PATCH: _ssh (scp)

PATCH: _ssh (scp)

[Clint Adams]

So this is based on a patch from David Engel for something resembling
the current 4.0.x CVS.  Is it necessary to provide the echo behavior
as well, or can we safely assume that everything can do ls -d1F these
days?

Index: Completion/Unix/Command/_ssh
===================================================================
RCS file: /cvsroot/zsh/zsh/Completion/Unix/Command/_ssh,v
retrieving revision 1.5
diff -u -r1.5 _ssh
--- Completion/Unix/Command/_ssh	2002/01/16 16:29:52	1.5
+++ Completion/Unix/Command/_ssh	2002/02/04 17:25:18
@@ -3,11 +3,16 @@
 _remote_files () {
   # This is extremely simple-minded; could parse "ls -F" output to do
   # colorings and LIST_TYPES and so on, but I'm just not that ambitious.
-  local expl
+  local expl remfiles
 
   if zstyle -T ":completion:${curcontext}:" remote-access; then
+    remfiles=(${(f)"$(ssh -a -x ${words[CURRENT]%:*} ls -d1F ${words[CURRENT]#*:}\*)"})
+
     _wanted files expl 'remote files' \
-        compadd $(ssh -a -x ${words[CURRENT]%:*} echo ${words[CURRENT]#*:}\*)
+      compadd ${${remfiles:#*/}/[*=@](#e)/}
+
+    _wanted dirs expl 'remote directories' \
+      compadd -S/ ${${(M)remfiles:#*/}/\\/(#e)/}
   else
     _message 'remote files'
   fi

Re: PATCH: _ssh (scp)

[Bart Schaefer]

On Feb 4, 12:59pm, Clint Adams wrote:
} Subject: PATCH: _ssh (scp)
}
} So this is based on a patch from David Engel for something resembling
} the current 4.0.x CVS.  Is it necessary to provide the echo behavior
} as well, or can we safely assume that everything can do ls -d1F these
} days?

Assorted remarks:

This still needs the 2>/dev/null fix I described in 16533.

If you *are* going to parse "ls -F" output, you should remove the comment
that indicates otherwise.

Another possible approach would be to use ls (without -d) on the directory
parent of the remote path, and let the completion system filter out the
non-matching names, rather than the rather crude hack I did of appending
a `*' to the path.  This would support complete-in-word a bit better, but
means more network traffic.

-- 
Bart Schaefer                                 Brass Lantern Enterprises
http://www.well.com/user/barts              http://www.brasslantern.com

Zsh: http://www.zsh.org | PHPerl Project: http://phperl.sourceforge.net   

Re: PATCH: _ssh (scp)

[Clint Adams]

> This still needs the 2>/dev/null fix I described in 16533.

Right.  Patch below.  Also adds | to the bracket expression.
We still need a solution to the zle redisplay problem.

> If you *are* going to parse "ls -F" output, you should remove the comment
> that indicates otherwise.

Well, it's not doing colorings, but I'll modify the comment.

> Another possible approach would be to use ls (without -d) on the directory
> parent of the remote path, and let the completion system filter out the
> non-matching names, rather than the rather crude hack I did of appending
> a `*' to the path.  This would support complete-in-word a bit better, but
> means more network traffic.

That should probably be configurable via an adjunct to the remote-access
style then.

Index: Completion/Unix/Command/_ssh
===================================================================
RCS file: /cvsroot/zsh/zsh/Completion/Unix/Command/_ssh,v
retrieving revision 1.6
diff -u -r1.6 _ssh
--- Completion/Unix/Command/_ssh	2002/02/04 18:03:25	1.6
+++ Completion/Unix/Command/_ssh	2002/02/04 18:40:48
@@ -1,15 +1,15 @@
 #compdef ssh slogin=ssh scp ssh-add ssh-agent ssh-keygen
 
 _remote_files () {
-  # This is extremely simple-minded; could parse "ls -F" output to do
-  # colorings and LIST_TYPES and so on, but I'm just not that ambitious.
+  # Rather than simple directory discrimination, there should be
+  # coloring based on all the different ls -F classifiers.
   local expl remfiles
 
   if zstyle -T ":completion:${curcontext}:" remote-access; then
-    remfiles=(${(f)"$(ssh -a -x ${words[CURRENT]%:*} ls -d1F ${words[CURRENT]#*:}\*)"})
+    remfiles=(${(f)"$(ssh -a -x ${words[CURRENT]%:*} ls -d1F ${words[CURRENT]#*:}\* 2>/dev/null)"})
 
