git.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: "Nguyễn Thái Ngọc Duy" <pclouds@gmail.com>
To: git@vger.kernel.org
Cc: "Junio C Hamano" <gitster@pobox.com>,
	"Nguyễn Thái Ngọc Duy" <pclouds@gmail.com>
Subject: [PATCH v2] checkout: avoid unnecessary match_pathspec calls
Date: Sun, 24 Mar 2013 19:55:25 +0700	[thread overview]
Message-ID: <1364129725-29597-1-git-send-email-pclouds@gmail.com> (raw)
In-Reply-To: <7vr4j52t0z.fsf@alter.siamese.dyndns.org>

In checkout_paths() we do this

 - for all updated items, call match_pathspec
 - for all items, call match_pathspec (inside unmerge_cache)
 - for all items, call match_pathspec (for showing "path .. is unmerged)
 - for updated items, call match_pathspec and update paths

That's a lot of duplicate match_pathspec(s) and the function is not
exactly cheap to be called so many times, especially on large indexes.
This patch makes it call match_pathspec once per updated index entry,
save the result in ce_flags and reuse the results in the following
loops.

The changes in 0a1283b (checkout $tree $path: do not clobber local
changes in $path not in $tree - 2011-09-30) limit the affected paths
to ones we read from $tree. We do not do anything to other modified
entries in this case, so the "for all items" above could be modified
to "for all updated items". But..

The command's behavior now is modified slightly: unmerged entries that
match $path, but not updated by $tree, are now NOT touched.  Although
this should be considered a bug fix, not a regression.

And while at there, free ps_matched after use.

The following command is tested on webkit, 215k entries. The pattern
is chosen mainly to make match_pathspec sweat:

git checkout -- "*[a-zA-Z]*[a-zA-Z]*[a-zA-Z]*"

        before      after
real    0m3.493s    0m2.737s
user    0m2.239s    0m1.586s
sys     0m1.252s    0m1.151s

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 > and worry about larger ones later, so if there were another choice,
 > i.e.
 >
 >  - eject nd/magic-pathspecs for now, cook this (and other small
 >    independent and clear improvements you may come up with, some of
 >    which might come out of nd/magic-pathspecs itself) and let it
 >    graduate first, and later revisit rerolld nd/magic-pathspecs
 >
 > that would be the ideal among the given choices ;-).

 Whichever is easier for you.

 > The above is a faithful rewrite, but I have to wonder why you need
 > two separate loops.
 >
 > Do you understand what the original loop is doing with ps_matched,
 > and why the code excludes certain paths while doing so?

 Nope, I did not dig that deep. I expected you to do it ;-) j/k

 > After commenting on the above, it makes me wonder if we even need to
 > bother marking entries that were in the index that did not come from
 > the tree-ish we are checking paths out of, though.  What breaks if
 > you did not do the rewrite above and dropped the second loop in your
 > patch?

 The test suite says none. There is a behavior change regarding
 unmerged entries as mentioned in the commit message. But I think it's
 a good change.

 builtin/checkout.c | 34 +++++++++++++++++++++++++++-------
 cache.h            |  1 +
 resolve-undo.c     | 19 ++++++++++++++++++-
 resolve-undo.h     |  1 +
 4 files changed, 47 insertions(+), 8 deletions(-)

diff --git a/builtin/checkout.c b/builtin/checkout.c
index a9c1b5a..359b983 100644
--- a/builtin/checkout.c
+++ b/builtin/checkout.c
@@ -271,24 +271,46 @@ static int checkout_paths(const struct checkout_opts *opts,
 		;
 	ps_matched = xcalloc(1, pos);
 
+	/*
+	 * Make sure all pathspecs participated in locating the paths
+	 * to be checked out.
+	 */
 	for (pos = 0; pos < active_nr; pos++) {
 		struct cache_entry *ce = active_cache[pos];
+		ce->ce_flags &= ~CE_MATCHED;
 		if (opts->source_tree && !(ce->ce_flags & CE_UPDATE))
+			/*
+			 * "git checkout tree-ish -- path", but this entry
+			 * is in the original index; it will not be checked
+			 * out to the working tree and it does not matter
+			 * if pathspec matched this entry.  We will not do
+			 * anything to this entry at all.
+			 */
 			continue;
-		match_pathspec(opts->pathspec, ce->name, ce_namelen(ce), 0, ps_matched);
+		/*
+		 * Either this entry came from the tree-ish we are
+		 * checking the paths out of, or we are checking out
+		 * of the index.
+		 */
+		if (match_pathspec(opts->pathspec, ce->name, ce_namelen(ce),
+				   0, ps_matched))
+			ce->ce_flags |= CE_MATCHED;
 	}
 
-	if (report_path_error(ps_matched, opts->pathspec, opts->prefix))
+	if (report_path_error(ps_matched, opts->pathspec, opts->prefix)) {
+		free(ps_matched);
 		return 1;
+	}
+	free(ps_matched);
 
 	/* "checkout -m path" to recreate conflicted state */
 	if (opts->merge)
-		unmerge_cache(opts->pathspec);
+		unmerge_marked_index(&the_index);
 
