1 /* $NetBSD: util.c,v 1.9 2011/02/27 17:33:37 joerg Exp $ */
3 /* $OpenBSD: util.c,v 1.39 2010/07/02 22:18:03 tedu Exp $ */
6 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
8 * Copyright (c) 1999 James Howard and Dag-Erling Coïdan Smørgrav
9 * Copyright (C) 2008-2010 Gabor Kovesdan <gabor@FreeBSD.org>
10 * Copyright (C) 2017 Kyle Evans <kevans@FreeBSD.org>
11 * All rights reserved.
13 * Redistribution and use in source and binary forms, with or without
14 * modification, are permitted provided that the following conditions
16 * 1. Redistributions of source code must retain the above copyright
17 * notice, this list of conditions and the following disclaimer.
18 * 2. Redistributions in binary form must reproduce the above copyright
19 * notice, this list of conditions and the following disclaimer in the
20 * documentation and/or other materials provided with the distribution.
22 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 #include <sys/cdefs.h>
36 __FBSDID("$FreeBSD$");
39 #include <sys/types.h>
57 static bool first_match = true;
60 * Parsing context; used to hold things like matches made and
64 regmatch_t matches[MAX_MATCHES]; /* Matches made */
65 /* XXX TODO: This should be a chunk, not a line */
66 struct str ln; /* Current line */
67 size_t lnstart; /* Position in line */
68 size_t matchidx; /* Latest match index */
69 int printed; /* Metadata printed? */
70 bool binary; /* Binary file? */
74 * Match printing context
77 long long tail; /* Number of trailing lines to record */
78 int last_outed; /* Number of lines since last output */
79 bool doctx; /* Printing context? */
80 bool printmatch; /* Printing matches? */
81 bool same_file; /* Same file as previously printed? */
84 static void procmatch_match(struct mprintc *mc, struct parsec *pc);
85 static void procmatch_nomatch(struct mprintc *mc, struct parsec *pc);
86 static bool procmatches(struct mprintc *mc, struct parsec *pc, bool matched);
87 #ifdef WITH_INTERNAL_NOSPEC
88 static int litexec(const struct pat *pat, const char *string,
89 size_t nmatch, regmatch_t pmatch[]);
91 static int procline(struct parsec *pc);
92 static void printline(struct parsec *pc, int sep);
93 static void printline_metadata(struct str *line, int sep);
96 file_matching(const char *fname)
98 char *fname_base, *fname_buf;
101 ret = finclude ? false : true;
102 fname_buf = strdup(fname);
103 if (fname_buf == NULL)
105 fname_base = basename(fname_buf);
107 for (unsigned int i = 0; i < fpatterns; ++i) {
108 if (fnmatch(fpattern[i].pat, fname, 0) == 0 ||
109 fnmatch(fpattern[i].pat, fname_base, 0) == 0)
111 * The last pattern matched wins exclusion/inclusion
112 * rights, so we can't reasonably bail out early here.
114 ret = (fpattern[i].mode != EXCL_PAT);
121 dir_matching(const char *dname)
125 ret = dinclude ? false : true;
127 for (unsigned int i = 0; i < dpatterns; ++i) {
128 if (dname != NULL && fnmatch(dpattern[i].pat, dname, 0) == 0)
130 * The last pattern matched wins exclusion/inclusion
131 * rights, so we can't reasonably bail out early here.
133 ret = (dpattern[i].mode != EXCL_PAT);
139 * Processes a directory when a recursive search is performed with
140 * the -R option. Each appropriate file is passed to procfile().
143 grep_tree(char **argv)
149 const char *wd[] = { ".", NULL };
156 fts_flags = FTS_COMFOLLOW;
159 fts_flags = FTS_PHYSICAL;
162 fts_flags = FTS_LOGICAL;
165 fts_flags |= FTS_NOSTAT | FTS_NOCHDIR;
167 fts = fts_open((argv[0] == NULL) ?
