blob: 531bc223ff3241e7f969b2b2b4d68931e80fa92f [file] [log] [blame]
Eric Smitha9f7d622008-02-17 19:46:49 +00001/* implements the string, long, and float formatters. that is,
2 string.__format__, etc. */
3
4/* Before including this, you must include either:
5 stringlib/unicodedefs.h
6 stringlib/stringdefs.h
7
8 Also, you should define the names:
9 FORMAT_STRING
10 FORMAT_LONG
11 FORMAT_FLOAT
12 to be whatever you want the public names of these functions to
13 be. These are the only non-static functions defined here.
14*/
15
16#define ALLOW_PARENS_FOR_SIGN 0
17
18/*
19 get_integer consumes 0 or more decimal digit characters from an
20 input string, updates *result with the corresponding positive
21 integer, and returns the number of digits consumed.
22
23 returns -1 on error.
24*/
25static int
26get_integer(STRINGLIB_CHAR **ptr, STRINGLIB_CHAR *end,
27 Py_ssize_t *result)
28{
29 Py_ssize_t accumulator, digitval, oldaccumulator;
30 int numdigits;
31 accumulator = numdigits = 0;
32 for (;;(*ptr)++, numdigits++) {
33 if (*ptr >= end)
34 break;
35 digitval = STRINGLIB_TODECIMAL(**ptr);
36 if (digitval < 0)
37 break;
38 /*
39 This trick was copied from old Unicode format code. It's cute,
40 but would really suck on an old machine with a slow divide
41 implementation. Fortunately, in the normal case we do not
42 expect too many digits.
43 */
44 oldaccumulator = accumulator;
45 accumulator *= 10;
46 if ((accumulator+10)/10 != oldaccumulator+1) {
47 PyErr_Format(PyExc_ValueError,
48 "Too many decimal digits in format string");
49 return -1;
50 }
51 accumulator += digitval;
52 }
53 *result = accumulator;
54 return numdigits;
55}
56
57/************************************************************************/
58/*********** standard format specifier parsing **************************/
59/************************************************************************/
60
61/* returns true if this character is a specifier alignment token */
62Py_LOCAL_INLINE(int)
63is_alignment_token(STRINGLIB_CHAR c)
64{
65 switch (c) {
66 case '<': case '>': case '=': case '^':
67 return 1;
68 default:
69 return 0;
70 }
71}
72
73/* returns true if this character is a sign element */
74Py_LOCAL_INLINE(int)
75is_sign_element(STRINGLIB_CHAR c)
76{
77 switch (c) {
78 case ' ': case '+': case '-':
79#if ALLOW_PARENS_FOR_SIGN
80 case '(':
81#endif
82 return 1;
83 default:
84 return 0;
85 }
86}
87
88
89typedef struct {
90 STRINGLIB_CHAR fill_char;
91 STRINGLIB_CHAR align;
92 STRINGLIB_CHAR sign;
93 Py_ssize_t width;
94 Py_ssize_t precision;
95 STRINGLIB_CHAR type;
96} InternalFormatSpec;
97
98/*
99 ptr points to the start of the format_spec, end points just past its end.
100 fills in format with the parsed information.
101 returns 1 on success, 0 on failure.
