3 * Copyright 2005, Lukas Sandstrom <lukass@etek.chalmers.se>
5 * This file is licensed under the GPL v2.
11 static const char pack_redundant_usage[] =
12 "git-pack-redundant [ --verbose ] [ --alt-odb ] < --all | <.pack filename> ...>";
14 static int load_all_packs = 0, verbose = 0, alt_odb = 0;
17 struct llist_item *next;
21 struct llist_item *front;
22 struct llist_item *back;
24 } *all_objects; /* all objects which must be present in local packfiles */
26 static struct pack_list {
27 struct pack_list *next;
28 struct packed_git *pack;
29 struct llist *unique_objects;
30 struct llist *all_objects;
31 } *local_packs = NULL, *altodb_packs = NULL;
39 static struct llist_item *free_nodes = NULL;
41 static inline struct llist_item *llist_item_get()
43 struct llist_item *new;
46 free_nodes = free_nodes->next;
48 new = xmalloc(sizeof(struct llist_item));
53 static inline void llist_item_put(struct llist_item *item)
55 item->next = free_nodes;
59 static void llist_free(struct llist *list)
61 while((list->back = list->front)) {
62 list->front = list->front->next;
63 llist_item_put(list->back);
68 static inline void llist_init(struct llist **list)
70 *list = xmalloc(sizeof(struct llist));
71 (*list)->front = (*list)->back = NULL;
75 static struct llist * llist_copy(struct llist *list)
78 struct llist_item *new, *old, *prev;
82 if ((ret->size = list->size) == 0)
85 new = ret->front = llist_item_get();
86 new->sha1 = list->front->sha1;
88 old = list->front->next;
91 new = llist_item_get();
93 new->sha1 = old->sha1;
102 static inline struct llist_item * llist_insert(struct llist *list,
103 struct llist_item *after, char *sha1)
105 struct llist_item *new = llist_item_get();
110 new->next = after->next;
112 if (after == list->back)
114 } else {/* insert in front */
118 new->next = list->front;
125 static inline struct llist_item * llist_insert_back(struct llist *list, char *sha1)
127 return llist_insert(list, list->back, sha1);
130 static inline struct llist_item * llist_insert_sorted_unique(struct llist *list,
131 char *sha1, struct llist_item *hint)
133 struct llist_item *prev = NULL, *l;
135 l = (hint == NULL) ? list->front : hint;
137 int cmp = memcmp(l->sha1, sha1, 20);
138 if (cmp > 0) { /* we insert before this entry */
139 return llist_insert(list, prev, sha1);
141 if(!cmp) { /* already exists */
147 /* insert at the end */
148 return llist_insert_back(list, sha1);
151 /* returns a pointer to an item in front of sha1 */
152 static inline struct llist_item * llist_sorted_remove(struct llist *list,
154 struct llist_item *hint)
156 struct llist_item *prev, *l;
159 l = (hint == NULL) ? list->front : hint;
162 int cmp = memcmp(l->sha1, sha1, 20);
163 if (cmp > 0) /* not in list, since sorted */
165 if(!cmp) { /* found */
167 if (hint != NULL && hint != list->front) {
168 /* we don't know the previous element */
170 goto redo_from_start;
172 list->front = l->next;
174 prev->next = l->next;
188 static void llist_sorted_difference_inplace(struct llist *A,
191 struct llist_item *hint, *b;
197 hint = llist_sorted_remove(A, b->sha1, hint);
202 static inline struct pack_list * pack_list_insert(struct pack_list **pl,
203 struct pack_list *entry)
205 struct pack_list *p = xmalloc(sizeof(struct pack_list));
206 memcpy(p, entry, sizeof(struct pack_list));
212 static inline size_t pack_list_size(struct pack_list *pl)
222 static struct pack_list * pack_list_difference(const struct pack_list *A,
223 const struct pack_list *B)
225 struct pack_list *ret;
226 const struct pack_list *pl;
233 if (A->pack == pl->pack)
234 return pack_list_difference(A->next, B);
237 ret = xmalloc(sizeof(struct pack_list));
238 memcpy(ret, A, sizeof(struct pack_list));
239 ret->next = pack_list_difference(A->next, B);
243 static void cmp_two_packs(struct pack_list *p1, struct pack_list *p2)
246 void *p1_base, *p2_base;
