diff options
| author | thegeorg <thegeorg@yandex-team.com> | 2023-03-05 12:50:38 +0300 |
|---|---|---|
| committer | thegeorg <thegeorg@yandex-team.com> | 2023-03-05 12:50:38 +0300 |
| commit | dc697e5cf6f0cd4d1ff44614a4b1c09a50583d94 (patch) | |
| tree | 151bc18c91b9bb7e7b26791e4d49d387a43f798b /contrib/restricted/abseil-cpp/absl/strings | |
| parent | ab17e559a95ccff2508caeca81d07daafaabf92b (diff) | |
| download | ydb-dc697e5cf6f0cd4d1ff44614a4b1c09a50583d94.tar.gz | |
Update contrib/restricted/abseil-cpp to 20230125.0
Diffstat (limited to 'contrib/restricted/abseil-cpp/absl/strings')
55 files changed, 2728 insertions, 1461 deletions
diff --git a/contrib/restricted/abseil-cpp/absl/strings/CMakeLists.darwin.txt b/contrib/restricted/abseil-cpp/absl/strings/CMakeLists.darwin.txt index 574ead4171..e3aa400b95 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/CMakeLists.darwin.txt +++ b/contrib/restricted/abseil-cpp/absl/strings/CMakeLists.darwin.txt @@ -20,6 +20,14 @@ target_link_libraries(abseil-cpp-absl-strings PUBLIC abseil-cpp-absl-numeric ) target_sources(abseil-cpp-absl-strings PRIVATE + ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/crc/crc32c.cc + ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/crc/internal/cpu_detect.cc + ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/crc/internal/crc.cc + ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/crc/internal/crc_cord_state.cc + ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/crc/internal/crc_memcpy_fallback.cc + ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/crc/internal/crc_memcpy_x86_64.cc + ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/crc/internal/crc_non_temporal_memcpy.cc + ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/crc/internal/crc_x86_arm_combined.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/status/statusor.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/ascii.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/charconv.cc @@ -40,6 +48,7 @@ target_sources(abseil-cpp-absl-strings PRIVATE ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/cordz_handle.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/cordz_info.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/cordz_sample_token.cc + ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/damerau_levenshtein_distance.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/escaping.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/memutil.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/ostringstream.cc @@ -49,6 +58,7 @@ target_sources(abseil-cpp-absl-strings PRIVATE ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/float_conversion.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/output.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/parser.cc + ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/stringify_sink.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/utf8.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/match.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/numbers.cc diff --git a/contrib/restricted/abseil-cpp/absl/strings/CMakeLists.linux-aarch64.txt b/contrib/restricted/abseil-cpp/absl/strings/CMakeLists.linux-aarch64.txt index 142fd5fd9c..2da01ab774 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/CMakeLists.linux-aarch64.txt +++ b/contrib/restricted/abseil-cpp/absl/strings/CMakeLists.linux-aarch64.txt @@ -21,6 +21,14 @@ target_link_libraries(abseil-cpp-absl-strings PUBLIC abseil-cpp-absl-numeric ) target_sources(abseil-cpp-absl-strings PRIVATE + ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/crc/crc32c.cc + ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/crc/internal/cpu_detect.cc + ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/crc/internal/crc.cc + ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/crc/internal/crc_cord_state.cc + ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/crc/internal/crc_memcpy_fallback.cc + ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/crc/internal/crc_memcpy_x86_64.cc + ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/crc/internal/crc_non_temporal_memcpy.cc + ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/crc/internal/crc_x86_arm_combined.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/status/statusor.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/ascii.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/charconv.cc @@ -41,6 +49,7 @@ target_sources(abseil-cpp-absl-strings PRIVATE ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/cordz_handle.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/cordz_info.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/cordz_sample_token.cc + ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/damerau_levenshtein_distance.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/escaping.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/memutil.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/ostringstream.cc @@ -50,6 +59,7 @@ target_sources(abseil-cpp-absl-strings PRIVATE ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/float_conversion.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/output.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/parser.cc + ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/stringify_sink.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/utf8.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/match.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/numbers.cc diff --git a/contrib/restricted/abseil-cpp/absl/strings/CMakeLists.linux.txt b/contrib/restricted/abseil-cpp/absl/strings/CMakeLists.linux.txt index 142fd5fd9c..2da01ab774 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/CMakeLists.linux.txt +++ b/contrib/restricted/abseil-cpp/absl/strings/CMakeLists.linux.txt @@ -21,6 +21,14 @@ target_link_libraries(abseil-cpp-absl-strings PUBLIC abseil-cpp-absl-numeric ) target_sources(abseil-cpp-absl-strings PRIVATE + ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/crc/crc32c.cc + ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/crc/internal/cpu_detect.cc + ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/crc/internal/crc.cc + ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/crc/internal/crc_cord_state.cc + ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/crc/internal/crc_memcpy_fallback.cc + ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/crc/internal/crc_memcpy_x86_64.cc + ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/crc/internal/crc_non_temporal_memcpy.cc + ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/crc/internal/crc_x86_arm_combined.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/status/statusor.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/ascii.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/charconv.cc @@ -41,6 +49,7 @@ target_sources(abseil-cpp-absl-strings PRIVATE ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/cordz_handle.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/cordz_info.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/cordz_sample_token.cc + ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/damerau_levenshtein_distance.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/escaping.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/memutil.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/ostringstream.cc @@ -50,6 +59,7 @@ target_sources(abseil-cpp-absl-strings PRIVATE ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/float_conversion.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/output.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/parser.cc + ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/stringify_sink.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/internal/utf8.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/match.cc ${CMAKE_SOURCE_DIR}/contrib/restricted/abseil-cpp/absl/strings/numbers.cc diff --git a/contrib/restricted/abseil-cpp/absl/strings/ascii.cc b/contrib/restricted/abseil-cpp/absl/strings/ascii.cc index 93bb03e958..868df2d102 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/ascii.cc +++ b/contrib/restricted/abseil-cpp/absl/strings/ascii.cc @@ -157,13 +157,13 @@ ABSL_DLL const char kToUpper[256] = { void AsciiStrToLower(std::string* s) { for (auto& ch : *s) { - ch = absl::ascii_tolower(ch); + ch = absl::ascii_tolower(static_cast<unsigned char>(ch)); } } void AsciiStrToUpper(std::string* s) { for (auto& ch : *s) { - ch = absl::ascii_toupper(ch); + ch = absl::ascii_toupper(static_cast<unsigned char>(ch)); } } @@ -183,17 +183,17 @@ void RemoveExtraAsciiWhitespace(std::string* str) { for (; input_it < input_end; ++input_it) { if (is_ws) { // Consecutive whitespace? Keep only the last. - is_ws = absl::ascii_isspace(*input_it); + is_ws = absl::ascii_isspace(static_cast<unsigned char>(*input_it)); if (is_ws) --output_it; } else { - is_ws = absl::ascii_isspace(*input_it); + is_ws = absl::ascii_isspace(static_cast<unsigned char>(*input_it)); } *output_it = *input_it; ++output_it; } - str->erase(output_it - &(*str)[0]); + str->erase(static_cast<size_t>(output_it - &(*str)[0])); } ABSL_NAMESPACE_END diff --git a/contrib/restricted/abseil-cpp/absl/strings/charconv.cc b/contrib/restricted/abseil-cpp/absl/strings/charconv.cc index fefcfc90a5..69d420bcea 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/charconv.cc +++ b/contrib/restricted/abseil-cpp/absl/strings/charconv.cc @@ -18,6 +18,7 @@ #include <cassert> #include <cmath> #include <cstring> +#include <limits> #include "absl/base/casts.h" #include "absl/numeric/bits.h" @@ -65,6 +66,14 @@ struct FloatTraits; template <> struct FloatTraits<double> { + using mantissa_t = uint64_t; + + // The number of bits in the given float type. + static constexpr int kTargetBits = 64; + + // The number of exponent bits in the given float type. + static constexpr int kTargetExponentBits = 11; + // The number of mantissa bits in the given float type. This includes the // implied high bit. static constexpr int kTargetMantissaBits = 53; @@ -83,6 +92,31 @@ struct FloatTraits<double> { // m * 2**kMinNormalExponent is exactly equal to DBL_MIN. static constexpr int kMinNormalExponent = -1074; + // The IEEE exponent bias. It equals ((1 << (kTargetExponentBits - 1)) - 1). + static constexpr int kExponentBias = 1023; + + // The Eisel-Lemire "Shifting to 54/25 Bits" adjustment. It equals (63 - 1 - + // kTargetMantissaBits). + static constexpr int kEiselLemireShift = 9; + + // The Eisel-Lemire high64_mask. It equals ((1 << kEiselLemireShift) - 1). + static constexpr uint64_t kEiselLemireMask = uint64_t{0x1FF}; + + // The smallest negative integer N (smallest negative means furthest from + // zero) such that parsing 9999999999999999999eN, with 19 nines, is still + // positive. Parsing a smaller (more negative) N will produce zero. + // + // Adjusting the decimal point and exponent, without adjusting the value, + // 9999999999999999999eN equals 9.999999999999999999eM where M = N + 18. + // + // 9999999999999999999, with 19 nines but no decimal point, is the largest + // "repeated nines" integer that fits in a uint64_t. + static constexpr int kEiselLemireMinInclusiveExp10 = -324 - 18; + + // The smallest positive integer N such that parsing 1eN produces infinity. + // Parsing a smaller N will produce something finite. + static constexpr int kEiselLemireMaxExclusiveExp10 = 309; + static double MakeNan(const char* tagp) { // Support nan no matter which namespace it's in. Some platforms // incorrectly don't put it in namespace std. @@ -103,7 +137,7 @@ struct FloatTraits<double> { // a normal value is made, or it must be less narrow than that, in which case // `exponent` must be exactly kMinNormalExponent, and a subnormal value is // made. - static double Make(uint64_t mantissa, int exponent, bool sign) { + static double Make(mantissa_t mantissa, int exponent, bool sign) { #ifndef ABSL_BIT_PACK_FLOATS // Support ldexp no matter which namespace it's in. Some platforms // incorrectly don't put it in namespace std. @@ -116,8 +150,10 @@ struct FloatTraits<double> { if (mantissa > kMantissaMask) { // Normal value. // Adjust by 1023 for the exponent representation bias, and an additional - // 52 due to the implied decimal point in the IEEE mantissa represenation. - dbl += uint64_t{exponent + 1023u + kTargetMantissaBits - 1} << 52; + // 52 due to the implied decimal point in the IEEE mantissa + // representation. + dbl += static_cast<uint64_t>(exponent + 1023 + kTargetMantissaBits - 1) + << 52; mantissa &= kMantissaMask; } else { // subnormal value @@ -134,16 +170,27 @@ struct FloatTraits<double> { // members and methods. template <> struct FloatTraits<float> { + using mantissa_t = uint32_t; + + static constexpr int kTargetBits = 32; + static constexpr int kTargetExponentBits = 8; static constexpr int kTargetMantissaBits = 24; static constexpr int kMaxExponent = 104; static constexpr int kMinNormalExponent = -149; + static constexpr int kExponentBias = 127; + static constexpr int kEiselLemireShift = 38; + static constexpr uint64_t kEiselLemireMask = uint64_t{0x3FFFFFFFFF}; + static constexpr int kEiselLemireMinInclusiveExp10 = -46 - 18; + static constexpr int kEiselLemireMaxExclusiveExp10 = 39; + static float MakeNan(const char* tagp) { // Support nanf no matter which namespace it's in. Some platforms // incorrectly don't put it in namespace std. using namespace std; // NOLINT return nanf(tagp); } - static float Make(uint32_t mantissa, int exponent, bool sign) { + + static float Make(mantissa_t mantissa, int exponent, bool sign) { #ifndef ABSL_BIT_PACK_FLOATS // Support ldexpf no matter which namespace it's in. Some platforms // incorrectly don't put it in namespace std. @@ -157,7 +204,8 @@ struct FloatTraits<float> { // Normal value. // Adjust by 127 for the exponent representation bias, and an additional // 23 due to the implied decimal point in the IEEE mantissa represenation. - flt += uint32_t{exponent + 127u + kTargetMantissaBits - 1} << 23; + flt += static_cast<uint32_t>(exponent + 127 + kTargetMantissaBits - 1) + << 23; mantissa &= kMantissaMask; } else { // subnormal value @@ -181,39 +229,45 @@ struct FloatTraits<float> { // // 2**63 <= Power10Mantissa(n) < 2**64. // +// See the "Table of powers of 10" comment below for a "1e60" example. +// // Lookups into the power-of-10 table must first check the Power10Overflow() and // Power10Underflow() functions, to avoid out-of-bounds table access. // -// Indexes into these tables are biased by -kPower10TableMin, and the table has -// values in the range [kPower10TableMin, kPower10TableMax]. -extern const uint64_t kPower10MantissaTable[]; -extern const int16_t kPower10ExponentTable[]; +// Indexes into these tables are biased by -kPower10TableMinInclusive. Valid +// indexes range from kPower10TableMinInclusive to kPower10TableMaxExclusive. +extern const uint64_t kPower10MantissaHighTable[]; // High 64 of 128 bits. +extern const uint64_t kPower10MantissaLowTable[]; // Low 64 of 128 bits. -// The smallest allowed value for use with the Power10Mantissa() and -// Power10Exponent() functions below. (If a smaller exponent is needed in +// The smallest (inclusive) allowed value for use with the Power10Mantissa() +// and Power10Exponent() functions below. (If a smaller exponent is needed in // calculations, the end result is guaranteed to underflow.) -constexpr int kPower10TableMin = -342; +constexpr int kPower10TableMinInclusive = -342; -// The largest allowed value for use with the Power10Mantissa() and -// Power10Exponent() functions below. (If a smaller exponent is needed in -// calculations, the end result is guaranteed to overflow.) -constexpr int kPower10TableMax = 308; +// The largest (exclusive) allowed value for use with the Power10Mantissa() and +// Power10Exponent() functions below. (If a larger-or-equal exponent is needed +// in calculations, the end result is guaranteed to overflow.) +constexpr int kPower10TableMaxExclusive = 309; uint64_t Power10Mantissa(int n) { - return kPower10MantissaTable[n - kPower10TableMin]; + return kPower10MantissaHighTable[n - kPower10TableMinInclusive]; } int Power10Exponent(int n) { - return kPower10ExponentTable[n - kPower10TableMin]; + // The 217706 etc magic numbers encode the results as a formula instead of a + // table. Their equivalence (over the kPower10TableMinInclusive .. + // kPower10TableMaxExclusive range) is confirmed by + // https://github.com/google/wuffs/blob/315b2e52625ebd7b02d8fac13e3cd85ea374fb80/script/print-mpb-powers-of-10.go + return (217706 * n >> 16) - 63; } // Returns true if n is large enough that 10**n always results in an IEEE // overflow. -bool Power10Overflow(int n) { return n > kPower10TableMax; } +bool Power10Overflow(int n) { return n >= kPower10TableMaxExclusive; } // Returns true if n is small enough that 10**n times a ParsedFloat mantissa // always results in an IEEE underflow. -bool Power10Underflow(int n) { return n < kPower10TableMin; } +bool Power10Underflow(int n) { return n < kPower10TableMinInclusive; } // Returns true if Power10Mantissa(n) * 2**Power10Exponent(n) is exactly equal // to 10**n numerically. Put another way, this returns true if there is no @@ -242,9 +296,11 @@ struct CalculatedFloat { // Returns the bit width of the given uint128. (Equivalently, returns 128 // minus the number of leading zero bits.) -unsigned BitWidth(uint128 value) { +int BitWidth(uint128 value) { if (Uint128High64(value) == 0) { - return static_cast<unsigned>(bit_width(Uint128Low64(value))); + // This static_cast is only needed when using a std::bit_width() + // implementation that does not have the fix for LWG 3656 applied. + return static_cast<int>(bit_width(Uint128Low64(value))); } return 128 - countl_zero(Uint128High64(value)); } @@ -285,14 +341,19 @@ template <typename FloatType> bool HandleEdgeCase(const strings_internal::ParsedFloat& input, bool negative, FloatType* value) { if (input.type == strings_internal::FloatType::kNan) { - // A bug in both clang and gcc would cause the compiler to optimize away the - // buffer we are building below. Declaring the buffer volatile avoids the - // issue, and has no measurable performance impact in microbenchmarks. + // A bug in both clang < 7 and gcc would cause the compiler to optimize + // away the buffer we are building below. Declaring the buffer volatile + // avoids the issue, and has no measurable performance impact in + // microbenchmarks. // // https://bugs.llvm.org/show_bug.cgi?id=37778 // https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86113 constexpr ptrdiff_t kNanBufferSize = 128; +#if defined(__GNUC__) || (defined(__clang__) && __clang_major__ < 7) volatile char n_char_sequence[kNanBufferSize]; +#else + char n_char_sequence[kNanBufferSize]; +#endif if (input.subrange_begin == nullptr) { n_char_sequence[0] = '\0'; } else { @@ -337,8 +398,10 @@ void EncodeResult(const CalculatedFloat& calculated, bool negative, *value = negative ? -0.0 : 0.0; return; } - *value = FloatTraits<FloatType>::Make(calculated.mantissa, - calculated.exponent, negative); + *value = FloatTraits<FloatType>::Make( + static_cast<typename FloatTraits<FloatType>::mantissa_t>( + calculated.mantissa), + calculated.exponent, negative); } // Returns the given uint128 shifted to the right by `shift` bits, and rounds @@ -519,7 +582,9 @@ CalculatedFloat CalculateFromParsedHexadecimal( const strings_internal::ParsedFloat& parsed_hex) { uint64_t mantissa = parsed_hex.mantissa; int exponent = parsed_hex.exponent; - auto mantissa_width = static_cast<unsigned>(bit_width(mantissa)); + // This static_cast is only needed when using a std::bit_width() + // implementation that does not have the fix for LWG 3656 applied. + int mantissa_width = static_cast<int>(bit_width(mantissa)); const int shift = NormalizedShiftSize<FloatType>(mantissa_width, exponent); bool result_exact; exponent += shift; @@ -595,6 +660,185 @@ CalculatedFloat CalculateFromParsedDecimal( binary_exponent); } +// As discussed in https://nigeltao.github.io/blog/2020/eisel-lemire.html the +// primary goal of the Eisel-Lemire algorithm is speed, for 99+% of the cases, +// not 100% coverage. As long as Eisel-Lemire doesn’t claim false positives, +// the combined approach (falling back to an alternative implementation when +// this function returns false) is both fast and correct. +template <typename FloatType> +bool EiselLemire(const strings_internal::ParsedFloat& input, bool negative, + FloatType* value, std::errc* ec) { + uint64_t man = input.mantissa; + int exp10 = input.exponent; + if (exp10 < FloatTraits<FloatType>::kEiselLemireMinInclusiveExp10) { + *value = negative ? -0.0 : 0.0; + *ec = std::errc::result_out_of_range; + return true; + } else if (exp10 >= FloatTraits<FloatType>::kEiselLemireMaxExclusiveExp10) { + // Return max (a finite value) consistent with from_chars and DR 3081. For + // SimpleAtod and SimpleAtof, post-processing will return infinity. + *value = negative ? -std::numeric_limits<FloatType>::max() + : std::numeric_limits<FloatType>::max(); + *ec = std::errc::result_out_of_range; + return true; + } + + // Assert kPower10TableMinInclusive <= exp10 < kPower10TableMaxExclusive. + // Equivalently, !Power10Underflow(exp10) and !Power10Overflow(exp10). + static_assert( + FloatTraits<FloatType>::kEiselLemireMinInclusiveExp10 >= + kPower10TableMinInclusive, + "(exp10-kPower10TableMinInclusive) in kPower10MantissaHighTable bounds"); + static_assert( + FloatTraits<FloatType>::kEiselLemireMaxExclusiveExp10 <= + kPower10TableMaxExclusive, + "(exp10-kPower10TableMinInclusive) in kPower10MantissaHighTable bounds"); + + // The terse (+) comments in this function body refer to sections of the + // https://nigeltao.github.io/blog/2020/eisel-lemire.html blog post. + // + // That blog post discusses double precision (11 exponent bits with a -1023 + // bias, 52 mantissa bits), but the same approach applies to single precision + // (8 exponent bits with a -127 bias, 23 mantissa bits). Either way, the + // computation here happens with 64-bit values (e.g. man) or 128-bit values + // (e.g. x) before finally converting to 64- or 32-bit floating point. + // + // See also "Number Parsing at a Gigabyte per Second, Software: Practice and + // Experience 51 (8), 2021" (https://arxiv.org/abs/2101.11408) for detail. + + // (+) Normalization. + int clz = countl_zero(man); + man <<= static_cast<unsigned int>(clz); + // The 217706 etc magic numbers are from the Power10Exponent function. + uint64_t ret_exp2 = + static_cast<uint64_t>((217706 * exp10 >> 16) + 64 + + FloatTraits<FloatType>::kExponentBias - clz); + + // (+) Multiplication. + uint128 x = static_cast<uint128>(man) * + static_cast<uint128>( + kPower10MantissaHighTable[exp10 - kPower10TableMinInclusive]); + + // (+) Wider Approximation. + static constexpr uint64_t high64_mask = + FloatTraits<FloatType>::kEiselLemireMask; + if (((Uint128High64(x) & high64_mask) == high64_mask) && + (man > (std::numeric_limits<uint64_t>::max() - Uint128Low64(x)))) { + uint128 y = + static_cast<uint128>(man) * + static_cast<uint128>( + kPower10MantissaLowTable[exp10 - kPower10TableMinInclusive]); + x += Uint128High64(y); + // For example, parsing "4503599627370497.5" will take the if-true + // branch here (for double precision), since: + // - x = 0x8000000000000BFF_FFFFFFFFFFFFFFFF + // - y = 0x8000000000000BFF_7FFFFFFFFFFFF400 + // - man = 0xA000000000000F00 + // Likewise, when parsing "0.0625" for single precision: + // - x = 0x7FFFFFFFFFFFFFFF_FFFFFFFFFFFFFFFF + // - y = 0x813FFFFFFFFFFFFF_8A00000000000000 + // - man = 0x9C40000000000000 + if (((Uint128High64(x) & high64_mask) == high64_mask) && + ((Uint128Low64(x) + 1) == 0) && + (man > (std::numeric_limits<uint64_t>::max() - Uint128Low64(y)))) { + return false; + } + } + + // (+) Shifting to 54 Bits (or for single precision, to 25 bits). + uint64_t msb = Uint128High64(x) >> 63; + uint64_t ret_man = + Uint128High64(x) >> (msb + FloatTraits<FloatType>::kEiselLemireShift); + ret_exp2 -= 1 ^ msb; + + // (+) Half-way Ambiguity. + // + // For example, parsing "1e+23" will take the if-true branch here (for double + // precision), since: + // - x = 0x54B40B1F852BDA00_0000000000000000 + // - ret_man = 0x002A5A058FC295ED + // Likewise, when parsing "20040229.0" for single precision: + // - x = 0x4C72894000000000_0000000000000000 + // - ret_man = 0x000000000131CA25 + if ((Uint128Low64(x) == 0) && ((Uint128High64(x) & high64_mask) == 0) && + ((ret_man & 3) == 1)) { + return false; + } + + // (+) From 54 to 53 Bits (or for single precision, from 25 to 24 bits). + ret_man += ret_man & 1; // Line From54a. + ret_man >>= 1; // Line From54b. + // Incrementing ret_man (at line From54a) may have overflowed 54 bits (53 + // bits after the right shift by 1 at line From54b), so adjust for that. + // + // For example, parsing "9223372036854775807" will take the if-true branch + // here (for double precision), since: + // - ret_man = 0x0020000000000000 = (1 << 53) + // Likewise, when parsing "2147483647.0" for single precision: + // - ret_man = 0x0000000001000000 = (1 << 24) + if ((ret_man >> FloatTraits<FloatType>::kTargetMantissaBits) > 0) { + ret_exp2 += 1; + // Conceptually, we need a "ret_man >>= 1" in this if-block to balance + // incrementing ret_exp2 in the line immediately above. However, we only + // get here when line From54a overflowed (after adding a 1), so ret_man + // here is (1 << 53). Its low 53 bits are therefore all zeroes. The only + // remaining use of ret_man is to mask it with ((1 << 52) - 1), so only its + // low 52 bits matter. A "ret_man >>= 1" would have no effect in practice. + // + // We omit the "ret_man >>= 1", even if it is cheap (and this if-branch is + // rarely taken) and technically 'more correct', so that mutation tests + // that would otherwise modify or omit that "ret_man >>= 1" don't complain + // that such code mutations have no observable effect. + } + + // ret_exp2 is a uint64_t. Zero or underflow means that we're in subnormal + // space. max_exp2 (0x7FF for double precision, 0xFF for single precision) or + // above means that we're in Inf/NaN space. + // + // The if block is equivalent to (but has fewer branches than): + // if ((ret_exp2 <= 0) || (ret_exp2 >= max_exp2)) { etc } + // + // For example, parsing "4.9406564584124654e-324" will take the if-true + // branch here, since ret_exp2 = -51. + static constexpr uint64_t max_exp2 = + (1 << FloatTraits<FloatType>::kTargetExponentBits) - 1; + if ((ret_exp2 - 1) >= (max_exp2 - 1)) { + return false; + } + +#ifndef ABSL_BIT_PACK_FLOATS + if (FloatTraits<FloatType>::kTargetBits == 64) { + *value = FloatTraits<FloatType>::Make( + (ret_man & 0x000FFFFFFFFFFFFFu) | 0x0010000000000000u, + static_cast<int>(ret_exp2) - 1023 - 52, negative); + return true; + } else if (FloatTraits<FloatType>::kTargetBits == 32) { + *value = FloatTraits<FloatType>::Make( + (static_cast<uint32_t>(ret_man) & 0x007FFFFFu) | 0x00800000u, + static_cast<int>(ret_exp2) - 127 - 23, negative); + return true; + } +#else + if (FloatTraits<FloatType>::kTargetBits == 64) { + uint64_t ret_bits = (ret_exp2 << 52) | (ret_man & 0x000FFFFFFFFFFFFFu); + if (negative) { + ret_bits |= 0x8000000000000000u; + } + *value = absl::bit_cast<double>(ret_bits); + return true; + } else if (FloatTraits<FloatType>::kTargetBits == 32) { + uint32_t ret_bits = (static_cast<uint32_t>(ret_exp2) << 23) | + (static_cast<uint32_t>(ret_man) & 0x007FFFFFu); + if (negative) { + ret_bits |= 0x80000000u; + } + *value = absl::bit_cast<float>(ret_bits); + return true; + } +#endif // ABSL_BIT_PACK_FLOATS + return false; +} + template <typename FloatType> from_chars_result FromCharsImpl(const char* first, const char* last, FloatType& value, chars_format fmt_flags) { @@ -668,6 +912,12 @@ from_chars_result FromCharsImpl(const char* first, const char* last, if (HandleEdgeCase(decimal_parse, negative, &value)) { return result; } + // A nullptr subrange_begin means that the decimal_parse.mantissa is exact + // (not truncated), a precondition of the Eisel-Lemire algorithm. + if ((decimal_parse.subrange_begin == nullptr) && + EiselLemire<FloatType>(decimal_parse, negative, &value, &result.ec)) { + return result; + } CalculatedFloat calculated = CalculateFromParsedDecimal<FloatType>(decimal_parse); EncodeResult(calculated, negative, &result, &value); @@ -688,15 +938,46 @@ from_chars_result from_chars(const char* first, const char* last, float& value, namespace { -// Table of powers of 10, from kPower10TableMin to kPower10TableMax. +// Table of powers of 10, from kPower10TableMinInclusive to +// kPower10TableMaxExclusive. +// +// kPower10MantissaHighTable[i - kPower10TableMinInclusive] stores the 64-bit +// mantissa. The high bit is always on. +// +// kPower10MantissaLowTable extends that 64-bit mantissa to 128 bits. +// +// Power10Exponent(i) calculates the power-of-two exponent. +// +// For a number i, this gives the unique mantissaHigh and exponent such that +// (mantissaHigh * 2**exponent) <= 10**i < ((mantissaHigh + 1) * 2**exponent). +// +// For example, Python can confirm that the exact hexadecimal value of 1e60 is: +// >>> a = 1000000000000000000000000000000000000000000000000000000000000 +// >>> hex(a) +// '0x9f4f2726179a224501d762422c946590d91000000000000000' +// Adding underscores at every 8th hex digit shows 50 hex digits: +// '0x9f4f2726_179a2245_01d76242_2c946590_d9100000_00000000_00'. +// In this case, the high bit of the first hex digit, 9, is coincidentally set, +// so we do not have to do further shifting to deduce the 128-bit mantissa: +// - kPower10MantissaHighTable[60 - kP10TMI] = 0x9f4f2726179a2245U +// - kPower10MantissaLowTable[ 60 - kP10TMI] = 0x01d762422c946590U +// where kP10TMI is kPower10TableMinInclusive. The low 18 of those 50 hex +// digits are truncated. +// +// 50 hex digits (with the high bit set) is 200 bits and mantissaHigh holds 64 +// bits, so Power10Exponent(60) = 200 - 64 = 136. Again, Python can confirm: +// >>> b = 0x9f4f2726179a2245 +// >>> ((b+0)<<136) <= a +// True +// >>> ((b+1)<<136) <= a +// False // -// kPower10MantissaTable[i - kPower10TableMin] stores the 64-bit mantissa (high -// bit always on), and kPower10ExponentTable[i - kPower10TableMin] stores the -// power-of-two exponent. For a given number i, this gives the unique mantissa -// and exponent such that mantissa * 2**exponent <= 10**i < (mantissa + 1) * -// 2**exponent. +// The tables were generated by +// https://github.com/google/wuffs/blob/315b2e52625ebd7b02d8fac13e3cd85ea374fb80/script/print-mpb-powers-of-10.go +// after re-formatting its output into two arrays of N uint64_t values (instead +// of an N element array of uint64_t pairs). -const uint64_t kPower10MantissaTable[] = { +const uint64_t kPower10MantissaHighTable[] = { 0xeef453d6923bd65aU, 0x9558b4661b6565f8U, 0xbaaee17fa23ebf76U, 0xe95a99df8ace6f53U, 0x91d8a02bb6c10594U, 0xb64ec836a47146f9U, 0xe3e27a444d8d98b7U, 0x8e6d8c6ab0787f72U, 0xb208ef855c969f4fU, @@ -916,67 +1197,224 @@ const uint64_t kPower10MantissaTable[] = { 0xb6472e511c81471dU, 0xe3d8f9e563a198e5U, 0x8e679c2f5e44ff8fU, }; -const int16_t kPower10ExponentTable[] = { - -1200, -1196, -1193, -1190, -1186, -1183, -1180, -1176, -1173, -1170, -1166, - -1163, -1160, -1156, -1153, -1150, -1146, -1143, -1140, -1136, -1133, -1130, - -1127, -1123, -1120, -1117, -1113, -1110, -1107, -1103, -1100, -1097, -1093, - -1090, -1087, -1083, -1080, -1077, -1073, -1070, -1067, -1063, -1060, -1057, - -1053, -1050, -1047, -1043, -1040, -1037, -1034, -1030, -1027, -1024, -1020, - -1017, -1014, -1010, -1007, -1004, -1000, -997, -994, -990, -987, -984, - -980, -977, -974, -970, -967, -964, -960, -957, -954, -950, -947, - -944, -940, -937, -934, -931, -927, -924, -921, -917, -914, -911, - -907, -904, -901, -897, -894, -891, -887, -884, -881, -877, -874, - -871, -867, -864, -861, -857, -854, -851, -847, -844, -841, -838, - -834, -831, -828, -824, -821, -818, -814, -811, -808, -804, -801, - -798, -794, -791, -788, -784, -781, -778, -774, -771, -768, -764, - -761, -758, -754, -751, -748, -744, -741, -738, -735, -731, -728, - -725, -721, -718, -715, -711, -708, -705, -701, -698, -695, -691, - -688, -685, -681, -678, -675, -671, -668, -665, -661, -658, -655, - -651, -648, -645, -642, -638, -635, -632, -628, -625, -622, -618, - -615, -612, -608, -605, -602, -598, -595, -592, -588, -585, -582, - -578, -575, -572, -568, -565, -562, -558, -555, -552, -549, -545, - -542, -539, -535, -532, -529, -525, -522, -519, -515, -512, -509, - -505, -502, -499, -495, -492, -489, -485, -482, -479, -475, -472, - -469, -465, -462, -459, -455, -452, -449, -446, -442, -439, -436, - -432, -429, -426, -422, -419, -416, -412, -409, -406, -402, -399, - -396, -392, -389, -386, -382, -379, -376, -372, -369, -366, -362, - -359, -356, -353, -349, -346, -343, -339, -336, -333, -329, -326, - -323, -319, -316, -313, -309, -306, -303, -299, -296, -293, -289, - -286, -283, -279, -276, -273, -269, -266, -263, -259, -256, -253, - -250, -246, -243, -240, -236, -233, -230, -226, -223, -220, -216, - -213, -210, -206, -203, -200, -196, -193, -190, -186, -183, -180, - -176, -173, -170, -166, -163, -160, -157, -153, -150, -147, -143, - -140, -137, -133, -130, -127, -123, -120, -117, -113, -110, -107, - -103, -100, -97, -93, -90, -87, -83, -80, -77, -73, -70, - -67, -63, -60, -57, -54, -50, -47, -44, -40, -37, -34, - -30, -27, -24, -20, -17, -14, -10, -7, -4, 0, 3, - 6, 10, 13, 16, 20, 23, 26, 30, 33, 36, 39, - 43, 46, 49, 53, 56, 59, 63, 66, 69, 73, 76, - 79, 83, 86, 89, 93, 96, 99, 103, 106, 109, 113, - 116, 119, 123, 126, 129, 132, 136, 139, 142, 146, 149, - 152, 156, 159, 162, 166, 169, 172, 176, 179, 182, 186, - 189, 192, 196, 199, 202, 206, 209, 212, 216, 219, 222, - 226, 229, 232, 235, 239, 242, 245, 249, 252, 255, 259, - 262, 265, 269, 272, 275, 279, 282, 285, 289, 292, 295, - 299, 302, 305, 309, 312, 315, 319, 322, 325, 328, 332, - 335, 338, 342, 345, 348, 352, 355, 358, 362, 365, 368, - 372, 375, 378, 382, 385, 388, 392, 395, 398, 402, 405, - 408, 412, 415, 418, 422, 425, 428, 431, 435, 438, 441, - 445, 448, 451, 455, 458, 461, 465, 468, 471, 475, 478, - 481, 485, 488, 491, 495, 498, 501, 505, 508, 511, 515, - 518, 521, 524, 528, 531, 534, 538, 541, 544, 548, 551, - 554, 558, 561, 564, 568, 571, 574, 578, 581, 584, 588, - 591, 594, 598, 601, 604, 608, 611, 614, 617, 621, 624, - 627, 631, 634, 637, 641, 644, 647, 651, 654, 657, 661, - 664, 667, 671, 674, 677, 681, 684, 687, 691, 694, 697, - 701, 704, 707, 711, 714, 717, 720, 724, 727, 730, 734, - 737, 740, 744, 747, 750, 754, 757, 760, 764, 767, 770, - 774, 777, 780, 784, 787, 790, 794, 797, 800, 804, 807, - 810, 813, 817, 820, 823, 827, 830, 833, 837, 840, 843, - 847, 850, 853, 857, 860, 863, 867, 870, 873, 877, 880, - 883, 887, 890, 893, 897, 900, 903, 907, 910, 913, 