git-grep: don't use sscanf
[git/jrn.git] / sha1_name.c
blobbede0e5b0659db31a1dfdaab3d17627199150ea7
1 #include "cache.h"
2 #include "tag.h"
3 #include "commit.h"
4 #include "tree.h"
5 #include "blob.h"
6 #include "tree-walk.h"
7 #include "refs.h"
9 static int find_short_object_filename(int len, const char *name, unsigned char *sha1)
11 struct alternate_object_database *alt;
12 char hex[40];
13 int found = 0;
14 static struct alternate_object_database *fakeent;
16 if (!fakeent) {
17 const char *objdir = get_object_directory();
18 int objdir_len = strlen(objdir);
19 int entlen = objdir_len + 43;
20 fakeent = xmalloc(sizeof(*fakeent) + entlen);
21 memcpy(fakeent->base, objdir, objdir_len);
22 fakeent->name = fakeent->base + objdir_len + 1;
23 fakeent->name[-1] = '/';
25 fakeent->next = alt_odb_list;
27 sprintf(hex, "%.2s", name);
28 for (alt = fakeent; alt && found < 2; alt = alt->next) {
29 struct dirent *de;
30 DIR *dir;
31 sprintf(alt->name, "%.2s/", name);
32 dir = opendir(alt->base);
33 if (!dir)
34 continue;
35 while ((de = readdir(dir)) != NULL) {
36 if (strlen(de->d_name) != 38)
37 continue;
38 if (memcmp(de->d_name, name + 2, len - 2))
39 continue;
40 if (!found) {
41 memcpy(hex + 2, de->d_name, 38);
42 found++;
44 else if (memcmp(hex + 2, de->d_name, 38)) {
45 found = 2;
46 break;
49 closedir(dir);
51 if (found == 1)
52 return get_sha1_hex(hex, sha1) == 0;
53 return found;
56 static int match_sha(unsigned len, const unsigned char *a, const unsigned char *b)
58 do {
59 if (*a != *b)
60 return 0;
61 a++;
62 b++;
63 len -= 2;
64 } while (len > 1);
65 if (len)
66 if ((*a ^ *b) & 0xf0)
67 return 0;
68 return 1;
71 static int find_short_packed_object(int len, const unsigned char *match, unsigned char *sha1)
73 struct packed_git *p;
74 unsigned char found_sha1[20];
75 int found = 0;
77 prepare_packed_git();
78 for (p = packed_git; p && found < 2; p = p->next) {
79 uint32_t num = num_packed_objects(p);
80 uint32_t first = 0, last = num;
81 while (first < last) {
82 uint32_t mid = (first + last) / 2;
83 unsigned char now[20];
84 int cmp;
86 nth_packed_object_sha1(p, mid, now);
87 cmp = hashcmp(match, now);
88 if (!cmp) {
89 first = mid;
90 break;
92 if (cmp > 0) {
93 first = mid+1;
94 continue;
96 last = mid;
98 if (first < num) {
99 unsigned char now[20], next[20];
100 nth_packed_object_sha1(p, first, now);
101 if (match_sha(len, match, now)) {
102 if (nth_packed_object_sha1(p, first+1, next) ||
103 !match_sha(len, match, next)) {
104 /* unique within this pack */
105 if (!found) {
106 hashcpy(found_sha1, now);
107 found++;
109 else if (hashcmp(found_sha1, now)) {
110 found = 2;
111 break;
114 else {
115 /* not even unique within this pack */
116 found = 2;
117 break;
122 if (found == 1)
123 hashcpy(sha1, found_sha1);
124 return found;
127 #define SHORT_NAME_NOT_FOUND (-1)
128 #define SHORT_NAME_AMBIGUOUS (-2)
130 static int find_unique_short_object(int len, char *canonical,
131 unsigned char *res, unsigned char *sha1)
133 int has_unpacked, has_packed;
134 unsigned char unpacked_sha1[20], packed_sha1[20];
136 has_unpacked = find_short_object_filename(len, canonical, unpacked_sha1);
137 has_packed = find_short_packed_object(len, res, packed_sha1);
138 if (!has_unpacked && !has_packed)
139 return SHORT_NAME_NOT_FOUND;
140 if (1 < has_unpacked || 1 < has_packed)
141 return SHORT_NAME_AMBIGUOUS;
142 if (has_unpacked != has_packed) {
143 hashcpy(sha1, (has_packed ? packed_sha1 : unpacked_sha1));
