1 /* tac - concatenate and print files in reverse
2 Copyright (C) 1988, 1989, 1990, 1991 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):
39 #if defined (CONFIG_BROKETS)
40 /* We use <config.h> instead of "config.h" so that a compilation
41 using -I. -I$srcdir will use ./config.h rather than $srcdir/config.h
42 (which it would do because it found this file in $srcdir). */
51 #include <sys/types.h>
62 /* The number of bytes per atomic read. */
63 #define INITIAL_READSIZE 8192
65 /* The number of bytes per atomic write. */
66 #define WRITESIZE 8192
70 static RETSIGTYPE cleanup ();
72 static int tac_file ();
73 static int tac_stdin ();
74 static char *xmalloc ();
75 static char *xrealloc ();
76 static void output ();
77 static void save_stdin ();
78 static void xwrite ();
84 /* The name this program was run with. */
87 /* The string that separates the records of the file. */
88 static char *separator;
90 /* If nonzero, print `separator' along with the record preceding it
91 in the file; otherwise with the record following it. */
92 static int separator_ends_record;
94 /* 0 if `separator' is to be matched as a regular expression;
95 otherwise, the length of `separator', used as a sentinel to
97 static int sentinel_length;
99 /* The length of a match with `separator'. If `sentinel_length' is 0,
100 `match_length' is computed every time a match succeeds;
101 otherwise, it is simply the length of `separator'. */
102 static int match_length;
104 /* The input buffer. */
107 /* The number of bytes to read at once into `buffer'. */
108 static unsigned read_size;
110 /* The size of `buffer'. This is read_size * 2 + sentinel_length + 2.
111 The extra 2 bytes allow `past_end' to have a value beyond the
112 end of `buffer' and `match_start' to run off the front of `buffer'. */
113 static unsigned buffer_size;
115 /* The compiled regular expression representing `separator'. */
116 static struct re_pattern_buffer compiled_separator;
118 /* If non-zero, display usage information and exit. */
119 static int show_help;
121 /* If non-zero, print the version on standard output then exit. */
122 static int show_version;
124 static struct option const longopts[] =
126 {"before", no_argument, &separator_ends_record, 0},
127 {"regex", no_argument, &sentinel_length, 0},
128 {"separator", required_argument, NULL, 's'},
129 {"help", no_argument, &show_help, 1},
130 {"version", no_argument, &show_version, 1},
139 fprintf (stderr, "Try `%s --help' for more information.\n",
144 Usage: %s [OPTION]... [FILE]...\n\
149 -b, --before attach the separator before instead of after\n\
150 -r, --regex interpret the separator as a regular expression\n\
151 -s, --separator=STRING use STRING as the separator instead of newline\n\
152 --help display this help and exit\n\
153 --version output version information and exit\n\
155 With no FILE, or when FILE is -, read standard input.\n\
166 const char *error_message; /* Return value from re_compile_pattern. */
168 int have_read_stdin = 0;
170 program_name = argv[0];
174 separator_ends_record = 1;
176 while ((optc = getopt_long (argc, argv, "brs:", longopts, (int *) 0))
184 separator_ends_record = 0;
192 error (1, 0, "separator cannot be empty");
201 printf ("%s\n", version_string);
208 if (sentinel_length == 0)
210 compiled_separator.allocated = 100;
211 compiled_separator.buffer = (unsigned char *)
212 xmalloc (compiled_separator.allocated);
213 compiled_separator.fastmap = xmalloc (256);
214 compiled_separator.translate = 0;
215 error_message = re_compile_pattern (separator, strlen (separator),
216 &compiled_separator);
218 error (1, 0, "%s", error_message);
221 match_length = sentinel_length = strlen (separator);
223 read_size = INITIAL_READSIZE;
224 /* A precaution that will probably never be needed. */
225 while (sentinel_length * 2 >= read_size)
227 buffer_size = read_size * 2 + sentinel_length + 2;
228 buffer = xmalloc (buffer_size);
231 strcpy (buffer, separator);
232 buffer += sentinel_length;
240 errors = tac_stdin ();
243 for (; optind < argc; ++optind)
245 if (strcmp (argv[optind], "-") == 0)
248 errors |= tac_stdin ();
251 errors |= tac_file (argv[optind]);
254 /* Flush the output buffer. */
255 output ((char *) NULL, (char *) NULL);
257 if (have_read_stdin && close (0) < 0)
258 error (1, errno, "-");
260 error (1, errno, "write error");
264 /* The name of a temporary file containing a copy of pipe input. */
267 /* Print the standard input in reverse, saving it to temporary
268 file `tempfile' first if it is a pipe.
269 Return 0 if ok, 1 if an error occurs. */
274 /* Previous values of signal handlers. */
275 RETSIGTYPE (*sigint) (), (*sighup) (), (*sigpipe) (), (*sigterm) ();
278 #ifdef _POSIX_VERSION
279 struct sigaction oldact, newact;
280 #endif /* _POSIX_VERSION */
282 /* No tempfile is needed for "tac < file".
