1 /* tac - concatenate and print files in reverse
2 Copyright (C) 1988, 1989, 1990, 1991, 1995 Free Software Foundation, Inc.
4 This program is free software; you can redistribute it and/or modify
5 it under the terms of the GNU General Public License as published by
6 the Free Software Foundation; either version 2, or (at your option)
9 This program is distributed in the hope that it will be useful,
10 but WITHOUT ANY WARRANTY; without even the implied warranty of
11 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 GNU General Public License for more details.
14 You should have received a copy of the GNU General Public License
15 along with this program; if not, write to the Free Software
16 Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */
18 /* Written by Jay Lepreau (lepreau@cs.utah.edu).
19 GNU enhancements by David MacKenzie (djm@gnu.ai.mit.edu). */
21 /* Copy each FILE, or the standard input if none are given or when a
22 FILE name of "-" is encountered, to the standard output with the
23 order of the records reversed. The records are separated by
24 instances of a string, or a newline if none is given. By default, the
25 separator string is attached to the end of the record that it
29 -b, --before The separator is attached to the beginning
30 of the record that it precedes in the file.
31 -r, --regex The separator is a regular expression.
32 -s, --separator=separator Use SEPARATOR as the record separator.
34 To reverse a file byte by byte, use (in bash, ksh, or sh):
42 #include <sys/types.h>
54 /* The number of bytes per atomic read. */
55 #define INITIAL_READSIZE 8192
57 /* The number of bytes per atomic write. */
58 #define WRITESIZE 8192
62 static RETSIGTYPE cleanup ();
64 static int tac_file ();
65 static int tac_stdin ();
66 static char *xmalloc ();
67 static char *xrealloc ();
68 static void output ();
69 static void save_stdin ();
70 static void xwrite ();
75 /* The name this program was run with. */
78 /* The string that separates the records of the file. */
79 static char *separator;
81 /* If nonzero, print `separator' along with the record preceding it
82 in the file; otherwise with the record following it. */
83 static int separator_ends_record;
85 /* 0 if `separator' is to be matched as a regular expression;
86 otherwise, the length of `separator', used as a sentinel to
88 static int sentinel_length;
90 /* The length of a match with `separator'. If `sentinel_length' is 0,
91 `match_length' is computed every time a match succeeds;
92 otherwise, it is simply the length of `separator'. */
93 static int match_length;
95 /* The input buffer. */
98 /* The number of bytes to read at once into `buffer'. */
99 static unsigned read_size;
101 /* The size of `buffer'. This is read_size * 2 + sentinel_length + 2.
102 The extra 2 bytes allow `past_end' to have a value beyond the
103 end of `buffer' and `match_start' to run off the front of `buffer'. */
104 static unsigned buffer_size;
106 /* The compiled regular expression representing `separator'. */
107 static struct re_pattern_buffer compiled_separator;
109 /* If non-zero, display usage information and exit. */
110 static int show_help;
112 /* If non-zero, print the version on standard output then exit. */
113 static int show_version;
115 static struct option const longopts[] =
117 {"before", no_argument, &separator_ends_record, 0},
118 {"regex", no_argument, &sentinel_length, 0},
119 {"separator", required_argument, NULL, 's'},
120 {"help", no_argument, &show_help, 1},
121 {"version", no_argument, &show_version, 1},
130 fprintf (stderr, "Try `%s --help' for more information.\n",
