libstdc++
charconv
Go to the documentation of this file.
1 // Primitive numeric conversions (to_chars and from_chars) -*- C++ -*-
2 
3 // Copyright (C) 2017-2020 Free Software Foundation, Inc.
4 //
5 // This file is part of the GNU ISO C++ Library. This library is free
6 // software; you can redistribute it and/or modify it under the
7 // terms of the GNU General Public License as published by the
8 // Free Software Foundation; either version 3, or (at your option)
9 // any later version.
10 
11 // This library is distributed in the hope that it will be useful,
12 // but WITHOUT ANY WARRANTY; without even the implied warranty of
13 // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 // GNU General Public License for more details.
15 
16 // Under Section 7 of GPL version 3, you are granted additional
17 // permissions described in the GCC Runtime Library Exception, version
18 // 3.1, as published by the Free Software Foundation.
19 
20 // You should have received a copy of the GNU General Public License and
21 // a copy of the GCC Runtime Library Exception along with this program;
22 // see the files COPYING3 and COPYING.RUNTIME respectively. If not, see
23 // <http://www.gnu.org/licenses/>.
24 
25 /** @file include/charconv
26  * This is a Standard C++ Library header.
27  */
28 
29 #ifndef _GLIBCXX_CHARCONV
30 #define _GLIBCXX_CHARCONV 1
31 
32 #pragma GCC system_header
33 
34 // As an extension we support <charconv> in C++14, but this header should not
35 // be included by any other library headers in C++14 mode. This ensures that
36 // the names defined in this header are not added to namespace std unless a
37 // user explicitly includes <charconv> in C++14 code.
38 #if __cplusplus >= 201402L
39 
40 #include <type_traits>
41 #include <bit> // for __bit_width
42 #include <cctype> // for isdigit
43 #include <bits/charconv.h> // for __to_chars_len, __to_chars_10_impl
44 #include <bits/error_constants.h> // for std::errc
45 #include <bits/int_limits.h>
46 
47 // FIXME: Define when floating point is supported:
48 // #define __cpp_lib_to_chars 201611L
49 
50 namespace std _GLIBCXX_VISIBILITY(default)
51 {
52 _GLIBCXX_BEGIN_NAMESPACE_VERSION
53 
54  /// Result type of std::to_chars
55  struct to_chars_result
56  {
57  char* ptr;
58  errc ec;
59 
60 #if __cplusplus > 201703L && __cpp_impl_three_way_comparison >= 201907L
61  friend bool
62  operator==(const to_chars_result&, const to_chars_result&) = default;
63 #endif
64  };
65 
66  /// Result type of std::from_chars
67  struct from_chars_result
68  {
69  const char* ptr;
70  errc ec;
71 
72 #if __cplusplus > 201703L && __cpp_impl_three_way_comparison >= 201907L
73  friend bool
74  operator==(const from_chars_result&, const from_chars_result&) = default;
75 #endif
76  };
77 
78 namespace __detail
79 {
80  template<typename _Tp>
81  using __integer_to_chars_result_type
82  = enable_if_t<__or_<__is_signed_integer<_Tp>,
83  __is_unsigned_integer<_Tp>,
84  is_same<char, remove_cv_t<_Tp>>>::value,
85  to_chars_result>;
86 
87  // Pick an unsigned type of suitable size. This is used to reduce the
88  // number of specializations of __to_chars_len, __to_chars etc. that
89  // get instantiated. For example, to_chars<char> and to_chars<short>
90  // and to_chars<unsigned> will all use the same code, and so will
91  // to_chars<long> when sizeof(int) == sizeof(long).
92  template<typename _Tp>
93  struct __to_chars_unsigned_type : __make_unsigned_selector_base
94  {
95  using _UInts = _List<unsigned int, unsigned long, unsigned long long
96 #if _GLIBCXX_USE_INT128
97  , unsigned __int128
98 #endif
99  >;
100  using type = typename __select<sizeof(_Tp), _UInts>::__type;
101  };
102 
103  template<typename _Tp>
104  using __unsigned_least_t = typename __to_chars_unsigned_type<_Tp>::type;
105 
106  // Generic implementation for arbitrary bases.
