Allow '(no author)' in git-svn's authors file.
[git] / sha1_name.c
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
9 static int find_short_object_filename(int len, const char *name, unsigned char *sha1)
10 {
11         struct alternate_object_database *alt;
12         char hex[40];
13         int found = 0;
14         static struct alternate_object_database *fakeent;
15
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] = '/';
24         }
25         fakeent->next = alt_odb_list;
26
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++;
43                         }
44                         else if (memcmp(hex + 2, de->d_name, 38)) {
45                                 found = 2;
46                                 break;
47                         }
48                 }
49                 closedir(dir);
50         }
51         if (found == 1)
52                 return get_sha1_hex(hex, sha1) == 0;
53         return found;
54 }
55
56 static int match_sha(unsigned len, const unsigned char *a, const unsigned char *b)
57 {
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;
69 }
70
71 static int find_short_packed_object(int len, const unsigned char *match, unsigned char *sha1)
72 {
73         struct packed_git *p;
74         unsigned char found_sha1[20];
75         int found = 0;
76
77         prepare_packed_git();
78         for (p = packed_git; p && found < 2; p = p->next) {
79                 unsigned num = num_packed_objects(p);
80                 unsigned first = 0, last = num;
81                 while (first < last) {
82                         unsigned mid = (first + last) / 2;
83                         unsigned char now[20];
84                         int cmp;
85
86                         nth_packed_object_sha1(p, mid, now);
87                         cmp = hashcmp(match, now);
88                         if (!cmp) {
89                                 first = mid;
90                                 break;
91                         }
92                         if (cmp > 0) {
93                                 first = mid+1;
94                                 continue;
95                         }
96                         last = mid;
97                 }
98                 if (first < num) {
99                         unsigned char now[20], next[20];
100                         nth_packed_object_sha1(p, first, now);
101                         if (match_sha(len, match, now)) {
102                                 if (nth_packed_object_sha1(p, first+1, next) ||
103                                     !match_sha(len, match, next)) {
104                                         /* unique within this pack */
105                                         if (!found) {
106                                                 hashcpy(found_sha1, now);
107                                                 found++;
108                                         }
109                                         else if (hashcmp(found_sha1, now)) {
110                                                 found = 2;
111                                                 break;
112                                         }
113                                 }
114                                 else {
115                                         /* not even unique within this pack */
116                                         found = 2;
117                                         break;
118                                 }
119                         }
120                 }
121         }
122         if (found == 1)
123                 hashcpy(sha1, found_sha1);
124         return found;
125 }
126
127 #define SHORT_NAME_NOT_FOUND (-1)
128 #define SHORT_NAME_AMBIGUOUS (-2)
129
130 static int find_unique_short_object(int len, char *canonical,
131                                     unsigned char *res, unsigned char *sha1)
132 {
133         int has_unpacked, has_packed;
134         unsigned char unpacked_sha1[20], packed_sha1[20];
135
136         has_unpacked = find_short_object_filename(len, canonical, unpacked_sha1);
137         has_packed = find_short_packed_object(len, res, packed_sha1);
138         if (!has_unpacked && !has_packed)
139                 return SHORT_NAME_NOT_FOUND;
140         if (1 < has_unpacked || 1 < has_packed)
141                 return SHORT_NAME_AMBIGUOUS;
142         if (has_unpacked != has_packed) {
143                 hashcpy(sha1, (has_packed ? packed_sha1 : unpacked_sha1));
144                 return 0;
145         }
146         /* Both have unique ones -- do they match? */
147         if (hashcmp(packed_sha1, unpacked_sha1))
148                 return SHORT_NAME_AMBIGUOUS;
149         hashcpy(sha1, packed_sha1);
150         return 0;
151 }
152
153 static int get_short_sha1(const char *name, int len, unsigned char *sha1,
154                           int quietly)
155 {
156         int i, status;
157         char canonical[40];
158         unsigned char res[20];
159
160         if (len < MINIMUM_ABBREV)
161                 return -1;
162         hashclr(res);
163         memset(canonical, 'x', 40);
164         for (i = 0; i < len ;i++) {
165                 unsigned char c = name[i];
