Warning: This file is not a C or C++ file. It does not have highlighting.
| 1 | // -*- C++ -*- |
|---|---|
| 2 | //===----------------------------------------------------------------------===// |
| 3 | // |
| 4 | // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. |
| 5 | // See https://llvm.org/LICENSE.txt for license information. |
| 6 | // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception |
| 7 | // |
| 8 | //===----------------------------------------------------------------------===// |
| 9 | |
| 10 | #ifndef _LIBCPP___FORMAT_PARSER_STD_FORMAT_SPEC_H |
| 11 | #define _LIBCPP___FORMAT_PARSER_STD_FORMAT_SPEC_H |
| 12 | |
| 13 | /// \file Contains the std-format-spec parser. |
| 14 | /// |
| 15 | /// Most of the code can be reused in the chrono-format-spec. |
| 16 | /// This header has some support for the chrono-format-spec since it doesn't |
| 17 | /// affect the std-format-spec. |
| 18 | |
| 19 | #include <__algorithm/copy_n.h> |
| 20 | #include <__algorithm/min.h> |
| 21 | #include <__assert> |
| 22 | #include <__concepts/arithmetic.h> |
| 23 | #include <__concepts/same_as.h> |
| 24 | #include <__config> |
| 25 | #include <__format/format_arg.h> |
| 26 | #include <__format/format_error.h> |
| 27 | #include <__format/format_parse_context.h> |
| 28 | #include <__format/format_string.h> |
| 29 | #include <__format/unicode.h> |
| 30 | #include <__format/width_estimation_table.h> |
| 31 | #include <__iterator/concepts.h> |
| 32 | #include <__iterator/iterator_traits.h> // iter_value_t |
| 33 | #include <__memory/addressof.h> |
| 34 | #include <__type_traits/common_type.h> |
| 35 | #include <__type_traits/is_constant_evaluated.h> |
| 36 | #include <__type_traits/is_trivially_copyable.h> |
| 37 | #include <__variant/monostate.h> |
| 38 | #include <cstdint> |
| 39 | #include <string> |
| 40 | #include <string_view> |
| 41 | |
| 42 | #if !defined(_LIBCPP_HAS_NO_PRAGMA_SYSTEM_HEADER) |
| 43 | # pragma GCC system_header |
| 44 | #endif |
| 45 | |
| 46 | _LIBCPP_PUSH_MACROS |
| 47 | #include <__undef_macros> |
| 48 | |
| 49 | _LIBCPP_BEGIN_NAMESPACE_STD |
| 50 | |
| 51 | #if _LIBCPP_STD_VER >= 20 |
| 52 | |
| 53 | namespace __format_spec { |
| 54 | |
| 55 | [[noreturn]] _LIBCPP_HIDE_FROM_ABI inline void |
| 56 | __throw_invalid_option_format_error(const char* __id, const char* __option) { |
| 57 | std::__throw_format_error( |
| 58 | (string("The format specifier for ") + __id + " does not allow the " + __option + " option").c_str()); |
| 59 | } |
| 60 | |
| 61 | [[noreturn]] _LIBCPP_HIDE_FROM_ABI inline void __throw_invalid_type_format_error(const char* __id) { |
| 62 | std::__throw_format_error( |
| 63 | (string("The type option contains an invalid value for ") + __id + " formatting argument").c_str()); |
| 64 | } |
| 65 | |
| 66 | template <contiguous_iterator _Iterator, class _ParseContext> |
| 67 | _LIBCPP_HIDE_FROM_ABI constexpr __format::__parse_number_result<_Iterator> |
| 68 | __parse_arg_id(_Iterator __begin, _Iterator __end, _ParseContext& __ctx) { |
| 69 | using _CharT = iter_value_t<_Iterator>; |
| 70 | // This function is a wrapper to call the real parser. But it does the |
| 71 | // validation for the pre-conditions and post-conditions. |
| 72 | if (__begin == __end) |
| 73 | std::__throw_format_error("End of input while parsing an argument index"); |
| 74 | |
| 75 | __format::__parse_number_result __r = __format::__parse_arg_id(__begin, __end, __ctx); |
| 76 | |
| 77 | if (__r.__last == __end || *__r.__last != _CharT('}')) |
| 78 | std::__throw_format_error("The argument index is invalid"); |
| 79 | |
| 80 | ++__r.__last; |
| 81 | return __r; |
| 82 | } |
| 83 | |
| 84 | template <class _Context> |
| 85 | _LIBCPP_HIDE_FROM_ABI constexpr uint32_t __substitute_arg_id(basic_format_arg<_Context> __format_arg) { |
| 86 | // [format.string.std]/8 |
| 87 | // If the corresponding formatting argument is not of integral type... |
| 88 | // This wording allows char and bool too. LWG-3720 changes the wording to |
| 89 | // If the corresponding formatting argument is not of standard signed or |
| 90 | // unsigned integer type, |
| 91 | // This means the 128-bit will not be valid anymore. |
| 92 | // TODO FMT Verify this resolution is accepted and add a test to verify |
| 93 | // 128-bit integrals fail and switch to visit_format_arg. |
| 94 | return std::__visit_format_arg( |
| 95 | [](auto __arg) -> uint32_t { |
| 96 | using _Type = decltype(__arg); |
| 97 | if constexpr (same_as<_Type, monostate>) |
| 98 | std::__throw_format_error("The argument index value is too large for the number of arguments supplied"); |
| 99 | |
| 100 | // [format.string.std]/8 |
| 101 | // If { arg-idopt } is used in a width or precision, the value of the |
| 102 | // corresponding formatting argument is used in its place. If the |
| 103 | // corresponding formatting argument is not of standard signed or unsigned |
| 104 | // integer type, or its value is negative for precision or non-positive for |
| 105 | // width, an exception of type format_error is thrown. |
| 106 | // |
| 107 | // When an integral is used in a format function, it is stored as one of |
| 108 | // the types checked below. Other integral types are promoted. For example, |
| 109 | // a signed char is stored as an int. |
| 110 | if constexpr (same_as<_Type, int> || same_as<_Type, unsigned int> || // |
| 111 | same_as<_Type, long long> || same_as<_Type, unsigned long long>) { |
| 112 | if constexpr (signed_integral<_Type>) { |
| 113 | if (__arg < 0) |
| 114 | std::__throw_format_error("An argument index may not have a negative value"); |
| 115 | } |
| 116 | |
| 117 | using _CT = common_type_t<_Type, decltype(__format::__number_max)>; |
| 118 | if (static_cast<_CT>(__arg) > static_cast<_CT>(__format::__number_max)) |
| 119 | std::__throw_format_error("The value of the argument index exceeds its maximum value"); |
| 120 | |
| 121 | return __arg; |
| 122 | } else |
| 123 | std::__throw_format_error("Replacement argument isn't a standard signed or unsigned integer type"); |
| 124 | }, |
| 125 | __format_arg); |
| 126 | } |
| 127 | |
| 128 | /// These fields are a filter for which elements to parse. |
| 129 | /// |
| 130 | /// They default to false so when a new field is added it needs to be opted in |
| 131 | /// explicitly. |
| 132 | struct _LIBCPP_HIDE_FROM_ABI __fields { |
| 133 | uint16_t __sign_ : 1 {false}; |
