IntToStringT { 24 static STR IntToString(INT value) { 25 // log10(2) ~= 0.3 bytes needed per bit or per byte log10(2**8) ~= 2.4. 26 // So round up to allocate 3 output characters per byte, plus 1 for '-'. 27 const size_t kOutputBufSize = 28 3 * sizeof(INT) + std::numeric_limits<INT>::is_signed; 29 30 // Create the string in a temporary buffer, write it back to front, and 31 // then return the substr of what we ended up using. 32 using CHR = typename STR::value_type; 33 CHR outbuf[kOutputBufSize]; 34 35 // The ValueOrDie call below can never fail, because UnsignedAbs is valid 36 // for all valid inputs. 37 auto res = CheckedNumeric<INT>(value).UnsignedAbs().ValueOrDie(); 38 39 CHR* end = outbuf + kOutputBufSize; 40 CHR* i = end; 41 do { 42 --i; 43 DCHECK(i != outbuf); 44 *i = static_cast<CHR>((res % 10) + '0'); 45 res /= 10; 46 } while (res != 0); 47 if (IsValueNegative(value)) { 48 --i; 49 DCHECK(i != outbuf); 50 *i = static_cast<CHR>('-'); 51 } 52 return STR(i, end); 53 } 54 }; 55 56 // Utility to convert a character to a digit in a given base 57 template<typename CHAR, int BASE, bool BASE_LTE_10> class BaseCharToDigit { 58 }; 59 60 // Faster specialization for bases <= 10 61 template<typename CHAR, int BASE> class BaseCharToDigit<CHAR, BASE, true> { 62 public: 63 static bool Convert(CHAR c, uint8_t* digit) { 64 if (c >= '0' && c < '0' + BASE) { 65 *digit = static_cast<uint8_t>(c - '0'); 66 return true; 67 } 68 return false; 69 } 70 }; 71 72 // Specialization for bases where 10 < base <= 36 73 template<typename CHAR, int BASE> class BaseCharToDigit<CHAR, BASE, false> { 74 public: 75 static bool Convert(CHAR c, uint8_t* digit) { 76 if (c >= '0' && c <= '9') { 77 *digit = c - '0'; 78 } else if (c >= 'a' && c < 'a' + BASE - 10) { 79 *digit = c - 'a' + 10; 80 } else if (c >= 'A' && c < 'A' + BASE - 10) { 81 *digit = c - 'A' + 10; 82 } else { 83 return false; 84 } 85 return true; 86 } 87 }; 88 89 template <int BASE, typename CHAR> 90 bool CharToDigit(CHAR c, uint8_t* digit) { 91 return BaseCharToDigit<CHAR, BASE, BASE <= 10>::Convert(c, digit); 92 } 93 94 // There is an IsUnicodeWhitespace for wchars defined in string_util.h, but it 95 // is locale independent, whereas the functions we are replacing were 96 // locale-dependent. TBD what is desired, but for the moment let's not 97 // introduce a change in behaviour. 98 template<typename CHAR> class WhitespaceHelper { 99 }; 100 101 template<> class WhitespaceHelper<char> { 102 public: 103 static bool Invoke(char c) { 104 return 0 != isspace(static_cast<unsigned char>(c)); 105 } 106 }; 107 108 template<typename CHAR> bool LocalIsWhitespace(CHAR c) { 109 return WhitespaceHelper<CHAR>::Invoke(c); 110 } 111 112 // IteratorRangeToNumberTraits should provide: 113 // - a typedef for iterator_type, the iterator type used as input. 114 // - a typedef for value_type, the target numeric type. 115 // - static functions min, max (returning the minimum and maximum permitted 116 // values) 117 // - constant kBase, the base in which to interpret the input 118 template<typename IteratorRangeToNumberTraits> 119 class IteratorRangeToNumber { 120 public: 121 typedef IteratorRangeToNumberTraits traits; 122 typedef typename traits::iterator_type const_iterator; 123 typedef typename traits::value_type value_type; 124 125 // Generalized iterator-range-to-number conversion. 126 // 127 static bool Invoke(const_iterator begin, 128 const_iterator end, 129 value_type* output) { 130 bool valid = true; 131 132 while (begin != end && LocalIsWhitespace(*begin)) { 133 valid = false; 134 ++begin; 135 } 136 137 if (begin != end && *begin == '-') { 138 if (!std::numeric_limits<value_type>::is_signed) { 139 valid = false; 140 } else if (!Negative::Invoke(begin + 1, end, output)) { 141 valid = false; 142 } 143 } else { 144 if (begin != end && *begin == '+') { 145 ++begin; 146 } 147 if (!Positive::Invoke(begin, end, output)) { 148 valid = false; 149 } 150 } 151 152 return valid; 153 } 154 155 private: 156 // Sign provides: 157 // - a static function, CheckBounds, that determines whether the next digit 158 // causes an overflow/underflow 159 // - a static function, Increment, that appends the next digit appropriately 160 // according to the sign of the number being parsed. 