168 __DECONST(char * const *, wd) : argv, fts_flags, NULL);
171 while ((p = fts_read(fts)) != NULL) {
172 switch (p->fts_info) {
178 warnx("%s: %s", p->fts_path, strerror(p->fts_errno));
183 if (dexclude || dinclude)
184 if (!dir_matching(p->fts_name) ||
185 !dir_matching(p->fts_path))
186 fts_set(fts, p, FTS_SKIP);
189 /* Print a warning for recursive directory loop */
190 warnx("warning: %s: recursive directory loop",
194 /* Check for file exclusion/inclusion */
196 if (fexclude || finclude)
197 ok &= file_matching(p->fts_path);
199 if (ok && procfile(p->fts_path))
210 procmatch_match(struct mprintc *mc, struct parsec *pc)
214 if (!first_match && (!mc->same_file || mc->last_outed > 0))
221 /* Print the matching line, but only if not quiet/binary */
222 if (mc->printmatch) {
224 while (pc->matchidx >= MAX_MATCHES) {
225 /* Reset matchidx and try again */
227 if (procline(pc) == 0)
233 mc->same_file = true;
239 procmatch_nomatch(struct mprintc *mc, struct parsec *pc)
242 /* Deal with any -A context as needed */
244 grep_printline(&pc->ln, '-');
248 } else if (Bflag == 0 || (Bflag > 0 && enqueue(&pc->ln)))
250 * Enqueue non-matching lines for -B context. If we're not
251 * actually doing -B context or if the enqueue resulted in a
252 * line being rotated out, then go ahead and increment
253 * last_outed to signify a gap between context/match.
259 * Process any matches in the current parsing context, return a boolean
260 * indicating whether we should halt any further processing or not. 'true' to
261 * continue processing, 'false' to halt.
264 procmatches(struct mprintc *mc, struct parsec *pc, bool matched)
268 * XXX TODO: This should loop over pc->matches and handle things on a
269 * line-by-line basis, setting up a `struct str` as needed.
271 /* Deal with any -B context or context separators */
273 procmatch_match(mc, pc);
275 /* Count the matches if we have a match limit */
277 /* XXX TODO: Decrement by number of matched lines */
279 if (mflag && mcount <= 0)
282 } else if (mc->doctx)
283 procmatch_nomatch(mc, pc);
289 * Opens a file and processes it. Each file is processed line-by-line
290 * passing the lines to procline().
293 procfile(const char *fn)
302 if (strcmp(fn, "-") == 0) {
303 fn = label != NULL ? label : errstr[1];
306 if (stat(fn, &sb) == 0) {
307 /* Check if we need to process the file */
308 s = sb.st_mode & S_IFMT;
309 if (dirbehave == DIR_SKIP && s == S_IFDIR)
311 if (devbehave == DEV_SKIP && (s == S_IFIFO ||
312 s == S_IFCHR || s == S_IFBLK || s == S_IFSOCK))
324 pc.ln.file = grep_strdup(fn);
329 pc.binary = f->binary;
330 memset(&mc, 0, sizeof(mc));
331 mc.printmatch = true;
332 if ((pc.binary && binbehave == BINFILE_BIN) || cflag || qflag ||
334 mc.printmatch = false;
335 if (mc.printmatch && (Aflag != 0 || Bflag != 0))
339 for (lines = 0; lines == 0 || !(lflag || qflag); ) {
341 * XXX TODO: We need to revisit this in a chunking world. We're
342 * not going to be doing per-line statistics because of the
343 * overhead involved. procmatches can figure that stuff out as
345 /* Reset per-line statistics */
350 pc.ln.off += pc.ln.len + 1;
351 /* XXX TODO: Grab a chunk */
352 if ((pc.ln.dat = grep_fgetln(f, &pc.ln.len)) == NULL ||
356 if (pc.ln.len > 0 && pc.ln.dat[pc.ln.len - 1] == fileeol)
360 /* Return if we need to skip a binary file */
361 if (pc.binary && binbehave == BINFILE_SKIP) {
368 if ((t = procline(&pc)) == 0)
371 /* Halt processing if we hit our match limit */
372 if (!procmatches(&mc, &pc, t == 0))
381 printf("%s:", pc.ln.file);
382 printf("%u\n", lines);
384 if (lflag && !qflag && lines != 0)
385 printf("%s%c", fn, nullflag ? 0 : '\n');
386 if (Lflag && !qflag && lines == 0)
387 printf("%s%c", fn, nullflag ? 0 : '\n');
388 if (lines != 0 && !cflag && !lflag && !Lflag &&
389 binbehave == BINFILE_BIN && f->binary && !qflag)
390 printf(errstr[7], fn);
397 #ifdef WITH_INTERNAL_NOSPEC
399 * Internal implementation of literal string search within a string, modeled
400 * after regexec(3), for use when the regex(3) implementation doesn't offer
401 * either REG_NOSPEC or REG_LITERAL. This does not apply in the default FreeBSD
402 * config, but in other scenarios such as building against libgnuregex or on
403 * some non-FreeBSD OSes.
