Zsh Mailing List Archive
Messages sorted by:
Re: Completion for gtar
- X-seq: zsh-users 1757
- From: Bruce Stephens <b.stephens@xxxxxxxxx>
- To: zsh-users@xxxxxxxxxxxxxxx
- Subject: Re: Completion for gtar
- Date: 26 Aug 1998 11:12:31 +0100
- In-reply-to: "Bart Schaefer"'s message of "Tue, 25 Aug 1998 10:48:31 -0700"
- References: <vbd89pnkp3.fsf@xxxxxxxxxxxxxxx> <980825104831.ZM15886@xxxxxxxxxxxxxxxxxxxxxxx>
- Sender: B.Stephens@xxxxxxxxx
"Bart Schaefer" <schaefer@xxxxxxxxxxxxxxxx> writes:
> Have you looked at PWS's tar completion in Misc/compctl-examples in
> 3.1.4? It even has a function to extract completable file names
> from the tar file.
No, I hadn't. Partly deliberately---this was as much an exercise as
anything. Now I look at it, and the two look reassuringly similar,
although PWS's provides more.
> You can just do
> compctl -f -x 'p' -k taropts -- gtar
> unless you want to unset taropts and not have it hanging around.
Yes. Or rather, I didn't want to rely on taropts sticking around.
> I don't think that means what you think it means. If you put the "-"
> in the list of -g patterns, then it's treated as a file name. It just
> so happens that, because it has no wildcards, it never gets dropped
> from the list of completions. So you'll find that you can complete
> e.g. like this:
> zsh% gtar zxf ~/<TAB>
> zsh% gtar zxf ~/-
> even when there's no file named "-" in your home dir.
That's clearly not what I want. You're right, there's no advantage in
having - complete to itself, so I ought to remove it.
> compctl -f -x 'p' -k taropts \
> - 'p W[1,*z*]' -/g '*.tar.(gz|Z)' -k '(-)' \
> -- gtar
> BTW, don't you want to add W[1,*f*] to that? You don't want to complete
> file names if there's no "f" option.
I don't care about not having the f option---I never use gtar except
on files. (Backups and so on are done differently, and by somebody
> That is what alternation is about, but alternation works differently.
> The -x branches are short-circuited on whether their patterns match, so
> once you take a -x branch you can never see completions from any of the
> other branches, even if no completions result from the branch where the
> patterns matched.
> Alternation with "+", on the other hand, short-circuits on whether the
> list of possible completions is empty, so it'll keep trying until it
> runs out of alternatives or finds one that produces something.
OK. I don't think this is clear from the documentation. Not as clear
as the two paragraphs you've just given, anyway.
Messages sorted by: