/src/abseil-cpp/absl/strings/numbers.h
Line | Count | Source (jump to first uncovered line) |
1 | | // Copyright 2017 The Abseil Authors. |
2 | | // |
3 | | // Licensed under the Apache License, Version 2.0 (the "License"); |
4 | | // you may not use this file except in compliance with the License. |
5 | | // You may obtain a copy of the License at |
6 | | // |
7 | | // https://www.apache.org/licenses/LICENSE-2.0 |
8 | | // |
9 | | // Unless required by applicable law or agreed to in writing, software |
10 | | // distributed under the License is distributed on an "AS IS" BASIS, |
11 | | // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
12 | | // See the License for the specific language governing permissions and |
13 | | // limitations under the License. |
14 | | // |
15 | | // ----------------------------------------------------------------------------- |
16 | | // File: numbers.h |
17 | | // ----------------------------------------------------------------------------- |
18 | | // |
19 | | // This package contains functions for converting strings to numbers. For |
20 | | // converting numbers to strings, use `StrCat()` or `StrAppend()` in str_cat.h, |
21 | | // which automatically detect and convert most number values appropriately. |
22 | | |
23 | | #ifndef ABSL_STRINGS_NUMBERS_H_ |
24 | | #define ABSL_STRINGS_NUMBERS_H_ |
25 | | |
26 | | #ifdef __SSSE3__ |
27 | | #include <tmmintrin.h> |
28 | | #endif |
29 | | |
30 | | #ifdef _MSC_VER |
31 | | #include <intrin.h> |
32 | | #endif |
33 | | |
34 | | #include <cstddef> |
35 | | #include <cstdint> |
36 | | #include <cstdlib> |
37 | | #include <cstring> |
38 | | #include <ctime> |
39 | | #include <limits> |
40 | | #include <string> |
41 | | #include <type_traits> |
42 | | |
43 | | #include "absl/base/attributes.h" |
44 | | #include "absl/base/config.h" |
45 | | #include "absl/base/internal/endian.h" |
46 | | #include "absl/base/macros.h" |
47 | | #include "absl/base/nullability.h" |
48 | | #include "absl/base/port.h" |
49 | | #include "absl/numeric/bits.h" |
50 | | #include "absl/numeric/int128.h" |
51 | | #include "absl/strings/string_view.h" |
52 | | |
53 | | namespace absl { |
54 | | ABSL_NAMESPACE_BEGIN |
55 | | |
56 | | // SimpleAtoi() |
57 | | // |
58 | | // Converts the given string (optionally followed or preceded by ASCII |
59 | | // whitespace) into an integer value, returning `true` if successful. The string |
60 | | // must reflect a base-10 integer whose value falls within the range of the |
61 | | // integer type (optionally preceded by a `+` or `-`). If any errors are |
62 | | // encountered, this function returns `false`, leaving `out` in an unspecified |
63 | | // state. |
64 | | template <typename int_type> |
65 | | [[nodiscard]] bool SimpleAtoi(absl::string_view str, |
66 | | int_type* absl_nonnull out); |
67 | | |
68 | | // SimpleAtof() |
69 | | // |
70 | | // Converts the given string (optionally followed or preceded by ASCII |
71 | | // whitespace) into a float, which may be rounded on overflow or underflow, |
72 | | // returning `true` if successful. |
73 | | // See https://en.cppreference.com/w/c/string/byte/strtof for details about the |
74 | | // allowed formats for `str`, except SimpleAtof() is locale-independent and will |
75 | | // always use the "C" locale. If any errors are encountered, this function |
76 | | // returns `false`, leaving `out` in an unspecified state. |
77 | | [[nodiscard]] bool SimpleAtof(absl::string_view str, float* absl_nonnull out); |
78 | | |
79 | | // SimpleAtod() |
80 | | // |
81 | | // Converts the given string (optionally followed or preceded by ASCII |