107  // Defined in <bits/charconv.h>.
108  template<typename _Tp>
109  constexpr unsigned
110  __to_chars_len(_Tp __value, int __base /* = 10 */) noexcept;
111 
112  template<typename _Tp>
113  constexpr unsigned
114  __to_chars_len_2(_Tp __value) noexcept
115  { return std::__bit_width(__value); }
116 
117  // Generic implementation for arbitrary bases.
118  template<typename _Tp>
119  to_chars_result
120  __to_chars(char* __first, char* __last, _Tp __val, int __base) noexcept
121  {
122  static_assert(is_integral<_Tp>::value, "implementation bug");
123  static_assert(is_unsigned<_Tp>::value, "implementation bug");
124 
125  to_chars_result __res;
126 
127  const unsigned __len = __to_chars_len(__val, __base);
128 
129  if (__builtin_expect((__last - __first) < __len, 0))
130  {
131  __res.ptr = __last;
132  __res.ec = errc::value_too_large;
133  return __res;
134  }
135 
136  unsigned __pos = __len - 1;
137 
138  static constexpr char __digits[] = {
139  '0', '1', '2', '3', '4', '5', '6', '7', '8', '9',
140  'a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i', 'j',
141  'k', 'l', 'm', 'n', 'o', 'p', 'q', 'r', 's', 't',
142  'u', 'v', 'w', 'x', 'y', 'z'
143  };
144 
145  while (__val >= (unsigned)__base)
146  {
147  auto const __quo = __val / __base;
148  auto const __rem = __val % __base;
149  __first[__pos--] = __digits[__rem];
150  __val = __quo;
151  }
152  *__first = __digits[__val];
153 
154  __res.ptr = __first + __len;
155  __res.ec = {};
156  return __res;
157  }
158 
159  template<typename _Tp>
160  __integer_to_chars_result_type<_Tp>
161  __to_chars_16(char* __first, char* __last, _Tp __val) noexcept
162  {
163  static_assert(is_integral<_Tp>::value, "implementation bug");
164  static_assert(is_unsigned<_Tp>::value, "implementation bug");
165 
166  to_chars_result __res;
167 
168  const unsigned __len = (__to_chars_len_2(__val) + 3) / 4;
169 
170  if (__builtin_expect((__last - __first) < __len, 0))
171  {
172  __res.ptr = __last;
173  __res.ec = errc::value_too_large;
174  return __res;
175  }
176 
177  static constexpr char __digits[] = {
178  '0', '1', '2', '3', '4', '5', '6', '7', '8', '9',
179  'a', 'b', 'c', 'd', 'e', 'f'
180  };
181  unsigned __pos = __len - 1;
182  while (__val >= 0x100)
183  {
184  auto __num = __val & 0xF;
185  __val >>= 4;
186  __first[__pos] = __digits[__num];
187  __num = __val & 0xF;
188  __val >>= 4;
189  __first[__pos - 1] = __digits[__num];
190  __pos -= 2;
191  }
192  if (__val >= 0x10)
193  {
194  const auto __num = __val & 0xF;
195  __val >>= 4;
196  __first[1] = __digits[__num];
197  __first[0] = __digits[__val];
198  }
199  else
200  __first[0] = __digits[__val];
201  __res.ptr = __first + __len;
202  __res.ec = {};
203  return __res;
204  }
205 
206  template<typename _Tp>