| 134 | uint16_t __alternate_form_ : 1 {false}; |
| 135 | uint16_t __zero_padding_ : 1 {false}; |
| 136 | uint16_t __precision_ : 1 {false}; |
| 137 | uint16_t __locale_specific_form_ : 1 {false}; |
| 138 | uint16_t __type_ : 1 {false}; |
| 139 | // Determines the valid values for fill. |
| 140 | // |
| 141 | // Originally the fill could be any character except { and }. Range-based |
| 142 | // formatters use the colon to mark the beginning of the |
| 143 | // underlying-format-spec. To avoid parsing ambiguities these formatter |
| 144 | // specializations prohibit the use of the colon as a fill character. |
| 145 | uint16_t __use_range_fill_ : 1 {false}; |
| 146 | uint16_t __clear_brackets_ : 1 {false}; |
| 147 | uint16_t __consume_all_ : 1 {false}; |
| 148 | }; |
| 149 | |
| 150 | // By not placing this constant in the formatter class it's not duplicated for |
| 151 | // char and wchar_t. |
| 152 | inline constexpr __fields __fields_bool{.__locale_specific_form_ = true, .__type_ = true, .__consume_all_ = true}; |
| 153 | inline constexpr __fields __fields_integral{ |
| 154 | .__sign_ = true, |
| 155 | .__alternate_form_ = true, |
| 156 | .__zero_padding_ = true, |
| 157 | .__locale_specific_form_ = true, |
| 158 | .__type_ = true, |
| 159 | .__consume_all_ = true}; |
| 160 | inline constexpr __fields __fields_floating_point{ |
| 161 | .__sign_ = true, |
| 162 | .__alternate_form_ = true, |
| 163 | .__zero_padding_ = true, |
| 164 | .__precision_ = true, |
| 165 | .__locale_specific_form_ = true, |
| 166 | .__type_ = true, |
| 167 | .__consume_all_ = true}; |
| 168 | inline constexpr __fields __fields_string{.__precision_ = true, .__type_ = true, .__consume_all_ = true}; |
| 169 | inline constexpr __fields __fields_pointer{.__zero_padding_ = true, .__type_ = true, .__consume_all_ = true}; |
| 170 | |
| 171 | # if _LIBCPP_STD_VER >= 23 |
| 172 | inline constexpr __fields __fields_tuple{.__use_range_fill_ = true, .__clear_brackets_ = true}; |
| 173 | inline constexpr __fields __fields_range{.__use_range_fill_ = true, .__clear_brackets_ = true}; |
| 174 | inline constexpr __fields __fields_fill_align_width{}; |
| 175 | # endif |
| 176 | |
| 177 | enum class __alignment : uint8_t { |
| 178 | /// No alignment is set in the format string. |
| 179 | __default, |
| 180 | __left, |
| 181 | __center, |
| 182 | __right, |
| 183 | __zero_padding |
| 184 | }; |
| 185 | |
| 186 | enum class __sign : uint8_t { |
| 187 | /// No sign is set in the format string. |
| 188 | /// |
| 189 | /// The sign isn't allowed for certain format-types. By using this value |
| 190 | /// it's possible to detect whether or not the user explicitly set the sign |
| 191 | /// flag. For formatting purposes it behaves the same as \ref __minus. |
| 192 | __default, |
| 193 | __minus, |
| 194 | __plus, |
| 195 | __space |
| 196 | }; |
| 197 | |
| 198 | enum class __type : uint8_t { |
| 199 | __default = 0, |
| 200 | __string, |
| 201 | __binary_lower_case, |
| 202 | __binary_upper_case, |
| 203 | __octal, |
| 204 | __decimal, |
| 205 | __hexadecimal_lower_case, |
| 206 | __hexadecimal_upper_case, |
| 207 | __pointer_lower_case, |
| 208 | __pointer_upper_case, |
| 209 | __char, |
| 210 | __hexfloat_lower_case, |
| 211 | __hexfloat_upper_case, |
| 212 | __scientific_lower_case, |
| 213 | __scientific_upper_case, |
| 214 | __fixed_lower_case, |
| 215 | __fixed_upper_case, |
| 216 | __general_lower_case, |
| 217 | __general_upper_case, |
| 218 | __debug |
| 219 | }; |
| 220 | |
| 221 | _LIBCPP_HIDE_FROM_ABI inline constexpr uint32_t __create_type_mask(__type __t) { |
| 222 | uint32_t __shift = static_cast<uint32_t>(__t); |
| 223 | if (__shift == 0) |
| 224 | return 1; |
| 225 | |
| 226 | if (__shift > 31) |
| 227 | std::__throw_format_error("The type does not fit in the mask"); |
| 228 | |
| 229 | return 1 << __shift; |
| 230 | } |
| 231 | |
| 232 | inline constexpr uint32_t __type_mask_integer = |
| 233 | __create_type_mask(__type::__binary_lower_case) | // |
| 234 | __create_type_mask(__type::__binary_upper_case) | // |
| 235 | __create_type_mask(__type::__decimal) | // |
| 236 | __create_type_mask(__type::__octal) | // |
| 237 | __create_type_mask(__type::__hexadecimal_lower_case) | // |
| 238 | __create_type_mask(__type::__hexadecimal_upper_case); |
| 239 | |
| 240 | struct __std { |
| 241 | __alignment __alignment_ : 3; |
| 242 | __sign __sign_ : 2; |
| 243 | bool __alternate_form_ : 1; |
| 244 | bool __locale_specific_form_ : 1; |
| 245 | __type __type_; |
| 246 | }; |
| 247 | |
| 248 | struct __chrono { |
| 249 | __alignment __alignment_ : 3; |
| 250 | bool __locale_specific_form_ : 1; |
| 251 | bool __hour_ : 1; |
| 252 | bool __weekday_name_ : 1; |
| 253 | bool __weekday_ : 1; |
| 254 | bool __day_of_year_ : 1; |
| 255 | bool __week_of_year_ : 1; |
| 256 | bool __month_name_ : 1; |
| 257 | }; |
| 258 | |
| 259 | // The fill UCS scalar value. |
| 260 | // |
| 261 | // This is always an array, with 1, 2, or 4 elements. |
| 262 | // The size of the data structure is always 32-bits. |
| 263 | template <class _CharT> |
| 264 | struct __code_point; |
| 265 | |
| 266 | template <> |
| 267 | struct __code_point<char> { |
| 268 | char __data[4] = {' '}; |
| 269 | }; |
| 270 | |
| 271 | # if _LIBCPP_HAS_WIDE_CHARACTERS |
| 272 | template <> |
| 273 | struct __code_point<wchar_t> { |
| 274 | wchar_t __data[4 / sizeof(wchar_t)] = {L' '}; |
| 275 | }; |
| 276 | # endif |
| 277 | |
| 278 | /// Contains the parsed formatting specifications. |
| 279 | /// |
| 280 | /// This contains information for both the std-format-spec and the |
| 281 | /// chrono-format-spec. This results in some unused members for both |
| 282 | /// specifications. However these unused members don't increase the size |
| 283 | /// of the structure. |
| 284 | /// |
| 285 | /// This struct doesn't cross ABI boundaries so its layout doesn't need to be |
| 286 | /// kept stable. |
| 287 | template <class _CharT> |
| 288 | struct __parsed_specifications { |
| 289 | union { |
| 290 | // The field __alignment_ is the first element in __std_ and __chrono_. |
| 291 | // This allows the code to always inspect this value regards which member |
| 292 | // of the union is the active member [class.union.general]/2. |
| 293 | // |
| 294 | // This is needed since the generic output routines handle the alignment of |
| 295 | // the output. |