916, - 920, 923, 926, 930, 933, 936, 940, 943, 946, 950, 953, - 956, 960, +const uint64_t kPower10MantissaLowTable[] = { + 0x113faa2906a13b3fU, 0x4ac7ca59a424c507U, 0x5d79bcf00d2df649U, + 0xf4d82c2c107973dcU, 0x79071b9b8a4be869U, 0x9748e2826cdee284U, + 0xfd1b1b2308169b25U, 0xfe30f0f5e50e20f7U, 0xbdbd2d335e51a935U, + 0xad2c788035e61382U, 0x4c3bcb5021afcc31U, 0xdf4abe242a1bbf3dU, + 0xd71d6dad34a2af0dU, 0x8672648c40e5ad68U, 0x680efdaf511f18c2U, + 0x0212bd1b2566def2U, 0x014bb630f7604b57U, 0x419ea3bd35385e2dU, + 0x52064cac828675b9U, 0x7343efebd1940993U, 0x1014ebe6c5f90bf8U, + 0xd41a26e077774ef6U, 0x8920b098955522b4U, 0x55b46e5f5d5535b0U, + 0xeb2189f734aa831dU, 0xa5e9ec7501d523e4U, 0x47b233c92125366eU, + 0x999ec0bb696e840aU, 0xc00670ea43ca250dU, 0x380406926a5e5728U, + 0xc605083704f5ecf2U, 0xf7864a44c633682eU, 0x7ab3ee6afbe0211dU, + 0x5960ea05bad82964U, 0x6fb92487298e33bdU, 0xa5d3b6d479f8e056U, + 0x8f48a4899877186cU, 0x331acdabfe94de87U, 0x9ff0c08b7f1d0b14U, + 0x07ecf0ae5ee44dd9U, 0xc9e82cd9f69d6150U, 0xbe311c083a225cd2U, + 0x6dbd630a48aaf406U, 0x092cbbccdad5b108U, 0x25bbf56008c58ea5U, + 0xaf2af2b80af6f24eU, 0x1af5af660db4aee1U, 0x50d98d9fc890ed4dU, + 0xe50ff107bab528a0U, 0x1e53ed49a96272c8U, 0x25e8e89c13bb0f7aU, + 0x77b191618c54e9acU, 0xd59df5b9ef6a2417U, 0x4b0573286b44ad1dU, + 0x4ee367f9430aec32U, 0x229c41f793cda73fU, 0x6b43527578c1110fU, + 0x830a13896b78aaa9U, 0x23cc986bc656d553U, 0x2cbfbe86b7ec8aa8U, + 0x7bf7d71432f3d6a9U, 0xdaf5ccd93fb0cc53U, 0xd1b3400f8f9cff68U, + 0x23100809b9c21fa1U, 0xabd40a0c2832a78aU, 0x16c90c8f323f516cU, + 0xae3da7d97f6792e3U, 0x99cd11cfdf41779cU, 0x40405643d711d583U, + 0x482835ea666b2572U, 0xda3243650005eecfU, 0x90bed43e40076a82U, + 0x5a7744a6e804a291U, 0x711515d0a205cb36U, 0x0d5a5b44ca873e03U, + 0xe858790afe9486c2U, 0x626e974dbe39a872U, 0xfb0a3d212dc8128fU, + 0x7ce66634bc9d0b99U, 0x1c1fffc1ebc44e80U, 0xa327ffb266b56220U, + 0x4bf1ff9f0062baa8U, 0x6f773fc3603db4a9U, 0xcb550fb4384d21d3U, + 0x7e2a53a146606a48U, 0x2eda7444cbfc426dU, 0xfa911155fefb5308U, + 0x793555ab7eba27caU, 0x4bc1558b2f3458deU, 0x9eb1aaedfb016f16U, + 0x465e15a979c1cadcU, 0x0bfacd89ec191ec9U, 0xcef980ec671f667bU, + 0x82b7e12780e7401aU, 0xd1b2ecb8b0908810U, 0x861fa7e6dcb4aa15U, + 0x67a791e093e1d49aU, 0xe0c8bb2c5c6d24e0U, 0x58fae9f773886e18U, + 0xaf39a475506a899eU, 0x6d8406c952429603U, 0xc8e5087ba6d33b83U, + 0xfb1e4a9a90880a64U, 0x5cf2eea09a55067fU, 0xf42faa48c0ea481eU, + 0xf13b94daf124da26U, 0x76c53d08d6b70858U, 0x54768c4b0c64ca6eU, + 0xa9942f5dcf7dfd09U, 0xd3f93b35435d7c4cU, 0xc47bc5014a1a6dafU, + 0x359ab6419ca1091bU, 0xc30163d203c94b62U, 0x79e0de63425dcf1dU, + 0x985915fc12f542e4U, 0x3e6f5b7b17b2939dU, 0xa705992ceecf9c42U, + 0x50c6ff782a838353U, 0xa4f8bf5635246428U, 0x871b7795e136be99U, + 0x28e2557b59846e3fU, 0x331aeada2fe589cfU, 0x3ff0d2c85def7621U, + 0x0fed077a756b53a9U, 0xd3e8495912c62894U, 0x64712dd7abbbd95cU, + 0xbd8d794d96aacfb3U, 0xecf0d7a0fc5583a0U, 0xf41686c49db57244U, + 0x311c2875c522ced5U, 0x7d633293366b828bU, 0xae5dff9c02033197U, + 0xd9f57f830283fdfcU, 0xd072df63c324fd7bU, 0x4247cb9e59f71e6dU, + 0x52d9be85f074e608U, 0x67902e276c921f8bU, 0x00ba1cd8a3db53b6U, + 0x80e8a40eccd228a4U, 0x6122cd128006b2cdU, 0x796b805720085f81U, + 0xcbe3303674053bb0U, 0xbedbfc4411068a9cU, 0xee92fb5515482d44U, + 0x751bdd152d4d1c4aU, 0xd262d45a78a0635dU, 0x86fb897116c87c34U, + 0xd45d35e6ae3d4da0U, 0x8974836059cca109U, 0x2bd1a438703fc94bU, + 0x7b6306a34627ddcfU, 0x1a3bc84c17b1d542U, 0x20caba5f1d9e4a93U, + 0x547eb47b7282ee9cU, 0xe99e619a4f23aa43U, 0x6405fa00e2ec94d4U, + 0xde83bc408dd3dd04U, 0x9624ab50b148d445U, 0x3badd624dd9b0957U, + 0xe54ca5d70a80e5d6U, 0x5e9fcf4ccd211f4cU, 0x7647c3200069671fU, + 0x29ecd9f40041e073U, 0xf468107100525890U, 0x7182148d4066eeb4U, + 0xc6f14cd848405530U, 0xb8ada00e5a506a7cU, 0xa6d90811f0e4851cU, + 0x908f4a166d1da663U, 0x9a598e4e043287feU, 0x40eff1e1853f29fdU, + 0xd12bee59e68ef47cU, 0x82bb74f8301958ceU, 0xe36a52363c1faf01U, + 0xdc44e6c3cb279ac1U, 0x29ab103a5ef8c0b9U, 0x7415d448f6b6f0e7U, + 0x111b495b3464ad21U, 0xcab10dd900beec34U, 0x3d5d514f40eea742U, + 0x0cb4a5a3112a5112U, 0x47f0e785eaba72abU, 0x59ed216765690f56U, + 0x306869c13ec3532cU, 0x1e414218c73a13fbU, 0xe5d1929ef90898faU, + 0xdf45f746b74abf39U, 0x6b8bba8c328eb783U, 0x066ea92f3f326564U, + 0xc80a537b0efefebdU, 0xbd06742ce95f5f36U, 0x2c48113823b73704U, + 0xf75a15862ca504c5U, 0x9a984d73dbe722fbU, 0xc13e60d0d2e0ebbaU, + 0x318df905079926a8U, 0xfdf17746497f7052U, 0xfeb6ea8bedefa633U, + 0xfe64a52ee96b8fc0U, 0x3dfdce7aa3c673b0U, 0x06bea10ca65c084eU, + 0x486e494fcff30a62U, 0x5a89dba3c3efccfaU, 0xf89629465a75e01cU, + 0xf6bbb397f1135823U, 0x746aa07ded582e2cU, 0xa8c2a44eb4571cdcU, + 0x92f34d62616ce413U, 0x77b020baf9c81d17U, 0x0ace1474dc1d122eU, + 0x0d819992132456baU, 0x10e1fff697ed6c69U, 0xca8d3ffa1ef463c1U, + 0xbd308ff8a6b17cb2U, 0xac7cb3f6d05ddbdeU, 0x6bcdf07a423aa96bU, + 0x86c16c98d2c953c6U, 0xe871c7bf077ba8b7U, 0x11471cd764ad4972U, + 0xd598e40d3dd89bcfU, 0x4aff1d108d4ec2c3U, 0xcedf722a585139baU, + 0xc2974eb4ee658828U, 0x733d226229feea32U, 0x0806357d5a3f525fU, + 0xca07c2dcb0cf26f7U, 0xfc89b393dd02f0b5U, 0xbbac2078d443ace2U, + 0xd54b944b84aa4c0dU, 0x0a9e795e65d4df11U, 0x4d4617b5ff4a16d5U, + 0x504bced1bf8e4e45U, 0xe45ec2862f71e1d6U, 0x5d767327bb4e5a4cU, + 0x3a6a07f8d510f86fU, 0x890489f70a55368bU, 0x2b45ac74ccea842eU, + 0x3b0b8bc90012929dU, 0x09ce6ebb40173744U, 0xcc420a6a101d0515U, + 0x9fa946824a12232dU, 0x47939822dc96abf9U, 0x59787e2b93bc56f7U, + 0x57eb4edb3c55b65aU, 0xede622920b6b23f1U, 0xe95fab368e45ecedU, + 0x11dbcb0218ebb414U, 0xd652bdc29f26a119U, 0x4be76d3346f0495fU, + 0x6f70a4400c562ddbU, 0xcb4ccd500f6bb952U, 0x7e2000a41346a7a7U, + 0x8ed400668c0c28c8U, 0x728900802f0f32faU, 0x4f2b40a03ad2ffb9U, + 0xe2f610c84987bfa8U, 0x0dd9ca7d2df4d7c9U, 0x91503d1c79720dbbU, + 0x75a44c6397ce912aU, 0xc986afbe3ee11abaU, 0xfbe85badce996168U, + 0xfae27299423fb9c3U, 0xdccd879fc967d41aU, 0x5400e987bbc1c920U, + 0x290123e9aab23b68U, 0xf9a0b6720aaf6521U, 0xf808e40e8d5b3e69U, + 0xb60b1d1230b20e04U, 0xb1c6f22b5e6f48c2U, 0x1e38aeb6360b1af3U, + 0x25c6da63c38de1b0U, 0x579c487e5a38ad0eU, 0x2d835a9df0c6d851U, + 0xf8e431456cf88e65U, 0x1b8e9ecb641b58ffU, 0xe272467e3d222f3fU, + 0x5b0ed81dcc6abb0fU, 0x98e947129fc2b4e9U, 0x3f2398d747b36224U, + 0x8eec7f0d19a03aadU, 0x1953cf68300424acU, 0x5fa8c3423c052dd7U, + 0x3792f412cb06794dU, 0xe2bbd88bbee40bd0U, 0x5b6aceaeae9d0ec4U, + 0xf245825a5a445275U, 0xeed6e2f0f0d56712U, 0x55464dd69685606bU, + 0xaa97e14c3c26b886U, 0xd53dd99f4b3066a8U, 0xe546a8038efe4029U, + 0xde98520472bdd033U, 0x963e66858f6d4440U, 0xdde7001379a44aa8U, + 0x5560c018580d5d52U, 0xaab8f01e6e10b4a6U, 0xcab3961304ca70e8U, + 0x3d607b97c5fd0d22U, 0x8cb89a7db77c506aU, 0x77f3608e92adb242U, + 0x55f038b237591ed3U, 0x6b6c46dec52f6688U, 0x2323ac4b3b3da015U, + 0xabec975e0a0d081aU, 0x96e7bd358c904a21U, 0x7e50d64177da2e54U, + 0xdde50bd1d5d0b9e9U, 0x955e4ec64b44e864U, 0xbd5af13bef0b113eU, + 0xecb1ad8aeacdd58eU, 0x67de18eda5814af2U, 0x80eacf948770ced7U, + 0xa1258379a94d028dU, 0x096ee45813a04330U, 0x8bca9d6e188853fcU, + 0x775ea264cf55347dU, 0x95364afe032a819dU, 0x3a83ddbd83f52204U, + 0xc4926a9672793542U, 0x75b7053c0f178293U, 0x5324c68b12dd6338U, + 0xd3f6fc16ebca5e03U, 0x88f4bb1ca6bcf584U, 0x2b31e9e3d06c32e5U, + 0x3aff322e62439fcfU, 0x09befeb9fad487c2U, 0x4c2ebe687989a9b3U, + 0x0f9d37014bf60a10U, 0x538484c19ef38c94U, 0x2865a5f206b06fb9U, + 0xf93f87b7442e45d3U, 0xf78f69a51539d748U, 0xb573440e5a884d1bU, + 0x31680a88f8953030U, 0xfdc20d2b36ba7c3dU, 0x3d32907604691b4cU, + 0xa63f9a49c2c1b10fU, 0x0fcf80dc33721d53U, 0xd3c36113404ea4a8U, + 0x645a1cac083126e9U, 0x3d70a3d70a3d70a3U, 0xccccccccccccccccU, + 0x0000000000000000U, 0x0000000000000000U, 0x0000000000000000U, + 0x0000000000000000U, 0x0000000000000000U, 0x0000000000000000U, + 0x0000000000000000U, 0x0000000000000000U, 0x0000000000000000U, + 0x0000000000000000U, 0x0000000000000000U, 0x0000000000000000U, + 0x0000000000000000U, 0x0000000000000000U, 0x0000000000000000U, + 0x0000000000000000U, 0x0000000000000000U, 0x0000000000000000U, + 0x0000000000000000U, 0x0000000000000000U, 0x0000000000000000U, + 0x0000000000000000U, 0x0000000000000000U, 0x0000000000000000U, + 0x0000000000000000U, 0x0000000000000000U, 0x0000000000000000U, + 0x0000000000000000U, 0x4000000000000000U, 0x5000000000000000U, + 0xa400000000000000U, 0x4d00000000000000U, 0xf020000000000000U, + 0x6c28000000000000U, 0xc732000000000000U, 0x3c7f400000000000U, + 0x4b9f100000000000U, 0x1e86d40000000000U, 0x1314448000000000U, + 0x17d955a000000000U, 0x5dcfab0800000000U, 0x5aa1cae500000000U, + 0xf14a3d9e40000000U, 0x6d9ccd05d0000000U, 0xe4820023a2000000U, + 0xdda2802c8a800000U, 0xd50b2037ad200000U, 0x4526f422cc340000U, + 0x9670b12b7f410000U, 0x3c0cdd765f114000U, 0xa5880a69fb6ac800U, + 0x8eea0d047a457a00U, 0x72a4904598d6d880U, 0x47a6da2b7f864750U, + 0x999090b65f67d924U, 0xfff4b4e3f741cf6dU, 0xbff8f10e7a8921a4U, + 0xaff72d52192b6a0dU, 0x9bf4f8a69f764490U, 0x02f236d04753d5b4U, + 0x01d762422c946590U, 0x424d3ad2b7b97ef5U, 0xd2e0898765a7deb2U, + 0x63cc55f49f88eb2fU, 0x3cbf6b71c76b25fbU, 0x8bef464e3945ef7aU, + 0x97758bf0e3cbb5acU, 0x3d52eeed1cbea317U, 0x4ca7aaa863ee4bddU, + 0x8fe8caa93e74ef6aU, 0xb3e2fd538e122b44U, 0x60dbbca87196b616U, + 0xbc8955e946fe31cdU, 0x6babab6398bdbe41U, 0xc696963c7eed2dd1U, + 0xfc1e1de5cf543ca2U, 0x3b25a55f43294bcbU, 0x49ef0eb713f39ebeU, + 0x6e3569326c784337U, 0x49c2c37f07965404U, 0xdc33745ec97be906U, + 0x69a028bb3ded71a3U, 0xc40832ea0d68ce0cU, 0xf50a3fa490c30190U, + 0x792667c6da79e0faU, 0x577001b891185938U, 0xed4c0226b55e6f86U, + 0x544f8158315b05b4U, 0x696361ae3db1c721U, 0x03bc3a19cd1e38e9U, + 0x04ab48a04065c723U, 0x62eb0d64283f9c76U, 0x3ba5d0bd324f8394U, + 0xca8f44ec7ee36479U, 0x7e998b13cf4e1ecbU, 0x9e3fedd8c321a67eU, + 0xc5cfe94ef3ea101eU, 0xbba1f1d158724a12U, 0x2a8a6e45ae8edc97U, + 0xf52d09d71a3293bdU, 0x593c2626705f9c56U, 0x6f8b2fb00c77836cU, + 0x0b6dfb9c0f956447U, 0x4724bd4189bd5eacU, 0x58edec91ec2cb657U, + 0x2f2967b66737e3edU, 0xbd79e0d20082ee74U, 0xecd8590680a3aa11U, + 0xe80e6f4820cc9495U, 0x3109058d147fdcddU, 0xbd4b46f0599fd415U, + 0x6c9e18ac7007c91aU, 0x03e2cf6bc604ddb0U, 0x84db8346b786151cU, + 0xe612641865679a63U, 0x4fcb7e8f3f60c07eU, 0xe3be5e330f38f09dU, + 0x5cadf5bfd3072cc5U, 0x73d9732fc7c8f7f6U, 0x2867e7fddcdd9afaU, + 0xb281e1fd541501b8U, 0x1f225a7ca91a4226U, 0x3375788de9b06958U, + 0x0052d6b1641c83aeU, 0xc0678c5dbd23a49aU, 0xf840b7ba963646e0U, + 0xb650e5a93bc3d898U, 0xa3e51f138ab4cebeU, 0xc66f336c36b10137U, + 0xb80b0047445d4184U, 0xa60dc059157491e5U, 0x87c89837ad68db2fU, + 0x29babe4598c311fbU, 0xf4296dd6fef3d67aU, 0x1899e4a65f58660cU, + 0x5ec05dcff72e7f8fU, 0x76707543f4fa1f73U, 0x6a06494a791c53a8U, + 0x0487db9d17636892U, 0x45a9d2845d3c42b6U, 0x0b8a2392ba45a9b2U, + 0x8e6cac7768d7141eU, 0x3207d795430cd926U, 0x7f44e6bd49e807b8U, + 0x5f16206c9c6209a6U, 0x36dba887c37a8c0fU, 0xc2494954da2c9789U, + 0xf2db9baa10b7bd6cU, 0x6f92829494e5acc7U, 0xcb772339ba1f17f9U, + 0xff2a760414536efbU, 0xfef5138519684abaU, 0x7eb258665fc25d69U, + 0xef2f773ffbd97a61U, 0xaafb550ffacfd8faU, 0x95ba2a53f983cf38U, + 0xdd945a747bf26183U, 0x94f971119aeef9e4U, 0x7a37cd5601aab85dU, + 0xac62e055c10ab33aU, 0x577b986b314d6009U, 0xed5a7e85fda0b80bU, + 0x14588f13be847307U, 0x596eb2d8ae258fc8U, 0x6fca5f8ed9aef3bbU, + 0x25de7bb9480d5854U, 0xaf561aa79a10ae6aU, 0x1b2ba1518094da04U, + 0x90fb44d2f05d0842U, 0x353a1607ac744a53U, 0x42889b8997915ce8U, + 0x69956135febada11U, 0x43fab9837e699095U, 0x94f967e45e03f4bbU, + 0x1d1be0eebac278f5U, 0x6462d92a69731732U, 0x7d7b8f7503cfdcfeU, + 0x5cda735244c3d43eU, 0x3a0888136afa64a7U, 0x088aaa1845b8fdd0U, + 0x8aad549e57273d45U, 0x36ac54e2f678864bU, 0x84576a1bb416a7ddU, + 0x656d44a2a11c51d5U, 0x9f644ae5a4b1b325U, 0x873d5d9f0dde1feeU, + 0xa90cb506d155a7eaU, 0x09a7f12442d588f2U, 0x0c11ed6d538aeb2fU, + 0x8f1668c8a86da5faU, 0xf96e017d694487bcU, 0x37c981dcc395a9acU, + 0x85bbe253f47b1417U, 0x93956d7478ccec8eU, 0x387ac8d1970027b2U, + 0x06997b05fcc0319eU, 0x441fece3bdf81f03U, 0xd527e81cad7626c3U, + 0x8a71e223d8d3b074U, 0xf6872d5667844e49U, 0xb428f8ac016561dbU, + 0xe13336d701beba52U, 0xecc0024661173473U, 0x27f002d7f95d0190U, + 0x31ec038df7b441f4U, 0x7e67047175a15271U, 0x0f0062c6e984d386U, + 0x52c07b78a3e60868U, 0xa7709a56ccdf8a82U, 0x88a66076400bb691U, + 0x6acff893d00ea435U, 0x0583f6b8c4124d43U, 0xc3727a337a8b704aU, + 0x744f18c0592e4c5cU, 0x1162def06f79df73U, 0x8addcb5645ac2ba8U, + 0x6d953e2bd7173692U, 0xc8fa8db6ccdd0437U, 0x1d9c9892400a22a2U, + 0x2503beb6d00cab4bU, 0x2e44ae64840fd61dU, 0x5ceaecfed289e5d2U, + 0x7425a83e872c5f47U, 0xd12f124e28f77719U, 0x82bd6b70d99aaa6fU, + 0x636cc64d1001550bU, 0x3c47f7e05401aa4eU, 0x65acfaec34810a71U, + 0x7f1839a741a14d0dU, 0x1ede48111209a050U, 0x934aed0aab460432U, + 0xf81da84d5617853fU, 0x36251260ab9d668eU, 0xc1d72b7c6b426019U, + 0xb24cf65b8612f81fU, 0xdee033f26797b627U, 0x169840ef017da3b1U, + 0x8e1f289560ee864eU, 0xf1a6f2bab92a27e2U, 0xae10af696774b1dbU, + 0xacca6da1e0a8ef29U, 0x17fd090a58d32af3U, 0xddfc4b4cef07f5b0U, + 0x4abdaf101564f98eU, 0x9d6d1ad41abe37f1U, 0x84c86189216dc5edU, + 0x32fd3cf5b4e49bb4U, 0x3fbc8c33221dc2a1U, 0x0fabaf3feaa5334aU, + 0x29cb4d87f2a7400eU, 0x743e20e9ef511012U, 0x914da9246b255416U, + 0x1ad089b6c2f7548eU, 0xa184ac2473b529b1U, 0xc9e5d72d90a2741eU, + 0x7e2fa67c7a658892U, 0xddbb901b98feeab7U, 0x552a74227f3ea565U, + 0xd53a88958f87275fU, 0x8a892abaf368f137U, 0x2d2b7569b0432d85U, + 0x9c3b29620e29fc73U, 0x8349f3ba91b47b8fU, 0x241c70a936219a73U, + 0xed238cd383aa0110U, 0xf4363804324a40aaU, 0xb143c6053edcd0d5U, + 0xdd94b7868e94050aU, 0xca7cf2b4191c8326U, 0xfd1c2f611f63a3f0U, + 0xbc633b39673c8cecU, 0xd5be0503e085d813U, 0x4b2d8644d8a74e18U, + 0xddf8e7d60ed1219eU, 0xcabb90e5c942b503U, 0x3d6a751f3b936243U, + 0x0cc512670a783ad4U, 0x27fb2b80668b24c5U, 0xb1f9f660802dedf6U, + 0x5e7873f8a0396973U, 0xdb0b487b6423e1e8U, 0x91ce1a9a3d2cda62U, + 0x7641a140cc7810fbU, 0xa9e904c87fcb0a9dU, 0x546345fa9fbdcd44U, + 0xa97c177947ad4095U, 0x49ed8eabcccc485dU, 0x5c68f256bfff5a74U, + 0x73832eec6fff3111U, 0xc831fd53c5ff7eabU, 0xba3e7ca8b77f5e55U, + 0x28ce1bd2e55f35ebU, 0x7980d163cf5b81b3U, 0xd7e105bcc332621fU, + 0x8dd9472bf3fefaa7U, 0xb14f98f6f0feb951U, 0x6ed1bf9a569f33d3U, + 0x0a862f80ec4700c8U, 0xcd27bb612758c0faU, 0x8038d51cb897789cU, + 0xe0470a63e6bd56c3U, 0x1858ccfce06cac74U, 0x0f37801e0c43ebc8U, + 0xd30560258f54e6baU, 0x47c6b82ef32a2069U, 0x4cdc331d57fa5441U, + 0xe0133fe4adf8e952U, 0x58180fddd97723a6U, 0x570f09eaa7ea7648U, }; } // namespace diff --git a/contrib/restricted/abseil-cpp/absl/strings/cord.cc b/contrib/restricted/abseil-cpp/absl/strings/cord.cc index 85a67a0810..1d33dd8317 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/cord.cc +++ b/contrib/restricted/abseil-cpp/absl/strings/cord.cc @@ -20,6 +20,7 @@ #include <cstdio> #include <cstdlib> #include <iomanip> +#include <ios> #include <iostream> #include <limits> #include <ostream> @@ -34,6 +35,7 @@ #include "absl/base/port.h" #include "absl/container/fixed_array.h" #include "absl/container/inlined_vector.h" +#include "absl/crc/internal/crc_cord_state.h" #include "absl/strings/cord_buffer.h" #include "absl/strings/escaping.h" #include "absl/strings/internal/cord_data_edge.h" @@ -166,9 +168,7 @@ constexpr unsigned char Cord::InlineRep::kMaxInline; inline void Cord::InlineRep::set_data(const char* data, size_t n) { static_assert(kMaxInline == 15, "set_data is hard-coded for a length of 15"); - - cord_internal::SmallMemmove<true>(data_.as_chars(), data, n); - set_inline_size(n); + data_.set_inline_data(data, n); } inline char* Cord::InlineRep::set_data(size_t n) { @@ -184,7 +184,7 @@ inline void Cord::InlineRep::reduce_size(size_t n) { assert(tag >= n); tag -= n; memset(data_.as_chars() + tag, 0, n); - set_inline_size(static_cast<char>(tag)); + set_inline_size(tag); } inline void Cord::InlineRep::remove_prefix(size_t n) { @@ -419,6 +419,7 @@ Cord& Cord::operator=(absl::string_view src) { // we keep it here to make diffs easier. void Cord::InlineRep::AppendArray(absl::string_view src, MethodIdentifier method) { + MaybeRemoveEmptyCrcNode(); if (src.empty()) return; // memcpy(_, nullptr, 0) is undefined. size_t appended = 0; @@ -436,8 +437,8 @@ void Cord::InlineRep::AppendArray(absl::string_view src, size_t inline_length = inline_size(); if (src.size() <= kMaxInline - inline_length) { // Append new data to embedded array - memcpy(data_.as_chars() + inline_length, src.data(), src.size()); set_inline_size(inline_length + src.size()); + memcpy(data_.as_chars() + inline_length, src.data(), src.size()); return; } @@ -478,6 +479,10 @@ inline CordRep* Cord::TakeRep() && { template <typename C> inline void Cord::AppendImpl(C&& src) { auto constexpr method = CordzUpdateTracker::kAppendCord; + + contents_.MaybeRemoveEmptyCrcNode(); + if (src.empty()) return; + if (empty()) { // Since destination is empty, we can avoid allocating a node, if (src.contents_.is_tree()) { @@ -537,18 +542,23 @@ static CordRep::ExtractResult ExtractAppendBuffer(CordRep* rep, } } -static CordBuffer CreateAppendBuffer(InlineData& data, size_t capacity) { +static CordBuffer CreateAppendBuffer(InlineData& data, size_t block_size, + size_t capacity) { // Watch out for overflow, people can ask for size_t::max(). const size_t size = data.inline_size(); - capacity = (std::min)(std::numeric_limits<size_t>::max() - size, capacity); - CordBuffer buffer = CordBuffer::CreateWithDefaultLimit(size + capacity); + const size_t max_capacity = std::numeric_limits<size_t>::max() - size; + capacity = (std::min)(max_capacity, capacity) + size; + CordBuffer buffer = + block_size ? CordBuffer::CreateWithCustomLimit(block_size, capacity) + : CordBuffer::CreateWithDefaultLimit(capacity); cord_internal::SmallMemmove(buffer.data(), data.as_chars(), size); buffer.SetLength(size); data = {}; return buffer; } -CordBuffer Cord::GetAppendBufferSlowPath(size_t capacity, size_t min_capacity) { +CordBuffer Cord::GetAppendBufferSlowPath(size_t block_size, size_t capacity, + size_t min_capacity) { auto constexpr method = CordzUpdateTracker::kGetAppendBuffer; CordRep* tree = contents_.tree(); if (tree != nullptr) { @@ -558,9 +568,10 @@ CordBuffer Cord::GetAppendBufferSlowPath(size_t capacity, size_t min_capacity) { contents_.SetTreeOrEmpty(result.tree, scope); return CordBuffer(result.extracted->flat()); } - return CordBuffer::CreateWithDefaultLimit(capacity); + return block_size ? CordBuffer::CreateWithCustomLimit(block_size, capacity) + : CordBuffer::CreateWithDefaultLimit(capacity); } - return CreateAppendBuffer(contents_.data_, capacity); + return CreateAppendBuffer(contents_.data_, block_size, capacity); } void Cord::Append(const Cord& src) { @@ -584,6 +595,9 @@ void Cord::Append(T&& src) { template void Cord::Append(std::string&& src); void Cord::Prepend(const Cord& src) { + contents_.MaybeRemoveEmptyCrcNode(); + if (src.empty()) return; + CordRep* src_tree = src.contents_.tree(); if (src_tree != nullptr) { CordRep::Ref(src_tree); @@ -598,16 +612,18 @@ void Cord::Prepend(const Cord& src) { } void Cord::PrependArray(absl::string_view src, MethodIdentifier method) { + contents_.MaybeRemoveEmptyCrcNode(); if (src.empty()) return; // memcpy(_, nullptr, 0) is undefined. + if (!contents_.is_tree()) { size_t cur_size = contents_.inline_size(); if (cur_size + src.size() <= InlineRep::kMaxInline) { // Use embedded storage. - char data[InlineRep::kMaxInline + 1] = {0}; - memcpy(data, src.data(), src.size()); - memcpy(data + src.size(), contents_.data(), cur_size); - memcpy(contents_.data_.as_chars(), data, InlineRep::kMaxInline + 1); - contents_.set_inline_size(cur_size + src.size()); + InlineData data; + data.set_inline_size(cur_size + src.size()); + memcpy(data.as_chars(), src.data(), src.size()); + memcpy(data.as_chars() + src.size(), contents_.data(), cur_size); + contents_.data_ = data; return; } } @@ -620,8 +636,8 @@ void Cord::AppendPrecise(absl::string_view src, MethodIdentifier method) { assert(src.size() <= cord_internal::kMaxFlatLength); if (contents_.remaining_inline_capacity() >= src.size()) { const size_t inline_length = contents_.inline_size(); - memcpy(contents_.data_.as_chars() + inline_length, src.data(), src.size()); contents_.set_inline_size(inline_length + src.size()); + memcpy(contents_.data_.as_chars() + inline_length, src.data(), src.size()); } else { contents_.AppendTree(CordRepFlat::Create(src), method); } @@ -631,12 +647,12 @@ void Cord::PrependPrecise(absl::string_view src, MethodIdentifier method) { assert(!src.empty()); assert(src.size() <= cord_internal::kMaxFlatLength); if (contents_.remaining_inline_capacity() >= src.size()) { - const size_t inline_length = contents_.inline_size(); - char data[InlineRep::kMaxInline + 1] = {0}; - memcpy(data, src.data(), src.size()); - memcpy(data + src.size(), contents_.data(), inline_length); - memcpy(contents_.data_.as_chars(), data, InlineRep::kMaxInline + 1); - contents_.set_inline_size(inline_length + src.size()); + const size_t cur_size = contents_.inline_size(); + InlineData data; + data.set_inline_size(cur_size + src.size()); + memcpy(data.as_chars(), src.data(), src.size()); + memcpy(data.as_chars() + src.size(), contents_.data(), cur_size); + contents_.data_ = data; } else { contents_.PrependTree(CordRepFlat::Create(src), method); } @@ -658,6 +674,7 @@ void Cord::RemovePrefix(size_t n) { ABSL_INTERNAL_CHECK(n <= size(), absl::StrCat("Requested prefix size ", n, " exceeds Cord's size ", size())); + contents_.MaybeRemoveEmptyCrcNode(); CordRep* tree = contents_.tree(); if (tree == nullptr) { contents_.remove_prefix(n); @@ -688,6 +705,7 @@ void Cord::RemoveSuffix(size_t n) { ABSL_INTERNAL_CHECK(n <= size(), absl::StrCat("Requested suffix size ", n, " exceeds Cord's size ", size())); + contents_.MaybeRemoveEmptyCrcNode(); CordRep* tree = contents_.tree(); if (tree == nullptr) { contents_.reduce_size(n); @@ -726,6 +744,7 @@ Cord Cord::Subcord(size_t pos, size_t new_size) const { } if (new_size <= InlineRep::kMaxInline) { + sub_cord.contents_.set_inline_size(new_size); char* dest = sub_cord.contents_.data_.as_chars(); Cord::ChunkIterator it = chunk_begin(); it.AdvanceBytes(pos); @@ -737,7 +756,6 @@ Cord Cord::Subcord(size_t pos, size_t new_size) const { ++it; } cord_internal::SmallMemmove(dest, it->data(), remaining_size); - sub_cord.contents_.set_inline_size(new_size); return sub_cord; } @@ -835,26 +853,44 @@ inline absl::string_view Cord::InlineRep::FindFlatStartPiece() const { return absl::string_view(node->external()->base + offset, length); } -void Cord::SetExpectedChecksum(uint32_t crc) { +void Cord::SetCrcCordState(crc_internal::CrcCordState state) { auto constexpr method = CordzUpdateTracker::kSetExpectedChecksum; - if (empty()) return; - - if (!contents_.is_tree()) { + if (empty()) { + contents_.MaybeRemoveEmptyCrcNode(); + CordRep* rep = CordRepCrc::New(nullptr, std::move(state)); + contents_.EmplaceTree(rep, method); + } else if (!contents_.is_tree()) { CordRep* rep = contents_.MakeFlatWithExtraCapacity(0); - rep = CordRepCrc::New(rep, crc); + rep = CordRepCrc::New(rep, std::move(state)); contents_.EmplaceTree(rep, method); } else { const CordzUpdateScope scope(contents_.data_.cordz_info(), method); - CordRep* rep = CordRepCrc::New(contents_.data_.as_tree(), crc); + CordRep* rep = CordRepCrc::New(contents_.data_.as_tree(), std::move(state)); contents_.SetTree(rep, scope); } } +void Cord::SetExpectedChecksum(uint32_t crc) { + // Construct a CrcCordState with a single chunk. + crc_internal::CrcCordState state; + state.mutable_rep()->prefix_crc.push_back( + crc_internal::CrcCordState::PrefixCrc(size(), absl::crc32c_t{crc})); + SetCrcCordState(std::move(state)); +} + +const crc_internal::CrcCordState* Cord::MaybeGetCrcCordState() const { + if (!contents_.is_tree() || !contents_.tree()->IsCrc()) { + return nullptr; + } + return &contents_.tree()->crc()->crc_cord_state; +} + absl::optional<uint32_t> Cord::ExpectedChecksum() const { if (!contents_.is_tree() || !contents_.tree()->IsCrc()) { return absl::nullopt; } - return contents_.tree()->crc()->crc; + return static_cast<uint32_t>( + contents_.tree()->crc()->crc_cord_state.Checksum()); } inline int Cord::CompareSlowPath(absl::string_view rhs, size_t compared_size, @@ -922,6 +958,7 @@ inline int Cord::CompareSlowPath(const Cord& rhs, size_t compared_size, } inline absl::string_view Cord::GetFirstChunk(const Cord& c) { + if (c.empty()) return {}; return c.contents_.FindFlatStartPiece(); } inline absl::string_view Cord::GetFirstChunk(absl::string_view sv) { @@ -1092,7 +1129,7 @@ Cord Cord::ChunkIterator::AdvanceAndReadBytes(size_t n) { : current_leaf_; const char* data = payload->IsExternal() ? payload->external()->base : payload->flat()->Data(); - const size_t offset = current_chunk_.data() - data; + const size_t offset = static_cast<size_t>(current_chunk_.data() - data); auto* tree = CordRepSubstring::Substring(payload, offset, n); subcord.contents_.EmplaceTree(VerifyTree(tree), method); @@ -1159,6 +1196,10 @@ absl::string_view Cord::FlattenSlowPath() { /* static */ bool Cord::GetFlatAux(CordRep* rep, absl::string_view* fragment) { assert(rep != nullptr); + if (rep->length == 0) { + *fragment = absl::string_view(); + return true; + } rep = cord_internal::SkipCrcNode(rep); if (rep->IsFlat()) { *fragment = absl::string_view(rep->flat()->Data(), rep->length); @@ -1190,6 +1231,7 @@ absl::string_view Cord::FlattenSlowPath() { absl::cord_internal::CordRep* rep, absl::FunctionRef<void(absl::string_view)> callback) { assert(rep != nullptr); + if (rep->length == 0) return; rep = cord_internal::SkipCrcNode(rep); if (rep->IsBtree()) { @@ -1223,8 +1265,12 @@ static void DumpNode(CordRep* rep, bool include_data, std::ostream* os, if (include_data) *os << static_cast<void*>(rep); *os << "]"; *os << " " << std::setw(indent) << ""; - if (rep->IsCrc()) { - *os << "CRC crc=" << rep->crc()->crc << "\n"; + bool leaf = false; + if (rep == nullptr) { + *os << "NULL\n"; + leaf = true; + } else if (rep->IsCrc()) { + *os << "CRC crc=" << rep->crc()->crc_cord_state.Checksum() << "\n"; indent += kIndentStep; rep = rep->crc()->child; } else if (rep->IsSubstring()) { @@ -1232,6 +1278,7 @@ static void DumpNode(CordRep* rep, bool include_data, std::ostream* os, indent += kIndentStep; rep = rep->substring()->child; } else { // Leaf or ring + leaf = true; if (rep->IsExternal()) { *os << "EXTERNAL ["; if (include_data) @@ -1245,6 +1292,8 @@ static void DumpNode(CordRep* rep, bool include_data, std::ostream* os, } else { CordRepBtree::Dump(rep, /*label=*/ "", include_data, *os); } + } + if (leaf) { if (stack.empty()) break; rep = stack.back(); stack.pop_back(); @@ -1290,11 +1339,14 @@ static bool VerifyNode(CordRep* root, CordRep* start_node, node->substring()->child->length, ReportError(root, node)); } else if (node->IsCrc()) { - ABSL_INTERNAL_CHECK(node->crc()->child != nullptr, - ReportError(root, node)); - ABSL_INTERNAL_CHECK(node->crc()->length == node->crc()->child->length, - ReportError(root, node)); - worklist.push_back(node->crc()->child); + ABSL_INTERNAL_CHECK( + node->crc()->child != nullptr || node->crc()->length == 0, + ReportError(root, node)); + if (node->crc()->child != nullptr) { + ABSL_INTERNAL_CHECK(node->crc()->length == node->crc()->child->length, + ReportError(root, node)); + worklist.push_back(node->crc()->child); + } } } while (!worklist.empty()); return true; @@ -1302,7 +1354,7 @@ static bool VerifyNode(CordRep* root, CordRep* start_node, std::ostream& operator<<(std::ostream& out, const Cord& cord) { for (absl::string_view chunk : cord.Chunks()) { - out.write(chunk.data(), chunk.size()); + out.write(chunk.data(), static_cast<std::streamsize>(chunk.size())); } return out; } diff --git a/contrib/restricted/abseil-cpp/absl/strings/cord.h b/contrib/restricted/abseil-cpp/absl/strings/cord.h index 18d6ab852f..c4a0d5aaae 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/cord.h +++ b/contrib/restricted/abseil-cpp/absl/strings/cord.h @@ -20,8 +20,7 @@ // structure. A Cord is a string-like sequence of characters optimized for // specific use cases. Unlike a `std::string`, which stores an array of // contiguous characters, Cord data is stored in a structure consisting of -// separate, reference-counted "chunks." (Currently, this implementation is a -// tree structure, though that implementation may change.) +// separate, reference-counted "chunks." // // Because a Cord consists of these chunks, data can be added to or removed from // a Cord during its lifetime. Chunks may also be shared between Cords. Unlike a @@ -77,6 +76,7 @@ #include "absl/base/macros.h" #include "absl/base/port.h" #include "absl/container/inlined_vector.h" +#include "absl/crc/internal/crc_cord_state.h" #include "absl/functional/function_ref.h" #include "absl/meta/type_traits.h" #include "absl/strings/cord_analysis.h" @@ -284,6 +284,19 @@ class Cord { // } CordBuffer GetAppendBuffer(size_t capacity, size_t min_capacity = 16); + // Returns a CordBuffer, re-using potential existing capacity in this cord. + // + // This function is identical to `GetAppendBuffer`, except that in the case + // where a new `CordBuffer` is allocated, it is allocated using the provided + // custom limit instead of the default limit. `GetAppendBuffer` will default + // to `CordBuffer::CreateWithDefaultLimit(capacity)` whereas this method + // will default to `CordBuffer::CreateWithCustomLimit(block_size, capacity)`. + // This method is equivalent to `GetAppendBuffer` if `block_size` is zero. + // See the documentation for `CreateWithCustomLimit` for more details on the + // restrictions and legal values for `block_size`. + CordBuffer GetCustomAppendBuffer(size_t block_size, size_t capacity, + size_t min_capacity = 16); + // Cord::Prepend() // // Prepends data to the Cord, which may come from another Cord or other string @@ -802,7 +815,7 @@ class Cord { InlineRep& operator=(const InlineRep& src); InlineRep& operator=(InlineRep&& src) noexcept; - explicit constexpr InlineRep(cord_internal::InlineData data); + explicit constexpr InlineRep(absl::string_view sv, CordRep* rep); void Swap(InlineRep* rhs); bool empty() const; @@ -861,33 +874,14 @@ class Cord { void PrependTreeToTree(CordRep* tree, MethodIdentifier method); void PrependTree(CordRep* tree, MethodIdentifier method); - template <bool has_length> - void GetAppendRegion(char** region, size_t* size, size_t length); + bool IsSame(const InlineRep& other) const { return data_ == other.data_; } - bool IsSame(const InlineRep& other) const { - return memcmp(&data_, &other.data_, sizeof(data_)) == 0; - } - int BitwiseCompare(const InlineRep& other) const { - uint64_t x, y; - // Use memcpy to avoid aliasing issues. - memcpy(&x, &data_, sizeof(x)); - memcpy(&y, &other.data_, sizeof(y)); - if (x == y) { - memcpy(&x, reinterpret_cast<const char*>(&data_) + 8, sizeof(x)); - memcpy(&y, reinterpret_cast<const char*>(&other.data_) + 8, sizeof(y)); - if (x == y) return 0; - } - return absl::big_endian::FromHost64(x) < absl::big_endian::FromHost64(y) - ? -1 - : 1; - } void CopyTo(std::string* dst) const { // memcpy is much faster when operating on a known size. On most supported // platforms, the small string optimization is large enough that resizing // to 15 bytes does not cause a memory allocation. - absl::strings_internal::STLStringResizeUninitialized(dst, - sizeof(data_) - 1); - memcpy(&(*dst)[0], &data_, sizeof(data_) - 1); + absl::strings_internal::STLStringResizeUninitialized(dst, kMaxInline); + data_.copy_max_inline_to(&(*dst)[0]); // erase is faster than resize because the logic for memory allocation is // not needed. dst->erase(inline_size()); @@ -932,6 +926,13 @@ class Cord { void set_inline_size(size_t size) { data_.set_inline_size(size); } size_t inline_size() const { return data_.inline_size(); } + // Empty cords that carry a checksum have a CordRepCrc node with a null + // child node. The code can avoid lots of special cases where it would + // otherwise transition from tree to inline storage if we just remove the + // CordRepCrc node before mutations. Must never be called inside a + // CordzUpdateScope since it untracks the cordz info. + void MaybeRemoveEmptyCrcNode(); + cord_internal::InlineData data_; }; InlineRep contents_; @@ -980,7 +981,8 @@ class Cord { void AppendPrecise(absl::string_view src, MethodIdentifier method); void PrependPrecise(absl::string_view src, MethodIdentifier method); - CordBuffer GetAppendBufferSlowPath(size_t capacity, size_t min_capacity); + CordBuffer GetAppendBufferSlowPath(size_t block_size, size_t capacity, + size_t min_capacity); // Prepends the provided data to this instance. `method` contains the public // API method for this action which is tracked for Cordz sampling purposes. @@ -1000,6 +1002,10 @@ class Cord { }); return H::combine(combiner.finalize(std::move(hash_state)), size()); } + + friend class CrcCord; + void SetCrcCordState(crc_internal::CrcCordState state); + const crc_internal::CrcCordState* MaybeGetCrcCordState() const; }; ABSL_NAMESPACE_END @@ -1016,46 +1022,6 @@ extern std::ostream& operator<<(std::ostream& out, const Cord& cord); namespace cord_internal { -// Fast implementation of memmove for up to 15 bytes. This implementation is -// safe for overlapping regions. If nullify_tail is true, the destination is -// padded with '\0' up to 16 bytes. -template <bool nullify_tail = false> -inline void SmallMemmove(char* dst, const char* src, size_t n) { - if (n >= 8) { - assert(n <= 16); - uint64_t buf1; - uint64_t buf2; - memcpy(&buf1, src, 8); - memcpy(&buf2, src + n - 8, 8); - if (nullify_tail) { - memset(dst + 8, 0, 8); - } - memcpy(dst, &buf1, 8); - memcpy(dst + n - 8, &buf2, 8); - } else if (n >= 4) { - uint32_t buf1; - uint32_t buf2; - memcpy(&buf1, src, 4); - memcpy(&buf2, src + n - 4, 4); - if (nullify_tail) { - memset(dst + 4, 0, 4); - memset(dst + 8, 0, 8); - } - memcpy(dst, &buf1, 4); - memcpy(dst + n - 4, &buf2, 4); - } else { - if (n != 0) { - dst[0] = src[0]; - dst[n / 2] = src[n / 2]; - dst[n - 1] = src[n - 1]; - } - if (nullify_tail) { - memset(dst + 8, 0, 8); - memset(dst + n, 0, 8); - } - } -} - // Does non-template-specific `CordRepExternal` initialization. // Requires `data` to be non-empty. void InitializeCordRepExternal(absl::string_view data, CordRepExternal* rep); @@ -1099,8 +1065,8 @@ Cord MakeCordFromExternal(absl::string_view data, Releaser&& releaser) { return cord; } -constexpr Cord::InlineRep::InlineRep(cord_internal::InlineData data) - : data_(data) {} +constexpr Cord::InlineRep::InlineRep(absl::string_view sv, CordRep* rep) + : data_(sv, rep) {} inline Cord::InlineRep::InlineRep(const Cord::InlineRep& src) : data_(InlineData::kDefaultInit) { @@ -1179,7 +1145,7 @@ inline cord_internal::CordRepFlat* Cord::InlineRep::MakeFlatWithExtraCapacity( size_t len = data_.inline_size(); auto* result = CordRepFlat::New(len + extra); result->length = len; - memcpy(result->Data(), data_.as_chars(), sizeof(data_)); + data_.copy_max_inline_to(result->Data()); return result; } @@ -1241,6 +1207,18 @@ inline void Cord::InlineRep::CopyToArray(char* dst) const { cord_internal::SmallMemmove(dst, data_.as_chars(), n); } +inline void Cord::InlineRep::MaybeRemoveEmptyCrcNode() { + CordRep* rep = tree(); + if (rep == nullptr || ABSL_PREDICT_TRUE(rep->length > 0)) { + return; + } + assert(rep->IsCrc()); + assert(rep->crc()->child == nullptr); + CordzInfo::MaybeUntrackCord(cordz_info()); + CordRep::Unref(rep); + ResetToEmpty(); +} + constexpr inline Cord::Cord() noexcept {} inline Cord::Cord(absl::string_view src) @@ -1248,13 +1226,12 @@ inline Cord::Cord(absl::string_view src) template <typename T> constexpr Cord::Cord(strings_internal::StringConstant<T>) - : contents_(strings_internal::StringConstant<T>::value.size() <= + : contents_(strings_internal::StringConstant<T>::value, + strings_internal::StringConstant<T>::value.size() <= cord_internal::kMaxInline - ? cord_internal::InlineData( - strings_internal::StringConstant<T>::value) - : cord_internal::InlineData( - &cord_internal::ConstInitExternalStorage< - strings_internal::StringConstant<T>>::value)) {} + ? nullptr + : &cord_internal::ConstInitExternalStorage< + strings_internal::StringConstant<T>>::value) {} inline Cord& Cord::operator=(const Cord& x) { contents_ = x.contents_; @@ -1290,7 +1267,7 @@ inline size_t Cord::size() const { return contents_.size(); } -inline bool Cord::empty() const { return contents_.empty(); } +inline bool Cord::empty() const { return size() == 0; } inline size_t Cord::EstimatedMemoryUsage( CordMemoryAccounting accounting_method) const { @@ -1360,7 +1337,17 @@ inline void Cord::Prepend(CordBuffer buffer) { inline CordBuffer Cord::GetAppendBuffer(size_t capacity, size_t min_capacity) { if (empty()) return CordBuffer::CreateWithDefaultLimit(capacity); - return GetAppendBufferSlowPath(capacity, min_capacity); + return GetAppendBufferSlowPath(0, capacity, min_capacity); +} + +inline CordBuffer Cord::GetCustomAppendBuffer(size_t block_size, + size_t capacity, + size_t min_capacity) { + if (empty()) { + return block_size ? CordBuffer::CreateWithCustomLimit(block_size, capacity) + : CordBuffer::CreateWithDefaultLimit(capacity); + } + return GetAppendBufferSlowPath(block_size, capacity, min_capacity); } extern template void Cord::Append(std::string&& src); @@ -1368,7 +1355,7 @@ extern template void Cord::Prepend(std::string&& src); inline int Cord::Compare(const Cord& rhs) const { if (!contents_.is_tree() && !rhs.contents_.is_tree()) { - return contents_.BitwiseCompare(rhs.contents_); + return contents_.data_.Compare(rhs.contents_.data_); } return CompareImpl(rhs); @@ -1406,7 +1393,11 @@ inline Cord::ChunkIterator::ChunkIterator(cord_internal::CordRep* tree) { inline Cord::ChunkIterator::ChunkIterator(const Cord* cord) { if (CordRep* tree = cord->contents_.tree()) { bytes_remaining_ = tree->length; - InitTree(tree); + if (ABSL_PREDICT_TRUE(bytes_remaining_ != 0)) { + InitTree(tree); + } else { + current_chunk_ = {}; + } } else { bytes_remaining_ = cord->contents_.inline_size(); current_chunk_ = {cord->contents_.data(), bytes_remaining_}; @@ -1575,7 +1566,7 @@ inline void Cord::ForEachChunk( if (rep == nullptr) { callback(absl::string_view(contents_.data(), contents_.size())); } else { - return ForEachChunkAux(rep, callback); + ForEachChunkAux(rep, callback); } } diff --git a/contrib/restricted/abseil-cpp/absl/strings/cord_buffer.h b/contrib/restricted/abseil-cpp/absl/strings/cord_buffer.h index 56a6ce6f6e..15494b31e0 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/cord_buffer.h +++ b/contrib/restricted/abseil-cpp/absl/strings/cord_buffer.h @@ -330,8 +330,7 @@ class CordBuffer { // Returns the available area of the internal SSO data absl::Span<char> short_available() { - assert(is_short()); - const size_t length = (short_rep.raw_size >> 1); + const size_t length = short_length(); return absl::Span<char>(short_rep.data + length, kInlineCapacity - length); } @@ -347,7 +346,7 @@ class CordBuffer { // Returns the length of the internal SSO data. size_t short_length() const { assert(is_short()); - return short_rep.raw_size >> 1; + return static_cast<size_t>(short_rep.raw_size >> 1); } // Sets the length of the internal SSO data. @@ -412,8 +411,12 @@ class CordBuffer { // Power2 functions static bool IsPow2(size_t size) { return absl::has_single_bit(size); } - static size_t Log2Floor(size_t size) { return absl::bit_width(size) - 1; } - static size_t Log2Ceil(size_t size) { return absl::bit_width(size - 1); } + static size_t Log2Floor(size_t size) { + return static_cast<size_t>(absl::bit_width(size) - 1); + } + static size_t Log2Ceil(size_t size) { + return static_cast<size_t>(absl::bit_width(size - 1)); + } // Implementation of `CreateWithCustomLimit()`. // This implementation allows for future memory allocation hints to diff --git a/contrib/restricted/abseil-cpp/absl/strings/escaping.cc b/contrib/restricted/abseil-cpp/absl/strings/escaping.cc index 18b20b83fd..939668466b 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/escaping.cc +++ b/contrib/restricted/abseil-cpp/absl/strings/escaping.cc @@ -42,11 +42,11 @@ constexpr bool kUnescapeNulls = false; inline bool is_octal_digit(char c) { return ('0' <= c) && (c <= '7'); } -inline int hex_digit_to_int(char c) { +inline unsigned int hex_digit_to_int(char c) { static_assert('0' == 0x30 && 'A' == 0x41 && 'a' == 0x61, "Character set must be ASCII."); - assert(absl::ascii_isxdigit(c)); - int x = static_cast<unsigned char>(c); + assert(absl::ascii_isxdigit(static_cast<unsigned char>(c))); + unsigned int x = static_cast<unsigned char>(c); if (x > '9') { x += 9; } @@ -121,27 +121,29 @@ bool CUnescapeInternal(absl::string_view source, bool leave_nulls_escaped, case '7': { // octal digit: 1 to 3 digits const char* octal_start = p; - unsigned int ch = *p - '0'; - if (p < last_byte && is_octal_digit(p[1])) ch = ch * 8 + *++p - '0'; + unsigned int ch = static_cast<unsigned int>(*p - '0'); // digit 1 if (p < last_byte && is_octal_digit(p[1])) - ch = ch * 8 + *++p - '0'; // now points at last digit + ch = ch * 8 + static_cast<unsigned int>(*++p - '0'); // digit 2 + if (p < last_byte && is_octal_digit(p[1])) + ch = ch * 8 + static_cast<unsigned int>(*++p - '0'); // digit 3 if (ch > 0xff) { if (error) { *error = "Value of \\" + - std::string(octal_start, p + 1 - octal_start) + + std::string(octal_start, + static_cast<size_t>(p + 1 - octal_start)) + " exceeds 0xff"; } return false; } if ((ch == 0) && leave_nulls_escaped) { // Copy the escape sequence for the null character - const ptrdiff_t octal_size = p + 1 - octal_start; + const size_t octal_size = static_cast<size_t>(p + 1 - octal_start); *d++ = '\\'; memmove(d, octal_start, octal_size); d += octal_size; break; } - *d++ = ch; + *d++ = static_cast<char>(ch); break; } case 'x': @@ -149,32 +151,34 @@ bool CUnescapeInternal(absl::string_view source, bool leave_nulls_escaped, if (p >= last_byte) { if (error) *error = "String cannot end with \\x"; return false; - } else if (!absl::ascii_isxdigit(p[1])) { + } else if (!absl::ascii_isxdigit(static_cast<unsigned char>(p[1]))) { if (error) *error = "\\x cannot be followed by a non-hex digit"; return false; } unsigned int ch = 0; const char* hex_start = p; - while (p < last_byte && absl::ascii_isxdigit(p[1])) + while (p < last_byte && + absl::ascii_isxdigit(static_cast<unsigned char>(p[1]))) // Arbitrarily many hex digits ch = (ch << 4) + hex_digit_to_int(*++p); if (ch > 0xFF) { if (error) { *error = "Value of \\" + - std::string(hex_start, p + 1 - hex_start) + + std::string(hex_start, + static_cast<size_t>(p + 1 - hex_start)) + " exceeds 0xff"; } return false; } if ((ch == 0) && leave_nulls_escaped) { // Copy the escape sequence for the null character - const ptrdiff_t hex_size = p + 1 - hex_start; + const size_t hex_size = static_cast<size_t>(p + 1 - hex_start); *d++ = '\\'; memmove(d, hex_start, hex_size); d += hex_size; break; } - *d++ = ch; + *d++ = static_cast<char>(ch); break; } case 'u': { @@ -184,18 +188,20 @@ bool CUnescapeInternal(absl::string_view source, bool leave_nulls_escaped, if (p + 4 >= end) { if (error) { *error = "\\u must be followed by 4 hex digits: \\" + - std::string(hex_start, p + 1 - hex_start); + std::string(hex_start, + static_cast<size_t>(p + 1 - hex_start)); } return false; } for (int i = 0; i < 4; ++i) { // Look one char ahead. - if (absl::ascii_isxdigit(p[1])) { + if (absl::ascii_isxdigit(static_cast<unsigned char>(p[1]))) { rune = (rune << 4) + hex_digit_to_int(*++p); // Advance p. } else { if (error) { *error = "\\u must be followed by 4 hex digits: \\" + - std::string(hex_start, p + 1 - hex_start); + std::string(hex_start, + static_cast<size_t>(p + 1 - hex_start)); } return false; } @@ -220,20 +226,22 @@ bool CUnescapeInternal(absl::string_view source, bool leave_nulls_escaped, if (p + 8 >= end) { if (error) { *error = "\\U must be followed by 8 hex digits: \\" + - std::string(hex_start, p + 1 - hex_start); + std::string(hex_start, + static_cast<size_t>(p + 1 - hex_start)); } return false; } for (int i = 0; i < 8; ++i) { // Look one char ahead. - if (absl::ascii_isxdigit(p[1])) { + if (absl::ascii_isxdigit(static_cast<unsigned char>(p[1]))) { // Don't change rune until we're sure this // is within the Unicode limit, but do advance p. uint32_t newrune = (rune << 4) + hex_digit_to_int(*++p); if (newrune > 0x10FFFF) { if (error) { *error = "Value of \\" + - std::string(hex_start, p + 1 - hex_start) + + std::string(hex_start, + static_cast<size_t>(p + 1 - hex_start)) + " exceeds Unicode limit (0x10FFFF)"; } return false; @@ -243,7 +251,8 @@ bool CUnescapeInternal(absl::string_view source, bool leave_nulls_escaped, } else { if (error) { *error = "\\U must be followed by 8 hex digits: \\" + - std::string(hex_start, p + 1 - hex_start); + std::string(hex_start, + static_cast<size_t>(p + 1 - hex_start)); } return false; } @@ -291,7 +300,7 @@ bool CUnescapeInternal(absl::string_view source, bool leave_nulls_escaped, error)) { return false; } - dest->erase(dest_size); + dest->erase(static_cast<size_t>(dest_size)); return true; } @@ -311,7 +320,7 @@ std::string CEscapeInternal(absl::string_view src, bool use_hex, std::string dest; bool last_hex_escape = false; // true if last output char was \xNN. - for (unsigned char c : src) { + for (char c : src) { bool is_hex_escape = false; switch (c) { case '\n': dest.append("\\" "n"); break; @@ -320,28 +329,30 @@ std::string CEscapeInternal(absl::string_view src, bool use_hex, case '\"': dest.append("\\" "\""); break; case '\'': dest.append("\\" "'"); break; case '\\': dest.append("\\" "\\"); break; - default: + default: { // Note that if we emit \xNN and the src character after that is a hex // digit then that digit must be escaped too to prevent it being // interpreted as part of the character code by C. - if ((!utf8_safe || c < 0x80) && - (!absl::ascii_isprint(c) || - (last_hex_escape && absl::ascii_isxdigit(c)))) { + const unsigned char uc = static_cast<unsigned char>(c); + if ((!utf8_safe || uc < 0x80) && + (!absl::ascii_isprint(uc) || + (last_hex_escape && absl::ascii_isxdigit(uc)))) { if (use_hex) { dest.append("\\" "x"); - dest.push_back(numbers_internal::kHexChar[c / 16]); - dest.push_back(numbers_internal::kHexChar[c % 16]); + dest.push_back(numbers_internal::kHexChar[uc / 16]); + dest.push_back(numbers_internal::kHexChar[uc % 16]); is_hex_escape = true; } else { dest.append("\\"); - dest.push_back(numbers_internal::kHexChar[c / 64]); - dest.push_back(numbers_internal::kHexChar[(c % 64) / 8]); - dest.push_back(numbers_internal::kHexChar[c % 8]); + dest.push_back(numbers_internal::kHexChar[uc / 64]); + dest.push_back(numbers_internal::kHexChar[(uc % 64) / 8]); + dest.push_back(numbers_internal::kHexChar[uc % 8]); } } else { dest.push_back(c); break; } + } } last_hex_escape = is_hex_escape; } @@ -350,7 +361,7 @@ std::string CEscapeInternal(absl::string_view src, bool use_hex, } /* clang-format off */ -constexpr char c_escaped_len[256] = { +constexpr unsigned char c_escaped_len[256] = { 4, 4, 4, 4, 4, 4, 4, 4, 4, 2, 2, 4, 4, 2, 4, 4, // \t, \n, \r 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 1, 1, 2, 1, 1, 1, 1, 2, 1, 1, 1, 1, 1, 1, 1, 1, // ", ' @@ -375,7 +386,8 @@ constexpr char c_escaped_len[256] = { // that UTF-8 bytes are not handled specially. inline size_t CEscapedLength(absl::string_view src) { size_t escaped_len = 0; - for (unsigned char c : src) escaped_len += c_escaped_len[c]; + for (char c : src) + escaped_len += c_escaped_len[static_cast<unsigned char>(c)]; return escaped_len; } @@ -391,8 +403,8 @@ void CEscapeAndAppendInternal(absl::string_view src, std::string* dest) { cur_dest_len + escaped_len); char* append_ptr = &(*dest)[cur_dest_len]; - for (unsigned char c : src) { - int char_len = c_escaped_len[c]; + for (char c : src) { + size_t char_len = c_escaped_len[static_cast<unsigned char>(c)]; if (char_len == 1) { *append_ptr++ = c; } else if (char_len == 2) { @@ -424,9 +436,9 @@ void CEscapeAndAppendInternal(absl::string_view src, std::string* dest) { } } else { *append_ptr++ = '\\'; - *append_ptr++ = '0' + c / 64; - *append_ptr++ = '0' + (c % 64) / 8; - *append_ptr++ = '0' + c % 8; + *append_ptr++ = '0' + static_cast<unsigned char>(c) / 64; + *append_ptr++ = '0' + (static_cast<unsigned char>(c) % 64) / 8; + *append_ptr++ = '0' + static_cast<unsigned char>(c) % 8; } } } @@ -440,7 +452,7 @@ bool Base64UnescapeInternal(const char* src_param, size_t szsrc, char* dest, size_t destidx = 0; int decode = 0; int state = 0; - unsigned int ch = 0; + unsigned char ch = 0; unsigned int temp = 0; // If "char" is signed by default, using *src as an array index results in @@ -500,13 +512,13 @@ bool Base64UnescapeInternal(const char* src_param, size_t szsrc, char* dest, // how to handle those cases. GET_INPUT(first, 4); - temp = decode; + temp = static_cast<unsigned char>(decode); GET_INPUT(second, 3); - temp = (temp << 6) | decode; + temp = (temp << 6) | static_cast<unsigned char>(decode); GET_INPUT(third, 2); - temp = (temp << 6) | decode; + temp = (temp << 6) | static_cast<unsigned char>(decode); GET_INPUT(fourth, 1); - temp = (temp << 6) | decode; + temp = (temp << 6) | static_cast<unsigned char>(decode); } else { // We really did have four good data bytes, so advance four // characters in the string. @@ -518,11 +530,11 @@ bool Base64UnescapeInternal(const char* src_param, size_t szsrc, char* dest, // temp has 24 bits of input, so write that out as three bytes. if (destidx + 3 > szdest) return false; - dest[destidx + 2] = temp; + dest[destidx + 2] = static_cast<char>(temp); temp >>= 8; - dest[destidx + 1] = temp; + dest[destidx + 1] = static_cast<char>(temp); temp >>= 8; - dest[destidx] = temp; + dest[destidx] = static_cast<char>(temp); destidx += 3; } } else { @@ -583,18 +595,18 @@ bool Base64UnescapeInternal(const char* src_param, size_t szsrc, char* dest, } // Each input character gives us six bits of output. - temp = (temp << 6) | decode; + temp = (temp << 6) | static_cast<unsigned char>(decode); ++state; if (state == 4) { // If we've accumulated 24 bits of output, write that out as // three bytes. if (dest) { if (destidx + 3 > szdest) return false; - dest[destidx + 2] = temp; + dest[destidx + 2] = static_cast<char>(temp); temp >>= 8; - dest[destidx + 1] = temp; + dest[destidx + 1] = static_cast<char>(temp); temp >>= 8; - dest[destidx] = temp; + dest[destidx] = static_cast<char>(temp); } destidx += 3; state = 0; @@ -619,7 +631,7 @@ bool Base64UnescapeInternal(const char* src_param, size_t szsrc, char* dest, if (dest) { if (destidx + 1 > szdest) return false; temp >>= 4; - dest[destidx] = temp; + dest[destidx] = static_cast<char>(temp); } ++destidx; expected_equals = 2; @@ -630,9 +642,9 @@ bool Base64UnescapeInternal(const char* src_param, size_t szsrc, char* dest, if (dest) { if (destidx + 2 > szdest) return false; temp >>= 2; - dest[destidx + 1] = temp; + dest[destidx + 1] = static_cast<char>(temp); temp >>= 8; - dest[destidx] = temp; + dest[destidx] = static_cast<char>(temp); } destidx += 2; expected_equals = 1; @@ -772,8 +784,7 @@ template <typename String> bool Base64UnescapeInternal(const char* src, size_t slen, String* dest, const signed char* unbase64) { // Determine the size of the output string. Base64 encodes every 3 bytes into - // 4 characters. any leftover chars are added directly for good measure. - // This is documented in the base64 RFC: http://tools.ietf.org/html/rfc3548 + // 4 characters. Any leftover chars are added directly for good measure. const size_t dest_len = 3 * (slen / 4) + (slen % 4); strings_internal::STLStringResizeUninitialized(dest, dest_len); @@ -821,9 +832,9 @@ constexpr char kHexValueLenient[256] = { // or a string. This works because we use the [] operator to access // individual characters at a time. template <typename T> -void HexStringToBytesInternal(const char* from, T to, ptrdiff_t num) { - for (int i = 0; i < num; i++) { - to[i] = (kHexValueLenient[from[i * 2] & 0xFF] << 4) + +void HexStringToBytesInternal(const char* from, T to, size_t num) { + for (size_t i = 0; i < num; i++) { + to[i] = static_cast<char>(kHexValueLenient[from[i * 2] & 0xFF] << 4) + (kHexValueLenient[from[i * 2 + 1] & 0xFF]); } } @@ -831,7 +842,7 @@ void HexStringToBytesInternal(const char* from, T to, ptrdiff_t num) { // This is a templated function so that T can be either a char* or a // std::string. template <typename T> -void BytesToHexStringInternal(const unsigned char* src, T dest, ptrdiff_t num) { +void BytesToHexStringInternal(const unsigned char* src, T dest, size_t num) { auto dest_ptr = &dest[0]; for (auto src_ptr = src; src_ptr != (src + num); ++src_ptr, dest_ptr += 2) { const char* hex_p = &numbers_internal::kHexTable[*src_ptr * 2]; @@ -876,8 +887,8 @@ std::string Utf8SafeCHexEscape(absl::string_view src) { // WebSafeBase64Escape() - Google's variation of base64 encoder // // Check out -// http://tools.ietf.org/html/rfc2045 for formal description, but what we -// care about is that... +// https://datatracker.ietf.org/doc/html/rfc2045 for formal description, but +// what we care about is that... // Take the encoded stuff in groups of 4 characters and turn each // character into a code 0 to 63 thus: // A-Z map to 0 to 25 diff --git a/contrib/restricted/abseil-cpp/absl/strings/escaping.h b/contrib/restricted/abseil-cpp/absl/strings/escaping.h index f5ca26c5da..7c082fef3d 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/escaping.h +++ b/contrib/restricted/abseil-cpp/absl/strings/escaping.h @@ -117,35 +117,40 @@ std::string Utf8SafeCEscape(absl::string_view src); // conversion. std::string Utf8SafeCHexEscape(absl::string_view src); -// Base64Unescape() -// -// Converts a `src` string encoded in Base64 to its binary equivalent, writing -// it to a `dest` buffer, returning `true` on success. If `src` contains invalid -// characters, `dest` is cleared and returns `false`. -bool Base64Unescape(absl::string_view src, std::string* dest); - -// WebSafeBase64Unescape() -// -// Converts a `src` string encoded in Base64 to its binary equivalent, writing -// it to a `dest` buffer, but using '-' instead of '+', and '_' instead of '/'. -// If `src` contains invalid characters, `dest` is cleared and returns `false`. -bool WebSafeBase64Unescape(absl::string_view src, std::string* dest); - // Base64Escape() // -// Encodes a `src` string into a base64-encoded string, with padding characters. -// This function conforms with RFC 4648 section 4 (base64). +// Encodes a `src` string into a base64-encoded 'dest' string with padding +// characters. This function conforms with RFC 4648 section 4 (base64) and RFC +// 2045. See also CalculateBase64EscapedLen(). void Base64Escape(absl::string_view src, std::string* dest); std::string Base64Escape(absl::string_view src); // WebSafeBase64Escape() // -// Encodes a `src` string into a base64-like string, using '-' instead of '+' -// and '_' instead of '/', and without padding. This function conforms with RFC -// 4648 section 5 (base64url). +// Encodes a `src` string into a base64 string, like Base64Escape() does, but +// outputs '-' instead of '+' and '_' instead of '/', and does not pad 'dest'. +// This function conforms with RFC 4648 section 5 (base64url). void WebSafeBase64Escape(absl::string_view src, std::string* dest); std::string WebSafeBase64Escape(absl::string_view src); +// Base64Unescape() +// +// Converts a `src` string encoded in Base64 (RFC 4648 section 4) to its binary +// equivalent, writing it to a `dest` buffer, returning `true` on success. If +// `src` contains invalid characters, `dest` is cleared and returns `false`. +// If padding is included (note that `Base64Escape()` does produce it), it must +// be correct. In the padding, '=' and '.' are treated identically. +bool Base64Unescape(absl::string_view src, std::string* dest); + +// WebSafeBase64Unescape() +// +// Converts a `src` string encoded in "web safe" Base64 (RFC 4648 section 5) to +// its binary equivalent, writing it to a `dest` buffer. If `src` contains +// invalid characters, `dest` is cleared and returns `false`. If padding is +// included (note that `WebSafeBase64Escape()` does not produce it), it must be +// correct. In the padding, '=' and '.' are treated identically. +bool WebSafeBase64Unescape(absl::string_view src, std::string* dest); + // HexStringToBytes() // // Converts an ASCII hex string into bytes, returning binary data of length diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/char_map.h b/contrib/restricted/abseil-cpp/absl/strings/internal/char_map.h index 61484de0b7..70a9034336 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/internal/char_map.h +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/char_map.h @@ -73,10 +73,10 @@ class Charmap { } // Containing all the chars in the C-string 's'. - // Note that this is expensively recursive because of the C++11 constexpr - // formulation. Use only in constexpr initializers. static constexpr Charmap FromString(const char* s) { - return *s == 0 ? Charmap() : (Char(*s) | FromString(s + 1)); + Charmap ret; + while (*s) ret = ret | Char(*s++); + return ret; } // Containing all the chars in the closed interval [lo,hi]. @@ -103,10 +103,9 @@ class Charmap { constexpr Charmap(uint64_t b0, uint64_t b1, uint64_t b2, uint64_t b3) : m_{b0, b1, b2, b3} {} - static constexpr uint64_t RangeForWord(unsigned char lo, unsigned char hi, - uint64_t word) { - return OpenRangeFromZeroForWord(hi + 1, word) & - ~OpenRangeFromZeroForWord(lo, word); + static constexpr uint64_t RangeForWord(char lo, char hi, uint64_t word) { + return OpenRangeFromZeroForWord(static_cast<unsigned char>(hi) + 1, word) & + ~OpenRangeFromZeroForWord(static_cast<unsigned char>(lo), word); } // All the chars in the specified word of the range [0, upper). @@ -119,13 +118,16 @@ class Charmap { : (~static_cast<uint64_t>(0) >> (64 - upper % 64)); } - static constexpr uint64_t CharMaskForWord(unsigned char x, uint64_t word) { - return (x / 64 == word) ? (static_cast<uint64_t>(1) << (x % 64)) : 0; + static constexpr uint64_t CharMaskForWord(char x, uint64_t word) { + const auto unsigned_x = static_cast<unsigned char>(x); + return (unsigned_x / 64 == word) + ? (static_cast<uint64_t>(1) << (unsigned_x % 64)) + : 0; } - private: - void SetChar(unsigned char c) { - m_[c / 64] |= static_cast<uint64_t>(1) << (c % 64); + void SetChar(char c) { + const auto unsigned_c = static_cast<unsigned char>(c); + m_[unsigned_c / 64] |= static_cast<uint64_t>(1) << (unsigned_c % 64); } uint64_t m_[4]; diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/charconv_bigint.cc b/contrib/restricted/abseil-cpp/absl/strings/internal/charconv_bigint.cc index ebf8c0791a..282b639eb2 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/internal/charconv_bigint.cc +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/charconv_bigint.cc @@ -242,7 +242,7 @@ int BigUnsigned<max_words>::ReadDigits(const char* begin, const char* end, // decimal exponent to compensate. --exponent_adjust; } - int digit = (*begin - '0'); + char digit = (*begin - '0'); --significant_digits; if (significant_digits == 0 && std::next(begin) != end && (digit == 0 || digit == 5)) { @@ -255,7 +255,7 @@ int BigUnsigned<max_words>::ReadDigits(const char* begin, const char* end, // 500000...000000000001 to correctly round up, rather than to nearest. ++digit; } - queued = 10 * queued + digit; + queued = 10 * queued + static_cast<uint32_t>(digit); ++digits_queued; if (digits_queued == kMaxSmallPowerOfTen) { MultiplyBy(kTenToNth[kMaxSmallPowerOfTen]); @@ -341,8 +341,8 @@ std::string BigUnsigned<max_words>::ToString() const { std::string result; // Build result in reverse order while (copy.size() > 0) { - int next_digit = copy.DivMod<10>(); - result.push_back('0' + next_digit); + uint32_t next_digit = copy.DivMod<10>(); + result.push_back('0' + static_cast<char>(next_digit)); } if (result.empty()) { result.push_back('0'); diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/charconv_parse.cc b/contrib/restricted/abseil-cpp/absl/strings/internal/charconv_parse.cc index d29acaf462..98823def83 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/internal/charconv_parse.cc +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/charconv_parse.cc @@ -190,11 +190,11 @@ bool IsDigit<16>(char ch) { template <> unsigned ToDigit<10>(char ch) { - return ch - '0'; + return static_cast<unsigned>(ch - '0'); } template <> unsigned ToDigit<16>(char ch) { - return kAsciiToInt[static_cast<unsigned char>(ch)]; + return static_cast<unsigned>(kAsciiToInt[static_cast<unsigned char>(ch)]); } template <> diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/cord_internal.h b/contrib/restricted/abseil-cpp/absl/strings/internal/cord_internal.h index b50fb79a55..63a81f4f16 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/internal/cord_internal.h +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/cord_internal.h @@ -27,9 +27,20 @@ #include "absl/base/internal/invoke.h" #include "absl/base/optimization.h" #include "absl/container/internal/compressed_tuple.h" +#include "absl/container/internal/container_memory.h" #include "absl/meta/type_traits.h" #include "absl/strings/string_view.h" +// We can only add poisoning if we can detect consteval executions. +#if defined(ABSL_HAVE_CONSTANT_EVALUATED) && \ + (defined(ABSL_HAVE_ADDRESS_SANITIZER) || \ + defined(ABSL_HAVE_MEMORY_SANITIZER)) +#define ABSL_INTERNAL_CORD_HAVE_SANITIZER 1 +#endif + +#define ABSL_CORD_INTERNAL_NO_SANITIZE \ + ABSL_ATTRIBUTE_NO_SANITIZE_ADDRESS ABSL_ATTRIBUTE_NO_SANITIZE_MEMORY + namespace absl { ABSL_NAMESPACE_BEGIN namespace cord_internal { @@ -91,6 +102,46 @@ enum Constants { // Emits a fatal error "Unexpected node type: xyz" and aborts the program. ABSL_ATTRIBUTE_NORETURN void LogFatalNodeType(CordRep* rep); +// Fast implementation of memmove for up to 15 bytes. This implementation is +// safe for overlapping regions. If nullify_tail is true, the destination is +// padded with '\0' up to 15 bytes. +template <bool nullify_tail = false> +inline void SmallMemmove(char* dst, const char* src, size_t n) { + if (n >= 8) { + assert(n <= 15); + uint64_t buf1; + uint64_t buf2; + memcpy(&buf1, src, 8); + memcpy(&buf2, src + n - 8, 8); + if (nullify_tail) { + memset(dst + 7, 0, 8); + } + memcpy(dst, &buf1, 8); + memcpy(dst + n - 8, &buf2, 8); + } else if (n >= 4) { + uint32_t buf1; + uint32_t buf2; + memcpy(&buf1, src, 4); + memcpy(&buf2, src + n - 4, 4); + if (nullify_tail) { + memset(dst + 4, 0, 4); + memset(dst + 7, 0, 8); + } + memcpy(dst, &buf1, 4); + memcpy(dst + n - 4, &buf2, 4); + } else { + if (n != 0) { + dst[0] = src[0]; + dst[n / 2] = src[n / 2]; + dst[n - 1] = src[n - 1]; + } + if (nullify_tail) { + memset(dst + 7, 0, 8); + memset(dst + n, 0, 8); + } + } +} + // Compact class for tracking the reference count and state flags for CordRep // instances. Data is stored in an atomic int32_t for compactness and speed. class RefcountAndFlags { @@ -129,8 +180,9 @@ class RefcountAndFlags { } // Returns the current reference count using acquire semantics. - inline int32_t Get() const { - return count_.load(std::memory_order_acquire) >> kNumFlags; + inline size_t Get() const { + return static_cast<size_t>(count_.load(std::memory_order_acquire) >> + kNumFlags); } // Returns whether the atomic integer is 1. @@ -224,7 +276,11 @@ struct CordRep { : length(l), refcount(immortal), tag(EXTERNAL), storage{} {} // The following three fields have to be less than 32 bytes since - // that is the smallest supported flat node size. + // that is the smallest supported flat node size. Some code optimizations rely + // on the specific layout of these fields. Notably: the non-trivial field + // `refcount` being preceded by `length`, and being tailed by POD data + // members only. + // # LINT.IfChange size_t length; RefcountAndFlags refcount; // If tag < FLAT, it represents CordRepKind and indicates the type of node. @@ -240,6 +296,7 @@ struct CordRep { // allocate room for these in the derived class, as not all compilers reuse // padding space from the base class (clang and gcc do, MSVC does not, etc) uint8_t storage[3]; + // # LINT.ThenChange(cord_rep_btree.h:copy_raw) // Returns true if this instance's tag matches the requested type. constexpr bool IsRing() const { return tag == RING; } @@ -422,25 +479,25 @@ constexpr char GetOrNull(absl::string_view data, size_t pos) { return pos < data.size() ? data[pos] : '\0'; } -// We store cordz_info as 64 bit pointer value in big endian format. This -// guarantees that the least significant byte of cordz_info matches the last -// byte of the inline data representation in as_chars_, which holds the inlined +// We store cordz_info as 64 bit pointer value in little endian format. This +// guarantees that the least significant byte of cordz_info matches the first +// byte of the inline data representation in `data`, which holds the inlined // size or the 'is_tree' bit. using cordz_info_t = int64_t; // Assert that the `cordz_info` pointer value perfectly overlaps the last half -// of `as_chars_` and can hold a pointer value. +// of `data` and can hold a pointer value. static_assert(sizeof(cordz_info_t) * 2 == kMaxInline + 1, ""); static_assert(sizeof(cordz_info_t) >= sizeof(intptr_t), ""); -// BigEndianByte() creates a big endian representation of 'value', i.e.: a big -// endian value where the last byte in the host's representation holds 'value`, -// with all other bytes being 0. -static constexpr cordz_info_t BigEndianByte(unsigned char value) { +// LittleEndianByte() creates a little endian representation of 'value', i.e.: +// a little endian value where the first byte in the host's representation +// holds 'value`, with all other bytes being 0. +static constexpr cordz_info_t LittleEndianByte(unsigned char value) { #if defined(ABSL_IS_BIG_ENDIAN) - return value; -#else return static_cast<cordz_info_t>(value) << ((sizeof(cordz_info_t) - 1) * 8); +#else + return value; #endif } @@ -449,38 +506,80 @@ class InlineData { // DefaultInitType forces the use of the default initialization constructor. enum DefaultInitType { kDefaultInit }; - // kNullCordzInfo holds the big endian representation of intptr_t(1) + // kNullCordzInfo holds the little endian representation of intptr_t(1) // This is the 'null' / initial value of 'cordz_info'. The null value // is specifically big endian 1 as with 64-bit pointers, the last // byte of cordz_info overlaps with the last byte holding the tag. - static constexpr cordz_info_t kNullCordzInfo = BigEndianByte(1); - - constexpr InlineData() : as_chars_{0} {} - explicit InlineData(DefaultInitType) {} - explicit constexpr InlineData(CordRep* rep) : as_tree_(rep) {} - explicit constexpr InlineData(absl::string_view chars) - : as_chars_{ - GetOrNull(chars, 0), GetOrNull(chars, 1), - GetOrNull(chars, 2), GetOrNull(chars, 3), - GetOrNull(chars, 4), GetOrNull(chars, 5), - GetOrNull(chars, 6), GetOrNull(chars, 7), - GetOrNull(chars, 8), GetOrNull(chars, 9), - GetOrNull(chars, 10), GetOrNull(chars, 11), - GetOrNull(chars, 12), GetOrNull(chars, 13), - GetOrNull(chars, 14), static_cast<char>((chars.size() << 1))} {} + static constexpr cordz_info_t kNullCordzInfo = LittleEndianByte(1); + + // kTagOffset contains the offset of the control byte / tag. This constant is + // intended mostly for debugging purposes: do not remove this constant as it + // is actively inspected and used by gdb pretty printing code. + static constexpr size_t kTagOffset = 0; + + // Implement `~InlineData()` conditionally: we only need this destructor to + // unpoison poisoned instances under *SAN, and it will only compile correctly + // if the current compiler supports `absl::is_constant_evaluated()`. +#ifdef ABSL_INTERNAL_CORD_HAVE_SANITIZER + ~InlineData() noexcept { unpoison(); } +#endif + + constexpr InlineData() noexcept { poison_this(); } + + explicit InlineData(DefaultInitType) noexcept : rep_(kDefaultInit) { + poison_this(); + } + + explicit InlineData(CordRep* rep) noexcept : rep_(rep) { + ABSL_ASSERT(rep != nullptr); + } + + // Explicit constexpr constructor to create a constexpr InlineData + // value. Creates an inlined SSO value if `rep` is null, otherwise + // creates a tree instance value. + constexpr InlineData(absl::string_view sv, CordRep* rep) noexcept + : rep_(rep ? Rep(rep) : Rep(sv)) { + poison(); + } + + constexpr InlineData(const InlineData& rhs) noexcept; + InlineData& operator=(const InlineData& rhs) noexcept; + + friend bool operator==(const InlineData& lhs, const InlineData& rhs) { +#ifdef ABSL_INTERNAL_CORD_HAVE_SANITIZER + const Rep l = lhs.rep_.SanitizerSafeCopy(); + const Rep r = rhs.rep_.SanitizerSafeCopy(); + return memcmp(&l, &r, sizeof(l)) == 0; +#else + return memcmp(&lhs, &rhs, sizeof(lhs)) == 0; +#endif + } + friend bool operator!