[PATCH 0/5] drop "struct name_path" and path_name()

[PATCH 0/5] drop "struct name_path" and path_name()

[Jeff King]

The graph traversal code in list-objects.c uses "struct name_path" to
build a linked list of path components, which it then feeds to the
callbacks. This is meant to be efficient, because we keep pointers into
the actual tree data for each name. However, there are two things that
work against this:

  1. In some cases, we keep in parallel a strbuf with the running
     pathname, so that we can feed it to tree_entry_interesting().

  2. The ultimate fate of this linked list is often to get concatenated
     into a single buffer anyway, via path_name().

So it's really not buying us much efficiency over just using
strbuf_addstr() and strbuf_setlen() in the first place. And it's extra
code that is slightly tricky to get right, especially with respect to
size_t and integer overflow (compare path_name() and
show_object_with_path() before and after).

This series drops the whole thing in favor of using a strbuf. Because I
wanted to make sure we weren't regressing performance, I measured two
cases before and after (and of course verified that they produce
identical output):

  1. "git rev-list --objects --all", which prints the name of each tree
     and blob we find directly from the linked list (without ever
     constructing the whole string), and does not use
     tree_entry_interesting (and so does not otherwise need to keep the
     running strbuf). So we'd see any negative effects of the strategy
     here.

  2. "git prune --dry-run", which walks the complete graph but does
     not do anything useful with the pathnames. So it would not
     otherwise need to assemble or look at the path components at all.

Both of them showed no measurable difference in their best-of-five times
when run on git.git. I didn't measure peak memory usage. For the reasons
explained in patch 3, it will actually be slightly _better_ for a normal
repo like git.git. But you could construct a pathological case where it
is worse (e.g., if you had a tree with a 500MB path-name, the old code
would need 500MB to run rev-list, and the new code will need 2*500MB
during the callback). I think the cleanup is worth it.

  [1/5]: http-push: stop using name_path
  [2/5]: show_object_with_name: simplify by using path_name()
  [3/5]: list-objects: convert name_path to a strbuf
  [4/5]: list-objects: drop name_path entirely
  [5/5]: list-objects: pass full pathname to callbacks

-Peff

[PATCH 1/5] http-push: stop using name_path

[Jeff King]

The graph traversal code here passes along a name_path to
build up the pathname at which we find each blob. But we
never actually do anything with the resulting names, making
it a waste of code and memory.

This usage came in aa1dbc9 (Update http-push functionality,
2006-03-07), and originally the result was passed to
"add_object" (which stored it, but didn't really use it,
either). But we stopped using that function in 1f1e895 (Add
"named object array" concept, 2006-06-19) in favor of
storing just the objects themselves.

Moreover, the generation of the name in process_tree() is
buggy. It sticks "name" onto the end of the name_path linked
list, and then passes it down again as it recurses (instead
of "entry.path"). So it's a good thing this was unused, as
the resulting path for "a/b/c/d" would end up as "a/a/a/a".

Signed-off-by: Jeff King <peff@peff.net>
---
 http-push.c | 23 +++++++----------------
 1 file changed, 7 insertions(+), 16 deletions(-)

diff --git a/http-push.c b/http-push.c
index d857b13..bd60668 100644
--- a/http-push.c
+++ b/http-push.c
@@ -1277,9 +1277,7 @@ static struct object_list **add_one_object(struct object *obj, struct object_lis
 }
 
 static struct object_list **process_blob(struct blob *blob,
-					 struct object_list **p,
-					 struct name_path *path,
-					 const char *name)
+					 struct object_list **p)
 {
 	struct object *obj = &blob->object;
 
@@ -1293,14 +1291,11 @@ static struct object_list **process_blob(struct blob *blob,
 }
 
 static struct object_list **process_tree(struct tree *tree,
-					 struct object_list **p,
-					 struct name_path *path,
-					 const char *name)
+					 struct object_list **p)
 {
 	struct object *obj = &tree->object;
 	struct tree_desc desc;
 	struct name_entry entry;
-	struct name_path me;
 
 	obj->flags |= LOCAL;
 
