Hey Taylor,

On 11 May 2023, at 20:00, Taylor Blau wrote:

On Thu, May 11, 2023 at 06:10:32PM +0000, John Cai via GitGitGadget wrote:

quoted

From: John Cai <redacted>

At GitLab, we have a system that creates ephemeral internal refs that
don't live long before getting deleted. Having an option to exclude
certain refs from a packed-refs file allows these internal references to
be deleted much more efficiently.

Add an --exclude option to the pack-refs builtin, and use the ref
exclusions API to exclude certain refs from being packed into the final
packed-refs file

Signed-off-by: John Cai <redacted>
---
 Documentation/git-pack-refs.txt | 12 +++++++++++-
 builtin/pack-refs.c             | 20 ++++++++++++++++----
 refs.c                          |  4 ++--
 refs.h                          |  7 ++++++-
 refs/debug.c                    |  4 ++--
 refs/files-backend.c            | 16 ++++++++++------
 refs/packed-backend.c           |  2 +-
 refs/refs-internal.h            |  3 ++-
 revision.h                      |  2 +-
 t/helper/test-ref-store.c       |  3 ++-
 t/t3210-pack-refs.sh            | 16 ++++++++++++++++
 11 files changed, 69 insertions(+), 20 deletions(-)

diff --git a/Documentation/git-pack-refs.txt b/Documentation/git-pack-refs.txt
index e011e5fead3..c0f7426e519 100644
--- a/Documentation/git-pack-refs.txt
+++ b/Documentation/git-pack-refs.txt

@@ -8,7 +8,7 @@ git-pack-refs - Pack heads and tags for efficient repository access
 SYNOPSIS
 --------
 [verse]
-'git pack-refs' [--all] [--no-prune]
+'git pack-refs' [--all] [--no-prune] [--exclude <pattern>]

 DESCRIPTION
 -----------

@@ -60,6 +60,16 @@ interests.
 The command usually removes loose refs under `$GIT_DIR/refs`
 hierarchy after packing them.  This option tells it not to.

+--exclude <pattern>::
+
+Do not pack refs matching the given `glob(7)` pattern. Repetitions of this option
+accumulate exclusion patterns. Use `--no-exclude` to clear and reset the list of
+patterns. If a ref is already packed, including it with `--exclude` will not
+unpack it.
+
+When used with `--all`, it will use the difference between the set of all refs,
+and what is provided to `--exclude`.
+

I think this last paragraph could be simplified, though feel free to
discard my suggestion if you think it makes things less clear.

  When used with `--all`, pack only loose refs which do not match any of
  the provided `--exclude` patterns.

I like the wording here, thanks

quoted

 int cmd_pack_refs(int argc, const char **argv, const char *prefix)
 {
 	unsigned int flags = PACK_REFS_PRUNE;
+	static struct ref_exclusions excludes = REF_EXCLUSIONS_INIT;
+	struct pack_refs_opts pack_refs_opts = {.exclusions = &excludes, .flags = flags};
+	static struct string_list option_excluded_refs = STRING_LIST_INIT_NODUP;
+	struct string_list_item *item;

Since this list does not appear to be sensitive to its order, have you
considered using the strvec API instead of the string_list one?

Thanks for this suggestion--you're right in that the order doesn't matter here.
The only thing is, the only option parsing macro I could find that operates on
strvec is OPT_PASSTHRU_ARGV. I tried it out, and it seem to work just fine.

quoted

diff --git a/refs.c b/refs.c
index d2a98e1c21f..881a0da65cf 100644
--- a/refs.c
+++ b/refs.c

@@ -2132,9 +2132,9 @@ void base_ref_store_init(struct ref_store *refs, struct repository *repo,
 }

 /* backend functions */
-int refs_pack_refs(struct ref_store *refs, unsigned int flags)
+int refs_pack_refs(struct ref_store *refs, struct pack_refs_opts *opts)
 {
-	return refs->be->pack_refs(refs, flags);
+	return refs->be->pack_refs(refs, opts);
 }

 int peel_iterated_oid(const struct object_id *base, struct object_id *peeled)

diff --git a/refs.h b/refs.h
index 123cfa44244..46020bd335c 100644
--- a/refs.h
+++ b/refs.h

@@ -63,6 +63,11 @@ struct worktree;
 #define RESOLVE_REF_NO_RECURSE 0x02
 #define RESOLVE_REF_ALLOW_BAD_NAME 0x04

+struct pack_refs_opts {
+	unsigned int flags;
+	struct ref_exclusions *exclusions;

I think this would be OK to include directly in the struct instead of
via a pointer, but either is fine.

quoted

@@ -1175,15 +1176,18 @@ static void prune_refs(struct files_ref_store *refs, struct ref_to_prune **refs_
  */
 static int should_pack_ref(const char *refname,
 			   const struct object_id *oid, unsigned int ref_flags,
-			   unsigned int pack_flags)
+			   struct pack_refs_opts *opts)
 {
 	/* Do not pack per-worktree refs: */
 	if (parse_worktree_ref(refname, NULL, NULL, NULL) !=
 	    REF_WORKTREE_SHARED)
 		return 0;

+	if (opts->exclusions && ref_excluded(opts->exclusions, refname))
+		return 0;

Looks good, here is where we throw out refs that we don't want. I wonder
if ref_excluded() does the right thing with a zero-initialized argument
(i.e. that it behaves as if nothing matches).

Yes, I think we can skip checking if opt->exclusions is not null. Junio had
feedback around this as well.

I wonder if it's possible to skip over certain loose references by
avoiding traversal into the sub-directories for simple prefixes. That
may be a premature optimization, though, so I don't think you
necessarily need to worry about it in this round.

quoted

+test_expect_success 'test excluded refs are not packed' '
+	git branch dont_pack1 &&
+	git branch dont_pack2 &&
+	git branch pack_this &&
+	git pack-refs --all --exclude "refs/heads/dont_pack*" &&
+	test -f .git/refs/heads/dont_pack1 &&
+	test -f .git/refs/heads/dont_pack2 &&
+	! test -f ./git/refs/heads/pack_this'
+
+test_expect_success 'test --no-exclude refs clears excluded refs' '
+	git branch dont_pack3 &&
+	git branch dont_pack4 &&
+	git pack-refs --all --exclude "refs/heads/dont_pack*" --no-exclude &&
+	! test -f .git/refs/heads/dont_pack3 &&
+	! test -f .git/refs/heads/dont_pack4'

Tests look good. The trailing quote is a little odd to be placed on the
last line of the test instead of off on its own, but I suppose that is
imitating existing style, which is OK.

thanks for the feedback!
John

Thanks,
Taylor