split-index: the writing part
[git/jrn.git] / sha1_name.c
blob2b6322fad064845f4c782286fe9764f6b958f30d
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 satisfy 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(parse_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;
346 int for_each_abbrev(const char *prefix, each_abbrev_fn fn, void *cb_data)
348 char hex_pfx[40];
349 unsigned char bin_pfx[20];
350 struct disambiguate_state ds;
351 int len = strlen(prefix);
353 if (len < MINIMUM_ABBREV || len > 40)
354 return -1;
355 if (prepare_prefixes(prefix, len, bin_pfx, hex_pfx) < 0)
356 return -1;
358 prepare_alt_odb();
360 memset(&ds, 0, sizeof(ds));
361 ds.always_call_fn = 1;
362 ds.cb_data = cb_data;
363 ds.fn = fn;
365 find_short_object_filename(len, hex_pfx, &ds);
366 find_short_packed_object(len, bin_pfx, &ds);
367 return ds.ambiguous;
370 const char *find_unique_abbrev(const unsigned char *sha1, int len)
372 int status, exists;
373 static char hex[41];
375 exists = has_sha1_file(sha1);
376 memcpy(hex, sha1_to_hex(sha1), 40);
377 if (len == 40 || !len)
378 return hex;
379 while (len < 40) {
380 unsigned char sha1_ret[20];
381 status = get_short_sha1(hex, len, sha1_ret, GET_SHA1_QUIETLY);
382 if (exists
383 ? !status
384 : status == SHORT_NAME_NOT_FOUND) {
385 hex[len] = 0;
386 return hex;
388 len++;
390 return hex;
393 static int ambiguous_path(const char *path, int len)
395 int slash = 1;
396 int cnt;
398 for (cnt = 0; cnt < len; cnt++) {
399 switch (*path++) {
400 case '\0':
401 break;
402 case '/':
403 if (slash)
404 break;
405 slash = 1;
406 continue;
407 case '.':
408 continue;
409 default:
410 slash = 0;
411 continue;
413 break;
415 return slash;
418 static inline int upstream_mark(const char *string, int len)
420 const char *suffix[] = { "@{upstream}", "@{u}" };
421 int i;
423 for (i = 0; i < ARRAY_SIZE(suffix); i++) {
424 int suffix_len = strlen(suffix[i]);
425 if (suffix_len <= len
426 && !memcmp(string, suffix[i], suffix_len))
427 return suffix_len;
429 return 0;
432 static int get_sha1_1(const char *name, int len, unsigned char *sha1, unsigned lookup_flags);
433 static int interpret_nth_prior_checkout(const char *name, int namelen, struct strbuf *buf);
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 static const char *object_name_msg = N_(
439 "Git normally never creates a ref that ends with 40 hex characters\n"
440 "because it will be ignored when you just specify 40-hex. These refs\n"
441 "may be created by mistake. For example,\n"
442 "\n"
443 " git checkout -b $br $(git rev-parse ...)\n"
444 "\n"
445 "where \"$br\" is somehow empty and a 40-hex ref is created. Please\n"
446 "examine these refs and maybe delete them. Turn this message off by\n"
447 "running \"git config advice.objectNameWarning false\"");
448 unsigned char tmp_sha1[20];
449 char *real_ref = NULL;
450 int refs_found = 0;
451 int at, reflog_len, nth_prior = 0;
453 if (len == 40 && !get_sha1_hex(str, sha1)) {
454 if (warn_ambiguous_refs && warn_on_object_refname_ambiguity) {
455 refs_found = dwim_ref(str, len, tmp_sha1, &real_ref);
456 if (refs_found > 0) {
457 warning(warn_msg, len, str);
458 if (advice_object_name_warning)
459 fprintf(stderr, "%s\n", _(object_name_msg));
461 free(real_ref);
463 return 0;
466 /* basic@{time or number or -number} format to query ref-log */
467 reflog_len = at = 0;
468 if (len && str[len-1] == '}') {
469 for (at = len-4; at >= 0; at--) {
470 if (str[at] == '@' && str[at+1] == '{') {
471 if (str[at+2] == '-') {
472 if (at != 0)
473 /* @{-N} not at start */
474 return -1;
475 nth_prior = 1;
476 continue;
478 if (!upstream_mark(str + at, len - at)) {
479 reflog_len = (len-1) - (at+2);
480 len = at;
482 break;
487 /* Accept only unambiguous ref paths. */
488 if (len && ambiguous_path(str, len))
489 return -1;
491 if (nth_prior) {
492 struct strbuf buf = STRBUF_INIT;
493 int detached;
495 if (interpret_nth_prior_checkout(str, len, &buf) > 0) {
496 detached = (buf.len == 40 && !get_sha1_hex(buf.buf, sha1));
