peel_onion: disambiguate to favor tree-ish when we know we want a tree-ish
[git/jnareb-git.git] / sha1_name.c
blob45788df8bfb79d48c19ef6e1092a4785c6ce525e
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"
8 #include "remote.h"
10 static int get_sha1_oneline(const char *, unsigned char *, struct commit_list *);
12 typedef int (*disambiguate_hint_fn)(const unsigned char *, void *);
14 struct disambiguate_state {
15 disambiguate_hint_fn fn;
16 void *cb_data;
17 unsigned char candidate[20];
18 unsigned candidate_exists:1;
19 unsigned candidate_checked:1;
20 unsigned candidate_ok:1;
21 unsigned disambiguate_fn_used:1;
22 unsigned ambiguous:1;
23 unsigned always_call_fn:1;
26 static void update_candidates(struct disambiguate_state *ds, const unsigned char *current)
28 if (ds->always_call_fn) {
29 ds->ambiguous = ds->fn(current, ds->cb_data) ? 1 : 0;
30 return;
32 if (!ds->candidate_exists) {
33 /* this is the first candidate */
34 hashcpy(ds->candidate, current);
35 ds->candidate_exists = 1;
36 return;
37 } else if (!hashcmp(ds->candidate, current)) {
38 /* the same as what we already have seen */
39 return;
42 if (!ds->fn) {
43 /* cannot disambiguate between ds->candidate and current */
44 ds->ambiguous = 1;
45 return;
48 if (!ds->candidate_checked) {
49 ds->candidate_ok = ds->fn(ds->candidate, ds->cb_data);
50 ds->disambiguate_fn_used = 1;
51 ds->candidate_checked = 1;
54 if (!ds->candidate_ok) {
55 /* discard the candidate; we know it does not satisify fn */
56 hashcpy(ds->candidate, current);
57 ds->candidate_checked = 0;
58 return;
61 /* if we reach this point, we know ds->candidate satisfies fn */
62 if (ds->fn(current, ds->cb_data)) {
64 * if both current and candidate satisfy fn, we cannot
65 * disambiguate.
67 ds->candidate_ok = 0;
68 ds->ambiguous = 1;
71 /* otherwise, current can be discarded and candidate is still good */
74 static void find_short_object_filename(int len, const char *hex_pfx, struct disambiguate_state *ds)
76 struct alternate_object_database *alt;
77 char hex[40];
78 static struct alternate_object_database *fakeent;
80 if (!fakeent) {
82 * Create a "fake" alternate object database that
83 * points to our own object database, to make it
84 * easier to get a temporary working space in
85 * alt->name/alt->base while iterating over the
86 * object databases including our own.
88 const char *objdir = get_object_directory();
89 int objdir_len = strlen(objdir);
90 int entlen = objdir_len + 43;
91 fakeent = xmalloc(sizeof(*fakeent) + entlen);
92 memcpy(fakeent->base, objdir, objdir_len);
93 fakeent->name = fakeent->base + objdir_len + 1;
94 fakeent->name[-1] = '/';
96 fakeent->next = alt_odb_list;
98 sprintf(hex, "%.2s", hex_pfx);
99 for (alt = fakeent; alt && !ds->ambiguous; alt = alt->next) {
100 struct dirent *de;
101 DIR *dir;
102 sprintf(alt->name, "%.2s/", hex_pfx);
103 dir = opendir(alt->base);
104 if (!dir)
105 continue;
107 while (!ds->ambiguous && (de = readdir(dir)) != NULL) {
108 unsigned char sha1[20];
110 if (strlen(de->d_name) != 38)
111 continue;
112 if (memcmp(de->d_name, hex_pfx + 2, len - 2))
113 continue;
114 memcpy(hex + 2, de->d_name, 38);
115 if (!get_sha1_hex(hex, sha1))
116 update_candidates(ds, sha1);
118 closedir(dir);
122 static int match_sha(unsigned len, const unsigned char *a, const unsigned char *b)
124 do {
125 if (*a != *b)
126 return 0;
127 a++;
128 b++;
129 len -= 2;
130 } while (len > 1);
131 if (len)
132 if ((*a ^ *b) & 0xf0)
133 return 0;
134 return 1;
137 static void unique_in_pack(int len,
138 const unsigned char *bin_pfx,
139 struct packed_git *p,
140 struct disambiguate_state *ds)
142 uint32_t num, last, i, first = 0;
143 const unsigned char *current = NULL;
145 open_pack_index(p);
146 num = p->num_objects;
147 last = num;
148 while (first < last) {
149 uint32_t mid = (first + last) / 2;
150 const unsigned char *current;
151 int cmp;
153 current = nth_packed_object_sha1(p, mid);
154 cmp = hashcmp(bin_pfx, current);
155 if (!cmp) {
156 first = mid;
157 break;
159 if (cmp > 0) {
160 first = mid+1;
161 continue;
163 last = mid;
167 * At this point, "first" is the location of the lowest object
168 * with an object name that could match "bin_pfx". See if we have
169 * 0, 1 or more objects that actually match(es).
