Attempt to delay prepare_alt_odb during get_sha1
[git/jrn.git] / sha1_name.c
blob8dfceb2d7fc769c554e4b2d59cd41e7b04973533
1 #include "cache.h"
2 #include "tag.h"
3 #include "commit.h"
4 #include "tree.h"
5 #include "blob.h"
6 #include "tree-walk.h"
7 #include "refs.h"
9 static int find_short_object_filename(int len, const char *name, unsigned char *sha1)
11 struct alternate_object_database *alt;
12 char hex[40];
13 int found = 0;
14 static struct alternate_object_database *fakeent;
16 if (!fakeent) {
17 const char *objdir = get_object_directory();
18 int objdir_len = strlen(objdir);
19 int entlen = objdir_len + 43;
20 fakeent = xmalloc(sizeof(*fakeent) + entlen);
21 memcpy(fakeent->base, objdir, objdir_len);
22 fakeent->name = fakeent->base + objdir_len + 1;
23 fakeent->name[-1] = '/';
25 fakeent->next = alt_odb_list;
27 sprintf(hex, "%.2s", name);
28 for (alt = fakeent; alt && found < 2; alt = alt->next) {
29 struct dirent *de;
30 DIR *dir;
31 sprintf(alt->name, "%.2s/", name);
32 dir = opendir(alt->base);
33 if (!dir)
34 continue;
35 while ((de = readdir(dir)) != NULL) {
36 if (strlen(de->d_name) != 38)
37 continue;
38 if (memcmp(de->d_name, name + 2, len - 2))
39 continue;
40 if (!found) {
41 memcpy(hex + 2, de->d_name, 38);
42 found++;
44 else if (memcmp(hex + 2, de->d_name, 38)) {
45 found = 2;
46 break;
49 closedir(dir);
51 if (found == 1)
52 return get_sha1_hex(hex, sha1) == 0;
53 return found;
56 static int match_sha(unsigned len, const unsigned char *a, const unsigned char *b)
58 do {
59 if (*a != *b)
60 return 0;
61 a++;
62 b++;
63 len -= 2;
64 } while (len > 1);
65 if (len)
66 if ((*a ^ *b) & 0xf0)
67 return 0;
68 return 1;
71 static int find_short_packed_object(int len, const unsigned char *match, unsigned char *sha1)
73 struct packed_git *p;
74 const unsigned char *found_sha1 = NULL;
75 int found = 0;
77 prepare_packed_git();
78 for (p = packed_git; p && found < 2; p = p->next) {
79 uint32_t num = p->num_objects;
80 uint32_t first = 0, last = num;
81 while (first < last) {
82 uint32_t mid = (first + last) / 2;
83 const unsigned char *now;
84 int cmp;
86 now = nth_packed_object_sha1(p, mid);
87 cmp = hashcmp(match, now);
88 if (!cmp) {
89 first = mid;
90 break;
92 if (cmp > 0) {
93 first = mid+1;
94 continue;
96 last = mid;
98 if (first < num) {
99 const unsigned char *now, *next;
100 now = nth_packed_object_sha1(p, first);
101 if (match_sha(len, match, now)) {
102 next = nth_packed_object_sha1(p, first+1);
103 if (!next|| !match_sha(len, match, next)) {
104 /* unique within this pack */
105 if (!found) {
106 found_sha1 = now;
107 found++;
109 else if (hashcmp(found_sha1, now)) {
110 found = 2;
111 break;
114 else {
115 /* not even unique within this pack */
116 found = 2;
117 break;
122 if (found == 1)
123 hashcpy(sha1, found_sha1);
124 return found;
127 #define SHORT_NAME_NOT_FOUND (-1)
128 #define SHORT_NAME_AMBIGUOUS (-2)
130 static int find_unique_short_object(int len, char *canonical,
131 unsigned char *res, unsigned char *sha1)
133 int has_unpacked, has_packed;
134 unsigned char unpacked_sha1[20], packed_sha1[20];
136 prepare_alt_odb();
137 has_unpacked = find_short_object_filename(len, canonical, unpacked_sha1);
138 has_packed = find_short_packed_object(len, res, packed_sha1);
139 if (!has_unpacked && !has_packed)
140 return SHORT_NAME_NOT_FOUND;
141 if (1 < has_unpacked || 1 < has_packed)
142 return SHORT_NAME_AMBIGUOUS;
143 if (has_unpacked != has_packed) {
