Python code coverage for Python/formatter_unicode.c

#	count	content
1	n/a	/* implements the unicode (as opposed to string) version of the
2	n/a	built-in formatters for string, int, float. that is, the versions
3	n/a	of int.__float__, etc., that take and return unicode objects */
4	n/a
5	n/a	#include "Python.h"
6	n/a	#include <locale.h>
7	n/a
8	n/a	/* Raises an exception about an unknown presentation type for this
9	n/a	* type. */
10	n/a
11	n/a	static void
12	n/a	unknown_presentation_type(Py_UCS4 presentation_type,
13	n/a	const char* type_name)
14	n/a	{
15	n/a	/* %c might be out-of-range, hence the two cases. */
16	n/a	if (presentation_type > 32 && presentation_type < 128)
17	n/a	PyErr_Format(PyExc_ValueError,
18	n/a	"Unknown format code '%c' "
19	n/a	"for object of type '%.200s'",
20	n/a	(char)presentation_type,
21	n/a	type_name);
22	n/a	else
23	n/a	PyErr_Format(PyExc_ValueError,
24	n/a	"Unknown format code '\\x%x' "
25	n/a	"for object of type '%.200s'",
26	n/a	(unsigned int)presentation_type,
27	n/a	type_name);
28	n/a	}
29	n/a
30	n/a	static void
31	n/a	invalid_comma_type(Py_UCS4 presentation_type)
32	n/a	{
33	n/a	if (presentation_type > 32 && presentation_type < 128)
34	n/a	PyErr_Format(PyExc_ValueError,
35	n/a	"Cannot specify ',' or '_' with '%c'.",
36	n/a	(char)presentation_type);
37	n/a	else
38	n/a	PyErr_Format(PyExc_ValueError,
39	n/a	"Cannot specify ',' or '_' with '\\x%x'.",
40	n/a	(unsigned int)presentation_type);
41	n/a	}
42	n/a
43	n/a	static void
44	n/a	invalid_comma_and_underscore(void)
45	n/a	{
46	n/a	PyErr_Format(PyExc_ValueError, "Cannot specify both ',' and '_'.");
47	n/a	}
48	n/a
49	n/a	/*
50	n/a	get_integer consumes 0 or more decimal digit characters from an
51	n/a	input string, updates *result with the corresponding positive
52	n/a	integer, and returns the number of digits consumed.
53	n/a
54	n/a	returns -1 on error.
55	n/a	*/
56	n/a	static int
57	n/a	get_integer(PyObject str, Py_ssize_t ppos, Py_ssize_t end,
58	n/a	Py_ssize_t *result)
59	n/a	{
60	n/a	Py_ssize_t accumulator, digitval, pos = *ppos;
61	n/a	int numdigits;
62	n/a	int kind = PyUnicode_KIND(str);
63	n/a	void *data = PyUnicode_DATA(str);
64	n/a
65	n/a	accumulator = numdigits = 0;
66	n/a	for (; pos < end; pos++, numdigits++) {
67	n/a	digitval = Py_UNICODE_TODECIMAL(PyUnicode_READ(kind, data, pos));
68	n/a	if (digitval < 0)
69	n/a	break;
70	n/a	/*
71	n/a	Detect possible overflow before it happens:
72	n/a
73	n/a	accumulator * 10 + digitval > PY_SSIZE_T_MAX if and only if
74	n/a	accumulator > (PY_SSIZE_T_MAX - digitval) / 10.
75	n/a	*/
76	n/a	if (accumulator > (PY_SSIZE_T_MAX - digitval) / 10) {
77	n/a	PyErr_Format(PyExc_ValueError,
78	n/a	"Too many decimal digits in format string");
79	n/a	*ppos = pos;
80	n/a	return -1;
81	n/a	}
82	n/a	accumulator = accumulator * 10 + digitval;
83	n/a	}
84	n/a	*ppos = pos;
85	n/a	*result = accumulator;
86	n/a	return numdigits;
87	n/a	}
88	n/a
89	n/a	/************************************************************************/
90	n/a	/********* standard format specifier parsing ************************/
91	n/a	/************************************************************************/
92	n/a
93	n/a	/* returns true if this character is a specifier alignment token */
94	n/a	Py_LOCAL_INLINE(int)
95	n/a	is_alignment_token(Py_UCS4 c)
96	n/a	{
97	n/a	switch (c) {
98	n/a	case '<': case '>': case '=': case '^':
99	n/a	return 1;
100	n/a	default:
101	n/a	return 0;
102	n/a	}
103	n/a	}
104	n/a
105	n/a	/* returns true if this character is a sign element */
106	n/a	Py_LOCAL_INLINE(int)
107	n/a	is_sign_element(Py_UCS4 c)
108	n/a	{
109	n/a	switch (c) {
110	n/a	case ' ': case '+': case '-':
111	n/a	return 1;
112	n/a	default:
113	n/a	return 0;
114	n/a	}
115	n/a	}
116	n/a
117	n/a	/* Locale type codes. LT_NO_LOCALE must be zero. */
118	n/a	enum LocaleType {
119	n/a	LT_NO_LOCALE = 0,
120	n/a	LT_DEFAULT_LOCALE,
121	n/a	LT_UNDERSCORE_LOCALE,
122	n/a	LT_UNDER_FOUR_LOCALE,
123	n/a	LT_CURRENT_LOCALE
124	n/a	};
125	n/a
126	n/a	typedef struct {
127	n/a	Py_UCS4 fill_char;
128	n/a	Py_UCS4 align;
129	n/a	int alternate;
130	n/a	Py_UCS4 sign;
131	n/a	Py_ssize_t width;
132	n/a	enum LocaleType thousands_separators;
133	n/a	Py_ssize_t precision;
134	n/a	Py_UCS4 type;
135	n/a	} InternalFormatSpec;
136	n/a
137	n/a	#if 0
138	n/a	/* Occasionally useful for debugging. Should normally be commented out. */
139	n/a	static void
140	n/a	DEBUG_PRINT_FORMAT_SPEC(InternalFormatSpec *format)
141	n/a	{
142	n/a	printf("internal format spec: fill_char %d\n", format->fill_char);
143	n/a	printf("internal format spec: align %d\n", format->align);
144	n/a	printf("internal format spec: alternate %d\n", format->alternate);
145	n/a	printf("internal format spec: sign %d\n", format->sign);
146	n/a	printf("internal format spec: width %zd\n", format->width);
147	n/a	printf("internal format spec: thousands_separators %d\n",
148	n/a	format->thousands_separators);
149	n/a	printf("internal format spec: precision %zd\n", format->precision);
150	n/a	printf("internal format spec: type %c\n", format->type);
151	n/a	printf("\n");
152	n/a	}
153	n/a	#endif
154	n/a
155	n/a
156	n/a	/*
157	n/a	ptr points to the start of the format_spec, end points just past its end.
158	n/a	fills in format with the parsed information.
159	n/a	returns 1 on success, 0 on failure.
