bug#5849: 23.1.95; completion-auto-help blocks icomplete-mode

bug#5849: 23.1.95; completion-auto-help blocks icomplete-mode

[Leo]

1. Emacs -q -nw
2. (icomplete-mode t)
3. (setq completion-auto-help 'lazy)
4. M-x w TAB

The message from completion [Next char not unique] will block the hint
from icomplete mode.

Leo

bug#5849: 23.1.95; completion-auto-help blocks icomplete-mode

[Chong Yidong]

> 1. Emacs -q -nw
> 2. (icomplete-mode t)
> 3. (setq completion-auto-help 'lazy)
> 4. M-x w TAB
>
> The message from completion [Next char not unique] will block the hint
> from icomplete mode.

Maybe completion-auto-help should do nothing if icomplete-mode is on.
What do you think?

bug#5849: 23.1.95; completion-auto-help blocks icomplete-mode

[Drew Adams]

> > 1. Emacs -q -nw
> > 2. (icomplete-mode t)
> > 3. (setq completion-auto-help 'lazy)
> > 4. M-x w TAB
> >
> > The message from completion [Next char not unique] will 
> > block the hint from icomplete mode.
> 
> Maybe completion-auto-help should do nothing if icomplete-mode is on.
> What do you think?

That would be a terrible idea. That would mean not showing *Completions*.

If the value is `lazy' and you see [Next char not unique], that message just
lets you know that there are completions; hitting TAB a second time then shows
*Completions*.

If your suggestion were to just not show the message [Next char not unique] when
icomplete-mode is on, that would be different. In that case, OK. Yes, that
message is not really needed if icomplete-mode is showing possible completions.

But we certainly do not want to suppress display of *Completions* just because
icomplete-mode is on. That would be crazy.

bug#5849: 23.1.95; completion-auto-help blocks icomplete-mode

[Chong Yidong]

"Drew Adams" <drew.adams@oracle.com> writes:

> If your suggestion were to just not show the message [Next char not
> unique] when icomplete-mode is on

Yes, obviously this is what I meant.  Like this:

=== modified file 'lisp/minibuffer.el'
*** lisp/minibuffer.el	2010-03-24 18:02:56 +0000
--- lisp/minibuffer.el	2010-04-10 18:49:41 +0000
***************
*** 528,536 ****
                ;; Show the completion table, if requested.
                (cond
                 ((not exact)
!                 (if (case completion-auto-help
!                       (lazy (eq this-command last-command))
!                       (t completion-auto-help))
                      (minibuffer-completion-help)
                    (minibuffer-message "Next char not unique")))
                 ;; If the last exact completion and this one were the same, it
--- 528,537 ----
                ;; Show the completion table, if requested.
                (cond
                 ((not exact)
!                 (if (cond (icomplete-mode t)
! 			  ((eq completion-auto-help 'lazy)
! 			   (eq this-command last-command))
! 			  (t completion-auto-help))
                      (minibuffer-completion-help)
                    (minibuffer-message "Next char not unique")))
                 ;; If the last exact completion and this one were the same, it

bug#5849: 23.1.95; completion-auto-help blocks icomplete-mode

[Drew Adams]

> > If your suggestion were to just not show the message
> > [Next char not unique] when icomplete-mode is on
> 
> Yes, obviously this is what I meant.  Like this:
> 
> (if (cond (icomplete-mode t)
>           ((eq completion-auto-help 'lazy)
>            (eq this-command last-command))
>           (t completion-auto-help))
>     (minibuffer-completion-help)
>   (minibuffer-message "Next char not unique")))

Yes, that's OK by me.
But either of these (equivalent) forms is clearer, IMO:

(if (or icomplete-mode
        (and completion-auto-help
             (or (not (eq completion-auto-help 'lazy))
                 (eq this-command last-command))))
    (minibuffer-completion-help)
  (minibuffer-message "Next char not unique"))

(if (or icomplete-mode
        (if (eq completion-auto-help 'lazy)
            (eq this-command last-command)
          completion-auto-help))
    (minibuffer-completion-help)
  (minibuffer-message "Next char not unique"))

---

And if we didn't care whether a non-t, non-`lazy', non-nil value calls
`minibuffer-completion-help', then this (not equivalent) would be OK too:

(if (or icomplete-mode
        (eq t completion-auto-help)
        (and (eq completion-auto-help 'lazy)
             (eq this-command last-command)))
    (minibuffer-completion-help)
  (minibuffer-message "Next char not unique"))

bug#5849: 23.1.95; completion-auto-help blocks icomplete-mode

[Leo]

On 2010-04-10 19:51 +0100, Chong Yidong wrote:
>> If your suggestion were to just not show the message [Next char not
>> unique] when icomplete-mode is on
>
> Yes, obviously this is what I meant.

The behaviour described is desirable. But it seems it will be better to
make minibuffer completion code more flexible instead of special-case
icomplete.

Leo

bug#5849: 23.1.95; completion-auto-help blocks icomplete-mode

[Leo]

On 2010-04-10 19:51 +0100, Chong Yidong wrote:
> "Drew Adams" <drew.adams@oracle.com> writes:
>
>> If your suggestion were to just not show the message [Next char not
>> unique] when icomplete-mode is on
>
> Yes, obviously this is what I meant.  Like this:
[...]

