Warning: This file is not a C or C++ file. It does not have highlighting.
| 1 | //===-- Format string parser for printf -------------------------*- C++ -*-===// |
|---|---|
| 2 | // |
| 3 | // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. |
| 4 | // See https://llvm.org/LICENSE.txt for license information. |
| 5 | // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception |
| 6 | // |
| 7 | //===----------------------------------------------------------------------===// |
| 8 | |
| 9 | #ifndef LLVM_LIBC_SRC_STDIO_PRINTF_CORE_PARSER_H |
| 10 | #define LLVM_LIBC_SRC_STDIO_PRINTF_CORE_PARSER_H |
| 11 | |
| 12 | #include "include/llvm-libc-macros/stdfix-macros.h" |
| 13 | #include "src/__support/CPP/algorithm.h" // max |
| 14 | #include "src/__support/CPP/limits.h" |
| 15 | #include "src/__support/CPP/optional.h" |
| 16 | #include "src/__support/CPP/type_traits.h" |
| 17 | #include "src/__support/macros/config.h" |
| 18 | #include "src/__support/str_to_integer.h" |
| 19 | #include "src/stdio/printf_core/core_structs.h" |
| 20 | #include "src/stdio/printf_core/printf_config.h" |
| 21 | |
| 22 | #include <stddef.h> |
| 23 | |
| 24 | #ifdef LIBC_INTERNAL_PRINTF_HAS_FIXED_POINT |
| 25 | #include "src/__support/fixed_point/fx_rep.h" |
| 26 | #endif // LIBC_INTERNAL_PRINTF_HAS_FIXED_POINT |
| 27 | #ifndef LIBC_COPT_PRINTF_DISABLE_STRERROR |
| 28 | #include "src/__support/libc_errno.h" |
| 29 | #endif // LIBC_COPT_PRINTF_DISABLE_STRERROR |
| 30 | |
| 31 | namespace LIBC_NAMESPACE_DECL { |
| 32 | namespace printf_core { |
| 33 | |
| 34 | template <typename T> struct int_type_of { |
| 35 | using type = T; |
| 36 | }; |
| 37 | template <> struct int_type_of<double> { |
| 38 | using type = fputil::FPBits<double>::StorageType; |
| 39 | }; |
| 40 | template <> struct int_type_of<long double> { |
| 41 | using type = fputil::FPBits<long double>::StorageType; |
| 42 | }; |
| 43 | |
| 44 | #ifdef LIBC_INTERNAL_PRINTF_HAS_FIXED_POINT |
| 45 | template <typename T> |
| 46 | struct int_type_of<cpp::enable_if<cpp::is_fixed_point_v<T>, T>> { |
| 47 | using type = typename fixed_point::FXRep<T>::StorageType; |
| 48 | }; |
| 49 | #endif // LIBC_INTERNAL_PRINTF_HAS_FIXED_POINT |
| 50 | |
| 51 | template <typename T> using int_type_of_v = typename int_type_of<T>::type; |
| 52 | |
| 53 | #ifndef LIBC_COPT_PRINTF_DISABLE_INDEX_MODE |
| 54 | #define WRITE_ARG_VAL_SIMPLEST(dst, arg_type, index) \ |
| 55 | { \ |
| 56 | auto temp = get_arg_value<arg_type>(index); \ |
| 57 | if (!temp.has_value()) { \ |
| 58 | section.has_conv = false; \ |
| 59 | } else { \ |
| 60 | dst = static_cast<decltype(dst)>( \ |
| 61 | cpp::bit_cast<int_type_of_v<arg_type>>(temp.value())); \ |
| 62 | } \ |
| 63 | } |
| 64 | #else |
| 65 | #define WRITE_ARG_VAL_SIMPLEST(dst, arg_type, _) \ |
| 66 | dst = cpp::bit_cast<int_type_of_v<arg_type>>(get_next_arg_value<arg_type>()) |
| 67 | #endif // LIBC_COPT_PRINTF_DISABLE_INDEX_MODE |
| 68 | |
| 69 | template <typename ArgProvider> class Parser { |
| 70 | const char *__restrict str; |
| 71 | |
| 72 | size_t cur_pos = 0; |
| 73 | ArgProvider args_cur; |
| 74 | |
| 75 | #ifndef LIBC_COPT_PRINTF_DISABLE_INDEX_MODE |
| 76 | // args_start stores the start of the va_args, which is allows getting the |
| 77 | // value of arguments that have already been passed. args_index is tracked so |
| 78 | // that we know which argument args_cur is on. |
| 79 | ArgProvider args_start; |
| 80 | size_t args_index = 1; |
| 81 | |
| 82 | // Defined in printf_config.h |