160	n/a	if failure, sets the exception
161	n/a	*/
162	n/a	static int
163	n/a	parse_internal_render_format_spec(PyObject *format_spec,
164	n/a	Py_ssize_t start, Py_ssize_t end,
165	n/a	InternalFormatSpec *format,
166	n/a	char default_type,
167	n/a	char default_align)
168	n/a	{
169	n/a	Py_ssize_t pos = start;
170	n/a	int kind = PyUnicode_KIND(format_spec);
171	n/a	void *data = PyUnicode_DATA(format_spec);
172	n/a	/* end-pos is used throughout this code to specify the length of
173	n/a	the input string */
174	n/a	#define READ_spec(index) PyUnicode_READ(kind, data, index)
175	n/a
176	n/a	Py_ssize_t consumed;
177	n/a	int align_specified = 0;
178	n/a	int fill_char_specified = 0;
179	n/a
180	n/a	format->fill_char = ' ';
181	n/a	format->align = default_align;
182	n/a	format->alternate = 0;
183	n/a	format->sign = '\0';
184	n/a	format->width = -1;
185	n/a	format->thousands_separators = LT_NO_LOCALE;
186	n/a	format->precision = -1;
187	n/a	format->type = default_type;
188	n/a
189	n/a	/* If the second char is an alignment token,
190	n/a	then parse the fill char */
191	n/a	if (end-pos >= 2 && is_alignment_token(READ_spec(pos+1))) {
192	n/a	format->align = READ_spec(pos+1);
193	n/a	format->fill_char = READ_spec(pos);
194	n/a	fill_char_specified = 1;
195	n/a	align_specified = 1;
196	n/a	pos += 2;
197	n/a	}
198	n/a	else if (end-pos >= 1 && is_alignment_token(READ_spec(pos))) {
199	n/a	format->align = READ_spec(pos);
200	n/a	align_specified = 1;
201	n/a	++pos;
202	n/a	}
203	n/a
204	n/a	/* Parse the various sign options */
205	n/a	if (end-pos >= 1 && is_sign_element(READ_spec(pos))) {
206	n/a	format->sign = READ_spec(pos);
207	n/a	++pos;
208	n/a	}
209	n/a
210	n/a	/* If the next character is #, we're in alternate mode. This only
211	n/a	applies to integers. */
212	n/a	if (end-pos >= 1 && READ_spec(pos) == '#') {
213	n/a	format->alternate = 1;
214	n/a	++pos;
215	n/a	}
216	n/a
217	n/a	/* The special case for 0-padding (backwards compat) */
218	n/a	if (!fill_char_specified && end-pos >= 1 && READ_spec(pos) == '0') {
219	n/a	format->fill_char = '0';
220	n/a	if (!align_specified) {
221	n/a	format->align = '=';
222	n/a	}
223	n/a	++pos;
224	n/a	}
225	n/a
226	n/a	consumed = get_integer(format_spec, &pos, end, &format->width);
227	n/a	if (consumed == -1)
228	n/a	/* Overflow error. Exception already set. */
229	n/a	return 0;
230	n/a
231	n/a	/* If consumed is 0, we didn't consume any characters for the
232	n/a	width. In that case, reset the width to -1, because
233	n/a	get_integer() will have set it to zero. -1 is how we record
234	n/a	that the width wasn't specified. */
235	n/a	if (consumed == 0)
236	n/a	format->width = -1;
237	n/a
238	n/a	/* Comma signifies add thousands separators */
239	n/a	if (end-pos && READ_spec(pos) == ',') {
240	n/a	format->thousands_separators = LT_DEFAULT_LOCALE;
241	n/a	++pos;
242	n/a	}
243	n/a	/* Underscore signifies add thousands separators */
244	n/a	if (end-pos && READ_spec(pos) == '_') {
245	n/a	if (format->thousands_separators != LT_NO_LOCALE) {
246	n/a	invalid_comma_and_underscore();
247	n/a	return 0;
248	n/a	}
249	n/a	format->thousands_separators = LT_UNDERSCORE_LOCALE;
250	n/a	++pos;
251	n/a	}
252	n/a	if (end-pos && READ_spec(pos) == ',') {
253	n/a	invalid_comma_and_underscore();
254	n/a	return 0;
255	n/a	}
256	n/a
257	n/a	/* Parse field precision */
258	n/a	if (end-pos && READ_spec(pos) == '.') {
259	n/a	++pos;
260	n/a
261	n/a	consumed = get_integer(format_spec, &pos, end, &format->precision);
262	n/a	if (consumed == -1)
263	n/a	/* Overflow error. Exception already set. */
264	n/a	return 0;
265	n/a
266	n/a	/* Not having a precision after a dot is an error. */
267	n/a	if (consumed == 0) {
268	n/a	PyErr_Format(PyExc_ValueError,
269	n/a	"Format specifier missing precision");
270	n/a	return 0;
271	n/a	}
272	n/a
273	n/a	}
274	n/a
275	n/a	/* Finally, parse the type field. */
276	n/a
277	n/a	if (end-pos > 1) {
278	n/a	/* More than one char remain, invalid format specifier. */
279	n/a	PyErr_Format(PyExc_ValueError, "Invalid format specifier");
280	n/a	return 0;
281	n/a	}
282	n/a
283	n/a	if (end-pos == 1) {
284	n/a	format->type = READ_spec(pos);
285	n/a	++pos;
286	n/a	}
287	n/a
288	n/a	/* Do as much validating as we can, just by looking at the format
289	n/a	specifier. Do not take into account what type of formatting
290	n/a	we're doing (int, float, string). */
291	n/a
292	n/a	if (format->thousands_separators) {
293	n/a	switch (format->type) {
294	n/a	case 'd':
295	n/a	case 'e':
296	n/a	case 'f':
297	n/a	case 'g':
298	n/a	case 'E':
299	n/a	case 'G':
300	n/a	case '%':
301	n/a	case 'F':
302	n/a	case '\0':
303	n/a	/* These are allowed. See PEP 378.*/
304	n/a	break;
305	n/a	case 'b':
306	n/a	case 'o':
307	n/a	case 'x':
308	n/a	case 'X':
309	n/a	/* Underscores are allowed in bin/oct/hex. See PEP 515. */
310	n/a	if (format->thousands_separators == LT_UNDERSCORE_LOCALE) {
311	n/a	/* Every four digits, not every three, in bin/oct/hex. */
312	n/a	format->thousands_separators = LT_UNDER_FOUR_LOCALE;
313	n/a	break;
314	n/a	}
315	n/a	default:
316	n/a	invalid_comma_type(format->type);
317	n/a	return 0;
318	n/a	}
319	n/a	}
320	n/a
321	n/a	assert (format->align <= 127);
322	n/a	assert (format->sign <= 127);
323	n/a	return 1;
324	n/a	}
325	n/a
326	n/a	/* Calculate the padding needed. */
327	n/a	static void
328	n/a	calc_padding(Py_ssize_t nchars, Py_ssize_t width, Py_UCS4 align,
329	n/a	Py_ssize_t n_lpadding, Py_ssize_t n_rpadding,
330	n/a	Py_ssize_t *n_total)
331	n/a	{
332	n/a	if (width >= 0) {
333	n/a	if (nchars > width)
334	n/a	*n_total = nchars;
335	n/a	else
336	n/a	*n_total = width;
337	n/a	}
338	n/a	else {
339	n/a	/* not specified, use all of the chars and no more */
340	n/a	*n_total = nchars;
341	n/a	}
342	n/a
343	n/a	/* Figure out how much leading space we need, based on the
344	n/a	aligning */
345	n/a	if (align == '>')
346	n/a	n_lpadding = n_total - nchars;
347	n/a	else if (align == '^')
348	n/a	n_lpadding = (n_total - nchars) / 2;
349	n/a	else if (align == '<' \|\| align == '=')
350	n/a	*n_lpadding = 0;
351	n/a	else {
352	n/a	/* We should never have an unspecified alignment. */
353	n/a	*n_lpadding = 0;
354	n/a	assert(0);
355	n/a	}
356	n/a
357	n/a	n_rpadding = n_total - nchars - *n_lpadding;
358	n/a	}
359	n/a
360	n/a	/* Do the padding, and return a pointer to where the caller-supplied
361	n/a	content goes. */
362	n/a	static int
363	n/a	fill_padding(_PyUnicodeWriter *writer,
364	n/a	Py_ssize_t nchars,
365	n/a	Py_UCS4 fill_char, Py_ssize_t n_lpadding,
366	n/a	Py_ssize_t n_rpadding)
367	n/a	{
368	n/a	Py_ssize_t pos;
369	n/a
370	n/a	/* Pad on left. */
371	n/a	if (n_lpadding) {
372	n/a	pos = writer->pos;
373	n/a	_PyUnicode_FastFill(writer->buffer, pos, n_lpadding, fill_char);
374	n/a	}
375	n/a
376	n/a	/* Pad on right. */
377	n/a	if (n_rpadding) {
378	n/a	pos = writer->pos + nchars + n_lpadding;
379	n/a	_PyUnicode_FastFill(writer->buffer, pos, n_rpadding, fill_char);
380	n/a	}
381	n/a
382	n/a	/* Pointer to the user content. */
383	n/a	writer->pos += n_lpadding;
384	n/a	return 0;
385	n/a	}
386	n/a
387	n/a	/************************************************************************/
388	n/a	/********* common routines for numeric formatting *******************/
389	n/a	/************************************************************************/
390	n/a
391	n/a	/* Locale info needed for formatting integers and the part of floats
392	n/a	before and including the decimal. Note that locales only support
393	n/a	8-bit chars, not unicode. */
394	n/a	typedef struct {
395	n/a	PyObject *decimal_point;
396	n/a	PyObject *thousands_sep;
397	n/a	const char *grouping;
398	n/a	} LocaleInfo;
399	n/a
400	n/a	#define STATIC_LOCALE_INFO_INIT {0, 0, 0}
401	n/a
402	n/a	/* describes the layout for an integer, see the comment in
403	n/a	calc_number_widths() for details */
404	n/a	typedef struct {
405	n/a	Py_ssize_t n_lpadding;
406	n/a	Py_ssize_t n_prefix;
407	n/a	Py_ssize_t n_spadding;
408	n/a	Py_ssize_t n_rpadding;
409	n/a	char sign;
410	n/a	Py_ssize_t n_sign; /* number of digits needed for sign (0/1) */
411	n/a	Py_ssize_t n_grouped_digits; /* Space taken up by the digits, including
412	n/a	any grouping chars. */
413	n/a	Py_ssize_t n_decimal; /* 0 if only an integer */
414	n/a	Py_ssize_t n_remainder; /* Digits in decimal and/or exponent part,
415	n/a	excluding the decimal itself, if
416	n/a	present. */
417	n/a
418	n/a	/* These 2 are not the widths of fields, but are needed by
419	n/a	STRINGLIB_GROUPING. */
420	n/a	Py_ssize_t n_digits; /* The number of digits before a decimal
421	n/a	or exponent. */
422	n/a	Py_ssize_t n_min_width; /* The min_width we used when we computed
423	n/a	the n_grouped_digits width. */
424	n/a	} NumberFieldWidths;
425	n/a
426	n/a
427	n/a	/* Given a number of the form:
428	n/a	digits[remainder]
429	n/a	where ptr points to the start and end points to the end, find where
430	n/a	the integer part ends. This could be a decimal, an exponent, both,
431	n/a	or neither.