=(const InlineData& lhs, const InlineData& rhs) { + return !operator==(lhs, rhs); + } + + // Poisons the unused inlined SSO data if the current instance + // is inlined, else un-poisons the entire instance. + constexpr void poison(); + + // Un-poisons this instance. + constexpr void unpoison(); + + // Poisons the current instance. This is used on default initialization. + constexpr void poison_this(); // Returns true if the current instance is empty. // The 'empty value' is an inlined data value of zero length. - bool is_empty() const { return tag() == 0; } + bool is_empty() const { return rep_.tag() == 0; } // Returns true if the current instance holds a tree value. - bool is_tree() const { return (tag() & 1) != 0; } + bool is_tree() const { return (rep_.tag() & 1) != 0; } // Returns true if the current instance holds a cordz_info value. // Requires the current instance to hold a tree value. bool is_profiled() const { assert(is_tree()); - return as_tree_.cordz_info != kNullCordzInfo; + return rep_.cordz_info() != kNullCordzInfo; } // Returns true if either of the provided instances hold a cordz_info value. @@ -489,7 +588,7 @@ class InlineData { static bool is_either_profiled(const InlineData& data1, const InlineData& data2) { assert(data1.is_tree() && data2.is_tree()); - return (data1.as_tree_.cordz_info | data2.as_tree_.cordz_info) != + return (data1.rep_.cordz_info() | data2.rep_.cordz_info()) != kNullCordzInfo; } @@ -498,8 +597,8 @@ class InlineData { // Requires the current instance to hold a tree value. CordzInfo* cordz_info() const { assert(is_tree()); - intptr_t info = static_cast<intptr_t>( - absl::big_endian::ToHost64(static_cast<uint64_t>(as_tree_.cordz_info))); + intptr_t info = static_cast<intptr_t>(absl::little_endian::ToHost64( + static_cast<uint64_t>(rep_.cordz_info()))); assert(info & 1); return reinterpret_cast<CordzInfo*>(info - 1); } @@ -510,21 +609,21 @@ class InlineData { void set_cordz_info(CordzInfo* cordz_info) { assert(is_tree()); uintptr_t info = reinterpret_cast<uintptr_t>(cordz_info) | 1; - as_tree_.cordz_info = - static_cast<cordz_info_t>(absl::big_endian::FromHost64(info)); + rep_.set_cordz_info( + static_cast<cordz_info_t>(absl::little_endian::FromHost64(info))); } // Resets the current cordz_info to null / empty. void clear_cordz_info() { assert(is_tree()); - as_tree_.cordz_info = kNullCordzInfo; + rep_.set_cordz_info(kNullCordzInfo); } // Returns a read only pointer to the character data inside this instance. // Requires the current instance to hold inline data. const char* as_chars() const { assert(!is_tree()); - return as_chars_; + return rep_.as_chars(); } // Returns a mutable pointer to the character data inside this instance. @@ -542,20 +641,33 @@ class InlineData { // // It's an error to read from the returned pointer without a preceding write // if the current instance does not hold inline data, i.e.: is_tree() == true. - char* as_chars() { return as_chars_; } + char* as_chars() { return rep_.as_chars(); } // Returns the tree value of this value. // Requires the current instance to hold a tree value. CordRep* as_tree() const { assert(is_tree()); - return as_tree_.rep; + return rep_.tree(); + } + + void set_inline_data(const char* data, size_t n) { + ABSL_ASSERT(n <= kMaxInline); + unpoison(); + rep_.set_tag(static_cast<int8_t>(n << 1)); + SmallMemmove<true>(rep_.as_chars(), data, n); + poison(); + } + + void copy_max_inline_to(char* dst) const { + assert(!is_tree()); + memcpy(dst, rep_.SanitizerSafeCopy().as_chars(), kMaxInline); } // Initialize this instance to holding the tree value `rep`, // initializing the cordz_info to null, i.e.: 'not profiled'. void make_tree(CordRep* rep) { - as_tree_.rep = rep; - as_tree_.cordz_info = kNullCordzInfo; + unpoison(); + rep_.make_tree(rep); } // Set the tree value of this instance to 'rep`. @@ -563,54 +675,198 @@ class InlineData { // Does not affect the value of cordz_info. void set_tree(CordRep* rep) { assert(is_tree()); - as_tree_.rep = rep; + rep_.set_tree(rep); } // Returns the size of the inlined character data inside this instance. // Requires the current instance to hold inline data. - size_t inline_size() const { - assert(!is_tree()); - return tag() >> 1; - } + size_t inline_size() const { return rep_.inline_size(); } // Sets the size of the inlined character data inside this instance. // Requires `size` to be <= kMaxInline. // See the documentation on 'as_chars()' for more information and examples. void set_inline_size(size_t size) { - ABSL_ASSERT(size <= kMaxInline); - tag() = static_cast<char>(size << 1); + unpoison(); + rep_.set_inline_size(size); + poison(); + } + + // Compares 'this' inlined data with rhs. The comparison is a straightforward + // lexicographic comparison. `Compare()` returns values as follows: + // + // -1 'this' InlineData instance is smaller + // 0 the InlineData instances are equal + // 1 'this' InlineData instance larger + int Compare(const InlineData& rhs) const { + return Compare(rep_.SanitizerSafeCopy(), rhs.rep_.SanitizerSafeCopy()); } private: - // See cordz_info_t for forced alignment and size of `cordz_info` details. - struct AsTree { - explicit constexpr AsTree(absl::cord_internal::CordRep* tree) - : rep(tree), cordz_info(kNullCordzInfo) {} - // This union uses up extra space so that whether rep is 32 or 64 bits, - // cordz_info will still start at the eighth byte, and the last - // byte of cordz_info will still be the last byte of InlineData. - union { + struct Rep { + // See cordz_info_t for forced alignment and size of `cordz_info` details. + struct AsTree { + explicit constexpr AsTree(absl::cord_internal::CordRep* tree) + : rep(tree) {} + cordz_info_t cordz_info = kNullCordzInfo; absl::cord_internal::CordRep* rep; - cordz_info_t unused_aligner; }; - cordz_info_t cordz_info; - }; - char& tag() { return reinterpret_cast<char*>(this)[kMaxInline]; } - char tag() const { return reinterpret_cast<const char*>(this)[kMaxInline]; } + explicit Rep(DefaultInitType) {} + constexpr Rep() : data{0} {} + constexpr Rep(const Rep&) = default; + constexpr Rep& operator=(const Rep&) = default; + + explicit constexpr Rep(CordRep* rep) : as_tree(rep) {} + + explicit constexpr Rep(absl::string_view chars) + : data{static_cast<char>((chars.size() << 1)), + GetOrNull(chars, 0), + GetOrNull(chars, 1), + GetOrNull(chars, 2), + GetOrNull(chars, 3), + GetOrNull(chars, 4), + GetOrNull(chars, 5), + GetOrNull(chars, 6), + GetOrNull(chars, 7), + GetOrNull(chars, 8), + GetOrNull(chars, 9), + GetOrNull(chars, 10), + GetOrNull(chars, 11), + GetOrNull(chars, 12), + GetOrNull(chars, 13), + GetOrNull(chars, 14)} {} + + // Disable sanitizer as we must always be able to read `tag`. + ABSL_CORD_INTERNAL_NO_SANITIZE + int8_t tag() const { return reinterpret_cast<const int8_t*>(this)[0]; } + void set_tag(int8_t rhs) { reinterpret_cast<int8_t*>(this)[0] = rhs; } + + char* as_chars() { return data + 1; } + const char* as_chars() const { return data + 1; } + + bool is_tree() const { return (tag() & 1) != 0; } + + size_t inline_size() const { + ABSL_ASSERT(!is_tree()); + return static_cast<size_t>(tag()) >> 1; + } + + void set_inline_size(size_t size) { + ABSL_ASSERT(size <= kMaxInline); + set_tag(static_cast<int8_t>(size << 1)); + } + + CordRep* tree() const { return as_tree.rep; } + void set_tree(CordRep* rhs) { as_tree.rep = rhs; } + + cordz_info_t cordz_info() const { return as_tree.cordz_info; } + void set_cordz_info(cordz_info_t rhs) { as_tree.cordz_info = rhs; } + + void make_tree(CordRep* tree) { + as_tree.rep = tree; + as_tree.cordz_info = kNullCordzInfo; + } + +#ifdef ABSL_INTERNAL_CORD_HAVE_SANITIZER + Rep SanitizerSafeCopy() const { + Rep res; + if (is_tree()) { + res = *this; + } else { + res.set_tag(tag()); + memcpy(res.as_chars(), as_chars(), inline_size()); + } + return res; + } +#else + const Rep& SanitizerSafeCopy() const { return *this; } +#endif - // If the data has length <= kMaxInline, we store it in `as_chars_`, and - // store the size in the last char of `as_chars_` shifted left + 1. - // Else we store it in a tree and store a pointer to that tree in - // `as_tree_.rep` and store a tag in `tagged_size`. - union { - char as_chars_[kMaxInline + 1]; - AsTree as_tree_; + // If the data has length <= kMaxInline, we store it in `data`, and + // store the size in the first char of `data` shifted left + 1. + // Else we store it in a tree and store a pointer to that tree in + // `as_tree.rep` with a tagged pointer to make `tag() & 1` non zero. + union { + char data[kMaxInline + 1]; + AsTree as_tree; + }; }; + + // Private implementation of `Compare()` + static inline int Compare(const Rep& lhs, const Rep& rhs) { + uint64_t x, y; + memcpy(&x, lhs.as_chars(), sizeof(x)); + memcpy(&y, rhs.as_chars(), sizeof(y)); + if (x == y) { + memcpy(&x, lhs.as_chars() + 7, sizeof(x)); + memcpy(&y, rhs.as_chars() + 7, sizeof(y)); + if (x == y) { + if (lhs.inline_size() == rhs.inline_size()) return 0; + return lhs.inline_size() < rhs.inline_size() ? -1 : 1; + } + } + x = absl::big_endian::FromHost64(x); + y = absl::big_endian::FromHost64(y); + return x < y ? -1 : 1; + } + + Rep rep_; }; static_assert(sizeof(InlineData) == kMaxInline + 1, ""); +#ifdef ABSL_INTERNAL_CORD_HAVE_SANITIZER + +constexpr InlineData::InlineData(const InlineData& rhs) noexcept + : rep_(rhs.rep_.SanitizerSafeCopy()) { + poison(); +} + +inline InlineData& InlineData::operator=(const InlineData& rhs) noexcept { + unpoison(); + rep_ = rhs.rep_.SanitizerSafeCopy(); + poison(); + return *this; +} + +constexpr void InlineData::poison_this() { + if (!absl::is_constant_evaluated()) { + container_internal::SanitizerPoisonObject(this); + } +} + +constexpr void InlineData::unpoison() { + if (!absl::is_constant_evaluated()) { + container_internal::SanitizerUnpoisonObject(this); + } +} + +constexpr void InlineData::poison() { + if (!absl::is_constant_evaluated()) { + if (is_tree()) { + container_internal::SanitizerUnpoisonObject(this); + } else if (const size_t size = inline_size()) { + if (size < kMaxInline) { + const char* end = rep_.as_chars() + size; + container_internal::SanitizerPoisonMemoryRegion(end, kMaxInline - size); + } + } else { + container_internal::SanitizerPoisonObject(this); + } + } +} + +#else // ABSL_INTERNAL_CORD_HAVE_SANITIZER + +constexpr InlineData::InlineData(const InlineData&) noexcept = default; +inline InlineData& InlineData::operator=(const InlineData&) noexcept = default; + +constexpr void InlineData::poison_this() {} +constexpr void InlineData::unpoison() {} +constexpr void InlineData::poison() {} + +#endif // ABSL_INTERNAL_CORD_HAVE_SANITIZER + inline CordRepSubstring* CordRep::substring() { assert(IsSubstring()); return static_cast<CordRepSubstring*>(this); diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/cord_rep_btree.cc b/contrib/restricted/abseil-cpp/absl/strings/internal/cord_rep_btree.cc index cacbf3da30..a86fdc0b4e 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/internal/cord_rep_btree.cc +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/cord_rep_btree.cc @@ -17,11 +17,13 @@ #include <cassert> #include <cstdint> #include <iostream> +#include <ostream> #include <string> #include "absl/base/attributes.h" #include "absl/base/config.h" #include "absl/base/internal/raw_logging.h" +#include "absl/base/optimization.h" #include "absl/strings/internal/cord_data_edge.h" #include "absl/strings/internal/cord_internal.h" #include "absl/strings/internal/cord_rep_consume.h" @@ -55,8 +57,10 @@ inline bool exhaustive_validation() { // Prints the entire tree structure or 'rep'. External callers should // not specify 'depth' and leave it to its default (0) value. // Rep may be a CordRepBtree tree, or a SUBSTRING / EXTERNAL / FLAT node. -void DumpAll(const CordRep* rep, bool include_contents, std::ostream& stream, - int depth = 0) { +void DumpAll(const CordRep* rep, + bool include_contents, + std::ostream& stream, + size_t depth = 0) { // Allow for full height trees + substring -> flat / external nodes. assert(depth <= CordRepBtree::kMaxDepth + 2); std::string sharing = const_cast<CordRep*>(rep)->refcount.IsOne() @@ -283,7 +287,7 @@ struct StackOperations { case CordRepBtree::kSelf: return result.tree; } - ABSL_INTERNAL_UNREACHABLE; + ABSL_UNREACHABLE(); return result.tree; } @@ -499,7 +503,7 @@ OpResult CordRepBtree::SetEdge(bool owned, CordRep* edge, size_t delta) { // open interval [begin, back) or [begin + 1, end) depending on `edge_type`. // We conveniently cover both case using a constexpr `shift` being 0 or 1 // as `end :== back + 1`. - result = {CopyRaw(), kCopied}; + result = {CopyRaw(length), kCopied}; constexpr int shift = edge_type == kFront ? 1 : 0; for (CordRep* r : Edges(begin() + shift, back() + shift)) { CordRep::Ref(r); diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/cord_rep_btree.h b/contrib/restricted/abseil-cpp/absl/strings/internal/cord_rep_btree.h index 2cbc09eca1..4209e51256 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/internal/cord_rep_btree.h +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/cord_rep_btree.h @@ -95,8 +95,9 @@ class CordRepBtree : public CordRep { // local stack variable compared to Cord's current near 400 bytes stack use. // The maximum `height` value of a node is then `kMaxDepth - 1` as node height // values start with a value of 0 for leaf nodes. - static constexpr int kMaxDepth = 12; - static constexpr int kMaxHeight = kMaxDepth - 1; + static constexpr size_t kMaxDepth = 12; + // See comments on height() for why this is an int and not a size_t. + static constexpr int kMaxHeight = static_cast<int>(kMaxDepth - 1); // `Action` defines the action for unwinding changes done at the btree's leaf // level that need to be propagated up to the parent node(s). Each operation @@ -445,9 +446,9 @@ class CordRepBtree : public CordRep { template <EdgeType edge_type> static CordRepBtree* NewLeaf(absl::string_view data, size_t extra); - // Creates a raw copy of this Btree node, copying all properties, but - // without adding any references to existing edges. - CordRepBtree* CopyRaw() const; + // Creates a raw copy of this Btree node with the specified length, copying + // all properties, but without adding any references to existing edges. + CordRepBtree* CopyRaw(size_t new_length) const; // Creates a full copy of this Btree node, adding a reference on all edges. CordRepBtree* Copy() const; @@ -665,15 +666,28 @@ inline void CordRepBtree::Unref(absl::Span<CordRep* const> edges) { } } -inline CordRepBtree* CordRepBtree::CopyRaw() const { - auto* tree = static_cast<CordRepBtree*>(::operator new(sizeof(CordRepBtree))); - memcpy(static_cast<void*>(tree), this, sizeof(CordRepBtree)); - new (&tree->refcount) RefcountAndFlags; +inline CordRepBtree* CordRepBtree::CopyRaw(size_t new_length) const { + CordRepBtree* tree = new CordRepBtree; + + // `length` and `refcount` are the first members of `CordRepBtree`. + // We initialize `length` using the given length, have `refcount` be set to + // ref = 1 through its default constructor, and copy all data beyond + // 'refcount' which starts with `tag` using a single memcpy: all contents + // except `refcount` is trivially copyable, and the compiler does not + // efficiently coalesce member-wise copy of these members. + // See https://gcc.godbolt.org/z/qY8zsca6z + // # LINT.IfChange(copy_raw) + tree->length = new_length; + uint8_t* dst = &tree->tag; + const uint8_t* src = &tag; + const ptrdiff_t offset = src - reinterpret_cast<const uint8_t*>(this); + memcpy(dst, src, sizeof(CordRepBtree) - static_cast<size_t>(offset)); return tree; + // # LINT.ThenChange() } inline CordRepBtree* CordRepBtree::Copy() const { - CordRepBtree* tree = CopyRaw(); + CordRepBtree* tree = CopyRaw(length); for (CordRep* rep : Edges()) CordRep::Ref(rep); return tree; } @@ -682,8 +696,7 @@ inline CordRepBtree* CordRepBtree::CopyToEndFrom(size_t begin, size_t new_length) const { assert(begin >= this->begin()); assert(begin <= this->end()); - CordRepBtree* tree = CopyRaw(); - tree->length = new_length; + CordRepBtree* tree = CopyRaw(new_length); tree->set_begin(begin); for (CordRep* edge : tree->Edges()) CordRep::Ref(edge); return tree; @@ -693,8 +706,7 @@ inline CordRepBtree* CordRepBtree::CopyBeginTo(size_t end, size_t new_length) const { assert(end <= capacity()); assert(end >= this->begin()); - CordRepBtree* tree = CopyRaw(); - tree->length = new_length; + CordRepBtree* tree = CopyRaw(new_length); tree->set_end(end); for (CordRep* edge : tree->Edges()) CordRep::Ref(edge); return tree; diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/cord_rep_btree_navigator.cc b/contrib/restricted/abseil-cpp/absl/strings/internal/cord_rep_btree_navigator.cc index 9b896a3d09..6ed20c23a7 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/internal/cord_rep_btree_navigator.cc +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/cord_rep_btree_navigator.cc @@ -90,7 +90,7 @@ CordRepBtreeNavigator::Position CordRepBtreeNavigator::Skip(size_t n) { // edges that must be skipped. while (height > 0) { node = edge->btree(); - index_[height] = index; + index_[height] = static_cast<uint8_t>(index); node_[--height] = node; index = node->begin(); edge = node->Edge(index); @@ -101,7 +101,7 @@ CordRepBtreeNavigator::Position CordRepBtreeNavigator::Skip(size_t n) { edge = node->Edge(index); } } - index_[0] = index; + index_[0] = static_cast<uint8_t>(index); return {edge, n}; } @@ -126,7 +126,7 @@ ReadResult CordRepBtreeNavigator::Read(size_t edge_offset, size_t n) { do { length -= edge->length; while (++index == node->end()) { - index_[height] = index; + index_[height] = static_cast<uint8_t>(index); if (++height > height_) { subtree->set_end(subtree_end); if (length == 0) return {subtree, 0}; @@ -154,7 +154,7 @@ ReadResult CordRepBtreeNavigator::Read(size_t edge_offset, size_t n) { // edges that must be read, adding 'down' nodes to `subtree`. while (height > 0) { node = edge->btree(); - index_[height] = index; + index_[height] = static_cast<uint8_t>(index); node_[--height] = node; index = node->begin(); edge = node->Edge(index); @@ -178,7 +178,7 @@ ReadResult CordRepBtreeNavigator::Read(size_t edge_offset, size_t n) { subtree->edges_[subtree_end++] = Substring(edge, 0, length); } subtree->set_end(subtree_end); - index_[0] = index; + index_[0] = static_cast<uint8_t>(index); return {tree, length}; } diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/cord_rep_crc.cc b/contrib/restricted/abseil-cpp/absl/strings/internal/cord_rep_crc.cc index ee14035410..dbe54cc400 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/internal/cord_rep_crc.cc +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/cord_rep_crc.cc @@ -16,6 +16,7 @@ #include <cassert> #include <cstdint> +#include <utility> #include "absl/base/config.h" #include "absl/strings/internal/cord_internal.h" @@ -24,11 +25,10 @@ namespace absl { ABSL_NAMESPACE_BEGIN namespace cord_internal { -CordRepCrc* CordRepCrc::New(CordRep* child, uint32_t crc) { - assert(child != nullptr); - if (child->IsCrc()) { +CordRepCrc* CordRepCrc::New(CordRep* child, crc_internal::CrcCordState state) { + if (child != nullptr && child->IsCrc()) { if (child->refcount.IsOne()) { - child->crc()->crc = crc; + child->crc()->crc_cord_state = std::move(state); return child->crc(); } CordRep* old = child; @@ -37,15 +37,17 @@ CordRepCrc* CordRepCrc::New(CordRep* child, uint32_t crc) { CordRep::Unref(old); } auto* new_cordrep = new CordRepCrc; - new_cordrep->length = child->length; + new_cordrep->length = child != nullptr ? child->length : 0; new_cordrep->tag = cord_internal::CRC; new_cordrep->child = child; - new_cordrep->crc = crc; + new_cordrep->crc_cord_state = std::move(state); return new_cordrep; } void CordRepCrc::Destroy(CordRepCrc* node) { - CordRep::Unref(node->child); + if (node->child != nullptr) { + CordRep::Unref(node->child); + } delete node; } diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/cord_rep_crc.h b/contrib/restricted/abseil-cpp/absl/strings/internal/cord_rep_crc.h index 5294b0d133..379d7a604e 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/internal/cord_rep_crc.h +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/cord_rep_crc.h @@ -20,6 +20,7 @@ #include "absl/base/config.h" #include "absl/base/optimization.h" +#include "absl/crc/internal/crc_cord_state.h" #include "absl/strings/internal/cord_internal.h" namespace absl { @@ -34,14 +35,14 @@ namespace cord_internal { // the contained checksum is the user's responsibility. struct CordRepCrc : public CordRep { CordRep* child; - uint32_t crc; + absl::crc_internal::CrcCordState crc_cord_state; // Consumes `child` and returns a CordRepCrc prefixed tree containing `child`. // If the specified `child` is itself a CordRepCrc node, then this method - // either replaces the existing node, or directly updates the crc value in it + // either replaces the existing node, or directly updates the crc state in it // depending on the node being shared or not, i.e.: refcount.IsOne(). - // `child` must not be null. Never returns null. - static CordRepCrc* New(CordRep* child, uint32_t crc); + // `child` must only be null if the Cord is empty. Never returns null. + static CordRepCrc* New(CordRep* child, crc_internal::CrcCordState state); // Destroys (deletes) the provided node. `node` must not be null. static void Destroy(CordRepCrc* node); diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/cordz_functions.h b/contrib/restricted/abseil-cpp/absl/strings/internal/cordz_functions.h index c9ba14508a..ed108bf1b4 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/internal/cordz_functions.h +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/cordz_functions.h @@ -12,8 +12,8 @@ // See the License for the specific language governing permissions and // limitations under the License. -#ifndef ABSL_STRINGS_CORDZ_FUNCTIONS_H_ -#define ABSL_STRINGS_CORDZ_FUNCTIONS_H_ +#ifndef ABSL_STRINGS_INTERNAL_CORDZ_FUNCTIONS_H_ +#define ABSL_STRINGS_INTERNAL_CORDZ_FUNCTIONS_H_ #include <stdint.h> @@ -32,18 +32,10 @@ int32_t get_cordz_mean_interval(); // Sets the sample rate with the average interval between samples. void set_cordz_mean_interval(int32_t mean_interval); -// Enable cordz unless any of the following applies: -// - no thread local support -// - MSVC build -// - Android build -// - Apple build -// - DLL build -// Hashtablez is turned off completely in opensource builds. -// MSVC's static atomics are dynamically initialized in debug mode, which breaks -// sampling. -#if defined(ABSL_HAVE_THREAD_LOCAL) && !defined(_MSC_VER) && \ - !defined(ABSL_BUILD_DLL) && !defined(ABSL_CONSUME_DLL) && \ - !defined(__ANDROID__) && !defined(__APPLE__) +// Cordz is only enabled on Linux with thread_local support. +#if defined(ABSL_INTERNAL_CORDZ_ENABLED) +#error ABSL_INTERNAL_CORDZ_ENABLED cannot be set directly +#elif defined(__linux__) && defined(ABSL_HAVE_THREAD_LOCAL) #define ABSL_INTERNAL_CORDZ_ENABLED 1 #endif @@ -82,4 +74,4 @@ inline void cordz_set_next_sample_for_testing(int64_t) {} ABSL_NAMESPACE_END } // namespace absl -#endif // ABSL_STRINGS_CORDZ_FUNCTIONS_H_ +#endif // ABSL_STRINGS_INTERNAL_CORDZ_FUNCTIONS_H_ diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/cordz_handle.h b/contrib/restricted/abseil-cpp/absl/strings/internal/cordz_handle.h index 5df53c782a..3c800b433f 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/internal/cordz_handle.h +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/cordz_handle.h @@ -12,8 +12,8 @@ // See the License for the specific language governing permissions and // limitations under the License. -#ifndef ABSL_STRINGS_CORDZ_HANDLE_H_ -#define ABSL_STRINGS_CORDZ_HANDLE_H_ +#ifndef ABSL_STRINGS_INTERNAL_CORDZ_HANDLE_H_ +#define ABSL_STRINGS_INTERNAL_CORDZ_HANDLE_H_ #include <atomic> #include <vector> @@ -128,4 +128,4 @@ class CordzSnapshot : public CordzHandle { ABSL_NAMESPACE_END } // namespace absl -#endif // ABSL_STRINGS_CORDZ_HANDLE_H_ +#endif // ABSL_STRINGS_INTERNAL_CORDZ_HANDLE_H_ diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/cordz_info.cc b/contrib/restricted/abseil-cpp/absl/strings/internal/cordz_info.cc index dac3fd8b1b..530f33bed4 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/internal/cordz_info.cc +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/cordz_info.cc @@ -35,7 +35,7 @@ namespace cord_internal { using ::absl::base_internal::SpinLockHolder; #ifdef ABSL_INTERNAL_NEED_REDUNDANT_CONSTEXPR_DECL -constexpr int CordzInfo::kMaxStackDepth; +constexpr size_t CordzInfo::kMaxStackDepth; #endif ABSL_CONST_INIT CordzInfo::List CordzInfo::global_list_{absl::kConstInit}; @@ -291,7 +291,7 @@ CordzInfo::MethodIdentifier CordzInfo::GetParentMethod(const CordzInfo* src) { : src->method_; } -int CordzInfo::FillParentStack(const CordzInfo* src, void** stack) { +size_t CordzInfo::FillParentStack(const CordzInfo* src, void** stack) { assert(stack); if (src == nullptr) return 0; if (src->parent_stack_depth_) { @@ -302,11 +302,14 @@ int CordzInfo::FillParentStack(const CordzInfo* src, void** stack) { return src->stack_depth_; } -CordzInfo::CordzInfo(CordRep* rep, const CordzInfo* src, +CordzInfo::CordzInfo(CordRep* rep, + const CordzInfo* src, MethodIdentifier method) : rep_(rep), - stack_depth_(absl::GetStackTrace(stack_, /*max_depth=*/kMaxStackDepth, - /*skip_count=*/1)), + stack_depth_( + static_cast<size_t>(absl::GetStackTrace(stack_, + /*max_depth=*/kMaxStackDepth, + /*skip_count=*/1))), parent_stack_depth_(FillParentStack(src, parent_stack_)), method_(method), parent_method_(GetParentMethod(src)), diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/cordz_info.h b/contrib/restricted/abseil-cpp/absl/strings/internal/cordz_info.h index 026d5b9981..17eaa91c77 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/internal/cordz_info.h +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/cordz_info.h @@ -12,8 +12,8 @@ // See the License for the specific language governing permissions and // limitations under the License. -#ifndef ABSL_STRINGS_CORDZ_INFO_H_ -#define ABSL_STRINGS_CORDZ_INFO_H_ +#ifndef ABSL_STRINGS_INTERNAL_CORDZ_INFO_H_ +#define ABSL_STRINGS_INTERNAL_CORDZ_INFO_H_ #include <atomic> #include <cstdint> @@ -196,7 +196,7 @@ class ABSL_LOCKABLE CordzInfo : public CordzHandle { std::atomic<CordzInfo*> head ABSL_GUARDED_BY(mutex){nullptr}; }; - static constexpr int kMaxStackDepth = 64; + static constexpr size_t kMaxStackDepth = 64; explicit CordzInfo(CordRep* rep, const CordzInfo* src, MethodIdentifier method); @@ -216,7 +216,7 @@ class ABSL_LOCKABLE CordzInfo : public CordzHandle { // `stack_` depending on `parent_stack_` being empty, returning the size of // the parent stack. // Returns 0 if `src` is null. - static int FillParentStack(const CordzInfo* src, void** stack); + static size_t FillParentStack(const CordzInfo* src, void** stack); void ODRCheck() const { #ifndef NDEBUG @@ -244,8 +244,8 @@ class ABSL_LOCKABLE CordzInfo : public CordzHandle { void* stack_[kMaxStackDepth]; void* parent_stack_[kMaxStackDepth]; - const int stack_depth_; - const int parent_stack_depth_; + const size_t stack_depth_; + const size_t parent_stack_depth_; const MethodIdentifier method_; const MethodIdentifier parent_method_; CordzUpdateTracker update_tracker_; @@ -295,4 +295,4 @@ inline CordRep* CordzInfo::RefCordRep() const ABSL_LOCKS_EXCLUDED(mutex_) { ABSL_NAMESPACE_END } // namespace absl -#endif // ABSL_STRINGS_CORDZ_INFO_H_ +#endif // ABSL_STRINGS_INTERNAL_CORDZ_INFO_H_ diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/cordz_sample_token.h b/contrib/restricted/abseil-cpp/absl/strings/internal/cordz_sample_token.h index 28a1d70ccc..b58022c3f9 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/internal/cordz_sample_token.h +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/cordz_sample_token.h @@ -16,8 +16,8 @@ #include "absl/strings/internal/cordz_handle.h" #include "absl/strings/internal/cordz_info.h" -#ifndef ABSL_STRINGS_CORDZ_SAMPLE_TOKEN_H_ -#define ABSL_STRINGS_CORDZ_SAMPLE_TOKEN_H_ +#ifndef ABSL_STRINGS_INTERNAL_CORDZ_SAMPLE_TOKEN_H_ +#define ABSL_STRINGS_INTERNAL_CORDZ_SAMPLE_TOKEN_H_ namespace absl { ABSL_NAMESPACE_BEGIN @@ -94,4 +94,4 @@ class CordzSampleToken : public CordzSnapshot { ABSL_NAMESPACE_END } // namespace absl -#endif // ABSL_STRINGS_CORDZ_SAMPLE_TOKEN_H_ +#endif // ABSL_STRINGS_INTERNAL_CORDZ_SAMPLE_TOKEN_H_ diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/cordz_statistics.h b/contrib/restricted/abseil-cpp/absl/strings/internal/cordz_statistics.h index 5707190577..9f558df494 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/internal/cordz_statistics.h +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/cordz_statistics.h @@ -45,12 +45,12 @@ struct CordzStatistics { }; // The size of the cord in bytes. This matches the result of Cord::size(). - int64_t size = 0; + size_t size = 0; // The estimated memory used by the sampled cord. This value matches the // value as reported by Cord::EstimatedMemoryUsage(). // A value of 0 implies the property has not been recorded. - int64_t estimated_memory_usage = 0; + size_t estimated_memory_usage = 0; // The effective memory used by the sampled cord, inversely weighted by the // effective indegree of each allocated node. This is a representation of the @@ -59,14 +59,14 @@ struct CordzStatistics { // by multiple Cord instances, and for cases where a Cord includes the same // node multiple times (either directly or indirectly). // A value of 0 implies the property has not been recorded. - int64_t estimated_fair_share_memory_usage = 0; + size_t estimated_fair_share_memory_usage = 0; // The total number of nodes referenced by this cord. // For ring buffer Cords, this includes the 'ring buffer' node. // For btree Cords, this includes all 'CordRepBtree' tree nodes as well as all // the substring, flat and external nodes referenced by the tree. // A value of 0 implies the property has not been recorded. - int64_t node_count = 0; + size_t node_count = 0; // Detailed node counts per type NodeCounts node_counts; diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/damerau_levenshtein_distance.cc b/contrib/restricted/abseil-cpp/absl/strings/internal/damerau_levenshtein_distance.cc new file mode 100644 index 0000000000..a084568fa8 --- /dev/null +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/damerau_levenshtein_distance.cc @@ -0,0 +1,93 @@ +// Copyright 2022 The Abseil Authors +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// https://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#include "absl/strings/internal/damerau_levenshtein_distance.h" + +#include <algorithm> +#include <array> +#include <numeric> + +#include "absl/strings/string_view.h" +namespace absl { +ABSL_NAMESPACE_BEGIN +namespace strings_internal { +// Calculate DamerauLevenshtein (adjacent transpositions) distance +// between two strings, +// https://en.wikipedia.org/wiki/Damerau%E2%80%93Levenshtein_distance. The +// algorithm follows the condition that no substring is edited more than once. +// While this can reduce is larger distance, it's a) a much simpler algorithm +// and b) more realistic for the case that typographic mistakes should be +// detected. +// When the distance is larger than cutoff, or one of the strings has more +// than MAX_SIZE=100 characters, the code returns min(MAX_SIZE, cutoff) + 1. +uint8_t CappedDamerauLevenshteinDistance(absl::string_view s1, + absl::string_view s2, uint8_t cutoff) { + const uint8_t MAX_SIZE = 100; + const uint8_t _cutoff = std::min(MAX_SIZE, cutoff); + const uint8_t cutoff_plus_1 = static_cast<uint8_t>(_cutoff + 1); + + if (s1.size() > s2.size()) std::swap(s1, s2); + if (s1.size() + _cutoff < s2.size() || s2.size() > MAX_SIZE) + return cutoff_plus_1; + + if (s1.empty()) + return static_cast<uint8_t>(s2.size()); + + // Lower diagonal bound: y = x - lower_diag + const uint8_t lower_diag = + _cutoff - static_cast<uint8_t>(s2.size() - s1.size()); + // Upper diagonal bound: y = x + upper_diag + const uint8_t upper_diag = _cutoff; + + // d[i][j] is the number of edits required to convert s1[0, i] to s2[0, j] + std::array<std::array<uint8_t, MAX_SIZE + 2>, MAX_SIZE + 2> d; + std::iota(d[0].begin(), d[0].begin() + upper_diag + 1, 0); + d[0][cutoff_plus_1] = cutoff_plus_1; + for (size_t i = 1; i <= s1.size(); ++i) { + // Deduce begin of relevant window. + size_t j_begin = 1; + if (i > lower_diag) { + j_begin = i - lower_diag; + d[i][j_begin - 1] = cutoff_plus_1; + } else { + d[i][0] = static_cast<uint8_t>(i); + } + + // Deduce end of relevant window. + size_t j_end = i + upper_diag; + if (j_end > s2.size()) { + j_end = s2.size(); + } else { + d[i][j_end + 1] = cutoff_plus_1; + } + + for (size_t j = j_begin; j <= j_end; ++j) { + const uint8_t deletion_distance = d[i - 1][j] + 1; + const uint8_t insertion_distance = d[i][j - 1] + 1; + const uint8_t mismatched_tail_cost = s1[i - 1] == s2[j - 1] ? 