interpret_nth_last_branch(): avoid traversing the reflog twice
[git/dscho.git] / sha1_name.c
blobd6972f2d6a2fd5af38b9babbacd97c9388962bb4
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 const unsigned char *found_sha1 = NULL;
75 int found = 0;
77 prepare_packed_git();
78 for (p = packed_git; p && found < 2; p = p->next) {
79 uint32_t num, last;
80 uint32_t first = 0;
81 open_pack_index(p);
82 num = p->num_objects;
83 last = num;
84 while (first < last) {
85 uint32_t mid = (first + last) / 2;
86 const unsigned char *now;
87 int cmp;
89 now = nth_packed_object_sha1(p, mid);
90 cmp = hashcmp(match, now);
91 if (!cmp) {
92 first = mid;
93 break;
95 if (cmp > 0) {
96 first = mid+1;
97 continue;
99 last = mid;
101 if (first < num) {
102 const unsigned char *now, *next;
103 now = nth_packed_object_sha1(p, first);
104 if (match_sha(len, match, now)) {
105 next = nth_packed_object_sha1(p, first+1);
106 if (!next|| !match_sha(len, match, next)) {
107 /* unique within this pack */
108 if (!found) {
109 found_sha1 = now;
110 found++;
112 else if (hashcmp(found_sha1, now)) {
113 found = 2;
114 break;
117 else {
118 /* not even unique within this pack */
119 found = 2;
120 break;
125 if (found == 1)
126 hashcpy(sha1, found_sha1);
127 return found;
130 #define SHORT_NAME_NOT_FOUND (-1)
131 #define SHORT_NAME_AMBIGUOUS (-2)
133 static int find_unique_short_object(int len, char *canonical,
134 unsigned char *res, unsigned char *sha1)
136 int has_unpacked, has_packed;
137 unsigned char unpacked_sha1[20], packed_sha1[20];
139 prepare_alt_odb();
140 has_unpacked = find_short_object_filename(len, canonical, unpacked_sha1);
141 has_packed = find_short_packed_object(len, res, packed_sha1);
142 if (!has_unpacked && !has_packed)
143 return SHORT_NAME_NOT_FOUND;
144 if (1 < has_unpacked || 1 < has_packed)
145 return SHORT_NAME_AMBIGUOUS;
146 if (has_unpacked != has_packed) {
147 hashcpy(sha1, (has_packed ? packed_sha1 : unpacked_sha1));
148 return 0;
150 /* Both have unique ones -- do they match? */
151 if (hashcmp(packed_sha1, unpacked_sha1))
152 return SHORT_NAME_AMBIGUOUS;
153 hashcpy(sha1, packed_sha1);
154 return 0;
157 static int get_short_sha1(const char *name, int len, unsigned char *sha1,
158 int quietly)
160 int i, status;
161 char canonical[40];
162 unsigned char res[20];
164 if (len < MINIMUM_ABBREV || len > 40)
165 return -1;
166 hashclr(res);
167 memset(canonical, 'x', 40);
168 for (i = 0; i < len ;i++) {
169 unsigned char c = name[i];
170 unsigned char val;
171 if (c >= '0' && c <= '9')
172 val = c - '0';
173 else if (c >= 'a' && c <= 'f')
174 val = c - 'a' + 10;
175 else if (c >= 'A' && c <='F') {
176 val = c - 'A' + 10;
177 c -= 'A' - 'a';
179 else
180 return -1;
181 canonical[i] = c;
182 if (!(i & 1))
183 val <<= 4;
184 res[i >> 1] |= val;
187 status = find_unique_short_object(i, canonical, res, sha1);
188 if (!quietly && (status == SHORT_NAME_AMBIGUOUS))
189 return error("short SHA1 %.*s is ambiguous.", len, canonical);
190 return status;
193 const char *find_unique_abbrev(const unsigned char *sha1, int len)
195 int status, exists;
196 static char hex[41];
198 exists = has_sha1_file(sha1);
199 memcpy(hex, sha1_to_hex(sha1), 40);
200 if (len == 40 || !len)
201 return hex;
202 while (len < 40) {
203 unsigned char sha1_ret[20];
204 status = get_short_sha1(hex, len, sha1_ret, 1);
205 if (exists
206 ? !status
207 : status == SHORT_NAME_NOT_FOUND) {
208 hex[len] = 0;
209 return hex;
211 len++;
213 return hex;
216 static int ambiguous_path(const char *path, int len)
218 int slash = 1;
219 int cnt;
221 for (cnt = 0; cnt < len; cnt++) {
222 switch (*path++) {
223 case '\0':
224 break;
225 case '/':