497 strbuf_release(&buf);
498 if (detached)
499 return 0;
503 if (!len && reflog_len)
504 /* allow "@{...}" to mean the current branch reflog */
505 refs_found = dwim_ref("HEAD", 4, sha1, &real_ref);
506 else if (reflog_len)
507 refs_found = dwim_log(str, len, sha1, &real_ref);
508 else
509 refs_found = dwim_ref(str, len, sha1, &real_ref);
511 if (!refs_found)
512 return -1;
514 if (warn_ambiguous_refs &&
515 (refs_found > 1 ||
516 !get_short_sha1(str, len, tmp_sha1, GET_SHA1_QUIETLY)))
517 warning(warn_msg, len, str);
519 if (reflog_len) {
520 int nth, i;
521 unsigned long at_time;
522 unsigned long co_time;
523 int co_tz, co_cnt;
525 /* Is it asking for N-th entry, or approxidate? */
526 for (i = nth = 0; 0 <= nth && i < reflog_len; i++) {
527 char ch = str[at+2+i];
528 if ('0' <= ch && ch <= '9')
529 nth = nth * 10 + ch - '0';
530 else
531 nth = -1;
533 if (100000000 <= nth) {
534 at_time = nth;
535 nth = -1;
536 } else if (0 <= nth)
537 at_time = 0;
538 else {
539 int errors = 0;
540 char *tmp = xstrndup(str + at + 2, reflog_len);
541 at_time = approxidate_careful(tmp, &errors);
542 free(tmp);
543 if (errors)
544 return -1;
546 if (read_ref_at(real_ref, at_time, nth, sha1, NULL,
547 &co_time, &co_tz, &co_cnt)) {
548 if (!len) {
549 if (starts_with(real_ref, "refs/heads/")) {
550 str = real_ref + 11;
551 len = strlen(real_ref + 11);
552 } else {
553 /* detached HEAD */
554 str = "HEAD";
555 len = 4;
558 if (at_time)
559 warning("Log for '%.*s' only goes "
560 "back to %s.", len, str,
561 show_date(co_time, co_tz, DATE_RFC2822));
562 else {
563 die("Log for '%.*s' only has %d entries.",
564 len, str, co_cnt);
569 free(real_ref);
570 return 0;
573 static int get_parent(const char *name, int len,
574 unsigned char *result, int idx)
576 unsigned char sha1[20];
577 int ret = get_sha1_1(name, len, sha1, GET_SHA1_COMMITTISH);
578 struct commit *commit;
579 struct commit_list *p;
581 if (ret)
582 return ret;
583 commit = lookup_commit_reference(sha1);
584 if (parse_commit(commit))
585 return -1;
586 if (!idx) {
587 hashcpy(result, commit->object.sha1);
588 return 0;
590 p = commit->parents;
591 while (p) {
592 if (!--idx) {
593 hashcpy(result, p->item->object.sha1);
594 return 0;
596 p = p->next;
598 return -1;
601 static int get_nth_ancestor(const char *name, int len,
602 unsigned char *result, int generation)
604 unsigned char sha1[20];
605 struct commit *commit;
606 int ret;
608 ret = get_sha1_1(name, len, sha1, GET_SHA1_COMMITTISH);
609 if (ret)
610 return ret;
611 commit = lookup_commit_reference(sha1);
612 if (!commit)
613 return -1;
615 while (generation--) {
616 if (parse_commit(commit) || !commit->parents)
617 return -1;
618 commit = commit->parents->item;
620 hashcpy(result, commit->object.sha1);
621 return 0;
624 struct object *peel_to_type(const char *name, int namelen,
625 struct object *o, enum object_type expected_type)
627 if (name && !namelen)
628 namelen = strlen(name);
629 while (1) {
630 if (!o || (!o->parsed && !parse_object(o->sha1)))
631 return NULL;
632 if (expected_type == OBJ_ANY || o->type == expected_type)
633 return o;
634 if (o->type == OBJ_TAG)
635 o = ((struct tag*) o)->tagged;
636 else if (o->type == OBJ_COMMIT)
637 o = &(((struct commit *) o)->tree->object);
638 else {
639 if (name)
640 error("%.*s: expected %s type, but the object "
641 "dereferences to %s type",
642 namelen, name, typename(expected_type),
643 typename(o->type));
644 return NULL;
649 static int peel_onion(const char *name, int len, unsigned char *sha1)
651 unsigned char outer[20];
652 const char *sp;
653 unsigned int expected_type = 0;
654 unsigned lookup_flags = 0;
655 struct object *o;
658 * "ref^{type}" dereferences ref repeatedly until you cannot
659 * dereference anymore, or you get an object of given type,
660 * whichever comes first. "ref^{}" means just dereference
661 * tags until you get a non-tag. "ref^0" is a shorthand for
662 * "ref^{commit}". "commit^{tree}" could be used to find the
663 * top-level tree of the given commit.
