Merge branch 'jk/pack-objects-optim-mru'
[alt-git.git] / sha1_name.c
blob3b647fd7cf79127cfa6f684836e387568a87c310
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"
10 #include "sha1-array.h"
12 static int get_sha1_oneline(const char *, unsigned char *, struct commit_list *);
14 typedef int (*disambiguate_hint_fn)(const unsigned char *, void *);
16 struct disambiguate_state {
17 int len; /* length of prefix in hex chars */
18 char hex_pfx[GIT_SHA1_HEXSZ + 1];
19 unsigned char bin_pfx[GIT_SHA1_RAWSZ];
21 disambiguate_hint_fn fn;
22 void *cb_data;
23 unsigned char candidate[GIT_SHA1_RAWSZ];
24 unsigned candidate_exists:1;
25 unsigned candidate_checked:1;
26 unsigned candidate_ok:1;
27 unsigned disambiguate_fn_used:1;
28 unsigned ambiguous:1;
29 unsigned always_call_fn:1;
32 static void update_candidates(struct disambiguate_state *ds, const unsigned char *current)
34 if (ds->always_call_fn) {
35 ds->ambiguous = ds->fn(current, ds->cb_data) ? 1 : 0;
36 return;
38 if (!ds->candidate_exists) {
39 /* this is the first candidate */
40 hashcpy(ds->candidate, current);
41 ds->candidate_exists = 1;
42 return;
43 } else if (!hashcmp(ds->candidate, current)) {
44 /* the same as what we already have seen */
45 return;
48 if (!ds->fn) {
49 /* cannot disambiguate between ds->candidate and current */
50 ds->ambiguous = 1;
51 return;
54 if (!ds->candidate_checked) {
55 ds->candidate_ok = ds->fn(ds->candidate, ds->cb_data);
56 ds->disambiguate_fn_used = 1;
57 ds->candidate_checked = 1;
60 if (!ds->candidate_ok) {
61 /* discard the candidate; we know it does not satisfy fn */
62 hashcpy(ds->candidate, current);
63 ds->candidate_checked = 0;
64 return;
67 /* if we reach this point, we know ds->candidate satisfies fn */
68 if (ds->fn(current, ds->cb_data)) {
70 * if both current and candidate satisfy fn, we cannot
71 * disambiguate.
73 ds->candidate_ok = 0;
74 ds->ambiguous = 1;
77 /* otherwise, current can be discarded and candidate is still good */
80 static void find_short_object_filename(struct disambiguate_state *ds)
82 struct alternate_object_database *alt;
83 char hex[GIT_SHA1_HEXSZ];
84 static struct alternate_object_database *fakeent;
86 if (!fakeent) {
88 * Create a "fake" alternate object database that
89 * points to our own object database, to make it
90 * easier to get a temporary working space in
91 * alt->name/alt->base while iterating over the
92 * object databases including our own.
94 const char *objdir = get_object_directory();
95 size_t objdir_len = strlen(objdir);
96 fakeent = xmalloc(st_add3(sizeof(*fakeent), objdir_len, 43));
97 memcpy(fakeent->base, objdir, objdir_len);
98 fakeent->name = fakeent->base + objdir_len + 1;
99 fakeent->name[-1] = '/';
101 fakeent->next = alt_odb_list;
103 xsnprintf(hex, sizeof(hex), "%.2s", ds->hex_pfx);
104 for (alt = fakeent; alt && !ds->ambiguous; alt = alt->next) {
105 struct dirent *de;
106 DIR *dir;
108 * every alt_odb struct has 42 extra bytes after the base
109 * for exactly this purpose
111 xsnprintf(alt->name, 42, "%.2s/", ds->hex_pfx);
112 dir = opendir(alt->base);
113 if (!dir)
114 continue;
116 while (!ds->ambiguous && (de = readdir(dir)) != NULL) {
117 unsigned char sha1[20];
119 if (strlen(de->d_name) != 38)
120 continue;
121 if (memcmp(de->d_name, ds->hex_pfx + 2, ds->len - 2))
122 continue;
123 memcpy(hex + 2, de->d_name, 38);
124 if (!get_sha1_hex(hex, sha1))
125 update_candidates(ds, sha1);
127 closedir(dir);
131 static int match_sha(unsigned len, const unsigned char *a, const unsigned char *b)
133 do {
134 if (*a != *b)
135 return 0;
136 a++;
137 b++;
138 len -= 2;
139 } while (len > 1);
140 if (len)
141 if ((*a ^ *b) & 0xf0)
142 return 0;
143 return 1;
146 static void unique_in_pack(struct packed_git *p,
147 struct disambiguate_state *ds)
149 uint32_t num, last, i, first = 0;
150 const unsigned char *current = NULL;
152 open_pack_index(p);
153 num = p->num_objects;
154 last = num;
155 while (first < last) {
156 uint32_t mid = (first + last) / 2;
157 const unsigned char *current;
158 int cmp;
160 current = nth_packed_object_sha1(p, mid);
161 cmp = hashcmp(ds->bin_pfx, current);
162 if (!cmp) {
163 first = mid;
164 break;
166 if (cmp > 0) {
167 first = mid+1;
168 continue;
170 last = mid;
174 * At this point, "first" is the location of the lowest object
175 * with an object name that could match "bin_pfx". See if we have
176 * 0, 1 or more objects that actually match(es).
178 for (i = first; i < num && !ds->ambiguous; i++) {
179 current = nth_packed_object_sha1(p, i);
180 if (!match_sha(ds->len, ds->bin_pfx, current))
181 break;
182 update_candidates(ds, current);
186 static void find_short_packed_object(struct disambiguate_state *ds)
188 struct packed_git *p;
190 prepare_packed_git();
191 for (p = packed_git; p && !ds->ambiguous; p = p->next)
192 unique_in_pack(p, ds);
195 #define SHORT_NAME_NOT_FOUND (-1)
196 #define SHORT_NAME_AMBIGUOUS (-2)
198 static int finish_object_disambiguation(struct disambiguate_state *ds,
199 unsigned char *sha1)
201 if (ds->ambiguous)
202 return SHORT_NAME_AMBIGUOUS;
204 if (!ds->candidate_exists)
205 return SHORT_NAME_NOT_FOUND;
207 if (!ds->candidate_checked)
209 * If this is the only candidate, there is no point
210 * calling the disambiguation hint callback.
