1 /* grep.c - show lines matching regular expressions
2 *
3 * Copyright 2013 CE Strake <strake888 at gmail.com>
4 *
5 * See http://pubs.opengroup.org/onlinepubs/9699919799/utilities/grep.html
6 *
7 * Posix doesn't even specify -r, documenting deviations from it is silly.
8 * echo hello | grep -w ''
9 * echo '' | grep -w ''
10 * echo hello | grep -f </dev/null
11 *
12
13 USE_GREP(NEWTOY(grep, "(line-buffered)(color):;(exclude-dir)*S(exclude)*M(include)*ZzEFHIab(byte-offset)h(no-filename)ino(only-matching)rRsvwcl(files-with-matches)q(quiet)(silent)e*f*C#B#A#m#x[!wx][!EFw]", TOYFLAG_BIN|TOYFLAG_ARGFAIL(2)|TOYFLAG_LINEBUF))
14 USE_EGREP(OLDTOY(egrep, grep, TOYFLAG_BIN|TOYFLAG_ARGFAIL(2)|TOYFLAG_LINEBUF))
15 USE_FGREP(OLDTOY(fgrep, grep, TOYFLAG_BIN|TOYFLAG_ARGFAIL(2)|TOYFLAG_LINEBUF))
16
17 config GREP
18 bool "grep"
19 default y
20 help
21 usage: grep [-EFrivwcloqsHbhn] [-ABC NUM] [-m MAX] [-e REGEX]... [-MS PATTERN]... [-f REGFILE] [FILE]...
22
23 Show lines matching regular expressions. If no -e, first argument is
24 regular expression to match. With no files (or "-" filename) read stdin.
25 Returns 0 if matched, 1 if no match found, 2 for command errors.
26
27 -e Regex to match. (May be repeated.)
28 -f File listing regular expressions to match.
29
30 file search:
31 -r Recurse into subdirectories (defaults FILE to ".")
32 -R Recurse into subdirectories and symlinks to directories
33 -M Match filename pattern (--include)
34 -S Skip filename pattern (--exclude)
35 --exclude-dir=PATTERN Skip directory pattern
36 -I Ignore binary files
37
38 match type:
39 -A Show NUM lines after -B Show NUM lines before match
40 -C NUM lines context (A+B) -E extended regex syntax
41 -F fixed (literal match) -a always text (not binary)
42 -i case insensitive -m match MAX many lines
43 -v invert match -w whole word (implies -E)
44 -x whole line -z input NUL terminated
45
46 display modes: (default: matched line)
47 -c count of matching lines -l show only matching filenames
48 -o only matching part -q quiet (errors only)
49 -s silent (no error msg) -Z output NUL terminated
50
51 output prefix (default: filename if checking more than 1 file)
52 -H force filename -b byte offset of match
53 -h hide filename -n line number of match
54
55 config EGREP
56 bool
57 default y
58 depends on GREP
59
60 config FGREP
61 bool
62 default y
63 depends on GREP
64 */
65
66 #define FOR_grep
67 #include "toys.h"
68
69 GLOBALS(
70 long m, A, B, C;
71 struct arg_list *f, *e, *M, *S, *exclude_dir;
72 char *color;
73
74 char *purple, *cyan, *red, *green, *grey;
75 struct double_list *reg;
76 char indelim, outdelim;
77 int found, tried;
78 )
79
80 struct reg {
81 struct reg *next, *prev;
82 int rc;
83 regex_t r;
84 regmatch_t m;
85 };
86
numdash(long num,char dash)87 static void numdash(long num, char dash)
88 {
89 printf("%s%ld%s%c", TT.green, num, TT.cyan, dash);
90 }
91
92 // Emit line with various potential prefixes and delimiter
outline(char * line,char dash,char * name,long lcount,long bcount,unsigned trim)93 static void outline(char *line, char dash, char *name, long lcount, long bcount,
94 unsigned trim)
95 {
96 if (!trim && FLAG(o)) return;
97 if (name && FLAG(H)) printf("%s%s%s%c", TT.purple, name, TT.cyan, dash);
98 if (FLAG(c)) {
99 printf("%s%ld", TT.grey, lcount);
100 xputc(TT.outdelim);
101 } else if (lcount && FLAG(n)) numdash(lcount, dash);
102 if (bcount && FLAG(b)) numdash(bcount-1, dash);
103 if (line) {
104 if (FLAG(color)) xputsn(FLAG(o) ? TT.red : TT.grey);
105 // support embedded NUL bytes in output
106 xputsl(line, trim);
107 xputc(TT.outdelim);
108 }
109 }
110
111 // Show matches in one file
do_grep(int fd,char * name)112 static void do_grep(int fd, char *name)
113 {
114 long lcount = 0, mcount = 0, offset = 0, after = 0, before = 0;
115 struct double_list *dlb = 0;
116 char *bars = 0;
117 FILE *file;
118 int bin = 0;
119
120 if (!FLAG(r)) TT.tried++;
121 if (!fd) name = "(standard input)";
122
123 // Only run binary file check on lseekable files.