432	n/a	If a decimal point is present, set *has_decimal and increment
433	n/a	remainder beyond it.
434	n/a	Results are undefined (but shouldn't crash) for improperly
435	n/a	formatted strings.
436	n/a	*/
437	n/a	static void
438	n/a	parse_number(PyObject *s, Py_ssize_t pos, Py_ssize_t end,
439	n/a	Py_ssize_t n_remainder, int has_decimal)
440	n/a	{
441	n/a	Py_ssize_t remainder;
442	n/a	int kind = PyUnicode_KIND(s);
443	n/a	void *data = PyUnicode_DATA(s);
444	n/a
445	n/a	while (pos<end && Py_ISDIGIT(PyUnicode_READ(kind, data, pos)))
446	n/a	++pos;
447	n/a	remainder = pos;
448	n/a
449	n/a	/* Does remainder start with a decimal point? */
450	n/a	*has_decimal = pos<end && PyUnicode_READ(kind, data, remainder) == '.';
451	n/a
452	n/a	/* Skip the decimal point. */
453	n/a	if (*has_decimal)
454	n/a	remainder++;
455	n/a
456	n/a	*n_remainder = end - remainder;
457	n/a	}
458	n/a
459	n/a	/* not all fields of format are used. for example, precision is
460	n/a	unused. should this take discrete params in order to be more clear
461	n/a	about what it does? or is passing a single format parameter easier
462	n/a	and more efficient enough to justify a little obfuscation? */
463	n/a	static Py_ssize_t
464	n/a	calc_number_widths(NumberFieldWidths *spec, Py_ssize_t n_prefix,
465	n/a	Py_UCS4 sign_char, PyObject *number, Py_ssize_t n_start,
466	n/a	Py_ssize_t n_end, Py_ssize_t n_remainder,
467	n/a	int has_decimal, const LocaleInfo *locale,
468	n/a	const InternalFormatSpec format, Py_UCS4 maxchar)
469	n/a	{
470	n/a	Py_ssize_t n_non_digit_non_padding;
471	n/a	Py_ssize_t n_padding;
472	n/a
473	n/a	spec->n_digits = n_end - n_start - n_remainder - (has_decimal?1:0);
474	n/a	spec->n_lpadding = 0;
475	n/a	spec->n_prefix = n_prefix;
476	n/a	spec->n_decimal = has_decimal ? PyUnicode_GET_LENGTH(locale->decimal_point) : 0;
477	n/a	spec->n_remainder = n_remainder;
478	n/a	spec->n_spadding = 0;
479	n/a	spec->n_rpadding = 0;
480	n/a	spec->sign = '\0';
481	n/a	spec->n_sign = 0;
482	n/a
483	n/a	/* the output will look like:
484	n/a	\| \|
485	n/a	\| <lpadding> <sign> <prefix> <spadding> <grouped_digits> <decimal> <remainder> <rpadding> \|
486	n/a	\| \|
487	n/a
488	n/a	sign is computed from format->sign and the actual
489	n/a	sign of the number
490	n/a
491	n/a	prefix is given (it's for the '0x' prefix)
492	n/a
493	n/a	digits is already known
494	n/a
495	n/a	the total width is either given, or computed from the
496	n/a	actual digits
497	n/a
498	n/a	only one of lpadding, spadding, and rpadding can be non-zero,
499	n/a	and it's calculated from the width and other fields
500	n/a	*/
501	n/a
502	n/a	/* compute the various parts we're going to write */
503	n/a	switch (format->sign) {
504	n/a	case '+':
505	n/a	/* always put a + or - */
506	n/a	spec->n_sign = 1;
507	n/a	spec->sign = (sign_char == '-' ? '-' : '+');
508	n/a	break;
509	n/a	case ' ':
510	n/a	spec->n_sign = 1;
511	n/a	spec->sign = (sign_char == '-' ? '-' : ' ');
512	n/a	break;
513	n/a	default:
514	n/a	/* Not specified, or the default (-) */
515	n/a	if (sign_char == '-') {
516	n/a	spec->n_sign = 1;
517	n/a	spec->sign = '-';
518	n/a	}
519	n/a	}
520	n/a
521	n/a	/* The number of chars used for non-digits and non-padding. */
522	n/a	n_non_digit_non_padding = spec->n_sign + spec->n_prefix + spec->n_decimal +
523	n/a	spec->n_remainder;
524	n/a
525	n/a	/* min_width can go negative, that's okay. format->width == -1 means
526	n/a	we don't care. */
527	n/a	if (format->fill_char == '0' && format->align == '=')
528	n/a	spec->n_min_width = format->width - n_non_digit_non_padding;
529	n/a	else
530	n/a	spec->n_min_width = 0;
531	n/a
532	n/a	if (spec->n_digits == 0)
533	n/a	/* This case only occurs when using 'c' formatting, we need
534	n/a	to special case it because the grouping code always wants
535	n/a	to have at least one character. */
536	n/a	spec->n_grouped_digits = 0;
537	n/a	else {
538	n/a	Py_UCS4 grouping_maxchar;
539	n/a	spec->n_grouped_digits = _PyUnicode_InsertThousandsGrouping(
540	n/a	NULL, 0,
541	n/a	0, NULL,
542	n/a	spec->n_digits, spec->n_min_width,
543	n/a	locale->grouping, locale->thousands_sep, &grouping_maxchar);
544	n/a	maxchar = Py_MAX(maxchar, grouping_maxchar);
545	n/a	}
546	n/a
547	n/a	/* Given the desired width and the total of digit and non-digit
548	n/a	space we consume, see if we need any padding. format->width can
549	n/a	be negative (meaning no padding), but this code still works in
550	n/a	that case. */
551	n/a	n_padding = format->width -
552	n/a	(n_non_digit_non_padding + spec->n_grouped_digits);
553	n/a	if (n_padding > 0) {
554	n/a	/* Some padding is needed. Determine if it's left, space, or right. */
555	n/a	switch (format->align) {
556	n/a	case '<':
557	n/a	spec->n_rpadding = n_padding;
558	n/a	break;
559	n/a	case '^':
560	n/a	spec->n_lpadding = n_padding / 2;
561	n/a	spec->n_rpadding = n_padding - spec->n_lpadding;
562	n/a	break;
563	n/a	case '=':
564	n/a	spec->n_spadding = n_padding;
565	n/a	break;
566	n/a	case '>':
567	n/a	spec->n_lpadding = n_padding;
568	n/a	break;
569	n/a	default:
570	n/a	/* Shouldn't get here, but treat it as '>' */
571	n/a	spec->n_lpadding = n_padding;
572	n/a	assert(0);
573	n/a	break;
574	n/a	}
575	n/a	}
576	n/a
577	n/a	if (spec->n_lpadding \|\| spec->n_spadding \|\| spec->n_rpadding)
578	n/a	maxchar = Py_MAX(maxchar, format->fill_char);
579	n/a
580	n/a	if (spec->n_decimal)
581	n/a	maxchar = Py_MAX(maxchar, PyUnicode_MAX_CHAR_VALUE(locale->decimal_point));
582	n/a
583	n/a	return spec->n_lpadding + spec->n_sign + spec->n_prefix +
584	n/a	spec->n_spadding + spec->n_grouped_digits + spec->n_decimal +
585	n/a	spec->n_remainder + spec->n_rpadding;
586	n/a	}
587	n/a
588	n/a	/* Fill in the digit parts of a numbers's string representation,
589	n/a	as determined in calc_number_widths().