102 if failure, sets the exception
103*/
104static int
105parse_internal_render_format_spec(PyObject *format_spec,
106 InternalFormatSpec *format,
107 char default_type)
108{
109 STRINGLIB_CHAR *ptr = STRINGLIB_STR(format_spec);
110 STRINGLIB_CHAR *end = ptr + STRINGLIB_LEN(format_spec);
111
112 /* end-ptr is used throughout this code to specify the length of
113 the input string */
114
115 Py_ssize_t specified_width;
116
117 format->fill_char = '\0';
118 format->align = '\0';
119 format->sign = '\0';
120 format->width = -1;
121 format->precision = -1;
122 format->type = default_type;
123
124 /* If the second char is an alignment token,
125 then parse the fill char */
126 if (end-ptr >= 2 && is_alignment_token(ptr[1])) {
127 format->align = ptr[1];
128 format->fill_char = ptr[0];
129 ptr += 2;
130 }
131 else if (end-ptr >= 1 && is_alignment_token(ptr[0])) {
132 format->align = ptr[0];
Eric Smith8a803dd2008-02-20 23:39:28 +0000133 ++ptr;
Eric Smitha9f7d622008-02-17 19:46:49 +0000134 }
135
136 /* Parse the various sign options */
137 if (end-ptr >= 1 && is_sign_element(ptr[0])) {
138 format->sign = ptr[0];
Eric Smith8a803dd2008-02-20 23:39:28 +0000139 ++ptr;
Eric Smitha9f7d622008-02-17 19:46:49 +0000140#if ALLOW_PARENS_FOR_SIGN
141 if (end-ptr >= 1 && ptr[0] == ')') {
Eric Smith8a803dd2008-02-20 23:39:28 +0000142 ++ptr;
Eric Smitha9f7d622008-02-17 19:46:49 +0000143 }
144#endif
145 }
146
147 /* The special case for 0-padding (backwards compat) */
148 if (format->fill_char == '\0' && end-ptr >= 1 && ptr[0] == '0') {
149 format->fill_char = '0';
150 if (format->align == '\0') {
151 format->align = '=';
152 }
Eric Smith8a803dd2008-02-20 23:39:28 +0000153 ++ptr;
Eric Smitha9f7d622008-02-17 19:46:49 +0000154 }
155
156 /* XXX add error checking */
157 specified_width = get_integer(&ptr, end, &format->width);
158
159 /* if specified_width is 0, we didn't consume any characters for
160 the width. in that case, reset the width to -1, because
161 get_integer() will have set it to zero */
162 if (specified_width == 0) {
163 format->width = -1;
164 }
165
166 /* Parse field precision */
167 if (end-ptr && ptr[0] == '.') {
Eric Smith8a803dd2008-02-20 23:39:28 +0000168 ++ptr;
Eric Smitha9f7d622008-02-17 19:46:49 +0000169
170 /* XXX add error checking */
171 specified_width = get_integer(&ptr, end, &format->precision);
172
173 /* not having a precision after a dot is an error */
174 if (specified_width == 0) {
175 PyErr_Format(PyExc_ValueError,
176 "Format specifier missing precision");
177 return 0;
178 }
179
180 }
181
182 /* Finally, parse the type field */
183
184 if (end-ptr > 1) {
185 /* invalid conversion spec */
186 PyErr_Format(PyExc_ValueError, "Invalid conversion specification");
187 return 0;
188 }
189
190 if (end-ptr == 1) {
191 format->type = ptr[0];
Eric Smith8a803dd2008-02-20 23:39:28 +0000192 ++ptr;
Eric Smitha9f7d622008-02-17 19:46:49 +0000193 }
194
195 return 1;
196}
197
198#if defined FORMAT_FLOAT || defined FORMAT_LONG
199/************************************************************************/
200/*********** common routines for numeric formatting *********************/
201/************************************************************************/
202
203/* describes the layout for an integer, see the comment in
204 _calc_integer_widths() for details */
205typedef struct {
206 Py_ssize_t n_lpadding;
207 Py_ssize_t n_spadding;
208 Py_ssize_t n_rpadding;
209 char lsign;
210 Py_ssize_t n_lsign;
211 char rsign;
212 Py_ssize_t n_rsign;
213 Py_ssize_t n_total; /* just a convenience, it's derivable from the
214 other fields */
215} NumberFieldWidths;
216
217/* not all fields of format are used. for example, precision is
218 unused. should this take discrete params in order to be more clear
219 about what it does? or is passing a single format parameter easier
220 and more efficient enough to justify a little obfuscation? */
