Feature request: Limit git-status reports to a directory

Feature request: Limit git-status reports to a directory

[Michel Marti]

I am sometimes interested in only seeing the status for a specific 
directory (and its sub-directories), but git-status is no help in this 
case - passing a directory does some sort of "git-commit --dry-run". I 
first thought that this is a bug until I saw in the man-page that this 
is actually a feature...

What I would like to have is something like "git status --limit <path>" 
and the reported filenames should then be relative to the current 
directory for easier copy/pasting them to e.g. git-add, git-diff, etc.

Cheers,

- Michel

Re: Feature request: Limit git-status reports to a directory

[Yin Ping]

On 10/25/07, Michel Marti <mma@objectxp.com> wrote:
> I am sometimes interested in only seeing the status for a specific
> directory (and its sub-directories), but git-status is no help in this
> case - passing a directory does some sort of "git-commit --dry-run". I
> first thought that this is a bug until I saw in the man-page that this
> is actually a feature...
It's also painful for me. IMHO, the behaviour of "git-status" should
keep consistent with "git-diff" and "git-log" which allow for the
path.

Another point, It will be helpful to add a config item to change the
default behaviour for 'git-diff" and "git-log". For example,
'diff.defaultcurrentpath=true' to let git only show difference in
current directory instead of difference in top directory when typing
'git-diff'
>
> - Michel
>
> -
> To unsubscribe from this list: send the line "unsubscribe git" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>


-- 
franky

Re: Feature request: Limit git-status reports to a directory

[Johannes Schindelin]

Hi,

On Thu, 25 Oct 2007, Yin Ping wrote:

> On 10/25/07, Michel Marti <mma@objectxp.com> wrote:
> > I am sometimes interested in only seeing the status for a specific 
> > directory (and its sub-directories), but git-status is no help in this 
> > case - passing a directory does some sort of "git-commit --dry-run". I 
> > first thought that this is a bug until I saw in the man-page that this 
> > is actually a feature...
>
> It's also painful for me. IMHO, the behaviour of "git-status" should 
> keep consistent with "git-diff" and "git-log" which allow for the path.

I am not so sure.  In other SCMs, "git status" may be a way to do "git 
diff --name-only" or "git ls-files", but not in git.  Here, it strictly 
means "what would be happening if I were to commit _right_ _now_?".

> Another point, It will be helpful to add a config item to change the 
> default behaviour for 'git-diff" and "git-log". For example, 
> 'diff.defaultcurrentpath=true' to let git only show difference in 
> current directory instead of difference in top directory when typing 
> 'git-diff'

IMHO it is not asking users too much when you say "git diff ." is for the 
current directory, and "git diff" is for the whole working tree.

Besides, we cannot really change the default behaviour, since some 
porcelains use "git log" (and certainly there are some which use "git 
diff", too).  They would suffer from this unexpected -- and indeed 
inconsistent, since the setting can differ between repositories -- output.

Hth,
Dscho

Re: Feature request: Limit git-status reports to a directory

[Wincent Colaiuta]

El 25/10/2007, a las 11:55, Johannes Schindelin escribió:

> Hi,
>
> On Thu, 25 Oct 2007, Yin Ping wrote:
>
>> On 10/25/07, Michel Marti <mma@objectxp.com> wrote:
>>
>> It's also painful for me. IMHO, the behaviour of "git-status" should
>> keep consistent with "git-diff" and "git-log" which allow for the  
>> path.
>
> I am not so sure.  In other SCMs, "git status" may be a way to do "git
> diff --name-only" or "git ls-files", but not in git.  Here, it  
> strictly
> means "what would be happening if I were to commit _right_ _now_?".

Yes, but there's no reason why the user shouldn't be able to scope  
that down to a specific path, just as they currently can for git-diff  
(as you point out):

> IMHO it is not asking users too much when you say "git diff ." is  
> for the
> current directory, and "git diff" is for the whole working tree.

Sometimes if you have a dirty tree with lots of modified files and  
potentially lots of things added to the index the output of git- 
status can be quite long, and perhaps all you want to know about is  
what is the status of *this* directory or *that* file rather than  
having to visually scan through the entire git-status output.  
Accepting path info would therefore be a nice usability improvement.

Allowing git-status to accept a path would be consistent with how  
other git commands (like git-diff) already work, and with other SCMs  
too. The user is expected to know that what's in the index is what  
will be committed, and that if he/she types "git-status foo" then he/ 
she may only be seeing a subset of what's staged in the index.

But the way git-status currently behaves when supplied path info is  
puzzling to say the least. As the man page says:

> "The command takes the same set of options as git-commit; it shows  
> what would be committed if the same options are given to git-commit."

