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

[PATCH] diff: restrict when prefetching occurs

From
JTJonathan Tan <jonathantanmy@google.com>
Date
Mar 31, 2020, 02:04 UTC
Message-ID
<20200331020418.55640-1-jonathantanmy@google.com>

Commit 7fbbcb21b1 ("diff: batch fetching of missing blobs", 2019-04-08) optimized "diff" by prefetching blobs in a partial clone, but there are some cases wherein blobs do not need to be prefetched. In particular, if (1) no blob data is included in the output of the diff, (2) break-rewrite detection is not requested, and (3) no inexact rename detection is needed, then no blobs are read at all.

Therefore, in such a case, do not prefetch. Change diffcore_std() to only prefetch if (1) and/or (2) is not true (currently, it always prefetches); change diffcore_rename() to prefetch if (3) is not true and no prefetch has yet occurred.

Helped-by: Jeff King <peff@peff.net>
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
---
I decided to revisit [1] because at $DAYJOB there was a new case that
required this. I used Peff's code from [2] for the rebase part, and also
automatically prefetch if blob data will be included in the output of
the diff and/or break-rewrite detection is requested.
[1] https://lore.kernel.org/git/20200128213508.31661-1-jonathantanmy@google.com/
[2] https://lore.kernel.org/git/20200130055136.GA2184413@coredump.intra.peff.net/
---
 diff.c                        | 26 +++++++++++++++----
 diffcore-rename.c             | 40 +++++++++++++++++++++++++++-
 diffcore.h                    |  2 +-
 t/t4067-diff-partial-clone.sh | 49 +++++++++++++++++++++++++++++++++++
 4 files changed, 110 insertions(+), 7 deletions(-)
diff --git a/diff.c b/diff.c
index 1010d806f5..19c5d638d6 100644
--- a/diff.c
+++ b/diff.c
@@ -6507,10 +6507,24 @@ static void add_if_missing(struct repository *r,
 
 void diffcore_std(struct diff_options *options)
 {
-	if (options->repo == the_repository && has_promisor_remote()) {
-		/*
-		 * Prefetch the diff pairs that are about to be flushed.
-		 */
+	int prefetched = 0;
+	int output_formats_to_prefetch = DIFF_FORMAT_DIFFSTAT |
+		DIFF_FORMAT_NUMSTAT |
+		DIFF_FORMAT_PATCH |
+		DIFF_FORMAT_SHORTSTAT |
+		DIFF_FORMAT_DIRSTAT;
+
+	/*
+	 * Check if the user requested a blob-data-requiring diff output and/or
+	 * break-rewrite detection (which requires blob data). If yes, prefetch
+	 * the diff pairs.
+	 *
+	 * If no prefetching occurs, diffcore_rename() will prefetch if it
+	 * decides that it needs inexact rename detection.
+	 */
+	if (options->repo == the_repository && has_promisor_remote() &&
+	    (options->output_format & output_formats_to_prefetch ||
+	     (!options->found_follow && options->break_opt != -1))) {
 		int i;
 		struct diff_queue_struct *q = &diff_queued_diff;
 		struct oid_array to_fetch = OID_ARRAY_INIT;
@@ -6520,6 +6534,8 @@ void diffcore_std(struct diff_options *options)
 			add_if_missing(options->repo, &to_fetch, p->one);
 			add_if_missing(options->repo, &to_fetch, p->two);
 		}
+		prefetched = 1;
+
 		if (to_fetch.nr)
 			/*
 			 * NEEDSWORK: Consider deduplicating the OIDs sent.
@@ -6538,7 +6554,7 @@ void diffcore_std(struct diff_options *options)
 			diffcore_break(options->repo,
 				       options->break_opt);
 		if (options->detect_rename)
-			diffcore_rename(options);
+			diffcore_rename(options, prefetched);
 		if (options->break_opt != -1)
 			diffcore_merge_broken();
 	}
diff --git a/diffcore-rename.c b/diffcore-rename.c
index e189f407af..962565f066 100644
--- a/diffcore-rename.c
+++ b/diffcore-rename.c
@@ -7,6 +7,7 @@
 #include "object-store.h"
 #include "hashmap.h"
 #include "progress.h"
+#include "promisor-remote.h"
 
