Re: compctl -y problem.

Re: compctl -y problem.

[Bart Schaefer]

On May 24, 11:44am, Cesar Crusius wrote:
} Subject: compctl -y problem.
}
} compctl -K project_changes -y '$change_descriptions' aecd
} 
} When I press <Tab> after aecd, the table shows as expected, but the
} cursor does not return to where it should. It goes to one line below
} where I originally typed <Tab>.

I mocked up a test case for this and there seems to be something pretty
badly wrong with the -y option.

Here's the test case:

  text80 () {
    local x=''
    repeat 8 x=1234567890$x
    text=$x$'\n'$text
    reply=($((++count))x80 $last)
    last=($reply)
  }
  text70 () {
    local x=''
    repeat 7 x=1234567890$x
    text=$x$'\n'$text
    reply=($((++count))x80 $last)
    last=($reply)
  }
  array () {
    local x=''
    repeat 8 x=1234567890$x
    text=($x $text)
    reply=($((++count))x80 $last)
    last=($reply)
  }

  compctl -K text80 -y '$text' try80
  compctl -K text70 -y '$text' try70
  compctl -K array -y '$text' tryA

Before each test, initialze the globals:

  count=0 text='' last=()

My results were as follows:

Completion after `try80' puts the cursor N-1 lines ABOVE the correct line,
where N is the number of lines in $text.

Completion after `try70' puts the cursor N-1 lines BELOW the correct line.

Completion after `tryA' puts the shell into an infinite loop; it has to
be killed with `kill -9' from another terminal.

I know Sven won't be thrilled about being asked to fix anything in the
old compctl code ... perhaps we should just remove the `-y' option, as
it's new since 3.0 anyway, and document the correct way to achieve the
same effect by using the new completion system (which, incidentally, I
don't know how to do ... `compadd -d' doesn't quite cut it).

-- 
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: compctl -y problem.

[Bart Schaefer]

(Part of this thread went to zsh-workers ... returning now to zsh-users
with a workaround)

On May 25,  3:11pm, Bart Schaefer wrote:
}
} I know Sven won't be thrilled about being asked to fix anything in the
} old compctl code ... perhaps we should just remove the `-y' option, as
} it's new since 3.0 anyway, and document the correct way to achieve the
} same effect by using the new completion system (which, incidentally, I
} don't know how to do ... `compadd -d' doesn't quite cut it).

On further reading of the compadd doc, `compadd -ld ...' appears to be
the thing to do, though it's a bit convoluted.

autoload -U compinit
compinit

function _aecd {
  local change_descriptions reply
  # Force change_descriptions to be a one-element array
  change_descriptions=("$(aereport changes -unf | grep -v completed)")
  # The name "reply" is not special here, I just happened to use it
  reply=($(aereport changes -terse | grep -v completed | sed -e 's/ .*$//'))
  # Display the change_descriptions in place of the first completion
  compadd -ld change_descriptions $reply[1]
  # Add the rest of the completions, but don't display them in the list
  compadd -na 'reply[2,-1]'
  # Force the list to appear so the descriptions are visible
  compstate[list]='list force'
}
compdef _aecd aecd

It does appear that "compadd -d" still produces a sub-optimal display in
this case if there are any lines in $change_descriptions that have exactly
$COLUMNS characters; but it properly returns the cursor to the right spot
even so.

I don't know what the output of `aereport changes` looks like, so it may
be that there's a simpler solution.  For example, if the same number of
lines will be returned by `aereport changes -unf` and `... -terse`, this
will work:

function _aecd {
  local change_descriptions reply
  change_descriptions=(${(f)"$(aereport changes -unf | grep -v completed)"})
  reply=($(aereport changes -terse | grep -v completed | sed -e 's/ .*$//'))
  compadd -ld change_descriptions -a reply
  compstate[list]='list force'
}

You may also wish to drop that compstate[list] assignment and instead use
the force-list style to control it:

zstyle ':completion:*:aecd' force-list always

See the manual for more.

