builtin/apply: move 'state' init into init_apply_state()
[git/gitweb.git] / sha1_name.c
blobca7ddd6f2cf7074a1d74e011c2bd260c2ae6ffe6
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"
9 #include "dir.h"
11 static int get_sha1_oneline(const char *, unsigned char *, struct commit_list *);
13 typedef int (*disambiguate_hint_fn)(const unsigned char *, void *);
15 struct disambiguate_state {
16 disambiguate_hint_fn fn;
17 void *cb_data;
18 unsigned char candidate[20];
19 unsigned candidate_exists:1;
20 unsigned candidate_checked:1;
21 unsigned candidate_ok:1;
22 unsigned disambiguate_fn_used:1;
23 unsigned ambiguous:1;
24 unsigned always_call_fn:1;
27 static void update_candidates(struct disambiguate_state *ds, const unsigned char *current)
29 if (ds->always_call_fn) {
30 ds->ambiguous = ds->fn(current, ds->cb_data) ? 1 : 0;
31 return;
33 if (!ds->candidate_exists) {
34 /* this is the first candidate */
35 hashcpy(ds->candidate, current);
36 ds->candidate_exists = 1;
37 return;
38 } else if (!hashcmp(ds->candidate, current)) {
39 /* the same as what we already have seen */
40 return;
43 if (!ds->fn) {
44 /* cannot disambiguate between ds->candidate and current */
45 ds->ambiguous = 1;
46 return;
49 if (!ds->candidate_checked) {
50 ds->candidate_ok = ds->fn(ds->candidate, ds->cb_data);
51 ds->disambiguate_fn_used = 1;
52 ds->candidate_checked = 1;
55 if (!ds->candidate_ok) {
56 /* discard the candidate; we know it does not satisfy fn */
57 hashcpy(ds->candidate, current);
58 ds->candidate_checked = 0;
59 return;
62 /* if we reach this point, we know ds->candidate satisfies fn */
63 if (ds->fn(current, ds->cb_data)) {
65 * if both current and candidate satisfy fn, we cannot
66 * disambiguate.
68 ds->candidate_ok = 0;
69 ds->ambiguous = 1;
72 /* otherwise, current can be discarded and candidate is still good */
75 static void find_short_object_filename(int len, const char *hex_pfx, struct disambiguate_state *ds)
77 struct alternate_object_database *alt;
78 char hex[40];
79 static struct alternate_object_database *fakeent;
81 if (!fakeent) {
83 * Create a "fake" alternate object database that
84 * points to our own object database, to make it
85 * easier to get a temporary working space in
86 * alt->name/alt->base while iterating over the
87 * object databases including our own.
89 const char *objdir = get_object_directory();
90 size_t objdir_len = strlen(objdir);
91 fakeent = xmalloc(st_add3(sizeof(*fakeent), objdir_len, 43));
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 xsnprintf(hex, sizeof(hex), "%.2s", hex_pfx);
99 for (alt = fakeent; alt && !ds->ambiguous; alt = alt->next) {
100 struct dirent *de;
101 DIR *dir;
103 * every alt_odb struct has 42 extra bytes after the base
104 * for exactly this purpose
106 xsnprintf(alt->name, 42, "%.2s/", hex_pfx);
107 dir = opendir(alt->base);
108 if (!dir)
109 continue;
111 while (!ds->ambiguous && (de = readdir(dir)) != NULL) {
112 unsigned char sha1[20];
114 if (strlen(de->d_name) != 38)
115 continue;
116 if (memcmp(de->d_name, hex_pfx + 2, len - 2))
117 continue;
118 memcpy(hex + 2, de->d_name, 38);
119 if (!get_sha1_hex(hex, sha1))
120 update_candidates(ds, sha1);
122 closedir(dir);
126 static int match_sha(unsigned len, const unsigned char *a, const unsigned char *b)
128 do {
129 if (*a != *b)
130 return 0;
131 a++;
132 b++;
133 len -= 2;
134 } while (len > 1);
135 if (len)
136 if ((*a ^ *b) & 0xf0)
137 return 0;
138 return 1;
141 static void unique_in_pack(int len,
142 const unsigned char *bin_pfx,
143 struct packed_git *p,
144 struct disambiguate_state *ds)
146 uint32_t num, last, i, first = 0;
147 const unsigned char *current = NULL;
149 open_pack_index(p);
150 num = p->num_objects;
151 last = num;
152 while (first < last) {
153 uint32_t mid = (first + last) / 2;
154 const unsigned char *current;
155 int cmp;
157 current = nth_packed_object_sha1(p, mid);
158 cmp = hashcmp(bin_pfx, current);
159 if (!cmp) {
160 first = mid;
161 break;
163 if (cmp > 0) {
164 first = mid+1;
165 continue;
167 last = mid;
171 * At this point, "first" is the location of the lowest object
172 * with an object name that could match "bin_pfx". See if we have
173 * 0, 1 or more objects that actually match(es).
175 for (i = first; i < num && !ds->ambiguous; i++) {
176 current = nth_packed_object_sha1(p, i);
177 if (!match_sha(len, bin_pfx, current))
178 break;
179 update_candidates(ds, current);
183 static void find_short_packed_object(int len, const unsigned char *bin_pfx,
184 struct disambiguate_state *ds)
186 struct packed_git *p;
188 prepare_packed_git();
189 for (p = packed_git; p && !ds->ambiguous; p = p->next)
190 unique_in_pack(len, bin_pfx, p, ds);
193 #define SHORT_NAME_NOT_FOUND (-1)
194 #define SHORT_NAME_AMBIGUOUS (-2)
196 static int finish_object_disambiguation(struct disambiguate_state *ds,
197 unsigned char *sha1)
199 if (ds->ambiguous)
200 return SHORT_NAME_AMBIGUOUS;
202 if (!ds->candidate_exists)
203 return SHORT_NAME_NOT_FOUND;
205 if (!ds->candidate_checked)
207 * If this is the only candidate, there is no point
208 * calling the disambiguation hint callback.