@@ -1310,21 +1305,17 @@ static struct object_list **process_tree(struct tree *tree,
 		die("bad tree object %s", oid_to_hex(&obj->oid));
 
 	obj->flags |= SEEN;
-	name = xstrdup(name);
 	p = add_one_object(obj, p);
-	me.up = path;
-	me.elem = name;
-	me.elem_len = strlen(name);
 
 	init_tree_desc(&desc, tree->buffer, tree->size);
 
 	while (tree_entry(&desc, &entry))
 		switch (object_type(entry.mode)) {
 		case OBJ_TREE:
-			p = process_tree(lookup_tree(entry.sha1), p, &me, name);
+			p = process_tree(lookup_tree(entry.sha1), p);
 			break;
 		case OBJ_BLOB:
-			p = process_blob(lookup_blob(entry.sha1), p, &me, name);
+			p = process_blob(lookup_blob(entry.sha1), p);
 			break;
 		default:
 			/* Subproject commit - not in this repository */
@@ -1343,7 +1334,7 @@ static int get_delta(struct rev_info *revs, struct remote_lock *lock)
 	int count = 0;
 
 	while ((commit = get_revision(revs)) != NULL) {
-		p = process_tree(commit->tree, p, NULL, "");
+		p = process_tree(commit->tree, p);
 		commit->object.flags |= LOCAL;
 		if (!(commit->object.flags & UNINTERESTING))
 			count += add_send_request(&commit->object, lock);
@@ -1362,11 +1353,11 @@ static int get_delta(struct rev_info *revs, struct remote_lock *lock)
 			continue;
 		}
 		if (obj->type == OBJ_TREE) {
-			p = process_tree((struct tree *)obj, p, NULL, name);
+			p = process_tree((struct tree *)obj, p);
 			continue;
 		}
 		if (obj->type == OBJ_BLOB) {
-			p = process_blob((struct blob *)obj, p, NULL, name);
+			p = process_blob((struct blob *)obj, p);
 			continue;
 		}
 		die("unknown pending object %s (%s)", oid_to_hex(&obj->oid), name);
-- 
2.7.1.550.gf5fcbd3

[PATCH 2/5] show_object_with_name: simplify by using path_name()

[Jeff King]

When "git rev-list" shows an object with its associated path
name, it does so by walking the name_path linked list and
printing each component (stopping at any embedded NULs or
newlines).

We'd like to eventually get rid of name_path entirely in
favor of a single buffer, and dropping this custom printing
code is part of that. As a first step, let's use path_name()
to format the list into a single buffer, and print that.
This is strictly less efficient than the original, but it's
a temporary step in the refactoring; our end game will be to
get the fully formatted name in the first place.

Signed-off-by: Jeff King <peff@peff.net>
---
 revision.c | 40 ++++++----------------------------------
 1 file changed, 6 insertions(+), 34 deletions(-)

diff --git a/revision.c b/revision.c
index f24ead5..6387068 100644
--- a/revision.c
+++ b/revision.c
@@ -49,46 +49,18 @@ char *path_name(const struct name_path *path, const char *name)
 	return n;
 }
 
-static int show_path_component_truncated(FILE *out, const char *name, int len)
-{
-	int cnt;
-	for (cnt = 0; cnt < len; cnt++) {
-		int ch = name[cnt];
-		if (!ch || ch == '\n')
-			return -1;
-		fputc(ch, out);
-	}
-	return len;
-}
-
-static int show_path_truncated(FILE *out, const struct name_path *path)
-{
-	int emitted, ours;
-
-	if (!path)
-		return 0;
-	emitted = show_path_truncated(out, path->up);
-	if (emitted < 0)
-		return emitted;
-	if (emitted)
-		fputc('/', out);
-	ours = show_path_component_truncated(out, path->elem, path->elem_len);
-	if (ours < 0)
-		return ours;
-	return ours || emitted;
-}
-
 void show_object_with_name(FILE *out, struct object *obj,
 			   const struct name_path *path, const char *component)
 {
-	struct name_path leaf;
-	leaf.up = (struct name_path *)path;
-	leaf.elem = component;
-	leaf.elem_len = strlen(component);
+	char *name = path_name(path, component);
+	char *p;
 
 	fprintf(out, "%s ", oid_to_hex(&obj->oid));
-	show_path_truncated(out, &leaf);
+	for (p = name; *p && *p != '\n'; p++)
+		fputc(*p, out);
 	fputc('\n', out);
+
+	free(name);
 }
 
 static void mark_blob_uninteresting(struct blob *blob)
-- 
2.7.1.550.gf5fcbd3

[PATCH 3/5] list-objects: convert name_path to a strbuf

[Jeff King]

The "struct name_path" data is examined in only two places:
we generate it in process_tree(), and we convert it to a
single string in path_name(). Everyone else just passes it
through to those functions.