665 if (len < 4 || name[len-1] != '}')
666 return -1;
668 for (sp = name + len - 1; name <= sp; sp--) {
669 int ch = *sp;
670 if (ch == '{' && name < sp && sp[-1] == '^')
671 break;
673 if (sp <= name)
674 return -1;
676 sp++; /* beginning of type name, or closing brace for empty */
677 if (starts_with(sp, "commit}"))
678 expected_type = OBJ_COMMIT;
679 else if (starts_with(sp, "tag}"))
680 expected_type = OBJ_TAG;
681 else if (starts_with(sp, "tree}"))
682 expected_type = OBJ_TREE;
683 else if (starts_with(sp, "blob}"))
684 expected_type = OBJ_BLOB;
685 else if (starts_with(sp, "object}"))
686 expected_type = OBJ_ANY;
687 else if (sp[0] == '}')
688 expected_type = OBJ_NONE;
689 else if (sp[0] == '/')
690 expected_type = OBJ_COMMIT;
691 else
692 return -1;
694 if (expected_type == OBJ_COMMIT)
695 lookup_flags = GET_SHA1_COMMITTISH;
696 else if (expected_type == OBJ_TREE)
697 lookup_flags = GET_SHA1_TREEISH;
699 if (get_sha1_1(name, sp - name - 2, outer, lookup_flags))
700 return -1;
702 o = parse_object(outer);
703 if (!o)
704 return -1;
705 if (!expected_type) {
706 o = deref_tag(o, name, sp - name - 2);
707 if (!o || (!o->parsed && !parse_object(o->sha1)))
708 return -1;
709 hashcpy(sha1, o->sha1);
710 return 0;
714 * At this point, the syntax look correct, so
715 * if we do not get the needed object, we should
716 * barf.
718 o = peel_to_type(name, len, o, expected_type);
719 if (!o)
720 return -1;
722 hashcpy(sha1, o->sha1);
723 if (sp[0] == '/') {
724 /* "$commit^{/foo}" */
725 char *prefix;
726 int ret;
727 struct commit_list *list = NULL;
730 * $commit^{/}. Some regex implementation may reject.
731 * We don't need regex anyway. '' pattern always matches.
733 if (sp[1] == '}')
734 return 0;
736 prefix = xstrndup(sp + 1, name + len - 1 - (sp + 1));
737 commit_list_insert((struct commit *)o, &list);
738 ret = get_sha1_oneline(prefix, sha1, list);
739 free(prefix);
740 return ret;
742 return 0;
745 static int get_describe_name(const char *name, int len, unsigned char *sha1)
747 const char *cp;
748 unsigned flags = GET_SHA1_QUIETLY | GET_SHA1_COMMIT;
750 for (cp = name + len - 1; name + 2 <= cp; cp--) {
751 char ch = *cp;
752 if (hexval(ch) & ~0377) {
753 /* We must be looking at g in "SOMETHING-g"
754 * for it to be describe output.
756 if (ch == 'g' && cp[-1] == '-') {
757 cp++;
758 len -= cp - name;
759 return get_short_sha1(cp, len, sha1, flags);
763 return -1;
766 static int get_sha1_1(const char *name, int len, unsigned char *sha1, unsigned lookup_flags)
768 int ret, has_suffix;
769 const char *cp;
772 * "name~3" is "name^^^", "name~" is "name~1", and "name^" is "name^1".