210 * On the other hand, if the current candidate
211 * replaced an earlier candidate that did _not_ pass
212 * the disambiguation hint callback, then we do have
213 * more than one objects that match the short name
214 * given, so we should make sure this one matches;
215 * otherwise, if we discovered this one and the one
216 * that we previously discarded in the reverse order,
217 * we would end up showing different results in the
218 * same repository!
220 ds->candidate_ok = (!ds->disambiguate_fn_used ||
221 ds->fn(ds->candidate, ds->cb_data));
223 if (!ds->candidate_ok)
224 return SHORT_NAME_AMBIGUOUS;
226 hashcpy(sha1, ds->candidate);
227 return 0;
230 static int disambiguate_commit_only(const unsigned char *sha1, void *cb_data_unused)
232 int kind = sha1_object_info(sha1, NULL);
233 return kind == OBJ_COMMIT;
236 static int disambiguate_committish_only(const unsigned char *sha1, void *cb_data_unused)
238 struct object *obj;
239 int kind;
241 kind = sha1_object_info(sha1, NULL);
242 if (kind == OBJ_COMMIT)
243 return 1;
244 if (kind != OBJ_TAG)
245 return 0;
247 /* We need to do this the hard way... */
248 obj = deref_tag(parse_object(sha1), NULL, 0);
249 if (obj && obj->type == OBJ_COMMIT)
250 return 1;
251 return 0;
254 static int disambiguate_tree_only(const unsigned char *sha1, void *cb_data_unused)
256 int kind = sha1_object_info(sha1, NULL);
257 return kind == OBJ_TREE;
260 static int disambiguate_treeish_only(const unsigned char *sha1, void *cb_data_unused)
262 struct object *obj;
263 int kind;
265 kind = sha1_object_info(sha1, NULL);
266 if (kind == OBJ_TREE || kind == OBJ_COMMIT)
267 return 1;
268 if (kind != OBJ_TAG)
269 return 0;
271 /* We need to do this the hard way... */
272 obj = deref_tag(lookup_object(sha1), NULL, 0);
273 if (obj && (obj->type == OBJ_TREE || obj->type == OBJ_COMMIT))
274 return 1;
275 return 0;
278 static int disambiguate_blob_only(const unsigned char *sha1, void *cb_data_unused)
280 int kind = sha1_object_info(sha1, NULL);
281 return kind == OBJ_BLOB;
284 static int prepare_prefixes(const char *name, int len,
285 unsigned char *bin_pfx,
286 char *hex_pfx)
288 int i;
290 hashclr(bin_pfx);
291 memset(hex_pfx, 'x', 40);
292 for (i = 0; i < len ;i++) {
293 unsigned char c = name[i];
294 unsigned char val;
295 if (c >= '0' && c <= '9')
296 val = c - '0';
297 else if (c >= 'a' && c <= 'f')
298 val = c - 'a' + 10;
299 else if (c >= 'A' && c <='F') {
300 val = c - 'A' + 10;
301 c -= 'A' - 'a';
303 else
304 return -1;
305 hex_pfx[i] = c;
306 if (!(i & 1))
307 val <<= 4;
308 bin_pfx[i >> 1] |= val;
310 return 0;
313 static int get_short_sha1(const char *name, int len, unsigned char *sha1,
314 unsigned flags)
316 int status;
317 char hex_pfx[40];
318 unsigned char bin_pfx[20];
319 struct disambiguate_state ds;
320 int quietly = !!(flags & GET_SHA1_QUIETLY);
322 if (len < MINIMUM_ABBREV || len > 40)
323 return -1;
324 if (prepare_prefixes(name, len, bin_pfx, hex_pfx) < 0)
325 return -1;
327 prepare_alt_odb();
329 memset(&ds, 0, sizeof(ds));
330 if (flags & GET_SHA1_COMMIT)
331 ds.fn = disambiguate_commit_only;
332 else if (flags & GET_SHA1_COMMITTISH)
333 ds.fn = disambiguate_committish_only;
334 else if (flags & GET_SHA1_TREE)
335 ds.fn = disambiguate_tree_only;
336 else if (flags & GET_SHA1_TREEISH)
337 ds.fn = disambiguate_treeish_only;
338 else if (flags & GET_SHA1_BLOB)
339 ds.fn = disambiguate_blob_only;
341 find_short_object_filename(len, hex_pfx, &ds);
342 find_short_packed_object(len, bin_pfx, &ds);
343 status = finish_object_disambiguation(&ds, sha1);
345 if (!quietly && (status == SHORT_NAME_AMBIGUOUS))
346 return error("short SHA1 %.*s is ambiguous.", len, hex_pfx);
347 return status;
350 int for_each_abbrev(const char *prefix, each_abbrev_fn fn, void *cb_data)
352 char hex_pfx[40];
353 unsigned char bin_pfx[20];
354 struct disambiguate_state ds;
355 int len = strlen(prefix);
357 if (len < MINIMUM_ABBREV || len > 40)
358 return -1;
359 if (prepare_prefixes(prefix, len, bin_pfx, hex_pfx) < 0)
360 return -1;
362 prepare_alt_odb();
364 memset(&ds, 0, sizeof(ds));
365 ds.always_call_fn = 1;
366 ds.cb_data = cb_data;
367 ds.fn = fn;
369 find_short_object_filename(len, hex_pfx, &ds);
370 find_short_packed_object(len, bin_pfx, &ds);
371 return ds.ambiguous;
374 int find_unique_abbrev_r(char *hex, const unsigned char *sha1, int len)
376 int status, exists;
378 sha1_to_hex_r(hex, sha1);
379 if (len == 40 || !len)
380 return 40;
381 exists = has_sha1_file(sha1);
382 while (len < 40) {
383 unsigned char sha1_ret[20];
384 status = get_short_sha1(hex, len, sha1_ret, GET_SHA1_QUIETLY);
385 if (exists
386 ? !status