0 : 1; + const uint8_t mismatch_distance = d[i - 1][j - 1] + mismatched_tail_cost; + uint8_t transposition_distance = _cutoff + 1; + if (i > 1 && j > 1 && s1[i - 1] == s2[j - 2] && s1[i - 2] == s2[j - 1]) + transposition_distance = d[i - 2][j - 2] + 1; + d[i][j] = std::min({cutoff_plus_1, deletion_distance, insertion_distance, + mismatch_distance, transposition_distance}); + } + } + return d[s1.size()][s2.size()]; +} + +} // namespace strings_internal + +ABSL_NAMESPACE_END +} // namespace absl diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/damerau_levenshtein_distance.h b/contrib/restricted/abseil-cpp/absl/strings/internal/damerau_levenshtein_distance.h new file mode 100644 index 0000000000..7a4bd64462 --- /dev/null +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/damerau_levenshtein_distance.h @@ -0,0 +1,34 @@ +// Copyright 2022 The Abseil Authors +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// https://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#ifndef ABSL_STRINGS_INTERNAL_DAMERAU_LEVENSHTEIN_DISTANCE_H_ +#define ABSL_STRINGS_INTERNAL_DAMERAU_LEVENSHTEIN_DISTANCE_H_ + +#include <cstdint> + +#include "absl/strings/string_view.h" + +namespace absl { +ABSL_NAMESPACE_BEGIN +namespace strings_internal { +// Calculate DamerauLevenshtein distance between two strings. +// When the distance is larger than cutoff, the code just returns cutoff + 1. +uint8_t CappedDamerauLevenshteinDistance(absl::string_view s1, + absl::string_view s2, uint8_t cutoff); + +} // namespace strings_internal +ABSL_NAMESPACE_END +} // namespace absl + +#endif // ABSL_STRINGS_INTERNAL_DAMERAU_LEVENSHTEIN_DISTANCE_H_ diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/escaping.cc b/contrib/restricted/abseil-cpp/absl/strings/internal/escaping.cc index cfea096111..8bd0890d60 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/internal/escaping.cc +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/escaping.cc @@ -28,19 +28,11 @@ size_t CalculateBase64EscapedLenInternal(size_t input_len, bool do_padding) { // Base64 encodes three bytes of input at a time. If the input is not // divisible by three, we pad as appropriate. // - // (from https://tools.ietf.org/html/rfc3548) - // Special processing is performed if fewer than 24 bits are available - // at the end of the data being encoded. A full encoding quantum is - // always completed at the end of a quantity. When fewer than 24 input - // bits are available in an input group, zero bits are added (on the - // right) to form an integral number of 6-bit groups. Padding at the - // end of the data is performed using the '=' character. Since all base - // 64 input is an integral number of octets, only the following cases - // can arise: - // Base64 encodes each three bytes of input into four bytes of output. size_t len = (input_len / 3) * 4; + // Since all base 64 input is an integral number of octets, only the following + // cases can arise: if (input_len % 3 == 0) { // (from https://tools.ietf.org/html/rfc3548) // (1) the final quantum of encoding input is an integral multiple of 24 @@ -83,6 +75,16 @@ size_t Base64EscapeInternal(const unsigned char* src, size_t szsrc, char* dest, char* const limit_dest = dest + szdest; const unsigned char* const limit_src = src + szsrc; + // (from https://tools.ietf.org/html/rfc3548) + // Special processing is performed if fewer than 24 bits are available + // at the end of the data being encoded. A full encoding quantum is + // always completed at the end of a quantity. When fewer than 24 input + // bits are available in an input group, zero bits are added (on the + // right) to form an integral number of 6-bit groups. + // + // If do_padding is true, padding at the end of the data is performed. This + // output padding uses the '=' character. + // Three bytes of data encodes to four characters of cyphertext. // So we can pump through three-byte chunks atomically. if (szsrc >= 3) { // "limit_src - 3" is UB if szsrc < 3. diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/escaping.h b/contrib/restricted/abseil-cpp/absl/strings/internal/escaping.h index 6a9ce602d9..b04033ffc9 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/internal/escaping.h +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/escaping.h @@ -25,19 +25,17 @@ namespace strings_internal { ABSL_CONST_INIT extern const char kBase64Chars[]; -// Calculates how long a string will be when it is base64 encoded given its -// length and whether or not the result should be padded. +// Calculates the length of a Base64 encoding (RFC 4648) of a string of length +// `input_len`, with or without padding per `do_padding`. Note that 'web-safe' +// encoding (section 5 of the RFC) does not change this length. size_t CalculateBase64EscapedLenInternal(size_t input_len, bool do_padding); -// Base64-encodes `src` using the alphabet provided in `base64` and writes the -// result to `dest`. If `do_padding` is true, `dest` is padded with '=' chars -// until its length is a multiple of 3. Returns the length of `dest`. +// Base64-encodes `src` using the alphabet provided in `base64` (which +// determines whether to do web-safe encoding or not) and writes the result to +// `dest`. If `do_padding` is true, `dest` is padded with '=' chars until its +// length is a multiple of 3. Returns the length of `dest`. size_t Base64EscapeInternal(const unsigned char* src, size_t szsrc, char* dest, size_t szdest, const char* base64, bool do_padding); - -// Base64-encodes `src` using the alphabet provided in `base64` and writes the -// result to `dest`. If `do_padding` is true, `dest` is padded with '=' chars -// until its length is a multiple of 3. template <typename String> void Base64EscapeInternal(const unsigned char* src, size_t szsrc, String* dest, bool do_padding, const char* base64_chars) { diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/has_absl_stringify.h b/contrib/restricted/abseil-cpp/absl/strings/internal/has_absl_stringify.h new file mode 100644 index 0000000000..55a0850829 --- /dev/null +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/has_absl_stringify.h @@ -0,0 +1,55 @@ +// Copyright 2022 The Abseil Authors +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// https://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#ifndef ABSL_STRINGS_INTERNAL_HAS_ABSL_STRINGIFY_H_ +#define ABSL_STRINGS_INTERNAL_HAS_ABSL_STRINGIFY_H_ +#include <string> +#include <type_traits> +#include <utility> + +#include "absl/strings/string_view.h" + +namespace absl { +ABSL_NAMESPACE_BEGIN + +namespace strings_internal { + +// This is an empty class not intended to be used. It exists so that +// `HasAbslStringify` can reference a universal class rather than needing to be +// copied for each new sink. +class UnimplementedSink { + public: + void Append(size_t count, char ch); + + void Append(string_view v); + + // Support `absl::Format(&sink, format, args...)`. + friend void AbslFormatFlush(UnimplementedSink* sink, absl::string_view v); +}; + +template <typename T, typename = void> +struct HasAbslStringify : std::false_type {}; + +template <typename T> +struct HasAbslStringify< + T, std::enable_if_t<std::is_void<decltype(AbslStringify( + std::declval<strings_internal::UnimplementedSink&>(), + std::declval<const T&>()))>::value>> : std::true_type {}; + +} // namespace strings_internal + +ABSL_NAMESPACE_END +} // namespace absl + +#endif // ABSL_STRINGS_INTERNAL_HAS_ABSL_STRINGIFY_H_ diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/memutil.cc b/contrib/restricted/abseil-cpp/absl/strings/internal/memutil.cc index 2519c6881e..44996a7549 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/internal/memutil.cc +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/memutil.cc @@ -54,10 +54,11 @@ size_t memspn(const char* s, size_t slen, const char* accept) { cont: c = *p++; - if (slen-- == 0) return p - 1 - s; + if (slen-- == 0) + return static_cast<size_t>(p - 1 - s); for (spanp = accept; (sc = *spanp++) != '\0';) if (sc == c) goto cont; - return p - 1 - s; + return static_cast<size_t>(p - 1 - s); } size_t memcspn(const char* s, size_t slen, const char* reject) { @@ -68,9 +69,10 @@ size_t memcspn(const char* s, size_t slen, const char* reject) { while (slen-- != 0) { c = *p++; for (spanp = reject; (sc = *spanp++) != '\0';) - if (sc == c) return p - 1 - s; + if (sc == c) + return static_cast<size_t>(p - 1 - s); } - return p - s; + return static_cast<size_t>(p - s); } char* mempbrk(const char* s, size_t slen, const char* accept) { @@ -97,8 +99,9 @@ const char* memmatch(const char* phaystack, size_t haylen, const char* pneedle, const char* hayend = phaystack + haylen - neelen + 1; // A static cast is used here to work around the fact that memchr returns // a void* on Posix-compliant systems and const void* on Windows. - while ((match = static_cast<const char*>( - memchr(phaystack, pneedle[0], hayend - phaystack)))) { + while ( + (match = static_cast<const char*>(memchr( + phaystack, pneedle[0], static_cast<size_t>(hayend - phaystack))))) { if (memcmp(match, pneedle, neelen) == 0) return match; else diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/ostringstream.cc b/contrib/restricted/abseil-cpp/absl/strings/internal/ostringstream.cc index dc6cfe1686..a0e5ec08c2 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/internal/ostringstream.cc +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/ostringstream.cc @@ -14,20 +14,27 @@ #include "absl/strings/internal/ostringstream.h" +#include <cassert> +#include <cstddef> +#include <ios> +#include <streambuf> + namespace absl { ABSL_NAMESPACE_BEGIN namespace strings_internal { -OStringStream::Buf::int_type OStringStream::overflow(int c) { - assert(s_); - if (!Buf::traits_type::eq_int_type(c, Buf::traits_type::eof())) - s_->push_back(static_cast<char>(c)); +OStringStream::Streambuf::int_type OStringStream::Streambuf::overflow(int c) { + assert(str_); + if (!std::streambuf::traits_type::eq_int_type( + c, std::streambuf::traits_type::eof())) + str_->push_back(static_cast<char>(c)); return 1; } -std::streamsize OStringStream::xsputn(const char* s, std::streamsize n) { - assert(s_); - s_->append(s, static_cast<size_t>(n)); +std::streamsize OStringStream::Streambuf::xsputn(const char* s, + std::streamsize n) { + assert(str_); + str_->append(s, static_cast<size_t>(n)); return n; } diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/ostringstream.h b/contrib/restricted/abseil-cpp/absl/strings/internal/ostringstream.h index d25d60473f..c0e237dbe8 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/internal/ostringstream.h +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/ostringstream.h @@ -16,11 +16,13 @@ #define ABSL_STRINGS_INTERNAL_OSTRINGSTREAM_H_ #include <cassert> +#include <ios> #include <ostream> #include <streambuf> #include <string> +#include <utility> -#include "absl/base/port.h" +#include "absl/base/config.h" namespace absl { ABSL_NAMESPACE_BEGIN @@ -60,26 +62,49 @@ namespace strings_internal { // strm << 3.14; // // Note: flush() has no effect. No reason to call it. -class OStringStream : private std::basic_streambuf<char>, public std::ostream { +class OStringStream final : public std::ostream { public: // The argument can be null, in which case you'll need to call str(p) with a // non-null argument before you can write to the stream. // // The destructor of OStringStream doesn't use the std::string. It's OK to // destroy the std::string before the stream. - explicit OStringStream(std::string* s) : std::ostream(this), s_(s) {} + explicit OStringStream(std::string* str) + : std::ostream(&buf_), buf_(str) {} + OStringStream(OStringStream&& that) + : std::ostream(std::move(static_cast<std::ostream&>(that))), + buf_(that.buf_) { + rdbuf(&buf_); + } + OStringStream& operator=(OStringStream&& that) { + std::ostream::operator=(std::move(static_cast<std::ostream&>(that))); + buf_ = that.buf_; + rdbuf(&buf_); + return *this; + } - std::string* str() { return s_; } - const std::string* str() const { return s_; } - void str(std::string* s) { s_ = s; } + std::string* str() { return buf_.str(); } + const std::string* str() const { return buf_.str(); } + void str(std::string* str) { buf_.str(str); } private: - using Buf = std::basic_streambuf<char>; + class Streambuf final : public std::streambuf { + public: + explicit Streambuf(std::string* str) : str_(str) {} + Streambuf(const Streambuf&) = default; + Streambuf& operator=(const Streambuf&) = default; - Buf::int_type overflow(int c) override; - std::streamsize xsputn(const char* s, std::streamsize n) override; + std::string* str() { return str_; } + const std::string* str() const { return str_; } + void str(std::string* str) { str_ = str; } - std::string* s_; + protected: + int_type overflow(int c) override; + std::streamsize xsputn(const char* s, std::streamsize n) override; + + private: + std::string* str_; + } buf_; }; } // namespace strings_internal diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/arg.cc b/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/arg.cc index 02aeeebec7..018dd05235 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/arg.cc +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/arg.cc @@ -77,7 +77,7 @@ class IntDigits { v >>= 3; } while (v); start_ = p; - size_ = storage_ + sizeof(storage_) - p; + size_ = static_cast<size_t>(storage_ + sizeof(storage_) - p); } // Print the signed or unsigned integer as decimal. @@ -86,7 +86,8 @@ class IntDigits { void PrintAsDec(T v) { static_assert(std::is_integral<T>::value, ""); start_ = storage_; - size_ = numbers_internal::FastIntToBuffer(v, storage_) - storage_; + size_ = static_cast<size_t>(numbers_internal::FastIntToBuffer(v, storage_) - + storage_); } void PrintAsDec(int128 v) { @@ -115,7 +116,7 @@ class IntDigits { if (add_neg) { *--p = '-'; } - size_ = storage_ + sizeof(storage_) - p; + size_ = static_cast<size_t>(storage_ + sizeof(storage_) - p); start_ = p; } @@ -138,7 +139,7 @@ class IntDigits { ++p; } start_ = p; - size_ = storage_ + sizeof(storage_) - p; + size_ = static_cast<size_t>(storage_ + sizeof(storage_) - p); } // Print the unsigned integer as hex using uppercase. @@ -154,7 +155,7 @@ class IntDigits { v >>= 4; } while (v); start_ = p; - size_ = storage_ + sizeof(storage_) - p; + size_ = static_cast<size_t>(storage_ + sizeof(storage_) - p); } // The printed value including the '-' sign if available. @@ -208,10 +209,12 @@ string_view SignColumn(bool neg, const FormatConversionSpecImpl conv) { return {}; } -bool ConvertCharImpl(unsigned char v, const FormatConversionSpecImpl conv, - FormatSinkImpl *sink) { +bool ConvertCharImpl(char v, + const FormatConversionSpecImpl conv, + FormatSinkImpl* sink) { size_t fill = 0; - if (conv.width() >= 0) fill = conv.width(); + if (conv.width() >= 0) + fill = static_cast<size_t>(conv.width()); ReducePadding(1, &fill); if (!conv.has_left_flag()) sink->Append(fill, ' '); sink->Append(1, v); @@ -225,7 +228,8 @@ bool ConvertIntImplInnerSlow(const IntDigits &as_digits, // Print as a sequence of Substrings: // [left_spaces][sign][base_indicator][zeroes][formatted][right_spaces] size_t fill = 0; - if (conv.width() >= 0) fill = conv.width(); + if (conv.width() >= 0) + fill = static_cast<size_t>(conv.width()); string_view formatted = as_digits.without_neg_or_zero(); ReducePadding(formatted, &fill); @@ -236,10 +240,9 @@ bool ConvertIntImplInnerSlow(const IntDigits &as_digits, string_view base_indicator = BaseIndicator(as_digits, conv); ReducePadding(base_indicator, &fill); - int precision = conv.precision(); - bool precision_specified = precision >= 0; - if (!precision_specified) - precision = 1; + bool precision_specified = conv.precision() >= 0; + size_t precision = + precision_specified ? static_cast<size_t>(conv.precision()) : size_t{1}; if (conv.has_alt_flag() && conv.conversion_char() == FormatConversionCharInternal::o) { @@ -247,7 +250,7 @@ bool ConvertIntImplInnerSlow(const IntDigits &as_digits, // "For o conversion, it increases the precision (if necessary) to // force the first digit of the result to be zero." if (formatted.empty() || *formatted.begin() != '0') { - int needed = static_cast<int>(formatted.size()) + 1; + size_t needed = formatted.size() + 1; precision = std::max(precision, needed); } } @@ -275,19 +278,71 @@ bool ConvertIntImplInnerSlow(const IntDigits &as_digits, return true; } +template <typename T, + typename std::enable_if<(std::is_integral<T>::value && + std::is_signed<T>::value) || + std::is_same<T, int128>::value, + int>::type = 0> +constexpr auto ConvertV(T) { + return FormatConversionCharInternal::d; +} + +template <typename T, + typename std::enable_if<(std::is_integral<T>::value && + std::is_unsigned<T>::value) || + std::is_same<T, uint128>::value, + int>::type = 0> +constexpr auto ConvertV(T) { + return FormatConversionCharInternal::u; +} + +template <typename T> +bool ConvertFloatArg(T v, FormatConversionSpecImpl conv, FormatSinkImpl *sink) { + if (conv.conversion_char() == FormatConversionCharInternal::v) { + conv.set_conversion_char(FormatConversionCharInternal::g); + } + + return FormatConversionCharIsFloat(conv.conversion_char()) && + ConvertFloatImpl(v, conv, sink); +} + +inline bool ConvertStringArg(string_view v, const FormatConversionSpecImpl conv, + FormatSinkImpl *sink) { + if (conv.is_basic()) { + sink->Append(v); + return true; + } + return sink->PutPaddedString(v, conv.width(), conv.precision(), + conv.has_left_flag()); +} + +} // namespace + +bool ConvertBoolArg(bool v, FormatSinkImpl *sink) { + if (v) { + sink->Append("true"); + } else { + sink->Append("false"); + } + return true; +} + template <typename T> -bool ConvertIntArg(T v, const FormatConversionSpecImpl conv, - FormatSinkImpl *sink) { +bool ConvertIntArg(T v, FormatConversionSpecImpl conv, FormatSinkImpl *sink) { using U = typename MakeUnsigned<T>::type; IntDigits as_digits; + if (conv.conversion_char() == FormatConversionCharInternal::v) { + conv.set_conversion_char(ConvertV(T{})); + } + // This odd casting is due to a bug in -Wswitch behavior in gcc49 which causes // it to complain about a switch/case type mismatch, even though both are // FormatConverionChar. Likely this is because at this point // FormatConversionChar is declared, but not defined. switch (static_cast<uint8_t>(conv.conversion_char())) { case static_cast<uint8_t>(FormatConversionCharInternal::c): - return ConvertCharImpl(static_cast<unsigned char>(v), conv, sink); + return ConvertCharImpl(static_cast<char>(v), conv, sink); case static_cast<uint8_t>(FormatConversionCharInternal::o): as_digits.PrintAsOct(static_cast<U>(v)); @@ -320,7 +375,7 @@ bool ConvertIntArg(T v, const FormatConversionSpecImpl conv, return ConvertFloatImpl(static_cast<double>(v), conv, sink); default: - ABSL_ASSUME(false); + ABSL_ASSUME(false); } if (conv.is_basic()) { @@ -330,24 +385,37 @@ bool ConvertIntArg(T v, const FormatConversionSpecImpl conv, return ConvertIntImplInnerSlow(as_digits, conv, sink); } -template <typename T> -bool ConvertFloatArg(T v, const FormatConversionSpecImpl conv, - FormatSinkImpl *sink) { - return FormatConversionCharIsFloat(conv.conversion_char()) && - ConvertFloatImpl(v, conv, sink); -} - -inline bool ConvertStringArg(string_view v, const FormatConversionSpecImpl conv, - FormatSinkImpl *sink) { - if (conv.is_basic()) { - sink->Append(v); - return true; - } - return sink->PutPaddedString(v, conv.width(), conv.precision(), - conv.has_left_flag()); -} - -} // namespace +template bool ConvertIntArg<char>(char v, FormatConversionSpecImpl conv, + FormatSinkImpl *sink); +template bool ConvertIntArg<signed char>(signed char v, + FormatConversionSpecImpl conv, + FormatSinkImpl *sink); +template bool ConvertIntArg<unsigned char>(unsigned char v, + FormatConversionSpecImpl conv, + FormatSinkImpl *sink); +template bool ConvertIntArg<short>(short v, // NOLINT + FormatConversionSpecImpl conv, + FormatSinkImpl *sink); +template bool ConvertIntArg<unsigned short>(unsigned short v, // NOLINT + FormatConversionSpecImpl conv, + FormatSinkImpl *sink); +template bool ConvertIntArg<int>(int v, FormatConversionSpecImpl conv, + FormatSinkImpl *sink); +template bool ConvertIntArg<unsigned int>(unsigned int v, + FormatConversionSpecImpl conv, + FormatSinkImpl *sink); +template bool ConvertIntArg<long>(long v, // NOLINT + FormatConversionSpecImpl conv, + FormatSinkImpl *sink); +template bool ConvertIntArg<unsigned long>(unsigned long v, // NOLINT + FormatConversionSpecImpl conv, + FormatSinkImpl *sink); +template bool ConvertIntArg<long long>(long long v, // NOLINT + FormatConversionSpecImpl conv, + FormatSinkImpl *sink); +template bool ConvertIntArg<unsigned long long>(unsigned long long v, // NOLINT + FormatConversionSpecImpl conv, + FormatSinkImpl *sink); // ==================== Strings ==================== StringConvertResult FormatConvertImpl(const std::string &v, @@ -375,7 +443,7 @@ FormatConvertImpl(const char *v, const FormatConversionSpecImpl conv, len = std::strlen(v); } else { // If precision is set, we look for the NUL-terminator on the valid range. - len = std::find(v, v + conv.precision(), '\0') - v; + len = static_cast<size_t>(std::find(v, v + conv.precision(), '\0') - v); } return {ConvertStringArg(string_view(v, len), conv, sink)}; } @@ -410,19 +478,18 @@ FloatingConvertResult FormatConvertImpl(long double v, } // ==================== Chars ==================== -IntegralConvertResult FormatConvertImpl(char v, - const FormatConversionSpecImpl conv, - FormatSinkImpl *sink) { +CharConvertResult FormatConvertImpl(char v, const FormatConversionSpecImpl conv, + FormatSinkImpl *sink) { return {ConvertIntArg(v, conv, sink)}; } -IntegralConvertResult FormatConvertImpl(signed char v, - const FormatConversionSpecImpl conv, - FormatSinkImpl *sink) { +CharConvertResult FormatConvertImpl(signed char v, + const FormatConversionSpecImpl conv, + FormatSinkImpl *sink) { return {ConvertIntArg(v, conv, sink)}; } -IntegralConvertResult FormatConvertImpl(unsigned char v, - const FormatConversionSpecImpl conv, - FormatSinkImpl *sink) { +CharConvertResult FormatConvertImpl(unsigned char v, + const FormatConversionSpecImpl conv, + FormatSinkImpl *sink) { return {ConvertIntArg(v, conv, sink)}; } diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/arg.h b/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/arg.h index b9dda90901..e4b1662827 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/arg.h +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/arg.h @@ -18,6 +18,7 @@ #include <string.h> #include <wchar.h> +#include <algorithm> #include <cstdio> #include <iomanip> #include <limits> @@ -25,10 +26,12 @@ #include <sstream> #include <string> #include <type_traits> +#include <utility> #include "absl/base/port.h" #include "absl/meta/type_traits.h" #include "absl/numeric/int128.h" +#include "absl/strings/internal/has_absl_stringify.h" #include "absl/strings/internal/str_format/extension.h" #include "absl/strings/string_view.h" @@ -45,6 +48,24 @@ class FormatConversionSpec; namespace str_format_internal { +template <FormatConversionCharSet C> +struct ArgConvertResult { + bool value; +}; + +using IntegralConvertResult = ArgConvertResult<FormatConversionCharSetUnion( + FormatConversionCharSetInternal::c, + FormatConversionCharSetInternal::kNumeric, + FormatConversionCharSetInternal::kStar, + FormatConversionCharSetInternal::v)>; +using FloatingConvertResult = ArgConvertResult<FormatConversionCharSetUnion( + FormatConversionCharSetInternal::kFloating, + FormatConversionCharSetInternal::v)>; +using CharConvertResult = ArgConvertResult<FormatConversionCharSetUnion( + FormatConversionCharSetInternal::c, + FormatConversionCharSetInternal::kNumeric, + FormatConversionCharSetInternal::kStar)>; + template <typename T, typename = void> struct HasUserDefinedConvert : std::false_type {}; @@ -55,7 +76,50 @@ struct HasUserDefinedConvert<T, void_t<decltype(AbslFormatConvert( std::declval<FormatSink*>()))>> : std::true_type {}; -void AbslFormatConvert(); // Stops the lexical name lookup +// These declarations prevent ADL lookup from continuing in absl namespaces, +// we are deliberately using these as ADL hooks and want them to consider +// non-absl namespaces only. +void AbslFormatConvert(); +void AbslStringify(); + +template <typename T> +bool ConvertIntArg(T v, FormatConversionSpecImpl conv, FormatSinkImpl* sink); + +// Forward declarations of internal `ConvertIntArg` function template +// instantiations are here to avoid including the template body in the headers +// and instantiating it in large numbers of translation units. Explicit +// instantiations can be found in "absl/strings/internal/str_format/arg.cc" +extern template bool ConvertIntArg<char>(char v, FormatConversionSpecImpl conv, + FormatSinkImpl* sink); +extern template bool ConvertIntArg<signed char>(signed char v, + FormatConversionSpecImpl conv, + FormatSinkImpl* sink); +extern template bool ConvertIntArg<unsigned char>(unsigned char v, + FormatConversionSpecImpl conv, + FormatSinkImpl* sink); +extern template bool ConvertIntArg<short>(short v, // NOLINT + FormatConversionSpecImpl conv, + FormatSinkImpl* sink); +extern template bool ConvertIntArg<unsigned short>( // NOLINT + unsigned short v, FormatConversionSpecImpl conv, // NOLINT + FormatSinkImpl* sink); +extern template bool ConvertIntArg<int>(int v, FormatConversionSpecImpl conv, + FormatSinkImpl* sink); +extern template bool ConvertIntArg<unsigned int>(unsigned int v, + FormatConversionSpecImpl conv, + FormatSinkImpl* sink); +extern template bool ConvertIntArg<long>( // NOLINT + long v, FormatConversionSpecImpl conv, FormatSinkImpl* sink); // NOLINT +extern template bool ConvertIntArg<unsigned long>(unsigned long v, // NOLINT + FormatConversionSpecImpl conv, + FormatSinkImpl* sink); +extern template bool ConvertIntArg<long long>(long long v, // NOLINT + FormatConversionSpecImpl conv, + FormatSinkImpl* sink); +extern template bool ConvertIntArg<unsigned long long>( // NOLINT + unsigned long long v, FormatConversionSpecImpl conv, // NOLINT + FormatSinkImpl* sink); + template <typename T> auto FormatConvertImpl(const T& v, FormatConversionSpecImpl conv, FormatSinkImpl* sink) @@ -72,6 +136,39 @@ auto FormatConvertImpl(const T& v, FormatConversionSpecImpl conv, } template <typename T> +auto FormatConvertImpl(const T& v, FormatConversionSpecImpl conv, + FormatSinkImpl* sink) + -> std::enable_if_t<std::is_enum<T>::value && + std::is_void<decltype(AbslStringify( + std::declval<FormatSink&>(), v))>::value, + IntegralConvertResult> { + if (conv.conversion_char() == FormatConversionCharInternal::v) { + using FormatSinkT = + absl::enable_if_t<sizeof(const T& (*)()) != 0, FormatSink>; + auto fs = sink->Wrap<FormatSinkT>(); + AbslStringify(fs, v); + return {true}; + } else { + return {ConvertIntArg( + static_cast<typename std::underlying_type<T>::type>(v), conv, sink)}; + } +} + +template <typename T> +auto FormatConvertImpl(const T& v, FormatConversionSpecImpl, + FormatSinkImpl* sink) + -> std::enable_if_t<!std::is_enum<T>::value && + std::is_void<decltype(AbslStringify( + std::declval<FormatSink&>(), v))>::value, + ArgConvertResult<FormatConversionCharSetInternal::v>> { + using FormatSinkT = + absl::enable_if_t<sizeof(const T& (*)()) != 0, FormatSink>; + auto fs = sink->Wrap<FormatSinkT>(); + AbslStringify(fs, v); + return {true}; +} + +template <typename T> class StreamedWrapper; // If 'v' can be converted (in the printf sense) according to 'conv', @@ -96,11 +193,6 @@ struct VoidPtr { }; template <FormatConversionCharSet C> -struct ArgConvertResult { - bool value; -}; - -template <FormatConversionCharSet C> constexpr FormatConversionCharSet ExtractCharSet(FormatConvertResult<C>) { return C; } @@ -110,8 +202,8 @@ constexpr FormatConversionCharSet ExtractCharSet(ArgConvertResult<C>) { return C; } -using StringConvertResult = - ArgConvertResult<FormatConversionCharSetInternal::s>; +using StringConvertResult = ArgConvertResult<FormatConversionCharSetUnion( + FormatConversionCharSetInternal::s, FormatConversionCharSetInternal::v)>; ArgConvertResult<FormatConversionCharSetInternal::p> FormatConvertImpl( VoidPtr v, FormatConversionSpecImpl conv, FormatSinkImpl* sink); @@ -173,12 +265,7 @@ StringConvertResult FormatConvertImpl(const AbslCord& value, return {true}; } -using IntegralConvertResult = ArgConvertResult<FormatConversionCharSetUnion( - FormatConversionCharSetInternal::c, - FormatConversionCharSetInternal::kNumeric, - FormatConversionCharSetInternal::kStar)>; -using FloatingConvertResult = - ArgConvertResult<FormatConversionCharSetInternal::kFloating>; +bool ConvertBoolArg(bool v, FormatSinkImpl* sink); // Floats. FloatingConvertResult FormatConvertImpl(float v, FormatConversionSpecImpl conv, @@ -190,14 +277,14 @@ FloatingConvertResult FormatConvertImpl(long double v, FormatSinkImpl* sink); // Chars. -IntegralConvertResult FormatConvertImpl(char v, FormatConversionSpecImpl conv, - FormatSinkImpl* sink); -IntegralConvertResult FormatConvertImpl(signed char v, - FormatConversionSpecImpl conv, - FormatSinkImpl* sink); -IntegralConvertResult FormatConvertImpl(unsigned char v, - FormatConversionSpecImpl conv, - FormatSinkImpl* sink); +CharConvertResult FormatConvertImpl(char v, FormatConversionSpecImpl conv, + FormatSinkImpl* sink); +CharConvertResult FormatConvertImpl(signed char v, + FormatConversionSpecImpl conv, + FormatSinkImpl* sink); +CharConvertResult FormatConvertImpl(unsigned char v, + FormatConversionSpecImpl conv, + FormatSinkImpl* sink); // Ints. IntegralConvertResult FormatConvertImpl(short v, // NOLINT @@ -228,9 +315,16 @@ IntegralConvertResult FormatConvertImpl(int128 v, FormatConversionSpecImpl conv, IntegralConvertResult FormatConvertImpl(uint128 v, FormatConversionSpecImpl conv, FormatSinkImpl* sink); + +// This function needs to be a template due to ambiguity regarding type +// conversions. template <typename T, enable_if_t<std::is_same<T, bool>::value, int> = 0> IntegralConvertResult FormatConvertImpl(T v, FormatConversionSpecImpl conv, FormatSinkImpl* sink) { + if (conv.conversion_char() == FormatConversionCharInternal::v) { + return {ConvertBoolArg(v, sink)}; + } + return FormatConvertImpl(static_cast<int>(v), conv, sink); } @@ -238,7 +332,8 @@ IntegralConvertResult FormatConvertImpl(T v, FormatConversionSpecImpl conv, // FormatArgImpl will use the underlying Convert functions instead. template <typename T> typename std::enable_if<std::is_enum<T>::value && - !HasUserDefinedConvert<T>::value, + !HasUserDefinedConvert<T>::value && + !strings_internal::HasAbslStringify<T>::value, IntegralConvertResult>::type FormatConvertImpl(T v, FormatConversionSpecImpl conv, FormatSinkImpl* sink); @@ -301,11 +396,11 @@ struct FormatArgImplFriend { template <typename Arg> constexpr FormatConversionCharSet ArgumentToConv() { - return absl::str_format_internal::ExtractCharSet( - decltype(str_format_internal::FormatConvertImpl( - std::declval<const Arg&>(), - std::declval<const FormatConversionSpecImpl&>(), - std::declval<FormatSinkImpl*>())){}); + using ConvResult = decltype(str_format_internal::FormatConvertImpl( + std::declval<const Arg&>(), + std::declval<const FormatConversionSpecImpl&>(), + std::declval<FormatSinkImpl*>())); + return absl::str_format_internal::ExtractCharSet(ConvResult{}); } // A type-erased handle to a format argument. @@ -351,7 +446,8 @@ class FormatArgImpl { template <typename T, typename = void> struct DecayType { static constexpr bool kHasUserDefined = - str_format_internal::HasUserDefinedConvert<T>::value; + str_format_internal::HasUserDefinedConvert<T>::value || + strings_internal::HasAbslStringify<T>::value; using type = typename std::conditional< !kHasUserDefined && std::is_convertible<T, const char*>::value, const char*, @@ -363,6 +459,7 @@ class FormatArgImpl { struct DecayType<T, typename std::enable_if< !str_format_internal::HasUserDefinedConvert<T>::value && + !strings_internal::HasAbslStringify<T>::value && std::is_enum<T>::value>::type> { using type = typename std::underlying_type<T>::type; }; diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/bind.cc b/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/bind.cc index c988ba8fd2..77a4222337 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/bind.cc +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/bind.cc @@ -32,7 +32,8 @@ inline bool BindFromPosition(int position, int* value, return false; } // -1 because positions are 1-based - return FormatArgImplFriend::ToInt(pack[position - 1], value); + return FormatArgImplFriend::ToInt(pack[static_cast<size_t>(position) - 1], + value); } class ArgContext { @@ -56,7 +57,7 @@ inline bool ArgContext::Bind(const UnboundConversion* unbound, const FormatArgImpl* arg = nullptr; int arg_position = unbound->arg_position; if (static_cast<size_t>(arg_position - 1) >= pack_.size()) return false; - arg = &pack_[arg_position - 1]; // 1-based + arg = &pack_[static_cast<size_t>(arg_position - 1)]; // 1-based if (unbound->flags != Flags::kBasic) { int width = unbound->width.value(); diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/bind.h b/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/bind.h index 80f29654b3..b73c50287c 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/bind.h +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/bind.h @@ -235,9 +235,10 @@ class StreamedWrapper { private: template <typename S> - friend ArgConvertResult<FormatConversionCharSetInternal::s> FormatConvertImpl( - const StreamedWrapper<S>& v, FormatConversionSpecImpl conv, - FormatSinkImpl* out); + friend ArgConvertResult<FormatConversionCharSetUnion( + FormatConversionCharSetInternal::s, FormatConversionCharSetInternal::v)> + FormatConvertImpl(const StreamedWrapper<S>& v, FormatConversionSpecImpl conv, + FormatSinkImpl* out); const T& v_; }; diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/checker.h b/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/checker.h index 4fd19d13ae..eab6ab9d00 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/checker.h +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/checker.h @@ -15,8 +15,11 @@ #ifndef ABSL_STRINGS_INTERNAL_STR_FORMAT_CHECKER_H_ #define ABSL_STRINGS_INTERNAL_STR_FORMAT_CHECKER_H_ +#include <algorithm> + #include "absl/base/attributes.h" #include "absl/strings/internal/str_format/arg.h" +#include "absl/strings/internal/str_format/constexpr_parser.h" #include "absl/strings/internal/str_format/extension.h" // Compile time check support for entry points. @@ -36,297 +39,56 @@ namespace absl { ABSL_NAMESPACE_BEGIN namespace str_format_internal { -constexpr bool AllOf() { return true; } - -template <typename... T> -constexpr bool AllOf(bool b, T... t) { - return b && AllOf(t...); -} - #ifdef ABSL_INTERNAL_ENABLE_FORMAT_CHECKER -constexpr bool ContainsChar(const char* chars, char c) { - return *chars == c || (*chars && ContainsChar(chars + 1, c)); -} - -// A constexpr compatible list of Convs. -struct ConvList { - const FormatConversionCharSet* array; - int count; - - // We do the bound check here to avoid having to do it on the callers. - // Returning an empty FormatConversionCharSet has the same effect as - // short circuiting because it will never match any conversion. - constexpr FormatConversionCharSet operator[](int i) const { - return i < count ? array[i] : FormatConversionCharSet{}; - } - - constexpr ConvList without_front() const { - return count != 0 ? ConvList{array + 1, count - 1} : *this; - } -}; - -template <size_t count> -struct ConvListT { - // Make sure the array has size > 0. - FormatConversionCharSet list[count ? count : 1]; -}; - -constexpr char GetChar(string_view str, size_t index) { - return index < str.size() ? str[index] : char{}; -} - -constexpr string_view ConsumeFront(string_view str, size_t len = 1) { - return len <= str.size() ? string_view(str.data() + len, str.size() - len) - : string_view(); -} - -constexpr string_view ConsumeAnyOf(string_view format, const char* chars) { - return ContainsChar(chars, GetChar(format, 0)) - ? ConsumeAnyOf(ConsumeFront(format), chars) - : format; -} - -constexpr bool IsDigit(char c) { return c >= '0' && c <= '9'; } - -// Helper class for the ParseDigits function. -// It encapsulates the two return values we need there. -struct Integer { - string_view format; - int value; - - // If the next character is a '$', consume it. - // Otherwise, make `this` an invalid positional argument. - constexpr Integer ConsumePositionalDollar() const { - return GetChar(format, 0) == '$' ? Integer{ConsumeFront(format), value} - : Integer{format, 0}; - } -}; - -constexpr Integer ParseDigits(string_view format, int value = 0) { - return IsDigit(GetChar(format, 0)) - ? ParseDigits(ConsumeFront(format), - 10 * value + GetChar(format, 0) - '0') - : Integer{format, value}; -} - -// Parse digits for a positional argument. -// The parsing also consumes the '$'. -constexpr Integer ParsePositional(string_view format) { - return ParseDigits(format).ConsumePositionalDollar(); -} - -// Parses a single conversion specifier. -// See ConvParser::Run() for post conditions. -class ConvParser { - constexpr ConvParser SetFormat(string_view format) const { - return ConvParser(format, args_, error_, arg_position_, is_positional_); - } - - constexpr ConvParser SetArgs(ConvList args) const { - return ConvParser(format_, args, error_, arg_position_, is_positional_); - } - - constexpr ConvParser SetError(bool error) const { - return ConvParser(format_, args_, error_ || error, arg_position_, - is_positional_); - } - - constexpr ConvParser SetArgPosition(int arg_position) const { - return ConvParser(format_, args_, error_, arg_position, is_positional_); - } - - // Consumes the next arg and verifies that it matches `conv`. - // `error_` is set if there is no next arg or if it doesn't match `conv`. - constexpr ConvParser ConsumeNextArg(char conv) const { - return SetArgs(args_.without_front()).SetError(!Contains(args_[0], conv)); - } - - // Verify that positional argument `i.value` matches `conv`. - // `error_` is set if `i.value` is not a valid argument or if it doesn't - // match. - constexpr ConvParser VerifyPositional(Integer i, char conv) const { - return SetFormat(i.format).SetError(!Contains(args_[i.value - 1], conv)); - } - - // Parse the position of the arg and store it in `arg_position_`. - constexpr ConvParser ParseArgPosition(Integer arg) const { - return SetFormat(arg.format).SetArgPosition(arg.value); - } - - // Consume the flags. - constexpr ConvParser ParseFlags() const { - return SetFormat(ConsumeAnyOf(format_, "-+ #0")); - } - - // Consume the width. - // If it is '*', we verify that it matches `args_`. `error_` is set if it - // doesn't match. - constexpr ConvParser ParseWidth() const { - return IsDigit(GetChar(format_, 0)) - ? SetFormat(ParseDigits(format_).format) - : GetChar(format_, 0) == '*' - ? is_positional_ - ? VerifyPositional( - ParsePositional(ConsumeFront(format_)), '*') - : SetFormat(ConsumeFront(format_)) - .ConsumeNextArg('*') - : *this; - } - - // Consume the precision. - // If it is '*', we verify that it matches `args_`. `error_` is set if it - // doesn't match. - constexpr ConvParser ParsePrecision() const { - return GetChar(format_, 0) != '.' - ? *this - : GetChar(format_, 1) == '*' - ? is_positional_ - ? VerifyPositional( - ParsePositional(ConsumeFront(format_, 2)), '*') - : SetFormat(ConsumeFront(format_, 2)) - .ConsumeNextArg('*') - : SetFormat(ParseDigits(ConsumeFront(format_)).format); - } - - // Consume the length characters. - constexpr ConvParser ParseLength() const { - return SetFormat(ConsumeAnyOf(format_, "lLhjztq")); - } - - // Consume the conversion character and verify that it matches `args_`. - // `error_` is set if it doesn't match. - constexpr ConvParser ParseConversion() const { - return is_positional_ - ? VerifyPositional({ConsumeFront(format_), arg_position_}, - GetChar(format_, 0)) - : ConsumeNextArg(GetChar(format_, 0)) - .SetFormat(ConsumeFront(format_)); - } - - constexpr ConvParser(string_view format, ConvList args, bool error, - int arg_position, bool is_positional) - : format_(format), - args_(args), - error_(error), - arg_position_(arg_position), - is_positional_(is_positional) {} - - public: - constexpr ConvParser(string_view format, ConvList args, bool is_positional) - : format_(format), - args_(args), - error_(false), - arg_position_(0), - is_positional_(is_positional) {} - - // Consume the whole conversion specifier. - // `format()` will be set to the character after the conversion character. - // `error()` will be set if any of the arguments do not match. - constexpr ConvParser Run() const { - return (is_positional_ ? ParseArgPosition(ParsePositional(format_)) : *this) - .ParseFlags() - .ParseWidth() - .ParsePrecision() - .ParseLength() - .ParseConversion(); - } - - constexpr string_view format() const { return format_; } - constexpr ConvList args() const { return args_; } - constexpr bool error() const { return error_; } - constexpr bool is_positional() const { return is_positional_; } - - private: - string_view format_; - // Current list of arguments. If we are not in positional mode we will consume - // from the front. - ConvList args_; - bool error_; - // Holds the argument position of the conversion character, if we are in - // positional mode. Otherwise, it is unspecified. - int arg_position_; - // Whether we are in positional mode. - // It changes the behavior of '*' and where to find the converted argument. - bool is_positional_; -}; - -// Parses a whole format expression. -// See FormatParser::Run(). -class FormatParser { - static constexpr bool FoundPercent(string_view format) { - return format.empty() || - (GetChar(format, 0) == '%' && GetChar(format, 1) != '%'); - } - - // We use an inner function to increase the recursion limit. - // The inner function consumes up to `limit` characters on every run. - // This increases the limit from 512 to ~512*limit. - static constexpr string_view ConsumeNonPercentInner(string_view format, - int limit = 20) { - return FoundPercent(format) || !limit - ? format - : ConsumeNonPercentInner( - ConsumeFront(format, GetChar(format, 0) == '%' && - GetChar(format, 1) == '%' - ? 