774 has_suffix = 0;
775 for (cp = name + len - 1; name <= cp; cp--) {
776 int ch = *cp;
777 if ('0' <= ch && ch <= '9')
778 continue;
779 if (ch == '~' || ch == '^')
780 has_suffix = ch;
781 break;
784 if (has_suffix) {
785 int num = 0;
786 int len1 = cp - name;
787 cp++;
788 while (cp < name + len)
789 num = num * 10 + *cp++ - '0';
790 if (!num && len1 == len - 1)
791 num = 1;
792 if (has_suffix == '^')
793 return get_parent(name, len1, sha1, num);
794 /* else if (has_suffix == '~') -- goes without saying */
795 return get_nth_ancestor(name, len1, sha1, num);
798 ret = peel_onion(name, len, sha1);
799 if (!ret)
800 return 0;
802 ret = get_sha1_basic(name, len, sha1);
803 if (!ret)
804 return 0;
806 /* It could be describe output that is "SOMETHING-gXXXX" */
807 ret = get_describe_name(name, len, sha1);
808 if (!ret)
809 return 0;
811 return get_short_sha1(name, len, sha1, lookup_flags);
815 * This interprets names like ':/Initial revision of "git"' by searching
816 * through history and returning the first commit whose message starts
817 * the given regular expression.
819 * For future extension, ':/!' is reserved. If you want to match a message
820 * beginning with a '!', you have to repeat the exclamation mark.
823 /* Remember to update object flag allocation in object.h */
824 #define ONELINE_SEEN (1u<<20)
826 static int handle_one_ref(const char *path,
827 const unsigned char *sha1, int flag, void *cb_data)
829 struct commit_list **list = cb_data;
830 struct object *object = parse_object(sha1);
831 if (!object)
832 return 0;
833 if (object->type == OBJ_TAG) {
834 object = deref_tag(object, path, strlen(path));
835 if (!object)
836 return 0;
838 if (object->type != OBJ_COMMIT)
839 return 0;
840 commit_list_insert_by_date((struct commit *)object, list);
841 return 0;
844 static int get_sha1_oneline(const char *prefix, unsigned char *sha1,
845 struct commit_list *list)
847 struct commit_list *backup = NULL, *l;
848 int found = 0;
849 regex_t regex;
851 if (prefix[0] == '!') {
852 if (prefix[1] != '!')
853 die ("Invalid search pattern: %s", prefix);
854 prefix++;
857 if (regcomp(&regex, prefix, REG_EXTENDED))
858 die("Invalid search pattern: %s", prefix);
860 for (l = list; l; l = l->next) {
861 l->item->object.flags |= ONELINE_SEEN;
862 commit_list_insert(l->item, &backup);
864 while (list) {
865 char *p, *to_free = NULL;
866 struct commit *commit;
867 enum object_type type;
868 unsigned long size;
869 int matches;
871 commit = pop_most_recent_commit(&list, ONELINE_SEEN);
872 if (!parse_object(commit->object.sha1))
873 continue;
874 if (commit->buffer)
875 p = commit->buffer;
876 else {
877 p = read_sha1_file(commit->object.sha1, &type, &size);
878 if (!p)
879 continue;
880 to_free = p;
883 p = strstr(p, "\n\n");
884 matches = p && !regexec(&regex, p + 2, 0, NULL, 0);
885 free(to_free);
887 if (matches) {
888 hashcpy(sha1, commit->object.sha1);
889 found = 1;
890 break;
893 regfree(&regex);
894 free_commit_list(list);
895 for (l = backup; l; l = l->next)
896 clear_commit_marks(l->item, ONELINE_SEEN);
897 free_commit_list(backup);
898 return found ? 0 : -1;
901 struct grab_nth_branch_switch_cbdata {
902 int remaining;
903 struct strbuf buf;
906 static int grab_nth_branch_switch(unsigned char *osha1, unsigned char *nsha1,
907 const char *email, unsigned long timestamp, int tz,
908 const char *message, void *cb_data)
910 struct grab_nth_branch_switch_cbdata *cb = cb_data;
911 const char *match = NULL, *target = NULL;
912 size_t len;
914 if (starts_with(message, "checkout: moving from ")) {
915 match = message + strlen("checkout: moving from ");
916 target = strstr(match, " to ");
919 if (!match || !target)
920 return 0;
921 if (--(cb->remaining) == 0) {
922 len = target - match;
923 strbuf_reset(&cb->buf);
924 strbuf_add(&cb->buf, match, len);
925 return 1; /* we are done */
927 return 0;
931 * Parse @{-N} syntax, return the number of characters parsed
932 * if successful; otherwise signal an error with negative value.
