]> git.ipfire.org Git - thirdparty/glibc.git/blob - stdio-common/printf_fp.c
Fri Mar 8 22:16:48 1996 Roland McGrath <roland@charlie-brown.gnu.ai.mit.edu>
[thirdparty/glibc.git] / stdio-common / printf_fp.c
1 /* Floating point output for `printf'.
2 Copyright (C) 1995, 1996 Free Software Foundation, Inc.
3 Written by Ulrich Drepper.
4
5 This file is part of the GNU C Library.
6
7 The GNU C Library is free software; you can redistribute it and/or
8 modify it under the terms of the GNU Library General Public License as
9 published by the Free Software Foundation; either version 2 of the
10 License, or (at your option) any later version.
11
12 The GNU C Library is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 Library General Public License for more details.
16
17 You should have received a copy of the GNU Library General Public
18 License along with the GNU C Library; see the file COPYING.LIB. If
19 not, write to the Free Software Foundation, Inc., 675 Mass Ave,
20 Cambridge, MA 02139, USA. */
21
22 /* The gmp headers need some configuration frobs. */
23 #define HAVE_ALLOCA 1
24
25 #ifdef USE_IN_LIBIO
26 # include <libioP.h>
27 #else
28 # include <stdio.h>
29 #endif
30 #include <alloca.h>
31 #include <ansidecl.h>
32 #include <ctype.h>
33 #include <float.h>
34 #include <gmp-mparam.h>
35 #include "../stdlib/gmp.h"
36 #include "../stdlib/gmp-impl.h"
37 #include "../stdlib/longlong.h"
38 #include "../stdlib/fpioconst.h"
39 #include "../locale/localeinfo.h"
40 #include <limits.h>
41 #include <math.h>
42 #include <printf.h>
43 #include <string.h>
44 #include <unistd.h>
45 #include <stdlib.h>
46
47 #define NDEBUG /* Undefine this for debugging assertions. */
48 #include <assert.h>
49
50 /* This defines make it possible to use the same code for GNU C library and
51 the GNU I/O library. */
52 #ifdef USE_IN_LIBIO
53 # define PUT(f, s, n) _IO_sputn (f, s, n)
54 # define PAD(f, c, n) _IO_padn (f, c, n)
55 /* We use this file GNU C library and GNU I/O library. So make
56 names equal. */
57 # undef putc
58 # define putc(c, f) _IO_putc (c, f)
59 # define size_t _IO_size_t
60 # define FILE _IO_FILE
61 #else /* ! USE_IN_LIBIO */
62 # define PUT(f, s, n) fwrite (s, 1, n, f)
63 # define PAD(f, c, n) __printf_pad (f, c, n)
64 ssize_t __printf_pad __P ((FILE *, char pad, int n)); /* In vfprintf.c. */
65 #endif /* USE_IN_LIBIO */
66 \f
67 /* Macros for doing the actual output. */
68
69 #define outchar(ch) \
70 do \
71 { \
72 register CONST int outc = (ch); \
73 if (putc (outc, fp) == EOF) \
74 return -1; \
75 ++done; \
76 } while (0)
77
78 #define PRINT(ptr, len) \
79 do \
80 { \
81 register size_t outlen = (len); \
82 if (len > 20) \
83 { \
84 if (PUT (fp, ptr, outlen) != outlen) \
85 return -1; \
86 ptr += outlen; \
87 done += outlen; \
88 } \
89 else \
90 { \
91 while (outlen-- > 0) \
92 outchar (*ptr++); \
93 } \
94 } while (0)
95
96 #define PADN(ch, len) \
97 do \
98 { \
99 if (PAD (fp, ch, len) != len) \
100 return -1; \
101 done += len; \
102 } \
103 while (0)
104 \f
105 /* We use the GNU MP library to handle large numbers.
