Merge branch 'pack-path-walk'

Here is where I think the meat of the RFC really lies. There are still
some rough edges, but the data will show that 'git pack-objects
--path-walk' has the potential to be an extremely effective way to pack
objects. (Caveats will come later in the analysis section.)

Documentation/config/pack.txt

@@ -155,6 +155,14 @@ pack.useSparse::
 	commits contain certain types of direct renames. Default is
 	`true`.
 
+pack.usePathWalk::
+	When true, git will default to using the '--path-walk' option in
+	'git pack-objects' when the '--revs' option is present. This
+	algorithm groups objects by path to maximize the ability to
+	compute delta chains across historical versions of the same
+	object. This may disable other options, such as using bitmaps to
+	enumerate objects.
+
 pack.preferBitmapTips::
 	When selecting which commits will receive bitmaps, prefer a
 	commit at the tip of any reference that is a suffix of any value

builtin/pack-objects.c

@@ -39,6 +39,9 @@
 #include "promisor-remote.h"
 #include "pack-mtimes.h"
 #include "parse-options.h"
+#include "blob.h"
+#include "tree.h"
+#include "path-walk.h"
 
 /*
  * Objects we are going to pack are collected in the `to_pack` structure.
@@ -215,6 +218,7 @@ static int delta_search_threads;
 static int pack_to_stdout;
 static int sparse;
 static int thin;
+static int path_walk;
 static int num_preferred_base;
 static struct progress *progress_state;
 
@@ -3139,6 +3143,35 @@ static int add_ref_tag(const char *tag UNUSED, const struct object_id *oid,
 	return 0;
 }
 
+static int should_attempt_deltas(struct object_entry *entry)
+{
+	if (DELTA(entry))
+		/* This happens if we decided to reuse existing
+		 * delta from a pack.  "reuse_delta &&" is implied.
+		 */
+		return 0;
+
+	if (!entry->type_valid || oe_size_less_than(&to_pack, entry, 50))
+		return 0;
+
+	if (entry->no_try_delta)
+		return 0;
+
+	if (!entry->preferred_base) {
+		if (oe_type(entry) < 0)
+			die(_("unable to get type of object %s"),
+				oid_to_hex(&entry->idx.oid));
+	} else if (oe_type(entry) < 0) {
+		/*
+		 * This object is not found, but we
+		 * don't have to include it anyway.
+		 */
+		return 0;
+	}
+
+	return 1;
+}
+
 static void prepare_pack(int window, int depth)
 {
 	struct object_entry **delta_list;
@@ -3169,33 +3202,11 @@ static void prepare_pack(int window, int depth)
 	for (i = 0; i < to_pack.nr_objects; i++) {
 		struct object_entry *entry = to_pack.objects + i;
 
-		if (DELTA(entry))
-			/* This happens if we decided to reuse existing
-			 * delta from a pack.  "reuse_delta &&" is implied.
-			 */
-			continue;
-
-		if (!entry->type_valid ||
-		    oe_size_less_than(&to_pack, entry, 50))
-			continue;
-
-		if (entry->no_try_delta)
+		if (!should_attempt_deltas(entry))
 			continue;
 
-		if (!entry->preferred_base) {
+		if (!entry->preferred_base)
 			nr_deltas++;
-			if (oe_type(entry) < 0)
-				die(_("unable to get type of object %s"),
-				    oid_to_hex(&entry->idx.oid));
-		} else {
-			if (oe_type(entry) < 0) {
-				/*
-				 * This object is not found, but we
-				 * don't have to include it anyway.
-				 */
-				continue;
-			}
-		}
 
 		delta_list[n++] = entry;
 	}
@@ -4109,6 +4120,117 @@ static void mark_bitmap_preferred_tips(void)
 	}
 }
 
