gitweb: Add charset info to "raw" output of 'text/plain' blobs
[git/jnareb-git/bp-gitweb.git] / sha1_name.c
blobb0b2167578a7baebeba676af0b33161fb688bae0
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, last;
80 uint32_t first = 0;
81 open_pack_index(p);
82 num = p->num_objects;
83 last = num;
84 while (first < last) {
85 uint32_t mid = (first + last) / 2;
86 const unsigned char *now;
87 int cmp;
89 now = nth_packed_object_sha1(p, mid);
90 cmp = hashcmp(match, now);
91 if (!cmp) {
92 first = mid;
93 break;
95 if (cmp > 0) {
96 first = mid+1;
97 continue;
99 last = mid;
101 if (first < num) {
102 const unsigned char *now, *next;
103 now = nth_packed_object_sha1(p, first);
104 if (match_sha(len, match, now)) {
105 next = nth_packed_object_sha1(p, first+1);
106 if (!next|| !match_sha(len, match, next)) {
107 /* unique within this pack */
108 if (!found) {
109 found_sha1 = now;
110 found++;
112 else if (hashcmp(found_sha1, now)) {
113 found = 2;
114 break;
117 else {
118 /* not even unique within this pack */
119 found = 2;
120 break;
125 if (found == 1)
126 hashcpy(sha1, found_sha1);
127 return found;
130 #define SHORT_NAME_NOT_FOUND (-1)
131 #define SHORT_NAME_AMBIGUOUS (-2)
133 static int find_unique_short_object(int len, char *canonical,
134 unsigned char *res, unsigned char *sha1)
136 int has_unpacked, has_packed;
137 unsigned char unpacked_sha1[20], packed_sha1[20];
139 prepare_alt_odb();
140 has_unpacked = find_short_object_filename(len, canonical, unpacked_sha1);
141 has_packed = find_short_packed_object(len, res, packed_sha1);
142 if (!has_unpacked && !has_packed)
143 return SHORT_NAME_NOT_FOUND;
144 if (1 < has_unpacked || 1 < has_packed)
145 return SHORT_NAME_AMBIGUOUS;
146 if (has_unpacked != has_packed) {
147 hashcpy(sha1, (has_packed ? packed_sha1 : unpacked_sha1));
148 return 0;
150 /* Both have unique ones -- do they match? */
151 if (hashcmp(packed_sha1, unpacked_sha1))
152 return SHORT_NAME_AMBIGUOUS;
153 hashcpy(sha1, packed_sha1);
154 return 0;
157 static int get_short_sha1(const char *name, int len, unsigned char *sha1,
158 int quietly)
160 int i, status;
161 char canonical[40];
162 unsigned char res[20];
164 if (len < MINIMUM_ABBREV || len > 40)
165 return -1;
166 hashclr(res);
167 memset(canonical, 'x', 40);
168 for (i = 0; i < len ;i++) {
169 unsigned char c = name[i];
170 unsigned char val;
171 if (c >= '0' && c <= '9')
172 val = c - '0';
173 else if (c >= 'a' && c <= 'f')
174 val = c - 'a' + 10;
175 else if (c >= 'A' && c <='F') {
176 val = c - 'A' + 10;
177 c -= 'A' - 'a';
179 else
180 return -1;
181 canonical[i] = c;
182 if (!(i & 1))
183 val <<= 4;
184 res[i >> 1] |= val;
187 status = find_unique_short_object(i, canonical, res, sha1);
188 if (!quietly && (status == SHORT_NAME_AMBIGUOUS))
189 return error("short SHA1 %.*s is ambiguous.", len, canonical);
190 return status;
193 const char *find_unique_abbrev(const unsigned char *sha1, int len)
195 int status, exists;
196 static char hex[41];
198 exists = has_sha1_file(sha1);
199 memcpy(hex, sha1_to_hex(sha1), 40);
200 if (len == 40 || !len)
201 return hex;
202 while (len < 40) {