387 : status == SHORT_NAME_NOT_FOUND) {
388 hex[len] = 0;
389 return len;
391 len++;
393 return len;
396 const char *find_unique_abbrev(const unsigned char *sha1, int len)
398 static char hex[GIT_SHA1_HEXSZ + 1];
399 find_unique_abbrev_r(hex, sha1, len);
400 return hex;
403 static int ambiguous_path(const char *path, int len)
405 int slash = 1;
406 int cnt;
408 for (cnt = 0; cnt < len; cnt++) {
409 switch (*path++) {
410 case '\0':
411 break;
412 case '/':
413 if (slash)
414 break;
415 slash = 1;
416 continue;
417 case '.':
418 continue;
419 default:
420 slash = 0;
421 continue;
423 break;
425 return slash;
428 static inline int at_mark(const char *string, int len,
429 const char **suffix, int nr)
431 int i;
433 for (i = 0; i < nr; i++) {
434 int suffix_len = strlen(suffix[i]);
435 if (suffix_len <= len
436 && !memcmp(string, suffix[i], suffix_len))
437 return suffix_len;
439 return 0;
442 static inline int upstream_mark(const char *string, int len)
444 const char *suffix[] = { "@{upstream}", "@{u}" };
445 return at_mark(string, len, suffix, ARRAY_SIZE(suffix));
448 static inline int push_mark(const char *string, int len)
450 const char *suffix[] = { "@{push}" };
451 return at_mark(string, len, suffix, ARRAY_SIZE(suffix));
454 static int get_sha1_1(const char *name, int len, unsigned char *sha1, unsigned lookup_flags);
455 static int interpret_nth_prior_checkout(const char *name, int namelen, struct strbuf *buf);
457 static int get_sha1_basic(const char *str, int len, unsigned char *sha1,
458 unsigned int flags)
460 static const char *warn_msg = "refname '%.*s' is ambiguous.";
461 static const char *object_name_msg = N_(
462 "Git normally never creates a ref that ends with 40 hex characters\n"
463 "because it will be ignored when you just specify 40-hex. These refs\n"
464 "may be created by mistake. For example,\n"
465 "\n"
466 " git checkout -b $br $(git rev-parse ...)\n"
467 "\n"
468 "where \"$br\" is somehow empty and a 40-hex ref is created. Please\n"
469 "examine these refs and maybe delete them. Turn this message off by\n"
470 "running \"git config advice.objectNameWarning false\"");
471 unsigned char tmp_sha1[20];
472 char *real_ref = NULL;
473 int refs_found = 0;
474 int at, reflog_len, nth_prior = 0;
476 if (len == 40 && !get_sha1_hex(str, sha1)) {
477 if (warn_ambiguous_refs && warn_on_object_refname_ambiguity) {
478 refs_found = dwim_ref(str, len, tmp_sha1, &real_ref);
479 if (refs_found > 0) {
480 warning(warn_msg, len, str);
481 if (advice_object_name_warning)
482 fprintf(stderr, "%s\n", _(object_name_msg));
484 free(real_ref);
486 return 0;
489 /* basic@{time or number or -number} format to query ref-log */
490 reflog_len = at = 0;
491 if (len && str[len-1] == '}') {
492 for (at = len-4; at >= 0; at--) {
493 if (str[at] == '@' && str[at+1] == '{') {
494 if (str[at+2] == '-') {
495 if (at != 0)
496 /* @{-N} not at start */
497 return -1;
498 nth_prior = 1;
499 continue;
501 if (!upstream_mark(str + at, len - at) &&
502 !push_mark(str + at, len - at)) {
503 reflog_len = (len-1) - (at+2);
504 len = at;
506 break;
511 /* Accept only unambiguous ref paths. */
512 if (len && ambiguous_path(str, len))
513 return -1;
515 if (nth_prior) {
516 struct strbuf buf = STRBUF_INIT;
517 int detached;
519 if (interpret_nth_prior_checkout(str, len, &buf) > 0) {
520 detached = (buf.len == 40 && !get_sha1_hex(buf.buf, sha1));
521 strbuf_release(&buf);
522 if (detached)
523 return 0;
527 if (!len && reflog_len)
528 /* allow "@{...}" to mean the current branch reflog */
529 refs_found = dwim_ref("HEAD", 4, sha1, &real_ref);
530 else if (reflog_len)
531 refs_found = dwim_log(str, len, sha1, &real_ref);
532 else
533 refs_found = dwim_ref(str, len, sha1, &real_ref);
535 if (!refs_found)
536 return -1;
538 if (warn_ambiguous_refs && !(flags & GET_SHA1_QUIETLY) &&
539 (refs_found > 1 ||
540 !get_short_sha1(str, len, tmp_sha1, GET_SHA1_QUIETLY)))
541 warning(warn_msg, len, str);
543 if (reflog_len) {
544 int nth, i;
545 unsigned long at_time;
546 unsigned long co_time;
547 int co_tz, co_cnt;
549 /* Is it asking for N-th entry, or approxidate? */
550 for (i = nth = 0; 0 <= nth && i < reflog_len; i++) {
551 char ch = str[at+2+i];
552 if ('0' <= ch && ch <= '9')
553 nth = nth * 10 + ch - '0';
554 else
555 nth = -1;
557 if (100000000 <= nth) {
558 at_time = nth;
559 nth = -1;
560 } else if (0 <= nth)
561 at_time = 0;
562 else {
563 int errors = 0;
564 char *tmp = xstrndup(str + at + 2, reflog_len);
565 at_time = approxidate_careful(tmp, &errors);
566 free(tmp);
567 if (errors) {
568 free(real_ref);
569 return -1;
572 if (read_ref_at(real_ref, flags, at_time, nth, sha1, NULL,
