diffcore-rename: avoid processing duplicate destinations
[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(const char *str)
37 {
38         int i;
39
40         for (i = 1; i < ARRAY_SIZE(object_type_strings); i++)
41                 if (!strcmp(str, object_type_strings[i]))
42                         return i;
43         die("invalid object type \"%s\"", str);
44 }
45
46 /*
47  * Return a numerical hash value between 0 and n-1 for the object with
48  * the specified sha1.  n must be a power of 2.  Please note that the
49  * return value is *not* consistent across computer architectures.
50  */
51 static unsigned int hash_obj(const unsigned char *sha1, unsigned int n)
52 {
53         unsigned int hash;
54
55         /*
56          * Since the sha1 is essentially random, we just take the
57          * required number of bits directly from the first
58          * sizeof(unsigned int) bytes of sha1.  First we have to copy
59          * the bytes into a properly aligned integer.  If we cared
60          * about getting consistent results across architectures, we
61          * would have to call ntohl() here, too.
62          */
63         memcpy(&hash, sha1, sizeof(unsigned int));
64         return hash & (n - 1);
65 }
66
67 /*
68  * Insert obj into the hash table hash, which has length size (which
69  * must be a power of 2).  On collisions, simply overflow to the next
70  * empty bucket.
71  */
72 static void insert_obj_hash(struct object *obj, struct object **hash, unsigned int size)
73 {
74         unsigned int j = hash_obj(obj->sha1, size);
75
76         while (hash[j]) {
77                 j++;
78                 if (j >= size)
79                         j = 0;
80         }
81         hash[j] = obj;
82 }
83
84 /*
85  * Look up the record for the given sha1 in the hash map stored in
86  * obj_hash.  Return NULL if it was not found.
87  */
88 struct object *lookup_object(const unsigned char *sha1)
89 {
90         unsigned int i, first;
91         struct object *obj;
92
93         if (!obj_hash)
94                 return NULL;
95
96         first = i = hash_obj(sha1, obj_hash_size);
97         while ((obj = obj_hash[i]) != NULL) {
98                 if (!hashcmp(sha1, obj->sha1))
99                         break;
100                 i++;
101                 if (i == obj_hash_size)
102                         i = 0;
103         }
104         if (obj && i != first) {
105                 /*
106                  * Move object to where we started to look for it so
107                  * that we do not need to walk the hash table the next
108                  * time we look for it.
109                  */
110                 struct object *tmp = obj_hash[i];
111                 obj_hash[i] = obj_hash[first];
112                 obj_hash[first] = tmp;
113         }
114         return obj;
115 }
116
117 /*
118  * Increase the size of the hash map stored in obj_hash to the next
119  * power of 2 (but at least 32).  Copy the existing values to the new
120  * hash map.
121  */
122 static void grow_object_hash(void)
123 {
124         int i;
125         /*
126          * Note that this size must always be power-of-2 to match hash_obj
127          * above.
128          */
129         int new_hash_size = obj_hash_size < 32 ? 32 : 2 * obj_hash_size;
130         struct object **new_hash;
131
132         new_hash = xcalloc(new_hash_size, sizeof(struct object *));
133         for (i = 0; i < obj_hash_size; i++) {
134                 struct object *obj = obj_hash[i];
135                 if (!obj)
136                         continue;
137                 insert_obj_hash(obj, new_hash, new_hash_size);
138         }
139         free(obj_hash);
140         obj_hash = new_hash;
141         obj_hash_size = new_hash_size;
142 }
143
144 void *create_object(const unsigned char *sha1, void *o)
145 {
146         struct object *obj = o;
147
148         obj->parsed = 0;
149         obj->used = 0;
150         obj->flags = 0;
151         hashcpy(obj->sha1, sha1);
152
153         if (obj_hash_size - 1 <= nr_objs * 2)
154                 grow_object_hash();
155
156         insert_obj_hash(obj, obj_hash, obj_hash_size);
157         nr_objs++;
158         return obj;
159 }
160
161 void *object_as_type(struct object *obj, enum object_type type, int quiet)
162 {
163         if (obj->type == type)
164                 return obj;
165         else if (obj->type == OBJ_NONE) {
166                 if (type == OBJ_COMMIT)
167                         ((struct commit *)obj)->index = alloc_commit_index();
168                 obj->type = type;
169                 return obj;
170         }
171         else {
172                 if (!quiet)
173                         error("object %s is a %s, not a %s",
174                               sha1_to_hex(obj->sha1),
175                               typename(obj->type), typename(type));
176                 return NULL;
177         }
178 }
179
180 struct object *lookup_unknown_object(const unsigned char *sha1)
181 {
182         struct object *obj = lookup_object(sha1);
183         if (!obj)
184                 obj = create_object(sha1, alloc_object_node());
185         return obj;
186 }
187
188 struct object *parse_object_buffer(const unsigned char *sha1, 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(sha1);
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(sha1);
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(sha1);
215                 if (commit) {
216                         if (parse_commit_buffer(commit, buffer, size))
217                                 return NULL;
218                         if (!get_cached_commit_buffer(commit, NULL)) {
219                                 set_commit_buffer(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(sha1);
226                 if (tag) {
227                         if (parse_tag_buffer(tag, buffer, size))
228                                return NULL;
229                         obj = &tag->object;
230                 }
231         } else {
232                 warning("object %s has unknown type id %d", sha1_to_hex(sha1), type);
233                 obj = NULL;
234         }