226 if (slash)
227 break;
228 slash = 1;
229 continue;
230 case '.':
231 continue;
232 default:
233 slash = 0;
234 continue;
236 break;
238 return slash;
241 int dwim_ref(const char *str, int len, unsigned char *sha1, char **ref)
243 const char **p, *r;
244 int refs_found = 0;
246 *ref = NULL;
247 for (p = ref_rev_parse_rules; *p; p++) {
248 char fullref[PATH_MAX];
249 unsigned char sha1_from_ref[20];
250 unsigned char *this_result;
252 this_result = refs_found ? sha1_from_ref : sha1;
253 mksnpath(fullref, sizeof(fullref), *p, len, str);
254 r = resolve_ref(fullref, this_result, 1, NULL);
255 if (r) {
256 if (!refs_found++)
257 *ref = xstrdup(r);
258 if (!warn_ambiguous_refs)
259 break;
262 return refs_found;
265 int dwim_log(const char *str, int len, unsigned char *sha1, char **log)
267 const char **p;
268 int logs_found = 0;
270 *log = NULL;
271 for (p = ref_rev_parse_rules; *p; p++) {
272 struct stat st;
273 unsigned char hash[20];
274 char path[PATH_MAX];
275 const char *ref, *it;
277 mksnpath(path, sizeof(path), *p, len, str);
278 ref = resolve_ref(path, hash, 1, NULL);
279 if (!ref)
280 continue;
281 if (!stat(git_path("logs/%s", path), &st) &&
282 S_ISREG(st.st_mode))
283 it = path;
284 else if (strcmp(ref, path) &&
285 !stat(git_path("logs/%s", ref), &st) &&
286 S_ISREG(st.st_mode))
287 it = ref;
288 else
289 continue;
290 if (!logs_found++) {
291 *log = xstrdup(it);
292 hashcpy(sha1, hash);
294 if (!warn_ambiguous_refs)
295 break;
297 return logs_found;
300 static int get_sha1_1(const char *name, int len, unsigned char *sha1);
302 static int get_sha1_basic(const char *str, int len, unsigned char *sha1)
304 static const char *warning = "warning: refname '%.*s' is ambiguous.\n";
305 char *real_ref = NULL;
306 int refs_found = 0;
307 int at, reflog_len;
309 if (len == 40 && !get_sha1_hex(str, sha1))
310 return 0;
312 /* basic@{time or number or -number} format to query ref-log */
313 reflog_len = at = 0;
314 if (str[len-1] == '}') {
315 for (at = 0; at < len - 1; at++) {
316 if (str[at] == '@' && str[at+1] == '{') {
317 reflog_len = (len-1) - (at+2);
318 len = at;
319 break;
324 /* Accept only unambiguous ref paths. */
325 if (len && ambiguous_path(str, len))
326 return -1;
328 if (!len && reflog_len) {
329 struct strbuf buf = STRBUF_INIT;
330 int ret;
331 /* try the @{-N} syntax for n-th checkout */
332 ret = interpret_nth_last_branch(str+at, &buf);
333 if (ret > 0) {
334 /* substitute this branch name and restart */
335 return get_sha1_1(buf.buf, buf.len, sha1);
336 } else if (ret == 0) {
337 return -1;
339 /* allow "@{...}" to mean the current branch reflog */
340 refs_found = dwim_ref("HEAD", 4, sha1, &real_ref);
341 } else if (reflog_len)
342 refs_found = dwim_log(str, len, sha1, &real_ref);
343 else
344 refs_found = dwim_ref(str, len, sha1, &real_ref);
346 if (!refs_found)
347 return -1;
349 if (warn_ambiguous_refs && refs_found > 1)
350 fprintf(stderr, warning, len, str);
352 if (reflog_len) {
353 int nth, i;
354 unsigned long at_time;
355 unsigned long co_time;
356 int co_tz, co_cnt;
358 /* Is it asking for N-th entry, or approxidate? */
359 for (i = nth = 0; 0 <= nth && i < reflog_len; i++) {
360 char ch = str[at+2+i];
361 if ('0' <= ch && ch <= '9')
362 nth = nth * 10 + ch - '0';
363 else
364 nth = -1;
366 if (100000000 <= nth) {
367 at_time = nth;
368 nth = -1;
369 } else if (0 <= nth)
370 at_time = 0;
371 else {
372 char *tmp = xstrndup(str + at + 2, reflog_len);
373 at_time = approxidate(tmp);
374 free(tmp);
376 if (read_ref_at(real_ref, at_time, nth, sha1, NULL,
377 &co_time, &co_tz, &co_cnt)) {
378 if (at_time)
379 fprintf(stderr,
380 "warning: Log for '%.*s' only goes "