203 unsigned char sha1_ret[20];
204 status = get_short_sha1(hex, len, sha1_ret, 1);
205 if (exists
206 ? !status
207 : status == SHORT_NAME_NOT_FOUND) {
208 hex[len] = 0;
209 return hex;
211 len++;
213 return hex;
216 static int ambiguous_path(const char *path, int len)
218 int slash = 1;
219 int cnt;
221 for (cnt = 0; cnt < len; cnt++) {
222 switch (*path++) {
223 case '\0':
224 break;
225 case '/':
226 if (slash)
227 break;
228 slash = 1;
229 continue;
230 case '.':
231 continue;
232 default:
233 slash = 0;
234 continue;
236 break;
238 return slash;
241 int dwim_ref(const char *str, int len, unsigned char *sha1, char **ref)
243 const char **p, *r;
244 int refs_found = 0;
246 *ref = NULL;
247 for (p = ref_rev_parse_rules; *p; p++) {
248 unsigned char sha1_from_ref[20];
249 unsigned char *this_result;
251 this_result = refs_found ? sha1_from_ref : sha1;
252 r = resolve_ref(mkpath(*p, len, str), this_result, 1, NULL);
253 if (r) {
254 if (!refs_found++)
255 *ref = xstrdup(r);
256 if (!warn_ambiguous_refs)
257 break;
260 return refs_found;
263 int dwim_log(const char *str, int len, unsigned char *sha1, char **log)
265 const char **p;
266 int logs_found = 0;
268 *log = NULL;
269 for (p = ref_rev_parse_rules; *p; p++) {
270 struct stat st;
271 unsigned char hash[20];
272 char path[PATH_MAX];
273 const char *ref, *it;
275 strcpy(path, mkpath(*p, len, str));
276 ref = resolve_ref(path, hash, 0, NULL);
277 if (!ref)
278 continue;
279 if (!stat(git_path("logs/%s", path), &st) &&
280 S_ISREG(st.st_mode))
281 it = path;
282 else if (strcmp(ref, path) &&
283 !stat(git_path("logs/%s", ref), &st) &&
284 S_ISREG(st.st_mode))
285 it = ref;
286 else
287 continue;
288 if (!logs_found++) {
289 *log = xstrdup(it);
290 hashcpy(sha1, hash);
292 if (!warn_ambiguous_refs)
293 break;
295 return logs_found;
298 static int get_sha1_basic(const char *str, int len, unsigned char *sha1)
300 static const char *warning = "warning: refname '%.*s' is ambiguous.\n";
301 char *real_ref = NULL;
302 int refs_found = 0;
303 int at, reflog_len;
305 if (len == 40 && !get_sha1_hex(str, sha1))
306 return 0;
308 /* basic@{time or number} format to query ref-log */
309 reflog_len = at = 0;
310 if (str[len-1] == '}') {
311 for (at = 0; at < len - 1; at++) {
312 if (str[at] == '@' && str[at+1] == '{') {
313 reflog_len = (len-1) - (at+2);
314 len = at;
315 break;
320 /* Accept only unambiguous ref paths. */
321 if (len && ambiguous_path(str, len))
322 return -1;
324 if (!len && reflog_len) {
325 /* allow "@{...}" to mean the current branch reflog */
326 refs_found = dwim_ref("HEAD", 4, sha1, &real_ref);
327 } else if (reflog_len)
328 refs_found = dwim_log(str, len, sha1, &real_ref);
329 else
330 refs_found = dwim_ref(str, len, sha1, &real_ref);
332 if (!refs_found)
333 return -1;
335 if (warn_ambiguous_refs && refs_found > 1)
336 fprintf(stderr, warning, len, str);
338 if (reflog_len) {
339 int nth, i;
340 unsigned long at_time;
341 unsigned long co_time;
342 int co_tz, co_cnt;
344 /* Is it asking for N-th entry, or approxidate? */
345 for (i = nth = 0; 0 <= nth && i < reflog_len; i++) {
