]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * Builtin "git grep" | |
3 | * | |
4 | * Copyright (c) 2006 Junio C Hamano | |
5 | */ | |
6 | #include "cache.h" | |
7 | #include "repository.h" | |
8 | #include "config.h" | |
9 | #include "blob.h" | |
10 | #include "tree.h" | |
11 | #include "commit.h" | |
12 | #include "tag.h" | |
13 | #include "tree-walk.h" | |
14 | #include "builtin.h" | |
15 | #include "parse-options.h" | |
16 | #include "string-list.h" | |
17 | #include "run-command.h" | |
18 | #include "userdiff.h" | |
19 | #include "grep.h" | |
20 | #include "quote.h" | |
21 | #include "dir.h" | |
22 | #include "pathspec.h" | |
23 | #include "submodule.h" | |
24 | #include "submodule-config.h" | |
25 | ||
26 | static char const * const grep_usage[] = { | |
27 | N_("git grep [<options>] [-e] <pattern> [<rev>...] [[--] <path>...]"), | |
28 | NULL | |
29 | }; | |
30 | ||
31 | static int recurse_submodules; | |
32 | ||
33 | #define GREP_NUM_THREADS_DEFAULT 8 | |
34 | static int num_threads; | |
35 | ||
36 | #ifndef NO_PTHREADS | |
37 | static pthread_t *threads; | |
38 | ||
39 | /* We use one producer thread and THREADS consumer | |
40 | * threads. The producer adds struct work_items to 'todo' and the | |
41 | * consumers pick work items from the same array. | |
42 | */ | |
43 | struct work_item { | |
44 | struct grep_source source; | |
45 | char done; | |
46 | struct strbuf out; | |
47 | }; | |
48 | ||
49 | /* In the range [todo_done, todo_start) in 'todo' we have work_items | |
50 | * that have been or are processed by a consumer thread. We haven't | |
51 | * written the result for these to stdout yet. | |
52 | * | |
53 | * The work_items in [todo_start, todo_end) are waiting to be picked | |
54 | * up by a consumer thread. | |
55 | * | |
56 | * The ranges are modulo TODO_SIZE. | |
57 | */ | |
58 | #define TODO_SIZE 128 | |
59 | static struct work_item todo[TODO_SIZE]; | |
60 | static int todo_start; | |
61 | static int todo_end; | |
62 | static int todo_done; | |
63 | ||
64 | /* Has all work items been added? */ | |
65 | static int all_work_added; | |
66 | ||
67 | /* This lock protects all the variables above. */ | |
68 | static pthread_mutex_t grep_mutex; | |
69 | ||
70 | static inline void grep_lock(void) | |
71 | { | |
72 | assert(num_threads); | |
73 | pthread_mutex_lock(&grep_mutex); | |
74 | } | |
75 | ||
76 | static inline void grep_unlock(void) | |
77 | { | |
78 | assert(num_threads); | |
79 | pthread_mutex_unlock(&grep_mutex); | |
80 | } | |
81 | ||
82 | /* Signalled when a new work_item is added to todo. */ | |
83 | static pthread_cond_t cond_add; | |
84 | ||
85 | /* Signalled when the result from one work_item is written to | |
86 | * stdout. | |
87 | */ | |
88 | static pthread_cond_t cond_write; | |
89 | ||
90 | /* Signalled when we are finished with everything. */ | |
91 | static pthread_cond_t cond_result; | |
92 | ||
93 | static int skip_first_line; | |
94 | ||
95 | static void add_work(struct grep_opt *opt, const struct grep_source *gs) | |
96 | { | |
97 | grep_lock(); | |
98 | ||
99 | while ((todo_end+1) % ARRAY_SIZE(todo) == todo_done) { | |
100 | pthread_cond_wait(&cond_write, &grep_mutex); | |
101 | } | |
102 | ||
103 | todo[todo_end].source = *gs; | |
104 | if (opt->binary != GREP_BINARY_TEXT) | |
105 | grep_source_load_driver(&todo[todo_end].source); | |
106 | todo[todo_end].done = 0; | |
107 | strbuf_reset(&todo[todo_end].out); | |
108 | todo_end = (todo_end + 1) % ARRAY_SIZE(todo); | |
109 | ||
110 | pthread_cond_signal(&cond_add); | |
111 | grep_unlock(); | |
112 | } | |
113 | ||
114 | static struct work_item *get_work(void) | |
115 | { | |
116 | struct work_item *ret; | |
117 | ||
118 | grep_lock(); | |
119 | while (todo_start == todo_end && !all_work_added) { | |
120 | pthread_cond_wait(&cond_add, &grep_mutex); | |
121 | } | |
122 | ||
123 | if (todo_start == todo_end && all_work_added) { | |
124 | ret = NULL; | |
125 | } else { | |
126 | ret = &todo[todo_start]; | |
127 | todo_start = (todo_start + 1) % ARRAY_SIZE(todo); | |
128 | } | |
129 | grep_unlock(); | |
130 | return ret; | |
131 | } | |
132 | ||
133 | static void work_done(struct work_item *w) | |
134 | { | |
135 | int old_done; | |
136 | ||
137 | grep_lock(); | |
138 | w->done = 1; | |
139 | old_done = todo_done; | |
140 | for(; todo[todo_done].done && todo_done != todo_start; | |
141 | todo_done = (todo_done+1) % ARRAY_SIZE(todo)) { | |
142 | w = &todo[todo_done]; | |
143 | if (w->out.len) { | |
144 | const char *p = w->out.buf; | |
145 | size_t len = w->out.len; | |
146 | ||
147 | /* Skip the leading hunk mark of the first file. */ | |
148 | if (skip_first_line) { | |
149 | while (len) { | |
150 | len--; | |
151 | if (*p++ == '\n') | |
152 | break; | |
153 | } | |
154 | skip_first_line = 0; | |
155 | } | |
156 | ||
157 | write_or_die(1, p, len); | |
158 | } | |
159 | grep_source_clear(&w->source); | |
160 | } | |
161 | ||