207  inline __integer_to_chars_result_type<_Tp>
208  __to_chars_10(char* __first, char* __last, _Tp __val) noexcept
209  {
210  static_assert(is_integral<_Tp>::value, "implementation bug");
211  static_assert(is_unsigned<_Tp>::value, "implementation bug");
212 
213  to_chars_result __res;
214 
215  const unsigned __len = __to_chars_len(__val, 10);
216 
217  if (__builtin_expect((__last - __first) < __len, 0))
218  {
219  __res.ptr = __last;
220  __res.ec = errc::value_too_large;
221  return __res;
222  }
223 
224  __detail::__to_chars_10_impl(__first, __len, __val);
225  __res.ptr = __first + __len;
226  __res.ec = {};
227  return __res;
228  }
229 
230  template<typename _Tp>
231  __integer_to_chars_result_type<_Tp>
232  __to_chars_8(char* __first, char* __last, _Tp __val) noexcept
233  {
234  static_assert(is_integral<_Tp>::value, "implementation bug");
235  static_assert(is_unsigned<_Tp>::value, "implementation bug");
236 
237  to_chars_result __res;
238  unsigned __len;
239 
240  if _GLIBCXX17_CONSTEXPR (__detail::__int_limits<_Tp>::digits <= 16)
241  {
242  __len = __val > 077777u ? 6u
243  : __val > 07777u ? 5u
244  : __val > 0777u ? 4u
245  : __val > 077u ? 3u
246  : __val > 07u ? 2u
247  : 1u;
248  }
249  else
250  __len = (__to_chars_len_2(__val) + 2) / 3;
251 
252  if (__builtin_expect((__last - __first) < __len, 0))
253  {
254  __res.ptr = __last;
255  __res.ec = errc::value_too_large;
256  return __res;
257  }
258 
259  unsigned __pos = __len - 1;
260  while (__val >= 0100)
261  {
262  auto __num = __val & 7;
263  __val >>= 3;
264  __first[__pos] = '0' + __num;
265  __num = __val & 7;
266  __val >>= 3;
267  __first[__pos - 1] = '0' + __num;
268  __pos -= 2;
269  }
270  if (__val >= 010)
271  {
272  auto const __num = __val & 7;
273  __val >>= 3;
274  __first[1] = '0' + __num;
275  __first[0] = '0' + __val;
276  }
277  else
278  __first[0] = '0' + __val;
279  __res.ptr = __first + __len;
280  __res.ec = {};
281  return __res;
282  }
283 
284  template<typename _Tp>
285  __integer_to_chars_result_type<_Tp>
286  __to_chars_2(char* __first, char* __last, _Tp __val) noexcept
287  {
288  static_assert(is_integral<_Tp>::value, "implementation bug");
289  static_assert(is_unsigned<_Tp>::value, "implementation bug");
290 
291  to_chars_result __res;
292 
293  const unsigned __len = __to_chars_len_2(__val);
294 
295  if (__builtin_expect((__last - __first) < __len, 0))
296  {
297  __res.ptr = __last;
298  __res.ec = errc::value_too_large;
299  return __res;
300  }
301 
302  unsigned __pos = __len - 1;
303 
304  while (__pos)
305  {
306  __first[__pos--] = '0' + (__val & 1);
307  __val >>= 1;
308  }
309  // First digit is always '1' because __to_chars_len_2 skips
310  // leading zero bits and std::to_chars handles zero values
311  // directly.
