Making built-in completion recognize : as word boundary

Making built-in completion recognize : as word boundary

[Matthias B.]

Something that's been bothering me for years now with zsh is that the
built-in completion can't complete

PATH=/usr/bin:/sb<TAB>

presumably because it doesn't consider the colon a word boundary. Could
someone please point me at the code responsible so that I can fix this.

MSB

p.s.: I know that the function-based completion system doesn't have this
problem, but I don't want to use it.

-- 
2+2=5, for moderately large values of two.

Re: Making built-in completion recognize : as word boundary

[Mikael Magnusson]

On 29/02/2008, Matthias B. <msbREMOVE-THIS@winterdrache.de> wrote:
> Something that's been bothering me for years now with zsh is that the
>  built-in completion can't complete
>
>  PATH=/usr/bin:/sb<TAB>
>
>  presumably because it doesn't consider the colon a word boundary. Could
>  someone please point me at the code responsible so that I can fix this.
>
>  MSB
>
>  p.s.: I know that the function-based completion system doesn't have this
>  problem, but I don't want to use it.

path=(/usr/bin /sb<tab>
should work fine i think.

You can also make more of these automagic paired variables with some
argument to typeset.
typeset -T MYVAR myvar
myvar=(do re mi)
echo $MYVAR
do:re:mi

I know that doesn't answer your actual question, but maybe it's useful
as a workaround.

-- 
Mikael Magnusson

Re: Making built-in completion recognize : as word boundary

[Matthias B.]

On Fri, 29 Feb 2008 22:23:38 +0100 "Mikael Magnusson" <mikachu@gmail.com>
wrote:

> I know that doesn't answer your actual question, but maybe it's useful
> as a workaround.

Thanks, but the PATH and : issue was just an example. The same problem
happens with

program --file=/us<TAB>

and in other cases. zsh does not seem to take anything but whitespace as a
word boundary except for special cases such as variable assignment (e.g.
PATH=/us<TAB> works). I really want to fix this for good, but I don't feel
like digging through the complete source, so I'll need at least some
directions.

MSB

-- 
Ambition is a poor excuse for not having enough sense to be lazy.

Re: Making built-in completion recognize : as word boundary

[Oliver Kiddle]

"Matthias B." wrote:
> Something that's been bothering me for years now with zsh is that the
> built-in completion can't complete
> 
> PATH=/usr/bin:/sb<TAB>

Try:
  setopt magicequalsubst

Re: Making built-in completion recognize : as word boundary

[Andrey Borzenkov]

[-- Attachment #1: Type: text/plain, Size: 481 bytes --]

On Saturday 01 March 2008, Oliver Kiddle wrote:
> 
> "Matthias B." wrote:
> > Something that's been bothering me for years now with zsh is that the
> > built-in completion can't complete
> > 
> > PATH=/usr/bin:/sb<TAB>
> 
> Try:
>   setopt magicequalsubst
> 

Well, as it works with "new" completion ouf-of-the-box, is there any reason to
use compctl now?

{pts/1}% print $options[magicequalsubst]
off
{pts/1}% PATH=/usr/bin:/sbTAB
{pts/1}% PATH=/usr/bin:/sbin/

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 197 bytes --]

Re: Making built-in completion recognize : as word boundary

[Matthias B.]

On Sat, 01 Mar 2008 01:37:02 +0100 Oliver Kiddle <okiddle@yahoo.co.uk>
wrote:

> "Matthias B." wrote:
> > Something that's been bothering me for years now with zsh is that the
> > built-in completion can't complete
> > 
> > PATH=/usr/bin:/sb<TAB>
> 
> Try:
>   setopt magicequalsubst

Doesn't help. Nor would I expect it to based on the documentation. It says
nothing about effects on TAB-completion.

On Sat, 1 Mar 2008 09:06:54 +0300 Andrey Borzenkov <arvidjaar@newmail.ru>
wrote:

> Well, as it works with "new" completion ouf-of-the-box, is there any
> reason to use compctl now?

Yes. The shell function based completion is slow (ever tried it on a
500MHz machine? That's what I'm writing this message on.) and more
importantly it's too complex for my taste. I don't want all those
completions and I
don't have the time to prune down the default set or start from scratch
to write my own. All I want is some decent filename completion.

MSB

-- 
I will continue to be an impossible person
as long as those who are now possible remain possible.

Re: Making built-in completion recognize : as word boundary

[Bart Schaefer]

On Mar 1,  2:17pm, Matthias B. wrote:
}
} > > PATH=/usr/bin:/sb<TAB>
} > 
} > Try:
} >   setopt magicequalsubst
} 
} Doesn't help. Nor would I expect it to based on the documentation. It says
} nothing about effects on TAB-completion.

Try:

	compctl -D -f -x 'W[0,*=*] N[-1,:]' -/

See also Misc/compctl-examples in the zsh source.

Re: Making built-in completion recognize : as word boundary

[Matthias B.]

On Sat, 01 Mar 2008 11:08:18 -0800
Bart Schaefer <schaefer@brasslantern.com> wrote:

> Try:
> 
> 	compctl -D -f -x 'W[0,*=*] N[-1,:]' -/
> 
> See also Misc/compctl-examples in the zsh source.

Thanks a lot. The following seems to do what I want, namely enable file
completion after : and = anywhere on the command line.

compctl -D -f -x 'N[-1,:=]' -f

MSB

-- 
Democracy is when the people have the right to make the wrong choices.

Re: Making built-in completion recognize : as word boundary

[Clint Adams]

On Sat, Mar 01, 2008 at 02:17:47PM +0100, Matthias B. wrote:
> Yes. The shell function based completion is slow (ever tried it on a
> 500MHz machine? That's what I'm writing this message on.) and more

I used it religiously on a 440MHz machine until last year.

Re: Making built-in completion recognize : as word boundary

[Oliver Kiddle]

"Matthias B." wrote:
> > Try:
> >   setopt magicequalsubst
> 
> Doesn't help. Nor would I expect it to based on the documentation. It says
> nothing about effects on TAB-completion.

Sorry, I was getting mixed up between colons and equals. It does affect
the shell function based completion for colons. I expect Bart's answer
will be more helpful.

> Yes. The shell function based completion is slow (ever tried it on a
> 500MHz machine? That's what I'm writing this message on.) and more

I regularly use it on a NAS box that is probably slower (600MHz, low
power, ARM-based) and I've used it on much slower machines in the past.
You can speed it up by cutting down the number of completers and by
configuring it in areas where you find it slow. However I appreciate
that there's a lot of complexity to learn about to do that.

Oliver