124 if (!FLAG(a) && !lseek(fd, 0, SEEK_CUR)) {
125 char buf[256];
126 int len, i = 0;
127 wchar_t wc;
128
129 // If the first 256 bytes don't parse as utf8, call it binary.
130 if (0<(len = read(fd, buf, 256))) {
131 lseek(fd, -len, SEEK_CUR);
132 while (i<len) {
133 bin = utf8towc(&wc, buf+i, len-i);
134 if (bin == -2) i = len;
135 if (bin<1) break;
136 i += bin;
137 }
138 bin = i!=len;
139 }
140 if (bin && FLAG(I)) return;
141 }
142
143 if (!(file = fdopen(fd, "r"))) return perror_msg("%s", name);
144
145 // Loop through lines of input
146 for (;;) {
147 char *line = 0, *start;
148 struct reg *shoe;
149 size_t ulen;
150 long len;
151 int matched = 0, rc = 1;
152
153 // get next line, check and trim delimiter
154 lcount++;
155 errno = 0;
156 ulen = len = getdelim(&line, &ulen, TT.indelim, file);
157 if (len == -1 && errno) perror_msg("%s", name);
158 if (len<1) break;
159 if (line[ulen-1] == TT.indelim) line[--ulen] = 0;
160
161 // Prepare for next line
162 start = line;
163 if (TT.reg) for (shoe = (void *)TT.reg; shoe; shoe = shoe->next)
164 shoe->rc = 0;
165
166 // Loop to handle multiple matches in same line
167 do {
168 regmatch_t *mm = (void *)toybuf;
169
170 // Handle "fixed" (literal) matches
171 if (FLAG(F)) {
172 struct arg_list *seek, fseek;
173 char *s = 0;
174
175 for (seek = TT.e; seek; seek = seek->next) {
176 if (FLAG(x)) {
177 if (!(FLAG(i) ? strcasecmp : strcmp)(seek->arg, line)) s = line;
178 } else if (!*seek->arg) {
179 // No need to set fseek.next because this will match every line.
180 seek = &fseek;
181 fseek.arg = s = line;
182 } else if (FLAG(i)) s = strcasestr(start, seek->arg);
183 else s = strstr(start, seek->arg);
184
185 if (s) break;
186 }
187
188 if (s) {
189 rc = 0;
190 mm->rm_so = (s-start);
191 mm->rm_eo = (s-start)+strlen(seek->arg);
192 } else rc = 1;
193
194 // Handle regex matches
195 } else {
196 int baseline = mm->rm_eo;
197
198 mm->rm_so = mm->rm_eo = INT_MAX;
199 rc = 1;
200 for (shoe = (void *)TT.reg; shoe; shoe = shoe->next) {
201
202 // Do we need to re-check this regex?
203 if (!shoe->rc) {
204 shoe->m.rm_so -= baseline;
205 shoe->m.rm_eo -= baseline;
206 if (!matched || shoe->m.rm_so<0)
207 shoe->rc = regexec0(&shoe->r, start, ulen-(start-line), 1,
208 &shoe->m, start==line ? 0 : REG_NOTBOL);
209 }
210
211 // If we got a match, is it a _better_ match?