406 litexec(const struct pat *pat, const char *string, size_t nmatch,
409 char *(*strstr_fn)(const char *, const char *);
412 size_t idx, n, ofs, stringlen;
414 if (cflags & REG_ICASE)
415 strstr_fn = strcasestr;
419 ofs = pmatch[0].rm_so;
420 stringlen = pmatch[0].rm_eo;
421 if (ofs >= stringlen)
422 return (REG_NOMATCH);
423 subject = strndup(string, stringlen);
426 for (n = 0; ofs < stringlen;) {
427 search = (subject + ofs);
428 if ((unsigned long)pat->len > strlen(search))
430 sub = strstr_fn(search, pat->pat);
432 * Ignoring the empty string possibility due to context: grep optimizes
433 * for empty patterns and will never reach this point.
438 /* Fill in pmatch if necessary */
440 pmatch[idx].rm_so = ofs + (sub - search);
441 pmatch[idx].rm_eo = pmatch[idx].rm_so + pat->len;
444 ofs = pmatch[idx].rm_so + 1;
446 /* We only needed to know if we match or not */
450 if (n > 0 && nmatch > 0)
451 for (n = idx; n < nmatch; ++n)
452 pmatch[n].rm_so = pmatch[n].rm_eo = -1;
454 return (n > 0 ? 0 : REG_NOMATCH);
456 #endif /* WITH_INTERNAL_NOSPEC */
458 #define iswword(x) (iswalnum((x)) || (x) == L'_')
461 * Processes a line comparing it with the specified patterns. Each pattern
462 * is looped to be compared along with the full string, saving each and every
463 * match, which is necessary to colorize the output and to count the
464 * matches. The matching lines are passed to printline() to display the
465 * appropriate output.
468 procline(struct parsec *pc)
470 regmatch_t pmatch, lastmatch, chkmatch;
471 wchar_t wbegin, wend;
474 int c = 0, r = 0, lastmatches = 0, leflags = eflags;
475 size_t startm = 0, matchidx;
478 matchidx = pc->matchidx;
480 /* Special case: empty pattern with -w flag, check first character */
481 if (matchall && wflag) {
485 if (sscanf(&pc->ln.dat[0], "%lc", &wend) != 1 || iswword(wend))
494 /* Initialize to avoid a false positive warning from GCC. */
495 lastmatch.rm_so = lastmatch.rm_eo = 0;
497 /* Loop to process the whole line */
498 while (st <= pc->ln.len) {
502 if (st > 0 && pc->ln.dat[st - 1] != fileeol)
503 leflags |= REG_NOTBOL;
504 /* Loop to compare with all the patterns */
505 for (i = 0; i < patterns; i++) {
507 pmatch.rm_eo = pc->ln.len;
508 #ifdef WITH_INTERNAL_NOSPEC
509 if (grepbehave == GREP_FIXED)
510 r = litexec(&pattern[i], pc->ln.dat, 1, &pmatch);
513 r = regexec(&r_pattern[i], pc->ln.dat, 1, &pmatch,
517 /* Check for full match */
518 if (xflag && (pmatch.rm_so != 0 ||
519 (size_t)pmatch.rm_eo != pc->ln.len))
521 /* Check for whole word match */
523 wbegin = wend = L' ';
524 if (pmatch.rm_so != 0 &&
525 sscanf(&pc->ln.dat[pmatch.rm_so - 1],
526 "%lc", &wbegin) != 1)
528 else if ((size_t)pmatch.rm_eo !=
530 sscanf(&pc->ln.dat[pmatch.rm_eo],
533 else if (iswword(wbegin) ||
537 * If we're doing whole word matching and we
538 * matched once, then we should try the pattern
539 * again after advancing just past the start of
540 * the earliest match. This allows the pattern
541 * to match later on in the line and possibly
542 * still match a whole word.
