detail/charconv/detail/parser.hpp
0.0% Lines (0/162)
0.0% Functions (0/4)
0.0% Branches (0/174)
detail/charconv/detail/parser.hpp
| Line | Hits | Source Code |
|---|---|---|
| 1 | // Copyright 2023 Matt Borland | |
| 2 | // Distributed under the Boost Software License, Version 1.0. | |
| 3 | // https://www.boost.org/LICENSE_1_0.txt | |
| 4 | ||
| 5 | #ifndef BOOST_JSON_DETAIL_CHARCONV_DETAIL_PARSER_HPP | |
| 6 | #define BOOST_JSON_DETAIL_CHARCONV_DETAIL_PARSER_HPP | |
| 7 | ||
| 8 | #include <boost/json/detail/charconv/detail/config.hpp> | |
| 9 | #include <boost/json/detail/charconv/detail/from_chars_result.hpp> | |
| 10 | #include <boost/json/detail/charconv/detail/from_chars_integer_impl.hpp> | |
| 11 | #include <boost/json/detail/charconv/detail/integer_search_trees.hpp> | |
| 12 | #include <boost/json/detail/charconv/limits.hpp> | |
| 13 | #include <boost/json/detail/charconv/chars_format.hpp> | |
| 14 | #include <system_error> | |
| 15 | #include <type_traits> | |
| 16 | #include <limits> | |
| 17 | #include <cerrno> | |
| 18 | #include <cstdint> | |
| 19 | #include <cstring> | |
| 20 | ||
| 21 | #if defined(__GNUC__) && __GNUC__ < 5 && !defined(__clang__) | |
| 22 | # pragma GCC diagnostic push | |
| 23 | # pragma GCC diagnostic ignored "-Wmissing-field-initializers" | |
| 24 | #endif | |
| 25 | ||
| 26 | namespace boost { namespace json { namespace detail { namespace charconv { namespace detail { | |
| 27 | ||
| 28 | ✗ | inline bool is_integer_char(char c) noexcept |
| 29 | { | |
| 30 | ✗ | return (c >= '0') && (c <= '9'); |
| 31 | } | |
| 32 | ||
| 33 | ✗ | inline bool is_hex_char(char c) noexcept |
| 34 | { | |
| 35 | ✗ | return is_integer_char(c) || (((c >= 'a') && (c <= 'f')) || ((c >= 'A') && (c <= 'F'))); |
| 36 | } | |
| 37 | ||
| 38 | ✗ | inline bool is_delimiter(char c, chars_format fmt) noexcept |
| 39 | { | |
| 40 | ✗ | if (fmt != chars_format::hex) |
| 41 | { | |
| 42 | ✗ | return !is_integer_char(c) && c != 'e' && c != 'E'; |
| 43 | } | |
| 44 | ||
| 45 | ✗ | return !is_hex_char(c) && c != 'p' && c != 'P'; |
| 46 | } | |
| 47 | ||
| 48 | template <typename Unsigned_Integer, typename Integer> | |
| 49 | ✗ | inline from_chars_result parser(const char* first, const char* last, bool& sign, Unsigned_Integer& significand, Integer& exponent, chars_format fmt = chars_format::general) noexcept |
| 50 | { | |
| 51 | ✗ | if (first > last) |
| 52 | { | |
| 53 | ✗ | return {first, std::errc::invalid_argument}; |
| 54 | } | |
| 55 | ||
| 56 | ✗ | auto next = first; |
| 57 | ✗ | bool all_zeros = true; |
| 58 | ||
| 59 | // First extract the sign | |
| 60 | ✗ | if (*next == '-') |
| 61 | { | |
| 62 | ✗ | sign = true; |
| 63 | ✗ | ++next; |
| 64 | } | |
| 65 | ✗ | else if (*next == '+') |
| 66 | { | |
| 67 | ✗ | return {next, std::errc::invalid_argument}; |
| 68 | } | |
| 69 | else | |
| 70 | { | |
| 71 | ✗ | sign = false; |
| 72 | } | |
| 73 | ||
| 74 | // Ignore leading zeros (e.g. 00005 or -002.3e+5) | |
| 75 | ✗ | while (*next == '0' && next != last) |
| 76 | { | |
| 77 | ✗ | ++next; |
| 78 | } | |
| 79 | ||
| 80 | // If the number is 0 we can abort now | |
| 81 | char exp_char; | |
| 82 | char capital_exp_char; | |
| 83 | ✗ | if (fmt != chars_format::hex) |
| 84 | { | |
| 85 | ✗ | exp_char = 'e'; |
| 86 | ✗ | capital_exp_char = 'E'; |
| 87 | } | |
| 88 | else | |
| 89 | { | |
| 90 | ✗ | exp_char = 'p'; |
| 91 | ✗ | capital_exp_char = 'P'; |