212 * On the other hand, if the current candidate
213 * replaced an earlier candidate that did _not_ pass
214 * the disambiguation hint callback, then we do have
215 * more than one objects that match the short name
216 * given, so we should make sure this one matches;
217 * otherwise, if we discovered this one and the one
218 * that we previously discarded in the reverse order,
219 * we would end up showing different results in the
220 * same repository!
222 ds->candidate_ok = (!ds->disambiguate_fn_used ||
223 ds->fn(ds->candidate, ds->cb_data));
225 if (!ds->candidate_ok)
226 return SHORT_NAME_AMBIGUOUS;
228 hashcpy(sha1, ds->candidate);
229 return 0;
232 static int disambiguate_commit_only(const unsigned char *sha1, void *cb_data_unused)
234 int kind = sha1_object_info(sha1, NULL);
235 return kind == OBJ_COMMIT;
238 static int disambiguate_committish_only(const unsigned char *sha1, void *cb_data_unused)
240 struct object *obj;
241 int kind;
243 kind = sha1_object_info(sha1, NULL);
244 if (kind == OBJ_COMMIT)
245 return 1;
246 if (kind != OBJ_TAG)
247 return 0;
249 /* We need to do this the hard way... */
250 obj = deref_tag(parse_object(sha1), NULL, 0);
251 if (obj && obj->type == OBJ_COMMIT)
252 return 1;
253 return 0;
256 static int disambiguate_tree_only(const unsigned char *sha1, void *cb_data_unused)
258 int kind = sha1_object_info(sha1, NULL);
259 return kind == OBJ_TREE;
262 static int disambiguate_treeish_only(const unsigned char *sha1, void *cb_data_unused)
264 struct object *obj;
265 int kind;
267 kind = sha1_object_info(sha1, NULL);
268 if (kind == OBJ_TREE || kind == OBJ_COMMIT)
269 return 1;
270 if (kind != OBJ_TAG)
271 return 0;
273 /* We need to do this the hard way... */
274 obj = deref_tag(parse_object(sha1), NULL, 0);
275 if (obj && (obj->type == OBJ_TREE || obj->type == OBJ_COMMIT))
276 return 1;
277 return 0;
280 static int disambiguate_blob_only(const unsigned char *sha1, void *cb_data_unused)
282 int kind = sha1_object_info(sha1, NULL);
283 return kind == OBJ_BLOB;
286 static disambiguate_hint_fn default_disambiguate_hint;
288 int set_disambiguate_hint_config(const char *var, const char *value)
290 static const struct {
291 const char *name;
292 disambiguate_hint_fn fn;
293 } hints[] = {
294 { "none", NULL },
295 { "commit", disambiguate_commit_only },
296 { "committish", disambiguate_committish_only },
297 { "tree", disambiguate_tree_only },
298 { "treeish", disambiguate_treeish_only },
299 { "blob", disambiguate_blob_only }
301 int i;
303 if (!value)
304 return config_error_nonbool(var);
306 for (i = 0; i < ARRAY_SIZE(hints); i++) {
307 if (!strcasecmp(value, hints[i].name)) {
308 default_disambiguate_hint = hints[i].fn;
309 return 0;
313 return error("unknown hint type for '%s': %s", var, value);
316 static int init_object_disambiguation(const char *name, int len,
317 struct disambiguate_state *ds)
319 int i;
321 if (len < MINIMUM_ABBREV || len > GIT_SHA1_HEXSZ)
322 return -1;
324 memset(ds, 0, sizeof(*ds));
326 for (i = 0; i < len ;i++) {
327 unsigned char c = name[i];
328 unsigned char val;
329 if (c >= '0' && c <= '9')
330 val = c - '0';
331 else if (c >= 'a' && c <= 'f')
332 val = c - 'a' + 10;
333 else if (c >= 'A' && c <='F') {
334 val = c - 'A' + 10;
335 c -= 'A' - 'a';
337 else
338 return -1;
339 ds->hex_pfx[i] = c;
340 if (!(i & 1))
341 val <<= 4;
342 ds->bin_pfx[i >> 1] |= val;
345 ds->len = len;
346 ds->hex_pfx[len] = '\0';
347 prepare_alt_odb();
348 return 0;
351 static int show_ambiguous_object(const unsigned char *sha1, void *data)
353 const struct disambiguate_state *ds = data;
354 struct strbuf desc = STRBUF_INIT;
355 int type;
357 if (ds->fn && !ds->fn(sha1, ds->cb_data))
358 return 0;
360 type = sha1_object_info(sha1, NULL);
361 if (type == OBJ_COMMIT) {
362 struct commit *commit = lookup_commit(sha1);
363 if (commit) {
364 struct pretty_print_context pp = {0};
365 pp.date_mode.type = DATE_SHORT;
366 format_commit_message(commit, " %ad - %s", &desc, &pp);
368 } else if (type == OBJ_TAG) {
369 struct tag *tag = lookup_tag(sha1);
370 if (!parse_tag(tag) && tag->tag)
371 strbuf_addf(&desc, " %s", tag->tag);
374 advise(" %s %s%s",
375 find_unique_abbrev(sha1, DEFAULT_ABBREV),
376 typename(type) ? typename(type) : "unknown type",
377 desc.buf);
379 strbuf_release(&desc);
380 return 0;
383 static int get_short_sha1(const char *name, int len, unsigned char *sha1,
384 unsigned flags)
386 int status;
387 struct disambiguate_state ds;
388 int quietly = !!(flags & GET_SHA1_QUIETLY);
390 if (init_object_disambiguation(name, len, &ds) < 0)
391 return -1;
393 if (HAS_MULTI_BITS(flags & GET_SHA1_DISAMBIGUATORS))
394 die("BUG: multiple get_short_sha1 disambiguator flags");
396 if (flags & GET_SHA1_COMMIT)
397 ds.fn = disambiguate_commit_only;
398 else if (flags & GET_SHA1_COMMITTISH)
399 ds.fn = disambiguate_committish_only;
400 else if (flags & GET_SHA1_TREE)
401 ds.fn = disambiguate_tree_only;
402 else if (flags & GET_SHA1_TREEISH)
403 ds.fn = disambiguate_treeish_only;
404 else if (flags & GET_SHA1_BLOB)
405 ds.fn = disambiguate_blob_only;
406 else
407 ds.fn = default_disambiguate_hint;
409 find_short_object_filename(&ds);
410 find_short_packed_object(&ds);
411 status = finish_object_disambiguation(&ds, sha1);
413 if (!quietly && (status == SHORT_NAME_AMBIGUOUS)) {
414 error(_("short SHA1 %s is ambiguous"), ds.hex_pfx);
417 * We may still have ambiguity if we simply saw a series of
418 * candidates that did not satisfy our hint function. In
419 * that case, we still want to show them, so disable the hint
420 * function entirely.
