source: mainline/uspace/lib/posix/stdio/scanf.c@ a12f7f1

lfn serial ticket/834-toolchain-update topic/msim-upgrade topic/simplify-dev-export
Last change on this file since a12f7f1 was a12f7f1, checked in by Petr Koupy <petr.koupy@…>, 14 years ago

All occurences of call to native free() secured from passing NULL pointer.
  • Property mode set to 100644
File size: 35.3 KB
Line 
1/*
2 * Copyright (c) 2011 Petr Koupy
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 *
9 * - Redistributions of source code must retain the above copyright
10 * notice, this list of conditions and the following disclaimer.
11 * - Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
14 * - The name of the author may not be used to endorse or promote products
15 * derived from this software without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
18 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
19 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
20 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
21 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
22 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
26 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
27 */
28
29/** @addtogroup libposix
30 * @{
31 */
32/** @file Implementation of the scanf backend.
33 */
34
35#define LIBPOSIX_INTERNAL
36
37/* Must be first. */
38#include "../stdbool.h"
39
40#include "../assert.h"
41#include "../errno.h"
42
43#include "../stdio.h"
44#include "../stdlib.h"
45#include "../stddef.h"
46#include "../string.h"
47#include "../ctype.h"
48#include "../sys/types.h"
49
50#include "../internal/common.h"
51#include "../libc/malloc.h"
52
53/** Unified data type for possible data sources for scanf. */
54typedef union __data_source {
55 FILE *stream; /**< Input file stream. */
56 const char *string; /**< Input string. */
57} _data_source;
58
59/** Internal state of the input provider. */
60enum {
61 /** Partly constructed but not yet functional. */
62 _PROV_CONSTRUCTED,
63 /** Ready to serve any request. */
64 _PROV_READY,
65 /** Cursor is temporarily lent to the external entity. No action is
66 * possible until the cursor is returned. */
67 _PROV_CURSOR_LENT,
68};
69
70/** Universal abstraction over data input for scanf. */
71typedef struct __input_provider {
72 /** Source of data elements. */
73 _data_source source;
74 /** How many elements was already processed. */
75 int consumed;
76 /** How many elements was already fetched from the source. */
77 int fetched;
78 /** Elements are fetched from the source in batches (e.g. by getline())
79 * to allow using strtol/strtod family even on streams. */
80 char *window;
81 /** Size of the current window. */
82 size_t window_size;
83 /** Points to the next element to be processed inside the current window. */
84 const char *cursor;
85 /** Internal state of the provider. */
86 int state;
87
88 /** Take control over data source. Finish initialization of the internal
89 * structures (e.g. allocation of window). */
90 void (*capture)(struct __input_provider *);
91 /** Get a single element from the source and update the internal structures
92 * accordingly (e.g. greedy update of the window). Return -1 if the
93 * element cannot be obtained. */
94 int (*pop)(struct __input_provider *);
95 /** Undo the most recent not-undone pop operation. Might be necesarry to
96 * flush current window and seek data source backwards. Return 0 if the
97 * pop history is exhausted, non-zero on success. */
98 int (*undo)(struct __input_provider *);
99 /** Lend the cursor to the caller. */
100 const char * (*borrow_cursor)(struct __input_provider *);
101 /** Take control over possibly incremented cursor and update the internal
102 * structures if necessary. */
103 void (*return_cursor)(struct __input_provider *, const char *);
104 /** Release the control over the source. That is, synchronize any
105 * fetched but non-consumed elements (e.g. by seeking) and destruct
106 * internal structures (e.g. window deallocation). */
107 void (*release)(struct __input_provider *);
108} _input_provider;
109
110/** @see __input_provider */
111static void _capture_stream(_input_provider *self)
112{
113 assert(self->source.stream);
114 assert(self->state == _PROV_CONSTRUCTED);
115 /* Caller could already pre-allocated the window. */
116 assert((self->window == NULL && self->window_size == 0) ||
117 (self->window && self->window_size > 0));
118
119 /* Initialize internal structures. */
120 self->consumed = 0;
121 ssize_t fetched = posix_getline(
122 &self->window, &self->window_size, self->source.stream);
123 if (fetched != -1) {
124 self->fetched = fetched;
125 self->cursor = self->window;
126 } else {
127 /* EOF encountered. */
128 self->fetched = 0;
129 self->cursor = NULL;
130 }
131 self->state = _PROV_READY;
132}
133
134/** @see __input_provider */
135static void _capture_string(_input_provider *self)
136{
137 assert(self->source.string);
138 assert(self->state == _PROV_CONSTRUCTED);
139
140 /* Initialize internal structures. */
141 self->consumed = 0;
142 self->fetched = posix_strlen(self->source.string);
143 self->window = (char *) self->source.string;
144 self->window_size = self->fetched + 1;
145 self->cursor = self->window;
146 self->state = _PROV_READY;
147}
148
149/** @see __input_provider */