| 92 | } | |
| 93 | ||
| 94 | ✗ | if (next == last || *next == exp_char || *next == -capital_exp_char) |
| 95 | { | |
| 96 | ✗ | significand = 0; |
| 97 | ✗ | exponent = 0; |
| 98 | ✗ | return {next, std::errc()}; |
| 99 | } | |
| 100 | ||
| 101 | // Next we get the significand | |
| 102 | ✗ | constexpr std::size_t significand_buffer_size = limits<Unsigned_Integer>::max_chars10 - 1; // Base 10 or 16 |
| 103 | ✗ | char significand_buffer[significand_buffer_size] {}; |
| 104 | ✗ | std::size_t i = 0; |
| 105 | ✗ | std::size_t dot_position = 0; |
| 106 | ✗ | Integer extra_zeros = 0; |
| 107 | ✗ | Integer leading_zero_powers = 0; |
| 108 | ✗ | const auto char_validation_func = (fmt != charconv::chars_format::hex) ? is_integer_char : is_hex_char; |
| 109 | ✗ | const int base = (fmt != charconv::chars_format::hex) ? 10 : 16; |
| 110 | ||
| 111 | ✗ | while (char_validation_func(*next) && next != last && i < significand_buffer_size) |
| 112 | { | |
| 113 | ✗ | all_zeros = false; |
| 114 | ✗ | significand_buffer[i] = *next; |
| 115 | ✗ | ++next; |
| 116 | ✗ | ++i; |
| 117 | } | |
| 118 | ||
| 119 | ✗ | bool fractional = false; |
| 120 | ✗ | if (next == last) |
| 121 | { | |
| 122 | // if fmt is chars_format::scientific the e is required | |
| 123 | ✗ | if (fmt == chars_format::scientific) |
| 124 | { | |
| 125 | ✗ | return {first, std::errc::invalid_argument}; |
| 126 | } | |
| 127 | ||
| 128 | ✗ | exponent = 0; |
| 129 | ✗ | std::size_t offset = i; |
| 130 | ||
| 131 | ✗ | from_chars_result r = from_chars(significand_buffer, significand_buffer + offset, significand, base); |
| 132 | ✗ | switch (r.ec) |
| 133 | { | |
| 134 | ✗ | case std::errc::invalid_argument: |
| 135 | ✗ | return {first, std::errc::invalid_argument}; |
| 136 | ✗ | case std::errc::result_out_of_range: |
| 137 | ✗ | return {next, std::errc::result_out_of_range}; |
| 138 | ✗ | default: |
| 139 | ✗ | return {next, std::errc()}; |
| 140 | } | |
| 141 | } | |
| 142 | ✗ | else if (*next == '.') |
| 143 | { | |
| 144 | ✗ | ++next; |
| 145 | ✗ | fractional = true; |
| 146 | ✗ | dot_position = i; |
| 147 | ||
| 148 | // Process the fractional part if we have it | |
| 149 | // | |
| 150 | // if fmt is chars_format::scientific the e is required | |
| 151 | // if fmt is chars_format::fixed and not scientific the e is disallowed | |
| 152 | // if fmt is chars_format::general (which is scientific and fixed) the e is optional | |
| 153 | ||
| 154 | // If we have the value 0.00001 we can continue to chop zeros and adjust the exponent | |
| 155 | // so that we get the useful parts of the fraction | |
| 156 | ✗ | if (all_zeros) |
| 157 | { | |
| 158 | ✗ | while (*next == '0' && next != last) |
| 159 | { | |
| 160 | ✗ | ++next; |
| 161 | ✗ | --leading_zero_powers; |
| 162 | } | |
| 163 | ||
| 164 | ✗ | if (next == last) |
| 165 | { | |
| 166 | ✗ | return {last, std::errc()}; |
| 167 | } | |
| 168 | } | |
| 169 | ||
| 170 | ✗ | while (char_validation_func(*next) && next != last && i < significand_buffer_size) |
| 171 | { | |
| 172 | ✗ | significand_buffer[i] = *next; |
| 173 | ✗ | ++next; |
| 174 | ✗ | ++i; |
| 175 | } | |
| 176 | } | |
| 177 | ||
| 178 | ✗ | if (i == significand_buffer_size) |
| 179 | { | |
| 180 | // We can not process any more significant figures into the significand so skip to the end | |
| 181 | // or the exponent part and capture the additional orders of magnitude for the exponent | |
