Documentation updates and cleanups.
[paraslash.git] / string.c
1 /*
2 * Copyright (C) 2004-2008 Andre Noll <maan@systemlinux.org>
3 *
4 * Licensed under the GPL v2. For licencing details see COPYING.
5 */
6
7 /** \file string.c Memory allocation and string handling functions. */
8
9 #include "para.h"
10 #include "string.h"
11
12 #include <sys/time.h> /* gettimeofday */
13 #include <pwd.h>
14 #include <sys/utsname.h> /* uname() */
15 #include <string.h>
16
17 #include "error.h"
18
19 /**
20 * Paraslash's version of realloc().
21 *
22 * \param p Pointer to the memory block, may be \p NULL.
23 * \param size The desired new size.
24 *
25 * A wrapper for realloc(3). It calls \p exit(\p EXIT_FAILURE) on errors,
26 * i.e. there is no need to check the return value in the caller.
27 *
28 * \return A pointer to the newly allocated memory, which is suitably aligned
29 * for any kind of variable and may be different from \a p.
30 *
31 * \sa realloc(3).
32 */
33 __must_check __malloc void *para_realloc(void *p, size_t size)
34 {
35 /*
36 * No need to check for NULL pointers: If p is NULL, the call
37 * to realloc is equivalent to malloc(size)
38 */
39 assert(size);
40 if (!(p = realloc(p, size))) {
41 PARA_EMERG_LOG("realloc failed (size = %zu), aborting\n",
42 size);
43 exit(EXIT_FAILURE);
44 }
45 return p;
46 }
47
48 /**
49 * Paraslash's version of malloc().
50 *
51 * \param size The desired new size.
52 *
53 * A wrapper for malloc(3) which exits on errors.
54 *
55 * \return A pointer to the allocated memory, which is suitably aligned for any
56 * kind of variable.
57 *
58 * \sa malloc(3).
59 */
60 __must_check __malloc void *para_malloc(size_t size)
61 {
62 assert(size);
63 void *p = malloc(size);
64
65 if (!p) {
66 PARA_EMERG_LOG("malloc failed (size = %zu), aborting\n",
67 size);
68 exit(EXIT_FAILURE);
69 }
70 return p;
71 }
72
73 /**
74 * Paraslash's version of calloc().
75 *
76 * \param size The desired new size.
77 *
78 * A wrapper for calloc(3) which exits on errors.
79 *
80 * \return A pointer to the allocated and zeroed-out memory, which is suitably
81 * aligned for any kind of variable.
82 *
83 * \sa calloc(3)
84 */
85 __must_check __malloc void *para_calloc(size_t size)
86 {
87 void *ret = para_malloc(size);
88
89 memset(ret, 0, size);
90 return ret;
91 }
92
93 /**
94 * Paraslash's version of strdup().
95 *
96 * \param s The string to be duplicated.
97 *
98 * A wrapper for strdup(3). It calls \p exit(EXIT_FAILURE) on errors, i.e.
99 * there is no need to check the return value in the caller.
100 *
101 * \return A pointer to the duplicated string. If \p s was the NULL pointer,
102 * an pointer to an empty string is returned.
103 *
104 * \sa strdup(3)
105 */
106 __must_check __malloc char *para_strdup(const char *s)
107 {
108 char *ret;
109
110 if ((ret = strdup(s? s: "")))
111 return ret;
112 PARA_EMERG_LOG("strdup failed, aborting\n");
113 exit(EXIT_FAILURE);
114 }
115
116 /**
117 * Allocate a sufficiently large string and print into it.
118 *
119 * \param fmt A usual format string.
120 *
121 * Produce output according to \p fmt. No artificial bound on the length of the
122 * resulting string is imposed.
123 *
124 * \return This function either returns a pointer to a string that must be
125 * freed by the caller or aborts without returning.
126 *
127 * \sa printf(3).
128 */
129 __must_check __printf_1_2 __malloc char *make_message(const char *fmt, ...)
130 {
131 char *msg;
132
133 PARA_VSPRINTF(fmt, msg);
134 return msg;
135 }
136
137 /**
138 * Paraslash's version of strcat().
139 *
140 * \param a String to be appended to.
141 * \param b String to append.
142 *
143 * Append \p b to \p a.
144 *
145 * \return If \a a is \p NULL, return a pointer to a copy of \a b, i.e.
146 * para_strcat(NULL, b) is equivalent to para_strdup(b). If \a b is \p NULL,
147 * return \a a without making a copy of \a a. Otherwise, construct the
148 * concatenation \a c, free \a a (but not \a b) and return \a c.