2 - : 1), - limit - 1); - } - - // Consume characters until the next conversion spec %. - // It skips %%. - static constexpr string_view ConsumeNonPercent(string_view format) { - return FoundPercent(format) - ? format - : ConsumeNonPercent(ConsumeNonPercentInner(format)); - } - - static constexpr bool IsPositional(string_view format) { - return IsDigit(GetChar(format, 0)) ? IsPositional(ConsumeFront(format)) - : GetChar(format, 0) == '$'; - } - - constexpr bool RunImpl(bool is_positional) const { - // In non-positional mode we require all arguments to be consumed. - // In positional mode just reaching the end of the format without errors is - // enough. - return (format_.empty() && (is_positional || args_.count == 0)) || - (!format_.empty() && - ValidateArg( - ConvParser(ConsumeFront(format_), args_, is_positional).Run())); - } - - constexpr bool ValidateArg(ConvParser conv) const { - return !conv.error() && FormatParser(conv.format(), conv.args()) - .RunImpl(conv.is_positional()); - } - - public: - constexpr FormatParser(string_view format, ConvList args) - : format_(ConsumeNonPercent(format)), args_(args) {} - - // Runs the parser for `format` and `args`. - // It verifies that the format is valid and that all conversion specifiers - // match the arguments passed. - // In non-positional mode it also verfies that all arguments are consumed. - constexpr bool Run() const { - return RunImpl(!format_.empty() && IsPositional(ConsumeFront(format_))); - } - - private: - string_view format_; - // Current list of arguments. - // If we are not in positional mode we will consume from the front and will - // have to be empty in the end. - ConvList args_; -}; - template <FormatConversionCharSet... C> constexpr bool ValidFormatImpl(string_view format) { - return FormatParser(format, - {ConvListT<sizeof...(C)>{{C...}}.list, sizeof...(C)}) - .Run(); + int next_arg = 0; + const char* p = format.data(); + const char* const end = p + format.size(); + constexpr FormatConversionCharSet + kAllowedConvs[(std::max)(sizeof...(C), size_t{1})] = {C...}; + bool used[(std::max)(sizeof...(C), size_t{1})]{}; + constexpr int kNumArgs = sizeof...(C); + while (p != end) { + while (p != end && *p != '%') ++p; + if (p == end) { + break; + } + if (p + 1 >= end) return false; + if (p[1] == '%') { + // %% + p += 2; + continue; + } + + UnboundConversion conv(absl::kConstInit); + p = ConsumeUnboundConversion(p + 1, end, &conv, &next_arg); + if (p == nullptr) return false; + if (conv.arg_position <= 0 || conv.arg_position > kNumArgs) { + return false; + } + if (!Contains(kAllowedConvs[conv.arg_position - 1], conv.conv)) { + return false; + } + used[conv.arg_position - 1] = true; + for (auto extra : {conv.width, conv.precision}) { + if (extra.is_from_arg()) { + int pos = extra.get_from_arg(); + if (pos <= 0 || pos > kNumArgs) return false; + used[pos - 1] = true; + if (!Contains(kAllowedConvs[pos - 1], '*')) { + return false; + } + } + } + } + if (sizeof...(C) != 0) { + for (bool b : used) { + if (!b) return false; + } + } + return true; } #endif // ABSL_INTERNAL_ENABLE_FORMAT_CHECKER diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/constexpr_parser.h b/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/constexpr_parser.h new file mode 100644 index 0000000000..3dc1776b42 --- /dev/null +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/constexpr_parser.h @@ -0,0 +1,351 @@ +// Copyright 2022 The Abseil Authors +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// https://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#ifndef ABSL_STRINGS_INTERNAL_STR_FORMAT_CONSTEXPR_PARSER_H_ +#define ABSL_STRINGS_INTERNAL_STR_FORMAT_CONSTEXPR_PARSER_H_ + +#include <cassert> +#include <cstdint> +#include <limits> + +#include "absl/base/const_init.h" +#include "absl/strings/internal/str_format/extension.h" + +namespace absl { +ABSL_NAMESPACE_BEGIN +namespace str_format_internal { + +enum class LengthMod : std::uint8_t { h, hh, l, ll, L, j, z, t, q, none }; + +// The analyzed properties of a single specified conversion. +struct UnboundConversion { + // This is a user defined default constructor on purpose to skip the + // initialization of parts of the object that are not necessary. + UnboundConversion() {} // NOLINT + + // This constructor is provided for the static checker. We don't want to do + // the unnecessary initialization in the normal case. + explicit constexpr UnboundConversion(absl::ConstInitType) + : arg_position{}, width{}, precision{} {} + + class InputValue { + public: + constexpr void set_value(int value) { + assert(value >= 0); + value_ = value; + } + constexpr int value() const { return value_; } + + // Marks the value as "from arg". aka the '*' format. + // Requires `value >= 1`. + // When set, is_from_arg() return true and get_from_arg() returns the + // original value. + // `value()`'s return value is unspecified in this state. + constexpr void set_from_arg(int value) { + assert(value > 0); + value_ = -value - 1; + } + constexpr bool is_from_arg() const { return value_ < -1; } + constexpr int get_from_arg() const { + assert(is_from_arg()); + return -value_ - 1; + } + + private: + int value_ = -1; + }; + + // No need to initialize. It will always be set in the parser. + int arg_position; + + InputValue width; + InputValue precision; + + Flags flags = Flags::kBasic; + LengthMod length_mod = LengthMod::none; + FormatConversionChar conv = FormatConversionCharInternal::kNone; +}; + +// Helper tag class for the table below. +// It allows fast `char -> ConversionChar/LengthMod/Flags` checking and +// conversions. +class ConvTag { + public: + constexpr ConvTag(FormatConversionChar conversion_char) // NOLINT + : tag_(static_cast<uint8_t>(conversion_char)) {} + constexpr ConvTag(LengthMod length_mod) // NOLINT + : tag_(0x80 | static_cast<uint8_t>(length_mod)) {} + constexpr ConvTag(Flags flags) // NOLINT + : tag_(0xc0 | static_cast<uint8_t>(flags)) {} + constexpr ConvTag() : tag_(0xFF) {} + + constexpr bool is_conv() const { return (tag_ & 0x80) == 0; } + constexpr bool is_length() const { return (tag_ & 0xC0) == 0x80; } + constexpr bool is_flags() const { return (tag_ & 0xE0) == 0xC0; } + + constexpr FormatConversionChar as_conv() const { + assert(is_conv()); + assert(!is_length()); + assert(!is_flags()); + return static_cast<FormatConversionChar>(tag_); + } + constexpr LengthMod as_length() const { + assert(!is_conv()); + assert(is_length()); + assert(!is_flags()); + return static_cast<LengthMod>(tag_ & 0x3F); + } + constexpr Flags as_flags() const { + assert(!is_conv()); + assert(!is_length()); + assert(is_flags()); + return static_cast<Flags>(tag_ & 0x1F); + } + + private: + uint8_t tag_; +}; + +struct ConvTagHolder { + using CC = FormatConversionCharInternal; + using LM = LengthMod; + + // Abbreviations to fit in the table below. + static constexpr auto kFSign = Flags::kSignCol; + static constexpr auto kFAlt = Flags::kAlt; + static constexpr auto kFPos = Flags::kShowPos; + static constexpr auto kFLeft = Flags::kLeft; + static constexpr auto kFZero = Flags::kZero; + + static constexpr ConvTag value[256] = { + {}, {}, {}, {}, {}, {}, {}, {}, // 00-07 + {}, {}, {}, {}, {}, {}, {}, {}, // 08-0f + {}, {}, {}, {}, {}, {}, {}, {}, // 10-17 + {}, {}, {}, {}, {}, {}, {}, {}, // 18-1f + kFSign, {}, {}, kFAlt, {}, {}, {}, {}, // !"#$%&' + {}, {}, {}, kFPos, {}, kFLeft, {}, {}, // ()*+,-./ + kFZero, {}, {}, {}, {}, {}, {}, {}, // 01234567 + {}, {}, {}, {}, {}, {}, {}, {}, // 89:;<=>? + {}, CC::A, {}, {}, {}, CC::E, CC::F, CC::G, // @ABCDEFG + {}, {}, {}, {}, LM::L, {}, {}, {}, // HIJKLMNO + {}, {}, {}, {}, {}, {}, {}, {}, // PQRSTUVW + CC::X, {}, {}, {}, {}, {}, {}, {}, // XYZ[\]^_ + {}, CC::a, {}, CC::c, CC::d, CC::e, CC::f, CC::g, // `abcdefg + LM::h, CC::i, LM::j, {}, LM::l, {}, CC::n, CC::o, // hijklmno + CC::p, LM::q, {}, CC::s, LM::t, CC::u, CC::v, {}, // pqrstuvw + CC::x, {}, LM::z, {}, {}, {}, {}, {}, // xyz{|}! + {}, {}, {}, {}, {}, {}, {}, {}, // 80-87 + {}, {}, {}, {}, {}, {}, {}, {}, // 88-8f + {}, {}, {}, {}, {}, {}, {}, {}, // 90-97 + {}, {}, {}, {}, {}, {}, {}, {}, // 98-9f + {}, {}, {}, {}, {}, {}, {}, {}, // a0-a7 + {}, {}, {}, {}, {}, {}, {}, {}, // a8-af + {}, {}, {}, {}, {}, {}, {}, {}, // b0-b7 + {}, {}, {}, {}, {}, {}, {}, {}, // b8-bf + {}, {}, {}, {}, {}, {}, {}, {}, // c0-c7 + {}, {}, {}, {}, {}, {}, {}, {}, // c8-cf + {}, {}, {}, {}, {}, {}, {}, {}, // d0-d7 + {}, {}, {}, {}, {}, {}, {}, {}, // d8-df + {}, {}, {}, {}, {}, {}, {}, {}, // e0-e7 + {}, {}, {}, {}, {}, {}, {}, {}, // e8-ef + {}, {}, {}, {}, {}, {}, {}, {}, // f0-f7 + {}, {}, {}, {}, {}, {}, {}, {}, // f8-ff + }; +}; + +// Keep a single table for all the conversion chars and length modifiers. +constexpr ConvTag GetTagForChar(char c) { + return ConvTagHolder::value[static_cast<unsigned char>(c)]; +} + +constexpr bool CheckFastPathSetting(const UnboundConversion& conv) { + bool width_precision_needed = + conv.width.value() >= 0 || conv.precision.value() >= 0; + if (width_precision_needed && conv.flags == Flags::kBasic) { +#if defined(__clang__) + // Some compilers complain about this in constexpr even when not executed, + // so only enable the error dump in clang. + fprintf(stderr, + "basic=%d left=%d show_pos=%d sign_col=%d alt=%d zero=%d " + "width=%d precision=%d\n", + conv.flags == Flags::kBasic ? 1 : 0, + FlagsContains(conv.flags, Flags::kLeft) ? 1 : 0, + FlagsContains(conv.flags, Flags::kShowPos) ? 1 : 0, + FlagsContains(conv.flags, Flags::kSignCol) ? 1 : 0, + FlagsContains(conv.flags, Flags::kAlt) ? 1 : 0, + FlagsContains(conv.flags, Flags::kZero) ? 1 : 0, conv.width.value(), + conv.precision.value()); +#endif // defined(__clang__) + return false; + } + return true; +} + +constexpr int ParseDigits(char& c, const char*& pos, const char* const end) { + int digits = c - '0'; + // We do not want to overflow `digits` so we consume at most digits10 + // digits. If there are more digits the parsing will fail later on when the + // digit doesn't match the expected characters. + int num_digits = std::numeric_limits<int>::digits10; + for (;;) { + if (ABSL_PREDICT_FALSE(pos == end)) break; + c = *pos++; + if ('0' > c || c > '9') break; + --num_digits; + if (ABSL_PREDICT_FALSE(!num_digits)) break; + digits = 10 * digits + c - '0'; + } + return digits; +} + +template <bool is_positional> +constexpr const char* ConsumeConversion(const char* pos, const char* const end, + UnboundConversion* conv, + int* next_arg) { + const char* const original_pos = pos; + char c = 0; + // Read the next char into `c` and update `pos`. Returns false if there are + // no more chars to read. +#define ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR() \ + do { \ + if (ABSL_PREDICT_FALSE(pos == end)) return nullptr; \ + c = *pos++; \ + } while (0) + + if (is_positional) { + ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR(); + if (ABSL_PREDICT_FALSE(c < '1' || c > '9')) return nullptr; + conv->arg_position = ParseDigits(c, pos, end); + assert(conv->arg_position > 0); + if (ABSL_PREDICT_FALSE(c != '$')) return nullptr; + } + + ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR(); + + // We should start with the basic flag on. + assert(conv->flags == Flags::kBasic); + + // Any non alpha character makes this conversion not basic. + // This includes flags (-+ #0), width (1-9, *) or precision (.). + // All conversion characters and length modifiers are alpha characters. + if (c < 'A') { + while (c <= '0') { + auto tag = GetTagForChar(c); + if (tag.is_flags()) { + conv->flags = conv->flags | tag.as_flags(); + ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR(); + } else { + break; + } + } + + if (c <= '9') { + if (c >= '0') { + int maybe_width = ParseDigits(c, pos, end); + if (!is_positional && c == '$') { + if (ABSL_PREDICT_FALSE(*next_arg != 0)) return nullptr; + // Positional conversion. + *next_arg = -1; + return ConsumeConversion<true>(original_pos, end, conv, next_arg); + } + conv->flags = conv->flags | Flags::kNonBasic; + conv->width.set_value(maybe_width); + } else if (c == '*') { + conv->flags = conv->flags | Flags::kNonBasic; + ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR(); + if (is_positional) { + if (ABSL_PREDICT_FALSE(c < '1' || c > '9')) return nullptr; + conv->width.set_from_arg(ParseDigits(c, pos, end)); + if (ABSL_PREDICT_FALSE(c != '$')) return nullptr; + ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR(); + } else { + conv->width.set_from_arg(++*next_arg); + } + } + } + + if (c == '.') { + conv->flags = conv->flags | Flags::kNonBasic; + ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR(); + if ('0' <= c && c <= '9') { + conv->precision.set_value(ParseDigits(c, pos, end)); + } else if (c == '*') { + ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR(); + if (is_positional) { + if (ABSL_PREDICT_FALSE(c < '1' || c > '9')) return nullptr; + conv->precision.set_from_arg(ParseDigits(c, pos, end)); + if (c != '$') return nullptr; + ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR(); + } else { + conv->precision.set_from_arg(++*next_arg); + } + } else { + conv->precision.set_value(0); + } + } + } + + auto tag = GetTagForChar(c); + + if (ABSL_PREDICT_FALSE(c == 'v' && conv->flags != Flags::kBasic)) { + return nullptr; + } + + if (ABSL_PREDICT_FALSE(!tag.is_conv())) { + if (ABSL_PREDICT_FALSE(!tag.is_length())) return nullptr; + + // It is a length modifier. + using str_format_internal::LengthMod; + LengthMod length_mod = tag.as_length(); + ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR(); + if (c == 'h' && length_mod == LengthMod::h) { + conv->length_mod = LengthMod::hh; + ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR(); + } else if (c == 'l' && length_mod == LengthMod::l) { + conv->length_mod = LengthMod::ll; + ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR(); + } else { + conv->length_mod = length_mod; + } + tag = GetTagForChar(c); + + if (ABSL_PREDICT_FALSE(c == 'v')) return nullptr; + if (ABSL_PREDICT_FALSE(!tag.is_conv())) return nullptr; + } + + assert(CheckFastPathSetting(*conv)); + (void)(&CheckFastPathSetting); + + conv->conv = tag.as_conv(); + if (!is_positional) conv->arg_position = ++*next_arg; + return pos; +} + +// Consume conversion spec prefix (not including '%') of [p, end) if valid. +// Examples of valid specs would be e.g.: "s", "d", "-12.6f". +// If valid, it returns the first character following the conversion spec, +// and the spec part is broken down and returned in 'conv'. +// If invalid, returns nullptr. +constexpr const char* ConsumeUnboundConversion(const char* p, const char* end, + UnboundConversion* conv, + int* next_arg) { + if (*next_arg < 0) return ConsumeConversion<true>(p, end, conv, next_arg); + return ConsumeConversion<false>(p, end, conv, next_arg); +} + +} // namespace str_format_internal +ABSL_NAMESPACE_END +} // namespace absl + +#endif // ABSL_STRINGS_INTERNAL_STR_FORMAT_CONSTEXPR_PARSER_H_ diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/extension.cc b/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/extension.cc index f93153d5c0..2a0ceb13d7 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/extension.cc +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/extension.cc @@ -58,7 +58,8 @@ constexpr FormatConversionCharSet FormatConversionCharSetInternal::kPointer; bool FormatSinkImpl::PutPaddedString(string_view value, int width, int precision, bool left) { size_t space_remaining = 0; - if (width >= 0) space_remaining = width; + if (width >= 0) + space_remaining = static_cast<size_t>(width); size_t n = value.size(); if (precision >= 0) n = std::min(n, static_cast<size_t>(precision)); string_view shown(value.data(), n); diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/extension.h b/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/extension.h index 55e8ac882d..603bd49d18 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/extension.h +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/extension.h @@ -169,7 +169,7 @@ inline std::ostream& operator<<(std::ostream& os, Flags v) { X_VAL(f) X_SEP X_VAL(F) X_SEP X_VAL(e) X_SEP X_VAL(E) X_SEP \ X_VAL(g) X_SEP X_VAL(G) X_SEP X_VAL(a) X_SEP X_VAL(A) X_SEP \ /* misc */ \ - X_VAL(n) X_SEP X_VAL(p) + X_VAL(n) X_SEP X_VAL(p) X_SEP X_VAL(v) // clang-format on // This type should not be referenced, it exists only to provide labels @@ -191,7 +191,7 @@ struct FormatConversionCharInternal { c, s, // text d, i, o, u, x, X, // int f, F, e, E, g, G, a, A, // float - n, p, // misc + n, p, v, // misc kNone }; // clang-format on @@ -292,6 +292,8 @@ class FormatConversionSpecImpl { return conv_; } + void set_conversion_char(FormatConversionChar c) { conv_ = c; } + // Returns the specified width. If width is unspecfied, it returns a negative // value. int width() const { return width_; } diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/float_conversion.cc b/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/float_conversion.cc index b1c4068475..8e497852bb 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/float_conversion.cc +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/float_conversion.cc @@ -92,27 +92,30 @@ class StackArray { // Calculates `10 * (*v) + carry` and stores the result in `*v` and returns // the carry. +// Requires: `0 <= carry <= 9` template <typename Int> -inline Int MultiplyBy10WithCarry(Int *v, Int carry) { +inline char MultiplyBy10WithCarry(Int* v, char carry) { using BiggerInt = absl::conditional_t<sizeof(Int) == 4, uint64_t, uint128>; - BiggerInt tmp = 10 * static_cast<BiggerInt>(*v) + carry; + BiggerInt tmp = + 10 * static_cast<BiggerInt>(*v) + static_cast<BiggerInt>(carry); *v = static_cast<Int>(tmp); - return static_cast<Int>(tmp >> (sizeof(Int) * 8)); + return static_cast<char>(tmp >> (sizeof(Int) * 8)); } // Calculates `(2^64 * carry + *v) / 10`. // Stores the quotient in `*v` and returns the remainder. // Requires: `0 <= carry <= 9` -inline uint64_t DivideBy10WithCarry(uint64_t *v, uint64_t carry) { +inline char DivideBy10WithCarry(uint64_t* v, char carry) { constexpr uint64_t divisor = 10; // 2^64 / divisor = chunk_quotient + chunk_remainder / divisor constexpr uint64_t chunk_quotient = (uint64_t{1} << 63) / (divisor / 2); constexpr uint64_t chunk_remainder = uint64_t{} - chunk_quotient * divisor; + const uint64_t carry_u64 = static_cast<uint64_t>(carry); const uint64_t mod = *v % divisor; - const uint64_t next_carry = chunk_remainder * carry + mod; - *v = *v / divisor + carry * chunk_quotient + next_carry / divisor; - return next_carry % divisor; + const uint64_t next_carry = chunk_remainder * carry_u64 + mod; + *v = *v / divisor + carry_u64 * chunk_quotient + next_carry / divisor; + return static_cast<char>(next_carry % divisor); } using MaxFloatType = @@ -125,11 +128,11 @@ using MaxFloatType = // // Requires `0 <= exp` and `exp <= numeric_limits<MaxFloatType>::max_exponent`. class BinaryToDecimal { - static constexpr int ChunksNeeded(int exp) { + static constexpr size_t ChunksNeeded(int exp) { // We will left shift a uint128 by `exp` bits, so we need `128+exp` total // bits. Round up to 32. // See constructor for details about adding `10%` to the value. - return (128 + exp + 31) / 32 * 11 / 10; + return static_cast<size_t>((128 + exp + 31) / 32 * 11 / 10); } public: @@ -140,7 +143,7 @@ class BinaryToDecimal { assert(exp > 0); assert(exp <= std::numeric_limits<MaxFloatType>::max_exponent); static_assert( - static_cast<int>(StackArray::kMaxCapacity) >= + StackArray::kMaxCapacity >= ChunksNeeded(std::numeric_limits<MaxFloatType>::max_exponent), ""); @@ -149,9 +152,9 @@ class BinaryToDecimal { [=](absl::Span<uint32_t> input) { f(BinaryToDecimal(input, v, exp)); }); } - int TotalDigits() const { - return static_cast<int>((decimal_end_ - decimal_start_) * kDigitsPerChunk + - CurrentDigits().size()); + size_t TotalDigits() const { + return (decimal_end_ - decimal_start_) * kDigitsPerChunk + + CurrentDigits().size(); } // See the current block of digits. @@ -190,30 +193,31 @@ class BinaryToDecimal { // the decimal representation is around 7% less efficient in space than the // binary one. We allocate an extra 10% memory to account for this. See // ChunksNeeded for this calculation. - int chunk_index = exp / 32; + size_t after_chunk_index = static_cast<size_t>(exp / 32 + 1); decimal_start_ = decimal_end_ = ChunksNeeded(exp); const int offset = exp % 32; // Left shift v by exp bits. - data_[chunk_index] = static_cast<uint32_t>(v << offset); + data_[after_chunk_index - 1] = static_cast<uint32_t>(v << offset); for (v >>= (32 - offset); v; v >>= 32) - data_[++chunk_index] = static_cast<uint32_t>(v); + data_[++after_chunk_index - 1] = static_cast<uint32_t>(v); - while (chunk_index >= 0) { + while (after_chunk_index > 0) { // While we have more than one chunk available, go in steps of 1e9. - // `data_[chunk_index]` holds the highest non-zero binary chunk, so keep - // the variable updated. + // `data_[after_chunk_index - 1]` holds the highest non-zero binary chunk, + // so keep the variable updated. uint32_t carry = 0; - for (int i = chunk_index; i >= 0; --i) { - uint64_t tmp = uint64_t{data_[i]} + (uint64_t{carry} << 32); - data_[i] = static_cast<uint32_t>(tmp / uint64_t{1000000000}); + for (size_t i = after_chunk_index; i > 0; --i) { + uint64_t tmp = uint64_t{data_[i - 1]} + (uint64_t{carry} << 32); + data_[i - 1] = static_cast<uint32_t>(tmp / uint64_t{1000000000}); carry = static_cast<uint32_t>(tmp % uint64_t{1000000000}); } // If the highest chunk is now empty, remove it from view. - if (data_[chunk_index] == 0) --chunk_index; + if (data_[after_chunk_index - 1] == 0) + --after_chunk_index; --decimal_start_; - assert(decimal_start_ != chunk_index); + assert(decimal_start_ != after_chunk_index - 1); data_[decimal_start_] = carry; } @@ -225,13 +229,13 @@ class BinaryToDecimal { } private: - static constexpr int kDigitsPerChunk = 9; + static constexpr size_t kDigitsPerChunk = 9; - int decimal_start_; - int decimal_end_; + size_t decimal_start_; + size_t decimal_end_; char digits_[kDigitsPerChunk]; - int size_ = 0; + size_t size_ = 0; absl::Span<uint32_t> data_; }; @@ -251,25 +255,26 @@ class FractionalDigitGenerator { static_assert(StackArray::kMaxCapacity >= (Limits::digits + 128 - Limits::min_exponent + 31) / 32, ""); - StackArray::RunWithCapacity((Limits::digits + exp + 31) / 32, - [=](absl::Span<uint32_t> input) { - f(FractionalDigitGenerator(input, v, exp)); - }); + StackArray::RunWithCapacity( + static_cast<size_t>((Limits::digits + exp + 31) / 32), + [=](absl::Span<uint32_t> input) { + f(FractionalDigitGenerator(input, v, exp)); + }); } // Returns true if there are any more non-zero digits left. - bool HasMoreDigits() const { return next_digit_ != 0 || chunk_index_ >= 0; } + bool HasMoreDigits() const { return next_digit_ != 0 || after_chunk_index_; } // Returns true if the remainder digits are greater than 5000... bool IsGreaterThanHalf() const { - return next_digit_ > 5 || (next_digit_ == 5 && chunk_index_ >= 0); + return next_digit_ > 5 || (next_digit_ == 5 && after_chunk_index_); } // Returns true if the remainder digits are exactly 5000... - bool IsExactlyHalf() const { return next_digit_ == 5 && chunk_index_ < 0; } + bool IsExactlyHalf() const { return next_digit_ == 5 && !after_chunk_index_; } struct Digits { - int digit_before_nine; - int num_nines; + char digit_before_nine; + size_t num_nines; }; // Get the next set of digits. @@ -288,35 +293,37 @@ class FractionalDigitGenerator { private: // Return the next digit. - int GetOneDigit() { - if (chunk_index_ < 0) return 0; + char GetOneDigit() { + if (!after_chunk_index_) + return 0; - uint32_t carry = 0; - for (int i = chunk_index_; i >= 0; --i) { - carry = MultiplyBy10WithCarry(&data_[i], carry); + char carry = 0; + for (size_t i = after_chunk_index_; i > 0; --i) { + carry = MultiplyBy10WithCarry(&data_[i - 1], carry); } // If the lowest chunk is now empty, remove it from view. - if (data_[chunk_index_] == 0) --chunk_index_; + if (data_[after_chunk_index_ - 1] == 0) + --after_chunk_index_; return carry; } FractionalDigitGenerator(absl::Span<uint32_t> data, uint128 v, int exp) - : chunk_index_(exp / 32), data_(data) { + : after_chunk_index_(static_cast<size_t>(exp / 32 + 1)), data_(data) { const int offset = exp % 32; // Right shift `v` by `exp` bits. - data_[chunk_index_] = static_cast<uint32_t>(v << (32 - offset)); + data_[after_chunk_index_ - 1] = static_cast<uint32_t>(v << (32 - offset)); v >>= offset; // Make sure we don't overflow the data. We already calculated that // non-zero bits fit, so we might not have space for leading zero bits. - for (int pos = chunk_index_; v; v >>= 32) + for (size_t pos = after_chunk_index_ - 1; v; v >>= 32) data_[--pos] = static_cast<uint32_t>(v); // Fill next_digit_, as GetDigits expects it to be populated always. next_digit_ = GetOneDigit(); } - int next_digit_; - int chunk_index_; + char next_digit_; + size_t after_chunk_index_; absl::Span<uint32_t> data_; }; @@ -362,7 +369,7 @@ char *PrintIntegralDigitsFromRightFast(uint128 v, char *p) { auto low = static_cast<uint64_t>(v); while (high != 0) { - uint64_t carry = DivideBy10WithCarry(&high, 0); + char carry = DivideBy10WithCarry(&high, 0); carry = DivideBy10WithCarry(&low, carry); *--p = carry + '0'; } @@ -373,13 +380,15 @@ char *PrintIntegralDigitsFromRightFast(uint128 v, char *p) { // shifting. // Performs rounding if necessary to fit within `precision`. // Returns the pointer to one after the last character written. -char *PrintFractionalDigitsFast(uint64_t v, char *start, int exp, - int precision) { +char* PrintFractionalDigitsFast(uint64_t v, + char* start, + int exp, + size_t precision) { char *p = start; v <<= (64 - exp); while (precision > 0) { if (!v) return p; - *p++ = MultiplyBy10WithCarry(&v, uint64_t{0}) + '0'; + *p++ = MultiplyBy10WithCarry(&v, 0) + '0'; --precision; } @@ -393,8 +402,6 @@ char *PrintFractionalDigitsFast(uint64_t v, char *start, int exp, RoundToEven(p - 1); } - assert(precision == 0); - // Precision can only be zero here. return p; } @@ -402,8 +409,10 @@ char *PrintFractionalDigitsFast(uint64_t v, char *start, int exp, // after shifting. // Performs rounding if necessary to fit within `precision`. // Returns the pointer to one after the last character written. -char *PrintFractionalDigitsFast(uint128 v, char *start, int exp, - int precision) { +char* PrintFractionalDigitsFast(uint128 v, + char* start, + int exp, + size_t precision) { char *p = start; v <<= (128 - exp); auto high = static_cast<uint64_t>(v >> 64); @@ -412,7 +421,7 @@ char *PrintFractionalDigitsFast(uint128 v, char *start, int exp, // While we have digits to print and `low` is not empty, do the long // multiplication. while (precision > 0 && low != 0) { - uint64_t carry = MultiplyBy10WithCarry(&low, uint64_t{0}); + char carry = MultiplyBy10WithCarry(&low, 0); carry = MultiplyBy10WithCarry(&high, carry); *p++ = carry + '0'; @@ -424,7 +433,7 @@ char *PrintFractionalDigitsFast(uint128 v, char *start, int exp, // above. while (precision > 0) { if (!high) return p; - *p++ = MultiplyBy10WithCarry(&high, uint64_t{0}) + '0'; + *p++ = MultiplyBy10WithCarry(&high, 0) + '0'; --precision; } @@ -438,14 +447,12 @@ char *PrintFractionalDigitsFast(uint128 v, char *start, int exp, RoundToEven(p - 1); } - assert(precision == 0); - // Precision can only be zero here. return p; } struct FormatState { char sign_char; - int precision; + size_t precision; const FormatConversionSpecImpl &conv; FormatSinkImpl *sink; @@ -455,9 +462,9 @@ struct FormatState { }; struct Padding { - int left_spaces; - int zeros; - int right_spaces; + size_t left_spaces; + size_t zeros; + size_t right_spaces; }; Padding ExtraWidthToPadding(size_t total_size, const FormatState &state) { @@ -465,7 +472,7 @@ Padding ExtraWidthToPadding(size_t total_size, const FormatState &state) { static_cast<size_t>(state.conv.width()) <= total_size) { return {0, 0, 0}; } - int missing_chars = state.conv.width() - total_size; + size_t missing_chars = static_cast<size_t>(state.conv.width()) - total_size; if (state.conv.has_left_flag()) { return {0, 0, missing_chars}; } else if (state.conv.has_zero_flag()) { @@ -475,8 +482,10 @@ Padding ExtraWidthToPadding(size_t total_size, const FormatState &state) { } } -void FinalPrint(const FormatState &state, absl::string_view data, - int padding_offset, int trailing_zeros, +void FinalPrint(const FormatState& state, + absl::string_view data, + size_t padding_offset, + size_t trailing_zeros, absl::string_view data_postfix) { if (state.conv.width() < 0) { // No width specified. Fast-path. @@ -487,10 +496,10 @@ void FinalPrint(const FormatState &state, absl::string_view data, return; } - auto padding = ExtraWidthToPadding((state.sign_char != '\0' ? 1 : 0) + - data.size() + data_postfix.size() + - static_cast<size_t>(trailing_zeros), - state); + auto padding = + ExtraWidthToPadding((state.sign_char != '\0' ? 1 : 0) + data.size() + + data_postfix.size() + trailing_zeros, + state); state.sink->Append(padding.left_spaces, ' '); if (state.sign_char != '\0') state.sink->Append(1, state.sign_char); @@ -547,15 +556,16 @@ void FormatFFast(Int v, int exp, const FormatState &state) { if (integral_digits_start[-1] != '0') --integral_digits_start; } - size_t size = fractional_digits_end - integral_digits_start; + size_t size = + static_cast<size_t>(fractional_digits_end - integral_digits_start); // In `alt` mode (flag #) we keep the `.` even if there are no fractional // digits. In non-alt mode, we strip it. if (!state.ShouldPrintDot()) --size; FinalPrint(state, absl::string_view(integral_digits_start, size), /*padding_offset=*/0, - static_cast<int>(state.precision - (fractional_digits_end - - fractional_digits_start)), + state.precision - static_cast<size_t>(fractional_digits_end - + fractional_digits_start), /*data_postfix=*/""); } @@ -567,21 +577,22 @@ void FormatFFast(Int v, int exp, const FormatState &state) { void FormatFPositiveExpSlow(uint128 v, int exp, const FormatState &state) { BinaryToDecimal::RunConversion(v, exp, [&](BinaryToDecimal btd) { const size_t total_digits = - btd.TotalDigits() + - (state.ShouldPrintDot() ? static_cast<size_t>(state.precision) + 1 : 0); + btd.TotalDigits() + (state.ShouldPrintDot() ? state.precision + 1 : 0); const auto padding = ExtraWidthToPadding( total_digits + (state.sign_char != '\0' ? 