 	/* Any unmerged paths? */
 	for (pos = 0; pos < active_nr; pos++) {
 		struct cache_entry *ce = active_cache[pos];
-		if (match_pathspec(opts->pathspec, ce->name, ce_namelen(ce), 0, NULL)) {
+		if (ce->ce_flags & CE_MATCHED) {
 			if (!ce_stage(ce))
 				continue;
 			if (opts->force) {
@@ -313,9 +335,7 @@ static int checkout_paths(const struct checkout_opts *opts,
 	state.refresh_cache = 1;
 	for (pos = 0; pos < active_nr; pos++) {
 		struct cache_entry *ce = active_cache[pos];
-		if (opts->source_tree && !(ce->ce_flags & CE_UPDATE))
-			continue;
-		if (match_pathspec(opts->pathspec, ce->name, ce_namelen(ce), 0, NULL)) {
+		if (ce->ce_flags & CE_MATCHED) {
 			if (!ce_stage(ce)) {
 				errs |= checkout_entry(ce, &state, NULL);
 				continue;
diff --git a/cache.h b/cache.h
index c56315c..04e6090 100644
--- a/cache.h
+++ b/cache.h
@@ -161,6 +161,7 @@ struct cache_entry {
 
 #define CE_UNPACKED          (1 << 24)
 #define CE_NEW_SKIP_WORKTREE (1 << 25)
+#define CE_MATCHED           (1 << 26)
 
 /*
  * Extended on-disk flags
diff --git a/resolve-undo.c b/resolve-undo.c
index 72b4612..639eb9c 100644
--- a/resolve-undo.c
+++ b/resolve-undo.c
@@ -118,7 +118,7 @@ int unmerge_index_entry_at(struct index_state *istate, int pos)
 	struct cache_entry *ce;
 	struct string_list_item *item;
 	struct resolve_undo_info *ru;
-	int i, err = 0;
+	int i, err = 0, matched;
 
 	if (!istate->resolve_undo)
 		return pos;
@@ -137,6 +137,7 @@ int unmerge_index_entry_at(struct index_state *istate, int pos)
 	ru = item->util;
 	if (!ru)
 		return pos;
+	matched = ce->ce_flags & CE_MATCHED;
 	remove_index_entry_at(istate, pos);
 	for (i = 0; i < 3; i++) {
 		struct cache_entry *nce;
@@ -144,6 +145,8 @@ int unmerge_index_entry_at(struct index_state *istate, int pos)
 			continue;
 		nce = make_cache_entry(ru->mode[i], ru->sha1[i],
 				       ce->name, i + 1, 0);
+		if (matched)
+			nce->ce_flags |= CE_MATCHED;
 		if (add_index_entry(istate, nce, ADD_CACHE_OK_TO_ADD)) {
 			err = 1;
 			error("cannot unmerge '%s'", ce->name);
@@ -156,6 +159,20 @@ int unmerge_index_entry_at(struct index_state *istate, int pos)
 	return unmerge_index_entry_at(istate, pos);
 }
 
+void unmerge_marked_index(struct index_state *istate)
+{
+	int i;
+
+	if (!istate->resolve_undo)
+		return;
+
+	for (i = 0; i < istate->cache_nr; i++) {
+		struct cache_entry *ce = istate->cache[i];
+		if (ce->ce_flags & CE_MATCHED)
+			i = unmerge_index_entry_at(istate, i);
+	}
+}
+
 void unmerge_index(struct index_state *istate, const char **pathspec)
 {
 	int i;
diff --git a/resolve-undo.h b/resolve-undo.h
index 8458769..7a30206 100644
--- a/resolve-undo.h
+++ b/resolve-undo.h
@@ -12,5 +12,6 @@ extern struct string_list *resolve_undo_read(const char *, unsigned long);
 extern void resolve_undo_clear_index(struct index_state *);
 extern int unmerge_index_entry_at(struct index_state *, int);
 extern void unmerge_index(struct index_state *, const char **);
+extern void unmerge_marked_index(struct index_state *);
 
 #endif
-- 
1.8.2.83.gc99314b

  reply	other threads:[~2013-03-24 12:55 UTC|newest]

Thread overview: 7+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2013-03-23 10:55 [PATCH] checkout: avoid unncessary match_pathspec calls Nguyễn Thái Ngọc Duy
2013-03-24  2:45 ` Eric Sunshine
2013-03-24  6:47 ` Junio C Hamano
2013-03-24 12:55   ` Nguyễn Thái Ngọc Duy [this message]
2013-03-25 16:26     ` [PATCH v2] checkout: avoid unnecessary " Junio C Hamano
2013-03-27  5:58       ` [PATCH v3] " Nguyễn Thái Ngọc Duy
2013-03-28 22:32         ` Junio C Hamano

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=1364129725-29597-1-git-send-email-pclouds@gmail.com \
    --to=pclouds@gmail.com \
    --cc=git@vger.kernel.org \
    --cc=gitster@pobox.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).