9 static const char pack_usage[] = "git-pack-objects [-q] [--no-reuse-delta] [--non-empty] [--local] [--incremental] [--window=N] [--depth=N] {--stdout | base-name} < object-list";
12 unsigned char sha1[20];
13 unsigned long size; /* uncompressed size */
14 unsigned long offset; /* offset into the final pack file;
15 * nonzero if already written.
17 unsigned int depth; /* delta depth */
18 unsigned int delta_limit; /* base adjustment for in-pack delta */
19 unsigned int hash; /* name hint hash */
20 enum object_type type;
21 enum object_type in_pack_type; /* could be delta */
22 unsigned long delta_size; /* delta data size (uncompressed) */
23 struct object_entry *delta; /* delta base object */
24 struct packed_git *in_pack; /* already in pack */
25 unsigned int in_pack_offset;
26 struct object_entry *delta_child; /* delitified objects who bases me */
27 struct object_entry *delta_sibling; /* other deltified objects who
28 * uses the same base as me
33 * Objects we are going to pack are colected in objects array (dynamically
34 * expanded). nr_objects & nr_alloc controls this array. They are stored
35 * in the order we see -- typically rev-list --objects order that gives us
36 * nice "minimum seek" order.
38 * sorted-by-sha ans sorted-by-type are arrays of pointers that point at
39 * elements in the objects array. The former is used to build the pack
40 * index (lists object names in the ascending order to help offset lookup),
41 * and the latter is used to group similar things together by try_delta()
45 static unsigned char object_list_sha1[20];
46 static int non_empty = 0;
47 static int no_reuse_delta = 0;
49 static int incremental = 0;
50 static struct object_entry **sorted_by_sha, **sorted_by_type;
51 static struct object_entry *objects = NULL;
52 static int nr_objects = 0, nr_alloc = 0;
53 static const char *base_name;
54 static unsigned char pack_file_sha1[20];
55 static int progress = 1;
56 static volatile int progress_update = 0;
59 * The object names in objects array are hashed with this hashtable,
60 * to help looking up the entry by object name. Binary search from
61 * sorted_by_sha is also possible but this was easier to code and faster.
62 * This hashtable is built after all the objects are seen.
64 static int *object_ix = NULL;
65 static int object_ix_hashsz = 0;
68 * Pack index for existing packs give us easy access to the offsets into
69 * corresponding pack file where each object's data starts, but the entries
70 * do not store the size of the compressed representation (uncompressed
71 * size is easily available by examining the pack entry header). We build
72 * a hashtable of existing packs (pack_revindex), and keep reverse index
73 * here -- pack index file is sorted by object name mapping to offset; this
74 * pack_revindex[].revindex array is an ordered list of offsets, so if you
75 * know the offset of an object, next offset is where its packed
76 * representation ends.
78 struct pack_revindex {
80 unsigned long *revindex;
81 } *pack_revindex = NULL;
82 static int pack_revindex_hashsz = 0;
87 static int written = 0;
88 static int written_delta = 0;
89 static int reused = 0;
90 static int reused_delta = 0;
92 static int pack_revindex_ix(struct packed_git *p)
94 unsigned int ui = (unsigned int) p;
97 ui = ui ^ (ui >> 16); /* defeat structure alignment */
98 i = (int)(ui % pack_revindex_hashsz);
99 while (pack_revindex[i].p) {
100 if (pack_revindex[i].p == p)
102 if (++i == pack_revindex_hashsz)
108 static void prepare_pack_ix(void)
111 struct packed_git *p;
112 for (num = 0, p = packed_git; p; p = p->next)
116 pack_revindex_hashsz = num * 11;
117 pack_revindex = xcalloc(sizeof(*pack_revindex), pack_revindex_hashsz);
118 for (p = packed_git; p; p = p->next) {
119 num = pack_revindex_ix(p);
121 pack_revindex[num].p = p;
123 /* revindex elements are lazily initialized */
126 static int cmp_offset(const void *a_, const void *b_)
128 unsigned long a = *(unsigned long *) a_;
129 unsigned long b = *(unsigned long *) b_;
139 * Ordered list of offsets of objects in the pack.
