]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * Builtin "git grep" | |
3 | * | |
4 | * Copyright (c) 2006 Junio C Hamano | |
5 | */ | |
6 | #include "cache.h" | |
7 | #include "blob.h" | |
8 | #include "tree.h" | |
9 | #include "commit.h" | |
10 | #include "tag.h" | |
11 | #include "tree-walk.h" | |
12 | #include "builtin.h" | |
13 | #include "parse-options.h" | |
14 | #include "userdiff.h" | |
15 | #include "grep.h" | |
16 | #include "quote.h" | |
17 | #include "dir.h" | |
18 | ||
19 | #ifndef NO_PTHREADS | |
20 | #include "thread-utils.h" | |
21 | #include <pthread.h> | |
22 | #endif | |
23 | ||
24 | static char const * const grep_usage[] = { | |
25 | "git grep [options] [-e] <pattern> [<rev>...] [[--] path...]", | |
26 | NULL | |
27 | }; | |
28 | ||
29 | static int use_threads = 1; | |
30 | ||
31 | #ifndef NO_PTHREADS | |
32 | #define THREADS 8 | |
33 | static pthread_t threads[THREADS]; | |
34 | ||
35 | static void *load_sha1(const unsigned char *sha1, unsigned long *size, | |
36 | const char *name); | |
37 | static void *load_file(const char *filename, size_t *sz); | |
38 | ||
39 | enum work_type {WORK_SHA1, WORK_FILE}; | |
40 | ||
41 | /* We use one producer thread and THREADS consumer | |
42 | * threads. The producer adds struct work_items to 'todo' and the | |
43 | * consumers pick work items from the same array. | |
44 | */ | |
45 | struct work_item | |
46 | { | |
47 | enum work_type type; | |
48 | char *name; | |
49 | ||
50 | /* if type == WORK_SHA1, then 'identifier' is a SHA1, | |
51 | * otherwise type == WORK_FILE, and 'identifier' is a NUL | |
52 | * terminated filename. | |
53 | */ | |
54 | void *identifier; | |
55 | char done; | |
56 | struct strbuf out; | |
57 | }; | |
58 | ||
59 | /* In the range [todo_done, todo_start) in 'todo' we have work_items | |
60 | * that have been or are processed by a consumer thread. We haven't | |
61 | * written the result for these to stdout yet. | |
62 | * | |
63 | * The work_items in [todo_start, todo_end) are waiting to be picked | |
64 | * up by a consumer thread. | |
65 | * | |
66 | * The ranges are modulo TODO_SIZE. | |
67 | */ | |
68 | #define TODO_SIZE 128 | |
69 | static struct work_item todo[TODO_SIZE]; | |
70 | static int todo_start; | |
71 | static int todo_end; | |
72 | static int todo_done; | |
73 | ||
74 | /* Has all work items been added? */ | |
75 | static int all_work_added; | |
76 | ||
77 | /* This lock protects all the variables above. */ | |
78 | static pthread_mutex_t grep_mutex; | |
79 | ||
80 | /* Used to serialize calls to read_sha1_file. */ | |
81 | static pthread_mutex_t read_sha1_mutex; | |
82 | ||
83 | #define grep_lock() pthread_mutex_lock(&grep_mutex) | |
84 | #define grep_unlock() pthread_mutex_unlock(&grep_mutex) | |
85 | #define read_sha1_lock() pthread_mutex_lock(&read_sha1_mutex) | |
86 | #define read_sha1_unlock() pthread_mutex_unlock(&read_sha1_mutex) | |
87 | ||
88 | /* Signalled when a new work_item is added to todo. */ | |
89 | static pthread_cond_t cond_add; | |
90 | ||
91 | /* Signalled when the result from one work_item is written to | |
92 | * stdout. | |
93 | */ | |
94 | static pthread_cond_t cond_write; | |
95 | ||
96 | /* Signalled when we are finished with everything. */ | |
97 | static pthread_cond_t cond_result; | |
98 | ||
99 | static void add_work(enum work_type type, char *name, void *id) | |
100 | { | |
101 | grep_lock(); | |
102 | ||
103 | while ((todo_end+1) % ARRAY_SIZE(todo) == todo_done) { | |
104 | pthread_cond_wait(&cond_write, &grep_mutex); | |
105 | } | |
106 | ||
107 | todo[todo_end].type = type; | |
108 | todo[todo_end].name = name; | |
109 | todo[todo_end].identifier = id; | |
110 | todo[todo_end].done = 0; | |
111 | strbuf_reset(&todo[todo_end].out); | |
112 | todo_end = (todo_end + 1) % ARRAY_SIZE(todo); | |
113 | ||
114 | pthread_cond_signal(&cond_add); | |
115 | grep_unlock(); | |
116 | } | |
117 | ||
118 | static struct work_item *get_work(void) | |
119 | { | |
120 | struct work_item *ret; | |
121 | ||
122 | grep_lock(); | |
123 | while (todo_start == todo_end && !all_work_added) { | |
124 | pthread_cond_wait(&cond_add, &grep_mutex); | |
125 | } | |
126 | ||