106
107 An MP variable occupies a varying number of entries in its array. We keep
108 track of this number for efficiency reasons. Otherwise we would always
109 have to process the whole array. */
110 #define MPN_VAR(name) mp_limb *name; mp_size_t name##size
111
112 #define MPN_ASSIGN(dst,src) \
113 memcpy (dst, src, (dst##size = src##size) * sizeof (mp_limb))
114 #define MPN_GE(u,v) \
115 (u##size > v##size || (u##size == v##size && __mpn_cmp (u, v, u##size) >= 0))
116
117 extern int __isinfl (long double), __isnanl (long double);
118
119 extern mp_size_t __mpn_extract_double (mp_ptr res_ptr, mp_size_t size,
120 int *expt, int *is_neg,
121 double value);
122 extern mp_size_t __mpn_extract_long_double (mp_ptr res_ptr, mp_size_t size,
123 int *expt, int *is_neg,
124 long double value);
125
126
127 static unsigned int guess_grouping (unsigned int intdig_max,
128 const char *grouping, wchar_t sepchar);
129 static char *group_number (char *buf, char *bufend, unsigned int intdig_no,
130 const char *grouping, wchar_t thousands_sep);
131
132
133 int
134 __printf_fp (FILE *fp,
135 const struct printf_info *info,
136 const *const *args)
137 {
138 /* The floating-point value to output. */
139 union
140 {
141 double dbl;
142 LONG_DOUBLE ldbl;
143 }
144 fpnum;
145
146 /* Locale-dependent representation of decimal point. */
147 wchar_t decimal;
148
149 /* Locale-dependent thousands separator and grouping specification. */
150 wchar_t thousands_sep;
151 const char *grouping;
152
153 /* "NaN" or "Inf" for the special cases. */
154 CONST char *special = NULL;
155
156 /* We need just a few limbs for the input before shifting to the right
157 position. */
158 mp_limb fp_input[(LDBL_MANT_DIG + BITS_PER_MP_LIMB - 1) / BITS_PER_MP_LIMB];
159 /* We need to shift the contents of fp_input by this amount of bits. */
160 int to_shift;
161
162 /* The significant of the floting-point value in question */
163 MPN_VAR(frac);
164 /* and the exponent. */
165 int exponent;
166 /* Sign of the exponent. */
167 int expsign = 0;
168 /* Sign of float number. */
169 int is_neg = 0;
170
171 /* Scaling factor. */
172 MPN_VAR(scale);
173
174 /* Temporary bignum value. */
175 MPN_VAR(tmp);
176
177 /* Digit which is result of last hack_digit() call. */
178 int digit;
179
180 /* The type of output format that will be used: 'e'/'E' or 'f'. */
181 int type;
182
183 /* Counter for number of written characters. */
184 int done = 0;
185
186 /* General helper (carry limb). */
187 mp_limb cy;
188
189 char hack_digit (void)
190 {
191 mp_limb hi;
192
193 if (expsign != 0 && type == 'f' && exponent-- > 0)
194 hi = 0;
195 else if (scalesize == 0)
196 {
197 hi = frac[fracsize - 1];
198 cy = __mpn_mul_1 (frac, frac, fracsize - 1, 10);
199 frac[fracsize - 1] = cy;
200 }
201 else
202 {
203 if (fracsize < scalesize)
204 hi = 0;
205 else
206 {
207 hi = __mpn_divmod (tmp, frac, fracsize, scale, scalesize);
208 tmp[fracsize - scalesize] = hi;
209 hi = tmp[0];
210
211 fracsize = __mpn_normal_size (frac, scalesize);
212 if (fracsize == 0)
213 {
214 /* We're not prepared for an mpn variable with zero
215 limbs. */
216 fracsize = 1;
217 return '0' + hi;
218 }
219 }
220
221 cy = __mpn_mul_1 (frac, frac, fracsize, 10);
222 if (cy != 0)
223 frac[fracsize++] = cy;
224 }
225
226 return '0' + hi;
227 }
228
229
230 /* Figure out the decimal point character. */
231 if (mbtowc (&decimal, _NL_CURRENT (LC_NUMERIC, DECIMAL_POINT),
