[PATCH 0/4] Add git-pack-intersect

[PATCH 0/4] Add git-pack-intersect

[Lukas Sandström]

This patch series adds git-pack-intersect. It finds redundant packs
by calculating the union of all objects present in .git/objects/pack
and then computing the smallest set of packs which contain all the
objects in this union.

It is quite fast, if I may say so myself. On my AMD3200+ it manages
to minimize a linux-2.6 repository with 9 pack files totaling 430MB
in ~0.7 seconds. The remaining packfiles total 102MB, which is a nice
reduction.

git-fsck-objects reports no errors after pruning, but the algorithm is 
not proven correct so backups might be in order before this gets more
testing.

/Lukas

[PATCH 1/4] Add git-pack-intersect

[Lukas Sandström]

Add git-pack-intersect

This patch adds the program git-pack-intersect. It is
used to find redundant packs in git repositories.

Signed-off-by: Lukas Sandström <lukass@etek.chalmers.se>

---

 Makefile         |    2 
 pack-intersect.c |  577 ++++++++++++++++++++++++++++++++++++++++++++++++++++++
 sha1_file.c      |    3 
 3 files changed, 581 insertions(+), 1 deletions(-)
 create mode 100644 pack-intersect.c

applies-to: bb7dd65e1d945edbe0137a761ebc388c7394067a
cb2f9b435d8101fd29454b77fb5047b7edf847dc
diff --git a/Makefile b/Makefile
index b202be1..4c646c9 100644
--- a/Makefile
+++ b/Makefile
@@ -122,7 +122,7 @@ PROGRAMS = \
 	git-unpack-objects$X git-update-index$X git-update-server-info$X \
 	git-upload-pack$X git-verify-pack$X git-write-tree$X \
 	git-update-ref$X git-symbolic-ref$X git-check-ref-format$X \
-	git-name-rev$X $(SIMPLE_PROGRAMS)
+	git-name-rev$X git-pack-intersect$X $(SIMPLE_PROGRAMS)
 
