identifying meta-key bound functions from last-command-char?

identifying meta-key bound functions from last-command-char?

[Ken Manheimer]

[-- Attachment #1: Type: text/plain, Size: 1223 bytes --]

in a recent emacs build a technique i've been using to deliberately fetch
key bindings according to the last-command-char is failing for meta-modified
keys.  the problem is that, for meta-modified keys,

  (characterp last-command-char) => nil

this breaks an approach like:

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

i need some way to fetch bindings according to last-command-char for a
special key handler that i use in overlays and, (for example) in
allout.elallout-hotspot-key-handler, a pre-command-hook function.
though probably
overly complicated, it has worked through the released emacs 22.  in the
recent cvs build (23.0.60), i get an error:

  wrong-type-argument characterp 134217760

(that's for "\M- " meta space, but the same things happens for other meta
keys.)  i'm failing to figure out how to translate the meta-modified key to
something that i can use for key-binding, and presume i'm missing something
obvious.  can someone steer me in the right direction?

(incidentally, i've  been unable to build todays HEAD checkout throughout
the day:
  doc-view.el:770:43:Error: Wrong number of arguments: #[nil "\300\207"
[(image-mode-window-get (quote page))] 1], 1
)
-- 
ken
http://myriadicity.net

[-- Attachment #2: Type: text/html, Size: 1439 bytes --]

Re: identifying meta-key bound functions from last-command-char?

[Stefan Monnier]

> this breaks an approach like:

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

Try (key-binding (vector last-command-char))


        Stefan

Re: identifying meta-key bound functions from last-command-char?

[Kenichi Handa]

In article <2cd46e7f0802251630h2cc2ae5di1d77ed3b87611416@mail.gmail.com>, "Ken Manheimer" <ken.manheimer@gmail.com> writes:

> [1  <text/plain; ISO-8859-1 (7bit)>]
> in a recent emacs build a technique i've been using to deliberately fetch
> key bindings according to the last-command-char is failing for meta-modified
> keys.  the problem is that, for meta-modified keys,

>   (characterp last-command-char) => nil

> this breaks an approach like:

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

> i need some way to fetch bindings according to last-command-char for a
> special key handler that i use in overlays and, (for example) in
> allout.elallout-hotspot-key-handler, a pre-command-hook function.
> though probably
> overly complicated, it has worked through the released emacs 22.  in the
> recent cvs build (23.0.60), i get an error:

>   wrong-type-argument characterp 134217760

> (that's for "\M- " meta space, but the same things happens for other meta
> keys.)  i'm failing to figure out how to translate the meta-modified key to
> something that i can use for key-binding, and presume i'm missing something
> obvious.  can someone steer me in the right direction?

Unfortunately, I think Emacs doesn't have a clear view about
how to handle an integer which contains a character code and
meta-modifier masks.  Is it a character or not?

I think it's not a character but an event object, so the
above code should be:
  (key-binding (vector last-command-char))

The related problem is when to change, for instance, (' ' |
CTRL_MASK) to '\0'.  Now Emacs signals an error when we type
C-q C-SPC.

---
Kenichi Handa
handa@ni.aist.go.jp

Re: identifying meta-key bound functions from last-command-char?

[Stefan Monnier]

> The related problem is when to change, for instance, (' ' |
> CTRL_MASK) to '\0'.

Traditionally Emacs has mapped CTRL_MASK + CHAR to the corresponding
ASCII control char, if it exists.  Looks like this mapping is now
missing somewhere.

It's not like this mapping is clear: e.g. the distinction between C-a
and C-A.


        Stefan

Re: identifying meta-key bound functions from last-command-char?

[Kenichi Handa]

In article <jwvhcfwih53.fsf-monnier+emacs@gnu.org>, Stefan Monnier <monnier@iro.umontreal.ca> writes:

> > The related problem is when to change, for instance, (' ' |
> > CTRL_MASK) to '\0'.

> Traditionally Emacs has mapped CTRL_MASK + CHAR to the corresponding
> ASCII control char, if it exists.  Looks like this mapping is now
> missing somewhere.

The routine doing that mapping still exists; char_string().
But, before reaching that function, the macro
CHECK_CHARACTER rejects such an integer value.

Should we implement something like
CHECK_CHARACTER_RESOLVE_MODIFER and use it instead of
CHECK_CHARACTER at proper places?