135 Usage: %s [OPTION]... [FILE]...\n\
140 -b, --before attach the separator before instead of after\n\
141 -r, --regex interpret the separator as a regular expression\n\
142 -s, --separator=STRING use STRING as the separator instead of newline\n\
143 --help display this help and exit\n\
144 --version output version information and exit\n\
146 With no FILE, or when FILE is -, read standard input.\n\
157 const char *error_message; /* Return value from re_compile_pattern. */
159 int have_read_stdin = 0;
161 program_name = argv[0];
165 separator_ends_record = 1;
167 while ((optc = getopt_long (argc, argv, "brs:", longopts, (int *) 0))
175 separator_ends_record = 0;
183 error (1, 0, "separator cannot be empty");
192 printf ("tac - %s\n", version_string);
199 if (sentinel_length == 0)
201 compiled_separator.allocated = 100;
202 compiled_separator.buffer = (unsigned char *)
203 xmalloc (compiled_separator.allocated);
204 compiled_separator.fastmap = xmalloc (256);
205 compiled_separator.translate = 0;
206 error_message = re_compile_pattern (separator, strlen (separator),
207 &compiled_separator);
209 error (1, 0, "%s", error_message);
212 match_length = sentinel_length = strlen (separator);
214 read_size = INITIAL_READSIZE;
215 /* A precaution that will probably never be needed. */
216 while (sentinel_length * 2 >= read_size)
218 buffer_size = read_size * 2 + sentinel_length + 2;
219 buffer = xmalloc (buffer_size);
222 strcpy (buffer, separator);
223 buffer += sentinel_length;
231 errors = tac_stdin ();
234 for (; optind < argc; ++optind)
236 if (strcmp (argv[optind], "-") == 0)
239 errors |= tac_stdin ();
242 errors |= tac_file (argv[optind]);
245 /* Flush the output buffer. */
246 output ((char *) NULL, (char *) NULL);
248 if (have_read_stdin && close (0) < 0)
249 error (1, errno, "-");
251 error (1, errno, "write error");
255 /* The name of a temporary file containing a copy of pipe input. */
258 /* Print the standard input in reverse, saving it to temporary
259 file `tempfile' first if it is a pipe.
260 Return 0 if ok, 1 if an error occurs. */
265 /* Previous values of signal handlers. */
266 RETSIGTYPE (*sigint) (), (*sighup) (), (*sigpipe) (), (*sigterm) ();
269 #ifdef _POSIX_VERSION
270 struct sigaction oldact, newact;
271 #endif /* _POSIX_VERSION */
273 /* No tempfile is needed for "tac < file".
274 Use fstat instead of checking for errno == ESPIPE because
275 lseek doesn't work on some special files but doesn't return an
277 if (fstat (0, &stats))
279 error (0, errno, "standard input");
282 if (S_ISREG (stats.st_mode))
283 return tac (0, "standard input");
285 #ifdef _POSIX_VERSION
286 newact.sa_handler = cleanup;
287 sigemptyset (&newact.sa_mask);
290 sigaction (SIGINT, NULL, &oldact);
291 sigint = oldact.sa_handler;
292 if (sigint != SIG_IGN)
293 sigaction (SIGINT, &newact, NULL);
295 sigaction (SIGHUP, NULL, &oldact);
296 sighup = oldact.sa_handler;
297 if (sighup != SIG_IGN)
298 sigaction (SIGHUP, &newact, NULL);
300 sigaction (SIGPIPE, NULL, &oldact);
301 sigpipe = oldact.sa_handler;
302 if (sigpipe != SIG_IGN)
303 sigaction (SIGPIPE, &newact, NULL);
305 sigaction (SIGTERM, NULL, &oldact);
306 sigterm = oldact.sa_handler;
307 if (sigterm != SIG_IGN)
308 sigaction (SIGTERM, &newact, NULL);
309 #else /* !_POSIX_VERSION */
310 sigint = signal (SIGINT, SIG_IGN);
311 if (sigint != SIG_IGN)
312 signal (SIGINT, cleanup);
314 sighup = signal (SIGHUP, SIG_IGN);
315 if (sighup != SIG_IGN)
316 signal (SIGHUP, cleanup);
318 sigpipe = signal (SIGPIPE, SIG_IGN);
319 if (sigpipe != SIG_IGN)
320 signal (SIGPIPE, cleanup);
322 sigterm = signal (SIGTERM, SIG_IGN);
323 if (sigterm != SIG_IGN)
