History !# substitutions

History !# substitutions

[Bart Schaefer]

The !# substitution, which shuffles words from the current command line,
has always been a bit flaky in zsh, but now it seems to be broken entirely.

zagzig% echo foo bar !#:1
zsh: no words available from current command

(It should have echoed "foo bar foo".)

I'm not sure when this broke, but it was sometime before beta15-hzoli14.

I suspect it's related to the behavior of "fc" in the "fed" function that
was posted here a few days ago, because they both want the "current"
history event, which in some sense hasn't happened yet.

-- 
Bart Schaefer                             Brass Lantern Enterprises
http://www.well.com/user/barts            http://www.nbn.com/people/lantern

New male in /home/schaefer:
>N  2 Justin William Schaefer  Sat May 11 03:43  53/4040  "Happy Birthday"

Re: History !# substitutions

[Peter Stephenson]

> The !# substitution, which shuffles words from the current command line,
> has always been a bit flaky in zsh, but now it seems to be broken entirely.
> 
> zagzig% echo foo bar !#:1
> zsh: no words available from current command

Profuse apologies:  this no doubt came in why I upgraded the
word-selection part of the history mechanism and didn't know enough
about !# to make it work.

The following seems to do the trick; do let me know of unforeseen
behaviour.  (I've deleted that particular error message since `no such
word in event' seems to describe the remaining occasions when there is
an error entirely adequately.)  I hope flakiness can now be avoided
entirely.

*** Src/hist.c.curhist	Fri May 10 22:22:30 1996
--- Src/hist.c	Wed May 29 10:12:15 1996
***************
*** 1111,1131 ****
  getargs(Histent elist, int arg1, int arg2)
  {
      char *ret;
!     int pos1;
  
!     if (arg1 >= elist->nwords || arg2 >= elist->nwords) {
  	/* remember, argN is indexed from 0, nwords is total no. of words */
  	inerrflush();
! 	if (!elist->nwords)
! 	    zerr("no words available from current command", NULL, 0);
! 	else
! 	    zerr("no such word in event", NULL, 0);
  	return NULL;
      }
  
!     pos1 = elist->words[2*arg1];
      ret = dupstring(elist->text + pos1);
!     ret[elist->words[2*arg2+1] - pos1] = '\0';
      
      return ret;
  }
--- 1111,1137 ----
  getargs(Histent elist, int arg1, int arg2)
  {
      char *ret;
!     short *words;
!     int pos1, nwords;
  
!     if (elist == curhistent) {
!       words = chwords;
!       nwords = chwordpos/2;
!     } else {
!       words = elist->words;
!       nwords = elist->nwords;
!     }
! 
!     if (arg1 >= nwords || arg2 >= nwords) {
  	/* remember, argN is indexed from 0, nwords is total no. of words */
  	inerrflush();
! 	zerr("no such word in event", NULL, 0);
  	return NULL;
      }
  
!     pos1 = words[2*arg1];
      ret = dupstring(elist->text + pos1);
!     ret[words[2*arg2+1] - pos1] = '\0';
      
      return ret;
  }

-- 
Peter Stephenson <pws@ifh.de>       Tel: +49 33762 77366
WWW:  http://www.ifh.de/~pws/       Fax: +49 33762 77330
Deutches Electronen-Synchrotron --- Institut fuer Hochenergiephysik Zeuthen
DESY-IfH, 15735 Zeuthen, Germany.

Re: History !# substitutions

[Peter Stephenson]

pws@ifh.de wrote:
> I hope flakiness can now be avoided entirely.

Did I really say that?

This additional patch fixes up things like

% print !#:$:s/i/o/o
pronto

which even after the last patch was choking because the number of
words available was not being counted properly, so it couldn't count
back from the end of the line.  

At this point, the end of the line is the last word to be slurped,
which may or may not be obvious.  I think this is useful, since you
quite often need to do things to the previous word on the line.  It's
not consistent with lines other than the current one, but if it does
anything, it has to be this.  I added to the manual entry in case it's
not clear.  (If the manual entry is wrong it's a bug.)