82 | | // whitespace) into a double, which may be rounded on overflow or underflow, |
83 | | // returning `true` if successful. |
84 | | // See https://en.cppreference.com/w/c/string/byte/strtof for details about the |
85 | | // allowed formats for `str`, except SimpleAtod is locale-independent and will |
86 | | // always use the "C" locale. If any errors are encountered, this function |
87 | | // returns `false`, leaving `out` in an unspecified state. |
88 | | [[nodiscard]] bool SimpleAtod(absl::string_view str, double* absl_nonnull out); |
89 | | |
90 | | // SimpleAtob() |
91 | | // |
92 | | // Converts the given string into a boolean, returning `true` if successful. |
93 | | // The following case-insensitive strings are interpreted as boolean `true`: |
94 | | // "true", "t", "yes", "y", "1". The following case-insensitive strings |
95 | | // are interpreted as boolean `false`: "false", "f", "no", "n", "0". If any |
96 | | // errors are encountered, this function returns `false`, leaving `out` in an |
97 | | // unspecified state. |
98 | | [[nodiscard]] bool SimpleAtob(absl::string_view str, bool* absl_nonnull out); |
99 | | |
100 | | // SimpleHexAtoi() |
101 | | // |
102 | | // Converts a hexadecimal string (optionally followed or preceded by ASCII |
103 | | // whitespace) to an integer, returning `true` if successful. Only valid base-16 |
104 | | // hexadecimal integers whose value falls within the range of the integer type |
105 | | // (optionally preceded by a `+` or `-`) can be converted. A valid hexadecimal |
106 | | // value may include both upper and lowercase character symbols, and may |
107 | | // optionally include a leading "0x" (or "0X") number prefix, which is ignored |
108 | | // by this function. If any errors are encountered, this function returns |
109 | | // `false`, leaving `out` in an unspecified state. |
110 | | template <typename int_type> |
111 | | [[nodiscard]] bool SimpleHexAtoi(absl::string_view str, |
112 | | int_type* absl_nonnull out); |
113 | | |
114 | | // Overloads of SimpleHexAtoi() for 128 bit integers. |
115 | | [[nodiscard]] inline bool SimpleHexAtoi(absl::string_view str, |
116 | | absl::int128* absl_nonnull out); |
117 | | [[nodiscard]] inline bool SimpleHexAtoi(absl::string_view str, |
118 | | absl::uint128* absl_nonnull out); |
119 | | |
120 | | ABSL_NAMESPACE_END |
121 | | } // namespace absl |
122 | | |
123 | | // End of public API. Implementation details follow. |
124 | | |
125 | | namespace absl { |
126 | | ABSL_NAMESPACE_BEGIN |
127 | | namespace numbers_internal { |
128 | | |
129 | | template <typename int_type> |
130 | 0 | constexpr bool is_signed() { |
131 | 0 | if constexpr (std::is_arithmetic<int_type>::value) { |
132 | 0 | // Use std::numeric_limits<T>::is_signed where it's defined to work. |
133 | 0 | return std::numeric_limits<int_type>::is_signed; |
134 | 0 | } |
135 | 0 | // TODO(jorg): This signed-ness check is used because it works correctly |
136 | 0 | // with enums, and it also serves to check that int_type is not a pointer. |
137 | 0 | // If one day something like std::is_signed<enum E> works, switch to it. |
138 | 0 | return static_cast<int_type>(1) - 2 < 0; |
139 | 0 | } Unexecuted instantiation: bool absl::numbers_internal::is_signed<long long>() Unexecuted instantiation: bool absl::numbers_internal::is_signed<unsigned long long>() Unexecuted instantiation: bool absl::numbers_internal::is_signed<unsigned char>() Unexecuted instantiation: bool absl::numbers_internal::is_signed<char>() Unexecuted instantiation: bool absl::numbers_internal::is_signed<signed char>() Unexecuted instantiation: bool absl::numbers_internal::is_signed<wchar_t>() Unexecuted instantiation: bool absl::numbers_internal::is_signed<unsigned short>() Unexecuted instantiation: bool absl::numbers_internal::is_signed<short>() |