590	n/a	Return -1 on error, or 0 on success. */
591	n/a	static int
592	n/a	fill_number(_PyUnicodeWriter writer, const NumberFieldWidths spec,
593	n/a	PyObject *digits, Py_ssize_t d_start, Py_ssize_t d_end,
594	n/a	PyObject *prefix, Py_ssize_t p_start,
595	n/a	Py_UCS4 fill_char,
596	n/a	LocaleInfo *locale, int toupper)
597	n/a	{
598	n/a	/* Used to keep track of digits, decimal, and remainder. */
599	n/a	Py_ssize_t d_pos = d_start;
600	n/a	const unsigned int kind = writer->kind;
601	n/a	const void *data = writer->data;
602	n/a	Py_ssize_t r;
603	n/a
604	n/a	if (spec->n_lpadding) {
605	n/a	_PyUnicode_FastFill(writer->buffer,
606	n/a	writer->pos, spec->n_lpadding, fill_char);
607	n/a	writer->pos += spec->n_lpadding;
608	n/a	}
609	n/a	if (spec->n_sign == 1) {
610	n/a	PyUnicode_WRITE(kind, data, writer->pos, spec->sign);
611	n/a	writer->pos++;
612	n/a	}
613	n/a	if (spec->n_prefix) {
614	n/a	_PyUnicode_FastCopyCharacters(writer->buffer, writer->pos,
615	n/a	prefix, p_start,
616	n/a	spec->n_prefix);
617	n/a	if (toupper) {
618	n/a	Py_ssize_t t;
619	n/a	for (t = 0; t < spec->n_prefix; t++) {
620	n/a	Py_UCS4 c = PyUnicode_READ(kind, data, writer->pos + t);
621	n/a	c = Py_TOUPPER(c);
622	n/a	assert (c <= 127);
623	n/a	PyUnicode_WRITE(kind, data, writer->pos + t, c);
624	n/a	}
625	n/a	}
626	n/a	writer->pos += spec->n_prefix;
627	n/a	}
628	n/a	if (spec->n_spadding) {
629	n/a	_PyUnicode_FastFill(writer->buffer,
630	n/a	writer->pos, spec->n_spadding, fill_char);
631	n/a	writer->pos += spec->n_spadding;
632	n/a	}
633	n/a
634	n/a	/* Only for type 'c' special case, it has no digits. */
635	n/a	if (spec->n_digits != 0) {
636	n/a	/* Fill the digits with InsertThousandsGrouping. */
637	n/a	char *pdigits;
638	n/a	if (PyUnicode_READY(digits))
639	n/a	return -1;
640	n/a	pdigits = PyUnicode_DATA(digits);
641	n/a	if (PyUnicode_KIND(digits) < kind) {
642	n/a	pdigits = _PyUnicode_AsKind(digits, kind);
643	n/a	if (pdigits == NULL)
644	n/a	return -1;
645	n/a	}
646	n/a	r = _PyUnicode_InsertThousandsGrouping(
647	n/a	writer->buffer, writer->pos,
648	n/a	spec->n_grouped_digits,
649	n/a	pdigits + kind * d_pos,
650	n/a	spec->n_digits, spec->n_min_width,
651	n/a	locale->grouping, locale->thousands_sep, NULL);
652	n/a	if (r == -1)
653	n/a	return -1;
654	n/a	assert(r == spec->n_grouped_digits);
655	n/a	if (PyUnicode_KIND(digits) < kind)
656	n/a	PyMem_Free(pdigits);
657	n/a	d_pos += spec->n_digits;
658	n/a	}
659	n/a	if (toupper) {
660	n/a	Py_ssize_t t;
661	n/a	for (t = 0; t < spec->n_grouped_digits; t++) {
662	n/a	Py_UCS4 c = PyUnicode_READ(kind, data, writer->pos + t);
663	n/a	c = Py_TOUPPER(c);
664	n/a	if (c > 127) {
665	n/a	PyErr_SetString(PyExc_SystemError, "non-ascii grouped digit");
666	n/a	return -1;
667	n/a	}
668	n/a	PyUnicode_WRITE(kind, data, writer->pos + t, c);
669	n/a	}
670	n/a	}
671	n/a	writer->pos += spec->n_grouped_digits;
672	n/a
673	n/a	if (spec->n_decimal) {
674	n/a	_PyUnicode_FastCopyCharacters(
675	n/a	writer->buffer, writer->pos,
676	n/a	locale->decimal_point, 0, spec->n_decimal);
677	n/a	writer->pos += spec->n_decimal;
678	n/a	d_pos += 1;
679	n/a	}
680	n/a
681	n/a	if (spec->n_remainder) {
682	n/a	_PyUnicode_FastCopyCharacters(
683	n/a	writer->buffer, writer->pos,
684	n/a	digits, d_pos, spec->n_remainder);
685	n/a	writer->pos += spec->n_remainder;
686	n/a	/* d_pos += spec->n_remainder; */
687	n/a	}
688	n/a
689	n/a	if (spec->n_rpadding) {
690	n/a	_PyUnicode_FastFill(writer->buffer,
691	n/a	writer->pos, spec->n_rpadding,
692	n/a	fill_char);
693	n/a	writer->pos += spec->n_rpadding;
694	n/a	}
695	n/a	return 0;
696	n/a	}
697	n/a
698	n/a	static const char no_grouping[1] = {CHAR_MAX};
699	n/a
700	n/a	/* Find the decimal point character(s?), thousands_separator(s?), and
701	n/a	grouping description, either for the current locale if type is
702	n/a	LT_CURRENT_LOCALE, a hard-coded locale if LT_DEFAULT_LOCALE or
703	n/a	LT_UNDERSCORE_LOCALE/LT_UNDER_FOUR_LOCALE, or none if LT_NO_LOCALE. */
704	n/a	static int
705	n/a	get_locale_info(enum LocaleType type, LocaleInfo *locale_info)
706	n/a	{
707	n/a	switch (type) {
708	n/a	case LT_CURRENT_LOCALE: {
709	n/a	struct lconv *locale_data = localeconv();
710	n/a	locale_info->decimal_point = PyUnicode_DecodeLocale(
711	n/a	locale_data->decimal_point,
712	n/a	NULL);
713	n/a	if (locale_info->decimal_point == NULL)
714	n/a	return -1;
715	n/a	locale_info->thousands_sep = PyUnicode_DecodeLocale(
716	n/a	locale_data->thousands_sep,
717	n/a	NULL);
718	n/a	if (locale_info->thousands_sep == NULL)
719	n/a	return -1;
720	n/a	locale_info->grouping = locale_data->grouping;
721	n/a	break;
722	n/a	}
723	n/a	case LT_DEFAULT_LOCALE:
724	n/a	case LT_UNDERSCORE_LOCALE:
725	n/a	case LT_UNDER_FOUR_LOCALE:
726	n/a	locale_info->decimal_point = PyUnicode_FromOrdinal('.');
727	n/a	locale_info->thousands_sep = PyUnicode_FromOrdinal(
728	n/a	type == LT_DEFAULT_LOCALE ? ',' : '_');
729	n/a	if (!locale_info->decimal_point \|\| !locale_info->thousands_sep)
730	n/a	return -1;
731	n/a	if (type != LT_UNDER_FOUR_LOCALE)
732	n/a	locale_info->grouping = "\3"; /* Group every 3 characters. The
733	n/a	(implicit) trailing 0 means repeat
734	n/a	infinitely. */
735	n/a	else
736	n/a	locale_info->grouping = "\4"; /* Bin/oct/hex group every four. */
737	n/a	break;
738	n/a	case LT_NO_LOCALE:
739	n/a	locale_info->decimal_point = PyUnicode_FromOrdinal('.');
740	n/a	locale_info->thousands_sep = PyUnicode_New(0, 0);
741	n/a	if (!locale_info->decimal_point \|\| !locale_info->thousands_sep)
742	n/a	return -1;
743	n/a	locale_info->grouping = no_grouping;
744	n/a	break;
745	n/a	}
746	n/a	return 0;
747	n/a	}
748	n/a
749	n/a	static void
750	n/a	free_locale_info(LocaleInfo *locale_info)
751	n/a	{
752	n/a	Py_XDECREF(locale_info->decimal_point);
753	n/a	Py_XDECREF(locale_info->thousands_sep);
754	n/a	}
755	n/a
756	n/a	/************************************************************************/
757	n/a	/********* string formatting ****************************************/
758	n/a	/************************************************************************/
759	n/a
760	n/a	static int
761	n/a	format_string_internal(PyObject value, const InternalFormatSpec format,
762	n/a	_PyUnicodeWriter *writer)
763	n/a	{
764	n/a	Py_ssize_t lpad;
765	n/a	Py_ssize_t rpad;
766	n/a	Py_ssize_t total;
767	n/a	Py_ssize_t len;
768	n/a	int result = -1;
769	n/a	Py_UCS4 maxchar;
770	n/a
771	n/a	assert(PyUnicode_IS_READY(value));
772	n/a	len = PyUnicode_GET_LENGTH(value);
773	n/a
774	n/a	/* sign is not allowed on strings */
775	n/a	if (format->sign != '\0') {
776	n/a	PyErr_SetString(PyExc_ValueError,
777	n/a	"Sign not allowed in string format specifier");
778	n/a	goto done;
779	n/a	}
780	n/a
781	n/a	/* alternate is not allowed on strings */
