Key-binding without minor mode!

Key-binding without minor mode!

[Phillip Lord]

I am trying to find what a keybinding would have been if a minor mode
had not been active, by switching the minor mode off with a let binding.
But I don't know how to find the keybinding. I have tried...

(key-binding (char-to-string last-command-event))

Works quite nicely keys such alphabetical keys, but fails for return 
which return a symbol.


I've also tried 

(key-binding (this-command-keys))
(key-binding (this-command-keys-vector))

and also sorts of other alternatives.

I'd appreciate any help!

Phil

Re: Key-binding without minor mode!

[Stefan Monnier]

> (key-binding (this-command-keys-vector))

This should work.


        Stefan

Re: Key-binding without minor mode!

[Phillip Lord]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

>> (key-binding (this-command-keys-vector))
>
> This should work.


Unfortunately, this fails for me.

This problem seems to be this. If I define keys as follows:

(define-key pabbrev-mode-map [tab] 'pabbrev-expand-maybe)


Then in lisp mode this looks up looks up the binding for
[tab] which returns nil, presumably because of this...

(define-key global-map "\t" 'indent-for-tab-command)

which binds "\t" and not [tab] even though the [tab] keybinding
overrides the "\t" one.

Previously I've dealt with this by checking for either, but this
hard codes [tab] as my keypress, so it can't be changed.

Phil

Re: Key-binding without minor mode!

[Stefan Monnier]

> This problem seems to be this. If I define keys as follows:
> (define-key pabbrev-mode-map [tab] 'pabbrev-expand-maybe)

If you don't want to distinguish between [tab] and "\t", then use "\t",
which will cover both.


        Stefan

Re: Key-binding without minor mode!

[Phillip Lord]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

>> This problem seems to be this. If I define keys as follows:
>> (define-key pabbrev-mode-map [tab] 'pabbrev-expand-maybe)
>
> If you don't want to distinguish between [tab] and "\t", then use "\t",
> which will cover both.


It's a nice idea, but it fails, unfortunate, because I am defining for a
minor mode. If the major mode uses `\t` then all is good. The minor-mode
will override the major. But if the major-mode defines binds to [tab]
then it will override the minor mode.

I think I have no choice but to handle tab explicitly. Is [tab] the only
key that works like this?

Phil

Re: Key-binding without minor mode!

[Stefan Monnier]

> But if the major-mode defines binds to [tab] then it will override the
> minor mode.

Indeed.  It might be a bug in the major-mode, tho (which manifests
for example under ttys where the TAB key sends a \t rather than a `tab').

> I think I have no choice but to handle tab explicitly. Is [tab] the only
> key that works like this?

No.  What you're trying to do is inherently tricky/fiddly/unreliable
because Emacs's key-binding infrastructure is not designed for that.

You might want to report a bug asking for some feature that lets you do
what you want (which I suspect is something like "let my command
delegate to the next command bound to this key").


        Stefan

Re: Key-binding without minor mode!

[Phillip Lord]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

>> But if the major-mode defines binds to [tab] then it will override the
>> minor mode.
>
> Indeed.  It might be a bug in the major-mode, tho (which manifests
> for example under ttys where the TAB key sends a \t rather than a `tab').

Oh, dear, yes. So binding to [tab] in my minor mode is a mistake. I need
to bind to both \t and [tab]. If I bind to just \t then a major mode can
override this by binding [tab]. If I bind, though, to just [tab] then I
have no binding in tty.


>
>> I think I have no choice but to handle tab explicitly. Is [tab] the only
>> key that works like this?
>
> No.  What you're trying to do is inherently tricky/fiddly/unreliable
> because Emacs's key-binding infrastructure is not designed for that.
>

Yep. I've ended up with this. I've had to do something similar for
return which does the same thing.


(defun pabbrev-get-previous-binding ()
  "Show the binding of tab if pabbrev were not active.
The command `pabbrev-show-previous-binding' prints this out."
  (let ((pabbrev-mode nil))
    (let ((tckv
           (if pabbrev-xemacs-p
               (this-command-keys)
             (this-command-keys-vector))))
      (cond
       ((or (equal tckv [tab]) 
            (equal tckv [9]))
        (or (key-binding [tab]) (key-binding "\t")))
       ((or (equal tckv [return])
            (equal tckv [32]))
        (or (key-binding [return]) (key-binding "\n")))
       (t
        (key-binding tckv))))))


> You might want to report a bug asking for some feature that lets you do
> what you want (which I suspect is something like "let my command
> delegate to the next command bound to this key").

Will do!

Phil

Re: Key-binding without minor mode!

[Stefan Monnier]

> I need to bind to both \t and [tab].

Preferably no.

> If I bind to just \t then a major mode can override this by binding [tab].

That'd be a bug in the major mode, which would be better to fix there.


        Stefan

Re: Key-binding without minor mode!

[Phillip Lord]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

>> I need to bind to both \t and [tab].
>
> Preferably no.
>
>> If I bind to just \t then a major mode can override this by binding [tab].
>
> That'd be a bug in the major mode, which would be better to fix there.

I am confused then. Should the major mode be binding to \t always then,
and never [tab]? In which case, what is the point of being able to bind
to [tab]?

Phil

Re: Key-binding without minor mode!

[Stefan Monnier]

> I am confused then. Should the major mode be binding to \t always then,
> and never [tab]? In which case, what is the point of being able to bind
> to [tab]?

The major mode should only bind to [tab] if it wants to behave
differently under a GUI than under a tty (which seems highly unlikely),
or if it does not care to work under a tty and wants to distinguish the
TAB key from the C-i key.


        Stefan

Re: Key-binding without minor mode!

[Phillip Lord]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

>> I am confused then. Should the major mode be binding to \t always then,
>> and never [tab]? In which case, what is the point of being able to bind
>> to [tab]?
>
> The major mode should only bind to [tab] if it wants to behave
> differently under a GUI than under a tty (which seems highly unlikely),
> or if it does not care to work under a tty and wants to distinguish the
> TAB key from the C-i key.



org does I think -- it binds tab, and C-i but not "\t".

(org-defkey org-mode-map "\C-i"       'org-cycle)
(org-defkey org-mode-map [(tab)]      'org-cycle)

Phil

Re: Key-binding without minor mode!

[Stefan Monnier]

> org does I think -- it binds tab, and C-i but not "\t".

\C-i is another way to write \t

> (org-defkey org-mode-map "\C-i"       'org-cycle)
> (org-defkey org-mode-map [(tab)]      'org-cycle)

The second binding looks like a misfeature to me.  You might want to
report it as a bug.


        Stefan

Re: Key-binding without minor mode!

[Phillip Lord]

Stefan Monnier <monnier@iro.umontreal.ca> writes:
>> org does I think -- it binds tab, and C-i but not "\t".
>
> \C-i is another way to write \t
>
>> (org-defkey org-mode-map "\C-i"       'org-cycle)
>> (org-defkey org-mode-map [(tab)]      'org-cycle)
>
> The second binding looks like a misfeature to me.  You might want to
> report it as a bug.


I'll find out why. They are obviously aware of it, as it breaks
yasnippet at least according to their faq. Although yasnippet binds:


(define-key map [(tab)] 'yas-expand) 
(define-key map (kbd "TAB") 'yas-expand)


I'm really glad that it wasn't just me that was confused!

Phil