Re: Fwd: completion-ui.el question: my vimpulse-show-completion-menu function always gives me an empty menu

[Toby Cubitt <t.s.cubitt.98@cantab.net>]

Jason Spiro wrote:
> 2007/4/13, Toby Cubitt <tsc25@cantab.net> wrote:
 >> (defun vimpulse-try-hippie-expand-listed-functions
 >>        (prefix &optional maxnum)
 >> ;  (interactive "*MString to expand:
 >> ;nMaximum number of expansions, or nil for unlimited: ")
 >> (setq expansions nil)
 >> (with-temp-buffer
 >>   (insert-string prefix)
 >>   (hippie-expand nil)
 >>   (unless (string-equal (current-message) "No expansion found")
 >>     (while (and
 >>             (not (string-equal
 >>                   (current-message)
 >>                   "No further expansions found"))
 >>             (if maxnum (not (= (length expansions) maxnum))))
 >>       ;; Hippie-expand was designed to be run interactively.  If
 >>       ;; this-command and last-command are equal, it assumes it's
 >>       ;; being called as a "repeated use" (additional call without
 >>       ;; moving the point) and searches for additional expansions
 >>       ;; with the same prefix.  So, make this-command and
 >>       ;; last-command equal.
 >>       (setq this-command last-command)
 >>       (hippie-expand nil)
 >>       (add-to-list 'expansions (buffer-string))
 >>       (buffer-string)
 >>     ))
 >>   expansions))

 > This is almost OK. Completion-UI expects the completion-function to
 > return a list of completion strings *without* the prefix. So you need
 > to change "expansions" in the last line to:
 >   (mapcar (lambda (str) (substring str 1)) expansions)

Sorry, this isn't quite correct. It should of course be:

   (mapcar (lambda (str) (substring str (length prefix))) expansions)


>> How exactly does this multiple-word completion work? I don't follow.
> 
> For example, type:  The quick brown fox jumped over the lazy dog. My
> favorite foM-/ SPC M-/ SPC M-/ SPC M-/ SPC M-/ SPC M-/
> 
> It should write: The quick brown fox jumped over the lazy dog. My
> favorite fox jumped over the lazy dog"

I see. This relies on the fact that the first dabbrev expansion is the 
one that is closest in the buffer. If you're completion function (or 
rather, hippie-expand's) returned the completions in this order, you 
could get the same behaviour using completion-UI. My personal favourite 
is to enable 'completion-use-dynamic', so that you don't even need to 
type the M-/ equivalent (M-TAB).

Toby