782	n/a	if (format->alternate) {
783	n/a	PyErr_SetString(PyExc_ValueError,
784	n/a	"Alternate form (#) not allowed in string format "
785	n/a	"specifier");
786	n/a	goto done;
787	n/a	}
788	n/a
789	n/a	/* '=' alignment not allowed on strings */
790	n/a	if (format->align == '=') {
791	n/a	PyErr_SetString(PyExc_ValueError,
792	n/a	"'=' alignment not allowed "
793	n/a	"in string format specifier");
794	n/a	goto done;
795	n/a	}
796	n/a
797	n/a	if ((format->width == -1 \|\| format->width <= len)
798	n/a	&& (format->precision == -1 \|\| format->precision >= len)) {
799	n/a	/* Fast path */
800	n/a	return _PyUnicodeWriter_WriteStr(writer, value);
801	n/a	}
802	n/a
803	n/a	/* if precision is specified, output no more that format.precision
804	n/a	characters */
805	n/a	if (format->precision >= 0 && len >= format->precision) {
806	n/a	len = format->precision;
807	n/a	}
808	n/a
809	n/a	calc_padding(len, format->width, format->align, &lpad, &rpad, &total);
810	n/a
811	n/a	maxchar = writer->maxchar;
812	n/a	if (lpad != 0 \|\| rpad != 0)
813	n/a	maxchar = Py_MAX(maxchar, format->fill_char);
814	n/a	if (PyUnicode_MAX_CHAR_VALUE(value) > maxchar) {
815	n/a	Py_UCS4 valmaxchar = _PyUnicode_FindMaxChar(value, 0, len);
816	n/a	maxchar = Py_MAX(maxchar, valmaxchar);
817	n/a	}
818	n/a
819	n/a	/* allocate the resulting string */
820	n/a	if (_PyUnicodeWriter_Prepare(writer, total, maxchar) == -1)
821	n/a	goto done;
822	n/a
823	n/a	/* Write into that space. First the padding. */
824	n/a	result = fill_padding(writer, len, format->fill_char, lpad, rpad);
825	n/a	if (result == -1)
826	n/a	goto done;
827	n/a
828	n/a	/* Then the source string. */
829	n/a	if (len) {
830	n/a	_PyUnicode_FastCopyCharacters(writer->buffer, writer->pos,
831	n/a	value, 0, len);
832	n/a	}
833	n/a	writer->pos += (len + rpad);
834	n/a	result = 0;
835	n/a
836	n/a	done:
837	n/a	return result;
838	n/a	}
839	n/a
840	n/a
841	n/a	/************************************************************************/
842	n/a	/********* long formatting ******************************************/
843	n/a	/************************************************************************/
844	n/a
845	n/a	static int
846	n/a	format_long_internal(PyObject value, const InternalFormatSpec format,
847	n/a	_PyUnicodeWriter *writer)
848	n/a	{
849	n/a	int result = -1;
850	n/a	Py_UCS4 maxchar = 127;
851	n/a	PyObject *tmp = NULL;
852	n/a	Py_ssize_t inumeric_chars;
853	n/a	Py_UCS4 sign_char = '\0';
854	n/a	Py_ssize_t n_digits; /* count of digits need from the computed
855	n/a	string */
856	n/a	Py_ssize_t n_remainder = 0; /* Used only for 'c' formatting, which
857	n/a	produces non-digits */
858	n/a	Py_ssize_t n_prefix = 0; /* Count of prefix chars, (e.g., '0x') */
859	n/a	Py_ssize_t n_total;
860	n/a	Py_ssize_t prefix = 0;
861	n/a	NumberFieldWidths spec;
862	n/a	long x;
863	n/a
864	n/a	/* Locale settings, either from the actual locale or
865	n/a	from a hard-code pseudo-locale */
866	n/a	LocaleInfo locale = STATIC_LOCALE_INFO_INIT;
867	n/a
868	n/a	/* no precision allowed on integers */
869	n/a	if (format->precision != -1) {
870	n/a	PyErr_SetString(PyExc_ValueError,
871	n/a	"Precision not allowed in integer format specifier");
872	n/a	goto done;
873	n/a	}
874	n/a
875	n/a	/* special case for character formatting */
876	n/a	if (format->type == 'c') {
877	n/a	/* error to specify a sign */
878	n/a	if (format->sign != '\0') {
879	n/a	PyErr_SetString(PyExc_ValueError,
880	n/a	"Sign not allowed with integer"
881	n/a	" format specifier 'c'");
882	n/a	goto done;
883	n/a	}
884	n/a	/* error to request alternate format */
885	n/a	if (format->alternate) {
886	n/a	PyErr_SetString(PyExc_ValueError,
887	n/a	"Alternate form (#) not allowed with integer"
888	n/a	" format specifier 'c'");
889	n/a	goto done;
890	n/a	}
891	n/a
892	n/a	/* taken from unicodeobject.c formatchar() */
893	n/a	/* Integer input truncated to a character */
894	n/a	x = PyLong_AsLong(value);
895	n/a	if (x == -1 && PyErr_Occurred())
896	n/a	goto done;
897	n/a	if (x < 0 \|\| x > 0x10ffff) {
898	n/a	PyErr_SetString(PyExc_OverflowError,
899	n/a	"%c arg not in range(0x110000)");
900	n/a	goto done;
901	n/a	}
902	n/a	tmp = PyUnicode_FromOrdinal(x);
903	n/a	inumeric_chars = 0;
904	n/a	n_digits = 1;
905	n/a	maxchar = Py_MAX(maxchar, (Py_UCS4)x);
906	n/a
907	n/a	/* As a sort-of hack, we tell calc_number_widths that we only
908	n/a	have "remainder" characters. calc_number_widths thinks
909	n/a	these are characters that don't get formatted, only copied
910	n/a	into the output string. We do this for 'c' formatting,
911	n/a	because the characters are likely to be non-digits. */
912	n/a	n_remainder = 1;
913	n/a	}
914	n/a	else {
915	n/a	int base;
916	n/a	int leading_chars_to_skip = 0; /* Number of characters added by
917	n/a	PyNumber_ToBase that we want to
918	n/a	skip over. */
919	n/a
920	n/a	/* Compute the base and how many characters will be added by
921	n/a	PyNumber_ToBase */
922	n/a	switch (format->type) {
923	n/a	case 'b':
924	n/a	base = 2;
925	n/a	leading_chars_to_skip = 2; /* 0b */
926	n/a	break;
927	n/a	case 'o':
928	n/a	base = 8;
929	n/a	leading_chars_to_skip = 2; /* 0o */
930	n/a	break;
931	n/a	case 'x':
932	n/a	case 'X':
933	n/a	base = 16;
934	n/a	leading_chars_to_skip = 2; /* 0x */
935	n/a	break;
936	n/a	default: /* shouldn't be needed, but stops a compiler warning */
937	n/a	case 'd':
938	n/a	case 'n':
939	n/a	base = 10;
940	n/a	break;
941	n/a	}
942	n/a
943	n/a	if (format->sign != '+' && format->sign != ' '
944	n/a	&& format->width == -1
945	n/a	&& format->type != 'X' && format->type != 'n'
946	n/a	&& !format->thousands_separators
947	n/a	&& PyLong_CheckExact(value))
948	n/a	{
949	n/a	/* Fast path */
950	n/a	return _PyLong_FormatWriter(writer, value, base, format->alternate);
951	n/a	}
952	n/a
953	n/a	/* The number of prefix chars is the same as the leading
954	n/a	chars to skip */
955	n/a	if (format->alternate)
956	n/a	n_prefix = leading_chars_to_skip;
957	n/a
958	n/a	/* Do the hard part, converting to a string in a given base */
959	n/a	tmp = _PyLong_Format(value, base);
960	n/a	if (tmp == NULL \|\| PyUnicode_READY(tmp) == -1)
961	n/a	goto done;
962	n/a
963	n/a	inumeric_chars = 0;
964	n/a	n_digits = PyUnicode_GET_LENGTH(tmp);
965	n/a
966	n/a	prefix = inumeric_chars;
967	n/a
968	n/a	/* Is a sign character present in the output? If so, remember it
969	n/a	and skip it */
970	n/a	if (PyUnicode_READ_CHAR(tmp, inumeric_chars) == '-') {
971	n/a	sign_char = '-';
972	n/a	++prefix;
973	n/a	++leading_chars_to_skip;
974	n/a	}
975	n/a
976	n/a	/* Skip over the leading chars (0x, 0b, etc.) */
977	n/a	n_digits -= leading_chars_to_skip;
978	n/a	inumeric_chars += leading_chars_to_skip;
979	n/a	}
980	n/a
981	n/a	/* Determine the grouping, separator, and decimal point, if any. */