149 *
150 * \sa strcat(3)
151 */
152 __must_check __malloc char *para_strcat(char *a, const char *b)
153 {
154 char *tmp;
155
156 if (!a)
157 return para_strdup(b);
158 if (!b)
159 return a;
160 tmp = make_message("%s%s", a, b);
161 free(a);
162 return tmp;
163 }
164
165 /**
166 * Paraslash's version of dirname().
167 *
168 * \param name Pointer to the full path.
169 *
170 * Compute the directory component of \p name.
171 *
172 * \return If \a name is \p NULL or the empty string, return \p NULL.
173 * Otherwise, Make a copy of \a name and return its directory component. Caller
174 * is responsible to free the result.
175 */
176 __must_check __malloc char *para_dirname(const char *name)
177 {
178 char *p, *ret;
179
180 if (!name || !*name)
181 return NULL;
182 ret = para_strdup(name);
183 p = strrchr(ret, '/');
184 if (!p)
185 *ret = '\0';
186 else
187 *p = '\0';
188 return ret;
189 }
190
191 /**
192 * Paraslash's version of basename().
193 *
194 * \param name Pointer to the full path.
195 *
196 * Compute the filename component of \a name.
197 *
198 * \return \p NULL if (a) \a name is the empty string or \p NULL, or (b) name
199 * ends with a slash. Otherwise, a pointer within \a name is returned. Caller
200 * must not free the result.
201 */
202 __must_check const char *para_basename(const char *name)
203 {
204 const char *ret;
205
206 if (!name || !*name)
207 return NULL;
208 ret = strrchr(name, '/');
209 if (!ret)
210 return name;
211 ret++;
212 return ret;
213 }
214
215 /**
216 * Cut trailing newline.
217 *
218 * \param buf The string to be chopped.
219 *
220 * Replace the last character in \p buf by zero if it is euqal to
221 * the newline character.
222 */
223 void chop(char *buf)
224 {
225 int n = strlen(buf);
226 if (!n)
227 return;
228 if (buf[n - 1] == '\n')
229 buf[n - 1] = '\0';
230 }
231
232 /**
233 * Get a random filename.
234 *
235 * This is by no means a secure way to create temporary files in a hostile
236 * direcory like \p /tmp. However, it is OK to use for temp files, fifos,
237 * sockets that are created in ~/.paraslash. Result must be freed by the
238 * caller.
239 *
240 * \return A pointer to a random filename.
241 */
242 __must_check __malloc char *para_tmpname(void)
243 {
244 struct timeval now;
245 unsigned int seed;
246
247 gettimeofday(&now, NULL);
248 seed = now.tv_usec;
249 srand(seed);
250 return make_message("%08i", rand());
251 }
252
253 /**
254 * Get the logname of the current user.
255 *
256 * \return A dynammically allocated string that must be freed by the caller. On
257 * errors, the string "unknown user" is returned, i.e. this function never
258 * returns \p NULL.
259 *
260 * \sa getpwuid(3).
261 */
262 __must_check __malloc char *para_logname(void)
263 {
264 struct passwd *pw = getpwuid(getuid());
265 return para_strdup(pw? pw->pw_name : "unknown_user");
266 }
267
268 /**
269 * Get the home directory of the current user.
270 *
271 * \return A dynammically allocated string that must be freed by the caller. If
272 * the home directory could not be found, this function returns "/tmp".
273 */
274 __must_check __malloc char *para_homedir(void)
275 {
276 struct passwd *pw = getpwuid(getuid());
277 return para_strdup(pw? pw->pw_dir : "/tmp");
278 }
279
280 /**
281 * Split string and return pointers to its parts.
282 *
283 * \param args The string to be split.
284 * \param argv_ptr Pointer to the list of substrings.
285 * \param delim Delimiter.
286 *
287 * This function modifies \a args by replacing each occurance of \a delim by
288 * zero. A \p NULL-terminated array of pointers to char* is allocated dynamically
289 * and these pointers are initialized to point to the broken-up substrings
290 * within \a args. A pointer to this array is returned via \a argv_ptr.
291 *
292 * \return The number of substrings found in \a args.
293 */
294 __must_check unsigned split_args(char *args, char *** const argv_ptr, const char *delim)
295 {
296 char *p = args;
297 char **argv;
298 size_t n = 0, i, j;
299
300 p = args + strspn(args, delim);
301 for (;;) {
302 i = strcspn(p, delim);
303 if (!i)
304 break;
305 p += i;
306 n++;
307 p += strspn(p, delim);
308 }
309 *argv_ptr = para_malloc((n + 1) * sizeof(char *));
310 argv = *argv_ptr;
311 i = 0;
312 p = args + strspn(args, delim);
313 while (p) {
314 argv[i] = p;
315 j = strcspn(p, delim);
316 if (!j)
317 break;
318 p += strcspn(p, delim);
319 if (*p) {
320 *p = '\0';
321 p++;
322 p += strspn(p, delim);
323 }
324 i++;
325 }
326 argv[n] = NULL;
327 return n;
328 }
329
330 /**
331 * Ensure that file descriptors 0, 1, and 2 are valid.