| 83 | static constexpr size_t DESC_ARR_LEN = LIBC_COPT_PRINTF_INDEX_ARR_LEN; |
| 84 | |
| 85 | // desc_arr stores the sizes of the variables in the ArgProvider. This is used |
| 86 | // in index mode to reduce repeated string parsing. The sizes are stored as |
| 87 | // TypeDesc objects, which store the size as well as minimal type information. |
| 88 | // This is necessary because some systems separate the floating point and |
| 89 | // integer values in va_args. |
| 90 | TypeDesc desc_arr[DESC_ARR_LEN] = {type_desc_from_type<void>()}; |
| 91 | |
| 92 | // TODO: Look into object stores for optimization. |
| 93 | |
| 94 | #endif // LIBC_COPT_PRINTF_DISABLE_INDEX_MODE |
| 95 | |
| 96 | public: |
| 97 | #ifndef LIBC_COPT_PRINTF_DISABLE_INDEX_MODE |
| 98 | LIBC_INLINE Parser(const char *__restrict new_str, ArgProvider &args) |
| 99 | : str(new_str), args_cur(args), args_start(args) {} |
| 100 | #else |
| 101 | LIBC_INLINE Parser(const char *__restrict new_str, ArgProvider &args) |
| 102 | : str(new_str), args_cur(args) {} |
| 103 | #endif // LIBC_COPT_PRINTF_DISABLE_INDEX_MODE |
| 104 | |
| 105 | // get_next_section will parse the format string until it has a fully |
| 106 | // specified format section. This can either be a raw format section with no |
| 107 | // conversion, or a format section with a conversion that has all of its |
| 108 | // variables stored in the format section. |
| 109 | LIBC_INLINE FormatSection get_next_section() { |
| 110 | FormatSection section; |
| 111 | size_t starting_pos = cur_pos; |
| 112 | if (str[cur_pos] == '%') { |
| 113 | // format section |
| 114 | section.has_conv = true; |
| 115 | |
| 116 | ++cur_pos; |
| 117 | [[maybe_unused]] size_t conv_index = 0; |
| 118 | |
| 119 | #ifndef LIBC_COPT_PRINTF_DISABLE_INDEX_MODE |
| 120 | conv_index = parse_index(&cur_pos); |
| 121 | #endif // LIBC_COPT_PRINTF_DISABLE_INDEX_MODE |
| 122 | |
| 123 | section.flags = parse_flags(&cur_pos); |
| 124 | |
| 125 | // handle width |
| 126 | section.min_width = 0; |
| 127 | if (str[cur_pos] == '*') { |
| 128 | ++cur_pos; |
| 129 | |
| 130 | WRITE_ARG_VAL_SIMPLEST(section.min_width, int, parse_index(&cur_pos)); |
| 131 | } else if (internal::isdigit(str[cur_pos])) { |
| 132 | auto result = internal::strtointeger<int>(str + cur_pos, 10); |
| 133 | section.min_width = result.value; |
| 134 | cur_pos = cur_pos + static_cast<size_t>(result.parsed_len); |
| 135 | } |
| 136 | if (section.min_width < 0) { |
| 137 | section.min_width = |
| 138 | (section.min_width == INT_MIN) ? INT_MAX : -section.min_width; |
| 139 | section.flags = static_cast<FormatFlags>(section.flags | |
| 140 | FormatFlags::LEFT_JUSTIFIED); |
| 141 | } |
| 142 | |
| 143 | // handle precision |
| 144 | section.precision = -1; // negative precisions are ignored. |
| 145 | if (str[cur_pos] == '.') { |
| 146 | ++cur_pos; |
| 147 | section.precision = 0; // if there's a . but no specified precision, the |
| 148 | // precision is implicitly 0. |
| 149 | if (str[cur_pos] == '*') { |
| 150 | ++cur_pos; |
| 151 | |
| 152 | WRITE_ARG_VAL_SIMPLEST(section.precision, int, parse_index(&cur_pos)); |
| 153 | |
| 154 | } else if (internal::isdigit(str[cur_pos])) { |
| 155 | auto result = internal::strtointeger<int>(str + cur_pos, 10); |
| 156 | section.precision = result.value; |
| 157 | cur_pos = cur_pos + static_cast<size_t>(result.parsed_len); |
| 158 | } |
| 159 | } |
| 160 | |
| 161 | auto [lm, bw] = parse_length_modifier(&cur_pos); |
| 162 | section.length_modifier = lm; |