232 strlen (_NL_CURRENT (LC_NUMERIC, DECIMAL_POINT))) <= 0)
233 decimal = (wchar_t) *_NL_CURRENT (LC_NUMERIC, DECIMAL_POINT);
234
235
236 if (info->group)
237 {
238 grouping = _NL_CURRENT (LC_NUMERIC, GROUPING);
239 if (*grouping <= 0 || *grouping == CHAR_MAX)
240 grouping = NULL;
241 else
242 {
243 /* Figure out the thousands seperator character. */
244 if (mbtowc (&thousands_sep, _NL_CURRENT (LC_NUMERIC, THOUSANDS_SEP),
245 strlen (_NL_CURRENT (LC_NUMERIC, THOUSANDS_SEP))) <= 0)
246 thousands_sep = (wchar_t) *_NL_CURRENT (LC_NUMERIC, THOUSANDS_SEP);
247 if (thousands_sep == L'\0')
248 grouping = NULL;
249 }
250 }
251 else
252 grouping = NULL;
253
254 /* Fetch the argument value. */
255 if (info->is_long_double && sizeof (long double) > sizeof (double))
256 {
257 fpnum.ldbl = *(const long double *) args[0];
258
259 /* Check for special values: not a number or infinity. */
260 if (__isnanl (fpnum.ldbl))
261 {
262 special = "NaN";
263 is_neg = 0;
264 }
265 else if (__isinfl (fpnum.ldbl))
266 {
267 special = "Inf";
268 is_neg = fpnum.ldbl < 0;
269 }
270 else
271 {
272 fracsize = __mpn_extract_long_double (fp_input,
273 (sizeof (fp_input) /
274 sizeof (fp_input[0])),
275 &exponent, &is_neg,
276 fpnum.ldbl);
277 to_shift = 1 + fracsize * BITS_PER_MP_LIMB - LDBL_MANT_DIG;
278 }
279 }
280 else
281 {
282 fpnum.dbl = *(const double *) args[0];
283
284 /* Check for special values: not a number or infinity. */
285 if (__isnan (fpnum.dbl))
286 {
287 special = "NaN";
288 is_neg = 0;
289 }
290 else if (__isinf (fpnum.dbl))
291 {
292 special = "Inf";
293 is_neg = fpnum.dbl < 0;
294 }
295 else
296 {
297 fracsize = __mpn_extract_double (fp_input,
298 (sizeof (fp_input)
299 / sizeof (fp_input[0])),
300 &exponent, &is_neg, fpnum.dbl);
301 to_shift = 1 + fracsize * BITS_PER_MP_LIMB - DBL_MANT_DIG;
302 }
303 }
304
305 if (special)
306 {
307 int width = info->prec > info->width ? info->prec : info->width;
308
309 if (is_neg || info->showsign || info->space)
310 --width;
311 width -= 3;
312
313 if (!info->left && width > 0)
314 PADN (' ', width);
315
316 if (is_neg)
317 outchar ('-');
318 else if (info->showsign)
319 outchar ('+');
320 else if (info->space)
321 outchar (' ');
322
323 PRINT (special, 3);
324
325 if (info->left && width > 0)
326 PADN (' ', width);
327
328 return done;
329 }
330
331
332 /* We need three multiprecision variables. Now that we have the exponent
333 of the number we can allocate the needed memory. It would be more
334 efficient to use variables of the fixed maximum size but because this
335 would be really big it could lead to memory problems. */
336 {
337 mp_size_t bignum_size = ((ABS (exponent) + BITS_PER_MP_LIMB - 1)
338 / BITS_PER_MP_LIMB + 3) * sizeof (mp_limb);
339 frac = (mp_limb *) alloca (bignum_size);
340 tmp = (mp_limb *) alloca (bignum_size);
341 scale = (mp_limb *) alloca (bignum_size);
342 }
343
344 /* We now have to distinguish between numbers with positive and negative
345 exponents because the method used for the one is not applicable/efficient
346 for the other. */
347 scalesize = 0;
348 if (exponent > 2)
349 {
350 /* |FP| >= 8.0. */
351 int scaleexpo = 0;
352 int explog = LDBL_MAX_10_EXP_LOG;
353 int exp10 = 0;
354 const struct mp_power *tens = &_fpioconst_pow10[explog + 1];
355 int cnt_h, cnt_l, i;
356
357 if ((exponent + to_shift) % BITS_PER_MP_LIMB == 0)
358 {
359 MPN_COPY_DECR (frac + (exponent + to_shift) / BITS_PER_MP_LIMB,