283 Use fstat instead of checking for errno == ESPIPE because
284 lseek doesn't work on some special files but doesn't return an
286 if (fstat (0, &stats))
288 error (0, errno, "standard input");
291 if (S_ISREG (stats.st_mode))
292 return tac (0, "standard input");
294 #ifdef _POSIX_VERSION
295 newact.sa_handler = cleanup;
296 sigemptyset (&newact.sa_mask);
299 sigaction (SIGINT, NULL, &oldact);
300 sigint = oldact.sa_handler;
301 if (sigint != SIG_IGN)
302 sigaction (SIGINT, &newact, NULL);
304 sigaction (SIGHUP, NULL, &oldact);
305 sighup = oldact.sa_handler;
306 if (sighup != SIG_IGN)
307 sigaction (SIGHUP, &newact, NULL);
309 sigaction (SIGPIPE, NULL, &oldact);
310 sigpipe = oldact.sa_handler;
311 if (sigpipe != SIG_IGN)
312 sigaction (SIGPIPE, &newact, NULL);
314 sigaction (SIGTERM, NULL, &oldact);
315 sigterm = oldact.sa_handler;
316 if (sigterm != SIG_IGN)
317 sigaction (SIGTERM, &newact, NULL);
318 #else /* !_POSIX_VERSION */
319 sigint = signal (SIGINT, SIG_IGN);
320 if (sigint != SIG_IGN)
321 signal (SIGINT, cleanup);
323 sighup = signal (SIGHUP, SIG_IGN);
324 if (sighup != SIG_IGN)
325 signal (SIGHUP, cleanup);
327 sigpipe = signal (SIGPIPE, SIG_IGN);
328 if (sigpipe != SIG_IGN)
329 signal (SIGPIPE, cleanup);
331 sigterm = signal (SIGTERM, SIG_IGN);
332 if (sigterm != SIG_IGN)
333 signal (SIGTERM, cleanup);
334 #endif /* _POSIX_VERSION */
338 errors = tac_file (tempfile);
342 #ifdef _POSIX_VERSION
343 newact.sa_handler = sigint;
344 sigaction (SIGINT, &newact, NULL);
345 newact.sa_handler = sighup;
346 sigaction (SIGHUP, &newact, NULL);
347 newact.sa_handler = sigterm;
348 sigaction (SIGTERM, &newact, NULL);
349 newact.sa_handler = sigpipe;
350 sigaction (SIGPIPE, &newact, NULL);
351 #else /* !_POSIX_VERSION */
352 signal (SIGINT, sigint);
353 signal (SIGHUP, sighup);
354 signal (SIGTERM, sigterm);
355 signal (SIGPIPE, sigpipe);
356 #endif /* _POSIX_VERSION */
361 /* Make a copy of the standard input in `tempfile'. */
366 static char *template = NULL;
367 static char *tempdir;
371 if (template == NULL)
373 tempdir = getenv ("TMPDIR");
376 template = xmalloc (strlen (tempdir) + 11);
378 sprintf (template, "%s/tacXXXXXX", tempdir);
379 tempfile = mktemp (template);
381 fd = creat (tempfile, 0600);
384 error (0, errno, "%s", tempfile);
387 while ((bytes_read = safe_read (0, buffer, read_size)) > 0)
388 if (full_write (fd, buffer, bytes_read) < 0)
390 error (0, errno, "%s", tempfile);
395 error (0, errno, "%s", tempfile);
398 if (bytes_read == -1)
400 error (0, errno, "read error");
405 /* Print FILE in reverse.
406 Return 0 if ok, 1 if an error occurs. */
417 error (0, errno, "%s", file);
420 errors = tac (fd, file);
423 error (0, errno, "%s", file);
429 /* Print in reverse the file open on descriptor FD for reading FILE.
430 Return 0 if ok, 1 if an error occurs. */
437 /* Pointer to the location in `buffer' where the search for
438 the next separator will begin. */
440 /* Pointer to one past the rightmost character in `buffer' that
441 has not been printed yet. */
443 unsigned saved_record_size; /* Length of the record growing in `buffer'. */
444 off_t file_pos; /* Offset in the file of the next read. */
445 /* Nonzero if `output' has not been called yet for any file.
446 Only used when the separator is attached to the preceding record. */
448 char first_char = *separator; /* Speed optimization, non-regexp. */
449 char *separator1 = separator + 1; /* Speed optimization, non-regexp. */
450 int match_length1 = match_length - 1; /* Speed optimization, non-regexp. */
451 struct re_registers regs;
453 /* Find the size of the input file. */
454 file_pos = lseek (fd, (off_t) 0, SEEK_END);
456 return 0; /* It's an empty file. */
458 /* Arrange for the first read to lop off enough to leave the rest of the
459 file a multiple of `read_size'. Since `read_size' can change, this may
460 not always hold during the program run, but since it usually will, leave
461 it here for i/o efficiency (page/sector boundaries and all that).
