git-ls-files -o no recurse?

git-ls-files -o no recurse?

[Darrin Thompson]

git-ls-files -o reports _all_ the unknown files it finds in a work area.
Subversion and probably other systems "simply ignore all the files 
and directories inside an unknown directory and just note the directory
as unknown." (A quote from my porcelain's mailing list.)

I'd like to make this friendly behavior the default in my porcelain.
Unfortunately, getting the same result from a porcelain requires
implementing a non-trivial algorithm to discover common prefixes, worry
about performance on large projects, etc.

Would it be hard to make git-ls-files optionally do this?

--
Darrin

Re: git-ls-files -o no recurse?

[Linus Torvalds]

On Wed, 4 Jan 2006, Darrin Thompson wrote:
> 
> Would it be hard to make git-ls-files optionally do this?

Something like the appended may or may not be what you're looking for..

		Linus
---
diff --git a/ls-files.c b/ls-files.c
index 5e9ac71..2e3e2e8 100644
--- a/ls-files.c
+++ b/ls-files.c
@@ -19,6 +19,7 @@ static int show_stage = 0;
 static int show_unmerged = 0;
 static int show_modified = 0;
 static int show_killed = 0;
+static int show_ignored_directories = 0;
 static int line_terminator = '\n';
 
 static int prefix_len = 0, prefix_offset = 0;
@@ -233,6 +234,19 @@ static void add_name(const char *pathnam
 	dir[nr_dir++] = ent;
 }
 
+static int dir_exists(const char *dirname, int len)
+{
+	int pos = cache_name_pos(dirname, len);
+	if (pos >= 0)
+		return 1;
+	pos = -pos-1;
+	if (pos >= active_nr)
+		return 0;
+	if (strncmp(active_cache[pos]->name, dirname, len))
+		return 0;
+	return active_cache[pos]->name[len] == '/';
+}
+
 /*
  * Read a directory tree. We currently ignore anything but
  * directories, regular files and symlinks. That's because git
@@ -280,6 +294,10 @@ static void read_directory(const char *p
 					continue;
 				/* fallthrough */
 			case DT_DIR:
+				if (show_ignored_directories) {
+					if (!dir_exists(fullname, baselen + len))
+						break;
+				}
 				memcpy(fullname + baselen + len, "/", 2);
 				read_directory(fullname, fullname,
 					       baselen + len + 1);
@@ -622,6 +640,10 @@ int main(int argc, const char **argv)
 			show_killed = 1;
 			continue;
 		}
+		if (!strcmp(arg, "--directory")) {
+			show_ignored_directories = 1;
+			continue;
+		}
 		if (!strcmp(arg, "-u") || !strcmp(arg, "--unmerged")) {
 			/* There's no point in showing unmerged unless
 			 * you also show the stage information.

Re: git-ls-files -o no recurse?

[Darrin Thompson]

On Wed, 2006-01-04 at 13:31 -0800, Linus Torvalds wrote:
> 
> On Wed, 4 Jan 2006, Darrin Thompson wrote:
> > 
> > Would it be hard to make git-ls-files optionally do this?
> 
> Something like the appended may or may not be what you're looking for..
> 

Worked great!

> @@ -19,6 +19,7 @@ static int show_stage = 0;
>  static int show_unmerged = 0;
>  static int show_modified = 0;
>  static int show_killed = 0;
> +static int show_ignored_directories = 0;
>  static int line_terminator = '\n';

Did you perhaps mean s/ignored/other/g ? I thought maybe you were being
clever but I couldn't get it to do anything useful with -i, only -o.

--
Darrin

Re: git-ls-files -o no recurse?

[Linus Torvalds]

On Thu, 5 Jan 2006, Darrin Thompson wrote:
> 
> Worked great!

Ok. You can bug Junio about it when he returns, maybe he thinks it's a 
great feature.

> > +static int show_ignored_directories = 0;
> 
> Did you perhaps mean s/ignored/other/g ? I thought maybe you were being
> clever but I couldn't get it to do anything useful with -i, only -o.

Yeah. 

		Linus

Re: git-ls-files -o no recurse?

[Junio C Hamano]

Linus Torvalds <torvalds@osdl.org> writes:

> On Wed, 4 Jan 2006, Darrin Thompson wrote:
>> 
>> Would it be hard to make git-ls-files optionally do this?
>
> Something like the appended may or may not be what you're looking for..
>
> 		Linus

Not to mention test scripts, I suspect that making this the
default would break existing callers including the famous
"git-init-db && git-add ." pattern.  The other commonplace use
is git-status, which might benefit from this option.  If we are
going to use it in git-status, we may want to give them trailing
slashes to make them stand out, though.

Re: git-ls-files -o no recurse?

[Darrin Thompson]

This adds a test a test for the --directory option to git-ls-files.