324 signal (SIGTERM, cleanup);
325 #endif /* _POSIX_VERSION */
329 errors = tac_file (tempfile);
333 #ifdef _POSIX_VERSION
334 newact.sa_handler = sigint;
335 sigaction (SIGINT, &newact, NULL);
336 newact.sa_handler = sighup;
337 sigaction (SIGHUP, &newact, NULL);
338 newact.sa_handler = sigterm;
339 sigaction (SIGTERM, &newact, NULL);
340 newact.sa_handler = sigpipe;
341 sigaction (SIGPIPE, &newact, NULL);
342 #else /* !_POSIX_VERSION */
343 signal (SIGINT, sigint);
344 signal (SIGHUP, sighup);
345 signal (SIGTERM, sigterm);
346 signal (SIGPIPE, sigpipe);
347 #endif /* _POSIX_VERSION */
352 /* Make a copy of the standard input in `tempfile'. */
357 static char *template = NULL;
358 static char *tempdir;
362 if (template == NULL)
364 tempdir = getenv ("TMPDIR");
367 template = xmalloc (strlen (tempdir) + 11);
369 sprintf (template, "%s/tacXXXXXX", tempdir);
370 tempfile = mktemp (template);
372 fd = creat (tempfile, 0600);
375 error (0, errno, "%s", tempfile);
378 while ((bytes_read = safe_read (0, buffer, read_size)) > 0)
379 if (full_write (fd, buffer, bytes_read) < 0)
381 error (0, errno, "%s", tempfile);
386 error (0, errno, "%s", tempfile);
389 if (bytes_read == -1)
391 error (0, errno, "read error");
396 /* Print FILE in reverse.
397 Return 0 if ok, 1 if an error occurs. */
405 fd = open (file, O_RDONLY);
408 error (0, errno, "%s", file);
411 errors = tac (fd, file);
414 error (0, errno, "%s", file);
420 /* Print in reverse the file open on descriptor FD for reading FILE.
421 Return 0 if ok, 1 if an error occurs. */
428 /* Pointer to the location in `buffer' where the search for
429 the next separator will begin. */
431 /* Pointer to one past the rightmost character in `buffer' that
432 has not been printed yet. */
434 unsigned saved_record_size; /* Length of the record growing in `buffer'. */
435 off_t file_pos; /* Offset in the file of the next read. */
436 /* Nonzero if `output' has not been called yet for any file.
437 Only used when the separator is attached to the preceding record. */
439 char first_char = *separator; /* Speed optimization, non-regexp. */
440 char *separator1 = separator + 1; /* Speed optimization, non-regexp. */
441 int match_length1 = match_length - 1; /* Speed optimization, non-regexp. */
442 struct re_registers regs;
444 /* Find the size of the input file. */
445 file_pos = lseek (fd, (off_t) 0, SEEK_END);
447 return 0; /* It's an empty file. */
449 /* Arrange for the first read to lop off enough to leave the rest of the
450 file a multiple of `read_size'. Since `read_size' can change, this may
451 not always hold during the program run, but since it usually will, leave
452 it here for i/o efficiency (page/sector boundaries and all that).
453 Note: the efficiency gain has not been verified. */
454 saved_record_size = file_pos % read_size;
455 if (saved_record_size == 0)
456 saved_record_size = read_size;
457 file_pos -= saved_record_size;
458 /* `file_pos' now points to the start of the last (probably partial) block
459 in the input file. */
461 lseek (fd, file_pos, SEEK_SET);
462 if (safe_read (fd, buffer, saved_record_size) != saved_record_size)
464 error (0, errno, "%s", file);
468 match_start = past_end = buffer + saved_record_size;
469 /* For non-regexp search, move past impossible positions for a match. */
471 match_start -= match_length1;
475 /* Search backward from `match_start' - 1 to `buffer' for a match
476 with `separator'; for speed, use strncmp if `separator' contains no
478 If the match succeeds, set `match_start' to point to the start of
479 the match and `match_length' to the length of the match.