127 | if (todo_start == todo_end && all_work_added) { | |
128 | ret = NULL; | |
129 | } else { | |
130 | ret = &todo[todo_start]; | |
131 | todo_start = (todo_start + 1) % ARRAY_SIZE(todo); | |
132 | } | |
133 | grep_unlock(); | |
134 | return ret; | |
135 | } | |
136 | ||
137 | static void grep_sha1_async(struct grep_opt *opt, char *name, | |
138 | const unsigned char *sha1) | |
139 | { | |
140 | unsigned char *s; | |
141 | s = xmalloc(20); | |
142 | memcpy(s, sha1, 20); | |
143 | add_work(WORK_SHA1, name, s); | |
144 | } | |
145 | ||
146 | static void grep_file_async(struct grep_opt *opt, char *name, | |
147 | const char *filename) | |
148 | { | |
149 | add_work(WORK_FILE, name, xstrdup(filename)); | |
150 | } | |
151 | ||
152 | static void work_done(struct work_item *w) | |
153 | { | |
154 | int old_done; | |
155 | ||
156 | grep_lock(); | |
157 | w->done = 1; | |
158 | old_done = todo_done; | |
159 | for(; todo[todo_done].done && todo_done != todo_start; | |
160 | todo_done = (todo_done+1) % ARRAY_SIZE(todo)) { | |
161 | w = &todo[todo_done]; | |
162 | write_or_die(1, w->out.buf, w->out.len); | |
163 | free(w->name); | |
164 | free(w->identifier); | |
165 | } | |
166 | ||
167 | if (old_done != todo_done) | |
168 | pthread_cond_signal(&cond_write); | |
169 | ||
170 | if (all_work_added && todo_done == todo_end) | |
171 | pthread_cond_signal(&cond_result); | |
172 | ||
173 | grep_unlock(); | |
174 | } | |
175 | ||
176 | static void *run(void *arg) | |
177 | { | |
178 | int hit = 0; | |
179 | struct grep_opt *opt = arg; | |
180 | ||
181 | while (1) { | |
182 | struct work_item *w = get_work(); | |
183 | if (!w) | |
184 | break; | |
185 | ||
186 | opt->output_priv = w; | |
187 | if (w->type == WORK_SHA1) { | |
188 | unsigned long sz; | |
189 | void* data = load_sha1(w->identifier, &sz, w->name); | |
190 | ||
191 | if (data) { | |
192 | hit |= grep_buffer(opt, w->name, data, sz); | |
193 | free(data); | |
194 | } | |
195 | } else if (w->type == WORK_FILE) { | |
196 | size_t sz; | |
197 | void* data = load_file(w->identifier, &sz); | |
198 | if (data) { | |
199 | hit |= grep_buffer(opt, w->name, data, sz); | |
200 | free(data); | |
201 | } | |
202 | } else { | |
203 | assert(0); | |
204 | } | |
205 | ||
206 | work_done(w); | |
207 | } | |
208 | free_grep_patterns(arg); | |
209 | free(arg); | |
210 | ||
211 | return (void*) (intptr_t) hit; | |
212 | } | |
213 | ||
214 | static void strbuf_out(struct grep_opt *opt, const void *buf, size_t size) | |
215 | { | |
216 | struct work_item *w = opt->output_priv; | |
217 | strbuf_add(&w->out, buf, size); | |
218 | } | |
219 | ||
220 | static void start_threads(struct grep_opt *opt) | |
221 | { | |
222 | int i; | |
223 | ||
224 | pthread_mutex_init(&grep_mutex, NULL); | |
225 | pthread_mutex_init(&read_sha1_mutex, NULL); | |
226 | pthread_cond_init(&cond_add, NULL); | |
227 | pthread_cond_init(&cond_write, NULL); | |
228 | pthread_cond_init(&cond_result, NULL); | |
229 | ||
230 | for (i = 0; i < ARRAY_SIZE(todo); i++) { | |
231 | strbuf_init(&todo[i].out, 0); | |
232 | } | |
233 | ||
234 | for (i = 0; i < ARRAY_SIZE(threads); i++) { | |
235 | int err; | |
236 | struct grep_opt *o = grep_opt_dup(opt); | |
237 | o->output = strbuf_out; | |
238 | compile_grep_patterns(o); | |
239 | err = pthread_create(&threads[i], NULL, run, o); | |
240 | ||
241 | if (err) | |
242 | die("grep: failed to create thread: %s", | |
243 | strerror(err)); | |
244 | } | |
245 | } | |
246 | ||
247 | static int wait_all(void) | |
248 | { | |
249 | int hit = 0; | |
250 | int i; | |
251 | ||
252 | grep_lock(); | |
253 | all_work_added = 1; | |
254 | ||
255 | /* Wait until all work is done. */ | |
256 | while (todo_done != todo_end) | |
257 | pthread_cond_wait(&cond_result, &grep_mutex); | |
258 | ||
259 | /* Wake up all the consumer threads so they can see that there | |
260 | * is no more work to do. | |
261 | */ | |
262 | pthread_cond_broadcast(&cond_add); | |
263 | grep_unlock(); | |
264 | ||
265 | for (i = 0; i < ARRAY_SIZE(threads); i++) { | |
266 | void *h; | |
267 | pthread_join(threads[i], &h); | |
268 | hit |= (int) (intptr_t) h; | |
269 | } | |
270 | ||
271 | pthread_mutex_destroy(&grep_mutex); | |