I wonder if it makes sense to always have the message from completion
append to the text in the minibuffer. For example, in this case, the
icomplete hint will go first and the completion message last.

This would seem to fix another problem seen in M-:. When users try to
complete while in minibuffer for example when using M-:, it switches to
show the completion message and blocks users from seeing the text they
are editing in the minibuffer.

Leo

bug#5849: 23.1.95; completion-auto-help blocks icomplete-mode

[Stefan Monnier]

> Yes, obviously this is what I meant.  Like this:

> === modified file 'lisp/minibuffer.el'
> *** lisp/minibuffer.el	2010-03-24 18:02:56 +0000
> --- lisp/minibuffer.el	2010-04-10 18:49:41 +0000
> ***************
> *** 528,536 ****
>                 ;; Show the completion table, if requested.
>                 (cond
>                  ((not exact)
> !                 (if (case completion-auto-help
> !                       (lazy (eq this-command last-command))
> !                       (t completion-auto-help))
>                       (minibuffer-completion-help)
>                     (minibuffer-message "Next char not unique")))
>                  ;; If the last exact completion and this one were the same, it
> --- 528,537 ----
>                 ;; Show the completion table, if requested.
>                 (cond
>                  ((not exact)
> !                 (if (cond (icomplete-mode t)
> ! 			  ((eq completion-auto-help 'lazy)
> ! 			   (eq this-command last-command))
> ! 			  (t completion-auto-help))
>                       (minibuffer-completion-help)
>                     (minibuffer-message "Next char not unique")))
>                  ;; If the last exact completion and this one were the same, it

It's a good workaround, thanks.  Note that most other messages from the
completion code are similarly problematic, I think.  So maybe we should
have a buffer-local completion-inhibit-messages variable, which
icomplete could set in the minibuffer.


        Stefan

bug#5849: 23.1.95; completion-auto-help blocks icomplete-mode

[Chong Yidong]

I have checked in the patch that I posted earlier for this bug into the
emacs-23 branch, and a more complete fix into the trunk.  The latter
involves a new variable completion-show-inline-help, which
icomplete-mode binds to nil.

bug#5849: 23.1.95; completion-auto-help blocks icomplete-mode

[Johan Bockgård]

Chong Yidong <cyd@stupidchicken.com> writes:

> I have checked in the patch that I posted earlier for this bug into the
> emacs-23 branch, and a more complete fix into the trunk.  The latter
> involves a new variable completion-show-inline-help, which
> icomplete-mode binds to nil.

The fix in the trunk sets completion-show-inline-help globally which is
not quite right:

emacs -Q
M-x icomplete-mode RET
C-x C-f nonexistingfile TAB
=> no message
Expected: "[No match]"



This patch to the current trunk sets completion-show-inline-help locally
when icomplete is actually being used (simple completion).


2011-07-09  Johan Bockgård  <bojohan@gnu.org>

        Fix previous fix of Bug#5849.

	* icomplete.el (icomplete-mode): Don't set
	completion-show-inline-help.
	(icomplete-minibuffer-setup): Set completion-show-inline-help
	locally during icompletion.


=== modified file 'lisp/icomplete.el'
--- lisp/icomplete.el	2011-05-31 03:03:38 +0000
+++ lisp/icomplete.el	2011-07-09 14:01:48 +0000
@@ -179,11 +179,8 @@
   (if icomplete-mode
       ;; The following is not really necessary after first time -
       ;; no great loss.
-      (progn
-	(setq completion-show-inline-help nil)
-	(add-hook 'minibuffer-setup-hook 'icomplete-minibuffer-setup))
-    (remove-hook 'minibuffer-setup-hook 'icomplete-minibuffer-setup)
-    (setq completion-show-inline-help t)))
+      (add-hook 'minibuffer-setup-hook 'icomplete-minibuffer-setup)
+    (remove-hook 'minibuffer-setup-hook 'icomplete-minibuffer-setup)))
 
 ;;;_ > icomplete-simple-completing-p ()
 (defun icomplete-simple-completing-p ()
@@ -209,6 +206,7 @@
   "Run in minibuffer on activation to establish incremental completion.
 Usually run by inclusion in `minibuffer-setup-hook'."
   (when (and icomplete-mode (icomplete-simple-completing-p))
+    (set (make-local-variable 'completion-show-inline-help) nil)
     (add-hook 'pre-command-hook
 	      (lambda () (run-hooks 'icomplete-pre-command-hook))
 	      nil t)

bug#5849: 23.1.95; completion-auto-help blocks icomplete-mode

[Chong Yidong]

Johan Bockgård <bojohan@gnu.org> writes:

> The fix in the trunk sets completion-show-inline-help globally which is
> not quite right:
>
> emacs -Q
> M-x icomplete-mode RET
> C-x C-f nonexistingfile TAB
> => no message
> Expected: "[No match]"
>
> This patch to the current trunk sets completion-show-inline-help locally
> when icomplete is actually being used (simple completion).

Thanks, committed.