Merge branch 'mk/diff-delta-uint-may-be-shorter-than-ulong'
[git] / object.c
1 #include "cache.h"
2 #include "object.h"
3 #include "blob.h"
4 #include "tree.h"
5 #include "commit.h"
6 #include "tag.h"
7
8 static struct object **obj_hash;
9 static int nr_objs, obj_hash_size;
10
11 unsigned int get_max_object_index(void)
12 {
13         return obj_hash_size;
14 }
15
16 struct object *get_indexed_object(unsigned int idx)
17 {
18         return obj_hash[idx];
19 }
20
21 static const char *object_type_strings[] = {
22         NULL,           /* OBJ_NONE = 0 */
23         "commit",       /* OBJ_COMMIT = 1 */
24         "tree",         /* OBJ_TREE = 2 */
25         "blob",         /* OBJ_BLOB = 3 */
26         "tag",          /* OBJ_TAG = 4 */
27 };
28
29 const char *typename(unsigned int type)
30 {
31         if (type >= ARRAY_SIZE(object_type_strings))
32                 return NULL;
33         return object_type_strings[type];
34 }
35
36 int type_from_string_gently(const char *str, ssize_t len, int gentle)
37 {
38         int i;
39
40         if (len < 0)
41                 len = strlen(str);
42
43         for (i = 1; i < ARRAY_SIZE(object_type_strings); i++)
44                 if (!strncmp(str, object_type_strings[i], len) &&
45                     object_type_strings[i][len] == '\0')
46                         return i;
47
48         if (gentle)
49                 return -1;
50
51         die("invalid object type \"%s\"", str);
52 }
53
54 /*
55  * Return a numerical hash value between 0 and n-1 for the object with
56  * the specified sha1.  n must be a power of 2.  Please note that the
57  * return value is *not* consistent across computer architectures.
58  */
59 static unsigned int hash_obj(const unsigned char *sha1, unsigned int n)
60 {
61         return sha1hash(sha1) & (n - 1);
62 }
63
64 /*
65  * Insert obj into the hash table hash, which has length size (which
66  * must be a power of 2).  On collisions, simply overflow to the next
67  * empty bucket.
68  */
69 static void insert_obj_hash(struct object *obj, struct object **hash, unsigned int size)
70 {
71         unsigned int j = hash_obj(obj->oid.hash, size);
72
73         while (hash[j]) {
74                 j++;
75                 if (j >= size)
76                         j = 0;
77         }
78         hash[j] = obj;
79 }
80
81 /*
82  * Look up the record for the given sha1 in the hash map stored in
83  * obj_hash.  Return NULL if it was not found.
84  */
85 struct object *lookup_object(const unsigned char *sha1)
86 {
87         unsigned int i, first;
88         struct object *obj;
89
90         if (!obj_hash)
91                 return NULL;
92
93         first = i = hash_obj(sha1, obj_hash_size);
94         while ((obj = obj_hash[i]) != NULL) {
95                 if (!hashcmp(sha1, obj->oid.hash))
96                         break;
97                 i++;
98                 if (i == obj_hash_size)
99                         i = 0;
100         }
101         if (obj && i != first) {
102                 /*
103                  * Move object to where we started to look for it so
104                  * that we do not need to walk the hash table the next
105                  * time we look for it.
106                  */
107                 SWAP(obj_hash[i], obj_hash[first]);
108         }
109         return obj;
110 }
111
112 /*
113  * Increase the size of the hash map stored in obj_hash to the next
114  * power of 2 (but at least 32).  Copy the existing values to the new
115  * hash map.
116  */
117 static void grow_object_hash(void)
118 {
119         int i;
120         /*
121          * Note that this size must always be power-of-2 to match hash_obj
122          * above.
