From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <zsh-workers-return-6022-mason-zsh=primenet.com.au@sunsite.auc.dk>
Received: (qmail 15673 invoked from network); 13 Apr 1999 15:04:25 -0000
Received: from sunsite.auc.dk (130.225.51.30)
  by ns1.primenet.com.au with SMTP; 13 Apr 1999 15:04:25 -0000
Received: (qmail 25888 invoked by alias); 13 Apr 1999 15:04:14 -0000
Mailing-List: contact zsh-workers-help@sunsite.auc.dk; run by ezmlm
Precedence: bulk
X-No-Archive: yes
X-Seq: 6022
Received: (qmail 25876 invoked from network); 13 Apr 1999 15:04:12 -0000
Message-Id: <9904131439.AA40214@ibmth.df.unipi.it>
To: zsh-workers@sunsite.auc.dk
Subject: Re: PATCH: matching in the new completion system 
In-Reply-To: "Sven Wischnowsky"'s message of "Tue, 13 Apr 1999 15:18:50 DFT."
             <199904131318.PAA04262@beta.informatik.hu-berlin.de> 
Date: Tue, 13 Apr 1999 16:39:55 +0200
From: Peter Stephenson <pws@ibmth.df.unipi.it>

Sven Wischnowsky wrote:
> Peter Stephenson wrote:
> 
> > It gets worse after that.  If I now go back and do /u/s/<TAB>, the whole
> > shell hangs completely and has to be killed from somewhere else.  (I just
> > started off with /u/s/m<TAB>, but it all looks similar.)
> 
> Ick. I can't reproduce this, but...
> 
> Are you sure it hangs? Or is it just incredibly slow? For me, the lack
> of speed comes from the size of the directories `/usr/s*'. Also, there 
> was a problem with building a pattern that is used to remove unmatched 
> strings over which I stumbled when trying the same (`/u/s/<TAB>'). The 
> problem was that I used `${(@M)tmp1:#(${(j:|:)~tmp2})}' where I should
> have used `${(@M)tmp1:#(${(j:|:)~${(@)tmp2:q}})}' to make sure special
> characters in the filenames are not taken as pattern characters.

It's something to with the |'s.  It's trying to build a *huge* pattern like
that.  By the looks of it, it includes every single thing in the command
table.  It begins

*/(XNSquery|XNSrouted|acct|adfutil|allocp|arp|audit|auditbin|auditcat|
auditpr|auditselect|auditstream|automount|autopush|backbyinode|backbyname|
backup|bffcreate|biod|bootexpand|bootinfo|bootpd|bootrec...

These are the tail ends of the files in /u*/s*/*. The trace of _path_files
finishes up with

+ tmp1=( <all the files in /u*/s*/*, this looks reasonable> )
+ compadd -O tmp2 -F fignore - <all the commands from the last part of that>
+ [[ 423 -eq 0 ]]
+ [[ /usr/sbin/XNSquery == */* ]]

and then it stops.  The part it's doing corresponds to

    # Remove all files that weren't matched.

    if [[ "$tmp1[1]" = */* ]]; then
      tmp1=( "${(@M)tmp1:#*/(${(j:|:)~${(@)tmp2:q}})}" )
    else
      tmp1=( "${(@M)tmp1:#(${(j:|:)~${(@)tmp2:q}})}" )
    fi

It does seem to be due to speed.  In fact, it now takes 8 seconds just to
build all the files in the Src directory, irrespective of path completion,
which should be almost instantaneous.  I smell a rat somewhere.

-- 
Peter Stephenson <pws@ibmth.df.unipi.it>       Tel: +39 050 844536
WWW:  http://www.ifh.de/~pws/
Dipartimento di Fisica, Via Buonarroti 2, 56127 Pisa, Italy