object.c: allow parse_object to handle arbitrary repositories
[git] / object.c
1 #include "cache.h"
2 #include "object.h"
3 #include "replace-object.h"
4 #include "object-store.h"
5 #include "blob.h"
6 #include "tree.h"
7 #include "commit.h"
8 #include "tag.h"
9 #include "alloc.h"
10 #include "object-store.h"
11 #include "packfile.h"
12
13 unsigned int get_max_object_index(void)
14 {
15         return the_repository->parsed_objects->obj_hash_size;
16 }
17
18 struct object *get_indexed_object(unsigned int idx)
19 {
20         return the_repository->parsed_objects->obj_hash[idx];
21 }
22
23 static const char *object_type_strings[] = {
24         NULL,           /* OBJ_NONE = 0 */
25         "commit",       /* OBJ_COMMIT = 1 */
26         "tree",         /* OBJ_TREE = 2 */
27         "blob",         /* OBJ_BLOB = 3 */
28         "tag",          /* OBJ_TAG = 4 */
29 };
30
31 const char *type_name(unsigned int type)
32 {
33         if (type >= ARRAY_SIZE(object_type_strings))
34                 return NULL;
35         return object_type_strings[type];
36 }
37
38 int type_from_string_gently(const char *str, ssize_t len, int gentle)
39 {
40         int i;
41
42         if (len < 0)
43                 len = strlen(str);
44
45         for (i = 1; i < ARRAY_SIZE(object_type_strings); i++)
46                 if (!strncmp(str, object_type_strings[i], len) &&
47                     object_type_strings[i][len] == '\0')
48                         return i;
49
50         if (gentle)
51                 return -1;
52
53         die("invalid object type \"%s\"", str);
54 }
55
56 /*
57  * Return a numerical hash value between 0 and n-1 for the object with
58  * the specified sha1.  n must be a power of 2.  Please note that the
59  * return value is *not* consistent across computer architectures.
60  */
61 static unsigned int hash_obj(const unsigned char *sha1, unsigned int n)
62 {
63         return sha1hash(sha1) & (n - 1);
64 }
65
66 /*
67  * Insert obj into the hash table hash, which has length size (which
68  * must be a power of 2).  On collisions, simply overflow to the next
69  * empty bucket.
70  */
71 static void insert_obj_hash(struct object *obj, struct object **hash, unsigned int size)
72 {
73         unsigned int j = hash_obj(obj->oid.hash, size);
74
75         while (hash[j]) {
76                 j++;
77                 if (j >= size)
78                         j = 0;
79         }
80         hash[j] = obj;
81 }
82
83 /*
84  * Look up the record for the given sha1 in the hash map stored in
85  * obj_hash.  Return NULL if it was not found.
86  */
87 struct object *lookup_object(struct repository *r, const unsigned char *sha1)
88 {
89         unsigned int i, first;
90         struct object *obj;
91
92         if (!r->parsed_objects->obj_hash)
93                 return NULL;
94
95         first = i = hash_obj(sha1, r->parsed_objects->obj_hash_size);
96         while ((obj = r->parsed_objects->obj_hash[i]) != NULL) {
97                 if (!hashcmp(sha1, obj->oid.hash))
98                         break;
99                 i++;
100                 if (i == r->parsed_objects->obj_hash_size)
101                         i = 0;
102         }
103         if (obj && i != first) {
104                 /*
105                  * Move object to where we started to look for it so
106                  * that we do not need to walk the hash table the next
107                  * time we look for it.
108                  */
109                 SWAP(r->parsed_objects->obj_hash[i],
110                      r->parsed_objects->obj_hash[first]);
111         }
112         return obj;
113 }
114
115 /*
116  * Increase the size of the hash map stored in obj_hash to the next
117  * power of 2 (but at least 32).  Copy the existing values to the new
118  * hash map.
119  */
120 static void grow_object_hash(struct repository *r)
121 {
122         int i;
123         /*
124          * Note that this size must always be power-of-2 to match hash_obj
125          * above.