150static int _pop_stream(_input_provider *self)
151{
152 assert(self->state == _PROV_READY);
153
154 if (self->cursor) {
155 int c = *self->cursor;
156 ++self->consumed;
157 ++self->cursor;
158 /* Do we need to fetch a new line from the source? */
159 if (*self->cursor == '\0') {
160 ssize_t fetched = posix_getline(&self->window,
161 &self->window_size, self->source.stream);
162 if (fetched != -1) {
163 self->fetched += fetched;
164 self->cursor = self->window;
165 } else {
166 /* EOF encountered. */
167 self->cursor = NULL;
168 }
169 }
170 return c;
171 } else {
172 /* Already at EOF. */
173 return -1;
174 }
175}
176
177/** @see __input_provider */
178static int _pop_string(_input_provider *self)
179{
180 assert(self->state == _PROV_READY);
181
182 if (*self->cursor != '\0') {
183 int c = *self->cursor;
184 ++self->consumed;
185 ++self->cursor;
186 return c;
187 } else {
188 /* String depleted. */
189 return -1;
190 }
191}
192
193/** @see __input_provider */
194static int _undo_stream(_input_provider *self)
195{
196 assert(self->state == _PROV_READY);
197
198 if (self->consumed == 0) {
199 /* Undo history exhausted. */
200 return 0;
201 }
202
203 if (!self->cursor || self->window == self->cursor) {
204 /* Complex case. Either at EOF (cursor == NULL) or there is no more
205 * place to retreat to inside the window. Seek the source backwards
206 * and flush the window. Regarding the scanf, this could happend only
207 * when matching unbounded string (%s) or unbounded scanset (%[) not
208 * containing newline, while at the same time newline is the character
209 * that breaks the matching process. */
210 int rc = posix_fseek(
211 self->source.stream, -1, SEEK_CUR);
212 if (rc == -1) {
213 /* Seek failed. */
214 return 0;
215 }
216 ssize_t fetched = posix_getline(&self->window,
217 &self->window_size, self->source.stream);
218 if (fetched != -1) {
219 assert(fetched == 1);
220 self->fetched = self->consumed + 1;
221 self->cursor = self->window;
222 } else {
223 /* Stream is broken. */
224 return 0;
225 }
226 } else {
227 /* Simple case. Still inside window. */
228 --self->cursor;
229 }
230 --self->consumed;
231 return 1; /* Success. */
232}
233
234/** @see __input_provider */
235static int _undo_string(_input_provider *self)
236{
237 assert(self->state == _PROV_READY);
238
239 if (self->consumed > 0) {
240 --self->consumed;
241 --self->cursor;
242 } else {
243 /* Undo history exhausted. */
244 return 0;
245 }
246 return 1; /* Success. */
247}
248
249/** @see __input_provider */
250static const char *_borrow_cursor_universal(_input_provider *self)
251{
252 assert(self->state == _PROV_READY);
253
254 self->state = _PROV_CURSOR_LENT;
255 return self->cursor;
256}
257
258/** @see __input_provider */
259static void _return_cursor_stream(_input_provider *self, const char *cursor)
260{
261 assert(self->state == _PROV_CURSOR_LENT);
262
263 /* Check how much of the window did external entity consumed. */
264 self->consumed += cursor - self->cursor;
265 self->cursor = cursor;
266 if (*self->cursor == '\0') {
267 /* Window was completely consumed, fetch new data. */
268 ssize_t fetched = posix_getline(&self->window,
269 &self->window_size, self->source.stream);
270 if (fetched != -1) {
271 self->fetched += fetched;
272 self->cursor = self->window;
273 } else {
274 /* EOF encountered. */
275 self->cursor = NULL;
276 }
277 }
278 self->state = _PROV_READY;
279}
280
281/** @see __input_provider */
282static void _return_cursor_string(_input_provider *self, const char *cursor)
283{
284 assert(self->state == _PROV_CURSOR_LENT);
285
286 /* Check how much of the window did external entity consumed. */
287 self->consumed += cursor - self->cursor;
288 self->cursor = cursor;
289 self->state = _PROV_READY;
290}
291
292/** @see __input_provider */
293static void _release_stream(_input_provider *self)
294{
295 assert(self->state == _PROV_READY);
296 assert(self->consumed >= self->fetched);
297
298 /* Try to correct the difference between the stream position and what was
299 * actually consumed. If it is not possible, continue anyway. */
300 posix_fseek(self->source.stream, self->consumed - self->fetched, SEEK_CUR);
301
302 /* Destruct internal structures. */
303 self->fetched = 0;
304 self->cursor = NULL;
305 if (self->window) {
306 free(self->window);
307 self->window = NULL;
308 }
309 self->window_size = 0;
310 self->state = _PROV_CONSTRUCTED;
311}
312
313/** @see __input_provider */
314static void _release_string(_input_provider *self)
315{
316 assert(self->state == _PROV_READY);
317
318 /* Destruct internal structures. */
319 self->fetched = 0;
320 self->cursor = NULL;
321 self->window = NULL;
322 self->window_size = 0;
323 self->state = _PROV_CONSTRUCTED;
324}
325
326/** Length modifier values. */
327enum {
328 LMOD_NONE,
329 LMOD_hh,
330 LMOD_h,
331 LMOD_l,
332 LMOD_ll,
333 LMOD_j,
334 LMOD_z,
335 LMOD_t,
336 LMOD_L,
337 LMOD_p, /* Reserved for %p conversion. */
338};
339
340/**
341 * Decides whether provided characters specify length modifier. If so, the
342 * recognized modifier is stored through provider pointer.
343 *
344 * @param c Candidate on the length modifier.
345 * @param _c Next character (might be NUL).
346 * @param modifier Pointer to the modifier value.
347 * @return Whether the modifier was recognized or not.