422 if (!ds.ambiguous)
423 ds.fn = NULL;
425 advise(_("The candidates are:"));
426 for_each_abbrev(ds.hex_pfx, show_ambiguous_object, &ds);
429 return status;
432 static int collect_ambiguous(const unsigned char *sha1, void *data)
434 sha1_array_append(data, sha1);
435 return 0;
438 int for_each_abbrev(const char *prefix, each_abbrev_fn fn, void *cb_data)
440 struct sha1_array collect = SHA1_ARRAY_INIT;
441 struct disambiguate_state ds;
442 int ret;
444 if (init_object_disambiguation(prefix, strlen(prefix), &ds) < 0)
445 return -1;
447 ds.always_call_fn = 1;
448 ds.fn = collect_ambiguous;
449 ds.cb_data = &collect;
450 find_short_object_filename(&ds);
451 find_short_packed_object(&ds);
453 ret = sha1_array_for_each_unique(&collect, fn, cb_data);
454 sha1_array_clear(&collect);
455 return ret;
458 int find_unique_abbrev_r(char *hex, const unsigned char *sha1, int len)
460 int status, exists;
462 sha1_to_hex_r(hex, sha1);
463 if (len == 40 || !len)
464 return 40;
465 exists = has_sha1_file(sha1);
466 while (len < 40) {
467 unsigned char sha1_ret[20];
468 status = get_short_sha1(hex, len, sha1_ret, GET_SHA1_QUIETLY);
469 if (exists
470 ? !status
471 : status == SHORT_NAME_NOT_FOUND) {
472 hex[len] = 0;
473 return len;
475 len++;
477 return len;
480 const char *find_unique_abbrev(const unsigned char *sha1, int len)
482 static char hex[GIT_SHA1_HEXSZ + 1];
483 find_unique_abbrev_r(hex, sha1, len);
484 return hex;
487 static int ambiguous_path(const char *path, int len)
489 int slash = 1;
490 int cnt;
492 for (cnt = 0; cnt < len; cnt++) {
493 switch (*path++) {
494 case '\0':
495 break;
496 case '/':
497 if (slash)
498 break;
499 slash = 1;
500 continue;
501 case '.':
502 continue;
503 default:
504 slash = 0;
505 continue;
507 break;
509 return slash;
512 static inline int at_mark(const char *string, int len,
513 const char **suffix, int nr)
515 int i;
517 for (i = 0; i < nr; i++) {
518 int suffix_len = strlen(suffix[i]);
519 if (suffix_len <= len
520 && !memcmp(string, suffix[i], suffix_len))
521 return suffix_len;
523 return 0;
526 static inline int upstream_mark(const char *string, int len)
528 const char *suffix[] = { "@{upstream}", "@{u}" };
529 return at_mark(string, len, suffix, ARRAY_SIZE(suffix));
532 static inline int push_mark(const char *string, int len)
534 const char *suffix[] = { "@{push}" };
535 return at_mark(string, len, suffix, ARRAY_SIZE(suffix));
538 static int get_sha1_1(const char *name, int len, unsigned char *sha1, unsigned lookup_flags);
539 static int interpret_nth_prior_checkout(const char *name, int namelen, struct strbuf *buf);
541 static int get_sha1_basic(const char *str, int len, unsigned char *sha1,
542 unsigned int flags)
544 static const char *warn_msg = "refname '%.*s' is ambiguous.";
545 static const char *object_name_msg = N_(
546 "Git normally never creates a ref that ends with 40 hex characters\n"
547 "because it will be ignored when you just specify 40-hex. These refs\n"
548 "may be created by mistake. For example,\n"
549 "\n"
550 " git checkout -b $br $(git rev-parse ...)\n"
551 "\n"
552 "where \"$br\" is somehow empty and a 40-hex ref is created. Please\n"
553 "examine these refs and maybe delete them. Turn this message off by\n"
554 "running \"git config advice.objectNameWarning false\"");
555 unsigned char tmp_sha1[20];
556 char *real_ref = NULL;
557 int refs_found = 0;
558 int at, reflog_len, nth_prior = 0;
560 if (len == 40 && !get_sha1_hex(str, sha1)) {
561 if (warn_ambiguous_refs && warn_on_object_refname_ambiguity) {
562 refs_found = dwim_ref(str, len, tmp_sha1, &real_ref);
563 if (refs_found > 0) {
564 warning(warn_msg, len, str);
565 if (advice_object_name_warning)
566 fprintf(stderr, "%s\n", _(object_name_msg));
568 free(real_ref);
570 return 0;
573 /* basic@{time or number or -number} format to query ref-log */
574 reflog_len = at = 0;
575 if (len && str[len-1] == '}') {
576 for (at = len-4; at >= 0; at--) {
577 if (str[at] == '@' && str[at+1] == '{') {
578 if (str[at+2] == '-') {
579 if (at != 0)
580 /* @{-N} not at start */
581 return -1;
582 nth_prior = 1;
583 continue;
585 if (!upstream_mark(str + at, len - at) &&
586 !push_mark(str + at, len - at)) {
587 reflog_len = (len-1) - (at+2);
588 len = at;
590 break;
595 /* Accept only unambiguous ref paths. */
596 if (len && ambiguous_path(str, len))
597 return -1;
599 if (nth_prior) {
600 struct strbuf buf = STRBUF_INIT;
601 int detached;
603 if (interpret_nth_prior_checkout(str, len, &buf) > 0) {
604 detached = (buf.len == 40 && !get_sha1_hex(buf.buf, sha1));
605 strbuf_release(&buf);
606 if (detached)
607 return 0;
611 if (!len && reflog_len)
612 /* allow "@{...}" to mean the current branch reflog */
613 refs_found = dwim_ref("HEAD", 4, sha1, &real_ref);
614 else if (reflog_len)
615 refs_found = dwim_log(str, len, sha1, &real_ref);
616 else
617 refs_found = dwim_ref(str, len, sha1, &real_ref);
619 if (!refs_found)
620 return -1;
622 if (warn_ambiguous_refs && !(flags & GET_SHA1_QUIETLY) &&
623 (refs_found > 1 ||
624 !get_short_sha1(str, len, tmp_sha1, GET_SHA1_QUIETLY)))