982	n/a	if (get_locale_info(format->type == 'n' ? LT_CURRENT_LOCALE :
983	n/a	format->thousands_separators,
984	n/a	&locale) == -1)
985	n/a	goto done;
986	n/a
987	n/a	/* Calculate how much memory we'll need. */
988	n/a	n_total = calc_number_widths(&spec, n_prefix, sign_char, tmp, inumeric_chars,
989	n/a	inumeric_chars + n_digits, n_remainder, 0,
990	n/a	&locale, format, &maxchar);
991	n/a
992	n/a	/* Allocate the memory. */
993	n/a	if (_PyUnicodeWriter_Prepare(writer, n_total, maxchar) == -1)
994	n/a	goto done;
995	n/a
996	n/a	/* Populate the memory. */
997	n/a	result = fill_number(writer, &spec,
998	n/a	tmp, inumeric_chars, inumeric_chars + n_digits,
999	n/a	tmp, prefix, format->fill_char,
1000	n/a	&locale, format->type == 'X');
1001	n/a
1002	n/a	done:
1003	n/a	Py_XDECREF(tmp);
1004	n/a	free_locale_info(&locale);
1005	n/a	return result;
1006	n/a	}
1007	n/a
1008	n/a	/************************************************************************/
1009	n/a	/********* float formatting *****************************************/
1010	n/a	/************************************************************************/
1011	n/a
1012	n/a	/* much of this is taken from unicodeobject.c */
1013	n/a	static int
1014	n/a	format_float_internal(PyObject *value,
1015	n/a	const InternalFormatSpec *format,
1016	n/a	_PyUnicodeWriter *writer)
1017	n/a	{
1018	n/a	char buf = NULL; / buffer returned from PyOS_double_to_string */
1019	n/a	Py_ssize_t n_digits;
1020	n/a	Py_ssize_t n_remainder;
1021	n/a	Py_ssize_t n_total;
1022	n/a	int has_decimal;
1023	n/a	double val;
1024	n/a	int precision, default_precision = 6;
1025	n/a	Py_UCS4 type = format->type;
1026	n/a	int add_pct = 0;
1027	n/a	Py_ssize_t index;
1028	n/a	NumberFieldWidths spec;
1029	n/a	int flags = 0;
1030	n/a	int result = -1;
1031	n/a	Py_UCS4 maxchar = 127;
1032	n/a	Py_UCS4 sign_char = '\0';
1033	n/a	int float_type; /* Used to see if we have a nan, inf, or regular float. */
1034	n/a	PyObject *unicode_tmp = NULL;
1035	n/a
1036	n/a	/* Locale settings, either from the actual locale or
1037	n/a	from a hard-code pseudo-locale */
1038	n/a	LocaleInfo locale = STATIC_LOCALE_INFO_INIT;
1039	n/a
1040	n/a	if (format->precision > INT_MAX) {
1041	n/a	PyErr_SetString(PyExc_ValueError, "precision too big");
1042	n/a	goto done;
1043	n/a	}
1044	n/a	precision = (int)format->precision;
1045	n/a
1046	n/a	if (format->alternate)
1047	n/a	flags \|= Py_DTSF_ALT;
1048	n/a
1049	n/a	if (type == '\0') {
1050	n/a	/* Omitted type specifier. Behaves in the same way as repr(x)
1051	n/a	and str(x) if no precision is given, else like 'g', but with
1052	n/a	at least one digit after the decimal point. */
1053	n/a	flags \|= Py_DTSF_ADD_DOT_0;
1054	n/a	type = 'r';
1055	n/a	default_precision = 0;
1056	n/a	}
1057	n/a
1058	n/a	if (type == 'n')
1059	n/a	/* 'n' is the same as 'g', except for the locale used to
1060	n/a	format the result. We take care of that later. */
1061	n/a	type = 'g';
1062	n/a
1063	n/a	val = PyFloat_AsDouble(value);
1064	n/a	if (val == -1.0 && PyErr_Occurred())
1065	n/a	goto done;
1066	n/a
1067	n/a	if (type == '%') {
1068	n/a	type = 'f';
1069	n/a	val *= 100;
1070	n/a	add_pct = 1;
1071	n/a	}
1072	n/a
1073	n/a	if (precision < 0)
1074	n/a	precision = default_precision;
1075	n/a	else if (type == 'r')
1076	n/a	type = 'g';
1077	n/a
1078	n/a	/* Cast "type", because if we're in unicode we need to pass an
1079	n/a	8-bit char. This is safe, because we've restricted what "type"
1080	n/a	can be. */
1081	n/a	buf = PyOS_double_to_string(val, (char)type, precision, flags,
1082	n/a	&float_type);
1083	n/a	if (buf == NULL)
1084	n/a	goto done;
1085	n/a	n_digits = strlen(buf);
1086	n/a
1087	n/a	if (add_pct) {
1088	n/a	/* We know that buf has a trailing zero (since we just called
1089	n/a	strlen() on it), and we don't use that fact any more. So we
1090	n/a	can just write over the trailing zero. */
1091	n/a	buf[n_digits] = '%';
1092	n/a	n_digits += 1;
1093	n/a	}
1094	n/a
1095	n/a	if (format->sign != '+' && format->sign != ' '
1096	n/a	&& format->width == -1
1097	n/a	&& format->type != 'n'
1098	n/a	&& !format->thousands_separators)
1099	n/a	{
1100	n/a	/* Fast path */
1101	n/a	result = _PyUnicodeWriter_WriteASCIIString(writer, buf, n_digits);
1102	n/a	PyMem_Free(buf);
1103	n/a	return result;
1104	n/a	}
1105	n/a
1106	n/a	/* Since there is no unicode version of PyOS_double_to_string,
1107	n/a	just use the 8 bit version and then convert to unicode. */
1108	n/a	unicode_tmp = _PyUnicode_FromASCII(buf, n_digits);
1109	n/a	PyMem_Free(buf);
1110	n/a	if (unicode_tmp == NULL)
1111	n/a	goto done;
1112	n/a
1113	n/a	/* Is a sign character present in the output? If so, remember it
1114	n/a	and skip it */
1115	n/a	index = 0;
1116	n/a	if (PyUnicode_READ_CHAR(unicode_tmp, index) == '-') {
1117	n/a	sign_char = '-';
1118	n/a	++index;
1119	n/a	--n_digits;
1120	n/a	}
1121	n/a
1122	n/a	/* Determine if we have any "remainder" (after the digits, might include
1123	n/a	decimal or exponent or both (or neither)) */
1124	n/a	parse_number(unicode_tmp, index, index + n_digits, &n_remainder, &has_decimal);
1125	n/a
1126	n/a	/* Determine the grouping, separator, and decimal point, if any. */
1127	n/a	if (get_locale_info(format->type == 'n' ? LT_CURRENT_LOCALE :
1128	n/a	format->thousands_separators,
1129	n/a	&locale) == -1)
1130	n/a	goto done;
1131	n/a
1132	n/a	/* Calculate how much memory we'll need. */
1133	n/a	n_total = calc_number_widths(&spec, 0, sign_char, unicode_tmp, index,
1134	n/a	index + n_digits, n_remainder, has_decimal,
1135	n/a	&locale, format, &maxchar);
1136	n/a
1137	n/a	/* Allocate the memory. */
1138	n/a	if (_PyUnicodeWriter_Prepare(writer, n_total, maxchar) == -1)
1139	n/a	goto done;
1140	n/a
1141	n/a	/* Populate the memory. */
1142	n/a	result = fill_number(writer, &spec,
1143	n/a	unicode_tmp, index, index + n_digits,
1144	n/a	NULL, 0, format->fill_char,
1145	n/a	&locale, 0);
1146	n/a
1147	n/a	done:
1148	n/a	Py_XDECREF(unicode_tmp);
1149	n/a	free_locale_info(&locale);
1150	n/a	return result;
1151	n/a	}
1152	n/a
1153	n/a	/************************************************************************/
1154	n/a	/********* complex formatting ***************************************/
1155	n/a	/************************************************************************/
1156	n/a
1157	n/a	static int
1158	n/a	format_complex_internal(PyObject *value,
1159	n/a	const InternalFormatSpec *format,
1160	n/a	_PyUnicodeWriter *writer)
1161	n/a	{
1162	n/a	double re;
1163	n/a	double im;
1164	n/a	char re_buf = NULL; / buffer returned from PyOS_double_to_string */
1165	n/a	char im_buf = NULL; / buffer returned from PyOS_double_to_string */
1166	n/a
1167	n/a	InternalFormatSpec tmp_format = *format;