| 163 | section.conv_name = str[cur_pos]; |
| 164 | section.bit_width = bw; |
| 165 | switch (str[cur_pos]) { |
| 166 | case ('%'): |
| 167 | // Regardless of options, a % conversion is always safe. The standard |
| 168 | // says that "The complete conversion specification shall be %%" but it |
| 169 | // also says that "If a conversion specification is invalid, the |
| 170 | // behavior is undefined." Based on that we define that any conversion |
| 171 | // specification ending in '%' shall display as '%' regardless of any |
| 172 | // valid or invalid options. |
| 173 | section.has_conv = true; |
| 174 | break; |
| 175 | case ('c'): |
| 176 | WRITE_ARG_VAL_SIMPLEST(section.conv_val_raw, int, conv_index); |
| 177 | break; |
| 178 | case ('d'): |
| 179 | case ('i'): |
| 180 | case ('o'): |
| 181 | case ('x'): |
| 182 | case ('X'): |
| 183 | case ('u'): |
| 184 | case ('b'): |
| 185 | case ('B'): |
| 186 | switch (lm) { |
| 187 | case (LengthModifier::hh): |
| 188 | case (LengthModifier::h): |
| 189 | case (LengthModifier::none): |
| 190 | WRITE_ARG_VAL_SIMPLEST(section.conv_val_raw, int, conv_index); |
| 191 | break; |
| 192 | case (LengthModifier::l): |
| 193 | WRITE_ARG_VAL_SIMPLEST(section.conv_val_raw, long, conv_index); |
| 194 | break; |
| 195 | case (LengthModifier::ll): |
| 196 | case (LengthModifier::L): // This isn't in the standard, but is in other |
| 197 | // libc implementations. |
| 198 | |
| 199 | WRITE_ARG_VAL_SIMPLEST(section.conv_val_raw, long long, conv_index); |
| 200 | break; |
| 201 | case (LengthModifier::j): |
| 202 | |
| 203 | WRITE_ARG_VAL_SIMPLEST(section.conv_val_raw, intmax_t, conv_index); |
| 204 | break; |
| 205 | case (LengthModifier::z): |
| 206 | |
| 207 | WRITE_ARG_VAL_SIMPLEST(section.conv_val_raw, size_t, conv_index); |
| 208 | break; |
| 209 | case (LengthModifier::t): |
| 210 | |
| 211 | WRITE_ARG_VAL_SIMPLEST(section.conv_val_raw, ptrdiff_t, conv_index); |
| 212 | break; |
| 213 | |
| 214 | case (LengthModifier::w): |
| 215 | case (LengthModifier::wf): |
| 216 | if (bw == 0) { |
| 217 | section.has_conv = false; |
| 218 | } else if (bw <= cpp::numeric_limits<unsigned int>::digits) { |
| 219 | WRITE_ARG_VAL_SIMPLEST(section.conv_val_raw, int, conv_index); |
| 220 | } else if (bw <= cpp::numeric_limits<unsigned long>::digits) { |
| 221 | WRITE_ARG_VAL_SIMPLEST(section.conv_val_raw, long, conv_index); |
| 222 | } else if (bw <= cpp::numeric_limits<unsigned long long>::digits) { |
| 223 | WRITE_ARG_VAL_SIMPLEST(section.conv_val_raw, long long, conv_index); |
| 224 | } else { |
| 225 | WRITE_ARG_VAL_SIMPLEST(section.conv_val_raw, intmax_t, conv_index); |
| 226 | } |
| 227 | break; |
| 228 | } |
| 229 | break; |
| 230 | #ifndef LIBC_COPT_PRINTF_DISABLE_FLOAT |
| 231 | case ('f'): |
| 232 | case ('F'): |
| 233 | case ('e'): |
| 234 | case ('E'): |
| 235 | case ('a'): |
| 236 | case ('A'): |
| 237 | case ('g'): |
| 238 | case ('G'): |
| 239 | if (lm != LengthModifier::L) { |
| 240 | WRITE_ARG_VAL_SIMPLEST(section.conv_val_raw, double, conv_index); |
| 241 | } else { |
| 242 | WRITE_ARG_VAL_SIMPLEST(section.conv_val_raw, long double, conv_index); |
| 243 | } |
| 244 | break; |
| 245 | #endif // LIBC_COPT_PRINTF_DISABLE_FLOAT |
| 246 | #ifdef LIBC_INTERNAL_PRINTF_HAS_FIXED_POINT |
| 247 | // Capitalization represents sign, but we only need to get the right |
| 248 | // bitwidth here so we ignore that. |
| 249 | case ('r'): |
| 250 | case ('R'): |
