// Copyright (c) 2012 The Chromium Authors. All rights reserved. // Use of this source code is governed by a BSD-style license that can be // found in the LICENSE file. #include "base/strings/string_number_conversions.h" #include #include #include #include #include #include "base/logging.h" #include "base/numerics/safe_conversions.h" #include "base/numerics/safe_math.h" #include "base/scoped_clear_errno.h" #include "base/strings/utf_string_conversions.h" #include "base/third_party/dmg_fp/dmg_fp.h" namespace base { namespace { template struct IntToStringT { static STR IntToString(INT value) { // log10(2) ~= 0.3 bytes needed per bit or per byte log10(2**8) ~= 2.4. // So round up to allocate 3 output characters per byte, plus 1 for '-'. const size_t kOutputBufSize = 3 * sizeof(INT) + std::numeric_limits::is_signed; // Create the string in a temporary buffer, write it back to front, and // then return the substr of what we ended up using. using CHR = typename STR::value_type; CHR outbuf[kOutputBufSize]; // The ValueOrDie call below can never fail, because UnsignedAbs is valid // for all valid inputs. auto res = CheckedNumeric(value).UnsignedAbs().ValueOrDie(); CHR* end = outbuf + kOutputBufSize; CHR* i = end; do { --i; DCHECK(i != outbuf); *i = static_cast((res % 10) + '0'); res /= 10; } while (res != 0); if (IsValueNegative(value)) { --i; DCHECK(i != outbuf); *i = static_cast('-'); } return STR(i, end); } }; // Utility to convert a character to a digit in a given base template class BaseCharToDigit { }; // Faster specialization for bases <= 10 template class BaseCharToDigit { public: static bool Convert(CHAR c, uint8_t* digit) { if (c >= '0' && c < '0' + BASE) { *digit = static_cast(c - '0'); return true; } return false; } }; // Specialization for bases where 10 < base <= 36 template class BaseCharToDigit { public: static bool Convert(CHAR c, uint8_t* digit) { if (c >= '0' && c <= '9') { *digit = c - '0'; } else if (c >= 'a' && c < 'a' + BASE - 10) { *digit = c - 'a' + 10; } else if (c >= 'A' && c < 'A' + BASE - 10) { *digit = c - 'A' + 10; } else { return false; } return true; } }; template bool CharToDigit(CHAR c, uint8_t* digit) { return BaseCharToDigit::Convert(c, digit); } // There is an IsUnicodeWhitespace for wchars defined in string_util.h, but it // is locale independent, whereas the functions we are replacing were // locale-dependent. TBD what is desired, but for the moment let's not // introduce a change in behaviour. template class WhitespaceHelper { }; template<> class WhitespaceHelper { public: static bool Invoke(char c) { return 0 != isspace(static_cast(c)); } }; template<> class WhitespaceHelper { public: static bool Invoke(char16 c) { return 0 != iswspace(c); } }; template bool LocalIsWhitespace(CHAR c) { return WhitespaceHelper::Invoke(c); } // IteratorRangeToNumberTraits should provide: // - a typedef for iterator_type, the iterator type used as input. // - a typedef for value_type, the target numeric type. // - static functions min, max (returning the minimum and maximum permitted // values) // - constant kBase, the base in which to interpret the input template class IteratorRangeToNumber { public: typedef IteratorRangeToNumberTraits traits; typedef typename traits::iterator_type const_iterator; typedef typename traits::value_type value_type; // Generalized iterator-range-to-number conversion. // static bool Invoke(const_iterator begin, const_iterator end, value_type* output) { bool valid = true; while (begin != end && LocalIsWhitespace(*begin)) { valid = false; ++begin; } if (begin != end && *begin == '-') { if (!std::numeric_limits::is_signed) { valid = false; } else if (!Negative::Invoke(begin + 1, end, output)) { valid = false; } } else { if (begin != end && *begin == '+') { ++begin; } if (!Positive::Invoke(begin, end, output)) { valid = false; } } return valid; } private: // Sign provides: // - a static function, CheckBounds, that determines whether the next digit // causes an overflow/underflow // - a static function, Increment, that