summaryrefslogtreecommitdiff
path: root/lib/strtod.c
blob: 9b3a1422a852f4570f23fac169f626561e868bf6 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
/* Copyright (C) 1991-1992, 1997, 1999, 2003, 2006, 2008-2022 Free Software
   Foundation, Inc.

   This file is free software: you can redistribute it and/or modify
   it under the terms of the GNU Lesser General Public License as
   published by the Free Software Foundation, either version 3 of the
   License, or (at your option) any later version.

   This file is distributed in the hope that it will be useful,
   but WITHOUT ANY WARRANTY; without even the implied warranty of
   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
   GNU Lesser General Public License for more details.

   You should have received a copy of the GNU Lesser General Public License
   along with this program.  If not, see <https://www.gnu.org/licenses/>.  */

#if ! defined USE_LONG_DOUBLE
# include <config.h>
#endif

/* Specification.  */
#include <stdlib.h>

#include <ctype.h>      /* isspace() */
#include <errno.h>
#include <float.h>      /* {DBL,LDBL}_{MIN,MAX} */
#include <limits.h>     /* LONG_{MIN,MAX} */
#include <locale.h>     /* localeconv() */
#include <math.h>       /* NAN */
#include <stdbool.h>
#include <stdio.h>      /* sprintf() */
#include <string.h>     /* strdup() */
#if HAVE_NL_LANGINFO
# include <langinfo.h>
#endif

#include "c-ctype.h"

#undef MIN
#undef MAX
#ifdef USE_LONG_DOUBLE
# define STRTOD strtold
# define LDEXP ldexpl
# if defined __hpux && defined __hppa
   /* We cannot call strtold on HP-UX/hppa, because its return type is a struct,
      not a 'long double'.  */
#  define HAVE_UNDERLYING_STRTOD 0
# elif STRTOLD_HAS_UNDERFLOW_BUG
   /* strtold would not set errno=ERANGE upon underflow.  */
#  define HAVE_UNDERLYING_STRTOD 0
# else
#  define HAVE_UNDERLYING_STRTOD HAVE_STRTOLD
# endif
# define DOUBLE long double
# define MIN LDBL_MIN
# define MAX LDBL_MAX
# define L_(literal) literal##L
#else
# define STRTOD strtod
# define LDEXP ldexp
# define HAVE_UNDERLYING_STRTOD 1
# define DOUBLE double
# define MIN DBL_MIN
# define MAX DBL_MAX
# define L_(literal) literal
#endif

#if (defined USE_LONG_DOUBLE ? HAVE_LDEXPM_IN_LIBC : HAVE_LDEXP_IN_LIBC)
# define USE_LDEXP 1
#else
# define USE_LDEXP 0
#endif

/* Return true if C is a space in the current locale, avoiding
   problems with signed char and isspace.  */
static bool
locale_isspace (char c)
{
  unsigned char uc = c;
  return isspace (uc) != 0;
}

/* Determine the decimal-point character according to the current locale.  */
static char
decimal_point_char (void)
{
  const char *point;
  /* Determine it in a multithread-safe way.  We know nl_langinfo is
     multithread-safe on glibc systems and Mac OS X systems, but is not required
     to be multithread-safe by POSIX.  sprintf(), however, is multithread-safe.
     localeconv() is rarely multithread-safe.  */
#if HAVE_NL_LANGINFO && (__GLIBC__ || defined __UCLIBC__ || (defined __APPLE__ && defined __MACH__))
  point = nl_langinfo (RADIXCHAR);
#elif 1
  char pointbuf[5];
  sprintf (pointbuf, "%#.0f", 1.0);
  point = &pointbuf[1];
#else
  point = localeconv () -> decimal_point;
#endif
  /* The decimal point is always a single byte: either '.' or ','.  */
  return (point[0] != '\0' ? point[0] : '.');
}

#if !USE_LDEXP
 #undef LDEXP
 #define LDEXP dummy_ldexp
 /* A dummy definition that will never be invoked.  */
 static DOUBLE LDEXP (_GL_UNUSED DOUBLE x, _GL_UNUSED int exponent)
 {
   abort ();
   return L_(0.0);
 }
#endif

/* Return X * BASE**EXPONENT.  Return an extreme value and set errno
   to ERANGE if underflow or overflow occurs.  */
static DOUBLE
scale_radix_exp (DOUBLE x, int radix, long int exponent)
{
  /* If RADIX == 10, this code is neither precise nor fast; it is
     merely a straightforward and relatively portable approximation.
     If N == 2, this code is precise on a radix-2 implementation,
     albeit perhaps not fast if ldexp is not in libc.  */

  long int e = exponent;

  if (USE_LDEXP && radix == 2)
    return LDEXP (x, e < INT_MIN ? INT_MIN : INT_MAX < e ? INT_MAX : e);
  else
    {
      DOUBLE r = x;

