Re: compconfig[list_condition]

[Sven Wischnowsky <wischnow@xxxxxxxxxxxxxxxxxxxxxxx>]

[ List changed to -workers, I /think/ this is more appropriate. If you 
don't agree, forward it to -users... ]

Bart Schaefer wrote:

> I'm still fooling around with insert-and-predict.  In its present state, it
> attempts regular completion on any self-inserted key (other than space) that
> does not result in a history-beginning-search match.  With autolist set,
> however, this tends to produce things like this:
> 
> zagzig[28] echo $
> zsh: do you wish to see all 128 possibilities? 
> 
> This means I have to type an extra character, which if I'm typing quickly
> enough I fail to do, causing a character that I wanted to have on the line
> to be consumed instead.  This was previously discussed (in zsh-users/2641)
> and the result seems to have been the addition of compstate[list_lines] and
> BUFFERLINES (mis-documented as BLINES, by the way) in 8227.

Oops. I first used `BLINES' and then thought someone might complain
that it's too short or not different enough from `LINES'. The patch
below fixes this.

> So I've added
> 
> compconfig[list_condition]='compstate[list_lines]+BUFFERLINES < LINES'
> compconfig[completer]='_list:_complete'
> 
> to my completion configuration, and set LISTMAX very large.  However, this
> doesn't do what I thought it would.  I tried reversing the order of the
> completers, and setting and unsetting autolist, but either I get listings
> that scroll off the screen or I don't get any listings at all.

The list completer does something entirely different (I guess nobody
uses it, btw). I was thinking about changing it so that it would
support this kind of stuff, but it may be better to add another
completer for this, because: for its real purpose it has to be called
before `_complete' and such (because we have to make sure that it *is* 
called). But this decide-if-we-should-list stuff has to be called at
the end, because, of course, we can only calculate the number of lines 
needed *after* all matches have been added. That's what causing you
problems, here, `$[compstate[list_lines]+BUFFERLINES < LINES]' is
always true because there are no matches yet.

> What *should* I be doing to get listings if and only if they fit on the
> screen?

In the same patch that brought you `list_lines' (or was it one patch
after that?), I added the post-completion functions which are
guarenteed to be executed after the completers have been run. Just
have a look at `incremental-complete-word'. It adds a simple function:

icw-list-helper() {

  # +1 for the status line we will add...

  if [[ compstate[list_lines]+BUFFERLINES+1 -gt LINES ]]; then
    compstate[list]='list explanations'
    if [[ compstate[list_lines]+BUFFERLINES+1 -gt LINES ]]; then
      compstate[list]=''
      compstate[force_list]=yes
    fi
    toolong='...'
  fi
}

and starts completion with (the equivalent of):

  comppostfuncs=( icw-list-helper )
  zle $wid "$@"

The helper function can decide if the list fits on the screen and
shows it only then. The `toolong' is used in the i-c-w prompt (so you
won't need it) and the `+1' is needed to calculate the number of lines 
needed including the status line used by i-c-w (so you won't need it
either).

And remember to set `comppostfuncs' before every call to the
completion system, because it will be reset there.

Ok?

Bye
 Sven


--
Sven Wischnowsky                         wischnow@xxxxxxxxxxxxxxxxxxxxxxx