     _wanted files expl 'remote files' \
-      compadd ${${remfiles:#*/}/[*=@](#e)/}
+      compadd ${${remfiles:#*/}/[*=@|](#e)/}
 
     _wanted dirs expl 'remote directories' \
       compadd -S/ ${${(M)remfiles:#*/}/\\/(#e)/}

Re: PATCH: _ssh (scp)

[Oliver Kiddle]

Clint wrote:

>      _wanted files expl 'remote files' \
> -        compadd $(ssh -a -x ${words[CURRENT]%:*} echo
> ${words[CURRENT]#*:}\*)
> +      compadd ${${remfiles:#*/}/[*=@](#e)/}
> +
> +    _wanted dirs expl 'remote directories' \
> +      compadd -S/ ${${(M)remfiles:#*/}/\\/(#e)/}

It would be more consistent with normal file completion to add the
files and directories together with the same tag and description.
I'd also be tempted to setup separate display strings so that the * for
executables, @ for symlinks etc are preserved in the file list.

Is it possible to have colour coding work in this sort of case? When I
did _mtools I briefly looked but couldn't see how it could be done.

Oliver

__________________________________________________
Do You Yahoo!?
Everything you'll ever need on one web page
from News and Sport to Email and Music Charts
http://uk.my.yahoo.com

Re: PATCH: _ssh (scp)

[Clint Adams]

> It would be more consistent with normal file completion to add the
> files and directories together with the same tag and description.
> I'd also be tempted to setup separate display strings so that the * for
> executables, @ for symlinks etc are preserved in the file list.

This does that somewhat messily.

I intend to revert the bit where it checks PREFIX and in all cases
have it do an ls on the dirname.

However, I can't decide if it is better to
ls -a1F /dir (thereby getting all the dotfiles and potentially
respecting the local setting of GLOB_DOTS)
or
ls -d1F /dir/* (thereby respecting the remote setting of GLOB_DOTS or
equivalent)

Index: Completion/Unix/Command/_ssh
===================================================================
RCS file: /cvsroot/zsh/zsh/Completion/Unix/Command/_ssh,v
retrieving revision 1.7
diff -u -r1.7 _ssh
--- Completion/Unix/Command/_ssh	2002/02/04 18:44:23	1.7
+++ Completion/Unix/Command/_ssh	2002/02/05 17:35:42
@@ -1,18 +1,26 @@
 #compdef ssh slogin=ssh scp ssh-add ssh-agent ssh-keygen
 
 _remote_files () {
-  # Rather than simple directory discrimination, there should be
-  # coloring based on all the different ls -F classifiers.
-  local expl remfiles
+  # There should be coloring based on all the different ls -F classifiers.
+  local expl remfiles remdispf remdispd
 
   if zstyle -T ":completion:${curcontext}:" remote-access; then
-    remfiles=(${(f)"$(ssh -a -x ${words[CURRENT]%:*} ls -d1F ${words[CURRENT]#*:}\* 2>/dev/null)"})
+    if [[ $options[globdots] == "on" ]] && [[ "$PREFIX" == */ || -z "$PREFIX" ]];
+    then
+      print ssh -a -x ${words[CURRENT]%:*} ls -a1F ${words[CURRENT]#*:}
+      remfiles=(${(f)"$(ssh -a -x ${words[CURRENT]%:*} ls -a1F ${words[CURRENT]#*:} 2>/dev/null)"})
+    else
+      remfiles=(${(f)"$(ssh -a -x ${words[CURRENT]%:*} ls -d1F ${words[CURRENT]#*:}\* 2>/dev/null)"})
+    fi
 
-    _wanted files expl 'remote files' \
-      compadd ${${remfiles:#*/}/[*=@|](#e)/}
+    remdispf=(${remfiles:#*/})
+    remdispd=(${(M)remfiles:#*/})
 
-    _wanted dirs expl 'remote directories' \
-      compadd -S/ ${${(M)remfiles:#*/}/\\/(#e)/}
+    _wanted files expl 'remote files and directories' \
+      compadd -d remdispf ${${remfiles:#*/}/[*=@|](#e)/}
+      
+    _wanted files expl 'remote files and directories' \
+      compadd -S/ -d remdispd ${${(M)remfiles:#*/}/\\/(#e)/}
   else
     _message 'remote files'
   fi

Re: PATCH: _ssh (scp)

[Clint Adams]

> ls -a1F /dir (thereby getting all the dotfiles and potentially
> respecting the local setting of GLOB_DOTS)
> or
> ls -d1F /dir/* (thereby respecting the remote setting of GLOB_DOTS or
> equivalent)

I opted for the latter, but it doesn't quite work.