573 &co_time, &co_tz, &co_cnt)) {
574 if (!len) {
575 if (starts_with(real_ref, "refs/heads/")) {
576 str = real_ref + 11;
577 len = strlen(real_ref + 11);
578 } else {
579 /* detached HEAD */
580 str = "HEAD";
581 len = 4;
584 if (at_time) {
585 if (!(flags & GET_SHA1_QUIETLY)) {
586 warning("Log for '%.*s' only goes "
587 "back to %s.", len, str,
588 show_date(co_time, co_tz, DATE_MODE(RFC2822)));
590 } else {
591 if (flags & GET_SHA1_QUIETLY) {
592 exit(128);
594 die("Log for '%.*s' only has %d entries.",
595 len, str, co_cnt);
600 free(real_ref);
601 return 0;
604 static int get_parent(const char *name, int len,
605 unsigned char *result, int idx)
607 unsigned char sha1[20];
608 int ret = get_sha1_1(name, len, sha1, GET_SHA1_COMMITTISH);
609 struct commit *commit;
610 struct commit_list *p;
612 if (ret)
613 return ret;
614 commit = lookup_commit_reference(sha1);
615 if (parse_commit(commit))
616 return -1;
617 if (!idx) {
618 hashcpy(result, commit->object.oid.hash);
619 return 0;
621 p = commit->parents;
622 while (p) {
623 if (!--idx) {
624 hashcpy(result, p->item->object.oid.hash);
625 return 0;
627 p = p->next;
629 return -1;
632 static int get_nth_ancestor(const char *name, int len,
633 unsigned char *result, int generation)
635 unsigned char sha1[20];
636 struct commit *commit;
637 int ret;
639 ret = get_sha1_1(name, len, sha1, GET_SHA1_COMMITTISH);
640 if (ret)
641 return ret;
642 commit = lookup_commit_reference(sha1);
643 if (!commit)
644 return -1;
646 while (generation--) {
647 if (parse_commit(commit) || !commit->parents)
648 return -1;
649 commit = commit->parents->item;
651 hashcpy(result, commit->object.oid.hash);
652 return 0;
655 struct object *peel_to_type(const char *name, int namelen,
656 struct object *o, enum object_type expected_type)
658 if (name && !namelen)
659 namelen = strlen(name);
660 while (1) {
661 if (!o || (!o->parsed && !parse_object(o->oid.hash)))
662 return NULL;
663 if (expected_type == OBJ_ANY || o->type == expected_type)
664 return o;
665 if (o->type == OBJ_TAG)
666 o = ((struct tag*) o)->tagged;
667 else if (o->type == OBJ_COMMIT)
668 o = &(((struct commit *) o)->tree->object);
669 else {
670 if (name)
671 error("%.*s: expected %s type, but the object "
672 "dereferences to %s type",
673 namelen, name, typename(expected_type),
674 typename(o->type));
675 return NULL;
680 static int peel_onion(const char *name, int len, unsigned char *sha1)
682 unsigned char outer[20];
683 const char *sp;
684 unsigned int expected_type = 0;
685 unsigned lookup_flags = 0;
686 struct object *o;
689 * "ref^{type}" dereferences ref repeatedly until you cannot
690 * dereference anymore, or you get an object of given type,
691 * whichever comes first. "ref^{}" means just dereference
692 * tags until you get a non-tag. "ref^0" is a shorthand for
693 * "ref^{commit}". "commit^{tree}" could be used to find the
694 * top-level tree of the given commit.
696 if (len < 4 || name[len-1] != '}')
697 return -1;
699 for (sp = name + len - 1; name <= sp; sp--) {
700 int ch = *sp;
701 if (ch == '{' && name < sp && sp[-1] == '^')
702 break;
704 if (sp <= name)
705 return -1;
707 sp++; /* beginning of type name, or closing brace for empty */
708 if (starts_with(sp, "commit}"))
709 expected_type = OBJ_COMMIT;
710 else if (starts_with(sp, "tag}"))
711 expected_type = OBJ_TAG;
712 else if (starts_with(sp, "tree}"))
713 expected_type = OBJ_TREE;
714 else if (starts_with(sp, "blob}"))
715 expected_type = OBJ_BLOB;
716 else if (starts_with(sp, "object}"))
717 expected_type = OBJ_ANY;
718 else if (sp[0] == '}')
719 expected_type = OBJ_NONE;
720 else if (sp[0] == '/')
721 expected_type = OBJ_COMMIT;
722 else
723 return -1;
725 if (expected_type == OBJ_COMMIT)
726 lookup_flags = GET_SHA1_COMMITTISH;
727 else if (expected_type == OBJ_TREE)
728 lookup_flags = GET_SHA1_TREEISH;
730 if (get_sha1_1(name, sp - name - 2, outer, lookup_flags))
731 return -1;
733 o = parse_object(outer);
734 if (!o)
735 return -1;
736 if (!expected_type) {
737 o = deref_tag(o, name, sp - name - 2);
738 if (!o || (!o->parsed && !parse_object(o->oid.hash)))
739 return -1;
740 hashcpy(sha1, o->oid.hash);
741 return 0;
745 * At this point, the syntax look correct, so
746 * if we do not get the needed object, we should
747 * barf.
749 o = peel_to_type(name, len, o, expected_type);
750 if (!o)
751 return -1;
753 hashcpy(sha1, o->oid.hash);
754 if (sp[0] == '/') {
755 /* "$commit^{/foo}" */
756 char *prefix;
757 int ret;
758 struct commit_list *list = NULL;
761 * $commit^{/}. Some regex implementation may reject.
762 * We don't need regex anyway. '' pattern always matches.