360 fp_input, fracsize);
361 fracsize += (exponent + to_shift) / BITS_PER_MP_LIMB;
362 }
363 else
364 {
365 cy = __mpn_lshift (frac + (exponent + to_shift) / BITS_PER_MP_LIMB,
366 fp_input, fracsize,
367 (exponent + to_shift) % BITS_PER_MP_LIMB);
368 fracsize += (exponent + to_shift) / BITS_PER_MP_LIMB;
369 if (cy)
370 frac[fracsize++] = cy;
371 }
372 MPN_ZERO (frac, (exponent + to_shift) / BITS_PER_MP_LIMB);
373
374 assert (tens > &_fpioconst_pow10[0]);
375 do
376 {
377 --tens;
378
379 /* The number of the product of two binary numbers with n and m
380 bits respectively has m+n or m+n-1 bits. */
381 if (exponent >= scaleexpo + tens->p_expo - 1)
382 {
383 if (scalesize == 0)
384 MPN_ASSIGN (tmp, tens->array);
385 else
386 {
387 cy = __mpn_mul (tmp, scale, scalesize,
388 &tens->array[_FPIO_CONST_OFFSET],
389 tens->arraysize - _FPIO_CONST_OFFSET);
390 tmpsize = scalesize + tens->arraysize - _FPIO_CONST_OFFSET;
391 if (cy == 0)
392 --tmpsize;
393 }
394
395 if (MPN_GE (frac, tmp))
396 {
397 int cnt;
398 MPN_ASSIGN (scale, tmp);
399 count_leading_zeros (cnt, scale[scalesize - 1]);
400 scaleexpo = (scalesize - 2) * BITS_PER_MP_LIMB - cnt - 1;
401 exp10 |= 1 << explog;
402 }
403 }
404 --explog;
405 }
406 while (tens > &_fpioconst_pow10[0]);
407 exponent = exp10;
408
409 /* Optimize number representations. We want to represent the numbers
410 with the lowest number of bytes possible without losing any
411 bytes. Also the highest bit in the scaling factor has to be set
412 (this is a requirement of the MPN division routines). */
413 if (scalesize > 0)
414 {
415 /* Determine minimum number of zero bits at the end of
416 both numbers. */
417 for (i = 0; scale[i] == 0 && frac[i] == 0; i++)
418 ;
419
420 /* Determine number of bits the scaling factor is misplaced. */
421 count_leading_zeros (cnt_h, scale[scalesize - 1]);
422
423 if (cnt_h == 0)
424 {
425 /* The highest bit of the scaling factor is already set. So
426 we only have to remove the trailing empty limbs. */
427 if (i > 0)
428 {
429 MPN_COPY_INCR (scale, scale + i, scalesize - i);
430 scalesize -= i;
431 MPN_COPY_INCR (frac, frac + i, fracsize - i);
432 fracsize -= i;
433 }
434 }
435 else
436 {
437 if (scale[i] != 0)
438 {
439 count_trailing_zeros (cnt_l, scale[i]);
440 if (frac[i] != 0)
441 {
442 int cnt_l2;
443 count_trailing_zeros (cnt_l2, frac[i]);
444 if (cnt_l2 < cnt_l)
445 cnt_l = cnt_l2;
446 }
447 }
448 else
449 count_trailing_zeros (cnt_l, frac[i]);
450
451 /* Now shift the numbers to their optimal position. */
452 if (i == 0 && BITS_PER_MP_LIMB - cnt_h > cnt_l)
453 {
454 /* We cannot save any memory. So just roll both numbers
455 so that the scaling factor has its highest bit set. */
456
457 (void) __mpn_lshift (scale, scale, scalesize, cnt_h);
458 cy = __mpn_lshift (frac, frac, fracsize, cnt_h);
459 if (cy != 0)
460 frac[fracsize++] = cy;
461 }
462 else if (BITS_PER_MP_LIMB - cnt_h <= cnt_l)
463 {
464 /* We can save memory by removing the trailing zero limbs
465 and by packing the non-zero limbs which gain another
466 free one. */
467
468 (void) __mpn_rshift (scale, scale + i, scalesize - i,
469 BITS_PER_MP_LIMB - cnt_h);
470 scalesize -= i + 1;
471 (void) __mpn_rshift (frac, frac + i, fracsize - i,
472 BITS_PER_MP_LIMB - cnt_h);
473 fracsize -= frac[fracsize - i - 1] == 0 ? i + 1 : i;
474 }
475 else
476 {
477 /* We can only save the memory of the limbs which are zero.