235         return obj;
236 }
237
238 struct object *parse_object_or_die(const unsigned char *sha1,
239                                    const char *name)
240 {
241         struct object *o = parse_object(sha1);
242         if (o)
243                 return o;
244
245         die(_("unable to parse object: %s"), name ? name : sha1_to_hex(sha1));
246 }
247
248 struct object *parse_object(const unsigned char *sha1)
249 {
250         unsigned long size;
251         enum object_type type;
252         int eaten;
253         const unsigned char *repl = lookup_replace_object(sha1);
254         void *buffer;
255         struct object *obj;
256
257         obj = lookup_object(sha1);
258         if (obj && obj->parsed)
259                 return obj;
260
261         if ((obj && obj->type == OBJ_BLOB) ||
262             (!obj && has_sha1_file(sha1) &&
263              sha1_object_info(sha1, NULL) == OBJ_BLOB)) {
264                 if (check_sha1_signature(repl, NULL, 0, NULL) < 0) {
265                         error("sha1 mismatch %s", sha1_to_hex(repl));
266                         return NULL;
267                 }
268                 parse_blob_buffer(lookup_blob(sha1), NULL, 0);
269                 return lookup_object(sha1);
270         }
271
272         buffer = read_sha1_file(sha1, &type, &size);
273         if (buffer) {
274                 if (check_sha1_signature(repl, buffer, size, typename(type)) < 0) {
275                         free(buffer);
276                         error("sha1 mismatch %s", sha1_to_hex(repl));
277                         return NULL;
278                 }
279
280                 obj = parse_object_buffer(sha1, type, size, buffer, &eaten);
281                 if (!eaten)
282                         free(buffer);
283                 return obj;
284         }
285         return NULL;
286 }
287
288 struct object_list *object_list_insert(struct object *item,
289                                        struct object_list **list_p)
290 {
291         struct object_list *new_list = xmalloc(sizeof(struct object_list));
292         new_list->item = item;
293         new_list->next = *list_p;
294         *list_p = new_list;
295         return new_list;
296 }
297
298 int object_list_contains(struct object_list *list, struct object *obj)
299 {
300         while (list) {
301                 if (list->item == obj)
302                         return 1;
303                 list = list->next;
304         }
305         return 0;
306 }
307
308 /*
309  * A zero-length string to which object_array_entry::name can be
310  * initialized without requiring a malloc/free.
311  */
312 static char object_array_slopbuf[1];
313
314 static void add_object_array_with_mode_context(struct object *obj, const char *name,
315                                                struct object_array *array,
316                                                unsigned mode,
317                                                struct object_context *context)
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                 objects = xrealloc(objects, alloc * sizeof(*objects));
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         entry->context = context;
341         array->nr = ++nr;
342 }
343
344 void add_object_array(struct object *obj, const char *name, struct object_array *array)
345 {
346         add_object_array_with_mode(obj, name, array, S_IFINVALID);
347 }
348
349 void add_object_array_with_mode(struct object *obj, const char *name, struct object_array *array, unsigned mode)
350 {
351         add_object_array_with_mode_context(obj, name, array, mode, NULL);
352 }
353
354 void add_object_array_with_context(struct object *obj, const char *name, struct object_array *array, struct object_context *context)
355 {
356         if (context)
357                 add_object_array_with_mode_context(obj, name, array, context->mode, context);
358         else
359                 add_object_array_with_mode_context(obj, name, array, S_IFINVALID, context);
360 }
361
362 void object_array_filter(struct object_array *array,
363                          object_array_each_func_t want, void *cb_data)
364 {
365         unsigned nr = array->nr, src, dst;
366         struct object_array_entry *objects = array->objects;
367
368         for (src = dst = 0; src < nr; src++) {
369                 if (want(&objects[src], cb_data)) {
370                         if (src != dst)
371                                 objects[dst] = objects[src];
372                         dst++;
373                 } else {
374                         if (objects[src].name != object_array_slopbuf)
375                                 free(objects[src].name);
376                 }
377         }
378         array->nr = dst;
379 }
380
381 /*
382  * Return true iff array already contains an entry with name.
383  */
384 static int contains_name(struct object_array *array, const char *name)
385 {
386         unsigned nr = array->nr, i;
387         struct object_array_entry *object = array->objects;
388
389         for (i = 0; i < nr; i++, object++)
390                 if (!strcmp(object->name, name))
391                         return 1;
392         return 0;
393 }
394
395 void object_array_remove_duplicates(struct object_array *array)
396 {
397         unsigned nr = array->nr, src;
398         struct object_array_entry *objects = array->objects;
399
400         array->nr = 0;
401         for (src = 0; src < nr; src++) {
402                 if (!contains_name(array, objects[src].name)) {
403                         if (src != array->nr)
404                                 objects[array->nr] = objects[src];
405                         array->nr++;
406                 } else {
407                         if (objects[src].name != object_array_slopbuf)
408                                 free(objects[src].name);
409                 }
410         }
411 }
412
413 void clear_object_flags(unsigned flags)
414 {
415         int i;
416
417         for (i=0; i < obj_hash_size; i++) {
418                 struct object *obj = obj_hash[i];
419                 if (obj)
420                         obj->flags &= ~flags;
421         }
422 }