247 struct llist_item *p1_hint = NULL, *p2_hint = NULL;
249 p1_off = p2_off = 256 * 4 + 4;
250 p1_base = (void *)p1->pack->index_base;
251 p2_base = (void *)p2->pack->index_base;
253 while (p1_off <= p1->pack->index_size - 3 * 20 &&
254 p2_off <= p2->pack->index_size - 3 * 20)
256 int cmp = memcmp(p1_base + p1_off, p2_base + p2_off, 20);
259 p1_hint = llist_sorted_remove(p1->unique_objects,
260 p1_base + p1_off, p1_hint);
261 p2_hint = llist_sorted_remove(p2->unique_objects,
262 p1_base + p1_off, p2_hint);
267 if (cmp < 0) { /* p1 has the object, p2 doesn't */
269 } else { /* p2 has the object, p1 doesn't */
275 static void pll_insert(struct pll **pll, struct pll **hint_table)
278 int i = (*pll)->pl_size - 1;
280 if (hint_table[i] == NULL) {
281 hint_table[i--] = *pll;
282 for (; i >= 0; --i) {
283 if (hint_table[i] != NULL)
286 if (hint_table[i] == NULL) /* no elements in list */
287 die("Why did this happen?");
290 prev = hint_table[i];
291 while (prev->next && prev->next->pl_size < (*pll)->pl_size)
294 (*pll)->next = prev->next;
298 /* all the permutations have to be free()d at the same time,
299 * since they refer to each other
301 static struct pll * get_all_permutations(struct pack_list *list)
303 struct pll *subset, *pll, *new_pll = NULL; /*silence warning*/
304 static struct pll **hint = NULL;
306 hint = xcalloc(pack_list_size(list), sizeof(struct pll *));
311 if (list->next == NULL) {
312 new_pll = xmalloc(sizeof(struct pll));
314 new_pll->next = NULL;
316 new_pll->pl_size = 1;
320 pll = subset = get_all_permutations(list->next);
322 if (pll->pl->pack == list->pack) {
326 new_pll = xmalloc(sizeof(struct pll));
328 new_pll->pl = xmalloc(sizeof(struct pack_list));
329 memcpy(new_pll->pl, list, sizeof(struct pack_list));
330 new_pll->pl->next = pll->pl;
331 new_pll->pl_size = pll->pl_size + 1;
333 pll_insert(&new_pll, hint);
338 new_pll = xmalloc(sizeof(struct pll));
339 new_pll->pl = xmalloc(sizeof(struct pack_list));
340 memcpy(new_pll->pl, list, sizeof(struct pack_list));
341 new_pll->pl->next = NULL;
342 new_pll->pl_size = 1;
343 pll_insert(&new_pll, hint);
348 static int is_superset(struct pack_list *pl, struct llist *list)
352 diff = llist_copy(list);
355 llist_sorted_difference_inplace(diff, pl->all_objects);
356 if (diff->size == 0) { /* we're done */
366 static size_t sizeof_union(struct packed_git *p1, struct packed_git *p2)
370 void *p1_base, *p2_base;
372 p1_off = p2_off = 256 * 4 + 4;
373 p1_base = (void *)p1->index_base;
374 p2_base = (void *)p2->index_base;
376 while (p1_off <= p1->index_size - 3 * 20 &&
377 p2_off <= p2->index_size - 3 * 20)
379 int cmp = memcmp(p1_base + p1_off, p2_base + p2_off, 20);
387 if (cmp < 0) { /* p1 has the object, p2 doesn't */
389 } else { /* p2 has the object, p1 doesn't */
396 /* another O(n^2) function ... */
397 static size_t get_pack_redundancy(struct pack_list *pl)
399 struct pack_list *subset;
405 while ((subset = pl->next)) {
407 ret += sizeof_union(pl->pack, subset->pack);
408 subset = subset->next;
415 static inline size_t pack_set_bytecount(struct pack_list *pl)
419 ret += pl->pack->pack_size;
420 ret += pl->pack->index_size;
426 static void minimize(struct pack_list **min)
428 struct pack_list *pl, *unique = NULL,
429 *non_unique = NULL, *min_perm = NULL;
430 struct pll *perm, *perm_all, *perm_ok = NULL, *new_perm;
431 struct llist *missing;
432 size_t min_perm_size = (size_t)-1, perm_size;
436 if(pl->unique_objects->size)
437 pack_list_insert(&unique, pl);
439 pack_list_insert(&non_unique, pl);
442 /* find out which objects are missing from the set of unique packs */
443 missing = llist_copy(all_objects);
446 llist_sorted_difference_inplace(missing,
451 /* return if there are no objects missing from the unique set */
452 if (missing->size == 0) {
457 /* find the permutations which contain all missing objects */