934 static int interpret_nth_prior_checkout(const char *name, int namelen,
935 struct strbuf *buf)
937 long nth;
938 int retval;
939 struct grab_nth_branch_switch_cbdata cb;
940 const char *brace;
941 char *num_end;
943 if (namelen < 4)
944 return -1;
945 if (name[0] != '@' || name[1] != '{' || name[2] != '-')
946 return -1;
947 brace = memchr(name, '}', namelen);
948 if (!brace)
949 return -1;
950 nth = strtol(name + 3, &num_end, 10);
951 if (num_end != brace)
952 return -1;
953 if (nth <= 0)
954 return -1;
955 cb.remaining = nth;
956 strbuf_init(&cb.buf, 20);
958 retval = 0;
959 if (0 < for_each_reflog_ent_reverse("HEAD", grab_nth_branch_switch, &cb)) {
960 strbuf_reset(buf);
961 strbuf_add(buf, cb.buf.buf, cb.buf.len);
962 retval = brace - name + 1;
965 strbuf_release(&cb.buf);
966 return retval;
969 int get_sha1_mb(const char *name, unsigned char *sha1)
971 struct commit *one, *two;
972 struct commit_list *mbs;
973 unsigned char sha1_tmp[20];
974 const char *dots;
975 int st;
977 dots = strstr(name, "...");
978 if (!dots)
979 return get_sha1(name, sha1);
980 if (dots == name)
981 st = get_sha1("HEAD", sha1_tmp);
982 else {
983 struct strbuf sb;
984 strbuf_init(&sb, dots - name);
985 strbuf_add(&sb, name, dots - name);
986 st = get_sha1_committish(sb.buf, sha1_tmp);
987 strbuf_release(&sb);
989 if (st)
990 return st;
991 one = lookup_commit_reference_gently(sha1_tmp, 0);
992 if (!one)
993 return -1;
995 if (get_sha1_committish(dots[3] ? (dots + 3) : "HEAD", sha1_tmp))
996 return -1;
997 two = lookup_commit_reference_gently(sha1_tmp, 0);
998 if (!two)
999 return -1;
1000 mbs = get_merge_bases(one, two, 1);
1001 if (!mbs || mbs->next)
1002 st = -1;
1003 else {
1004 st = 0;
1005 hashcpy(sha1, mbs->item->object.sha1);
1007 free_commit_list(mbs);
1008 return st;
1011 /* parse @something syntax, when 'something' is not {.*} */
1012 static int interpret_empty_at(const char *name, int namelen, int len, struct strbuf *buf)
1014 const char *next;
1016 if (len || name[1] == '{')
1017 return -1;
1019 /* make sure it's a single @, or @@{.*}, not @foo */
1020 next = memchr(name + len + 1, '@', namelen - len - 1);
1021 if (next && next[1] != '{')
1022 return -1;
1023 if (!next)
1024 next = name + namelen;
1025 if (next != name + 1)
1026 return -1;
1028 strbuf_reset(buf);
1029 strbuf_add(buf, "HEAD", 4);
1030 return 1;
1033 static int reinterpret(const char *name, int namelen, int len, struct strbuf *buf)
1035 /* we have extra data, which might need further processing */
1036 struct strbuf tmp = STRBUF_INIT;
1037 int used = buf->len;
1038 int ret;
1040 strbuf_add(buf, name + len, namelen - len);
1041 ret = interpret_branch_name(buf->buf, buf->len, &tmp);
1042 /* that data was not interpreted, remove our cruft */
1043 if (ret < 0) {
1044 strbuf_setlen(buf, used);
1045 return len;
1047 strbuf_reset(buf);
1048 strbuf_addbuf(buf, &tmp);
1049 strbuf_release(&tmp);
1050 /* tweak for size of {-N} versus expanded ref name */
1051 return ret - used + len;
1054 static void set_shortened_ref(struct strbuf *buf, const char *ref)
1056 char *s = shorten_unambiguous_ref(ref, 0);
1057 strbuf_reset(buf);
1058 strbuf_addstr(buf, s);
1059 free(s);
1062 static const char *get_upstream_branch(const char *branch_buf, int len)
1064 char *branch = xstrndup(branch_buf, len);
1065 struct branch *upstream = branch_get(*branch ? branch : NULL);
1068 * Upstream can be NULL only if branch refers to HEAD and HEAD
1069 * points to something different than a branch.