140 | | |
141 | | // Digit conversion. |
142 | | ABSL_DLL extern const char kHexChar[17]; // 0123456789abcdef |
143 | | ABSL_DLL extern const char |
144 | | kHexTable[513]; // 000102030405060708090a0b0c0d0e0f1011... |
145 | | |
146 | | // Writes a two-character representation of 'i' to 'buf'. 'i' must be in the |
147 | | // range 0 <= i < 100, and buf must have space for two characters. Example: |
148 | | // char buf[2]; |
149 | | // PutTwoDigits(42, buf); |
150 | | // // buf[0] == '4' |
151 | | // // buf[1] == '2' |
152 | | void PutTwoDigits(uint32_t i, char* absl_nonnull buf); |
153 | | |
154 | | // safe_strto?() functions for implementing SimpleAtoi() |
155 | | |
156 | | bool safe_strto8_base(absl::string_view text, int8_t* absl_nonnull value, |
157 | | int base); |
158 | | bool safe_strto16_base(absl::string_view text, int16_t* absl_nonnull value, |
159 | | int base); |
160 | | bool safe_strto32_base(absl::string_view text, int32_t* absl_nonnull value, |
161 | | int base); |
162 | | bool safe_strto64_base(absl::string_view text, int64_t* absl_nonnull value, |
163 | | int base); |
164 | | bool safe_strto128_base(absl::string_view text, |
165 | | absl::int128* absl_nonnull value, int base); |
166 | | bool safe_strtou8_base(absl::string_view text, uint8_t* absl_nonnull value, |
167 | | int base); |
168 | | bool safe_strtou16_base(absl::string_view text, uint16_t* absl_nonnull value, |
169 | | int base); |
170 | | bool safe_strtou32_base(absl::string_view text, uint32_t* absl_nonnull value, |
171 | | int base); |
172 | | bool safe_strtou64_base(absl::string_view text, uint64_t* absl_nonnull value, |
173 | | int base); |
174 | | bool safe_strtou128_base(absl::string_view text, |
175 | | absl::uint128* absl_nonnull value, int base); |
176 | | |
177 | | static const int kFastToBufferSize = 32; |
178 | | static const int kSixDigitsToBufferSize = 16; |
179 | | |
180 | | // Helper function for fast formatting of floating-point values. |
181 | | // The result is the same as printf's "%g", a.k.a. "%.6g"; that is, six |
182 | | // significant digits are returned, trailing zeros are removed, and numbers |
183 | | // outside the range 0.0001-999999 are output using scientific notation |
184 | | // (1.23456e+06). This routine is heavily optimized. |
185 | | // Required buffer size is `kSixDigitsToBufferSize`. |
186 | | size_t SixDigitsToBuffer(double d, char* absl_nonnull buffer); |
187 | | |
188 | | // WARNING: These functions may write more characters than necessary, because |
189 | | // they are intended for speed. All functions take an output buffer |
190 | | // as an argument and return a pointer to the last byte they wrote, which is the |
191 | | // terminating '\0'. At most `kFastToBufferSize` bytes are written. |
192 | | char* absl_nonnull FastIntToBuffer(int32_t i, char* absl_nonnull buffer) |
193 | | ABSL_INTERNAL_NEED_MIN_SIZE(buffer, kFastToBufferSize); |
194 | | char* absl_nonnull FastIntToBuffer(uint32_t n, char* absl_nonnull out_str) |
195 | | ABSL_INTERNAL_NEED_MIN_SIZE(out_str, kFastToBufferSize); |
196 | | char* absl_nonnull FastIntToBuffer(int64_t i, char* absl_nonnull buffer) |
197 | | ABSL_INTERNAL_NEED_MIN_SIZE(buffer, kFastToBufferSize); |
198 | | char* absl_nonnull FastIntToBuffer(uint64_t i, char* absl_nonnull buffer) |
199 | | ABSL_INTERNAL_NEED_MIN_SIZE(buffer, kFastToBufferSize); |
200 | | |
201 | | // For enums and integer types that are not an exact match for the types above, |
202 | | // use templates to call the appropriate one of the four overloads above. |
