completing history words

completing history words

[Oliver Kiddle]

I'm a little bit mystified by the way _history_complete_word is
implemented as it seems to re-implement menu-completion while not
offering non-menu completion. Why does it not just complete history
words and rely on separate completers based on different builtin
widgets (menu-complete, complete-word etc)? I wanted to be able to
complete history in a fairly normal way but it behaves in a slightly
strange way, doing things like inserting a match for ambiguous
completions.

I created a small function (_complete_history) containing just the
line:
compadd -Q -a 'historywords[2,-1]'

Then, I defined three completion widgets based on it:
zle -C history-complete complete-word _complete_history
zle -C history-menu menu-complete _complete_history
zle -C history-reverse-menu reverse-menu-complete _complete_history

As far as I can tell, the latter two behave largely like
_history_complete_older and _history_complete_newer. The first one
provides roughly the behaviour which I wanted.

I just needed a few key bindings and it was all working:

bindkey '\e,' history-complete
bindkey '\e.' history-menu
bindkey '\e/' history-reverse-menu
bindkey -M menuselect '\e,' complete-word
bindkey -M menuselect '\e.'  menu-complete
bindkey -M menuselect '\e/' reverse-menu-complete

It actually seems to have a problem with strange characters (e.g.
'$') not being considered part of the word being completed but other
than that, it is fine.

Obviously, my _complete_history function could be much improved with
styles for sorting and to manipulate compstate so listing etc can be
turned off but this way of doing it seems better to me. It is much
simpler and configuration of it will be more similar to configuring for
other completion widgets. Can anyone please explain why it is currently
done differently?

Oliver

Re: completing history words

[Sven Wischnowsky]

Oliver Kiddle wrote:

> I'm a little bit mystified by the way _history_complete_word is
> implemented as it seems to re-implement menu-completion while not
> offering non-menu completion. Why does it not just complete history
> words and rely on separate completers based on different builtin
> widgets (menu-complete, complete-word etc)? I wanted to be able to
> complete history in a fairly normal way but it behaves in a slightly
> strange way, doing things like inserting a match for ambiguous
> completions.

It's rather old...

> I created a small function (_complete_history) containing just the
> line:
> compadd -Q -a 'historywords[2,-1]'

Have you seen the _history completer?

> ...
> 
> 
> It actually seems to have a problem with strange characters (e.g.
> '$') not being considered part of the word being completed but other
> than that, it is fine.

See some recent mails, _history and _h_c_w put $IPREFIX into $PREFIX
for that.

> Obviously, my _complete_history function could be much improved with
> styles for sorting and to manipulate compstate so listing etc can be
> turned off but this way of doing it seems better to me. It is much
> simpler and configuration of it will be more similar to configuring for
> other completion widgets. Can anyone please explain why it is currently
> done differently?

Well, Adam will have to have the final word about this...

Bye
 Sven


--
Sven Wischnowsky                         wischnow@informatik.hu-berlin.de