478 The non-zero parts occupy the same number of limbs. */
479
480 (void) __mpn_rshift (scale, scale + (i - 1),
481 scalesize - (i - 1),
482 BITS_PER_MP_LIMB - cnt_h);
483 scalesize -= i;
484 (void) __mpn_rshift (frac, frac + (i - 1),
485 fracsize - (i - 1),
486 BITS_PER_MP_LIMB - cnt_h);
487 fracsize -= frac[fracsize - (i - 1) - 1] == 0 ? i : i - 1;
488 }
489 }
490 }
491 }
492 else if (exponent < 0)
493 {
494 /* |FP| < 1.0. */
495 int exp10 = 0;
496 int explog = LDBL_MAX_10_EXP_LOG;
497 const struct mp_power *tens = &_fpioconst_pow10[explog + 1];
498 mp_size_t used_limbs = fracsize - 1;
499
500 /* Now shift the input value to its right place. */
501 cy = __mpn_lshift (frac, fp_input, fracsize, to_shift);
502 frac[fracsize++] = cy;
503 assert (cy == 1 || (frac[fracsize - 2] == 0 && frac[0] == 0));
504
505 expsign = 1;
506 exponent = -exponent;
507
508 assert (tens != &_fpioconst_pow10[0]);
509 do
510 {
511 --tens;
512
513 if (exponent >= tens->m_expo)
514 {
515 int i, incr, cnt_h, cnt_l;
516 mp_limb topval[2];
517
518 /* The __mpn_mul function expects the first argument to be
519 bigger than the second. */
520 if (fracsize < tens->arraysize - _FPIO_CONST_OFFSET)
521 cy = __mpn_mul (tmp, &tens->array[_FPIO_CONST_OFFSET],
522 tens->arraysize - _FPIO_CONST_OFFSET,
523 frac, fracsize);
524 else
525 cy = __mpn_mul (tmp, frac, fracsize,
526 &tens->array[_FPIO_CONST_OFFSET],
527 tens->arraysize - _FPIO_CONST_OFFSET);
528 tmpsize = fracsize + tens->arraysize - _FPIO_CONST_OFFSET;
529 if (cy == 0)
530 --tmpsize;
531
532 count_leading_zeros (cnt_h, tmp[tmpsize - 1]);
533 incr = (tmpsize - fracsize) * BITS_PER_MP_LIMB
534 + BITS_PER_MP_LIMB - 1 - cnt_h;
535
536 assert (incr <= tens->p_expo);
537
538 /* If we increased the exponent by exactly 3 we have to test
539 for overflow. This is done by comparing with 10 shifted
540 to the right position. */
541 if (incr == exponent + 3)
542 if (cnt_h <= BITS_PER_MP_LIMB - 4)
543 {
544 topval[0] = 0;
545 topval[1]
546 = ((mp_limb) 10) << (BITS_PER_MP_LIMB - 4 - cnt_h);
547 }
548 else
549 {
550 topval[0] = ((mp_limb) 10) << (BITS_PER_MP_LIMB - 4);
551 topval[1] = 0;
552 (void) __mpn_lshift (topval, topval, 2,
553 BITS_PER_MP_LIMB - cnt_h);
554 }
555
556 /* We have to be careful when multiplying the last factor.
557 If the result is greater than 1.0 be have to test it
558 against 10.0. If it is greater or equal to 10.0 the
559 multiplication was not valid. This is because we cannot
560 determine the number of bits in the result in advance. */
561 if (incr < exponent + 3
562 || (incr == exponent + 3 &&
563 (tmp[tmpsize - 1] < topval[1]
564 || (tmp[tmpsize - 1] == topval[1]
565 && tmp[tmpsize - 2] < topval[0]))))
566 {
567 /* The factor is right. Adapt binary and decimal
568 exponents. */
569 exponent -= incr;
570 exp10 |= 1 << explog;
571
572 /* If this factor yields a number greater or equal to
573 1.0, we must not shift the non-fractional digits down. */
574 if (exponent < 0)
575 cnt_h += -exponent;
576
577 /* Now we optimize the number representation. */
578 for (i = 0; tmp[i] == 0; ++i);
579 if (cnt_h == BITS_PER_MP_LIMB - 1)
580 {
581 MPN_COPY (frac, tmp + i, tmpsize - i);
582 fracsize = tmpsize - i;
583 }
584 else
585 {
586 count_trailing_zeros (cnt_l, tmp[i]);
587
588 /* Now shift the numbers to their optimal position. */
589 if (i == 0 && BITS_PER_MP_LIMB - 1 - cnt_h > cnt_l)
590 {