166                 unsigned char val;
167                 if (c >= '0' && c <= '9')
168                         val = c - '0';
169                 else if (c >= 'a' && c <= 'f')
170                         val = c - 'a' + 10;
171                 else if (c >= 'A' && c <='F') {
172                         val = c - 'A' + 10;
173                         c -= 'A' - 'a';
174                 }
175                 else
176                         return -1;
177                 canonical[i] = c;
178                 if (!(i & 1))
179                         val <<= 4;
180                 res[i >> 1] |= val;
181         }
182
183         status = find_unique_short_object(i, canonical, res, sha1);
184         if (!quietly && (status == SHORT_NAME_AMBIGUOUS))
185                 return error("short SHA1 %.*s is ambiguous.", len, canonical);
186         return status;
187 }
188
189 const char *find_unique_abbrev(const unsigned char *sha1, int len)
190 {
191         int status, is_null;
192         static char hex[41];
193
194         is_null = is_null_sha1(sha1);
195         memcpy(hex, sha1_to_hex(sha1), 40);
196         if (len == 40 || !len)
197                 return hex;
198         while (len < 40) {
199                 unsigned char sha1_ret[20];
200                 status = get_short_sha1(hex, len, sha1_ret, 1);
201                 if (!status ||
202                     (is_null && status != SHORT_NAME_AMBIGUOUS)) {
203                         hex[len] = 0;
204                         return hex;
205                 }
206                 if (status != SHORT_NAME_AMBIGUOUS)
207                         return NULL;
208                 len++;
209         }
210         return NULL;
211 }
212
213 static int ambiguous_path(const char *path, int len)
214 {
215         int slash = 1;
216         int cnt;
217
218         for (cnt = 0; cnt < len; cnt++) {
219                 switch (*path++) {
220                 case '\0':
221                         break;
222                 case '/':
223                         if (slash)
224                                 break;
225                         slash = 1;
226                         continue;
227                 case '.':
228                         continue;
229                 default:
230                         slash = 0;
231                         continue;
232                 }
233                 break;
234         }
235         return slash;
236 }
237
238 static int get_sha1_basic(const char *str, int len, unsigned char *sha1)
239 {
240         static const char *fmt[] = {
241                 "%.*s",
242                 "refs/%.*s",
243                 "refs/tags/%.*s",
244                 "refs/heads/%.*s",
245                 "refs/remotes/%.*s",
246                 "refs/remotes/%.*s/HEAD",
247                 NULL
248         };
249         static const char *warning = "warning: refname '%.*s' is ambiguous.\n";
250         const char **p, *pathname;
251         char *real_path = NULL;
252         int refs_found = 0, am;
253         unsigned long at_time = (unsigned long)-1;
254         unsigned char *this_result;
255         unsigned char sha1_from_ref[20];
256
257         if (len == 40 && !get_sha1_hex(str, sha1))
258                 return 0;
259
260         /* At a given period of time? "@{2 hours ago}" */
261         for (am = 1; am < len - 1; am++) {
262                 if (str[am] == '@' && str[am+1] == '{' && str[len-1] == '}') {
263                         int date_len = len - am - 3;
264                         char *date_spec = xmalloc(date_len + 1);
265                         strlcpy(date_spec, str + am + 2, date_len + 1);
266                         at_time = approxidate(date_spec);
267                         free(date_spec);
268                         len = am;
269                         break;
270                 }
271         }
272
273         /* Accept only unambiguous ref paths. */
274         if (ambiguous_path(str, len))
275                 return -1;
276
277         for (p = fmt; *p; p++) {
278                 this_result = refs_found ? sha1_from_ref : sha1;
279                 pathname = resolve_ref(git_path(*p, len, str), this_result, 1);
280                 if (pathname) {
281                         if (!refs_found++)
282                                 real_path = xstrdup(pathname);
283                         if (!warn_ambiguous_refs)
284                                 break;
285                 }
286         }
287
288         if (!refs_found)
289                 return -1;
290
291         if (warn_ambiguous_refs && refs_found > 1)
292                 fprintf(stderr, warning, len, str);
293
294         if (at_time != (unsigned long)-1) {
295                 read_ref_at(
296                         real_path + strlen(git_path(".")) - 1,
297                         at_time,
298                         sha1);
299         }
300
301         free(real_path);
302         return 0;
303 }
304
305 static int get_sha1_1(const char *name, int len, unsigned char *sha1);
306
307 static int get_parent(const char *name, int len,
308                       unsigned char *result, int idx)