625 warning(warn_msg, len, str);
627 if (reflog_len) {
628 int nth, i;
629 unsigned long at_time;
630 unsigned long co_time;
631 int co_tz, co_cnt;
633 /* Is it asking for N-th entry, or approxidate? */
634 for (i = nth = 0; 0 <= nth && i < reflog_len; i++) {
635 char ch = str[at+2+i];
636 if ('0' <= ch && ch <= '9')
637 nth = nth * 10 + ch - '0';
638 else
639 nth = -1;
641 if (100000000 <= nth) {
642 at_time = nth;
643 nth = -1;
644 } else if (0 <= nth)
645 at_time = 0;
646 else {
647 int errors = 0;
648 char *tmp = xstrndup(str + at + 2, reflog_len);
649 at_time = approxidate_careful(tmp, &errors);
650 free(tmp);
651 if (errors) {
652 free(real_ref);
653 return -1;
656 if (read_ref_at(real_ref, flags, at_time, nth, sha1, NULL,
657 &co_time, &co_tz, &co_cnt)) {
658 if (!len) {
659 if (starts_with(real_ref, "refs/heads/")) {
660 str = real_ref + 11;
661 len = strlen(real_ref + 11);
662 } else {
663 /* detached HEAD */
664 str = "HEAD";
665 len = 4;
668 if (at_time) {
669 if (!(flags & GET_SHA1_QUIETLY)) {
670 warning("Log for '%.*s' only goes "
671 "back to %s.", len, str,
672 show_date(co_time, co_tz, DATE_MODE(RFC2822)));
674 } else {
675 if (flags & GET_SHA1_QUIETLY) {
676 exit(128);
678 die("Log for '%.*s' only has %d entries.",
679 len, str, co_cnt);
684 free(real_ref);
685 return 0;
688 static int get_parent(const char *name, int len,
689 unsigned char *result, int idx)
691 unsigned char sha1[20];
692 int ret = get_sha1_1(name, len, sha1, GET_SHA1_COMMITTISH);
693 struct commit *commit;
694 struct commit_list *p;
696 if (ret)
697 return ret;
698 commit = lookup_commit_reference(sha1);
699 if (parse_commit(commit))
700 return -1;
701 if (!idx) {
702 hashcpy(result, commit->object.oid.hash);
703 return 0;
705 p = commit->parents;
706 while (p) {
707 if (!--idx) {
708 hashcpy(result, p->item->object.oid.hash);
709 return 0;
711 p = p->next;
713 return -1;
716 static int get_nth_ancestor(const char *name, int len,
717 unsigned char *result, int generation)
719 unsigned char sha1[20];
720 struct commit *commit;
721 int ret;
723 ret = get_sha1_1(name, len, sha1, GET_SHA1_COMMITTISH);
724 if (ret)
725 return ret;
726 commit = lookup_commit_reference(sha1);
727 if (!commit)
728 return -1;
730 while (generation--) {
731 if (parse_commit(commit) || !commit->parents)
732 return -1;
733 commit = commit->parents->item;
735 hashcpy(result, commit->object.oid.hash);
736 return 0;
739 struct object *peel_to_type(const char *name, int namelen,
740 struct object *o, enum object_type expected_type)
742 if (name && !namelen)
743 namelen = strlen(name);
744 while (1) {
745 if (!o || (!o->parsed && !parse_object(o->oid.hash)))
746 return NULL;
747 if (expected_type == OBJ_ANY || o->type == expected_type)
748 return o;
749 if (o->type == OBJ_TAG)
750 o = ((struct tag*) o)->tagged;
751 else if (o->type == OBJ_COMMIT)
752 o = &(((struct commit *) o)->tree->object);
753 else {
754 if (name)
755 error("%.*s: expected %s type, but the object "
756 "dereferences to %s type",
757 namelen, name, typename(expected_type),
758 typename(o->type));
759 return NULL;
764 static int peel_onion(const char *name, int len, unsigned char *sha1,
765 unsigned lookup_flags)
767 unsigned char outer[20];
768 const char *sp;
769 unsigned int expected_type = 0;
770 struct object *o;
773 * "ref^{type}" dereferences ref repeatedly until you cannot
774 * dereference anymore, or you get an object of given type,
775 * whichever comes first. "ref^{}" means just dereference
776 * tags until you get a non-tag. "ref^0" is a shorthand for
777 * "ref^{commit}". "commit^{tree}" could be used to find the
778 * top-level tree of the given commit.
780 if (len < 4 || name[len-1] != '}')
781 return -1;
783 for (sp = name + len - 1; name <= sp; sp--) {
784 int ch = *sp;
785 if (ch == '{' && name < sp && sp[-1] == '^')
786 break;
788 if (sp <= name)
789 return -1;
791 sp++; /* beginning of type name, or closing brace for empty */
792 if (starts_with(sp, "commit}"))
793 expected_type = OBJ_COMMIT;
794 else if (starts_with(sp, "tag}"))
795 expected_type = OBJ_TAG;
796 else if (starts_with(sp, "tree}"))
797 expected_type = OBJ_TREE;
798 else if (starts_with(sp, "blob}"))
799 expected_type = OBJ_BLOB;
800 else if (starts_with(sp, "object}"))
801 expected_type = OBJ_ANY;
802 else if (sp[0] == '}')
803 expected_type = OBJ_NONE;
804 else if (sp[0] == '/')
805 expected_type = OBJ_COMMIT;
806 else
807 return -1;
809 lookup_flags &= ~GET_SHA1_DISAMBIGUATORS;
810 if (expected_type == OBJ_COMMIT)
811 lookup_flags |= GET_SHA1_COMMITTISH;
812 else if (expected_type == OBJ_TREE)
813 lookup_flags |= GET_SHA1_TREEISH;
815 if (get_sha1_1(name, sp - name - 2, outer, lookup_flags))
816 return -1;
818 o = parse_object(outer);
819 if (!o)
820 return -1;
821 if (!expected_type) {
822 o = deref_tag(o, name, sp - name - 2);
823 if (!o || (!o->parsed && !parse_object(o->oid.hash)))
824 return -1;
825 hashcpy(sha1, o->oid.hash);
826 return 0;
830 * At this point, the syntax look correct, so
831 * if we do not get the needed object, we should
832 * barf.