171 for (i = first; i < num && !ds->ambiguous; i++) {
172 current = nth_packed_object_sha1(p, i);
173 if (!match_sha(len, bin_pfx, current))
174 break;
175 update_candidates(ds, current);
179 static void find_short_packed_object(int len, const unsigned char *bin_pfx,
180 struct disambiguate_state *ds)
182 struct packed_git *p;
184 prepare_packed_git();
185 for (p = packed_git; p && !ds->ambiguous; p = p->next)
186 unique_in_pack(len, bin_pfx, p, ds);
189 #define SHORT_NAME_NOT_FOUND (-1)
190 #define SHORT_NAME_AMBIGUOUS (-2)
192 static int finish_object_disambiguation(struct disambiguate_state *ds,
193 unsigned char *sha1)
195 if (ds->ambiguous)
196 return SHORT_NAME_AMBIGUOUS;
198 if (!ds->candidate_exists)
199 return SHORT_NAME_NOT_FOUND;
201 if (!ds->candidate_checked)
203 * If this is the only candidate, there is no point
204 * calling the disambiguation hint callback.
206 * On the other hand, if the current candidate
207 * replaced an earlier candidate that did _not_ pass
208 * the disambiguation hint callback, then we do have
209 * more than one objects that match the short name
210 * given, so we should make sure this one matches;
211 * otherwise, if we discovered this one and the one
212 * that we previously discarded in the reverse order,
213 * we would end up showing different results in the
214 * same repository!
216 ds->candidate_ok = (!ds->disambiguate_fn_used ||
217 ds->fn(ds->candidate, ds->cb_data));
219 if (!ds->candidate_ok)
220 return SHORT_NAME_AMBIGUOUS;
222 hashcpy(sha1, ds->candidate);
223 return 0;
226 static int disambiguate_commit_only(const unsigned char *sha1, void *cb_data_unused)
228 int kind = sha1_object_info(sha1, NULL);
229 return kind == OBJ_COMMIT;
232 static int disambiguate_committish_only(const unsigned char *sha1, void *cb_data_unused)
234 struct object *obj;
235 int kind;
237 kind = sha1_object_info(sha1, NULL);
238 if (kind == OBJ_COMMIT)
239 return 1;
240 if (kind != OBJ_TAG)
241 return 0;
243 /* We need to do this the hard way... */
244 obj = deref_tag(lookup_object(sha1), NULL, 0);
245 if (obj && obj->type == OBJ_COMMIT)
246 return 1;
247 return 0;
250 static int disambiguate_tree_only(const unsigned char *sha1, void *cb_data_unused)
252 int kind = sha1_object_info(sha1, NULL);
253 return kind == OBJ_TREE;
256 static int disambiguate_treeish_only(const unsigned char *sha1, void *cb_data_unused)
258 struct object *obj;
259 int kind;
261 kind = sha1_object_info(sha1, NULL);
262 if (kind == OBJ_TREE || kind == OBJ_COMMIT)
263 return 1;
264 if (kind != OBJ_TAG)
265 return 0;
267 /* We need to do this the hard way... */
268 obj = deref_tag(lookup_object(sha1), NULL, 0);
269 if (obj && (obj->type == OBJ_TREE || obj->type == OBJ_COMMIT))
270 return 1;
271 return 0;
274 static int disambiguate_blob_only(const unsigned char *sha1, void *cb_data_unused)
276 int kind = sha1_object_info(sha1, NULL);
277 return kind == OBJ_BLOB;
280 static int prepare_prefixes(const char *name, int len,
281 unsigned char *bin_pfx,
282 char *hex_pfx)
284 int i;
286 hashclr(bin_pfx);
287 memset(hex_pfx, 'x', 40);
288 for (i = 0; i < len ;i++) {
289 unsigned char c = name[i];
290 unsigned char val;
291 if (c >= '0' && c <= '9')
292 val = c - '0';
293 else if (c >= 'a' && c <= 'f')
294 val = c - 'a' + 10;
295 else if (c >= 'A' && c <='F') {
296 val = c - 'A' + 10;
297 c -= 'A' - 'a';
299 else
300 return -1;
301 hex_pfx[i] = c;
302 if (!(i & 1))
303 val <<= 4;
304 bin_pfx[i >> 1] |= val;
306 return 0;
309 static int get_short_sha1(const char *name, int len, unsigned char *sha1,
310 unsigned flags)
312 int status;
313 char hex_pfx[40];
314 unsigned char bin_pfx[20];
315 struct disambiguate_state ds;
316 int quietly = !!(flags & GET_SHA1_QUIETLY);
318 if (len < MINIMUM_ABBREV || len > 40)
319 return -1;
320 if (prepare_prefixes(name, len, bin_pfx, hex_pfx) < 0)
321 return -1;
323 prepare_alt_odb();
325 memset(&ds, 0, sizeof(ds));
326 if (flags & GET_SHA1_COMMIT)
327 ds.fn = disambiguate_commit_only;
328 else if (flags & GET_SHA1_COMMITTISH)
329 ds.fn = disambiguate_committish_only;
330 else if (flags & GET_SHA1_TREE)
331 ds.fn = disambiguate_tree_only;
332 else if (flags & GET_SHA1_TREEISH)
333 ds.fn = disambiguate_treeish_only;
334 else if (flags & GET_SHA1_BLOB)
335 ds.fn = disambiguate_blob_only;
337 find_short_object_filename(len, hex_pfx, &ds);
338 find_short_packed_object(len, bin_pfx, &ds);
339 status = finish_object_disambiguation(&ds, sha1);
341 if (!quietly && (status == SHORT_NAME_AMBIGUOUS))