| 296 | __alignment __alignment_ : 3; |
| 297 | __std __std_; |
| 298 | __chrono __chrono_; |
| 299 | }; |
| 300 | |
| 301 | /// The requested width. |
| 302 | /// |
| 303 | /// When the format-spec used an arg-id for this field it has already been |
| 304 | /// replaced with the value of that arg-id. |
| 305 | int32_t __width_; |
| 306 | |
| 307 | /// The requested precision. |
| 308 | /// |
| 309 | /// When the format-spec used an arg-id for this field it has already been |
| 310 | /// replaced with the value of that arg-id. |
| 311 | int32_t __precision_; |
| 312 | |
| 313 | __code_point<_CharT> __fill_; |
| 314 | |
| 315 | _LIBCPP_HIDE_FROM_ABI constexpr bool __has_width() const { return __width_ > 0; } |
| 316 | |
| 317 | _LIBCPP_HIDE_FROM_ABI constexpr bool __has_precision() const { return __precision_ >= 0; } |
| 318 | }; |
| 319 | |
| 320 | // Validate the struct is small and cheap to copy since the struct is passed by |
| 321 | // value in formatting functions. |
| 322 | static_assert(sizeof(__parsed_specifications<char>) == 16); |
| 323 | static_assert(is_trivially_copyable_v<__parsed_specifications<char>>); |
| 324 | # if _LIBCPP_HAS_WIDE_CHARACTERS |
| 325 | static_assert(sizeof(__parsed_specifications<wchar_t>) == 16); |
| 326 | static_assert(is_trivially_copyable_v<__parsed_specifications<wchar_t>>); |
| 327 | # endif |
| 328 | |
| 329 | /// The parser for the std-format-spec. |
| 330 | /// |
| 331 | /// Note this class is a member of std::formatter specializations. It's |
| 332 | /// expected developers will create their own formatter specializations that |
| 333 | /// inherit from the std::formatter specializations. This means this class |
| 334 | /// must be ABI stable. To aid the stability the unused bits in the class are |
| 335 | /// set to zero. That way they can be repurposed if a future revision of the |
| 336 | /// Standards adds new fields to std-format-spec. |
| 337 | template <class _CharT> |
| 338 | class __parser { |
| 339 | public: |
| 340 | // Parses the format specification. |
| 341 | // |
| 342 | // Depending on whether the parsing is done compile-time or run-time |
| 343 | // the method slightly differs. |
| 344 | // - Only parses a field when it is in the __fields. Accepting all |
| 345 | // fields and then validating the valid ones has a performance impact. |
| 346 | // This is faster but gives slighly worse error messages. |
| 347 | // - At compile-time when a field is not accepted the parser will still |
| 348 | // parse it and give an error when it's present. This gives a more |
| 349 | // accurate error. |
| 350 | // The idea is that most times the format instead of the vformat |
| 351 | // functions are used. In that case the error will be detected during |
| 352 | // compilation and there is no need to pay for the run-time overhead. |
| 353 | template <class _ParseContext> |
| 354 | _LIBCPP_HIDE_FROM_ABI constexpr typename _ParseContext::iterator __parse(_ParseContext& __ctx, __fields __fields) { |
| 355 | auto __begin = __ctx.begin(); |
| 356 | auto __end = __ctx.end(); |
| 357 | if (__begin == __end || *__begin == _CharT('}') || (__fields.__use_range_fill_ && *__begin == _CharT(':'))) |
| 358 | return __begin; |
| 359 | |
| 360 | if (__parse_fill_align(__begin, __end) && __begin == __end) |
| 361 | return __begin; |
| 362 | |
| 363 | if (__fields.__sign_) { |
| 364 | if (__parse_sign(__begin) && __begin == __end) |
| 365 | return __begin; |
| 366 | } else if (std::is_constant_evaluated() && __parse_sign(__begin)) { |
| 367 | std::__throw_format_error("The format specification does not allow the sign option"); |
| 368 | } |
| 369 | |
| 370 | if (__fields.__alternate_form_) { |
| 371 | if (__parse_alternate_form(__begin) && __begin == __end) |
| 372 | return __begin; |
| 373 | } else if (std::is_constant_evaluated() && __parse_alternate_form(__begin)) { |
| 374 | std::__throw_format_error("The format specifier does not allow the alternate form option"); |
| 375 | } |
| 376 | |
| 377 | if (__fields.__zero_padding_) { |
| 378 | if (__parse_zero_padding(__begin) && __begin == __end) |
| 379 | return __begin; |
| 380 | } else if (std::is_constant_evaluated() && __parse_zero_padding(__begin)) { |
| 381 | std::__throw_format_error("The format specifier does not allow the zero-padding option"); |
| 382 | } |
| 383 | |
| 384 | if (__parse_width(__begin, __end, __ctx) && __begin == __end) |
| 385 | return __begin; |
| 386 | |
| 387 | if (__fields.__precision_) { |
| 388 | if (__parse_precision(__begin, __end, __ctx) && __begin == __end) |
| 389 | return __begin; |
| 390 | } else if (std::is_constant_evaluated() && __parse_precision(__begin, __end, __ctx)) { |
| 391 | std::__throw_format_error("The format specifier does not allow the precision option"); |
| 392 | } |
| 393 | |
| 394 | if (__fields.__locale_specific_form_) { |
| 395 | if (__parse_locale_specific_form(__begin) && __begin == __end) |
| 396 | return __begin; |
| 397 | } else if (std::is_constant_evaluated() && __parse_locale_specific_form(__begin)) { |
| 398 | std::__throw_format_error("The format specifier does not allow the locale-specific form option"); |
| 399 | } |
| 400 | |
| 401 | if (__fields.__clear_brackets_) { |
| 402 | if (__parse_clear_brackets(__begin) && __begin == __end) |
| 403 | return __begin; |
| 404 | } else if (std::is_constant_evaluated() && __parse_clear_brackets(__begin)) { |
| 405 | std::__throw_format_error("The format specifier does not allow the n option"); |
| 406 | } |
| 407 | |
| 408 | if (__fields.__type_) |
| 409 | __parse_type(__begin); |
| 410 | |
| 411 | if (!__fields.__consume_all_) |
| 412 | return __begin; |
| 413 | |
| 414 | if (__begin != __end && *__begin != _CharT('}')) |
| 415 | std::__throw_format_error("The format specifier should consume the input or end with a '}'"); |
| 416 | |
| 417 | return __begin; |
| 418 | } |
| 419 | |
| 420 | // Validates the selected the parsed data. |
| 421 | // |
| 422 | // The valid fields in the parser may depend on the display type |
| 423 | // selected. But the type is the last optional field, so by the time |
| 424 | // it's known an option can't be used, it already has been parsed. |
| 425 | // This does the validation again. |
| 426 | // |
| 427 | // For example an integral may have a sign, zero-padding, or alternate |
| 428 | // form when the type option is not 'c'. So the generic approach is: |