272 | pthread_mutex_destroy(&read_sha1_mutex); | |
273 | pthread_cond_destroy(&cond_add); | |
274 | pthread_cond_destroy(&cond_write); | |
275 | pthread_cond_destroy(&cond_result); | |
276 | ||
277 | return hit; | |
278 | } | |
279 | #else /* !NO_PTHREADS */ | |
280 | #define read_sha1_lock() | |
281 | #define read_sha1_unlock() | |
282 | ||
283 | static int wait_all(void) | |
284 | { | |
285 | return 0; | |
286 | } | |
287 | #endif | |
288 | ||
289 | static int grep_config(const char *var, const char *value, void *cb) | |
290 | { | |
291 | struct grep_opt *opt = cb; | |
292 | ||
293 | switch (userdiff_config(var, value)) { | |
294 | case 0: break; | |
295 | case -1: return -1; | |
296 | default: return 0; | |
297 | } | |
298 | ||
299 | if (!strcmp(var, "color.grep")) { | |
300 | opt->color = git_config_colorbool(var, value, -1); | |
301 | return 0; | |
302 | } | |
303 | if (!strcmp(var, "color.grep.match")) { | |
304 | if (!value) | |
305 | return config_error_nonbool(var); | |
306 | color_parse(value, var, opt->color_match); | |
307 | return 0; | |
308 | } | |
309 | return git_color_default_config(var, value, cb); | |
310 | } | |
311 | ||
312 | /* | |
313 | * Return non-zero if max_depth is negative or path has no more then max_depth | |
314 | * slashes. | |
315 | */ | |
316 | static int accept_subdir(const char *path, int max_depth) | |
317 | { | |
318 | if (max_depth < 0) | |
319 | return 1; | |
320 | ||
321 | while ((path = strchr(path, '/')) != NULL) { | |
322 | max_depth--; | |
323 | if (max_depth < 0) | |
324 | return 0; | |
325 | path++; | |
326 | } | |
327 | return 1; | |
328 | } | |
329 | ||
330 | /* | |
331 | * Return non-zero if name is a subdirectory of match and is not too deep. | |
332 | */ | |
333 | static int is_subdir(const char *name, int namelen, | |
334 | const char *match, int matchlen, int max_depth) | |
335 | { | |
336 | if (matchlen > namelen || strncmp(name, match, matchlen)) | |
337 | return 0; | |
338 | ||
339 | if (name[matchlen] == '\0') /* exact match */ | |
340 | return 1; | |
341 | ||
342 | if (!matchlen || match[matchlen-1] == '/' || name[matchlen] == '/') | |
343 | return accept_subdir(name + matchlen + 1, max_depth); | |
344 | ||
345 | return 0; | |
346 | } | |
347 | ||
348 | /* | |
349 | * git grep pathspecs are somewhat different from diff-tree pathspecs; | |
350 | * pathname wildcards are allowed. | |
351 | */ | |
352 | static int pathspec_matches(const char **paths, const char *name, int max_depth) | |
353 | { | |
354 | int namelen, i; | |
355 | if (!paths || !*paths) | |
356 | return accept_subdir(name, max_depth); | |
357 | namelen = strlen(name); | |
358 | for (i = 0; paths[i]; i++) { | |
359 | const char *match = paths[i]; | |
360 | int matchlen = strlen(match); | |
361 | const char *cp, *meta; | |
362 | ||
363 | if (is_subdir(name, namelen, match, matchlen, max_depth)) | |
364 | return 1; | |
365 | if (!fnmatch(match, name, 0)) | |
366 | return 1; | |
367 | if (name[namelen-1] != '/') | |
368 | continue; | |
369 | ||
370 | /* We are being asked if the directory ("name") is worth | |
371 | * descending into. | |
372 | * | |
373 | * Find the longest leading directory name that does | |
374 | * not have metacharacter in the pathspec; the name | |
375 | * we are looking at must overlap with that directory. | |
376 | */ | |
377 | for (cp = match, meta = NULL; cp - match < matchlen; cp++) { | |
378 | char ch = *cp; | |
379 | if (ch == '*' || ch == '[' || ch == '?') { | |
380 | meta = cp; | |
381 | break; | |
382 | } | |
383 | } | |
384 | if (!meta) | |
385 | meta = cp; /* fully literal */ | |
386 | ||
387 | if (namelen <= meta - match) { | |
388 | /* Looking at "Documentation/" and | |
389 | * the pattern says "Documentation/howto/", or | |
390 | * "Documentation/diff*.txt". The name we | |
391 | * have should match prefix. | |
392 | */ | |
393 | if (!memcmp(match, name, namelen)) | |
394 | return 1; | |
395 | continue; | |
396 | } | |
397 | ||
398 | if (meta - match < namelen) { | |
399 | /* Looking at "Documentation/howto/" and | |
400 | * the pattern says "Documentation/h*"; | |
401 | * match up to "Do.../h"; this avoids descending | |
402 | * into "Documentation/technical/". | |