162 | if (old_done != todo_done) | |
163 | pthread_cond_signal(&cond_write); | |
164 | ||
165 | if (all_work_added && todo_done == todo_end) | |
166 | pthread_cond_signal(&cond_result); | |
167 | ||
168 | grep_unlock(); | |
169 | } | |
170 | ||
171 | static void *run(void *arg) | |
172 | { | |
173 | int hit = 0; | |
174 | struct grep_opt *opt = arg; | |
175 | ||
176 | while (1) { | |
177 | struct work_item *w = get_work(); | |
178 | if (!w) | |
179 | break; | |
180 | ||
181 | opt->output_priv = w; | |
182 | hit |= grep_source(opt, &w->source); | |
183 | grep_source_clear_data(&w->source); | |
184 | work_done(w); | |
185 | } | |
186 | free_grep_patterns(arg); | |
187 | free(arg); | |
188 | ||
189 | return (void*) (intptr_t) hit; | |
190 | } | |
191 | ||
192 | static void strbuf_out(struct grep_opt *opt, const void *buf, size_t size) | |
193 | { | |
194 | struct work_item *w = opt->output_priv; | |
195 | strbuf_add(&w->out, buf, size); | |
196 | } | |
197 | ||
198 | static void start_threads(struct grep_opt *opt) | |
199 | { | |
200 | int i; | |
201 | ||
202 | pthread_mutex_init(&grep_mutex, NULL); | |
203 | pthread_mutex_init(&grep_read_mutex, NULL); | |
204 | pthread_mutex_init(&grep_attr_mutex, NULL); | |
205 | pthread_cond_init(&cond_add, NULL); | |
206 | pthread_cond_init(&cond_write, NULL); | |
207 | pthread_cond_init(&cond_result, NULL); | |
208 | grep_use_locks = 1; | |
209 | ||
210 | for (i = 0; i < ARRAY_SIZE(todo); i++) { | |
211 | strbuf_init(&todo[i].out, 0); | |
212 | } | |
213 | ||
214 | threads = xcalloc(num_threads, sizeof(*threads)); | |
215 | for (i = 0; i < num_threads; i++) { | |
216 | int err; | |
217 | struct grep_opt *o = grep_opt_dup(opt); | |
218 | o->output = strbuf_out; | |
219 | if (i) | |
220 | o->debug = 0; | |
221 | compile_grep_patterns(o); | |
222 | err = pthread_create(&threads[i], NULL, run, o); | |
223 | ||
224 | if (err) | |
225 | die(_("grep: failed to create thread: %s"), | |
226 | strerror(err)); | |
227 | } | |
228 | } | |
229 | ||
230 | static int wait_all(void) | |
231 | { | |
232 | int hit = 0; | |
233 | int i; | |
234 | ||
235 | grep_lock(); | |
236 | all_work_added = 1; | |
237 | ||
238 | /* Wait until all work is done. */ | |
239 | while (todo_done != todo_end) | |
240 | pthread_cond_wait(&cond_result, &grep_mutex); | |
241 | ||
242 | /* Wake up all the consumer threads so they can see that there | |
243 | * is no more work to do. | |
244 | */ | |
245 | pthread_cond_broadcast(&cond_add); | |
246 | grep_unlock(); | |
247 | ||
248 | for (i = 0; i < num_threads; i++) { | |
249 | void *h; | |
250 | pthread_join(threads[i], &h); | |
251 | hit |= (int) (intptr_t) h; | |
252 | } | |
253 | ||
254 | free(threads); | |
255 | ||
256 | pthread_mutex_destroy(&grep_mutex); | |
257 | pthread_mutex_destroy(&grep_read_mutex); | |
258 | pthread_mutex_destroy(&grep_attr_mutex); | |
259 | pthread_cond_destroy(&cond_add); | |
260 | pthread_cond_destroy(&cond_write); | |
261 | pthread_cond_destroy(&cond_result); | |
262 | grep_use_locks = 0; | |
263 | ||
264 | return hit; | |
265 | } | |
266 | #else /* !NO_PTHREADS */ | |
267 | ||
268 | static int wait_all(void) | |
269 | { | |
270 | return 0; | |
271 | } | |
272 | #endif | |
273 | ||
274 | static int grep_cmd_config(const char *var, const char *value, void *cb) | |
275 | { | |
276 | int st = grep_config(var, value, cb); | |
277 | if (git_color_default_config(var, value, cb) < 0) | |
278 | st = -1; | |
279 | ||
280 | if (!strcmp(var, "grep.threads")) { | |
281 | num_threads = git_config_int(var, value); | |
282 | if (num_threads < 0) | |
283 | die(_("invalid number of threads specified (%d) for %s"), | |
284 | num_threads, var); | |
285 | #ifdef NO_PTHREADS | |
286 | else if (num_threads && num_threads != 1) { | |
287 | /* | |
288 | * TRANSLATORS: %s is the configuration | |
289 | * variable for tweaking threads, currently | |
290 | * grep.threads | |
291 | */ | |
292 | warning(_("no threads support, ignoring %s"), var); | |
293 | num_threads = 0; | |
294 | } | |
295 | #endif | |
296 | } | |
297 | ||
298 | if (!strcmp(var, "submodule.recurse")) | |
299 | recurse_submodules = git_config_bool(var, value); | |
300 | ||
301 | return st; | |
302 | } | |
303 | ||
304 | static void *lock_and_read_oid_file(const struct object_id *oid, enum object_type *type, unsigned long *size) | |
305 | { | |
306 | void *data; | |
307 | ||
308 | grep_read_lock(); | |
309 | data = read_sha1_file(oid->hash, type, size); | |
310 | grep_read_unlock(); | |
311 | return data; | |
312 | } | |
313 | ||
314 | static int grep_oid(struct grep_opt *opt, const struct object_id *oid, | |
315 | const char *filename, int tree_name_len, | |
316 | const char *path) | |
317 | { | |
318 | struct strbuf pathbuf = STRBUF_INIT; | |
319 | struct grep_source gs; | |
320 | ||
321 | if (opt->relative && opt->prefix_length) { | |
322 | quote_path_relative(filename + tree_name_len, opt->prefix, &pathbuf); | |