764 if (sp[1] == '}')
765 return 0;
767 prefix = xstrndup(sp + 1, name + len - 1 - (sp + 1));
768 commit_list_insert((struct commit *)o, &list);
769 ret = get_sha1_oneline(prefix, sha1, list);
770 free(prefix);
771 return ret;
773 return 0;
776 static int get_describe_name(const char *name, int len, unsigned char *sha1)
778 const char *cp;
779 unsigned flags = GET_SHA1_QUIETLY | GET_SHA1_COMMIT;
781 for (cp = name + len - 1; name + 2 <= cp; cp--) {
782 char ch = *cp;
783 if (!isxdigit(ch)) {
784 /* We must be looking at g in "SOMETHING-g"
785 * for it to be describe output.
787 if (ch == 'g' && cp[-1] == '-') {
788 cp++;
789 len -= cp - name;
790 return get_short_sha1(cp, len, sha1, flags);
794 return -1;
797 static int get_sha1_1(const char *name, int len, unsigned char *sha1, unsigned lookup_flags)
799 int ret, has_suffix;
800 const char *cp;
803 * "name~3" is "name^^^", "name~" is "name~1", and "name^" is "name^1".
805 has_suffix = 0;
806 for (cp = name + len - 1; name <= cp; cp--) {
807 int ch = *cp;
808 if ('0' <= ch && ch <= '9')
809 continue;
810 if (ch == '~' || ch == '^')
811 has_suffix = ch;
812 break;
815 if (has_suffix) {
816 int num = 0;
817 int len1 = cp - name;
818 cp++;
819 while (cp < name + len)
820 num = num * 10 + *cp++ - '0';
821 if (!num && len1 == len - 1)
822 num = 1;
823 if (has_suffix == '^')
824 return get_parent(name, len1, sha1, num);
825 /* else if (has_suffix == '~') -- goes without saying */
826 return get_nth_ancestor(name, len1, sha1, num);
829 ret = peel_onion(name, len, sha1);
830 if (!ret)
831 return 0;
833 ret = get_sha1_basic(name, len, sha1, lookup_flags);
834 if (!ret)
835 return 0;
837 /* It could be describe output that is "SOMETHING-gXXXX" */
838 ret = get_describe_name(name, len, sha1);
839 if (!ret)
840 return 0;
842 return get_short_sha1(name, len, sha1, lookup_flags);
846 * This interprets names like ':/Initial revision of "git"' by searching
847 * through history and returning the first commit whose message starts
848 * the given regular expression.
850 * For negative-matching, prefix the pattern-part with '!-', like: ':/!-WIP'.
852 * For a literal '!' character at the beginning of a pattern, you have to repeat
853 * that, like: ':/!!foo'
855 * For future extension, all other sequences beginning with ':/!' are reserved.
858 /* Remember to update object flag allocation in object.h */
859 #define ONELINE_SEEN (1u<<20)
861 static int handle_one_ref(const char *path, const struct object_id *oid,
862 int flag, void *cb_data)
864 struct commit_list **list = cb_data;
865 struct object *object = parse_object(oid->hash);
866 if (!object)
867 return 0;
868 if (object->type == OBJ_TAG) {
869 object = deref_tag(object, path, strlen(path));
870 if (!object)
871 return 0;
873 if (object->type != OBJ_COMMIT)
874 return 0;
875 commit_list_insert((struct commit *)object, list);
876 return 0;
879 static int get_sha1_oneline(const char *prefix, unsigned char *sha1,
880 struct commit_list *list)
882 struct commit_list *backup = NULL, *l;
883 int found = 0;
884 int negative = 0;
885 regex_t regex;
887 if (prefix[0] == '!') {
888 prefix++;
890 if (prefix[0] == '-') {
891 prefix++;
892 negative = 1;
893 } else if (prefix[0] != '!') {
894 return -1;
898 if (regcomp(&regex, prefix, REG_EXTENDED))
899 return -1;
901 for (l = list; l; l = l->next) {
902 l->item->object.flags |= ONELINE_SEEN;
903 commit_list_insert(l->item, &backup);
905 while (list) {
906 const char *p, *buf;
907 struct commit *commit;
908 int matches;
910 commit = pop_most_recent_commit(&list, ONELINE_SEEN);
911 if (!parse_object(commit->object.oid.hash))
912 continue;
913 buf = get_commit_buffer(commit, NULL);
914 p = strstr(buf, "\n\n");
915 matches = negative ^ (p && !regexec(&regex, p + 2, 0, NULL, 0));
916 unuse_commit_buffer(commit, buf);
918 if (matches) {
919 hashcpy(sha1, commit->object.oid.hash);
920 found = 1;
921 break;
924 regfree(&regex);
925 free_commit_list(list);
926 for (l = backup; l; l = l->next)
927 clear_commit_marks(l->item, ONELINE_SEEN);
928 free_commit_list(backup);
929 return found ? 0 : -1;
932 struct grab_nth_branch_switch_cbdata {
933 int remaining;
934 struct strbuf buf;
937 static int grab_nth_branch_switch(unsigned char *osha1, unsigned char *nsha1,
938 const char *email, unsigned long timestamp, int tz,
939 const char *message, void *cb_data)
941 struct grab_nth_branch_switch_cbdata *cb = cb_data;
942 const char *match = NULL, *target = NULL;
943 size_t len;
945 if (skip_prefix(message, "checkout: moving from ", &match))
946 target = strstr(match, " to ");
948 if (!match || !target)
949 return 0;
950 if (--(cb->remaining) == 0) {
951 len = target - match;
952 strbuf_reset(&cb->buf);
953 strbuf_add(&cb->buf, match, len);
954 return 1; /* we are done */
956 return 0;
960 * Parse @{-N} syntax, return the number of characters parsed
961 * if successful; otherwise signal an error with negative value.