403 | */ | |
404 | if (!memcmp(match, name, meta - match)) | |
405 | return 1; | |
406 | continue; | |
407 | } | |
408 | } | |
409 | return 0; | |
410 | } | |
411 | ||
412 | static void *lock_and_read_sha1_file(const unsigned char *sha1, enum object_type *type, unsigned long *size) | |
413 | { | |
414 | void *data; | |
415 | ||
416 | if (use_threads) { | |
417 | read_sha1_lock(); | |
418 | data = read_sha1_file(sha1, type, size); | |
419 | read_sha1_unlock(); | |
420 | } else { | |
421 | data = read_sha1_file(sha1, type, size); | |
422 | } | |
423 | return data; | |
424 | } | |
425 | ||
426 | static void *load_sha1(const unsigned char *sha1, unsigned long *size, | |
427 | const char *name) | |
428 | { | |
429 | enum object_type type; | |
430 | void *data = lock_and_read_sha1_file(sha1, &type, size); | |
431 | ||
432 | if (!data) | |
433 | error("'%s': unable to read %s", name, sha1_to_hex(sha1)); | |
434 | ||
435 | return data; | |
436 | } | |
437 | ||
438 | static int grep_sha1(struct grep_opt *opt, const unsigned char *sha1, | |
439 | const char *filename, int tree_name_len) | |
440 | { | |
441 | struct strbuf pathbuf = STRBUF_INIT; | |
442 | char *name; | |
443 | ||
444 | if (opt->relative && opt->prefix_length) { | |
445 | quote_path_relative(filename + tree_name_len, -1, &pathbuf, | |
446 | opt->prefix); | |
447 | strbuf_insert(&pathbuf, 0, filename, tree_name_len); | |
448 | } else { | |
449 | strbuf_addstr(&pathbuf, filename); | |
450 | } | |
451 | ||
452 | name = strbuf_detach(&pathbuf, NULL); | |
453 | ||
454 | #ifndef NO_PTHREADS | |
455 | if (use_threads) { | |
456 | grep_sha1_async(opt, name, sha1); | |
457 | return 0; | |
458 | } else | |
459 | #endif | |
460 | { | |
461 | int hit; | |
462 | unsigned long sz; | |
463 | void *data = load_sha1(sha1, &sz, name); | |
464 | if (!data) | |
465 | hit = 0; | |
466 | else | |
467 | hit = grep_buffer(opt, name, data, sz); | |
468 | ||
469 | free(data); | |
470 | free(name); | |
471 | return hit; | |
472 | } | |
473 | } | |
474 | ||
475 | static void *load_file(const char *filename, size_t *sz) | |
476 | { | |
477 | struct stat st; | |
478 | char *data; | |
479 | int i; | |
480 | ||
481 | if (lstat(filename, &st) < 0) { | |
482 | err_ret: | |
483 | if (errno != ENOENT) | |
484 | error("'%s': %s", filename, strerror(errno)); | |
485 | return 0; | |
486 | } | |
487 | if (!S_ISREG(st.st_mode)) | |
488 | return 0; | |
489 | *sz = xsize_t(st.st_size); | |
490 | i = open(filename, O_RDONLY); | |
491 | if (i < 0) | |
492 | goto err_ret; | |
493 | data = xmalloc(*sz + 1); | |
494 | if (st.st_size != read_in_full(i, data, *sz)) { | |
495 | error("'%s': short read %s", filename, strerror(errno)); | |
496 | close(i); | |
497 | free(data); | |
498 | return 0; | |
499 | } | |
500 | close(i); | |
501 | data[*sz] = 0; | |
502 | return data; | |
503 | } | |
504 | ||
505 | static int grep_file(struct grep_opt *opt, const char *filename) | |
506 | { | |
507 | struct strbuf buf = STRBUF_INIT; | |
508 | char *name; | |
509 | ||
510 | if (opt->relative && opt->prefix_length) | |
511 | quote_path_relative(filename, -1, &buf, opt->prefix); | |
512 | else | |
513 | strbuf_addstr(&buf, filename); | |
514 | name = strbuf_detach(&buf, NULL); | |
515 | ||
516 | #ifndef NO_PTHREADS | |
517 | if (use_threads) { | |
518 | grep_file_async(opt, name, filename); | |
519 | return 0; | |
520 | } else | |
521 | #endif | |
522 | { | |
523 | int hit; | |
524 | size_t sz; | |
525 | void *data = load_file(filename, &sz); | |
526 | if (!data) | |
527 | hit = 0; | |
528 | else | |
529 | hit = grep_buffer(opt, name, data, sz); | |
530 | ||
531 | free(data); | |
532 | free(name); | |
533 | return hit; | |
534 | } | |
535 | } | |
536 | ||
537 | static int grep_cache(struct grep_opt *opt, const char **paths, int cached) | |
538 | { | |
539 | int hit = 0; | |
540 | int nr; | |
541 | read_cache(); | |
542 | ||
543 | for (nr = 0; nr < active_nr; nr++) { | |
544 | struct cache_entry *ce = active_cache[nr]; | |
545 | if (!S_ISREG(ce->ce_mode)) | |
546 | continue; | |
547 | if (!pathspec_matches(paths, ce->name, opt->max_depth)) | |
548 | continue; | |
549 | /* | |
