diff options
author | BobTheBob9 <for.oliver.kirkham@gmail.com> | 2023-01-04 18:08:13 +0000 |
---|---|---|
committer | BobTheBob9 <for.oliver.kirkham@gmail.com> | 2023-01-04 18:08:13 +0000 |
commit | e9f8a34dc3ca736f307cf195390aee7c8f5cb456 (patch) | |
tree | e65d1c2b1888895baf65bbde73e0a1a02cbb091a /include/rapidjson/internal | |
parent | 6c45c7e94634c340019d12615b7eae8f3c0f49b3 (diff) | |
parent | 4fb1ae07d8f078e7abde99900c6a8286148a380a (diff) | |
download | NorthstarLauncher-e9f8a34dc3ca736f307cf195390aee7c8f5cb456.tar.gz NorthstarLauncher-e9f8a34dc3ca736f307cf195390aee7c8f5cb456.zip |
Merge remote-tracking branch 'origin/main' into experimental-bots-pr
Diffstat (limited to 'include/rapidjson/internal')
-rw-r--r-- | include/rapidjson/internal/biginteger.h | 290 | ||||
-rw-r--r-- | include/rapidjson/internal/diyfp.h | 258 | ||||
-rw-r--r-- | include/rapidjson/internal/dtoa.h | 245 | ||||
-rw-r--r-- | include/rapidjson/internal/ieee754.h | 78 | ||||
-rw-r--r-- | include/rapidjson/internal/itoa.h | 304 | ||||
-rw-r--r-- | include/rapidjson/internal/meta.h | 181 | ||||
-rw-r--r-- | include/rapidjson/internal/pow10.h | 55 | ||||
-rw-r--r-- | include/rapidjson/internal/regex.h | 701 | ||||
-rw-r--r-- | include/rapidjson/internal/stack.h | 230 | ||||
-rw-r--r-- | include/rapidjson/internal/strfunc.h | 55 | ||||
-rw-r--r-- | include/rapidjson/internal/strtod.h | 269 | ||||
-rw-r--r-- | include/rapidjson/internal/swap.h | 46 |
12 files changed, 2712 insertions, 0 deletions
diff --git a/include/rapidjson/internal/biginteger.h b/include/rapidjson/internal/biginteger.h new file mode 100644 index 00000000..9d3e88c9 --- /dev/null +++ b/include/rapidjson/internal/biginteger.h @@ -0,0 +1,290 @@ +// Tencent is pleased to support the open source community by making RapidJSON available. +// +// Copyright (C) 2015 THL A29 Limited, a Tencent company, and Milo Yip. All rights reserved. +// +// Licensed under the MIT License (the "License"); you may not use this file except +// in compliance with the License. You may obtain a copy of the License at +// +// http://opensource.org/licenses/MIT +// +// 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 RAPIDJSON_BIGINTEGER_H_ +#define RAPIDJSON_BIGINTEGER_H_ + +#include "../rapidjson.h" + +#if defined(_MSC_VER) && defined(_M_AMD64) +#include <intrin.h> // for _umul128 +#pragma intrinsic(_umul128) +#endif + +RAPIDJSON_NAMESPACE_BEGIN +namespace internal { + +class BigInteger { +public: + typedef uint64_t Type; + + BigInteger(const BigInteger& rhs) : count_(rhs.count_) { + std::memcpy(digits_, rhs.digits_, count_ * sizeof(Type)); + } + + explicit BigInteger(uint64_t u) : count_(1) { + digits_[0] = u; + } + + BigInteger(const char* decimals, size_t length) : count_(1) { + RAPIDJSON_ASSERT(length > 0); + digits_[0] = 0; + size_t i = 0; + const size_t kMaxDigitPerIteration = 19; // 2^64 = 18446744073709551616 > 10^19 + while (length >= kMaxDigitPerIteration) { + AppendDecimal64(decimals + i, decimals + i + kMaxDigitPerIteration); + length -= kMaxDigitPerIteration; + i += kMaxDigitPerIteration; + } + + if (length > 0) + AppendDecimal64(decimals + i, decimals + i + length); + } + + BigInteger& operator=(const BigInteger &rhs) + { + if (this != &rhs) { + count_ = rhs.count_; + std::memcpy(digits_, rhs.digits_, count_ * sizeof(Type)); + } + return *this; + } + + BigInteger& operator=(uint64_t u) { + digits_[0] = u; + count_ = 1; + return *this; + } + + BigInteger& operator+=(uint64_t u) { + Type backup = digits_[0]; + digits_[0] += u; + for (size_t i = 0; i < count_ - 1; i++) { + if (digits_[i] >= backup) + return *this; // no carry + backup = digits_[i + 1]; + digits_[i + 1] += 1; + } + + // Last carry + if (digits_[count_ - 1] < backup) + PushBack(1); + + return *this; + } + + BigInteger& operator*=(uint64_t u) { + if (u == 0) return *this = 0; + if (u == 1) return *this; + if (*this == 1) return *this = u; + + uint64_t k = 0; + for (size_t i = 0; i < count_; i++) { + uint64_t hi; + digits_[i] = MulAdd64(digits_[i], u, k, &hi); + k = hi; + } + + if (k > 0) + PushBack(k); + + return *this; + } + + BigInteger& operator*=(uint32_t u) { + if (u == 0) return *this = 0; + if (u == 1) return *this; + if (*this == 1) return *this = u; + + uint64_t k = 0; + for (size_t i = 0; i < count_; i++) { + const uint64_t c = digits_[i] >> 32; + const uint64_t d = digits_[i] & 0xFFFFFFFF; + const uint64_t uc = u * c; + const uint64_t ud = u * d; + const uint64_t p0 = ud + k; + const uint64_t p1 = uc + (p0 >> 32); + digits_[i] = (p0 & 0xFFFFFFFF) | (p1 << 32); + k = p1 >> 32; + } + + if (k > 0) + PushBack(k); + + return *this; + } + + BigInteger& operator<<=(size_t shift) { + if (IsZero() || shift == 0) return *this; + + size_t offset = shift / kTypeBit; + size_t interShift = shift % kTypeBit; + RAPIDJSON_ASSERT(count_ + offset <= kCapacity); + + if (interShift == 0) { + std::memmove(&digits_[count_ - 1 + offset], &digits_[count_ - 1], count_ * sizeof(Type)); + count_ += offset; + } + else { + digits_[count_] = 0; + for (size_t i = count_; i > 0; i--) + digits_[i + offset] = (digits_[i] << interShift) | (digits_[i - 1] >> (kTypeBit - interShift)); + digits_[offset] = digits_[0] << interShift; + count_ += offset; + if (digits_[count_]) + count_++; + } + + std::memset(digits_, 0, offset * sizeof(Type)); + + return *this; + } + + bool operator==(const BigInteger& rhs) const { + return count_ == rhs.count_ && std::memcmp(digits_, rhs.digits_, count_ * sizeof(Type)) == 0; + } + + bool operator==(const Type rhs) const { + return count_ == 1 && digits_[0] == rhs; + } + + BigInteger& MultiplyPow5(unsigned exp) { + static const uint32_t kPow5[12] = { + 5, + 5 * 5, + 5 * 5 * 5, + 5 * 5 * 5 * 5, + 5 * 5 * 5 * 5 * 5, + 5 * 5 * 5 * 5 * 5 * 5, + 5 * 5 * 5 * 5 * 5 * 5 * 5, + 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5, + 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5, + 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5, + 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5, + 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5 * 5 + }; + if (exp == 0) return *this; + for (; exp >= 27; exp -= 27) *this *= RAPIDJSON_UINT64_C2(0X6765C793, 0XFA10079D); // 5^27 + for (; exp >= 13; exp -= 13) *this *= static_cast<uint32_t>(1220703125u); // 5^13 + if (exp > 0) *this *= kPow5[exp - 1]; + return *this; + } + + // Compute absolute difference of this and rhs. + // Assume this != rhs + bool Difference(const BigInteger& rhs, BigInteger* out) const { + int cmp = Compare(rhs); + RAPIDJSON_ASSERT(cmp != 0); + const BigInteger *a, *b; // Makes a > b + bool ret; + if (cmp < 0) { a = &rhs; b = this; ret = true; } + else { a = this; b = &rhs; ret = false; } + + Type borrow = 0; + for (size_t i = 0; i < a->count_; i++) { + Type d = a->digits_[i] - borrow; + if (i < b->count_) + d -= b->digits_[i]; + borrow = (d > a->digits_[i]) ? 1 : 0; + out->digits_[i] = d; + if (d != 0) + out->count_ = i + 1; + } + + return ret; + } + + int Compare(const BigInteger& rhs) const { + if (count_ != rhs.count_) + return count_ < rhs.count_ ? -1 : 1; + + for (size_t i = count_; i-- > 0;) + if (digits_[i] != rhs.digits_[i]) + return digits_[i] < rhs.digits_[i] ? -1 : 1; + + return 0; + } + + size_t GetCount() const { return count_; } + Type GetDigit(size_t index) const { RAPIDJSON_ASSERT(index < count_); return digits_[index]; } + bool IsZero() const { return count_ == 1 && digits_[0] == 0; } + +private: + void AppendDecimal64(const char* begin, const char* end) { + uint64_t u = ParseUint64(begin, end); + if (IsZero()) + *this = u; + else { + unsigned exp = static_cast<unsigned>(end - begin); + (MultiplyPow5(exp) <<= exp) += u; // *this = *this * 10^exp + u + } + } + + void PushBack(Type digit) { + RAPIDJSON_ASSERT(count_ < kCapacity); + digits_[count_++] = digit; + } + + static uint64_t ParseUint64(const char* begin, const char* end) { + uint64_t r = 0; + for (const char* p = begin; p != end; ++p) { + RAPIDJSON_ASSERT(*p >= '0' && *p <= '9'); + r = r * 10u + static_cast<unsigned>(*p - '0'); + } + return r; + } + + // Assume a * b + k < 2^128 + static uint64_t MulAdd64(uint64_t a, uint64_t b, uint64_t k, uint64_t* outHigh) { +#if defined(_MSC_VER) && defined(_M_AMD64) + uint64_t low = _umul128(a, b, outHigh) + k; + if (low < k) + (*outHigh)++; + return low; +#elif (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 6)) && defined(__x86_64__) + __extension__ typedef unsigned __int128 uint128; + uint128 p = static_cast<uint128>(a) * static_cast<uint128>(b); + p += k; + *outHigh = static_cast<uint64_t>(p >> 64); + return static_cast<uint64_t>(p); +#else + const uint64_t a0 = a & 0xFFFFFFFF, a1 = a >> 32, b0 = b & 0xFFFFFFFF, b1 = b >> 32; + uint64_t x0 = a0 * b0, x1 = a0 * b1, x2 = a1 * b0, x3 = a1 * b1; + x1 += (x0 >> 32); // can't give carry + x1 += x2; + if (x1 < x2) + x3 += (static_cast<uint64_t>(1) << 32); + uint64_t lo = (x1 << 32) + (x0 & 0xFFFFFFFF); + uint64_t hi = x3 + (x1 >> 32); + + lo += k; + if (lo < k) + hi++; + *outHigh = hi; + return lo; +#endif + } + + static const size_t kBitCount = 3328; // 64bit * 54 > 10^1000 + static const size_t kCapacity = kBitCount / sizeof(Type); + static const size_t kTypeBit = sizeof(Type) * 8; + + Type digits_[kCapacity]; + size_t count_; +}; + +} // namespace internal +RAPIDJSON_NAMESPACE_END + +#endif // RAPIDJSON_BIGINTEGER_H_ diff --git a/include/rapidjson/internal/diyfp.h b/include/rapidjson/internal/diyfp.h new file mode 100644 index 00000000..c9fefdc6 --- /dev/null +++ b/include/rapidjson/internal/diyfp.h @@ -0,0 +1,258 @@ +// Tencent is pleased to support the open source community by making RapidJSON available. +// +// Copyright (C) 2015 THL A29 Limited, a Tencent company, and Milo Yip. All rights reserved. +// +// Licensed under the MIT License (the "License"); you may not use this file except +// in compliance with the License. You may obtain a copy of the License at +// +// http://opensource.org/licenses/MIT +// +// 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. + +// This is a C++ header-only implementation of Grisu2 algorithm from the publication: +// Loitsch, Florian. "Printing floating-point numbers quickly and accurately with +// integers." ACM Sigplan Notices 45.6 (2010): 233-243. + +#ifndef RAPIDJSON_DIYFP_H_ +#define RAPIDJSON_DIYFP_H_ + +#include "../rapidjson.h" + +#if defined(_MSC_VER) && defined(_M_AMD64) +#include <intrin.h> +#pragma intrinsic(_BitScanReverse64) +#pragma intrinsic(_umul128) +#endif + +RAPIDJSON_NAMESPACE_BEGIN +namespace internal { + +#ifdef __GNUC__ +RAPIDJSON_DIAG_PUSH +RAPIDJSON_DIAG_OFF(effc++) +#endif + +#ifdef __clang__ +RAPIDJSON_DIAG_PUSH +RAPIDJSON_DIAG_OFF(padded) +#endif + +struct DiyFp { + DiyFp() : f(), e() {} + + DiyFp(uint64_t fp, int exp) : f(fp), e(exp) {} + + explicit DiyFp(double d) { + union { + double d; + uint64_t u64; + } u = { d }; + + int biased_e = static_cast<int>((u.u64 & kDpExponentMask) >> kDpSignificandSize); + uint64_t significand = (u.u64 & kDpSignificandMask); + if (biased_e != 0) { + f = significand + kDpHiddenBit; + e = biased_e - kDpExponentBias; + } + else { + f = significand; + e = kDpMinExponent + 1; + } + } + + DiyFp operator-(const DiyFp& rhs) const { + return DiyFp(f - rhs.f, e); + } + + DiyFp operator*(const DiyFp& rhs) const { +#if defined(_MSC_VER) && defined(_M_AMD64) + uint64_t h; + uint64_t l = _umul128(f, rhs.f, &h); + if (l & (uint64_t(1) << 63)) // rounding + h++; + return DiyFp(h, e + rhs.e + 64); +#elif (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 6)) && defined(__x86_64__) + __extension__ typedef unsigned __int128 uint128; + uint128 p = static_cast<uint128>(f) * static_cast<uint128>(rhs.f); + uint64_t h = static_cast<uint64_t>(p >> 64); + uint64_t l = static_cast<uint64_t>(p); + if (l & (uint64_t(1) << 63)) // rounding + h++; + return DiyFp(h, e + rhs.e + 64); +#else + const uint64_t M32 = 0xFFFFFFFF; + const uint64_t a = f >> 32; + const uint64_t b = f & M32; + const uint64_t c = rhs.f >> 32; + const uint64_t d = rhs.f & M32; + const uint64_t ac = a * c; + const uint64_t bc = b * c; + const uint64_t ad = a * d; + const uint64_t bd = b * d; + uint64_t tmp = (bd >> 32) + (ad & M32) + (bc & M32); + tmp += 1U << 31; /// mult_round + return DiyFp(ac + (ad >> 32) + (bc >> 32) + (tmp >> 32), e + rhs.e + 64); +#endif + } + + DiyFp Normalize() const { +#if defined(_MSC_VER) && defined(_M_AMD64) + unsigned long index; + _BitScanReverse64(&index, f); + return DiyFp(f << (63 - index), e - (63 - index)); +#elif defined(__GNUC__) && __GNUC__ >= 4 + int s = __builtin_clzll(f); + return DiyFp(f << s, e - s); +#else + DiyFp res = *this; + while (!(res.f & (static_cast<uint64_t>(1) << 63))) { + res.f <<= 1; + res.e--; + } + return res; +#endif + } + + DiyFp NormalizeBoundary() const { + DiyFp res = *this; + while (!(res.f & (kDpHiddenBit << 1))) { + res.f <<= 1; + res.e--; + } + res.f <<= (kDiySignificandSize - kDpSignificandSize - 2); + res.e = res.e - (kDiySignificandSize - kDpSignificandSize - 2); + return res; + } + + void NormalizedBoundaries(DiyFp* minus, DiyFp* plus) const { + DiyFp pl = DiyFp((f << 1) + 1, e - 1).NormalizeBoundary(); + DiyFp mi = (f == kDpHiddenBit) ? DiyFp((f << 2) - 1, e - 2) : DiyFp((f << 1) - 1, e - 1); + mi.f <<= mi.e - pl.e; + mi.e = pl.e; + *plus = pl; + *minus = mi; + } + + double ToDouble() const { + union { + double d; + uint64_t u64; + }u; + const uint64_t be = (e == kDpDenormalExponent && (f & kDpHiddenBit) == 0) ? 0 : + static_cast<uint64_t>(e + kDpExponentBias); + u.u64 = (f & kDpSignificandMask) | (be << kDpSignificandSize); + return u.d; + } + + static const int kDiySignificandSize = 64; + static const int kDpSignificandSize = 52; + static const int kDpExponentBias = 0x3FF + kDpSignificandSize; + static const int kDpMaxExponent = 0x7FF - kDpExponentBias; + static const int kDpMinExponent = -kDpExponentBias; + static const int kDpDenormalExponent = -kDpExponentBias + 1; + static const uint64_t kDpExponentMask = RAPIDJSON_UINT64_C2(0x7FF00000, 0x00000000); + static const uint64_t kDpSignificandMask = RAPIDJSON_UINT64_C2(0x000FFFFF, 0xFFFFFFFF); + static const uint64_t kDpHiddenBit = RAPIDJSON_UINT64_C2(0x00100000, 0x00000000); + + uint64_t f; + int e; +}; + +inline DiyFp GetCachedPowerByIndex(size_t index) { + // 10^-348, 10^-340, ..., 10^340 + static const uint64_t kCachedPowers_F[] = { + RAPIDJSON_UINT64_C2(0xfa8fd5a0, 0x081c0288), RAPIDJSON_UINT64_C2(0xbaaee17f, 0xa23ebf76), + RAPIDJSON_UINT64_C2(0x8b16fb20, 0x3055ac76), RAPIDJSON_UINT64_C2(0xcf42894a, 0x5dce35ea), + RAPIDJSON_UINT64_C2(0x9a6bb0aa, 0x55653b2d), RAPIDJSON_UINT64_C2(0xe61acf03, 0x3d1a45df), + RAPIDJSON_UINT64_C2(0xab70fe17, 0xc79ac6ca), RAPIDJSON_UINT64_C2(0xff77b1fc, 0xbebcdc4f), + RAPIDJSON_UINT64_C2(0xbe5691ef, 0x416bd60c), RAPIDJSON_UINT64_C2(0x8dd01fad, 0x907ffc3c), + RAPIDJSON_UINT64_C2(0xd3515c28, 0x31559a83), RAPIDJSON_UINT64_C2(0x9d71ac8f, 0xada6c9b5), + RAPIDJSON_UINT64_C2(0xea9c2277, 0x23ee8bcb), RAPIDJSON_UINT64_C2(0xaecc4991, 0x4078536d), + RAPIDJSON_UINT64_C2(0x823c1279, 0x5db6ce57), RAPIDJSON_UINT64_C2(0xc2109436, 0x4dfb5637), + RAPIDJSON_UINT64_C2(0x9096ea6f, 0x3848984f), RAPIDJSON_UINT64_C2(0xd77485cb, 0x25823ac7), + RAPIDJSON_UINT64_C2(0xa086cfcd, 0x97bf97f4), RAPIDJSON_UINT64_C2(0xef340a98, 0x172aace5), + RAPIDJSON_UINT64_C2(0xb23867fb, 0x2a35b28e), RAPIDJSON_UINT64_C2(0x84c8d4df, 0xd2c63f3b), + RAPIDJSON_UINT64_C2(0xc5dd4427, 0x1ad3cdba), RAPIDJSON_UINT64_C2(0x936b9fce, 0xbb25c996), + RAPIDJSON_UINT64_C2(0xdbac6c24, 0x7d62a584), RAPIDJSON_UINT64_C2(0xa3ab6658, 0x0d5fdaf6), + RAPIDJSON_UINT64_C2(0xf3e2f893, 0xdec3f126), RAPIDJSON_UINT64_C2(0xb5b5ada8, 0xaaff80b8), + RAPIDJSON_UINT64_C2(0x87625f05, 0x6c7c4a8b), RAPIDJSON_UINT64_C2(0xc9bcff60, 0x34c13053), + RAPIDJSON_UINT64_C2(0x964e858c, 0x91ba2655), RAPIDJSON_UINT64_C2(0xdff97724, 0x70297ebd), + RAPIDJSON_UINT64_C2(0xa6dfbd9f, 0xb8e5b88f), RAPIDJSON_UINT64_C2(0xf8a95fcf, 0x88747d94), + RAPIDJSON_UINT64_C2(0xb9447093, 0x8fa89bcf), RAPIDJSON_UINT64_C2(0x8a08f0f8, 0xbf0f156b), + RAPIDJSON_UINT64_C2(0xcdb02555, 0x653131b6), RAPIDJSON_UINT64_C2(0x993fe2c6, 0xd07b7fac), + RAPIDJSON_UINT64_C2(0xe45c10c4, 0x2a2b3b06), RAPIDJSON_UINT64_C2(0xaa242499, 0x697392d3), + RAPIDJSON_UINT64_C2(0xfd87b5f2, 0x8300ca0e), RAPIDJSON_UINT64_C2(0xbce50864, 0x92111aeb), + RAPIDJSON_UINT64_C2(0x8cbccc09, 0x6f5088cc), RAPIDJSON_UINT64_C2(0xd1b71758, 0xe219652c), + RAPIDJSON_UINT64_C2(0x9c400000, 0x00000000), RAPIDJSON_UINT64_C2(0xe8d4a510, 0x00000000), + RAPIDJSON_UINT64_C2(0xad78ebc5, 0xac620000), RAPIDJSON_UINT64_C2(0x813f3978, 0xf8940984), + RAPIDJSON_UINT64_C2(0xc097ce7b, 0xc90715b3), RAPIDJSON_UINT64_C2(0x8f7e32ce, 0x7bea5c70), + RAPIDJSON_UINT64_C2(0xd5d238a4, 0xabe98068), RAPIDJSON_UINT64_C2(0x9f4f2726, 0x179a2245), + RAPIDJSON_UINT64_C2(0xed63a231, 0xd4c4fb27), RAPIDJSON_UINT64_C2(0xb0de6538, 0x8cc8ada8), + RAPIDJSON_UINT64_C2(0x83c7088e, 0x1aab65db), RAPIDJSON_UINT64_C2(0xc45d1df9, 0x42711d9a), + RAPIDJSON_UINT64_C2(0x924d692c, 0xa61be758), RAPIDJSON_UINT64_C2(0xda01ee64, 0x1a708dea), + RAPIDJSON_UINT64_C2(0xa26da399, 0x9aef774a), RAPIDJSON_UINT64_C2(0xf209787b, 0xb47d6b85), + RAPIDJSON_UINT64_C2(0xb454e4a1, 0x79dd1877), RAPIDJSON_UINT64_C2(0x865b8692, 0x5b9bc5c2), + RAPIDJSON_UINT64_C2(0xc83553c5, 0xc8965d3d), RAPIDJSON_UINT64_C2(0x952ab45c, 0xfa97a0b3), + RAPIDJSON_UINT64_C2(0xde469fbd, 0x99a05fe3), RAPIDJSON_UINT64_C2(0xa59bc234, 0xdb398c25), + RAPIDJSON_UINT64_C2(0xf6c69a72, 0xa3989f5c), RAPIDJSON_UINT64_C2(0xb7dcbf53, 0x54e9bece), + RAPIDJSON_UINT64_C2(0x88fcf317, 0xf22241e2), RAPIDJSON_UINT64_C2(0xcc20ce9b, 0xd35c78a5), + RAPIDJSON_UINT64_C2(0x98165af3, 0x7b2153df), RAPIDJSON_UINT64_C2(0xe2a0b5dc, 0x971f303a), + RAPIDJSON_UINT64_C2(0xa8d9d153, 0x5ce3b396), RAPIDJSON_UINT64_C2(0xfb9b7cd9, 0xa4a7443c), + RAPIDJSON_UINT64_C2(0xbb764c4c, 0xa7a44410), RAPIDJSON_UINT64_C2(0x8bab8eef, 0xb6409c1a), + RAPIDJSON_UINT64_C2(0xd01fef10, 0xa657842c), RAPIDJSON_UINT64_C2(0x9b10a4e5, 0xe9913129), + RAPIDJSON_UINT64_C2(0xe7109bfb, 0xa19c0c9d), RAPIDJSON_UINT64_C2(0xac2820d9, 0x623bf429), + RAPIDJSON_UINT64_C2(0x80444b5e, 0x7aa7cf85), RAPIDJSON_UINT64_C2(0xbf21e440, 0x03acdd2d), + RAPIDJSON_UINT64_C2(0x8e679c2f, 0x5e44ff8f), RAPIDJSON_UINT64_C2(0xd433179d, 0x9c8cb841), + RAPIDJSON_UINT64_C2(0x9e19db92, 0xb4e31ba9), RAPIDJSON_UINT64_C2(0xeb96bf6e, 0xbadf77d9), + RAPIDJSON_UINT64_C2(0xaf87023b, 0x9bf0ee6b) + }; + static const int16_t kCachedPowers_E[] = { + -1220, -1193, -1166, -1140, -1113, -1087, -1060, -1034, -1007, -980, + -954, -927, -901, -874, -847, -821, -794, -768, -741, -715, + -688, -661, -635, -608, -582, -555, -529, -502, -475, -449, + -422, -396, -369, -343, -316, -289, -263, -236, -210, -183, + -157, -130, -103, -77, -50, -24, 3, 30, 56, 83, + 109, 136, 162, 189, 216, 242, 269, 295, 322, 348, + 375, 402, 428, 455, 481, 508, 534, 561, 588, 614, + 641, 667, 694, 720, 747, 774, 800, 827, 853, 880, + 907, 933, 960, 986, 1013, 1039, 1066 + }; + return DiyFp(kCachedPowers_F[index], kCachedPowers_E[index]); +} + +inline DiyFp GetCachedPower(int e, int* K) { + + //int k = static_cast<int>(ceil((-61 - e) * 0.30102999566398114)) + 374; + double dk = (-61 - e) * 0.30102999566398114 + 347; // dk must be positive, so can do ceiling in positive + int k = static_cast<int>(dk); + if (dk - k > 0.0) + k++; + + unsigned index = static_cast<unsigned>((k >> 3) + 1); + *K = -(-348 + static_cast<int>(index << 3)); // decimal exponent no need lookup table + + return GetCachedPowerByIndex(index); +} + +inline DiyFp GetCachedPower10(int exp, int *outExp) { + unsigned index = (static_cast<unsigned>(exp) + 348u) / 8u; + *outExp = -348 + static_cast<int>(index) * 8; + return GetCachedPowerByIndex(index); + } + +#ifdef __GNUC__ +RAPIDJSON_DIAG_POP +#endif + +#ifdef __clang__ +RAPIDJSON_DIAG_POP +RAPIDJSON_DIAG_OFF(padded) +#endif + +} // namespace internal +RAPIDJSON_NAMESPACE_END + +#endif // RAPIDJSON_DIYFP_H_ diff --git a/include/rapidjson/internal/dtoa.h b/include/rapidjson/internal/dtoa.h new file mode 100644 index 00000000..8d6350e6 --- /dev/null +++ b/include/rapidjson/internal/dtoa.h @@ -0,0 +1,245 @@ +// Tencent is pleased to support the open source community by making RapidJSON available. +// +// Copyright (C) 2015 THL A29 Limited, a Tencent company, and Milo Yip. All rights reserved. +// +// Licensed under the MIT License (the "License"); you may not use this file except +// in compliance with the License. You may obtain a copy of the License at +// +// http://opensource.org/licenses/MIT +// +// 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. + +// This is a C++ header-only implementation of Grisu2 algorithm from the publication: +// Loitsch, Florian. "Printing floating-point numbers quickly and accurately with +// integers." ACM Sigplan Notices 45.6 (2010): 233-243. + +#ifndef RAPIDJSON_DTOA_ +#define RAPIDJSON_DTOA_ + +#include "itoa.h" // GetDigitsLut() +#include "diyfp.h" +#include "ieee754.h" + +RAPIDJSON_NAMESPACE_BEGIN +namespace internal { + +#ifdef __GNUC__ +RAPIDJSON_DIAG_PUSH +RAPIDJSON_DIAG_OFF(effc++) +RAPIDJSON_DIAG_OFF(array-bounds) // some gcc versions generate wrong warnings https://gcc.gnu.org/bugzilla/show_bug.cgi?id=59124 +#endif + +inline void GrisuRound(char* buffer, int len, uint64_t delta, uint64_t rest, uint64_t ten_kappa, uint64_t wp_w) { + while (rest < wp_w && delta - rest >= ten_kappa && + (rest + ten_kappa < wp_w || /// closer + wp_w - rest > rest + ten_kappa - wp_w)) { + buffer[len - 1]--; + rest += ten_kappa; + } +} + +inline unsigned CountDecimalDigit32(uint32_t n) { + // Simple pure C++ implementation was faster than __builtin_clz version in this situation. + if (n < 10) return 1; + if (n < 100) return 2; + if (n < 1000) return 3; + if (n < 10000) return 4; + if (n < 100000) return 5; + if (n < 1000000) return 6; + if (n < 10000000) return 7; + if (n < 100000000) return 8; + // Will not reach 10 digits in DigitGen() + //if (n < 1000000000) return 9; + //return 10; + return 9; +} + +inline void DigitGen(const DiyFp& W, const DiyFp& Mp, uint64_t delta, char* buffer, int* len, int* K) { + static const uint32_t kPow10[] = { 1, 10, 100, 1000, 10000, 100000, 1000000, 10000000, 100000000, 1000000000 }; + const DiyFp one(uint64_t(1) << -Mp.e, Mp.e); + const DiyFp wp_w = Mp - W; + uint32_t p1 = static_cast<uint32_t>(Mp.f >> -one.e); + uint64_t p2 = Mp.f & (one.f - 1); + unsigned kappa = CountDecimalDigit32(p1); // kappa in [0, 9] + *len = 0; + + while (kappa > 0) { + uint32_t d = 0; + switch (kappa) { + case 9: d = p1 / 100000000; p1 %= 100000000; break; + case 8: d = p1 / 10000000; p1 %= 10000000; break; + case 7: d = p1 / 1000000; p1 %= 1000000; break; + case 6: d = p1 / 100000; p1 %= 100000; break; + case 5: d = p1 / 10000; p1 %= 10000; break; + case 4: d = p1 / 1000; p1 %= 1000; break; + case 3: d = p1 / 100; p1 %= 100; break; + case 2: d = p1 / 10; p1 %= 10; break; + case 1: d = p1; p1 = 0; break; + default:; + } + if (d || *len) + buffer[(*len)++] = static_cast<char>('0' + static_cast<char>(d)); + kappa--; + uint64_t tmp = (static_cast<uint64_t>(p1) << -one.e) + p2; + if (tmp <= delta) { + *K += kappa; + GrisuRound(buffer, *len, delta, tmp, static_cast<uint64_t>(kPow10[kappa]) << -one.e, wp_w.f); + return; + } + } + + // kappa = 0 + for (;;) { + p2 *= 10; + delta *= 10; + char d = static_cast<char>(p2 >> -one.e); + if (d || *len) + buffer[(*len)++] = static_cast<char>('0' + d); + p2 &= one.f - 1; + kappa--; + if (p2 < delta) { + *K += kappa; + int index = -static_cast<int>(kappa); + GrisuRound(buffer, *len, delta, p2, one.f, wp_w.f * (index < 9 ? kPow10[-static_cast<int>(kappa)] : 0)); + return; + } + } +} + +inline void Grisu2(double value, char* buffer, int* length, int* K) { + const DiyFp v(value); + DiyFp w_m, w_p; + v.NormalizedBoundaries(&w_m, &w_p); + + const DiyFp c_mk = GetCachedPower(w_p.e, K); + const DiyFp W = v.Normalize() * c_mk; + DiyFp Wp = w_p * c_mk; + DiyFp Wm = w_m * c_mk; + Wm.f++; + Wp.f--; + DigitGen(W, Wp, Wp.f - Wm.f, buffer, length, K); +} + +inline char* WriteExponent(int K, char* buffer) { + if (K < 0) { + *buffer++ = '-'; + K = -K; + } + + if (K >= 100) { + *buffer++ = static_cast<char>('0' + static_cast<char>(K / 100)); + K %= 100; + const char* d = GetDigitsLut() + K * 2; + *buffer++ = d[0]; + *buffer++ = d[1]; + } + else if (K >= 10) { + const char* d = GetDigitsLut() + K * 2; + *buffer++ = d[0]; + *buffer++ = d[1]; + } + else + *buffer++ = static_cast<char>('0' + static_cast<char>(K)); + + return buffer; +} + +inline char* Prettify(char* buffer, int length, int k, int maxDecimalPlaces) { + const int kk = length + k; // 10^(kk-1) <= v < 10^kk + + if (0 <= k && kk <= 21) { + // 1234e7 -> 12340000000 + for (int i = length; i < kk; i++) + buffer[i] = '0'; + buffer[kk] = '.'; + buffer[kk + 1] = '0'; + return &buffer[kk + 2]; + } + else if (0 < kk && kk <= 21) { + // 1234e-2 -> 12.34 + std::memmove(&buffer[kk + 1], &buffer[kk], static_cast<size_t>(length - kk)); + buffer[kk] = '.'; + if (0 > k + maxDecimalPlaces) { + // When maxDecimalPlaces = 2, 1.2345 -> 1.23, 1.102 -> 1.1 + // Remove extra trailing zeros (at least one) after truncation. + for (int i = kk + maxDecimalPlaces; i > kk + 1; i--) + if (buffer[i] != '0') + return &buffer[i + 1]; + return &buffer[kk + 2]; // Reserve one zero + } + else + return &buffer[length + 1]; + } + else if (-6 < kk && kk <= 0) { + // 1234e-6 -> 0.001234 + const int offset = 2 - kk; + std::memmove(&buffer[offset], &buffer[0], static_cast<size_t>(length)); + buffer[0] = '0'; + buffer[1] = '.'; + for (int i = 2; i < offset; i++) + buffer[i] = '0'; + if (length - kk > maxDecimalPlaces) { + // When maxDecimalPlaces = 2, 0.123 -> 0.12, 0.102 -> 0.1 + // Remove extra trailing zeros (at least one) after truncation. + for (int i = maxDecimalPlaces + 1; i > 2; i--) + if (buffer[i] != '0') + return &buffer[i + 1]; + return &buffer[3]; // Reserve one zero + } + else + return &buffer[length + offset]; + } + else if (kk < -maxDecimalPlaces) { + // Truncate to zero + buffer[0] = '0'; + buffer[1] = '.'; + buffer[2] = '0'; + return &buffer[3]; + } + else if (length == 1) { + // 1e30 + buffer[1] = 'e'; + return WriteExponent(kk - 1, &buffer[2]); + } + else { + // 1234e30 -> 1.234e33 + std::memmove(&buffer[2], &buffer[1], static_cast<size_t>(length - 1)); + buffer[1] = '.'; + buffer[length + 1] = 'e'; + return WriteExponent(kk - 1, &buffer[0 + length + 2]); + } +} + +inline char* dtoa(double value, char* buffer, int maxDecimalPlaces = 324) { + RAPIDJSON_ASSERT(maxDecimalPlaces >= 1); + Double d(value); + if (d.IsZero()) { + if (d.Sign()) + *buffer++ = '-'; // -0.0, Issue #289 + buffer[0] = '0'; + buffer[1] = '.'; + buffer[2] = '0'; + return &buffer[3]; + } + else { + if (value < 0) { + *buffer++ = '-'; + value = -value; + } + int length, K; + Grisu2(value, buffer, &length, &K); + return Prettify(buffer, length, K, maxDecimalPlaces); + } +} + +#ifdef __GNUC__ +RAPIDJSON_DIAG_POP +#endif + +} // namespace internal +RAPIDJSON_NAMESPACE_END + +#endif // RAPIDJSON_DTOA_ diff --git a/include/rapidjson/internal/ieee754.h b/include/rapidjson/internal/ieee754.h new file mode 100644 index 00000000..82bb0b99 --- /dev/null +++ b/include/rapidjson/internal/ieee754.h @@ -0,0 +1,78 @@ +// Tencent is pleased to support the open source community by making RapidJSON available. +// +// Copyright (C) 2015 THL A29 Limited, a Tencent company, and Milo Yip. All rights reserved. +// +// Licensed under the MIT License (the "License"); you may not use this file except +// in compliance with the License. You may obtain a copy of the License at +// +// http://opensource.org/licenses/MIT +// +// 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 RAPIDJSON_IEEE754_ +#define RAPIDJSON_IEEE754_ + +#include "../rapidjson.h" + +RAPIDJSON_NAMESPACE_BEGIN +namespace internal { + +class Double { +public: + Double() {} + Double(double d) : d_(d) {} + Double(uint64_t u) : u_(u) {} + + double Value() const { return d_; } + uint64_t Uint64Value() const { return u_; } + + double NextPositiveDouble() const { + RAPIDJSON_ASSERT(!Sign()); + return Double(u_ + 1).Value(); + } + + bool Sign() const { return (u_ & kSignMask) != 0; } + uint64_t Significand() const { return u_ & kSignificandMask; } + int Exponent() const { return static_cast<int>(((u_ & kExponentMask) >> kSignificandSize) - kExponentBias); } + + bool IsNan() const { return (u_ & kExponentMask) == kExponentMask && Significand() != 0; } + bool IsInf() const { return (u_ & kExponentMask) == kExponentMask && Significand() == 0; } + bool IsNanOrInf() const { return (u_ & kExponentMask) == kExponentMask; } + bool IsNormal() const { return (u_ & kExponentMask) != 0 || Significand() == 0; } + bool IsZero() const { return (u_ & (kExponentMask | kSignificandMask)) == 0; } + + uint64_t IntegerSignificand() const { return IsNormal() ? Significand() | kHiddenBit : Significand(); } + int IntegerExponent() const { return (IsNormal() ? Exponent() : kDenormalExponent) - kSignificandSize; } + uint64_t ToBias() const { return (u_ & kSignMask) ? ~u_ + 1 : u_ | kSignMask; } + + static unsigned EffectiveSignificandSize(int order) { + if (order >= -1021) + return 53; + else if (order <= -1074) + return 0; + else + return static_cast<unsigned>(order) + 1074; + } + +private: + static const int kSignificandSize = 52; + static const int kExponentBias = 0x3FF; + static const int kDenormalExponent = 1 - kExponentBias; + static const uint64_t kSignMask = RAPIDJSON_UINT64_C2(0x80000000, 0x00000000); + static const uint64_t kExponentMask = RAPIDJSON_UINT64_C2(0x7FF00000, 0x00000000); + static const uint64_t kSignificandMask = RAPIDJSON_UINT64_C2(0x000FFFFF, 0xFFFFFFFF); + static const uint64_t kHiddenBit = RAPIDJSON_UINT64_C2(0x00100000, 0x00000000); + + union { + double d_; + uint64_t u_; + }; +}; + +} // namespace internal +RAPIDJSON_NAMESPACE_END + +#endif // RAPIDJSON_IEEE754_ diff --git a/include/rapidjson/internal/itoa.h b/include/rapidjson/internal/itoa.h new file mode 100644 index 00000000..01a4e7e7 --- /dev/null +++ b/include/rapidjson/internal/itoa.h @@ -0,0 +1,304 @@ +// Tencent is pleased to support the open source community by making RapidJSON available. +// +// Copyright (C) 2015 THL A29 Limited, a Tencent company, and Milo Yip. All rights reserved. +// +// Licensed under the MIT License (the "License"); you may not use this file except +// in compliance with the License. You may obtain a copy of the License at +// +// http://opensource.org/licenses/MIT +// +// 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 RAPIDJSON_ITOA_ +#define RAPIDJSON_ITOA_ + +#include "../rapidjson.h" + +RAPIDJSON_NAMESPACE_BEGIN +namespace internal { + +inline const char* GetDigitsLut() { + static const char cDigitsLut[200] = { + '0','0','0','1','0','2','0','3','0','4','0','5','0','6','0','7','0','8','0','9', + '1','0','1','1','1','2','1','3','1','4','1','5','1','6','1','7','1','8','1','9', + '2','0','2','1','2','2','2','3','2','4','2','5','2','6','2','7','2','8','2','9', + '3','0','3','1','3','2','3','3','3','4','3','5','3','6','3','7','3','8','3','9', + '4','0','4','1','4','2','4','3','4','4','4','5','4','6','4','7','4','8','4','9', + '5','0','5','1','5','2','5','3','5','4','5','5','5','6','5','7','5','8','5','9', + '6','0','6','1','6','2','6','3','6','4','6','5','6','6','6','7','6','8','6','9', + '7','0','7','1','7','2','7','3','7','4','7','5','7','6','7','7','7','8','7','9', + '8','0','8','1','8','2','8','3','8','4','8','5','8','6','8','7','8','8','8','9', + '9','0','9','1','9','2','9','3','9','4','9','5','9','6','9','7','9','8','9','9' + }; + return cDigitsLut; +} + +inline char* u32toa(uint32_t value, char* buffer) { + const char* cDigitsLut = GetDigitsLut(); + + if (value < 10000) { + const uint32_t d1 = (value / 100) << 1; + const uint32_t d2 = (value % 100) << 1; + + if (value >= 1000) + *buffer++ = cDigitsLut[d1]; + if (value >= 100) + *buffer++ = cDigitsLut[d1 + 1]; + if (value >= 10) + *buffer++ = cDigitsLut[d2]; + *buffer++ = cDigitsLut[d2 + 1]; + } + else if (value < 100000000) { + // value = bbbbcccc + const uint32_t b = value / 10000; + const uint32_t c = value % 10000; + + const uint32_t d1 = (b / 100) << 1; + const uint32_t d2 = (b % 100) << 1; + + const uint32_t d3 = (c / 100) << 1; + const uint32_t d4 = (c % 100) << 1; + + if (value >= 10000000) + *buffer++ = cDigitsLut[d1]; + if (value >= 1000000) + *buffer++ = cDigitsLut[d1 + 1]; + if (value >= 100000) + *buffer++ = cDigitsLut[d2]; + *buffer++ = cDigitsLut[d2 + 1]; + + *buffer++ = cDigitsLut[d3]; + *buffer++ = cDigitsLut[d3 + 1]; + *buffer++ = cDigitsLut[d4]; + *buffer++ = cDigitsLut[d4 + 1]; + } + else { + // value = aabbbbcccc in decimal + + const uint32_t a = value / 100000000; // 1 to 42 + value %= 100000000; + + if (a >= 10) { + const unsigned i = a << 1; + *buffer++ = cDigitsLut[i]; + *buffer++ = cDigitsLut[i + 1]; + } + else + *buffer++ = static_cast<char>('0' + static_cast<char>(a)); + + const uint32_t b = value / 10000; // 0 to 9999 + const uint32_t c = value % 10000; // 0 to 9999 + + const uint32_t d1 = (b / 100) << 1; + const uint32_t d2 = (b % 100) << 1; + + const uint32_t d3 = (c / 100) << 1; + const uint32_t d4 = (c % 100) << 1; + + *buffer++ = cDigitsLut[d1]; + *buffer++ = cDigitsLut[d1 + 1]; + *buffer++ = cDigitsLut[d2]; + *buffer++ = cDigitsLut[d2 + 1]; + *buffer++ = cDigitsLut[d3]; + *buffer++ = cDigitsLut[d3 + 1]; + *buffer++ = cDigitsLut[d4]; + *buffer++ = cDigitsLut[d4 + 1]; + } + return buffer; +} + +inline char* i32toa(int32_t value, char* buffer) { + uint32_t u = static_cast<uint32_t>(value); + if (value < 0) { + *buffer++ = '-'; + u = ~u + 1; + } + + return u32toa(u, buffer); +} + +inline char* u64toa(uint64_t value, char* buffer) { + const char* cDigitsLut = GetDigitsLut(); + const uint64_t kTen8 = 100000000; + const uint64_t kTen9 = kTen8 * 10; + const uint64_t kTen10 = kTen8 * 100; + const uint64_t kTen11 = kTen8 * 1000; + const uint64_t kTen12 = kTen8 * 10000; + const uint64_t kTen13 = kTen8 * 100000; + const uint64_t kTen14 = kTen8 * 1000000; + const uint64_t kTen15 = kTen8 * 10000000; + const uint64_t kTen16 = kTen8 * kTen8; + + if (value < kTen8) { + uint32_t v = static_cast<uint32_t>(value); + if (v < 10000) { + const uint32_t d1 = (v / 100) << 1; + const uint32_t d2 = (v % 100) << 1; + + if (v >= 1000) + *buffer++ = cDigitsLut[d1]; + if (v >= 100) + *buffer++ = cDigitsLut[d1 + 1]; + if (v >= 10) + *buffer++ = cDigitsLut[d2]; + *buffer++ = cDigitsLut[d2 + 1]; + } + else { + // value = bbbbcccc + const uint32_t b = v / 10000; + const uint32_t c = v % 10000; + + const uint32_t d1 = (b / 100) << 1; + const uint32_t d2 = (b % 100) << 1; + + const uint32_t d3 = (c / 100) << 1; + const uint32_t d4 = (c % 100) << 1; + + if (value >= 10000000) + *buffer++ = cDigitsLut[d1]; + if (value >= 1000000) + *buffer++ = cDigitsLut[d1 + 1]; + if (value >= 100000) + *buffer++ = cDigitsLut[d2]; + *buffer++ = cDigitsLut[d2 + 1]; + + *buffer++ = cDigitsLut[d3]; + *buffer++ = cDigitsLut[d3 + 1]; + *buffer++ = cDigitsLut[d4]; + *buffer++ = cDigitsLut[d4 + 1]; + } + } + else if (value < kTen16) { + const uint32_t v0 = static_cast<uint32_t>(value / kTen8); + const uint32_t v1 = static_cast<uint32_t>(value % kTen8); + + const uint32_t b0 = v0 / 10000; + const uint32_t c0 = v0 % 10000; + + const uint32_t d1 = (b0 / 100) << 1; + const uint32_t d2 = (b0 % 100) << 1; + + const uint32_t d3 = (c0 / 100) << 1; + const uint32_t d4 = (c0 % 100) << 1; + + const uint32_t b1 = v1 / 10000; + const uint32_t c1 = v1 % 10000; + + const uint32_t d5 = (b1 / 100) << 1; + const uint32_t d6 = (b1 % 100) << 1; + + const uint32_t d7 = (c1 / 100) << 1; + const uint32_t d8 = (c1 % 100) << 1; + + if (value >= kTen15) + *buffer++ = cDigitsLut[d1]; + if (value >= kTen14) + *buffer++ = cDigitsLut[d1 + 1]; + if (value >= kTen13) + *buffer++ = cDigitsLut[d2]; + if (value >= kTen12) + *buffer++ = cDigitsLut[d2 + 1]; + if (value >= kTen11) + *buffer++ = cDigitsLut[d3]; + if (value >= kTen10) + *buffer++ = cDigitsLut[d3 + 1]; + if (value >= kTen9) + *buffer++ = cDigitsLut[d4]; + if (value >= kTen8) + *buffer++ = cDigitsLut[d4 + 1]; + + *buffer++ = cDigitsLut[d5]; + *buffer++ = cDigitsLut[d5 + 1]; + *buffer++ = cDigitsLut[d6]; + *buffer++ = cDigitsLut[d6 + 1]; + *buffer++ = cDigitsLut[d7]; + *buffer++ = cDigitsLut[d7 + 1]; + *buffer++ = cDigitsLut[d8]; + *buffer++ = cDigitsLut[d8 + 1]; + } + else { + const uint32_t a = static_cast<uint32_t>(value / kTen16); // 1 to 1844 + value %= kTen16; + + if (a < 10) + *buffer++ = static_cast<char>('0' + static_cast<char>(a)); + else if (a < 100) { + const uint32_t i = a << 1; + *buffer++ = cDigitsLut[i]; + *buffer++ = cDigitsLut[i + 1]; + } + else if (a < 1000) { + *buffer++ = static_cast<char>('0' + static_cast<char>(a / 100)); + + const uint32_t i = (a % 100) << 1; + *buffer++ = cDigitsLut[i]; + *buffer++ = cDigitsLut[i + 1]; + } + else { + const uint32_t i = (a / 100) << 1; + const uint32_t j = (a % 100) << 1; + *buffer++ = cDigitsLut[i]; + *buffer++ = cDigitsLut[i + 1]; + *buffer++ = cDigitsLut[j]; + *buffer++ = cDigitsLut[j + 1]; + } + + const uint32_t v0 = static_cast<uint32_t>(value / kTen8); + const uint32_t v1 = static_cast<uint32_t>(value % kTen8); + + const uint32_t b0 = v0 / 10000; + const uint32_t c0 = v0 % 10000; + + const uint32_t d1 = (b0 / 100) << 1; + const uint32_t d2 = (b0 % 100) << 1; + + const uint32_t d3 = (c0 / 100) << 1; + const uint32_t d4 = (c0 % 100) << 1; + + const uint32_t b1 = v1 / 10000; + const uint32_t c1 = v1 % 10000; + + const uint32_t d5 = (b1 / 100) << 1; + const uint32_t d6 = (b1 % 100) << 1; + + const uint32_t d7 = (c1 / 100) << 1; + const uint32_t d8 = (c1 % 100) << 1; + + *buffer++ = cDigitsLut[d1]; + *buffer++ = cDigitsLut[d1 + 1]; + *buffer++ = cDigitsLut[d2]; + *buffer++ = cDigitsLut[d2 + 1]; + *buffer++ = cDigitsLut[d3]; + *buffer++ = cDigitsLut[d3 + 1]; + *buffer++ = cDigitsLut[d4]; + *buffer++ = cDigitsLut[d4 + 1]; + *buffer++ = cDigitsLut[d5]; + *buffer++ = cDigitsLut[d5 + 1]; + *buffer++ = cDigitsLut[d6]; + *buffer++ = cDigitsLut[d6 + 1]; + *buffer++ = cDigitsLut[d7]; + *buffer++ = cDigitsLut[d7 + 1]; + *buffer++ = cDigitsLut[d8]; + *buffer++ = cDigitsLut[d8 + 1]; + } + + return buffer; +} + +inline char* i64toa(int64_t value, char* buffer) { + uint64_t u = static_cast<uint64_t>(value); + if (value < 0) { + *buffer++ = '-'; + u = ~u + 1; + } + + return u64toa(u, buffer); +} + +} // namespace internal +RAPIDJSON_NAMESPACE_END + +#endif // RAPIDJSON_ITOA_ diff --git a/include/rapidjson/internal/meta.h b/include/rapidjson/internal/meta.h new file mode 100644 index 00000000..5a9aaa42 --- /dev/null +++ b/include/rapidjson/internal/meta.h @@ -0,0 +1,181 @@ +// Tencent is pleased to support the open source community by making RapidJSON available. +// +// Copyright (C) 2015 THL A29 Limited, a Tencent company, and Milo Yip. All rights reserved. +// +// Licensed under the MIT License (the "License"); you may not use this file except +// in compliance with the License. You may obtain a copy of the License at +// +// http://opensource.org/licenses/MIT +// +// 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 RAPIDJSON_INTERNAL_META_H_ +#define RAPIDJSON_INTERNAL_META_H_ + +#include "../rapidjson.h" + +#ifdef __GNUC__ +RAPIDJSON_DIAG_PUSH +RAPIDJSON_DIAG_OFF(effc++) +#endif +#if defined(_MSC_VER) +RAPIDJSON_DIAG_PUSH +RAPIDJSON_DIAG_OFF(6334) +#endif + +#if RAPIDJSON_HAS_CXX11_TYPETRAITS +#include <type_traits> +#endif + +//@cond RAPIDJSON_INTERNAL +RAPIDJSON_NAMESPACE_BEGIN +namespace internal { + +// Helper to wrap/convert arbitrary types to void, useful for arbitrary type matching +template <typename T> struct Void { typedef void Type; }; + +/////////////////////////////////////////////////////////////////////////////// +// BoolType, TrueType, FalseType +// +template <bool Cond> struct BoolType { + static const bool Value = Cond; + typedef BoolType Type; +}; +typedef BoolType<true> TrueType; +typedef BoolType<false> FalseType; + + +/////////////////////////////////////////////////////////////////////////////// +// SelectIf, BoolExpr, NotExpr, AndExpr, OrExpr +// + +template <bool C> struct SelectIfImpl { template <typename T1, typename T2> struct Apply { typedef T1 Type; }; }; +template <> struct SelectIfImpl<false> { template <typename T1, typename T2> struct Apply { typedef T2 Type; }; }; +template <bool C, typename T1, typename T2> struct SelectIfCond : SelectIfImpl<C>::template Apply<T1,T2> {}; +template <typename C, typename T1, typename T2> struct SelectIf : SelectIfCond<C::Value, T1, T2> {}; + +template <bool Cond1, bool Cond2> struct AndExprCond : FalseType {}; +template <> struct AndExprCond<true, true> : TrueType {}; +template <bool Cond1, bool Cond2> struct OrExprCond : TrueType {}; +template <> struct OrExprCond<false, false> : FalseType {}; + +template <typename C> struct BoolExpr : SelectIf<C,TrueType,FalseType>::Type {}; +template <typename C> struct NotExpr : SelectIf<C,FalseType,TrueType>::Type {}; +template <typename C1, typename C2> struct AndExpr : AndExprCond<C1::Value, C2::Value>::Type {}; +template <typename C1, typename C2> struct OrExpr : OrExprCond<C1::Value, C2::Value>::Type {}; + + +/////////////////////////////////////////////////////////////////////////////// +// AddConst, MaybeAddConst, RemoveConst +template <typename T> struct AddConst { typedef const T Type; }; +template <bool Constify, typename T> struct MaybeAddConst : SelectIfCond<Constify, const T, T> {}; +template <typename T> struct RemoveConst { typedef T Type; }; +template <typename T> struct RemoveConst<const T> { typedef T Type; }; + + +/////////////////////////////////////////////////////////////////////////////// +// IsSame, IsConst, IsMoreConst, IsPointer +// +template <typename T, typename U> struct IsSame : FalseType {}; +template <typename T> struct IsSame<T, T> : TrueType {}; + +template <typename T> struct IsConst : FalseType {}; +template <typename T> struct IsConst<const T> : TrueType {}; + +template <typename CT, typename T> +struct IsMoreConst + : AndExpr<IsSame<typename RemoveConst<CT>::Type, typename RemoveConst<T>::Type>, + BoolType<IsConst<CT>::Value >= IsConst<T>::Value> >::Type {}; + +template <typename T> struct IsPointer : FalseType {}; +template <typename T> struct IsPointer<T*> : TrueType {}; + +/////////////////////////////////////////////////////////////////////////////// +// IsBaseOf +// +#if RAPIDJSON_HAS_CXX11_TYPETRAITS + +template <typename B, typename D> struct IsBaseOf + : BoolType< ::std::is_base_of<B,D>::value> {}; + +#else // simplified version adopted from Boost + +template<typename B, typename D> struct