| 429 | // |
| 430 | // typename _ParseContext::iterator __result = __parser_.__parse(__ctx, __format_spec::__fields_integral); |
| 431 | // if (__parser.__type_ == __format_spec::__type::__char) { |
| 432 | // __parser.__validate((__format_spec::__fields_bool, "an integer"); |
| 433 | // ... // more char adjustments |
| 434 | // } else { |
| 435 | // ... // validate an integral type. |
| 436 | // } |
| 437 | // |
| 438 | // For some types all valid options need a second validation run, like |
| 439 | // boolean types. |
| 440 | // |
| 441 | // Depending on whether the validation is done at compile-time or |
| 442 | // run-time the error differs |
| 443 | // - run-time the exception is thrown and contains the type of field |
| 444 | // being validated. |
| 445 | // - at compile-time the line with `std::__throw_format_error` is shown |
| 446 | // in the output. In that case it's important for the error to be on one |
| 447 | // line. |
| 448 | // Note future versions of C++ may allow better compile-time error |
| 449 | // reporting. |
| 450 | _LIBCPP_HIDE_FROM_ABI constexpr void |
| 451 | __validate(__fields __fields, const char* __id, uint32_t __type_mask = -1) const { |
| 452 | if (!__fields.__sign_ && __sign_ != __sign::__default) { |
| 453 | if (std::is_constant_evaluated()) |
| 454 | std::__throw_format_error("The format specifier does not allow the sign option"); |
| 455 | else |
| 456 | __format_spec::__throw_invalid_option_format_error(__id, "sign"); |
| 457 | } |
| 458 | |
| 459 | if (!__fields.__alternate_form_ && __alternate_form_) { |
| 460 | if (std::is_constant_evaluated()) |
| 461 | std::__throw_format_error("The format specifier does not allow the alternate form option"); |
| 462 | else |
| 463 | __format_spec::__throw_invalid_option_format_error(__id, "alternate form"); |
| 464 | } |
| 465 | |
| 466 | if (!__fields.__zero_padding_ && __alignment_ == __alignment::__zero_padding) { |
| 467 | if (std::is_constant_evaluated()) |
| 468 | std::__throw_format_error("The format specifier does not allow the zero-padding option"); |
| 469 | else |
| 470 | __format_spec::__throw_invalid_option_format_error(__id, "zero-padding"); |
| 471 | } |
| 472 | |
| 473 | if (!__fields.__precision_ && __precision_ != -1) { // Works both when the precision has a value or an arg-id. |
| 474 | if (std::is_constant_evaluated()) |
| 475 | std::__throw_format_error("The format specifier does not allow the precision option"); |
| 476 | else |
| 477 | __format_spec::__throw_invalid_option_format_error(__id, "precision"); |
| 478 | } |
| 479 | |
| 480 | if (!__fields.__locale_specific_form_ && __locale_specific_form_) { |
| 481 | if (std::is_constant_evaluated()) |
| 482 | std::__throw_format_error("The format specifier does not allow the locale-specific form option"); |
| 483 | else |
| 484 | __format_spec::__throw_invalid_option_format_error(__id, "locale-specific form"); |
| 485 | } |
| 486 | |
| 487 | if ((__create_type_mask(__type_) & __type_mask) == 0) { |
| 488 | if (std::is_constant_evaluated()) |
| 489 | std::__throw_format_error("The format specifier uses an invalid value for the type option"); |
| 490 | else |
| 491 | __format_spec::__throw_invalid_type_format_error(__id); |
| 492 | } |
| 493 | } |
| 494 | |
| 495 | /// \returns the `__parsed_specifications` with the resolved dynamic sizes.. |
| 496 | _LIBCPP_HIDE_FROM_ABI __parsed_specifications<_CharT> __get_parsed_std_specifications(auto& __ctx) const { |
| 497 | return __parsed_specifications<_CharT>{ |
| 498 | .__std_ = __std{.__alignment_ = __alignment_, |
| 499 | .__sign_ = __sign_, |
| 500 | .__alternate_form_ = __alternate_form_, |
| 501 | .__locale_specific_form_ = __locale_specific_form_, |
| 502 | .__type_ = __type_}, |
| 503 | .__width_{__get_width(__ctx)}, |
| 504 | .__precision_{__get_precision(__ctx)}, |
| 505 | .__fill_{__fill_}}; |
| 506 | } |
| 507 | |
| 508 | _LIBCPP_HIDE_FROM_ABI __parsed_specifications<_CharT> __get_parsed_chrono_specifications(auto& __ctx) const { |
| 509 | return __parsed_specifications<_CharT>{ |
| 510 | .__chrono_ = |
| 511 | __chrono{.__alignment_ = __alignment_, |
| 512 | .__locale_specific_form_ = __locale_specific_form_, |
| 513 | .__hour_ = __hour_, |
| 514 | .__weekday_name_ = __weekday_name_, |
| 515 | .__weekday_ = __weekday_, |
| 516 | .__day_of_year_ = __day_of_year_, |
| 517 | .__week_of_year_ = __week_of_year_, |
| 518 | .__month_name_ = __month_name_}, |
| 519 | .__width_{__get_width(__ctx)}, |
| 520 | .__precision_{__get_precision(__ctx)}, |
| 521 | .__fill_{__fill_}}; |
| 522 | } |
| 523 | |
| 524 | __alignment __alignment_ : 3 {__alignment::__default}; |
| 525 | __sign __sign_ : 2 {__sign::__default}; |
| 526 | bool __alternate_form_ : 1 {false}; |
| 527 | bool __locale_specific_form_ : 1 {false}; |
| 528 | bool __clear_brackets_ : 1 {false}; |
| 529 | __type __type_{__type::__default}; |
| 530 | |
| 531 | // These flags are only used for formatting chrono. Since the struct has |
| 532 | // padding space left it's added to this structure. |
| 533 | bool __hour_ : 1 {false}; |
| 534 | |
| 535 | bool __weekday_name_ : 1 {false}; |
| 536 | bool __weekday_ : 1 {false}; |
| 537 | |
| 538 | bool __day_of_year_ : 1 {false}; |
| 539 | bool __week_of_year_ : 1 {false}; |
| 540 | |
| 541 | bool __month_name_ : 1 {false}; |
| 542 | |
| 543 | uint8_t __reserved_0_ : 2 {0}; |
| 544 | uint8_t __reserved_1_ : 6 {0}; |
| 545 | // These two flags are only used internally and not part of the |
| 546 | // __parsed_specifications. Therefore put them at the end. |
| 547 | bool __width_as_arg_ : 1 {false}; |
| 548 | bool __precision_as_arg_ : 1 {false}; |
| 549 | |
| 550 | /// The requested width, either the value or the arg-id. |
| 551 | int32_t __width_{0}; |
| 552 | |
| 553 | /// The requested precision, either the value or the arg-id. |
| 554 | int32_t __precision_{-1}; |
| 555 | |
| 556 | __code_point<_CharT> __fill_{}; |
| 557 | |
| 558 | private: |
| 559 | _LIBCPP_HIDE_FROM_ABI constexpr bool __parse_alignment(_CharT __c) { |
| 560 | switch (__c) { |
| 561 | case _CharT('<'): |
| 562 | __alignment_ = __alignment::__left; |
| 563 | return true; |
| 564 | |
| 565 | case _CharT('^'): |
| 566 | __alignment_ = __alignment::__center; |
| 567 | return true; |
| 568 | |
| 569 | case _CharT('>'): |
| 570 | __alignment_ = __alignment::__right; |
| 571 | return true; |
| 572 | } |