221static void
222calc_number_widths(NumberFieldWidths *r, STRINGLIB_CHAR actual_sign,
223 Py_ssize_t n_digits, const InternalFormatSpec *format)
224{
225 r->n_lpadding = 0;
226 r->n_spadding = 0;
227 r->n_rpadding = 0;
228 r->lsign = '\0';
229 r->n_lsign = 0;
230 r->rsign = '\0';
231 r->n_rsign = 0;
232
233 /* the output will look like:
234 | |
235 | <lpadding> <lsign> <spadding> <digits> <rsign> <rpadding> |
236 | |
237
238 lsign and rsign are computed from format->sign and the actual
239 sign of the number
240
241 digits is already known
242
243 the total width is either given, or computed from the
244 actual digits
245
246 only one of lpadding, spadding, and rpadding can be non-zero,
247 and it's calculated from the width and other fields
248 */
249
250 /* compute the various parts we're going to write */
251 if (format->sign == '+') {
252 /* always put a + or - */
253 r->n_lsign = 1;
254 r->lsign = (actual_sign == '-' ? '-' : '+');
255 }
256#if ALLOW_PARENS_FOR_SIGN
257 else if (format->sign == '(') {
258 if (actual_sign == '-') {
259 r->n_lsign = 1;
260 r->lsign = '(';
261 r->n_rsign = 1;
262 r->rsign = ')';
263 }
264 }
265#endif
266 else if (format->sign == ' ') {
267 r->n_lsign = 1;
268 r->lsign = (actual_sign == '-' ? '-' : ' ');
269 }
270 else {
271 /* non specified, or the default (-) */
272 if (actual_sign == '-') {
273 r->n_lsign = 1;
274 r->lsign = '-';
275 }
276 }
277
278 /* now the number of padding characters */
279 if (format->width == -1) {
280 /* no padding at all, nothing to do */
281 }
282 else {
283 /* see if any padding is needed */
284 if (r->n_lsign + n_digits + r->n_rsign >= format->width) {
285 /* no padding needed, we're already bigger than the
286 requested width */
287 }
288 else {
289 /* determine which of left, space, or right padding is
290 needed */
291 Py_ssize_t padding = format->width -
292 (r->n_lsign + n_digits + r->n_rsign);
293 if (format->align == '<')
294 r->n_rpadding = padding;
295 else if (format->align == '>')
296 r->n_lpadding = padding;
297 else if (format->align == '^') {
298 r->n_lpadding = padding / 2;
299 r->n_rpadding = padding - r->n_lpadding;
300 }
301 else if (format->align == '=')
302 r->n_spadding = padding;
303 else
304 r->n_lpadding = padding;
305 }
306 }
307 r->n_total = r->n_lpadding + r->n_lsign + r->n_spadding +
308 n_digits + r->n_rsign + r->n_rpadding;
309}
310
311/* fill in the non-digit parts of a numbers's string representation,
312 as determined in _calc_integer_widths(). returns the pointer to
313 where the digits go. */
314static STRINGLIB_CHAR *
315fill_number(STRINGLIB_CHAR *p_buf, const NumberFieldWidths *spec,
316 Py_ssize_t n_digits, STRINGLIB_CHAR fill_char)
317{
318 STRINGLIB_CHAR* p_digits;
319
320 if (spec->n_lpadding) {
321 STRINGLIB_FILL(p_buf, fill_char, spec->n_lpadding);
322 p_buf += spec->n_lpadding;
323 }
324 if (spec->n_lsign == 1) {
325 *p_buf++ = spec->lsign;
326 }
327 if (spec->n_spadding) {
328 STRINGLIB_FILL(p_buf, fill_char, spec->n_spadding);
329 p_buf += spec->n_spadding;
330 }
331 p_digits = p_buf;
332 p_buf += n_digits;
333 if (spec->n_rsign == 1) {
334 *p_buf++ = spec->rsign;
335 }
336 if (spec->n_rpadding) {
337 STRINGLIB_FILL(p_buf, fill_char, spec->n_rpadding);
338 p_buf += spec->n_rpadding;
339 }
340 return p_digits;
341}
342#endif /* FORMAT_FLOAT || FORMAT_LONG */
343
344/************************************************************************/
345/*********** string formatting ******************************************/
346/************************************************************************/
347
348static PyObject *
349format_string_internal(PyObject *value, const InternalFormatSpec *format)
350{
351 Py_ssize_t width; /* total field width */
352 Py_ssize_t lpad;
353 STRINGLIB_CHAR *dst;
354 STRINGLIB_CHAR *src = STRINGLIB_STR(value);
355 Py_ssize_t len = STRINGLIB_LEN(value);