480 Otherwise, make `match_start' < `buffer'. */
481 if (sentinel_length == 0)
483 int i = match_start - buffer;
486 ret = re_search (&compiled_separator, buffer, i, i - 1, -i, ®s);
488 match_start = buffer - 1;
491 error (0, 0, "error in regular expression search");
496 match_start = buffer + regs.start[0];
497 match_length = regs.end[0] - regs.start[0];
502 /* `match_length' is constant for non-regexp boundaries. */
503 while (*--match_start != first_char
504 || (match_length1 && strncmp (match_start + 1, separator1,
509 /* Check whether we backed off the front of `buffer' without finding
510 a match for `separator'. */
511 if (match_start < buffer)
515 /* Hit the beginning of the file; print the remaining record. */
516 output (buffer, past_end);
520 saved_record_size = past_end - buffer;
521 if (saved_record_size > read_size)
523 /* `buffer_size' is about twice `read_size', so since
524 we want to read in another `read_size' bytes before
525 the data already in `buffer', we need to increase
528 int offset = sentinel_length ? sentinel_length : 1;
531 buffer_size = read_size * 2 + sentinel_length + 2;
532 newbuffer = xrealloc (buffer - offset, buffer_size) + offset;
533 /* Adjust the pointers for the new buffer location. */
534 match_start += newbuffer - buffer;
535 past_end += newbuffer - buffer;
539 /* Back up to the start of the next bufferfull of the file. */
540 if (file_pos >= read_size)
541 file_pos -= read_size;
544 read_size = file_pos;
547 lseek (fd, file_pos, SEEK_SET);
549 /* Shift the pending record data right to make room for the new.
550 The source and destination regions probably overlap. */
551 memmove (buffer + read_size, buffer, saved_record_size);
552 past_end = buffer + read_size + saved_record_size;
553 /* For non-regexp searches, avoid unneccessary scanning. */
555 match_start = buffer + read_size;
557 match_start = past_end;
559 if (safe_read (fd, buffer, read_size) != read_size)
561 error (0, errno, "%s", file);
567 /* Found a match of `separator'. */
568 if (separator_ends_record)
570 char *match_end = match_start + match_length;
572 /* If this match of `separator' isn't at the end of the
573 file, print the record. */
574 if (first_time == 0 || match_end != past_end)
575 output (match_end, past_end);
576 past_end = match_end;
581 output (match_start, past_end);
582 past_end = match_start;
584 match_start -= match_length - 1;
589 /* Print the characters from START to PAST_END - 1.
590 If START is NULL, just flush the buffer. */
593 output (start, past_end)
597 static char buffer[WRITESIZE];
598 static int bytes_in_buffer = 0;
599 int bytes_to_add = past_end - start;
600 int bytes_available = WRITESIZE - bytes_in_buffer;
604 xwrite (STDOUT_FILENO, buffer, bytes_in_buffer);
609 /* Write out as many full buffers as possible. */
610 while (bytes_to_add >= bytes_available)
612 memcpy (buffer + bytes_in_buffer, start, bytes_available);
613 bytes_to_add -= bytes_available;
614 start += bytes_available;
615 xwrite (STDOUT_FILENO, buffer, WRITESIZE);
617 bytes_available = WRITESIZE;
620 memcpy (buffer + bytes_in_buffer, start, bytes_to_add);
621 bytes_in_buffer += bytes_to_add;
632 xwrite (desc, buffer, size)
637 if (full_write (desc, buffer, size) < 0)
639 error (0, errno, "write error");
644 /* Allocate N bytes of memory dynamically, with error checking. */
655 error (0, 0, "virtual memory exhausted");
661 /* Change the size of memory area P to N bytes, with error checking. */
671 error (0, 0, "virtual memory exhausted");