| 573 | return false; |
| 574 | } |
| 575 | |
| 576 | _LIBCPP_HIDE_FROM_ABI constexpr void __validate_fill_character(_CharT __fill) { |
| 577 | // The forbidden fill characters all code points formed from a single code unit, thus the |
| 578 | // check can be omitted when more code units are used. |
| 579 | if (__fill == _CharT('{')) |
| 580 | std::__throw_format_error("The fill option contains an invalid value"); |
| 581 | } |
| 582 | |
| 583 | # if _LIBCPP_HAS_UNICODE |
| 584 | // range-fill and tuple-fill are identical |
| 585 | template <contiguous_iterator _Iterator> |
| 586 | requires same_as<_CharT, char> |
| 587 | # if _LIBCPP_HAS_WIDE_CHARACTERS |
| 588 | || (same_as<_CharT, wchar_t> && sizeof(wchar_t) == 2) |
| 589 | # endif |
| 590 | _LIBCPP_HIDE_FROM_ABI constexpr bool __parse_fill_align(_Iterator& __begin, _Iterator __end) { |
| 591 | _LIBCPP_ASSERT_VALID_ELEMENT_ACCESS( |
| 592 | __begin != __end, |
| 593 | "when called with an empty input the function will cause " |
| 594 | "undefined behavior by evaluating data not in the input"); |
| 595 | __unicode::__code_point_view<_CharT> __view{__begin, __end}; |
| 596 | __unicode::__consume_result __consumed = __view.__consume(); |
| 597 | if (__consumed.__status != __unicode::__consume_result::__ok) |
| 598 | std::__throw_format_error("The format specifier contains malformed Unicode characters"); |
| 599 | |
| 600 | if (__view.__position() < __end && __parse_alignment(*__view.__position())) { |
| 601 | ptrdiff_t __code_units = __view.__position() - __begin; |
| 602 | if (__code_units == 1) |
| 603 | // The forbidden fill characters all are code points encoded |
| 604 | // in one code unit, thus the check can be omitted when more |
| 605 | // code units are used. |
| 606 | __validate_fill_character(*__begin); |
| 607 | |
| 608 | std::copy_n(__begin, __code_units, std::addressof(__fill_.__data[0])); |
| 609 | __begin += __code_units + 1; |
| 610 | return true; |
| 611 | } |
| 612 | |
| 613 | if (!__parse_alignment(*__begin)) |
| 614 | return false; |
| 615 | |
| 616 | ++__begin; |
| 617 | return true; |
| 618 | } |
| 619 | |
| 620 | # if _LIBCPP_HAS_WIDE_CHARACTERS |
| 621 | template <contiguous_iterator _Iterator> |
| 622 | requires(same_as<_CharT, wchar_t> && sizeof(wchar_t) == 4) |
| 623 | _LIBCPP_HIDE_FROM_ABI constexpr bool __parse_fill_align(_Iterator& __begin, _Iterator __end) { |
| 624 | _LIBCPP_ASSERT_VALID_ELEMENT_ACCESS( |
| 625 | __begin != __end, |
| 626 | "when called with an empty input the function will cause " |
| 627 | "undefined behavior by evaluating data not in the input"); |
| 628 | if (__begin + 1 != __end && __parse_alignment(*(__begin + 1))) { |
| 629 | if (!__unicode::__is_scalar_value(*__begin)) |
| 630 | std::__throw_format_error("The fill option contains an invalid value"); |
| 631 | |
| 632 | __validate_fill_character(*__begin); |
| 633 | |
| 634 | __fill_.__data[0] = *__begin; |
| 635 | __begin += 2; |
| 636 | return true; |
| 637 | } |
| 638 | |
| 639 | if (!__parse_alignment(*__begin)) |
| 640 | return false; |
| 641 | |
| 642 | ++__begin; |
| 643 | return true; |
| 644 | } |
| 645 | |
| 646 | # endif // _LIBCPP_HAS_WIDE_CHARACTERS |
| 647 | |
| 648 | # else // _LIBCPP_HAS_UNICODE |
| 649 | // range-fill and tuple-fill are identical |
| 650 | template <contiguous_iterator _Iterator> |
| 651 | _LIBCPP_HIDE_FROM_ABI constexpr bool __parse_fill_align(_Iterator& __begin, _Iterator __end) { |
| 652 | _LIBCPP_ASSERT_VALID_ELEMENT_ACCESS( |
| 653 | __begin != __end, |
| 654 | "when called with an empty input the function will cause " |
| 655 | "undefined behavior by evaluating data not in the input"); |
| 656 | if (__begin + 1 != __end) { |
| 657 | if (__parse_alignment(*(__begin + 1))) { |
| 658 | __validate_fill_character(*__begin); |
| 659 | |
| 660 | __fill_.__data[0] = *__begin; |
| 661 | __begin += 2; |
| 662 | return true; |
| 663 | } |
| 664 | } |
| 665 | |
| 666 | if (!__parse_alignment(*__begin)) |
| 667 | return false; |
| 668 | |
| 669 | ++__begin; |
| 670 | return true; |
| 671 | } |
| 672 | |
| 673 | # endif // _LIBCPP_HAS_UNICODE |
| 674 | |
| 675 | template <contiguous_iterator _Iterator> |
| 676 | _LIBCPP_HIDE_FROM_ABI constexpr bool __parse_sign(_Iterator& __begin) { |
| 677 | switch (*__begin) { |
| 678 | case _CharT('-'): |
| 679 | __sign_ = __sign::__minus; |
| 680 | break; |
| 681 | case _CharT('+'): |
| 682 | __sign_ = __sign::__plus; |
| 683 | break; |
| 684 | case _CharT(' '): |
| 685 | __sign_ = __sign::__space; |
| 686 | break; |
| 687 | default: |
| 688 | return false; |
| 689 | } |
| 690 | ++__begin; |
| 691 | return true; |
| 692 | } |
| 693 | |
| 694 | template <contiguous_iterator _Iterator> |
| 695 | _LIBCPP_HIDE_FROM_ABI constexpr bool __parse_alternate_form(_Iterator& __begin) { |
| 696 | if (*__begin != _CharT('#')) |
| 697 | return false; |
| 698 | |
| 699 | __alternate_form_ = true; |
| 700 | ++__begin; |
| 701 | return true; |
| 702 | } |
| 703 | |
| 704 | template <contiguous_iterator _Iterator> |
| 705 | _LIBCPP_HIDE_FROM_ABI constexpr bool __parse_zero_padding(_Iterator& __begin) { |
| 706 | if (*__begin != _CharT('0')) |
| 707 | return false; |
| 708 | |
| 709 | if (__alignment_ == __alignment::__default) |
| 710 | __alignment_ = __alignment::__zero_padding; |
| 711 | ++__begin; |
| 712 | return true; |
| 713 | } |
| 714 | |
| 715 | template <contiguous_iterator _Iterator> |
| 716 | _LIBCPP_HIDE_FROM_ABI constexpr bool __parse_width(_Iterator& __begin, _Iterator __end, auto& __ctx) { |
| 717 | if (*__begin == _CharT('0')) |
| 718 | std::__throw_format_error("The width option should not have a leading zero"); |
| 719 | |
| 720 | if (*__begin == _CharT('{')) { |
| 721 | __format::__parse_number_result __r = __format_spec::__parse_arg_id(++__begin, __end, __ctx); |
| 722 | __width_as_arg_ = true; |
| 723 | __width_ = __r.__value; |
| 724 | __begin = __r.__last; |
| 725 | return true; |
| 726 | } |
| 727 | |
| 728 | if (*__begin < _CharT('0') || *__begin > _CharT('9')) |
| 729 | return false; |
| 730 | |
| 731 | __format::__parse_number_result __r = __format::__parse_number(__begin, __end); |
| 732 | __width_ = __r.__value; |
| 733 | _LIBCPP_ASSERT_INTERNAL(__width_ != 0, |
| 734 | "A zero value isn't allowed and should be impossible, " |
| 735 | "due to validations in this function"); |
| 736 | __begin = __r.__last; |
| 737 | return true; |