Is '--directory' really what we want?

 t/t3003-ls-files-others-directory.sh |   38
++++++++++++++++++++++++++++++++++
 1 files changed, 38 insertions(+), 0 deletions(-)
 create mode 100644 t/t3003-ls-files-others-directory.sh

applies-to: 5001806332081159b00c35291d6aea232ed7e909
d484a8477d430a5fb2cefd2cf7008a5973d1fea5
diff --git a/t/t3003-ls-files-others-directory.sh
b/t/t3003-ls-files-others-directory.sh
new file mode 100644
index 0000000..d1d3d86
--- /dev/null
+++ b/t/t3003-ls-files-others-directory.sh
@@ -0,0 +1,38 @@
+#!/bin/sh
+#
+# Copyright (c) 2005 Darrin Thompson
+# Based on an earlier test by Junio C Hamano
+#
+
+test_description='git-ls-files test (--others --directory)
+
+This test runs git-ls-files --others --directory with the following on
+the filesystem.
+
+    path0       - a file
+    path1	- a symlink
+    path2/file2 - a file in a directory
+
+The --directory option should cause path2 to be in the listing, but
+not path2/file2.
+'
+. ./test-lib.sh
+
+date >path0
+ln -s xyzzy path1
+mkdir path2
+date >path2/file2
+test_expect_success \
+    'git-ls-files --directory --others to show output.' \
+    'git-ls-files --directory --others >output'
+cat >expected <<EOF
+output
+path0
+path1
+path2
+EOF
+
+test_expect_success \
+    'git-ls-files --directory --others should not pick up dir
contents.' \
+    'diff output expected'
+test_done
---
0.99.9i

Re: git-ls-files -o no recurse?

[Darrin Thompson]

Here's a rejiggered version of the original Linus patch. The names have
been changed a little.

Enjoy.

It prevents git from recursing into "other" directories when used with
the -o option.

---

 ls-files.c |   22 ++++++++++++++++++++++
 1 files changed, 22 insertions(+), 0 deletions(-)

applies-to: 4e7e791ecc24975530de1f2855cf5f17f112140b
741ec8ad5b7f3717bb462c2becfd00974da7ec16
diff --git a/ls-files.c b/ls-files.c
index 5e9ac71..cba8ca1 100644
--- a/ls-files.c
+++ b/ls-files.c
@@ -19,6 +19,7 @@ static int show_stage = 0;
 static int show_unmerged = 0;
 static int show_modified = 0;
 static int show_killed = 0;
+static int show_other_directories = 0;
 static int line_terminator = '\n';
 
 static int prefix_len = 0, prefix_offset = 0;
@@ -233,6 +234,19 @@ static void add_name(const char *pathnam
 	dir[nr_dir++] = ent;
 }
 
+static int dir_exists(const char *dirname, int len)
+{
+	int pos = cache_name_pos(dirname, len);
+	if (pos >= 0)
+		return 1;
+	pos = -pos-1;
+	if (pos >= active_nr)
+		return 0;
+	if (strncmp(active_cache[pos]->name, dirname, len))
+		return 0;
+	return active_cache[pos]->name[len] == '/';
+}
+
 /*
  * Read a directory tree. We currently ignore anything but
  * directories, regular files and symlinks. That's because git
@@ -280,6 +294,10 @@ static void read_directory(const char *p
 					continue;
 				/* fallthrough */
 			case DT_DIR:
+				if (show_other_directories) {
+					if (!dir_exists(fullname, baselen + len))
+						break;
+				}
 				memcpy(fullname + baselen + len, "/", 2);
 				read_directory(fullname, fullname,
 					       baselen + len + 1);
@@ -622,6 +640,10 @@ int main(int argc, const char **argv)
 			show_killed = 1;
 			continue;
 		}
+		if (!strcmp(arg, "--directory")) {
+			show_other_directories = 1;
+			continue;
+		}
 		if (!strcmp(arg, "-u") || !strcmp(arg, "--unmerged")) {
 			/* There's no point in showing unmerged unless
 			 * you also show the stage information.
---
0.99.9i

Re: git-ls-files -o no recurse?

[Junio C Hamano]

Darrin Thompson <darrint@progeny.com> writes:

> Here's a rejiggered version of the original Linus patch. The names have
> been changed a little.
>
> Enjoy.
>
> It prevents git from recursing into "other" directories when used with
> the -o option.

Looks same to what I have in the proposed updates branch, so we
are on the same page.  Thanks.

I have been unsure about the name "--directory", like you said.
The most accurate one I thought of so far is:

  --dont-recurse-into-directories-without-tracked-files

but it is way too long.  We cannot even say "untracked
directories" (we do not track directories) to shorten it a bit.

Re: git-ls-files -o no recurse?

[Linus Torvalds]

On Sat, 7 Jan 2006, Junio C Hamano wrote:
> 
> I have been unsure about the name "--directory", like you said.

Well, I didn't like "--directory" either, but couldn't come up with 
anything better. How about just "--no-recurse", which is not technically 
accurate (we always recurse into directories we know about), but might be 
more understandable.

The reason I called it "--directory" was that it would talk about 
directories that it doesn't know about (as opposed to individual files it 
doesn't know about). So "--other --directory" kind of makes sense if you 
read it that way ("show files and directories we don't know about").

But yeah, it's not a great name.

		Linus