834 o = peel_to_type(name, len, o, expected_type);
835 if (!o)
836 return -1;
838 hashcpy(sha1, o->oid.hash);
839 if (sp[0] == '/') {
840 /* "$commit^{/foo}" */
841 char *prefix;
842 int ret;
843 struct commit_list *list = NULL;
846 * $commit^{/}. Some regex implementation may reject.
847 * We don't need regex anyway. '' pattern always matches.
849 if (sp[1] == '}')
850 return 0;
852 prefix = xstrndup(sp + 1, name + len - 1 - (sp + 1));
853 commit_list_insert((struct commit *)o, &list);
854 ret = get_sha1_oneline(prefix, sha1, list);
855 free(prefix);
856 return ret;
858 return 0;
861 static int get_describe_name(const char *name, int len, unsigned char *sha1)
863 const char *cp;
864 unsigned flags = GET_SHA1_QUIETLY | GET_SHA1_COMMIT;
866 for (cp = name + len - 1; name + 2 <= cp; cp--) {
867 char ch = *cp;
868 if (!isxdigit(ch)) {
869 /* We must be looking at g in "SOMETHING-g"
870 * for it to be describe output.
872 if (ch == 'g' && cp[-1] == '-') {
873 cp++;
874 len -= cp - name;
875 return get_short_sha1(cp, len, sha1, flags);
879 return -1;
882 static int get_sha1_1(const char *name, int len, unsigned char *sha1, unsigned lookup_flags)
884 int ret, has_suffix;
885 const char *cp;
888 * "name~3" is "name^^^", "name~" is "name~1", and "name^" is "name^1".
890 has_suffix = 0;
891 for (cp = name + len - 1; name <= cp; cp--) {
892 int ch = *cp;
893 if ('0' <= ch && ch <= '9')
894 continue;
895 if (ch == '~' || ch == '^')
896 has_suffix = ch;
897 break;
900 if (has_suffix) {
901 int num = 0;
902 int len1 = cp - name;
903 cp++;
904 while (cp < name + len)
905 num = num * 10 + *cp++ - '0';
906 if (!num && len1 == len - 1)
907 num = 1;
908 if (has_suffix == '^')
909 return get_parent(name, len1, sha1, num);
910 /* else if (has_suffix == '~') -- goes without saying */
911 return get_nth_ancestor(name, len1, sha1, num);
914 ret = peel_onion(name, len, sha1, lookup_flags);
915 if (!ret)
916 return 0;
918 ret = get_sha1_basic(name, len, sha1, lookup_flags);
919 if (!ret)
920 return 0;
922 /* It could be describe output that is "SOMETHING-gXXXX" */
923 ret = get_describe_name(name, len, sha1);
924 if (!ret)
925 return 0;
927 return get_short_sha1(name, len, sha1, lookup_flags);
931 * This interprets names like ':/Initial revision of "git"' by searching
932 * through history and returning the first commit whose message starts
933 * the given regular expression.
935 * For negative-matching, prefix the pattern-part with '!-', like: ':/!-WIP'.
937 * For a literal '!' character at the beginning of a pattern, you have to repeat
938 * that, like: ':/!!foo'
940 * For future extension, all other sequences beginning with ':/!' are reserved.
943 /* Remember to update object flag allocation in object.h */
944 #define ONELINE_SEEN (1u<<20)
946 static int handle_one_ref(const char *path, const struct object_id *oid,
947 int flag, void *cb_data)
949 struct commit_list **list = cb_data;
950 struct object *object = parse_object(oid->hash);
951 if (!object)
952 return 0;
953 if (object->type == OBJ_TAG) {
954 object = deref_tag(object, path, strlen(path));
955 if (!object)
956 return 0;
958 if (object->type != OBJ_COMMIT)
959 return 0;
960 commit_list_insert((struct commit *)object, list);
961 return 0;
964 static int get_sha1_oneline(const char *prefix, unsigned char *sha1,
965 struct commit_list *list)
967 struct commit_list *backup = NULL, *l;
968 int found = 0;
969 int negative = 0;
970 regex_t regex;
972 if (prefix[0] == '!') {
973 prefix++;
975 if (prefix[0] == '-') {
976 prefix++;
977 negative = 1;
978 } else if (prefix[0] != '!') {
979 return -1;
983 if (regcomp(&regex, prefix, REG_EXTENDED))
984 return -1;
986 for (l = list; l; l = l->next) {
987 l->item->object.flags |= ONELINE_SEEN;
988 commit_list_insert(l->item, &backup);
990 while (list) {
991 const char *p, *buf;
992 struct commit *commit;
993 int matches;
995 commit = pop_most_recent_commit(&list, ONELINE_SEEN);
996 if (!parse_object(commit->object.oid.hash))
997 continue;
998 buf = get_commit_buffer(commit, NULL);
999 p = strstr(buf, "\n\n");
1000 matches = negative ^ (p && !regexec(&regex, p + 2, 0, NULL, 0));
1001 unuse_commit_buffer(commit, buf);
1003 if (matches) {
1004 hashcpy(sha1, commit->object.oid.hash);
1005 found = 1;
1006 break;
1009 regfree(&regex);
1010 free_commit_list(list);
1011 for (l = backup; l; l = l->next)
1012 clear_commit_marks(l->item, ONELINE_SEEN);
1013 free_commit_list(backup);
1014 return found ? 0 : -1;
1017 struct grab_nth_branch_switch_cbdata {
1018 int remaining;
1019 struct strbuf buf;
1022 static int grab_nth_branch_switch(unsigned char *osha1, unsigned char *nsha1,
1023 const char *email, unsigned long timestamp, int tz,
1024 const char *message, void *cb_data)
1026 struct grab_nth_branch_switch_cbdata *cb = cb_data;
1027 const char *match = NULL, *target = NULL;
1028 size_t len;
1030 if (skip_prefix(message, "checkout: moving from ", &match))
1031 target = strstr(match, " to ");
1033 if (!match || !target)
1034 return 0;
1035 if (--(cb->remaining) == 0) {
1036 len = target - match;
1037 strbuf_reset(&cb->buf);
1038 strbuf_add(&cb->buf, match, len);
1039 return 1; /* we are done */
1041 return 0;
1045 * Parse @{-N} syntax, return the number of characters parsed
1046 * if successful; otherwise signal an error with negative value.