+static inline int is_oid_interesting(struct repository *repo,
+				     struct object_id *oid,
+				     enum object_type type)
+{
+	if (type == OBJ_TAG) {
+		struct tag *t = lookup_tag(repo, oid);
+		return t && !(t->object.flags & UNINTERESTING);
+	}
+
+	if (type == OBJ_COMMIT) {
+		struct commit *c = lookup_commit(repo, oid);
+		return c && !(c->object.flags & UNINTERESTING);
+	}
+
+	if (type == OBJ_TREE) {
+		struct tree *t = lookup_tree(repo, oid);
+		return t && !(t->object.flags & UNINTERESTING);
+	}
+
+	if (type == OBJ_BLOB) {
+		struct blob *b = lookup_blob(repo, oid);
+		return b && !(b->object.flags & UNINTERESTING);
+	}
+
+	return 0;
+}
+
+static int add_objects_by_path(const char *path,
+			       struct oid_array *oids,
+			       enum object_type type,
+			       void *data)
+{
+	struct object_entry **delta_list;
+	size_t oe_start = to_pack.nr_objects;
+	size_t oe_end;
+	unsigned int sub_list_size;
+	unsigned int *processed = data;
+
+	/*
+	 * First, add all objects to the packing data, including the ones
+	 * marked UNINTERESTING (translated to 'exclude') as they can be
+	 * used as delta bases.
+	 */
+	for (size_t i = 0; i < oids->nr; i++) {
+		struct object_id *oid = &oids->oid[i];
+		int exclude = !is_oid_interesting(the_repository, oid, type);
+		add_object_entry(oid, type, path, exclude);
+	}
+
+	oe_end = to_pack.nr_objects;
+
+	/* We can skip delta calculations if it is a no-op. */
+	if (oe_end == oe_start || !window)
+		return 0;
+
+	sub_list_size = 0;
+	ALLOC_ARRAY(delta_list, oe_end - oe_start);
+
+	for (size_t i = 0; i < oe_end - oe_start; i++) {
+		struct object_entry *entry = to_pack.objects + oe_start + i;
+
+		if (!should_attempt_deltas(entry))
+			continue;
+
+		delta_list[sub_list_size++] = entry;
+	}
+
+	/*
+	 * Find delta bases among this list of objects that all match the same
+	 * path. This causes the delta compression to be interleaved in the
+	 * object walk, which can lead to confusing progress indicators. This is
+	 * also incompatible with threaded delta calculations. In the future,
+	 * consider creating a list of regions in the full to_pack.objects array
+	 * that could be picked up by the threaded delta computation.
+	 */
+	if (sub_list_size && window) {
+		QSORT(delta_list, sub_list_size, type_size_sort);
+		find_deltas(delta_list, &sub_list_size, window, depth, processed);
+	}
+
+	free(delta_list);
+	return 0;
+}
+
+static void get_object_list_path_walk(struct rev_info *revs)
+{
+	struct path_walk_info info = PATH_WALK_INFO_INIT;
+	unsigned int processed = 0;
+
+	info.revs = revs;
+
+	info.revs->tag_objects = 1;
+	info.tags = 1;
+	info.commits = 1;
+	info.trees = 1;
+	info.blobs = 1;
+	info.path_fn = add_objects_by_path;
+	info.path_fn_data = &processed;
+
+	/*
+	 * Allow the --[no-]sparse option to be interesting here, if only
+	 * for testing purposes. Paths with no interesting objects will not
+	 * contribute to the resulting pack, but only create noisy preferred
+	 * base objects.
+	 */
+	info.prune_all_uninteresting = sparse;
+
+	if (walk_objects_by_path(&info))
+		die(_("failed to pack objects via path-walk"));
+}
+
 static void get_object_list(struct rev_info *revs, int ac, const char **av)
 {
 	struct setup_revision_opt s_r_opt = {
@@ -4155,7 +4277,7 @@ static void get_object_list(struct rev_info *revs, int ac, const char **av)
 
 	warn_on_object_refname_ambiguity = save_warning;
 
-	if (use_bitmap_index && !get_object_list_from_bitmap(revs))
+	if (use_bitmap_index && !path_walk && !get_object_list_from_bitmap(revs))
 		return;
 
 	if (use_delta_islands)
@@ -4164,15 +4286,19 @@ static void get_object_list(struct rev_info *revs, int ac, const char **av)
 	if (write_bitmap_index)
 		mark_bitmap_preferred_tips();
 