1 : 0), state); state.sink->Append(padding.left_spaces, ' '); - if (state.sign_char != '\0') state.sink->Append(1, state.sign_char); + if (state.sign_char != '\0') + state.sink->Append(1, state.sign_char); state.sink->Append(padding.zeros, '0'); do { state.sink->Append(btd.CurrentDigits()); } while (btd.AdvanceDigits()); - if (state.ShouldPrintDot()) state.sink->Append(1, '.'); + if (state.ShouldPrintDot()) + state.sink->Append(1, '.'); state.sink->Append(state.precision, '0'); state.sink->Append(padding.right_spaces, ' '); }); @@ -594,8 +605,7 @@ void FormatFPositiveExpSlow(uint128 v, int exp, const FormatState &state) { // digits. void FormatFNegativeExpSlow(uint128 v, int exp, const FormatState &state) { const size_t total_digits = - /* 0 */ 1 + - (state.ShouldPrintDot() ? static_cast<size_t>(state.precision) + 1 : 0); + /* 0 */ 1 + (state.ShouldPrintDot() ? state.precision + 1 : 0); auto padding = ExtraWidthToPadding(total_digits + (state.sign_char ? 1 : 0), state); padding.zeros += 1; @@ -606,7 +616,7 @@ void FormatFNegativeExpSlow(uint128 v, int exp, const FormatState &state) { if (state.ShouldPrintDot()) state.sink->Append(1, '.'); // Print digits - int digits_to_go = state.precision; + size_t digits_to_go = state.precision; FractionalDigitGenerator::RunConversion( v, exp, [&](FractionalDigitGenerator digit_gen) { @@ -666,7 +676,8 @@ void FormatFNegativeExpSlow(uint128 v, int exp, const FormatState &state) { template <typename Int> void FormatF(Int mantissa, int exp, const FormatState &state) { if (exp >= 0) { - const int total_bits = sizeof(Int) * 8 - LeadingZeros(mantissa) + exp; + const int total_bits = + static_cast<int>(sizeof(Int) * 8) - LeadingZeros(mantissa) + exp; // Fallback to the slow stack-based approach if we can't do it in a 64 or // 128 bit state. @@ -686,9 +697,9 @@ void FormatF(Int mantissa, int exp, const FormatState &state) { // Grab the group of four bits (nibble) from `n`. E.g., nibble 1 corresponds to // bits 4-7. template <typename Int> -uint8_t GetNibble(Int n, int nibble_index) { +uint8_t GetNibble(Int n, size_t nibble_index) { constexpr Int mask_low_nibble = Int{0xf}; - int shift = nibble_index * 4; + int shift = static_cast<int>(nibble_index * 4); n &= mask_low_nibble << shift; return static_cast<uint8_t>((n >> shift) & 0xf); } @@ -696,9 +707,9 @@ uint8_t GetNibble(Int n, int nibble_index) { // Add one to the given nibble, applying carry to higher nibbles. Returns true // if overflow, false otherwise. template <typename Int> -bool IncrementNibble(int nibble_index, Int *n) { - constexpr int kShift = sizeof(Int) * 8 - 1; - constexpr int kNumNibbles = sizeof(Int) * 8 / 4; +bool IncrementNibble(size_t nibble_index, Int* n) { + constexpr size_t kShift = sizeof(Int) * 8 - 1; + constexpr size_t kNumNibbles = sizeof(Int) * 8 / 4; Int before = *n >> kShift; // Here we essentially want to take the number 1 and move it into the requsted // nibble, then add it to *n to effectively increment the nibble. However, @@ -706,28 +717,32 @@ bool IncrementNibble(int nibble_index, Int *n) { // i.e., if the nibble_index is out of range. So therefore we check for this // and if we are out of range we just add 0 which leaves *n unchanged, which // seems like the reasonable thing to do in that case. - *n += ((nibble_index >= kNumNibbles) ? 0 : (Int{1} << (nibble_index * 4))); + *n += ((nibble_index >= kNumNibbles) + ? 0 + : (Int{1} << static_cast<int>(nibble_index * 4))); Int after = *n >> kShift; return (before && !after) || (nibble_index >= kNumNibbles); } // Return a mask with 1's in the given nibble and all lower nibbles. template <typename Int> -Int MaskUpToNibbleInclusive(int nibble_index) { - constexpr int kNumNibbles = sizeof(Int) * 8 / 4; +Int MaskUpToNibbleInclusive(size_t nibble_index) { + constexpr size_t kNumNibbles = sizeof(Int) * 8 / 4; static const Int ones = ~Int{0}; - return ones >> std::max(0, 4 * (kNumNibbles - nibble_index - 1)); + ++nibble_index; + return ones >> static_cast<int>( + 4 * (std::max(kNumNibbles, nibble_index) - nibble_index)); } // Return a mask with 1's below the given nibble. template <typename Int> -Int MaskUpToNibbleExclusive(int nibble_index) { - return nibble_index <= 0 ? 0 : MaskUpToNibbleInclusive<Int>(nibble_index - 1); +Int MaskUpToNibbleExclusive(size_t nibble_index) { + return nibble_index == 0 ? 0 : MaskUpToNibbleInclusive<Int>(nibble_index - 1); } template <typename Int> -Int MoveToNibble(uint8_t nibble, int nibble_index) { - return Int{nibble} << (4 * nibble_index); +Int MoveToNibble(uint8_t nibble, size_t nibble_index) { + return Int{nibble} << static_cast<int>(4 * nibble_index); } // Given mantissa size, find optimal # of mantissa bits to put in initial digit. @@ -744,10 +759,10 @@ Int MoveToNibble(uint8_t nibble, int nibble_index) { // a multiple of four. Once again, the goal is to have all fractional digits // represent real precision. template <typename Float> -constexpr int HexFloatLeadingDigitSizeInBits() { +constexpr size_t HexFloatLeadingDigitSizeInBits() { return std::numeric_limits<Float>::digits % 4 > 0 - ? std::numeric_limits<Float>::digits % 4 - : 4; + ? static_cast<size_t>(std::numeric_limits<Float>::digits % 4) + : size_t{4}; } // This function captures the rounding behavior of glibc for hex float @@ -757,16 +772,17 @@ constexpr int HexFloatLeadingDigitSizeInBits() { // point that is not followed by 800000..., it disregards the parity and rounds // up if > 8 and rounds down if < 8. template <typename Int> -bool HexFloatNeedsRoundUp(Int mantissa, int final_nibble_displayed, +bool HexFloatNeedsRoundUp(Int mantissa, + size_t final_nibble_displayed, uint8_t leading) { // If the last nibble (hex digit) to be displayed is the lowest on in the // mantissa then that means that we don't have any further nibbles to inform // rounding, so don't round. - if (final_nibble_displayed <= 0) { + if (final_nibble_displayed == 0) { return false; } - int rounding_nibble_idx = final_nibble_displayed - 1; - constexpr int kTotalNibbles = sizeof(Int) * 8 / 4; + size_t rounding_nibble_idx = final_nibble_displayed - 1; + constexpr size_t kTotalNibbles = sizeof(Int) * 8 / 4; assert(final_nibble_displayed <= kTotalNibbles); Int mantissa_up_to_rounding_nibble_inclusive = mantissa & MaskUpToNibbleInclusive<Int>(rounding_nibble_idx); @@ -793,7 +809,7 @@ struct HexFloatTypeParams { } int min_exponent; - int leading_digit_size_bits; + size_t leading_digit_size_bits; }; // Hex Float Rounding. First check if we need to round; if so, then we do that @@ -803,10 +819,12 @@ struct HexFloatTypeParams { template <typename Int> void FormatARound(bool precision_specified, const FormatState &state, uint8_t *leading, Int *mantissa, int *exp) { - constexpr int kTotalNibbles = sizeof(Int) * 8 / 4; + constexpr size_t kTotalNibbles = sizeof(Int) * 8 / 4; // Index of the last nibble that we could display given precision. - int final_nibble_displayed = - precision_specified ? std::max(0, (kTotalNibbles - state.precision)) : 0; + size_t final_nibble_displayed = + precision_specified + ? (std::max(kTotalNibbles, state.precision) - state.precision) + : 0; if (HexFloatNeedsRoundUp(*mantissa, final_nibble_displayed, *leading)) { // Need to round up. bool overflow = IncrementNibble(final_nibble_displayed, mantissa); @@ -830,9 +848,9 @@ void FormatARound(bool precision_specified, const FormatState &state, template <typename Int> void FormatANormalize(const HexFloatTypeParams float_traits, uint8_t *leading, Int *mantissa, int *exp) { - constexpr int kIntBits = sizeof(Int) * 8; + constexpr size_t kIntBits = sizeof(Int) * 8; static const Int kHighIntBit = Int{1} << (kIntBits - 1); - const int kLeadDigitBitsCount = float_traits.leading_digit_size_bits; + const size_t kLeadDigitBitsCount = float_traits.leading_digit_size_bits; // Normalize mantissa so that highest bit set is in MSB position, unless we // get interrupted by the exponent threshold. while (*mantissa && !(*mantissa & kHighIntBit)) { @@ -846,18 +864,18 @@ void FormatANormalize(const HexFloatTypeParams float_traits, uint8_t *leading, } // Extract bits for leading digit then shift them away leaving the // fractional part. - *leading = - static_cast<uint8_t>(*mantissa >> (kIntBits - kLeadDigitBitsCount)); - *exp -= (*mantissa != 0) ? kLeadDigitBitsCount : *exp; - *mantissa <<= kLeadDigitBitsCount; + *leading = static_cast<uint8_t>( + *mantissa >> static_cast<int>(kIntBits - kLeadDigitBitsCount)); + *exp -= (*mantissa != 0) ? static_cast<int>(kLeadDigitBitsCount) : *exp; + *mantissa <<= static_cast<int>(kLeadDigitBitsCount); } template <typename Int> void FormatA(const HexFloatTypeParams float_traits, Int mantissa, int exp, bool uppercase, const FormatState &state) { // Int properties. - constexpr int kIntBits = sizeof(Int) * 8; - constexpr int kTotalNibbles = sizeof(Int) * 8 / 4; + constexpr size_t kIntBits = sizeof(Int) * 8; + constexpr size_t kTotalNibbles = sizeof(Int) * 8 / 4; // Did the user specify a precision explicitly? const bool precision_specified = state.conv.precision() >= 0; @@ -903,16 +921,19 @@ void FormatA(const HexFloatTypeParams float_traits, Int mantissa, int exp, } // ============ Fractional Digits ============ - int digits_emitted = 0; + size_t digits_emitted = 0; while (mantissa > 0) { *digits_iter++ = digits[GetNibble(mantissa, kTotalNibbles - 1)]; mantissa <<= 4; ++digits_emitted; } - int trailing_zeros = - precision_specified ? state.precision - digits_emitted : 0; - assert(trailing_zeros >= 0); - auto digits_result = string_view(digits_buffer, digits_iter - digits_buffer); + size_t trailing_zeros = 0; + if (precision_specified) { + assert(state.precision >= digits_emitted); + trailing_zeros = state.precision - digits_emitted; + } + auto digits_result = string_view( + digits_buffer, static_cast<size_t>(digits_iter - digits_buffer)); // =============== Exponent ================== constexpr size_t kBufSizeForExpDecRepr = @@ -925,11 +946,11 @@ void FormatA(const HexFloatTypeParams float_traits, Int mantissa, int exp, numbers_internal::FastIntToBuffer(exp < 0 ? -exp : exp, exp_buffer + 2); // ============ Assemble Result ============== - FinalPrint(state, // - digits_result, // 0xN.NNN... - 2, // offset in `data` to start padding if needed. - trailing_zeros, // num remaining mantissa padding zeros - exp_buffer); // exponent + FinalPrint(state, + digits_result, // 0xN.NNN... + 2, // offset of any padding + static_cast<size_t>(trailing_zeros), // remaining mantissa padding + exp_buffer); // exponent } char *CopyStringTo(absl::string_view v, char *out) { @@ -961,10 +982,10 @@ bool FallbackToSnprintf(const Float v, const FormatConversionSpecImpl &conv, int n = snprintf(&space[0], space.size(), fmt, w, p, v); if (n < 0) return false; if (static_cast<size_t>(n) < space.size()) { - result = absl::string_view(space.data(), n); + result = absl::string_view(space.data(), static_cast<size_t>(n)); break; } - space.resize(n + 1); + space.resize(static_cast<size_t>(n) + 1); } sink->Append(result); return true; @@ -972,13 +993,13 @@ bool FallbackToSnprintf(const Float v, const FormatConversionSpecImpl &conv, // 128-bits in decimal: ceil(128*log(2)/log(10)) // or std::numeric_limits<__uint128_t>::digits10 -constexpr int kMaxFixedPrecision = 39; +constexpr size_t kMaxFixedPrecision = 39; -constexpr int kBufferLength = /*sign*/ 1 + - /*integer*/ kMaxFixedPrecision + - /*point*/ 1 + - /*fraction*/ kMaxFixedPrecision + - /*exponent e+123*/ 5; +constexpr size_t kBufferLength = /*sign*/ 1 + + /*integer*/ kMaxFixedPrecision + + /*point*/ 1 + + /*fraction*/ kMaxFixedPrecision + + /*exponent e+123*/ 5; struct Buffer { void push_front(char c) { @@ -1001,7 +1022,7 @@ struct Buffer { char last_digit() const { return end[-1] == '.' ? end[-2] : end[-1]; } - int size() const { return static_cast<int>(end - begin); } + size_t size() const { return static_cast<size_t>(end - begin); } char data[kBufferLength]; char *begin; @@ -1030,8 +1051,9 @@ bool ConvertNonNumericFloats(char sign_char, Float v, return false; } - return sink->PutPaddedString(string_view(text, ptr - text), conv.width(), -1, - conv.has_left_flag()); + return sink->PutPaddedString( + string_view(text, static_cast<size_t>(ptr - text)), conv.width(), -1, + conv.has_left_flag()); } // Round up the last digit of the value. @@ -1068,12 +1090,12 @@ void PrintExponent(int exp, char e, Buffer *out) { } // Exponent digits. if (exp > 99) { - out->push_back(exp / 100 + '0'); - out->push_back(exp / 10 % 10 + '0'); - out->push_back(exp % 10 + '0'); + out->push_back(static_cast<char>(exp / 100 + '0')); + out->push_back(static_cast<char>(exp / 10 % 10 + '0')); + out->push_back(static_cast<char>(exp % 10 + '0')); } else { - out->push_back(exp / 10 + '0'); - out->push_back(exp % 10 + '0'); + out->push_back(static_cast<char>(exp / 10 + '0')); + out->push_back(static_cast<char>(exp % 10 + '0')); } } @@ -1115,8 +1137,8 @@ Decomposed<Float> Decompose(Float v) { // In Fixed mode, we add a '.' at the end. // In Precision mode, we add a '.' after the first digit. template <FormatStyle mode, typename Int> -int PrintIntegralDigits(Int digits, Buffer *out) { - int printed = 0; +size_t PrintIntegralDigits(Int digits, Buffer* out) { + size_t printed = 0; if (digits) { for (; digits; digits /= 10) out->push_front(digits % 10 + '0'); printed = out->size(); @@ -1135,10 +1157,10 @@ int PrintIntegralDigits(Int digits, Buffer *out) { } // Back out 'extra_digits' digits and round up if necessary. -bool RemoveExtraPrecision(int extra_digits, bool has_leftover_value, - Buffer *out, int *exp_out) { - if (extra_digits <= 0) return false; - +void RemoveExtraPrecision(size_t extra_digits, + bool has_leftover_value, + Buffer* out, + int* exp_out) { // Back out the extra digits out->end -= extra_digits; @@ -1158,15 +1180,17 @@ bool RemoveExtraPrecision(int extra_digits, bool has_leftover_value, if (needs_to_round_up) { RoundUp<FormatStyle::Precision>(out, exp_out); } - return true; } // Print the value into the buffer. // This will not include the exponent, which will be returned in 'exp_out' for // Precision mode. template <typename Int, typename Float, FormatStyle mode> -bool FloatToBufferImpl(Int int_mantissa, int exp, int precision, Buffer *out, - int *exp_out) { +bool FloatToBufferImpl(Int int_mantissa, + int exp, + size_t precision, + Buffer* out, + int* exp_out) { assert((CanFitMantissa<Float, Int>())); const int int_bits = std::numeric_limits<Int>::digits; @@ -1182,14 +1206,16 @@ bool FloatToBufferImpl(Int int_mantissa, int exp, int precision, Buffer *out, // The value will overflow the Int return false; } - int digits_printed = PrintIntegralDigits<mode>(int_mantissa << exp, out); - int digits_to_zero_pad = precision; + size_t digits_printed = PrintIntegralDigits<mode>(int_mantissa << exp, out); + size_t digits_to_zero_pad = precision; if (mode == FormatStyle::Precision) { - *exp_out = digits_printed - 1; - digits_to_zero_pad -= digits_printed - 1; - if (RemoveExtraPrecision(-digits_to_zero_pad, false, out, exp_out)) { + *exp_out = static_cast<int>(digits_printed - 1); + if (digits_to_zero_pad < digits_printed - 1) { + RemoveExtraPrecision(digits_printed - 1 - digits_to_zero_pad, false, + out, exp_out); return true; } + digits_to_zero_pad -= digits_printed - 1; } for (; digits_to_zero_pad-- > 0;) out->push_back('0'); return true; @@ -1203,10 +1229,10 @@ bool FloatToBufferImpl(Int int_mantissa, int exp, int precision, Buffer *out, const Int mask = (Int{1} << exp) - 1; // Print the integral part first. - int digits_printed = PrintIntegralDigits<mode>(int_mantissa >> exp, out); + size_t digits_printed = PrintIntegralDigits<mode>(int_mantissa >> exp, out); int_mantissa &= mask; - int fractional_count = precision; + size_t fractional_count = precision; if (mode == FormatStyle::Precision) { if (digits_printed == 0) { // Find the first non-zero digit, when in Precision mode. @@ -1222,20 +1248,21 @@ bool FloatToBufferImpl(Int int_mantissa, int exp, int precision, Buffer *out, int_mantissa &= mask; } else { // We already have a digit, and a '.' - *exp_out = digits_printed - 1; - fractional_count -= *exp_out; - if (RemoveExtraPrecision(-fractional_count, int_mantissa != 0, out, - exp_out)) { + *exp_out = static_cast<int>(digits_printed - 1); + if (fractional_count < digits_printed - 1) { // If we had enough digits, return right away. // The code below will try to round again otherwise. + RemoveExtraPrecision(digits_printed - 1 - fractional_count, + int_mantissa != 0, out, exp_out); return true; } + fractional_count -= digits_printed - 1; } } auto get_next_digit = [&] { int_mantissa *= 10; - int digit = static_cast<int>(int_mantissa >> exp); + char digit = static_cast<char>(int_mantissa >> exp); int_mantissa &= mask; return digit; }; @@ -1245,7 +1272,7 @@ bool FloatToBufferImpl(Int int_mantissa, int exp, int precision, Buffer *out, out->push_back(get_next_digit() + '0'); } - int next_digit = get_next_digit(); + char next_digit = get_next_digit(); if (next_digit > 5 || (next_digit == 5 && (int_mantissa || out->last_digit() % 2 == 1))) { RoundUp<mode>(out, exp_out); @@ -1255,24 +1282,25 @@ bool FloatToBufferImpl(Int int_mantissa, int exp, int precision, Buffer *out, } template <FormatStyle mode, typename Float> -bool FloatToBuffer(Decomposed<Float> decomposed, int precision, Buffer *out, - int *exp) { +bool FloatToBuffer(Decomposed<Float> decomposed, + size_t precision, + Buffer* out, + int* exp) { if (precision > kMaxFixedPrecision) return false; // Try with uint64_t. if (CanFitMantissa<Float, std::uint64_t>() && FloatToBufferImpl<std::uint64_t, Float, mode>( - static_cast<std::uint64_t>(decomposed.mantissa), - static_cast<std::uint64_t>(decomposed.exponent), precision, out, exp)) + static_cast<std::uint64_t>(decomposed.mantissa), decomposed.exponent, + precision, out, exp)) return true; #if defined(ABSL_HAVE_INTRINSIC_INT128) // If that is not enough, try with __uint128_t. return CanFitMantissa<Float, __uint128_t>() && FloatToBufferImpl<__uint128_t, Float, mode>( - static_cast<__uint128_t>(decomposed.mantissa), - static_cast<__uint128_t>(decomposed.exponent), precision, out, - exp); + static_cast<__uint128_t>(decomposed.mantissa), decomposed.exponent, + precision, out, exp); #endif return false; } @@ -1280,12 +1308,15 @@ bool FloatToBuffer(Decomposed<Float> decomposed, int precision, Buffer *out, void WriteBufferToSink(char sign_char, absl::string_view str, const FormatConversionSpecImpl &conv, FormatSinkImpl *sink) { - int left_spaces = 0, zeros = 0, right_spaces = 0; - int missing_chars = - conv.width() >= 0 ? std::max(conv.width() - static_cast<int>(str.size()) - - static_cast<int>(sign_char != 0), - 0) - : 0; + size_t left_spaces = 0, zeros = 0, right_spaces = 0; + size_t missing_chars = 0; + if (conv.width() >= 0) { + const size_t conv_width_size_t = static_cast<size_t>(conv.width()); + const size_t existing_chars = + str.size() + static_cast<size_t>(sign_char != 0); + if (conv_width_size_t > existing_chars) + missing_chars = conv_width_size_t - existing_chars; + } if (conv.has_left_flag()) { right_spaces = missing_chars; } else if (conv.has_zero_flag()) { @@ -1321,7 +1352,8 @@ bool FloatToSink(const Float v, const FormatConversionSpecImpl &conv, return true; } - int precision = conv.precision() < 0 ? 6 : conv.precision(); + size_t precision = + conv.precision() < 0 ? 6 : static_cast<size_t>(conv.precision()); int exp = 0; @@ -1348,12 +1380,12 @@ bool FloatToSink(const Float v, const FormatConversionSpecImpl &conv, &buffer); } else if (c == FormatConversionCharInternal::g || c == FormatConversionCharInternal::G) { - precision = std::max(0, precision - 1); + precision = std::max(precision, size_t{1}) - 1; if (!FloatToBuffer<FormatStyle::Precision>(decomposed, precision, &buffer, &exp)) { return FallbackToSnprintf(v, conv, sink); } - if (precision + 1 > exp && exp >= -4) { + if ((exp < 0 || precision + 1 > static_cast<size_t>(exp)) && exp >= -4) { if (exp < 0) { // Have 1.23456, needs 0.00123456 // Move the first digit @@ -1388,9 +1420,11 @@ bool FloatToSink(const Float v, const FormatConversionSpecImpl &conv, return false; } - WriteBufferToSink(sign_char, - absl::string_view(buffer.begin, buffer.end - buffer.begin), - conv, sink); + WriteBufferToSink( + sign_char, + absl::string_view(buffer.begin, + static_cast<size_t>(buffer.end - buffer.begin)), + conv, sink); return true; } diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/parser.cc b/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/parser.cc index 2c9c07dacc..5aaab6983f 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/parser.cc +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/parser.cc @@ -31,207 +31,14 @@ namespace absl { ABSL_NAMESPACE_BEGIN namespace str_format_internal { -using CC = FormatConversionCharInternal; -using LM = LengthMod; +// Define the array for non-constexpr uses. +constexpr ConvTag ConvTagHolder::value[256]; -// Abbreviations to fit in the table below. -constexpr auto f_sign = Flags::kSignCol; -constexpr auto f_alt = Flags::kAlt; -constexpr auto f_pos = Flags::kShowPos; -constexpr auto f_left = Flags::kLeft; -constexpr auto f_zero = Flags::kZero; - -ABSL_CONST_INIT const ConvTag kTags[256] = { - {}, {}, {}, {}, {}, {}, {}, {}, // 00-07 - {}, {}, {}, {}, {}, {}, {}, {}, // 08-0f - {}, {}, {}, {}, {}, {}, {}, {}, // 10-17 - {}, {}, {}, {}, {}, {}, {}, {}, // 18-1f - f_sign, {}, {}, f_alt, {}, {}, {}, {}, // !"#$%&' - {}, {}, {}, f_pos, {}, f_left, {}, {}, // ()*+,-./ - f_zero, {}, {}, {}, {}, {}, {}, {}, // 01234567 - {}, {}, {}, {}, {}, {}, {}, {}, // 89:;<=>? - {}, CC::A, {}, {}, {}, CC::E, CC::F, CC::G, // @ABCDEFG - {}, {}, {}, {}, LM::L, {}, {}, {}, // HIJKLMNO - {}, {}, {}, {}, {}, {}, {}, {}, // PQRSTUVW - CC::X, {}, {}, {}, {}, {}, {}, {}, // XYZ[\]^_ - {}, CC::a, {}, CC::c, CC::d, CC::e, CC::f, CC::g, // `abcdefg - LM::h, CC::i, LM::j, {}, LM::l, {}, CC::n, CC::o, // hijklmno - CC::p, LM::q, {}, CC::s, LM::t, CC::u, {}, {}, // pqrstuvw - CC::x, {}, LM::z, {}, {}, {}, {}, {}, // xyz{|}! - {}, {}, {}, {}, {}, {}, {}, {}, // 80-87 - {}, {}, {}, {}, {}, {}, {}, {}, // 88-8f - {}, {}, {}, {}, {}, {}, {}, {}, // 90-97 - {}, {}, {}, {}, {}, {}, {}, {}, // 98-9f - {}, {}, {}, {}, {}, {}, {}, {}, // a0-a7 - {}, {}, {}, {}, {}, {}, {}, {}, // a8-af - {}, {}, {}, {}, {}, {}, {}, {}, // b0-b7 - {}, {}, {}, {}, {}, {}, {}, {}, // b8-bf - {}, {}, {}, {}, {}, {}, {}, {}, // c0-c7 - {}, {}, {}, {}, {}, {}, {}, {}, // c8-cf - {}, {}, {}, {}, {}, {}, {}, {}, // d0-d7 - {}, {}, {}, {}, {}, {}, {}, {}, // d8-df - {}, {}, {}, {}, {}, {}, {}, {}, // e0-e7 - {}, {}, {}, {}, {}, {}, {}, {}, // e8-ef - {}, {}, {}, {}, {}, {}, {}, {}, // f0-f7 - {}, {}, {}, {}, {}, {}, {}, {}, // f8-ff -}; - -namespace { - -bool CheckFastPathSetting(const UnboundConversion& conv) { - bool width_precision_needed = - conv.width.value() >= 0 || conv.precision.value() >= 0; - if (width_precision_needed && conv.flags == Flags::kBasic) { - fprintf(stderr, - "basic=%d left=%d show_pos=%d sign_col=%d alt=%d zero=%d " - "width=%d precision=%d\n", - conv.flags == Flags::kBasic ? 1 : 0, - FlagsContains(conv.flags, Flags::kLeft) ? 1 : 0, - FlagsContains(conv.flags, Flags::kShowPos) ? 1 : 0, - FlagsContains(conv.flags, Flags::kSignCol) ? 1 : 0, - FlagsContains(conv.flags, Flags::kAlt) ? 1 : 0, - FlagsContains(conv.flags, Flags::kZero) ? 1 : 0, conv.width.value(), - conv.precision.value()); - return false; - } - return true; -} - -template <bool is_positional> -const char *ConsumeConversion(const char *pos, const char *const end, - UnboundConversion *conv, int *next_arg) { - const char* const original_pos = pos; - char c; - // Read the next char into `c` and update `pos`. Returns false if there are - // no more chars to read. -#define ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR() \ - do { \ - if (ABSL_PREDICT_FALSE(pos == end)) return nullptr; \ - c = *pos++; \ - } while (0) - - const auto parse_digits = [&] { - int digits = c - '0'; - // We do not want to overflow `digits` so we consume at most digits10 - // digits. If there are more digits the parsing will fail later on when the - // digit doesn't match the expected characters. - int num_digits = std::numeric_limits<int>::digits10; - for (;;) { - if (ABSL_PREDICT_FALSE(pos == end)) break; - c = *pos++; - if (!std::isdigit(c)) break; - --num_digits; - if (ABSL_PREDICT_FALSE(!num_digits)) break; - digits = 10 * digits + c - '0'; - } - return digits; - }; - - if (is_positional) { - ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR(); - if (ABSL_PREDICT_FALSE(c < '1' || c > '9')) return nullptr; - conv->arg_position = parse_digits(); - assert(conv->arg_position > 0); - if (ABSL_PREDICT_FALSE(c != '$')) return nullptr; - } - - ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR(); - - // We should start with the basic flag on. - assert(conv->flags == Flags::kBasic); - - // Any non alpha character makes this conversion not basic. - // This includes flags (-+ #0), width (1-9, *) or precision (.). - // All conversion characters and length modifiers are alpha characters. - if (c < 'A') { - while (c <= '0') { - auto tag = GetTagForChar(c); - if (tag.is_flags()) { - conv->flags = conv->flags | tag.as_flags(); - ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR(); - } else { - break; - } - } - - if (c <= '9') { - if (c >= '0') { - int maybe_width = parse_digits(); - if (!is_positional && c == '$') { - if (ABSL_PREDICT_FALSE(*next_arg != 0)) return nullptr; - // Positional conversion. - *next_arg = -1; - return ConsumeConversion<true>(original_pos, end, conv, next_arg); - } - conv->flags = conv->flags | Flags::kNonBasic; - conv->width.set_value(maybe_width); - } else if (c == '*') { - conv->flags = conv->flags | Flags::kNonBasic; - ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR(); - if (is_positional) { - if (ABSL_PREDICT_FALSE(c < '1' || c > '9')) return nullptr; - conv->width.set_from_arg(parse_digits()); - if (ABSL_PREDICT_FALSE(c != '$')) return nullptr; - ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR(); - } else { - conv->width.set_from_arg(++*next_arg); - } - } - } - - if (c == '.') { - conv->flags = conv->flags | Flags::kNonBasic; - ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR(); - if (std::isdigit(c)) { - conv->precision.set_value(parse_digits()); - } else if (c == '*') { - ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR(); - if (is_positional) { - if (ABSL_PREDICT_FALSE(c < '1' || c > '9')) return nullptr; - conv->precision.set_from_arg(parse_digits()); - if (c != '$') return nullptr; - ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR(); - } else { - conv->precision.set_from_arg(++*next_arg); - } - } else { - conv->precision.set_value(0); - } - } - } - - auto tag = GetTagForChar(c); - - if (ABSL_PREDICT_FALSE(!tag.is_conv())) { - if (ABSL_PREDICT_FALSE(!tag.is_length())) return nullptr; - - // It is a length modifier. - using str_format_internal::LengthMod; - LengthMod length_mod = tag.as_length(); - ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR(); - if (c == 'h' && length_mod == LengthMod::h) { - conv->length_mod = LengthMod::hh; - ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR(); - } else if (c == 'l' && length_mod == LengthMod::l) { - conv->length_mod = LengthMod::ll; - ABSL_FORMAT_PARSER_INTERNAL_GET_CHAR(); - } else { - conv->length_mod = length_mod; - } - tag = GetTagForChar(c); - if (ABSL_PREDICT_FALSE(!tag.is_conv())) return nullptr; - } - - assert(CheckFastPathSetting(*conv)); - (void)(&CheckFastPathSetting); - - conv->conv = tag.as_conv(); - if (!is_positional) conv->arg_position = ++*next_arg; - return pos; +ABSL_ATTRIBUTE_NOINLINE const char* ConsumeUnboundConversionNoInline( + const char* p, const char* end, UnboundConversion* conv, int* next_arg) { + return ConsumeUnboundConversion(p, end, conv, next_arg); } -} // namespace - std::string LengthModToString(LengthMod v) { switch (v) { case LengthMod::h: @@ -258,12 +65,6 @@ std::string LengthModToString(LengthMod v) { return ""; } -const char *ConsumeUnboundConversion(const char *p, const char *end, - UnboundConversion *conv, int *next_arg) { - if (*next_arg < 0) return ConsumeConversion<true>(p, end, conv, next_arg); - return ConsumeConversion<false>(p, end, conv, next_arg); -} - struct ParsedFormatBase::ParsedFormatConsumer { explicit ParsedFormatConsumer(ParsedFormatBase *parsedformat) : parsed(parsedformat), data_pos(parsedformat->data_.get()) {} @@ -312,11 +113,11 @@ bool ParsedFormatBase::MatchesConversions( std::initializer_list<FormatConversionCharSet> convs) const { std::unordered_set<int> used; auto add_if_valid_conv = [&](int pos, char c) { - if (static_cast<size_t>(pos) > convs.size() || - !Contains(convs.begin()[pos - 1], c)) - return false; - used.insert(pos); - return true; + if (static_cast<size_t>(pos) > convs.size() || + !Contains(convs.begin()[pos - 1], c)) + return false; + used.insert(pos); + return true; }; for (const ConversionItem &item : items_) { if (!item.is_conversion) continue; diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/parser.h b/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/parser.h index 32b91d034d..35b6d49c14 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/parser.h +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/str_format/parser.h @@ -29,111 +29,18 @@ #include <vector> #include "absl/strings/internal/str_format/checker.h" +#include "absl/strings/internal/str_format/constexpr_parser.h" #include "absl/strings/internal/str_format/extension.h" namespace absl { ABSL_NAMESPACE_BEGIN namespace str_format_internal { -enum class LengthMod : std::uint8_t { h, hh, l, ll, L, j, z, t, q, none }; - std::string LengthModToString(LengthMod v); -// The analyzed properties of a single specified conversion. -struct UnboundConversion { - UnboundConversion() {} - - class InputValue { - public: - void set_value(int value) { - assert(value >= 0); - value_ = value; - } - int value() const { return value_; } - - // Marks the value as "from arg". aka the '*' format. - // Requires `value >= 1`. - // When set, is_from_arg() return true and get_from_arg() returns the - // original value. - // `value()`'s return value is unspecfied in this state. - void set_from_arg(int value) { - assert(value > 0); - value_ = -value - 1; - } - bool is_from_arg() const { return value_ < -1; } - int get_from_arg() const { - assert(is_from_arg()); - return -value_ - 1; - } - - private: - int value_ = -1; - }; - - // No need to initialize. It will always be set in the parser. - int arg_position; - - InputValue width; - InputValue precision; - - Flags flags = Flags::kBasic; - LengthMod length_mod = LengthMod::none; - FormatConversionChar conv = FormatConversionCharInternal::kNone; -}; - -// Consume conversion spec prefix (not including '%') of [p, end) if valid. -// Examples of valid specs would be e.g.: "s", "d", "-12.6f". -// If valid, it returns the first character following the conversion spec, -// and the spec part is broken down and returned in 'conv'. -// If invalid, returns nullptr. -const char* ConsumeUnboundConversion(const char* p, const char* end, - UnboundConversion* conv, int* next_arg); - -// Helper tag class for the table below. -// It allows fast `char -> ConversionChar/LengthMod/Flags` checking and -// conversions. -class ConvTag { - public: - constexpr ConvTag(FormatConversionChar conversion_char) // NOLINT - : tag_(static_cast<uint8_t>(conversion_char)) {} - constexpr ConvTag(LengthMod length_mod) // NOLINT - : tag_(0x80 | static_cast<uint8_t>(length_mod)) {} - constexpr ConvTag(Flags flags) // NOLINT - : tag_(0xc0 | static_cast<uint8_t>(flags)) {} - constexpr ConvTag() : tag_(0xFF) {} - - bool is_conv() const { return (tag_ & 0x80) == 0; } - bool is_length() const { return (tag_ & 0xC0) == 0x80; } - bool is_flags() const { return (tag_ & 0xE0) == 0xC0; } - - FormatConversionChar as_conv() const { - assert(is_conv()); - assert(!is_length()); - assert(!is_flags()); - return static_cast<FormatConversionChar>(tag_); - } - LengthMod as_length() const { - assert(!is_conv()); - assert(is_length()); - assert(!is_flags()); - return static_cast<LengthMod>(tag_ & 0x3F); - } - Flags as_flags() const { - assert(!is_conv()); - assert(!is_length()); - assert(is_flags()); - return static_cast<Flags>(tag_ & 0x1F); - } - - private: - uint8_t tag_; -}; - -extern const ConvTag kTags[256]; -// Keep a single table for all the conversion chars and length modifiers. -inline ConvTag GetTagForChar(char c) { - return kTags[static_cast<unsigned char>(c)]; -} +const char* ConsumeUnboundConversionNoInline(const char* p, const char* end, + UnboundConversion* conv, + int* next_arg); // Parse the format string provided in 'src' and pass the identified items into // 'consumer'. @@ -155,10 +62,11 @@ bool ParseFormatString(string_view src, Consumer consumer) { static_cast<const char*>(memchr(p, '%', static_cast<size_t>(end - p))); if (!percent) { // We found the last substring. - return consumer.Append(string_view(p, end - p)); + return consumer.Append(string_view(p, static_cast<size_t>(end - p))); } // We found a percent, so push the text run then process the percent. - if (ABSL_PREDICT_FALSE(!consumer.Append(string_view(p, percent - p)))) { + if (ABSL_PREDICT_FALSE(!consumer.Append( + string_view(p, static_cast<size_t>(percent - p))))) { return false; } if (ABSL_PREDICT_FALSE(percent + 1 >= end)) return false; @@ -186,10 +94,11 @@ bool ParseFormatString(string_view src, Consumer consumer) { } } else if (percent[1] != '%') { UnboundConversion conv; - p = ConsumeUnboundConversion(percent + 1, end, &conv, &next_arg); + p = ConsumeUnboundConversionNoInline(percent + 1, end, &conv, &next_arg); if (ABSL_PREDICT_FALSE(p == nullptr)) return false; if (ABSL_PREDICT_FALSE(!consumer.ConvertOne( - conv, string_view(percent + 1, p - (percent + 1))))) { + conv, string_view(percent + 1, + static_cast<size_t>(p - (percent + 1)))))) { return false; } } else { diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/str_split_internal.h b/contrib/restricted/abseil-cpp/absl/strings/internal/str_split_internal.h index e766421617..35edf3aa43 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/internal/str_split_internal.h +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/str_split_internal.h @@ -132,7 +132,8 @@ class SplitIterator { const absl::string_view text = splitter_->text(); const absl::string_view d = delimiter_.Find(text, pos_); if (d.data() == text.data() + text.size()) state_ = kLastState; - curr_ = text.substr(pos_, d.data() - (text.data() + pos_)); + curr_ = text.substr(pos_, + static_cast<size_t>(d.data() - (text.data() + pos_))); pos_ += curr_.size() + d.size(); } while (!predicate_(curr_)); return *this; diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/stringify_sink.cc b/contrib/restricted/abseil-cpp/absl/strings/internal/stringify_sink.cc new file mode 100644 index 0000000000..7c6995abb1 --- /dev/null +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/stringify_sink.cc @@ -0,0 +1,28 @@ +// Copyright 2022 The Abseil Authors +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// https://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#include "absl/strings/internal/stringify_sink.h" +namespace absl { +ABSL_NAMESPACE_BEGIN +namespace strings_internal { + +void StringifySink::Append(size_t count, char ch) { buffer_.append(count, ch); } + +void StringifySink::Append(string_view v) { + buffer_.append(v.data(), v.size()); +} + +} // namespace strings_internal +ABSL_NAMESPACE_END +} // namespace absl diff --git a/contrib/restricted/abseil-cpp/absl/strings/internal/stringify_sink.h b/contrib/restricted/abseil-cpp/absl/strings/internal/stringify_sink.h new file mode 100644 index 0000000000..fc3747bb72 --- /dev/null +++ b/contrib/restricted/abseil-cpp/absl/strings/internal/stringify_sink.h @@ -0,0 +1,57 @@ +// Copyright 2022 The Abseil Authors +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// https://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#ifndef ABSL_STRINGS_INTERNAL_STRINGIFY_SINK_H_ +#define ABSL_STRINGS_INTERNAL_STRINGIFY_SINK_H_ + +#include <string> +#include <type_traits> +#include <utility> + +#include "absl/strings/string_view.h" + +namespace absl { +ABSL_NAMESPACE_BEGIN + +namespace strings_internal { +class StringifySink { + public: + void Append(size_t count, char ch); + + void Append(string_view v); + + // Support `absl::Format(&sink, format, args...)