141 static void prepare_pack_revindex(struct pack_revindex *rix)
143 struct packed_git *p = rix->p;
144 int num_ent = num_packed_objects(p);
146 void *index = p->index_base + 256;
148 rix->revindex = xmalloc(sizeof(unsigned long) * (num_ent + 1));
149 for (i = 0; i < num_ent; i++) {
150 long hl = *((long *)(index + 24 * i));
151 rix->revindex[i] = ntohl(hl);
153 /* This knows the pack format -- the 20-byte trailer
154 * follows immediately after the last object data.
156 rix->revindex[num_ent] = p->pack_size - 20;
157 qsort(rix->revindex, num_ent, sizeof(unsigned long), cmp_offset);
160 static unsigned long find_packed_object_size(struct packed_git *p,
165 struct pack_revindex *rix;
166 unsigned long *revindex;
167 num = pack_revindex_ix(p);
169 die("internal error: pack revindex uninitialized");
170 rix = &pack_revindex[num];
172 prepare_pack_revindex(rix);
173 revindex = rix->revindex;
175 hi = num_packed_objects(p) + 1;
177 int mi = (lo + hi) / 2;
178 if (revindex[mi] == ofs) {
179 return revindex[mi+1] - ofs;
181 else if (ofs < revindex[mi])
186 die("internal error: pack revindex corrupt");
189 static void *delta_against(void *buf, unsigned long size, struct object_entry *entry)
191 unsigned long othersize, delta_size;
193 void *otherbuf = read_sha1_file(entry->delta->sha1, type, &othersize);
197 die("unable to read %s", sha1_to_hex(entry->delta->sha1));
198 delta_buf = diff_delta(otherbuf, othersize,
199 buf, size, &delta_size, 0);
200 if (!delta_buf || delta_size != entry->delta_size)
201 die("delta size changed");
208 * The per-object header is a pretty dense thing, which is
209 * - first byte: low four bits are "size", then three bits of "type",
210 * and the high bit is "size continues".
211 * - each byte afterwards: low seven bits are size continuation,
212 * with the high bit being "size continues"
214 static int encode_header(enum object_type type, unsigned long size, unsigned char *hdr)
219 if (type < OBJ_COMMIT || type > OBJ_DELTA)
220 die("bad type %d", type);
222 c = (type << 4) | (size & 15);
234 static unsigned long write_object(struct sha1file *f, struct object_entry *entry)
239 unsigned char header[10];
240 unsigned hdrlen, datalen;
241 enum object_type obj_type;
244 obj_type = entry->type;
245 if (! entry->in_pack)
246 to_reuse = 0; /* can't reuse what we don't have */
247 else if (obj_type == OBJ_DELTA)
248 to_reuse = 1; /* check_object() decided it for us */
249 else if (obj_type != entry->in_pack_type)
250 to_reuse = 0; /* pack has delta which is unusable */
251 else if (entry->delta)
252 to_reuse = 0; /* we want to pack afresh */
254 to_reuse = 1; /* we have it in-pack undeltified,
255 * and we do not need to deltify it.