IsBaseOfImpl { + RAPIDJSON_STATIC_ASSERT(sizeof(B) != 0); + RAPIDJSON_STATIC_ASSERT(sizeof(D) != 0); + + typedef char (&Yes)[1]; + typedef char (&No) [2]; + + template <typename T> + static Yes Check(const D*, T); + static No Check(const B*, int); + + struct Host { + operator const B*() const; + operator const D*(); + }; + + enum { Value = (sizeof(Check(Host(), 0)) == sizeof(Yes)) }; +}; + +template <typename B, typename D> struct IsBaseOf + : OrExpr<IsSame<B, D>, BoolExpr<IsBaseOfImpl<B, D> > >::Type {}; + +#endif // RAPIDJSON_HAS_CXX11_TYPETRAITS + + +////////////////////////////////////////////////////////////////////////// +// EnableIf / DisableIf +// +template <bool Condition, typename T = void> struct EnableIfCond { typedef T Type; }; +template <typename T> struct EnableIfCond<false, T> { /* empty */ }; + +template <bool Condition, typename T = void> struct DisableIfCond { typedef T Type; }; +template <typename T> struct DisableIfCond<true, T> { /* empty */ }; + +template <typename Condition, typename T = void> +struct EnableIf : EnableIfCond<Condition::Value, T> {}; + +template <typename Condition, typename T = void> +struct DisableIf : DisableIfCond<Condition::Value, T> {}; + +// SFINAE helpers +struct SfinaeTag {}; +template <typename T> struct RemoveSfinaeTag; +template <typename T> struct RemoveSfinaeTag<SfinaeTag&(*)(T)> { typedef T Type; }; + +#define RAPIDJSON_REMOVEFPTR_(type) \ + typename ::RAPIDJSON_NAMESPACE::internal::RemoveSfinaeTag \ + < ::RAPIDJSON_NAMESPACE::internal::SfinaeTag&(*) type>::Type + +#define RAPIDJSON_ENABLEIF(cond) \ + typename ::RAPIDJSON_NAMESPACE::internal::EnableIf \ + <RAPIDJSON_REMOVEFPTR_(cond)>::Type * = NULL + +#define RAPIDJSON_DISABLEIF(cond) \ + typename ::RAPIDJSON_NAMESPACE::internal::DisableIf \ + <RAPIDJSON_REMOVEFPTR_(cond)>::Type * = NULL + +#define RAPIDJSON_ENABLEIF_RETURN(cond,returntype) \ + typename ::RAPIDJSON_NAMESPACE::internal::EnableIf \ + <RAPIDJSON_REMOVEFPTR_(cond), \ + RAPIDJSON_REMOVEFPTR_(returntype)>::Type + +#define RAPIDJSON_DISABLEIF_RETURN(cond,returntype) \ + typename ::RAPIDJSON_NAMESPACE::internal::DisableIf \ + <RAPIDJSON_REMOVEFPTR_(cond), \ + RAPIDJSON_REMOVEFPTR_(returntype)>::Type + +} // namespace internal +RAPIDJSON_NAMESPACE_END +//@endcond + +#if defined(__GNUC__) || defined(_MSC_VER) +RAPIDJSON_DIAG_POP +#endif + +#endif // RAPIDJSON_INTERNAL_META_H_ diff --git a/include/rapidjson/internal/pow10.h b/include/rapidjson/internal/pow10.h new file mode 100644 index 00000000..02f475d7 --- /dev/null +++ b/include/rapidjson/internal/pow10.h @@ -0,0 +1,55 @@ +// Tencent is pleased to support the open source community by making RapidJSON available. +// +// Copyright (C) 2015 THL A29 Limited, a Tencent company, and Milo Yip. All rights reserved. +// +// Licensed under the MIT License (the "License"); you may not use this file except +// in compliance with the License. You may obtain a copy of the License at +// +// http://opensource.org/licenses/MIT +// +// 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 RAPIDJSON_POW10_ +#define RAPIDJSON_POW10_ + +#include "../rapidjson.h" + +RAPIDJSON_NAMESPACE_BEGIN +namespace internal { + +//! Computes integer powers of 10 in double (10.0^n). +/*! This function uses lookup table for fast and accurate results. + \param n non-negative exponent. Must <= 308. + \return 10.0^n +*/ +inline double Pow10(int n) { + static const double e[] = { // 1e-0...1e308: 309 * 8 bytes = 2472 bytes + 1e+0, + 1e+1, 1e+2, 1e+3, 1e+4, 1e+5, 1e+6, 1e+7, 1e+8, 1e+9, 1e+10, 1e+11, 1e+12, 1e+13, 1e+14, 1e+15, 1e+16, 1e+17, 1e+18, 1e+19, 1e+20, + 1e+21, 1e+22, 1e+23, 1e+24, 1e+25, 1e+26, 1e+27, 1e+28, 1e+29, 1e+30, 1e+31, 1e+32, 1e+33, 1e+34, 1e+35, 1e+36, 1e+37, 1e+38, 1e+39, 1e+40, + 1e+41, 1e+42, 1e+43, 1e+44, 1e+45, 1e+46, 1e+47, 1e+48, 1e+49, 1e+50, 1e+51, 1e+52, 1e+53, 1e+54, 1e+55, 1e+56, 1e+57, 1e+58, 1e+59, 1e+60, + 1e+61, 1e+62, 1e+63, 1e+64, 1e+65, 1e+66, 1e+67, 1e+68, 1e+69, 1e+70, 1e+71, 1e+72, 1e+73, 1e+74, 1e+75, 1e+76, 1e+77, 1e+78, 1e+79, 1e+80, + 1e+81, 1e+82, 1e+83, 1e+84, 1e+85, 1e+86, 1e+87, 1e+88, 1e+89, 1e+90, 1e+91, 1e+92, 1e+93, 1e+94, 1e+95, 1e+96, 1e+97, 1e+98, 1e+99, 1e+100, + 1e+101,1e+102,1e+103,1e+104,1e+105,1e+106,1e+107,1e+108,1e+109,1e+110,1e+111,1e+112,1e+113,1e+114,1e+115,1e+116,1e+117,1e+118,1e+119,1e+120, + 1e+121,1e+122,1e+123,1e+124,1e+125,1e+126,1e+127,1e+128,1e+129,1e+130,1e+131,1e+132,1e+133,1e+134,1e+135,1e+136,1e+137,1e+138,1e+139,1e+140, + 1e+141,1e+142,1e+143,1e+144,1e+145,1e+146,1e+147,1e+148,1e+149,1e+150,1e+151,1e+152,1e+153,1e+154,1e+155,1e+156,1e+157,1e+158,1e+159,1e+160, + 1e+161,1e+162,1e+163,1e+164,1e+165,1e+166,1e+167,1e+168,1e+169,1e+170,1e+171,1e+172,1e+173,1e+174,1e+175,1e+176,1e+177,1e+178,1e+179,1e+180, + 1e+181,1e+182,1e+183,1e+184,1e+185,1e+186,1e+187,1e+188,1e+189,1e+190,1e+191,1e+192,1e+193,1e+194,1e+195,1e+196,1e+197,1e+198,1e+199,1e+200, + 1e+201,1e+202,1e+203,1e+204,1e+205,1e+206,1e+207,1e+208,1e+209,1e+210,1e+211,1e+212,1e+213,1e+214,1e+215,1e+216,1e+217,1e+218,1e+219,1e+220, + 1e+221,1e+222,1e+223,1e+224,1e+225,1e+226,1e+227,1e+228,1e+229,1e+230,1e+231,1e+232,1e+233,1e+234,1e+235,1e+236,1e+237,1e+238,1e+239,1e+240, + 1e+241,1e+242,1e+243,1e+244,1e+245,1e+246,1e+247,1e+248,1e+249,1e+250,1e+251,1e+252,1e+253,1e+254,1e+255,1e+256,1e+257,1e+258,1e+259,1e+260, + 1e+261,1e+262,1e+263,1e+264,1e+265,1e+266,1e+267,1e+268,1e+269,1e+270,1e+271,1e+272,1e+273,1e+274,1e+275,1e+276,1e+277,1e+278,1e+279,1e+280, + 1e+281,1e+282,1e+283,1e+284,1e+285,1e+286,1e+287,1e+288,1e+289,1e+290,1e+291,1e+292,1e+293,1e+294,1e+295,1e+296,1e+297,1e+298,1e+299,1e+300, + 1e+301,1e+302,1e+303,1e+304,1e+305,1e+306,1e+307,1e+308 + }; + RAPIDJSON_ASSERT(n >= 0 && n <= 308); + return e[n]; +} + +} // namespace internal +RAPIDJSON_NAMESPACE_END + +#endif // RAPIDJSON_POW10_ diff --git a/include/rapidjson/internal/regex.h b/include/rapidjson/internal/regex.h new file mode 100644 index 00000000..422a5240 --- /dev/null +++ b/include/rapidjson/internal/regex.h @@ -0,0 +1,701 @@ +// Tencent is pleased to support the open source community by making RapidJSON available. +// +// Copyright (C) 2015 THL A29 Limited, a Tencent company, and Milo Yip. All rights reserved. +// +// Licensed under the MIT License (the "License"); you may not use this file except +// in compliance with the License. You may obtain a copy of the License at +// +// http://opensource.org/licenses/MIT +// +// 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 RAPIDJSON_INTERNAL_REGEX_H_ +#define RAPIDJSON_INTERNAL_REGEX_H_ + +#include "../allocators.h" +#include "../stream.h" +#include "stack.h" + +#ifdef __clang__ +RAPIDJSON_DIAG_PUSH +RAPIDJSON_DIAG_OFF(padded) +RAPIDJSON_DIAG_OFF(switch-enum) +RAPIDJSON_DIAG_OFF(implicit-fallthrough) +#endif + +#ifdef __GNUC__ +RAPIDJSON_DIAG_PUSH +RAPIDJSON_DIAG_OFF(effc++) +#endif + +#ifdef _MSC_VER +RAPIDJSON_DIAG_PUSH +RAPIDJSON_DIAG_OFF(4512) // assignment operator could not be generated +#endif + +#ifndef RAPIDJSON_REGEX_VERBOSE +#define RAPIDJSON_REGEX_VERBOSE 0 +#endif + +RAPIDJSON_NAMESPACE_BEGIN +namespace internal { + +/////////////////////////////////////////////////////////////////////////////// +// GenericRegex + +static const SizeType kRegexInvalidState = ~SizeType(0); //!< Represents an invalid index in GenericRegex::State::out, out1 +static const SizeType kRegexInvalidRange = ~SizeType(0); + +//! Regular expression engine with subset of ECMAscript grammar. +/*! + Supported regular expression syntax: + - \c ab Concatenation + - \c a|b Alternation + - \c a? Zero or one + - \c a* Zero or more + - \c a+ One or more + - \c a{3} Exactly 3 times + - \c a{3,} At least 3 times + - \c a{3,5} 3 to 5 times + - \c (ab) Grouping + - \c ^a At the beginning + - \c a$ At the end + - \c . Any character + - \c [abc] Character classes + - \c [a-c] Character class range + - \c [a-z0-9_] Character class combination + - \c [^abc] Negated character classes + - \c [^a-c] Negated character class range + - \c [\b] Backspace (U+0008) + - \c \\| \\\\ ... Escape characters + - \c \\f Form feed (U+000C) + - \c \\n Line feed (U+000A) + - \c \\r Carriage return (U+000D) + - \c \\t Tab (U+0009) + - \c \\v Vertical tab (U+000B) + + \note This is a Thompson NFA engine, implemented with reference to + Cox, Russ. "Regular Expression Matching Can Be Simple And Fast (but is slow in Java, Perl, PHP, Python, Ruby,...).", + https://swtch.com/~rsc/regexp/regexp1.html +*/ +template <typename Encoding, typename Allocator = CrtAllocator> +class GenericRegex { +public: + typedef typename Encoding::Ch Ch; + + GenericRegex(const Ch* source, Allocator* allocator = 0) : + states_(allocator, 256), ranges_(allocator, 256), root_(kRegexInvalidState), stateCount_(), rangeCount_(), + stateSet_(), state0_(allocator, 0), state1_(allocator, 0), anchorBegin_(), anchorEnd_() + { + GenericStringStream<Encoding> ss(source); + DecodedStream<GenericStringStream<Encoding> > ds(ss); + Parse(ds); + } + + ~GenericRegex() { + Allocator::Free(stateSet_); + } + + bool IsValid() const { + return root_ != kRegexInvalidState; + } + + template <typename InputStream> + bool Match(InputStream& is) const { + return SearchWithAnchoring(is, true, true); + } + + bool Match(const Ch* s) const { + GenericStringStream<Encoding> is(s); + return Match(is); + } + + template <typename InputStream> + bool Search(InputStream& is) const { + return SearchWithAnchoring(is, anchorBegin_, anchorEnd_); + } + + bool Search(const Ch* s) const { + GenericStringStream<Encoding> is(s); + return Search(is); + } + +private: + enum Operator { + kZeroOrOne, + kZeroOrMore, + kOneOrMore, + kConcatenation, + kAlternation, + kLeftParenthesis + }; + + static const unsigned kAnyCharacterClass = 0xFFFFFFFF; //!< For '.' + static const unsigned kRangeCharacterClass = 0xFFFFFFFE; + static const unsigned kRangeNegationFlag = 0x80000000; + + struct Range { + unsigned start; // + unsigned end; + SizeType next; + }; + + struct State { + SizeType out; //!< Equals to kInvalid for matching state + SizeType out1; //!< Equals to non-kInvalid for split + SizeType rangeStart; + unsigned codepoint; + }; + + struct Frag { + Frag(SizeType s, SizeType o, SizeType m) : start(s), out(o), minIndex(m) {} + SizeType start; + SizeType out; //!