312  __first[0] = '1';
313 
314  __res.ptr = __first + __len;
315  __res.ec = {};
316  return __res;
317  }
318 
319 } // namespace __detail
320 
321  template<typename _Tp>
322  __detail::__integer_to_chars_result_type<_Tp>
323  __to_chars_i(char* __first, char* __last, _Tp __value, int __base = 10)
324  {
325  __glibcxx_assert(2 <= __base && __base <= 36);
326 
327  using _Up = __detail::__unsigned_least_t<_Tp>;
328  _Up __unsigned_val = __value;
329 
330  if (__first == __last) [[__unlikely__]]
331  return { __last, errc::value_too_large };
332 
333  if (__value == 0)
334  {
335  *__first = '0';
336  return { __first + 1, errc{} };
337  }
338 
339  if _GLIBCXX17_CONSTEXPR (std::is_signed<_Tp>::value)
340  if (__value < 0)
341  {
342  if (__builtin_expect(__first != __last, 1))
343  *__first++ = '-';
344  __unsigned_val = _Up(~__value) + _Up(1);
345  }
346 
347  switch (__base)
348  {
349  case 16:
350  return __detail::__to_chars_16(__first, __last, __unsigned_val);
351  case 10:
352  return __detail::__to_chars_10(__first, __last, __unsigned_val);
353  case 8:
354  return __detail::__to_chars_8(__first, __last, __unsigned_val);
355  case 2:
356  return __detail::__to_chars_2(__first, __last, __unsigned_val);
357  default:
358  return __detail::__to_chars(__first, __last, __unsigned_val, __base);
359  }
360  }
361 
362 #define _GLIBCXX_TO_CHARS(T) \
363  inline to_chars_result \
364  to_chars(char* __first, char* __last, T __value, int __base = 10) \
365  { return std::__to_chars_i<T>(__first, __last, __value, __base); }
366 _GLIBCXX_TO_CHARS(char)
367 _GLIBCXX_TO_CHARS(signed char)
368 _GLIBCXX_TO_CHARS(unsigned char)
369 _GLIBCXX_TO_CHARS(signed short)
370 _GLIBCXX_TO_CHARS(unsigned short)
371 _GLIBCXX_TO_CHARS(signed int)
372 _GLIBCXX_TO_CHARS(unsigned int)
373 _GLIBCXX_TO_CHARS(signed long)
374 _GLIBCXX_TO_CHARS(unsigned long)
375 _GLIBCXX_TO_CHARS(signed long long)
376 _GLIBCXX_TO_CHARS(unsigned long long)
377 #if defined(__GLIBCXX_TYPE_INT_N_0)
378 _GLIBCXX_TO_CHARS(signed __GLIBCXX_TYPE_INT_N_0)
379 _GLIBCXX_TO_CHARS(unsigned __GLIBCXX_TYPE_INT_N_0)
380 #endif
381 #if defined(__GLIBCXX_TYPE_INT_N_1)
382 _GLIBCXX_TO_CHARS(signed __GLIBCXX_TYPE_INT_N_1)
383 _GLIBCXX_TO_CHARS(unsigned __GLIBCXX_TYPE_INT_N_1)
384 #endif
385 #if defined(__GLIBCXX_TYPE_INT_N_2)
386 _GLIBCXX_TO_CHARS(signed __GLIBCXX_TYPE_INT_N_2)
387 _GLIBCXX_TO_CHARS(unsigned __GLIBCXX_TYPE_INT_N_2)
388 #endif
389 #if defined(__GLIBCXX_TYPE_INT_N_3)
390 _GLIBCXX_TO_CHARS(signed __GLIBCXX_TYPE_INT_N_3)
391 _GLIBCXX_TO_CHARS(unsigned __GLIBCXX_TYPE_INT_N_3)
392 #endif
393 #undef _GLIBCXX_TO_CHARS
394 
395  // _GLIBCXX_RESOLVE_LIB_DEFECTS
396  // 3266. to_chars(bool) should be deleted
397  to_chars_result to_chars(char*, char*, bool, int = 10) = delete;
398 
399 namespace __detail
400 {
401  template<typename _Tp>
402  bool
403  __raise_and_add(_Tp& __val, int __base, unsigned char __c)
404  {
405  if (__builtin_mul_overflow(__val, __base, &__val)
406  || __builtin_add_overflow(__val, __c, &__val))
407  return false;
408  return true;
409  }
410 
411  /// std::from_chars implementation for integers in base 2.