---
Kenichi Handa
handa@ni.aist.go.jp

Re: identifying meta-key bound functions from last-command-char?

[Stefan Monnier]

>> > The related problem is when to change, for instance, (' ' |
>> > CTRL_MASK) to '\0'.

>> Traditionally Emacs has mapped CTRL_MASK + CHAR to the corresponding
>> ASCII control char, if it exists.  Looks like this mapping is now
>> missing somewhere.

> The routine doing that mapping still exists; char_string().

The question is: where is it called from and is there a place where it
should be called but isn't (or is, but too late)?

> But, before reaching that function, the macro
> CHECK_CHARACTER rejects such an integer value.

> Should we implement something like
> CHECK_CHARACTER_RESOLVE_MODIFER and use it instead of
> CHECK_CHARACTER at proper places?

Maybe, or rather move the CHECK_CHARACTER to some later time when the
mapping has taken place.


        Stefan

Re: identifying meta-key bound functions from last-command-char?

[Miles Bader]

"Ken Manheimer" <ken.manheimer@gmail.com> writes:
> the problem is that, for meta-modified keys,
>
>   (characterp last-command-char) => nil
>
> this breaks an approach like:
>
>   (key-binding (char-to-string last-command-char))

It wants a vector, so give it a vector:

  (key-binding (vector last-command-char))

-Miles

-- 
Discriminate, v.i. To note the particulars in which one person or thing is,
if possible, more objectionable than another.

Re: identifying meta-key bound functions from last-command-char?

[Ken Manheimer]

using 'vector' instead of char-to-string solved my problem - thanks, all!
-- 
ken
http://myriadicity.net

Re: identifying meta-key bound functions from last-command-char?

[Stefan Monnier]

> using 'vector' instead of char-to-string solved my problem - thanks, all!

Using strings for "sequence of events" is a hack that happened to work
back in the tty days and was made obsolete (tho sadly still supported)
when X11 support was added.


        Stefan

Re: identifying meta-key bound functions from last-command-char?

[Ken Manheimer]

On 2/26/08, Stefan Monnier <monnier@iro.umontreal.ca> wrote:
> > using 'vector' instead of char-to-string solved my problem - thanks, all!
>
> Using strings for "sequence of events" is a hack that happened to work
>  back in the tty days and was made obsolete (tho sadly still supported)
>  when X11 support was added.

the closest lead i found in the elisp info manual was 21.6.14 _Putting
Keyboard Events in Strings_, and it distinctly does not address my
need.  i think the details are missing from the manual, and that could
be repaired.

as i see it, the purpose is not too obscure or unusual: deliberately
identifying a key binding for user-entered keys from within a handler
that's intervening in the command loop, eg, on pre-command-hook,
overlay modification-hooks, or something like that.  last-command-char
/ last-command-event is the only data structure i found for getting
the input, and the only details about using that are very brief, in
21.4, _Information from the Command Loop_.

i expect that the details - using `(keyboard-binding (vector
last-command-char))' - belong somewhere in the 21.6 _Input Events_
section, perhaps in 21.6.1 _Keyboard Events_?  there could then be a
footnote / link from 21.6.14 _Putting Keyboard Events in Strings_ and
the entry for last-command-char / -event in 21.4.

the text could go below what currently exists in 21.6.1.  i don't have
enough grasp of input events to be complete, but here's a description
of the context i do know:

  To handle command-loop input from a hook function, like a pre-command-hook
  or modification-hooks entry, you can get the input event from
last-command-event
  (see _Keyboard Events_).  You can identify a keybinding for that input using:

    ``(keyboard-binding (vector last-command-event))''

can anyone help flesh this out?  i would have spent a lot less time
thrashing if the info were easier to find.  (it still may make sense
to repair char-to-string, though.)
--
ken
http://myriadicity.net

Re: identifying meta-key bound functions from last-command-char?

[Stefan Monnier]

> as i see it, the purpose is not too obscure or unusual: deliberately
> identifying a key binding for user-entered keys from within a handler
> that's intervening in the command loop, eg, on pre-command-hook,
> overlay modification-hooks, or something like that.  last-command-char
> / last-command-event is the only data structure i found for getting
> the input, and the only details about using that are very brief, in
> 21.4, _Information from the Command Loop_.

So maybe you'd rather use (key-binding (this-command-keys-vector)) so it
also works for commands bound to multi-key sequences.


        Stefan