read-char doesn't use input-decode-map

read-char doesn't use input-decode-map

[Rob Giardina]

Hi,

I've noticed that read-char doesn't use input-decode-map. I'm
doing some fun things with input-decode-map (code attached) to
convert all the right-side keys to the left using a super prefix
on OSX (allows one hand kdb, one hand mouse). It works nicely in
most cases except for the read-char family of functions. Is this
a bug or the expected behavior?

Thanks,
Rob

;;crazy experiment to mirror right characters onto left hand
(let ((from   "`12345qwertasdfgzxcvb~!@#$%QWERTASDFGZXCVB")
      (to     "-09876poiuy:lkjh/.,mn_)(*&^POIUY;LKJH/<>MN")
      (from-v (vconcat ["right" "left"] (mapcar 'char-to-string from)))
      (to-v   (vconcat [" "     " "]    (mapcar 'char-to-string to))))
  (map 'list
       #'(lambda(k to)
           (let* ((base-k (list 'super (intern k)))
                  (base-to (list (intern to))))
             (mapc
              #'(lambda(pre)
                  (define-key input-decode-map
                    (vector (append pre base-k))
                    (vector (append pre base-to))))
              '(() (control) (meta) (control meta)))))
       from-v
       to-v)

Re: read-char doesn't use input-decode-map

[Stefan Monnier]

> I've noticed that read-char doesn't use input-decode-map.

Indeed.  And changing that can introduce bugs elsewhere.
That's why Emacs-23.2 introduced read-key which obeys input-decode-map
and friends (to the extent possible).


        Stefan

Re: read-char doesn't use input-decode-map

[Rob Giardina]

Stefan Monnier <monnier <at> IRO.UMontreal.CA> writes:
 
> > I've noticed that read-char doesn't use input-decode-map.
> 
> Indeed.  And changing that can introduce bugs elsewhere.
> That's why Emacs-23.2 introduced read-key which obeys input-decode-map
> and friends (to the extent possible).

Wow, read-key is pretty wild. I didn't realize you could do that
with idle timers.  Perhaps that's vanilla if you're used to
emacs' event handling :)

This apparently means that code like I attached has no way to
reliably translate all of a certain key into another key, at
least for existing code that uses read-char.  For example,
y-or-n-p.

-Rob

Re: read-char doesn't use input-decode-map

[Stefan Monnier]

> Wow, read-key is pretty wild. I didn't realize you could do that
> with idle timers.  Perhaps that's vanilla if you're used to
> emacs' event handling :)

It's definitely not vanilla: it took me a while to find something that
worked OK, and even the current solution is pretty hackish since it
relies on time to detect "key press boundaries", which is fundamentally
unreliable.  Sadly, I think there is fundamentally no reliable way to
solve this problem, so it's probably about as good as it gets.

> This apparently means that code like I attached has no way to
> reliably translate all of a certain key into another key, at
> least for existing code that uses read-char.  For example,
> y-or-n-p.

Using read-char makes it pretty much hopeless, yes.
Most uses of reac-char should be replaced by read-key instead.


        Stefan

Re: read-char doesn't use input-decode-map

[Lennart Borgman]

On Thu, Dec 31, 2009 at 3:51 AM, Stefan Monnier
<monnier@iro.umontreal.ca> wrote:
>
> Most uses of reac-char should be replaced by read-key instead.


Then it is perhaps the time to use the prompt face for read-key too?
It is a bit confusing now. I suggest something like:

        (progn
	  (use-global-map read-key-empty-map)
          (message (concat (apply 'propertize prompt (member 'face
minibuffer-prompt-properties))
                           (propertize " " 'face 'cursor)))
	  (aref (catch 'read-key (read-key-sequence-vector nil nil t)) 0))

Re: read-char doesn't use input-decode-map

[Stefan Monnier]

>> Most uses of reac-char should be replaced by read-key instead.
> Then it is perhaps the time to use the prompt face for read-key too?

I guess so, yes.  But if so, the problem should probably be fixed
directly in read-key-sequence(-vector) rather than only in read-key.


        Stefan

Re: read-char doesn't use input-decode-map

[Lennart Borgman]

On Mon, Jan 4, 2010 at 3:18 AM, Stefan Monnier <monnier@iro.umontreal.ca> wrote:
>>> Most uses of reac-char should be replaced by read-key instead.
>> Then it is perhaps the time to use the prompt face for read-key too?
>
> I guess so, yes.  But if so, the problem should probably be fixed
> directly in read-key-sequence(-vector) rather than only in read-key.


Ok, there should be code for this in y-or-n-p for example. (If anyone
has time to write it right now.)