1168	n/a	Py_ssize_t n_re_digits;
1169	n/a	Py_ssize_t n_im_digits;
1170	n/a	Py_ssize_t n_re_remainder;
1171	n/a	Py_ssize_t n_im_remainder;
1172	n/a	Py_ssize_t n_re_total;
1173	n/a	Py_ssize_t n_im_total;
1174	n/a	int re_has_decimal;
1175	n/a	int im_has_decimal;
1176	n/a	int precision, default_precision = 6;
1177	n/a	Py_UCS4 type = format->type;
1178	n/a	Py_ssize_t i_re;
1179	n/a	Py_ssize_t i_im;
1180	n/a	NumberFieldWidths re_spec;
1181	n/a	NumberFieldWidths im_spec;
1182	n/a	int flags = 0;
1183	n/a	int result = -1;
1184	n/a	Py_UCS4 maxchar = 127;
1185	n/a	enum PyUnicode_Kind rkind;
1186	n/a	void *rdata;
1187	n/a	Py_UCS4 re_sign_char = '\0';
1188	n/a	Py_UCS4 im_sign_char = '\0';
1189	n/a	int re_float_type; /* Used to see if we have a nan, inf, or regular float. */
1190	n/a	int im_float_type;
1191	n/a	int add_parens = 0;
1192	n/a	int skip_re = 0;
1193	n/a	Py_ssize_t lpad;
1194	n/a	Py_ssize_t rpad;
1195	n/a	Py_ssize_t total;
1196	n/a	PyObject *re_unicode_tmp = NULL;
1197	n/a	PyObject *im_unicode_tmp = NULL;
1198	n/a
1199	n/a	/* Locale settings, either from the actual locale or
1200	n/a	from a hard-code pseudo-locale */
1201	n/a	LocaleInfo locale = STATIC_LOCALE_INFO_INIT;
1202	n/a
1203	n/a	if (format->precision > INT_MAX) {
1204	n/a	PyErr_SetString(PyExc_ValueError, "precision too big");
1205	n/a	goto done;
1206	n/a	}
1207	n/a	precision = (int)format->precision;
1208	n/a
1209	n/a	/* Zero padding is not allowed. */
1210	n/a	if (format->fill_char == '0') {
1211	n/a	PyErr_SetString(PyExc_ValueError,
1212	n/a	"Zero padding is not allowed in complex format "
1213	n/a	"specifier");
1214	n/a	goto done;
1215	n/a	}
1216	n/a
1217	n/a	/* Neither is '=' alignment . */
1218	n/a	if (format->align == '=') {
1219	n/a	PyErr_SetString(PyExc_ValueError,
1220	n/a	"'=' alignment flag is not allowed in complex format "
1221	n/a	"specifier");
1222	n/a	goto done;
1223	n/a	}
1224	n/a
1225	n/a	re = PyComplex_RealAsDouble(value);
1226	n/a	if (re == -1.0 && PyErr_Occurred())
1227	n/a	goto done;
1228	n/a	im = PyComplex_ImagAsDouble(value);
1229	n/a	if (im == -1.0 && PyErr_Occurred())
1230	n/a	goto done;
1231	n/a
1232	n/a	if (format->alternate)
1233	n/a	flags \|= Py_DTSF_ALT;
1234	n/a
1235	n/a	if (type == '\0') {
1236	n/a	/* Omitted type specifier. Should be like str(self). */
1237	n/a	type = 'r';
1238	n/a	default_precision = 0;
1239	n/a	if (re == 0.0 && copysign(1.0, re) == 1.0)
1240	n/a	skip_re = 1;
1241	n/a	else
1242	n/a	add_parens = 1;
1243	n/a	}
1244	n/a
1245	n/a	if (type == 'n')
1246	n/a	/* 'n' is the same as 'g', except for the locale used to
1247	n/a	format the result. We take care of that later. */
1248	n/a	type = 'g';
1249	n/a
1250	n/a	if (precision < 0)
1251	n/a	precision = default_precision;
1252	n/a	else if (type == 'r')
1253	n/a	type = 'g';
1254	n/a
1255	n/a	/* Cast "type", because if we're in unicode we need to pass an
1256	n/a	8-bit char. This is safe, because we've restricted what "type"
1257	n/a	can be. */
1258	n/a	re_buf = PyOS_double_to_string(re, (char)type, precision, flags,
1259	n/a	&re_float_type);
1260	n/a	if (re_buf == NULL)
1261	n/a	goto done;
1262	n/a	im_buf = PyOS_double_to_string(im, (char)type, precision, flags,
1263	n/a	&im_float_type);
1264	n/a	if (im_buf == NULL)
1265	n/a	goto done;
1266	n/a
1267	n/a	n_re_digits = strlen(re_buf);
1268	n/a	n_im_digits = strlen(im_buf);
1269	n/a
1270	n/a	/* Since there is no unicode version of PyOS_double_to_string,
1271	n/a	just use the 8 bit version and then convert to unicode. */
1272	n/a	re_unicode_tmp = _PyUnicode_FromASCII(re_buf, n_re_digits);
1273	n/a	if (re_unicode_tmp == NULL)
1274	n/a	goto done;
1275	n/a	i_re = 0;
1276	n/a
1277	n/a	im_unicode_tmp = _PyUnicode_FromASCII(im_buf, n_im_digits);
1278	n/a	if (im_unicode_tmp == NULL)
1279	n/a	goto done;
1280	n/a	i_im = 0;
1281	n/a
1282	n/a	/* Is a sign character present in the output? If so, remember it
1283	n/a	and skip it */
1284	n/a	if (PyUnicode_READ_CHAR(re_unicode_tmp, i_re) == '-') {
1285	n/a	re_sign_char = '-';
1286	n/a	++i_re;
1287	n/a	--n_re_digits;
1288	n/a	}
1289	n/a	if (PyUnicode_READ_CHAR(im_unicode_tmp, i_im) == '-') {
1290	n/a	im_sign_char = '-';
1291	n/a	++i_im;
1292	n/a	--n_im_digits;
1293	n/a	}
1294	n/a
1295	n/a	/* Determine if we have any "remainder" (after the digits, might include
1296	n/a	decimal or exponent or both (or neither)) */
1297	n/a	parse_number(re_unicode_tmp, i_re, i_re + n_re_digits,
1298	n/a	&n_re_remainder, &re_has_decimal);
1299	n/a	parse_number(im_unicode_tmp, i_im, i_im + n_im_digits,
1300	n/a	&n_im_remainder, &im_has_decimal);
1301	n/a
1302	n/a	/* Determine the grouping, separator, and decimal point, if any. */
1303	n/a	if (get_locale_info(format->type == 'n' ? LT_CURRENT_LOCALE :
1304	n/a	format->thousands_separators,
1305	n/a	&locale) == -1)
1306	n/a	goto done;
1307	n/a
1308	n/a	/* Turn off any padding. We'll do it later after we've composed
1309	n/a	the numbers without padding. */
1310	n/a	tmp_format.fill_char = '\0';
1311	n/a	tmp_format.align = '<';
1312	n/a	tmp_format.width = -1;
1313	n/a
1314	n/a	/* Calculate how much memory we'll need. */
1315	n/a	n_re_total = calc_number_widths(&re_spec, 0, re_sign_char, re_unicode_tmp,
1316	n/a	i_re, i_re + n_re_digits, n_re_remainder,
1317	n/a	re_has_decimal, &locale, &tmp_format,
1318	n/a	&maxchar);
1319	n/a
1320	n/a	/* Same formatting, but always include a sign, unless the real part is
1321	n/a	* going to be omitted, in which case we use whatever sign convention was
1322	n/a	* requested by the original format. */
1323	n/a	if (!skip_re)
1324	n/a	tmp_format.sign = '+';
1325	n/a	n_im_total = calc_number_widths(&im_spec, 0, im_sign_char, im_unicode_tmp,
1326	n/a	i_im, i_im + n_im_digits, n_im_remainder,
1327	n/a	im_has_decimal, &locale, &tmp_format,
1328	n/a	&maxchar);
1329	n/a
1330	n/a	if (skip_re)
1331	n/a	n_re_total = 0;
1332	n/a
1333	n/a	/* Add 1 for the 'j', and optionally 2 for parens. */
1334	n/a	calc_padding(n_re_total + n_im_total + 1 + add_parens * 2,
1335	n/a	format->width, format->align, &lpad, &rpad, &total);
1336	n/a
1337	n/a	if (lpad \|\| rpad)
1338	n/a	maxchar = Py_MAX(maxchar, format->fill_char);
1339	n/a
1340	n/a	if (_PyUnicodeWriter_Prepare(writer, total, maxchar) == -1)
1341	n/a	goto done;
1342	n/a	rkind = writer->kind;
1343	n/a	rdata = writer->data;
1344	n/a
1345	n/a	/* Populate the memory. First, the padding. */
1346	n/a	result = fill_padding(writer,
1347	n/a	n_re_total + n_im_total + 1 + add_parens * 2,
1348	n/a	format->fill_char, lpad, rpad);
1349	n/a	if (result == -1)
1350	n/a	goto done;
1351	n/a
1352	n/a	if (add_parens) {
1353	n/a	PyUnicode_WRITE(rkind, rdata, writer->pos, '(');
1354	n/a	writer->pos++;
1355	n/a	}
1356	n/a
1357	n/a	if (!skip_re) {
1358	n/a	result = fill_number(writer, &re_spec,