1048 static int interpret_nth_prior_checkout(const char *name, int namelen,
1049 struct strbuf *buf)
1051 long nth;
1052 int retval;
1053 struct grab_nth_branch_switch_cbdata cb;
1054 const char *brace;
1055 char *num_end;
1057 if (namelen < 4)
1058 return -1;
1059 if (name[0] != '@' || name[1] != '{' || name[2] != '-')
1060 return -1;
1061 brace = memchr(name, '}', namelen);
1062 if (!brace)
1063 return -1;
1064 nth = strtol(name + 3, &num_end, 10);
1065 if (num_end != brace)
1066 return -1;
1067 if (nth <= 0)
1068 return -1;
1069 cb.remaining = nth;
1070 strbuf_init(&cb.buf, 20);
1072 retval = 0;
1073 if (0 < for_each_reflog_ent_reverse("HEAD", grab_nth_branch_switch, &cb)) {
1074 strbuf_reset(buf);
1075 strbuf_addbuf(buf, &cb.buf);
1076 retval = brace - name + 1;
1079 strbuf_release(&cb.buf);
1080 return retval;
1083 int get_oid_mb(const char *name, struct object_id *oid)
1085 struct commit *one, *two;
1086 struct commit_list *mbs;
1087 struct object_id oid_tmp;
1088 const char *dots;
1089 int st;
1091 dots = strstr(name, "...");
1092 if (!dots)
1093 return get_oid(name, oid);
1094 if (dots == name)
1095 st = get_oid("HEAD", &oid_tmp);
1096 else {
1097 struct strbuf sb;
1098 strbuf_init(&sb, dots - name);
1099 strbuf_add(&sb, name, dots - name);
1100 st = get_sha1_committish(sb.buf, oid_tmp.hash);
1101 strbuf_release(&sb);
1103 if (st)
1104 return st;
1105 one = lookup_commit_reference_gently(oid_tmp.hash, 0);
1106 if (!one)
1107 return -1;
1109 if (get_sha1_committish(dots[3] ? (dots + 3) : "HEAD", oid_tmp.hash))
1110 return -1;
1111 two = lookup_commit_reference_gently(oid_tmp.hash, 0);
1112 if (!two)
1113 return -1;
1114 mbs = get_merge_bases(one, two);
1115 if (!mbs || mbs->next)
1116 st = -1;
1117 else {
1118 st = 0;
1119 oidcpy(oid, &mbs->item->object.oid);
1121 free_commit_list(mbs);
1122 return st;
1125 /* parse @something syntax, when 'something' is not {.*} */
1126 static int interpret_empty_at(const char *name, int namelen, int len, struct strbuf *buf)
1128 const char *next;
1130 if (len || name[1] == '{')
1131 return -1;
1133 /* make sure it's a single @, or @@{.*}, not @foo */
1134 next = memchr(name + len + 1, '@', namelen - len - 1);
1135 if (next && next[1] != '{')
1136 return -1;
1137 if (!next)
1138 next = name + namelen;
1139 if (next != name + 1)
1140 return -1;
1142 strbuf_reset(buf);
1143 strbuf_add(buf, "HEAD", 4);
1144 return 1;
1147 static int reinterpret(const char *name, int namelen, int len, struct strbuf *buf)
1149 /* we have extra data, which might need further processing */
1150 struct strbuf tmp = STRBUF_INIT;
1151 int used = buf->len;
1152 int ret;
1154 strbuf_add(buf, name + len, namelen - len);
1155 ret = interpret_branch_name(buf->buf, buf->len, &tmp);
1156 /* that data was not interpreted, remove our cruft */
1157 if (ret < 0) {
1158 strbuf_setlen(buf, used);
1159 return len;
1161 strbuf_reset(buf);
1162 strbuf_addbuf(buf, &tmp);
1163 strbuf_release(&tmp);
1164 /* tweak for size of {-N} versus expanded ref name */
1165 return ret - used + len;
1168 static void set_shortened_ref(struct strbuf *buf, const char *ref)
1170 char *s = shorten_unambiguous_ref(ref, 0);
1171 strbuf_reset(buf);
1172 strbuf_addstr(buf, s);
1173 free(s);
1176 static int interpret_branch_mark(const char *name, int namelen,
1177 int at, struct strbuf *buf,
1178 int (*get_mark)(const char *, int),
1179 const char *(*get_data)(struct branch *,
1180 struct strbuf *))
1182 int len;
1183 struct branch *branch;
1184 struct strbuf err = STRBUF_INIT;
1185 const char *value;
1187 len = get_mark(name + at, namelen - at);
1188 if (!len)
1189 return -1;
1191 if (memchr(name, ':', at))
1192 return -1;
1194 if (at) {
1195 char *name_str = xmemdupz(name, at);
1196 branch = branch_get(name_str);
1197 free(name_str);
1198 } else
1199 branch = branch_get(NULL);
1201 value = get_data(branch, &err);
1202 if (!value)
1203 die("%s", err.buf);
1205 set_shortened_ref(buf, value);
1206 return len + at;
1210 * This reads short-hand syntax that not only evaluates to a commit
1211 * object name, but also can act as if the end user spelled the name
1212 * of the branch from the command line.
1214 * - "@{-N}" finds the name of the Nth previous branch we were on, and
1215 * places the name of the branch in the given buf and returns the
1216 * number of characters parsed if successful.
1218 * - "<branch>@{upstream}" finds the name of the other ref that
1219 * <branch> is configured to merge with (missing <branch> defaults
1220 * to the current branch), and places the name of the branch in the
1221 * given buf and returns the number of characters parsed if
1222 * successful.
1224 * If the input is not of the accepted format, it returns a negative
1225 * number to signal an error.
1227 * If the input was ok but there are not N branch switches in the
1228 * reflog, it returns 0.
