summaryrefslogtreecommitdiff
path: root/lib/strtod.c
blob: 7810e7cde8307798317f3a66583be77ec79ff215 (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
/* Copyright (C) 1991-1992, 1997, 1999, 2003, 2006, 2008-2017 Free Software
   Foundation, Inc.

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

   This program 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 General Public License for more details.

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

#include <config.h>

#include <stdlib.h>

#include <ctype.h>
#include <errno.h>
#include <float.h>
#include <limits.h>
#include <math.h>
#include <stdbool.h>
#include <string.h>

#include "c-ctype.h"

#ifndef HAVE_LDEXP_IN_LIBC
#define HAVE_LDEXP_IN_LIBC 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;
}

#if !HAVE_LDEXP_IN_LIBC
 #define ldexp dummy_ldexp
 /* A dummy definition that will never be invoked.  */
 static double ldexp (double x _GL_UNUSED, int exponent _GL_UNUSED)
 {
   abort ();
   return 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 (HAVE_LDEXP_IN_LIBC && 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 < -DBL_MAX / radix)
                    {
                      errno = ERANGE;
                      return -HUGE_VAL;
                    }
                  else if (DBL_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.  To convert from a number of digits to a radix exponent,
   multiply by RADIX_MULTIPLIER (either 1 or 4).  */
static double
parse_number (const char *nptr,
              int base, int radix, int radix_multiplier, char expchar,
              char **endptr)
{
  const char *s = nptr;
  bool got_dot = false;
  long int exponent = 0;
  double num = 0;

  for (;; ++s)
    {
      int digit;
      if (c_isdigit (*s))
        digit = *s - '0';
      else if (base == 16 && c_isxdigit (*s))
        digit = c_tolower (*s) - ('a' - 10);
      else if (! got_dot && *s == '.')
        {
          /* Record that we have found the decimal point.  */
          got_dot = true;
          continue;
        }
      else
        /* Any other character terminates the number.  */
        break;

      /* Make sure that multiplication by base will not overflow.  */
      if (num <= DBL_MAX / base)
        num = num * base + digit;
      else
        {
          /* The value of the digit doesn't matter, since we have already
             gotten as many digits as can be represented in a 'double'.
             This doesn't necessarily mean the result will overflow.
             The exponent may reduce it to within range.

             We just need to record that there was another
             digit so that we can multiply by 10 later.  */
          exponent += radix_multiplier;
        }

      /* Keep track of the number of digits after the decimal point.
         If we just divided by base here, we might lose precision.  */
      if (got_dot)
        exponent -= radix_multiplier;
    }

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

      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);
}

static double underlying_strtod (const char *, char **);

/* 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 -DBL_MIN * DBL_MIN does not work when cross-compiling
   to PowerPC on Mac OS X 10.5.  */
#if defined __hpux || defined __sgi || defined __ICC
static double
compute_minus_zero (void)
{
  return -DBL_MIN * DBL_MIN;
}
# define minus_zero compute_minus_zero ()
#else
double minus_zero = -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)
{
  bool negative = false;

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

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

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

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

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

  if (c_isdigit (s[*s == '.']))
    {
      /* 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] == '.')]))
            {
              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, '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] == '+')]))
                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, 'e', &endbuf);
                }
              else
                {
                  dup[e - s] = '\0';
                  num = underlying_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;
}

/* The underlying strtod implementation.  This must be defined
   after strtod because it #undefs strtod.  */
static double
underlying_strtod (const char *nptr, char **endptr)
{
#undef strtod
  return strtod (nptr, endptr);
}