1071 if (!upstream)
1072 die(_("HEAD does not point to a branch"));
1073 if (!upstream->merge || !upstream->merge[0]->dst) {
1074 if (!ref_exists(upstream->refname))
1075 die(_("No such branch: '%s'"), branch);
1076 if (!upstream->merge) {
1077 die(_("No upstream configured for branch '%s'"),
1078 upstream->name);
1080 die(
1081 _("Upstream branch '%s' not stored as a remote-tracking branch"),
1082 upstream->merge[0]->src);
1084 free(branch);
1086 return upstream->merge[0]->dst;
1089 static int interpret_upstream_mark(const char *name, int namelen,
1090 int at, struct strbuf *buf)
1092 int len;
1094 len = upstream_mark(name + at, namelen - at);
1095 if (!len)
1096 return -1;
1098 if (memchr(name, ':', at))
1099 return -1;
1101 set_shortened_ref(buf, get_upstream_branch(name, at));
1102 return len + at;
1106 * This reads short-hand syntax that not only evaluates to a commit
1107 * object name, but also can act as if the end user spelled the name
1108 * of the branch from the command line.
1110 * - "@{-N}" finds the name of the Nth previous branch we were on, and
1111 * places the name of the branch in the given buf and returns the
1112 * number of characters parsed if successful.
1114 * - "<branch>@{upstream}" finds the name of the other ref that
1115 * <branch> is configured to merge with (missing <branch> defaults
1116 * to the current branch), and places the name of the branch in the
1117 * given buf and returns the number of characters parsed if
1118 * successful.
1120 * If the input is not of the accepted format, it returns a negative
1121 * number to signal an error.
1123 * If the input was ok but there are not N branch switches in the
1124 * reflog, it returns 0.
1126 int interpret_branch_name(const char *name, int namelen, struct strbuf *buf)
1128 char *at;
1129 const char *start;
1130 int len = interpret_nth_prior_checkout(name, namelen, buf);
1132 if (!namelen)
1133 namelen = strlen(name);
1135 if (!len) {
1136 return len; /* syntax Ok, not enough switches */
1137 } else if (len > 0) {
1138 if (len == namelen)
1139 return len; /* consumed all */
1140 else
1141 return reinterpret(name, namelen, len, buf);
1144 for (start = name;
1145 (at = memchr(start, '@', namelen - (start - name)));
1146 start = at + 1) {
1148 len = interpret_empty_at(name, namelen, at - name, buf);
1149 if (len > 0)
1150 return reinterpret(name, namelen, len, buf);
1152 len = interpret_upstream_mark(name, namelen, at - name, buf);
1153 if (len > 0)
1154 return len;
1157 return -1;
1160 int strbuf_branchname(struct strbuf *sb, const char *name)
1162 int len = strlen(name);
1163 int used = interpret_branch_name(name, len, sb);
1165 if (used == len)
1166 return 0;
1167 if (used < 0)
1168 used = 0;
1169 strbuf_add(sb, name + used, len - used);
1170 return len;
1173 int strbuf_check_branch_ref(struct strbuf *sb, const char *name)
1175 strbuf_branchname(sb, name);
1176 if (name[0] == '-')
1177 return -1;
1178 strbuf_splice(sb, 0, 0, "refs/heads/", 11);
1179 return check_refname_format(sb->buf, 0);
1183 * This is like "get_sha1_basic()", except it allows "sha1 expressions",
1184 * notably "xyz^" for "parent of xyz"
1186 int get_sha1(const char *name, unsigned char *sha1)
1188 struct object_context unused;
1189 return get_sha1_with_context(name, 0, sha1, &unused);
1193 * Many callers know that the user meant to name a commit-ish by
1194 * syntactical positions where the object name appears. Calling this
1195 * function allows the machinery to disambiguate shorter-than-unique
1196 * abbreviated object names between commit-ish and others.
1198 * Note that this does NOT error out when the named object is not a
1199 * commit-ish. It is merely to give a hint to the disambiguation
1200 * machinery.