309 {
310         unsigned char sha1[20];
311         int ret = get_sha1_1(name, len, sha1);
312         struct commit *commit;
313         struct commit_list *p;
314
315         if (ret)
316                 return ret;
317         commit = lookup_commit_reference(sha1);
318         if (!commit)
319                 return -1;
320         if (parse_commit(commit))
321                 return -1;
322         if (!idx) {
323                 hashcpy(result, commit->object.sha1);
324                 return 0;
325         }
326         p = commit->parents;
327         while (p) {
328                 if (!--idx) {
329                         hashcpy(result, p->item->object.sha1);
330                         return 0;
331                 }
332                 p = p->next;
333         }
334         return -1;
335 }
336
337 static int get_nth_ancestor(const char *name, int len,
338                             unsigned char *result, int generation)
339 {
340         unsigned char sha1[20];
341         int ret = get_sha1_1(name, len, sha1);
342         if (ret)
343                 return ret;
344
345         while (generation--) {
346                 struct commit *commit = lookup_commit_reference(sha1);
347
348                 if (!commit || parse_commit(commit) || !commit->parents)
349                         return -1;
350                 hashcpy(sha1, commit->parents->item->object.sha1);
351         }
352         hashcpy(result, sha1);
353         return 0;
354 }
355
356 static int peel_onion(const char *name, int len, unsigned char *sha1)
357 {
358         unsigned char outer[20];
359         const char *sp;
360         unsigned int expected_type = 0;
361         struct object *o;
362
363         /*
364          * "ref^{type}" dereferences ref repeatedly until you cannot
365          * dereference anymore, or you get an object of given type,
366          * whichever comes first.  "ref^{}" means just dereference
367          * tags until you get a non-tag.  "ref^0" is a shorthand for
368          * "ref^{commit}".  "commit^{tree}" could be used to find the
369          * top-level tree of the given commit.
370          */
371         if (len < 4 || name[len-1] != '}')
372                 return -1;
373
374         for (sp = name + len - 1; name <= sp; sp--) {
375                 int ch = *sp;
376                 if (ch == '{' && name < sp && sp[-1] == '^')
377                         break;
378         }
379         if (sp <= name)
380                 return -1;
381
382         sp++; /* beginning of type name, or closing brace for empty */
383         if (!strncmp(commit_type, sp, 6) && sp[6] == '}')
384                 expected_type = OBJ_COMMIT;
385         else if (!strncmp(tree_type, sp, 4) && sp[4] == '}')
386                 expected_type = OBJ_TREE;
387         else if (!strncmp(blob_type, sp, 4) && sp[4] == '}')
388                 expected_type = OBJ_BLOB;
389         else if (sp[0] == '}')
390                 expected_type = OBJ_NONE;
391         else
392                 return -1;
393
394         if (get_sha1_1(name, sp - name - 2, outer))
395                 return -1;
396
397         o = parse_object(outer);
398         if (!o)
399                 return -1;
400         if (!expected_type) {
401                 o = deref_tag(o, name, sp - name - 2);
402                 if (!o || (!o->parsed && !parse_object(o->sha1)))
403                         return -1;
404                 hashcpy(sha1, o->sha1);
405         }
406         else {
407                 /* At this point, the syntax look correct, so
408                  * if we do not get the needed object, we should
409                  * barf.
410                  */
411
412                 while (1) {
413                         if (!o || (!o->parsed && !parse_object(o->sha1)))
414                                 return -1;
415                         if (o->type == expected_type) {
416                                 hashcpy(sha1, o->sha1);
417                                 return 0;
418                         }
419                         if (o->type == OBJ_TAG)
420                                 o = ((struct tag*) o)->tagged;
421                         else if (o->type == OBJ_COMMIT)
422                                 o = &(((struct commit *) o)->tree->object);
423                         else
424                                 return error("%.*s: expected %s type, but the object dereferences to %s type",
425                                              len, name, typename(expected_type),
426                                              typename(o->type));
427                         if (!o->parsed)
428                                 parse_object(o->sha1);
429                 }
430         }
431         return 0;
432 }
433
434 static int get_describe_name(const char *name, int len, unsigned char *sha1)
435 {
436         const char *cp;
437
438         for (cp = name + len - 1; name + 2 <= cp; cp--) {
439                 char ch = *cp;
440                 if (hexval(ch) & ~0377) {
441                         /* We must be looking at g in "SOMETHING-g"
442                          * for it to be describe output.
