summaryrefslogtreecommitdiffstats
path: root/Python/pystrtod.c
blob: 16efa9d3ab8c2987d56a0c6436316ab3ee121e09 (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
/* -*- Mode: C; c-file-style: "python" -*- */

#include <Python.h>
#include <locale.h>

/* ascii character tests (as opposed to locale tests) */
#define ISSPACE(c)  ((c) == ' ' || (c) == '\f' || (c) == '\n' || \
                     (c) == '\r' || (c) == '\t' || (c) == '\v')
#define ISDIGIT(c)  ((c) >= '0' && (c) <= '9')
#define ISXDIGIT(c) (ISDIGIT(c) || ((c) >= 'a' && (c) <= 'f') || ((c) >= 'A' && (c) <= 'F'))


/**
 * PyOS_ascii_strtod:
 * @nptr:    the string to convert to a numeric value.
 * @endptr:  if non-%NULL, it returns the character after
 *           the last character used in the conversion.
 * 
 * Converts a string to a #gdouble value.
 * This function behaves like the standard strtod() function
 * does in the C locale. It does this without actually
 * changing the current locale, since that would not be
 * thread-safe.
 *
 * This function is typically used when reading configuration
 * files or other non-user input that should be locale independent.
 * To handle input from the user you should normally use the
 * locale-sensitive system strtod() function.
 *
 * If the correct value would cause overflow, plus or minus %HUGE_VAL
 * is returned (according to the sign of the value), and %ERANGE is
 * stored in %errno. If the correct value would cause underflow,
 * zero is returned and %ERANGE is stored in %errno.
 * If memory allocation fails, %ENOMEM is stored in %errno.
 * 
 * This function resets %errno before calling strtod() so that
 * you can reliably detect overflow and underflow.
 *
 * Return value: the #gdouble value.
 **/
double
PyOS_ascii_strtod(const char *nptr, char **endptr)
{
	char *fail_pos;
	double val = -1.0;
	struct lconv *locale_data;
	const char *decimal_point;
	size_t decimal_point_len;
	const char *p, *decimal_point_pos;
	const char *end = NULL; /* Silence gcc */
	const char *digits_pos = NULL;
	int negate = 0;

	assert(nptr != NULL);

	fail_pos = NULL;

	locale_data = localeconv();
	decimal_point = locale_data->decimal_point;
	decimal_point_len = strlen(decimal_point);

	assert(decimal_point_len != 0);

	decimal_point_pos = NULL;

	/* We process any leading whitespace and the optional sign manually,
	   then pass the remainder to the system strtod.  This ensures that
	   the result of an underflow has the correct sign. (bug #1725)  */

	p = nptr;
	/* Skip leading space */
	while (ISSPACE(*p))
		p++;

	/* Process leading sign, if present */
	if (*p == '-') {
		negate = 1;
		p++;
	} else if (*p == '+') {
		p++;
	}

	/* What's left should begin with a digit, a decimal point, or one of
	   the letters i, I, n, N. It should not begin with 0x or 0X */
	if ((!ISDIGIT(*p) &&
	     *p != '.' && *p != 'i' && *p != 'I' && *p != 'n' && *p != 'N')
	    ||
	    (*p == '0' && (p[1] == 'x' || p[1] == 'X')))
	{
		if (endptr)
			*endptr = (char*)nptr;
		errno = EINVAL;
		return val;
	}
	digits_pos = p;

	if (decimal_point[0] != '.' || 
	    decimal_point[1] != 0)
	{
		while (ISDIGIT(*p))
			p++;

		if (*p == '.')
		{
			decimal_point_pos = p++;

			while (ISDIGIT(*p))
				p++;

			if (*p == 'e' || *p == 'E')
				p++;
			if (*p == '+' || *p == '-')
				p++;
			while (ISDIGIT(*p))
				p++;
			end = p;
		}
		else if (strncmp(p, decimal_point, decimal_point_len) == 0)
		{
			/* Python bug #1417699 */
			if (endptr)
				*endptr = (char*)nptr;
			errno = EINVAL;
			return val;
		}
		/* For the other cases, we need not convert the decimal point */
	}