544 if (r == REG_NOMATCH &&
545 (retry == pc->lnstart ||
546 (unsigned int)pmatch.rm_so + 1 < retry))
547 retry = pmatch.rm_so + 1;
548 if (r == REG_NOMATCH)
558 * Replace previous match if the new one is earlier
559 * and/or longer. This will lead to some amount of
560 * extra work if -o/--color are specified, but it's
561 * worth it from a correctness point of view.
563 if (matchidx > startm) {
564 chkmatch = pc->matches[matchidx - 1];
565 if (pmatch.rm_so < chkmatch.rm_so ||
566 (pmatch.rm_so == chkmatch.rm_so &&
567 (pmatch.rm_eo - pmatch.rm_so) >
568 (chkmatch.rm_eo - chkmatch.rm_so))) {
569 pc->matches[matchidx - 1] = pmatch;
573 /* Advance as normal if not */
574 pc->matches[matchidx++] = pmatch;
577 /* avoid excessive matching - skip further patterns */
578 if ((color == NULL && !oflag) || qflag || lflag ||
579 matchidx >= MAX_MATCHES) {
587 * Advance to just past the start of the earliest match, try
588 * again just in case we still have a chance to match later in
591 if (lastmatches == 0 && retry > pc->lnstart) {
596 /* XXX TODO: We will need to keep going, since we're chunky */
597 /* One pass if we are not recording matches */
598 if (!wflag && ((color == NULL && !oflag) || qflag || lflag || Lflag))
601 /* If we didn't have any matches or REG_NOSUB set */
602 if (lastmatches == 0 || (cflags & REG_NOSUB))
605 if (lastmatches == 0)
608 else if (st == nst && lastmatch.rm_so == lastmatch.rm_eo)
609 /* Zero-length match -- advance one more so we don't get stuck */
612 /* Advance st based on previous matches */
617 /* Reflect the new matchidx in the context */
618 pc->matchidx = matchidx;
625 * Safe malloc() for internal use.
628 grep_malloc(size_t size)
632 if ((ptr = malloc(size)) == NULL)
638 * Safe calloc() for internal use.
641 grep_calloc(size_t nmemb, size_t size)
645 if ((ptr = calloc(nmemb, size)) == NULL)
651 * Safe realloc() for internal use.
654 grep_realloc(void *ptr, size_t size)
657 if ((ptr = realloc(ptr, size)) == NULL)
663 * Safe strdup() for internal use.
666 grep_strdup(const char *str)
670 if ((ret = strdup(str)) == NULL)
676 * Print an entire line as-is, there are no inline matches to consider. This is
677 * used for printing context.
679 void grep_printline(struct str *line, int sep) {
680 printline_metadata(line, sep);
681 fwrite(line->dat, line->len, 1, stdout);
686 printline_metadata(struct str *line, int sep)
693 fputs(line->file, stdout);
696 printf("%s", line->file);
703 printf("%d", line->line_no);
709 printf("%lld", (long long)(line->off + line->boff));
717 * Prints a matching line according to the command line options.
720 printline(struct parsec *pc, int sep)
726 /* If matchall, everything matches but don't actually print for -o */
727 if (oflag && matchall)
730 matchidx = pc->matchidx;
733 if ((oflag || color) && matchidx > 0) {
734 /* Only print metadata once per line if --color */
735 if (!oflag && pc->printed == 0)
736 printline_metadata(&pc->ln, sep);
737 for (i = 0; i < matchidx; i++) {
738 match = pc->matches[i];
739 /* Don't output zero length matches */
740 if (match.rm_so == match.rm_eo)
743 * Metadata is printed on a per-line basis, so every
744 * match gets file metadata with the -o flag.
747 pc->ln.boff = match.rm_so;
748 printline_metadata(&pc->ln, sep);
750 fwrite(pc->ln.dat + a, match.rm_so - a, 1,
753 fprintf(stdout, "\33[%sm\33[K", color);
754 fwrite(pc->ln.dat + match.rm_so,
755 match.rm_eo - match.rm_so, 1, stdout);
757 fprintf(stdout, "\33[m\33[K");
763 if (pc->ln.len - a > 0)
764 fwrite(pc->ln.dat + a, pc->ln.len - a, 1,
769 grep_printline(&pc->ln, sep);