-	if (prepare_revision_walk(revs))
-		die(_("revision walk setup failed"));
-	mark_edges_uninteresting(revs, show_edge, sparse);
-
 	if (!fn_show_object)
 		fn_show_object = show_object;
-	traverse_commit_list(revs,
-			     show_commit, fn_show_object,
-			     NULL);
+
+	if (path_walk) {
+		get_object_list_path_walk(revs);
+	} else {
+		if (prepare_revision_walk(revs))
+			die(_("revision walk setup failed"));
+		mark_edges_uninteresting(revs, show_edge, sparse);
+		traverse_commit_list(revs,
+				show_commit, fn_show_object,
+				NULL);
+	}
 
 	if (unpack_unreachable_expiration) {
 		revs->ignore_missing_links = 1;
@@ -4367,6 +4493,8 @@ int cmd_pack_objects(int argc, const char **argv, const char *prefix)
 			 N_("use the sparse reachability algorithm")),
 		OPT_BOOL(0, "thin", &thin,
 			 N_("create thin packs")),
+		OPT_BOOL(0, "path-walk", &path_walk,
+			 N_("use the path-walk API to walk objects when possible")),
 		OPT_BOOL(0, "shallow", &shallow,
 			 N_("create packs suitable for shallow fetches")),
 		OPT_BOOL(0, "honor-pack-keep", &ignore_packed_keep_on_disk,
@@ -4405,11 +4533,14 @@ int cmd_pack_objects(int argc, const char **argv, const char *prefix)
 
 	disable_replace_refs();
 
+	path_walk = git_env_bool("GIT_TEST_PACK_PATH_WALK", -1);
 	sparse = git_env_bool("GIT_TEST_PACK_SPARSE", -1);
 	if (the_repository->gitdir) {
 		prepare_repo_settings(the_repository);
 		if (sparse < 0)
 			sparse = the_repository->settings.pack_use_sparse;
+		if (path_walk < 0)
+			path_walk = the_repository->settings.pack_use_path_walk;
 		if (the_repository->settings.pack_use_multi_pack_reuse)
 			allow_pack_reuse = MULTI_PACK_REUSE;
 	}
@@ -4447,7 +4578,19 @@ int cmd_pack_objects(int argc, const char **argv, const char *prefix)
 		window = 0;
 
 	strvec_push(&rp, "pack-objects");
-	if (thin) {
+
+	if (path_walk && filter_options.choice) {
+		warning(_("cannot use --filter with --path-walk"));
+		path_walk = 0;
+	}
+	if (path_walk) {
+		strvec_push(&rp, "--boundary");
+		 /*
+		  * We must disable the bitmaps because we are removing
+		  * the --objects / --objects-edge[-aggressive] options.
+		  */
+		use_bitmap_index = 0;
+	} else if (thin) {
 		use_internal_rev_list = 1;
 		strvec_push(&rp, shallow
 				? "--objects-edge-aggressive"

builtin/repack.c

@@ -57,6 +57,7 @@ struct pack_objects_args {
 	int no_reuse_object;
 	int quiet;
 	int local;
+	int path_walk;
 	struct list_objects_filter_options filter_options;
 };
 
@@ -288,6 +289,8 @@ static void prepare_pack_objects(struct child_process *cmd,
 		strvec_pushf(&cmd->args, "--no-reuse-delta");
 	if (args->no_reuse_object)
 		strvec_pushf(&cmd->args, "--no-reuse-object");
+	if (args->path_walk)
+		strvec_pushf(&cmd->args, "--path-walk");
 	if (args->local)
 		strvec_push(&cmd->args,  "--local");
 	if (args->quiet)
@@ -1157,6 +1160,8 @@ int cmd_repack(int argc, const char **argv, const char *prefix)
 				N_("pass --no-reuse-delta to git-pack-objects")),
 		OPT_BOOL('F', NULL, &po_args.no_reuse_object,
 				N_("pass --no-reuse-object to git-pack-objects")),
+		OPT_BOOL(0, "path-walk", &po_args.path_walk,
+				N_("pass --path-walk to git-pack-objects")),
 		OPT_NEGBIT('n', NULL, &run_update_server_info,
 				N_("do not run git-update-server-info"), 1),
 		OPT__QUIET(&po_args.quiet, N_("be quiet")),