550 | * If CE_VALID is on, we assume worktree file and its cache entry | |
551 | * are identical, even if worktree file has been modified, so use | |
552 | * cache version instead | |
553 | */ | |
554 | if (cached || (ce->ce_flags & CE_VALID) || ce_skip_worktree(ce)) { | |
555 | if (ce_stage(ce)) | |
556 | continue; | |
557 | hit |= grep_sha1(opt, ce->sha1, ce->name, 0); | |
558 | } | |
559 | else | |
560 | hit |= grep_file(opt, ce->name); | |
561 | if (ce_stage(ce)) { | |
562 | do { | |
563 | nr++; | |
564 | } while (nr < active_nr && | |
565 | !strcmp(ce->name, active_cache[nr]->name)); | |
566 | nr--; /* compensate for loop control */ | |
567 | } | |
568 | if (hit && opt->status_only) | |
569 | break; | |
570 | } | |
571 | free_grep_patterns(opt); | |
572 | return hit; | |
573 | } | |
574 | ||
575 | static int grep_tree(struct grep_opt *opt, const char **paths, | |
576 | struct tree_desc *tree, | |
577 | const char *tree_name, const char *base) | |
578 | { | |
579 | int len; | |
580 | int hit = 0; | |
581 | struct name_entry entry; | |
582 | char *down; | |
583 | int tn_len = strlen(tree_name); | |
584 | struct strbuf pathbuf; | |
585 | ||
586 | strbuf_init(&pathbuf, PATH_MAX + tn_len); | |
587 | ||
588 | if (tn_len) { | |
589 | strbuf_add(&pathbuf, tree_name, tn_len); | |
590 | strbuf_addch(&pathbuf, ':'); | |
591 | tn_len = pathbuf.len; | |
592 | } | |
593 | strbuf_addstr(&pathbuf, base); | |
594 | len = pathbuf.len; | |
595 | ||
596 | while (tree_entry(tree, &entry)) { | |
597 | int te_len = tree_entry_len(entry.path, entry.sha1); | |
598 | pathbuf.len = len; | |
599 | strbuf_add(&pathbuf, entry.path, te_len); | |
600 | ||
601 | if (S_ISDIR(entry.mode)) | |
602 | /* Match "abc/" against pathspec to | |
603 | * decide if we want to descend into "abc" | |
604 | * directory. | |
605 | */ | |
606 | strbuf_addch(&pathbuf, '/'); | |
607 | ||
608 | down = pathbuf.buf + tn_len; | |
609 | if (!pathspec_matches(paths, down, opt->max_depth)) | |
610 | ; | |
611 | else if (S_ISREG(entry.mode)) | |
612 | hit |= grep_sha1(opt, entry.sha1, pathbuf.buf, tn_len); | |
613 | else if (S_ISDIR(entry.mode)) { | |
614 | enum object_type type; | |
615 | struct tree_desc sub; | |
616 | void *data; | |
617 | unsigned long size; | |
618 | ||
619 | data = lock_and_read_sha1_file(entry.sha1, &type, &size); | |
620 | if (!data) | |
621 | die("unable to read tree (%s)", | |
622 | sha1_to_hex(entry.sha1)); | |
623 | init_tree_desc(&sub, data, size); | |
624 | hit |= grep_tree(opt, paths, &sub, tree_name, down); | |
625 | free(data); | |
626 | } | |
627 | if (hit && opt->status_only) | |
628 | break; | |
629 | } | |
630 | strbuf_release(&pathbuf); | |
631 | return hit; | |
632 | } | |
633 | ||
634 | static int grep_object(struct grep_opt *opt, const char **paths, | |
635 | struct object *obj, const char *name) | |
636 | { | |
637 | if (obj->type == OBJ_BLOB) | |
638 | return grep_sha1(opt, obj->sha1, name, 0); | |
639 | if (obj->type == OBJ_COMMIT || obj->type == OBJ_TREE) { | |
640 | struct tree_desc tree; | |
641 | void *data; | |
642 | unsigned long size; | |
643 | int hit; | |
644 | data = read_object_with_reference(obj->sha1, tree_type, | |
645 | &size, NULL); | |
646 | if (!data) | |
647 | die("unable to read tree (%s)", sha1_to_hex(obj->sha1)); | |
648 | init_tree_desc(&tree, data, size); | |
649 | hit = grep_tree(opt, paths, &tree, name, ""); | |
650 | free(data); | |
651 | return hit; | |
652 | } | |
653 | die("unable to grep from object of type %s", typename(obj->type)); | |
654 | } | |
655 | ||
656 | static int grep_directory(struct grep_opt *opt, const char **paths) | |
657 | { | |
658 | struct dir_struct dir; | |
659 | int i, hit = 0; | |
660 | ||
661 | memset(&dir, 0, sizeof(dir)); | |
662 | setup_standard_excludes(&dir); | |
663 | ||
664 | fill_directory(&dir, paths); | |
665 | for (i = 0; i < dir.nr; i++) { | |
666 | hit |= grep_file(opt, dir.entries[i]->name); | |
667 | if (hit && opt->status_only) | |
668 | break; | |
669 | } | |
670 | free_grep_patterns(opt); | |
671 | return hit; | |
672 | } | |
673 | ||
674 | static int context_callback(const struct option *opt, const char *arg, | |
675 | int unset) | |
676 | { | |