963 static int interpret_nth_prior_checkout(const char *name, int namelen,
964 struct strbuf *buf)
966 long nth;
967 int retval;
968 struct grab_nth_branch_switch_cbdata cb;
969 const char *brace;
970 char *num_end;
972 if (namelen < 4)
973 return -1;
974 if (name[0] != '@' || name[1] != '{' || name[2] != '-')
975 return -1;
976 brace = memchr(name, '}', namelen);
977 if (!brace)
978 return -1;
979 nth = strtol(name + 3, &num_end, 10);
980 if (num_end != brace)
981 return -1;
982 if (nth <= 0)
983 return -1;
984 cb.remaining = nth;
985 strbuf_init(&cb.buf, 20);
987 retval = 0;
988 if (0 < for_each_reflog_ent_reverse("HEAD", grab_nth_branch_switch, &cb)) {
989 strbuf_reset(buf);
990 strbuf_addbuf(buf, &cb.buf);
991 retval = brace - name + 1;
994 strbuf_release(&cb.buf);
995 return retval;
998 int get_sha1_mb(const char *name, unsigned char *sha1)
1000 struct commit *one, *two;
1001 struct commit_list *mbs;
1002 unsigned char sha1_tmp[20];
1003 const char *dots;
1004 int st;
1006 dots = strstr(name, "...");
1007 if (!dots)
1008 return get_sha1(name, sha1);
1009 if (dots == name)
1010 st = get_sha1("HEAD", sha1_tmp);
1011 else {
1012 struct strbuf sb;
1013 strbuf_init(&sb, dots - name);
1014 strbuf_add(&sb, name, dots - name);
1015 st = get_sha1_committish(sb.buf, sha1_tmp);
1016 strbuf_release(&sb);
1018 if (st)
1019 return st;
1020 one = lookup_commit_reference_gently(sha1_tmp, 0);
1021 if (!one)
1022 return -1;
1024 if (get_sha1_committish(dots[3] ? (dots + 3) : "HEAD", sha1_tmp))
1025 return -1;
1026 two = lookup_commit_reference_gently(sha1_tmp, 0);
1027 if (!two)
1028 return -1;
1029 mbs = get_merge_bases(one, two);
1030 if (!mbs || mbs->next)
1031 st = -1;
1032 else {
1033 st = 0;
1034 hashcpy(sha1, mbs->item->object.oid.hash);
1036 free_commit_list(mbs);
1037 return st;
1040 /* parse @something syntax, when 'something' is not {.*} */
1041 static int interpret_empty_at(const char *name, int namelen, int len, struct strbuf *buf)
1043 const char *next;
1045 if (len || name[1] == '{')
1046 return -1;
1048 /* make sure it's a single @, or @@{.*}, not @foo */
1049 next = memchr(name + len + 1, '@', namelen - len - 1);
1050 if (next && next[1] != '{')
1051 return -1;
1052 if (!next)
1053 next = name + namelen;
1054 if (next != name + 1)
1055 return -1;
1057 strbuf_reset(buf);
1058 strbuf_add(buf, "HEAD", 4);
1059 return 1;
1062 static int reinterpret(const char *name, int namelen, int len, struct strbuf *buf)
1064 /* we have extra data, which might need further processing */
1065 struct strbuf tmp = STRBUF_INIT;
1066 int used = buf->len;
1067 int ret;
1069 strbuf_add(buf, name + len, namelen - len);
1070 ret = interpret_branch_name(buf->buf, buf->len, &tmp);
1071 /* that data was not interpreted, remove our cruft */
1072 if (ret < 0) {
1073 strbuf_setlen(buf, used);
1074 return len;
1076 strbuf_reset(buf);
1077 strbuf_addbuf(buf, &tmp);
1078 strbuf_release(&tmp);
1079 /* tweak for size of {-N} versus expanded ref name */
1080 return ret - used + len;
1083 static void set_shortened_ref(struct strbuf *buf, const char *ref)
1085 char *s = shorten_unambiguous_ref(ref, 0);
1086 strbuf_reset(buf);
1087 strbuf_addstr(buf, s);
1088 free(s);
1091 static int interpret_branch_mark(const char *name, int namelen,
1092 int at, struct strbuf *buf,
1093 int (*get_mark)(const char *, int),
1094 const char *(*get_data)(struct branch *,
1095 struct strbuf *))
1097 int len;
1098 struct branch *branch;
1099 struct strbuf err = STRBUF_INIT;
1100 const char *value;
1102 len = get_mark(name + at, namelen - at);
1103 if (!len)
1104 return -1;
1106 if (memchr(name, ':', at))
1107 return -1;
1109 if (at) {
1110 char *name_str = xmemdupz(name, at);
1111 branch = branch_get(name_str);
1112 free(name_str);
1113 } else
1114 branch = branch_get(NULL);
1116 value = get_data(branch, &err);
1117 if (!value)
1118 die("%s", err.buf);
1120 set_shortened_ref(buf, value);
1121 return len + at;
1125 * This reads short-hand syntax that not only evaluates to a commit
1126 * object name, but also can act as if the end user spelled the name
1127 * of the branch from the command line.
1129 * - "@{-N}" finds the name of the Nth previous branch we were on, and
1130 * places the name of the branch in the given buf and returns the
1131 * number of characters parsed if successful.
1133 * - "<branch>@{upstream}" finds the name of the other ref that
1134 * <branch> is configured to merge with (missing <branch> defaults
1135 * to the current branch), and places the name of the branch in the
1136 * given buf and returns the number of characters parsed if
1137 * successful.
1139 * If the input is not of the accepted format, it returns a negative
1140 * number to signal an error.
1142 * If the input was ok but there are not N branch switches in the
1143 * reflog, it returns 0.