 # Backward compatibility -- to be removed after 1.0
 PROGRAMS += git-ssh-pull$X git-ssh-push$X
diff --git a/pack-intersect.c b/pack-intersect.c
new file mode 100644
index 0000000..2267478
--- /dev/null
+++ b/pack-intersect.c
@@ -0,0 +1,577 @@
+/*
+*
+* Copyright 2005, Lukas Sandstrom <lukass@etek.chalmers.se>
+*
+* This file is licensed under the GPL v2.
+*
+*/
+
+#include "cache.h"
+
+static const char pack_intersect_usage[] =
+"git-pack-intersect [ -v ]  < -a | <.pack filename> ...>";
+
+int all = 0, verbose = 0;
+
+struct llist_item {
+	struct llist_item *next;
+	char *sha1;
+};
+struct llist {
+	struct llist_item *front;
+	struct llist_item *back;
+	size_t size;
+} *all_objects;
+
+struct pack_list {
+	struct pack_list *next;
+	struct packed_git *pack;
+	struct llist *unique_objects;
+	struct llist *all_objects;
+} *pack_list;
+
+struct pll {
+	struct pll *next;
+	struct pack_list *pl;
+};
+
+inline void llist_free(struct llist *list)
+{
+	while((list->back = list->front)) {
+		list->front = list->front->next;
+		free(list->back);
+	}
+	free(list);
+}
+
+inline void llist_init(struct llist **list)
+{
+	*list = xmalloc(sizeof(struct llist));
+	(*list)->front = (*list)->back = NULL;
+	(*list)->size = 0;
+}
+
+struct llist * llist_copy(struct llist *list)
+{
+	struct llist *ret;
+	struct llist_item *new, *old, *prev;
+	
+	llist_init(&ret);
+
+	if ((ret->size = list->size) == 0)
+		return ret;
+
+	new = ret->front = xmalloc(sizeof(struct llist_item));
+	new->sha1 = list->front->sha1;
+
+	old = list->front->next;
+	while (old) {
+		prev = new;
+		new = xmalloc(sizeof(struct llist_item));
+		prev->next = new;
+		new->sha1 = old->sha1;
+		old = old->next;
+	}
+	new->next = NULL;
+	ret->back = new;
+	
+	return ret;
+}
+
+inline struct llist_item * llist_insert(struct llist *list,
+					struct llist_item *after, char *sha1)
+{
+	struct llist_item *new = xmalloc(sizeof(struct llist_item));
+	new->sha1 = sha1;
+	new->next = NULL;
+
+	if (after != NULL) {
+		new->next = after->next;
+		after->next = new;
+		if (after == list->back)
+			list->back = new;
+	} else {/* insert in front */
+		if (list->size == 0)
+			list->back = new;
+		else
+			new->next = list->front;
+		list->front = new;
+	}
+	list->size++;
+	return new;
+}
+
+inline struct llist_item * llist_insert_back(struct llist *list, char *sha1)
+{
+	return llist_insert(list, list->back, sha1);
+}
+
+inline struct llist_item * llist_insert_sorted_unique(struct llist *list,
+					char *sha1, struct llist_item *hint)
+{
+	struct llist_item *prev = NULL, *l;
+
+	l = (hint == NULL) ? list->front : hint;
+	while (l) {
+		int cmp = memcmp(l->sha1, sha1, 20);
+		if (cmp > 0) { /* we insert before this entry */
+			return llist_insert(list, prev, sha1);
+		}
+		if(!cmp) { /* already exists */
+			return l;
+		}
+		prev = l;
+		l = l->next;
+	}
+	/* insert at the end */
+	return llist_insert_back(list, sha1);
+}
+
+/* computes A\B */
+struct llist * llist_sorted_difference(struct llist_item *A,
+				       struct llist_item *B)
+{
+	struct llist *ret;
+	llist_init(&ret);
+
+	while (A != NULL && B != NULL) {
+		int cmp = memcmp(A->sha1, B->sha1, 20);
+		if (!cmp) {
+			A = A->next;
+			B = B->next;
+			continue;
+		}
+		if(cmp > 0) { /* we'll never find this B */
+			B = B->next;
+			continue;
+		}
+		/* A has the object, B doesn't */
+		llist_insert_back(ret, A->sha1);
+		A = A->next;
+	}
+	while (A != NULL) {
+		llist_insert_back(ret, A->sha1);
+		A = A->next;
+	}
+	return ret;
+}
+
+/* returns a pointer to an item in front of sha1 */
+inline struct llist_item * llist_sorted_remove(struct llist *list, char *sha1,
+					       struct llist_item *hint)
+{
+	struct llist_item *prev, *l;
+
+redo_from_start:
+	l = (hint == NULL) ? list->front : hint;
+	prev = NULL;
+	while (l) {
+		int cmp = memcmp(l->sha1, sha1, 20);
+		if (cmp > 0) /* not in list, since sorted */
+			return prev;
+		if(!cmp) { /* found */
+			if (prev == NULL) {
+				if (hint != NULL && hint != list->front) {
+					/* we don't know the previous element */
+					hint = NULL;
+					goto redo_from_start;
+				}
+				list->front = l->next;
+			} else
+				prev->next = l->next;
+			if (l == list->back)
+				list->back = prev;
+			free(l);
+			list->size--;
+			return prev;
+		}
+		prev = l;
+		l = l->next;
+	}
+	return prev;
+}
+
+inline struct pack_list * pack_list_insert(struct pack_list **pl,