346 char ch = str[at+2+i];
347 if ('0' <= ch && ch <= '9')
348 nth = nth * 10 + ch - '0';
349 else
350 nth = -1;
352 if (0 <= nth)
353 at_time = 0;
354 else {
355 char *tmp = xstrndup(str + at + 2, reflog_len);
356 at_time = approxidate(tmp);
357 free(tmp);
359 if (read_ref_at(real_ref, at_time, nth, sha1, NULL,
360 &co_time, &co_tz, &co_cnt)) {
361 if (at_time)
362 fprintf(stderr,
363 "warning: Log for '%.*s' only goes "
364 "back to %s.\n", len, str,
365 show_date(co_time, co_tz, DATE_RFC2822));
366 else
367 fprintf(stderr,
368 "warning: Log for '%.*s' only has "
369 "%d entries.\n", len, str, co_cnt);
373 free(real_ref);
374 return 0;
377 static int get_sha1_1(const char *name, int len, unsigned char *sha1);
379 static int get_parent(const char *name, int len,
380 unsigned char *result, int idx)
382 unsigned char sha1[20];
383 int ret = get_sha1_1(name, len, sha1);
384 struct commit *commit;
385 struct commit_list *p;
387 if (ret)
388 return ret;
389 commit = lookup_commit_reference(sha1);
390 if (!commit)
391 return -1;
392 if (parse_commit(commit))
393 return -1;
394 if (!idx) {
395 hashcpy(result, commit->object.sha1);
396 return 0;
398 p = commit->parents;
399 while (p) {
400 if (!--idx) {
401 hashcpy(result, p->item->object.sha1);
402 return 0;
404 p = p->next;
406 return -1;
409 static int get_nth_ancestor(const char *name, int len,
410 unsigned char *result, int generation)
412 unsigned char sha1[20];
413 struct commit *commit;
414 int ret;
416 ret = get_sha1_1(name, len, sha1);
417 if (ret)
418 return ret;
419 commit = lookup_commit_reference(sha1);
420 if (!commit)
421 return -1;
423 while (generation--) {
424 if (parse_commit(commit) || !commit->parents)
425 return -1;
426 commit = commit->parents->item;
428 hashcpy(result, commit->object.sha1);
429 return 0;
432 struct object *peel_to_type(const char *name, int namelen,
433 struct object *o, enum object_type expected_type)
435 if (name && !namelen)
436 namelen = strlen(name);
437 if (!o) {
438 unsigned char sha1[20];
439 if (get_sha1_1(name, namelen, sha1))
440 return NULL;
441 o = parse_object(sha1);
443 while (1) {
444 if (!o || (!o->parsed && !parse_object(o->sha1)))
445 return NULL;
446 if (o->type == expected_type)
447 return o;
448 if (o->type == OBJ_TAG)
449 o = ((struct tag*) o)->tagged;
450 else if (o->type == OBJ_COMMIT)
451 o = &(((struct commit *) o)->tree->object);
452 else {
453 if (name)
454 error("%.*s: expected %s type, but the object "
455 "dereferences to %s type",
456 namelen, name, typename(expected_type),
457 typename(o->type));
458 return NULL;
463 static int peel_onion(const char *name, int len, unsigned char *sha1)
465 unsigned char outer[20];
466 const char *sp;
467 unsigned int expected_type = 0;
468 struct object *o;
471 * "ref^{type}" dereferences ref repeatedly until you cannot
472 * dereference anymore, or you get an object of given type,
473 * whichever comes first. "ref^{}" means just dereference
474 * tags until you get a non-tag. "ref^0" is a shorthand for
475 * "ref^{commit}". "commit^{tree}" could be used to find the
476 * top-level tree of the given commit.