591 /* We cannot save any memory. Just roll the
592 number so that the leading digit is in a
593 seperate limb. */
594
595 cy = __mpn_lshift (frac, tmp, tmpsize, cnt_h + 1);
596 fracsize = tmpsize + 1;
597 frac[fracsize - 1] = cy;
598 }
599 else if (BITS_PER_MP_LIMB - 1 - cnt_h <= cnt_l)
600 {
601 (void) __mpn_rshift (frac, tmp + i, tmpsize - i,
602 BITS_PER_MP_LIMB - 1 - cnt_h);
603 fracsize = tmpsize - i;
604 }
605 else
606 {
607 /* We can only save the memory of the limbs which
608 are zero. The non-zero parts occupy the same
609 number of limbs. */
610
611 (void) __mpn_rshift (frac, tmp + (i - 1),
612 tmpsize - (i - 1),
613 BITS_PER_MP_LIMB - 1 - cnt_h);
614 fracsize = tmpsize - (i - 1);
615 }
616 }
617 used_limbs = fracsize - 1;
618 }
619 }
620 --explog;
621 }
622 while (tens != &_fpioconst_pow10[1] && exponent > 0);
623 /* All factors but 10^-1 are tested now. */
624 if (exponent > 0)
625 {
626 cy = __mpn_mul_1 (tmp, frac, fracsize, 10);
627 tmpsize = fracsize;
628 assert (cy == 0 || tmp[tmpsize - 1] < 20);
629
630 (void) __mpn_rshift (frac, tmp, tmpsize, MIN (4, exponent));
631 fracsize = tmpsize;
632 exp10 |= 1;
633 assert (frac[fracsize - 1] < 10);
634 }
635 exponent = exp10;
636 }
637 else
638 {
639 /* This is a special case. We don't need a factor because the
640 numbers are in the range of 0.0 <= fp < 8.0. We simply
641 shift it to the right place and divide it by 1.0 to get the
642 leading digit. (Of course this division is not really made.) */
643 assert (0 <= exponent && exponent < 3 &&
644 exponent + to_shift < BITS_PER_MP_LIMB);
645
646 /* Now shift the input value to its right place. */
647 cy = __mpn_lshift (frac, fp_input, fracsize, (exponent + to_shift));
648 frac[fracsize++] = cy;
649 exponent = 0;
650 }
651
652 {
653 int width = info->width;
654 char *buffer, *startp, *cp;
655 int chars_needed;
656 int expscale;
657 int intdig_max, intdig_no = 0;
658 int fracdig_min, fracdig_max, fracdig_no = 0;
659 int dig_max;
660 int significant;
661
662 if (tolower (info->spec) == 'e')
663 {
664 type = info->spec;
665 intdig_max = 1;
666 fracdig_min = fracdig_max = info->prec < 0 ? 6 : info->prec;
667 chars_needed = 1 + 1 + fracdig_max + 1 + 1 + 4;
668 /* d . ddd e +- ddd */
669 dig_max = INT_MAX; /* Unlimited. */
670 significant = 1; /* Does not matter here. */
671 }
672 else if (info->spec == 'f')
673 {
674 type = 'f';
675 fracdig_min = fracdig_max = info->prec < 0 ? 6 : info->prec;
676 if (expsign == 0)
677 {
678 intdig_max = exponent + 1;
679 /* This can be really big! */ /* XXX Maybe malloc if too big? */
680 chars_needed = exponent + 1 + 1 + fracdig_max;
681 }
682 else
683 {
684 intdig_max = 1;
685 chars_needed = 1 + 1 + fracdig_max;
686 }
687 dig_max = INT_MAX; /* Unlimited. */
688 significant = 1; /* Does not matter here. */
689 }
690 else
691 {
692 dig_max = info->prec < 0 ? 6 : (info->prec == 0 ? 1 : info->prec);
693 if ((expsign == 0 && exponent >= dig_max)
694 || (expsign != 0 && exponent > 4))
695 {
696 type = isupper (info->spec) ? 'E' : 'e';
697 fracdig_max = dig_max - 1;
698 intdig_max = 1;
699 chars_needed = 1 + 1 + fracdig_max + 1 + 1 + 4;
700 }
701 else
702 {
703 type = 'f';
704 intdig_max = expsign == 0 ? exponent + 1 : 0;
705 fracdig_max = dig_max - intdig_max;
706 /* We need space for the significant digits and perhaps for
707 leading zeros when < 1.0. Pessimistic guess: dig_max. */