      if (r != 0)
        {
          if (e < 0)
            {
              while (e++ != 0)
                {
                  r /= radix;
                  if (r == 0 && x != 0)
                    {
                      errno = ERANGE;
                      break;
                    }
                }
            }
          else
            {
              while (e-- != 0)
                {
                  if (r < -MAX / radix)
                    {
                      errno = ERANGE;
                      return -HUGE_VAL;
                    }
                  else if (MAX / radix < r)
                    {
                      errno = ERANGE;
                      return HUGE_VAL;
                    }
                  else
                    r *= radix;
                }
            }
        }

      return r;
    }
}

/* Parse a number at NPTR; this is a bit like strtol (NPTR, ENDPTR)
   except there are no leading spaces or signs or "0x", and ENDPTR is
   nonnull.  The number uses a base BASE (either 10 or 16) fraction, a
   radix RADIX (either 10 or 2) exponent, and exponent character
   EXPCHAR.  BASE is RADIX**RADIX_MULTIPLIER.  */
static DOUBLE
parse_number (const char *nptr,
              int base, int radix, int radix_multiplier, char radixchar,
              char expchar,
              char **endptr)
{
  const char *s = nptr;
  const char *digits_start;
  const char *digits_end;
  const char *radixchar_ptr;
  long int exponent;
  DOUBLE num;

  /* First, determine the start and end of the digit sequence.  */
  digits_start = s;
  radixchar_ptr = NULL;
  for (;; ++s)
    {
      if (base == 16 ? c_isxdigit (*s) : c_isdigit (*s))
        ;
      else if (radixchar_ptr == NULL && *s == radixchar)
        {
          /* Record that we have found the decimal point.  */
          radixchar_ptr = s;
        }
      else
        /* Any other character terminates the digit sequence.  */
        break;
    }
  digits_end = s;
  /* Now radixchar_ptr == NULL or
     digits_start <= radixchar_ptr < digits_end.  */

  if (false)
    { /* Unoptimized.  */
      exponent =
        (radixchar_ptr != NULL
         ? - (long int) (digits_end - radixchar_ptr - 1)
         : 0);
    }
  else
    { /* Remove trailing zero digits.  This reduces rounding errors for
         inputs such as 1.0000000000 or 10000000000e-10.  */
      while (digits_end > digits_start)
        {
          if (digits_end - 1 == radixchar_ptr || *(digits_end - 1) == '0')
            digits_end--;
          else
            break;
        }
      exponent =
        (radixchar_ptr != NULL
         ? (digits_end > radixchar_ptr
            ? - (long int) (digits_end - radixchar_ptr - 1)
            : (long int) (radixchar_ptr - digits_end))
         : (long int) (s - digits_end));
    }

  /* Then, convert the digit sequence to a number.  */
  {
    const char *dp;
    num = 0;
    for (dp = digits_start; dp < digits_end; dp++)
      if (dp != radixchar_ptr)
        {
          int digit;

          /* Make sure that multiplication by BASE will not overflow.  */
          if (!(num <= MAX / base))
            {
              /* The value of the digit and all subsequent digits don't matter,
                 since we have already gotten as many digits as can be
                 represented in a 'DOUBLE'.  This doesn't necessarily mean that
                 the result will overflow: The exponent may reduce it to within
                 range.  */
              exponent +=
                (digits_end - dp)
                - (radixchar_ptr >= dp && radixchar_ptr < digits_end ? 1 : 0);
              break;
            }

          /* Eat the next digit.  */
          if (c_isdigit (*dp))
            digit = *dp - '0';
          else if (base == 16 && c_isxdigit (*dp))
            digit = c_tolower (*dp) - ('a' - 10);
          else
            abort ();
          num = num * base + digit;
        }
  }

  exponent = exponent * radix_multiplier;

  /* Finally, parse the exponent.  */
  if (c_tolower (*s) == expchar && ! locale_isspace (s[1]))
    {
      /* Add any given exponent to the implicit one.  */
      int saved_errno = errno;
      char *end;
      long int value = strtol (s + 1, &end, 10);
      errno = saved_errno;

      if (s + 1 != end)
        {
          /* Skip past the exponent, and add in the implicit exponent,
             resulting in an extreme value on overflow.  */
          s = end;
          exponent =
            (exponent < 0
             ? (value < LONG_MIN - exponent ? LONG_MIN : exponent + value)
             : (LONG_MAX - exponent < value ? LONG_MAX : exponent + value));
        }
    }

  *endptr = (char *) s;
  return scale_radix_exp (num, radix, exponent);
}

/* HP cc on HP-UX 10.20 has a bug with the constant expression -0.0.
   ICC 10.0 has a bug when optimizing the expression -zero.
   The expression -MIN * MIN does not work when cross-compiling
   to PowerPC on Mac OS X 10.5.  */
static DOUBLE
minus_zero (void)
{
#if defined __hpux || defined __sgi || defined __ICC
  return -MIN * MIN;
#else
  return -0.0;
#endif
}