144 hashcpy(sha1, (has_packed ? packed_sha1 : unpacked_sha1));
145 return 0;
147 /* Both have unique ones -- do they match? */
148 if (hashcmp(packed_sha1, unpacked_sha1))
149 return SHORT_NAME_AMBIGUOUS;
150 hashcpy(sha1, packed_sha1);
151 return 0;
154 static int get_short_sha1(const char *name, int len, unsigned char *sha1,
155 int quietly)
157 int i, status;
158 char canonical[40];
159 unsigned char res[20];
161 if (len < MINIMUM_ABBREV || len > 40)
162 return -1;
163 hashclr(res);
164 memset(canonical, 'x', 40);
165 for (i = 0; i < len ;i++) {
166 unsigned char c = name[i];
167 unsigned char val;
168 if (c >= '0' && c <= '9')
169 val = c - '0';
170 else if (c >= 'a' && c <= 'f')
171 val = c - 'a' + 10;
172 else if (c >= 'A' && c <='F') {
173 val = c - 'A' + 10;
174 c -= 'A' - 'a';
176 else
177 return -1;
178 canonical[i] = c;
179 if (!(i & 1))
180 val <<= 4;
181 res[i >> 1] |= val;
184 status = find_unique_short_object(i, canonical, res, sha1);
185 if (!quietly && (status == SHORT_NAME_AMBIGUOUS))
186 return error("short SHA1 %.*s is ambiguous.", len, canonical);
187 return status;
190 const char *find_unique_abbrev(const unsigned char *sha1, int len)
192 int status, is_null;
193 static char hex[41];
195 is_null = is_null_sha1(sha1);
196 memcpy(hex, sha1_to_hex(sha1), 40);
197 if (len == 40 || !len)
198 return hex;
199 while (len < 40) {
200 unsigned char sha1_ret[20];
201 status = get_short_sha1(hex, len, sha1_ret, 1);
202 if (!status ||
203 (is_null && status != SHORT_NAME_AMBIGUOUS)) {
204 hex[len] = 0;
205 return hex;
207 if (status != SHORT_NAME_AMBIGUOUS)
208 return NULL;
209 len++;
211 return NULL;
214 static int ambiguous_path(const char *path, int len)
216 int slash = 1;
217 int cnt;
219 for (cnt = 0; cnt < len; cnt++) {
220 switch (*path++) {
221 case '\0':
222 break;
223 case '/':
224 if (slash)
225 break;
226 slash = 1;
227 continue;
228 case '.':
229 continue;
230 default:
231 slash = 0;
232 continue;
234 break;
236 return slash;
239 static const char *ref_fmt[] = {
240 "%.*s",
241 "refs/%.*s",
242 "refs/tags/%.*s",
243 "refs/heads/%.*s",
244 "refs/remotes/%.*s",
245 "refs/remotes/%.*s/HEAD",
246 NULL
249 int dwim_ref(const char *str, int len, unsigned char *sha1, char **ref)
251 const char **p, *r;
252 int refs_found = 0;
254 *ref = NULL;
255 for (p = ref_fmt; *p; p++) {
256 unsigned char sha1_from_ref[20];
257 unsigned char *this_result;
259 this_result = refs_found ? sha1_from_ref : sha1;
260 r = resolve_ref(mkpath(*p, len, str), this_result, 1, NULL);
261 if (r) {
262 if (!refs_found++)
263 *ref = xstrdup(r);
264 if (!warn_ambiguous_refs)
265 break;
268 return refs_found;
271 int dwim_log(const char *str, int len, unsigned char *sha1, char **log)
273 const char **p;
274 int logs_found = 0;
276 *log = NULL;
277 for (p = ref_fmt; *p; p++) {
278 struct stat st;
279 unsigned char hash[20];
280 char path[PATH_MAX];
281 const char *ref, *it;
283 strcpy(path, mkpath(*p, len, str));
284 ref = resolve_ref(path, hash, 0, NULL);
285 if (!ref)
286 continue;
287 if (!stat(git_path("logs/%s", path), &st) &&
288 S_ISREG(st.st_mode))
289 it = path;
290 else if (strcmp(ref, path) &&
291 !stat(git_path("logs/%s", ref), &st) &&
292 S_ISREG(st.st_mode))
293 it = ref;
294 else
295 continue;
296 if (!logs_found++) {
297 *log = xstrdup(it);
298 hashcpy(sha1, hash);
300 if (!warn_ambiguous_refs)
301 break;
303 return logs_found;
306 static int get_sha1_basic(const char *str, int len, unsigned char *sha1)