| 251 | // all fract sizes we support are less than 32 bits, and currently doing |
| 252 | // va_args with fixed point types just doesn't work. |
| 253 | // TODO: Move to fixed point types once va_args supports it. |
| 254 | WRITE_ARG_VAL_SIMPLEST(section.conv_val_raw, uint32_t, conv_index); |
| 255 | break; |
| 256 | case ('k'): |
| 257 | case ('K'): |
| 258 | if (lm == LengthModifier::l) { |
| 259 | WRITE_ARG_VAL_SIMPLEST(section.conv_val_raw, uint64_t, conv_index); |
| 260 | } else { |
| 261 | WRITE_ARG_VAL_SIMPLEST(section.conv_val_raw, uint32_t, conv_index); |
| 262 | } |
| 263 | break; |
| 264 | #endif // LIBC_INTERNAL_PRINTF_HAS_FIXED_POINT |
| 265 | #ifndef LIBC_COPT_PRINTF_DISABLE_STRERROR |
| 266 | case ('m'): |
| 267 | // %m is an odd conversion in that it doesn't consume an argument, it |
| 268 | // just takes the current value of errno as its argument. |
| 269 | section.conv_val_raw = |
| 270 | static_cast<fputil::FPBits<double>::StorageType>(libc_errno); |
| 271 | break; |
| 272 | #endif // LIBC_COPT_PRINTF_DISABLE_STRERROR |
| 273 | #ifndef LIBC_COPT_PRINTF_DISABLE_WRITE_INT |
| 274 | case ('n'): // Intentional fallthrough |
| 275 | #endif // LIBC_COPT_PRINTF_DISABLE_WRITE_INT |
| 276 | case ('p'): |
| 277 | WRITE_ARG_VAL_SIMPLEST(section.conv_val_ptr, void *, conv_index); |
| 278 | break; |
| 279 | case ('s'): |
| 280 | WRITE_ARG_VAL_SIMPLEST(section.conv_val_ptr, char *, conv_index); |
| 281 | break; |
| 282 | default: |
| 283 | // if the conversion is undefined, change this to a raw section. |
| 284 | section.has_conv = false; |
| 285 | break; |
| 286 | } |
| 287 | // If the end of the format section is on the '\0'. This means we need to |
| 288 | // not advance the cur_pos. |
| 289 | if (str[cur_pos] != '\0') |
| 290 | ++cur_pos; |
| 291 | |
| 292 | } else { |
| 293 | // raw section |
| 294 | section.has_conv = false; |
| 295 | while (str[cur_pos] != '%' && str[cur_pos] != '\0') |
| 296 | ++cur_pos; |
| 297 | } |
| 298 | section.raw_string = {str + starting_pos, cur_pos - starting_pos}; |
| 299 | return section; |
| 300 | } |
| 301 | |
| 302 | private: |
| 303 | // parse_flags parses the flags inside a format string. It assumes that |
| 304 | // str[*local_pos] is inside a format specifier, and parses any flags it |
| 305 | // finds. It returns a FormatFlags object containing the set of found flags |
| 306 | // arithmetically or'd together. local_pos will be moved past any flags found. |
| 307 | LIBC_INLINE FormatFlags parse_flags(size_t *local_pos) { |
| 308 | bool found_flag = true; |
| 309 | FormatFlags flags = FormatFlags(0); |
| 310 | while (found_flag) { |
| 311 | switch (str[*local_pos]) { |
| 312 | case '-': |
| 313 | flags = static_cast<FormatFlags>(flags | FormatFlags::LEFT_JUSTIFIED); |
| 314 | break; |
| 315 | case '+': |
| 316 | flags = static_cast<FormatFlags>(flags | FormatFlags::FORCE_SIGN); |
| 317 | break; |
| 318 | case ' ': |
| 319 | flags = static_cast<FormatFlags>(flags | FormatFlags::SPACE_PREFIX); |
| 320 | break; |
| 321 | case '#': |
| 322 | flags = static_cast<FormatFlags>(flags | FormatFlags::ALTERNATE_FORM); |
| 323 | break; |
| 324 | case '0': |
| 325 | flags = static_cast<FormatFlags>(flags | FormatFlags::LEADING_ZEROES); |
| 326 | break; |
| 327 | default: |
| 328 | found_flag = false; |
| 329 | } |
| 330 | if (found_flag) |
| 331 | ++*local_pos; |
| 332 | } |
| 333 | return flags; |
| 334 | } |
| 335 | |
| 336 | // parse_length_modifier parses the length modifier inside a format string. It |