*** Src/hist.c.nwords	Wed May 29 10:12:15 1996
--- Src/hist.c	Wed May 29 16:58:24 1996
***************
*** 181,186 ****
--- 181,198 ----
      return 0;
  }
  
+ /* Get the maximum no. of words for a history entry. */
+ 
+ /**/
+ int
+ getargc(Histent ehist)
+ {
+   if (ehist == curhistent)
+     return chwordpos ? chwordpos/2-1 : 0;
+   else
+     return ehist->nwords-1;
+ }
+ 
  /* Perform history substitution, returning the next character afterwards. */
  
  /**/
***************
*** 199,205 ****
  	isfirstch = 0;
  	inungetc(hatchar);
  	if (!(ehist = gethist(defev))
! 	    || !(sline = getargs(ehist, 0, ehist->nwords-1))
  	    || getsubsargs(sline) || !hsubl)
  	    return -1;
  	subst(&sline, hsubl, hsubr, 0);
--- 211,217 ----
  	isfirstch = 0;
  	inungetc(hatchar);
  	if (!(ehist = gethist(defev))
! 	    || !(sline = getargs(ehist, 0, getargc(ehist)))
  	    || getsubsargs(sline) || !hsubl)
  	    return -1;
  	subst(&sline, hsubl, hsubr, 0);
***************
*** 318,331 ****
  
  	/* extract the relevant arguments */
  
! 	argc = ehist->nwords - 1;
  	if (c == ':') {
  	    cflag = 1;
  	    c = ingetc();
  	    if (c == '%' && marg != -1) {
  		if (!evset) {
  		    ehist = gethist(defev = mev);
! 		    argc = ehist->nwords-1;
  		} else {
  		    zerr("Ambiguous history reference", NULL, 0);
  		    while (c != '\n' && !lexstop)
--- 330,343 ----
  
  	/* extract the relevant arguments */
  
! 	argc = getargc(ehist);
  	if (c == ':') {
  	    cflag = 1;
  	    c = ingetc();
  	    if (c == '%' && marg != -1) {
  		if (!evset) {
  		    ehist = gethist(defev = mev);
! 		    argc = getargc(ehist);
  		} else {
  		    zerr("Ambiguous history reference", NULL, 0);
  		    while (c != '\n' && !lexstop)
*** Doc/zshexpn.man.nwords	Wed May 29 17:14:23 1996
--- Doc/zshexpn.man	Wed May 29 17:16:09 1996
***************
*** 804,810 ****
  .IR str .
  .TP
  .B !#
! Refer to the current command line typed in so far.
  .TP
  .BR !{ .\|.\|. }
  Insulate a history reference from adjacent characters (if necessary).
--- 804,812 ----
  .IR str .
  .TP
  .B !#
! Refer to the current command line typed in so far.  The line is
! treated as if it were complete up to and including the word before the
! one with the !# reference.
  .TP
  .BR !{ .\|.\|. }
  Insulate a history reference from adjacent characters (if necessary).

-- 
Peter Stephenson <pws@ifh.de>       Tel: +49 33762 77366
WWW:  http://www.ifh.de/~pws/       Fax: +49 33762 77330
Deutches Electronen-Synchrotron --- Institut fuer Hochenergiephysik Zeuthen
DESY-IfH, 15735 Zeuthen, Germany.

Re: History !# substitutions

[Bart Schaefer]

On May 29, 10:19am, Peter Stephenson wrote:
} Subject: Re: History !# substitutions
}
} > The !# substitution, which shuffles words from the current command line,
} > has always been a bit flaky in zsh, but now it seems to be broken entirely.

On May 29,  5:20pm, Peter Stephenson wrote:
} Subject: Re: History !# substitutions
}
} pws@ifh.de wrote:
} > I hope flakiness can now be avoided entirely.
} 
} Did I really say that?

You did, but thank you anyway.  This is the least flaky I can ever recall
it being.  The only thing that still doesn't work is completion of a
reference into the current command, and even that now works a lot better
than it used to.

Here's zsh 2.4.300:

zipx1[22] echo foo bar !#:1 ba<TAB>
zsh: no such word in event
zipx1[22] echo foo bar
                       ^
		cursor here

Here's beta19 with your two patches:

zigzag% echo foo bar !#:1 ba<TAB>
[BEEP][cursor doesn't move]

Much better, but still not perfect.  I realize it probably can't expand
to the correct word, but could it at least ignore the history error and
continue completing stuff on the rest of the line?

-- 
Bart Schaefer                             Brass Lantern Enterprises
http://www.well.com/user/barts            http://www.nbn.com/people/lantern

New male in /home/schaefer:
>N  2 Justin William Schaefer  Sat May 11 03:43  53/4040  "Happy Birthday"

Re: History !# substitutions

[Peter Stephenson]

schaefer@candle.brasslantern.com wrote:
> zigzag% echo foo bar !#:1 ba<TAB>
> [BEEP][cursor doesn't move]
> 
> Much better, but still not perfect.  I realize it probably can't expand
> to the correct word, but could it at least ignore the history error and
> continue completing stuff on the rest of the line?