126          */
127         int new_hash_size = r->parsed_objects->obj_hash_size < 32 ? 32 : 2 * r->parsed_objects->obj_hash_size;
128         struct object **new_hash;
129
130         new_hash = xcalloc(new_hash_size, sizeof(struct object *));
131         for (i = 0; i < r->parsed_objects->obj_hash_size; i++) {
132                 struct object *obj = r->parsed_objects->obj_hash[i];
133
134                 if (!obj)
135                         continue;
136                 insert_obj_hash(obj, new_hash, new_hash_size);
137         }
138         free(r->parsed_objects->obj_hash);
139         r->parsed_objects->obj_hash = new_hash;
140         r->parsed_objects->obj_hash_size = new_hash_size;
141 }
142
143 void *create_object(struct repository *r, const unsigned char *sha1, void *o)
144 {
145         struct object *obj = o;
146
147         obj->parsed = 0;
148         obj->flags = 0;
149         hashcpy(obj->oid.hash, sha1);
150
151         if (r->parsed_objects->obj_hash_size - 1 <= r->parsed_objects->nr_objs * 2)
152                 grow_object_hash(r);
153
154         insert_obj_hash(obj, r->parsed_objects->obj_hash,
155                         r->parsed_objects->obj_hash_size);
156         r->parsed_objects->nr_objs++;
157         return obj;
158 }
159
160 void *object_as_type(struct repository *r, struct object *obj, enum object_type type, int quiet)
161 {
162         if (obj->type == type)
163                 return obj;
164         else if (obj->type == OBJ_NONE) {
165                 if (type == OBJ_COMMIT)
166                         ((struct commit *)obj)->index = alloc_commit_index(r);
167                 obj->type = type;
168                 return obj;
169         }
170         else {
171                 if (!quiet)
172                         error("object %s is a %s, not a %s",
173                               oid_to_hex(&obj->oid),
174                               type_name(obj->type), type_name(type));
175                 return NULL;
176         }
177 }
178
179 struct object *lookup_unknown_object(const unsigned char *sha1)
180 {
181         struct object *obj = lookup_object(the_repository, sha1);
182         if (!obj)
183                 obj = create_object(the_repository, sha1,
184                                     alloc_object_node(the_repository));
185         return obj;
186 }
187
188 struct object *parse_object_buffer(struct repository *r, const struct object_id *oid, enum object_type type, unsigned long size, void *buffer, int *eaten_p)
189 {
190         struct object *obj;
191         *eaten_p = 0;
192
193         obj = NULL;
194         if (type == OBJ_BLOB) {
195                 struct blob *blob = lookup_blob(r, oid);
196                 if (blob) {
197                         if (parse_blob_buffer(blob, buffer, size))
198                                 return NULL;
199                         obj = &blob->object;
200                 }
201         } else if (type == OBJ_TREE) {
202                 struct tree *tree = lookup_tree(r, oid);
203                 if (tree) {
204                         obj = &tree->object;
205                         if (!tree->buffer)
206                                 tree->object.parsed = 0;
207                         if (!tree->object.parsed) {
208                                 if (parse_tree_buffer(tree, buffer, size))
209                                         return NULL;
210                                 *eaten_p = 1;
211                         }
212                 }
213         } else if (type == OBJ_COMMIT) {
214                 struct commit *commit = lookup_commit(r, oid);
215                 if (commit) {
216                         if (parse_commit_buffer(r, commit, buffer, size, 1))
217                                 return NULL;
218                         if (!get_cached_commit_buffer(r, commit, NULL)) {
219                                 set_commit_buffer(r, commit, buffer, size);
220                                 *eaten_p = 1;
221                         }
222                         obj = &commit->object;
223                 }
224         } else if (type == OBJ_TAG) {
225                 struct tag *tag = lookup_tag(r, oid);
226                 if (tag) {
227                         if (parse_tag_buffer(r, tag, buffer, size))
228                                return NULL;
229                         obj = &tag->object;
230                 }
231         } else {
232                 warning("object %s has unknown type id %d", oid_to_hex(oid), type);
233                 obj = NULL;
234         }
235         return obj;
236 }
237
238 struct object *parse_object_or_die(const struct object_id *oid,
239                                    const char *name)
240 {
241         struct object *o = parse_object(the_repository, oid);
242         if (o)
243                 return o;
244
245         die(_("unable to parse object: %s"), name ? name : oid_to_hex(oid));
246 }
247
248 struct object *parse_object(struct repository *r, const struct object_id *oid)
249 {
250         unsigned long size;
251         enum object_type type;
252         int eaten;
253         const struct object_id *repl = lookup_replace_object(r, oid);
254         void *buffer;
255         struct object *obj;
256
257         obj = lookup_object(r, oid->hash);
258         if (obj && obj->parsed)
259                 return obj;
260
261         if ((obj && obj->type == OBJ_BLOB && has_object_file(oid)) ||
262             (!obj && has_object_file(oid) &&
263              oid_object_info(r, oid, NULL) == OBJ_BLOB)) {
264                 if (check_object_signature(repl, NULL, 0, NULL) < 0) {
265                         error("sha1 mismatch %s", oid_to_hex(oid));
266                         return NULL;
267                 }
268                 parse_blob_buffer(lookup_blob(r, oid), NULL, 0);
269                 return lookup_object(r, oid->hash);
270         }
271
272         buffer = read_object_file(oid, &type, &size);
273         if (buffer) {
274                 if (check_object_signature(repl, buffer, size, type_name(type)) < 0) {
275                         free(buffer);
276                         error("sha1 mismatch %s", oid_to_hex(repl));
277                         return NULL;
278                 }
279
280                 obj = parse_object_buffer(r, oid, type, size,
281                                           buffer, &eaten);
282                 if (!eaten)
283                         free(buffer);
284                 return obj;
285         }
286         return NULL;
287 }
288
289 struct object_list *object_list_insert(struct object *item,
290                                        struct object_list **list_p)
291 {
292         struct object_list *new_list = xmalloc(sizeof(struct object_list));
293         new_list->item = item;
294         new_list->next = *list_p;
295         *list_p = new_list;
296         return new_list;
297 }
298
299 int object_list_contains(struct object_list *list, struct object *obj)
300 {
301         while (list) {
302                 if (list->item == obj)
303                         return 1;
304                 list = list->next;
305         }
306         return 0;
307 }
308
309 /*
310  * A zero-length string to which object_array_entry::name can be
311  * initialized without requiring a malloc/free.