123          */
124         int new_hash_size = obj_hash_size < 32 ? 32 : 2 * obj_hash_size;
125         struct object **new_hash;
126
127         new_hash = xcalloc(new_hash_size, sizeof(struct object *));
128         for (i = 0; i < obj_hash_size; i++) {
129                 struct object *obj = obj_hash[i];
130                 if (!obj)
131                         continue;
132                 insert_obj_hash(obj, new_hash, new_hash_size);
133         }
134         free(obj_hash);
135         obj_hash = new_hash;
136         obj_hash_size = new_hash_size;
137 }
138
139 void *create_object(const unsigned char *sha1, void *o)
140 {
141         struct object *obj = o;
142
143         obj->parsed = 0;
144         obj->flags = 0;
145         hashcpy(obj->oid.hash, sha1);
146
147         if (obj_hash_size - 1 <= nr_objs * 2)
148                 grow_object_hash();
149
150         insert_obj_hash(obj, obj_hash, obj_hash_size);
151         nr_objs++;
152         return obj;
153 }
154
155 void *object_as_type(struct object *obj, enum object_type type, int quiet)
156 {
157         if (obj->type == type)
158                 return obj;
159         else if (obj->type == OBJ_NONE) {
160                 if (type == OBJ_COMMIT)
161                         ((struct commit *)obj)->index = alloc_commit_index();
162                 obj->type = type;
163                 return obj;
164         }
165         else {
166                 if (!quiet)
167                         error("object %s is a %s, not a %s",
168                               oid_to_hex(&obj->oid),
169                               typename(obj->type), typename(type));
170                 return NULL;
171         }
172 }
173
174 struct object *lookup_unknown_object(const unsigned char *sha1)
175 {
176         struct object *obj = lookup_object(sha1);
177         if (!obj)
178                 obj = create_object(sha1, alloc_object_node());
179         return obj;
180 }
181
182 struct object *parse_object_buffer(const struct object_id *oid, enum object_type type, unsigned long size, void *buffer, int *eaten_p)
183 {
184         struct object *obj;
185         *eaten_p = 0;
186
187         obj = NULL;
188         if (type == OBJ_BLOB) {
189                 struct blob *blob = lookup_blob(oid);
190                 if (blob) {
191                         if (parse_blob_buffer(blob, buffer, size))
192                                 return NULL;
193                         obj = &blob->object;
194                 }
195         } else if (type == OBJ_TREE) {
196                 struct tree *tree = lookup_tree(oid);
197                 if (tree) {
198                         obj = &tree->object;
199                         if (!tree->buffer)
200                                 tree->object.parsed = 0;
201                         if (!tree->object.parsed) {
202                                 if (parse_tree_buffer(tree, buffer, size))
203                                         return NULL;
204                                 *eaten_p = 1;
205                         }
206                 }
207         } else if (type == OBJ_COMMIT) {
208                 struct commit *commit = lookup_commit(oid);
209                 if (commit) {
210                         if (parse_commit_buffer(commit, buffer, size))
211                                 return NULL;
212                         if (!get_cached_commit_buffer(commit, NULL)) {
213                                 set_commit_buffer(commit, buffer, size);
214                                 *eaten_p = 1;
215                         }
216                         obj = &commit->object;
217                 }
218         } else if (type == OBJ_TAG) {
219                 struct tag *tag = lookup_tag(oid);
220                 if (tag) {
221                         if (parse_tag_buffer(tag, buffer, size))
222                                return NULL;
223                         obj = &tag->object;
224                 }
225         } else {
226                 warning("object %s has unknown type id %d", oid_to_hex(oid), type);
227                 obj = NULL;
228         }
229         return obj;
230 }
231
232 struct object *parse_object_or_die(const struct object_id *oid,
233                                    const char *name)
234 {
235         struct object *o = parse_object(oid);
236         if (o)
237                 return o;
238
239         die(_("unable to parse object: %s"), name ? name : oid_to_hex(oid));
240 }
241
242 struct object *parse_object(const struct object_id *oid)
243 {
244         unsigned long size;
245         enum object_type type;
246         int eaten;
247         const unsigned char *repl = lookup_replace_object(oid->hash);
248         void *buffer;
249         struct object *obj;
250
251         obj = lookup_object(oid->hash);
252         if (obj && obj->parsed)
253                 return obj;
254
255         if ((obj && obj->type == OBJ_BLOB) ||
256             (!obj && has_object_file(oid) &&
257              sha1_object_info(oid->hash, NULL) == OBJ_BLOB)) {
258                 if (check_sha1_signature(repl, NULL, 0, NULL) < 0) {
259                         error("sha1 mismatch %s", oid_to_hex(oid));
260                         return NULL;
261                 }
262                 parse_blob_buffer(lookup_blob(oid), NULL, 0);
263                 return lookup_object(oid->hash);
264         }
265
266         buffer = read_sha1_file(oid->hash, &type, &size);
267         if (buffer) {
268                 if (check_sha1_signature(repl, buffer, size, typename(type)) < 0) {
269                         free(buffer);
270                         error("sha1 mismatch %s", sha1_to_hex(repl));
271                         return NULL;
272                 }
273
274                 obj = parse_object_buffer(oid, type, size, buffer, &eaten);
275                 if (!eaten)
276                         free(buffer);
277                 return obj;
278         }
279         return NULL;
280 }
281
282 struct object_list *object_list_insert(struct object *item,
283                                        struct object_list **list_p)
284 {
285         struct object_list *new_list = xmalloc(sizeof(struct object_list));
286         new_list->item = item;
287         new_list->next = *list_p;
288         *list_p = new_list;
289         return new_list;
290 }
291
292 int object_list_contains(struct object_list *list, struct object *obj)
293 {
294         while (list) {
295                 if (list->item == obj)
296                         return 1;
297                 list = list->next;
298         }
299         return 0;
300 }
301
302 /*
303  * A zero-length string to which object_array_entry::name can be
304  * initialized without requiring a malloc/free.