308 static const char *warning = "warning: refname '%.*s' is ambiguous.\n";
309 char *real_ref = NULL;
310 int refs_found = 0;
311 int at, reflog_len;
313 if (len == 40 && !get_sha1_hex(str, sha1))
314 return 0;
316 /* basic@{time or number} format to query ref-log */
317 reflog_len = at = 0;
318 if (str[len-1] == '}') {
319 for (at = 0; at < len - 1; at++) {
320 if (str[at] == '@' && str[at+1] == '{') {
321 reflog_len = (len-1) - (at+2);
322 len = at;
323 break;
328 /* Accept only unambiguous ref paths. */
329 if (len && ambiguous_path(str, len))
330 return -1;
332 if (!len && reflog_len) {
333 /* allow "@{...}" to mean the current branch reflog */
334 refs_found = dwim_ref("HEAD", 4, sha1, &real_ref);
335 } else if (reflog_len)
336 refs_found = dwim_log(str, len, sha1, &real_ref);
337 else
338 refs_found = dwim_ref(str, len, sha1, &real_ref);
340 if (!refs_found)
341 return -1;
343 if (warn_ambiguous_refs && refs_found > 1)
344 fprintf(stderr, warning, len, str);
346 if (reflog_len) {
347 int nth, i;
348 unsigned long at_time;
349 unsigned long co_time;
350 int co_tz, co_cnt;
352 /* Is it asking for N-th entry, or approxidate? */
353 for (i = nth = 0; 0 <= nth && i < reflog_len; i++) {
354 char ch = str[at+2+i];
355 if ('0' <= ch && ch <= '9')
356 nth = nth * 10 + ch - '0';
357 else
358 nth = -1;
360 if (0 <= nth)
361 at_time = 0;
362 else
363 at_time = approxidate(str + at + 2);
364 if (read_ref_at(real_ref, at_time, nth, sha1, NULL,
365 &co_time, &co_tz, &co_cnt)) {
366 if (at_time)
367 fprintf(stderr,
368 "warning: Log for '%.*s' only goes "
369 "back to %s.\n", len, str,
370 show_rfc2822_date(co_time, co_tz));
371 else
372 fprintf(stderr,
373 "warning: Log for '%.*s' only has "
374 "%d entries.\n", len, str, co_cnt);
378 free(real_ref);
379 return 0;
382 static int get_sha1_1(const char *name, int len, unsigned char *sha1);
384 static int get_parent(const char *name, int len,
385 unsigned char *result, int idx)
387 unsigned char sha1[20];
388 int ret = get_sha1_1(name, len, sha1);
389 struct commit *commit;
390 struct commit_list *p;
392 if (ret)
393 return ret;
394 commit = lookup_commit_reference(sha1);
395 if (!commit)
396 return -1;
397 if (parse_commit(commit))
398 return -1;
399 if (!idx) {
400 hashcpy(result, commit->object.sha1);
401 return 0;
403 p = commit->parents;
404 while (p) {
405 if (!--idx) {
406 hashcpy(result, p->item->object.sha1);
407 return 0;
409 p = p->next;
411 return -1;
414 static int get_nth_ancestor(const char *name, int len,
415 unsigned char *result, int generation)
417 unsigned char sha1[20];
418 int ret = get_sha1_1(name, len, sha1);
419 if (ret)
420 return ret;
422 while (generation--) {
423 struct commit *commit = lookup_commit_reference(sha1);
425 if (!commit || parse_commit(commit) || !commit->parents)
426 return -1;
427 hashcpy(sha1, commit->parents->item->object.sha1);
429 hashcpy(result, sha1);
430 return 0;
433 static int peel_onion(const char *name, int len, unsigned char *sha1)
435 unsigned char outer[20];
436 const char *sp;
437 unsigned int expected_type = 0;
438 struct object *o;
441 * "ref^{type}" dereferences ref repeatedly until you cannot
442 * dereference anymore, or you get an object of given type,
443 * whichever comes first. "ref^{}" means just dereference
444 * tags until you get a non-tag. "ref^0" is a shorthand for
445 * "ref^{commit}". "commit^{tree}" could be used to find the
446 * top-level tree of the given commit.
