[RFC] Add --index to git-commit: just commit current index

[RFC] Add --index to git-commit: just commit current index

[Alex Riesen]

[-- Attachment #1: Type: text/plain, Size: 579 bytes --]

Refreshing index takes a long time on big repositories with many files,
especially if the developer was unlucky enough to stick to a slow filesystem
or a broken OS. In this situation explicit git-update-index with
git-commit --index will speedup the workflow.
Giving either --all, -o, or -i silently turns --index off (these have to
refresh index).
In case of unmodified index no status message is shown, for all
the same reasons: it takes too long.
---

First use of new --quiet :)

 git-commit.sh |   21 +++++++++++++++++----
 1 files changed, 17 insertions(+), 4 deletions(-)

[-- Attachment #2: 0001-Add-index-to-git-commit-just-commit-current-index.txt --]
[-- Type: text/plain, Size: 2319 bytes --]

From 49b846a654d2f1ef540d6438356bffeafa97ba62 Mon Sep 17 00:00:00 2001
From: Alex Riesen <raa.lkml@gmail.com>
Date: Thu, 15 Mar 2007 10:39:36 +0100
Subject: [PATCH] Add --index to git-commit: just commit current index

Refreshing index takes a long time on big repositories with many files,
especially if the developer is unlucky enough to stick to a slow filesystem
or a broken OS. In this situation explicit git-update-index with
git-commit --index will speedup the workflow.
Giving either --all, --only, or --include silently turns --index off.
---
 git-commit.sh |   21 +++++++++++++++++----
 1 files changed, 17 insertions(+), 4 deletions(-)

diff --git a/git-commit.sh b/git-commit.sh
index 3656d60..56543c9 100755
--- a/git-commit.sh
+++ b/git-commit.sh
@@ -3,7 +3,7 @@
 # Copyright (c) 2005 Linus Torvalds
 # Copyright (c) 2006 Junio C Hamano
 
-USAGE='[-a | --interactive] [-s] [-v] [--no-verify] [-m <message> | -F <logfile> | (-C|-c) <commit> | --amend] [-u] [-e] [--author <author>] [[-i | -o] <path>...]'
+USAGE='[-a | --interactive] [-s] [-v] [--no-verify] [-m <message> | -F <logfile> | (-C|-c) <commit> | --amend] [-u] [-e] [--author <author>] [--index] [[-i | -o] <path>...]'
 SUBDIRECTORY_OK=Yes
 . git-sh-setup
 require_work_tree
@@ -87,6 +87,7 @@ signoff=
 force_author=
 only_include_assumed=
 untracked_files=
+index_only=
 while case "$#" in 0) break;; esac
 do
 	case "$1" in
@@ -262,6 +263,10 @@ $1"
 		untracked_files=t
 		shift
 		;;
+	--index)
+		index_only=t
+		shift
+		;;
 	--)
 		shift
 		break
@@ -275,6 +280,7 @@ $1"
 	esac
 done
 case "$edit_flag" in t) no_edit= ;; esac
+case "$all$also$only" in t*) index_only= ;; esac
 
 ################################################################
 # Sanity check options
@@ -404,15 +410,22 @@ else
 	USE_INDEX="$THIS_INDEX"
 fi
 
-case "$status_only" in
-t)
+case "$status_only,$index_only" in
+t,*)
 	# This will silently fail in a read-only repository, which is
 	# what we want.
 	GIT_INDEX_FILE="$USE_INDEX" git-update-index -q --unmerged --refresh
 	run_status
 	exit $?
 	;;
-'')
+,t)
+	GIT_INDEX_FILE="$USE_INDEX" \
+	git-diff-index --cached --quiet --exit-code HEAD && {
+		echo >&2 "No changes in the index"
+		exit 1
+	}
+	;;
+,)
 	GIT_INDEX_FILE="$USE_INDEX" git-update-index -q --refresh || exit
 	;;
 esac
-- 
1.5.0.4.439.gce5f

Re: [RFC] Add --index to git-commit: just commit current index

[Junio C Hamano]

"Alex Riesen" <raa.lkml@gmail.com> writes:

> First use of new --quiet :)

You do not need to say --exit-code if you use --quiet.

> Refreshing index takes a long time on big repositories with
> many files, especially if the developer was unlucky enough to
> stick to a slow filesystem or a broken OS. In this situation
> explicit git-update-index with git-commit --index will speedup
> the workflow.

Does it?

A typical workflow would go something like this:

	- repeat from here
	- "edit foo"
        - "edit bar"
        - "git diff" to help me see what I changed
        - "git add foo" as the change is sane
        - test and see breakage
        - "git diff HEAD" to help me see what I broke
        - go back to 'here' to fix it up
        - "git diff HEAD" to help me see what I changed
        - "git add foo bar" to include what I changed
        - "git commit"

If I have a large project on a filesystem with slow lstat(2), I
would imagine your development is slowed anyway because you
would use diff far more often than commit.  I wonder if it may
be a better idea to use (and extend if needed) existing 'assume
unchanged' on such a system, exactly because "diff" side would
take more time than final "commit", and if you do use 'assume
unchanged', then it also makes --refresh a no-op.

In any case, I think your --index is a misnomer, as we do commit
the current index.  If the sole purpose of your patch is to omit
refreshing it, then it should be named as such.

Re: [RFC] Add --index to git-commit: just commit current index

[Alex Riesen]

On 3/16/07, Junio C Hamano <junkio@cox.net> wrote:
>
> > First use of new --quiet :)
>
> You do not need to say --exit-code if you use --quiet.

Just forgot it in

> > Refreshing index takes a long time on big repositories with
> > many files, especially if the developer was unlucky enough to
> > stick to a slow filesystem or a broken OS. In this situation
> > explicit git-update-index with git-commit --index will speedup
> > the workflow.
>
> Does it?

IFF you use git-update-index, yes

> A typical workflow would go something like this:
>
>         - repeat from here
>         - "edit foo"
>         - "edit bar"
>         - "git diff" to help me see what I changed
>         - "git add foo" as the change is sane
>         - test and see breakage
>         - "git diff HEAD" to help me see what I broke
>         - go back to 'here' to fix it up
>         - "git diff HEAD" to help me see what I changed
>         - "git add foo bar" to include what I changed
>         - "git commit"

- edit
- test
- edit
- test
- update-index (or add) file(s)
- git commit --index <--- this is faster than before
- repeat

> If I have a large project on a filesystem with slow lstat(2), I
> would imagine your development is slowed anyway because you
> would use diff far more often than commit.  I wonder if it may

I avoid using diff-files for the whole project. diff-index --cached is
fast always. It is ok even on windows

> be a better idea to use (and extend if needed) existing 'assume
> unchanged' on such a system, exactly because "diff" side would
> take more time than final "commit", and if you do use 'assume
> unchanged', then it also makes --refresh a no-op.

Forgot about it too. It didn't seem to work properly, the last time
I tried (for a long time, I admit). Besides, sometimes I want to do a
refresh, and having to switch the option on and off is annoying.

> In any case, I think your --index is a misnomer, as we do commit
> the current index.  If the sole purpose of your patch is to omit
> refreshing it, then it should be named as such.

--no-refresh?