332 *
333 * Common approach that opens /dev/null until it gets a file descriptor greater
334 * than two.
335 *
336 * \sa okir's Black Hats Manual.
337 */
338 void valid_fd_012(void)
339 {
340 while (1) {
341 int fd = open("/dev/null", O_RDWR);
342 if (fd < 0)
343 exit(EXIT_FAILURE);
344 if (fd > 2) {
345 close(fd);
346 break;
347 }
348 }
349 }
350
351 /**
352 * Get the own hostname.
353 *
354 * \return A dynammically allocated string containing the hostname.
355 *
356 * \sa uname(2).
357 */
358 __malloc char *para_hostname(void)
359 {
360 struct utsname u;
361
362 uname(&u);
363 return para_strdup(u.nodename);
364 }
365
366 /**
367 * Used to distinguish between read-only and read-write mode.
368 *
369 * \sa for_each_line(), for_each_line_ro().
370 */
371 enum for_each_line_modes{
372 /** Activate read-only mode. */
373 LINE_MODE_RO,
374 /** Activate read-write mode. */
375 LINE_MODE_RW
376 };
377
378 static int for_each_complete_line(enum for_each_line_modes mode, char *buf,
379 size_t size, line_handler_t *line_handler, void *private_data)
380 {
381 char *start = buf, *end;
382 int ret, i, num_lines = 0;
383
384 // PARA_NOTICE_LOG("buf: %s\n", buf);
385 while (start < buf + size) {
386 char *next_null;
387 char *next_cr;
388
389 next_cr = memchr(start, '\n', buf + size - start);
390 next_null = memchr(start, '\0', buf + size - start);
391 if (!next_cr && !next_null)
392 break;
393 if (next_cr && next_null) {
394 end = next_cr < next_null? next_cr : next_null;
395 } else if (next_null) {
396 end = next_null;
397 } else
398 end = next_cr;
399 num_lines++;
400 if (!line_handler) {
401 start = ++end;
402 continue;
403 }
404 if (mode == LINE_MODE_RO) {
405 size_t s = end - start;
406 char *b = para_malloc(s + 1);
407 memcpy(b, start, s);
408 b[s] = '\0';
409 // PARA_NOTICE_LOG("b: %s, start: %s\n", b, start);
410 ret = line_handler(b, private_data);
411 free(b);
412 } else {
413 *end = '\0';
414 ret = line_handler(start, private_data);
415 }
416 if (ret < 0)
417 return ret;
418 start = ++end;
419 }
420 if (!line_handler || mode == LINE_MODE_RO)
421 return num_lines;
422 i = buf + size - start;
423 if (i && i != size)
424 memmove(buf, start, i);
425 return i;
426 }
427
428 /**
429 * Call a custom function for each complete line.
430 *
431 * \param buf The buffer containing data seperated by newlines.
432 * \param size The number of bytes in \a buf.
433 * \param line_handler The custom function.
434 * \param private_data Pointer passed to \a line_handler.
435 *
436 * If \p line_handler is \p NULL, the function returns the number of complete
437 * lines in \p buf. Otherwise, \p line_handler is called for each complete
438 * line in \p buf. The first argument to \p line_handler is the current line,
439 * and \p private_data is passed as the second argument. The function returns
440 * if \p line_handler returns a negative value or no more lines are in the
441 * buffer. The rest of the buffer (last chunk containing an incomplete line)
442 * is moved to the beginning of the buffer.
443 *
444 * \return If \p line_handler is not \p NULL, this function returns the number
445 * of bytes not handled to \p line_handler on success, or the negative return
446 * value of the \p line_handler on errors.
447 *
448 * \sa for_each_line_ro().
449 */
450 int for_each_line(char *buf, size_t size, line_handler_t *line_handler,
451 void *private_data)
452 {
453 return for_each_complete_line(LINE_MODE_RW, buf, size, line_handler,
454 private_data);
455 }
456
457 /**
458 * Call a custom function for each complete line.
459 *
460 * \param buf Same meaning as in \p for_each_line().
461 * \param size Same meaning as in \p for_each_line().
462 * \param line_handler Same meaning as in \p for_each_line().
463 * \param private_data Same meaning as in \p for_each_line().
464 *
465 * This function behaves like \p for_each_line(), but \a buf is left unchanged.
466 *
467 * \return On success, the function returns the number of complete lines in \p
468 * buf, otherwise the (negative) return value of \p line_handler is returned.
469 *
470 * \sa for_each_line().