448 if (len < 4 || name[len-1] != '}')
449 return -1;
451 for (sp = name + len - 1; name <= sp; sp--) {
452 int ch = *sp;
453 if (ch == '{' && name < sp && sp[-1] == '^')
454 break;
456 if (sp <= name)
457 return -1;
459 sp++; /* beginning of type name, or closing brace for empty */
460 if (!strncmp(commit_type, sp, 6) && sp[6] == '}')
461 expected_type = OBJ_COMMIT;
462 else if (!strncmp(tree_type, sp, 4) && sp[4] == '}')
463 expected_type = OBJ_TREE;
464 else if (!strncmp(blob_type, sp, 4) && sp[4] == '}')
465 expected_type = OBJ_BLOB;
466 else if (sp[0] == '}')
467 expected_type = OBJ_NONE;
468 else
469 return -1;
471 if (get_sha1_1(name, sp - name - 2, outer))
472 return -1;
474 o = parse_object(outer);
475 if (!o)
476 return -1;
477 if (!expected_type) {
478 o = deref_tag(o, name, sp - name - 2);
479 if (!o || (!o->parsed && !parse_object(o->sha1)))
480 return -1;
481 hashcpy(sha1, o->sha1);
483 else {
484 /* At this point, the syntax look correct, so
485 * if we do not get the needed object, we should
486 * barf.
489 while (1) {
490 if (!o || (!o->parsed && !parse_object(o->sha1)))
491 return -1;
492 if (o->type == expected_type) {
493 hashcpy(sha1, o->sha1);
494 return 0;
496 if (o->type == OBJ_TAG)
497 o = ((struct tag*) o)->tagged;
498 else if (o->type == OBJ_COMMIT)
499 o = &(((struct commit *) o)->tree->object);
500 else
501 return error("%.*s: expected %s type, but the object dereferences to %s type",
502 len, name, typename(expected_type),
503 typename(o->type));
504 if (!o->parsed)
505 parse_object(o->sha1);
508 return 0;
511 static int get_describe_name(const char *name, int len, unsigned char *sha1)
513 const char *cp;
515 for (cp = name + len - 1; name + 2 <= cp; cp--) {
516 char ch = *cp;
517 if (hexval(ch) & ~0377) {
518 /* We must be looking at g in "SOMETHING-g"
519 * for it to be describe output.
521 if (ch == 'g' && cp[-1] == '-') {
522 cp++;
523 len -= cp - name;
524 return get_short_sha1(cp, len, sha1, 1);
528 return -1;
531 static int get_sha1_1(const char *name, int len, unsigned char *sha1)
533 int ret, has_suffix;
534 const char *cp;
536 /* "name~3" is "name^^^",
537 * "name~" and "name~0" are name -- not "name^0"!
538 * "name^" is not "name^0"; it is "name^1".
540 has_suffix = 0;
541 for (cp = name + len - 1; name <= cp; cp--) {
542 int ch = *cp;
543 if ('0' <= ch && ch <= '9')
544 continue;
545 if (ch == '~' || ch == '^')
546 has_suffix = ch;
547 break;
550 if (has_suffix) {
551 int num = 0;
552 int len1 = cp - name;
553 cp++;
554 while (cp < name + len)
555 num = num * 10 + *cp++ - '0';
556 if (has_suffix == '^') {
557 if (!num && len1 == len - 1)
558 num = 1;
559 return get_parent(name, len1, sha1, num);
561 /* else if (has_suffix == '~') -- goes without saying */
562 return get_nth_ancestor(name, len1, sha1, num);
565 ret = peel_onion(name, len, sha1);
566 if (!ret)
567 return 0;
569 ret = get_sha1_basic(name, len, sha1);
570 if (!ret)
571 return 0;
573 /* It could be describe output that is "SOMETHING-gXXXX" */
574 ret = get_describe_name(name, len, sha1);
575 if (!ret)
576 return 0;
578 return get_short_sha1(name, len, sha1, 0);
581 static int handle_one_ref(const char *path,
582 const unsigned char *sha1, int flag, void *cb_data)
584 struct commit_list **list = cb_data;
585 struct object *object = parse_object(sha1);
586 if (!object)
587 return 0;
588 if (object->type == OBJ_TAG)
589 object = deref_tag(object, path, strlen(path));
590 if (object->type != OBJ_COMMIT)
591 return 0;
592 insert_by_date((struct commit *)object, list);
593 return 0;
597 * This interprets names like ':/Initial revision of "git"' by searching
598 * through history and returning the first commit whose message starts
599 * with the given string.