1145 int interpret_branch_name(const char *name, int namelen, struct strbuf *buf)
1147 char *at;
1148 const char *start;
1149 int len = interpret_nth_prior_checkout(name, namelen, buf);
1151 if (!namelen)
1152 namelen = strlen(name);
1154 if (!len) {
1155 return len; /* syntax Ok, not enough switches */
1156 } else if (len > 0) {
1157 if (len == namelen)
1158 return len; /* consumed all */
1159 else
1160 return reinterpret(name, namelen, len, buf);
1163 for (start = name;
1164 (at = memchr(start, '@', namelen - (start - name)));
1165 start = at + 1) {
1167 len = interpret_empty_at(name, namelen, at - name, buf);
1168 if (len > 0)
1169 return reinterpret(name, namelen, len, buf);
1171 len = interpret_branch_mark(name, namelen, at - name, buf,
1172 upstream_mark, branch_get_upstream);
1173 if (len > 0)
1174 return len;
1176 len = interpret_branch_mark(name, namelen, at - name, buf,
1177 push_mark, branch_get_push);
1178 if (len > 0)
1179 return len;
1182 return -1;
1185 int strbuf_branchname(struct strbuf *sb, const char *name)
1187 int len = strlen(name);
1188 int used = interpret_branch_name(name, len, sb);
1190 if (used == len)
1191 return 0;
1192 if (used < 0)
1193 used = 0;
1194 strbuf_add(sb, name + used, len - used);
1195 return len;
1198 int strbuf_check_branch_ref(struct strbuf *sb, const char *name)
1200 strbuf_branchname(sb, name);
1201 if (name[0] == '-')
1202 return -1;
1203 strbuf_splice(sb, 0, 0, "refs/heads/", 11);
1204 return check_refname_format(sb->buf, 0);
1208 * This is like "get_sha1_basic()", except it allows "sha1 expressions",
1209 * notably "xyz^" for "parent of xyz"
1211 int get_sha1(const char *name, unsigned char *sha1)
1213 struct object_context unused;
1214 return get_sha1_with_context(name, 0, sha1, &unused);
1218 * This is like "get_sha1()", but for struct object_id.
1220 int get_oid(const char *name, struct object_id *oid)
1222 return get_sha1(name, oid->hash);
1227 * Many callers know that the user meant to name a commit-ish by
1228 * syntactical positions where the object name appears. Calling this
1229 * function allows the machinery to disambiguate shorter-than-unique
1230 * abbreviated object names between commit-ish and others.
1232 * Note that this does NOT error out when the named object is not a
1233 * commit-ish. It is merely to give a hint to the disambiguation
1234 * machinery.
1236 int get_sha1_committish(const char *name, unsigned char *sha1)
1238 struct object_context unused;
1239 return get_sha1_with_context(name, GET_SHA1_COMMITTISH,
1240 sha1, &unused);
1243 int get_sha1_treeish(const char *name, unsigned char *sha1)
1245 struct object_context unused;
1246 return get_sha1_with_context(name, GET_SHA1_TREEISH,
1247 sha1, &unused);
1250 int get_sha1_commit(const char *name, unsigned char *sha1)
1252 struct object_context unused;
1253 return get_sha1_with_context(name, GET_SHA1_COMMIT,
1254 sha1, &unused);
1257 int get_sha1_tree(const char *name, unsigned char *sha1)
1259 struct object_context unused;
1260 return get_sha1_with_context(name, GET_SHA1_TREE,
1261 sha1, &unused);
1264 int get_sha1_blob(const char *name, unsigned char *sha1)
1266 struct object_context unused;
1267 return get_sha1_with_context(name, GET_SHA1_BLOB,
1268 sha1, &unused);
1271 /* Must be called only when object_name:filename doesn't exist. */
1272 static void diagnose_invalid_sha1_path(const char *prefix,
1273 const char *filename,
1274 const unsigned char *tree_sha1,
1275 const char *object_name,
1276 int object_name_len)
1278 unsigned char sha1[20];
1279 unsigned mode;
1281 if (!prefix)
1282 prefix = "";
1284 if (file_exists(filename))
1285 die("Path '%s' exists on disk, but not in '%.*s'.",
1286 filename, object_name_len, object_name);
1287 if (errno == ENOENT || errno == ENOTDIR) {
1288 char *fullname = xstrfmt("%s%s", prefix, filename);
1290 if (!get_tree_entry(tree_sha1, fullname,
1291 sha1, &mode)) {
1292 die("Path '%s' exists, but not '%s'.\n"
1293 "Did you mean '%.*s:%s' aka '%.*s:./%s'?",
1294 fullname,
1295 filename,
1296 object_name_len, object_name,
1297 fullname,
1298 object_name_len, object_name,
1299 filename);
1301 die("Path '%s' does not exist in '%.*s'",
1302 filename, object_name_len, object_name);
1306 /* Must be called only when :stage:filename doesn't exist. */
1307 static void diagnose_invalid_index_path(int stage,
1308 const char *prefix,
1309 const char *filename)
1311 const struct cache_entry *ce;
1312 int pos;
1313 unsigned namelen = strlen(filename);
1314 struct strbuf fullname = STRBUF_INIT;
1316 if (!prefix)
1317 prefix = "";
1319 /* Wrong stage number? */
1320 pos = cache_name_pos(filename, namelen);
1321 if (pos < 0)
1322 pos = -pos - 1;
1323 if (pos < active_nr) {
1324 ce = active_cache[pos];
1325 if (ce_namelen(ce) == namelen &&
1326 !memcmp(ce->name, filename, namelen))
1327 die("Path '%s' is in the index, but not at stage %d.\n"
1328 "Did you mean ':%d:%s'?",
1329 filename, stage,
1330 ce_stage(ce), filename);
1333 /* Confusion between relative and absolute filenames? */
1334 strbuf_addstr(&fullname, prefix);
1335 strbuf_addstr(&fullname, filename);
1336 pos = cache_name_pos(fullname.buf, fullname.len);