We can further note that process_tree() already keeps a
single strbuf with the leading tree path, for use with
tree_entry_interesting().

Instead of building a separate name_path linked list, let's
just use the one we already build in "base". This reduces
the amount of code (especially tricky code in path_name()
which did not check for integer overflows caused by deep
or large pathnames).

It is also more efficient in some instances.  Any time we
were using tree_entry_interesting, we were building up the
strbuf anyway, so this is an immediate and obvious win
there. In cases where we were not, we trade off storing
"pathname/" in a strbuf on the heap for each level of the
path, instead of two pointers and an int on the stack (with
one pointer into the tree object). On a 64-bit system, the
latter is 20 bytes; so if path components are less than that
on average, this has lower peak memory usage.  In practice
it probably doesn't matter either way; we are already
holding in memory all of the tree objects leading up to each
pathname, and for normal-depth pathnames, we are only
talking about hundreds of bytes.

This patch leaves "struct name_path" as a thin wrapper
around the strbuf, to avoid disrupting callbacks. We should
fix them, but leaving it out makes this diff easier to view.

Signed-off-by: Jeff King <peff@peff.net>
---
 list-objects.c | 22 +++++++++-------------
 revision.c     | 25 +++++--------------------
 revision.h     |  4 +---
 3 files changed, 15 insertions(+), 36 deletions(-)

diff --git a/list-objects.c b/list-objects.c
index 11732d9..4f60a3e 100644
--- a/list-objects.c
+++ b/list-objects.c
@@ -62,7 +62,6 @@ static void process_gitlink(struct rev_info *revs,
 static void process_tree(struct rev_info *revs,
 			 struct tree *tree,
 			 show_object_fn show,
-			 struct name_path *path,
 			 struct strbuf *base,
 			 const char *name,
 			 void *cb_data)
@@ -86,17 +85,14 @@ static void process_tree(struct rev_info *revs,
 			return;
 		die("bad tree object %s", oid_to_hex(&obj->oid));
 	}
+
 	obj->flags |= SEEN;
-	show(obj, path, name, cb_data);
-	me.up = path;
-	me.elem = name;
-	me.elem_len = strlen(name);
-
-	if (!match) {
-		strbuf_addstr(base, name);
-		if (base->len)
-			strbuf_addch(base, '/');
-	}
+	me.base = base;
+	show(obj, &me, name, cb_data);
+
+	strbuf_addstr(base, name);
+	if (base->len)
+		strbuf_addch(base, '/');
 
 	init_tree_desc(&desc, tree->buffer, tree->size);
 
@@ -113,7 +109,7 @@ static void process_tree(struct rev_info *revs,
 		if (S_ISDIR(entry.mode))
 			process_tree(revs,
 				     lookup_tree(entry.sha1),
-				     show, &me, base, entry.path,
+				     show, base, entry.path,
 				     cb_data);
 		else if (S_ISGITLINK(entry.mode))
 			process_gitlink(revs, entry.sha1,
@@ -220,7 +216,7 @@ void traverse_commit_list(struct rev_info *revs,
 			path = "";
 		if (obj->type == OBJ_TREE) {
 			process_tree(revs, (struct tree *)obj, show_object,
-				     NULL, &base, path, data);
+				     &base, path, data);
 			continue;
 		}
 		if (obj->type == OBJ_BLOB) {
diff --git a/revision.c b/revision.c
index 6387068..8dd0950 100644
--- a/revision.c
+++ b/revision.c
@@ -27,26 +27,11 @@ static const char *term_good;
 
 char *path_name(const struct name_path *path, const char *name)
 {
-	const struct name_path *p;
-	char *n, *m;
-	int nlen = strlen(name);
-	int len = nlen + 1;
-
-	for (p = path; p; p = p->up) {
-		if (p->elem_len)
-			len += p->elem_len + 1;
-	}
-	n = xmalloc(len);
-	m = n + len - (nlen + 1);
-	memcpy(m, name, nlen + 1);
-	for (p = path; p; p = p->up) {
-		if (p->elem_len) {
-			m -= p->elem_len + 1;
-			memcpy(m, p->elem, p->elem_len);
-			m[p->elem_len] = '/';
-		}
-	}
-	return n;
+	struct strbuf ret = STRBUF_INIT;
+	if (path)
+		strbuf_addbuf(&ret, path->base);
+	strbuf_addstr(&ret, name);
+	return strbuf_detach(&ret, NULL);
 }
 