458 perm_all = perm = get_all_permutations(non_unique);
460 if (perm_ok && perm->pl_size > perm_ok->pl_size)
461 break; /* ignore all larger permutations */
462 if (is_superset(perm->pl, missing)) {
463 new_perm = xmalloc(sizeof(struct pll));
464 memcpy(new_perm, perm, sizeof(struct pll));
465 new_perm->next = perm_ok;
472 die("Internal error: No complete sets found!\n");
474 /* find the permutation with the smallest size */
477 perm_size = pack_set_bytecount(perm->pl);
478 if (min_perm_size > perm_size) {
479 min_perm_size = perm_size;
485 /* add the unique packs to the list */
488 pack_list_insert(min, pl);
493 static void load_all_objects(void)
495 struct pack_list *pl = local_packs;
496 struct llist_item *hint, *l;
498 llist_init(&all_objects);
502 l = pl->all_objects->front;
504 hint = llist_insert_sorted_unique(all_objects,
510 /* remove objects present in remote packs */
513 llist_sorted_difference_inplace(all_objects, pl->all_objects);
518 /* this scales like O(n^2) */
519 static void cmp_local_packs(void)
521 struct pack_list *subset, *pl = local_packs;
523 while ((subset = pl)) {
524 while((subset = subset->next))
525 cmp_two_packs(pl, subset);
530 static void scan_alt_odb_packs(void)
532 struct pack_list *local, *alt;
538 llist_sorted_difference_inplace(local->unique_objects,
546 static struct pack_list * add_pack(struct packed_git *p)
552 if (!p->pack_local && !(alt_odb || verbose))
556 llist_init(&l.all_objects);
559 base = (void *)p->index_base;
560 while (off <= p->index_size - 3 * 20) {
561 llist_insert_back(l.all_objects, base + off);
564 /* this list will be pruned in cmp_two_packs later */
565 l.unique_objects = llist_copy(l.all_objects);
567 return pack_list_insert(&local_packs, &l);
569 return pack_list_insert(&altodb_packs, &l);
572 static struct pack_list * add_pack_file(char *filename)
574 struct packed_git *p = packed_git;
576 if (strlen(filename) < 40)
577 die("Bad pack filename: %s\n", filename);
580 if (strstr(p->pack_name, filename))
584 die("Filename %s not found in packed_git\n", filename);
587 static void load_all(void)
589 struct packed_git *p = packed_git;
597 int main(int argc, char **argv)
600 struct pack_list *min, *red, *pl;
601 struct llist *ignore;
602 char *sha1, buf[42]; /* 40 byte sha1 + \n + \0 */
604 for (i = 1; i < argc; i++) {
605 const char *arg = argv[i];
606 if(!strcmp(arg, "--")) {
610 if(!strcmp(arg, "--all")) {
614 if(!strcmp(arg, "--verbose")) {
618 if(!strcmp(arg, "--alt-odb")) {
623 usage(pack_redundant_usage);
628 prepare_packed_git();
633 while (*(argv + i) != NULL)
634 add_pack_file(*(argv + i++));
636 if (local_packs == NULL)
637 die("Zero packs found!\n");
643 scan_alt_odb_packs();
645 /* ignore objects given on stdin */
648 while (fgets(buf, sizeof(buf), stdin)) {
650 if (get_sha1_hex(buf, sha1))
651 die("Bad sha1 on stdin: %s", buf);
652 llist_insert_sorted_unique(ignore, sha1, NULL);
655 llist_sorted_difference_inplace(all_objects, ignore);
658 llist_sorted_difference_inplace(pl->unique_objects, ignore);
665 fprintf(stderr, "There are %lu packs available in alt-odbs.\n",
666 (unsigned long)pack_list_size(altodb_packs));
667 fprintf(stderr, "The smallest (bytewise) set of packs is:\n");
670 fprintf(stderr, "\t%s\n", pl->pack->pack_name);
673 fprintf(stderr, "containing %lu duplicate objects "
674 "with a total size of %lukb.\n",
675 (unsigned long)get_pack_redundancy(min),
676 (unsigned long)pack_set_bytecount(min)/1024);
677 fprintf(stderr, "A total of %lu unique objects were considered.\n",
678 (unsigned long)all_objects->size);
679 fprintf(stderr, "Redundant packs (with indexes):\n");
681 pl = red = pack_list_difference(local_packs, min);
684 sha1_pack_index_name(pl->pack->sha1),
685 pl->pack->pack_name);
689 fprintf(stderr, "%luMB of redundant packs in total.\n", pack_set_bytecount(red)/(1024*1024));