123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303 |
- // Copyright 2020 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.
- #ifndef BASE_STRINGS_STRING_NUMBER_CONVERSIONS_INTERNAL_H_
- #define BASE_STRINGS_STRING_NUMBER_CONVERSIONS_INTERNAL_H_
- #include <ctype.h>
- #include <errno.h>
- #include <stdlib.h>
- #include <wctype.h>
- #include <limits>
- #include "base/check_op.h"
- #include "base/logging.h"
- #include "base/no_destructor.h"
- #include "base/numerics/safe_math.h"
- #include "base/strings/string_util.h"
- #include "base/third_party/double_conversion/double-conversion/double-conversion.h"
- namespace base {
- namespace internal {
- template <typename STR, typename INT>
- static STR IntToStringT(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<INT>::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.
- std::make_unsigned_t<INT> res =
- CheckedNumeric<INT>(value).UnsignedAbs().ValueOrDie();
- CHR* end = outbuf + kOutputBufSize;
- CHR* i = end;
- do {
- --i;
- DCHECK(i != outbuf);
- *i = static_cast<CHR>((res % 10) + '0');
- res /= 10;
- } while (res != 0);
- if (IsValueNegative(value)) {
- --i;
- DCHECK(i != outbuf);
- *i = static_cast<CHR>('-');
- }
- return STR(i, end);
- }
- // Utility to convert a character to a digit in a given base
- template <int BASE, typename CHAR>
- Optional<uint8_t> CharToDigit(CHAR c) {
- static_assert(1 <= BASE && BASE <= 36, "BASE needs to be in [1, 36]");
- if (c >= '0' && c < '0' + std::min(BASE, 10))
- return c - '0';
- if (c >= 'a' && c < 'a' + BASE - 10)
- return c - 'a' + 10;
- if (c >= 'A' && c < 'A' + BASE - 10)
- return c - 'A' + 10;
- return base::nullopt;
- }
- // 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 <typename CHAR>
- class WhitespaceHelper {};
- template <>
- class WhitespaceHelper<char> {
- public:
- static bool Invoke(char c) {
- return 0 != isspace(static_cast<unsigned char>(c));
- }
- };
- template <>
- class WhitespaceHelper<char16> {
- public:
- static bool Invoke(char16 c) { return 0 != iswspace(c); }
- };
- template <typename CHAR>
- bool LocalIsWhitespace(CHAR c) {
- return WhitespaceHelper<CHAR>::Invoke(c);
- }
- template <typename Number, int kBase>
- class StringToNumberParser {
- public:
- struct Result {
- Number value = 0;
- bool valid = false;
- };
- static constexpr Number kMin = std::numeric_limits<Number>::min();
- static constexpr Number kMax = std::numeric_limits<Number>::max();
- // 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 <typename Sign>
- class Base {
- public:
- template <typename Iter>
- static Result Invoke(Iter begin, Iter end) {
- Number value = 0;
- if (begin == end) {
- return {value, false};
- }
- // Note: no performance difference was found when using template
- // specialization to remove this check in bases other than 16
- if (kBase == 16 && end - begin > 2 && *begin == '0' &&
- (*(begin + 1) == 'x' || *(begin + 1) == 'X')) {
- begin += 2;
- }
- for (Iter current = begin; current != end; ++current) {
- Optional<uint8_t> new_digit = CharToDigit<kBase>(*current);
- if (!new_digit) {
- return {value, false};
- }
- if (current != begin) {
- Result result = Sign::CheckBounds(value, *new_digit);
- if (!result.valid)
- return result;
- value *= kBase;
- }
- value = Sign::Increment(value, *new_digit);
- }
- return {value, true};
- }
- };
- class Positive : public Base<Positive> {
- public:
- static Result CheckBounds(Number value, uint8_t new_digit) {
- if (value > static_cast<Number>(kMax / kBase) ||
- (value == static_cast<Number>(kMax / kBase) &&
- new_digit > kMax % kBase)) {
- return {kMax, false};
- }
- return {value, true};
- }
- static Number Increment(Number lhs, uint8_t rhs) { return lhs + rhs; }
- };
- class Negative : public Base<Negative> {
- public:
- static Result CheckBounds(Number value, uint8_t new_digit) {
- if (value < kMin / kBase ||
- (value == kMin / kBase && new_digit > 0 - kMin % kBase)) {
- return {kMin, false};
- }
- return {value, true};
- }
- static Number Increment(Number lhs, uint8_t rhs) { return lhs - rhs; }