 /* Table of rename/copy destinations */
 
@@ -448,7 +449,18 @@ static int find_renames(struct diff_score *mx, int dst_cnt, int minimum_score, i
 	return count;
 }
 
-void diffcore_rename(struct diff_options *options)
+static void add_if_missing(struct repository *r,
+			   struct oid_array *to_fetch,
+			   const struct diff_filespec *filespec)
+{
+	if (filespec && filespec->oid_valid &&
+	    !S_ISGITLINK(filespec->mode) &&
+	    oid_object_info_extended(r, &filespec->oid, NULL,
+				     OBJECT_INFO_FOR_PREFETCH))
+		oid_array_append(to_fetch, &filespec->oid);
+}
+
+void diffcore_rename(struct diff_options *options, int prefetched)
 {
 	int detect_rename = options->detect_rename;
 	int minimum_score = options->rename_score;
@@ -538,6 +550,32 @@ void diffcore_rename(struct diff_options *options)
 		break;
 	}
 
+	if (!prefetched) {
+		/*
+		 * At this point we know there's actual work to do: we have rename
+		 * destinations that didn't find an exact match, and we have potential
+		 * sources. So we'll have to do inexact rename detection, which
+		 * requires looking at the blobs.
+		 *
+		 * If we haven't already prefetched, it's worth pre-fetching
+		 * them as a group now.
+		 */
+		int i;
+		struct oid_array to_fetch = OID_ARRAY_INIT;
+
+		for (i = 0; i < rename_dst_nr; i++) {
+			if (rename_dst[i].pair)
+				continue; /* already found exact match */
+			add_if_missing(options->repo, &to_fetch, rename_dst[i].two);
+		}
+		for (i = 0; i < rename_src_nr; i++)
+			add_if_missing(options->repo, &to_fetch, rename_src[i].p->one);
+		if (to_fetch.nr)
+			promisor_remote_get_direct(options->repo,
+						   to_fetch.oid, to_fetch.nr);
+		oid_array_clear(&to_fetch);
+	}
+
 	if (options->show_rename_progress) {
 		progress = start_delayed_progress(
 				_("Performing inexact rename detection"),
diff --git a/diffcore.h b/diffcore.h
index 7c07347e42..9f69506574 100644
--- a/diffcore.h
+++ b/diffcore.h
@@ -144,7 +144,7 @@ struct diff_filepair *diff_queue(struct diff_queue_struct *,
 void diff_q(struct diff_queue_struct *, struct diff_filepair *);
 
 void diffcore_break(struct repository *, int);
-void diffcore_rename(struct diff_options *);
+void diffcore_rename(struct diff_options *, int prefetched);
 void diffcore_merge_broken(void);
 void diffcore_pickaxe(struct diff_options *);
 void diffcore_order(const char *orderfile);
diff --git a/t/t4067-diff-partial-clone.sh b/t/t4067-diff-partial-clone.sh
index 4831ad35e6..7acb64727d 100755
--- a/t/t4067-diff-partial-clone.sh
+++ b/t/t4067-diff-partial-clone.sh
@@ -131,4 +131,53 @@ test_expect_success 'diff with rename detection batches blobs' '
 	test_line_count = 1 done_lines
 '
 