342 return error("short SHA1 %.*s is ambiguous.", len, hex_pfx);
343 return status;
347 int for_each_abbrev(const char *prefix, each_abbrev_fn fn, void *cb_data)
349 char hex_pfx[40];
350 unsigned char bin_pfx[20];
351 struct disambiguate_state ds;
352 int len = strlen(prefix);
354 if (len < MINIMUM_ABBREV || len > 40)
355 return -1;
356 if (prepare_prefixes(prefix, len, bin_pfx, hex_pfx) < 0)
357 return -1;
359 prepare_alt_odb();
361 memset(&ds, 0, sizeof(ds));
362 ds.always_call_fn = 1;
363 ds.cb_data = cb_data;
364 ds.fn = fn;
366 find_short_object_filename(len, hex_pfx, &ds);
367 find_short_packed_object(len, bin_pfx, &ds);
368 return ds.ambiguous;
371 const char *find_unique_abbrev(const unsigned char *sha1, int len)
373 int status, exists;
374 static char hex[41];
376 exists = has_sha1_file(sha1);
377 memcpy(hex, sha1_to_hex(sha1), 40);
378 if (len == 40 || !len)
379 return hex;
380 while (len < 40) {
381 unsigned char sha1_ret[20];
382 status = get_short_sha1(hex, len, sha1_ret, GET_SHA1_QUIETLY);
383 if (exists
384 ? !status
385 : status == SHORT_NAME_NOT_FOUND) {
386 hex[len] = 0;
387 return hex;
389 len++;
391 return hex;
394 static int ambiguous_path(const char *path, int len)
396 int slash = 1;
397 int cnt;
399 for (cnt = 0; cnt < len; cnt++) {
400 switch (*path++) {
401 case '\0':
402 break;
403 case '/':
404 if (slash)
405 break;
406 slash = 1;
407 continue;
408 case '.':
409 continue;
410 default:
411 slash = 0;
412 continue;
414 break;
416 return slash;
419 static inline int upstream_mark(const char *string, int len)
421 const char *suffix[] = { "@{upstream}", "@{u}" };
422 int i;
424 for (i = 0; i < ARRAY_SIZE(suffix); i++) {
425 int suffix_len = strlen(suffix[i]);
426 if (suffix_len <= len
427 && !memcmp(string, suffix[i], suffix_len))
428 return suffix_len;
430 return 0;
433 static int get_sha1_1(const char *name, int len, unsigned char *sha1, unsigned lookup_flags);
435 static int get_sha1_basic(const char *str, int len, unsigned char *sha1)
437 static const char *warn_msg = "refname '%.*s' is ambiguous.";
438 char *real_ref = NULL;
439 int refs_found = 0;
440 int at, reflog_len;
442 if (len == 40 && !get_sha1_hex(str, sha1))
443 return 0;
445 /* basic@{time or number or -number} format to query ref-log */
446 reflog_len = at = 0;
447 if (len && str[len-1] == '}') {
448 for (at = len-2; at >= 0; at--) {
449 if (str[at] == '@' && str[at+1] == '{') {
450 if (!upstream_mark(str + at, len - at)) {
451 reflog_len = (len-1) - (at+2);
452 len = at;
454 break;
459 /* Accept only unambiguous ref paths. */
460 if (len && ambiguous_path(str, len))
461 return -1;
463 if (!len && reflog_len) {
464 struct strbuf buf = STRBUF_INIT;
465 int ret;
466 /* try the @{-N} syntax for n-th checkout */
467 ret = interpret_branch_name(str+at, &buf);
468 if (ret > 0) {
469 /* substitute this branch name and restart */
470 return get_sha1_1(buf.buf, buf.len, sha1, 0);
471 } else if (ret == 0) {
472 return -1;
474 /* allow "@{...}" to mean the current branch reflog */
475 refs_found = dwim_ref("HEAD", 4, sha1, &real_ref);
476 } else if (reflog_len)
477 refs_found = dwim_log(str, len, sha1, &real_ref);
478 else
479 refs_found = dwim_ref(str, len, sha1, &real_ref);
481 if (!refs_found)
482 return -1;
484 if (warn_ambiguous_refs && refs_found > 1)
485 warning(warn_msg, len, str);
487 if (reflog_len) {
488 int nth, i;
489 unsigned long at_time;
490 unsigned long co_time;
491 int co_tz, co_cnt;
493 /* a @{-N} placed anywhere except the start is an error */
494 if (str[at+2] == '-')
495 return -1;
497 /* Is it asking for N-th entry, or approxidate? */
498 for (i = nth = 0; 0 <= nth && i < reflog_len; i++) {
499 char ch = str[at+2+i];
500 if ('0' <= ch && ch <= '9')
501 nth = nth * 10 + ch - '0';
502 else
503 nth = -1;
505 if (100000000 <= nth) {
506 at_time = nth;
507 nth = -1;
508 } else if (0 <= nth)
509 at_time = 0;
510 else {
511 int errors = 0;
512 char *tmp = xstrndup(str + at + 2, reflog_len);
513 at_time = approxidate_careful(tmp, &errors);
514 free(tmp);
515 if (errors)
516 return -1;
518 if (read_ref_at(real_ref, at_time, nth, sha1, NULL,
519 &co_time, &co_tz, &co_cnt)) {
520 if (at_time)
521 warning("Log for '%.*s' only goes "
522 "back to %s.", len, str,
523 show_date(co_time, co_tz, DATE_RFC2822));
524 else {
525 free(real_ref);
526 die("Log for '%.*s' only has %d entries.",
527 len, str, co_cnt);
532 free(real_ref);
533 return 0;
536 static int get_parent(const char *name, int len,