- };
- };
- template <typename Number, int kBase, typename Str>
- auto StringToNumber(BasicStringPiece<Str> input) {
- using Parser = StringToNumberParser<Number, kBase>;
- using Result = typename Parser::Result;
- bool has_leading_whitespace = false;
- auto begin = input.begin();
- auto end = input.end();
- while (begin != end && LocalIsWhitespace(*begin)) {
- has_leading_whitespace = true;
- ++begin;
- }
- if (begin != end && *begin == '-') {
- if (!std::numeric_limits<Number>::is_signed) {
- return Result{0, false};
- }
- Result result = Parser::Negative::Invoke(begin + 1, end);
- result.valid &= !has_leading_whitespace;
- return result;
- }
- if (begin != end && *begin == '+') {
- ++begin;
- }
- Result result = Parser::Positive::Invoke(begin, end);
- result.valid &= !has_leading_whitespace;
- return result;
- }
- template <typename STR, typename VALUE>
- bool StringToIntImpl(BasicStringPiece<STR> input, VALUE& output) {
- auto result = StringToNumber<VALUE, 10>(input);
- output = result.value;
- return result.valid;
- }
- template <typename STR, typename VALUE>
- bool HexStringToIntImpl(BasicStringPiece<STR> input, VALUE& output) {
- auto result = StringToNumber<VALUE, 16>(input);
- output = result.value;
- return result.valid;
- }
- static const double_conversion::DoubleToStringConverter*
- GetDoubleToStringConverter() {
- static NoDestructor<double_conversion::DoubleToStringConverter> converter(
- double_conversion::DoubleToStringConverter::EMIT_POSITIVE_EXPONENT_SIGN,
- nullptr, nullptr, 'e', -6, 12, 0, 0);
- return converter.get();
- }
- // Converts a given (data, size) pair to a desired string type. For
- // performance reasons, this dispatches to a different constructor if the
- // passed-in data matches the string's value_type.
- template <typename StringT>
- StringT ToString(const typename StringT::value_type* data, size_t size) {
- return StringT(data, size);
- }
- template <typename StringT, typename CharT>
- StringT ToString(const CharT* data, size_t size) {
- return StringT(data, data + size);
- }
- template <typename StringT>
- StringT DoubleToStringT(double value) {
- char buffer[32];
- double_conversion::StringBuilder builder(buffer, sizeof(buffer));
- GetDoubleToStringConverter()->ToShortest(value, &builder);
- return ToString<StringT>(buffer, builder.position());
- }
- template <typename STRING, typename CHAR>
- bool StringToDoubleImpl(STRING input, const CHAR* data, double& output) {
- static NoDestructor<double_conversion::StringToDoubleConverter> converter(
- double_conversion::StringToDoubleConverter::ALLOW_LEADING_SPACES |
- double_conversion::StringToDoubleConverter::ALLOW_TRAILING_JUNK,
- 0.0, 0, nullptr, nullptr);
- int processed_characters_count;
- output = converter->StringToDouble(data, input.size(),
- &processed_characters_count);
- // Cases to return false:
- // - If the input string is empty, there was nothing to parse.
- // - If the value saturated to HUGE_VAL.
- // - If the entire string was not processed, there are either characters
- // remaining in the string after a parsed number, or the string does not
- // begin with a parseable number.
- // - If the first character is a space, there was leading whitespace
- return !input.empty() && output != HUGE_VAL && output != -HUGE_VAL &&
- static_cast<size_t>(processed_characters_count) == input.size() &&
- !IsUnicodeWhitespace(input[0]);
- }
- template <typename OutIter>
- static bool HexStringToByteContainer(StringPiece input, OutIter output) {
- size_t count = input.size();
- if (count == 0 || (count % 2) != 0)
- return false;
- for (uintptr_t i = 0; i < count / 2; ++i) {
- // most significant 4 bits
- Optional<uint8_t> msb = CharToDigit<16>(input[i * 2]);
- // least significant 4 bits
- Optional<uint8_t> lsb = CharToDigit<16>(input[i * 2 + 1]);
- if (!msb || !lsb) {
- return false;
- }
- *(output++) = (*msb << 4) | *lsb;
- }
- return true;
- }
- } // namespace internal
- } // namespace base
- #endif // BASE_STRINGS_STRING_NUMBER_CONVERSIONS_INTERNAL_H_
|