1202 int get_sha1_committish(const char *name, unsigned char *sha1)
1204 struct object_context unused;
1205 return get_sha1_with_context(name, GET_SHA1_COMMITTISH,
1206 sha1, &unused);
1209 int get_sha1_treeish(const char *name, unsigned char *sha1)
1211 struct object_context unused;
1212 return get_sha1_with_context(name, GET_SHA1_TREEISH,
1213 sha1, &unused);
1216 int get_sha1_commit(const char *name, unsigned char *sha1)
1218 struct object_context unused;
1219 return get_sha1_with_context(name, GET_SHA1_COMMIT,
1220 sha1, &unused);
1223 int get_sha1_tree(const char *name, unsigned char *sha1)
1225 struct object_context unused;
1226 return get_sha1_with_context(name, GET_SHA1_TREE,
1227 sha1, &unused);
1230 int get_sha1_blob(const char *name, unsigned char *sha1)
1232 struct object_context unused;
1233 return get_sha1_with_context(name, GET_SHA1_BLOB,
1234 sha1, &unused);
1237 /* Must be called only when object_name:filename doesn't exist. */
1238 static void diagnose_invalid_sha1_path(const char *prefix,
1239 const char *filename,
1240 const unsigned char *tree_sha1,
1241 const char *object_name,
1242 int object_name_len)
1244 struct stat st;
1245 unsigned char sha1[20];
1246 unsigned mode;
1248 if (!prefix)
1249 prefix = "";
1251 if (!lstat(filename, &st))
1252 die("Path '%s' exists on disk, but not in '%.*s'.",
1253 filename, object_name_len, object_name);
1254 if (errno == ENOENT || errno == ENOTDIR) {
1255 char *fullname = xmalloc(strlen(filename)
1256 + strlen(prefix) + 1);
1257 strcpy(fullname, prefix);
1258 strcat(fullname, filename);
1260 if (!get_tree_entry(tree_sha1, fullname,
1261 sha1, &mode)) {
1262 die("Path '%s' exists, but not '%s'.\n"
1263 "Did you mean '%.*s:%s' aka '%.*s:./%s'?",
1264 fullname,
1265 filename,
1266 object_name_len, object_name,
1267 fullname,
1268 object_name_len, object_name,
1269 filename);
1271 die("Path '%s' does not exist in '%.*s'",
1272 filename, object_name_len, object_name);
1276 /* Must be called only when :stage:filename doesn't exist. */
1277 static void diagnose_invalid_index_path(int stage,
1278 const char *prefix,
1279 const char *filename)
1281 struct stat st;
1282 const struct cache_entry *ce;
1283 int pos;
1284 unsigned namelen = strlen(filename);
1285 unsigned fullnamelen;
1286 char *fullname;
1288 if (!prefix)
1289 prefix = "";
1291 /* Wrong stage number? */
1292 pos = cache_name_pos(filename, namelen);
1293 if (pos < 0)
1294 pos = -pos - 1;
1295 if (pos < active_nr) {
1296 ce = active_cache[pos];
1297 if (ce_namelen(ce) == namelen &&
1298 !memcmp(ce->name, filename, namelen))
1299 die("Path '%s' is in the index, but not at stage %d.\n"
1300 "Did you mean ':%d:%s'?",
1301 filename, stage,
1302 ce_stage(ce), filename);
1305 /* Confusion between relative and absolute filenames? */
1306 fullnamelen = namelen + strlen(prefix);
1307 fullname = xmalloc(fullnamelen + 1);
1308 strcpy(fullname, prefix);
1309 strcat(fullname, filename);
1310 pos = cache_name_pos(fullname, fullnamelen);
1311 if (pos < 0)
1312 pos = -pos - 1;
1313 if (pos < active_nr) {
1314 ce = active_cache[pos];
1315 if (ce_namelen(ce) == fullnamelen &&
1316 !memcmp(ce->name, fullname, fullnamelen))
1317 die("Path '%s' is in the index, but not '%s'.\n"
1318 "Did you mean ':%d:%s' aka ':%d:./%s'?",
1319 fullname, filename,
1320 ce_stage(ce), fullname,
1321 ce_stage(ce), filename);
1324 if (!lstat(filename, &st))
1325 die("Path '%s' exists on disk, but not in the index.", filename);
1326 if (errno == ENOENT || errno == ENOTDIR)
1327 die("Path '%s' does not exist (neither on disk nor in the index).",
1328 filename);
1330 free(fullname);
1334 static char *resolve_relative_path(const char *rel)
1336 if (!starts_with(rel, "./") && !starts_with(rel, "../"))
1337 return NULL;
1339 if (!startup_info)
1340 die("BUG: startup_info struct is not initialized.");
1342 if (!is_inside_work_tree())
1343 die("relative path syntax can't be used outside working tree.");
