git/list[1] front-page[2] threads[3] people[4] search[5] about
 

[PATCH 1/3] prune: lazily perform reachability traversal

From
Jeff King <peff@peff.net>
Date
Feb 14, 2019, 04:35 UTC
Message-ID
<20190214043522.GA19183@sigill.intra.peff.net>
In-Reply-To
<20190214043127.GA19019@sigill.intra.peff.net>

The general strategy of "git prune" is to do a full reachability walk, then for each loose object see if we found it in our walk. But if we don't have any loose objects, we don't need to do the expensive walk in the first place.

This patch postpones that walk until the first time we need to see its results.

Note that this is really a specific case of a more general optimization, which is that we could traverse only far enough to find the object under consideration (i.e., stop the traversal when we find it, then pick up again when asked about the next object, etc). That could save us in some instances from having to do a full walk. But it's actually a bit tricky to do with our traversal code, and you'd need to do a full walk anyway if you have even a single unreachable object (which you generally do, if any objects are actually left after running git-repack).

So in practice this lazy-load of the full walk catches one easy but common case (i.e., you've just repacked via git-gc, and there's nothing unreachable).

The perf script is fairly contrived, but it does show off the improvement:

  Test                            HEAD^             HEAD
  -------------------------------------------------------------------------
  5304.4: prune with no objects   3.66(3.60+0.05)   0.00(0.00+0.00) -100.0%
and would let us know if we accidentally regress this optimization.

Note also that we need to take special care with prune_shallow(), which relies on us having performed the traversal. So this optimization can only kick in for a non-shallow repository. Since this is easy to get wrong and is not covered by existing tests, let's add an extra test to t5304 that covers this case explicitly.

Signed-off-by: Jeff King <peff@peff.net>
---
The diff looks nice with --color-moved. I wish there was a way to
communicate that information in a plaintext email.
 builtin/prune.c       | 43 ++++++++++++++++++++++++++++++++-----------
 t/perf/p5304-prune.sh | 24 ++++++++++++++++++++++++
 t/t5304-prune.sh      | 12 ++++++++++++
 3 files changed, 68 insertions(+), 11 deletions(-)
 create mode 100755 t/perf/p5304-prune.sh
diff --git a/builtin/prune.c b/builtin/prune.c
index 1ec9ddd751..04b6573945 100644
--- a/builtin/prune.c
+++ b/builtin/prune.c
@@ -31,16 +31,40 @@ static int prune_tmp_file(const char *fullpath)
 	return 0;
 }
 
-static int prune_object(const struct object_id *oid, const char *fullpath,
-			void *data)
+static void perform_reachability_traversal(struct rev_info *revs)
 {
-	struct stat st;
+	static int initialized;
+	struct progress *progress = NULL;
+
+	if (initialized)
+		return;
+
+	if (show_progress)
+		progress = start_delayed_progress(_("Checking connectivity"), 0);
+	mark_reachable_objects(revs, 1, expire, progress);
+	stop_progress(&progress);
+	initialized = 1;
+}
+
+static int is_object_reachable(const struct object_id *oid,
+			       struct rev_info *revs)
+{
+	perform_reachability_traversal(revs);
 
 	/*
 	 * Do we know about this object?
 	 * It must have been reachable
 	 */
-	if (lookup_object(the_repository, oid->hash))
+	return !!lookup_object(the_repository, oid->hash);
+}
+
+static int prune_object(const struct object_id *oid, const char *fullpath,
+			void *data)
+{
+	struct rev_info *revs = data;
+	struct stat st;
+
+	if (is_object_reachable(oid, revs))
 		return 0;
 
 	if (lstat(fullpath, &st)) {
@@ -102,7 +126,6 @@ static void remove_temporary_files(const char *path)
 int cmd_prune(int argc, const char **argv, const char *prefix)
 {
 	struct rev_info revs;
-	struct progress *progress = NULL;
 	int exclude_promisor_objects = 0;
 	const struct option options[] = {
 		OPT__DRY_RUN(&show_only, N_("do not remove, show only")),
@@ -142,17 +165,13 @@ int cmd_prune(int argc, const char **argv, const char *prefix)
 
 	if (show_progress == -1)
 		show_progress = isatty(2);
-	if (show_progress)
-		progress = start_delayed_progress(_("Checking connectivity"), 0);
 	if (exclude_promisor_objects) {
 		fetch_if_missing = 0;
 		revs.exclude_promisor_objects = 1;
 	}
 
-	mark_reachable_objects(&revs, 1, expire, progress);
-	stop_progress(&progress);
 	for_each_loose_file_in_objdir(get_object_directory(), prune_object,
-				      prune_cruft, prune_subdir, NULL);
+				      prune_cruft, prune_subdir, &revs);
 
 	prune_packed_objects(show_only ? PRUNE_PACKED_DRY_RUN : 0);
 	remove_temporary_files(get_object_directory());
@@ -160,8 +179,10 @@ int cmd_prune(int argc, const char **argv, const char *prefix)
 	remove_temporary_files(s);
 	free(s);
 