1230 int interpret_branch_name(const char *name, int namelen, struct strbuf *buf)
1232 char *at;
1233 const char *start;
1234 int len = interpret_nth_prior_checkout(name, namelen, buf);
1236 if (!namelen)
1237 namelen = strlen(name);
1239 if (!len) {
1240 return len; /* syntax Ok, not enough switches */
1241 } else if (len > 0) {
1242 if (len == namelen)
1243 return len; /* consumed all */
1244 else
1245 return reinterpret(name, namelen, len, buf);
1248 for (start = name;
1249 (at = memchr(start, '@', namelen - (start - name)));
1250 start = at + 1) {
1252 len = interpret_empty_at(name, namelen, at - name, buf);
1253 if (len > 0)
1254 return reinterpret(name, namelen, len, buf);
1256 len = interpret_branch_mark(name, namelen, at - name, buf,
1257 upstream_mark, branch_get_upstream);
1258 if (len > 0)
1259 return len;
1261 len = interpret_branch_mark(name, namelen, at - name, buf,
1262 push_mark, branch_get_push);
1263 if (len > 0)
1264 return len;
1267 return -1;
1270 int strbuf_branchname(struct strbuf *sb, const char *name)
1272 int len = strlen(name);
1273 int used = interpret_branch_name(name, len, sb);
1275 if (used == len)
1276 return 0;
1277 if (used < 0)
1278 used = 0;
1279 strbuf_add(sb, name + used, len - used);
1280 return len;
1283 int strbuf_check_branch_ref(struct strbuf *sb, const char *name)
1285 strbuf_branchname(sb, name);
1286 if (name[0] == '-')
1287 return -1;
1288 strbuf_splice(sb, 0, 0, "refs/heads/", 11);
1289 return check_refname_format(sb->buf, 0);
1293 * This is like "get_sha1_basic()", except it allows "sha1 expressions",
1294 * notably "xyz^" for "parent of xyz"
1296 int get_sha1(const char *name, unsigned char *sha1)
1298 struct object_context unused;
1299 return get_sha1_with_context(name, 0, sha1, &unused);
1303 * This is like "get_sha1()", but for struct object_id.
1305 int get_oid(const char *name, struct object_id *oid)
1307 return get_sha1(name, oid->hash);
1312 * Many callers know that the user meant to name a commit-ish by
1313 * syntactical positions where the object name appears. Calling this
1314 * function allows the machinery to disambiguate shorter-than-unique
1315 * abbreviated object names between commit-ish and others.
1317 * Note that this does NOT error out when the named object is not a
1318 * commit-ish. It is merely to give a hint to the disambiguation
1319 * machinery.
1321 int get_sha1_committish(const char *name, unsigned char *sha1)
1323 struct object_context unused;
1324 return get_sha1_with_context(name, GET_SHA1_COMMITTISH,
1325 sha1, &unused);
1328 int get_sha1_treeish(const char *name, unsigned char *sha1)
1330 struct object_context unused;
1331 return get_sha1_with_context(name, GET_SHA1_TREEISH,
1332 sha1, &unused);
1335 int get_sha1_commit(const char *name, unsigned char *sha1)
1337 struct object_context unused;
1338 return get_sha1_with_context(name, GET_SHA1_COMMIT,
1339 sha1, &unused);
1342 int get_sha1_tree(const char *name, unsigned char *sha1)
1344 struct object_context unused;
1345 return get_sha1_with_context(name, GET_SHA1_TREE,
1346 sha1, &unused);
1349 int get_sha1_blob(const char *name, unsigned char *sha1)
1351 struct object_context unused;
1352 return get_sha1_with_context(name, GET_SHA1_BLOB,
1353 sha1, &unused);
1356 /* Must be called only when object_name:filename doesn't exist. */
1357 static void diagnose_invalid_sha1_path(const char *prefix,
1358 const char *filename,
1359 const unsigned char *tree_sha1,
1360 const char *object_name,
1361 int object_name_len)
1363 unsigned char sha1[20];
1364 unsigned mode;
1366 if (!prefix)
1367 prefix = "";
1369 if (file_exists(filename))
1370 die("Path '%s' exists on disk, but not in '%.*s'.",
1371 filename, object_name_len, object_name);
1372 if (errno == ENOENT || errno == ENOTDIR) {
1373 char *fullname = xstrfmt("%s%s", prefix, filename);
1375 if (!get_tree_entry(tree_sha1, fullname,
1376 sha1, &mode)) {
1377 die("Path '%s' exists, but not '%s'.\n"
1378 "Did you mean '%.*s:%s' aka '%.*s:./%s'?",
1379 fullname,
1380 filename,
1381 object_name_len, object_name,
1382 fullname,
1383 object_name_len, object_name,
1384 filename);
1386 die("Path '%s' does not exist in '%.*s'",
1387 filename, object_name_len, object_name);
1391 /* Must be called only when :stage:filename doesn't exist. */
1392 static void diagnose_invalid_index_path(int stage,
1393 const char *prefix,
1394 const char *filename)
1396 const struct cache_entry *ce;
1397 int pos;
1398 unsigned namelen = strlen(filename);
1399 struct strbuf fullname = STRBUF_INIT;
1401 if (!prefix)
1402 prefix = "";
1404 /* Wrong stage number? */
1405 pos = cache_name_pos(filename, namelen);
1406 if (pos < 0)
1407 pos = -pos - 1;
1408 if (pos < active_nr) {
1409 ce = active_cache[pos];
1410 if (ce_namelen(ce) == namelen &&
1411 !memcmp(ce->name, filename, namelen))
1412 die("Path '%s' is in the index, but not at stage %d.\n"
1413 "Did you mean ':%d:%s'?",
1414 filename, stage,
1415 ce_stage(ce), filename);
1418 /* Confusion between relative and absolute filenames? */
1419 strbuf_addstr(&fullname, prefix);
1420 strbuf_addstr(&fullname, filename);
1421 pos = cache_name_pos(fullname.buf, fullname.len);
1422 if (pos < 0)
1423 pos = -pos - 1;
1424 if (pos < active_nr) {