212 if (!shoe->rc && (shoe->m.rm_so < mm->rm_so ||
213 (shoe->m.rm_so == mm->rm_so && shoe->m.rm_eo >= mm->rm_eo)))
214 {
215 mm = &shoe->m;
216 rc = 0;
217 }
218 }
219 }
220
221 if (!rc && FLAG(o) && !mm->rm_eo && ulen>start-line) {
222 start++;
223 continue;
224 }
225
226 if (!rc && FLAG(x) && (mm->rm_so || ulen-(start-line)!=mm->rm_eo)) rc = 1;
227
228 if (!rc && FLAG(w)) {
229 char c = 0;
230
231 if ((start+mm->rm_so)!=line) {
232 c = start[mm->rm_so-1];
233 if (!isalnum(c) && c != '_') c = 0;
234 }
235 if (!c) {
236 c = start[mm->rm_eo];
237 if (!isalnum(c) && c != '_') c = 0;
238 }
239 if (c) {
240 start += mm->rm_so+1;
241 continue;
242 }
243 }
244
245 if (FLAG(v)) {
246 if (FLAG(o)) {
247 if (rc) mm->rm_eo = ulen-(start-line);
248 else if (!mm->rm_so) {
249 start += mm->rm_eo;
250 continue;
251 } else mm->rm_eo = mm->rm_so;
252 } else {
253 if (!rc) break;
254 mm->rm_eo = ulen-(start-line);
255 }
256 mm->rm_so = 0;
257 } else if (rc) break;
258
259 // At least one line we didn't print since match while -ABC active
260 if (bars) {
261 xputs(bars);
262 bars = 0;
263 }
264 matched++;
265 TT.found = 1;
266 if (FLAG(q)) {
267 toys.exitval = 0;
268 xexit();
269 }
270 if (FLAG(l)) {
271 xprintf("%s%c", name, TT.outdelim);
272 free(line);
273 fclose(file);
274 return;
275 }
276
277 if (!FLAG(c)) {
278 long bcount = 1 + offset + (start-line) + (FLAG(o) ? mm->rm_so : 0);
279
280 if (bin) printf("Binary file %s matches\n", name);
281 else if (FLAG(o))
282 outline(start+mm->rm_so, ':', name, lcount, bcount,
283 mm->rm_eo-mm->rm_so);
284 else {
285 while (dlb) {
286 struct double_list *dl = dlist_pop(&dlb);
287 unsigned *uu = (void *)(dl->data+(strlen(dl->data)|3)+1);
288
289 outline(dl->data, '-', name, lcount-before, uu[0]+1, uu[1]);
290 free(dl->data);
291 free(dl);
292 before--;
293 }
294
295 if (matched==1)
296 outline(FLAG(color) ? 0 : line, ':', name, lcount, bcount, ulen);
297 if (FLAG(color)) {
298 xputsn(TT.grey);
299 if (mm->rm_so) xputsl(line, mm->rm_so);
300 xputsn(TT.red);
301 xputsl(line+mm->rm_so, mm->rm_eo-mm->rm_so);
302 }
303
304 if (TT.A) after = TT.A+1;
305 }
306 }
307
308 start += mm->rm_eo;
309 if (mm->rm_so == mm->rm_eo) break;
310 if (!FLAG(o) && FLAG(color)) break;
311 } while (*start);
312 offset += len;
313
314 if (matched) {
315 // Finish off pending line color fragment.
316 if (FLAG(color) && !FLAG(o)) {
317 xputsn(TT.grey);
318 if (ulen > start-line) xputsl(start, ulen-(start-line));
319 xputc(TT.outdelim);
320 }
321 mcount++;
322 } else {
323 int discard = (after || TT.B);
324
325 if (after && --after) {
326 outline(line, '-', name, lcount, 0, ulen);
327 discard = 0;
328 }
329 if (discard && TT.B) {
330 unsigned *uu, ul = (ulen|3)+1;
331
332 line = xrealloc(line, ul+8);
333 uu = (void *)(line+ul);
334 uu[0] = offset-len;
335 uu[1] = ulen;
336 dlist_add(&dlb, line);
337 line = 0;
338 if (++before>TT.B) {
339 struct double_list *dl;
340
341 dl = dlist_pop(&dlb);
342 free(dl->data);
343 free(dl);
344 before--;
345 } else discard = 0;
346 }
347 // If we discarded a line while displaying context, show bars before next
348 // line (but don't show them now in case that was last match in file)
349 if (discard && mcount) bars = "--";
350 }
351 free(line);
352
353 if (FLAG(m) && mcount >= TT.m) break;
354 }
355
356 if (FLAG(c)) outline(0, ':', name, mcount, 0, 1);
357
358 // loopfiles will also close the fd, but this frees an (opaque) struct.
359 fclose(file);
360 while (dlb) {
361 struct double_list *dl = dlist_pop(&dlb);
362
363 free(dl->data);
364 free(dl);
365 }
366 }
367
parse_regex(void)368 static void parse_regex(void)
369 {
370 struct arg_list *al, *new, *list = NULL;
371 char *s, *ss;
372
373 // Add all -f lines to -e list. (Yes, this is leaking allocation context for
374 // exit to free. Not supporting nofork for this command any time soon.)
375 al = TT.f ? TT.f : TT.e;
376 while (al) {
377 if (TT.f) {
378 if (!*(s = ss = xreadfile(al->arg, 0, 0))) {
379 free(ss);
380 s = 0;
381 }
382 } else s = ss = al->arg;
383
384 // Advance, when we run out of -f switch to -e.