348 */
349static inline int is_length_mod(int c, int _c, int *modifier)
350{
351 assert(modifier);
352
353 switch (c) {
354 case 'h':
355 /* Check whether the modifier was not already recognized. */
356 if (*modifier == LMOD_NONE) {
357 *modifier = _c == 'h' ? LMOD_hh : LMOD_h;
358 } else {
359 /* Format string is invalid. Notify the caller. */
360 *modifier = LMOD_NONE;
361 }
362 return 1;
363 case 'l':
364 if (*modifier == LMOD_NONE) {
365 *modifier = _c == 'l' ? LMOD_ll : LMOD_l;
366 } else {
367 *modifier = LMOD_NONE;
368 }
369 return 1;
370 case 'j':
371 *modifier = *modifier == LMOD_NONE ? LMOD_j : LMOD_NONE;
372 return 1;
373 case 'z':
374 *modifier = *modifier == LMOD_NONE ? LMOD_z : LMOD_NONE;
375 return 1;
376 case 't':
377 *modifier = *modifier == LMOD_NONE ? LMOD_t : LMOD_NONE;
378 return 1;
379 case 'L':
380 *modifier = *modifier == LMOD_NONE ? LMOD_L : LMOD_NONE;
381 return 1;
382 default:
383 return 0;
384 }
385}
386
387/**
388 * Decides whether provided character specifies integer conversion. If so, the
389 * semantics of the conversion is stored through provided pointers..
390 *
391 * @param c Candidate on the integer conversion.
392 * @param is_unsigned Pointer to store whether the conversion is signed or not.
393 * @param base Pointer to store the base of the integer conversion.
394 * @return Whether the conversion was recognized or not.
395 */
396static inline int is_int_conv(int c, bool *is_unsigned, int *base)
397{
398 assert(is_unsigned && base);
399
400 switch (c) {
401 case 'd':
402 *is_unsigned = false;
403 *base = 10;
404 return 1;
405 case 'i':
406 *is_unsigned = false;
407 *base = 0;
408 return 1;
409 case 'o':
410 *is_unsigned = true;
411 *base = 8;
412 return 1;
413 case 'u':
414 *is_unsigned = true;
415 *base = 10;
416 return 1;
417 case 'p': /* According to POSIX, %p modifier is implementation defined but
418 * must correspond to its printf counterpart. */
419 case 'x':
420 case 'X':
421 *is_unsigned = true;
422 *base = 16;
423 return 1;
424 return 1;
425 default:
426 return 0;
427 }
428}
429
430/**
431 * Decides whether provided character specifies conversion of the floating
432 * point number.
433 *
434 * @param c Candidate on the floating point conversion.
435 * @return Whether the conversion was recognized or not.
436 */
437static inline int is_float_conv(int c)
438{
439 switch (c) {
440 case 'a':
441 case 'A':
442 case 'e':
443 case 'E':
444 case 'f':
445 case 'F':
446 case 'g':
447 case 'G':
448 return 1;
449 default:
450 return 0;
451 }
452}
453
454/**
455 * Decides whether provided character specifies conversion of the character
456 * sequence.
457 *
458 * @param c Candidate on the character sequence conversion.
459 * @param modifier Pointer to store length modifier for wide chars.
460 * @return Whether the conversion was recognized or not.
461 */
462static inline int is_seq_conv(int c, int *modifier)
463{
464 assert(modifier);
465
466 switch (c) {
467 case 'S':
468 *modifier = LMOD_l;
469 /* fallthrough */
470 case 's':
471 return 1;
472 case 'C':
473 *modifier = LMOD_l;
474 /* fallthrough */
475 case 'c':
476 return 1;
477 case '[':
478 return 1;
479 default:
480 return 0;
481 }
482}
483
484/**
485 * Backend for the whole family of scanf functions. Uses input provider
486 * to abstract over differences between strings and streams. Should be
487 * POSIX compliant (apart from the not supported stuff).
488 *
489 * NOT SUPPORTED: locale (see strtold), wide chars, numbered output arguments
490 *
491 * @param in Input provider.
492 * @param fmt Format description.
493 * @param arg Output arguments.
494 * @return The number of converted output items or EOF on failure.