 void show_object_with_name(FILE *out, struct object *obj,
diff --git a/revision.h b/revision.h
index 23857c0..2a26310 100644
--- a/revision.h
+++ b/revision.h
@@ -258,9 +258,7 @@ extern void mark_parents_uninteresting(struct commit *commit);
 extern void mark_tree_uninteresting(struct tree *tree);
 
 struct name_path {
-	struct name_path *up;
-	int elem_len;
-	const char *elem;
+	struct strbuf *base;
 };
 
 char *path_name(const struct name_path *path, const char *name);
-- 
2.7.1.550.gf5fcbd3

[PATCH 4/5] list-objects: drop name_path entirely

[Jeff King]

In the previous commit, we left name_path as a thin wrapper
around a strbuf. This patch drops it entirely. As a result,
every show_object_fn callback needs to be adjusted. However,
none of their code needs to be changed at all, because the
only use was to pass it to path_name(), which now handles
the bare strbuf.

Signed-off-by: Jeff King <peff@peff.net>
---
 builtin/pack-objects.c |  4 ++--
 builtin/rev-list.c     |  4 ++--
 list-objects.c         | 12 +++++-------
 list-objects.h         |  2 +-
 pack-bitmap-write.c    |  2 +-
 pack-bitmap.c          |  4 ++--
 reachable.c            |  2 +-
 revision.c             |  6 +++---
 revision.h             |  8 ++------
 9 files changed, 19 insertions(+), 25 deletions(-)

diff --git a/builtin/pack-objects.c b/builtin/pack-objects.c
index 4dae5b1..8bbb9bd 100644
--- a/builtin/pack-objects.c
+++ b/builtin/pack-objects.c
@@ -2285,7 +2285,7 @@ static void show_commit(struct commit *commit, void *data)
 }
 
 static void show_object(struct object *obj,
-			const struct name_path *path, const char *last,
+			struct strbuf *path, const char *last,
 			void *data)
 {
 	char *name = path_name(path, last);
@@ -2480,7 +2480,7 @@ static int get_object_list_from_bitmap(struct rev_info *revs)
 }
 
 static void record_recent_object(struct object *obj,
-				 const struct name_path *path,
+				 struct strbuf *path,
 				 const char *last,
 				 void *data)
 {
diff --git a/builtin/rev-list.c b/builtin/rev-list.c
index 3aa89a1..a92c3ca 100644
--- a/builtin/rev-list.c
+++ b/builtin/rev-list.c
@@ -178,7 +178,7 @@ static void finish_commit(struct commit *commit, void *data)
 }
 
 static void finish_object(struct object *obj,
-			  const struct name_path *path, const char *name,
+			  struct strbuf *path, const char *name,
 			  void *cb_data)
 {
 	struct rev_list_info *info = cb_data;
@@ -189,7 +189,7 @@ static void finish_object(struct object *obj,
 }
 
 static void show_object(struct object *obj,
-			const struct name_path *path, const char *component,
+			struct strbuf *path, const char *component,
 			void *cb_data)
 {
 	struct rev_list_info *info = cb_data;
diff --git a/list-objects.c b/list-objects.c
index 4f60a3e..4397766 100644
--- a/list-objects.c
+++ b/list-objects.c
@@ -11,7 +11,7 @@
 static void process_blob(struct rev_info *revs,
 			 struct blob *blob,
 			 show_object_fn show,
-			 struct name_path *path,
+			 struct strbuf *path,
 			 const char *name,
 			 void *cb_data)
 {
@@ -52,7 +52,7 @@ static void process_blob(struct rev_info *revs,
 static void process_gitlink(struct rev_info *revs,
 			    const unsigned char *sha1,
 			    show_object_fn show,
-			    struct name_path *path,
+			    struct strbuf *path,
 			    const char *name,
 			    void *cb_data)
 {
@@ -69,7 +69,6 @@ static void process_tree(struct rev_info *revs,
 	struct object *obj = &tree->object;
 	struct tree_desc desc;
 	struct name_entry entry;
-	struct name_path me;
 	enum interesting match = revs->diffopt.pathspec.nr == 0 ?
 		all_entries_interesting: entry_not_interesting;
 	int baselen = base->len;
@@ -87,8 +86,7 @@ static void process_tree(struct rev_info *revs,
 	}
 