323 | strbuf_insert(&pathbuf, 0, filename, tree_name_len); | |
324 | } else { | |
325 | strbuf_addstr(&pathbuf, filename); | |
326 | } | |
327 | ||
328 | grep_source_init(&gs, GREP_SOURCE_OID, pathbuf.buf, path, oid); | |
329 | strbuf_release(&pathbuf); | |
330 | ||
331 | #ifndef NO_PTHREADS | |
332 | if (num_threads) { | |
333 | /* | |
334 | * add_work() copies gs and thus assumes ownership of | |
335 | * its fields, so do not call grep_source_clear() | |
336 | */ | |
337 | add_work(opt, &gs); | |
338 | return 0; | |
339 | } else | |
340 | #endif | |
341 | { | |
342 | int hit; | |
343 | ||
344 | hit = grep_source(opt, &gs); | |
345 | ||
346 | grep_source_clear(&gs); | |
347 | return hit; | |
348 | } | |
349 | } | |
350 | ||
351 | static int grep_file(struct grep_opt *opt, const char *filename) | |
352 | { | |
353 | struct strbuf buf = STRBUF_INIT; | |
354 | struct grep_source gs; | |
355 | ||
356 | if (opt->relative && opt->prefix_length) | |
357 | quote_path_relative(filename, opt->prefix, &buf); | |
358 | else | |
359 | strbuf_addstr(&buf, filename); | |
360 | ||
361 | grep_source_init(&gs, GREP_SOURCE_FILE, buf.buf, filename, filename); | |
362 | strbuf_release(&buf); | |
363 | ||
364 | #ifndef NO_PTHREADS | |
365 | if (num_threads) { | |
366 | /* | |
367 | * add_work() copies gs and thus assumes ownership of | |
368 | * its fields, so do not call grep_source_clear() | |
369 | */ | |
370 | add_work(opt, &gs); | |
371 | return 0; | |
372 | } else | |
373 | #endif | |
374 | { | |
375 | int hit; | |
376 | ||
377 | hit = grep_source(opt, &gs); | |
378 | ||
379 | grep_source_clear(&gs); | |
380 | return hit; | |
381 | } | |
382 | } | |
383 | ||
384 | static void append_path(struct grep_opt *opt, const void *data, size_t len) | |
385 | { | |
386 | struct string_list *path_list = opt->output_priv; | |
387 | ||
388 | if (len == 1 && *(const char *)data == '\0') | |
389 | return; | |
390 | string_list_append(path_list, xstrndup(data, len)); | |
391 | } | |
392 | ||
393 | static void run_pager(struct grep_opt *opt, const char *prefix) | |
394 | { | |
395 | struct string_list *path_list = opt->output_priv; | |
396 | struct child_process child = CHILD_PROCESS_INIT; | |
397 | int i, status; | |
398 | ||
399 | for (i = 0; i < path_list->nr; i++) | |
400 | argv_array_push(&child.args, path_list->items[i].string); | |
401 | child.dir = prefix; | |
402 | child.use_shell = 1; | |
403 | ||
404 | status = run_command(&child); | |
405 | if (status) | |
406 | exit(status); | |
407 | } | |
408 | ||
409 | static int grep_cache(struct grep_opt *opt, struct repository *repo, | |
410 | const struct pathspec *pathspec, int cached); | |
411 | static int grep_tree(struct grep_opt *opt, const struct pathspec *pathspec, | |
412 | struct tree_desc *tree, struct strbuf *base, int tn_len, | |
413 | int check_attr, struct repository *repo); | |
414 | ||
415 | static int grep_submodule(struct grep_opt *opt, struct repository *superproject, | |
416 | const struct pathspec *pathspec, | |
417 | const struct object_id *oid, | |
418 | const char *filename, const char *path) | |
419 | { | |
420 | struct repository submodule; | |
421 | int hit; | |
422 | ||
423 | if (!is_submodule_active(superproject, path)) | |
424 | return 0; | |
425 | ||
426 | if (repo_submodule_init(&submodule, superproject, path)) | |
427 | return 0; | |
428 | ||
429 | repo_read_gitmodules(&submodule); | |
430 | ||
431 | /* | |
432 | * NEEDSWORK: This adds the submodule's object directory to the list of | |
433 | * alternates for the single in-memory object store. This has some bad | |
434 | * consequences for memory (processed objects will never be freed) and | |
435 | * performance (this increases the number of pack files git has to pay | |
436 | * attention to, to the sum of the number of pack files in all the | |
437 | * repositories processed so far). This can be removed once the object | |
438 | * store is no longer global and instead is a member of the repository | |
439 | * object. | |
440 | */ | |
441 | grep_read_lock(); | |
442 | add_to_alternates_memory(submodule.objectdir); | |
443 | grep_read_unlock(); | |
444 | ||
445 | if (oid) { | |
446 | struct object *object; | |
447 | struct tree_desc tree; | |
448 | void *data; | |
449 | unsigned long size; | |
450 | struct strbuf base = STRBUF_INIT; | |
451 | ||
452 | object = parse_object_or_die(oid, oid_to_hex(oid)); | |
453 | ||
454 | grep_read_lock(); | |
455 | data = read_object_with_reference(&object->oid, tree_type, | |
456 | &size, NULL); | |
457 | grep_read_unlock(); | |
458 | ||
459 | if (!data) | |
460 | die(_("unable to read tree (%s)"), oid_to_hex(&object->oid)); | |
461 | ||
462 | strbuf_addstr(&base, filename); | |
463 | strbuf_addch(&base, '/'); | |
464 | ||
465 | init_tree_desc(&tree, data, size); | |
466 | hit = grep_tree(opt, pathspec, &tree, &base, base.len, | |
467 | object->type == OBJ_COMMIT, &submodule); | |