356 PyObject *result = NULL;
357
358 /* sign is not allowed on strings */
359 if (format->sign != '\0') {
360 PyErr_SetString(PyExc_ValueError,
361 "Sign not allowed in string format specifier");
362 goto done;
363 }
364
365 /* '=' alignment not allowed on strings */
366 if (format->align == '=') {
367 PyErr_SetString(PyExc_ValueError,
368 "'=' alignment not allowed "
369 "in string format specifier");
370 goto done;
371 }
372
373 /* if precision is specified, output no more that format.precision
374 characters */
375 if (format->precision >= 0 && len >= format->precision) {
376 len = format->precision;
377 }
378
379 if (format->width >= 0) {
380 width = format->width;
381
382 /* but use at least len characters */
383 if (len > width) {
384 width = len;
385 }
386 }
387 else {
388 /* not specified, use all of the chars and no more */
389 width = len;
390 }
391
392 /* allocate the resulting string */
393 result = STRINGLIB_NEW(NULL, width);
394 if (result == NULL)
395 goto done;
396
397 /* now write into that space */
398 dst = STRINGLIB_STR(result);
399
400 /* figure out how much leading space we need, based on the
401 aligning */
402 if (format->align == '>')
403 lpad = width - len;
404 else if (format->align == '^')
405 lpad = (width - len) / 2;
406 else
407 lpad = 0;
408
409 /* if right aligning, increment the destination allow space on the
410 left */
411 memcpy(dst + lpad, src, len * sizeof(STRINGLIB_CHAR));
412
413 /* do any padding */
414 if (width > len) {
415 STRINGLIB_CHAR fill_char = format->fill_char;
416 if (fill_char == '\0') {
417 /* use the default, if not specified */
418 fill_char = ' ';
419 }
420
421 /* pad on left */
422 if (lpad)
423 STRINGLIB_FILL(dst, fill_char, lpad);
424
425 /* pad on right */
426 if (width - len - lpad)
427 STRINGLIB_FILL(dst + len + lpad, fill_char, width - len - lpad);
428 }
429
430done:
431 return result;
432}
433
434
435/************************************************************************/
436/*********** long formatting ********************************************/
437/************************************************************************/
438
439#if defined FORMAT_LONG || defined FORMAT_INT
440typedef PyObject*
441(*IntOrLongToString)(PyObject *value, int base);
442
443static PyObject *
444format_int_or_long_internal(PyObject *value, const InternalFormatSpec *format,
445 IntOrLongToString tostring)
446{
447 PyObject *result = NULL;
448 PyObject *tmp = NULL;
449 STRINGLIB_CHAR *pnumeric_chars;
450 STRINGLIB_CHAR numeric_char;
451 STRINGLIB_CHAR sign = '\0';
452 STRINGLIB_CHAR *p;
453 Py_ssize_t n_digits; /* count of digits need from the computed
454 string */
455 Py_ssize_t n_leading_chars;
456 NumberFieldWidths spec;
457 long x;
458
459 /* no precision allowed on integers */
460 if (format->precision != -1) {
461 PyErr_SetString(PyExc_ValueError,
462 "Precision not allowed in integer format specifier");
463 goto done;
464 }
465
466
467 /* special case for character formatting */
468 if (format->type == 'c') {
469 /* error to specify a sign */
470 if (format->sign != '\0') {
471 PyErr_SetString(PyExc_ValueError,
472 "Sign not allowed with integer"
473 " format specifier 'c'");
474 goto done;
475 }
476
477 /* taken from unicodeobject.c formatchar() */
478 /* Integer input truncated to a character */
479/* XXX: won't work for int */
480 x = PyLong_AsLong(value);
481 if (x == -1 && PyErr_Occurred())
482 goto done;
483#ifdef Py_UNICODE_WIDE
484 if (x < 0 || x > 0x10ffff) {
485 PyErr_SetString(PyExc_OverflowError,
486 "%c arg not in range(0x110000) "
487 "(wide Python build)");
488 goto done;
489 }
490#else
491 if (x < 0 || x > 0xffff) {
492 PyErr_SetString(PyExc_OverflowError,
493 "%c arg not in range(0x10000) "
494 "(narrow Python build)");
495 goto done;
496 }
497#endif
498 numeric_char = (STRINGLIB_CHAR)x;
499 pnumeric_chars = &numeric_char;
500 n_digits = 1;