I think a judicious `stophist = 1' in get_comp_string() should fix this.
At this point, any valid history expansion has been done, and the
lexrestore() will fix up the real history, so this should be pretty
harmless otherwise.


I changed the code, then found there was a bug in menu completion:
hitting tab repeatedly added an extra space each time.  Then I tried
it with the old version and discovered it was there anyway.  Has this
been reported and fixed while I was asleep?

% setopt menucomplete
% print zle<TAB><TAB><TAB><TAB><TAB><Ctrl-e>

produces

% print zle_misc.c     _
                       ^ cursor here.

This is something to do with the code in do_single() around line 3343 of
zle_tricky.c where something called singlec is used to determine
whether a space or slash should be added.  I don't know enough about
what this does to fix it.


Here's the (probably correct, as far as patches to zle_tricky.c go
:-/) patch for ignoring previous history entries when completing.

*** Src/zle_tricky.c.hstop	Thu May 30 10:18:25 1996
--- Src/zle_tricky.c	Thu May 30 10:19:43 1996
***************
*** 936,941 ****
--- 936,942 ----
      lexsave();
      inpush(dupstrspace((char *) linptr), 0);
      strinbeg();
+     stophist = 1;
      pushheap();
      heapalloc();
      i = tt0 = cp = rd = 0;

-- 
Peter Stephenson <pws@ifh.de>       Tel: +49 33762 77366
WWW:  http://www.ifh.de/~pws/       Fax: +49 33762 77330
Deutches Electronen-Synchrotron --- Institut fuer Hochenergiephysik Zeuthen
DESY-IfH, 15735 Zeuthen, Germany.

Re: History !# substitutions

[Bart Schaefer]

On May 30, 10:33am, Peter Stephenson wrote:
} Subject: Re: History !# substitutions
}
} schaefer@candle.brasslantern.com wrote:
} > zigzag% echo foo bar !#:1 ba<TAB>
} > [BEEP][cursor doesn't move]
} > 
} > Much better, but still not perfect.  I realize it probably can't expand
} > to the correct word, but could it at least ignore the history error and
} > continue completing stuff on the rest of the line?
} 
} I think a judicious `stophist = 1' in get_comp_string() should fix this.

Works like a champ!  Thanks again!

} I changed the code, then found there was a bug in menu completion:
} hitting tab repeatedly added an extra space each time.  Then I tried
} it with the old version and discovered it was there anyway.  Has this
} been reported and fixed while I was asleep?

It's worse than that; look:

zagzig% echo foo co bar<C-b><C-b><C-b><C-b><TAB><TAB><TAB><TAB>
zagzig% echo foo config bar
zagzig% echo foo config.cache barr
zagzig% echo foo config.guess barrr
zagzig% echo foo config.h barrrr

Seems to be an off-by-one error in the length of the string, not just
a space or slash being appended.

-- 
Bart Schaefer                             Brass Lantern Enterprises
http://www.well.com/user/barts            http://www.nbn.com/people/lantern

New male in /home/schaefer:
>N  2 Justin William Schaefer  Sat May 11 03:43  53/4040  "Happy Birthday"

Re: History !# substitutions

[Zefram]

>I changed the code, then found there was a bug in menu completion:
>hitting tab repeatedly added an extra space each time.  Then I tried
>it with the old version and discovered it was there anyway.  Has this
>been reported and fixed while I was asleep?

Yes.  It's a result of the line no longer being NUL terminated when
editing normally -- when the completion code (which *does* have a NUL
at the end of the line) calls a normal editing function, the NUL
doesn't get moved appropriately.  I sent a patch for it a few days
ago.

-zefram

Re: History !# substitutions

[Barton E. Schaefer]

On May 30,  4:11pm, Zefram wrote:
} Subject: Re: History !# substitutions
}
} >I changed the code, then found there was a bug in menu completion:
} >hitting tab repeatedly added an extra space each time.  Then I tried
} >it with the old version and discovered it was there anyway.  Has this
} >been reported and fixed while I was asleep?
} 
} Yes.  It's a result of the line no longer being NUL terminated when
} editing normally -- when the completion code (which *does* have a NUL
} at the end of the line) calls a normal editing function, the NUL
} doesn't get moved appropriately.  I sent a patch for it a few days
} ago.

Sorry, Zefram, but I have that patch applied and I'm still seeing the
behavior I reported (duplication of the last character on the line
ever time TAB is pressed somewhere in the middle of the line).

-- 
Bart Schaefer                     Vice President, Technology, Z-Code Software
schaefer@z-code.com                  Division of NCD Software Corporation
http://www.well.com/www/barts           http://www.ncdsoft.com/ZMail/