468 | strbuf_release(&base); | |
469 | free(data); | |
470 | } else { | |
471 | hit = grep_cache(opt, &submodule, pathspec, 1); | |
472 | } | |
473 | ||
474 | repo_clear(&submodule); | |
475 | return hit; | |
476 | } | |
477 | ||
478 | static int grep_cache(struct grep_opt *opt, struct repository *repo, | |
479 | const struct pathspec *pathspec, int cached) | |
480 | { | |
481 | int hit = 0; | |
482 | int nr; | |
483 | struct strbuf name = STRBUF_INIT; | |
484 | int name_base_len = 0; | |
485 | if (repo->submodule_prefix) { | |
486 | name_base_len = strlen(repo->submodule_prefix); | |
487 | strbuf_addstr(&name, repo->submodule_prefix); | |
488 | } | |
489 | ||
490 | repo_read_index(repo); | |
491 | ||
492 | for (nr = 0; nr < repo->index->cache_nr; nr++) { | |
493 | const struct cache_entry *ce = repo->index->cache[nr]; | |
494 | strbuf_setlen(&name, name_base_len); | |
495 | strbuf_addstr(&name, ce->name); | |
496 | ||
497 | if (S_ISREG(ce->ce_mode) && | |
498 | match_pathspec(pathspec, name.buf, name.len, 0, NULL, | |
499 | S_ISDIR(ce->ce_mode) || | |
500 | S_ISGITLINK(ce->ce_mode))) { | |
501 | /* | |
502 | * If CE_VALID is on, we assume worktree file and its | |
503 | * cache entry are identical, even if worktree file has | |
504 | * been modified, so use cache version instead | |
505 | */ | |
506 | if (cached || (ce->ce_flags & CE_VALID) || | |
507 | ce_skip_worktree(ce)) { | |
508 | if (ce_stage(ce) || ce_intent_to_add(ce)) | |
509 | continue; | |
510 | hit |= grep_oid(opt, &ce->oid, name.buf, | |
511 | 0, name.buf); | |
512 | } else { | |
513 | hit |= grep_file(opt, name.buf); | |
514 | } | |
515 | } else if (recurse_submodules && S_ISGITLINK(ce->ce_mode) && | |
516 | submodule_path_match(pathspec, name.buf, NULL)) { | |
517 | hit |= grep_submodule(opt, repo, pathspec, NULL, ce->name, ce->name); | |
518 | } else { | |
519 | continue; | |
520 | } | |
521 | ||
522 | if (ce_stage(ce)) { | |
523 | do { | |
524 | nr++; | |
525 | } while (nr < repo->index->cache_nr && | |
526 | !strcmp(ce->name, repo->index->cache[nr]->name)); | |
527 | nr--; /* compensate for loop control */ | |
528 | } | |
529 | if (hit && opt->status_only) | |
530 | break; | |
531 | } | |
532 | ||
533 | strbuf_release(&name); | |
534 | return hit; | |
535 | } | |
536 | ||
537 | static int grep_tree(struct grep_opt *opt, const struct pathspec *pathspec, | |
538 | struct tree_desc *tree, struct strbuf *base, int tn_len, | |
539 | int check_attr, struct repository *repo) | |
540 | { | |
541 | int hit = 0; | |
542 | enum interesting match = entry_not_interesting; | |
543 | struct name_entry entry; | |
544 | int old_baselen = base->len; | |
545 | struct strbuf name = STRBUF_INIT; | |
546 | int name_base_len = 0; | |
547 | if (repo->submodule_prefix) { | |
548 | strbuf_addstr(&name, repo->submodule_prefix); | |
549 | name_base_len = name.len; | |
550 | } | |
551 | ||
552 | while (tree_entry(tree, &entry)) { | |
553 | int te_len = tree_entry_len(&entry); | |
554 | ||
555 | if (match != all_entries_interesting) { | |
556 | strbuf_addstr(&name, base->buf + tn_len); | |
557 | match = tree_entry_interesting(&entry, &name, | |
558 | 0, pathspec); | |
559 | strbuf_setlen(&name, name_base_len); | |
560 | ||
561 | if (match == all_entries_not_interesting) | |
562 | break; | |
563 | if (match == entry_not_interesting) | |
564 | continue; | |
565 | } | |
566 | ||
567 | strbuf_add(base, entry.path, te_len); | |
568 | ||
569 | if (S_ISREG(entry.mode)) { | |
570 | hit |= grep_oid(opt, entry.oid, base->buf, tn_len, | |
571 | check_attr ? base->buf + tn_len : NULL); | |
572 | } else if (S_ISDIR(entry.mode)) { | |
573 | enum object_type type; | |
574 | struct tree_desc sub; | |
575 | void *data; | |
576 | unsigned long size; | |
577 | ||
578 | data = lock_and_read_oid_file(entry.oid, &type, &size); | |
579 | if (!data) | |
580 | die(_("unable to read tree (%s)"), | |
581 | oid_to_hex(entry.oid)); | |
582 | ||
583 | strbuf_addch(base, '/'); | |
584 | init_tree_desc(&sub, data, size); | |
585 | hit |= grep_tree(opt, pathspec, &sub, base, tn_len, | |
586 | check_attr, repo); | |
587 | free(data); | |
588 | } else if (recurse_submodules && S_ISGITLINK(entry.mode)) { | |
589 | hit |= grep_submodule(opt, repo, pathspec, entry.oid, | |
590 | base->buf, base->buf + tn_len); | |
591 | } | |
592 | ||
593 | strbuf_setlen(base, old_baselen); | |
594 | ||
595 | if (hit && opt->status_only) | |
596 | break; | |
597 | } | |
598 | ||
599 | strbuf_release(&name); | |
600 | return hit; | |
601 | } | |
602 | ||
603 | static int grep_object(struct grep_opt *opt, const struct pathspec *pathspec, | |
604 | struct object *obj, const char *name, const char *path, | |
605 | struct repository *repo) | |
606 | { | |
607 | if (obj->type == OBJ_BLOB) | |
608 | return grep_oid(opt, &obj->oid, name, 0, path); | |
609 | if (obj->type == OBJ_COMMIT || obj->type == OBJ_TREE) { | |
610 | struct tree_desc tree; | |