	/* Set errno to zero, so that we can distinguish zero results
	   and underflows */
	errno = 0;

	if (decimal_point_pos)
	{
		char *copy, *c;

		/* We need to convert the '.' to the locale specific decimal point */
		copy = (char *)PyMem_MALLOC(end - digits_pos +
					    1 + decimal_point_len);
		if (copy == NULL) {
			if (endptr)
				*endptr = (char *)nptr;
			errno = ENOMEM;
			return val;
		}

		c = copy;
		memcpy(c, digits_pos, decimal_point_pos - digits_pos);
		c += decimal_point_pos - digits_pos;
		memcpy(c, decimal_point, decimal_point_len);
		c += decimal_point_len;
		memcpy(c, decimal_point_pos + 1, end - (decimal_point_pos + 1));
		c += end - (decimal_point_pos + 1);
		*c = 0;

		val = strtod(copy, &fail_pos);

		if (fail_pos)
		{
			if (fail_pos > decimal_point_pos)
				fail_pos = (char *)digits_pos +
					(fail_pos - copy) -
					(decimal_point_len - 1);
			else
				fail_pos = (char *)digits_pos +
					(fail_pos - copy);
		}

		PyMem_FREE(copy);

	}
	else {
		val = strtod(digits_pos, &fail_pos);
	}

	if (fail_pos == digits_pos)
		fail_pos = (char *)nptr;

	if (negate && fail_pos != nptr)
		val = -val;

	if (endptr)
		*endptr = fail_pos;

	return val;
}


/* From the C99 standard, section 7.19.6:
The exponent always contains at least two digits, and only as many more digits
as necessary to represent the exponent.
*/
#define MIN_EXPONENT_DIGITS 2

/* see FORMATBUFLEN in unicodeobject.c */
#define FLOAT_FORMATBUFLEN 120

/**
 * PyOS_ascii_formatd:
 * @buffer: A buffer to place the resulting string in
 * @buf_len: The length of the buffer.
 * @format: The printf()-style format to use for the
 *          code to use for converting. 
 * @d: The #gdouble to convert
 *
 * Converts a #gdouble to a string, using the '.' as
 * decimal point. To format the number you pass in
 * a printf()-style format string. Allowed conversion
 * specifiers are 'e', 'E', 'f', 'F', 'g', 'G', and 'n'.
 * 
 * 'n' is the same as 'g', except it uses the current locale.
 *
 * Return value: The pointer to the buffer with the converted string.
 **/
char *
PyOS_ascii_formatd(char       *buffer, 
		   size_t      buf_len, 
		   const char *format, 
		   double      d)
{
	char *p;
	char format_char;
	size_t format_len = strlen(format);

	/* For type 'n', we need to make a copy of the format string, because
	   we're going to modify 'n' -> 'g', and format is const char*, so we
	   can't modify it directly.  FLOAT_FORMATBUFLEN should be longer than
	   we ever need this to be.  There's an upcoming check to ensure it's
	   big enough. */
	char tmp_format[FLOAT_FORMATBUFLEN];

/* 	g_return_val_if_fail (buffer != NULL, NULL); */
/* 	g_return_val_if_fail (format[0] == '%', NULL); */
/* 	g_return_val_if_fail (strpbrk (format + 1, "'l%") == NULL, NULL); */

	/* The last character in the format string must be the format char */
	format_char = format[format_len - 1];

/* 	g_return_val_if_fail (format_char == 'e' || format_char == 'E' || */
/* 			      format_char == 'f' || format_char == 'F' || */
/* 			      format_char == 'g' || format_char == 'G', */
/* 			      NULL); */

	if (format[0] != '%')
		return NULL;

	/* I'm not sure why this test is here.  It's ensuring that the format
	   string after the first character doesn't have a single quote, a
	   lowercase l, or a percent. This is the reverse of the commented-out
	   test about 10 lines ago. */
	if (strpbrk(format + 1, "'l%"))
		return NULL;

	if (!(format_char == 'e' || format_char == 'E' || 
	      format_char == 'f' || format_char == 'F' || 
	      format_char == 'g' || format_char == 'G' ||
	      format_char == 'n'))
		return NULL;