677 | struct grep_opt *grep_opt = opt->value; | |
678 | int value; | |
679 | const char *endp; | |
680 | ||
681 | if (unset) { | |
682 | grep_opt->pre_context = grep_opt->post_context = 0; | |
683 | return 0; | |
684 | } | |
685 | value = strtol(arg, (char **)&endp, 10); | |
686 | if (*endp) { | |
687 | return error("switch `%c' expects a numerical value", | |
688 | opt->short_name); | |
689 | } | |
690 | grep_opt->pre_context = grep_opt->post_context = value; | |
691 | return 0; | |
692 | } | |
693 | ||
694 | static int file_callback(const struct option *opt, const char *arg, int unset) | |
695 | { | |
696 | struct grep_opt *grep_opt = opt->value; | |
697 | FILE *patterns; | |
698 | int lno = 0; | |
699 | struct strbuf sb = STRBUF_INIT; | |
700 | ||
701 | patterns = fopen(arg, "r"); | |
702 | if (!patterns) | |
703 | die_errno("cannot open '%s'", arg); | |
704 | while (strbuf_getline(&sb, patterns, '\n') == 0) { | |
705 | /* ignore empty line like grep does */ | |
706 | if (sb.len == 0) | |
707 | continue; | |
708 | append_grep_pattern(grep_opt, strbuf_detach(&sb, NULL), arg, | |
709 | ++lno, GREP_PATTERN); | |
710 | } | |
711 | fclose(patterns); | |
712 | strbuf_release(&sb); | |
713 | return 0; | |
714 | } | |
715 | ||
716 | static int not_callback(const struct option *opt, const char *arg, int unset) | |
717 | { | |
718 | struct grep_opt *grep_opt = opt->value; | |
719 | append_grep_pattern(grep_opt, "--not", "command line", 0, GREP_NOT); | |
720 | return 0; | |
721 | } | |
722 | ||
723 | static int and_callback(const struct option *opt, const char *arg, int unset) | |
724 | { | |
725 | struct grep_opt *grep_opt = opt->value; | |
726 | append_grep_pattern(grep_opt, "--and", "command line", 0, GREP_AND); | |
727 | return 0; | |
728 | } | |
729 | ||
730 | static int open_callback(const struct option *opt, const char *arg, int unset) | |
731 | { | |
732 | struct grep_opt *grep_opt = opt->value; | |
733 | append_grep_pattern(grep_opt, "(", "command line", 0, GREP_OPEN_PAREN); | |
734 | return 0; | |
735 | } | |
736 | ||
737 | static int close_callback(const struct option *opt, const char *arg, int unset) | |
738 | { | |
739 | struct grep_opt *grep_opt = opt->value; | |
740 | append_grep_pattern(grep_opt, ")", "command line", 0, GREP_CLOSE_PAREN); | |
741 | return 0; | |
742 | } | |
743 | ||
744 | static int pattern_callback(const struct option *opt, const char *arg, | |
745 | int unset) | |
746 | { | |
747 | struct grep_opt *grep_opt = opt->value; | |
748 | append_grep_pattern(grep_opt, arg, "-e option", 0, GREP_PATTERN); | |
749 | return 0; | |
750 | } | |
751 | ||
752 | static int help_callback(const struct option *opt, const char *arg, int unset) | |
753 | { | |
754 | return -1; | |
755 | } | |
756 | ||
757 | int cmd_grep(int argc, const char **argv, const char *prefix) | |
758 | { | |
759 | int hit = 0; | |
760 | int cached = 0; | |
761 | int seen_dashdash = 0; | |
762 | int external_grep_allowed__ignored; | |
763 | struct grep_opt opt; | |
764 | struct object_array list = { 0, 0, NULL }; | |
765 | const char **paths = NULL; | |
766 | int i; | |
767 | int dummy; | |
768 | int nongit = 0, use_index = 1; | |
769 | struct option options[] = { | |
770 | OPT_BOOLEAN(0, "cached", &cached, | |
771 | "search in index instead of in the work tree"), | |
772 | OPT_BOOLEAN(0, "index", &use_index, | |
773 | "--no-index finds in contents not managed by git"), | |
774 | OPT_GROUP(""), | |
775 | OPT_BOOLEAN('v', "invert-match", &opt.invert, | |
776 | "show non-matching lines"), | |
777 | OPT_BOOLEAN('i', "ignore-case", &opt.ignore_case, | |
778 | "case insensitive matching"), | |
779 | OPT_BOOLEAN('w', "word-regexp", &opt.word_regexp, | |
780 | "match patterns only at word boundaries"), | |
781 | OPT_SET_INT('a', "text", &opt.binary, | |
782 | "process binary files as text", GREP_BINARY_TEXT), | |
783 | OPT_SET_INT('I', NULL, &opt.binary, | |
784 | "don't match patterns in binary files", | |
785 | GREP_BINARY_NOMATCH), | |
786 | { OPTION_INTEGER, 0, "max-depth", &opt.max_depth, "depth", | |
787 | "descend at most <depth> levels", PARSE_OPT_NONEG, | |
788 | NULL, 1 }, | |
789 | OPT_GROUP(""), | |