 	obj->flags |= SEEN;
-	me.base = base;
-	show(obj, &me, name, cb_data);
+	show(obj, base, name, cb_data);
 
 	strbuf_addstr(base, name);
 	if (base->len)
@@ -113,12 +111,12 @@ static void process_tree(struct rev_info *revs,
 				     cb_data);
 		else if (S_ISGITLINK(entry.mode))
 			process_gitlink(revs, entry.sha1,
-					show, &me, entry.path,
+					show, base, entry.path,
 					cb_data);
 		else
 			process_blob(revs,
 				     lookup_blob(entry.sha1),
-				     show, &me, entry.path,
+				     show, base, entry.path,
 				     cb_data);
 	}
 	strbuf_setlen(base, baselen);
diff --git a/list-objects.h b/list-objects.h
index 136a1da..69c4c7d 100644
--- a/list-objects.h
+++ b/list-objects.h
@@ -2,7 +2,7 @@
 #define LIST_OBJECTS_H
 
 typedef void (*show_commit_fn)(struct commit *, void *);
-typedef void (*show_object_fn)(struct object *, const struct name_path *, const char *, void *);
+typedef void (*show_object_fn)(struct object *, struct strbuf *, const char *, void *);
 void traverse_commit_list(struct rev_info *, show_commit_fn, show_object_fn, void *);
 
 typedef void (*show_edge_fn)(struct commit *);
diff --git a/pack-bitmap-write.c b/pack-bitmap-write.c
index 6bff970..65ed342 100644
--- a/pack-bitmap-write.c
+++ b/pack-bitmap-write.c
@@ -148,7 +148,7 @@ static uint32_t find_object_pos(const unsigned char *sha1)
 	return entry->in_pack_pos;
 }
 
-static void show_object(struct object *object, const struct name_path *path,
+static void show_object(struct object *object, struct strbuf *path,
 			const char *last, void *data)
 {
 	struct bitmap *base = data;
diff --git a/pack-bitmap.c b/pack-bitmap.c
index dd8dc16..51f790e 100644
--- a/pack-bitmap.c
+++ b/pack-bitmap.c
@@ -414,7 +414,7 @@ static int ext_index_add_object(struct object *object, const char *name)
 	return bitmap_pos + bitmap_git.pack->num_objects;
 }
 
-static void show_object(struct object *object, const struct name_path *path,
+static void show_object(struct object *object, struct strbuf *path,
 			const char *last, void *data)
 {
 	struct bitmap *base = data;
@@ -895,7 +895,7 @@ struct bitmap_test_data {
 };
 
 static void test_show_object(struct object *object,
-			     const struct name_path *path,
+			     struct strbuf *path,
 			     const char *last, void *data)
 {
 	struct bitmap_test_data *tdata = data;
diff --git a/reachable.c b/reachable.c
index 43616d4..e60f08d 100644
--- a/reachable.c
+++ b/reachable.c
@@ -43,7 +43,7 @@ static int add_one_ref(const char *path, const struct object_id *oid,
  * The traversal will have already marked us as SEEN, so we
  * only need to handle any progress reporting here.
  */
-static void mark_object(struct object *obj, const struct name_path *path,
+static void mark_object(struct object *obj, struct strbuf *path,
 			const char *name, void *data)
 {
 	update_progress(data);
diff --git a/revision.c b/revision.c
index 8dd0950..3c84781 100644
--- a/revision.c
+++ b/revision.c
@@ -25,17 +25,17 @@ volatile show_early_output_fn_t show_early_output;
 static const char *term_bad;
 static const char *term_good;
 
-char *path_name(const struct name_path *path, const char *name)
+char *path_name(struct strbuf *path, const char *name)
 {
 	struct strbuf ret = STRBUF_INIT;
 	if (path)
-		strbuf_addbuf(&ret, path->base);
+		strbuf_addbuf(&ret, path);
 	strbuf_addstr(&ret, name);
 	return strbuf_detach(&ret, NULL);
 }
 
 void show_object_with_name(FILE *out, struct object *obj,
-			   const struct name_path *path, const char *component)
+			   struct strbuf *path, const char *component)
 {
 	char *name = path_name(path, component);
 	char *p;
diff --git a/revision.h b/revision.h
index 2a26310..7beab15 100644
--- a/revision.h
+++ b/revision.h
@@ -257,14 +257,10 @@ extern void put_revision_mark(const struct rev_info *revs,
 extern void mark_parents_uninteresting(struct commit *commit);
 extern void mark_tree_uninteresting(struct tree *tree);
 