appends the next digit appropriately // according to the sign of the number being parsed. template class Base { public: static bool Invoke(const_iterator begin, const_iterator end, typename traits::value_type* output) { *output = 0; if (begin == end) { return false; } // Note: no performance difference was found when using template // specialization to remove this check in bases other than 16 if (traits::kBase == 16 && end - begin > 2 && *begin == '0' && (*(begin + 1) == 'x' || *(begin + 1) == 'X')) { begin += 2; } for (const_iterator current = begin; current != end; ++current) { uint8_t new_digit = 0; if (!CharToDigit(*current, &new_digit)) { return false; } if (current != begin) { if (!Sign::CheckBounds(output, new_digit)) { return false; } *output *= traits::kBase; } Sign::Increment(new_digit, output); } return true; } }; class Positive : public Base { public: static bool CheckBounds(value_type* output, uint8_t new_digit) { if (*output > static_cast(traits::max() / traits::kBase) || (*output == static_cast(traits::max() / traits::kBase) && new_digit > traits::max() % traits::kBase)) { *output = traits::max(); return false; } return true; } static void Increment(uint8_t increment, value_type* output) { *output += increment; } }; class Negative : public Base { public: static bool CheckBounds(value_type* output, uint8_t new_digit) { if (*output < traits::min() / traits::kBase || (*output == traits::min() / traits::kBase && new_digit > 0 - traits::min() % traits::kBase)) { *output = traits::min(); return false; } return true; } static void Increment(uint8_t increment, value_type* output) { *output -= increment; } }; }; template class BaseIteratorRangeToNumberTraits { public: typedef ITERATOR iterator_type; typedef VALUE value_type; static value_type min() { return std::numeric_limits::min(); } static value_type max() { return std::numeric_limits::max(); } static const int kBase = BASE; }; template class BaseHexIteratorRangeToIntTraits : public BaseIteratorRangeToNumberTraits { }; template class BaseHexIteratorRangeToUIntTraits : public BaseIteratorRangeToNumberTraits {}; template class BaseHexIteratorRangeToInt64Traits : public BaseIteratorRangeToNumberTraits {}; template class BaseHexIteratorRangeToUInt64Traits : public BaseIteratorRangeToNumberTraits {}; typedef BaseHexIteratorRangeToIntTraits HexIteratorRangeToIntTraits; typedef BaseHexIteratorRangeToUIntTraits HexIteratorRangeToUIntTraits; typedef BaseHexIteratorRangeToInt64Traits HexIteratorRangeToInt64Traits; typedef BaseHexIteratorRangeToUInt64Traits HexIteratorRangeToUInt64Traits; template bool HexStringToBytesT(const STR& input, std::vector* output) { DCHECK_EQ(output->size(), 0u); size_t count = input.size(); if (count == 0 || (count % 2) != 0) return false; for (uintptr_t i = 0; i < count / 2; ++i) { uint8_t msb = 0; // most significant 4 bits uint8_t lsb = 0; // least significant 4 bits if (!CharToDigit<16>(input[i * 2], &msb) || !CharToDigit<16>(input[i * 2 + 1], &lsb)) return false; output->push_back((msb << 4) | lsb); } return true; } template class StringPieceToNumberTraits : public BaseIteratorRangeToNumberTraits { }; template bool StringToIntImpl(const StringPiece& input, VALUE* output) { return IteratorRangeToNumber >::Invoke( input.begin(), input.end(), output); } template class StringPiece16ToNumberTraits : public BaseIteratorRangeToNumberTraits { }; template bool String16ToIntImpl(const StringPiece16& input, VALUE* output) { return IteratorRangeToNumber >::Invoke( input.begin(), input.end(), output); } } // namespace std::string IntToString(int value) { return IntToStringT::IntToString(value); } string16 IntToString16(int value) { return IntToStringT::IntToString(value); } std::string UintToString(unsigned int value) { return IntToStringT::IntToString(value); } string16 UintToString16(unsigned int value) { return IntToStringT::IntToString(value); } std::string Int64ToString(int64_t value) { return IntToStringT::IntToString(value); } string16 Int64ToString16(int64_t value) { return