381 "back to %s.\n", len, str,
382 show_date(co_time, co_tz, DATE_RFC2822));
383 else
384 fprintf(stderr,
385 "warning: Log for '%.*s' only has "
386 "%d entries.\n", len, str, co_cnt);
390 free(real_ref);
391 return 0;
394 static int get_parent(const char *name, int len,
395 unsigned char *result, int idx)
397 unsigned char sha1[20];
398 int ret = get_sha1_1(name, len, sha1);
399 struct commit *commit;
400 struct commit_list *p;
402 if (ret)
403 return ret;
404 commit = lookup_commit_reference(sha1);
405 if (!commit)
406 return -1;
407 if (parse_commit(commit))
408 return -1;
409 if (!idx) {
410 hashcpy(result, commit->object.sha1);
411 return 0;
413 p = commit->parents;
414 while (p) {
415 if (!--idx) {
416 hashcpy(result, p->item->object.sha1);
417 return 0;
419 p = p->next;
421 return -1;
424 static int get_nth_ancestor(const char *name, int len,
425 unsigned char *result, int generation)
427 unsigned char sha1[20];
428 struct commit *commit;
429 int ret;
431 ret = get_sha1_1(name, len, sha1);
432 if (ret)
433 return ret;
434 commit = lookup_commit_reference(sha1);
435 if (!commit)
436 return -1;
438 while (generation--) {
439 if (parse_commit(commit) || !commit->parents)
440 return -1;
441 commit = commit->parents->item;
443 hashcpy(result, commit->object.sha1);
444 return 0;
447 struct object *peel_to_type(const char *name, int namelen,
448 struct object *o, enum object_type expected_type)
450 if (name && !namelen)
451 namelen = strlen(name);
452 if (!o) {
453 unsigned char sha1[20];
454 if (get_sha1_1(name, namelen, sha1))
455 return NULL;
456 o = parse_object(sha1);
458 while (1) {
459 if (!o || (!o->parsed && !parse_object(o->sha1)))
460 return NULL;
461 if (o->type == expected_type)
462 return o;
463 if (o->type == OBJ_TAG)
464 o = ((struct tag*) o)->tagged;
465 else if (o->type == OBJ_COMMIT)
466 o = &(((struct commit *) o)->tree->object);
467 else {
468 if (name)
469 error("%.*s: expected %s type, but the object "
470 "dereferences to %s type",
471 namelen, name, typename(expected_type),
472 typename(o->type));
473 return NULL;
478 static int peel_onion(const char *name, int len, unsigned char *sha1)
480 unsigned char outer[20];
481 const char *sp;
482 unsigned int expected_type = 0;
483 struct object *o;
486 * "ref^{type}" dereferences ref repeatedly until you cannot
487 * dereference anymore, or you get an object of given type,
488 * whichever comes first. "ref^{}" means just dereference
489 * tags until you get a non-tag. "ref^0" is a shorthand for
490 * "ref^{commit}". "commit^{tree}" could be used to find the
491 * top-level tree of the given commit.
493 if (len < 4 || name[len-1] != '}')
494 return -1;
496 for (sp = name + len - 1; name <= sp; sp--) {
497 int ch = *sp;
498 if (ch == '{' && name < sp && sp[-1] == '^')
499 break;
501 if (sp <= name)
502 return -1;
504 sp++; /* beginning of type name, or closing brace for empty */
505 if (!strncmp(commit_type, sp, 6) && sp[6] == '}')
506 expected_type = OBJ_COMMIT;
507 else if (!strncmp(tree_type, sp, 4) && sp[4] == '}')
508 expected_type = OBJ_TREE;
509 else if (!strncmp(blob_type, sp, 4) && sp[4] == '}')
510 expected_type = OBJ_BLOB;
511 else if (sp[0] == '}')
512 expected_type = OBJ_NONE;
513 else
514 return -1;
516 if (get_sha1_1(name, sp - name - 2, outer))
517 return -1;
519 o = parse_object(outer);
520 if (!o)
521 return -1;
522 if (!expected_type) {
523 o = deref_tag(o, name, sp - name - 2);
524 if (!o || (!o->parsed && !parse_object(o->sha1)))
525 return -1;
526 hashcpy(sha1, o->sha1);
528 else {
530 * At this point, the syntax look correct, so
531 * if we do not get the needed object, we should
532 * barf.