203 | | template <typename int_type> |
204 | | char* absl_nonnull FastIntToBuffer(int_type i, char* absl_nonnull buffer) |
205 | 0 | ABSL_INTERNAL_NEED_MIN_SIZE(buffer, kFastToBufferSize) { |
206 | 0 | static_assert(sizeof(i) <= 64 / 8, |
207 | 0 | "FastIntToBuffer works only with 64-bit-or-less integers."); |
208 | | // These conditions are constexpr bools to suppress MSVC warning C4127. |
209 | 0 | constexpr bool kIsSigned = is_signed<int_type>(); |
210 | 0 | constexpr bool kUse64Bit = sizeof(i) > 32 / 8; |
211 | 0 | if (kIsSigned) { |
212 | 0 | if (kUse64Bit) { |
213 | 0 | return FastIntToBuffer(static_cast<int64_t>(i), buffer); |
214 | 0 | } else { |
215 | 0 | return FastIntToBuffer(static_cast<int32_t>(i), buffer); |
216 | 0 | } |
217 | 0 | } else { |
218 | 0 | if (kUse64Bit) { |
219 | 0 | return FastIntToBuffer(static_cast<uint64_t>(i), buffer); |
220 | 0 | } else { |
221 | 0 | return FastIntToBuffer(static_cast<uint32_t>(i), buffer); |
222 | 0 | } |
223 | 0 | } |
224 | 0 | } Unexecuted instantiation: char* absl::numbers_internal::FastIntToBuffer<long long>(long long, char*) Unexecuted instantiation: char* absl::numbers_internal::FastIntToBuffer<unsigned long long>(unsigned long long, char*) Unexecuted instantiation: char* absl::numbers_internal::FastIntToBuffer<unsigned char>(unsigned char, char*) Unexecuted instantiation: char* absl::numbers_internal::FastIntToBuffer<char>(char, char*) Unexecuted instantiation: char* absl::numbers_internal::FastIntToBuffer<signed char>(signed char, char*) Unexecuted instantiation: char* absl::numbers_internal::FastIntToBuffer<wchar_t>(wchar_t, char*) Unexecuted instantiation: char* absl::numbers_internal::FastIntToBuffer<unsigned short>(unsigned short, char*) Unexecuted instantiation: char* absl::numbers_internal::FastIntToBuffer<short>(short, char*) |
225 | | |
226 | | // Implementation of SimpleAtoi, generalized to support arbitrary base (used |
227 | | // with base different from 10 elsewhere in Abseil implementation). |
228 | | template <typename int_type> |
229 | | [[nodiscard]] bool safe_strtoi_base(absl::string_view s, |
230 | | int_type* absl_nonnull out, int base) { |
231 | | static_assert(sizeof(*out) == 1 || sizeof(*out) == 2 || sizeof(*out) == 4 || |
232 | | sizeof(*out) == 8, |
233 | | "SimpleAtoi works only with 8, 16, 32, or 64-bit integers."); |
234 | | static_assert(!std::is_floating_point<int_type>::value, |
235 | | "Use SimpleAtof or SimpleAtod instead."); |
236 | | bool parsed; |
237 | | // These conditions are constexpr bools to suppress MSVC warning C4127. |
238 | | constexpr bool kIsSigned = is_signed<int_type>(); |
239 | | constexpr int kIntTypeSize = sizeof(*out) * 8; |
240 | | if (kIsSigned) { |
241 | | if (kIntTypeSize == 64) { |
242 | | int64_t val; |
243 | | parsed = numbers_internal::safe_strto64_base(s, &val, base); |
244 | | *out = static_cast<int_type>(val); |
245 | | } else if (kIntTypeSize == 32) { |
246 | | int32_t val; |
247 | | parsed = numbers_internal::safe_strto32_base(s, &val, base); |
248 | | *out = static_cast<int_type>(val); |
249 | | } else if (kIntTypeSize == 16) { |
250 | | int16_t val; |
251 | | parsed = numbers_internal::safe_strto16_base(s, &val, base); |
252 | | *out = static_cast<int_type>(val); |
253 | | } else if (kIntTypeSize == 8) { |
254 | | int8_t val; |
255 | | parsed = numbers_internal::safe_strto8_base(s, &val, base); |
256 | | *out = static_cast<int_type>(val); |
257 | | } |
258 | | } else { |
259 | | if (kIntTypeSize == 64) { |
260 | | uint64_t val; |
261 | | parsed = numbers_internal::safe_strtou64_base(s, &val, base); |
262 | | *out = static_cast<int_type>(val); |
263 | | } else if (kIntTypeSize == 32) { |