790 | OPT_BIT('E', "extended-regexp", &opt.regflags, | |
791 | "use extended POSIX regular expressions", REG_EXTENDED), | |
792 | OPT_NEGBIT('G', "basic-regexp", &opt.regflags, | |
793 | "use basic POSIX regular expressions (default)", | |
794 | REG_EXTENDED), | |
795 | OPT_BOOLEAN('F', "fixed-strings", &opt.fixed, | |
796 | "interpret patterns as fixed strings"), | |
797 | OPT_GROUP(""), | |
798 | OPT_BOOLEAN('n', NULL, &opt.linenum, "show line numbers"), | |
799 | OPT_NEGBIT('h', NULL, &opt.pathname, "don't show filenames", 1), | |
800 | OPT_BIT('H', NULL, &opt.pathname, "show filenames", 1), | |
801 | OPT_NEGBIT(0, "full-name", &opt.relative, | |
802 | "show filenames relative to top directory", 1), | |
803 | OPT_BOOLEAN('l', "files-with-matches", &opt.name_only, | |
804 | "show only filenames instead of matching lines"), | |
805 | OPT_BOOLEAN(0, "name-only", &opt.name_only, | |
806 | "synonym for --files-with-matches"), | |
807 | OPT_BOOLEAN('L', "files-without-match", | |
808 | &opt.unmatch_name_only, | |
809 | "show only the names of files without match"), | |
810 | OPT_BOOLEAN('z', "null", &opt.null_following_name, | |
811 | "print NUL after filenames"), | |
812 | OPT_BOOLEAN('c', "count", &opt.count, | |
813 | "show the number of matches instead of matching lines"), | |
814 | OPT_SET_INT(0, "color", &opt.color, "highlight matches", 1), | |
815 | OPT_GROUP(""), | |
816 | OPT_CALLBACK('C', NULL, &opt, "n", | |
817 | "show <n> context lines before and after matches", | |
818 | context_callback), | |
819 | OPT_INTEGER('B', NULL, &opt.pre_context, | |
820 | "show <n> context lines before matches"), | |
821 | OPT_INTEGER('A', NULL, &opt.post_context, | |
822 | "show <n> context lines after matches"), | |
823 | OPT_NUMBER_CALLBACK(&opt, "shortcut for -C NUM", | |
824 | context_callback), | |
825 | OPT_BOOLEAN('p', "show-function", &opt.funcname, | |
826 | "show a line with the function name before matches"), | |
827 | OPT_GROUP(""), | |
828 | OPT_CALLBACK('f', NULL, &opt, "file", | |
829 | "read patterns from file", file_callback), | |
830 | { OPTION_CALLBACK, 'e', NULL, &opt, "pattern", | |
831 | "match <pattern>", PARSE_OPT_NONEG, pattern_callback }, | |
832 | { OPTION_CALLBACK, 0, "and", &opt, NULL, | |
833 | "combine patterns specified with -e", | |
834 | PARSE_OPT_NOARG | PARSE_OPT_NONEG, and_callback }, | |
835 | OPT_BOOLEAN(0, "or", &dummy, ""), | |
836 | { OPTION_CALLBACK, 0, "not", &opt, NULL, "", | |
837 | PARSE_OPT_NOARG | PARSE_OPT_NONEG, not_callback }, | |
838 | { OPTION_CALLBACK, '(', NULL, &opt, NULL, "", | |
839 | PARSE_OPT_NOARG | PARSE_OPT_NONEG | PARSE_OPT_NODASH, | |
840 | open_callback }, | |
841 | { OPTION_CALLBACK, ')', NULL, &opt, NULL, "", | |
842 | PARSE_OPT_NOARG | PARSE_OPT_NONEG | PARSE_OPT_NODASH, | |
843 | close_callback }, | |
844 | OPT_BOOLEAN('q', "quiet", &opt.status_only, | |
845 | "indicate hit with exit status without output"), | |
846 | OPT_BOOLEAN(0, "all-match", &opt.all_match, | |
847 | "show only matches from files that match all patterns"), | |
848 | OPT_GROUP(""), | |
849 | OPT_BOOLEAN(0, "ext-grep", &external_grep_allowed__ignored, | |
850 | "allow calling of grep(1) (ignored by this build)"), | |
851 | { OPTION_CALLBACK, 0, "help-all", &options, NULL, "show usage", | |
852 | PARSE_OPT_HIDDEN | PARSE_OPT_NOARG, help_callback }, | |
853 | OPT_END() | |
854 | }; | |
855 | ||
856 | prefix = setup_git_directory_gently(&nongit); | |
857 | ||
858 | /* | |
859 | * 'git grep -h', unlike 'git grep -h <pattern>', is a request | |
860 | * to show usage information and exit. | |
861 | */ | |
862 | if (argc == 2 && !strcmp(argv[1], "-h")) | |
863 | usage_with_options(grep_usage, options); | |
864 | ||
865 | memset(&opt, 0, sizeof(opt)); | |
866 | opt.prefix = prefix; | |
867 | opt.prefix_length = (prefix && *prefix) ? strlen(prefix) : 0; | |
868 | opt.relative = 1; | |
869 | opt.pathname = 1; | |
870 | opt.pattern_tail = &opt.pattern_list; | |
871 | opt.regflags = REG_NEWLINE; | |
872 | opt.max_depth = -1; | |
873 | ||
874 | strcpy(opt.color_match, GIT_COLOR_BOLD_RED); | |
875 | opt.color = -1; | |
876 | git_config(grep_config, &opt); | |
877 | if (opt.color == -1) | |