161 template<typename Sign> 162 class Base { 163 public: 164 static bool Invoke(const_iterator begin, const_iterator end, 165 typename traits::value_type* output) { 166 *output = 0; 167 168 if (begin == end) { 169 return false; 170 } 171 172 // Note: no performance difference was found when using template 173 // specialization to remove this check in bases other than 16 174 if (traits::kBase == 16 && end - begin > 2 && *begin == '0' && 175 (*(begin + 1) == 'x' || *(begin + 1) == 'X')) { 176 begin += 2; 177 } 178 179 for (const_iterator current = begin; current != end; ++current) { 180 uint8_t new_digit = 0; 181 182 if (!CharToDigit<traits::kBase>(*current, &new_digit)) { 183 return false; 184 } 185 186 if (current != begin) { 187 if (!Sign::CheckBounds(output, new_digit)) { 188 return false; 189 } 190 *output *= traits::kBase; 191 } 192 193 Sign::Increment(new_digit, output); 194 } 195 return true; 196 } 197 }; 198 199 class Positive : public Base<Positive> { 200 public: 201 static bool CheckBounds(value_type* output, uint8_t new_digit) { 202 if (*output > static_cast<value_type>(traits::max() / traits::kBase) || 203 (*output == static_cast<value_type>(traits::max() / traits::kBase) && 204 new_digit > traits::max() % traits::kBase)) { 205 *output = traits::max(); 206 return false; 207 } 208 return true; 209 } 210 static void Increment(uint8_t increment, value_type* output) { 211 *output += increment; 212 } 213 }; 214 215 class Negative : public Base<Negative> { 216 public: 217 static bool CheckBounds(value_type* output, uint8_t new_digit) { 218 if (*output < traits::min() / traits::kBase || 219 (*output == traits::min() / traits::kBase && 220 new_digit > 0 - traits::min() % traits::kBase)) { 221 *output = traits::min(); 222 return false; 223 } 224 return true; 225 } 226 static void Increment(uint8_t increment, value_type* output) { 227 *output -= increment; 228 } 229 }; 230 }; 231 232 template<typename ITERATOR, typename VALUE, int BASE> 233 class BaseIteratorRangeToNumberTraits { 234 public: 235 typedef ITERATOR iterator_type; 236 typedef VALUE value_type; 237 static value_type min() { 238 return std::numeric_limits<value_type>::min(); 239 } 240 static value_type max() { 241 return std::numeric_limits<value_type>::max(); 242 } 243 static const int kBase = BASE; 244 }; 245 246 template<typename ITERATOR> 247 class BaseHexIteratorRangeToIntTraits 248 : public BaseIteratorRangeToNumberTraits<ITERATOR, int, 16> { 249 }; 250 251 template <typename ITERATOR> 252 class BaseHexIteratorRangeToUIntTraits 253 : public BaseIteratorRangeToNumberTraits<ITERATOR, uint32_t, 16> {}; 254 255 template <typename ITERATOR> 256 class BaseHexIteratorRangeToInt64Traits 257 : public BaseIteratorRangeToNumberTraits<ITERATOR, int64_t, 16> {}; 258 259 template <typename ITERATOR> 260 class BaseHexIteratorRangeToUInt64Traits 261 : public BaseIteratorRangeToNumberTraits<ITERATOR, uint64_t, 16> {}; 262 263 typedef BaseHexIteratorRangeToIntTraits<StringPiece::const_iterator> 264 HexIteratorRangeToIntTraits; 265 266 typedef BaseHexIteratorRangeToUIntTraits<StringPiece::const_iterator> 267 HexIteratorRangeToUIntTraits; 268 269 typedef BaseHexIteratorRangeToInt64Traits<StringPiece::const_iterator> 270 HexIteratorRangeToInt64Traits; 271 272 typedef BaseHexIteratorRangeToUInt64Traits<StringPiece::const_iterator> 273 HexIteratorRangeToUInt64Traits; 274 275 template <typename STR> 276 bool HexStringToBytesT(const STR& input, std::vector<uint8_t>* output) { 277 DCHECK_EQ(output->size(), 0u); 278 size_t count = input.size(); 279 if (count == 0 || (count % 2) != 0) 280 return false; 281 for (uintptr_t i = 0; i < count / 2; ++i) { 282 uint8_t msb = 0; // most significant 4 bits 283 uint8_t lsb = 0; // least significant 4 bits 284 if (!CharToDigit<16>(input[i * 2], &msb) || 285 !CharToDigit<16>(input[i * 2 + 1], &lsb)) 286 return false; 287 output->push_back((msb << 4) | lsb); 288 } 289 return true; 290 } 291 292 template <typename VALUE, int BASE> 293 class StringPieceToNumberTraits 294 : public BaseIteratorRangeToNumberTraits<StringPiece::const_iterator, 295 VALUE, 296 BASE> { 297 }; 298 299 template <typename VALUE> 300 bool StringToIntImpl(const StringPiece& input, VALUE* output) { 301 return IteratorRangeToNumber<StringPieceToNumberTraits<VALUE, 10> >::Invoke( 302 input.begin(), input.end(), output); 303 } 304 305 } // namespace 306 307 std::string IntToString(int value) { 308 return