611 | void *data; | |
612 | unsigned long size; | |
613 | struct strbuf base; | |
614 | int hit, len; | |
615 | ||
616 | grep_read_lock(); | |
617 | data = read_object_with_reference(&obj->oid, tree_type, | |
618 | &size, NULL); | |
619 | grep_read_unlock(); | |
620 | ||
621 | if (!data) | |
622 | die(_("unable to read tree (%s)"), oid_to_hex(&obj->oid)); | |
623 | ||
624 | len = name ? strlen(name) : 0; | |
625 | strbuf_init(&base, PATH_MAX + len + 1); | |
626 | if (len) { | |
627 | strbuf_add(&base, name, len); | |
628 | strbuf_addch(&base, ':'); | |
629 | } | |
630 | init_tree_desc(&tree, data, size); | |
631 | hit = grep_tree(opt, pathspec, &tree, &base, base.len, | |
632 | obj->type == OBJ_COMMIT, repo); | |
633 | strbuf_release(&base); | |
634 | free(data); | |
635 | return hit; | |
636 | } | |
637 | die(_("unable to grep from object of type %s"), type_name(obj->type)); | |
638 | } | |
639 | ||
640 | static int grep_objects(struct grep_opt *opt, const struct pathspec *pathspec, | |
641 | struct repository *repo, | |
642 | const struct object_array *list) | |
643 | { | |
644 | unsigned int i; | |
645 | int hit = 0; | |
646 | const unsigned int nr = list->nr; | |
647 | ||
648 | for (i = 0; i < nr; i++) { | |
649 | struct object *real_obj; | |
650 | real_obj = deref_tag(list->objects[i].item, NULL, 0); | |
651 | ||
652 | /* load the gitmodules file for this rev */ | |
653 | if (recurse_submodules) { | |
654 | submodule_free(); | |
655 | gitmodules_config_oid(&real_obj->oid); | |
656 | } | |
657 | if (grep_object(opt, pathspec, real_obj, list->objects[i].name, list->objects[i].path, | |
658 | repo)) { | |
659 | hit = 1; | |
660 | if (opt->status_only) | |
661 | break; | |
662 | } | |
663 | } | |
664 | return hit; | |
665 | } | |
666 | ||
667 | static int grep_directory(struct grep_opt *opt, const struct pathspec *pathspec, | |
668 | int exc_std, int use_index) | |
669 | { | |
670 | struct dir_struct dir; | |
671 | int i, hit = 0; | |
672 | ||
673 | memset(&dir, 0, sizeof(dir)); | |
674 | if (!use_index) | |
675 | dir.flags |= DIR_NO_GITLINKS; | |
676 | if (exc_std) | |
677 | setup_standard_excludes(&dir); | |
678 | ||
679 | fill_directory(&dir, &the_index, pathspec); | |
680 | for (i = 0; i < dir.nr; i++) { | |
681 | if (!dir_path_match(dir.entries[i], pathspec, 0, NULL)) | |
682 | continue; | |
683 | hit |= grep_file(opt, dir.entries[i]->name); | |
684 | if (hit && opt->status_only) | |
685 | break; | |
686 | } | |
687 | return hit; | |
688 | } | |
689 | ||
690 | static int context_callback(const struct option *opt, const char *arg, | |
691 | int unset) | |
692 | { | |
693 | struct grep_opt *grep_opt = opt->value; | |
694 | int value; | |
695 | const char *endp; | |
696 | ||
697 | if (unset) { | |
698 | grep_opt->pre_context = grep_opt->post_context = 0; | |
699 | return 0; | |
700 | } | |
701 | value = strtol(arg, (char **)&endp, 10); | |
702 | if (*endp) { | |
703 | return error(_("switch `%c' expects a numerical value"), | |
704 | opt->short_name); | |
705 | } | |
706 | grep_opt->pre_context = grep_opt->post_context = value; | |
707 | return 0; | |
708 | } | |
709 | ||
710 | static int file_callback(const struct option *opt, const char *arg, int unset) | |
711 | { | |
712 | struct grep_opt *grep_opt = opt->value; | |
713 | int from_stdin = !strcmp(arg, "-"); | |
714 | FILE *patterns; | |
715 | int lno = 0; | |
716 | struct strbuf sb = STRBUF_INIT; | |
717 | ||
718 | patterns = from_stdin ? stdin : fopen(arg, "r"); | |
719 | if (!patterns) | |
720 | die_errno(_("cannot open '%s'"), arg); | |
721 | while (strbuf_getline(&sb, patterns) == 0) { | |
722 | /* ignore empty line like grep does */ | |
723 | if (sb.len == 0) | |
724 | continue; | |
725 | ||
726 | append_grep_pat(grep_opt, sb.buf, sb.len, arg, ++lno, | |
727 | GREP_PATTERN); | |
728 | } | |
729 | if (!from_stdin) | |
730 | fclose(patterns); | |
731 | strbuf_release(&sb); | |
732 | return 0; | |
733 | } | |
734 | ||
735 | static int not_callback(const struct option *opt, const char *arg, int unset) | |
736 | { | |
737 | struct grep_opt *grep_opt = opt->value; | |
738 | append_grep_pattern(grep_opt, "--not", "command line", 0, GREP_NOT); | |
739 | return 0; | |
740 | } | |
741 | ||
742 | static int and_callback(const struct option *opt, const char *arg, int unset) | |
743 | { | |
744 | struct grep_opt *grep_opt = opt->value; | |
745 | append_grep_pattern(grep_opt, "--and", "command line", 0, GREP_AND); | |
746 | return 0; | |
747 | } | |
748 | ||
749 | static int open_callback(const struct option *opt, const char *arg, int unset) | |
750 | { | |
751 | struct grep_opt *grep_opt = opt->value; | |
752 | append_grep_pattern(grep_opt, "(", "command line", 0, GREP_OPEN_PAREN); | |
753 | return 0; | |
754 | } | |
755 | ||
756 | static int close_callback(const struct option *opt, const char *arg, int unset) | |
757 | { | |
758 | struct grep_opt *grep_opt = opt->value; | |