385 al = al->next;
386 if (!al && TT.f) {
387 TT.f = 0;
388 al = TT.e;
389 }
390 if (!s) continue;
391
392 // Split lines at \n, add individual lines to new list.
393 do {
394 ss = FLAG(z) ? 0 : strchr(s, '\n');
395 if (ss) *(ss++) = 0;
396 new = xmalloc(sizeof(struct arg_list));
397 new->next = list;
398 new->arg = s;
399 list = new;
400 s = ss;
401 } while (ss && *s);
402 }
403 TT.e = list;
404
405 if (!FLAG(F)) {
406 // Convert regex list
407 for (al = TT.e; al; al = al->next) {
408 struct reg *shoe;
409
410 if (FLAG(o) && !*al->arg) continue;
411 dlist_add_nomalloc(&TT.reg, (void *)(shoe = xmalloc(sizeof(struct reg))));
412 xregcomp(&shoe->r, al->arg,
413 (REG_EXTENDED*!!FLAG(E))|(REG_ICASE*!!FLAG(i)));
414 }
415 dlist_terminate(TT.reg);
416 }
417 }
418
do_grep_r(struct dirtree * new)419 static int do_grep_r(struct dirtree *new)
420 {
421 struct arg_list *al;
422 char *name;
423
424 if (!new->parent) TT.tried++;
425 if (!dirtree_notdotdot(new)) return 0;
426 if (S_ISDIR(new->st.st_mode)) {
427 for (al = TT.exclude_dir; al; al = al->next)
428 if (!fnmatch(al->arg, new->name, 0)) return 0;
429 return DIRTREE_RECURSE|(FLAG(R)?DIRTREE_SYMFOLLOW:0);
430 }
431 if (TT.S || TT.M) {
432 for (al = TT.S; al; al = al->next)
433 if (!fnmatch(al->arg, new->name, 0)) return 0;
434
435 if (TT.M) {
436 for (al = TT.M; al; al = al->next)
437 if (!fnmatch(al->arg, new->name, 0)) break;
438
439 if (!al) return 0;
440 }
441 }
442
443 // "grep -r onefile" doesn't show filenames, but "grep -r onedir" should.
444 if (new->parent && !FLAG(h)) toys.optflags |= FLAG_H;
445
446 name = dirtree_path(new, 0);
447 do_grep(openat(dirtree_parentfd(new), new->name, 0), name);
448 free(name);
449
450 return 0;
451 }
452
grep_main(void)453 void grep_main(void)
454 {
455 char **ss = toys.optargs;
456
457 if (FLAG(color) && (!TT.color || !strcmp(TT.color, "auto")) && !isatty(1))
458 toys.optflags &= ~FLAG_color;
459
460 if (FLAG(color)) {
461 TT.purple = "\033[35m";
462 TT.cyan = "\033[36m";
463 TT.red = "\033[1;31m";
464 TT.green = "\033[32m";
465 TT.grey = "\033[0m";
466 } else TT.purple = TT.cyan = TT.red = TT.green = TT.grey = "";
467
468 if (FLAG(R)) toys.optflags |= FLAG_r;
469
470 // Grep exits with 2 for errors
471 toys.exitval = 2;
472
473 if (!TT.A) TT.A = TT.C;
474 if (!TT.B) TT.B = TT.C;
475
476 TT.indelim = '\n' * !FLAG(z);
477 TT.outdelim = '\n' * !FLAG(Z);
478
479 // Handle egrep and fgrep
480 if (*toys.which->name == 'e') toys.optflags |= FLAG_E;
481 if (*toys.which->name == 'f') toys.optflags |= FLAG_F;
482
483 if (!TT.e && !TT.f) {
484 if (!*ss) error_exit("no REGEX");
485 TT.e = xzalloc(sizeof(struct arg_list));
486 TT.e->arg = *(ss++);
487 toys.optc--;
488 }
489
490 parse_regex();
491
492 if (!FLAG(h) && toys.optc>1) toys.optflags |= FLAG_H;
493
494 if (FLAG(s)) {
495 close(2);
496 xopen_stdio("/dev/null", O_RDWR);
497 }
498
499 if (FLAG(r)) {
500 // Iterate through -r arguments. Use "." as default if none provided.
501 for (ss = *ss ? ss : (char *[]){".", 0}; *ss; ss++) {
502 if (!strcmp(*ss, "-")) do_grep(0, *ss);
503 else dirtree_read(*ss, do_grep_r);
504 }
505 } else loopfiles_rw(ss, O_RDONLY|WARN_ONLY, 0, do_grep);
506 if (TT.tried >= toys.optc || (FLAG(q)&&TT.found)) toys.exitval = !TT.found;
507 }
508