144 return 0;
146 /* Both have unique ones -- do they match? */
147 if (hashcmp(packed_sha1, unpacked_sha1))
148 return SHORT_NAME_AMBIGUOUS;
149 hashcpy(sha1, packed_sha1);
150 return 0;
153 static int get_short_sha1(const char *name, int len, unsigned char *sha1,
154 int quietly)
156 int i, status;
157 char canonical[40];
158 unsigned char res[20];
160 if (len < MINIMUM_ABBREV || len > 40)
161 return -1;
162 hashclr(res);
163 memset(canonical, 'x', 40);
164 for (i = 0; i < len ;i++) {
165 unsigned char c = name[i];
166 unsigned char val;
167 if (c >= '0' && c <= '9')
168 val = c - '0';
169 else if (c >= 'a' && c <= 'f')
170 val = c - 'a' + 10;
171 else if (c >= 'A' && c <='F') {
172 val = c - 'A' + 10;
173 c -= 'A' - 'a';
175 else
176 return -1;
177 canonical[i] = c;
178 if (!(i & 1))
179 val <<= 4;
180 res[i >> 1] |= val;
183 status = find_unique_short_object(i, canonical, res, sha1);
184 if (!quietly && (status == SHORT_NAME_AMBIGUOUS))
185 return error("short SHA1 %.*s is ambiguous.", len, canonical);
186 return status;
189 const char *find_unique_abbrev(const unsigned char *sha1, int len)
191 int status, is_null;
192 static char hex[41];
194 is_null = is_null_sha1(sha1);
195 memcpy(hex, sha1_to_hex(sha1), 40);
196 if (len == 40 || !len)
197 return hex;
198 while (len < 40) {
199 unsigned char sha1_ret[20];
200 status = get_short_sha1(hex, len, sha1_ret, 1);
201 if (!status ||
202 (is_null && status != SHORT_NAME_AMBIGUOUS)) {
203 hex[len] = 0;
204 return hex;
206 if (status != SHORT_NAME_AMBIGUOUS)
207 return NULL;
208 len++;
210 return NULL;
213 static int ambiguous_path(const char *path, int len)
215 int slash = 1;
216 int cnt;
218 for (cnt = 0; cnt < len; cnt++) {
219 switch (*path++) {
220 case '\0':
221 break;
222 case '/':
223 if (slash)
224 break;
225 slash = 1;
226 continue;
227 case '.':
228 continue;
229 default:
230 slash = 0;
231 continue;
233 break;
235 return slash;
238 static const char *ref_fmt[] = {
239 "%.*s",
240 "refs/%.*s",
241 "refs/tags/%.*s",
242 "refs/heads/%.*s",
243 "refs/remotes/%.*s",
244 "refs/remotes/%.*s/HEAD",
245 NULL
248 int dwim_ref(const char *str, int len, unsigned char *sha1, char **ref)
250 const char **p, *r;
251 int refs_found = 0;
253 *ref = NULL;
254 for (p = ref_fmt; *p; p++) {
255 unsigned char sha1_from_ref[20];
256 unsigned char *this_result;
258 this_result = refs_found ? sha1_from_ref : sha1;
259 r = resolve_ref(mkpath(*p, len, str), this_result, 1, NULL);
260 if (r) {
261 if (!refs_found++)
262 *ref = xstrdup(r);
263 if (!warn_ambiguous_refs)
264 break;
267 return refs_found;
270 int dwim_log(const char *str, int len, unsigned char *sha1, char **log)
272 const char **p;
273 int logs_found = 0;
275 *log = NULL;
276 for (p = ref_fmt; *p; p++) {
277 struct stat st;
278 unsigned char hash[20];
279 char path[PATH_MAX];
280 const char *ref, *it;
282 strcpy(path, mkpath(*p, len, str));
283 ref = resolve_ref(path, hash, 0, NULL);
284 if (!ref)
285 continue;
286 if (!stat(git_path("logs/%s", path), &st) &&
287 S_ISREG(st.st_mode))
288 it = path;
289 else if (strcmp(ref, path) &&
290 !stat(git_path("logs/%s", ref), &st) &&
291 S_ISREG(st.st_mode))
292 it = ref;
293 else
294 continue;
295 if (!logs_found++) {
296 *log = xstrdup(it);
297 hashcpy(sha1, hash);
299 if (!warn_ambiguous_refs)
300 break;
302 return logs_found;
305 static int get_sha1_basic(const char *str, int len, unsigned char *sha1)