443                          */
444                         if (ch == 'g' && cp[-1] == '-') {
445                                 cp++;
446                                 len -= cp - name;
447                                 return get_short_sha1(cp, len, sha1, 1);
448                         }
449                 }
450         }
451         return -1;
452 }
453
454 static int get_sha1_1(const char *name, int len, unsigned char *sha1)
455 {
456         int ret, has_suffix;
457         const char *cp;
458
459         /* "name~3" is "name^^^",
460          * "name~" and "name~0" are name -- not "name^0"!
461          * "name^" is not "name^0"; it is "name^1".
462          */
463         has_suffix = 0;
464         for (cp = name + len - 1; name <= cp; cp--) {
465                 int ch = *cp;
466                 if ('0' <= ch && ch <= '9')
467                         continue;
468                 if (ch == '~' || ch == '^')
469                         has_suffix = ch;
470                 break;
471         }
472
473         if (has_suffix) {
474                 int num = 0;
475                 int len1 = cp - name;
476                 cp++;
477                 while (cp < name + len)
478                         num = num * 10 + *cp++ - '0';
479                 if (has_suffix == '^') {
480                         if (!num && len1 == len - 1)
481                                 num = 1;
482                         return get_parent(name, len1, sha1, num);
483                 }
484                 /* else if (has_suffix == '~') -- goes without saying */
485                 return get_nth_ancestor(name, len1, sha1, num);
486         }
487
488         ret = peel_onion(name, len, sha1);
489         if (!ret)
490                 return 0;
491
492         ret = get_sha1_basic(name, len, sha1);
493         if (!ret)
494                 return 0;
495
496         /* It could be describe output that is "SOMETHING-gXXXX" */
497         ret = get_describe_name(name, len, sha1);
498         if (!ret)
499                 return 0;
500
501         return get_short_sha1(name, len, sha1, 0);
502 }
503
504 /*
505  * This is like "get_sha1_basic()", except it allows "sha1 expressions",
506  * notably "xyz^" for "parent of xyz"
507  */
508 int get_sha1(const char *name, unsigned char *sha1)
509 {
510         int ret, bracket_depth;
511         unsigned unused;
512         int namelen = strlen(name);
513         const char *cp;
514
515         prepare_alt_odb();
516         ret = get_sha1_1(name, namelen, sha1);
517         if (!ret)
518                 return ret;
519         /* sha1:path --> object name of path in ent sha1
520          * :path -> object name of path in index
521          * :[0-3]:path -> object name of path in index at stage
522          */
523         if (name[0] == ':') {
524                 int stage = 0;
525                 struct cache_entry *ce;
526                 int pos;
527                 if (namelen < 3 ||
528                     name[2] != ':' ||
529                     name[1] < '0' || '3' < name[1])
530                         cp = name + 1;
531                 else {
532                         stage = name[1] - '0';
533                         cp = name + 3;
534                 }
535                 namelen = namelen - (cp - name);
536                 if (!active_cache)
537                         read_cache();
538                 if (active_nr < 0)
539                         return -1;
540                 pos = cache_name_pos(cp, namelen);
541                 if (pos < 0)
542                         pos = -pos - 1;
543                 while (pos < active_nr) {
544                         ce = active_cache[pos];
545                         if (ce_namelen(ce) != namelen ||
546                             memcmp(ce->name, cp, namelen))
547                                 break;
548                         if (ce_stage(ce) == stage) {
549                                 hashcpy(sha1, ce->sha1);
550                                 return 0;
551                         }
552                         pos++;
553                 }
554                 return -1;
555         }
556         for (cp = name, bracket_depth = 0; *cp; cp++) {
557                 if (*cp == '{')
558                         bracket_depth++;
559                 else if (bracket_depth && *cp == '}')
560                         bracket_depth--;
561                 else if (!bracket_depth && *cp == ':')
562                         break;
563         }
564         if (*cp == ':') {
565                 unsigned char tree_sha1[20];
566                 if (!get_sha1_1(name, cp-name, tree_sha1))
567                         return get_tree_entry(tree_sha1, cp+1, sha1,
568                                               &unused);
569         }
570         return ret;
571 }