Re: Viewing untracked+stashed files in git stash show

[Jeff King <peff@peff.net>]

On Fri, Mar 17, 2017 at 09:51:44AM -0400, Jeff King wrote:

> > Okay that's fine as long as you know. But the first thing that comes
> > to your mind is that you've lost those untracked files. Is there a
> > reason why git stash show doesn't show those files?
> 
> No, I don't think there is a good reason. The "untracked" option was
> added to git-stash much later, and nobody considered how it interacted
> with "stash show".
> 
> Changes from the index have the same problem. Probably "stash show"
> should have options "--index" and "--untracked" to show:
> 
>   # index
>   git diff $stash^1 $stash^2
> 
>   # untracked
>   git diff $stash^1 $stash^3
> 
> Possibly it should even show those by default when they are non-empty.

Actually, $stash^1 is already the parent of the other commits, so we can
just show them as-is. Doing this:

diff --git a/git-stash.sh b/git-stash.sh
index 9c70662cc..868d9027b 100755
--- a/git-stash.sh
+++ b/git-stash.sh
@@ -388,7 +388,16 @@ show_stash () {
 		fi
 	fi
 
-	git diff ${FLAGS} $b_commit $w_commit
+	{
+		# use --pretty=format instead of tformat, because it omits the
+		# blank line between the subject and the diff; use "."
+		# to enable history simplification and drop empty commits.
+		git show --first-parent -m --pretty=format:'==> %s <==' \
+			${FLAGS} $w_commit -- .
+		git show --pretty=format:'%n==> %s <==' ${FLAGS} $i_commit -- .
+		test -n "$u_commit" &&
+			git show --pretty=format:'%n==> %s <==' ${FLAGS} $u_commit -- .
+	} | git_pager
 }
 
 show_help () {
actually looks OK to me, though I don't know if people really like the
existing format. With a setup like:

  git init
  echo base >file
  git add file
  git commit -m base
  echo index >file
  git add file
  echo working tree >file
  echo untracked >other
  git stash -u

it looks like:

  $ git stash show
  ==> WIP on master: 8aa1ec9 base <==
   file | 2 +-
   1 file changed, 1 insertion(+), 1 deletion(-)
  
  ==> index on master: 8aa1ec9 base <==
   file | 2 +-
   1 file changed, 1 insertion(+), 1 deletion(-)
  
  ==> untracked files on master: 8aa1ec9 base <==
   other | 1 +
   1 file changed, 1 insertion(+)

or:

  $ git stash show -p
  ==> WIP on master: 8aa1ec9 base <==
  diff --git a/file b/file
  index df967b9..ace4bb6 100644
  --- a/file
  +++ b/file
  @@ -1 +1 @@
  -base
  +working tree
  
  ==> index on master: 8aa1ec9 base <==
  diff --git a/file b/file
  index df967b9..9015a7a 100644
  --- a/file
  +++ b/file
  @@ -1 +1 @@
  -base
  +index
  
  ==> untracked files on master: 8aa1ec9 base <==
  diff --git a/other b/other
  new file mode 100644
  index 0000000..5a72eb2
  --- /dev/null
  +++ b/other
  @@ -0,0 +1 @@
  +untracked

And if you have only a single change, it omits the others:

  $ echo working tree >file
  $ git stash
  $ git stash show
  ==> WIP on master: 8aa1ec9 base <==
   file | 2 +-
   1 file changed, 1 insertion(+), 1 deletion(-)


If I were designing "stash show" from scratch today, I'd probably do
something like that. But like I said, I don't know if people are
invested in the current output (which _can_ be piped to "git apply" if
need be, though my output could, too, if you do not have index or
untracked files).

-Peff