307 static const char *warning = "warning: refname '%.*s' is ambiguous.\n";
308 char *real_ref = NULL;
309 int refs_found = 0;
310 int at, reflog_len;
312 if (len == 40 && !get_sha1_hex(str, sha1))
313 return 0;
315 /* basic@{time or number} format to query ref-log */
316 reflog_len = at = 0;
317 if (str[len-1] == '}') {
318 for (at = 0; at < len - 1; at++) {
319 if (str[at] == '@' && str[at+1] == '{') {
320 reflog_len = (len-1) - (at+2);
321 len = at;
322 break;
327 /* Accept only unambiguous ref paths. */
328 if (len && ambiguous_path(str, len))
329 return -1;
331 if (!len && reflog_len) {
332 /* allow "@{...}" to mean the current branch reflog */
333 refs_found = dwim_ref("HEAD", 4, sha1, &real_ref);
334 } else if (reflog_len)
335 refs_found = dwim_log(str, len, sha1, &real_ref);
336 else
337 refs_found = dwim_ref(str, len, sha1, &real_ref);
339 if (!refs_found)
340 return -1;
342 if (warn_ambiguous_refs && refs_found > 1)
343 fprintf(stderr, warning, len, str);
345 if (reflog_len) {
346 int nth, i;
347 unsigned long at_time;
348 unsigned long co_time;
349 int co_tz, co_cnt;
351 /* Is it asking for N-th entry, or approxidate? */
352 for (i = nth = 0; 0 <= nth && i < reflog_len; i++) {
353 char ch = str[at+2+i];
354 if ('0' <= ch && ch <= '9')
355 nth = nth * 10 + ch - '0';
356 else
357 nth = -1;
359 if (0 <= nth)
360 at_time = 0;
361 else
362 at_time = approxidate(str + at + 2);
363 if (read_ref_at(real_ref, at_time, nth, sha1, NULL,
364 &co_time, &co_tz, &co_cnt)) {
365 if (at_time)
366 fprintf(stderr,
367 "warning: Log for '%.*s' only goes "
368 "back to %s.\n", len, str,
369 show_rfc2822_date(co_time, co_tz));
370 else
371 fprintf(stderr,
372 "warning: Log for '%.*s' only has "
373 "%d entries.\n", len, str, co_cnt);
377 free(real_ref);
378 return 0;
381 static int get_sha1_1(const char *name, int len, unsigned char *sha1);
383 static int get_parent(const char *name, int len,
384 unsigned char *result, int idx)
386 unsigned char sha1[20];
387 int ret = get_sha1_1(name, len, sha1);
388 struct commit *commit;
389 struct commit_list *p;
391 if (ret)
392 return ret;
393 commit = lookup_commit_reference(sha1);
394 if (!commit)
395 return -1;
396 if (parse_commit(commit))
397 return -1;
398 if (!idx) {
399 hashcpy(result, commit->object.sha1);
400 return 0;
402 p = commit->parents;
403 while (p) {
404 if (!--idx) {
405 hashcpy(result, p->item->object.sha1);
406 return 0;
408 p = p->next;
410 return -1;
413 static int get_nth_ancestor(const char *name, int len,
414 unsigned char *result, int generation)
416 unsigned char sha1[20];
417 int ret = get_sha1_1(name, len, sha1);
418 if (ret)
419 return ret;
421 while (generation--) {
422 struct commit *commit = lookup_commit_reference(sha1);
424 if (!commit || parse_commit(commit) || !commit->parents)
425 return -1;
426 hashcpy(sha1, commit->parents->item->object.sha1);
428 hashcpy(result, sha1);
429 return 0;
432 static int peel_onion(const char *name, int len, unsigned char *sha1)
434 unsigned char outer[20];
435 const char *sp;
436 unsigned int expected_type = 0;
437 struct object *o;
440 * "ref^{type}" dereferences ref repeatedly until you cannot
441 * dereference anymore, or you get an object of given type,
442 * whichever comes first. "ref^{}" means just dereference
443 * tags until you get a non-tag. "ref^0" is a shorthand for
444 * "ref^{commit}". "commit^{tree}" could be used to find the
445 * top-level tree of the given commit.