601 * For future extension, ':/!' is reserved. If you want to match a message
602 * beginning with a '!', you have to repeat the exclamation mark.
605 #define ONELINE_SEEN (1u<<20)
606 static int get_sha1_oneline(const char *prefix, unsigned char *sha1)
608 struct commit_list *list = NULL, *backup = NULL, *l;
609 int retval = -1;
611 if (prefix[0] == '!') {
612 if (prefix[1] != '!')
613 die ("Invalid search pattern: %s", prefix);
614 prefix++;
616 if (!save_commit_buffer)
617 return error("Could not expand oneline-name.");
618 for_each_ref(handle_one_ref, &list);
619 for (l = list; l; l = l->next)
620 commit_list_insert(l->item, &backup);
621 while (list) {
622 char *p;
623 struct commit *commit;
625 commit = pop_most_recent_commit(&list, ONELINE_SEEN);
626 parse_object(commit->object.sha1);
627 if (!commit->buffer || !(p = strstr(commit->buffer, "\n\n")))
628 continue;
629 if (!prefixcmp(p + 2, prefix)) {
630 hashcpy(sha1, commit->object.sha1);
631 retval = 0;
632 break;
635 free_commit_list(list);
636 for (l = backup; l; l = l->next)
637 clear_commit_marks(l->item, ONELINE_SEEN);
638 return retval;
642 * This is like "get_sha1_basic()", except it allows "sha1 expressions",
643 * notably "xyz^" for "parent of xyz"
645 int get_sha1(const char *name, unsigned char *sha1)
647 unsigned unused;
648 return get_sha1_with_mode(name, sha1, &unused);
651 int get_sha1_with_mode(const char *name, unsigned char *sha1, unsigned *mode)
653 int ret, bracket_depth;
654 int namelen = strlen(name);
655 const char *cp;
657 *mode = S_IFINVALID;
658 ret = get_sha1_1(name, namelen, sha1);
659 if (!ret)
660 return ret;
661 /* sha1:path --> object name of path in ent sha1
662 * :path -> object name of path in index
663 * :[0-3]:path -> object name of path in index at stage
665 if (name[0] == ':') {
666 int stage = 0;
667 struct cache_entry *ce;
668 int pos;
669 if (namelen > 2 && name[1] == '/')
670 return get_sha1_oneline(name + 2, sha1);
671 if (namelen < 3 ||
672 name[2] != ':' ||
673 name[1] < '0' || '3' < name[1])
674 cp = name + 1;
675 else {
676 stage = name[1] - '0';
677 cp = name + 3;
679 namelen = namelen - (cp - name);
680 if (!active_cache)
681 read_cache();
682 if (active_nr < 0)
683 return -1;
684 pos = cache_name_pos(cp, namelen);
685 if (pos < 0)
686 pos = -pos - 1;
687 while (pos < active_nr) {
688 ce = active_cache[pos];
689 if (ce_namelen(ce) != namelen ||
690 memcmp(ce->name, cp, namelen))
691 break;
692 if (ce_stage(ce) == stage) {
693 hashcpy(sha1, ce->sha1);
694 *mode = ntohl(ce->ce_mode);
695 return 0;
697 pos++;
699 return -1;
701 for (cp = name, bracket_depth = 0; *cp; cp++) {
702 if (*cp == '{')
703 bracket_depth++;
704 else if (bracket_depth && *cp == '}')
705 bracket_depth--;
706 else if (!bracket_depth && *cp == ':')
707 break;
709 if (*cp == ':') {
710 unsigned char tree_sha1[20];
711 if (!get_sha1_1(name, cp-name, tree_sha1))
712 return get_tree_entry(tree_sha1, cp+1, sha1,
713 mode);
715 return ret;