478 if (len < 4 || name[len-1] != '}')
479 return -1;
481 for (sp = name + len - 1; name <= sp; sp--) {
482 int ch = *sp;
483 if (ch == '{' && name < sp && sp[-1] == '^')
484 break;
486 if (sp <= name)
487 return -1;
489 sp++; /* beginning of type name, or closing brace for empty */
490 if (!strncmp(commit_type, sp, 6) && sp[6] == '}')
491 expected_type = OBJ_COMMIT;
492 else if (!strncmp(tree_type, sp, 4) && sp[4] == '}')
493 expected_type = OBJ_TREE;
494 else if (!strncmp(blob_type, sp, 4) && sp[4] == '}')
495 expected_type = OBJ_BLOB;
496 else if (sp[0] == '}')
497 expected_type = OBJ_NONE;
498 else
499 return -1;
501 if (get_sha1_1(name, sp - name - 2, outer))
502 return -1;
504 o = parse_object(outer);
505 if (!o)
506 return -1;
507 if (!expected_type) {
508 o = deref_tag(o, name, sp - name - 2);
509 if (!o || (!o->parsed && !parse_object(o->sha1)))
510 return -1;
511 hashcpy(sha1, o->sha1);
513 else {
515 * At this point, the syntax look correct, so
516 * if we do not get the needed object, we should
517 * barf.
519 o = peel_to_type(name, len, o, expected_type);
520 if (o) {
521 hashcpy(sha1, o->sha1);
522 return 0;
524 return -1;
526 return 0;
529 static int get_describe_name(const char *name, int len, unsigned char *sha1)
531 const char *cp;
533 for (cp = name + len - 1; name + 2 <= cp; cp--) {
534 char ch = *cp;
535 if (hexval(ch) & ~0377) {
536 /* We must be looking at g in "SOMETHING-g"
537 * for it to be describe output.
539 if (ch == 'g' && cp[-1] == '-') {
540 cp++;
541 len -= cp - name;
542 return get_short_sha1(cp, len, sha1, 1);
546 return -1;
549 static int get_sha1_1(const char *name, int len, unsigned char *sha1)
551 int ret, has_suffix;
552 const char *cp;
555 * "name~3" is "name^^^", "name~" is "name~1", and "name^" is "name^1".
557 has_suffix = 0;
558 for (cp = name + len - 1; name <= cp; cp--) {
559 int ch = *cp;
560 if ('0' <= ch && ch <= '9')
561 continue;
562 if (ch == '~' || ch == '^')
563 has_suffix = ch;
564 break;
567 if (has_suffix) {
568 int num = 0;
569 int len1 = cp - name;
570 cp++;
571 while (cp < name + len)
572 num = num * 10 + *cp++ - '0';
573 if (!num && len1 == len - 1)
574 num = 1;
575 if (has_suffix == '^')
576 return get_parent(name, len1, sha1, num);
577 /* else if (has_suffix == '~') -- goes without saying */
578 return get_nth_ancestor(name, len1, sha1, num);
581 ret = peel_onion(name, len, sha1);
582 if (!ret)
583 return 0;
585 ret = get_sha1_basic(name, len, sha1);
586 if (!ret)
587 return 0;
589 /* It could be describe output that is "SOMETHING-gXXXX" */
590 ret = get_describe_name(name, len, sha1);
591 if (!ret)
592 return 0;
594 return get_short_sha1(name, len, sha1, 0);
597 static int handle_one_ref(const char *path,
598 const unsigned char *sha1, int flag, void *cb_data)
600 struct commit_list **list = cb_data;
601 struct object *object = parse_object(sha1);
602 if (!object)
603 return 0;
604 if (object->type == OBJ_TAG) {
605 object = deref_tag(object, path, strlen(path));
606 if (!object)
607 return 0;
609 if (object->type != OBJ_COMMIT)
610 return 0;
611 insert_by_date((struct commit *)object, list);
612 return 0;
616 * This interprets names like ':/Initial revision of "git"' by searching
617 * through history and returning the first commit whose message starts
618 * with the given string.