305  */
306 static char object_array_slopbuf[1];
307
308 void add_object_array_with_path(struct object *obj, const char *name,
309                                 struct object_array *array,
310                                 unsigned mode, const char *path)
311 {
312         unsigned nr = array->nr;
313         unsigned alloc = array->alloc;
314         struct object_array_entry *objects = array->objects;
315         struct object_array_entry *entry;
316
317         if (nr >= alloc) {
318                 alloc = (alloc + 32) * 2;
319                 REALLOC_ARRAY(objects, alloc);
320                 array->alloc = alloc;
321                 array->objects = objects;
322         }
323         entry = &objects[nr];
324         entry->item = obj;
325         if (!name)
326                 entry->name = NULL;
327         else if (!*name)
328                 /* Use our own empty string instead of allocating one: */
329                 entry->name = object_array_slopbuf;
330         else
331                 entry->name = xstrdup(name);
332         entry->mode = mode;
333         if (path)
334                 entry->path = xstrdup(path);
335         else
336                 entry->path = NULL;
337         array->nr = ++nr;
338 }
339
340 void add_object_array(struct object *obj, const char *name, struct object_array *array)
341 {
342         add_object_array_with_path(obj, name, array, S_IFINVALID, NULL);
343 }
344
345 /*
346  * Free all memory associated with an entry; the result is
347  * in an unspecified state and should not be examined.
348  */
349 static void object_array_release_entry(struct object_array_entry *ent)
350 {
351         if (ent->name != object_array_slopbuf)
352                 free(ent->name);
353         free(ent->path);
354 }
355
356 void object_array_filter(struct object_array *array,
357                          object_array_each_func_t want, void *cb_data)
358 {
359         unsigned nr = array->nr, src, dst;
360         struct object_array_entry *objects = array->objects;
361
362         for (src = dst = 0; src < nr; src++) {
363                 if (want(&objects[src], cb_data)) {
364                         if (src != dst)
365                                 objects[dst] = objects[src];
366                         dst++;
367                 } else {
368                         object_array_release_entry(&objects[src]);
369                 }
370         }
371         array->nr = dst;
372 }
373
374 void object_array_clear(struct object_array *array)
375 {
376         int i;
377         for (i = 0; i < array->nr; i++)
378                 object_array_release_entry(&array->objects[i]);
379         FREE_AND_NULL(array->objects);
380         array->nr = array->alloc = 0;
381 }
382
383 /*
384  * Return true iff array already contains an entry with name.
385  */
386 static int contains_name(struct object_array *array, const char *name)
387 {
388         unsigned nr = array->nr, i;
389         struct object_array_entry *object = array->objects;
390
391         for (i = 0; i < nr; i++, object++)
392                 if (!strcmp(object->name, name))
393                         return 1;
394         return 0;
395 }
396
397 void object_array_remove_duplicates(struct object_array *array)
398 {
399         unsigned nr = array->nr, src;
400         struct object_array_entry *objects = array->objects;
401
402         array->nr = 0;
403         for (src = 0; src < nr; src++) {
404                 if (!contains_name(array, objects[src].name)) {
405                         if (src != array->nr)
406                                 objects[array->nr] = objects[src];
407                         array->nr++;
408                 } else {
409                         object_array_release_entry(&objects[src]);
410                 }
411         }
412 }
413
414 void clear_object_flags(unsigned flags)
415 {
416         int i;
417
418         for (i=0; i < obj_hash_size; i++) {
419                 struct object *obj = obj_hash[i];
420                 if (obj)
421                         obj->flags &= ~flags;
422         }
423 }