495 */
496static inline int _internal_scanf(
497 _input_provider *in, const char *restrict fmt, va_list arg)
498{
499 int c = -1;
500 int converted_cnt = 0;
501 bool converting = false;
502 bool matching_failure = false;
503
504 bool assign_supress = false;
505 bool assign_alloc = false;
506 long width = -1;
507 int length_mod = LMOD_NONE;
508 bool int_conv_unsigned = false;
509 int int_conv_base = 0;
510
511 /* Buffers allocated by scanf for optional 'm' specifier must be remembered
512 * to deallocaate them in case of an error. Because each of those buffers
513 * corresponds to one of the argument from va_list, there is an upper bound
514 * on the number of those arguments. In case of C99, this uppper bound is
515 * 127 arguments. */
516 char *buffers[127];
517 for (int i = 0; i < 127; ++i) {
518 buffers[i] = NULL;
519 }
520 int next_unused_buffer_idx = 0;
521
522 in->capture(in);
523
524 /* Interpret format string. Control shall prematurely jump from the cycle
525 * on input failure, matching failure or illegal format string. In order
526 * to keep error reporting simple enough and to keep input consistent,
527 * error condition shall be always manifested as jump from the cycle,
528 * not function return. Format string pointer shall be updated specifically
529 * for each sub-case (i.e. there shall be no loop-wide increment).*/
530 while (*fmt) {
531
532 if (converting) {
533
534 /* Processing inside conversion specifier. Either collect optional
535 * parameters or execute the conversion. When the conversion
536 * is successfully completed, increment conversion count and switch
537 * back to normal mode. */
538 if (*fmt == '*') {
539 /* Assignment-supression (optional). */
540 if (assign_supress) {
541 /* Already set. Illegal format string. */
542 break;
543 }
544 assign_supress = true;
545 ++fmt;
546 } else if (*fmt == 'm') {
547 /* Assignment-allocation (optional). */
548 if (assign_alloc) {
549 /* Already set. Illegal format string. */
550 break;
551 }
552 assign_alloc = true;
553 ++fmt;
554 } else if (*fmt == '$') {
555 /* Reference to numbered output argument. */
556 // TODO
557 not_implemented();
558 } else if (isdigit(*fmt)) {
559 /* Maximum field length (optional). */
560 if (width != -1) {
561 /* Already set. Illegal format string. */
562 break;
563 }
564 char *fmt_new = NULL;
565 width = posix_strtol(fmt, &fmt_new, 10);
566 if (width != 0) {
567 fmt = fmt_new;
568 } else {
569 /* Since POSIX requires width to be non-zero, it is
570 * sufficient to interpret zero width as error without
571 * referring to errno. */
572 break;
573 }
574 } else if (is_length_mod(*fmt, *(fmt + 1), &length_mod)) {
575 /* Length modifier (optional). */
576 if (length_mod == LMOD_NONE) {
577 /* Already set. Illegal format string. The actual detection
578 * is carried out in the is_length_mod(). */
579 break;
580 }
581 if (length_mod == LMOD_hh || length_mod == LMOD_ll) {
582 /* Modifier was two characters long. */
583 ++fmt;
584 }
585 ++fmt;
586 } else if (is_int_conv(*fmt, &int_conv_unsigned, &int_conv_base)) {
587 /* Integer conversion. */
588
589 /* Check sanity of optional parts of conversion specifier. */
590 if (assign_alloc || length_mod == LMOD_L) {
591 /* Illegal format string. */
592 break;
593 }
594
595 /* Conversion of the integer with %p specifier needs special
596 * handling, because it is not allowed to have arbitrary
597 * length modifier. */
598 if (*fmt == 'p') {
599 if (length_mod == LMOD_NONE) {
600 length_mod = LMOD_p;
601 } else {
602 /* Already set. Illegal format string. */
603 break;
604 }
605 }
606
607 /* First consume any white spaces, so we can borrow cursor
608 * from the input provider. This way, the cursor will either
609 * point to the non-white space while the input will be
610 * prefetched up to the newline (which is suitable for strtol),
611 * or the input will be at EOF. */
612 do {
613 c = in->pop(in);
614 } while (isspace(c));
615
616 /* After skipping the white spaces, can we actually continue? */
617 if (c == -1) {
618 /* Input failure. */
619 break;
620 } else {
621 /* Everything is OK, just undo the last pop, so the cursor
622 * can be borrowed. */
623 in->undo(in);
624 }
625
626 const char *cur_borrowed = NULL;
627 const char *cur_limited = NULL;
628 char *cur_updated = NULL;
629
630 /* Borrow the cursor. Until it is returned to the provider
631 * we cannot jump from the cycle, because it would leave
632 * the input inconsistent. */
633 cur_borrowed = in->borrow_cursor(in);
634
635 /* If the width is limited, the cursor horizont must be
636 * decreased accordingly. Otherwise the strtol could read more
637 * than allowed by width. */
638 if (width != -1) {
639 cur_limited = posix_strndup(cur_borrowed, width);
640 } else {
641 cur_limited = cur_borrowed;
642 }
643 cur_updated = (char *) cur_limited;
644
645 long long sres = 0;
646 unsigned long long ures = 0;
647 errno = 0; /* Reset errno to recognize error later. */
648 /* Try to convert the integer. */
649 if (int_conv_unsigned) {
650 ures = posix_strtoull(cur_limited, &cur_updated, int_conv_base);
651 } else {
652 sres = posix_strtoll(cur_limited, &cur_updated, int_conv_base);
653 }
654
655 /* Update the cursor so it can be returned to the provider. */
656 cur_borrowed += cur_updated - cur_limited;
657 if (width != -1 && cur_limited != NULL) {
658 /* Deallocate duplicated part of the cursor view. */
659 free(cur_limited);
660 }
661 cur_limited = NULL;
662 cur_updated = NULL;
663 /* Return the cursor to the provider. Input consistency is again
664 * the job of the provider, so we can report errors from
665 * now on. */
666 in->return_cursor(in, cur_borrowed);
667 cur_borrowed = NULL;
668
669 /* Check whether the conversion was successful. */
670 if (errno != EOK) {
671 matching_failure = true;
672 break;
673 }
674
675 /* If not supressed, assign the converted integer into
676 * the next output argument. */
677 if (!assign_supress) {
678 if (int_conv_unsigned) {
679 switch (length_mod) {
680 case LMOD_hh: ; /* Label cannot be part of declaration. */
681 unsigned char *phh = va_arg(arg, unsigned char *);
682 *phh = (unsigned char) ures;
683 break;
684 case LMOD_h: ;
685 unsigned short *ph = va_arg(arg, unsigned short *);
686 *ph = (unsigned short) ures;
687 break;
688 case LMOD_NONE: ;
689 unsigned *pdef = va_arg(arg, unsigned *);
690 *pdef = (unsigned) ures;
691 break;
692 case LMOD_l: ;
693 unsigned long *pl = va_arg(arg, unsigned long *);
694 *pl = (unsigned long) ures;
695 break;
696 case LMOD_ll: ;
697 unsigned long long *pll = va_arg(arg, unsigned long long *);
698 *pll = (unsigned long long) ures;
699 break;
700 case LMOD_j: ;
701 posix_uintmax_t *pj = va_arg(arg, posix_uintmax_t *);
702 *pj = (posix_uintmax_t) ures;
703 break;
704 case LMOD_z: ;
705 size_t *pz = va_arg(arg, size_t *);
706 *pz = (size_t) ures;
707 break;
708 case LMOD_t: ;
709 // FIXME: What is unsigned counterpart of the ptrdiff_t?