878 | opt.color = git_use_color_default; | |
879 | ||
880 | /* | |
881 | * If there is no -- then the paths must exist in the working | |
882 | * tree. If there is no explicit pattern specified with -e or | |
883 | * -f, we take the first unrecognized non option to be the | |
884 | * pattern, but then what follows it must be zero or more | |
885 | * valid refs up to the -- (if exists), and then existing | |
886 | * paths. If there is an explicit pattern, then the first | |
887 | * unrecognized non option is the beginning of the refs list | |
888 | * that continues up to the -- (if exists), and then paths. | |
889 | */ | |
890 | argc = parse_options(argc, argv, prefix, options, grep_usage, | |
891 | PARSE_OPT_KEEP_DASHDASH | | |
892 | PARSE_OPT_STOP_AT_NON_OPTION | | |
893 | PARSE_OPT_NO_INTERNAL_HELP); | |
894 | ||
895 | if (use_index && nongit) | |
896 | /* die the same way as if we did it at the beginning */ | |
897 | setup_git_directory(); | |
898 | ||
899 | /* | |
900 | * skip a -- separator; we know it cannot be | |
901 | * separating revisions from pathnames if | |
902 | * we haven't even had any patterns yet | |
903 | */ | |
904 | if (argc > 0 && !opt.pattern_list && !strcmp(argv[0], "--")) { | |
905 | argv++; | |
906 | argc--; | |
907 | } | |
908 | ||
909 | /* First unrecognized non-option token */ | |
910 | if (argc > 0 && !opt.pattern_list) { | |
911 | append_grep_pattern(&opt, argv[0], "command line", 0, | |
912 | GREP_PATTERN); | |
913 | argv++; | |
914 | argc--; | |
915 | } | |
916 | ||
917 | if (!opt.pattern_list) | |
918 | die("no pattern given."); | |
919 | if (!opt.fixed && opt.ignore_case) | |
920 | opt.regflags |= REG_ICASE; | |
921 | if ((opt.regflags != REG_NEWLINE) && opt.fixed) | |
922 | die("cannot mix --fixed-strings and regexp"); | |
923 | ||
924 | #ifndef NO_PTHREADS | |
925 | if (online_cpus() == 1 || !grep_threads_ok(&opt)) | |
926 | use_threads = 0; | |
927 | ||
928 | if (use_threads) | |
929 | start_threads(&opt); | |
930 | #else | |
931 | use_threads = 0; | |
932 | #endif | |
933 | ||
934 | compile_grep_patterns(&opt); | |
935 | ||
936 | /* Check revs and then paths */ | |
937 | for (i = 0; i < argc; i++) { | |
938 | const char *arg = argv[i]; | |
939 | unsigned char sha1[20]; | |
940 | /* Is it a rev? */ | |
941 | if (!get_sha1(arg, sha1)) { | |
942 | struct object *object = parse_object(sha1); | |
943 | if (!object) | |
944 | die("bad object %s", arg); | |
945 | add_object_array(object, arg, &list); | |
946 | continue; | |
947 | } | |
948 | if (!strcmp(arg, "--")) { | |
949 | i++; | |
950 | seen_dashdash = 1; | |
951 | } | |
952 | break; | |
953 | } | |
954 | ||
955 | /* The rest are paths */ | |
956 | if (!seen_dashdash) { | |
957 | int j; | |
958 | for (j = i; j < argc; j++) | |
959 | verify_filename(prefix, argv[j]); | |
960 | } | |
961 | ||
962 | if (i < argc) | |
963 | paths = get_pathspec(prefix, argv + i); | |
964 | else if (prefix) { | |
965 | paths = xcalloc(2, sizeof(const char *)); | |
966 | paths[0] = prefix; | |
967 | paths[1] = NULL; | |
968 | } | |
969 | ||
970 | if (!use_index) { | |
971 | int hit; | |
972 | if (cached) | |
973 | die("--cached cannot be used with --no-index."); | |
974 | if (list.nr) | |
975 | die("--no-index cannot be used with revs."); | |
976 | hit = grep_directory(&opt, paths); | |
977 | if (use_threads) | |
978 | hit |= wait_all(); | |
979 | return !hit; | |
980 | } | |
981 | ||
982 | if (!list.nr) { | |
983 | int hit; | |
984 | if (!cached) | |
985 | setup_work_tree(); | |
986 | ||
987 | hit = grep_cache(&opt, paths, cached); | |
988 | if (use_threads) | |
989 | hit |= wait_all(); | |
990 | return !hit; | |
991 | } | |
992 | ||
993 | if (cached) | |
994 | die("both --cached and trees are given."); | |
995 | ||
996 | for (i = 0; i < list.nr; i++) { | |
997 | struct object *real_obj; | |
998 | real_obj = deref_tag(list.objects[i].item, NULL, 0); | |
999 | if (grep_object(&opt, paths, real_obj, list.objects[i].name)) { | |
1000 | hit = 1; | |
1001 | if (opt.status_only) | |
1002 | break; | |
1003 | } | |
1004 | } | |
1005 | ||
1006 | if (use_threads) | |
1007 | hit |= wait_all(); | |
1008 | free_grep_patterns(&opt); | |
1009 | return !hit; | |
1010 | } |