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 ();
82 /* The name this program was run with. */
85 /* The string that separates the records of the file. */
86 static char *separator;
88 /* If nonzero, print `separator' along with the record preceding it
89 in the file; otherwise with the record following it. */
90 static int separator_ends_record;
92 /* 0 if `separator' is to be matched as a regular expression;
93 otherwise, the length of `separator', used as a sentinel to
95 static int sentinel_length;
97 /* The length of a match with `separator'. If `sentinel_length' is 0,
98 `match_length' is computed every time a match succeeds;
99 otherwise, it is simply the length of `separator'. */
100 static int match_length;
102 /* The input buffer. */
105 /* The number of bytes to read at once into `buffer'. */
106 static unsigned read_size;
108 /* The size of `buffer'. This is read_size * 2 + sentinel_length + 2.
109 The extra 2 bytes allow `past_end' to have a value beyond the
110 end of `buffer' and `match_start' to run off the front of `buffer'. */
111 static unsigned buffer_size;
113 /* The compiled regular expression representing `separator'. */
114 static struct re_pattern_buffer compiled_separator;
116 /* If non-zero, display usage information and exit. */
117 static int show_help;
119 /* If non-zero, print the version on standard output then exit. */
120 static int show_version;
122 static struct option const longopts[] =
124 {"before", no_argument, &separator_ends_record, 0},
125 {"regex", no_argument, &sentinel_length, 0},
126 {"separator", required_argument, NULL, 's'},
127 {"help", no_argument, &show_help, 1},
128 {"version", no_argument, &show_version, 1},
137 fprintf (stderr, "Try `%s --help' for more information.\n",
142 Usage: %s [OPTION]... [FILE]...\n\
147 -b, --before attach the separator before instead of after\n\
148 -r, --regex intepret the separator as a regular expression\n\
149 -s, --separator STRING use STRING as the separator instead of newline\n\
150 --help display this help and exit\n\
151 --version output version information and exit\n\
153 With no FILE, or when FILE is -, read standard input.\n\
164 const char *error_message; /* Return value from re_compile_pattern. */
166 int have_read_stdin = 0;
168 program_name = argv[0];
172 separator_ends_record = 1;
174 while ((optc = getopt_long (argc, argv, "brs:", longopts, (int *) 0))
182 separator_ends_record = 0;
190 error (1, 0, "separator cannot be empty");
199 printf ("%s\n", version_string);
206 if (sentinel_length == 0)
208 compiled_separator.allocated = 100;
209 compiled_separator.buffer = (unsigned char *)
210 xmalloc (compiled_separator.allocated);
211 compiled_separator.fastmap = xmalloc (256);
212 compiled_separator.translate = 0;
213 error_message = re_compile_pattern (separator, strlen (separator),
214 &compiled_separator);
216 error (1, 0, "%s", error_message);
219 match_length = sentinel_length = strlen (separator);
221 read_size = INITIAL_READSIZE;
222 /* A precaution that will probably never be needed. */
223 while (sentinel_length * 2 >= read_size)
225 buffer_size = read_size * 2 + sentinel_length + 2;
226 buffer = xmalloc (buffer_size);
229 strcpy (buffer, separator);
230 buffer += sentinel_length;
238 errors = tac_stdin ();
241 for (; optind < argc; ++optind)
243 if (strcmp (argv[optind], "-") == 0)
246 errors |= tac_stdin ();
249 errors |= tac_file (argv[optind]);
252 /* Flush the output buffer. */
253 output ((char *) NULL, (char *) NULL);
255 if (have_read_stdin && close (0) < 0)
256 error (1, errno, "-");
258 error (1, errno, "write error");
262 /* The name of a temporary file containing a copy of pipe input. */
265 /* Print the standard input in reverse, saving it to temporary
266 file `tempfile' first if it is a pipe.
267 Return 0 if ok, 1 if an error occurs. */
272 /* Previous values of signal handlers. */
273 RETSIGTYPE (*sigint) (), (*sighup) (), (*sigpipe) (), (*sigterm) ();
276 #ifdef _POSIX_VERSION
277 struct sigaction oldact, newact;
278 #endif /* _POSIX_VERSION */
280 /* No tempfile is needed for "tac < file".