537 unsigned char *result, int idx)
539 unsigned char sha1[20];
540 int ret = get_sha1_1(name, len, sha1, GET_SHA1_COMMITTISH);
541 struct commit *commit;
542 struct commit_list *p;
544 if (ret)
545 return ret;
546 commit = lookup_commit_reference(sha1);
547 if (!commit)
548 return -1;
549 if (parse_commit(commit))
550 return -1;
551 if (!idx) {
552 hashcpy(result, commit->object.sha1);
553 return 0;
555 p = commit->parents;
556 while (p) {
557 if (!--idx) {
558 hashcpy(result, p->item->object.sha1);
559 return 0;
561 p = p->next;
563 return -1;
566 static int get_nth_ancestor(const char *name, int len,
567 unsigned char *result, int generation)
569 unsigned char sha1[20];
570 struct commit *commit;
571 int ret;
573 ret = get_sha1_1(name, len, sha1, GET_SHA1_COMMITTISH);
574 if (ret)
575 return ret;
576 commit = lookup_commit_reference(sha1);
577 if (!commit)
578 return -1;
580 while (generation--) {
581 if (parse_commit(commit) || !commit->parents)
582 return -1;
583 commit = commit->parents->item;
585 hashcpy(result, commit->object.sha1);
586 return 0;
589 struct object *peel_to_type(const char *name, int namelen,
590 struct object *o, enum object_type expected_type)
592 if (name && !namelen)
593 namelen = strlen(name);
594 while (1) {
595 if (!o || (!o->parsed && !parse_object(o->sha1)))
596 return NULL;
597 if (o->type == expected_type)
598 return o;
599 if (o->type == OBJ_TAG)
600 o = ((struct tag*) o)->tagged;
601 else if (o->type == OBJ_COMMIT)
602 o = &(((struct commit *) o)->tree->object);
603 else {
604 if (name)
605 error("%.*s: expected %s type, but the object "
606 "dereferences to %s type",
607 namelen, name, typename(expected_type),
608 typename(o->type));
609 return NULL;
614 static int peel_onion(const char *name, int len, unsigned char *sha1)
616 unsigned char outer[20];
617 const char *sp;
618 unsigned int expected_type = 0;
619 unsigned lookup_flags = 0;
620 struct object *o;
623 * "ref^{type}" dereferences ref repeatedly until you cannot
624 * dereference anymore, or you get an object of given type,
625 * whichever comes first. "ref^{}" means just dereference
626 * tags until you get a non-tag. "ref^0" is a shorthand for
627 * "ref^{commit}". "commit^{tree}" could be used to find the
628 * top-level tree of the given commit.
630 if (len < 4 || name[len-1] != '}')
631 return -1;
633 for (sp = name + len - 1; name <= sp; sp--) {
634 int ch = *sp;
635 if (ch == '{' && name < sp && sp[-1] == '^')
636 break;
638 if (sp <= name)
639 return -1;
641 sp++; /* beginning of type name, or closing brace for empty */
642 if (!strncmp(commit_type, sp, 6) && sp[6] == '}')
643 expected_type = OBJ_COMMIT;
644 else if (!strncmp(tree_type, sp, 4) && sp[4] == '}')
645 expected_type = OBJ_TREE;
646 else if (!strncmp(blob_type, sp, 4) && sp[4] == '}')
647 expected_type = OBJ_BLOB;
648 else if (sp[0] == '}')
649 expected_type = OBJ_NONE;
650 else if (sp[0] == '/')
651 expected_type = OBJ_COMMIT;
652 else
653 return -1;
655 if (expected_type == OBJ_COMMIT)
656 lookup_flags = GET_SHA1_COMMITTISH;
657 else if (expected_type == OBJ_TREE)
658 lookup_flags = GET_SHA1_TREEISH;
660 if (get_sha1_1(name, sp - name - 2, outer, lookup_flags))
661 return -1;
663 o = parse_object(outer);
664 if (!o)
665 return -1;
666 if (!expected_type) {
667 o = deref_tag(o, name, sp - name - 2);
668 if (!o || (!o->parsed && !parse_object(o->sha1)))
669 return -1;
670 hashcpy(sha1, o->sha1);
671 return 0;
675 * At this point, the syntax look correct, so
676 * if we do not get the needed object, we should
677 * barf.
679 o = peel_to_type(name, len, o, expected_type);
680 if (!o)
681 return -1;
683 hashcpy(sha1, o->sha1);
684 if (sp[0] == '/') {
685 /* "$commit^{/foo}" */
686 char *prefix;
687 int ret;
688 struct commit_list *list = NULL;
691 * $commit^{/}. Some regex implementation may reject.
692 * We don't need regex anyway. '' pattern always matches.
694 if (sp[1] == '}')
695 return 0;
697 prefix = xstrndup(sp + 1, name + len - 1 - (sp + 1));
698 commit_list_insert((struct commit *)o, &list);
699 ret = get_sha1_oneline(prefix, sha1, list);
700 free(prefix);
701 return ret;
703 return 0;
706 static int get_describe_name(const char *name, int len, unsigned char *sha1)
708 const char *cp;
709 unsigned flags = GET_SHA1_QUIETLY | GET_SHA1_COMMIT;
711 for (cp = name + len - 1; name + 2 <= cp; cp--) {
712 char ch = *cp;
713 if (hexval(ch) & ~0377) {
714 /* We must be looking at g in "SOMETHING-g"
715 * for it to be describe output.