1425 ce = active_cache[pos];
1426 if (ce_namelen(ce) == fullname.len &&
1427 !memcmp(ce->name, fullname.buf, fullname.len))
1428 die("Path '%s' is in the index, but not '%s'.\n"
1429 "Did you mean ':%d:%s' aka ':%d:./%s'?",
1430 fullname.buf, filename,
1431 ce_stage(ce), fullname.buf,
1432 ce_stage(ce), filename);
1435 if (file_exists(filename))
1436 die("Path '%s' exists on disk, but not in the index.", filename);
1437 if (errno == ENOENT || errno == ENOTDIR)
1438 die("Path '%s' does not exist (neither on disk nor in the index).",
1439 filename);
1441 strbuf_release(&fullname);
1445 static char *resolve_relative_path(const char *rel)
1447 if (!starts_with(rel, "./") && !starts_with(rel, "../"))
1448 return NULL;
1450 if (!is_inside_work_tree())
1451 die("relative path syntax can't be used outside working tree.");
1453 /* die() inside prefix_path() if resolved path is outside worktree */
1454 return prefix_path(startup_info->prefix,
1455 startup_info->prefix ? strlen(startup_info->prefix) : 0,
1456 rel);
1459 static int get_sha1_with_context_1(const char *name,
1460 unsigned flags,
1461 const char *prefix,
1462 unsigned char *sha1,
1463 struct object_context *oc)
1465 int ret, bracket_depth;
1466 int namelen = strlen(name);
1467 const char *cp;
1468 int only_to_die = flags & GET_SHA1_ONLY_TO_DIE;
1470 if (only_to_die)
1471 flags |= GET_SHA1_QUIETLY;
1473 memset(oc, 0, sizeof(*oc));
1474 oc->mode = S_IFINVALID;
1475 ret = get_sha1_1(name, namelen, sha1, flags);
1476 if (!ret)
1477 return ret;
1479 * sha1:path --> object name of path in ent sha1
1480 * :path -> object name of absolute path in index
1481 * :./path -> object name of path relative to cwd in index
1482 * :[0-3]:path -> object name of path in index at stage
1483 * :/foo -> recent commit matching foo
1485 if (name[0] == ':') {
1486 int stage = 0;
1487 const struct cache_entry *ce;
1488 char *new_path = NULL;
1489 int pos;
1490 if (!only_to_die && namelen > 2 && name[1] == '/') {
1491 struct commit_list *list = NULL;
1493 for_each_ref(handle_one_ref, &list);
1494 commit_list_sort_by_date(&list);
1495 return get_sha1_oneline(name + 2, sha1, list);
1497 if (namelen < 3 ||
1498 name[2] != ':' ||
1499 name[1] < '0' || '3' < name[1])
1500 cp = name + 1;
1501 else {
1502 stage = name[1] - '0';
1503 cp = name + 3;
1505 new_path = resolve_relative_path(cp);
1506 if (!new_path) {
1507 namelen = namelen - (cp - name);
1508 } else {
1509 cp = new_path;
1510 namelen = strlen(cp);
1513 strlcpy(oc->path, cp, sizeof(oc->path));
1515 if (!active_cache)
1516 read_cache();
1517 pos = cache_name_pos(cp, namelen);
1518 if (pos < 0)
1519 pos = -pos - 1;
1520 while (pos < active_nr) {
1521 ce = active_cache[pos];
1522 if (ce_namelen(ce) != namelen ||
1523 memcmp(ce->name, cp, namelen))
1524 break;
1525 if (ce_stage(ce) == stage) {
1526 hashcpy(sha1, ce->oid.hash);
1527 oc->mode = ce->ce_mode;
1528 free(new_path);
1529 return 0;
1531 pos++;
1533 if (only_to_die && name[1] && name[1] != '/')
1534 diagnose_invalid_index_path(stage, prefix, cp);
1535 free(new_path);
1536 return -1;
1538 for (cp = name, bracket_depth = 0; *cp; cp++) {
1539 if (*cp == '{')
1540 bracket_depth++;
1541 else if (bracket_depth && *cp == '}')
1542 bracket_depth--;
1543 else if (!bracket_depth && *cp == ':')
1544 break;
1546 if (*cp == ':') {
1547 unsigned char tree_sha1[20];
1548 int len = cp - name;
1549 unsigned sub_flags = flags;
1551 sub_flags &= ~GET_SHA1_DISAMBIGUATORS;
1552 sub_flags |= GET_SHA1_TREEISH;
1554 if (!get_sha1_1(name, len, tree_sha1, sub_flags)) {
1555 const char *filename = cp+1;
1556 char *new_filename = NULL;
1558 new_filename = resolve_relative_path(filename);
1559 if (new_filename)
1560 filename = new_filename;
1561 if (flags & GET_SHA1_FOLLOW_SYMLINKS) {
1562 ret = get_tree_entry_follow_symlinks(tree_sha1,
1563 filename, sha1, &oc->symlink_path,
1564 &oc->mode);
1565 } else {
1566 ret = get_tree_entry(tree_sha1, filename,
1567 sha1, &oc->mode);
1568 if (ret && only_to_die) {
1569 diagnose_invalid_sha1_path(prefix,
1570 filename,
1571 tree_sha1,
1572 name, len);
1575 hashcpy(oc->tree, tree_sha1);
1576 strlcpy(oc->path, filename, sizeof(oc->path));
1578 free(new_filename);
1579 return ret;
1580 } else {
1581 if (only_to_die)
1582 die("Invalid object name '%.*s'.", len, name);
1585 return ret;
1589 * Call this function when you know "name" given by the end user must
1590 * name an object but it doesn't; the function _may_ die with a better
1591 * diagnostic message than "no such object 'name'", e.g. "Path 'doc' does not
1592 * exist in 'HEAD'" when given "HEAD:doc", or it may return in which case
1593 * you have a chance to diagnose the error further.
1595 void maybe_die_on_misspelt_object_name(const char *name, const char *prefix)
1597 struct object_context oc;
1598 unsigned char sha1[20];
1599 get_sha1_with_context_1(name, GET_SHA1_ONLY_TO_DIE, prefix, sha1, &oc);
1602 int get_sha1_with_context(const char *str, unsigned flags, unsigned char *sha1, struct object_context *orc)
1604 if (flags & GET_SHA1_FOLLOW_SYMLINKS && flags & GET_SHA1_ONLY_TO_DIE)
1605 die("BUG: incompatible flags for get_sha1_with_context");
1606 return get_sha1_with_context_1(str, flags, NULL, sha1, orc);