-	if (is_repository_shallow(the_repository))
+	if (is_repository_shallow(the_repository)) {
+		perform_reachability_traversal(&revs);
 		prune_shallow(show_only ? PRUNE_SHOW_ONLY : 0);
+	}
 
 	return 0;
 }
diff --git a/t/perf/p5304-prune.sh b/t/perf/p5304-prune.sh
new file mode 100755
index 0000000000..3c852084eb
--- /dev/null
+++ b/t/perf/p5304-prune.sh
@@ -0,0 +1,24 @@
+#!/bin/sh
+
+test_description='performance tests of prune'
+. ./perf-lib.sh
+
+test_perf_default_repo
+
+test_expect_success 'remove reachable loose objects' '
+	git repack -ad
+'
+
+test_expect_success 'remove unreachable loose objects' '
+	git prune
+'
+
+test_expect_success 'confirm there are no loose objects' '
+	git count-objects | grep ^0
+'
+
+test_perf 'prune with no objects' '
+	git prune
+'
+
+test_done
diff --git a/t/t5304-prune.sh b/t/t5304-prune.sh
index 270da21ac3..2c19a790c1 100755
--- a/t/t5304-prune.sh
+++ b/t/t5304-prune.sh
@@ -274,6 +274,18 @@ test_expect_success 'prune .git/shallow' '
 	test_path_is_missing .git/shallow
 '
 
+test_expect_success 'prune .git/shallow when there are no loose objects' '
+	SHA1=$(echo hi|git commit-tree HEAD^{tree}) &&
+	echo $SHA1 >.git/shallow &&
+	git update-ref refs/heads/shallow-tip $SHA1 &&
+	git repack -ad &&
+	# verify assumption that all loose objects are gone
+	git count-objects | grep ^0 &&
+	git prune &&
+	echo $SHA1 >expect &&
+	test_cmp expect .git/shallow
+'
+
 test_expect_success 'prune: handle alternate object database' '
 	test_create_repo A &&
 	git -C A commit --allow-empty -m "initial commit" &&
-- 
2.21.0.rc0.586.gffba1126a0
Previous: Jeff KingNext: Eric Sunshine
Message 2 of 57 in “some prune optimizations”
  1. 0/3 some prune optimizationsJeff King, Feb 14, 2019
  2. 1/3 prune: lazily perform reachability traversalJeff King, Feb 14, 2019
  3. Eric SunshineFeb 14, 2019
  4. Jeff KingFeb 14, 2019
  5. 2/3 prune: use bitmaps for reachability traversalJeff King, Feb 14, 2019
  6. bitmaps by default? [was: prune: use bitmaps for reachability traversal]Eric Wong, Mar 9, 2019
  7. Jeff KingMar 10, 2019
  8. repack: enable bitmaps by default on bare reposEric Wong, Mar 12, 2019
  9. Ævar Arnfjörð BjarmasonMar 12, 2019
  10. Jeff KingMar 12, 2019
  11. Jeff KingMar 12, 2019
  12. Eric WongMar 13, 2019
  13. Jeff KingMar 13, 2019
  14. repack: enable bitmaps by default on bare reposEric Wong, Mar 14, 2019
  15. Jeff KingMar 14, 2019
  16. 0/2 enable bitmap hash-cache by defaultJeff King, Mar 15, 2019
  17. 1/2 t5310: correctly remove bitmaps for jgit testJeff King, Mar 15, 2019
  18. SZEDER GáborMar 15, 2019
  19. Jeff KingMar 15, 2019
  20. 2/2 pack-objects: default to writing bitmap hash-cacheJeff King, Mar 15, 2019
  21. Ævar Arnfjörð BjarmasonApr 9, 2019
  22. Jeff KingApr 10, 2019
  23. Junio C HamanoApr 25, 2019
  24. Jeff KingMay 4, 2019
  25. Ævar Arnfjörð BjarmasonMay 4, 2019
  26. SZEDER GáborMay 4, 2019
  27. Ævar Arnfjörð BjarmasonMay 8, 2019
  28. Junio C HamanoMay 9, 2019
  29. Jeff KingMay 7, 2019
  30. Ævar Arnfjörð BjarmasonMay 7, 2019
  31. Jeff KingMay 8, 2019
  32. Derrick StoleeMay 8, 2019
  33. Ævar Arnfjörð BjarmasonMay 8, 2019
  34. Jeff KingMay 8, 2019
  35. Jeff KingMay 23, 2019
  36. Derrick StoleeMay 23, 2019
  37. Jeff KingMay 24, 2019
  38. Derrick StoleeMay 24, 2019
  39. Ævar Arnfjörð BjarmasonMay 23, 2019
  40. Jeff KingMay 24, 2019
  41. Ævar Arnfjörð BjarmasonMay 24, 2019
  42. Jeff KingMay 24, 2019
  43. Ævar Arnfjörð BjarmasonMay 24, 2019
  44. SZEDER GáborMay 24, 2019
  45. Ævar Arnfjörð BjarmasonMay 24, 2019
  46. SZEDER GáborMay 24, 2019
  47. Ævar Arnfjörð BjarmasonMay 24, 2019
  48. SZEDER GáborMay 24, 2019
  49. Ævar Arnfjörð BjarmasonMay 24, 2019
  50. pack-bitmap: look for an uninteresting bitmapDerrick Stolee, May 24, 2019
  51. Derrick StoleeApr 15, 2019
  52. Jeff KingApr 18, 2019
  53. t5304: add a test for pruning with bitmapsJeff King, Apr 18, 2019
  54. Derrick StoleeApr 20, 2019
  55. Jeff KingApr 20, 2019
  56. Derrick StoleeApr 20, 2019
  57. 3/3 prune: check SEEN flag for reachabilityJeff King, Feb 14, 2019

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.