+					   struct pack_list *entry)
+{
+	struct pack_list *p = xmalloc(sizeof(struct pack_list));
+	memcpy(p, entry, sizeof(struct pack_list));
+	p->next = *pl;
+	*pl = p;
+	return p;
+}
+
+struct pack_list * pack_list_difference(struct pack_list *A,
+					struct pack_list *B)
+{
+	struct pack_list *ret, *pl;
+
+	if (A == NULL)
+		return NULL;
+
+	pl = B;
+	while (pl != NULL) {
+		if (A->pack == pl->pack)
+			return pack_list_difference(A->next, B);
+		pl = pl->next;
+	}
+	ret = xmalloc(sizeof(struct pack_list));
+	memcpy(ret, A, sizeof(struct pack_list));
+	ret->next = pack_list_difference(A->next, B);
+	return ret;
+}
+
+void cmp_two_packs(struct pack_list *p1, struct pack_list *p2)
+{
+	int p1_off, p2_off;
+	void *p1_base, *p2_base;
+	struct llist_item *p1_hint = NULL, *p2_hint = NULL;
+	
+	p1_off = p2_off = 256 * 4 + 4;
+	p1_base = (void *)p1->pack->index_base;
+	p2_base = (void *)p2->pack->index_base;
+
+	while (p1_off <= p1->pack->index_size - 3 * 20 &&
+	       p2_off <= p2->pack->index_size - 3 * 20)
+	{
+		int cmp = memcmp(p1_base + p1_off, p2_base + p2_off, 20);
+		/* cmp ~ p1 - p2 */
+		if (cmp == 0) {
+			p1_hint = llist_sorted_remove(p1->unique_objects,
+					p1_base + p1_off, p1_hint);
+			p2_hint = llist_sorted_remove(p2->unique_objects,
+					p1_base + p1_off, p2_hint);
+			p1_off+=24;
+			p2_off+=24;
+			continue;
+		}
+		if (cmp < 0) { /* p1 has the object, p2 doesn't */
+			p1_off+=24;
+		} else { /* p2 has the object, p1 doesn't */
+			p2_off+=24;
+		}
+	}
+}
+
+/* all the permutations have to be free()d at the same time,
+ * since they refer to each other
+ */
+struct pll * get_all_permutations(struct pack_list *list)
+{
+	struct pll *subset, *pll, *new_pll = NULL; /*silence warning*/
+
+	if (list == NULL)
+		return NULL;
+
+	if (list->next == NULL) {
+		new_pll = xmalloc(sizeof(struct pll));
+		new_pll->next = NULL;
+		new_pll->pl = list;
+		return new_pll;
+	}
+
+	pll = subset = get_all_permutations(list->next);
+	while (pll) {
+		new_pll = xmalloc(sizeof(struct pll));
+		new_pll->next = pll->next;
+		pll->next = new_pll;
+
+		new_pll->pl = xmalloc(sizeof(struct pack_list));
+		memcpy(new_pll->pl, list, sizeof(struct pack_list));
+		new_pll->pl->next = pll->pl;
+
+		pll = new_pll->next;
+	}
+	/* add ourself to the end */
+	new_pll->next = xmalloc(sizeof(struct pll));
+	new_pll->next->pl = xmalloc(sizeof(struct pack_list));
+	new_pll->next->next = NULL;
+	memcpy(new_pll->next->pl, list, sizeof(struct pack_list));
+	new_pll->next->pl->next = NULL;
+
+	return subset;
+}
+
+int is_superset(struct pack_list *pl, struct llist *list)
+{
+	struct llist *diff, *old;
+
+	diff = llist_copy(list);
+
+	while (pl) {
+		old = diff;
+		diff = llist_sorted_difference(diff->front,
+					       pl->all_objects->front);
+		llist_free(old);
+		if (diff->size == 0) { /* we're done */
+			llist_free(diff);
+			return 1;
+		}
+		pl = pl->next;
+	}
+	llist_free(diff);
+	return 0;
+}
+
+size_t sizeof_union(struct packed_git *p1, struct packed_git *p2)
+{
+	size_t ret = 0;
+	int p1_off, p2_off;
+	void *p1_base, *p2_base;
+
+	p1_off = p2_off = 256 * 4 + 4;
+	p1_base = (void *)p1->index_base;
+	p2_base = (void *)p2->index_base;
+
+	while (p1_off <= p1->index_size - 3 * 20 &&
+	       p2_off <= p2->index_size - 3 * 20)
+	{
+		int cmp = memcmp(p1_base + p1_off, p2_base + p2_off, 20);
+		/* cmp ~ p1 - p2 */
+		if (cmp == 0) {
+			ret++;
+			p1_off+=24;
+			p2_off+=24;
+			continue;
+		}
+		if (cmp < 0) { /* p1 has the object, p2 doesn't */
+			p1_off+=24;
+		} else { /* p2 has the object, p1 doesn't */
+			p2_off+=24;
+		}
+	}
+	return ret;
+}
+
+/* another O(n^2) function ... */
+size_t get_pack_redundancy(struct pack_list *pl)
+{
+	struct pack_list *subset;
+	size_t ret = 0;
+	while ((subset = pl->next)) {
+		while(subset) {
+			ret += sizeof_union(pl->pack, subset->pack);
+			subset = subset->next;
+		}
+		pl = pl->next;
+	}
+	return ret;
+}
+
+inline size_t pack_set_bytecount(struct pack_list *pl)
+{
+	size_t ret = 0;
+	while (pl) {
+		ret += pl->pack->pack_size;
+		ret += pl->pack->index_size;
+		pl = pl->next;
+	}
+	return ret;
+}
+
+void minimize(struct pack_list **min)
+{
+	struct pack_list *pl, *unique = NULL,
+		*non_unique = NULL, *min_perm = NULL;
+	struct pll *perm, *perm_all, *perm_ok = NULL, *new_perm;
+	struct llist *missing, *old;
+	size_t min_perm_size = (size_t)-1, perm_size;
+
+	pl = pack_list;