447 if (len < 4 || name[len-1] != '}')
448 return -1;
450 for (sp = name + len - 1; name <= sp; sp--) {
451 int ch = *sp;
452 if (ch == '{' && name < sp && sp[-1] == '^')
453 break;
455 if (sp <= name)
456 return -1;
458 sp++; /* beginning of type name, or closing brace for empty */
459 if (!strncmp(commit_type, sp, 6) && sp[6] == '}')
460 expected_type = OBJ_COMMIT;
461 else if (!strncmp(tree_type, sp, 4) && sp[4] == '}')
462 expected_type = OBJ_TREE;
463 else if (!strncmp(blob_type, sp, 4) && sp[4] == '}')
464 expected_type = OBJ_BLOB;
465 else if (sp[0] == '}')
466 expected_type = OBJ_NONE;
467 else
468 return -1;
470 if (get_sha1_1(name, sp - name - 2, outer))
471 return -1;
473 o = parse_object(outer);
474 if (!o)
475 return -1;
476 if (!expected_type) {
477 o = deref_tag(o, name, sp - name - 2);
478 if (!o || (!o->parsed && !parse_object(o->sha1)))
479 return -1;
480 hashcpy(sha1, o->sha1);
482 else {
483 /* At this point, the syntax look correct, so
484 * if we do not get the needed object, we should
485 * barf.
488 while (1) {
489 if (!o || (!o->parsed && !parse_object(o->sha1)))
490 return -1;
491 if (o->type == expected_type) {
492 hashcpy(sha1, o->sha1);
493 return 0;
495 if (o->type == OBJ_TAG)
496 o = ((struct tag*) o)->tagged;
497 else if (o->type == OBJ_COMMIT)
498 o = &(((struct commit *) o)->tree->object);
499 else
500 return error("%.*s: expected %s type, but the object dereferences to %s type",
501 len, name, typename(expected_type),
502 typename(o->type));
503 if (!o->parsed)
504 parse_object(o->sha1);
507 return 0;
510 static int get_describe_name(const char *name, int len, unsigned char *sha1)
512 const char *cp;
514 for (cp = name + len - 1; name + 2 <= cp; cp--) {
515 char ch = *cp;
516 if (hexval(ch) & ~0377) {
517 /* We must be looking at g in "SOMETHING-g"
518 * for it to be describe output.
520 if (ch == 'g' && cp[-1] == '-') {
521 cp++;
522 len -= cp - name;
523 return get_short_sha1(cp, len, sha1, 1);
527 return -1;
530 static int get_sha1_1(const char *name, int len, unsigned char *sha1)
532 int ret, has_suffix;
533 const char *cp;
535 /* "name~3" is "name^^^",
536 * "name~" and "name~0" are name -- not "name^0"!
537 * "name^" is not "name^0"; it is "name^1".
539 has_suffix = 0;
540 for (cp = name + len - 1; name <= cp; cp--) {
541 int ch = *cp;
542 if ('0' <= ch && ch <= '9')
543 continue;
544 if (ch == '~' || ch == '^')
545 has_suffix = ch;
546 break;
549 if (has_suffix) {
550 int num = 0;
551 int len1 = cp - name;
552 cp++;
553 while (cp < name + len)
554 num = num * 10 + *cp++ - '0';
555 if (has_suffix == '^') {
556 if (!num && len1 == len - 1)
557 num = 1;
558 return get_parent(name, len1, sha1, num);
560 /* else if (has_suffix == '~') -- goes without saying */
561 return get_nth_ancestor(name, len1, sha1, num);
564 ret = peel_onion(name, len, sha1);
565 if (!ret)
566 return 0;
568 ret = get_sha1_basic(name, len, sha1);
569 if (!ret)
570 return 0;
572 /* It could be describe output that is "SOMETHING-gXXXX" */
573 ret = get_describe_name(name, len, sha1);
574 if (!ret)
575 return 0;
577 return get_short_sha1(name, len, sha1, 0);
580 static int handle_one_ref(const char *path,
581 const unsigned char *sha1, int flag, void *cb_data)
583 struct commit_list **list = cb_data;
584 struct object *object = parse_object(sha1);
585 if (!object)
586 return 0;
587 if (object->type == OBJ_TAG)
588 object = deref_tag(object, path, strlen(path));
589 if (object->type != OBJ_COMMIT)
590 return 0;
591 insert_by_date((struct commit *)object, list);
592 return 0;
596 * This interprets names like ':/Initial revision of "git"' by searching
597 * through history and returning the first commit whose message starts
598 * with the given string.