< link-list of all output states + SizeType minIndex; + }; + + template <typename SourceStream> + class DecodedStream { + public: + DecodedStream(SourceStream& ss) : ss_(ss), codepoint_() { Decode(); } + unsigned Peek() { return codepoint_; } + unsigned Take() { + unsigned c = codepoint_; + if (c) // No further decoding when '\0' + Decode(); + return c; + } + + private: + void Decode() { + if (!Encoding::Decode(ss_, &codepoint_)) + codepoint_ = 0; + } + + SourceStream& ss_; + unsigned codepoint_; + }; + + State& GetState(SizeType index) { + RAPIDJSON_ASSERT(index < stateCount_); + return states_.template Bottom<State>()[index]; + } + + const State& GetState(SizeType index) const { + RAPIDJSON_ASSERT(index < stateCount_); + return states_.template Bottom<State>()[index]; + } + + Range& GetRange(SizeType index) { + RAPIDJSON_ASSERT(index < rangeCount_); + return ranges_.template Bottom<Range>()[index]; + } + + const Range& GetRange(SizeType index) const { + RAPIDJSON_ASSERT(index < rangeCount_); + return ranges_.template Bottom<Range>()[index]; + } + + template <typename InputStream> + void Parse(DecodedStream<InputStream>& ds) { + Allocator allocator; + Stack<Allocator> operandStack(&allocator, 256); // Frag + Stack<Allocator> operatorStack(&allocator, 256); // Operator + Stack<Allocator> atomCountStack(&allocator, 256); // unsigned (Atom per parenthesis) + + *atomCountStack.template Push<unsigned>() = 0; + + unsigned codepoint; + while (ds.Peek() != 0) { + switch (codepoint = ds.Take()) { + case '^': + anchorBegin_ = true; + break; + + case '$': + anchorEnd_ = true; + break; + + case '|': + while (!operatorStack.Empty() && *operatorStack.template Top<Operator>() < kAlternation) + if (!Eval(operandStack, *operatorStack.template Pop<Operator>(1))) + return; + *operatorStack.template Push<Operator>() = kAlternation; + *atomCountStack.template Top<unsigned>() = 0; + break; + + case '(': + *operatorStack.template Push<Operator>() = kLeftParenthesis; + *atomCountStack.template Push<unsigned>() = 0; + break; + + case ')': + while (!operatorStack.Empty() && *operatorStack.template Top<Operator>() != kLeftParenthesis) + if (!Eval(operandStack, *operatorStack.template Pop<Operator>(1))) + return; + if (operatorStack.Empty()) + return; + operatorStack.template Pop<Operator>(1); + atomCountStack.template Pop<unsigned>(1); + ImplicitConcatenation(atomCountStack, operatorStack); + break; + + case '?': + if (!Eval(operandStack, kZeroOrOne)) + return; + break; + + case '*': + if (!Eval(operandStack, kZeroOrMore)) + return; + break; + + case '+': + if (!Eval(operandStack, kOneOrMore)) + return; + break; + + case '{': + { + unsigned n, m; + if (!ParseUnsigned(ds, &n)) + return; + + if (ds.Peek() == ',') { + ds.Take(); + if (ds.Peek() == '}') + m = kInfinityQuantifier; + else if (!ParseUnsigned(ds, &m) || m < n) + return; + } + else + m = n; + + if (!EvalQuantifier(operandStack, n, m) || ds.Peek() != '}') + return; + ds.Take(); + } + break; + + case '.': + PushOperand(operandStack, kAnyCharacterClass); + ImplicitConcatenation(atomCountStack, operatorStack); + break; + + case '[': + { + SizeType range; + if (!ParseRange(ds, &range)) + return; + SizeType s = NewState(kRegexInvalidState, kRegexInvalidState, kRangeCharacterClass); + GetState(s).rangeStart = range; + *operandStack.template Push<Frag>() = Frag(s, s, s); + } + ImplicitConcatenation(atomCountStack, operatorStack); + break; + + case '\\': // Escape character + if (!CharacterEscape(ds, &codepoint)) + return; // Unsupported escape character + // fall through to default + + default: // Pattern character + PushOperand(operandStack, codepoint); + ImplicitConcatenation(atomCountStack, operatorStack); + } + } + + while (!operatorStack.Empty()) + if (!Eval(operandStack, *operatorStack.template Pop<Operator>(1))) + return; + + // Link the operand to matching state. + if (operandStack.GetSize() == sizeof(Frag)) { + Frag* e = operandStack.template Pop<Frag>(1); + Patch(e->out, NewState(kRegexInvalidState, kRegexInvalidState, 0)); + root_ = e->start; + +#if RAPIDJSON_REGEX_VERBOSE + printf("root: %d\n", root_); + for (SizeType i = 0; i < stateCount_ ; i++) { + State& s = GetState(i); + printf("[%2d] out: %2d out1: %2d c: '%c'\n", i, s.out, s.out1, (char)s.codepoint); + } + printf("\n"); +#endif + } + + // Preallocate buffer for SearchWithAnchoring() + RAPIDJSON_ASSERT(stateSet_ == 0); + if (stateCount_ > 0) { + stateSet_ = static_cast<unsigned*>(states_.GetAllocator().Malloc(GetStateSetSize())); + state0_.template Reserve<SizeType>(stateCount_); + state1_.template Reserve<SizeType>(stateCount_); + } + } + + SizeType NewState(SizeType out, SizeType out1, unsigned codepoint) { + State* s = states_.template Push<State>(); + s->out = out; + s->out1 = out1; + s->codepoint = codepoint; + s->rangeStart = kRegexInvalidRange; + return stateCount_++; + } + + void PushOperand(Stack<Allocator>& operandStack, unsigned codepoint) { + SizeType s = NewState(kRegexInvalidState, kRegexInvalidState, codepoint); + *operandStack.template Push<Frag>() = Frag(s, s, s); + } + + void ImplicitConcatenation(Stack<Allocator>& atomCountStack, Stack<Allocator>& operatorStack) { + if (*atomCountStack.template Top<unsigned>()) + *operatorStack.template Push<Operator>() = kConcatenation; + (*atomCountStack.template Top<unsigned>())++; + } + + SizeType Append(SizeType l1, SizeType l2) { + SizeType old = l1; + while (GetState(l1).out != kRegexInvalidState) + l1 = GetState(l1).out; + GetState(l1).out = l2; + return old; + } + + void Patch(SizeType l, SizeType s) { + for (SizeType next; l != kRegexInvalidState; l = next) { + next = GetState(l).out; + GetState(l).out = s; + } + } + + bool Eval(Stack<Allocator>& operandStack, Operator op) { + switch (op) { + case kConcatenation: + RAPIDJSON_ASSERT(operandStack.GetSize() >= sizeof(Frag) * 2); + { + Frag e2 = *operandStack.template Pop<Frag>(1); + Frag e1 = *operandStack.template Pop<Frag>(1); + Patch(e1.out, e2.start); + *operandStack.template Push<Frag>() = Frag(e1.start, e2.out, Min(e1.minIndex, e2.minIndex)); + } + return true; + + case kAlternation: + if (operandStack.GetSize() >= sizeof(Frag) * 2) { + Frag e2 = *operandStack.template Pop<Frag>(1); + Frag e1 = *operandStack.template Pop<Frag>(1); + SizeType s = NewState(e1.start, e2.start, 0); + *operandStack.template Push<Frag>() = Frag(s, Append(e1.out, e2.out), Min(e1.minIndex, e2.minIndex)); + return true; + } + return false; + + case kZeroOrOne: + if (operandStack.GetSize() >= sizeof(Frag)) { + Frag e = *operandStack.template Pop<Frag>(1); + SizeType s = NewState(kRegexInvalidState, e.start, 0); + *operandStack.template Push<Frag>() = Frag(s, Append(e.out, s), e.minIndex); + return true; + } + return false; + + case kZeroOrMore: + if (operandStack.GetSize() >= sizeof(Frag)) { + Frag e = *operandStack.template Pop<Frag>(1); + SizeType s = NewState(kRegexInvalidState, e.start, 0); + Patch(e.out, s); + *operandStack.template Push<Frag>() = Frag(s, s, e.minIndex); + return true; + } + return false; + + default: + RAPIDJSON_ASSERT(op == kOneOrMore); + if (operandStack.GetSize() >= sizeof(Frag)) { + Frag e = *operandStack.template Pop<Frag>(1); + SizeType s = NewState(kRegexInvalidState, e.start, 0); + Patch(e.out, s); + *operandStack.template Push<Frag>() = Frag(e.start, s, e.minIndex); + return true; + } + return false; + } + } + + bool EvalQuantifier(Stack<Allocator>& operandStack, unsigned n, unsigned m) { + RAPIDJSON_ASSERT(n <= m); + RAPIDJSON_ASSERT(operandStack.GetSize() >= sizeof(Frag)); + + if (n == 0) { + if (m == 0) // a{0} not support + return false; + else if (m == kInfinityQuantifier) + Eval(operandStack, kZeroOrMore); // a{0,} -> a* + else { + Eval(operandStack, kZeroOrOne); // a{0,5} -> a? + for (unsigned i = 0; i < m - 1; i++) + CloneTopOperand(operandStack); // a{0,5} -> a? a? a? a? a? + for (unsigned i = 0; i < m - 1; i++) + Eval(operandStack, kConcatenation); // a{0,5} -> a?a?a?a?a? + } + return true; + } + + for (unsigned i = 0; i < n - 1; i++) // a{3} -> a a a + CloneTopOperand(operandStack); + + if (m == kInfinityQuantifier) + Eval(operandStack, kOneOrMore); // a{3,} -> a a a+ + else if (m > n) { + CloneTopOperand(operandStack); // a{3,5} -> a a a a + Eval(operandStack, kZeroOrOne); // a{3,5} -> a a a a? + for (unsigned i = n; i < m - 1; i++) + CloneTopOperand(operandStack); // a{3,5} -> a a a a? a? + for (unsigned i = n; i < m; i++) + Eval(operandStack, kConcatenation); // a{3,5} -> a a aa?a? + } + + for (unsigned i = 0; i < n - 1; i++) + Eval(operandStack, kConcatenation); // a{3} -> aaa, a{3,} -> aaa+, a{3.5} -> aaaa?a? + + return true; + } + + static SizeType Min(SizeType a, SizeType b) { return a < b ? a : b; } + + void CloneTopOperand(Stack<Allocator>& operandStack) { + const Frag src = *operandStack.template Top<Frag>(); // Copy constructor to prevent invalidation + SizeType count = stateCount_ - src.minIndex; // Assumes top operand contains states in [src->minIndex, stateCount_) + State* s = states_.template Push<State>(count); + memcpy(s, &GetState(src.minIndex), count * sizeof(State)); + for (SizeType j = 0; j < count; j++) { + if (s[j].out != kRegexInvalidState) + s[j].out += count; + if (s[j].out1 != kRegexInvalidState) + s[j].out1 += count; + } + *operandStack.template Push<Frag>() = Frag(src.start + count, src.out + count, src.minIndex + count); + stateCount_ += count; + } + + template <typename InputStream> + bool ParseUnsigned(DecodedStream<InputStream>& ds, unsigned* u) { + unsigned r = 0; + if (ds.Peek() < '0' || ds.Peek() > '9') + return false; + while (ds.Peek() >= '0' && ds.Peek() <= '9') { + if (r >= 429496729 && ds.Peek() > '5') // 2^32 - 1 = 4294967295 + return false; // overflow + r = r * 10 + (ds.Take() - '0'); + } + *u = r; + return true; + } + + template <typename InputStream> + bool ParseRange(DecodedStream<InputStream>& ds, SizeType* range) { + bool isBegin = true; + bool negate = false; + int step = 0; + SizeType start = kRegexInvalidRange; + SizeType current = kRegexInvalidRange; + unsigned codepoint; + while ((codepoint = ds.Take()) != 0) { + if (isBegin) { + isBegin = false; + if (codepoint == '^') { + negate = true; + continue; + } + } + + switch (codepoint) { + case ']': + if (start == kRegexInvalidRange) + return false; // Error: nothing inside [] + if (step == 2) { // Add trailing '-' + SizeType r = NewRange('-'); + RAPIDJSON_ASSERT(current != kRegexInvalidRange); + GetRange(current).next = r; + } + if (negate) + GetRange(start).start |= kRangeNegationFlag; + *range = start; + return true; + + case '\\': + if (ds.Peek() == 'b') { + ds.Take(); + codepoint = 0x0008; // Escape backspace character + } + else if (!CharacterEscape(ds, &codepoint)) + return false; + // fall through to default + + default: + switch (step) { + case 1: + if (codepoint == '-') { + step++; + break; + } + // fall through to step 0 for other characters + + case 0: + { + SizeType r = NewRange(codepoint); + if (current != kRegexInvalidRange) + GetRange(current).next = r; + if (start == kRegexInvalidRange) + start = r; + current = r; + } + step = 1; + break; + + default: + RAPIDJSON_ASSERT(step == 2); + GetRange(current).end = codepoint; + step = 0; + } + } + } + return false; + } + + SizeType NewRange(unsigned codepoint) { + Range* r = ranges_.template Push<Range>(); + r->start = r->end = codepoint; + r->next = kRegexInvalidRange; + return rangeCount_++; + } + + template <typename InputStream> + bool CharacterEscape(DecodedStream<InputStream>& ds, unsigned* escapedCodepoint) { + unsigned codepoint; + switch (codepoint = ds.Take()) { + case '^': + case '$': + case '|': + case '(': + case ')': + case '?': + case '*': + case '+': + case '.': + case '[': + case ']': + case '{': + case '}': + case '\\': + *escapedCodepoint = codepoint; return true; + case 'f': *escapedCodepoint = 0x000C; return true; + case 'n': *escapedCodepoint = 0x000A; return true; + case 'r': *escapedCodepoint = 0x000D; return true; + case 't': *escapedCodepoint = 0x0009; return true; + case 'v': *escapedCodepoint = 0x000B; return true; + default: + return false; // Unsupported escape character + } + } + + template <typename InputStream> + bool SearchWithAnchoring(InputStream& is, bool anchorBegin, bool anchorEnd) const { + RAPIDJSON_ASSERT(IsValid()); + DecodedStream<InputStream> ds(is); + + state0_.Clear(); + Stack<Allocator> *current = &state0_, *next = &state1_; + const size_t stateSetSize = GetStateSetSize(); + std::memset(stateSet_, 0, stateSetSize); + + bool matched = AddState(*current, root_); + unsigned codepoint; + while (!current->Empty() && (codepoint = ds.Take()) != 0) { + std::memset(stateSet_, 0, stateSetSize); + next->Clear(); + matched = false; + for (const SizeType* s = current->template Bottom<SizeType>(); s != current->template End<SizeType>(); ++s) { + const State& sr = GetState(*s); + if (sr.codepoint == codepoint || + sr.codepoint == kAnyCharacterClass || + (sr.codepoint == kRangeCharacterClass && MatchRange(sr.rangeStart, codepoint))) + { + matched = AddState(*next, sr.out) || matched; + if (!anchorEnd && matched) + return true; + } + if (!anchorBegin) + AddState(*next, root_); + } + internal::Swap(current, next); + } + + return matched; + } + + size_t GetStateSetSize() const { + return (stateCount_ + 31) / 32 * 4; + } + + // Return whether the added states is a match state + bool AddState(Stack<Allocator>& l, SizeType index) const { + RAPIDJSON_ASSERT(index != kRegexInvalidState); + + const State& s = GetState(index); + if (s.out1 != kRegexInvalidState) { // Split + bool matched = AddState(l, s.out); + return AddState(l, s.out1) || matched; + } + else if (!(stateSet_[index >> 5] & (1 << (index & 31)))) { + stateSet_[index >> 5] |= (1 << (index & 31)); + *l.template PushUnsafe<SizeType>() = index; + } + return s.out == kRegexInvalidState; // by using PushUnsafe() above, we can ensure s is not validated due to reallocation. + } + + bool MatchRange(SizeType rangeIndex, unsigned codepoint) const { + bool yes = (GetRange(rangeIndex).start & kRangeNegationFlag) == 0; + while (rangeIndex != kRegexInvalidRange) { + const Range& r = GetRange(rangeIndex); + if (codepoint >= (r.start & ~kRangeNegationFlag) && codepoint <= r.end) + return yes; + rangeIndex = r.next; + } + return !yes; + } + + Stack<Allocator> states_; + Stack<Allocator> ranges_; + SizeType root_; + SizeType stateCount_; + SizeType rangeCount_; + + static const unsigned kInfinityQuantifier = ~0u; + + // For SearchWithAnchoring() + uint32_t* stateSet_; // allocated by states_.GetAllocator() + mutable Stack<Allocator> state0_; + mutable Stack<Allocator> state1_; + bool anchorBegin_; + bool anchorEnd_; +}; + +typedef GenericRegex<UTF8<> > Regex; + +} // namespace internal +RAPIDJSON_NAMESPACE_END + +#ifdef __clang__ +RAPIDJSON_DIAG_POP +#endif + +#ifdef _MSC_VER +RAPIDJSON_DIAG_POP +#endif + +#endif // RAPIDJSON_INTERNAL_REGEX_H_ diff --git a/include/rapidjson/internal/stack.h b/include/rapidjson/internal/stack.h new file mode 100644 index 00000000..022c9aab --- /dev/null +++ b/include/rapidjson/internal/stack.h @@ -0,0 +1,230 @@ +// Tencent is pleased to support the open source community by making RapidJSON available. +// +// Copyright (C) 2015 THL A29 Limited, a Tencent company, and Milo Yip. All rights reserved. +// +// Licensed under the MIT License (the "License"); you may not use this file except +// in compliance with the License. You may obtain a copy of the License at +// +// http://opensource.org/licenses/MIT +// +// 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 RAPIDJSON_INTERNAL_STACK_H_ +#define RAPIDJSON_INTERNAL_STACK_H_ + +#include "../allocators.h" +#include "swap.h" + +#if defined(__clang__) +RAPIDJSON_DIAG_PUSH +RAPIDJSON_DIAG_OFF(c++98-compat) +#endif + +RAPIDJSON_NAMESPACE_BEGIN +namespace internal { + +/////////////////////////////////////////////////////////////////////////////// +// Stack + +//! A type-unsafe stack for storing different types of data. +/*! \tparam Allocator Allocator for allocating stack memory. +*/ +template <typename Allocator> +class Stack { +public: + // Optimization note: Do not allocate memory for stack_ in constructor. + // Do it lazily when first Push() -> Expand() -> Resize(). + Stack(Allocator* allocator, size_t stackCapacity) : allocator_(allocator), ownAllocator_(0), stack_(0), stackTop_(0), stackEnd_(0), initialCapacity_(stackCapacity) { + } + +#if RAPIDJSON_HAS_CXX11_RVALUE_REFS + Stack(Stack&& rhs) + : allocator_(rhs.allocator_), + ownAllocator_(rhs.ownAllocator_), + stack_(rhs.stack_), + stackTop_(rhs.stackTop_), + stackEnd_(rhs.stackEnd_), + initialCapacity_(rhs.initialCapacity_) + { + rhs.allocator_ = 0; + rhs.ownAllocator_ = 0; + rhs.stack_ = 0; + rhs.stackTop_ = 0; + rhs.stackEnd_ = 0; + rhs.initialCapacity_ = 0; + } +#endif + + ~Stack() { + Destroy(); + } + +#if RAPIDJSON_HAS_CXX11_RVALUE_REFS + Stack& operator=(Stack&& rhs) { + if (&rhs != this) + { + Destroy(); + + allocator_ = rhs.allocator_; + ownAllocator_ = rhs.ownAllocator_; + stack_ = rhs.stack_; + stackTop_ = rhs.stackTop_; + stackEnd_ = rhs.stackEnd_; + initialCapacity_ = rhs.initialCapacity_; + + rhs.allocator_ = 0; + rhs.ownAllocator_ = 0; + rhs.stack_ = 0; + rhs.stackTop_ = 0; + rhs.stackEnd_ = 0; + rhs.initialCapacity_ = 0; + } + return *this; + } +#endif + + void Swap(Stack& rhs) RAPIDJSON_NOEXCEPT { + internal::Swap(allocator_, rhs.allocator_); + internal::Swap(ownAllocator_, rhs.ownAllocator_); + internal::Swap(stack_, rhs.stack_); + internal::Swap(stackTop_, rhs.stackTop_); + internal::Swap(stackEnd_, rhs.stackEnd_); + internal::Swap(initialCapacity_, rhs.initialCapacity_); + } + + void Clear() { stackTop_ = stack_; } + + void ShrinkToFit() { + if (Empty()) { + // If the stack is empty, completely deallocate the memory. + Allocator::Free(stack_); + stack_ = 0; + stackTop_ = 0; + stackEnd_ = 0; + } + else + Resize(GetSize()); + } + + // Optimization note: try to minimize the size of this function for force inline. + // Expansion is run very infrequently, so it is moved to another (probably non-inline) function. + template<typename T> + RAPIDJSON_FORCEINLINE void Reserve(size_t count = 1) { + // Expand the stack if needed + if (RAPIDJSON_UNLIKELY(stackTop_ + sizeof(T) * count > stackEnd_)) + Expand<T>(count); + } + + template<typename T> + RAPIDJSON_FORCEINLINE T* Push(size_t count = 1) { + Reserve<T>(count); + return PushUnsafe<T>(count); + } + + template<typename T> + RAPIDJSON_FORCEINLINE T* PushUnsafe(size_t count = 1) { + RAPIDJSON_ASSERT(stackTop_ + sizeof(T) * count <= stackEnd_); + T* ret = reinterpret_cast<T*>(stackTop_); + stackTop_ += sizeof(T) * count; + return ret; + } + + template<typename T> + T* Pop(size_t count) { + RAPIDJSON_ASSERT(GetSize() >= count * sizeof(T)); + stackTop_ -= count * sizeof(T); + return reinterpret_cast<T*>(stackTop_); + } + + template<typename T> + T* Top() { + RAPIDJSON_ASSERT(GetSize() >= sizeof(T)); + return reinterpret_cast<T*>(stackTop_ - sizeof(T)); + } + + template<typename T> + const T* Top() const { + RAPIDJSON_ASSERT(GetSize() >= sizeof(T)); + return reinterpret_cast<T*>(stackTop_ - sizeof(T)); + } + + template<typename T> + T* End() { return reinterpret_cast<T*>(stackTop_); } + + template<typename T> + const T* End() const { return reinterpret_cast<T*>(stackTop_); } + + template<typename T> + T* Bottom() { return reinterpret_cast<T*>(stack_); } + + template<typename T> + const T* Bottom() const { return reinterpret_cast<T*>(stack_); } + + bool HasAllocator() const { + return allocator_ != 0; + } + + Allocator& GetAllocator() { + RAPIDJSON_ASSERT(allocator_); + return *allocator_; + } + + bool Empty() const { return stackTop_ == stack_; } + size_t GetSize() const { return static_cast<size_t>(stackTop_ - stack_); } + size_t GetCapacity() const { return static_cast<size_t>(stackEnd_ - stack_); } + +private: + template<typename T> + void Expand(size_t count) { + // Only expand the capacity if the current stack exists. Otherwise just create a stack with initial capacity. + size_t newCapacity; + if (stack_ == 0) { + if (!allocator_) + ownAllocator_ = allocator_ = RAPIDJSON_NEW(Allocator()); + newCapacity = initialCapacity_; + } else { + newCapacity = GetCapacity(); + newCapacity += (newCapacity + 1) / 2; + } + size_t newSize = GetSize() + sizeof(T) * count; + if (newCapacity < newSize) + newCapacity = newSize; + + Resize(newCapacity); + } + + void Resize(size_t newCapacity) { + const size_t size = GetSize(); // Backup the current size + stack_ = static_cast<char*>(allocator_->Realloc(stack_, GetCapacity(), newCapacity)); + stackTop_ = stack_ + size; + stackEnd_ = stack_ + newCapacity; + } + + void Destroy() { + Allocator::Free(stack_); + RAPIDJSON_DELETE(ownAllocator_); // Only delete if it is owned by the stack + } + + // Prohibit copy constructor & assignment operator. + Stack(const Stack&); + Stack& operator=(const Stack&); + + Allocator* allocator_; + Allocator* ownAllocator_; + char *stack_; + char *stackTop_; + char *stackEnd_; + size_t initialCapacity_; +}; + +} // namespace internal +RAPIDJSON_NAMESPACE_END + +#if defined(__clang__) +RAPIDJSON_DIAG_POP +#endif + +#endif // RAPIDJSON_STACK_H_ diff --git a/include/rapidjson/internal/strfunc.h b/include/rapidjson/internal/strfunc.h new file mode 100644 index 00000000..2edfae52 --- /dev/null +++ b/include/rapidjson/internal/strfunc.h @@ -0,0 +1,55 @@ +// Tencent is pleased to support the open source community by making RapidJSON available. +// +// Copyright (C) 2015 THL A29 Limited, a Tencent company, and Milo Yip. All rights reserved. +// +// Licensed under the MIT License (the "License"); you may not use this file except +// in