501 }
502 else {
503 int base;
504 int leading_chars_to_skip; /* Number of characters added by
505 PyNumber_ToBase that we want to
506 skip over. */
507
508 /* Compute the base and how many characters will be added by
509 PyNumber_ToBase */
510 switch (format->type) {
511 case 'b':
512 base = 2;
513 leading_chars_to_skip = 2; /* 0b */
514 break;
515 case 'o':
516 base = 8;
517 leading_chars_to_skip = 2; /* 0o */
518 break;
519 case 'x':
520 case 'X':
521 base = 16;
522 leading_chars_to_skip = 2; /* 0x */
523 break;
524 default: /* shouldn't be needed, but stops a compiler warning */
525 case 'd':
526 base = 10;
527 leading_chars_to_skip = 0;
528 break;
529 }
530
531 /* Do the hard part, converting to a string in a given base */
532 tmp = tostring(value, base);
533 if (tmp == NULL)
534 goto done;
535
536 pnumeric_chars = STRINGLIB_STR(tmp);
537 n_digits = STRINGLIB_LEN(tmp);
538
539 /* Remember not to modify what pnumeric_chars points to. it
540 might be interned. Only modify it after we copy it into a
541 newly allocated output buffer. */
542
543 /* Is a sign character present in the output? If so, remember it
544 and skip it */
545 sign = pnumeric_chars[0];
546 if (sign == '-') {
547 ++leading_chars_to_skip;
548 }
549
550 /* Skip over the leading chars (0x, 0b, etc.) */
551 n_digits -= leading_chars_to_skip;
552 pnumeric_chars += leading_chars_to_skip;
553 }
554
555 /* Calculate the widths of the various leading and trailing parts */
556 calc_number_widths(&spec, sign, n_digits, format);
557
558 /* Allocate a new string to hold the result */
559 result = STRINGLIB_NEW(NULL, spec.n_total);
560 if (!result)
561 goto done;
562 p = STRINGLIB_STR(result);
563
564 /* Fill in the digit parts */
565 n_leading_chars = spec.n_lpadding + spec.n_lsign + spec.n_spadding;
566 memmove(p + n_leading_chars,
567 pnumeric_chars,
568 n_digits * sizeof(STRINGLIB_CHAR));
569
570 /* if X, convert to uppercase */
571 if (format->type == 'X') {
572 Py_ssize_t t;
Eric Smith8a803dd2008-02-20 23:39:28 +0000573 for (t = 0; t < n_digits; ++t)
Eric Smitha9f7d622008-02-17 19:46:49 +0000574 p[t + n_leading_chars] = STRINGLIB_TOUPPER(p[t + n_leading_chars]);
575 }
576
577 /* Fill in the non-digit parts */
578 fill_number(p, &spec, n_digits,
579 format->fill_char == '\0' ? ' ' : format->fill_char);
580
581done:
582 Py_XDECREF(tmp);
583 return result;
584}
585#endif /* defined FORMAT_LONG || defined FORMAT_INT */
586
587/************************************************************************/
588/*********** float formatting *******************************************/
589/************************************************************************/
590
591#ifdef FORMAT_FLOAT
592#if STRINGLIB_IS_UNICODE
593/* taken from unicodeobject.c */
594static Py_ssize_t
595strtounicode(Py_UNICODE *buffer, const char *charbuffer)
596{
597 register Py_ssize_t i;
598 Py_ssize_t len = strlen(charbuffer);
Eric Smith8a803dd2008-02-20 23:39:28 +0000599 for (i = len - 1; i >= 0; --i)
Eric Smitha9f7d622008-02-17 19:46:49 +0000600 buffer[i] = (Py_UNICODE) charbuffer[i];
601
602 return len;
603}
604#endif
605
Eric Smitha9f7d622008-02-17 19:46:49 +0000606/* see FORMATBUFLEN in unicodeobject.c */
607#define FLOAT_FORMATBUFLEN 120
608
609/* much of this is taken from unicodeobject.c */
Eric Smitha9f7d622008-02-17 19:46:49 +0000610static PyObject *
Eric Smith8a803dd2008-02-20 23:39:28 +0000611format_float_internal(PyObject *value,
612 const InternalFormatSpec *format)
Eric Smitha9f7d622008-02-17 19:46:49 +0000613{
614 /* fmt = '%.' + `prec` + `type` + '%%'
615 worst case length = 2 + 10 (len of INT_MAX) + 1 + 2 = 15 (use 20)*/
616 char fmt[20];
617
618 /* taken from unicodeobject.c */
619 /* Worst case length calc to ensure no buffer overrun:
620
621 'g' formats:
622 fmt = %#.<prec>g
623 buf = '-' + [0-9]*prec + '.' + 'e+' + (longest exp
624 for any double rep.)