708 chars_needed = dig_max + dig_max + 1;
709 }
710 fracdig_min = info->alt ? fracdig_max : 0;
711 significant = 0; /* We count significant digits. */
712 }
713
714 if (grouping)
715 /* Guess the number of groups we will make, and thus how
716 many spaces we need for separator characters. */
717 chars_needed += guess_grouping (intdig_max, grouping, thousands_sep);
718
719 /* Allocate buffer for output. We need two more because while rounding
720 it is possible that we need two more characters in front of all the
721 other output. */
722 buffer = alloca (2 + chars_needed);
723 cp = startp = buffer + 2; /* Let room for rounding. */
724
725 /* Do the real work: put digits in allocated buffer. */
726 if (expsign == 0 || type != 'f')
727 {
728 assert (expsign == 0 || intdig_max == 1);
729 while (intdig_no < intdig_max)
730 {
731 ++intdig_no;
732 *cp++ = hack_digit ();
733 }
734 significant = 1;
735 if (info->alt
736 || fracdig_min > 0
737 || (fracdig_max > 0 && (fracsize > 1 || frac[0] != 0)))
738 *cp++ = decimal;
739 }
740 else
741 {
742 /* |fp| < 1.0 and the selected type is 'f', so put "0."
743 in the buffer. */
744 *cp++ = '0';
745 --exponent;
746 *cp++ = decimal;
747 }
748
749 /* Generate the needed number of fractional digits. */
750 while (fracdig_no < fracdig_min
751 || (fracdig_no < fracdig_max && (fracsize > 1 || frac[0] != 0)))
752 {
753 ++fracdig_no;
754 *cp = hack_digit ();
755 if (*cp != '0')
756 significant = 1;
757 else if (significant == 0)
758 {
759 ++fracdig_max;
760 if (fracdig_min > 0)
761 ++fracdig_min;
762 }
763 ++cp;
764 }
765
766 /* Do rounding. */
767 digit = hack_digit ();
768 if (digit > '4')
769 {
770 char *tp = cp;
771
772 if (digit == '5')
773 /* This is the critical case. */
774 if (fracsize == 1 && frac[0] == 0)
775 /* Rest of the number is zero -> round to even.
776 (IEEE 754-1985 4.1 says this is the default rounding.) */
777 if ((*(cp - 1) & 1) == 0)
778 goto do_expo;
779
780 if (fracdig_no > 0)
781 {
782 /* Process fractional digits. Terminate if not rounded or
783 radix character is reached. */
784 while (*--tp != decimal && *tp == '9')
785 *tp = '0';
786 if (*tp != decimal)
787 /* Round up. */
788 (*tp)++;
789 }
790
791 if (fracdig_no == 0 || *tp == decimal)
792 {
793 /* Round the integer digits. */
794 if (*(tp - 1) == decimal)
795 --tp;
796
797 while (--tp >= startp && *tp == '9')
798 *tp = '0';
799
800 if (tp >= startp)
801 /* Round up. */
802 (*tp)++;
803 else
804 /* It is more citical. All digits were 9's. */
805 {
806 if (type != 'f')
807 {
808 *startp = '1';
809 exponent += expsign == 0 ? 1 : -1;
810 }
811 else if (intdig_no == dig_max)
812 {
813 /* This is the case where for type %g the number fits
814 really in the range for %f output but after rounding
815 the number of digits is too big. */
816 *--startp = decimal;
817 *--startp = '1';
818
819 if (info->alt || fracdig_no > 0)
820 {
821 /* Overwrite the old radix character. */
822 startp[intdig_no + 2] = '0';
823 ++fracdig_no;
824 }
825
826 fracdig_no += intdig_no;
827 intdig_no = 1;
828 fracdig_max = intdig_max - intdig_no;
829 ++exponent;
830 /* Now we must print the exponent. */
831 type = isupper (info->spec) ? 'E' : 'e';
832 }
833 else
834 {
835 /* We can simply add another another digit before the
836 radix. */
837 *--startp = '1';
838 ++intdig_no;
839 }
840
841 /* While rounding the number of digits can change.
