lib/c-strtod.c

   1 /* Convert string to floating-point number, using the C locale.
   2
   3    Copyright (C) 2003-2004, 2006, 2009-2024 Free Software Foundation, Inc.
   4
   5    This program is free software: you can redistribute it and/or modify
   6    it under the terms of the GNU General Public License as published by
   7    the Free Software Foundation, either version 3 of the License, or
   8    (at your option) any later version.
   9
  10    This program is distributed in the hope that it will be useful,
  11    but WITHOUT ANY WARRANTY; without even the implied warranty of
  12    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  13    GNU General Public License for more details.
  14
  15    You should have received a copy of the GNU General Public License
  16    along with this program.  If not, see <https://www.gnu.org/licenses/>.  */
  17
  18 /* Written by Paul Eggert and Bruno Haible.  */
  19
  20 #include <config.h>
  21
  22 #include "c-strtod.h"
  23
  24 #include <errno.h>
  25 #include <locale.h>
  26 #include <stdio.h>
  27 #include <stdlib.h>
  28 #include <string.h>
  29
  30 #if FLOAT
  31 # define C_STRTOD c_strtof
  32 # define DOUBLE float
  33 # define STRTOD_L strtof_l
  34 # define HAVE_GOOD_STRTOD_L (HAVE_STRTOF_L && !GNULIB_defined_strtof_function)
  35 # define STRTOD strtof
  36 #elif LONG
  37 # define C_STRTOD c_strtold
  38 # define DOUBLE long double
  39 # define STRTOD_L strtold_l
  40 # define HAVE_GOOD_STRTOD_L (HAVE_STRTOLD_L && !GNULIB_defined_strtold_function)
  41 # define STRTOD strtold
  42 #else
  43 # define C_STRTOD c_strtod
  44 # define DOUBLE double
  45 # define STRTOD_L strtod_l
  46 # define HAVE_GOOD_STRTOD_L (HAVE_STRTOD_L && !GNULIB_defined_strtod_function)
  47 # define STRTOD strtod
  48 #endif
  49
  50 /* Here we don't need the newlocale override that supports gl_locale_name().  */
  51 #undef newlocale
  52
  53 #if defined LC_ALL_MASK && (HAVE_GOOD_STRTOD_L || HAVE_WORKING_USELOCALE)
  54
  55 /* Cache for the C locale object.
  56    Marked volatile so that different threads see the same value
  57    (avoids locking).  */
  58 static volatile locale_t c_locale_cache;
  59
  60 /* Return the C locale object, or (locale_t) 0 with errno set
  61    if it cannot be created.  */
  62 static locale_t
  63 c_locale (void)
  64 {
  65   if (!c_locale_cache)
  66     c_locale_cache = newlocale (LC_ALL_MASK, "C", (locale_t) 0);
  67   return c_locale_cache;
  68 }
  69
  70 #else
  71
  72 # if HAVE_NL_LANGINFO
  73 #  include <langinfo.h>
  74 # endif
  75 # include "c-ctype.h"
  76
  77 /* Determine the decimal-point character according to the current locale.  */
  78 static char
  79 decimal_point_char (void)
  80 {
  81   const char *point;
  82   /* Determine it in a multithread-safe way.  We know nl_langinfo is
  83      multithread-safe on glibc systems and Mac OS X systems, but is not required
  84      to be multithread-safe by POSIX.  sprintf(), however, is multithread-safe.