759 | append_grep_pattern(grep_opt, ")", "command line", 0, GREP_CLOSE_PAREN); | |
760 | return 0; | |
761 | } | |
762 | ||
763 | static int pattern_callback(const struct option *opt, const char *arg, | |
764 | int unset) | |
765 | { | |
766 | struct grep_opt *grep_opt = opt->value; | |
767 | append_grep_pattern(grep_opt, arg, "-e option", 0, GREP_PATTERN); | |
768 | return 0; | |
769 | } | |
770 | ||
771 | int cmd_grep(int argc, const char **argv, const char *prefix) | |
772 | { | |
773 | int hit = 0; | |
774 | int cached = 0, untracked = 0, opt_exclude = -1; | |
775 | int seen_dashdash = 0; | |
776 | int external_grep_allowed__ignored; | |
777 | const char *show_in_pager = NULL, *default_pager = "dummy"; | |
778 | struct grep_opt opt; | |
779 | struct object_array list = OBJECT_ARRAY_INIT; | |
780 | struct pathspec pathspec; | |
781 | struct string_list path_list = STRING_LIST_INIT_NODUP; | |
782 | int i; | |
783 | int dummy; | |
784 | int use_index = 1; | |
785 | int pattern_type_arg = GREP_PATTERN_TYPE_UNSPECIFIED; | |
786 | int allow_revs; | |
787 | ||
788 | struct option options[] = { | |
789 | OPT_BOOL(0, "cached", &cached, | |
790 | N_("search in index instead of in the work tree")), | |
791 | OPT_NEGBIT(0, "no-index", &use_index, | |
792 | N_("find in contents not managed by git"), 1), | |
793 | OPT_BOOL(0, "untracked", &untracked, | |
794 | N_("search in both tracked and untracked files")), | |
795 | OPT_SET_INT(0, "exclude-standard", &opt_exclude, | |
796 | N_("ignore files specified via '.gitignore'"), 1), | |
797 | OPT_BOOL(0, "recurse-submodules", &recurse_submodules, | |
798 | N_("recursively search in each submodule")), | |
799 | OPT_GROUP(""), | |
800 | OPT_BOOL('v', "invert-match", &opt.invert, | |
801 | N_("show non-matching lines")), | |
802 | OPT_BOOL('i', "ignore-case", &opt.ignore_case, | |
803 | N_("case insensitive matching")), | |
804 | OPT_BOOL('w', "word-regexp", &opt.word_regexp, | |
805 | N_("match patterns only at word boundaries")), | |
806 | OPT_SET_INT('a', "text", &opt.binary, | |
807 | N_("process binary files as text"), GREP_BINARY_TEXT), | |
808 | OPT_SET_INT('I', NULL, &opt.binary, | |
809 | N_("don't match patterns in binary files"), | |
810 | GREP_BINARY_NOMATCH), | |
811 | OPT_BOOL(0, "textconv", &opt.allow_textconv, | |
812 | N_("process binary files with textconv filters")), | |
813 | { OPTION_INTEGER, 0, "max-depth", &opt.max_depth, N_("depth"), | |
814 | N_("descend at most <depth> levels"), PARSE_OPT_NONEG, | |
815 | NULL, 1 }, | |
816 | OPT_GROUP(""), | |
817 | OPT_SET_INT('E', "extended-regexp", &pattern_type_arg, | |
818 | N_("use extended POSIX regular expressions"), | |
819 | GREP_PATTERN_TYPE_ERE), | |
820 | OPT_SET_INT('G', "basic-regexp", &pattern_type_arg, | |
821 | N_("use basic POSIX regular expressions (default)"), | |
822 | GREP_PATTERN_TYPE_BRE), | |
823 | OPT_SET_INT('F', "fixed-strings", &pattern_type_arg, | |
824 | N_("interpret patterns as fixed strings"), | |
825 | GREP_PATTERN_TYPE_FIXED), | |
826 | OPT_SET_INT('P', "perl-regexp", &pattern_type_arg, | |
827 | N_("use Perl-compatible regular expressions"), | |
828 | GREP_PATTERN_TYPE_PCRE), | |
829 | OPT_GROUP(""), | |
830 | OPT_BOOL('n', "line-number", &opt.linenum, N_("show line numbers")), | |
831 | OPT_NEGBIT('h', NULL, &opt.pathname, N_("don't show filenames"), 1), | |
832 | OPT_BIT('H', NULL, &opt.pathname, N_("show filenames"), 1), | |
833 | OPT_NEGBIT(0, "full-name", &opt.relative, | |
834 | N_("show filenames relative to top directory"), 1), | |
835 | OPT_BOOL('l', "files-with-matches", &opt.name_only, | |
836 | N_("show only filenames instead of matching lines")), | |
837 | OPT_BOOL(0, "name-only", &opt.name_only, | |
838 | N_("synonym for --files-with-matches")), | |
839 | OPT_BOOL('L', "files-without-match", | |
840 | &opt.unmatch_name_only, | |
841 | N_("show only the names of files without match")), | |
842 | OPT_BOOL('z', "null", &opt.null_following_name, | |
843 | N_("print NUL after filenames")), | |
844 | OPT_BOOL('c', "count", &opt.count, | |
845 | N_("show the number of matches instead of matching lines")), | |
846 | OPT__COLOR(&opt.color, N_("highlight matches")), | |
847 | OPT_BOOL(0, "break", &opt.file_break, | |
848 | N_("print empty line between matches from different files")), | |
849 | OPT_BOOL(0, "heading", &opt.heading, | |
850 | N_("show filename only once above matches from same file")), | |
851 | OPT_GROUP(""), | |
852 | OPT_CALLBACK('C', "context", &opt, N_("n"), | |
853 | N_("show <n> context lines before and after matches"), | |
854 | context_callback), | |
855 | OPT_INTEGER('B', "before-context", &opt.pre_context, | |
856 | N_("show <n> context lines before matches")), | |
857 | OPT_INTEGER('A', "after-context", &opt.post_context, | |
858 | N_("show <n> context lines after matches")), | |
859 | OPT_INTEGER(0, "threads", &num_threads, | |
860 | N_("use <n> worker threads")), | |
