Re: Default arg for `T m'?

[Lars Magne Ingebrigtsen <larsi@gnus.org>]

Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:

> So we don't use the INITIAL-INPUT argument of completing-read;
> instead, we use the DEF argument and also put the default in the
> history.

Right.  I've now modified `T m' to use the function below.  I think
the "normal" `completing-read' is rather to horrible to work with,
since you have to fiddle so much to get it to work in a sensible way.
But I may have overlooked something.

(defun gnus-completing-read (prompt table &optional predicate require-match
				    history inherit-input-method)
  (when (and history
	     (not (boundp history)))
    (set history nil))
  (completing-read
   (if (symbol-value history)
       (concat prompt " (" (car (symbol-value history)) "): ")
     (concat prompt ": "))
   table
   predicate
   require-match
   nil
   history
   (car (symbol-value history))
   inherit-input-method))

And used thusly:

(gnus-completing-read "Move to topic" gnus-topic-alist
                      nil t 'gnus-topic-history)

Any comments?  And anybody volunteering to go through the Gnus sources
and converting all calls to `completing-read' to
`gnus-completing-read' to get Emacs-compliant default prompt value
behavior?
                                            
-- 
(domestic pets only, the antidote for overdose, milk.)
   larsi@gnus.org * Lars Magne Ingebrigtsen