IntToStringT::IntToString(value); } std::string Uint64ToString(uint64_t value) { return IntToStringT::IntToString(value); } string16 Uint64ToString16(uint64_t value) { return IntToStringT::IntToString(value); } std::string SizeTToString(size_t value) { return IntToStringT::IntToString(value); } string16 SizeTToString16(size_t value) { return IntToStringT::IntToString(value); } std::string DoubleToString(double value) { // According to g_fmt.cc, it is sufficient to declare a buffer of size 32. char buffer[32]; dmg_fp::g_fmt(buffer, value); return std::string(buffer); } bool StringToInt(const StringPiece& input, int* output) { return StringToIntImpl(input, output); } bool StringToInt(const StringPiece16& input, int* output) { return String16ToIntImpl(input, output); } bool StringToUint(const StringPiece& input, unsigned* output) { return StringToIntImpl(input, output); } bool StringToUint(const StringPiece16& input, unsigned* output) { return String16ToIntImpl(input, output); } bool StringToInt64(const StringPiece& input, int64_t* output) { return StringToIntImpl(input, output); } bool StringToInt64(const StringPiece16& input, int64_t* output) { return String16ToIntImpl(input, output); } bool StringToUint64(const StringPiece& input, uint64_t* output) { return StringToIntImpl(input, output); } bool StringToUint64(const StringPiece16& input, uint64_t* output) { return String16ToIntImpl(input, output); } bool StringToSizeT(const StringPiece& input, size_t* output) { return StringToIntImpl(input, output); } bool StringToSizeT(const StringPiece16& input, size_t* output) { return String16ToIntImpl(input, output); } bool StringToDouble(const std::string& input, double* output) { // Thread-safe? It is on at least Mac, Linux, and Windows. ScopedClearErrno clear_errno; char* endptr = NULL; *output = dmg_fp::strtod(input.c_str(), &endptr); // Cases to return false: // - If errno is ERANGE, there was an overflow or underflow. // - If the input string is empty, there was nothing to parse. // - If endptr does not point to the end of the string, there are either // characters remaining in the string after a parsed number, or the string // does not begin with a parseable number. endptr is compared to the // expected end given the string's stated length to correctly catch cases // where the string contains embedded NUL characters. // - If the first character is a space, there was leading whitespace return errno == 0 && !input.empty() && input.c_str() + input.length() == endptr && !isspace(input[0]); } // Note: if you need to add String16ToDouble, first ask yourself if it's // really necessary. If it is, probably the best implementation here is to // convert to 8-bit and then use the 8-bit version. // Note: if you need to add an iterator range version of StringToDouble, first // ask yourself if it's really necessary. If it is, probably the best // implementation here is to instantiate a string and use the string version. std::string HexEncode(const void* bytes, size_t size) { static const char kHexChars[] = "0123456789ABCDEF"; // Each input byte creates two output hex characters. std::string ret(size * 2, '\0'); for (size_t i = 0; i < size; ++i) { char b = reinterpret_cast(bytes)[i]; ret[(i * 2)] = kHexChars[(b >> 4) & 0xf]; ret[(i * 2) + 1] = kHexChars[b & 0xf]; } return ret; } bool HexStringToInt(const StringPiece& input, int* output) { return IteratorRangeToNumber::Invoke( input.begin(), input.end(), output); } bool HexStringToUInt(const StringPiece& input, uint32_t* output) { return IteratorRangeToNumber::Invoke( input.begin(), input.end(), output); } bool HexStringToInt64(const StringPiece& input, int64_t* output) { return IteratorRangeToNumber::Invoke( input.begin(), input.end(), output); } bool HexStringToUInt64(const StringPiece& input, uint64_t* output) { return IteratorRangeToNumber::Invoke( input.begin(), input.end(), output); } bool HexStringToBytes(const std::string& input, std::vector* output) { return HexStringToBytesT(input, output); } } // namespace base