281 Use fstat instead of checking for errno == ESPIPE because
282 lseek doesn't work on some special files but doesn't return an
284 if (fstat (0, &stats))
286 error (0, errno, "standard input");
289 if (S_ISREG (stats.st_mode))
290 return tac (0, "standard input");
292 #ifdef _POSIX_VERSION
293 newact.sa_handler = cleanup;
294 sigemptyset (&newact.sa_mask);
297 sigaction (SIGINT, NULL, &oldact);
298 sigint = oldact.sa_handler;
299 if (sigint != SIG_IGN)
300 sigaction (SIGINT, &newact, NULL);
302 sigaction (SIGHUP, NULL, &oldact);
303 sighup = oldact.sa_handler;
304 if (sighup != SIG_IGN)
305 sigaction (SIGHUP, &newact, NULL);
307 sigaction (SIGPIPE, NULL, &oldact);
308 sigpipe = oldact.sa_handler;
309 if (sigpipe != SIG_IGN)
310 sigaction (SIGPIPE, &newact, NULL);
312 sigaction (SIGTERM, NULL, &oldact);
313 sigterm = oldact.sa_handler;
314 if (sigterm != SIG_IGN)
315 sigaction (SIGTERM, &newact, NULL);
316 #else /* !_POSIX_VERSION */
317 sigint = signal (SIGINT, SIG_IGN);
318 if (sigint != SIG_IGN)
319 signal (SIGINT, cleanup);
321 sighup = signal (SIGHUP, SIG_IGN);
322 if (sighup != SIG_IGN)
323 signal (SIGHUP, cleanup);
325 sigpipe = signal (SIGPIPE, SIG_IGN);
326 if (sigpipe != SIG_IGN)
327 signal (SIGPIPE, cleanup);
329 sigterm = signal (SIGTERM, SIG_IGN);
330 if (sigterm != SIG_IGN)
331 signal (SIGTERM, cleanup);
332 #endif /* _POSIX_VERSION */
336 errors = tac_file (tempfile);
340 #ifdef _POSIX_VERSION
341 newact.sa_handler = sigint;
342 sigaction (SIGINT, &newact, NULL);
343 newact.sa_handler = sighup;
344 sigaction (SIGHUP, &newact, NULL);
345 newact.sa_handler = sigterm;
346 sigaction (SIGTERM, &newact, NULL);
347 newact.sa_handler = sigpipe;
348 sigaction (SIGPIPE, &newact, NULL);
349 #else /* !_POSIX_VERSION */
350 signal (SIGINT, sigint);
351 signal (SIGHUP, sighup);
352 signal (SIGTERM, sigterm);
353 signal (SIGPIPE, sigpipe);
354 #endif /* _POSIX_VERSION */
359 /* Make a copy of the standard input in `tempfile'. */
364 static char *template = NULL;
365 static char *tempdir;
369 if (template == NULL)
371 tempdir = getenv ("TMPDIR");
374 template = xmalloc (strlen (tempdir) + 11);
376 sprintf (template, "%s/tacXXXXXX", tempdir);
377 tempfile = mktemp (template);
379 fd = creat (tempfile, 0600);
382 error (0, errno, "%s", tempfile);
385 while ((bytes_read = read (0, buffer, read_size)) > 0)
386 if (write (fd, buffer, bytes_read) != bytes_read)
388 error (0, errno, "%s", tempfile);
393 error (0, errno, "%s", tempfile);
396 if (bytes_read == -1)
398 error (0, errno, "read error");
403 /* Print FILE in reverse.
404 Return 0 if ok, 1 if an error occurs. */
415 error (0, errno, "%s", file);
418 errors = tac (fd, file);
421 error (0, errno, "%s", file);
427 /* Print in reverse the file open on descriptor FD for reading FILE.
428 Return 0 if ok, 1 if an error occurs. */
435 /* Pointer to the location in `buffer' where the search for
436 the next separator will begin. */
438 /* Pointer to one past the rightmost character in `buffer' that
439 has not been printed yet. */
441 unsigned saved_record_size; /* Length of the record growing in `buffer'. */
442 off_t file_pos; /* Offset in the file of the next read. */
443 /* Nonzero if `output' has not been called yet for any file.
444 Only used when the separator is attached to the preceding record. */
446 char first_char = *separator; /* Speed optimization, non-regexp. */
447 char *separator1 = separator + 1; /* Speed optimization, non-regexp. */
448 int match_length1 = match_length - 1; /* Speed optimization, non-regexp. */
449 struct re_registers regs;
451 /* Find the size of the input file. */
452 file_pos = lseek (fd, (off_t) 0, SEEK_END);
454 return 0; /* It's an empty file. */
456 /* Arrange for the first read to lop off enough to leave the rest of the
457 file a multiple of `read_size'. Since `read_size' can change, this may
458 not always hold during the program run, but since it usually will, leave
459 it here for i/o efficiency (page/sector boundaries and all that).