620 * For future extension, ':/!' is reserved. If you want to match a message
621 * beginning with a '!', you have to repeat the exclamation mark.
624 #define ONELINE_SEEN (1u<<20)
625 static int get_sha1_oneline(const char *prefix, unsigned char *sha1)
627 struct commit_list *list = NULL, *backup = NULL, *l;
628 int retval = -1;
629 char *temp_commit_buffer = NULL;
631 if (prefix[0] == '!') {
632 if (prefix[1] != '!')
633 die ("Invalid search pattern: %s", prefix);
634 prefix++;
636 for_each_ref(handle_one_ref, &list);
637 for (l = list; l; l = l->next)
638 commit_list_insert(l->item, &backup);
639 while (list) {
640 char *p;
641 struct commit *commit;
642 enum object_type type;
643 unsigned long size;
645 commit = pop_most_recent_commit(&list, ONELINE_SEEN);
646 if (!parse_object(commit->object.sha1))
647 continue;
648 free(temp_commit_buffer);
649 if (commit->buffer)
650 p = commit->buffer;
651 else {
652 p = read_sha1_file(commit->object.sha1, &type, &size);
653 if (!p)
654 continue;
655 temp_commit_buffer = p;
657 if (!(p = strstr(p, "\n\n")))
658 continue;
659 if (!prefixcmp(p + 2, prefix)) {
660 hashcpy(sha1, commit->object.sha1);
661 retval = 0;
662 break;
665 free(temp_commit_buffer);
666 free_commit_list(list);
667 for (l = backup; l; l = l->next)
668 clear_commit_marks(l->item, ONELINE_SEEN);
669 return retval;
673 * This is like "get_sha1_basic()", except it allows "sha1 expressions",
674 * notably "xyz^" for "parent of xyz"
676 int get_sha1(const char *name, unsigned char *sha1)
678 unsigned unused;
679 return get_sha1_with_mode(name, sha1, &unused);
682 int get_sha1_with_mode(const char *name, unsigned char *sha1, unsigned *mode)
684 int ret, bracket_depth;
685 int namelen = strlen(name);
686 const char *cp;
688 *mode = S_IFINVALID;
689 ret = get_sha1_1(name, namelen, sha1);
690 if (!ret)
691 return ret;
692 /* sha1:path --> object name of path in ent sha1
693 * :path -> object name of path in index
694 * :[0-3]:path -> object name of path in index at stage
696 if (name[0] == ':') {
697 int stage = 0;
698 struct cache_entry *ce;
699 int pos;
700 if (namelen > 2 && name[1] == '/')
701 return get_sha1_oneline(name + 2, sha1);
702 if (namelen < 3 ||
703 name[2] != ':' ||
704 name[1] < '0' || '3' < name[1])
705 cp = name + 1;
706 else {
707 stage = name[1] - '0';
708 cp = name + 3;
710 namelen = namelen - (cp - name);
711 if (!active_cache)
712 read_cache();
713 pos = cache_name_pos(cp, namelen);
714 if (pos < 0)
715 pos = -pos - 1;
716 while (pos < active_nr) {
717 ce = active_cache[pos];
718 if (ce_namelen(ce) != namelen ||
719 memcmp(ce->name, cp, namelen))
720 break;
721 if (ce_stage(ce) == stage) {
722 hashcpy(sha1, ce->sha1);
723 *mode = ce->ce_mode;
724 return 0;
726 pos++;
728 return -1;
730 for (cp = name, bracket_depth = 0; *cp; cp++) {
731 if (*cp == '{')
732 bracket_depth++;
733 else if (bracket_depth && *cp == '}')
734 bracket_depth--;
735 else if (!bracket_depth && *cp == ':')
736 break;
738 if (*cp == ':') {
739 unsigned char tree_sha1[20];
740 if (!get_sha1_1(name, cp-name, tree_sha1))
741 return get_tree_entry(tree_sha1, cp+1, sha1,
742 mode);
744 return ret;