+	while (pl) {
+		if(pl->unique_objects->size)
+			pack_list_insert(&unique, pl);
+		else
+			pack_list_insert(&non_unique, pl);
+		pl = pl->next;
+	}
+	/* find out which objects are missing from the set of unique packs */
+	missing = llist_copy(all_objects);
+	pl = unique;
+	while (pl) {
+		old = missing;
+		missing = llist_sorted_difference(missing->front,
+						  pl->all_objects->front);
+		llist_free(old);
+		pl = pl->next;
+	}
+
+	if (missing->size == 0) {
+		*min = unique;
+		return;
+	}
+
+	/* find the permutations which contain all missing objects */
+	perm_all = perm = get_all_permutations(non_unique);
+	while (perm) {
+		if (is_superset(perm->pl, missing)) {
+			new_perm = xmalloc(sizeof(struct pll));
+			new_perm->pl = perm->pl;
+			new_perm->next = perm_ok;
+			perm_ok = new_perm;
+		}
+		perm = perm->next;
+	}
+	
+	if (perm_ok == NULL)
+		die("Internal error: No complete sets found!\n");
+
+	/* find the permutation with the smallest size */
+	perm = perm_ok;
+	while (perm) {
+		perm_size = pack_set_bytecount(perm->pl);
+		if (min_perm_size > perm_size) {
+			min_perm_size = perm_size;
+			min_perm = perm->pl;
+		}
+		perm = perm->next;
+	}
+	*min = min_perm;
+	/* add the unique packs to the list */
+	pl = unique;
+	while(pl) {
+		pack_list_insert(min, pl);
+		pl = pl->next;
+	}
+}
+
+void load_all_objects()
+{
+	struct pack_list *pl = pack_list;
+	struct llist_item *hint, *l;
+	int i;
+
+	llist_init(&all_objects);
+
+	while (pl) {
+		i = 0;
+		hint = NULL;
+		l = pl->all_objects->front;
+		while (l) {
+			hint = llist_insert_sorted_unique(all_objects,
+							  l->sha1, hint);
+			l = l->next;
+		}
+		pl = pl->next;
+	}
+}
+
+/* this scales like O(n^2) */
+void cmp_packs()
+{
+	struct pack_list *subset, *curr = pack_list;
+
+	while ((subset = curr)) {
+		while((subset = subset->next))
+			cmp_two_packs(curr, subset);
+		curr = curr->next;
+	}
+}
+
+struct pack_list * add_pack(struct packed_git *p)
+{
+	struct pack_list l;
+	size_t off;
+	void *base;
+
+	l.pack = p;
+	llist_init(&l.all_objects);
+
+	off = 256 * 4 + 4;
+	base = (void *)p->index_base;
+	while (off <= p->index_size - 3 * 20) {
+		llist_insert_back(l.all_objects, base + off);
+		off+=24;
+	}
+	/* this list will be pruned in cmp_two_packs later */
+	l.unique_objects = llist_copy(l.all_objects);
+	return pack_list_insert(&pack_list, &l);
+}
+
+struct pack_list * add_pack_file(char *filename)
+{
+	struct packed_git *p = packed_git;
+
+	if (strlen(filename) < 40)
+		die("Bad pack filename: %s\n", filename);
+
+	while (p) {
+		if (strstr(p->pack_name, filename))
+			/* this will silently ignore packs in alt-odb */
+			return add_pack(p);
+		p = p->next;
+	}
+	die("Filename %s not found in packed_git\n", filename);
+}
+
+void load_all()
+{
+	struct packed_git *p = packed_git;
+
+	while (p) {
+		if (p->pack_local) /* ignore alt-odb for now */
+			add_pack(p);
+		p = p->next;
+	}
+}
+
+int main(int argc, char **argv)
+{
+	int i;
+	struct pack_list *min, *red, *pl;
+
+	for (i = 1; i < argc; i++) {
+		const char *arg = argv[i];
+		if(!strcmp(arg, "--"))
+			break;
+		if(!strcmp(arg, "-a")) {
+			all = 1;
+			continue;
+		}
+		if(!strcmp(arg, "-v")) {
+			verbose = 1;
+			continue;
+		}
+		if(*arg == '-')
+			usage(pack_intersect_usage);
+		else
+			break;
+	}
+
+	prepare_packed_git();
+
+	if(all)
+		load_all();
+	else
+		while (*(argv + i) != NULL)
+			add_pack_file(*(argv + i++));
+
+	if (pack_list == NULL)
+		die("Zero packs found!\n");
+
+	cmp_packs();
+
+	load_all_objects();
+
+	minimize(&min);
+	if (verbose) {
+		fprintf(stderr, "The smallest (bytewise) set of packs is:\n");
+		pl = min;
+		while (pl) {
+			fprintf(stderr, "\t%s\n", pl->pack->pack_name);
+			pl = pl->next;
+		}
+		fprintf(stderr, "containing %ld duplicate objects "
+				"with a total size of %ldkb.\n",
+			get_pack_redundancy(min), pack_set_bytecount(min)/1024);
+		fprintf(stderr, "Redundant packs (with indexes):\n");
+	}
+	pl = red = pack_list_difference(pack_list, min);
+	while (pl) {
+		printf("%s\n%s\n",
+		       sha1_pack_index_name(pl->pack->sha1), pl->pack->pack_name);
+		pl = pl->next;
+	}
+
+	return 0;
+}
diff --git a/sha1_file.c b/sha1_file.c
index 946a353..cd814d7 100644
--- a/sha1_file.c
+++ b/sha1_file.c
@@ -424,6 +424,7 @@ struct packed_git *add_packed_git(char *
 	struct packed_git *p;
 	unsigned long idx_size;
 	void *idx_map;
+	char sha1[20];
 