625 len = 1 + prec + 1 + 2 + 5 = 9 + prec
626
627 'f' formats:
628 buf = '-' + [0-9]*x + '.' + [0-9]*prec (with x < 50)
629 len = 1 + 50 + 1 + prec = 52 + prec
630
631 If prec=0 the effective precision is 1 (the leading digit is
632 always given), therefore increase the length by one.
633
634 */
635 char charbuf[FLOAT_FORMATBUFLEN];
636 Py_ssize_t n_digits;
637 double x;
638 Py_ssize_t precision = format->precision;
639 PyObject *result = NULL;
640 STRINGLIB_CHAR sign;
641 char* trailing = "";
642 STRINGLIB_CHAR *p;
643 NumberFieldWidths spec;
Eric Smith8a803dd2008-02-20 23:39:28 +0000644 STRINGLIB_CHAR type = format->type;
Eric Smitha9f7d622008-02-17 19:46:49 +0000645
646#if STRINGLIB_IS_UNICODE
647 Py_UNICODE unicodebuf[FLOAT_FORMATBUFLEN];
648#endif
649
650 /* first, do the conversion as 8-bit chars, using the platform's
651 snprintf. then, if needed, convert to unicode. */
652
653 /* 'F' is the same as 'f', per the PEP */
654 if (type == 'F')
655 type = 'f';
656
657 x = PyFloat_AsDouble(value);
658
659 if (x == -1.0 && PyErr_Occurred())
660 goto done;
661
662 if (type == '%') {
663 type = 'f';
664 x *= 100;
665 trailing = "%";
666 }
667
668 if (precision < 0)
669 precision = 6;
670 if (type == 'f' && (fabs(x) / 1e25) >= 1e25)
671 type = 'g';
672
673 /* cast "type", because if we're in unicode we need to pass a
674 8-bit char. this is safe, because we've restricted what "type"
675 can be */
676 PyOS_snprintf(fmt, sizeof(fmt), "%%.%" PY_FORMAT_SIZE_T "d%c", precision,
677 (char)type);
678
Eric Smith8a803dd2008-02-20 23:39:28 +0000679 /* do the actual formatting */
680 PyOS_ascii_formatd(charbuf, sizeof(charbuf), fmt, x);
Eric Smitha9f7d622008-02-17 19:46:49 +0000681
682 /* adding trailing to fmt with PyOS_snprintf doesn't work, not
683 sure why. we'll just concatentate it here, no harm done. we
684 know we can't have a buffer overflow from the fmt size
685 analysis */
686 strcat(charbuf, trailing);
687
688 /* rather than duplicate the code for snprintf for both unicode
689 and 8 bit strings, we just use the 8 bit version and then
690 convert to unicode in a separate code path. that's probably
691 the lesser of 2 evils. */
692#if STRINGLIB_IS_UNICODE
693 n_digits = strtounicode(unicodebuf, charbuf);
694 p = unicodebuf;
695#else
696 /* compute the length. I believe this is done because the return
697 value from snprintf above is unreliable */
698 n_digits = strlen(charbuf);
699 p = charbuf;
700#endif
701
702 /* is a sign character present in the output? if so, remember it
703 and skip it */
704 sign = p[0];
705 if (sign == '-') {
Eric Smith8a803dd2008-02-20 23:39:28 +0000706 ++p;
707 --n_digits;
Eric Smitha9f7d622008-02-17 19:46:49 +0000708 }
709
710 calc_number_widths(&spec, sign, n_digits, format);
711
712 /* allocate a string with enough space */
713 result = STRINGLIB_NEW(NULL, spec.n_total);
714 if (result == NULL)
715 goto done;
716
717 /* fill in the non-digit parts */
718 fill_number(STRINGLIB_STR(result), &spec, n_digits,
719 format->fill_char == '\0' ? ' ' : format->fill_char);
720
721 /* fill in the digit parts */
722 memmove(STRINGLIB_STR(result) +
723 (spec.n_lpadding + spec.n_lsign + spec.n_spadding),
724 p,
725 n_digits * sizeof(STRINGLIB_CHAR));
726
727done:
728 return result;
729}
Eric Smitha9f7d622008-02-17 19:46:49 +0000730#endif /* FORMAT_FLOAT */
731
732/************************************************************************/