710 size_t *pt = va_arg(arg, size_t *);
711 *pt = (size_t) ures;
712 break;
713 case LMOD_p: ;
714 void **pp = va_arg(arg, void **);
715 *pp = (void *) (uintptr_t) ures;
716 break;
717 default:
718 assert(false);
719 }
720 } else {
721 switch (length_mod) {
722 case LMOD_hh: ; /* Label cannot be part of declaration. */
723 signed char *phh = va_arg(arg, signed char *);
724 *phh = (signed char) sres;
725 break;
726 case LMOD_h: ;
727 short *ph = va_arg(arg, short *);
728 *ph = (short) sres;
729 break;
730 case LMOD_NONE: ;
731 int *pdef = va_arg(arg, int *);
732 *pdef = (int) sres;
733 break;
734 case LMOD_l: ;
735 long *pl = va_arg(arg, long *);
736 *pl = (long) sres;
737 break;
738 case LMOD_ll: ;
739 long long *pll = va_arg(arg, long long *);
740 *pll = (long long) sres;
741 break;
742 case LMOD_j: ;
743 posix_intmax_t *pj = va_arg(arg, posix_intmax_t *);
744 *pj = (posix_intmax_t) sres;
745 break;
746 case LMOD_z: ;
747 ssize_t *pz = va_arg(arg, ssize_t *);
748 *pz = (ssize_t) sres;
749 break;
750 case LMOD_t: ;
751 posix_ptrdiff_t *pt = va_arg(arg, posix_ptrdiff_t *);
752 *pt = (posix_ptrdiff_t) sres;
753 break;
754 default:
755 assert(false);
756 }
757 }
758 ++converted_cnt;
759 }
760
761 converting = false;
762 ++fmt;
763 } else if (is_float_conv(*fmt)) {
764 /* Floating point number conversion. */
765
766 /* Check sanity of optional parts of conversion specifier. */
767 if (assign_alloc) {
768 /* Illegal format string. */
769 break;
770 }
771 if (length_mod != LMOD_NONE &&
772 length_mod != LMOD_l &&
773 length_mod != LMOD_L) {
774 /* Illegal format string. */
775 break;
776 }
777
778 /* First consume any white spaces, so we can borrow cursor
779 * from the input provider. This way, the cursor will either
780 * point to the non-white space while the input will be
781 * prefetched up to the newline (which is suitable for strtof),
782 * or the input will be at EOF. */
783 do {
784 c = in->pop(in);
785 } while (isspace(c));
786
787 /* After skipping the white spaces, can we actually continue? */
788 if (c == -1) {
789 /* Input failure. */
790 break;
791 } else {
792 /* Everything is OK, just undo the last pop, so the cursor
793 * can be borrowed. */
794 in->undo(in);
795 }
796
797 const char *cur_borrowed = NULL;
798 const char *cur_limited = NULL;
799 char *cur_updated = NULL;
800
801 /* Borrow the cursor. Until it is returned to the provider
802 * we cannot jump from the cycle, because it would leave
803 * the input inconsistent. */
804 cur_borrowed = in->borrow_cursor(in);
805
806 /* If the width is limited, the cursor horizont must be
807 * decreased accordingly. Otherwise the strtof could read more
808 * than allowed by width. */
809 if (width != -1) {
810 cur_limited = posix_strndup(cur_borrowed, width);
811 } else {
812 cur_limited = cur_borrowed;
813 }
814 cur_updated = (char *) cur_limited;
815
816 float fres = 0.0;
817 double dres = 0.0;
818 long double ldres = 0.0;
819 errno = 0; /* Reset errno to recognize error later. */
820 /* Try to convert the floating point nubmer. */
821 switch (length_mod) {
822 case LMOD_NONE:
823 fres = posix_strtof(cur_limited, &cur_updated);
824 break;
825 case LMOD_l:
826 dres = posix_strtod(cur_limited, &cur_updated);
827 break;
828 case LMOD_L:
829 ldres = posix_strtold(cur_limited, &cur_updated);
830 break;
831 default:
832 assert(false);
833 }
834
835 /* Update the cursor so it can be returned to the provider. */
836 cur_borrowed += cur_updated - cur_limited;
837 if (width != -1 && cur_limited != NULL) {
838 /* Deallocate duplicated part of the cursor view. */
839 free(cur_limited);
840 }
841 cur_limited = NULL;
842 cur_updated = NULL;
843 /* Return the cursor to the provider. Input consistency is again
844 * the job of the provider, so we can report errors from
845 * now on. */
846 in->return_cursor(in, cur_borrowed);
847 cur_borrowed = NULL;
848
849 /* Check whether the conversion was successful. */
850 if (errno != EOK) {
851 matching_failure = true;
852 break;
853 }
854
855 /* If nto supressed, assign the converted floating point number