861 | OPT_NUMBER_CALLBACK(&opt, N_("shortcut for -C NUM"), | |
862 | context_callback), | |
863 | OPT_BOOL('p', "show-function", &opt.funcname, | |
864 | N_("show a line with the function name before matches")), | |
865 | OPT_BOOL('W', "function-context", &opt.funcbody, | |
866 | N_("show the surrounding function")), | |
867 | OPT_GROUP(""), | |
868 | OPT_CALLBACK('f', NULL, &opt, N_("file"), | |
869 | N_("read patterns from file"), file_callback), | |
870 | { OPTION_CALLBACK, 'e', NULL, &opt, N_("pattern"), | |
871 | N_("match <pattern>"), PARSE_OPT_NONEG, pattern_callback }, | |
872 | { OPTION_CALLBACK, 0, "and", &opt, NULL, | |
873 | N_("combine patterns specified with -e"), | |
874 | PARSE_OPT_NOARG | PARSE_OPT_NONEG, and_callback }, | |
875 | OPT_BOOL(0, "or", &dummy, ""), | |
876 | { OPTION_CALLBACK, 0, "not", &opt, NULL, "", | |
877 | PARSE_OPT_NOARG | PARSE_OPT_NONEG, not_callback }, | |
878 | { OPTION_CALLBACK, '(', NULL, &opt, NULL, "", | |
879 | PARSE_OPT_NOARG | PARSE_OPT_NONEG | PARSE_OPT_NODASH, | |
880 | open_callback }, | |
881 | { OPTION_CALLBACK, ')', NULL, &opt, NULL, "", | |
882 | PARSE_OPT_NOARG | PARSE_OPT_NONEG | PARSE_OPT_NODASH, | |
883 | close_callback }, | |
884 | OPT__QUIET(&opt.status_only, | |
885 | N_("indicate hit with exit status without output")), | |
886 | OPT_BOOL(0, "all-match", &opt.all_match, | |
887 | N_("show only matches from files that match all patterns")), | |
888 | { OPTION_SET_INT, 0, "debug", &opt.debug, NULL, | |
889 | N_("show parse tree for grep expression"), | |
890 | PARSE_OPT_NOARG | PARSE_OPT_HIDDEN, NULL, 1 }, | |
891 | OPT_GROUP(""), | |
892 | { OPTION_STRING, 'O', "open-files-in-pager", &show_in_pager, | |
893 | N_("pager"), N_("show matching files in the pager"), | |
894 | PARSE_OPT_OPTARG, NULL, (intptr_t)default_pager }, | |
895 | OPT_BOOL(0, "ext-grep", &external_grep_allowed__ignored, | |
896 | N_("allow calling of grep(1) (ignored by this build)")), | |
897 | OPT_END() | |
898 | }; | |
899 | ||
900 | init_grep_defaults(); | |
901 | git_config(grep_cmd_config, NULL); | |
902 | grep_init(&opt, prefix); | |
903 | ||
904 | /* | |
905 | * If there is no -- then the paths must exist in the working | |
906 | * tree. If there is no explicit pattern specified with -e or | |
907 | * -f, we take the first unrecognized non option to be the | |
908 | * pattern, but then what follows it must be zero or more | |
909 | * valid refs up to the -- (if exists), and then existing | |
910 | * paths. If there is an explicit pattern, then the first | |
911 | * unrecognized non option is the beginning of the refs list | |
912 | * that continues up to the -- (if exists), and then paths. | |
913 | */ | |
914 | argc = parse_options(argc, argv, prefix, options, grep_usage, | |
915 | PARSE_OPT_KEEP_DASHDASH | | |
916 | PARSE_OPT_STOP_AT_NON_OPTION); | |
917 | grep_commit_pattern_type(pattern_type_arg, &opt); | |
918 | ||
919 | if (use_index && !startup_info->have_repository) { | |
920 | int fallback = 0; | |
921 | git_config_get_bool("grep.fallbacktonoindex", &fallback); | |
922 | if (fallback) | |
923 | use_index = 0; | |
924 | else | |
925 | /* die the same way as if we did it at the beginning */ | |
926 | setup_git_directory(); | |
927 | } | |
928 | ||
929 | /* | |
930 | * skip a -- separator; we know it cannot be | |
931 | * separating revisions from pathnames if | |
932 | * we haven't even had any patterns yet | |
933 | */ | |
934 | if (argc > 0 && !opt.pattern_list && !strcmp(argv[0], "--")) { | |
935 | argv++; | |
936 | argc--; | |
937 | } | |
938 | ||
939 | /* First unrecognized non-option token */ | |
940 | if (argc > 0 && !opt.pattern_list) { | |
941 | append_grep_pattern(&opt, argv[0], "command line", 0, | |
942 | GREP_PATTERN); | |
943 | argv++; | |
944 | argc--; | |
945 | } | |
946 | ||
947 | if (show_in_pager == default_pager) | |
948 | show_in_pager = git_pager(1); | |
949 | if (show_in_pager) { | |
950 | opt.color = 0; | |
951 | opt.name_only = 1; | |
952 | opt.null_following_name = 1; | |
953 | opt.output_priv = &path_list; | |
954 | opt.output = append_path; | |
955 | string_list_append(&path_list, show_in_pager); | |
956 | } | |
957 | ||
958 | if (!opt.pattern_list) | |
959 | die(_("no pattern given.")); | |
960 | ||
961 | /* | |
962 | * We have to find "--" in a separate pass, because its presence | |
963 | * influences how we will parse arguments that come before it. | |
964 | */ | |
965 | for (i = 0; i < argc; i++) { | |
966 | if (!strcmp(argv[i], "--")) { | |
967 | seen_dashdash = 1; | |
968 | break; | |
969 | } | |
970 | } | |
971 | ||
972 | /* | |
973 | * Resolve any rev arguments. If we have a dashdash, then everything up | |
974 | * to it must resolve as a rev. If not, then we stop at the first | |
975 | * non-rev and assume everything else is a path. | |
976 | */ | |
977 | allow_revs = use_index && !untracked; | |
978 | for (i = 0; i < argc; i++) { | |