717 if (ch == 'g' && cp[-1] == '-') {
718 cp++;
719 len -= cp - name;
720 return get_short_sha1(cp, len, sha1, flags);
724 return -1;
727 static int get_sha1_1(const char *name, int len, unsigned char *sha1, unsigned lookup_flags)
729 int ret, has_suffix;
730 const char *cp;
733 * "name~3" is "name^^^", "name~" is "name~1", and "name^" is "name^1".
735 has_suffix = 0;
736 for (cp = name + len - 1; name <= cp; cp--) {
737 int ch = *cp;
738 if ('0' <= ch && ch <= '9')
739 continue;
740 if (ch == '~' || ch == '^')
741 has_suffix = ch;
742 break;
745 if (has_suffix) {
746 int num = 0;
747 int len1 = cp - name;
748 cp++;
749 while (cp < name + len)
750 num = num * 10 + *cp++ - '0';
751 if (!num && len1 == len - 1)
752 num = 1;
753 if (has_suffix == '^')
754 return get_parent(name, len1, sha1, num);
755 /* else if (has_suffix == '~') -- goes without saying */
756 return get_nth_ancestor(name, len1, sha1, num);
759 ret = peel_onion(name, len, sha1);
760 if (!ret)
761 return 0;
763 ret = get_sha1_basic(name, len, sha1);
764 if (!ret)
765 return 0;
767 /* It could be describe output that is "SOMETHING-gXXXX" */
768 ret = get_describe_name(name, len, sha1);
769 if (!ret)
770 return 0;
772 return get_short_sha1(name, len, sha1, lookup_flags);
776 * This interprets names like ':/Initial revision of "git"' by searching
777 * through history and returning the first commit whose message starts
778 * the given regular expression.
780 * For future extension, ':/!' is reserved. If you want to match a message
781 * beginning with a '!', you have to repeat the exclamation mark.
783 #define ONELINE_SEEN (1u<<20)
785 static int handle_one_ref(const char *path,
786 const unsigned char *sha1, int flag, void *cb_data)
788 struct commit_list **list = cb_data;
789 struct object *object = parse_object(sha1);
790 if (!object)
791 return 0;
792 if (object->type == OBJ_TAG) {
793 object = deref_tag(object, path, strlen(path));
794 if (!object)
795 return 0;
797 if (object->type != OBJ_COMMIT)
798 return 0;
799 commit_list_insert_by_date((struct commit *)object, list);
800 return 0;
803 static int get_sha1_oneline(const char *prefix, unsigned char *sha1,
804 struct commit_list *list)
806 struct commit_list *backup = NULL, *l;
807 int found = 0;
808 regex_t regex;
810 if (prefix[0] == '!') {
811 if (prefix[1] != '!')
812 die ("Invalid search pattern: %s", prefix);
813 prefix++;
816 if (regcomp(&regex, prefix, REG_EXTENDED))
817 die("Invalid search pattern: %s", prefix);
819 for (l = list; l; l = l->next) {
820 l->item->object.flags |= ONELINE_SEEN;
821 commit_list_insert(l->item, &backup);
823 while (list) {
824 char *p, *to_free = NULL;
825 struct commit *commit;
826 enum object_type type;
827 unsigned long size;
828 int matches;
830 commit = pop_most_recent_commit(&list, ONELINE_SEEN);
831 if (!parse_object(commit->object.sha1))
832 continue;
833 if (commit->buffer)
834 p = commit->buffer;
835 else {
836 p = read_sha1_file(commit->object.sha1, &type, &size);
837 if (!p)
838 continue;
839 to_free = p;
842 p = strstr(p, "\n\n");
843 matches = p && !regexec(&regex, p + 2, 0, NULL, 0);
844 free(to_free);
846 if (matches) {
847 hashcpy(sha1, commit->object.sha1);
848 found = 1;
849 break;
852 regfree(&regex);
853 free_commit_list(list);
854 for (l = backup; l; l = l->next)
855 clear_commit_marks(l->item, ONELINE_SEEN);
856 free_commit_list(backup);
857 return found ? 0 : -1;
860 struct grab_nth_branch_switch_cbdata {
861 long cnt, alloc;
862 struct strbuf *buf;
865 static int grab_nth_branch_switch(unsigned char *osha1, unsigned char *nsha1,
866 const char *email, unsigned long timestamp, int tz,
867 const char *message, void *cb_data)
869 struct grab_nth_branch_switch_cbdata *cb = cb_data;
870 const char *match = NULL, *target = NULL;
871 size_t len;
872 int nth;
874 if (!prefixcmp(message, "checkout: moving from ")) {
875 match = message + strlen("checkout: moving from ");
876 target = strstr(match, " to ");
879 if (!match || !target)
880 return 0;
882 len = target - match;
883 nth = cb->cnt++ % cb->alloc;
884 strbuf_reset(&cb->buf[nth]);
885 strbuf_add(&cb->buf[nth], match, len);
886 return 0;
890 * Parse @{-N} syntax, return the number of characters parsed
891 * if successful; otherwise signal an error with negative value.