| 337 | // assumes that str[*local_pos] is inside a format specifier. It returns a |
| 338 | // LengthModifier with the length modifier it found. It will advance local_pos |
| 339 | // after the format specifier if one is found. |
| 340 | LIBC_INLINE LengthSpec parse_length_modifier(size_t *local_pos) { |
| 341 | switch (str[*local_pos]) { |
| 342 | case ('l'): |
| 343 | if (str[*local_pos + 1] == 'l') { |
| 344 | *local_pos += 2; |
| 345 | return {LengthModifier::ll, 0}; |
| 346 | } else { |
| 347 | ++*local_pos; |
| 348 | return {LengthModifier::l, 0}; |
| 349 | } |
| 350 | case ('w'): { |
| 351 | LengthModifier lm; |
| 352 | if (str[*local_pos + 1] == 'f') { |
| 353 | *local_pos += 2; |
| 354 | lm = LengthModifier::wf; |
| 355 | } else { |
| 356 | ++*local_pos; |
| 357 | lm = LengthModifier::w; |
| 358 | } |
| 359 | if (internal::isdigit(str[*local_pos])) { |
| 360 | const auto result = internal::strtointeger<int>(str + *local_pos, 10); |
| 361 | *local_pos += static_cast<size_t>(result.parsed_len); |
| 362 | return {lm, static_cast<size_t>(cpp::max(0, result.value))}; |
| 363 | } |
| 364 | return {lm, 0}; |
| 365 | } |
| 366 | case ('h'): |
| 367 | if (str[*local_pos + 1] == 'h') { |
| 368 | *local_pos += 2; |
| 369 | return {LengthModifier::hh, 0}; |
| 370 | } else { |
| 371 | ++*local_pos; |
| 372 | return {LengthModifier::h, 0}; |
| 373 | } |
| 374 | case ('L'): |
| 375 | ++*local_pos; |
| 376 | return {LengthModifier::L, 0}; |
| 377 | case ('j'): |
| 378 | ++*local_pos; |
| 379 | return {LengthModifier::j, 0}; |
| 380 | case ('z'): |
| 381 | ++*local_pos; |
| 382 | return {LengthModifier::z, 0}; |
| 383 | case ('t'): |
| 384 | ++*local_pos; |
| 385 | return {LengthModifier::t, 0}; |
| 386 | default: |
| 387 | return {LengthModifier::none, 0}; |
| 388 | } |
| 389 | } |
| 390 | |
| 391 | // get_next_arg_value gets the next value from the arg list as type T. |
| 392 | template <class T> LIBC_INLINE T get_next_arg_value() { |
| 393 | return args_cur.template next_var<T>(); |
| 394 | } |
| 395 | |
| 396 | //---------------------------------------------------- |
| 397 | // INDEX MODE ONLY FUNCTIONS AFTER HERE: |
| 398 | //---------------------------------------------------- |
| 399 | |
| 400 | #ifndef LIBC_COPT_PRINTF_DISABLE_INDEX_MODE |
| 401 | |
| 402 | // parse_index parses the index of a value inside a format string. It |
| 403 | // assumes that str[*local_pos] points to character after a '%' or '*', and |
| 404 | // returns 0 if there is no closing $, or if it finds no number. If it finds a |
| 405 | // number, it will move local_pos past the end of the $, else it will not move |
| 406 | // local_pos. |
| 407 | LIBC_INLINE size_t parse_index(size_t *local_pos) { |
| 408 | if (internal::isdigit(str[*local_pos])) { |
| 409 | auto result = internal::strtointeger<int>(str + *local_pos, 10); |
| 410 | size_t index = static_cast<size_t>(result.value); |
| 411 | if (str[*local_pos + static_cast<size_t>(result.parsed_len)] != '$') |
| 412 | return 0; |
| 413 | *local_pos = static_cast<size_t>(1 + result.parsed_len) + *local_pos; |
| 414 | return index; |
| 415 | } |
| 416 | return 0; |
| 417 | } |
| 418 | |
| 419 | LIBC_INLINE void set_type_desc(size_t index, TypeDesc value) { |
| 420 | if (index != 0 && index <= DESC_ARR_LEN) |
| 421 | desc_arr[index - 1] = value; |
| 422 | } |
| 423 | |
| 424 | // get_arg_value gets the value from the arg list at index (starting at 1). |
| 425 | // This may require parsing the format string. An index of 0 is interpreted as |