979 | const char *arg = argv[i]; | |
980 | struct object_id oid; | |
981 | struct object_context oc; | |
982 | struct object *object; | |
983 | ||
984 | if (!strcmp(arg, "--")) { | |
985 | i++; | |
986 | break; | |
987 | } | |
988 | ||
989 | if (!allow_revs) { | |
990 | if (seen_dashdash) | |
991 | die(_("--no-index or --untracked cannot be used with revs")); | |
992 | break; | |
993 | } | |
994 | ||
995 | if (get_oid_with_context(arg, GET_OID_RECORD_PATH, | |
996 | &oid, &oc)) { | |
997 | if (seen_dashdash) | |
998 | die(_("unable to resolve revision: %s"), arg); | |
999 | break; | |
1000 | } | |
1001 | ||
1002 | object = parse_object_or_die(&oid, arg); | |
1003 | if (!seen_dashdash) | |
1004 | verify_non_filename(prefix, arg); | |
1005 | add_object_array_with_path(object, arg, &list, oc.mode, oc.path); | |
1006 | free(oc.path); | |
1007 | } | |
1008 | ||
1009 | /* | |
1010 | * Anything left over is presumed to be a path. But in the non-dashdash | |
1011 | * "do what I mean" case, we verify and complain when that isn't true. | |
1012 | */ | |
1013 | if (!seen_dashdash) { | |
1014 | int j; | |
1015 | for (j = i; j < argc; j++) | |
1016 | verify_filename(prefix, argv[j], j == i && allow_revs); | |
1017 | } | |
1018 | ||
1019 | parse_pathspec(&pathspec, 0, | |
1020 | PATHSPEC_PREFER_CWD | | |
1021 | (opt.max_depth != -1 ? PATHSPEC_MAXDEPTH_VALID : 0), | |
1022 | prefix, argv + i); | |
1023 | pathspec.max_depth = opt.max_depth; | |
1024 | pathspec.recursive = 1; | |
1025 | pathspec.recurse_submodules = !!recurse_submodules; | |
1026 | ||
1027 | #ifndef NO_PTHREADS | |
1028 | if (list.nr || cached || show_in_pager) | |
1029 | num_threads = 0; | |
1030 | else if (num_threads == 0) | |
1031 | num_threads = GREP_NUM_THREADS_DEFAULT; | |
1032 | else if (num_threads < 0) | |
1033 | die(_("invalid number of threads specified (%d)"), num_threads); | |
1034 | if (num_threads == 1) | |
1035 | num_threads = 0; | |
1036 | #else | |
1037 | if (num_threads) | |
1038 | warning(_("no threads support, ignoring --threads")); | |
1039 | num_threads = 0; | |
1040 | #endif | |
1041 | ||
1042 | if (!num_threads) | |
1043 | /* | |
1044 | * The compiled patterns on the main path are only | |
1045 | * used when not using threading. Otherwise | |
1046 | * start_threads() below calls compile_grep_patterns() | |
1047 | * for each thread. | |
1048 | */ | |
1049 | compile_grep_patterns(&opt); | |
1050 | ||
1051 | #ifndef NO_PTHREADS | |
1052 | if (num_threads) { | |
1053 | if (!(opt.name_only || opt.unmatch_name_only || opt.count) | |
1054 | && (opt.pre_context || opt.post_context || | |
1055 | opt.file_break || opt.funcbody)) | |
1056 | skip_first_line = 1; | |
1057 | start_threads(&opt); | |
1058 | } | |
1059 | #endif | |
1060 | ||
1061 | if (show_in_pager && (cached || list.nr)) | |
1062 | die(_("--open-files-in-pager only works on the worktree")); | |
1063 | ||
1064 | if (show_in_pager && opt.pattern_list && !opt.pattern_list->next) { | |
1065 | const char *pager = path_list.items[0].string; | |
1066 | int len = strlen(pager); | |
1067 | ||
1068 | if (len > 4 && is_dir_sep(pager[len - 5])) | |
1069 | pager += len - 4; | |
1070 | ||
1071 | if (opt.ignore_case && !strcmp("less", pager)) | |
1072 | string_list_append(&path_list, "-I"); | |
1073 | ||
1074 | if (!strcmp("less", pager) || !strcmp("vi", pager)) { | |
1075 | struct strbuf buf = STRBUF_INIT; | |
1076 | strbuf_addf(&buf, "+/%s%s", | |
1077 | strcmp("less", pager) ? "" : "*", | |
1078 | opt.pattern_list->pattern); | |
1079 | string_list_append(&path_list, buf.buf); | |
1080 | strbuf_detach(&buf, NULL); | |
1081 | } | |
1082 | } | |
1083 | ||
1084 | if (recurse_submodules && (!use_index || untracked)) | |
1085 | die(_("option not supported with --recurse-submodules.")); | |
1086 | ||
1087 | if (!show_in_pager && !opt.status_only) | |
1088 | setup_pager(); | |
1089 | ||
1090 | if (!use_index && (untracked || cached)) | |
1091 | die(_("--cached or --untracked cannot be used with --no-index.")); | |
1092 | ||
1093 | if (!use_index || untracked) { | |
1094 | int use_exclude = (opt_exclude < 0) ? use_index : !!opt_exclude; | |
1095 | hit = grep_directory(&opt, &pathspec, use_exclude, use_index); | |
1096 | } else if (0 <= opt_exclude) { | |
1097 | die(_("--[no-]exclude-standard cannot be used for tracked contents.")); | |
1098 | } else if (!list.nr) { | |
1099 | if (!cached) | |
1100 | setup_work_tree(); | |
1101 | ||
1102 | hit = grep_cache(&opt, the_repository, &pathspec, cached); | |
1103 | } else { | |
1104 | if (cached) | |
1105 | die(_("both --cached and trees are given.")); | |
1106 | ||
1107 | hit = grep_objects(&opt, &pathspec, the_repository, &list); | |
1108 | } | |
1109 | ||
1110 | if (num_threads) | |
1111 | hit |= wait_all(); | |
1112 | if (hit && show_in_pager) | |
1113 | run_pager(&opt, prefix); | |
1114 | clear_pathspec(&pathspec); | |
1115 | free_grep_patterns(&opt); | |
1116 | return !hit; | |
1117 | } |