893 static int interpret_nth_prior_checkout(const char *name, struct strbuf *buf)
895 long nth;
896 int i, retval;
897 struct grab_nth_branch_switch_cbdata cb;
898 const char *brace;
899 char *num_end;
901 if (name[0] != '@' || name[1] != '{' || name[2] != '-')
902 return -1;
903 brace = strchr(name, '}');
904 if (!brace)
905 return -1;
906 nth = strtol(name+3, &num_end, 10);
907 if (num_end != brace)
908 return -1;
909 if (nth <= 0)
910 return -1;
911 cb.alloc = nth;
912 cb.buf = xmalloc(nth * sizeof(struct strbuf));
913 for (i = 0; i < nth; i++)
914 strbuf_init(&cb.buf[i], 20);
915 cb.cnt = 0;
916 retval = 0;
917 for_each_recent_reflog_ent("HEAD", grab_nth_branch_switch, 40960, &cb);
918 if (cb.cnt < nth) {
919 cb.cnt = 0;
920 for_each_reflog_ent("HEAD", grab_nth_branch_switch, &cb);
922 if (cb.cnt < nth)
923 goto release_return;
924 i = cb.cnt % nth;
925 strbuf_reset(buf);
926 strbuf_add(buf, cb.buf[i].buf, cb.buf[i].len);
927 retval = brace-name+1;
929 release_return:
930 for (i = 0; i < nth; i++)
931 strbuf_release(&cb.buf[i]);
932 free(cb.buf);
934 return retval;
937 int get_sha1_mb(const char *name, unsigned char *sha1)
939 struct commit *one, *two;
940 struct commit_list *mbs;
941 unsigned char sha1_tmp[20];
942 const char *dots;
943 int st;
945 dots = strstr(name, "...");
946 if (!dots)
947 return get_sha1(name, sha1);
948 if (dots == name)
949 st = get_sha1("HEAD", sha1_tmp);
950 else {
951 struct strbuf sb;
952 strbuf_init(&sb, dots - name);
953 strbuf_add(&sb, name, dots - name);
954 st = get_sha1_committish(sb.buf, sha1_tmp);
955 strbuf_release(&sb);
957 if (st)
958 return st;
959 one = lookup_commit_reference_gently(sha1_tmp, 0);
960 if (!one)
961 return -1;
963 if (get_sha1_committish(dots[3] ? (dots + 3) : "HEAD", sha1_tmp))
964 return -1;
965 two = lookup_commit_reference_gently(sha1_tmp, 0);
966 if (!two)
967 return -1;
968 mbs = get_merge_bases(one, two, 1);
969 if (!mbs || mbs->next)
970 st = -1;
971 else {
972 st = 0;
973 hashcpy(sha1, mbs->item->object.sha1);
975 free_commit_list(mbs);
976 return st;
980 * This reads short-hand syntax that not only evaluates to a commit
981 * object name, but also can act as if the end user spelled the name
982 * of the branch from the command line.
984 * - "@{-N}" finds the name of the Nth previous branch we were on, and
985 * places the name of the branch in the given buf and returns the
986 * number of characters parsed if successful.
988 * - "<branch>@{upstream}" finds the name of the other ref that
989 * <branch> is configured to merge with (missing <branch> defaults
990 * to the current branch), and places the name of the branch in the
991 * given buf and returns the number of characters parsed if
992 * successful.
994 * If the input is not of the accepted format, it returns a negative
995 * number to signal an error.
997 * If the input was ok but there are not N branch switches in the
998 * reflog, it returns 0.
1000 int interpret_branch_name(const char *name, struct strbuf *buf)
1002 char *cp;
1003 struct branch *upstream;
1004 int namelen = strlen(name);
1005 int len = interpret_nth_prior_checkout(name, buf);
1006 int tmp_len;
1008 if (!len)
1009 return len; /* syntax Ok, not enough switches */
1010 if (0 < len && len == namelen)
1011 return len; /* consumed all */
1012 else if (0 < len) {
1013 /* we have extra data, which might need further processing */
1014 struct strbuf tmp = STRBUF_INIT;
1015 int used = buf->len;
1016 int ret;
1018 strbuf_add(buf, name + len, namelen - len);
1019 ret = interpret_branch_name(buf->buf, &tmp);
1020 /* that data was not interpreted, remove our cruft */
1021 if (ret < 0) {
1022 strbuf_setlen(buf, used);
1023 return len;
1025 strbuf_reset(buf);
1026 strbuf_addbuf(buf, &tmp);
1027 strbuf_release(&tmp);
1028 /* tweak for size of {-N} versus expanded ref name */
1029 return ret - used + len;
1032 cp = strchr(name, '@');
1033 if (!cp)
1034 return -1;
1035 tmp_len = upstream_mark(cp, namelen - (cp - name));
1036 if (!tmp_len)
1037 return -1;
1038 len = cp + tmp_len - name;
1039 cp = xstrndup(name, cp - name);
1040 upstream = branch_get(*cp ? cp : NULL);
1042 * Upstream can be NULL only if cp refers to HEAD and HEAD
1043 * points to something different than a branch.