	/* Map 'n' format_char to 'g', by copying the format string and
	   replacing the final 'n' with a 'g' */
	if (format_char == 'n') {
		if (format_len + 1 >= sizeof(tmp_format)) {
			/* The format won't fit in our copy.  Error out.  In
			   practice, this will never happen and will be detected
			   by returning NULL */
			return NULL;
		}
		strcpy(tmp_format, format);
		tmp_format[format_len - 1] = 'g';
		format = tmp_format;
	}

	/* Have PyOS_snprintf do the hard work */
	PyOS_snprintf(buffer, buf_len, format, d);

	/* Get the current local, and find the decimal point character (or
	   string?).  Convert that string back to a dot.  Do not do this if
	   using the 'n' (number) format code. */
	if (format_char != 'n') {
		struct lconv *locale_data = localeconv();
		const char *decimal_point = locale_data->decimal_point;
		size_t decimal_point_len = strlen(decimal_point);
		size_t rest_len;

		assert(decimal_point_len != 0);

		if (decimal_point[0] != '.' || decimal_point[1] != 0) {
			p = buffer;

			if (*p == '+' || *p == '-')
				p++;

			while (isdigit(Py_CHARMASK(*p)))
				p++;

			if (strncmp(p, decimal_point, decimal_point_len) == 0) {
				*p = '.';
				p++;
				if (decimal_point_len > 1) {
					rest_len = strlen(p +
						      (decimal_point_len - 1));
					memmove(p, p + (decimal_point_len - 1),
						rest_len);
					p[rest_len] = 0;
				}
			}
		}
	}

	/* If an exponent exists, ensure that the exponent is at least
	   MIN_EXPONENT_DIGITS digits, providing the buffer is large enough
	   for the extra zeros.  Also, if there are more than
	   MIN_EXPONENT_DIGITS, remove as many zeros as possible until we get
	   back to MIN_EXPONENT_DIGITS */
	p = strpbrk(buffer, "eE");
	if (p && (*(p + 1) == '-' || *(p + 1) == '+')) {
		char *start = p + 2;
		int exponent_digit_cnt = 0;
		int leading_zero_cnt = 0;
		int in_leading_zeros = 1;
		int significant_digit_cnt;

		p += 2;
		while (*p && isdigit(Py_CHARMASK(*p))) {
			if (in_leading_zeros && *p == '0')
				++leading_zero_cnt;
			if (*p != '0')
				in_leading_zeros = 0;
			++p;
			++exponent_digit_cnt;
		}

		significant_digit_cnt = exponent_digit_cnt - leading_zero_cnt;
		if (exponent_digit_cnt == MIN_EXPONENT_DIGITS) {
			/* If there are 2 exactly digits, we're done,
			   regardless of what they contain */
		}
		else if (exponent_digit_cnt > MIN_EXPONENT_DIGITS) {
			int extra_zeros_cnt;

			/* There are more than 2 digits in the exponent.  See
			   if we can delete some of the leading zeros */
			if (significant_digit_cnt < MIN_EXPONENT_DIGITS)
				significant_digit_cnt = MIN_EXPONENT_DIGITS;
			extra_zeros_cnt = exponent_digit_cnt - significant_digit_cnt;

			/* Delete extra_zeros_cnt worth of characters from the
			   front of the exponent */
			assert(extra_zeros_cnt >= 0);

			/* Add one to significant_digit_cnt to copy the
			   trailing 0 byte, thus setting the length */
			memmove(start,
				start + extra_zeros_cnt,
				significant_digit_cnt + 1);
		}
		else {
			/* If there are fewer than 2 digits, add zeros
			   until there are 2, if there's enough room */
			int zeros = MIN_EXPONENT_DIGITS - exponent_digit_cnt;
			if (start + zeros + exponent_digit_cnt + 1
			      < buffer + buf_len) {
				memmove(start + zeros, start,
					exponent_digit_cnt + 1);
				memset(start, '0', zeros);
			}
		}
	}

	return buffer;
}

double
PyOS_ascii_atof(const char *nptr)
{
	return PyOS_ascii_strtod(nptr, NULL);
}