| 426 | // the next value. If the format string is not valid, it may have gaps in its |
| 427 | // indexes. Requesting the value for any index after a gap will fail, since |
| 428 | // the arg list must be read in order and with the correct types. |
| 429 | template <class T> LIBC_INLINE cpp::optional<T> get_arg_value(size_t index) { |
| 430 | if (!(index == 0 || index == args_index)) { |
| 431 | bool success = args_to_index(index); |
| 432 | if (!success) { |
| 433 | // If we can't get to this index, then the value of the arg can't be |
| 434 | // found. |
| 435 | return cpp::optional<T>(); |
| 436 | } |
| 437 | } |
| 438 | |
| 439 | set_type_desc(index, type_desc_from_type<T>()); |
| 440 | |
| 441 | ++args_index; |
| 442 | return get_next_arg_value<T>(); |
| 443 | } |
| 444 | |
| 445 | // the ArgProvider can only return the next item in the list. This function is |
| 446 | // used in index mode when the item that needs to be read is not the next one. |
| 447 | // It moves cur_args to the index requested so the appropriate value may |
| 448 | // be read. This may involve parsing the format string, and is in the worst |
| 449 | // case an O(n^2) operation. |
| 450 | LIBC_INLINE bool args_to_index(size_t index) { |
| 451 | if (args_index > index) { |
| 452 | args_index = 1; |
| 453 | args_cur = args_start; |
| 454 | } |
| 455 | |
| 456 | while (args_index < index) { |
| 457 | TypeDesc cur_type_desc = type_desc_from_type<void>(); |
| 458 | if (args_index <= DESC_ARR_LEN) |
| 459 | cur_type_desc = desc_arr[args_index - 1]; |
| 460 | |
| 461 | if (cur_type_desc == type_desc_from_type<void>()) |
| 462 | cur_type_desc = get_type_desc(args_index); |
| 463 | |
| 464 | // A type of void represents the type being unknown. If the type for the |
| 465 | // requested index isn't in the desc_arr and isn't found by parsing the |
| 466 | // string, then then advancing to the requested index is impossible. In |
| 467 | // that case the function returns false. |
| 468 | if (cur_type_desc == type_desc_from_type<void>()) |
| 469 | return false; |
| 470 | |
| 471 | if (cur_type_desc == type_desc_from_type<uint32_t>()) |
| 472 | args_cur.template next_var<uint32_t>(); |
| 473 | else if (cur_type_desc == type_desc_from_type<uint64_t>()) |
| 474 | args_cur.template next_var<uint64_t>(); |
| 475 | #ifndef LIBC_COPT_PRINTF_DISABLE_FLOAT |
| 476 | // Floating point numbers are stored separately from the other arguments. |
| 477 | else if (cur_type_desc == type_desc_from_type<double>()) |
| 478 | args_cur.template next_var<double>(); |
| 479 | else if (cur_type_desc == type_desc_from_type<long double>()) |
| 480 | args_cur.template next_var<long double>(); |
| 481 | #endif // LIBC_COPT_PRINTF_DISABLE_FLOAT |
| 482 | #ifdef LIBC_INTERNAL_PRINTF_HAS_FIXED_POINT |
| 483 | // Floating point numbers may be stored separately from the other |
| 484 | // arguments. |
| 485 | else if (cur_type_desc == type_desc_from_type<short fract>()) |
| 486 | args_cur.template next_var<short fract>(); |
| 487 | else if (cur_type_desc == type_desc_from_type<fract>()) |
| 488 | args_cur.template next_var<fract>(); |
| 489 | else if (cur_type_desc == type_desc_from_type<long fract>()) |
| 490 | args_cur.template next_var<long fract>(); |
| 491 | else if (cur_type_desc == type_desc_from_type<short accum>()) |
| 492 | args_cur.template next_var<short accum>(); |
| 493 | else if (cur_type_desc == type_desc_from_type<accum>()) |
| 494 | args_cur.template next_var<accum>(); |
