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):
40 #include <sys/types.h>
50 /* The number of bytes per atomic read. */
51 #define INITIAL_READSIZE 8192
53 /* The number of bytes per atomic write. */
54 #define WRITESIZE 8192
58 static RETSIGTYPE cleanup ();
60 static int tac_file ();
61 static int tac_stdin ();
62 static char *xmalloc ();
63 static char *xrealloc ();
64 static void output ();
65 static void save_stdin ();
66 static void xwrite ();
70 /* The name this program was run with. */
73 /* The string that separates the records of the file. */
74 static char *separator;
76 /* If nonzero, print `separator' along with the record preceding it
77 in the file; otherwise with the record following it. */
78 static int separator_ends_record;
80 /* 0 if `separator' is to be matched as a regular expression;
81 otherwise, the length of `separator', used as a sentinel to
83 static int sentinel_length;
85 /* The length of a match with `separator'. If `sentinel_length' is 0,
86 `match_length' is computed every time a match succeeds;
87 otherwise, it is simply the length of `separator'. */
88 static int match_length;
90 /* The input buffer. */
93 /* The number of bytes to read at once into `buffer'. */
94 static unsigned read_size;
96 /* The size of `buffer'. This is read_size * 2 + sentinel_length + 2.
97 The extra 2 bytes allow `past_end' to have a value beyond the
98 end of `buffer' and `match_start' to run off the front of `buffer'. */
99 static unsigned buffer_size;
101 /* The compiled regular expression representing `separator'. */
102 static struct re_pattern_buffer compiled_separator;
104 static struct option const longopts[] =
106 {"before", 0, &separator_ends_record, 0},
107 {"regex", 0, &sentinel_length, 0},
108 {"separator", 1, NULL, 's'},
117 const char *error_message; /* Return value from re_compile_pattern. */
119 int have_read_stdin = 0;
121 program_name = argv[0];
125 separator_ends_record = 1;
127 while ((optc = getopt_long (argc, argv, "brs:", longopts, (int *) 0))
135 separator_ends_record = 0;
143 error (1, 0, "separator cannot be empty");
147 Usage: %s [-br] [-s separator] [--before] [--regex] [--separator=separator]\n\
154 if (sentinel_length == 0)
156 compiled_separator.allocated = 100;
157 compiled_separator.buffer = (unsigned char *)
158 xmalloc (compiled_separator.allocated);
159 compiled_separator.fastmap = xmalloc (256);
160 compiled_separator.translate = 0;
161 error_message = re_compile_pattern (separator, strlen (separator),
162 &compiled_separator);
164 error (1, 0, "%s", error_message);
167 match_length = sentinel_length = strlen (separator);
169 read_size = INITIAL_READSIZE;
170 /* A precaution that will probably never be needed. */
171 while (sentinel_length * 2 >= read_size)
173 buffer_size = read_size * 2 + sentinel_length + 2;
174 buffer = xmalloc (buffer_size);
177 strcpy (buffer, separator);
178 buffer += sentinel_length;
186 errors = tac_stdin ();
189 for (; optind < argc; ++optind)
191 if (strcmp (argv[optind], "-") == 0)
194 errors |= tac_stdin ();
197 errors |= tac_file (argv[optind]);
200 /* Flush the output buffer. */
201 output ((char *) NULL, (char *) NULL);
203 if (have_read_stdin && close (0) < 0)
204 error (1, errno, "-");
206 error (1, errno, "write error");
210 /* The name of a temporary file containing a copy of pipe input. */
213 /* Print the standard input in reverse, saving it to temporary
214 file `tempfile' first if it is a pipe.
215 Return 0 if ok, 1 if an error occurs. */
220 /* Previous values of signal handlers. */
221 RETSIGTYPE (*sigint) (), (*sighup) (), (*sigpipe) (), (*sigterm) ();
224 #ifdef _POSIX_VERSION
225 struct sigaction oldact, newact;
226 #endif /* _POSIX_VERSION */
228 /* No tempfile is needed for "tac < file".