471 */
472 int for_each_line_ro(char *buf, size_t size, line_handler_t *line_handler,
473 void *private_data)
474 {
475 return for_each_complete_line(LINE_MODE_RO, buf, size, line_handler,
476 private_data);
477 }
478
479 /**
480 * Safely print into a buffer at a given offset
481 *
482 * \param b Determines the buffer, its size, and the offset.
483 * \param fmt The format string.
484 *
485 * This function prints into the buffer given by \a b at the offset which is
486 * also given by \a b. If there is not enough space to hold the result, the
487 * buffer size is doubled until the underlying call to vsnprintf() succeeds
488 * or the size of the buffer exceeds the maximal size specified in \a pb.
489 *
490 * In the latter case the unmodified \a buf and \a offset values as well as the
491 * private_data pointer of \a b are passed to the \a max_size_handler of \a b.
492 * If this function succeeds, i.e. returns a non-negative value, the offset of
493 * \a b is reset to zero and the given data is written to the beginning of the
494 * buffer.
495 *
496 * Upon return, the offset of \a b is adjusted accordingly so that subsequent
497 * calls to this function append data to what is already contained in the
498 * buffer.
499 *
500 * It's OK to call this function with \p b->buf being \p NULL. In this case, an
501 * initial buffer is allocated.
502 *
503 * \return The number of bytes printed into the buffer (not including the
504 * therminating \p NULL byte).
505 *
506 * \sa make_message(), vsnprintf(3).
507 */
508 __printf_2_3 int para_printf(struct para_buffer *b, const char *fmt, ...)
509 {
510 int ret;
511
512 if (!b->buf) {
513 b->buf = para_malloc(128);
514 b->size = 128;
515 b->offset = 0;
516 }
517 while (1) {
518 char *p = b->buf + b->offset;
519 size_t size = b->size - b->offset;
520 va_list ap;
521 if (size) {
522 va_start(ap, fmt);
523 ret = vsnprintf(p, size, fmt, ap);
524 va_end(ap);
525 if (ret > -1 && ret < size) { /* success */
526 b->offset += ret;
527 return ret;
528 }
529 }
530 /* check if we may grow the buffer */
531 if (!b->max_size || 2 * b->size < b->max_size) { /* yes */
532 /* try again with more space */
533 b->size *= 2;
534 b->buf = para_realloc(b->buf, b->size);
535 continue;
536 }
537 /* can't grow buffer */
538 if (!b->offset || !b->max_size_handler) /* message too large */
539 return -ERRNO_TO_PARA_ERROR(ENOSPC);
540 ret = b->max_size_handler(b->buf, b->offset, b->private_data);
541 if (ret < 0)
542 return ret;
543 b->offset = 0;
544 }
545 }
546
547 /** \cond LLONG_MAX and LLONG_LIN might not be defined. */
548 #ifndef LLONG_MAX
549 #define LLONG_MAX (1 << (sizeof(long) - 1))
550 #endif
551 #ifndef LLONG_MIN
552 #define LLONG_MIN (-LLONG_MAX - 1LL)
553 #endif
554 /** \endcond */
555
556 /**
557 * Convert a string to a 64-bit signed integer value.
558 *
559 * \param str The string to be converted.
560 * \param value Result pointer.
561 *
562 * \return Standard.
563 *
564 * \sa para_atoi32(), strtol(3), atoi(3).
565 */
566 int para_atoi64(const char *str, int64_t *value)
567 {
568 char *endptr;
569 long long tmp;
570
571 errno = 0; /* To distinguish success/failure after call */
572 tmp = strtoll(str, &endptr, 10);
573 if (errno == ERANGE && (tmp == LLONG_MAX || tmp == LLONG_MIN))
574 return -E_ATOI_OVERFLOW;
575 if (errno != 0 && tmp == 0) /* other error */
576 return -E_STRTOLL;
577 if (endptr == str)
578 return -E_ATOI_NO_DIGITS;
579 if (*endptr != '\0') /* Further characters after number */
580 return -E_ATOI_JUNK_AT_END;
581 *value = tmp;
582 return 1;
583 }
584
585 /**
586 * Convert a string to a 32-bit signed integer value.
587 *
588 * \param str The string to be converted.
589 * \param value Result pointer.
590 *
591 * \return Standard.
592 *
593 * \sa para_atoi64().
594 */
595 int para_atoi32(const char *str, int32_t *value)
596 {
597 int64_t tmp;
598 int ret;
599 const int32_t max = 2147483647;
600
601 ret = para_atoi64(str, &tmp);
602 if (ret < 0)
603 return ret;
604 if (tmp > max || tmp < -max - 1)
605 return -E_ATOI_OVERFLOW;
606 *value = tmp;
607 return 1;
608 }