1045 if (!upstream)
1046 return error(_("HEAD does not point to a branch"));
1047 if (!upstream->merge || !upstream->merge[0]->dst) {
1048 if (!ref_exists(upstream->refname))
1049 return error(_("No such branch: '%s'"), cp);
1050 if (!upstream->merge)
1051 return error(_("No upstream configured for branch '%s'"),
1052 upstream->name);
1053 return error(
1054 _("Upstream branch '%s' not stored as a remote-tracking branch"),
1055 upstream->merge[0]->src);
1057 free(cp);
1058 cp = shorten_unambiguous_ref(upstream->merge[0]->dst, 0);
1059 strbuf_reset(buf);
1060 strbuf_addstr(buf, cp);
1061 free(cp);
1062 return len;
1065 int strbuf_branchname(struct strbuf *sb, const char *name)
1067 int len = strlen(name);
1068 if (interpret_branch_name(name, sb) == len)
1069 return 0;
1070 strbuf_add(sb, name, len);
1071 return len;
1074 int strbuf_check_branch_ref(struct strbuf *sb, const char *name)
1076 strbuf_branchname(sb, name);
1077 if (name[0] == '-')
1078 return -1;
1079 strbuf_splice(sb, 0, 0, "refs/heads/", 11);
1080 return check_refname_format(sb->buf, 0);
1084 * This is like "get_sha1_basic()", except it allows "sha1 expressions",
1085 * notably "xyz^" for "parent of xyz"
1087 int get_sha1(const char *name, unsigned char *sha1)
1089 struct object_context unused;
1090 return get_sha1_with_context(name, 0, sha1, &unused);
1094 * Many callers know that the user meant to name a committish by
1095 * syntactical positions where the object name appears. Calling this
1096 * function allows the machinery to disambiguate shorter-than-unique
1097 * abbreviated object names between committish and others.
1099 * Note that this does NOT error out when the named object is not a
1100 * committish. It is merely to give a hint to the disambiguation
1101 * machinery.
1103 int get_sha1_committish(const char *name, unsigned char *sha1)
1105 struct object_context unused;
1106 return get_sha1_with_context(name, GET_SHA1_COMMITTISH,
1107 sha1, &unused);
1110 int get_sha1_treeish(const char *name, unsigned char *sha1)
1112 struct object_context unused;
1113 return get_sha1_with_context(name, GET_SHA1_TREEISH,
1114 sha1, &unused);
1117 int get_sha1_commit(const char *name, unsigned char *sha1)
1119 struct object_context unused;
1120 return get_sha1_with_context(name, GET_SHA1_COMMIT,
1121 sha1, &unused);
1124 int get_sha1_tree(const char *name, unsigned char *sha1)
1126 struct object_context unused;
1127 return get_sha1_with_context(name, GET_SHA1_TREE,
1128 sha1, &unused);
1131 int get_sha1_blob(const char *name, unsigned char *sha1)
1133 struct object_context unused;
1134 return get_sha1_with_context(name, GET_SHA1_BLOB,
1135 sha1, &unused);
1138 /* Must be called only when object_name:filename doesn't exist. */
1139 static void diagnose_invalid_sha1_path(const char *prefix,
1140 const char *filename,
1141 const unsigned char *tree_sha1,
1142 const char *object_name,
1143 int object_name_len)
1145 struct stat st;
1146 unsigned char sha1[20];
1147 unsigned mode;
1149 if (!prefix)
1150 prefix = "";
1152 if (!lstat(filename, &st))
1153 die("Path '%s' exists on disk, but not in '%.*s'.",
1154 filename, object_name_len, object_name);
1155 if (errno == ENOENT || errno == ENOTDIR) {
1156 char *fullname = xmalloc(strlen(filename)
1157 + strlen(prefix) + 1);
1158 strcpy(fullname, prefix);
1159 strcat(fullname, filename);
1161 if (!get_tree_entry(tree_sha1, fullname,
1162 sha1, &mode)) {
1163 die("Path '%s' exists, but not '%s'.\n"
1164 "Did you mean '%.*s:%s' aka '%.*s:./%s'?",
1165 fullname,
1166 filename,
1167 object_name_len, object_name,
1168 fullname,
1169 object_name_len, object_name,
1170 filename);
1172 die("Path '%s' does not exist in '%.*s'",
1173 filename, object_name_len, object_name);
1177 /* Must be called only when :stage:filename doesn't exist. */
1178 static void diagnose_invalid_index_path(int stage,
1179 const char *prefix,
1180 const char *filename)
1182 struct stat st;
1183 struct cache_entry *ce;
1184 int pos;
1185 unsigned namelen = strlen(filename);
1186 unsigned fullnamelen;
1187 char *fullname;
1189 if (!prefix)
1190 prefix = "";
1192 /* Wrong stage number? */
1193 pos = cache_name_pos(filename, namelen);
1194 if (pos < 0)
1195 pos = -pos - 1;
1196 if (pos < active_nr) {
1197 ce = active_cache[pos];
1198 if (ce_namelen(ce) == namelen &&
1199 !memcmp(ce->name, filename, namelen))
1200 die("Path '%s' is in the index, but not at stage %d.\n"
1201 "Did you mean ':%d:%s'?",
1202 filename, stage,
1203 ce_stage(ce), filename);
1206 /* Confusion between relative and absolute filenames? */
1207 fullnamelen = namelen + strlen(prefix);
1208 fullname = xmalloc(fullnamelen + 1);
1209 strcpy(fullname, prefix);
1210 strcat(fullname, filename);
1211 pos = cache_name_pos(fullname, fullnamelen);
1212 if (pos < 0)
1213 pos = -pos - 1;
1214 if (pos < active_nr) {