 	if (check_packed_git_idx(path, &idx_size, &idx_map))
 		return NULL;
@@ -447,6 +448,8 @@ struct packed_git *add_packed_git(char *
 	p->pack_last_used = 0;
 	p->pack_use_cnt = 0;
 	p->pack_local = local;
+	if (!get_sha1_hex(path + path_len - 40 - 4, sha1))
+		memcpy(p->sha1, sha1, 20);
 	return p;
 }
 
---
0.99.9.GIT

[PATCH 2/4] Add documentation for git-pack-intersect

[Lukas Sandström]

Add documentation for git-pack-intersect

Signed-off-by: Lukas Sandström <lukass@etek.chalmers.se>

---

 Documentation/git-pack-intersect.txt |   47 ++++++++++++++++++++++++++++++++++
 1 files changed, 47 insertions(+), 0 deletions(-)
 create mode 100644 Documentation/git-pack-intersect.txt

applies-to: 2746df1385345537edc41746b191c67ee98eea20
a1b6ab6c15c3b782478a524a7fd7791ba92960e6
diff --git a/Documentation/git-pack-intersect.txt b/Documentation/git-pack-intersect.txt
new file mode 100644
index 0000000..a73d9e3
--- /dev/null
+++ b/Documentation/git-pack-intersect.txt
@@ -0,0 +1,47 @@
+git-pack-intersect(1)
+=====================
+
+NAME
+----
+git-pack-intersect - Program used to find redundant pack files.
+
+
+SYNOPSIS
+--------
+'git-pack-intersect [ -v ] < -a | .pack filename ... >'
+
+DESCRIPTION
+-----------
+This program computes which packs in your repository
+are redundant. The output is suitable for piping to
+'xargs rm' if you are in the root of the repository.
+
+OPTIONS
+-------
+
+-v::
+	Verbose. Outputs some statistics to stderr.
+	Has a small performance penalty.
+
+-a::
+	All. Processes all the local packs. Any filenames on
+	the commandline are ignored.
+
+Author
+------
+Written by Lukas Sandström <lukass@etek.chalmers.se>
+
+Documentation
+--------------
+Documentation by Lukas Sandström <lukass@etek.chalmers.se>
+
+See-Also
+--------
+gitlink:git-pack-objects[1]
+gitlink:git-repack[1]
+gitlink:git-prune-packed[1]
+
+GIT
+---
+Part of the gitlink:git[7] suite
+
---
0.99.9.GIT