856 * into the next output argument. */
857 if (!assign_supress) {
858 switch (length_mod) {
859 case LMOD_NONE: ; /* Label cannot be part of declaration. */
860 float *pf = va_arg(arg, float *);
861 *pf = fres;
862 break;
863 case LMOD_l: ;
864 double *pd = va_arg(arg, double *);
865 *pd = dres;
866 break;
867 case LMOD_L: ;
868 long double *pld = va_arg(arg, long double *);
869 *pld = ldres;
870 break;
871 default:
872 assert(false);
873 }
874 ++converted_cnt;
875 }
876
877 converting = false;
878 ++fmt;
879 } else if (is_seq_conv(*fmt, &length_mod)) {
880 /* Character sequence conversion. */
881
882 /* Check sanity of optional parts of conversion specifier. */
883 if (length_mod != LMOD_NONE &&
884 length_mod != LMOD_l) {
885 /* Illegal format string. */
886 break;
887 }
888
889 if (length_mod == LMOD_l) {
890 /* Wide chars not supported. */
891 // TODO
892 not_implemented();
893 }
894
895 int term_size = 1; /* Size of the terminator (0 or 1)). */
896 if (*fmt == 'c') {
897 term_size = 0;
898 width = width == -1 ? 1 : width;
899 }
900
901 if (*fmt == 's') {
902 /* Skip white spaces. */
903 do {
904 c = in->pop(in);
905 } while (isspace(c));
906 } else {
907 /* Fetch a single character. */
908 c = in->pop(in);
909 }
910
911 /* Check whether there is still input to read. */
912 if (c == -1) {
913 /* Input failure. */
914 break;
915 }
916
917 /* Prepare scanset. */
918 char terminate_on[256];
919 for (int i = 0; i < 256; ++i) {
920 terminate_on[i] = 0;
921 }
922 if (*fmt == 'c') {
923 ++fmt;
924 } else if (*fmt == 's') {
925 terminate_on[' '] = 1;
926 terminate_on['\n'] = 1;
927 terminate_on['\t'] = 1;
928 terminate_on['\f'] = 1;
929 terminate_on['\r'] = 1;
930 terminate_on['\v'] = 1;
931 ++fmt;
932 } else {
933 assert(*fmt == '[');
934 bool not = false;
935 bool dash = false;
936 ++fmt;
937 /* Check for negation. */
938 if (*fmt == '^') {
939 not = true;
940 ++fmt;
941 }
942 /* Check for escape sequences. */
943 if (*fmt == '-' || *fmt == ']') {
944 terminate_on[(int) *fmt] = 1;
945 ++fmt;
946 }
947 /* Check for ordinary characters and ranges. */
948 while (*fmt != '\0' && *fmt != ']') {
949 if (dash) {
950 for (char chr = *(fmt - 2); chr <= *fmt; ++chr) {
951 terminate_on[(int) chr] = 1;
952 }
953 dash = false;
954 } else if (*fmt == '-') {
955 dash = true;
956 } else {
957 terminate_on[(int) *fmt] = 1;
958 }
959 ++fmt;
960 }
961 /* Check for escape sequence. */
962 if (dash == true) {
963 terminate_on['-'] = 1;
964 }
965 /* Check whether the specifier was correctly terminated.*/
966 if (*fmt == '\0') {
967 /* Illegal format string. */
968 break;
969 } else {
970 ++fmt;
971 }
972 /* Inverse the scanset if necessary. */
973 if (not == false) {
974 for (int i = 0; i < 256; ++i) {
975 terminate_on[i] = terminate_on[i] ? 0 : 1;
976 }
977 }
978 }
979
980 char * buf = NULL;
981 size_t buf_size = 0;
982 char * cur = NULL;
983 size_t alloc_step = 80; /* Buffer size gain during reallocation. */
984 int my_buffer_idx = 0;
985
986 /* Retrieve the buffer into which popped characters
987 * will be stored. */
988 if (!assign_supress) {
989 if (assign_alloc) {
990 /* We must allocate our own buffer. */
991 buf_size =
992 width == -1 ? alloc_step : (size_t) width + term_size;
993 buf = malloc(buf_size);
994 if (!buf) {
995 /* No memory. */
996 break;
997 }
998 my_buffer_idx = next_unused_buffer_idx;
999 ++next_unused_buffer_idx;
1000 buffers[my_buffer_idx] = buf;
1001 cur = buf;
1002 } else {
1003 /* Caller provided its buffer. */
1004 buf = va_arg(arg, char *);
1005 cur = buf;
1006 buf_size =
1007 width == -1 ? SIZE_MAX : (size_t) width + term_size;
1008 }
1009 }
1010
1011 /* Match the string. The next character is already popped. */
1012 while ((width == -1 || width > 0) && c != -1 && !terminate_on[c]) {
1013
1014 /* Check whether the buffer is still sufficiently large. */
1015 if (!assign_supress) {
1016 /* Always reserve space for the null terminator. */
1017 if (cur == buf + buf_size - term_size) {
1018 /* Buffer size must be increased. */
1019 buf = realloc(buf, buf_size + alloc_step);
1020 if (buf) {
1021 buffers[my_buffer_idx] = buf;