1215 ce = active_cache[pos];
1216 if (ce_namelen(ce) == fullnamelen &&
1217 !memcmp(ce->name, fullname, fullnamelen))
1218 die("Path '%s' is in the index, but not '%s'.\n"
1219 "Did you mean ':%d:%s' aka ':%d:./%s'?",
1220 fullname, filename,
1221 ce_stage(ce), fullname,
1222 ce_stage(ce), filename);
1225 if (!lstat(filename, &st))
1226 die("Path '%s' exists on disk, but not in the index.", filename);
1227 if (errno == ENOENT || errno == ENOTDIR)
1228 die("Path '%s' does not exist (neither on disk nor in the index).",
1229 filename);
1231 free(fullname);
1235 static char *resolve_relative_path(const char *rel)
1237 if (prefixcmp(rel, "./") && prefixcmp(rel, "../"))
1238 return NULL;
1240 if (!startup_info)
1241 die("BUG: startup_info struct is not initialized.");
1243 if (!is_inside_work_tree())
1244 die("relative path syntax can't be used outside working tree.");
1246 /* die() inside prefix_path() if resolved path is outside worktree */
1247 return prefix_path(startup_info->prefix,
1248 startup_info->prefix ? strlen(startup_info->prefix) : 0,
1249 rel);
1252 static int get_sha1_with_context_1(const char *name,
1253 unsigned flags,
1254 const char *prefix,
1255 unsigned char *sha1,
1256 struct object_context *oc)
1258 int ret, bracket_depth;
1259 int namelen = strlen(name);
1260 const char *cp;
1261 int only_to_die = flags & GET_SHA1_ONLY_TO_DIE;
1263 memset(oc, 0, sizeof(*oc));
1264 oc->mode = S_IFINVALID;
1265 ret = get_sha1_1(name, namelen, sha1, flags);
1266 if (!ret)
1267 return ret;
1269 * sha1:path --> object name of path in ent sha1
1270 * :path -> object name of absolute path in index
1271 * :./path -> object name of path relative to cwd in index
1272 * :[0-3]:path -> object name of path in index at stage
1273 * :/foo -> recent commit matching foo
1275 if (name[0] == ':') {
1276 int stage = 0;
1277 struct cache_entry *ce;
1278 char *new_path = NULL;
1279 int pos;
1280 if (!only_to_die && namelen > 2 && name[1] == '/') {
1281 struct commit_list *list = NULL;
1282 for_each_ref(handle_one_ref, &list);
1283 return get_sha1_oneline(name + 2, sha1, list);
1285 if (namelen < 3 ||
1286 name[2] != ':' ||
1287 name[1] < '0' || '3' < name[1])
1288 cp = name + 1;
1289 else {
1290 stage = name[1] - '0';
1291 cp = name + 3;
1293 new_path = resolve_relative_path(cp);
1294 if (!new_path) {
1295 namelen = namelen - (cp - name);
1296 } else {
1297 cp = new_path;
1298 namelen = strlen(cp);
1301 strncpy(oc->path, cp,
1302 sizeof(oc->path));
1303 oc->path[sizeof(oc->path)-1] = '\0';
1305 if (!active_cache)
1306 read_cache();
1307 pos = cache_name_pos(cp, namelen);
1308 if (pos < 0)
1309 pos = -pos - 1;
1310 while (pos < active_nr) {
1311 ce = active_cache[pos];
1312 if (ce_namelen(ce) != namelen ||
1313 memcmp(ce->name, cp, namelen))
1314 break;
1315 if (ce_stage(ce) == stage) {
1316 hashcpy(sha1, ce->sha1);
1317 oc->mode = ce->ce_mode;
1318 free(new_path);
1319 return 0;
1321 pos++;
1323 if (only_to_die && name[1] && name[1] != '/')
1324 diagnose_invalid_index_path(stage, prefix, cp);
1325 free(new_path);
1326 return -1;
1328 for (cp = name, bracket_depth = 0; *cp; cp++) {
1329 if (*cp == '{')
1330 bracket_depth++;
1331 else if (bracket_depth && *cp == '}')
1332 bracket_depth--;
1333 else if (!bracket_depth && *cp == ':')
1334 break;
1336 if (*cp == ':') {
1337 unsigned char tree_sha1[20];
1338 int len = cp - name;
1339 if (!get_sha1_1(name, len, tree_sha1, GET_SHA1_TREEISH)) {
1340 const char *filename = cp+1;
1341 char *new_filename = NULL;
1343 new_filename = resolve_relative_path(filename);
1344 if (new_filename)
1345 filename = new_filename;
1346 ret = get_tree_entry(tree_sha1, filename, sha1, &oc->mode);
1347 if (ret && only_to_die) {
1348 diagnose_invalid_sha1_path(prefix, filename,
1349 tree_sha1,
1350 name, len);
1352 hashcpy(oc->tree, tree_sha1);
1353 strncpy(oc->path, filename,
1354 sizeof(oc->path));
1355 oc->path[sizeof(oc->path)-1] = '\0';
1357 free(new_filename);
1358 return ret;
1359 } else {
1360 if (only_to_die)
1361 die("Invalid object name '%.*s'.", len, name);
1364 return ret;
1368 * Call this function when you know "name" given by the end user must
1369 * name an object but it doesn't; the function _may_ die with a better
1370 * diagnostic message than "no such object 'name'", e.g. "Path 'doc' does not
1371 * exist in 'HEAD'" when given "HEAD:doc", or it may return in which case
1372 * you have a chance to diagnose the error further.
1374 void maybe_die_on_misspelt_object_name(const char *name, const char *prefix)
1376 struct object_context oc;
1377 unsigned char sha1[20];
1378 get_sha1_with_context_1(name, GET_SHA1_ONLY_TO_DIE, prefix, sha1, &oc);
1381 int get_sha1_with_context(const char *str, unsigned flags, unsigned char *sha1, struct object_context *orc)
1383 return get_sha1_with_context_1(str, flags, NULL, sha1, orc);