842 If the number now exceeds the limits remove some
843 fractional digits. */
844 if (intdig_no + fracdig_no > dig_max)
845 {
846 cp -= intdig_no + fracdig_no - dig_max;
847 fracdig_no -= intdig_no + fracdig_no - dig_max;
848 }
849 }
850 }
851 }
852
853 do_expo:
854 /* Now remove unnecessary '0' at the end of the string. */
855 while (fracdig_no > fracdig_min && *(cp - 1) == '0')
856 {
857 --cp;
858 --fracdig_no;
859 }
860 /* If we eliminate all fractional digits we perhaps also can remove
861 the radix character. */
862 if (fracdig_no == 0 && !info->alt && *(cp - 1) == decimal)
863 --cp;
864
865 if (grouping)
866 /* Add in separator characters, overwriting the same buffer. */
867 cp = group_number (startp, cp, intdig_no, grouping, thousands_sep);
868
869 /* Write the exponent if it is needed. */
870 if (type != 'f')
871 {
872 *cp++ = type;
873 *cp++ = expsign ? '-' : '+';
874
875 /* Find the magnitude of the exponent. */
876 expscale = 10;
877 while (expscale <= exponent)
878 expscale *= 10;
879
880 if (exponent < 10)
881 /* Exponent always has at least two digits. */
882 *cp++ = '0';
883 else
884 do
885 {
886 expscale /= 10;
887 *cp++ = '0' + (exponent / expscale);
888 exponent %= expscale;
889 }
890 while (expscale > 10);
891 *cp++ = '0' + exponent;
892 }
893
894 /* Compute number of characters which must be filled with the padding
895 character. */
896 if (is_neg || info->showsign || info->space)
897 --width;
898 width -= cp - startp;
899
900 if (!info->left && info->pad != '0' && width > 0)
901 PADN (info->pad, width);
902
903 if (is_neg)
904 outchar ('-');
905 else if (info->showsign)
906 outchar ('+');
907 else if (info->space)
908 outchar (' ');
909
910 if (!info->left && info->pad == '0' && width > 0)
911 PADN ('0', width);
912
913 PRINT (startp, cp - startp);
914
915 if (info->left && width > 0)
916 PADN (info->pad, width);
917 }
918 return done;
919 }
920 \f
921 /* Return the number of extra grouping characters that will be inserted
922 into a number with INTDIG_MAX integer digits. */
923
924 static unsigned int
925 guess_grouping (unsigned int intdig_max, const char *grouping, wchar_t sepchar)
926 {
927 unsigned int groups;
928
929 /* We treat all negative values like CHAR_MAX. */
930
931 if (*grouping == CHAR_MAX || *grouping <= 0)
932 /* No grouping should be done. */
933 return 0;
934
935 groups = 0;
936 while (intdig_max > (unsigned int) *grouping)
937 {
938 ++groups;
939 intdig_max -= *grouping++;
940
941 if (*grouping == CHAR_MAX || *grouping < 0)
942 /* No more grouping should be done. */
943 break;
944 else if (*grouping == 0)
945 {
946 /* Same grouping repeats. */
947 groups += intdig_max / grouping[-1];
948 break;
949 }
950 }
951
952 return groups;
953 }
954
955 /* Group the INTDIG_NO integer digits of the number in [BUF,BUFEND).
956 There is guaranteed enough space past BUFEND to extend it.
957 Return the new end of buffer. */
958
959 static char *
960 group_number (char *buf, char *bufend, unsigned int intdig_no,
961 const char *grouping, wchar_t thousands_sep)
962 {
963 unsigned int groups = guess_grouping (intdig_no, grouping, thousands_sep);
964 char *p;
965
966 if (groups == 0)
967 return bufend;
968
969 /* Move the fractional part down. */
970 memmove (buf + intdig_no + groups, buf + intdig_no,
971 bufend - (buf + intdig_no));
972
973 p = buf + intdig_no + groups - 1;
974 do
975 {
976 unsigned int len = *grouping++;
977 do
978 *p-- = buf[--intdig_no];
979 while (--len > 0);
980 *p-- = thousands_sep;
981
982 if (*grouping == CHAR_MAX || *grouping < 0)
983 /* No more grouping should be done. */
984 break;
985 else if (*grouping == 0)
986 /* Same grouping repeats. */
987 --grouping;
988 } while (intdig_no > (unsigned int) *grouping);
989
990 /* Copy the remaining ungrouped digits. */
991 do
992 *p-- = buf[--intdig_no];
993 while (p > buf);
994
995 return bufend + groups;
996 }