1359	n/a	re_unicode_tmp, i_re, i_re + n_re_digits,
1360	n/a	NULL, 0,
1361	n/a	0,
1362	n/a	&locale, 0);
1363	n/a	if (result == -1)
1364	n/a	goto done;
1365	n/a	}
1366	n/a	result = fill_number(writer, &im_spec,
1367	n/a	im_unicode_tmp, i_im, i_im + n_im_digits,
1368	n/a	NULL, 0,
1369	n/a	0,
1370	n/a	&locale, 0);
1371	n/a	if (result == -1)
1372	n/a	goto done;
1373	n/a	PyUnicode_WRITE(rkind, rdata, writer->pos, 'j');
1374	n/a	writer->pos++;
1375	n/a
1376	n/a	if (add_parens) {
1377	n/a	PyUnicode_WRITE(rkind, rdata, writer->pos, ')');
1378	n/a	writer->pos++;
1379	n/a	}
1380	n/a
1381	n/a	writer->pos += rpad;
1382	n/a
1383	n/a	done:
1384	n/a	PyMem_Free(re_buf);
1385	n/a	PyMem_Free(im_buf);
1386	n/a	Py_XDECREF(re_unicode_tmp);
1387	n/a	Py_XDECREF(im_unicode_tmp);
1388	n/a	free_locale_info(&locale);
1389	n/a	return result;
1390	n/a	}
1391	n/a
1392	n/a	/************************************************************************/
1393	n/a	/********* built in formatters **************************************/
1394	n/a	/************************************************************************/
1395	n/a	static int
1396	n/a	format_obj(PyObject obj, _PyUnicodeWriter writer)
1397	n/a	{
1398	n/a	PyObject *str;
1399	n/a	int err;
1400	n/a
1401	n/a	str = PyObject_Str(obj);
1402	n/a	if (str == NULL)
1403	n/a	return -1;
1404	n/a	err = _PyUnicodeWriter_WriteStr(writer, str);
1405	n/a	Py_DECREF(str);
1406	n/a	return err;
1407	n/a	}
1408	n/a
1409	n/a	int
1410	n/a	_PyUnicode_FormatAdvancedWriter(_PyUnicodeWriter *writer,
1411	n/a	PyObject *obj,
1412	n/a	PyObject *format_spec,
1413	n/a	Py_ssize_t start, Py_ssize_t end)
1414	n/a	{
1415	n/a	InternalFormatSpec format;
1416	n/a
1417	n/a	assert(PyUnicode_Check(obj));
1418	n/a
1419	n/a	/* check for the special case of zero length format spec, make
1420	n/a	it equivalent to str(obj) */
1421	n/a	if (start == end) {
1422	n/a	if (PyUnicode_CheckExact(obj))
1423	n/a	return _PyUnicodeWriter_WriteStr(writer, obj);
1424	n/a	else
1425	n/a	return format_obj(obj, writer);
1426	n/a	}
1427	n/a
1428	n/a	/* parse the format_spec */
1429	n/a	if (!parse_internal_render_format_spec(format_spec, start, end,
1430	n/a	&format, 's', '<'))
1431	n/a	return -1;
1432	n/a
1433	n/a	/* type conversion? */
1434	n/a	switch (format.type) {
1435	n/a	case 's':
1436	n/a	/* no type conversion needed, already a string. do the formatting */
1437	n/a	return format_string_internal(obj, &format, writer);
1438	n/a	default:
1439	n/a	/* unknown */
1440	n/a	unknown_presentation_type(format.type, obj->ob_type->tp_name);
1441	n/a	return -1;
1442	n/a	}
1443	n/a	}
1444	n/a
1445	n/a	int
1446	n/a	_PyLong_FormatAdvancedWriter(_PyUnicodeWriter *writer,
1447	n/a	PyObject *obj,
1448	n/a	PyObject *format_spec,
1449	n/a	Py_ssize_t start, Py_ssize_t end)
1450	n/a	{
1451	n/a	PyObject tmp = NULL, str = NULL;
1452	n/a	InternalFormatSpec format;
1453	n/a	int result = -1;
1454	n/a
1455	n/a	/* check for the special case of zero length format spec, make
1456	n/a	it equivalent to str(obj) */
1457	n/a	if (start == end) {
1458	n/a	if (PyLong_CheckExact(obj))
1459	n/a	return _PyLong_FormatWriter(writer, obj, 10, 0);
1460	n/a	else
1461	n/a	return format_obj(obj, writer);
1462	n/a	}
1463	n/a
1464	n/a	/* parse the format_spec */
1465	n/a	if (!parse_internal_render_format_spec(format_spec, start, end,
1466	n/a	&format, 'd', '>'))
1467	n/a	goto done;
1468	n/a
1469	n/a	/* type conversion? */
1470	n/a	switch (format.type) {
1471	n/a	case 'b':
1472	n/a	case 'c':
1473	n/a	case 'd':
1474	n/a	case 'o':
1475	n/a	case 'x':
1476	n/a	case 'X':
1477	n/a	case 'n':
1478	n/a	/* no type conversion needed, already an int. do the formatting */
1479	n/a	result = format_long_internal(obj, &format, writer);
1480	n/a	break;
1481	n/a
1482	n/a	case 'e':
1483	n/a	case 'E':
1484	n/a	case 'f':
1485	n/a	case 'F':
1486	n/a	case 'g':
1487	n/a	case 'G':
1488	n/a	case '%':
1489	n/a	/* convert to float */
1490	n/a	tmp = PyNumber_Float(obj);
1491	n/a	if (tmp == NULL)
1492	n/a	goto done;
1493	n/a	result = format_float_internal(tmp, &format, writer);
1494	n/a	break;
1495	n/a
1496	n/a	default:
1497	n/a	/* unknown */
1498	n/a	unknown_presentation_type(format.type, obj->ob_type->tp_name);
1499	n/a	goto done;
1500	n/a	}
1501	n/a
1502	n/a	done:
1503	n/a	Py_XDECREF(tmp);
1504	n/a	Py_XDECREF(str);
1505	n/a	return result;
1506	n/a	}
1507	n/a
1508	n/a	int
1509	n/a	_PyFloat_FormatAdvancedWriter(_PyUnicodeWriter *writer,
1510	n/a	PyObject *obj,
1511	n/a	PyObject *format_spec,
1512	n/a	Py_ssize_t start, Py_ssize_t end)
1513	n/a	{
1514	n/a	InternalFormatSpec format;
1515	n/a
1516	n/a	/* check for the special case of zero length format spec, make
1517	n/a	it equivalent to str(obj) */
1518	n/a	if (start == end)
1519	n/a	return format_obj(obj, writer);
1520	n/a
1521	n/a	/* parse the format_spec */
1522	n/a	if (!parse_internal_render_format_spec(format_spec, start, end,
1523	n/a	&format, '\0', '>'))
1524	n/a	return -1;
1525	n/a
1526	n/a	/* type conversion? */
1527	n/a	switch (format.type) {
1528	n/a	case '\0': /* No format code: like 'g', but with at least one decimal. */
1529	n/a	case 'e':
1530	n/a	case 'E':
1531	n/a	case 'f':
1532	n/a	case 'F':
1533	n/a	case 'g':
1534	n/a	case 'G':
1535	n/a	case 'n':
1536	n/a	case '%':
1537	n/a	/* no conversion, already a float. do the formatting */
1538	n/a	return format_float_internal(obj, &format, writer);
1539	n/a
1540	n/a	default:
1541	n/a	/* unknown */
1542	n/a	unknown_presentation_type(format.type, obj->ob_type->tp_name);
1543	n/a	return -1;
1544	n/a	}
1545	n/a	}
1546	n/a
1547	n/a	int
1548	n/a	_PyComplex_FormatAdvancedWriter(_PyUnicodeWriter *writer,
1549	n/a	PyObject *obj,
1550	n/a	PyObject *format_spec,
1551	n/a	Py_ssize_t start, Py_ssize_t end)
1552	n/a	{
1553	n/a	InternalFormatSpec format;
1554	n/a
1555	n/a	/* check for the special case of zero length format spec, make
1556	n/a	it equivalent to str(obj) */
1557	n/a	if (start == end)
1558	n/a	return format_obj(obj, writer);
1559	n/a
1560	n/a	/* parse the format_spec */
1561	n/a	if (!parse_internal_render_format_spec(format_spec, start, end,
1562	n/a	&format, '\0', '>'))
1563	n/a	return -1;
1564	n/a
1565	n/a	/* type conversion? */
1566	n/a	switch (format.type) {
1567	n/a	case '\0': /* No format code: like 'g', but with at least one decimal. */
1568	n/a	case 'e':
1569	n/a	case 'E':
1570	n/a	case 'f':
1571	n/a	case 'F':
1572	n/a	case 'g':
1573	n/a	case 'G':
1574	n/a	case 'n':
1575	n/a	/* no conversion, already a complex. do the formatting */
1576	n/a	return format_complex_internal(obj, &format, writer);
1577	n/a
1578	n/a	default:
1579	n/a	/* unknown */
1580	n/a	unknown_presentation_type(format.type, obj->ob_type->tp_name);
1581	n/a	return -1;
1582	n/a	}
1583	n/a	}