259 buf = read_sha1_file(entry->sha1, type, &size);
261 die("unable to read %s", sha1_to_hex(entry->sha1));
262 if (size != entry->size)
263 die("object %s size inconsistency (%lu vs %lu)",
264 sha1_to_hex(entry->sha1), size, entry->size);
266 buf = delta_against(buf, size, entry);
267 size = entry->delta_size;
268 obj_type = OBJ_DELTA;
271 * The object header is a byte of 'type' followed by zero or
272 * more bytes of length. For deltas, the 20 bytes of delta
275 hdrlen = encode_header(obj_type, size, header);
276 sha1write(f, header, hdrlen);
279 sha1write(f, entry->delta, 20);
282 datalen = sha1write_compressed(f, buf, size);
286 struct packed_git *p = entry->in_pack;
289 datalen = find_packed_object_size(p, entry->in_pack_offset);
290 buf = p->pack_base + entry->in_pack_offset;
291 sha1write(f, buf, datalen);
293 hdrlen = 0; /* not really */
294 if (obj_type == OBJ_DELTA)
298 if (obj_type == OBJ_DELTA)
301 return hdrlen + datalen;
304 static unsigned long write_one(struct sha1file *f,
305 struct object_entry *e,
306 unsigned long offset)
309 /* offset starts from header size and cannot be zero
310 * if it is written already.
314 offset += write_object(f, e);
315 /* if we are deltified, write out its base object. */
317 offset = write_one(f, e->delta, offset);
321 static void write_pack_file(void)
325 unsigned long offset;
326 struct pack_header hdr;
329 f = sha1fd(1, "<stdout>");
331 f = sha1create("%s-%s.%s", base_name, sha1_to_hex(object_list_sha1), "pack");
332 hdr.hdr_signature = htonl(PACK_SIGNATURE);
333 hdr.hdr_version = htonl(PACK_VERSION);
334 hdr.hdr_entries = htonl(nr_objects);
335 sha1write(f, &hdr, sizeof(hdr));
336 offset = sizeof(hdr);
337 for (i = 0; i < nr_objects; i++) {
338 offset = write_one(f, objects + i, offset);
339 if (progress_update) {
340 fprintf(stderr, "Writing (%d %d%%)\r",
341 i+1, (i+1) * 100/nr_objects);
346 sha1close(f, pack_file_sha1, 1);
349 static void write_index_file(void)
352 struct sha1file *f = sha1create("%s-%s.%s", base_name, sha1_to_hex(object_list_sha1), "idx");
353 struct object_entry **list = sorted_by_sha;
354 struct object_entry **last = list + nr_objects;
355 unsigned int array[256];
358 * Write the first-level table (the list is sorted,
359 * but we use a 256-entry lookup to be able to avoid
360 * having to do eight extra binary search iterations).
362 for (i = 0; i < 256; i++) {
363 struct object_entry **next = list;
364 while (next < last) {
365 struct object_entry *entry = *next;
366 if (entry->sha1[0] != i)
370 array[i] = htonl(next - sorted_by_sha);
373 sha1write(f, array, 256 * sizeof(int));
376 * Write the actual SHA1 entries..
378 list = sorted_by_sha;
379 for (i = 0; i < nr_objects; i++) {
380 struct object_entry *entry = *list++;
381 unsigned int offset = htonl(entry->offset);
382 sha1write(f, &offset, 4);
383 sha1write(f, entry->sha1, 20);
385 sha1write(f, pack_file_sha1, 20);
386 sha1close(f, NULL, 1);
389 static int add_object_entry(unsigned char *sha1, unsigned int hash)
391 unsigned int idx = nr_objects;
392 struct object_entry *entry;
393 struct packed_git *p;
394 unsigned int found_offset = 0;
395 struct packed_git *found_pack = NULL;
397 for (p = packed_git; p; p = p->next) {
399 if (find_pack_entry_one(sha1, &e, p)) {
402 if (local && !p->pack_local)
405 found_offset = e.offset;
411 if (idx >= nr_alloc) {
412 unsigned int needed = (idx + 1024) * 3 / 2;
413 objects = xrealloc(objects, needed * sizeof(*entry));
416 entry = objects + idx;
417 memset(entry, 0, sizeof(*entry));
418 memcpy(entry->sha1, sha1, 20);
421 entry->in_pack = found_pack;
422 entry->in_pack_offset = found_offset;
428 static int locate_object_entry_hash(unsigned char *sha1)
432 memcpy(&ui, sha1, sizeof(unsigned int));
433 i = ui % object_ix_hashsz;
434 while (0 < object_ix[i]) {
435 if (!memcmp(sha1, objects[object_ix[i]-1].sha1, 20))
437 if (++i == object_ix_hashsz)
443 static struct object_entry *locate_object_entry(unsigned char *sha1)
445 int i = locate_object_entry_hash(sha1);
447 return &objects[object_ix[i]-1];
451 static void check_object(struct object_entry *entry)
455 if (entry->in_pack) {
456 unsigned char base[20];
458 struct object_entry *base_entry;
460 /* We want in_pack_type even if we do not reuse delta.