| 182 | ✗ | bool found_dot = false; |
| 183 | ✗ | while ((char_validation_func(*next) || *next == '.') && next != last) |
| 184 | { | |
| 185 | ✗ | ++next; |
| 186 | ✗ | if (!fractional && !found_dot) |
| 187 | { | |
| 188 | ✗ | ++extra_zeros; |
| 189 | } | |
| 190 | ✗ | if (*next == '.') |
| 191 | { | |
| 192 | ✗ | found_dot = true; |
| 193 | } | |
| 194 | } | |
| 195 | } | |
| 196 | ||
| 197 | ✗ | if (next == last || is_delimiter(*next, fmt)) |
| 198 | { | |
| 199 | ✗ | if (fmt == chars_format::scientific) |
| 200 | { | |
| 201 | ✗ | return {first, std::errc::invalid_argument}; |
| 202 | } | |
| 203 | ✗ | if (dot_position != 0 || fractional) |
| 204 | { | |
| 205 | ✗ | exponent = static_cast<Integer>(dot_position) - i + extra_zeros + leading_zero_powers; |
| 206 | } | |
| 207 | else | |
| 208 | { | |
| 209 | ✗ | exponent = extra_zeros + leading_zero_powers; |
| 210 | } | |
| 211 | ✗ | std::size_t offset = i; |
| 212 | ||
| 213 | ✗ | from_chars_result r = from_chars(significand_buffer, significand_buffer + offset, significand, base); |
| 214 | ✗ | switch (r.ec) |
| 215 | { | |
| 216 | ✗ | case std::errc::invalid_argument: |
| 217 | ✗ | return {first, std::errc::invalid_argument}; |
| 218 | ✗ | case std::errc::result_out_of_range: |
| 219 | ✗ | return {next, std::errc::result_out_of_range}; |
| 220 | ✗ | default: |
| 221 | ✗ | return {next, std::errc()}; |
| 222 | } | |
| 223 | } | |
| 224 | ✗ | else if (*next == exp_char || *next == capital_exp_char) |
| 225 | { | |
| 226 | // Would be a number without a significand e.g. e+03 | |
| 227 | ✗ | if (next == first) |
| 228 | { | |
| 229 | ✗ | return {next, std::errc::invalid_argument}; |
| 230 | } | |
| 231 | ||
| 232 | ✗ | ++next; |
| 233 | ✗ | if (fmt == chars_format::fixed) |
| 234 | { | |
| 235 | ✗ | return {first, std::errc::invalid_argument}; |
| 236 | } | |
| 237 | ||
| 238 | ✗ | exponent = i - 1; |
| 239 | ✗ | std::size_t offset = i; |
| 240 | ✗ | bool round = false; |
| 241 | // If more digits are present than representable in the significand of the target type | |
| 242 | // we set the maximum | |
| 243 | ✗ | if (offset > significand_buffer_size) |
| 244 | { | |
| 245 | ✗ | offset = significand_buffer_size - 1; |
| 246 | ✗ | i = significand_buffer_size; |
| 247 | ✗ | if (significand_buffer[offset] == '5' || |
| 248 | ✗ | significand_buffer[offset] == '6' || |
| 249 | ✗ | significand_buffer[offset] == '7' || |
| 250 | ✗ | significand_buffer[offset] == '8' || |
| 251 | ✗ | significand_buffer[offset] == '9') |
| 252 | { | |
| 253 | ✗ | round = true; |
| 254 | } | |
| 255 | } | |
| 256 | ||
| 257 | // If the significand is 0 from chars will return std::errc::invalid_argument because there is nothing in the buffer, | |
| 258 | // but it is a valid value. We need to continue parsing to get the correct value of ptr even | |
| 259 | // though we know we could bail now. | |
| 260 | // | |
| 261 | // See GitHub issue #29: https://github.com/cppalliance/charconv/issues/29 | |
| 262 | ✗ | if (offset != 0) |
| 263 | { | |
| 264 | ✗ | from_chars_result r = from_chars(significand_buffer, significand_buffer + offset, significand, base); |
| 265 | ✗ | switch (r.ec) |
| 266 | { | |
| 267 | ✗ | case std::errc::invalid_argument: |
| 268 | ✗ | return {first, std::errc::invalid_argument}; |
| 269 | ✗ | case std::errc::result_out_of_range: |
| 270 | ✗ | return {next, std::errc::result_out_of_range}; |