[PATCH 3/4] Add git-pack-intersect to .gitignore

[Lukas Sandström]

Add git-pack-intersect to .gitignore

Signed-off-by: Lukas Sandström <lukass@etek.chalmers.se>

---

 .gitignore |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)

applies-to: f4f7accfd9ee93f528c85bae514334fbc7a70be7
7b8da40e79aa09b43ba4590c6bf8169ed85e4871
diff --git a/.gitignore b/.gitignore
index 716c340..6ff2530 100644
--- a/.gitignore
+++ b/.gitignore
@@ -60,6 +60,7 @@ git-mktag
 git-name-rev
 git-mv
 git-octopus
+git-pack-intersect
 git-pack-objects
 git-parse-remote
 git-patch-id
---
0.99.9.GIT

[PATCH 4/4] Make git-repack use git-pack-intersect

[Lukas Sandström]

Make git-repack use git-pack-intersect.

Signed-off-by: Lukas Sandström <lukass@etek.chalmers.se>

---

 git-repack.sh |   30 ++++++++++++------------------
 1 files changed, 12 insertions(+), 18 deletions(-)

applies-to: 73e05dab832dd7320a5128fbf97e693f23ffb949
2cbd6ade19a768eca47f6f7313f6831226ee58b7
diff --git a/git-repack.sh b/git-repack.sh
index d341966..3f28300 100755
--- a/git-repack.sh
+++ b/git-repack.sh
@@ -32,10 +32,6 @@ case ",$all_into_one," in
 	rev_list=
 	rev_parse='--all'
 	pack_objects=
-	# This part is a stop-gap until we have proper pack redundancy
-	# checker.
-	existing=`cd "$PACKDIR" && \
-	    find . -type f \( -name '*.pack' -o -name '*.idx' \) -print`
 	;;
 esac
 if [ "$local" ]; then
@@ -46,6 +42,14 @@ name=$(git-rev-list --objects $rev_list 
 	exit 1
 if [ -z "$name" ]; then
 	echo Nothing new to pack.
+	if test "$remove_redandant" = t ; then
+		echo "Removing redundant packs."
+		sync
+		redundant=$(git-pack-intersect -a)
+		if test "$redundant" != "" ; then
+			echo $redundant | xargs rm
+		fi
+	fi
 	exit 0
 fi
 echo "Pack pack-$name created."
@@ -58,20 +62,10 @@ exit
 
 if test "$remove_redandant" = t
 then
-	# We know $existing are all redandant only when
-	# all-into-one is used.
-	if test "$all_into_one" != '' && test "$existing" != ''
-	then
-		sync
-		( cd "$PACKDIR" &&
-		  for e in $existing
-		  do
-			case "$e" in
-			./pack-$name.pack | ./pack-$name.idx) ;;
-			*)	rm -f $e ;;
-			esac
-		  done
-		)
+	sync
+	redundant=$(git-pack-intersect -a)
+	if test "$redundant" != "" ; then
+		echo $redundant | xargs rm
 	fi
 fi
 
---
0.99.9.GIT

Re: [PATCH 0/4] Add git-pack-intersect

[Petr Baudis]

Dear diary, on Wed, Nov 09, 2005 at 02:20:59AM CET, I got a letter
where Lukas Sandström <lukass@etek.chalmers.se> said that...
> This patch series adds git-pack-intersect. It finds redundant packs
> by calculating the union of all objects present in .git/objects/pack
> and then computing the smallest set of packs which contain all the
> objects in this union.

Sounds nice, except the name - it does something else than what the name
says, so perhaps something like 'git-pack-redundant' would be more
appropriate.

-- 
				Petr "Pasky" Baudis
Stuff: http://pasky.or.cz/
VI has two modes: the one in which it beeps and the one in which
it doesn't.