-struct name_path {
-	struct strbuf *base;
-};
-
-char *path_name(const struct name_path *path, const char *name);
+char *path_name(struct strbuf *path, const char *name);
 
 extern void show_object_with_name(FILE *, struct object *,
-				  const struct name_path *, const char *);
+				  struct strbuf *, const char *);
 
 extern void add_pending_object(struct rev_info *revs,
 			       struct object *obj, const char *name);
-- 
2.7.1.550.gf5fcbd3

[PATCH 5/5] list-objects: pass full pathname to callbacks

[Jeff King]

When we find a blob at "a/b/c", we currently pass this to
our show_object_fn callbacks as two components: "a/b/" and
"c". Callbacks which want the full value then call
path_name(), which concatenates the two. But this is an
inefficient interface; the path is a strbuf, and we could
simply append "c" to it temporarily, then roll back the
length, without creating a new copy.

So we could improve this by teaching the callsites of
path_name() this trick (and there are only 3). But we can
also notice that no callback actually cares about the
broken-down representation, and simply pass each callback
the full path "a/b/c" as a string. The callback code becomes
even simpler, then, as we do not have to worry about freeing
an allocated buffer, nor rolling back our modification to
the strbuf.

This is theoretically less efficient, as some callbacks
would not bother to format the final path component. But in
practice this is not measurable. Since we use the same
strbuf over and over, our work to grow it is amortized, and
we really only pay to memcpy a few bytes.

Signed-off-by: Jeff King <peff@peff.net>
---
 builtin/pack-objects.c | 15 ++-------------
 builtin/rev-list.c     | 12 ++++--------
 list-objects.c         | 14 +++++++++-----
 list-objects.h         |  2 +-
 pack-bitmap-write.c    |  3 +--
 pack-bitmap.c          | 13 ++++---------
 reachable.c            |  5 ++---
 revision.c             | 17 ++---------------
 revision.h             |  3 +--
 9 files changed, 26 insertions(+), 58 deletions(-)

diff --git a/builtin/pack-objects.c b/builtin/pack-objects.c
index 8bbb9bd..a6609f1 100644
--- a/builtin/pack-objects.c
+++ b/builtin/pack-objects.c
@@ -2284,21 +2284,11 @@ static void show_commit(struct commit *commit, void *data)
 		index_commit_for_bitmap(commit);
 }
 
-static void show_object(struct object *obj,
-			struct strbuf *path, const char *last,
-			void *data)
+static void show_object(struct object *obj, const char *name, void *data)
 {
-	char *name = path_name(path, last);
-
 	add_preferred_base_object(name);
 	add_object_entry(obj->oid.hash, obj->type, name, 0);
 	obj->flags |= OBJECT_ADDED;
-
-	/*
-	 * We will have generated the hash from the name,
-	 * but not saved a pointer to it - we can free it
-	 */
-	free((char *)name);
 }
 
 static void show_edge(struct commit *commit)
@@ -2480,8 +2470,7 @@ static int get_object_list_from_bitmap(struct rev_info *revs)
 }
 
 static void record_recent_object(struct object *obj,
-				 struct strbuf *path,
-				 const char *last,
+				 const char *name,
 				 void *data)
 {
 	sha1_array_append(&recent_objects, obj->oid.hash);
diff --git a/builtin/rev-list.c b/builtin/rev-list.c
index a92c3ca..275da0d 100644
--- a/builtin/rev-list.c
+++ b/builtin/rev-list.c
@@ -177,9 +177,7 @@ static void finish_commit(struct commit *commit, void *data)
 	free_commit_buffer(commit);
 }
 
-static void finish_object(struct object *obj,
-			  struct strbuf *path, const char *name,
-			  void *cb_data)
+static void finish_object(struct object *obj, const char *name, void *cb_data)
 {
 	struct rev_list_info *info = cb_data;
 	if (obj->type == OBJ_BLOB && !has_object_file(&obj->oid))
@@ -188,15 +186,13 @@ static void finish_object(struct object *obj,
 		parse_object(obj->oid.hash);
 }
 