IntToStringT<std::string, int>::IntToString(value); 309 } 310 311 std::string UintToString(unsigned int value) { 312 return IntToStringT<std::string, unsigned int>::IntToString(value); 313 } 314 315 std::string Int64ToString(int64_t value) { 316 return IntToStringT<std::string, int64_t>::IntToString(value); 317 } 318 319 std::string Uint64ToString(uint64_t value) { 320 return IntToStringT<std::string, uint64_t>::IntToString(value); 321 } 322 323 std::string SizeTToString(size_t value) { 324 return IntToStringT<std::string, size_t>::IntToString(value); 325 } 326 327 std::string DoubleToString(double value) { 328 auto ret = std::to_string(value); 329 // If this returned an integer, don't do anything. 330 if (ret.find('.') == std::string::npos) { 331 return ret; 332 } 333 // Otherwise, it has an annoying tendency to leave trailing zeros. 334 size_t len = ret.size(); 335 while (len >= 2 && ret[len - 1] == '0' && ret[len - 2] != '.') { 336 --len; 337 } 338 ret.erase(len); 339 return ret; 340 } 341 342 bool StringToInt(const StringPiece& input, int* output) { 343 return StringToIntImpl(input, output); 344 } 345 346 bool StringToUint(const StringPiece& input, unsigned* output) { 347 return StringToIntImpl(input, output); 348 } 349 350 bool StringToInt64(const StringPiece& input, int64_t* output) { 351 return StringToIntImpl(input, output); 352 } 353 354 bool StringToUint64(const StringPiece& input, uint64_t* output) { 355 return StringToIntImpl(input, output); 356 } 357 358 bool StringToSizeT(const StringPiece& input, size_t* output) { 359 return StringToIntImpl(input, output); 360 } 361 362 bool StringToDouble(const std::string& input, double* output) { 363 char* endptr = nullptr; 364 *output = strtod(input.c_str(), &endptr); 365 366 // Cases to return false: 367 // - If the input string is empty, there was nothing to parse. 368 // - If endptr does not point to the end of the string, there are either 369 // characters remaining in the string after a parsed number, or the string 370 // does not begin with a parseable number. endptr is compared to the 371 // expected end given the string's stated length to correctly catch cases 372 // where the string contains embedded NUL characters. 373 // - If the first character is a space, there was leading whitespace 374 return !input.empty() && 375 input.c_str() + input.length() == endptr && 376 !isspace(input[0]) && 377 *output != std::numeric_limits<double>::infinity() && 378 *output != -std::numeric_limits<double>::infinity(); 379 } 380 381 // Note: if you need to add String16ToDouble, first ask yourself if it's 382 // really necessary. If it is, probably the best implementation here is to 383 // convert to 8-bit and then use the 8-bit version. 384 385 // Note: if you need to add an iterator range version of StringToDouble, first 386 // ask yourself if it's really necessary. If it is, probably the best 387 // implementation here is to instantiate a string and use the string version. 388 389 std::string HexEncode(const void* bytes, size_t size) { 390 static const char kHexChars[] = "0123456789ABCDEF"; 391 392 // Each input byte creates two output hex characters. 393 std::string ret(size * 2, '\0'); 394 395 for (size_t i = 0; i < size; ++i) { 396 char b = reinterpret_cast<const char*>(bytes)[i]; 397 ret[(i * 2)] = kHexChars[(b >> 4) & 0xf]; 398 ret[(i * 2) + 1] = kHexChars[b & 0xf]; 399 } 400 return ret; 401 } 402 403 bool HexStringToInt(const StringPiece& input, int* output) { 404 return IteratorRangeToNumber<HexIteratorRangeToIntTraits>::Invoke( 405 input.begin(), input.end(), output); 406 } 407 408 bool HexStringToUInt(const StringPiece& input, uint32_t* output) { 409 return IteratorRangeToNumber<HexIteratorRangeToUIntTraits>::Invoke( 410 input.begin(), input.end(), output); 411 } 412 413 bool HexStringToInt64(const StringPiece& input, int64_t* output) { 414 return IteratorRangeToNumber<HexIteratorRangeToInt64Traits>::Invoke( 415 input.begin(), input.end(), output); 416 } 417 418 bool HexStringToUInt64(const StringPiece& input, uint64_t* output) { 419 return IteratorRangeToNumber<HexIteratorRangeToUInt64Traits>::Invoke( 420 input.begin(), input.end(), output); 421 } 422 423 bool HexStringToBytes(const std::string& input, std::vector<uint8_t>* output) { 424 return HexStringToBytesT(input, output); 425 } 426 427 } // namespace base 428