460 Note: the efficiency gain has not been verified. */
461 saved_record_size = file_pos % read_size;
462 if (saved_record_size == 0)
463 saved_record_size = read_size;
464 file_pos -= saved_record_size;
465 /* `file_pos' now points to the start of the last (probably partial) block
466 in the input file. */
468 lseek (fd, file_pos, SEEK_SET);
469 if (read (fd, buffer, saved_record_size) != saved_record_size)
471 error (0, 1, "%s", file);
475 match_start = past_end = buffer + saved_record_size;
476 /* For non-regexp search, move past impossible positions for a match. */
478 match_start -= match_length1;
482 /* Search backward from `match_start' - 1 to `buffer' for a match
483 with `separator'; for speed, use strncmp if `separator' contains no
485 If the match succeeds, set `match_start' to point to the start of
486 the match and `match_length' to the length of the match.
487 Otherwise, make `match_start' < `buffer'. */
488 if (sentinel_length == 0)
490 int i = match_start - buffer;
493 ret = re_search (&compiled_separator, buffer, i, i - 1, -i, ®s);
495 match_start = buffer - 1;
498 error (0, 0, "error in regular expression search");
503 match_start = buffer + regs.start[0];
504 match_length = regs.end[0] - regs.start[0];
509 /* `match_length' is constant for non-regexp boundaries. */
510 while (*--match_start != first_char
511 || (match_length1 && strncmp (match_start + 1, separator1,
516 /* Check whether we backed off the front of `buffer' without finding
517 a match for `separator'. */
518 if (match_start < buffer)
522 /* Hit the beginning of the file; print the remaining record. */
523 output (buffer, past_end);
527 saved_record_size = past_end - buffer;
528 if (saved_record_size > read_size)
530 /* `buffer_size' is about twice `read_size', so since
531 we want to read in another `read_size' bytes before
532 the data already in `buffer', we need to increase
535 int offset = sentinel_length ? sentinel_length : 1;
538 buffer_size = read_size * 2 + sentinel_length + 2;
539 newbuffer = xrealloc (buffer - offset, buffer_size) + offset;
540 /* Adjust the pointers for the new buffer location. */
541 match_start += newbuffer - buffer;
542 past_end += newbuffer - buffer;
546 /* Back up to the start of the next bufferfull of the file. */
547 if (file_pos >= read_size)
548 file_pos -= read_size;
551 read_size = file_pos;
554 lseek (fd, file_pos, SEEK_SET);
556 /* Shift the pending record data right to make room for the new. */
557 bcopy (buffer, buffer + read_size, saved_record_size);
558 past_end = buffer + read_size + saved_record_size;
559 /* For non-regexp searches, avoid unneccessary scanning. */
561 match_start = buffer + read_size;
563 match_start = past_end;
565 if (read (fd, buffer, read_size) != read_size)
567 error (0, errno, "%s", file);
573 /* Found a match of `separator'. */
574 if (separator_ends_record)
576 char *match_end = match_start + match_length;
578 /* If this match of `separator' isn't at the end of the
579 file, print the record. */
580 if (first_time == 0 || match_end != past_end)
581 output (match_end, past_end);
582 past_end = match_end;
587 output (match_start, past_end);
588 past_end = match_start;
590 match_start -= match_length - 1;
595 /* Print the characters from START to PAST_END - 1.
596 If START is NULL, just flush the buffer. */
599 output (start, past_end)
603 static char buffer[WRITESIZE];
604 static int bytes_in_buffer = 0;
605 int bytes_to_add = past_end - start;
606 int bytes_available = WRITESIZE - bytes_in_buffer;
610 xwrite (1, buffer, bytes_in_buffer);
615 /* Write out as many full buffers as possible. */
616 while (bytes_to_add >= bytes_available)
618 bcopy (start, buffer + bytes_in_buffer, bytes_available);
619 bytes_to_add -= bytes_available;
620 start += bytes_available;
621 xwrite (1, buffer, WRITESIZE);
623 bytes_available = WRITESIZE;
626 bcopy (start, buffer + bytes_in_buffer, bytes_to_add);
627 bytes_in_buffer += bytes_to_add;
638 xwrite (desc, buffer, size)
643 if (write (desc, buffer, size) != size)
645 error (0, errno, "write error");
650 /* Allocate N bytes of memory dynamically, with error checking. */
661 error (0, 0, "virtual memory exhausted");
667 /* Change the size of memory area P to N bytes, with error checking. */
677 error (0, 0, "virtual memory exhausted");