462 Note: the efficiency gain has not been verified. */
463 saved_record_size = file_pos % read_size;
464 if (saved_record_size == 0)
465 saved_record_size = read_size;
466 file_pos -= saved_record_size;
467 /* `file_pos' now points to the start of the last (probably partial) block
468 in the input file. */
470 lseek (fd, file_pos, SEEK_SET);
471 if (safe_read (fd, buffer, saved_record_size) != saved_record_size)
473 error (0, 1, "%s", file);
477 match_start = past_end = buffer + saved_record_size;
478 /* For non-regexp search, move past impossible positions for a match. */
480 match_start -= match_length1;
484 /* Search backward from `match_start' - 1 to `buffer' for a match
485 with `separator'; for speed, use strncmp if `separator' contains no
487 If the match succeeds, set `match_start' to point to the start of
488 the match and `match_length' to the length of the match.
489 Otherwise, make `match_start' < `buffer'. */
490 if (sentinel_length == 0)
492 int i = match_start - buffer;
495 ret = re_search (&compiled_separator, buffer, i, i - 1, -i, ®s);
497 match_start = buffer - 1;
500 error (0, 0, "error in regular expression search");
505 match_start = buffer + regs.start[0];
506 match_length = regs.end[0] - regs.start[0];
511 /* `match_length' is constant for non-regexp boundaries. */
512 while (*--match_start != first_char
513 || (match_length1 && strncmp (match_start + 1, separator1,
518 /* Check whether we backed off the front of `buffer' without finding
519 a match for `separator'. */
520 if (match_start < buffer)
524 /* Hit the beginning of the file; print the remaining record. */
525 output (buffer, past_end);
529 saved_record_size = past_end - buffer;
530 if (saved_record_size > read_size)
532 /* `buffer_size' is about twice `read_size', so since
533 we want to read in another `read_size' bytes before
534 the data already in `buffer', we need to increase
537 int offset = sentinel_length ? sentinel_length : 1;
540 buffer_size = read_size * 2 + sentinel_length + 2;
541 newbuffer = xrealloc (buffer - offset, buffer_size) + offset;
542 /* Adjust the pointers for the new buffer location. */
543 match_start += newbuffer - buffer;
544 past_end += newbuffer - buffer;
548 /* Back up to the start of the next bufferfull of the file. */
549 if (file_pos >= read_size)
550 file_pos -= read_size;
553 read_size = file_pos;
556 lseek (fd, file_pos, SEEK_SET);
558 /* Shift the pending record data right to make room for the new. */
559 bcopy (buffer, buffer + read_size, saved_record_size);
560 past_end = buffer + read_size + saved_record_size;
561 /* For non-regexp searches, avoid unneccessary scanning. */
563 match_start = buffer + read_size;
565 match_start = past_end;
567 if (safe_read (fd, buffer, read_size) != read_size)
569 error (0, errno, "%s", file);
575 /* Found a match of `separator'. */
576 if (separator_ends_record)
578 char *match_end = match_start + match_length;
580 /* If this match of `separator' isn't at the end of the
581 file, print the record. */
582 if (first_time == 0 || match_end != past_end)
583 output (match_end, past_end);
584 past_end = match_end;
589 output (match_start, past_end);
590 past_end = match_start;
592 match_start -= match_length - 1;
597 /* Print the characters from START to PAST_END - 1.
598 If START is NULL, just flush the buffer. */
601 output (start, past_end)
605 static char buffer[WRITESIZE];
606 static int bytes_in_buffer = 0;
607 int bytes_to_add = past_end - start;
608 int bytes_available = WRITESIZE - bytes_in_buffer;
612 xwrite (1, buffer, bytes_in_buffer);
617 /* Write out as many full buffers as possible. */
618 while (bytes_to_add >= bytes_available)
620 bcopy (start, buffer + bytes_in_buffer, bytes_available);
621 bytes_to_add -= bytes_available;
622 start += bytes_available;
623 xwrite (1, buffer, WRITESIZE);
625 bytes_available = WRITESIZE;
628 bcopy (start, buffer + bytes_in_buffer, bytes_to_add);
629 bytes_in_buffer += bytes_to_add;
640 xwrite (desc, buffer, size)
645 if (full_write (desc, buffer, size) < 0)
647 error (0, errno, "write error");
652 /* Allocate N bytes of memory dynamically, with error checking. */
663 error (0, 0, "virtual memory exhausted");
669 /* Change the size of memory area P to N bytes, with error checking. */
679 error (0, 0, "virtual memory exhausted");