412  template<typename _Tp>
413  bool
414  __from_chars_binary(const char*& __first, const char* __last, _Tp& __val)
415  {
416  static_assert(is_integral<_Tp>::value, "implementation bug");
417  static_assert(is_unsigned<_Tp>::value, "implementation bug");
418 
419  const ptrdiff_t __len = __last - __first;
420  ptrdiff_t __i = 0;
421  while (__i < __len && __first[__i] == '0')
422  ++__i;
423  const ptrdiff_t __leading_zeroes = __i;
424 
425  while (__i < __len)
426  {
427  const unsigned char __c = (unsigned)__first[__i] - '0';
428  if (__c < 2)
429  __val = (__val << 1) | __c;
430  else
431  break;
432  __i++;
433  }
434  __first += __i;
435  return (__i - __leading_zeroes) <= __detail::__int_limits<_Tp>::digits;
436  }
437 
438  /// std::from_chars implementation for integers in bases 3 to 10.
439  template<typename _Tp>
440  bool
441  __from_chars_digit(const char*& __first, const char* __last, _Tp& __val,
442  int __base)
443  {
444  static_assert(is_integral<_Tp>::value, "implementation bug");
445  static_assert(is_unsigned<_Tp>::value, "implementation bug");
446 
447  auto __matches = [__base](char __c) {
448  return '0' <= __c && __c <= ('0' + (__base - 1));
449  };
450 
451  while (__first != __last)
452  {
453  const char __c = *__first;
454  if (__matches(__c))
455  {
456  if (!__raise_and_add(__val, __base, __c - '0'))
457  {
458  while (++__first != __last && __matches(*__first))
459  ;
460  return false;
461  }
462  __first++;
463  }
464  else
465  return true;
466  }
467  return true;
468  }
469 
470  constexpr unsigned char
471  __from_chars_alpha_to_num(char __c)
472  {
473  switch (__c)
474  {
475  case 'a':
476  case 'A':
477  return 10;
478  case 'b':
479  case 'B':
480  return 11;
481  case 'c':
482  case 'C':
483  return 12;
484  case 'd':
485  case 'D':
486  return 13;
487  case 'e':
488  case 'E':
489  return 14;
490  case 'f':
491  case 'F':
492  return 15;
493  case 'g':
494  case 'G':
495  return 16;
496  case 'h':
497  case 'H':
498  return 17;
499  case 'i':
500  case 'I':
501  return 18;
502  case 'j':
503  case 'J':
504  return 19;
505  case 'k':
506  case 'K':
507  return 20;
508  case 'l':
509  case 'L':
510  return 21;
511  case 'm':
512  case 'M':
513  return 22;
514  case 'n':
515  case 'N':
516  return 23;
517  case 'o':
518  case 'O':
519  return 24;
520  case 'p':
521  case 'P':
522  return 25;
523  case 'q':
524  case 'Q':
525  return 26;
526  case 'r':
527  case 'R':
528  return 27;
529  case 's':
530  case 'S':
531  return 28;
532  case 't':
533  case 'T':
534  return 29;
535  case 'u':
536  case 'U':
537  return 30;
538  case 'v':
539  case 'V':
540  return 31;
541  case 'w':
542  case 'W':
543  return 32;
544  case 'x':
545  case 'X':
546  return 33;
547  case 'y':
548  case 'Y':
549  return 34;
550  case 'z':
551  case 'Z':
552  return 35;
553  }
554  return __detail::__int_limits<unsigned char>::max();
555  }
556 
557  /// std::from_chars implementation for integers in bases 11 to 26.
558  template<typename _Tp>
559  bool
560  __from_chars_alnum(const char*& __first, const char* __last, _Tp& __val,
561  int __base)
562  {
563  bool __valid = true;
564  while (__first != __last)
565  {
566  unsigned char __c = *__first;
567  if (std::isdigit(__c))
568  __c -= '0';
569  else
570  {
571  __c = __from_chars_alpha_to_num(__c);
572  if (__c >= __base)
573  break;
574  }
575 
576  if (__builtin_expect(__valid, 1))
577  __valid = __raise_and_add(__val, __base, __c);
578  __first++;
579  }
580  return __valid;
581  }
582 
583  template<typename _Tp>
584  using __integer_from_chars_result_type
585  = enable_if_t<__or_<__is_signed_integer<_Tp>,
586  __is_unsigned_integer<_Tp>,
587  is_same<char, remove_cv_t<_Tp>>>::value,
588  from_chars_result>;
589 
590 } // namespace __detail
591 
592  /// std::from_chars for integral types.