-static void show_object(struct object *obj,
-			struct strbuf *path, const char *component,
-			void *cb_data)
+static void show_object(struct object *obj, const char *name, void *cb_data)
 {
 	struct rev_list_info *info = cb_data;
-	finish_object(obj, path, component, cb_data);
+	finish_object(obj, name, cb_data);
 	if (info->flags & REV_LIST_QUIET)
 		return;
-	show_object_with_name(stdout, obj, path, component);
+	show_object_with_name(stdout, obj, name);
 }
 
 static void show_edge(struct commit *commit)
diff --git a/list-objects.c b/list-objects.c
index 4397766..917cc5d 100644
--- a/list-objects.c
+++ b/list-objects.c
@@ -16,6 +16,7 @@ static void process_blob(struct rev_info *revs,
 			 void *cb_data)
 {
 	struct object *obj = &blob->object;
+	size_t pathlen;
 
 	if (!revs->blob_objects)
 		return;
@@ -24,7 +25,11 @@ static void process_blob(struct rev_info *revs,
 	if (obj->flags & (UNINTERESTING | SEEN))
 		return;
 	obj->flags |= SEEN;
-	show(obj, path, name, cb_data);
+
+	pathlen = path->len;
+	strbuf_addstr(path, name);
+	show(obj, path->buf, cb_data);
+	strbuf_setlen(path, pathlen);
 }
 
 /*
@@ -86,9 +91,8 @@ static void process_tree(struct rev_info *revs,
 	}
 
 	obj->flags |= SEEN;
-	show(obj, base, name, cb_data);
-
 	strbuf_addstr(base, name);
+	show(obj, base->buf, cb_data);
 	if (base->len)
 		strbuf_addch(base, '/');
 
@@ -207,7 +211,7 @@ void traverse_commit_list(struct rev_info *revs,
 			continue;
 		if (obj->type == OBJ_TAG) {
 			obj->flags |= SEEN;
-			show_object(obj, NULL, name, data);
+			show_object(obj, name, data);
 			continue;
 		}
 		if (!path)
@@ -219,7 +223,7 @@ void traverse_commit_list(struct rev_info *revs,
 		}
 		if (obj->type == OBJ_BLOB) {
 			process_blob(revs, (struct blob *)obj, show_object,
-				     NULL, path, data);
+				     &base, path, data);
 			continue;
 		}
 		die("unknown pending object %s (%s)",
diff --git a/list-objects.h b/list-objects.h
index 69c4c7d..0cebf85 100644
--- a/list-objects.h
+++ b/list-objects.h
@@ -2,7 +2,7 @@
 #define LIST_OBJECTS_H
 
 typedef void (*show_commit_fn)(struct commit *, void *);
-typedef void (*show_object_fn)(struct object *, struct strbuf *, const char *, void *);
+typedef void (*show_object_fn)(struct object *, const char *, void *);
 void traverse_commit_list(struct rev_info *, show_commit_fn, show_object_fn, void *);
 
 typedef void (*show_edge_fn)(struct commit *);
diff --git a/pack-bitmap-write.c b/pack-bitmap-write.c
index 65ed342..c30bcd0 100644
--- a/pack-bitmap-write.c
+++ b/pack-bitmap-write.c
@@ -148,8 +148,7 @@ static uint32_t find_object_pos(const unsigned char *sha1)
 	return entry->in_pack_pos;
 }
 
-static void show_object(struct object *object, struct strbuf *path,
-			const char *last, void *data)
+static void show_object(struct object *object, const char *name, void *data)
 {
 	struct bitmap *base = data;
 	bitmap_set(base, find_object_pos(object->oid.hash));
diff --git a/pack-bitmap.c b/pack-bitmap.c
index 51f790e..b949e51 100644
--- a/pack-bitmap.c
+++ b/pack-bitmap.c
@@ -414,19 +414,15 @@ static int ext_index_add_object(struct object *object, const char *name)
 	return bitmap_pos + bitmap_git.pack->num_objects;
 }
 
-static void show_object(struct object *object, struct strbuf *path,
-			const char *last, void *data)
+static void show_object(struct object *object, const char *name, void *data)
 {
 	struct bitmap *base = data;
 	int bitmap_pos;
 
 	bitmap_pos = bitmap_position(object->oid.hash);
 