534 o = peel_to_type(name, len, o, expected_type);
535 if (o) {
536 hashcpy(sha1, o->sha1);
537 return 0;
539 return -1;
541 return 0;
544 static int get_describe_name(const char *name, int len, unsigned char *sha1)
546 const char *cp;
548 for (cp = name + len - 1; name + 2 <= cp; cp--) {
549 char ch = *cp;
550 if (hexval(ch) & ~0377) {
551 /* We must be looking at g in "SOMETHING-g"
552 * for it to be describe output.
554 if (ch == 'g' && cp[-1] == '-') {
555 cp++;
556 len -= cp - name;
557 return get_short_sha1(cp, len, sha1, 1);
561 return -1;
564 static int get_sha1_1(const char *name, int len, unsigned char *sha1)
566 int ret, has_suffix;
567 const char *cp;
570 * "name~3" is "name^^^", "name~" is "name~1", and "name^" is "name^1".
572 has_suffix = 0;
573 for (cp = name + len - 1; name <= cp; cp--) {
574 int ch = *cp;
575 if ('0' <= ch && ch <= '9')
576 continue;
577 if (ch == '~' || ch == '^')
578 has_suffix = ch;
579 break;
582 if (has_suffix) {
583 int num = 0;
584 int len1 = cp - name;
585 cp++;
586 while (cp < name + len)
587 num = num * 10 + *cp++ - '0';
588 if (!num && len1 == len - 1)
589 num = 1;
590 if (has_suffix == '^')
591 return get_parent(name, len1, sha1, num);
592 /* else if (has_suffix == '~') -- goes without saying */
593 return get_nth_ancestor(name, len1, sha1, num);
596 ret = peel_onion(name, len, sha1);
597 if (!ret)
598 return 0;
600 ret = get_sha1_basic(name, len, sha1);
601 if (!ret)
602 return 0;
604 /* It could be describe output that is "SOMETHING-gXXXX" */
605 ret = get_describe_name(name, len, sha1);
606 if (!ret)
607 return 0;
609 return get_short_sha1(name, len, sha1, 0);
612 static int handle_one_ref(const char *path,
613 const unsigned char *sha1, int flag, void *cb_data)
615 struct commit_list **list = cb_data;
616 struct object *object = parse_object(sha1);
617 if (!object)
618 return 0;
619 if (object->type == OBJ_TAG) {
620 object = deref_tag(object, path, strlen(path));
621 if (!object)
622 return 0;
624 if (object->type != OBJ_COMMIT)
625 return 0;
626 insert_by_date((struct commit *)object, list);
627 return 0;
631 * This interprets names like ':/Initial revision of "git"' by searching
632 * through history and returning the first commit whose message starts
633 * with the given string.
635 * For future extension, ':/!' is reserved. If you want to match a message
636 * beginning with a '!', you have to repeat the exclamation mark.
639 #define ONELINE_SEEN (1u<<20)
640 static int get_sha1_oneline(const char *prefix, unsigned char *sha1)
642 struct commit_list *list = NULL, *backup = NULL, *l;
643 int retval = -1;
644 char *temp_commit_buffer = NULL;
646 if (prefix[0] == '!') {
647 if (prefix[1] != '!')
648 die ("Invalid search pattern: %s", prefix);
649 prefix++;
651 for_each_ref(handle_one_ref, &list);
652 for (l = list; l; l = l->next)
653 commit_list_insert(l->item, &backup);
654 while (list) {
655 char *p;
656 struct commit *commit;
657 enum object_type type;
658 unsigned long size;
660 commit = pop_most_recent_commit(&list, ONELINE_SEEN);
661 if (!parse_object(commit->object.sha1))
662 continue;
663 free(temp_commit_buffer);
664 if (commit->buffer)
665 p = commit->buffer;
666 else {
667 p = read_sha1_file(commit->object.sha1, &type, &size);
668 if (!p)
669 continue;
670 temp_commit_buffer = p;
672 if (!(p = strstr(p, "\n\n")))
673 continue;
674 if (!prefixcmp(p + 2, prefix)) {
675 hashcpy(sha1, commit->object.sha1);
676 retval = 0;
677 break;
680 free(temp_commit_buffer);
681 free_commit_list(list);
682 for (l = backup; l; l = l->next)
683 clear_commit_marks(l->item, ONELINE_SEEN);
684 return retval;
687 struct grab_nth_branch_switch_cbdata {
688 long cnt, alloc;
689 struct strbuf *buf;
692 static int grab_nth_branch_switch(unsigned char *osha1, unsigned char *nsha1,
693 const char *email, unsigned long timestamp, int tz,
694 const char *message, void *cb_data)
696 struct grab_nth_branch_switch_cbdata *cb = cb_data;
697 const char *match = NULL, *target = NULL;
698 size_t len;
699 int nth;
701 if (!prefixcmp(message, "checkout: moving from ")) {
702 match = message + strlen("checkout: moving from ");
703 if ((target = strstr(match, " to ")) != NULL)
704 target += 4;
707 if (!match)
708 return 0;
710 len = target - match - 4;
711 if (target[len] == '\n' && !strncmp(match, target, len))
712 return 0;
714 nth = cb->cnt++ % cb->alloc;
715 strbuf_reset(&cb->buf[nth]);
716 strbuf_add(&cb->buf[nth], match, len);
717 return 0;
721 * This reads "@{-N}" syntax, finds the name of the Nth previous
722 * branch we were on, and places the name of the branch in the given
723 * buf and returns the number of characters parsed if successful.
