Zsh Mailing List Archive
Messages sorted by:
Reverse Date,
Date,
Thread,
Author
Re: how to customize _all_matches use?
- X-seq: zsh-users 11573
- From: Bart Schaefer <schaefer@xxxxxxxxxxxxxxxx>
- To: zsh-users@xxxxxxxxxx
- Subject: Re: how to customize _all_matches use?
- Date: Sat, 16 Jun 2007 10:14:01 -0700
- In-reply-to: <4672f5ca.216f420a.0ac9.ffffac6fSMTPIN_ADDED@xxxxxxxxxxxxx>
- Mailing-list: contact zsh-users-help@xxxxxxxxxx; run by ezmlm
- References: <4672f5ca.216f420a.0ac9.ffffac6fSMTPIN_ADDED@xxxxxxxxxxxxx>
On Jun 15, 10:25pm, Vadim Zeitlin wrote:
}
} I'd like to have a key binding which would complete all matches which I
} would get by pressing TAB at once.
}
} % zstyle ':completion:all-matches:*' completer _all_matches
}
} However this doesn't seem to work
There's an important bit in the docs under _all_matches:
Note that this does not generate completions by itself. First use
any of the standard ways of generating a list of completions, then
use ^Xa to show all matches.
} I do get the behaviour I want if I do
}
} % zstyle ':completion:all-matches:*' completer _all_matches _complete
}
} but with one extremely annoying side-effect:
}
} % ls ^Xa
} % ls bar baz foo # this is the same line completed by the shell
} bar baz foo
} bar baz foo
}
} i.e. although I do get all the completions I also get *2* lists of them
} below.
Nope, that's only one list. It's a list containing (1) the single match
that was built by _all_matches containing all the other strings at once,
and (2 through 4) the individual matches generated by _complete.
The reason you're getting the listing is because four matches are an
ambiguous situation: zsh doesn't know which of the four you wanted. It
also inserts the first one because of MENU_COMPLETE (or at least that's
my best guess as to what's happening based on other things you wrote).
} Also, I'd like to not see anything at all if possible. So what am I
} doing wrong?
As far as I can tell, there's no simple way to get what you want, because
you must first generate matches before you can add another match that
consists of all the previous matches; which means you're guaranteed to
get an ambiguous completion (unless there's only one match from _complete
in the first place). The tag-order style doesn't help here because it's
checked inside _tags and hence inside _complete where the all-matches
tag doesn't yet exist.
So you'll have to play a game like _approximate does, where compadd is
redefined as a function to capture all the matches that other completers
attempt to add, and then don't really add them until the end. The idea
is to check for the -O/-A/-D options, and if none are present, then add
a -O option for your own array, and in eiher case call through to
"builtin compadd ...". Finally, in a comppostfunc, remove the compadd
function and call the real compadd with the string resulting from a
join of your array.
} For example, it would be very useful to have for "cvs diff" [...]
} so that pressing TAB after "cvs di" normally completes only the
} modified files. But pressing ^Xa also completes the sub-directories --
} is there any way I could avoid this?
Unfortunately there you're stuck by _cvs_modified_entries which adds
all the entries and subdirectories under the single tag "files". It has
nothing to do with _all_matches.
Messages sorted by:
Reverse Date,
Date,
Thread,
Author