461 * There is no point not reusing non-delta representations.
463 check_reuse_pack_delta(entry->in_pack,
464 entry->in_pack_offset,
466 &entry->in_pack_type);
468 /* Check if it is delta, and the base is also an object
469 * we are going to pack. If so we will reuse the existing
472 if (!no_reuse_delta &&
473 entry->in_pack_type == OBJ_DELTA &&
474 (base_entry = locate_object_entry(base))) {
476 /* Depth value does not matter - find_deltas()
477 * will never consider reused delta as the
478 * base object to deltify other objects
479 * against, in order to avoid circular deltas.
482 /* uncompressed size of the delta data */
483 entry->size = entry->delta_size = size;
484 entry->delta = base_entry;
485 entry->type = OBJ_DELTA;
487 entry->delta_sibling = base_entry->delta_child;
488 base_entry->delta_child = entry;
492 /* Otherwise we would do the usual */
495 if (sha1_object_info(entry->sha1, type, &entry->size))
496 die("unable to get type of object %s",
497 sha1_to_hex(entry->sha1));
499 if (!strcmp(type, "commit")) {
500 entry->type = OBJ_COMMIT;
501 } else if (!strcmp(type, "tree")) {
502 entry->type = OBJ_TREE;
503 } else if (!strcmp(type, "blob")) {
504 entry->type = OBJ_BLOB;
505 } else if (!strcmp(type, "tag")) {
506 entry->type = OBJ_TAG;
508 die("unable to pack object %s of type %s",
509 sha1_to_hex(entry->sha1), type);
512 static void hash_objects(void)
515 struct object_entry *oe;
517 object_ix_hashsz = nr_objects * 2;
518 object_ix = xcalloc(sizeof(int), object_ix_hashsz);
519 for (i = 0, oe = objects; i < nr_objects; i++, oe++) {
520 int ix = locate_object_entry_hash(oe->sha1);
522 error("the same object '%s' added twice",
523 sha1_to_hex(oe->sha1));
527 object_ix[ix] = i + 1;
531 static unsigned int check_delta_limit(struct object_entry *me, unsigned int n)
533 struct object_entry *child = me->delta_child;
536 unsigned int c = check_delta_limit(child, n + 1);
539 child = child->delta_sibling;
544 static void get_object_details(void)
547 struct object_entry *entry;
551 for (i = 0, entry = objects; i < nr_objects; i++, entry++)
553 for (i = 0, entry = objects; i < nr_objects; i++, entry++)
554 if (!entry->delta && entry->delta_child)
556 check_delta_limit(entry, 1);
559 typedef int (*entry_sort_t)(const struct object_entry *, const struct object_entry *);
561 static entry_sort_t current_sort;
563 static int sort_comparator(const void *_a, const void *_b)
565 struct object_entry *a = *(struct object_entry **)_a;
566 struct object_entry *b = *(struct object_entry **)_b;
567 return current_sort(a,b);
570 static struct object_entry **create_sorted_list(entry_sort_t sort)
572 struct object_entry **list = xmalloc(nr_objects * sizeof(struct object_entry *));
575 for (i = 0; i < nr_objects; i++)
576 list[i] = objects + i;
578 qsort(list, nr_objects, sizeof(struct object_entry *), sort_comparator);
582 static int sha1_sort(const struct object_entry *a, const struct object_entry *b)
584 return memcmp(a->sha1, b->sha1, 20);
587 static int type_size_sort(const struct object_entry *a, const struct object_entry *b)
589 if (a->type < b->type)
591 if (a->type > b->type)
593 if (a->hash < b->hash)
595 if (a->hash > b->hash)
597 if (a->size < b->size)
599 if (a->size > b->size)
601 return a < b ? -1 : (a > b);
605 struct object_entry *entry;
610 * We search for deltas _backwards_ in a list sorted by type and
611 * by size, so that we see progressively smaller and smaller files.