Re: [PATCH 0/4] Add git-pack-intersect

[Andreas Ericsson]

Petr Baudis wrote:
> Dear diary, on Wed, Nov 09, 2005 at 02:20:59AM CET, I got a letter
> where Lukas Sandström <lukass@etek.chalmers.se> said that...
> 
>>This patch series adds git-pack-intersect. It finds redundant packs
>>by calculating the union of all objects present in .git/objects/pack
>>and then computing the smallest set of packs which contain all the
>>objects in this union.
> 
> 
> Sounds nice, except the name - it does something else than what the name
> says, so perhaps something like 'git-pack-redundant' would be more
> appropriate.
> 

It would be better if it was in git-prune or a default action for 
git-repack. I can't imagine a scenario where keeping redundant packfiles 
is useful.

-- 
Andreas Ericsson                   andreas.ericsson@op5.se
OP5 AB                             www.op5.se
Tel: +46 8-230225                  Fax: +46 8-230231

[PATCH] Rename git-pack-intersect to git-pack-redundant

[Lukas Sandström]

Petr Baudis wrote:
> Dear diary, on Wed, Nov 09, 2005 at 02:20:59AM CET, I got a letter
> where Lukas Sandström <lukass@etek.chalmers.se> said that...
> 
>>This patch series adds git-pack-intersect. It finds redundant packs
>>by calculating the union of all objects present in .git/objects/pack
>>and then computing the smallest set of packs which contain all the
>>objects in this union.
> 
> 
> Sounds nice, except the name - it does something else than what the name
> says, so perhaps something like 'git-pack-redundant' would be more
> appropriate.
> 

Yes, it would. git-pack-intersect is a working name from before
I knew what the program would actually do. In the beginning
it just computed the intersection of two pack-files...

/Lukas Sandström

-- >8 -- cut here -- >8 --
Subject: [PATCH] Rename git-pack-intersect to git-pack-redundant

This patch renames git-pack-intersect to git-pack-redundant
as suggested by Petr Baudis. The new name reflects what the
program does, rather than how it does it.

Also fix a small argument parsing bug.

Signed-off-by: Lukas Sandström <lukass@etek.chalmers.se>

---

 .gitignore                           |    2 +-
 Documentation/git-pack-redundant.txt |    6 +++---
 Makefile                             |    2 +-
 git-repack.sh                        |    4 ++--
 pack-redundant.c                     |   10 ++++++----
 5 files changed, 13 insertions(+), 11 deletions(-)
 rename Documentation/{git-pack-intersect.txt => git-pack-redundant.txt} (86%)
 rename pack-intersect.c => pack-redundant.c (98%)

applies-to: 85a2ca124a0579c98df5e6d9158a7ee358aeefef
4db829aa0d1811ccf3505aca232045d7970aec5d
diff --git a/.gitignore b/.gitignore
index 6ff2530..1d1aa57 100644
--- a/.gitignore
+++ b/.gitignore
@@ -60,7 +60,7 @@ git-mktag
 git-name-rev
 git-mv
 git-octopus
-git-pack-intersect
+git-pack-redundant
 git-pack-objects
 git-parse-remote
 git-patch-id
diff --git a/Documentation/git-pack-intersect.txt b/Documentation/git-pack-redundant.txt
similarity index 86%
rename from Documentation/git-pack-intersect.txt
rename to Documentation/git-pack-redundant.txt
index a73d9e3..3829616 100644
--- a/Documentation/git-pack-intersect.txt
+++ b/Documentation/git-pack-redundant.txt
@@ -1,14 +1,14 @@
-git-pack-intersect(1)
+git-pack-redundant(1)
 =====================
 
 NAME
 ----
-git-pack-intersect - Program used to find redundant pack files.
+git-pack-redundant - Program used to find redundant pack files.
 