-	if (bitmap_pos < 0) {
-		char *name = path_name(path, last);
+	if (bitmap_pos < 0)
 		bitmap_pos = ext_index_add_object(object, name);
-		free(name);
-	}
 
 	bitmap_set(base, bitmap_pos);
 }
@@ -894,9 +890,8 @@ struct bitmap_test_data {
 	size_t seen;
 };
 
-static void test_show_object(struct object *object,
-			     struct strbuf *path,
-			     const char *last, void *data)
+static void test_show_object(struct object *object, const char *name,
+			     void *data)
 {
 	struct bitmap_test_data *tdata = data;
 	int bitmap_pos;
diff --git a/reachable.c b/reachable.c
index e60f08d..ed35201 100644
--- a/reachable.c
+++ b/reachable.c
@@ -43,15 +43,14 @@ static int add_one_ref(const char *path, const struct object_id *oid,
  * The traversal will have already marked us as SEEN, so we
  * only need to handle any progress reporting here.
  */
-static void mark_object(struct object *obj, struct strbuf *path,
-			const char *name, void *data)
+static void mark_object(struct object *obj, const char *name, void *data)
 {
 	update_progress(data);
 }
 
 static void mark_commit(struct commit *c, void *data)
 {
-	mark_object(&c->object, NULL, NULL, data);
+	mark_object(&c->object, NULL, data);
 }
 
 struct recent_data {
diff --git a/revision.c b/revision.c
index 3c84781..82f3ca4 100644
--- a/revision.c
+++ b/revision.c
@@ -25,27 +25,14 @@ volatile show_early_output_fn_t show_early_output;
 static const char *term_bad;
 static const char *term_good;
 
-char *path_name(struct strbuf *path, const char *name)
+void show_object_with_name(FILE *out, struct object *obj, const char *name)
 {
-	struct strbuf ret = STRBUF_INIT;
-	if (path)
-		strbuf_addbuf(&ret, path);
-	strbuf_addstr(&ret, name);
-	return strbuf_detach(&ret, NULL);
-}
-
-void show_object_with_name(FILE *out, struct object *obj,
-			   struct strbuf *path, const char *component)
-{
-	char *name = path_name(path, component);
-	char *p;
+	const char *p;
 
 	fprintf(out, "%s ", oid_to_hex(&obj->oid));
 	for (p = name; *p && *p != '\n'; p++)
 		fputc(*p, out);
 	fputc('\n', out);
-
-	free(name);
 }
 
 static void mark_blob_uninteresting(struct blob *blob)
diff --git a/revision.h b/revision.h
index 7beab15..dca0d38 100644
--- a/revision.h
+++ b/revision.h
@@ -259,8 +259,7 @@ extern void mark_tree_uninteresting(struct tree *tree);
 
 char *path_name(struct strbuf *path, const char *name);
 
-extern void show_object_with_name(FILE *, struct object *,
-				  struct strbuf *, const char *);
+extern void show_object_with_name(FILE *, struct object *, const char *);
 
 extern void add_pending_object(struct rev_info *revs,
 			       struct object *obj, const char *name);
-- 
2.7.1.550.gf5fcbd3

Re: [PATCH 5/5] list-objects: pass full pathname to callbacks

[Jeff King]

On Thu, Feb 11, 2016 at 05:28:36PM -0500, Jeff King wrote:

> +void show_object_with_name(FILE *out, struct object *obj, const char *name)
> [...]
>  	fprintf(out, "%s ", oid_to_hex(&obj->oid));
>  	for (p = name; *p && *p != '\n'; p++)
>  		fputc(*p, out);
>  	fputc('\n', out);

By the way, since I was timing things, I wondered if we would see any
improvement from using putc_unlocked, like:

diff --git a/revision.c b/revision.c
index 82f3ca4..ab72247 100644
--- a/revision.c
+++ b/revision.c
@@ -30,9 +30,11 @@ void show_object_with_name(FILE *out, struct object *obj, const char *name)
 	const char *p;
 
 	fprintf(out, "%s ", oid_to_hex(&obj->oid));
+	flockfile(out);
 	for (p = name; *p && *p != '\n'; p++)
-		fputc(*p, out);
-	fputc('\n', out);
+		putc_unlocked(*p, out);
+	putc_unlocked('\n', out);
+	funlockfile(out);
 }
 
 static void mark_blob_uninteresting(struct blob *blob)

But I couldn't measure any speedup. I imagine if you had 500MB pathnames
you might see some improvement, but I don't think it is even worth the
extra lines of code to worry about such a pathological case.

-Peff