229 Use fstat instead of checking for errno == ESPIPE because
230 lseek doesn't work on some special files but doesn't return an
232 if (fstat (0, &stats))
234 error (0, errno, "standard input");
237 if (S_ISREG (stats.st_mode))
238 return tac (0, "standard input");
240 #ifdef _POSIX_VERSION
241 newact.sa_handler = cleanup;
242 sigemptyset (&newact.sa_mask);
245 sigaction (SIGINT, NULL, &oldact);
246 sigint = oldact.sa_handler;
247 if (sigint != SIG_IGN)
248 sigaction (SIGINT, &newact, NULL);
250 sigaction (SIGHUP, NULL, &oldact);
251 sighup = oldact.sa_handler;
252 if (sighup != SIG_IGN)
253 sigaction (SIGHUP, &newact, NULL);
255 sigaction (SIGPIPE, NULL, &oldact);
256 sigpipe = oldact.sa_handler;
257 if (sigpipe != SIG_IGN)
258 sigaction (SIGPIPE, &newact, NULL);
260 sigaction (SIGTERM, NULL, &oldact);
261 sigterm = oldact.sa_handler;
262 if (sigterm != SIG_IGN)
263 sigaction (SIGTERM, &newact, NULL);
264 #else /* !_POSIX_VERSION */
265 sigint = signal (SIGINT, SIG_IGN);
266 if (sigint != SIG_IGN)
267 signal (SIGINT, cleanup);
269 sighup = signal (SIGHUP, SIG_IGN);
270 if (sighup != SIG_IGN)
271 signal (SIGHUP, cleanup);
273 sigpipe = signal (SIGPIPE, SIG_IGN);
274 if (sigpipe != SIG_IGN)
275 signal (SIGPIPE, cleanup);
277 sigterm = signal (SIGTERM, SIG_IGN);
278 if (sigterm != SIG_IGN)
279 signal (SIGTERM, cleanup);
280 #endif /* _POSIX_VERSION */
284 errors = tac_file (tempfile);
288 #ifdef _POSIX_VERSION
289 newact.sa_handler = sigint;
290 sigaction (SIGINT, &newact, NULL);
291 newact.sa_handler = sighup;
292 sigaction (SIGHUP, &newact, NULL);
293 newact.sa_handler = sigterm;
294 sigaction (SIGTERM, &newact, NULL);
295 newact.sa_handler = sigpipe;
296 sigaction (SIGPIPE, &newact, NULL);
297 #else /* !_POSIX_VERSION */
298 signal (SIGINT, sigint);
299 signal (SIGHUP, sighup);
300 signal (SIGTERM, sigterm);
301 signal (SIGPIPE, sigpipe);
302 #endif /* _POSIX_VERSION */
307 /* Make a copy of the standard input in `tempfile'. */
312 static char *template = NULL;
313 static char *tempdir;
317 if (template == NULL)
319 tempdir = getenv ("TMPDIR");
322 template = xmalloc (strlen (tempdir) + 11);
324 sprintf (template, "%s/tacXXXXXX", tempdir);
325 tempfile = mktemp (template);
327 fd = creat (tempfile, 0600);
330 error (0, errno, "%s", tempfile);
333 while ((bytes_read = read (0, buffer, read_size)) > 0)
334 if (write (fd, buffer, bytes_read) != bytes_read)
336 error (0, errno, "%s", tempfile);
341 error (0, errno, "%s", tempfile);
344 if (bytes_read == -1)
346 error (0, errno, "read error");
351 /* Print FILE in reverse.
352 Return 0 if ok, 1 if an error occurs. */
363 error (0, errno, "%s", file);
366 errors = tac (fd, file);
369 error (0, errno, "%s", file);
375 /* Print in reverse the file open on descriptor FD for reading FILE.
376 Return 0 if ok, 1 if an error occurs. */
383 /* Pointer to the location in `buffer' where the search for
384 the next separator will begin. */
386 /* Pointer to one past the rightmost character in `buffer' that
387 has not been printed yet. */
389 unsigned saved_record_size; /* Length of the record growing in `buffer'. */
390 off_t file_pos; /* Offset in the file of the next read. */
391 /* Nonzero if `output' has not been called yet for any file.
392 Only used when the separator is attached to the preceding record. */
394 char first_char = *separator; /* Speed optimization, non-regexp. */
395 char *separator1 = separator + 1; /* Speed optimization, non-regexp. */
396 int match_length1 = match_length - 1; /* Speed optimization, non-regexp. */
397 struct re_registers regs;
399 /* Find the size of the input file. */
400 file_pos = lseek (fd, (off_t) 0, SEEK_END);
402 return 0; /* It's an empty file. */
404 /* Arrange for the first read to lop off enough to leave the rest of the
405 file a multiple of `read_size'. Since `read_size' can change, this may
406 not always hold during the program run, but since it usually will, leave
407 it here for i/o efficiency (page/sector boundaries and all that).