  85      localeconv() is rarely multithread-safe.  */
  86 # if HAVE_NL_LANGINFO && (__GLIBC__ || defined __UCLIBC__ || (defined __APPLE__ && defined __MACH__))
  87   point = nl_langinfo (RADIXCHAR);
  88 # elif 1
  89   char pointbuf[5];
  90   sprintf (pointbuf, "%#.0f", 1.0);
  91   point = &pointbuf[1];
  92 # else
  93   point = localeconv () -> decimal_point;
  94 # endif
  95   /* The decimal point is always a single byte: either '.' or ','.  */
  96   return (point[0] != '\0' ? point[0] : '.');
  97 }
  98
  99 #endif
 100
 101 DOUBLE
 102 C_STRTOD (char const *nptr, char **endptr)
 103 {
 104   DOUBLE r;
 105
 106 #if defined LC_ALL_MASK && (HAVE_GOOD_STRTOD_L || HAVE_WORKING_USELOCALE)
 107
 108   locale_t locale = c_locale ();
 109   if (!locale)
 110     {
 111       if (endptr)
 112         *endptr = (char *) nptr;
 113       return 0; /* errno is set here */
 114     }
 115
 116 # if HAVE_GOOD_STRTOD_L
 117
 118   r = STRTOD_L (nptr, endptr, locale);
 119
 120 # else /* HAVE_WORKING_USELOCALE */
 121
 122   locale_t old_locale = uselocale (locale);
 123   if (old_locale == (locale_t)0)
 124     {
 125       if (endptr)
 126         *endptr = (char *) nptr;
 127       return 0; /* errno is set here */
 128     }
 129
 130   r = STRTOD (nptr, endptr);
 131
 132   int saved_errno = errno;
 133   if (uselocale (old_locale) == (locale_t)0)
 134     /* We can't switch back to the old locale.  The thread is hosed.  */
 135     abort ();
 136   errno = saved_errno;
 137
 138 # endif
 139
 140 #else
 141
 142   char decimal_point = decimal_point_char ();
 143
 144   if (decimal_point == '.')
 145     /* In this locale, C_STRTOD and STRTOD behave the same way.  */
 146     r = STRTOD (nptr, endptr);
 147   else
 148     {
 149       /* Start and end of the floating-point number.  */
 150       char const *start;
 151       char const *end;
 152       /* Position of the decimal point '.' in the floating-point number.
 153          Either decimal_point_p == NULL or start <= decimal_point_p < end.  */
 154       char const *decimal_point_p = NULL;
 155       /* Set to true if we encountered decimal_point while parsing.  */
 156       int seen_decimal_point = 0;
 157
 158       /* Parse
 159          1. a sequence of white-space characters,
 160          2. a subject sequence possibly containing a floating-point number,
 161          as described in POSIX
 162          <https://pubs.opengroup.org/onlinepubs/9699919799/functions/strtod.html>.
 163        */
 164       {
 165         char const *p;
 166
 167         p = nptr;
 168
 169         /* Parse a sequence of white-space characters.  */
 170         while (*p != '\0' && c_isspace ((unsigned char) *p))
 171           p++;
 172         start = p;
 173         end = p;
 174
 175         /* Start to parse a subject sequence.  */
 176         if (*p == '+' || *p == '-')
 177           p++;
 178         if (*p == '0')
 179           {
 180             end = p + 1;
 181             if (p[1] == 'x' || p[1] == 'X')
 182               {
 183                 size_t num_hex_digits = 0;
 184                 p += 2;
 185                 /* Parse a non-empty sequence of hexadecimal digits optionally
 186                    containing the decimal point character '.'.  */
 187                 while (*p != '\0')
 188                   {
 189                     if (c_isxdigit ((unsigned char) *p))
 190                       {
 191                         num_hex_digits++;
 192                         p++;
 193                       }
 194                     else if (*p == '.')