-- 
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: compctl -y problem.

[Sven Wischnowsky]

Bart Schaefer wrote:

> ...
> 
> I know Sven won't be thrilled about being asked to fix anything in the
> old compctl code ...

Indeed, though the patch below seems to fix it for me.

> perhaps we should just remove the `-y' option, as
> it's new since 3.0 anyway,

I think I suggested this some time ago. It would be really nice because
currently it's the biggest piece of code dangling in the base completion
code (`complete' module) that's specific to compctl.

> and document the correct way to achieve the
> same effect by using the new completion system (which, incidentally, I
> don't know how to do ... `compadd -d' doesn't quite cut it).

You already mentioned `compadd -ld' in another mail. I prefer this a lot
because there we have a one-to-one correspondence between matches and
strings to display, meaning that things like menu selection work
properly.  If there are other strings one want to have displayed
together with the list one can always use `compadd -X' or `compadd -x'.

Bye
  Sven

Index: Src/Zle/compresult.c
===================================================================
RCS file: /cvsroot/zsh/zsh/Src/Zle/compresult.c,v
retrieving revision 1.36
diff -u -r1.36 compresult.c
--- Src/Zle/compresult.c	2001/03/27 09:14:52	1.36
+++ Src/Zle/compresult.c	2001/05/28 11:40:13
@@ -1398,9 +1398,12 @@
 	if (!onlyexpl && pp) {
 	    /* We have an ylist, lets see, if it contains newlines. */
 	    hidden = 1;
-	    while (!nl && *pp)
-		nl = !!strchr(*pp++, '\n');
-
+	    while (!nl && *pp) {
+                if (ztrlen(*pp) >= columns)
+                    nl = 1;
+                else
+                    nl = !!strchr(*pp++, '\n');
+            }
 	    pp = g->ylist;
 	    if (nl || !pp[1]) {
 		/* Yup, there are newlines, count lines. */
@@ -1411,17 +1414,17 @@
 		while ((sptr = *pp)) {
 		    while (sptr && *sptr) {
 			nlines += (nlptr = strchr(sptr, '\n'))
-			    ? 1 + (nlptr-sptr) / columns
-			    : strlen(sptr) / columns;
+			    ? 1 + (nlptr - sptr - 1) / columns
+			    : (ztrlen(sptr) - 1) / columns;
 			sptr = nlptr ? nlptr+1 : NULL;
 		    }
 		    nlines++;
 		    pp++;
 		}
-		nlines--;
+		/*** nlines--; */
 	    } else {
 		while (*pp) {
-		    l = strlen(*pp);
+		    l = ztrlen(*pp);
 		    ndisp++;
 		    if (l > glong)
 			glong = l;
@@ -1524,7 +1527,7 @@
 			g->width = 1;
 			
 			while (*pp)
-			    glines += 1 + (strlen(*pp++) / columns);
+			    glines += 1 + (ztrlen(*pp++) / columns);
 		    }
 		}
 	    } else {
@@ -1566,7 +1569,7 @@
 		    VARARR(int, ylens, yl);
 
 		    for (i = 0; *pp; i++, pp++)
-			ylens[i] = strlen(*pp) + add;
+			ylens[i] = ztrlen(*pp) + add;
 
 		    if (g->flags & CGF_ROWS) {
 			int count, tcol, first, maxlines = 0, llines;
@@ -1913,10 +1916,16 @@
 		}
 	    }
 	    if (g->flags & CGF_LINES) {
-		while (*pp) {
-		    zputs(*pp, shout);
-		    if (*++pp)
-			putc('\n', shout);
+                char *p;
+
+		while ((p = *pp++)) {
+		    zputs(p, shout);
+		    if (*pp) {
+                        if (ztrlen(p) % columns)
+                            putc('\n', shout);
+                        else
+                            fputs(" \010", shout);
+                    }
 		}
 	    } else {
 		int n = g->lcount, nl, nc, i, a;

-- 
Sven Wischnowsky                         wischnow@informatik.hu-berlin.de