| 271 | ✗ | default: |
| 272 | ✗ | break; |
| 273 | } | |
| 274 | ||
| 275 | ✗ | if (round) |
| 276 | { | |
| 277 | ✗ | significand += 1; |
| 278 | } | |
| 279 | } | |
| 280 | ✗ | } |
| 281 | else | |
| 282 | { | |
| 283 | ✗ | return {first, std::errc::invalid_argument}; |
| 284 | } | |
| 285 | ||
| 286 | // Finally we get the exponent | |
| 287 | ✗ | constexpr std::size_t exponent_buffer_size = 6; // Float128 min exp is −16382 |
| 288 | ✗ | char exponent_buffer[exponent_buffer_size] {}; |
| 289 | ✗ | Integer significand_digits = i; |
| 290 | ✗ | i = 0; |
| 291 | ||
| 292 | // Get the sign first | |
| 293 | ✗ | if (*next == '-') |
| 294 | { | |
| 295 | ✗ | exponent_buffer[i] = *next; |
| 296 | ✗ | ++next; |
| 297 | ✗ | ++i; |
| 298 | } | |
| 299 | ✗ | else if (*next == '+') |
| 300 | { | |
| 301 | ✗ | ++next; |
| 302 | } | |
| 303 | ||
| 304 | // Next strip any leading zeros | |
| 305 | ✗ | while (*next == '0') |
| 306 | { | |
| 307 | ✗ | ++next; |
| 308 | } | |
| 309 | ||
| 310 | // Process the significant values | |
| 311 | ✗ | while (is_integer_char(*next) && next != last && i < exponent_buffer_size) |
| 312 | { | |
| 313 | ✗ | exponent_buffer[i] = *next; |
| 314 | ✗ | ++next; |
| 315 | ✗ | ++i; |
| 316 | } | |
| 317 | ||
| 318 | // If the exponent can't fit in the buffer the number is not representable | |
| 319 | ✗ | if (next != last && i == exponent_buffer_size) |
| 320 | { | |
| 321 | ✗ | return {next, std::errc::result_out_of_range}; |
| 322 | } | |
| 323 | ||
| 324 | // If the exponent was e+00 or e-00 | |
| 325 | ✗ | if (i == 0 || (i == 1 && exponent_buffer[0] == '-')) |
| 326 | { | |
| 327 | ✗ | if (fractional) |
| 328 | { | |
| 329 | ✗ | exponent = static_cast<Integer>(dot_position) - significand_digits; |
| 330 | } | |
| 331 | else | |
| 332 | { | |
| 333 | ✗ | exponent = extra_zeros; |
| 334 | } | |
| 335 | ||
| 336 | ✗ | return {next, std::errc()}; |
| 337 | } | |
| 338 | ||
| 339 | ✗ | const auto r = from_chars(exponent_buffer, exponent_buffer + i, exponent); |
| 340 | ||
| 341 | ✗ | exponent += leading_zero_powers; |
| 342 | ||
| 343 | ✗ | switch (r.ec) |
| 344 | { | |
| 345 | ✗ | case std::errc::invalid_argument: |
| 346 | ✗ | return {first, std::errc::invalid_argument}; |
| 347 | ✗ | case std::errc::result_out_of_range: |
| 348 | ✗ | return {next, std::errc::result_out_of_range}; |
| 349 | ✗ | default: |
| 350 | ✗ | if (fractional) |
| 351 | { | |
| 352 | // Need to take the offset from 1.xxx because compute_floatXXX assumes the significand is an integer | |
| 353 | // so the exponent is off by the number of digits in the significand - 1 | |
| 354 | ✗ | if (fmt == chars_format::hex) |
| 355 | { | |
| 356 | // In hex the number of digits parsed is possibly less than the number of digits in base10 | |
| 357 | ✗ | exponent -= num_digits(significand) - dot_position; |
| 358 | } | |
| 359 | else | |
| 360 | { | |
| 361 | ✗ | exponent -= significand_digits - dot_position; |
| 362 | } | |
| 363 | } | |
| 364 | else | |
| 365 | { | |
| 366 | ✗ | exponent += extra_zeros; |
| 367 | } | |
| 368 | ✗ | return {next, std::errc()}; |
| 369 | } | |
| 370 | } | |
| 371 | ||
| 372 | }}}}} // Namespaces | |
| 373 | ||
| 374 | #if defined(__GNUC__) && __GNUC__ < 5 && !defined(__clang__) | |
| 375 | # pragma GCC diagnostic pop | |
| 376 | #endif | |
| 377 | ||
| 378 | #endif // BOOST_JSON_DETAIL_CHARCONV_DETAIL_PARSER_HPP | |
| 379 |