593  template<typename _Tp>
594  __detail::__integer_from_chars_result_type<_Tp>
595  from_chars(const char* __first, const char* __last, _Tp& __value,
596  int __base = 10)
597  {
598  __glibcxx_assert(2 <= __base && __base <= 36);
599 
600  from_chars_result __res{__first, {}};
601 
602  int __sign = 1;
603  if _GLIBCXX17_CONSTEXPR (std::is_signed<_Tp>::value)
604  if (__first != __last && *__first == '-')
605  {
606  __sign = -1;
607  ++__first;
608  }
609 
610  using _Up = __detail::__unsigned_least_t<_Tp>;
611  _Up __val = 0;
612 
613  const auto __start = __first;
614  bool __valid;
615  if (__base == 2)
616  __valid = __detail::__from_chars_binary(__first, __last, __val);
617  else if (__base <= 10)
618  __valid = __detail::__from_chars_digit(__first, __last, __val, __base);
619  else
620  __valid = __detail::__from_chars_alnum(__first, __last, __val, __base);
621 
622  if (__builtin_expect(__first == __start, 0))
623  __res.ec = errc::invalid_argument;
624  else
625  {
626  __res.ptr = __first;
627  if (!__valid)
628  __res.ec = errc::result_out_of_range;
629  else
630  {
631  if _GLIBCXX17_CONSTEXPR (std::is_signed<_Tp>::value)
632  {
633  _Tp __tmp;
634  if (__builtin_mul_overflow(__val, __sign, &__tmp))
635  __res.ec = errc::result_out_of_range;
636  else
637  __value = __tmp;
638  }
639  else
640  {
641  if _GLIBCXX17_CONSTEXPR (__detail::__int_limits<_Up>::max()
642  > __detail::__int_limits<_Tp>::max())
643  {
644  if (__val > __detail::__int_limits<_Tp>::max())
645  __res.ec = errc::result_out_of_range;
646  else
647  __value = __val;
648  }
649  else
650  __value = __val;
651  }
652  }
653  }
654  return __res;
655  }
656 
657  /// floating-point format for primitive numerical conversion
658  enum class chars_format
659  {
660  scientific = 1, fixed = 2, hex = 4, general = fixed | scientific
661  };
662 
663  constexpr chars_format
664  operator|(chars_format __lhs, chars_format __rhs) noexcept
665  { return (chars_format)((unsigned)__lhs | (unsigned)__rhs); }
666 
667  constexpr chars_format
668  operator&(chars_format __lhs, chars_format __rhs) noexcept
669  { return (chars_format)((unsigned)__lhs & (unsigned)__rhs); }
670 
671  constexpr chars_format
672  operator^(chars_format __lhs, chars_format __rhs) noexcept
673  { return (chars_format)((unsigned)__lhs ^ (unsigned)__rhs); }
674 
675  constexpr chars_format
676  operator~(chars_format __fmt) noexcept
677  { return (chars_format)~(unsigned)__fmt; }
678 
679  constexpr chars_format&
680  operator|=(chars_format& __lhs, chars_format __rhs) noexcept
681  { return __lhs = __lhs | __rhs; }
682 
683  constexpr chars_format&
684  operator&=(chars_format& __lhs, chars_format __rhs) noexcept
685  { return __lhs = __lhs & __rhs; }
686 
687  constexpr chars_format&
688  operator^=(chars_format& __lhs, chars_format __rhs) noexcept
689  { return __lhs = __lhs ^ __rhs; }
690 
691 _GLIBCXX_END_NAMESPACE_VERSION
692 } // namespace std
693 #endif // C++14
694 #endif // _GLIBCXX_CHARCONV