312  */
313 static char object_array_slopbuf[1];
314
315 void add_object_array_with_path(struct object *obj, const char *name,
316                                 struct object_array *array,
317                                 unsigned mode, const char *path)
318 {
319         unsigned nr = array->nr;
320         unsigned alloc = array->alloc;
321         struct object_array_entry *objects = array->objects;
322         struct object_array_entry *entry;
323
324         if (nr >= alloc) {
325                 alloc = (alloc + 32) * 2;
326                 REALLOC_ARRAY(objects, alloc);
327                 array->alloc = alloc;
328                 array->objects = objects;
329         }
330         entry = &objects[nr];
331         entry->item = obj;
332         if (!name)
333                 entry->name = NULL;
334         else if (!*name)
335                 /* Use our own empty string instead of allocating one: */
336                 entry->name = object_array_slopbuf;
337         else
338                 entry->name = xstrdup(name);
339         entry->mode = mode;
340         if (path)
341                 entry->path = xstrdup(path);
342         else
343                 entry->path = NULL;
344         array->nr = ++nr;
345 }
346
347 void add_object_array(struct object *obj, const char *name, struct object_array *array)
348 {
349         add_object_array_with_path(obj, name, array, S_IFINVALID, NULL);
350 }
351
352 /*
353  * Free all memory associated with an entry; the result is
354  * in an unspecified state and should not be examined.
355  */
356 static void object_array_release_entry(struct object_array_entry *ent)
357 {
358         if (ent->name != object_array_slopbuf)
359                 free(ent->name);
360         free(ent->path);
361 }
362
363 struct object *object_array_pop(struct object_array *array)
364 {
365         struct object *ret;
366
367         if (!array->nr)
368                 return NULL;
369
370         ret = array->objects[array->nr - 1].item;
371         object_array_release_entry(&array->objects[array->nr - 1]);
372         array->nr--;
373         return ret;
374 }
375
376 void object_array_filter(struct object_array *array,
377                          object_array_each_func_t want, void *cb_data)
378 {
379         unsigned nr = array->nr, src, dst;
380         struct object_array_entry *objects = array->objects;
381
382         for (src = dst = 0; src < nr; src++) {
383                 if (want(&objects[src], cb_data)) {
384                         if (src != dst)
385                                 objects[dst] = objects[src];
386                         dst++;
387                 } else {
388                         object_array_release_entry(&objects[src]);
389                 }
390         }
391         array->nr = dst;
392 }
393
394 void object_array_clear(struct object_array *array)
395 {
396         int i;
397         for (i = 0; i < array->nr; i++)
398                 object_array_release_entry(&array->objects[i]);
399         FREE_AND_NULL(array->objects);
400         array->nr = array->alloc = 0;
401 }
402
403 /*
404  * Return true iff array already contains an entry with name.