600 * For future extension, ':/!' is reserved. If you want to match a message
601 * beginning with a '!', you have to repeat the exclamation mark.
604 #define ONELINE_SEEN (1u<<20)
605 static int get_sha1_oneline(const char *prefix, unsigned char *sha1)
607 struct commit_list *list = NULL, *backup = NULL, *l;
608 int retval = -1;
610 if (prefix[0] == '!') {
611 if (prefix[1] != '!')
612 die ("Invalid search pattern: %s", prefix);
613 prefix++;
615 if (!save_commit_buffer)
616 return error("Could not expand oneline-name.");
617 for_each_ref(handle_one_ref, &list);
618 for (l = list; l; l = l->next)
619 commit_list_insert(l->item, &backup);
620 while (list) {
621 char *p;
622 struct commit *commit;
624 commit = pop_most_recent_commit(&list, ONELINE_SEEN);
625 parse_object(commit->object.sha1);
626 if (!commit->buffer || !(p = strstr(commit->buffer, "\n\n")))
627 continue;
628 if (!prefixcmp(p + 2, prefix)) {
629 hashcpy(sha1, commit->object.sha1);
630 retval = 0;
631 break;
634 free_commit_list(list);
635 for (l = backup; l; l = l->next)
636 clear_commit_marks(l->item, ONELINE_SEEN);
637 return retval;
641 * This is like "get_sha1_basic()", except it allows "sha1 expressions",
642 * notably "xyz^" for "parent of xyz"
644 int get_sha1(const char *name, unsigned char *sha1)
646 int ret, bracket_depth;
647 unsigned unused;
648 int namelen = strlen(name);
649 const char *cp;
651 prepare_alt_odb();
652 ret = get_sha1_1(name, namelen, sha1);
653 if (!ret)
654 return ret;
655 /* sha1:path --> object name of path in ent sha1
656 * :path -> object name of path in index
657 * :[0-3]:path -> object name of path in index at stage
659 if (name[0] == ':') {
660 int stage = 0;
661 struct cache_entry *ce;
662 int pos;
663 if (namelen > 2 && name[1] == '/')
664 return get_sha1_oneline(name + 2, sha1);
665 if (namelen < 3 ||
666 name[2] != ':' ||
667 name[1] < '0' || '3' < name[1])
668 cp = name + 1;
669 else {
670 stage = name[1] - '0';
671 cp = name + 3;
673 namelen = namelen - (cp - name);
674 if (!active_cache)
675 read_cache();
676 if (active_nr < 0)
677 return -1;
678 pos = cache_name_pos(cp, namelen);
679 if (pos < 0)
680 pos = -pos - 1;
681 while (pos < active_nr) {
682 ce = active_cache[pos];
683 if (ce_namelen(ce) != namelen ||
684 memcmp(ce->name, cp, namelen))
685 break;
686 if (ce_stage(ce) == stage) {
687 hashcpy(sha1, ce->sha1);
688 return 0;
690 pos++;
692 return -1;
694 for (cp = name, bracket_depth = 0; *cp; cp++) {
695 if (*cp == '{')
696 bracket_depth++;
697 else if (bracket_depth && *cp == '}')
698 bracket_depth--;
699 else if (!bracket_depth && *cp == ':')
700 break;
702 if (*cp == ':') {
703 unsigned char tree_sha1[20];
704 if (!get_sha1_1(name, cp-name, tree_sha1))
705 return get_tree_entry(tree_sha1, cp+1, sha1,
706 &unused);
708 return ret;