612 * That's because we prefer deltas to be from the bigger file
613 * to the smaller - deletes are potentially cheaper, but perhaps
614 * more importantly, the bigger file is likely the more recent
617 static int try_delta(struct unpacked *cur, struct unpacked *old, unsigned max_depth)
619 struct object_entry *cur_entry = cur->entry;
620 struct object_entry *old_entry = old->entry;
621 unsigned long size, oldsize, delta_size, sizediff;
625 /* Don't bother doing diffs between different types */
626 if (cur_entry->type != old_entry->type)
629 /* If the current object is at edge, take the depth the objects
630 * that depend on the current object into account -- otherwise
631 * they would become too deep.
633 if (cur_entry->delta_child) {
634 if (max_depth <= cur_entry->delta_limit)
636 max_depth -= cur_entry->delta_limit;
639 size = cur_entry->size;
642 oldsize = old_entry->size;
643 sizediff = oldsize > size ? oldsize - size : size - oldsize;
644 if (sizediff > size / 8)
646 if (old_entry->depth >= max_depth)
652 * We always delta from the bigger to the smaller, since that's
653 * more space-efficient (deletes don't have to say _what_ they
656 max_size = size / 2 - 20;
657 if (cur_entry->delta)
658 max_size = cur_entry->delta_size-1;
659 if (sizediff >= max_size)
661 delta_buf = diff_delta(old->data, oldsize,
662 cur->data, size, &delta_size, max_size);
665 cur_entry->delta = old_entry;
666 cur_entry->delta_size = delta_size;
667 cur_entry->depth = old_entry->depth + 1;
672 static void progress_interval(int signum)
674 signal(SIGALRM, progress_interval);
678 static void find_deltas(struct object_entry **list, int window, int depth)
681 unsigned int array_size = window * sizeof(struct unpacked);
682 struct unpacked *array = xmalloc(array_size);
684 memset(array, 0, array_size);
689 struct object_entry *entry = list[i];
690 struct unpacked *n = array + idx;
695 if (progress_update || i == 0) {
696 fprintf(stderr, "Deltifying (%d %d%%)\r",
697 nr_objects-i, (nr_objects-i) * 100/nr_objects);
702 /* This happens if we decided to reuse existing
703 * delta from a pack. "!no_reuse_delta &&" is implied.