 195                       {
 196                         if (decimal_point_p == NULL)
 197                           {
 198                             decimal_point_p = p;
 199                             p++;
 200                           }
 201                         else
 202                           break;
 203                       }
 204                     else
 205                       break;
 206                   }
 207                 seen_decimal_point = (*p == decimal_point);
 208                 if (num_hex_digits > 0)
 209                   {
 210                     end = p;
 211                     /* Parse the character 'p' or the character 'P', optionally
 212                        followed by a '+' or '-' character, and then followed by
 213                        one or more decimal digits.  */
 214                     if (*p == 'p' || *p == 'P')
 215                       {
 216                         p++;
 217                         if (*p == '+' || *p == '-')
 218                           p++;
 219                         if (*p != '\0' && c_isdigit ((unsigned char) *p))
 220                           {
 221                             p++;
 222                             while (*p != '\0' && c_isdigit ((unsigned char) *p))
 223                               p++;
 224                             end = p;
 225                           }
 226                       }
 227                   }
 228                 else
 229                   decimal_point_p = NULL;
 230                 goto done_parsing;
 231               }
 232           }
 233         {
 234           size_t num_digits = 0;
 235           /* Parse a non-empty sequence of decimal digits optionally containing
 236              the decimal point character '.'.  */
 237           while (*p != '\0')
 238             {
 239               if (c_isdigit ((unsigned char) *p))
 240                 {
 241                   num_digits++;
 242                   p++;
 243                 }
 244               else if (*p == '.')
 245                 {
 246                   if (decimal_point_p == NULL)
 247                     {
 248                       decimal_point_p = p;
 249                       p++;
 250                     }
 251                   else
 252                     break;
 253                 }
 254               else
 255                 break;
 256             }
 257           seen_decimal_point = (*p == decimal_point);
 258           if (num_digits > 0)
 259             {
 260               end = p;
 261               /* Parse the character 'e' or the character 'E', optionally
 262                  followed by a '+' or '-' character, and then followed by one or
 263                  more decimal digits.  */
 264               if (*p == 'e' || *p == 'E')
 265                 {
 266                   p++;
 267                   if (*p == '+' || *p == '-')
 268                     p++;
 269                   if (*p != '\0' && c_isdigit ((unsigned char) *p))
 270                     {
 271                       p++;
 272                       while (*p != '\0' && c_isdigit ((unsigned char) *p))
 273                         p++;
 274                       end = p;
 275                     }
 276                 }
 277             }
 278           else
 279             decimal_point_p = NULL;
 280         }
 281       }
 282      done_parsing:
 283       if (end == start || (decimal_point_p == NULL && !seen_decimal_point))
 284         /* If end == start, we have not parsed anything.  The given string
 285            might be "INF", "INFINITY", "NAN", "NAN(chars)", or invalid.
 286            We can pass it to STRTOD.
 287            If end > start and decimal_point_p == NULL, we have parsed a
 288            floating-point number, and it does not have a '.' (and not a ','
 289            either, of course).  If !seen_decimal_point, we did not
 290            encounter decimal_point while parsing.  In this case, the
 291            locale-dependent STRTOD function can handle it.  */
 292         r = STRTOD (nptr, endptr);
 293       else
 294         {
 295           /* Create a modified floating-point number, in which the character '.'
 296              is replaced with the locale-dependent decimal_point.  */
 297           size_t len = end - start;
 298           char *buf;
 299           char *buf_malloced = NULL;
 300           char stackbuf[1000];
 301           char *end_in_buf;
 302
 303           if (len < sizeof (stackbuf))
 304             buf = stackbuf;
 305           else
 306             {
 307               buf = malloc (len + 1);
 308               if (buf == NULL)
 309                 {
 310                   /* Out of memory.
 311                      Callers may not be prepared to see errno == ENOMEM.  But
 312                      they should be prepared to see errno == EINVAL.  */
 313                   errno = EINVAL;
 314                   if (endptr != NULL)
 315                     *endptr = (char *) nptr;
 316                   return 0;
 317                 }
 318               buf_malloced = buf;
 319             }
 320
 321           memcpy (buf, start, len);
 322           if (decimal_point_p != NULL)
 323             buf[decimal_point_p - start] = decimal_point;
 324           buf[len] = '\0';
 325
 326           r = STRTOD (buf, &end_in_buf);
 327           if (endptr != NULL)
 328             *endptr = (char *) (start + (end_in_buf - buf));
 329
 330           if (buf_malloced != NULL)
 331             {
 332               int saved_errno = errno;
 333               free (buf_malloced);
 334               errno = saved_errno;
 335             }
 336         }
 337     }
 338
 339 #endif
 340
 341   return r;
 342 }