| 738 | } |
| 739 | |
| 740 | template <contiguous_iterator _Iterator> |
| 741 | _LIBCPP_HIDE_FROM_ABI constexpr bool __parse_precision(_Iterator& __begin, _Iterator __end, auto& __ctx) { |
| 742 | if (*__begin != _CharT('.')) |
| 743 | return false; |
| 744 | |
| 745 | ++__begin; |
| 746 | if (__begin == __end) |
| 747 | std::__throw_format_error("End of input while parsing format specifier precision"); |
| 748 | |
| 749 | if (*__begin == _CharT('{')) { |
| 750 | __format::__parse_number_result __arg_id = __format_spec::__parse_arg_id(++__begin, __end, __ctx); |
| 751 | __precision_as_arg_ = true; |
| 752 | __precision_ = __arg_id.__value; |
| 753 | __begin = __arg_id.__last; |
| 754 | return true; |
| 755 | } |
| 756 | |
| 757 | if (*__begin < _CharT('0') || *__begin > _CharT('9')) |
| 758 | std::__throw_format_error("The precision option does not contain a value or an argument index"); |
| 759 | |
| 760 | __format::__parse_number_result __r = __format::__parse_number(__begin, __end); |
| 761 | __precision_ = __r.__value; |
| 762 | __precision_as_arg_ = false; |
| 763 | __begin = __r.__last; |
| 764 | return true; |
| 765 | } |
| 766 | |
| 767 | template <contiguous_iterator _Iterator> |
| 768 | _LIBCPP_HIDE_FROM_ABI constexpr bool __parse_locale_specific_form(_Iterator& __begin) { |
| 769 | if (*__begin != _CharT('L')) |
| 770 | return false; |
| 771 | |
| 772 | __locale_specific_form_ = true; |
| 773 | ++__begin; |
| 774 | return true; |
| 775 | } |
| 776 | |
| 777 | template <contiguous_iterator _Iterator> |
| 778 | _LIBCPP_HIDE_FROM_ABI constexpr bool __parse_clear_brackets(_Iterator& __begin) { |
| 779 | if (*__begin != _CharT('n')) |
| 780 | return false; |
| 781 | |
| 782 | __clear_brackets_ = true; |
| 783 | ++__begin; |
| 784 | return true; |
| 785 | } |
| 786 | |
| 787 | template <contiguous_iterator _Iterator> |
| 788 | _LIBCPP_HIDE_FROM_ABI constexpr void __parse_type(_Iterator& __begin) { |
| 789 | // Determines the type. It does not validate whether the selected type is |
| 790 | // valid. Most formatters have optional fields that are only allowed for |
| 791 | // certain types. These parsers need to do validation after the type has |
| 792 | // been parsed. So its easier to implement the validation for all types in |
| 793 | // the specific parse function. |
| 794 | switch (*__begin) { |
| 795 | case 'A': |
| 796 | __type_ = __type::__hexfloat_upper_case; |
| 797 | break; |
| 798 | case 'B': |
| 799 | __type_ = __type::__binary_upper_case; |
| 800 | break; |
| 801 | case 'E': |
| 802 | __type_ = __type::__scientific_upper_case; |
| 803 | break; |
| 804 | case 'F': |
| 805 | __type_ = __type::__fixed_upper_case; |
| 806 | break; |
| 807 | case 'G': |
| 808 | __type_ = __type::__general_upper_case; |
| 809 | break; |
| 810 | case 'X': |
| 811 | __type_ = __type::__hexadecimal_upper_case; |
| 812 | break; |
| 813 | case 'a': |
| 814 | __type_ = __type::__hexfloat_lower_case; |
| 815 | break; |
| 816 | case 'b': |
| 817 | __type_ = __type::__binary_lower_case; |
| 818 | break; |
| 819 | case 'c': |
| 820 | __type_ = __type::__char; |
| 821 | break; |
| 822 | case 'd': |
| 823 | __type_ = __type::__decimal; |
| 824 | break; |
| 825 | case 'e': |
| 826 | __type_ = __type::__scientific_lower_case; |
| 827 | break; |
| 828 | case 'f': |
| 829 | __type_ = __type::__fixed_lower_case; |
| 830 | break; |
| 831 | case 'g': |
| 832 | __type_ = __type::__general_lower_case; |
| 833 | break; |
| 834 | case 'o': |
| 835 | __type_ = __type::__octal; |
| 836 | break; |
| 837 | case 'p': |
| 838 | __type_ = __type::__pointer_lower_case; |
| 839 | break; |
| 840 | case 'P': |
| 841 | __type_ = __type::__pointer_upper_case; |
| 842 | break; |
| 843 | case 's': |
| 844 | __type_ = __type::__string; |
| 845 | break; |
| 846 | case 'x': |
| 847 | __type_ = __type::__hexadecimal_lower_case; |
| 848 | break; |
| 849 | # if _LIBCPP_STD_VER >= 23 |
| 850 | case '?': |
| 851 | __type_ = __type::__debug; |
| 852 | break; |
| 853 | # endif |
| 854 | default: |
| 855 | return; |
| 856 | } |
| 857 | ++__begin; |
| 858 | } |
| 859 | |
| 860 | _LIBCPP_HIDE_FROM_ABI int32_t __get_width(auto& __ctx) const { |
| 861 | if (!__width_as_arg_) |
| 862 | return __width_; |
| 863 | |
| 864 | return __format_spec::__substitute_arg_id(__ctx.arg(__width_)); |
| 865 | } |
| 866 | |
| 867 | _LIBCPP_HIDE_FROM_ABI int32_t __get_precision(auto& __ctx) const { |
| 868 | if (!__precision_as_arg_) |
| 869 | return __precision_; |
| 870 | |
| 871 | return __format_spec::__substitute_arg_id(__ctx.arg(__precision_)); |
| 872 | } |
| 873 | }; |
| 874 | |
| 875 | // Validates whether the reserved bitfields don't change the size. |
| 876 | static_assert(sizeof(__parser<char>) == 16); |
| 877 | # if _LIBCPP_HAS_WIDE_CHARACTERS |
| 878 | static_assert(sizeof(__parser<wchar_t>) == 16); |
| 879 | # endif |
| 880 | |
| 881 | _LIBCPP_HIDE_FROM_ABI constexpr void __process_display_type_string(__format_spec::__type __type) { |
| 882 | switch (__type) { |
| 883 | case __format_spec::__type::__default: |
| 884 | case __format_spec::__type::__string: |
| 885 | case __format_spec::__type::__debug: |
| 886 | break; |
| 887 | |
| 888 | default: |
| 889 | std::__throw_format_error("The type option contains an invalid value for a string formatting argument"); |
| 890 | } |
| 891 | } |
| 892 | |
| 893 | template <class _CharT> |
| 894 | _LIBCPP_HIDE_FROM_ABI constexpr void __process_display_type_bool_string(__parser<_CharT>& __parser, const char* __id) { |
| 895 | __parser.__validate(__format_spec::__fields_bool, __id); |
| 896 | if (__parser.__alignment_ == __alignment::__default) |
| 897 | __parser.__alignment_ = __alignment::__left; |
| 898 | } |
| 899 | |
| 900 | template <class _CharT> |
| 901 | _LIBCPP_HIDE_FROM_ABI constexpr void __process_display_type_char(__parser<_CharT>& __parser, const char* __id) { |
| 902 | __format_spec::__process_display_type_bool_string(__parser, __id); |
| 903 | } |
| 904 | |
| 905 | template <class _CharT> |
| 906 | _LIBCPP_HIDE_FROM_ABI constexpr void __process_parsed_bool(__parser<_CharT>& __parser, const char* __id) { |
| 907 | switch (__parser.__type_) { |
| 908 | case __format_spec::__type::__default: |
| 909 | case __format_spec::__type::__string: |
| 910 | __format_spec::__process_display_type_bool_string(__parser, __id); |
| 911 | break; |
| 912 | |