compliance with the License. You may obtain a copy of the License at +// +// http://opensource.org/licenses/MIT +// +// 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 RAPIDJSON_INTERNAL_STRFUNC_H_ +#define RAPIDJSON_INTERNAL_STRFUNC_H_ + +#include "../stream.h" + +RAPIDJSON_NAMESPACE_BEGIN +namespace internal { + +//! Custom strlen() which works on different character types. +/*! \tparam Ch Character type (e.g. char, wchar_t, short) + \param s Null-terminated input string. + \return Number of characters in the string. + \note This has the same semantics as strlen(), the return value is not number of Unicode codepoints. +*/ +template <typename Ch> +inline SizeType StrLen(const Ch* s) { + const Ch* p = s; + while (*p) ++p; + return SizeType(p - s); +} + +//! Returns number of code points in a encoded string. +template<typename Encoding> +bool CountStringCodePoint(const typename Encoding::Ch* s, SizeType length, SizeType* outCount) { + GenericStringStream<Encoding> is(s); + const typename Encoding::Ch* end = s + length; + SizeType count = 0; + while (is.src_ < end) { + unsigned codepoint; + if (!Encoding::Decode(is, &codepoint)) + return false; + count++; + } + *outCount = count; + return true; +} + +} // namespace internal +RAPIDJSON_NAMESPACE_END + +#endif // RAPIDJSON_INTERNAL_STRFUNC_H_ diff --git a/include/rapidjson/internal/strtod.h b/include/rapidjson/internal/strtod.h new file mode 100644 index 00000000..289c413b --- /dev/null +++ b/include/rapidjson/internal/strtod.h @@ -0,0 +1,269 @@ +// Tencent is pleased to support the open source community by making RapidJSON available. +// +// Copyright (C) 2015 THL A29 Limited, a Tencent company, and Milo Yip. All rights reserved. +// +// Licensed under the MIT License (the "License"); you may not use this file except +// in compliance with the License. You may obtain a copy of the License at +// +// http://opensource.org/licenses/MIT +// +// 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 RAPIDJSON_STRTOD_ +#define RAPIDJSON_STRTOD_ + +#include "ieee754.h" +#include "biginteger.h" +#include "diyfp.h" +#include "pow10.h" + +RAPIDJSON_NAMESPACE_BEGIN +namespace internal { + +inline double FastPath(double significand, int exp) { + if (exp < -308) + return 0.0; + else if (exp >= 0) + return significand * internal::Pow10(exp); + else + return significand / internal::Pow10(-exp); +} + +inline double StrtodNormalPrecision(double d, int p) { + if (p < -308) { + // Prevent expSum < -308, making Pow10(p) = 0 + d = FastPath(d, -308); + d = FastPath(d, p + 308); + } + else + d = FastPath(d, p); + return d; +} + +template <typename T> +inline T Min3(T a, T b, T c) { + T m = a; + if (m > b) m = b; + if (m > c) m = c; + return m; +} + +inline int CheckWithinHalfULP(double b, const BigInteger& d, int dExp) { + const Double db(b); + const uint64_t bInt = db.IntegerSignificand(); + const int bExp = db.IntegerExponent(); + const int hExp = bExp - 1; + + int dS_Exp2 = 0, dS_Exp5 = 0, bS_Exp2 = 0, bS_Exp5 = 0, hS_Exp2 = 0, hS_Exp5 = 0; + + // Adjust for decimal exponent + if (dExp >= 0) { + dS_Exp2 += dExp; + dS_Exp5 += dExp; + } + else { + bS_Exp2 -= dExp; + bS_Exp5 -= dExp; + hS_Exp2 -= dExp; + hS_Exp5 -= dExp; + } + + // Adjust for binary exponent + if (bExp >= 0) + bS_Exp2 += bExp; + else { + dS_Exp2 -= bExp; + hS_Exp2 -= bExp; + } + + // Adjust for half ulp exponent + if (hExp >= 0) + hS_Exp2 += hExp; + else { + dS_Exp2 -= hExp; + bS_Exp2 -= hExp; + } + + // Remove common power of two factor from all three scaled values + int common_Exp2 = Min3(dS_Exp2, bS_Exp2, hS_Exp2); + dS_Exp2 -= common_Exp2; + bS_Exp2 -= common_Exp2; + hS_Exp2 -= common_Exp2; + + BigInteger dS = d; + dS.MultiplyPow5(static_cast<unsigned>(dS_Exp5)) <<= static_cast<unsigned>(dS_Exp2); + + BigInteger bS(bInt); + bS.MultiplyPow5(static_cast<unsigned>(bS_Exp5)) <<= static_cast<unsigned>(bS_Exp2); + + BigInteger hS(1); + hS.MultiplyPow5(static_cast<unsigned>(hS_Exp5)) <<= static_cast<unsigned>(hS_Exp2); + + BigInteger delta(0); + dS.Difference(bS, &delta); + + return delta.Compare(hS); +} + +inline bool StrtodFast(double d, int p, double* result) { + // Use fast path for string-to-double conversion if possible + // see http://www.exploringbinary.com/fast-path-decimal-to-floating-point-conversion/ + if (p > 22 && p < 22 + 16) { + // Fast Path Cases In Disguise + d *= internal::Pow10(p - 22); + p = 22; + } + + if (p >= -22 && p <= 22 && d <= 9007199254740991.0) { // 2^53 - 1 + *result = FastPath(d, p); + return true; + } + else + return false; +} + +// Compute an approximation and see if it is within 1/2 ULP +inline bool StrtodDiyFp(const char* decimals, size_t length, size_t decimalPosition, int exp, double* result) { + uint64_t significand = 0; + size_t i = 0; // 2^64 - 1 = 18446744073709551615, 1844674407370955161 = 0x1999999999999999 + for (; i < length; i++) { + if (significand > RAPIDJSON_UINT64_C2(0x19999999, 0x99999999) || + (significand == RAPIDJSON_UINT64_C2(0x19999999, 0x99999999) && decimals[i] > '5')) + break; + significand = significand * 10u + static_cast<unsigned>(decimals[i] - '0'); + } + + if (i < length && decimals[i] >= '5') // Rounding + significand++; + + size_t remaining = length - i; + const unsigned kUlpShift = 3; + const unsigned kUlp = 1 << kUlpShift; + int64_t error = (remaining == 0) ? 0 : kUlp / 2; + + DiyFp v(significand, 0); + v = v.Normalize(); + error <<= -v.e; + + const int dExp = static_cast<int>(decimalPosition) - static_cast<int>(i) + exp; + + int actualExp; + DiyFp cachedPower = GetCachedPower10(dExp, &actualExp); + if (actualExp != dExp) { + static const DiyFp kPow10[] = { + DiyFp(RAPIDJSON_UINT64_C2(0xa0000000, 00000000), -60), // 10^1 + DiyFp(RAPIDJSON_UINT64_C2(0xc8000000, 00000000), -57), // 10^2 + DiyFp(RAPIDJSON_UINT64_C2(0xfa000000, 00000000), -54), // 10^3 + DiyFp(RAPIDJSON_UINT64_C2(0x9c400000, 00000000), -50), // 10^4 + DiyFp(RAPIDJSON_UINT64_C2(0xc3500000, 00000000), -47), // 10^5 + DiyFp(RAPIDJSON_UINT64_C2(0xf4240000, 00000000), -44), // 10^6 + DiyFp(RAPIDJSON_UINT64_C2(0x98968000, 00000000), -40) // 10^7 + }; + int adjustment = dExp - actualExp - 1; + RAPIDJSON_ASSERT(adjustment >= 0 && adjustment < 7); + v = v * kPow10[adjustment]; + if (length + static_cast<unsigned>(adjustment)> 19u) // has more digits than decimal digits in 64-bit + error += kUlp / 2; + } + + v = v * cachedPower; + + error += kUlp + (error == 0 ? 0 : 1); + + const int oldExp = v.e; + v = v.Normalize(); + error <<= oldExp - v.e; + + const unsigned effectiveSignificandSize = Double::EffectiveSignificandSize(64 + v.e); + unsigned precisionSize = 64 - effectiveSignificandSize; + if (precisionSize + kUlpShift >= 64) { + unsigned scaleExp = (precisionSize + kUlpShift) - 63; + v.f >>= scaleExp; + v.e += scaleExp; + error = (error >> scaleExp) + 1 + static_cast<int>(kUlp); + precisionSize -= scaleExp; + } + + DiyFp rounded(v.f >> precisionSize, v.e + static_cast<int>(precisionSize)); + const uint64_t precisionBits = (v.f & ((uint64_t(1) << precisionSize) - 1)) * kUlp; + const uint64_t halfWay = (uint64_t(1) << (precisionSize - 1)) * kUlp; + if (precisionBits >= halfWay + static_cast<unsigned>(error)) { + rounded.f++; + if (rounded.f & (DiyFp::kDpHiddenBit << 1)) { // rounding overflows mantissa (issue #340) + rounded.f >>= 1; + rounded.e++; + } + } + + *result = rounded.ToDouble(); + + return halfWay - static_cast<unsigned>(error) >= precisionBits || precisionBits >= halfWay + static_cast<unsigned>(error); +} + +inline double StrtodBigInteger(double approx, const char* decimals, size_t length, size_t decimalPosition, int exp) { + const BigInteger dInt(decimals, length); + const int dExp = static_cast<int>(decimalPosition) - static_cast<int>(length) + exp; + Double a(approx); + int cmp = CheckWithinHalfULP(a.Value(), dInt, dExp); + if (cmp < 0) + return a.Value(); // within half ULP + else if (cmp == 0) { + // Round towards even + if (a.Significand() & 1) + return a.NextPositiveDouble(); + else + return a.Value(); + } + else // adjustment + return a.NextPositiveDouble(); +} + +inline double StrtodFullPrecision(double d, int p, const char* decimals, size_t length, size_t decimalPosition, int exp) { + RAPIDJSON_ASSERT(d >= 0.0); + RAPIDJSON_ASSERT(length >= 1); + + double result; + if (StrtodFast(d, p, &result)) + return result; + + // Trim leading zeros + while (*decimals == '0' && length > 1) { + length--; + decimals++; + decimalPosition--; + } + + // Trim trailing zeros + while (decimals[length - 1] == '0' && length > 1) { + length--; + decimalPosition--; + exp++; + } + + // Trim right-most digits + const int kMaxDecimalDigit = 780; + if (static_cast<int>(length) > kMaxDecimalDigit) { + int delta = (static_cast<int>(length) - kMaxDecimalDigit); + exp += delta; + decimalPosition -= static_cast<unsigned>(delta); + length = kMaxDecimalDigit; + } + + // If too small, underflow to zero + if (int(length) + exp < -324) + return 0.0; + + if (StrtodDiyFp(decimals, length, decimalPosition, exp, &result)) + return result; + + // Use approximation from StrtodDiyFp and make adjustment with BigInteger comparison + return StrtodBigInteger(result, decimals, length, decimalPosition, exp); +} + +} // namespace internal +RAPIDJSON_NAMESPACE_END + +#endif // RAPIDJSON_STRTOD_ diff --git a/include/rapidjson/internal/swap.h b/include/rapidjson/internal/swap.h new file mode 100644 index 00000000..666e49f9 --- /dev/null +++ b/include/rapidjson/internal/swap.h @@ -0,0 +1,46 @@ +// Tencent is pleased to support the open source community by making RapidJSON available. +// +// Copyright (C) 2015 THL A29 Limited, a Tencent company, and Milo Yip. All rights reserved. +// +// Licensed under the MIT License (the "License"); you may not use this file except +// in compliance with the License. You may obtain a copy of the License at +// +// http://opensource.org/licenses/MIT +// +// 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 RAPIDJSON_INTERNAL_SWAP_H_ +#define RAPIDJSON_INTERNAL_SWAP_H_ + +#include "../rapidjson.h" + +#if defined(__clang__) +RAPIDJSON_DIAG_PUSH +RAPIDJSON_DIAG_OFF(c++98-compat) +#endif + +RAPIDJSON_NAMESPACE_BEGIN +namespace internal { + +//! Custom swap() to avoid dependency on C++ <algorithm> header +/*! \tparam T Type of the arguments to swap, should be instantiated with primitive C++ types only. + \note This has the same semantics as std::swap(). +*/ +template <typename T> +inline void Swap(T& a, T& b) RAPIDJSON_NOEXCEPT { + T tmp = a; + a = b; + b = tmp; +} + +} // namespace internal +RAPIDJSON_NAMESPACE_END + +#if defined(__clang__) +RAPIDJSON_DIAG_POP +#endif + +#endif // RAPIDJSON_INTERNAL_SWAP_H_ |