| 495 | else if (cur_type_desc == type_desc_from_type<long accum>()) |
| 496 | args_cur.template next_var<long accum>(); |
| 497 | #endif // LIBC_INTERNAL_PRINTF_HAS_FIXED_POINT |
| 498 | // pointers may be stored separately from normal values. |
| 499 | else if (cur_type_desc == type_desc_from_type<void *>()) |
| 500 | args_cur.template next_var<void *>(); |
| 501 | else |
| 502 | args_cur.template next_var<uint32_t>(); |
| 503 | |
| 504 | ++args_index; |
| 505 | } |
| 506 | return true; |
| 507 | } |
| 508 | |
| 509 | // get_type_desc assumes that this format string uses index mode. It iterates |
| 510 | // through the format string until it finds a format specifier that defines |
| 511 | // the type of index, and returns a TypeDesc describing that type. It does not |
| 512 | // modify cur_pos. |
| 513 | LIBC_INLINE TypeDesc get_type_desc(size_t index) { |
| 514 | // index mode is assumed, and the indices start at 1, so an index |
| 515 | // of 0 is invalid. |
| 516 | size_t local_pos = 0; |
| 517 | |
| 518 | while (str[local_pos]) { |
| 519 | if (str[local_pos] == '%') { |
| 520 | ++local_pos; |
| 521 | |
| 522 | size_t conv_index = parse_index(&local_pos); |
| 523 | |
| 524 | // the flags aren't relevant for this situation, but I need to skip past |
| 525 | // them so they're parsed but the result is discarded. |
| 526 | parse_flags(&local_pos); |
| 527 | |
| 528 | // handle width |
| 529 | if (str[local_pos] == '*') { |
| 530 | ++local_pos; |
| 531 | |
| 532 | size_t width_index = parse_index(&local_pos); |
| 533 | set_type_desc(width_index, type_desc_from_type<int>()); |
| 534 | if (width_index == index) |
| 535 | return type_desc_from_type<int>(); |
| 536 | |
| 537 | } else if (internal::isdigit(str[local_pos])) { |
| 538 | while (internal::isdigit(str[local_pos])) |
| 539 | ++local_pos; |
| 540 | } |
| 541 | |
| 542 | // handle precision |
| 543 | if (str[local_pos] == '.') { |
| 544 | ++local_pos; |
| 545 | if (str[local_pos] == '*') { |
| 546 | ++local_pos; |
| 547 | |
| 548 | size_t precision_index = parse_index(&local_pos); |
| 549 | set_type_desc(precision_index, type_desc_from_type<int>()); |
| 550 | if (precision_index == index) |
| 551 | return type_desc_from_type<int>(); |
| 552 | |
| 553 | } else if (internal::isdigit(str[local_pos])) { |
| 554 | while (internal::isdigit(str[local_pos])) |
| 555 | ++local_pos; |
| 556 | } |
| 557 | } |
| 558 | |
| 559 | auto [lm, bw] = parse_length_modifier(&local_pos); |
| 560 | |
| 561 | // if we don't have an index for this conversion, then its position is |
| 562 | // unknown and all this information is irrelevant. The rest of this |
| 563 | // logic has been for skipping past this conversion properly to avoid |
| 564 | // weirdness with %%. |
| 565 | if (conv_index == 0) { |
| 566 | if (str[local_pos] != '\0') |
| 567 | ++local_pos; |
| 568 | continue; |
| 569 | } |
| 570 | |
| 571 | TypeDesc conv_size = type_desc_from_type<void>(); |
| 572 | switch (str[local_pos]) { |
| 573 | case ('%'): |
| 574 | conv_size = type_desc_from_type<void>(); |
| 575 | break; |
| 576 | case ('c'): |
| 577 | conv_size = type_desc_from_type<int>(); |
| 578 | break; |
| 579 | case ('d'): |
| 580 | case ('i'): |
| 581 | case ('o'): |
| 582 | case ('x'): |
| 583 | case ('X'): |
| 584 | case ('u'): |
| 585 | case ('b'): |
| 586 | case ('B'): |
| 587 | switch (lm) { |
| 588 | case (LengthModifier::hh): |
| 589 | case (LengthModifier::h): |
| 590 | case (LengthModifier::none): |
| 591 | conv_size = type_desc_from_type<int>(); |