| 913 | case __format_spec::__type::__binary_lower_case: |
| 914 | case __format_spec::__type::__binary_upper_case: |
| 915 | case __format_spec::__type::__octal: |
| 916 | case __format_spec::__type::__decimal: |
| 917 | case __format_spec::__type::__hexadecimal_lower_case: |
| 918 | case __format_spec::__type::__hexadecimal_upper_case: |
| 919 | break; |
| 920 | |
| 921 | default: |
| 922 | __format_spec::__throw_invalid_type_format_error(__id); |
| 923 | } |
| 924 | } |
| 925 | |
| 926 | template <class _CharT> |
| 927 | _LIBCPP_HIDE_FROM_ABI constexpr void __process_parsed_char(__parser<_CharT>& __parser, const char* __id) { |
| 928 | switch (__parser.__type_) { |
| 929 | case __format_spec::__type::__default: |
| 930 | case __format_spec::__type::__char: |
| 931 | case __format_spec::__type::__debug: |
| 932 | __format_spec::__process_display_type_char(__parser, __id); |
| 933 | break; |
| 934 | |
| 935 | case __format_spec::__type::__binary_lower_case: |
| 936 | case __format_spec::__type::__binary_upper_case: |
| 937 | case __format_spec::__type::__octal: |
| 938 | case __format_spec::__type::__decimal: |
| 939 | case __format_spec::__type::__hexadecimal_lower_case: |
| 940 | case __format_spec::__type::__hexadecimal_upper_case: |
| 941 | break; |
| 942 | |
| 943 | default: |
| 944 | __format_spec::__throw_invalid_type_format_error(__id); |
| 945 | } |
| 946 | } |
| 947 | |
| 948 | template <class _CharT> |
| 949 | _LIBCPP_HIDE_FROM_ABI constexpr void __process_parsed_integer(__parser<_CharT>& __parser, const char* __id) { |
| 950 | switch (__parser.__type_) { |
| 951 | case __format_spec::__type::__default: |
| 952 | case __format_spec::__type::__binary_lower_case: |
| 953 | case __format_spec::__type::__binary_upper_case: |
| 954 | case __format_spec::__type::__octal: |
| 955 | case __format_spec::__type::__decimal: |
| 956 | case __format_spec::__type::__hexadecimal_lower_case: |
| 957 | case __format_spec::__type::__hexadecimal_upper_case: |
| 958 | break; |
| 959 | |
| 960 | case __format_spec::__type::__char: |
| 961 | __format_spec::__process_display_type_char(__parser, __id); |
| 962 | break; |
| 963 | |
| 964 | default: |
| 965 | __format_spec::__throw_invalid_type_format_error(__id); |
| 966 | } |
| 967 | } |
| 968 | |
| 969 | template <class _CharT> |
| 970 | _LIBCPP_HIDE_FROM_ABI constexpr void __process_parsed_floating_point(__parser<_CharT>& __parser, const char* __id) { |
| 971 | switch (__parser.__type_) { |
| 972 | case __format_spec::__type::__default: |
| 973 | case __format_spec::__type::__hexfloat_lower_case: |
| 974 | case __format_spec::__type::__hexfloat_upper_case: |
| 975 | // Precision specific behavior will be handled later. |
| 976 | break; |
| 977 | case __format_spec::__type::__scientific_lower_case: |
| 978 | case __format_spec::__type::__scientific_upper_case: |
| 979 | case __format_spec::__type::__fixed_lower_case: |
| 980 | case __format_spec::__type::__fixed_upper_case: |
| 981 | case __format_spec::__type::__general_lower_case: |
| 982 | case __format_spec::__type::__general_upper_case: |
| 983 | if (!__parser.__precision_as_arg_ && __parser.__precision_ == -1) |
| 984 | // Set the default precision for the call to to_chars. |
| 985 | __parser.__precision_ = 6; |
| 986 | break; |
| 987 | |
| 988 | default: |
| 989 | __format_spec::__throw_invalid_type_format_error(__id); |
| 990 | } |
| 991 | } |
| 992 | |
| 993 | _LIBCPP_HIDE_FROM_ABI constexpr void __process_display_type_pointer(__format_spec::__type __type, const char* __id) { |
| 994 | switch (__type) { |
| 995 | case __format_spec::__type::__default: |
| 996 | case __format_spec::__type::__pointer_lower_case: |
| 997 | case __format_spec::__type::__pointer_upper_case: |
| 998 | break; |
| 999 | |
| 1000 | default: |
| 1001 | __format_spec::__throw_invalid_type_format_error(__id); |
| 1002 | } |
| 1003 | } |
| 1004 | |
| 1005 | template <contiguous_iterator _Iterator> |
| 1006 | struct __column_width_result { |
| 1007 | /// The number of output columns. |
| 1008 | size_t __width_; |
| 1009 | /// One beyond the last code unit used in the estimation. |
| 1010 | /// |
| 1011 | /// This limits the original output to fit in the wanted number of columns. |
| 1012 | _Iterator __last_; |
| 1013 | }; |
| 1014 | |
| 1015 | template <contiguous_iterator _Iterator> |
| 1016 | __column_width_result(size_t, _Iterator) -> __column_width_result<_Iterator>; |
| 1017 | |
| 1018 | /// Since a column width can be two it's possible that the requested column |
| 1019 | /// width can't be achieved. Depending on the intended usage the policy can be |
| 1020 | /// selected. |
| 1021 | /// - When used as precision the maximum width may not be exceeded and the |
| 1022 | /// result should be "rounded down" to the previous boundary. |
| 1023 | /// - When used as a width we're done once the minimum is reached, but |
| 1024 | /// exceeding is not an issue. Rounding down is an issue since that will |
| 1025 | /// result in writing fill characters. Therefore the result needs to be |
| 1026 | /// "rounded up". |
| 1027 | enum class __column_width_rounding { __down, __up }; |
| 1028 | |
| 1029 | # if _LIBCPP_HAS_UNICODE |
| 1030 | |
| 1031 | namespace __detail { |
| 1032 | template <contiguous_iterator _Iterator> |
| 1033 | _LIBCPP_HIDE_FROM_ABI constexpr __column_width_result<_Iterator> __estimate_column_width_grapheme_clustering( |
| 1034 | _Iterator __first, _Iterator __last, size_t __maximum, __column_width_rounding __rounding) noexcept { |
| 1035 | using _CharT = iter_value_t<_Iterator>; |
| 1036 | __unicode::__extended_grapheme_cluster_view<_CharT> __view{__first, __last}; |
| 1037 | |
| 1038 | __column_width_result<_Iterator> __result{0, __first}; |
| 1039 | while (__result.__last_ != __last && __result.__width_ <= __maximum) { |
| 1040 | typename __unicode::__extended_grapheme_cluster_view<_CharT>::__cluster __cluster = __view.__consume(); |
| 1041 | int __width = __width_estimation_table::__estimated_width(__cluster.__code_point_); |
| 1042 | |
| 1043 | // When the next entry would exceed the maximum width the previous width |
| 1044 | // might be returned. For example when a width of 100 is requested the |
| 1045 | // returned width might be 99, since the next code point has an estimated |
| 1046 | // column width of 2. This depends on the rounding flag. |
| 1047 | // When the maximum is exceeded the loop will abort the next iteration. |