This means that if you do try passing a path to git-status (as surely  
many newcomers have done), you'll see the combined result of what is  
already staged in the index *plus* what would happen if you git-added  
the path(s) that you passed on the command line. I'd argue that this  
is counter-intuitive, and I think that most would expect that the  
paths would serve as scope *limiters* rather than indicators that  
something should be *added* to the index.

To illustrate this, an example; just say you have git-status output  
like this:

# On branch master
# Changed but not updated:
#   (use "git add <file>..." to update what will be committed)
#
#       modified:   foo/bar
#       modified:   baz
#
# Untracked files:
#   (use "git add <file>..." to include in what will be committed)
#
#       NOTES
no changes added to commit (use "git add" and/or "git commit -a")

And you type "git-status foo":

# Changes to be committed:
#   (use "git reset HEAD <file>..." to unstage)
#
#       modified:   foo/bar
#
# Changed but not updated:
#   (use "git add <file>..." to update what will be committed)
#
#       modified:   baz
#
# Untracked files:
#   (use "git add <file>..." to include in what will be committed)
#
#       NOTES

I understand why it works this way, and it's explained by the man  
page, but the behaviour is the very last thing I would expect.

Cheers,
Wincent

Re: Feature request: Limit git-status reports to a directory

[Michel Marti]

[-- Attachment #1: Type: text/plain, Size: 755 bytes --]

on 10/25/2007 11:55 AM Johannes Schindelin said the following:
> I am not so sure.  In other SCMs, "git status" may be a way to do "git 
> diff --name-only" or "git ls-files", but not in git.  Here, it strictly 
> means "what would be happening if I were to commit _right_ _now_?".
I somehow agree with your argument, but then again, sometimes I'm still interested in
*only* the status for a given directory.

IMHO, answering the question "what would be happening if...?" should be  git-commit's task
(e.g. git-commit --dry-run). And git-status should behave similar to git-log and git-diff.

For those interested: I have attached a little script that mimics current git-status
(except the "dry-run" stuff) but also can take a list of directories/files.


[-- Attachment #2: git-status-new --]
[-- Type: text/plain, Size: 1957 bytes --]

#!/bin/sh

USAGE='[--staged] [--changed] [--untracked]'
SUBDIRECTORY_OK=1 . git-sh-setup
require_work_tree

print_stat_line() {
	case "$1" in
		M)	echo "#       modified:  $2";;
		A)	echo "#       new file:  $2";;
		R*)	echo "#       renamed:   $2 -> $3";;
		D)	echo "#       deleted:   $2";;
		U)	echo "#       unmerged:  $2";;
		X)	echo "#       $2";;
		*)	echo "#       [$S]:      $2";;
	esac
}

STAGED= CHANGED= UNTRACKED= HP=

while test $# != 0
do
	case "$1" in
		-s|--staged) STAGED=1; shift;;
		-c|--changed) CHANGED=1; shift;;
		-u|--untracked) UNTRACKED=1; shift;;
		--) shift; break;;
		-*) usage;;
		 *) break;;
	esac
done

if BRANCH_NAME=$(git symbolic-ref -q HEAD)
then
	BRANCH_NAME="On branch $(expr "z$BRANCH_NAME" : 'zrefs/heads/\(.*\)')"
else
	BRANCH_NAME="Not currently on any branch"
fi

[ "$#" = 0 ] && cd_to_toplevel


if [ -z "$STAGED$CHANGED$UNTRACKED" ]; then
	STAGED=1; CHANGED=1; UNTRACKED=1
fi

SP=$(echo _/$(git rev-parse --show-cdup)|tr '/' ' '|wc -w)

echo "# $BRANCH_NAME"

# Changes to be commited
[ "$STAGED" ] && git-diff --name-status --cached -M -- "$@"|while read S F R
do
	if [ -z "$HP" ]; then
		echo '# Changes to be committed:'
		echo '#   (use "git reset HEAD <file>..." to unstage)'
		echo '#'
		HP=1
	fi
	F=$(echo $F|cut -d'/' -f$SP-)
	print_stat_line "$S" "$F" "$R"

done

# Changed but not updated
[ "$CHANGED" ] && git-diff --name-status -- "$@"|while read S F
do
	if [ -z "$HP" ]; then
		echo '#'
		echo '# Changed but not updated:'
		echo '#   (use "git add <file>..." to update what will be committed)'
		echo '#'
		HP=1
	fi
	F=$(echo $F|cut -d'/' -f$SP-)
	print_stat_line "$S" "$F"
done

# Untracked files
[ "$UNTRACKED" ] && git-ls-files --exclude-per-directory=.gitignore -o --directory -- "$@"|while read F
do
	if [ -z "$HP" ]; then
		echo '#'
		echo '# Untracked files:'
		echo '#   (use "git add <file>..." to include in what will be committed)'
		echo '#'
		HP=1
	fi
	print_stat_line "X" "$F"
done