264 | | uint32_t val; |
265 | | parsed = numbers_internal::safe_strtou32_base(s, &val, base); |
266 | | *out = static_cast<int_type>(val); |
267 | | } else if (kIntTypeSize == 16) { |
268 | | uint16_t val; |
269 | | parsed = numbers_internal::safe_strtou16_base(s, &val, base); |
270 | | *out = static_cast<int_type>(val); |
271 | | } else if (kIntTypeSize == 8) { |
272 | | uint8_t val; |
273 | | parsed = numbers_internal::safe_strtou8_base(s, &val, base); |
274 | | *out = static_cast<int_type>(val); |
275 | | } |
276 | | } |
277 | | return parsed; |
278 | | } |
279 | | |
280 | | // FastHexToBufferZeroPad16() |
281 | | // |
282 | | // Outputs `val` into `out` as if by `snprintf(out, 17, "%016x", val)` but |
283 | | // without the terminating null character. Thus `out` must be of length >= 16. |
284 | | // Returns the number of non-pad digits of the output (it can never be zero |
285 | | // since 0 has one digit). |
286 | 0 | inline size_t FastHexToBufferZeroPad16(uint64_t val, char* absl_nonnull out) { |
287 | 0 | #ifdef ABSL_INTERNAL_HAVE_SSSE3 |
288 | 0 | uint64_t be = absl::big_endian::FromHost64(val); |
289 | 0 | const auto kNibbleMask = _mm_set1_epi8(0xf); |
290 | 0 | const auto kHexDigits = _mm_setr_epi8('0', '1', '2', '3', '4', '5', '6', '7', |
291 | 0 | '8', '9', 'a', 'b', 'c', 'd', 'e', 'f'); |
292 | 0 | auto v = _mm_loadl_epi64(reinterpret_cast<__m128i*>(&be)); // load lo dword |
293 | 0 | auto v4 = _mm_srli_epi64(v, 4); // shift 4 right |
294 | 0 | auto il = _mm_unpacklo_epi8(v4, v); // interleave bytes |
295 | 0 | auto m = _mm_and_si128(il, kNibbleMask); // mask out nibbles |
296 | 0 | auto hexchars = _mm_shuffle_epi8(kHexDigits, m); // hex chars |
297 | 0 | _mm_storeu_si128(reinterpret_cast<__m128i*>(out), hexchars); |
298 | 0 | #else |
299 | 0 | for (int i = 0; i < 8; ++i) { |
300 | 0 | auto byte = (val >> (56 - 8 * i)) & 0xFF; |
301 | 0 | auto* hex = &absl::numbers_internal::kHexTable[byte * 2]; |
302 | 0 | std::memcpy(out + 2 * i, hex, 2); |
303 | 0 | } |
304 | 0 | #endif |
305 | 0 | // | 0x1 so that even 0 has 1 digit. |
306 | 0 | return 16 - static_cast<size_t>(countl_zero(val | 0x1) / 4); |
307 | 0 | } |
308 | | |
309 | | } // namespace numbers_internal |
310 | | |
311 | | template <typename int_type> |
312 | | [[nodiscard]] bool SimpleAtoi(absl::string_view str, |
313 | | int_type* absl_nonnull out) { |
314 | | return numbers_internal::safe_strtoi_base(str, out, 10); |
315 | | } |
316 | | |
317 | | [[nodiscard]] inline bool SimpleAtoi(absl::string_view str, |
318 | 0 | absl::int128* absl_nonnull out) { |
319 | 0 | return numbers_internal::safe_strto128_base(str, out, 10); |
320 | 0 | } |
321 | | |
322 | | [[nodiscard]] inline bool SimpleAtoi(absl::string_view str, |
323 | 0 | absl::uint128* absl_nonnull out) { |
324 | 0 | return numbers_internal::safe_strtou128_base(str, out, 10); |
325 | 0 | } |
326 | | |
327 | | template <typename int_type> |
328 | | [[nodiscard]] bool SimpleHexAtoi(absl::string_view str, |
329 | | int_type* absl_nonnull out) { |
330 | | return numbers_internal::safe_strtoi_base(str, out, 16); |
331 | | } |
332 | | |
333 | | [[nodiscard]] inline bool SimpleHexAtoi(absl::string_view str, |
334 | 0 | absl::int128* absl_nonnull out) { |
335 | 0 | return numbers_internal::safe_strto128_base(str, out, 16); |
336 | 0 | } |
337 | | |
338 | | [[nodiscard]] inline bool SimpleHexAtoi(absl::string_view str, |
339 | 0 | absl::uint128* absl_nonnull out) { |
340 | 0 | return numbers_internal::safe_strtou128_base(str, out, 16); |
341 | 0 | } |
342 | | |
343 | | ABSL_NAMESPACE_END |
344 | | } // namespace absl |
345 | | |
346 | | #endif // ABSL_STRINGS_NUMBERS_H_ |