/* Convert NPTR to a DOUBLE.  If ENDPTR is not NULL, a pointer to the
   character after the last one used in the number is put in *ENDPTR.  */
DOUBLE
STRTOD (const char *nptr, char **endptr)
#if HAVE_UNDERLYING_STRTOD
# ifdef USE_LONG_DOUBLE
#  undef strtold
# else
#  undef strtod
# endif
#else
# undef STRTOD
# define STRTOD(NPTR,ENDPTR) \
   parse_number (NPTR, 10, 10, 1, radixchar, 'e', ENDPTR)
#endif
/* From here on, STRTOD refers to the underlying implementation.  It needs
   to handle only finite unsigned decimal numbers with non-null ENDPTR.  */
{
  char radixchar;
  bool negative = false;

  /* The number so far.  */
  DOUBLE num;

  const char *s = nptr;
  const char *end;
  char *endbuf;
  int saved_errno = errno;

  radixchar = decimal_point_char ();

  /* Eat whitespace.  */
  while (locale_isspace (*s))
    ++s;

  /* Get the sign.  */
  negative = *s == '-';
  if (*s == '-' || *s == '+')
    ++s;

  num = STRTOD (s, &endbuf);
  end = endbuf;

  if (c_isdigit (s[*s == radixchar]))
    {
      /* If a hex float was converted incorrectly, do it ourselves.
         If the string starts with "0x" but does not contain digits,
         consume the "0" ourselves.  If a hex float is followed by a
         'p' but no exponent, then adjust the end pointer.  */
      if (*s == '0' && c_tolower (s[1]) == 'x')
        {
          if (! c_isxdigit (s[2 + (s[2] == radixchar)]))
            {
              end = s + 1;

              /* strtod() on z/OS returns ERANGE for "0x".  */
              errno = saved_errno;
            }
          else if (end <= s + 2)
            {
              num = parse_number (s + 2, 16, 2, 4, radixchar, 'p', &endbuf);
              end = endbuf;
            }
          else
            {
              const char *p = s + 2;
              while (p < end && c_tolower (*p) != 'p')
                p++;
              if (p < end && ! c_isdigit (p[1 + (p[1] == '-' || p[1] == '+')]))
                {
                  char *dup = strdup (s);
                  errno = saved_errno;
                  if (!dup)
                    {
                      /* Not really our day, is it.  Rounding errors are
                         better than outright failure.  */
                      num =
                        parse_number (s + 2, 16, 2, 4, radixchar, 'p', &endbuf);
                    }
                  else
                    {
                      dup[p - s] = '\0';
                      num = STRTOD (dup, &endbuf);
                      saved_errno = errno;
                      free (dup);
                      errno = saved_errno;
                    }
                  end = p;
                }
            }
        }
      else
        {
          /* If "1e 1" was misparsed as 10.0 instead of 1.0, re-do the
             underlying STRTOD on a copy of the original string
             truncated to avoid the bug.  */
          const char *e = s + 1;
          while (e < end && c_tolower (*e) != 'e')
            e++;
          if (e < end && ! c_isdigit (e[1 + (e[1] == '-' || e[1] == '+')]))
            {
              char *dup = strdup (s);
              errno = saved_errno;
              if (!dup)
                {
                  /* Not really our day, is it.  Rounding errors are
                     better than outright failure.  */
                  num = parse_number (s, 10, 10, 1, radixchar, 'e', &endbuf);
                }
              else
                {
                  dup[e - s] = '\0';
                  num = STRTOD (dup, &endbuf);
                  saved_errno = errno;
                  free (dup);
                  errno = saved_errno;
                }
              end = e;
            }
        }

      s = end;
    }

  /* Check for infinities and NaNs.  */
  else if (c_tolower (*s) == 'i'
           && c_tolower (s[1]) == 'n'
           && c_tolower (s[2]) == 'f')
    {
      s += 3;
      if (c_tolower (*s) == 'i'
          && c_tolower (s[1]) == 'n'
          && c_tolower (s[2]) == 'i'
          && c_tolower (s[3]) == 't'
          && c_tolower (s[4]) == 'y')
        s += 5;
      num = HUGE_VAL;
      errno = saved_errno;
    }
  else if (c_tolower (*s) == 'n'
           && c_tolower (s[1]) == 'a'
           && c_tolower (s[2]) == 'n')
    {
      s += 3;
      if (*s == '(')
        {
          const char *p = s + 1;
          while (c_isalnum (*p))
            p++;
          if (*p == ')')
            s = p + 1;
        }

      /* If the underlying implementation misparsed the NaN, assume
         its result is incorrect, and return a NaN.  Normally it's
         better to use the underlying implementation's result, since a
         nice implementation populates the bits of the NaN according
         to interpreting n-char-sequence as a hexadecimal number.  */
      if (s != end || num == num)
        num = NAN;
      errno = saved_errno;
    }
  else
    {
      /* No conversion could be performed.  */
      errno = EINVAL;
      s = nptr;
    }

  if (endptr != NULL)
    *endptr = (char *) s;
  /* Special case -0.0, since at least ICC miscompiles negation.  We
     can't use copysign(), as that drags in -lm on some platforms.  */
  if (!num && negative)
    return minus_zero ();
  return negative ? -num : num;
}