733/*********** built in formatters ****************************************/
734/************************************************************************/
735#ifdef FORMAT_STRING
736PyObject *
737FORMAT_STRING(PyObject* value, PyObject* args)
738{
739 PyObject *format_spec;
740 PyObject *result = NULL;
741#if PY_VERSION_HEX < 0x03000000
742 PyObject *tmp = NULL;
743#endif
744 InternalFormatSpec format;
745
746 /* If 2.x, we accept either str or unicode, and try to convert it
747 to the right type. In 3.x, we insist on only unicode */
748#if PY_VERSION_HEX >= 0x03000000
749 if (!PyArg_ParseTuple(args, STRINGLIB_PARSE_CODE ":__format__",
750 &format_spec))
751 goto done;
752#else
753 /* If 2.x, convert format_spec to the same type as value */
754 /* This is to allow things like u''.format('') */
755 if (!PyArg_ParseTuple(args, "O:__format__", &format_spec))
756 goto done;
757 if (!(PyString_Check(format_spec) || PyUnicode_Check(format_spec))) {
758 PyErr_Format(PyExc_TypeError, "__format__ arg must be str "
759 "or unicode, not %s", Py_TYPE(format_spec)->tp_name);
760 goto done;
761 }
762 tmp = STRINGLIB_TOSTR(format_spec);
763 if (tmp == NULL)
764 goto done;
765 format_spec = tmp;
766#endif
767
768 /* check for the special case of zero length format spec, make
769 it equivalent to str(value) */
770 if (STRINGLIB_LEN(format_spec) == 0) {
771 result = STRINGLIB_TOSTR(value);
772 goto done;
773 }
774
775
776 /* parse the format_spec */
777 if (!parse_internal_render_format_spec(format_spec, &format, 's'))
778 goto done;
779
780 /* type conversion? */
781 switch (format.type) {
782 case 's':
783 /* no type conversion needed, already a string. do the formatting */
784 result = format_string_internal(value, &format);
785 break;
786 default:
787 /* unknown */
Martin v. Löwisd918e4e2008-04-07 03:08:28 +0000788 #if STRINGLIB_IS_UNICODE
789 /* If STRINGLIB_CHAR is Py_UNICODE, %c might be out-of-range,
790 hence the two cases. If it is char, gcc complains that the
791 condition below is always true, hence the ifdef. */
792 if (format.type > 32 && format.type <128)
793 #endif
794 PyErr_Format(PyExc_ValueError, "Unknown conversion type %c",
795 (char)format.type);
796 #if STRINGLIB_IS_UNICODE
797 else
798 PyErr_Format(PyExc_ValueError, "Unknown conversion type '\\x%x'",
799 (unsigned int)format.type);
800 #endif
Eric Smitha9f7d622008-02-17 19:46:49 +0000801 goto done;
802 }
803
804done:
805#if PY_VERSION_HEX < 0x03000000
806 Py_XDECREF(tmp);
807#endif
808 return result;
809}
810#endif /* FORMAT_STRING */
811
812#if defined FORMAT_LONG || defined FORMAT_INT
813static PyObject*
814format_int_or_long(PyObject* value, PyObject* args, IntOrLongToString tostring)
815{
816 PyObject *format_spec;
817 PyObject *result = NULL;
818 PyObject *tmp = NULL;
819 InternalFormatSpec format;
820
821 if (!PyArg_ParseTuple(args, STRINGLIB_PARSE_CODE ":__format__",
822 &format_spec))
823 goto done;
824
825 /* check for the special case of zero length format spec, make
826 it equivalent to str(value) */
827 if (STRINGLIB_LEN(format_spec) == 0) {
828 result = STRINGLIB_TOSTR(value);
829 goto done;
830 }
831
832 /* parse the format_spec */
833 if (!parse_internal_render_format_spec(format_spec, &format, 'd'))
834 goto done;
835
836 /* type conversion? */
837 switch (format.type) {
838 case 'b':
839 case 'c':
840 case 'd':
841 case 'o':
842 case 'x':
843 case 'X':
844 /* no type conversion needed, already an int (or long). do