1022 cur = buf + buf_size - term_size;
1023 buf_size += alloc_step;
1024 } else {
1025 /* Break just from this tight loop. Errno will
1026 * be checked after it. */
1027 break;
1028 }
1029 }
1030 /* Store the input character. */
1031 *cur = c;
1032 }
1033
1034 width = width == -1 ? -1 : width - 1;
1035 ++cur;
1036 c = in->pop(in);
1037 }
1038 if (errno == ENOMEM) {
1039 /* No memory. */
1040 break;
1041 }
1042 if (c != -1) {
1043 /* There is still more input, so undo the last pop. */
1044 in->undo(in);
1045 }
1046
1047 /* Check for failures. */
1048 if (cur == buf) {
1049 /* Matching failure. Input failure was already checked
1050 * earlier. */
1051 matching_failure = true;
1052 if (!assign_supress && assign_alloc) {
1053 /* Roll back. */
1054 free(buf);
1055 buffers[my_buffer_idx] = NULL;
1056 --next_unused_buffer_idx;
1057 }
1058 break;
1059 }
1060
1061 /* Store the terminator. */
1062 if (!assign_supress && term_size > 0) {
1063 /* Space for the terminator was reserved. */
1064 *cur = '\0';
1065 }
1066
1067 /* Store the result if not already stored. */
1068 if (!assign_supress) {
1069 if (assign_alloc) {
1070 char **pbuf = va_arg(arg, char **);
1071 *pbuf = buf;
1072 }
1073 ++converted_cnt;
1074 }
1075
1076 converting = false;
1077 /* Format string pointer already incremented. */
1078 } else if (*fmt == 'n') {
1079 /* Report the number of consumed bytes so far. */
1080
1081 /* Sanity check. */
1082 bool sane =
1083 width == -1 &&
1084 length_mod == LMOD_NONE &&
1085 assign_alloc == false &&
1086 assign_supress == false;
1087
1088 if (sane) {
1089 int *pi = va_arg(arg, int *);
1090 *pi = in->consumed;
1091 } else {
1092 /* Illegal format string. */
1093 break;
1094 }
1095
1096 /* This shall not be counted as conversion. */
1097 converting = false;
1098 ++fmt;
1099 } else {
1100 /* Illegal format string. */
1101 break;
1102 }
1103
1104 } else {
1105
1106 /* Processing outside conversion specifier. Either skip white
1107 * spaces or match characters one by one. If conversion specifier
1108 * is detected, switch to coversion mode. */
1109 if (isspace(*fmt)) {
1110 /* Skip white spaces in the format string. */
1111 while (isspace(*fmt)) {
1112 ++fmt;
1113 }
1114 /* Skip white spaces in the input. */
1115 do {
1116 c = in->pop(in);
1117 } while (isspace(c));
1118 if (c != -1) {
1119 /* Input is not at EOF, so undo the last pop operation. */
1120 in->undo(in);
1121 }
1122 } else if (*fmt == '%' && *(fmt + 1) != '%') {
1123 /* Conversion specifier detected. Switch modes. */
1124 converting = true;
1125 /* Reset the conversion context. */
1126 assign_supress = false;
1127 assign_alloc = false;
1128 width = -1;
1129 length_mod = LMOD_NONE;
1130 int_conv_unsigned = false;
1131 int_conv_base = 0;
1132 ++fmt;
1133 } else {
1134 /* One by one matching. */
1135 if (*fmt == '%') {
1136 /* Escape sequence detected. */
1137 ++fmt;
1138 assert(*fmt == '%');
1139 }
1140 c = in->pop(in);
1141 if (c == -1) {
1142 /* Input failure. */
1143 break;
1144 } else if (c != *fmt) {
1145 /* Matching failure. */
1146 in->undo(in);
1147 matching_failure = true;
1148 break;
1149 } else {
1150 ++fmt;
1151 }
1152 }
1153
1154 }
1155
1156 }
1157
1158 in->release(in);
1159
1160 /* This somewhat complicated return value decision is required by POSIX. */
1161 int rc;
1162 if (matching_failure) {
1163 rc = converted_cnt;
1164 } else {
1165 if (errno == EOK) {
1166 rc = converted_cnt > 0 ? converted_cnt : EOF;
1167 } else {
1168 rc = EOF;
1169 }
1170 }
1171 if (rc == EOF) {
1172 /* Caller will not know how many arguments were successfully converted,
1173 * so the deallocation of buffers is our responsibility. */
1174 for (int i = 0; i < next_unused_buffer_idx; ++i) {
1175 free(buffers[i]);
1176 buffers[i] = NULL;
1177 }
1178 next_unused_buffer_idx = 0;
1179 }
1180 return rc;
1181}
1182
1183/**
1184 * Convert formatted input from the stream.
1185 *
1186 * @param stream Input stream.
1187 * @param format Format description.
1188 * @param arg Output items.
1189 * @return The number of converted output items or EOF on failure.
1190 */
1191int posix_vfscanf(
1192 FILE *restrict stream, const char *restrict format, va_list arg)
1193{
1194 _input_provider provider = {
1195 { 0 }, 0, 0, NULL, 0, NULL, _PROV_CONSTRUCTED,
1196 _capture_stream, _pop_stream, _undo_stream,
1197 _borrow_cursor_universal, _return_cursor_stream, _release_stream
1198 };
1199 provider.source.stream = stream;
1200 return _internal_scanf(&provider, format, arg);
1201}
1202
1203/**
1204 * Convert formatted input from the string.
1205 *
1206 * @param s Input string.
1207 * @param format Format description.
1208 * @param arg Output items.
1209 * @return The number of converted output items or EOF on failure.
1210 */
1211int posix_vsscanf(
1212 const char *restrict s, const char *restrict format, va_list arg)
1213{
1214 _input_provider provider = {
1215 { 0 }, 0, 0, NULL, 0, NULL, _PROV_CONSTRUCTED,
1216 _capture_string, _pop_string, _undo_string,
1217 _borrow_cursor_universal, _return_cursor_string, _release_string
1218 };
1219 provider.source.string = s;
1220 return _internal_scanf(&provider, format, arg);
1221}
1222
1223// FIXME: put the testcases somewhere else
1224
1225#if 0
1226
1227//#include <stdio.h>
1228//#include <malloc.h>
1229//#include <string.h>
1230
1231#define test_val(fmt, exp_val, act_val) \
1232 if (exp_val == act_val) { \
1233 printf("succ, expected "fmt", actual "fmt"\n", exp_val, act_val); \
1234 } else { \
1235 printf("fail, expected "fmt", actual "fmt"\n", exp_val, act_val); \
1236 ++fail; \
1237 }
1238
1239#define test_str(fmt, exp_str, act_str) \
1240 if (posix_strcmp(exp_str, act_str) == 0) { \
1241 printf("succ, expected "fmt", actual "fmt"\n", exp_str, act_str); \
1242 } else { \
1243 printf("fail, expected "fmt", actual "fmt"\n", exp_str, act_str); \
1244 ++fail; \
1245 }
1246
1247void __posix_scanf_test(void);
1248void __posix_scanf_test(void)
1249{
1250 int fail = 0;
1251
1252 int ret;
1253
1254 unsigned char uhh;
1255 signed char shh;
1256 unsigned short uh;
1257 short sh;
1258 unsigned udef;
1259 int sdef;
1260 unsigned long ul;
1261 long sl;
1262 unsigned long long ull;
1263 long long sll;
1264 void *p;
1265
1266 float f;
1267 double d;
1268 long double ld;
1269
1270 char str[20];
1271 char seq[20];
1272 char scanset[20];
1273
1274 char *pstr;
1275 char *pseq;
1276 char *pscanset;
1277
1278 ret = posix_sscanf(
1279 "\n j tt % \t -121314 98765 aqw 0765 0x77 0xABCDEF88 -99 884",
1280 " j tt %%%3hhd%1hhu%3hd %3hu%u aqw%n %lo%llx %p %li %lld",
1281 &shh, &uhh, &sh, &uh, &udef, &sdef, &ul, &ull, &p, &sl, &sll);
1282 test_val("%d", -12, shh);
1283 test_val("%u", 1, uhh);
1284 test_val("%d", 314, sh);
1285 test_val("%u", 987, uh);
1286 test_val("%u", 65, udef);
1287 test_val("%d", 28, sdef);
1288 test_val("%lo", (unsigned long) 0765, ul);
1289 test_val("%llx", (unsigned long long) 0x77, ull);
1290 test_val("%p", (void *) 0xABCDEF88, p);
1291 test_val("%ld", (long) -99, sl);
1292 test_val("%lld", (long long) 884, sll);
1293 test_val("%d", 10, ret);
1294
1295 ret = posix_sscanf(
1296 "\n \t\t1.0 -0x555.AP10 1234.5678e12",
1297 "%f %lf %Lf",
1298 &f, &d, &ld);
1299 test_val("%f", 1.0, f);
1300 test_val("%lf", (double) -0x555.AP10, d);
1301 test_val("%Lf", (long double) 1234.5678e12, ld);
1302 test_val("%d", 3, ret);
1303
1304 ret = posix_sscanf(
1305 "\n\n\thello world \n",
1306 "%5s %ms",
1307 str, &pstr);
1308 test_str("%s", "hello", str);
1309 test_str("%s", "world", pstr);
1310 test_val("%d", 2, ret);
1311 free(pstr);
1312
1313 ret = posix_sscanf(
1314 "\n\n\thello world \n",
1315 " %5c %mc",
1316 seq, &pseq);
1317 seq[5] = '\0';
1318 pseq[1] = '\0';
1319 test_str("%s", "hello", seq);
1320 test_str("%s", "w", pseq);
1321 test_val("%d", 2, ret);
1322 free(pseq);
1323
1324 ret = posix_sscanf(
1325 "\n\n\th-e-l-l-o world-] \n",
1326 " %9[-eh-o] %m[^]-]",
1327 scanset, &pscanset);
1328 test_str("%s", "h-e-l-l-o", scanset);
1329 test_str("%s", "world", pscanset);
1330 test_val("%d", 2, ret);
1331 free(pscanset);
1332
1333 printf("Failed: %d\n", fail);
1334}
1335
1336#endif
1337
1338/** @}
1339 */
Note: See TracBrowser for help on using the repository browser.