725 * If the input is not of the accepted format, it returns a negative
726 * number to signal an error.
728 * If the input was ok but there are not N branch switches in the
729 * reflog, it returns 0.
731 int interpret_nth_last_branch(const char *name, struct strbuf *buf)
733 long nth;
734 int i;
735 struct grab_nth_branch_switch_cbdata cb;
736 const char *brace;
737 char *num_end;
739 if (name[0] != '@' || name[1] != '{' || name[2] != '-')
740 return -1;
741 brace = strchr(name, '}');
742 if (!brace)
743 return -1;
744 nth = strtol(name+3, &num_end, 10);
745 if (num_end != brace)
746 return -1;
747 if (nth <= 0)
748 return -1;
749 cb.alloc = nth;
750 cb.buf = xmalloc(nth * sizeof(struct strbuf));
751 for (i = 0; i < nth; i++)
752 strbuf_init(&cb.buf[i], 20);
753 cb.cnt = 0;
754 for_each_reflog_ent("HEAD", grab_nth_branch_switch, &cb);
755 if (cb.cnt < nth)
756 return 0;
757 i = cb.cnt % nth;
758 strbuf_reset(buf);
759 strbuf_add(buf, cb.buf[i].buf, cb.buf[i].len);
760 for (i = 0; i < nth; i++)
761 strbuf_release(&cb.buf[i]);
762 free(cb.buf);
764 return brace-name+1;
768 * This is like "get_sha1_basic()", except it allows "sha1 expressions",
769 * notably "xyz^" for "parent of xyz"
771 int get_sha1(const char *name, unsigned char *sha1)
773 unsigned unused;
774 return get_sha1_with_mode(name, sha1, &unused);
777 int get_sha1_with_mode(const char *name, unsigned char *sha1, unsigned *mode)
779 int ret, bracket_depth;
780 int namelen = strlen(name);
781 const char *cp;
783 *mode = S_IFINVALID;
784 ret = get_sha1_1(name, namelen, sha1);
785 if (!ret)
786 return ret;
787 /* sha1:path --> object name of path in ent sha1
788 * :path -> object name of path in index
789 * :[0-3]:path -> object name of path in index at stage
791 if (name[0] == ':') {
792 int stage = 0;
793 struct cache_entry *ce;
794 int pos;
795 if (namelen > 2 && name[1] == '/')
796 return get_sha1_oneline(name + 2, sha1);
797 if (namelen < 3 ||
798 name[2] != ':' ||
799 name[1] < '0' || '3' < name[1])
800 cp = name + 1;
801 else {
802 stage = name[1] - '0';
803 cp = name + 3;
805 namelen = namelen - (cp - name);
806 if (!active_cache)
807 read_cache();
808 pos = cache_name_pos(cp, namelen);
809 if (pos < 0)
810 pos = -pos - 1;
811 while (pos < active_nr) {
812 ce = active_cache[pos];
813 if (ce_namelen(ce) != namelen ||
814 memcmp(ce->name, cp, namelen))
815 break;
816 if (ce_stage(ce) == stage) {
817 hashcpy(sha1, ce->sha1);
818 *mode = ce->ce_mode;
819 return 0;
821 pos++;
823 return -1;
825 for (cp = name, bracket_depth = 0; *cp; cp++) {
826 if (*cp == '{')
827 bracket_depth++;
828 else if (bracket_depth && *cp == '}')
829 bracket_depth--;
830 else if (!bracket_depth && *cp == ':')
831 break;
833 if (*cp == ':') {
834 unsigned char tree_sha1[20];
835 if (!get_sha1_1(name, cp-name, tree_sha1))
836 return get_tree_entry(tree_sha1, cp+1, sha1,
837 mode);
839 return ret;