+test_expect_success 'diff does not fetch anything if inexact rename detection is not needed' '
+	test_when_finished "rm -rf server client trace" &&
+
+	test_create_repo server &&
+	echo a >server/a &&
+	printf "b\nb\nb\nb\nb\n" >server/b &&
+	git -C server add a b &&
+	git -C server commit -m x &&
+	rm server/b &&
+	printf "b\nb\nb\nb\nb\n" >server/c &&
+	git -C server add c &&
+	git -C server commit -a -m x &&
+
+	test_config -C server uploadpack.allowfilter 1 &&
+	test_config -C server uploadpack.allowanysha1inwant 1 &&
+	git clone --bare --filter=blob:limit=0 "file://$(pwd)/server" client &&
+
+	# Ensure no fetches.
+	GIT_TRACE_PACKET="$(pwd)/trace" git -C client diff --raw -M HEAD^ HEAD &&
+	! test_path_exists trace
+'
+
+test_expect_success 'diff --break-rewrites fetches only if necessary, and batches blobs if it does' '
+	test_when_finished "rm -rf server client trace" &&
+
+	test_create_repo server &&
+	echo a >server/a &&
+	printf "b\nb\nb\nb\nb\n" >server/b &&
+	git -C server add a b &&
+	git -C server commit -m x &&
+	printf "c\nc\nc\nc\nc\n" >server/b &&
+	git -C server commit -a -m x &&
+
+	test_config -C server uploadpack.allowfilter 1 &&
+	test_config -C server uploadpack.allowanysha1inwant 1 &&
+	git clone --bare --filter=blob:limit=0 "file://$(pwd)/server" client &&
+
+	# Ensure no fetches.
+	GIT_TRACE_PACKET="$(pwd)/trace" git -C client diff --raw -M HEAD^ HEAD &&
+	! test_path_exists trace &&
+
+	# But with --break-rewrites, ensure that there is exactly 1 negotiation
+	# by checking that there is only 1 "done" line sent. ("done" marks the
+	# end of negotiation.)
+	GIT_TRACE_PACKET="$(pwd)/trace" git -C client diff --break-rewrites --raw -M HEAD^ HEAD &&
+	grep "git> done" trace >done_lines &&
+	test_line_count = 1 done_lines
+'
+
 test_done

-- 
2.26.0.rc2.310.g2932bb562d-goog
Next: Derrick Stolee
Message 1 of 26 in “diff: restrict when prefetching occurs”
  1. diff: restrict when prefetching occursJonathan Tan, Mar 31, 2020
  2. Derrick StoleeMar 31, 2020
  3. Jonathan TanMar 31, 2020
  4. Derrick StoleeMar 31, 2020
  5. Junio C HamanoMar 31, 2020
  6. Junio C HamanoMar 31, 2020
  7. 0/2 Restrict when prefetcing occursJonathan Tan, Apr 2, 2020
  8. 1/2 promisor-remote: accept 0 as oid_nr in functionJonathan Tan, Apr 2, 2020
  9. Junio C HamanoApr 2, 2020
  10. Jonathan TanApr 2, 2020
  11. 2/2 diff: restrict when prefetching occursJonathan Tan, Apr 2, 2020
  12. Junio C HamanoApr 2, 2020
  13. Jonathan TanApr 2, 2020
  14. Junio C HamanoApr 2, 2020
  15. Junio C HamanoApr 2, 2020
  16. Jonathan TanApr 3, 2020
  17. Junio C HamanoApr 3, 2020
  18. Junio C HamanoApr 2, 2020
  19. Derrick StoleeApr 6, 2020
  20. Garima SinghApr 6, 2020
  21. 0/4 Restrict when prefetcing occursJonathan Tan, Apr 7, 2020
  22. 1/4 promisor-remote: accept 0 as oid_nr in functionJonathan Tan, Apr 7, 2020
  23. 2/4 diff: make diff_populate_filespec_options structJonathan Tan, Apr 7, 2020
  24. Junio C HamanoApr 7, 2020
  25. 3/4 diff: refactor object readJonathan Tan, Apr 7, 2020
  26. 4/4 diff: restrict when prefetching occursJonathan Tan, Apr 7, 2020

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.