| 592 | break; |
| 593 | case (LengthModifier::l): |
| 594 | conv_size = type_desc_from_type<long>(); |
| 595 | break; |
| 596 | case (LengthModifier::ll): |
| 597 | case (LengthModifier::L): // This isn't in the standard, but is in |
| 598 | // other libc implementations. |
| 599 | conv_size = type_desc_from_type<long long>(); |
| 600 | break; |
| 601 | case (LengthModifier::j): |
| 602 | conv_size = type_desc_from_type<intmax_t>(); |
| 603 | break; |
| 604 | case (LengthModifier::z): |
| 605 | conv_size = type_desc_from_type<size_t>(); |
| 606 | break; |
| 607 | case (LengthModifier::t): |
| 608 | conv_size = type_desc_from_type<ptrdiff_t>(); |
| 609 | break; |
| 610 | case (LengthModifier::w): |
| 611 | case (LengthModifier::wf): |
| 612 | if (bw <= cpp::numeric_limits<unsigned int>::digits) { |
| 613 | conv_size = type_desc_from_type<int>(); |
| 614 | } else if (bw <= cpp::numeric_limits<unsigned long>::digits) { |
| 615 | conv_size = type_desc_from_type<long>(); |
| 616 | } else if (bw <= cpp::numeric_limits<unsigned long long>::digits) { |
| 617 | conv_size = type_desc_from_type<long long>(); |
| 618 | } else { |
| 619 | conv_size = type_desc_from_type<intmax_t>(); |
| 620 | } |
| 621 | break; |
| 622 | } |
| 623 | break; |
| 624 | #ifndef LIBC_COPT_PRINTF_DISABLE_FLOAT |
| 625 | case ('f'): |
| 626 | case ('F'): |
| 627 | case ('e'): |
| 628 | case ('E'): |
| 629 | case ('a'): |
| 630 | case ('A'): |
| 631 | case ('g'): |
| 632 | case ('G'): |
| 633 | if (lm != LengthModifier::L) |
| 634 | conv_size = type_desc_from_type<double>(); |
| 635 | else |
| 636 | conv_size = type_desc_from_type<long double>(); |
| 637 | break; |
| 638 | #endif // LIBC_COPT_PRINTF_DISABLE_FLOAT |
| 639 | #ifdef LIBC_INTERNAL_PRINTF_HAS_FIXED_POINT |
| 640 | // Capitalization represents sign, but we only need to get the right |
| 641 | // bitwidth here so we ignore that. |
| 642 | case ('r'): |
| 643 | case ('R'): |
| 644 | conv_size = type_desc_from_type<uint32_t>(); |
| 645 | break; |
| 646 | case ('k'): |
| 647 | case ('K'): |
| 648 | if (lm == LengthModifier::l) { |
| 649 | conv_size = type_desc_from_type<uint64_t>(); |
| 650 | } else { |
| 651 | conv_size = type_desc_from_type<uint32_t>(); |
| 652 | } |
| 653 | break; |
| 654 | #endif // LIBC_INTERNAL_PRINTF_HAS_FIXED_POINT |
| 655 | #ifndef LIBC_COPT_PRINTF_DISABLE_WRITE_INT |
| 656 | case ('n'): |
| 657 | #endif // LIBC_COPT_PRINTF_DISABLE_WRITE_INT |
| 658 | case ('p'): |
| 659 | case ('s'): |
| 660 | conv_size = type_desc_from_type<void *>(); |
| 661 | break; |
| 662 | default: |
| 663 | conv_size = type_desc_from_type<int>(); |
| 664 | break; |
| 665 | } |
| 666 | |
| 667 | set_type_desc(conv_index, conv_size); |
| 668 | if (conv_index == index) |
| 669 | return conv_size; |
| 670 | } |
| 671 | // If the end of the format section is on the '\0'. This means we need to |
| 672 | // not advance the local_pos. |
| 673 | if (str[local_pos] != '\0') |
| 674 | ++local_pos; |
| 675 | } |
| 676 | |
| 677 | // If there is no size for the requested index, then it's unknown. Return |
| 678 | // void. |
| 679 | return type_desc_from_type<void>(); |
| 680 | } |
| 681 | |
| 682 | #endif // LIBC_COPT_PRINTF_DISABLE_INDEX_MODE |
| 683 | }; |
| 684 | |
| 685 | } // namespace printf_core |
| 686 | } // namespace LIBC_NAMESPACE_DECL |
| 687 | |
| 688 | #endif // LLVM_LIBC_SRC_STDIO_PRINTF_CORE_PARSER_H |
| 689 |
Warning: This file is not a C or C++ file. It does not have highlighting.