1337 if (pos < 0)
1338 pos = -pos - 1;
1339 if (pos < active_nr) {
1340 ce = active_cache[pos];
1341 if (ce_namelen(ce) == fullname.len &&
1342 !memcmp(ce->name, fullname.buf, fullname.len))
1343 die("Path '%s' is in the index, but not '%s'.\n"
1344 "Did you mean ':%d:%s' aka ':%d:./%s'?",
1345 fullname.buf, filename,
1346 ce_stage(ce), fullname.buf,
1347 ce_stage(ce), filename);
1350 if (file_exists(filename))
1351 die("Path '%s' exists on disk, but not in the index.", filename);
1352 if (errno == ENOENT || errno == ENOTDIR)
1353 die("Path '%s' does not exist (neither on disk nor in the index).",
1354 filename);
1356 strbuf_release(&fullname);
1360 static char *resolve_relative_path(const char *rel)
1362 if (!starts_with(rel, "./") && !starts_with(rel, "../"))
1363 return NULL;
1365 if (!is_inside_work_tree())
1366 die("relative path syntax can't be used outside working tree.");
1368 /* die() inside prefix_path() if resolved path is outside worktree */
1369 return prefix_path(startup_info->prefix,
1370 startup_info->prefix ? strlen(startup_info->prefix) : 0,
1371 rel);
1374 static int get_sha1_with_context_1(const char *name,
1375 unsigned flags,
1376 const char *prefix,
1377 unsigned char *sha1,
1378 struct object_context *oc)
1380 int ret, bracket_depth;
1381 int namelen = strlen(name);
1382 const char *cp;
1383 int only_to_die = flags & GET_SHA1_ONLY_TO_DIE;
1385 memset(oc, 0, sizeof(*oc));
1386 oc->mode = S_IFINVALID;
1387 ret = get_sha1_1(name, namelen, sha1, flags);
1388 if (!ret)
1389 return ret;
1391 * sha1:path --> object name of path in ent sha1
1392 * :path -> object name of absolute path in index
1393 * :./path -> object name of path relative to cwd in index
1394 * :[0-3]:path -> object name of path in index at stage
1395 * :/foo -> recent commit matching foo
1397 if (name[0] == ':') {
1398 int stage = 0;
1399 const struct cache_entry *ce;
1400 char *new_path = NULL;
1401 int pos;
1402 if (!only_to_die && namelen > 2 && name[1] == '/') {
1403 struct commit_list *list = NULL;
1405 for_each_ref(handle_one_ref, &list);
1406 commit_list_sort_by_date(&list);
1407 return get_sha1_oneline(name + 2, sha1, list);
1409 if (namelen < 3 ||
1410 name[2] != ':' ||
1411 name[1] < '0' || '3' < name[1])
1412 cp = name + 1;
1413 else {
1414 stage = name[1] - '0';
1415 cp = name + 3;
1417 new_path = resolve_relative_path(cp);
1418 if (!new_path) {
1419 namelen = namelen - (cp - name);
1420 } else {
1421 cp = new_path;
1422 namelen = strlen(cp);
1425 strlcpy(oc->path, cp, sizeof(oc->path));
1427 if (!active_cache)
1428 read_cache();
1429 pos = cache_name_pos(cp, namelen);
1430 if (pos < 0)
1431 pos = -pos - 1;
1432 while (pos < active_nr) {
1433 ce = active_cache[pos];
1434 if (ce_namelen(ce) != namelen ||
1435 memcmp(ce->name, cp, namelen))
1436 break;
1437 if (ce_stage(ce) == stage) {
1438 hashcpy(sha1, ce->sha1);
1439 oc->mode = ce->ce_mode;
1440 free(new_path);
1441 return 0;
1443 pos++;
1445 if (only_to_die && name[1] && name[1] != '/')
1446 diagnose_invalid_index_path(stage, prefix, cp);
1447 free(new_path);
1448 return -1;
1450 for (cp = name, bracket_depth = 0; *cp; cp++) {
1451 if (*cp == '{')
1452 bracket_depth++;
1453 else if (bracket_depth && *cp == '}')
1454 bracket_depth--;
1455 else if (!bracket_depth && *cp == ':')
1456 break;
1458 if (*cp == ':') {
1459 unsigned char tree_sha1[20];
1460 int len = cp - name;
1461 if (!get_sha1_1(name, len, tree_sha1, GET_SHA1_TREEISH)) {
1462 const char *filename = cp+1;
1463 char *new_filename = NULL;
1465 new_filename = resolve_relative_path(filename);
1466 if (new_filename)
1467 filename = new_filename;
1468 if (flags & GET_SHA1_FOLLOW_SYMLINKS) {
1469 ret = get_tree_entry_follow_symlinks(tree_sha1,
1470 filename, sha1, &oc->symlink_path,
1471 &oc->mode);
1472 } else {
1473 ret = get_tree_entry(tree_sha1, filename,
1474 sha1, &oc->mode);
1475 if (ret && only_to_die) {
1476 diagnose_invalid_sha1_path(prefix,
1477 filename,
1478 tree_sha1,
1479 name, len);
1482 hashcpy(oc->tree, tree_sha1);
1483 strlcpy(oc->path, filename, sizeof(oc->path));
1485 free(new_filename);
1486 return ret;
1487 } else {
1488 if (only_to_die)
1489 die("Invalid object name '%.*s'.", len, name);
1492 return ret;
1496 * Call this function when you know "name" given by the end user must
1497 * name an object but it doesn't; the function _may_ die with a better
1498 * diagnostic message than "no such object 'name'", e.g. "Path 'doc' does not
1499 * exist in 'HEAD'" when given "HEAD:doc", or it may return in which case
1500 * you have a chance to diagnose the error further.
1502 void maybe_die_on_misspelt_object_name(const char *name, const char *prefix)
1504 struct object_context oc;
1505 unsigned char sha1[20];
1506 get_sha1_with_context_1(name, GET_SHA1_ONLY_TO_DIE, prefix, sha1, &oc);
1509 int get_sha1_with_context(const char *str, unsigned flags, unsigned char *sha1, struct object_context *orc)
1511 if (flags & GET_SHA1_FOLLOW_SYMLINKS && flags & GET_SHA1_ONLY_TO_DIE)
1512 die("BUG: incompatible flags for get_sha1_with_context");
1513 return get_sha1_with_context_1(str, flags, NULL, sha1, orc);