845 the formatting */
846 result = format_int_or_long_internal(value, &format, tostring);
847 break;
848
849 case 'e':
850 case 'E':
851 case 'f':
852 case 'F':
853 case 'g':
854 case 'G':
855 case 'n':
856 case '%':
857 /* convert to float */
858 tmp = PyNumber_Float(value);
859 if (tmp == NULL)
860 goto done;
861 result = format_float_internal(value, &format);
862 break;
863
864 default:
865 /* unknown */
866 PyErr_Format(PyExc_ValueError, "Unknown conversion type %c",
867 format.type);
868 goto done;
869 }
870
871done:
872 Py_XDECREF(tmp);
873 return result;
874}
875#endif /* FORMAT_LONG || defined FORMAT_INT */
876
877#ifdef FORMAT_LONG
878/* Need to define long_format as a function that will convert a long
879 to a string. In 3.0, _PyLong_Format has the correct signature. In
880 2.x, we need to fudge a few parameters */
881#if PY_VERSION_HEX >= 0x03000000
882#define long_format _PyLong_Format
883#else
884static PyObject*
885long_format(PyObject* value, int base)
886{
887 /* Convert to base, don't add trailing 'L', and use the new octal
888 format. We already know this is a long object */
889 assert(PyLong_Check(value));
890 /* convert to base, don't add 'L', and use the new octal format */
891 return _PyLong_Format(value, base, 0, 1);
892}
893#endif
894
895PyObject *
896FORMAT_LONG(PyObject* value, PyObject* args)
897{
898 return format_int_or_long(value, args, long_format);
899}
900#endif /* FORMAT_LONG */
901
902#ifdef FORMAT_INT
903/* this is only used for 2.x, not 3.0 */
904static PyObject*
905int_format(PyObject* value, int base)
906{
907 /* Convert to base, and use the new octal format. We already
908 know this is an int object */
909 assert(PyInt_Check(value));
910 return _PyInt_Format((PyIntObject*)value, base, 1);
911}
912
913PyObject *
914FORMAT_INT(PyObject* value, PyObject* args)
915{
916 return format_int_or_long(value, args, int_format);
917}
918#endif /* FORMAT_INT */
919
920#ifdef FORMAT_FLOAT
921PyObject *
922FORMAT_FLOAT(PyObject *value, PyObject *args)
923{
924 PyObject *format_spec;
925 PyObject *result = NULL;
926 InternalFormatSpec format;
927
928 if (!PyArg_ParseTuple(args, STRINGLIB_PARSE_CODE ":__format__", &format_spec))
929 goto done;
930
931 /* check for the special case of zero length format spec, make
932 it equivalent to str(value) */
933 if (STRINGLIB_LEN(format_spec) == 0) {
934 result = STRINGLIB_TOSTR(value);
935 goto done;
936 }
937
938 /* parse the format_spec */
Eric Smith8113ca62008-03-17 11:01:01 +0000939 if (!parse_internal_render_format_spec(format_spec, &format, '\0'))
Eric Smitha9f7d622008-02-17 19:46:49 +0000940 goto done;
941
942 /* type conversion? */
943 switch (format.type) {
Eric Smith8113ca62008-03-17 11:01:01 +0000944 case '\0':
945 /* 'Z' means like 'g', but with at least one decimal. See
946 PyOS_ascii_formatd */
947 format.type = 'Z';
948 /* Deliberate fall through to the next case statement */
Eric Smitha9f7d622008-02-17 19:46:49 +0000949 case 'e':
950 case 'E':
951 case 'f':
952 case 'F':
953 case 'g':
954 case 'G':
955 case 'n':
956 case '%':
957 /* no conversion, already a float. do the formatting */
958 result = format_float_internal(value, &format);
959 break;
960
961 default:
962 /* unknown */
963 PyErr_Format(PyExc_ValueError, "Unknown conversion type %c",
964 format.type);
965 goto done;
966 }
967
968done:
969 return result;
970}
971#endif /* FORMAT_FLOAT */