Binding C-i breaks TAB bindings (was bug#13861)

Binding C-i breaks TAB bindings (was bug#13861)

[Ivan Kanis]

March, 03 at 14:27 Ivan wrote:

> emacs -Q
> in the scratch buffer type: 
> 
> (global-set-key [?\C-i] 'jumpc-jump-forward)
> 
> press the TAB key. 
> 
> I get the following error: 

> command-execute: Symbol's function definition is void:
>  jumpc-jump-forward

March, 03 at 18:53 Glenn wrote:

> This is explicitly explained in the "Function Keys" section of the ELisp
> manual.

Hi Glenn,

I just read the manual at the given section. I see that <tab> is the
name of the corresponding Lisp symbols [1]. 

[?\C-i] is a vector, surely it shouldn't override <tab> ?

Ivan

[1] I don't see it in the variables, is it a special name space?

Re: Binding C-i breaks TAB bindings (was bug#13861)

[Stefan Monnier]

> [?\C-i] is a vector, surely it shouldn't override <tab> ?

The fact that it's a vector is irrelevant, because you're comparing two
different notations: either compare [?\C-i] with [tab], or compare
"C-i" with "<tab>".

In any case <tab> is not overridden.  It's just that you don't have any
binding for <tab> so function-key-map makes Emacs fallback on the
C-i binding.


        Stefan

Re: Binding C-i breaks TAB bindings (was bug#13861)

[Ivan Kanis]

March, 04 at 9:28 Stefan wrote:

> In any case <tab> is not overridden.  It's just that you don't have any
> binding for <tab> so function-key-map makes Emacs fallback on the
> C-i binding.

OK, I think I finally understand.

C-h k TAB => TAB (translated from <tab>) runs the command ...

I missed the 'translated from' bit.

I am trying to free up C-i for my nefarious purpose. Does that mean I
have to fix each mode I use to bind the <tab> key to the proper command?
Is there a programmatic way?
-- 
Art washes from the soul the dust of everyday life.
    -- Pablo Picasso

Re: Binding C-i breaks TAB bindings (was bug#13861)

[Ivan Andrus]

On Mar 4, 2013, at 10:59 AM, Ivan Kanis <banana@kanis.fr> wrote:

> March, 04 at 9:28 Stefan wrote:
> 
>> In any case <tab> is not overridden.  It's just that you don't have any
>> binding for <tab> so function-key-map makes Emacs fallback on the
>> C-i binding.
> 
> OK, I think I finally understand.
> 
> C-h k TAB => TAB (translated from <tab>) runs the command ...
> 
> I missed the 'translated from' bit.
> 
> I am trying to free up C-i for my nefarious purpose. Does that mean I
> have to fix each mode I use to bind the <tab> key to the proper command?
> Is there a programmatic way?

The way I do it is by creating a minor-mode (called `gvol-mode'), and bind C-i to whatever I want.  Then I bind <tab> to the function below which turns off gvol-mode and finds the "best" keybinding and runs it.  It's not perfect, and you sometimes run into problems that are somewhat difficult to diagnose which I have worked around with the setq's.

HTH,
Ivan

(defun gvol-indent-for-tab-command ()
  "This is to fix `indent-for-tab-command' for `gvol-mode'.
It runs [tab] or C-i with `gvol-mode' nil because `gvol-mode'
binds C-i to a different command.  Ideally this should take into
account window system so that it can DTRT in a terminal (whatever
the right thing is)."
  (interactive)
  (let* ((gvol-mode nil)
         (command (or (key-binding [tab])
                      (key-binding "\C-i"))))
    ;; This is to satisfy `python-indent-line' which checks
    ;; `this-command' to cycle
    (setq this-command 'indent-for-tab-command)
    ;; Make people think this was called with C-i.  This allows
    ;; `self-insert-command' to work
    (setq last-command-event 9)
    (call-interactively command)))

Re: Binding C-i breaks TAB bindings (was bug#13861)

[Stefan Monnier]

>> In any case <tab> is not overridden.  It's just that you don't have any
>> binding for <tab> so function-key-map makes Emacs fallback on the
>> C-i binding.
> I am trying to free up C-i for my nefarious purpose.

In a tty, TAB and C-i can't be distinguished (at least, most/all
terminal emulators will send the exact same byte sequences in both cases
by default), so there isn't much you can do there.
But in a GUI session, you could add something like

   (define-key input-decode-map [?\C-i] [control-i])

This way, `tab' will work the same as ever, but C-i will now generate
a new event `control-i' which you can bind to any command you like.


        Stefan