1345 /* die() inside prefix_path() if resolved path is outside worktree */
1346 return prefix_path(startup_info->prefix,
1347 startup_info->prefix ? strlen(startup_info->prefix) : 0,
1348 rel);
1351 static int get_sha1_with_context_1(const char *name,
1352 unsigned flags,
1353 const char *prefix,
1354 unsigned char *sha1,
1355 struct object_context *oc)
1357 int ret, bracket_depth;
1358 int namelen = strlen(name);
1359 const char *cp;
1360 int only_to_die = flags & GET_SHA1_ONLY_TO_DIE;
1362 memset(oc, 0, sizeof(*oc));
1363 oc->mode = S_IFINVALID;
1364 ret = get_sha1_1(name, namelen, sha1, flags);
1365 if (!ret)
1366 return ret;
1368 * sha1:path --> object name of path in ent sha1
1369 * :path -> object name of absolute path in index
1370 * :./path -> object name of path relative to cwd in index
1371 * :[0-3]:path -> object name of path in index at stage
1372 * :/foo -> recent commit matching foo
1374 if (name[0] == ':') {
1375 int stage = 0;
1376 const struct cache_entry *ce;
1377 char *new_path = NULL;
1378 int pos;
1379 if (!only_to_die && namelen > 2 && name[1] == '/') {
1380 struct commit_list *list = NULL;
1381 for_each_ref(handle_one_ref, &list);
1382 return get_sha1_oneline(name + 2, sha1, list);
1384 if (namelen < 3 ||
1385 name[2] != ':' ||
1386 name[1] < '0' || '3' < name[1])
1387 cp = name + 1;
1388 else {
1389 stage = name[1] - '0';
1390 cp = name + 3;
1392 new_path = resolve_relative_path(cp);
1393 if (!new_path) {
1394 namelen = namelen - (cp - name);
1395 } else {
1396 cp = new_path;
1397 namelen = strlen(cp);
1400 strncpy(oc->path, cp,
1401 sizeof(oc->path));
1402 oc->path[sizeof(oc->path)-1] = '\0';
1404 if (!active_cache)
1405 read_cache();
1406 pos = cache_name_pos(cp, namelen);
1407 if (pos < 0)
1408 pos = -pos - 1;
1409 while (pos < active_nr) {
1410 ce = active_cache[pos];
1411 if (ce_namelen(ce) != namelen ||
1412 memcmp(ce->name, cp, namelen))
1413 break;
1414 if (ce_stage(ce) == stage) {
1415 hashcpy(sha1, ce->sha1);
1416 oc->mode = ce->ce_mode;
1417 free(new_path);
1418 return 0;
1420 pos++;
1422 if (only_to_die && name[1] && name[1] != '/')
1423 diagnose_invalid_index_path(stage, prefix, cp);
1424 free(new_path);
1425 return -1;
1427 for (cp = name, bracket_depth = 0; *cp; cp++) {
1428 if (*cp == '{')
1429 bracket_depth++;
1430 else if (bracket_depth && *cp == '}')
1431 bracket_depth--;
1432 else if (!bracket_depth && *cp == ':')
1433 break;
1435 if (*cp == ':') {
1436 unsigned char tree_sha1[20];
1437 int len = cp - name;
1438 if (!get_sha1_1(name, len, tree_sha1, GET_SHA1_TREEISH)) {
1439 const char *filename = cp+1;
1440 char *new_filename = NULL;
1442 new_filename = resolve_relative_path(filename);
1443 if (new_filename)
1444 filename = new_filename;
1445 ret = get_tree_entry(tree_sha1, filename, sha1, &oc->mode);
1446 if (ret && only_to_die) {
1447 diagnose_invalid_sha1_path(prefix, filename,
1448 tree_sha1,
1449 name, len);
1451 hashcpy(oc->tree, tree_sha1);
1452 strncpy(oc->path, filename,
1453 sizeof(oc->path));
1454 oc->path[sizeof(oc->path)-1] = '\0';
1456 free(new_filename);
1457 return ret;
1458 } else {
1459 if (only_to_die)
1460 die("Invalid object name '%.*s'.", len, name);
1463 return ret;
1467 * Call this function when you know "name" given by the end user must
1468 * name an object but it doesn't; the function _may_ die with a better
1469 * diagnostic message than "no such object 'name'", e.g. "Path 'doc' does not
1470 * exist in 'HEAD'" when given "HEAD:doc", or it may return in which case
1471 * you have a chance to diagnose the error further.
1473 void maybe_die_on_misspelt_object_name(const char *name, const char *prefix)
1475 struct object_context oc;
1476 unsigned char sha1[20];
1477 get_sha1_with_context_1(name, GET_SHA1_ONLY_TO_DIE, prefix, sha1, &oc);
1480 int get_sha1_with_context(const char *str, unsigned flags, unsigned char *sha1, struct object_context *orc)
1482 return get_sha1_with_context_1(str, flags, NULL, sha1, orc);