3.0.8% SLASH=/
3.0.8% print $SLASH
/
3.0.8% print $SLASH:h
/

4.0.4% SLASH=/
4.0.4% print $SLASH
/
4.0.4% print $SLASH:h
.

Was this change intentional?

Index: Completion/Unix/Command/_ssh
===================================================================
RCS file: /cvsroot/zsh/zsh/Completion/Unix/Command/_ssh,v
retrieving revision 1.8
diff -u -r1.8 _ssh
--- Completion/Unix/Command/_ssh	2002/02/05 18:03:26	1.8
+++ Completion/Unix/Command/_ssh	2002/02/05 18:05:04
@@ -5,13 +5,7 @@
   local expl remfiles remdispf remdispd
 
   if zstyle -T ":completion:${curcontext}:" remote-access; then
-    if [[ $options[globdots] == "on" ]] && [[ "$PREFIX" == */ || -z "$PREFIX" ]];
-    then
-      print ssh -a -x ${words[CURRENT]%:*} ls -a1F ${words[CURRENT]#*:}
-      remfiles=(${(f)"$(ssh -a -x ${words[CURRENT]%:*} ls -a1F ${words[CURRENT]#*:} 2>/dev/null)"})
-    else
-      remfiles=(${(f)"$(ssh -a -x ${words[CURRENT]%:*} ls -d1F ${words[CURRENT]#*:}\* 2>/dev/null)"})
-    fi
+    remfiles=(${(f)"$(ssh -a -x ${words[CURRENT]%:*} ls -d1F ${${${words[CURRENT]#*:}:h}/\\/(#e)/}/\* 2>/dev/null)"})
 
     remdispf=(${remfiles:#*/})
     remdispd=(${(M)remfiles:#*/})

Re: PATCH: _ssh (scp)

[Bart Schaefer]

On Feb 5,  1:08pm, Clint Adams wrote:
}
} 4.0.4% SLASH=/
} 4.0.4% print $SLASH:h
} .
} 
} Was this change intentional?

There was a change to make :h behave more as if it knew about file structure,
but I'm quite sure that turning `/' into `.' was not an intended effect, and
in fact it looks like a serious bug to me.


-- 
Bart Schaefer                                 Brass Lantern Enterprises
http://www.well.com/user/barts              http://www.brasslantern.com

Zsh: http://www.zsh.org | PHPerl Project: http://phperl.sourceforge.net   

Re: PATCH: _ssh (scp)

[Clint Adams]

> There was a change to make :h behave more as if it knew about file structure,
> but I'm quite sure that turning `/' into `.' was not an intended effect, and
> in fact it looks like a serious bug to me.

Is the following acceptable behavior on Cygwin?

Index: Src/hist.c
===================================================================
RCS file: /cvsroot/zsh/zsh/Src/hist.c,v
retrieving revision 1.36
diff -u -r1.36 hist.c
--- Src/hist.c	2001/10/26 23:47:10	1.36
+++ Src/hist.c	2002/02/06 16:38:14
@@ -1378,7 +1378,11 @@
     while (str >= *junkptr && !IS_DIRSEP(*str))
 	--str;
     if (str < *junkptr) {
-	*junkptr = dupstring (".");
+	if (IS_DIRSEP(**junkptr))
+	    *junkptr = dupstring ("/");
+	else
+	    *junkptr = dupstring (".");
+
 	return 0;
     }
     /* repeated slashes are considered like a single slash */

RE: PATCH: _ssh (scp)

[Borsenkow Andrej]

> 
> Is the following acceptable behavior on Cygwin?
>

Assuming that for foo=/ ${foo:t} is empty - yes. Of course for full
support we should do something or //server/share and x:\path but it is
way too much currently. Actually at some point I guess we silently
decided to not discuss any issues with \ as directory separator :-)

-andrej

The :h modifier

[Brass Lantern Enterprises]

On Wed, 6 Feb 2002, Borsenkow Andrej wrote:

> Of course for full support we should do something or //server/share
> and x:\path but it is way too much currently. Actually at some point I
> guess we silently decided to not discuss any issues with \ as
> directory separator :-)

We didn't "silently" decide that -- as I recall it was pretty noisy.

It's a bit difficult to do a search for this sort of thing in the
archives, but see for example the thread from 2000 with ":r modifier"
in the subject.