405  */
406 static int contains_name(struct object_array *array, const char *name)
407 {
408         unsigned nr = array->nr, i;
409         struct object_array_entry *object = array->objects;
410
411         for (i = 0; i < nr; i++, object++)
412                 if (!strcmp(object->name, name))
413                         return 1;
414         return 0;
415 }
416
417 void object_array_remove_duplicates(struct object_array *array)
418 {
419         unsigned nr = array->nr, src;
420         struct object_array_entry *objects = array->objects;
421
422         array->nr = 0;
423         for (src = 0; src < nr; src++) {
424                 if (!contains_name(array, objects[src].name)) {
425                         if (src != array->nr)
426                                 objects[array->nr] = objects[src];
427                         array->nr++;
428                 } else {
429                         object_array_release_entry(&objects[src]);
430                 }
431         }
432 }
433
434 void clear_object_flags(unsigned flags)
435 {
436         int i;
437
438         for (i=0; i < the_repository->parsed_objects->obj_hash_size; i++) {
439                 struct object *obj = the_repository->parsed_objects->obj_hash[i];
440                 if (obj)
441                         obj->flags &= ~flags;
442         }
443 }
444
445 void clear_commit_marks_all(unsigned int flags)
446 {
447         int i;
448
449         for (i = 0; i < the_repository->parsed_objects->obj_hash_size; i++) {
450                 struct object *obj = the_repository->parsed_objects->obj_hash[i];
451                 if (obj && obj->type == OBJ_COMMIT)
452                         obj->flags &= ~flags;
453         }
454 }
455
456 struct parsed_object_pool *parsed_object_pool_new(void)
457 {
458         struct parsed_object_pool *o = xmalloc(sizeof(*o));
459         memset(o, 0, sizeof(*o));
460
461         o->blob_state = allocate_alloc_state();
462         o->tree_state = allocate_alloc_state();
463         o->commit_state = allocate_alloc_state();
464         o->tag_state = allocate_alloc_state();
465         o->object_state = allocate_alloc_state();
466
467         o->is_shallow = -1;
468         o->shallow_stat = xcalloc(1, sizeof(*o->shallow_stat));
469
470         o->buffer_slab = allocate_commit_buffer_slab();
471
472         return o;
473 }
474
475 struct raw_object_store *raw_object_store_new(void)
476 {
477         struct raw_object_store *o = xmalloc(sizeof(*o));
478
479         memset(o, 0, sizeof(*o));
480         INIT_LIST_HEAD(&o->packed_git_mru);
481         return o;
482 }
483
484 static void free_alt_odb(struct alternate_object_database *alt)
485 {
486         strbuf_release(&alt->scratch);
487         oid_array_clear(&alt->loose_objects_cache);
488         free(alt);
489 }
490
491 static void free_alt_odbs(struct raw_object_store *o)
492 {
493         while (o->alt_odb_list) {
494                 struct alternate_object_database *next;
495
496                 next = o->alt_odb_list->next;
497                 free_alt_odb(o->alt_odb_list);
498                 o->alt_odb_list = next;
499         }
500 }
501
502 void raw_object_store_clear(struct raw_object_store *o)
503 {
504         FREE_AND_NULL(o->objectdir);
505         FREE_AND_NULL(o->alternate_db);
506
507         oidmap_free(o->replace_map, 1);
508         FREE_AND_NULL(o->replace_map);
509
510         free_alt_odbs(o);
511         o->alt_odb_tail = NULL;
512
513         INIT_LIST_HEAD(&o->packed_git_mru);
514         close_all_packs(o);
515         o->packed_git = NULL;
516 }
517
518 void parsed_object_pool_clear(struct parsed_object_pool *o)
519 {
520         /*
521          * As objects are allocated in slabs (see alloc.c), we do
522          * not need to free each object, but each slab instead.
523          *
524          * Before doing so, we need to free any additional memory
525          * the objects may hold.
526          */
527         unsigned i;
528
529         for (i = 0; i < o->obj_hash_size; i++) {
530                 struct object *obj = o->obj_hash[i];
531
532                 if (!obj)
533                         continue;
534
535                 if (obj->type == OBJ_TREE)
536                         free_tree_buffer((struct tree*)obj);
537                 else if (obj->type == OBJ_COMMIT)
538                         release_commit_memory((struct commit*)obj);
539                 else if (obj->type == OBJ_TAG)
540                         release_tag_memory((struct tag*)obj);
541         }
542
543         FREE_AND_NULL(o->obj_hash);
544         o->obj_hash_size = 0;
545
546         free_commit_buffer_slab(o->buffer_slab);
547         o->buffer_slab = NULL;
548
549         clear_alloc_state(o->blob_state);
550         clear_alloc_state(o->tree_state);
551         clear_alloc_state(o->commit_state);
552         clear_alloc_state(o->tag_state);
553         clear_alloc_state(o->object_state);
554         FREE_AND_NULL(o->blob_state);
555         FREE_AND_NULL(o->tree_state);
556         FREE_AND_NULL(o->commit_state);
557         FREE_AND_NULL(o->tag_state);
558         FREE_AND_NULL(o->object_state);
559 }