| 1048 | if (__rounding == __column_width_rounding::__down && __result.__width_ + __width > __maximum) |
| 1049 | return __result; |
| 1050 | |
| 1051 | __result.__width_ += __width; |
| 1052 | __result.__last_ = __cluster.__last_; |
| 1053 | } |
| 1054 | |
| 1055 | return __result; |
| 1056 | } |
| 1057 | |
| 1058 | } // namespace __detail |
| 1059 | |
| 1060 | // Unicode can be stored in several formats: UTF-8, UTF-16, and UTF-32. |
| 1061 | // Depending on format the relation between the number of code units stored and |
| 1062 | // the number of output columns differs. The first relation is the number of |
| 1063 | // code units forming a code point. (The text assumes the code units are |
| 1064 | // unsigned.) |
| 1065 | // - UTF-8 The number of code units is between one and four. The first 127 |
| 1066 | // Unicode code points match the ASCII character set. When the highest bit is |
| 1067 | // set it means the code point has more than one code unit. |
| 1068 | // - UTF-16: The number of code units is between 1 and 2. When the first |
| 1069 | // code unit is in the range [0xd800,0xdfff) it means the code point uses two |
| 1070 | // code units. |
| 1071 | // - UTF-32: The number of code units is always one. |
| 1072 | // |
| 1073 | // The code point to the number of columns is specified in |
| 1074 | // [format.string.std]/11. This list might change in the future. |
| 1075 | // |
| 1076 | // Another thing to be taken into account is Grapheme clustering. This means |
| 1077 | // that in some cases multiple code points are combined one element in the |
| 1078 | // output. For example: |
| 1079 | // - an ASCII character with a combined diacritical mark |
| 1080 | // - an emoji with a skin tone modifier |
| 1081 | // - a group of combined people emoji to create a family |
| 1082 | // - a combination of flag emoji |
| 1083 | // |
| 1084 | // See also: |
| 1085 | // - [format.string.general]/11 |
| 1086 | // - https://en.wikipedia.org/wiki/UTF-8#Encoding |
| 1087 | // - https://en.wikipedia.org/wiki/UTF-16#U+D800_to_U+DFFF |
| 1088 | |
| 1089 | _LIBCPP_HIDE_FROM_ABI constexpr bool __is_ascii(char32_t __c) { return __c < 0x80; } |
| 1090 | |
| 1091 | /// Determines the number of output columns needed to render the input. |
| 1092 | /// |
| 1093 | /// \note When the scanner encounters malformed Unicode it acts as-if every |
| 1094 | /// code unit is a one column code point. Typically a terminal uses the same |
| 1095 | /// strategy and replaces every malformed code unit with a one column |
| 1096 | /// replacement character. |
| 1097 | /// |
| 1098 | /// \param __first Points to the first element of the input range. |
| 1099 | /// \param __last Points beyond the last element of the input range. |
| 1100 | /// \param __maximum The maximum number of output columns. The returned number |
| 1101 | /// of estimated output columns will not exceed this value. |
| 1102 | /// \param __rounding Selects the rounding method. |
| 1103 | /// \c __down result.__width_ <= __maximum |
| 1104 | /// \c __up result.__width_ <= __maximum + 1 |
| 1105 | template <class _CharT, class _Iterator = typename basic_string_view<_CharT>::const_iterator> |
| 1106 | _LIBCPP_HIDE_FROM_ABI constexpr __column_width_result<_Iterator> __estimate_column_width( |
| 1107 | basic_string_view<_CharT> __str, size_t __maximum, __column_width_rounding __rounding) noexcept { |
| 1108 | // The width estimation is done in two steps: |
| 1109 | // - Quickly process for the ASCII part. ASCII has the following properties |
| 1110 | // - One code unit is one code point |
| 1111 | // - Every code point has an estimated width of one |
| 1112 | // - When needed it will a Unicode Grapheme clustering algorithm to find |
| 1113 | // the proper place for truncation. |
| 1114 | |
| 1115 | if (__str.empty() || __maximum == 0) |
| 1116 | return {0, __str.begin()}; |
| 1117 | |
| 1118 | // ASCII has one caveat; when an ASCII character is followed by a non-ASCII |
| 1119 | // character they might be part of an extended grapheme cluster. For example: |
| 1120 | // an ASCII letter and a COMBINING ACUTE ACCENT |
| 1121 | // The truncate should happen after the COMBINING ACUTE ACCENT. Therefore we |
| 1122 | // need to scan one code unit beyond the requested precision. When this code |
| 1123 | // unit is non-ASCII we omit the current code unit and let the Grapheme |
| 1124 | // clustering algorithm do its work. |
| 1125 | auto __it = __str.begin(); |
| 1126 | if (__format_spec::__is_ascii(*__it)) { |
| 1127 | do { |
| 1128 | --__maximum; |
| 1129 | ++__it; |
| 1130 | if (__it == __str.end()) |
| 1131 | return {__str.size(), __str.end()}; |
| 1132 | |
| 1133 | if (__maximum == 0) { |
| 1134 | if (__format_spec::__is_ascii(*__it)) |
| 1135 | return {static_cast<size_t>(__it - __str.begin()), __it}; |
| 1136 | |
| 1137 | break; |
| 1138 | } |
| 1139 | } while (__format_spec::__is_ascii(*__it)); |
| 1140 | --__it; |
| 1141 | ++__maximum; |
| 1142 | } |
| 1143 | |
| 1144 | ptrdiff_t __ascii_size = __it - __str.begin(); |
| 1145 | __column_width_result __result = |
| 1146 | __detail::__estimate_column_width_grapheme_clustering(__it, __str.end(), __maximum, __rounding); |
| 1147 | |
| 1148 | __result.__width_ += __ascii_size; |
| 1149 | return __result; |
| 1150 | } |
| 1151 | # else // _LIBCPP_HAS_UNICODE |
| 1152 | template <class _CharT> |
| 1153 | _LIBCPP_HIDE_FROM_ABI constexpr __column_width_result<typename basic_string_view<_CharT>::const_iterator> |
| 1154 | __estimate_column_width(basic_string_view<_CharT> __str, size_t __maximum, __column_width_rounding) noexcept { |
| 1155 | // When Unicode isn't supported assume ASCII and every code unit is one code |
| 1156 | // point. In ASCII the estimated column width is always one. Thus there's no |
| 1157 | // need for rounding. |
| 1158 | size_t __width = std::min(__str.size(), __maximum); |
| 1159 | return {__width, __str.begin() + __width}; |
| 1160 | } |
| 1161 | |
| 1162 | # endif // _LIBCPP_HAS_UNICODE |
| 1163 | |
| 1164 | } // namespace __format_spec |
| 1165 | |
| 1166 | #endif // _LIBCPP_STD_VER >= 20 |
| 1167 | |
| 1168 | _LIBCPP_END_NAMESPACE_STD |
| 1169 | |
| 1170 | _LIBCPP_POP_MACROS |
| 1171 | |
| 1172 | #endif // _LIBCPP___FORMAT_PARSER_STD_FORMAT_SPEC_H |
| 1173 |
Warning: This file is not a C or C++ file. It does not have highlighting.