709 n->data = read_sha1_file(entry->sha1, type, &size);
710 if (size != entry->size)
711 die("object %s inconsistent object length (%lu vs %lu)", sha1_to_hex(entry->sha1), size, entry->size);
715 unsigned int other_idx = idx + j;
717 if (other_idx >= window)
719 m = array + other_idx;
722 if (try_delta(n, m, depth) < 0)
733 for (i = 0; i < window; ++i)
738 static void prepare_pack(int window, int depth)
740 get_object_details();
741 sorted_by_type = create_sorted_list(type_size_sort);
743 find_deltas(sorted_by_type, window+1, depth);
746 static int reuse_cached_pack(unsigned char *sha1, int pack_to_stdout)
748 static const char cache[] = "pack-cache/pack-%s.%s";
749 char *cached_pack, *cached_idx;
750 int ifd, ofd, ifd_ix = -1;
752 cached_pack = git_path(cache, sha1_to_hex(sha1), "pack");
753 ifd = open(cached_pack, O_RDONLY);
757 if (!pack_to_stdout) {
758 cached_idx = git_path(cache, sha1_to_hex(sha1), "idx");
759 ifd_ix = open(cached_idx, O_RDONLY);
767 fprintf(stderr, "Reusing %d objects pack %s\n", nr_objects,
770 if (pack_to_stdout) {
777 snprintf(name, sizeof(name),
778 "%s-%s.%s", base_name, sha1_to_hex(sha1), "pack");
779 ofd = open(name, O_CREAT | O_EXCL | O_WRONLY, 0666);
781 die("unable to open %s (%s)", name, strerror(errno));
782 if (copy_fd(ifd, ofd))
786 snprintf(name, sizeof(name),
787 "%s-%s.%s", base_name, sha1_to_hex(sha1), "idx");
788 ofd = open(name, O_CREAT | O_EXCL | O_WRONLY, 0666);
790 die("unable to open %s (%s)", name, strerror(errno));
791 if (copy_fd(ifd_ix, ofd))
794 puts(sha1_to_hex(sha1));
800 int main(int argc, char **argv)
803 char line[PATH_MAX + 20];
804 int window = 10, depth = 10, pack_to_stdout = 0;
805 struct object_entry **list;
808 setup_git_directory();
810 for (i = 1; i < argc; i++) {
811 const char *arg = argv[i];
814 if (!strcmp("--non-empty", arg)) {
818 if (!strcmp("--local", arg)) {
822 if (!strcmp("--incremental", arg)) {
826 if (!strncmp("--window=", arg, 9)) {
828 window = strtoul(arg+9, &end, 0);
833 if (!strncmp("--depth=", arg, 8)) {
835 depth = strtoul(arg+8, &end, 0);
840 if (!strcmp("-q", arg)) {
844 if (!strcmp("--no-reuse-delta", arg)) {
848 if (!strcmp("--stdout", arg)) {
859 if (pack_to_stdout != !base_name)
862 prepare_packed_git();
866 v.it_interval.tv_sec = 1;
867 v.it_interval.tv_usec = 0;
868 v.it_value = v.it_interval;
869 signal(SIGALRM, progress_interval);
870 setitimer(ITIMER_REAL, &v, NULL);
871 fprintf(stderr, "Generating pack...\n");
874 while (fgets(line, sizeof(line), stdin) != NULL) {
877 unsigned char sha1[20];
879 if (progress_update) {
880 fprintf(stderr, "Counting objects...%d\r", nr_objects);
883 if (get_sha1_hex(line, sha1))
884 die("expected sha1, got garbage:\n %s", line);
888 unsigned char c = *p++;
891 hash = hash * 11 + c;
893 add_object_entry(sha1, hash);
896 fprintf(stderr, "Done counting %d objects.\n", nr_objects);
897 if (non_empty && !nr_objects)
900 sorted_by_sha = create_sorted_list(sha1_sort);
902 list = sorted_by_sha;
903 for (i = 0; i < nr_objects; i++) {
904 struct object_entry *entry = *list++;
905 SHA1_Update(&ctx, entry->sha1, 20);
907 SHA1_Final(object_list_sha1, &ctx);
909 if (reuse_cached_pack(object_list_sha1, pack_to_stdout))
912 prepare_pack(window, depth);
913 if (progress && pack_to_stdout) {
914 /* the other end usually displays progress itself */
915 struct itimerval v = {{0,},};
916 setitimer(ITIMER_REAL, &v, NULL);
917 signal(SIGALRM, SIG_IGN );
921 if (!pack_to_stdout) {
923 puts(sha1_to_hex(object_list_sha1));
927 fprintf(stderr, "Total %d, written %d (delta %d), reused %d (delta %d)\n",
928 nr_objects, written, written_delta, reused, reused_delta);