408 Note: the efficiency gain has not been verified. */
409 saved_record_size = file_pos % read_size;
410 if (saved_record_size == 0)
411 saved_record_size = read_size;
412 file_pos -= saved_record_size;
413 /* `file_pos' now points to the start of the last (probably partial) block
414 in the input file. */
416 lseek (fd, file_pos, SEEK_SET);
417 if (read (fd, buffer, saved_record_size) != saved_record_size)
419 error (0, 1, "%s", file);
423 match_start = past_end = buffer + saved_record_size;
424 /* For non-regexp search, move past impossible positions for a match. */
426 match_start -= match_length1;
430 /* Search backward from `match_start' - 1 to `buffer' for a match
431 with `separator'; for speed, use strncmp if `separator' contains no
433 If the match succeeds, set `match_start' to point to the start of
434 the match and `match_length' to the length of the match.
435 Otherwise, make `match_start' < `buffer'. */
436 if (sentinel_length == 0)
438 int i = match_start - buffer;
441 ret = re_search (&compiled_separator, buffer, i, i - 1, -i, ®s);
443 match_start = buffer - 1;
446 error (0, 0, "error in regular expression search");
451 match_start = buffer + regs.start[0];
452 match_length = regs.end[0] - regs.start[0];
457 /* `match_length' is constant for non-regexp boundaries. */
458 while (*--match_start != first_char
459 || (match_length1 && strncmp (match_start + 1, separator1,
464 /* Check whether we backed off the front of `buffer' without finding
465 a match for `separator'. */
466 if (match_start < buffer)
470 /* Hit the beginning of the file; print the remaining record. */
471 output (buffer, past_end);
475 saved_record_size = past_end - buffer;
476 if (saved_record_size > read_size)
478 /* `buffer_size' is about twice `read_size', so since
479 we want to read in another `read_size' bytes before
480 the data already in `buffer', we need to increase
483 int offset = sentinel_length ? sentinel_length : 1;
486 buffer_size = read_size * 2 + sentinel_length + 2;
487 newbuffer = xrealloc (buffer - offset, buffer_size) + offset;
488 /* Adjust the pointers for the new buffer location. */
489 match_start += newbuffer - buffer;
490 past_end += newbuffer - buffer;
494 /* Back up to the start of the next bufferfull of the file. */
495 if (file_pos >= read_size)
496 file_pos -= read_size;
499 read_size = file_pos;
502 lseek (fd, file_pos, SEEK_SET);
504 /* Shift the pending record data right to make room for the new. */
505 bcopy (buffer, buffer + read_size, saved_record_size);
506 past_end = buffer + read_size + saved_record_size;
507 /* For non-regexp searches, avoid unneccessary scanning. */
509 match_start = buffer + read_size;
511 match_start = past_end;
513 if (read (fd, buffer, read_size) != read_size)
515 error (0, errno, "%s", file);
521 /* Found a match of `separator'. */
522 if (separator_ends_record)
524 char *match_end = match_start + match_length;
526 /* If this match of `separator' isn't at the end of the
527 file, print the record. */
528 if (first_time == 0 || match_end != past_end)
529 output (match_end, past_end);
530 past_end = match_end;
535 output (match_start, past_end);
536 past_end = match_start;
538 match_start -= match_length - 1;
543 /* Print the characters from START to PAST_END - 1.
544 If START is NULL, just flush the buffer. */
547 output (start, past_end)
551 static char buffer[WRITESIZE];
552 static int bytes_in_buffer = 0;
553 int bytes_to_add = past_end - start;
554 int bytes_available = WRITESIZE - bytes_in_buffer;
558 xwrite (1, buffer, bytes_in_buffer);
563 /* Write out as many full buffers as possible. */
564 while (bytes_to_add >= bytes_available)
566 bcopy (start, buffer + bytes_in_buffer, bytes_available);
567 bytes_to_add -= bytes_available;
568 start += bytes_available;
569 xwrite (1, buffer, WRITESIZE);
571 bytes_available = WRITESIZE;
574 bcopy (start, buffer + bytes_in_buffer, bytes_to_add);
575 bytes_in_buffer += bytes_to_add;
586 xwrite (desc, buffer, size)
591 if (write (desc, buffer, size) != size)
593 error (0, errno, "write error");
598 /* Allocate N bytes of memory dynamically, with error checking. */
609 error (0, 0, "virtual memory exhausted");
615 /* Change the size of memory area P to N bytes, with error checking. */
625 error (0, 0, "virtual memory exhausted");