`. + friend void AbslFormatFlush(StringifySink* sink, absl::string_view v) { + sink->Append(v); + } + + private: + template <typename T> + friend string_view ExtractStringification(StringifySink& sink, const T& v); + + std::string buffer_; +}; + +template <typename T> +string_view ExtractStringification(StringifySink& sink, const T& v) { + AbslStringify(sink, v); + return sink.buffer_; +} + +} // namespace strings_internal + +ABSL_NAMESPACE_END +} // namespace absl + +#endif // ABSL_STRINGS_INTERNAL_STRINGIFY_SINK_H_ diff --git a/contrib/restricted/abseil-cpp/absl/strings/numbers.cc b/contrib/restricted/abseil-cpp/absl/strings/numbers.cc index e798fc69f7..2987158e07 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/numbers.cc +++ b/contrib/restricted/abseil-cpp/absl/strings/numbers.cc @@ -190,32 +190,32 @@ char* numbers_internal::FastIntToBuffer(uint32_t i, char* buffer) { if (i >= 1000) goto lt10_000; digits = i / 100; i -= digits * 100; - *buffer++ = '0' + digits; + *buffer++ = '0' + static_cast<char>(digits); goto lt100; } if (i < 1000000) { // 1,000,000 if (i >= 100000) goto lt1_000_000; digits = i / 10000; // 10,000 i -= digits * 10000; - *buffer++ = '0' + digits; + *buffer++ = '0' + static_cast<char>(digits); goto lt10_000; } if (i < 100000000) { // 100,000,000 if (i >= 10000000) goto lt100_000_000; digits = i / 1000000; // 1,000,000 i -= digits * 1000000; - *buffer++ = '0' + digits; + *buffer++ = '0' + static_cast<char>(digits); goto lt1_000_000; } // we already know that i < 1,000,000,000 digits = i / 100000000; // 100,000,000 i -= digits * 100000000; - *buffer++ = '0' + digits; + *buffer++ = '0' + static_cast<char>(digits); goto lt100_000_000; } char* numbers_internal::FastIntToBuffer(int32_t i, char* buffer) { - uint32_t u = i; + uint32_t u = static_cast<uint32_t>(i); if (i < 0) { *buffer++ = '-'; // We need to do the negation in modular (i.e., "unsigned") @@ -268,7 +268,7 @@ char* numbers_internal::FastIntToBuffer(uint64_t i, char* buffer) { } char* numbers_internal::FastIntToBuffer(int64_t i, char* buffer) { - uint64_t u = i; + uint64_t u = static_cast<uint64_t>(i); if (i < 0) { *buffer++ = '-'; u = 0 - u; @@ -329,7 +329,7 @@ static std::pair<uint64_t, uint64_t> PowFive(uint64_t num, int expfive) { result = Mul32(result, 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5); expfive -= 13; } - constexpr int powers_of_five[13] = { + constexpr uint32_t powers_of_five[13] = { 1, 5, 5 * 5, @@ -404,14 +404,14 @@ static ExpDigits SplitToSix(const double value) { // we multiply it by 65536 and see if the fractional part is close to 32768. // (The number doesn't have to be a power of two,but powers of two are faster) uint64_t d64k = d * 65536; - int dddddd; // A 6-digit decimal integer. + uint32_t dddddd; // A 6-digit decimal integer. if ((d64k % 65536) == 32767 || (d64k % 65536) == 32768) { // OK, it's fairly likely that precision was lost above, which is // not a surprise given only 52 mantissa bits are available. Therefore // redo the calculation using 128-bit numbers. (64 bits are not enough). // Start out with digits rounded down; maybe add one below. - dddddd = static_cast<int>(d64k / 65536); + dddddd = static_cast<uint32_t>(d64k / 65536); // mantissa is a 64-bit integer representing M.mmm... * 2^63. The actual // value we're representing, of course, is M.mmm... * 2^exp2. @@ -461,7 +461,7 @@ static ExpDigits SplitToSix(const double value) { } } else { // Here, we are not close to the edge. - dddddd = static_cast<int>((d64k + 32768) / 65536); + dddddd = static_cast<uint32_t>((d64k + 32768) / 65536); } if (dddddd == 1000000) { dddddd = 100000; @@ -469,7 +469,7 @@ static ExpDigits SplitToSix(const double value) { } exp_dig.exponent = exp; - int two_digits = dddddd / 10000; + uint32_t two_digits = dddddd / 10000; dddddd -= two_digits * 10000; numbers_internal::PutTwoDigits(two_digits, &exp_dig.digits[0]); @@ -499,7 +499,7 @@ size_t numbers_internal::SixDigitsToBuffer(double d, char* const buffer) { if (std::signbit(d)) *out++ = '-'; *out++ = '0'; *out = 0; - return out - buffer; + return static_cast<size_t>(out - buffer); } if (d < 0) { *out++ = '-'; @@ -507,7 +507,7 @@ size_t numbers_internal::SixDigitsToBuffer(double d, char* const buffer) { } if (d > std::numeric_limits<double>::max()) { strcpy(out, "inf"); // NOLINT(runtime/printf) - return out + 3 - buffer; + return static_cast<size_t>(out + 3 - buffer); } auto exp_dig = SplitToSix(d); @@ -519,7 +519,7 @@ size_t numbers_internal::SixDigitsToBuffer(double d, char* const buffer) { case 5: memcpy(out, &digits[0], 6), out += 6; *out = 0; - return out - buffer; + return static_cast<size_t>(out - buffer); case 4: memcpy(out, &digits[0], 5), out += 5; if (digits[5] != '0') { @@ -527,7 +527,7 @@ size_t numbers_internal::SixDigitsToBuffer(double d, char* const buffer) { *out++ = digits[5]; } *out = 0; - return out - buffer; + return static_cast<size_t>(out - buffer); case 3: memcpy(out, &digits[0], 4), out += 4; if ((digits[5] | digits[4]) != '0') { @@ -536,7 +536,7 @@ size_t numbers_internal::SixDigitsToBuffer(double d, char* const buffer) { if (digits[5] != '0') *out++ = digits[5]; } *out = 0; - return out - buffer; + return static_cast<size_t>(out - buffer); case 2: memcpy(out, &digits[0], 3), out += 3; *out++ = '.'; @@ -545,7 +545,7 @@ size_t numbers_internal::SixDigitsToBuffer(double d, char* const buffer) { while (out[-1] == '0') --out; if (out[-1] == '.') --out; *out = 0; - return out - buffer; + return static_cast<size_t>(out - buffer); case 1: memcpy(out, &digits[0], 2), out += 2; *out++ = '.'; @@ -554,7 +554,7 @@ size_t numbers_internal::SixDigitsToBuffer(double d, char* const buffer) { while (out[-1] == '0') --out; if (out[-1] == '.') --out; *out = 0; - return out - buffer; + return static_cast<size_t>(out - buffer); case 0: memcpy(out, &digits[0], 1), out += 1; *out++ = '.'; @@ -563,7 +563,7 @@ size_t numbers_internal::SixDigitsToBuffer(double d, char* const buffer) { while (out[-1] == '0') --out; if (out[-1] == '.') --out; *out = 0; - return out - buffer; + return static_cast<size_t>(out - buffer); case -4: out[2] = '0'; ++out; @@ -582,7 +582,7 @@ size_t numbers_internal::SixDigitsToBuffer(double d, char* const buffer) { out += 6; while (out[-1] == '0') --out; *out = 0; - return out - buffer; + return static_cast<size_t>(out - buffer); } assert(exp < -4 || exp >= 6); out[0] = digits[0]; @@ -601,12 +601,12 @@ size_t numbers_internal::SixDigitsToBuffer(double d, char* const buffer) { if (exp > 99) { int dig1 = exp / 100; exp -= dig1 * 100; - *out++ = '0' + dig1; + *out++ = '0' + static_cast<char>(dig1); } - PutTwoDigits(exp, out); + PutTwoDigits(static_cast<uint32_t>(exp), out); out += 2; *out = 0; - return out - buffer; + return static_cast<size_t>(out - buffer); } namespace { @@ -642,10 +642,12 @@ inline bool safe_parse_sign_and_base(absl::string_view* text /*inout*/, int base = *base_ptr; // Consume whitespace. - while (start < end && absl::ascii_isspace(start[0])) { + while (start < end && + absl::ascii_isspace(static_cast<unsigned char>(start[0]))) { ++start; } - while (start < end && absl::ascii_isspace(end[-1])) { + while (start < end && + absl::ascii_isspace(static_cast<unsigned char>(end[-1]))) { --end; } if (start >= end) { @@ -694,7 +696,7 @@ inline bool safe_parse_sign_and_base(absl::string_view* text /*inout*/, } else { return false; } - *text = absl::string_view(start, end - start); + *text = absl::string_view(start, static_cast<size_t>(end - start)); *base_ptr = base; return true; } @@ -920,17 +922,18 @@ inline bool safe_parse_positive_int(absl::string_view text, int base, const IntType vmax = std::numeric_limits<IntType>::max(); assert(vmax > 0); assert(base >= 0); - assert(vmax >= static_cast<IntType>(base)); + const IntType base_inttype = static_cast<IntType>(base); + assert(vmax >= base_inttype); const IntType vmax_over_base = LookupTables<IntType>::kVmaxOverBase[base]; assert(base < 2 || - std::numeric_limits<IntType>::max() / base == vmax_over_base); + std::numeric_limits<IntType>::max() / base_inttype == vmax_over_base); const char* start = text.data(); const char* end = start + text.size(); // loop over digits for (; start < end; ++start) { unsigned char c = static_cast<unsigned char>(start[0]); - int digit = kAsciiToInt[c]; - if (digit >= base) { + IntType digit = static_cast<IntType>(kAsciiToInt[c]); + if (digit >= base_inttype) { *value_p = value; return false; } @@ -938,7 +941,7 @@ inline bool safe_parse_positive_int(absl::string_view text, int base, *value_p = vmax; return false; } - value *= base; + value *= base_inttype; if (value > vmax - digit) { *value_p = vmax; return false; diff --git a/contrib/restricted/abseil-cpp/absl/strings/str_cat.cc b/contrib/restricted/abseil-cpp/absl/strings/str_cat.cc index f4a77493a4..114a2ff284 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/str_cat.cc +++ b/contrib/restricted/abseil-cpp/absl/strings/str_cat.cc @@ -17,12 +17,15 @@ #include <assert.h> #include <algorithm> +#include <cstddef> #include <cstdint> #include <cstring> +#include <string> #include "absl/strings/ascii.h" #include "absl/strings/internal/resize_uninitialized.h" #include "absl/strings/numbers.h" +#include "absl/strings/string_view.h" namespace absl { ABSL_NAMESPACE_BEGIN @@ -56,7 +59,7 @@ AlphaNum::AlphaNum(Dec dec) { *--writer = '0' + (value % 10); value /= 10; } - *--writer = '0' + value; + *--writer = '0' + static_cast<char>(value); if (neg) *--writer = '-'; ptrdiff_t fillers = writer - minfill; @@ -73,7 +76,7 @@ AlphaNum::AlphaNum(Dec dec) { if (add_sign_again) *--writer = '-'; } - piece_ = absl::string_view(writer, end - writer); + piece_ = absl::string_view(writer, static_cast<size_t>(end - writer)); } // ---------------------------------------------------------------------- @@ -141,12 +144,12 @@ namespace strings_internal { std::string CatPieces(std::initializer_list<absl::string_view> pieces) { std::string result; size_t total_size = 0; - for (const absl::string_view& piece : pieces) total_size += piece.size(); + for (absl::string_view piece : pieces) total_size += piece.size(); strings_internal::STLStringResizeUninitialized(&result, total_size); char* const begin = &result[0]; char* out = begin; - for (const absl::string_view& piece : pieces) { + for (absl::string_view piece : pieces) { const size_t this_size = piece.size(); if (this_size != 0) { memcpy(out, piece.data(), this_size); @@ -170,7 +173,7 @@ void AppendPieces(std::string* dest, std::initializer_list<absl::string_view> pieces) { size_t old_size = dest->size(); size_t total_size = old_size; - for (const absl::string_view& piece : pieces) { + for (absl::string_view piece : pieces) { ASSERT_NO_OVERLAP(*dest, piece); total_size += piece.size(); } @@ -178,7 +181,7 @@ void AppendPieces(std::string* dest, char* const begin = &(*dest)[0]; char* out = begin + old_size; - for (const absl::string_view& piece : pieces) { + for (absl::string_view piece : pieces) { const size_t this_size = piece.size(); if (this_size != 0) { memcpy(out, piece.data(), this_size); diff --git a/contrib/restricted/abseil-cpp/absl/strings/str_cat.h b/contrib/restricted/abseil-cpp/absl/strings/str_cat.h index a94bc5df69..730b4d8c60 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/str_cat.h +++ b/contrib/restricted/abseil-cpp/absl/strings/str_cat.h @@ -48,6 +48,40 @@ // `StrCat()` or `StrAppend()`. You may specify a minimum hex field width using // a `PadSpec` enum. // +// User-defined types can be formatted with the `AbslStringify()` customization +// point. The API relies on detecting an overload in the user-defined type's +// namespace of a free (non-member) `AbslStringify()` function as a definition +// (typically declared as a friend and implemented in-line. +// with the following signature: +// +// class MyClass { ... }; +// +// template <typename Sink> +// void AbslStringify(Sink& sink, const MyClass& value); +// +// An `AbslStringify()` overload for a type should only be declared in the same +// file and namespace as said type. +// +// Note that `AbslStringify()` also supports use with `absl::StrFormat()` and +// `absl::Substitute()`. +// +// Example: +// +// struct Point { +// // To add formatting support to `Point`, we simply need to add a free +// // (non-member) function `AbslStringify()`. This method specifies how +// // Point should be printed when absl::StrCat() is called on it. You can add +// // such a free function using a friend declaration within the body of the +// // class. The sink parameter is a templated type to avoid requiring +// // dependencies. +// template <typename Sink> friend void AbslStringify(Sink& +// sink, const Point& p) { +// absl::Format(&sink, "(%v, %v)", p.x, p.y); +// } +// +// int x; +// int y; +// }; // ----------------------------------------------------------------------------- #ifndef ABSL_STRINGS_STR_CAT_H_ @@ -57,9 +91,12 @@ #include <cstdint> #include <string> #include <type_traits> +#include <utility> #include <vector> #include "absl/base/port.h" +#include "absl/strings/internal/has_absl_stringify.h" +#include "absl/strings/internal/stringify_sink.h" #include "absl/strings/numbers.h" #include "absl/strings/string_view.h" @@ -205,8 +242,10 @@ struct Dec { // ----------------------------------------------------------------------------- // // The `AlphaNum` class acts as the main parameter type for `StrCat()` and -// `StrAppend()`, providing efficient conversion of numeric, boolean, and -// hexadecimal values (through the `Hex` type) into strings. +// `StrAppend()`, providing efficient conversion of numeric, boolean, decimal, +// and hexadecimal values (through the `Dec` and `Hex` types) into strings. +// `AlphaNum` should only be used as a function parameter. Do not instantiate +// `AlphaNum` directly as a stack variable. class AlphaNum { public: @@ -255,6 +294,13 @@ class AlphaNum { : piece_(NullSafeStringView(c_str)) {} // NOLINT(runtime/explicit) AlphaNum(absl::string_view pc) : piece_(pc) {} // NOLINT(runtime/explicit) + template <typename T, typename = typename std::enable_if< + strings_internal::HasAbslStringify<T>::value>::type> + AlphaNum( // NOLINT(runtime/explicit) + const T& v, // NOLINT(runtime/explicit) + strings_internal::StringifySink&& sink = {}) // NOLINT(runtime/explicit) + : piece_(strings_internal::ExtractStringification(sink, v)) {} + template <typename Allocator> AlphaNum( // NOLINT(runtime/explicit) const std::basic_string<char, std::char_traits<char>, Allocator>& str) @@ -274,7 +320,8 @@ class AlphaNum { // This overload matches only scoped enums. template <typename T, typename = typename std::enable_if< - std::is_enum<T>{} && !std::is_convertible<T, int>{}>::type> + std::is_enum<T>{} && !std::is_convertible<T, int>{} && + !strings_internal::HasAbslStringify<T>::value>::type> AlphaNum(T e) // NOLINT(runtime/explicit) : AlphaNum(static_cast<typename std::underlying_type<T>::type>(e)) {} diff --git a/contrib/restricted/abseil-cpp/absl/strings/str_format.h b/contrib/restricted/abseil-cpp/absl/strings/str_format.h index 4b05c70c23..3536b70e9d 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/str_format.h +++ b/contrib/restricted/abseil-cpp/absl/strings/str_format.h @@ -191,9 +191,9 @@ class FormatCountCapture { // absl::StrFormat(formatString, "TheVillage", 6); // // A format string generally follows the POSIX syntax as used within the POSIX -// `printf` specification. +// `printf` specification. (Exceptions are noted below.) // -// (See http://pubs.opengroup.org/onlinepubs/9699919799/functions/fprintf.html.) +// (See http://pubs.opengroup.org/onlinepubs/9699919799/functions/fprintf.html) // // In specific, the `FormatSpec` supports the following type specifiers: // * `c` for characters @@ -211,6 +211,10 @@ class FormatCountCapture { // * `n` for the special case of writing out the number of characters // written to this point. The resulting value must be captured within an // `absl::FormatCountCapture` type. +// * `v` for values using the default format for a deduced type. These deduced +// types include many of the primitive types denoted here as well as +// user-defined types containing the proper extensions. (See below for more +// information.) // // Implementation-defined behavior: // * A null pointer provided to "%s" or "%p" is output as "(nil)". @@ -239,6 +243,15 @@ class FormatCountCapture { // "%s%d%n", "hello", 123, absl::FormatCountCapture(&n)); // EXPECT_EQ(8, n); // +// NOTE: the `v` specifier (for "value") is a type specifier not present in the +// POSIX specification. %v will format values according to their deduced type. +// `v` uses `d` for signed integer values, `u` for unsigned integer values, `g` +// for floating point values, and formats boolean values as "true"/"false" +// (instead of 1 or 0 for booleans formatted using d). `const char*` is not +// supported; please use `std:string` and `string_view`. `char` is also not +// supported due to ambiguity of the type. This specifier does not support +// modifiers. +// // The `FormatSpec` intrinsically supports all of these fundamental C++ types: // // * Characters: `char`, `signed char`, `unsigned char` @@ -570,6 +583,41 @@ ABSL_MUST_USE_RESULT inline bool FormatUntyped( // StrFormat Extensions //------------------------------------------------------------------------------ // +// AbslStringify() +// +// A simpler customization API for formatting user-defined types using +// absl::StrFormat(). The API relies on detecting an overload in the +// user-defined type's namespace of a free (non-member) `AbslStringify()` +// function as a friend definition with the following signature: +// +// template <typename Sink> +// void AbslStringify(Sink& sink, const X& value); +// +// An `AbslStringify()` overload for a type should only be declared in the same +// file and namespace as said type. +// +// Note that unlike with AbslFormatConvert(), AbslStringify() does not allow +// customization of allowed conversion characters. AbslStringify() uses `%v` as +// the underlying conversion specififer. Additionally, AbslStringify() supports +// use with absl::StrCat while AbslFormatConvert() does not. +// +// Example: +// +// struct Point { +// // To add formatting support to `Point`, we simply need to add a free +// // (non-member) function `AbslStringify()`. This method prints in the +// // request format using the underlying `%v` specifier. You can add such a +// // free function using a friend declaration within the body of the class. +// // The sink parameter is a templated type to avoid requiring dependencies. +// template <typename Sink> +// friend void AbslStringify(Sink& sink, const Point& p) { +// absl::Format(&sink, "(%v, %v)", p.x, p.y); +// } +// +// int x; +// int y; +// }; +// // AbslFormatConvert() // // The StrFormat library provides a customization API for formatting @@ -616,9 +664,9 @@ ABSL_MUST_USE_RESULT inline bool FormatUntyped( // AbslFormatConvert(const Point& p, const absl::FormatConversionSpec& spec, // absl::FormatSink* s) { // if (spec.conversion_char() == absl::FormatConversionChar::s) { -// s->Append(absl::StrCat("x=", p.x, " y=", p.y)); +// absl::Format(s, "x=%vy=%v", p.x, p.y); // } else { -// s->Append(absl::StrCat(p.x, ",", p.y)); +// absl::Format(s, "%v,%v", p.x, p.y); // } // return {true}; // } @@ -637,7 +685,7 @@ enum class FormatConversionChar : uint8_t { c, s, // text d, i, o, u, x, X, // int f, F, e, E, g, G, a, A, // float - n, p // misc + n, p, v // misc }; // clang-format on @@ -757,6 +805,7 @@ enum class FormatConversionCharSet : uint64_t { // misc n = str_format_internal::FormatConversionCharToConvInt('n'), p = str_format_internal::FormatConversionCharToConvInt('p'), + v = str_format_internal::FormatConversionCharToConvInt('v'), // Used for width/precision '*' specification. kStar = static_cast<uint64_t>( @@ -771,23 +820,36 @@ enum class FormatConversionCharSet : uint64_t { // FormatSink // -// An abstraction to which conversions write their string data. +// A format sink is a generic abstraction to which conversions may write their +// formatted string data. `absl::FormatConvert()` uses this sink to write its +// formatted string. // class FormatSink { public: - // Appends `count` copies of `ch`. + // FormatSink::Append() + // + // Appends `count` copies of `ch` to the format sink. void Append(size_t count, char ch) { sink_->Append(count, ch); } + // Overload of FormatSink::Append() for appending the characters of a string + // view to a format sink. void Append(string_view v) { sink_->Append(v); } - // Appends the first `precision` bytes of `v`. If this is less than - // `width`, spaces will be appended first (if `left` is false), or + // FormatSink::PutPaddedString() + // + // Appends `precision` number of bytes of `v` to the format sink. If this is + // less than `width`, spaces will be appended first (if `left` is false), or // after (if `left` is true) to ensure the total amount appended is // at least `width`. bool PutPaddedString(string_view v, int width, int precision, bool left) { return sink_->PutPaddedString(v, width, precision, left); } + // Support `absl::Format(&sink, format, args...)`. + friend void AbslFormatFlush(FormatSink* sink, absl::string_view v) { + sink->Append(v); + } + private: friend str_format_internal::FormatSinkImpl; explicit FormatSink(str_format_internal::FormatSinkImpl* s) : sink_(s) {} diff --git a/contrib/restricted/abseil-cpp/absl/strings/string_view.cc b/contrib/restricted/abseil-cpp/absl/strings/string_view.cc index adce3be94e..e2261625f9 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/string_view.cc +++ b/contrib/restricted/abseil-cpp/absl/strings/string_view.cc @@ -32,7 +32,7 @@ void WritePadding(std::ostream& o, size_t pad) { memset(fill_buf, o.fill(), sizeof(fill_buf)); while (pad) { size_t n = std::min(pad, sizeof(fill_buf)); - o.write(fill_buf, n); + o.write(fill_buf, static_cast<std::streamsize>(n)); pad -= n; } } @@ -63,7 +63,7 @@ std::ostream& operator<<(std::ostream& o, string_view piece) { size_t lpad = 0; size_t rpad = 0; if (static_cast<size_t>(o.width()) > piece.size()) { - size_t pad = o.width() - piece.size(); + size_t pad = static_cast<size_t>(o.width()) - piece.size(); if ((o.flags() & o.adjustfield) == o.left) { rpad = pad; } else { @@ -71,7 +71,7 @@ std::ostream& operator<<(std::ostream& o, string_view piece) { } } if (lpad) WritePadding(o, lpad); - o.write(piece.data(), piece.size()); + o.write(piece.data(), static_cast<std::streamsize>(piece.size())); if (rpad) WritePadding(o, rpad); o.width(0); } @@ -86,7 +86,7 @@ string_view::size_type string_view::find(string_view s, } const char* result = strings_internal::memmatch(ptr_ + pos, length_ - pos, s.ptr_, s.length_); - return result ? result - ptr_ : npos; + return result ? static_cast<size_type>(result - ptr_) : npos; } string_view::size_type string_view::find(char c, size_type pos) const noexcept { @@ -95,7 +95,7 @@ string_view::size_type string_view::find(char c, size_type pos) const noexcept { } const char* result = static_cast<const char*>(memchr(ptr_ + pos, c, length_ - pos)); - return result != nullptr ? result - ptr_ : npos; + return result != nullptr ? static_cast<size_type>(result - ptr_) : npos; } string_view::size_type string_view::rfind(string_view s, @@ -104,7 +104,7 @@ string_view::size_type string_view::rfind(string_view s, if (s.empty()) return std::min(length_, pos); const char* last = ptr_ + std::min(length_ - s.length_, pos) + s.length_; const char* result = std::find_end(ptr_, last, s.ptr_, s.ptr_ + s.length_); - return result != last ? result - ptr_ : npos; + return result != last ? static_cast<size_type>(result - ptr_) : npos; } // Search range is [0..pos] inclusive. If pos == npos, search everything. diff --git a/contrib/restricted/abseil-cpp/absl/strings/string_view.h b/contrib/restricted/abseil-cpp/absl/strings/string_view.h index d0c870696c..01867b6414 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/string_view.h +++ b/contrib/restricted/abseil-cpp/absl/strings/string_view.h @@ -65,12 +65,6 @@ ABSL_NAMESPACE_END #define ABSL_INTERNAL_STRING_VIEW_MEMCMP memcmp #endif // ABSL_HAVE_BUILTIN(__builtin_memcmp) -#if defined(__cplusplus) && __cplusplus >= 201402L -#define ABSL_INTERNAL_STRING_VIEW_CXX14_CONSTEXPR constexpr -#else -#define ABSL_INTERNAL_STRING_VIEW_CXX14_CONSTEXPR -#endif - namespace absl { ABSL_NAMESPACE_BEGIN @@ -343,7 +337,7 @@ class string_view { // // Removes the first `n` characters from the `string_view`. Note that the // underlying string is not changed, only the view. - ABSL_INTERNAL_STRING_VIEW_CXX14_CONSTEXPR void remove_prefix(size_type n) { + constexpr void remove_prefix(size_type n) { ABSL_HARDENING_ASSERT(n <= length_); ptr_ += n; length_ -= n; @@ -353,7 +347,7 @@ class string_view { // // Removes the last `n` characters from the `string_view`. Note that the // underlying string is not changed, only the view. - ABSL_INTERNAL_STRING_VIEW_CXX14_CONSTEXPR void remove_suffix(size_type n) { + constexpr void remove_suffix(size_type n) { ABSL_HARDENING_ASSERT(n <= length_); length_ -= n; } @@ -361,7 +355,7 @@ class string_view { // string_view::swap() // // Swaps this `string_view` with another `string_view`. - ABSL_INTERNAL_STRING_VIEW_CXX14_CONSTEXPR void swap(string_view& s) noexcept { + constexpr void swap(string_view& s) noexcept { auto t = *this; *this = s; s = t; @@ -680,7 +674,6 @@ std::ostream& operator<<(std::ostream& o, string_view piece); ABSL_NAMESPACE_END } // namespace absl -#undef ABSL_INTERNAL_STRING_VIEW_CXX14_CONSTEXPR #undef ABSL_INTERNAL_STRING_VIEW_MEMCMP #endif // ABSL_USES_STD_STRING_VIEW diff --git a/contrib/restricted/abseil-cpp/absl/strings/substitute.cc b/contrib/restricted/abseil-cpp/absl/strings/substitute.cc index 8980b198c2..33a39305db 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/substitute.cc +++ b/contrib/restricted/abseil-cpp/absl/strings/substitute.cc @@ -40,7 +40,8 @@ void SubstituteAndAppendArray(std::string* output, absl::string_view format, absl::CEscape(format).c_str()); #endif return; - } else if (absl::ascii_isdigit(format[i + 1])) { + } else if (absl::ascii_isdigit( + static_cast<unsigned char>(format[i + 1]))) { int index = format[i + 1] - '0'; if (static_cast<size_t>(index) >= num_args) { #ifndef NDEBUG @@ -80,7 +81,7 @@ void SubstituteAndAppendArray(std::string* output, absl::string_view format, char* target = &(*output)[original_size]; for (size_t i = 0; i < format.size(); i++) { if (format[i] == '$') { - if (absl::ascii_isdigit(format[i + 1])) { + if (absl::ascii_isdigit(static_cast<unsigned char>(format[i + 1]))) { const absl::string_view src = args_array[format[i + 1] - '0']; target = std::copy(src.begin(), src.end(), target); ++i; // Skip next char. @@ -110,7 +111,8 @@ Arg::Arg(const void* value) { } while (num != 0); *--ptr = 'x'; *--ptr = '0'; - piece_ = absl::string_view(ptr, scratch_ + sizeof(scratch_) - ptr); + piece_ = absl::string_view( + ptr, static_cast<size_t>(scratch_ + sizeof(scratch_) - ptr)); } } @@ -132,7 +134,7 @@ Arg::Arg(Hex hex) { beg = writer; } - piece_ = absl::string_view(beg, end - beg); + piece_ = absl::string_view(beg, static_cast<size_t>(end - beg)); } // TODO(jorg): Don't duplicate so much code between here and str_cat.cc @@ -147,7 +149,7 @@ Arg::Arg(Dec dec) { *--writer = '0' + (value % 10); value /= 10; } - *--writer = '0' + value; + *--writer = '0' + static_cast<char>(value); if (neg) *--writer = '-'; ptrdiff_t fillers = writer - minfill; @@ -164,7 +166,7 @@ Arg::Arg(Dec dec) { if (add_sign_again) *--writer = '-'; } - piece_ = absl::string_view(writer, end - writer); + piece_ = absl::string_view(writer, static_cast<size_t>(end - writer)); } } // namespace substitute_internal diff --git a/contrib/restricted/abseil-cpp/absl/strings/substitute.h b/contrib/restricted/abseil-cpp/absl/strings/substitute.h index 6d2b08abb9..d6a5a690d7 100644 --- a/contrib/restricted/abseil-cpp/absl/strings/substitute.h +++ b/contrib/restricted/abseil-cpp/absl/strings/substitute.h @@ -55,6 +55,8 @@ // * bool (Printed as "true" or "false") // * pointer types other than char* (Printed as "0x<lower case hex string>", // except that null is printed as "NULL") +// * user-defined types via the `AbslStringify()` customization point. See the +// documentation for `absl::StrCat` for an explanation on how to use this. // // If an invalid format string is provided, Substitute returns an empty string // and SubstituteAndAppend does not change the provided output string. @@ -79,6 +81,7 @@ #include "absl/base/port.h" #include "absl/strings/ascii.h" #include "absl/strings/escaping.h" +#include "absl/strings/internal/stringify_sink.h" #include "absl/strings/numbers.h" #include "absl/strings/str_cat.h" #include "absl/strings/str_split.h" @@ -102,14 +105,14 @@ class Arg { // Overloads for string-y things // // Explicitly overload `const char*` so the compiler doesn't cast to `bool`. - Arg(const char* value) // NOLINT(runtime/explicit) + Arg(const char* value) // NOLINT(google-explicit-constructor) : piece_(absl::NullSafeStringView(value)) {} template <typename Allocator> Arg( // NOLINT const std::basic_string<char, std::char_traits<char>, Allocator>& value) noexcept : piece_(value) {} - Arg(absl::string_view value) // NOLINT(runtime/explicit) + Arg(absl::string_view value) // NOLINT(google-explicit-constructor) : piece_(value) {} // Overloads for primitives @@ -119,45 +122,67 @@ class Arg { // probably using them as 8-bit integers and would probably prefer an integer // representation. However, we can't really know, so we make the caller decide // what to do. - Arg(char value) // NOLINT(runtime/explicit) + Arg(char value) // NOLINT(google-explicit-constructor) : piece_(scratch_, 1) { scratch_[0] = value; } Arg(short value) // NOLINT(*) : piece_(scratch_, - numbers_internal::FastIntToBuffer(value, scratch_) - scratch_) {} + static_cast<size_t>( + numbers_internal::FastIntToBuffer(value, scratch_) - + scratch_)) {} Arg(unsigned short value) // NOLINT(*) : piece_(scratch_, - numbers_internal::FastIntToBuffer(value, scratch_) - scratch_) {} - Arg(int value) // NOLINT(runtime/explicit) + static_cast<size_t>( + numbers_internal::FastIntToBuffer(value, scratch_) - + scratch_)) {} + Arg(int value) // NOLINT(google-explicit-constructor) : piece_(scratch_, - numbers_internal::FastIntToBuffer(value, scratch_) - scratch_) {} - Arg(unsigned int value) // NOLINT(runtime/explicit) + static_cast<size_t>( + numbers_internal::FastIntToBuffer(value, scratch_) - + scratch_)) {} + Arg(unsigned int value) // NOLINT(google-explicit-constructor) : piece_(scratch_, - numbers_internal::FastIntToBuffer(value, scratch_) - scratch_) {} + static_cast<size_t>( + numbers_internal::FastIntToBuffer(value, scratch_) - + scratch_)) {} Arg(long value) // NOLINT(*) : piece_(scratch_, - numbers_internal::FastIntToBuffer(value, scratch_) - scratch_) {} + static_cast<size_t>( + numbers_internal::FastIntToBuffer(value, scratch_) - + scratch_)) {} Arg(unsigned long value) // NOLINT(*) : piece_(scratch_, - numbers_internal::FastIntToBuffer(value, scratch_) - scratch_) {} + static_cast<size_t>( + numbers_internal::FastIntToBuffer(value, scratch_) - + scratch_)) {} Arg(long long value) // NOLINT(*) : piece_(scratch_, - numbers_internal::FastIntToBuffer(value, scratch_) - scratch_) {} + static_cast<size_t>( + numbers_internal::FastIntToBuffer(value, scratch_) - + scratch_)) {} Arg(unsigned long long value) // NOLINT(*) : piece_(scratch_, - numbers_internal::FastIntToBuffer(value, scratch_) - scratch_) {} - Arg(float value) // NOLINT(runtime/explicit) + static_cast<size_t>( + numbers_internal::FastIntToBuffer(value, scratch_) - + scratch_)) {} + Arg(float value) // NOLINT(google-explicit-constructor) : piece_(scratch_, numbers_internal::SixDigitsToBuffer(value, scratch_)) { } - Arg(double value) // NOLINT(runtime/explicit) + Arg(double value) // NOLINT(google-explicit-constructor) : piece_(scratch_, numbers_internal::SixDigitsToBuffer(value, scratch_)) { } - Arg(bool value) // NOLINT(runtime/explicit) + Arg(bool value) // NOLINT(google-explicit-constructor) : piece_(value ? "true" : "false") {} - Arg(Hex hex); // NOLINT(runtime/explicit) - Arg(Dec dec); // NOLINT(runtime/explicit) + template <typename T, typename = typename std::enable_if< + strings_internal::HasAbslStringify<T>::value>::type> + Arg( // NOLINT(google-explicit-constructor) + const T& v, strings_internal::StringifySink&& sink = {}) + : piece_(strings_internal::ExtractStringification(sink, v)) {} + + Arg(Hex hex); // NOLINT(google-explicit-constructor) + Arg(Dec dec); // NOLINT(google-explicit-constructor) // vector<bool>::reference and const_reference require special help to convert // to `Arg` because it requires two user defined conversions. @@ -172,13 +197,14 @@ class Arg { // `void*` values, with the exception of `char*`, are printed as // "0x<hex value>". However, in the case of `nullptr`, "NULL" is printed. - Arg(const void* value); // NOLINT(runtime/explicit) + Arg(const void* value); // NOLINT(google-explicit-constructor) // Normal enums are already handled by the integer formatters. // This overload matches only scoped enums. template <typename T, typename = typename std::enable_if< - std::is_enum<T>{} && !std::is_convertible<T, int>{}>::type> + std::is_enum<T>{} && !std::is_convertible<T, int>{} && + !strings_internal::HasAbslStringify<T>::value>::type> Arg(T value) // NOLINT(google-explicit-constructor) : Arg(static_cast<typename std::underlying_type<T>::type>(value)) {} |