 
 SYNOPSIS
 --------
-'git-pack-intersect [ -v ] < -a | .pack filename ... >'
+'git-pack-redundant [ -v ] < -a | .pack filename ... >'
 
 DESCRIPTION
 -----------
diff --git a/Makefile b/Makefile
index 4c646c9..b4dca5f 100644
--- a/Makefile
+++ b/Makefile
@@ -122,7 +122,7 @@ PROGRAMS = \
 	git-unpack-objects$X git-update-index$X git-update-server-info$X \
 	git-upload-pack$X git-verify-pack$X git-write-tree$X \
 	git-update-ref$X git-symbolic-ref$X git-check-ref-format$X \
-	git-name-rev$X git-pack-intersect$X $(SIMPLE_PROGRAMS)
+	git-name-rev$X git-pack-redundant$X $(SIMPLE_PROGRAMS)
 
 # Backward compatibility -- to be removed after 1.0
 PROGRAMS += git-ssh-pull$X git-ssh-push$X
diff --git a/git-repack.sh b/git-repack.sh
index 3f28300..4ce0022 100755
--- a/git-repack.sh
+++ b/git-repack.sh
@@ -45,7 +45,7 @@ if [ -z "$name" ]; then
 	if test "$remove_redandant" = t ; then
 		echo "Removing redundant packs."
 		sync
-		redundant=$(git-pack-intersect -a)
+		redundant=$(git-pack-redundant -a)
 		if test "$redundant" != "" ; then
 			echo $redundant | xargs rm
 		fi
@@ -63,7 +63,7 @@ exit
 if test "$remove_redandant" = t
 then
 	sync
-	redundant=$(git-pack-intersect -a)
+	redundant=$(git-pack-redundant -a)
 	if test "$redundant" != "" ; then
 		echo $redundant | xargs rm
 	fi
diff --git a/pack-intersect.c b/pack-redundant.c
similarity index 98%
rename from pack-intersect.c
rename to pack-redundant.c
index 2267478..db3dcde 100644
--- a/pack-intersect.c
+++ b/pack-redundant.c
@@ -8,8 +8,8 @@
 
 #include "cache.h"
 
-static const char pack_intersect_usage[] =
-"git-pack-intersect [ -v ]  < -a | <.pack filename> ...>";
+static const char pack_redundant_usage[] =
+"git-pack-redundant [ -v ]  < -a | <.pack filename> ...>";
 
 int all = 0, verbose = 0;
 
@@ -522,8 +522,10 @@ int main(int argc, char **argv)
 
 	for (i = 1; i < argc; i++) {
 		const char *arg = argv[i];
-		if(!strcmp(arg, "--"))
+		if(!strcmp(arg, "--")) {
+			i++;
 			break;
+		}
 		if(!strcmp(arg, "-a")) {
 			all = 1;
 			continue;
@@ -533,7 +535,7 @@ int main(int argc, char **argv)
 			continue;
 		}
 		if(*arg == '-')
-			usage(pack_intersect_usage);
+			usage(pack_redundant_usage);
 		else
 			break;
 	}
---
0.99.9.GIT

Re: [PATCH 0/4] Add git-pack-intersect

[Lukas Sandström]

Andreas Ericsson wrote:
> It would be better if it was in git-prune or a default action for
> git-repack. I can't imagine a scenario where keeping redundant packfiles
> is useful.
> 

Perhaps if git-daemon ever does caching of packfiles, removing the smaller
packfiles might not be optimal.

Integrating it with git-prune sounds like a good idea though, I'll look 
in to it.

/Lukas 

Re: [PATCH 0/4] Add git-pack-intersect

[Junio C Hamano]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset=iso-2022-jp-2, Size: 924 bytes --]

Lukas Sandstr^[.A^[Nvm <lukass@etek.chalmers.se> writes:

> Perhaps if git-daemon ever does caching of packfiles, removing the smaller
> packfiles might not be optimal.

Ah, that reminds me of something.

I did not advertise it too much, but you can put a prepackaged
packs in $GIT_DIR/pack-cache/ and it is used when upload-pack
notices it is creating that exact pack.

This is useless for ordinary repository, but can be useful for
historical repositories whose heads/tags never change (e.g. the
Linux kernel repository resurrected from bkcvs).  There you
could:

	$ git repack -a -d
	$ ln $GIT_DIR/objects/pack/pack-* $GIT_DIR/pack-cache/.

and a clone request would feed the pack, without regenerating
the pack data from scratch.  upload-pack still reads, sorts and
computes SHA1 hash of the list of objects to find the pack name,
but that is a fairly quick operation, compared to the rest of
the pack generation process.