repeatable vs. non-repeatable commands

repeatable vs. non-repeatable commands

[Michael Heerdegen]

Hi,

in el-search, the default key bindings are all of the form
control-shift-letter, like C-S (search-forward), C-R (search-backward),
C-A (go to the first match) etc.

Since these bindings are not available in a console, I want to provide
another configurable binding-scheme of the form prefix+letter, where
"prefix" is a prefix key the user can choose.  In this case, I want to
equip the commands with a transient map so that the prefix has to be
given only the first time one of the commands is invoked.  For example,
if the prefix is meta-s e,

   meta-s e s s a

should invoke a forward search, then go to the next match, and then to
the first.

The transient map is the same for all prefixes, but in the shift-letter
case, it's wrong to use the transient map; it would be surprising when
C-S followed by r would search backward instead of stopping the search
and invoking the self-insert-command.

My question is: is there a way I'm not seeing to achieve this without
defining two sets of commands (one set establishing the transient map,
another without)?

FWIW, my idea was consulting `this-command-keys', but due to bug#27470,
I can't use it in my case.

Any ideas?


Thanks,

Michael.

Re: repeatable vs. non-repeatable commands

[Emanuel Berg]

Michael Heerdegen wrote:

> I want to provide another configurable
> binding-scheme of the form prefix+letter,
> where "prefix" is a prefix key the user
> can choose.

(define-prefix-command        'C-o-prefix)
(global-set-key        "\C-o" 'C-o-prefix)

(global-set-key "\C-od" (lambda () (interactive) (message "d")))
(global-set-key "\C-oD" (lambda () (interactive) (message "D")))

... no?

-- 
underground experts united
http://user.it.uu.se/~embe8573

Re: repeatable vs. non-repeatable commands

[Emanuel Berg]

Michael Heerdegen wrote:

> C-A [...] Since these bindings are not
> available in a console

In a way they are, with the familiar
workaround. I suppose it doesn't count as user
configurable tho. Well, it depends who the user
is...

In /etc/console-setup/remap.inc

    # C-a and C-A
    # a/A is 30; use showkey(1)
    # C-a already works
    # use the ""Private Use Area": U+E000 up to and including U+F8FF
    control shift keycode 30 = U+E000

then with 'sudo' or possibly first 'chmod +s /bin/loadkeys'

    loadkeys -q -c -s /etc/console-setup/remap.inc

in Emacs

    (define-key input-decode-map [?\uE000] [C-A])
    (global-set-key [C-A]  (lambda () (interactive) (message "A")))
    (global-set-key "\C-a" (lambda () (interactive) (message "a")))

-- 
underground experts united
http://user.it.uu.se/~embe8573

Re: repeatable vs. non-repeatable commands

[Stefan Monnier]

> FWIW, my idea was consulting `this-command-keys', but due to bug#27470,
> I can't use it in my case.

I just sent a workaround for bug#27470 that you might be able to use
(i.e. call this-command-keys(-vector) at the very beginning).

E.g.

    (defun el-search--repeatable-binding-p ()
      "Return non-nil if the last command's binding should be repeatable.
    Beware: this should be called before any other user-interaction."
      (let ((keys (this-command-keys-vector)))
        (> (length keys)
           ;; Meta is often turned into a separate ESC byte in
           ;; terminal emulators.
           (if (eq ?\e (aref keys 0)) 2 1))))

    (defun my-el-search-foo (arg1 arg &optional repeat)
      (interactive
        (let ((repeat (el-search--repeatable-binding-p)))
          ...
          (list ... repeat)))
      [...]
      (when repeat
        (set-transient-map ...)))


        Stefan

Re: repeatable vs. non-repeatable commands

[Michael Heerdegen]

Emanuel Berg <moasen@zoho.com> writes:

> (define-prefix-command        'C-o-prefix)
> (global-set-key        "\C-o" 'C-o-prefix)
>
> (global-set-key "\C-od" (lambda () (interactive) (message "d")))
> (global-set-key "\C-oD" (lambda () (interactive) (message "D")))
>
> ... no?

That only solves the easy part (defining a prefix key), but not the
hard.  The heard part is:

  (global-set-key "\C-od" #'the-function)

should make C-o d d message "d" two times.  But with the same named
command, binding

  (global-set-key "\C-D" #'the-function)

(that is, control-shift-d) should not make typing C-D d message "d" two
times.  Instead, only C-D should call the command, and the following d
is not special (i.e. calls self-insert-command).  That means, I need to
decide whether I have to establish a transient map in the body of
`the-function' base on the keys hit, and how I can do that correctly is
my question.


Thanks,

Michael.

Re: repeatable vs. non-repeatable commands

[Michael Heerdegen]

Emanuel Berg <moasen@zoho.com> writes:

> In a way they are, with the familiar workaround. I suppose it doesn't
> count as user configurable tho. Well, it depends who the user is...
>
> In /etc/console-setup/remap.inc
>
>     # C-a and C-A
>     # a/A is 30; use showkey(1)
>     # C-a already works
>     # use the ""Private Use Area": U+E000 up to and including U+F8FF
>     control shift keycode 30 = U+E000
>
> then with 'sudo' or possibly first 'chmod +s /bin/loadkeys'
>
>     loadkeys -q -c -s /etc/console-setup/remap.inc
>
> in Emacs
>
>     (define-key input-decode-map [?\uE000] [C-A])
>     (global-set-key [C-A]  (lambda () (interactive) (message "A")))
>     (global-set-key "\C-a" (lambda () (interactive) (message "a")))

Interesting!  I'm not really a console user myself, but it's good to
have your explanation in this thread.


Thanks,

Michael.

Re: repeatable vs. non-repeatable commands

[Emanuel Berg]

Michael Heerdegen wrote:

> That only solves the easy part (defining
> a prefix key), but not the hard. The heard
> part is:
>
>   (global-set-key "\C-od" #'the-function)
>
> should make C-o d d message "d" two times.

Yes, I suspected there was more to your
question - like an extra "d"...

Should `C-o d' echo just one "d"? How does
Emacs know when the keystroke is terminated?
Some timeout?

Is this something that's around Emacs in
general? How do you set it up regardless of
what C-D does?

I know you mentioned search - I have my own
search [1] which is why possibly I never
noticed. Or is it more common than that?


[1] http://user.it.uu.se/~embe8573/conf/emacs-init/wrap-search.el

-- 
underground experts united
http://user.it.uu.se/~embe8573

Re: repeatable vs. non-repeatable commands

[Emanuel Berg]

Michael Heerdegen wrote:

> Interesting! I'm not really a console user
> myself, but it's good to have your
> explanation in this thread.

Here is a 2015 tutorial [1]. Back then, I got
tired of repeating myself so I decided to put
*everything* into one text file, like the old
BBS guys! What has been, will be again - double
meaning unintentional.

Ironically, this time around I didn't think of
it so what happened was I repeated myself.

Anyway, I made some very minor updates so hit
the reload button a couple of times to make
sure it says "2017" on line 180 :)

[1] http://user.it.uu.se/~embe8573/tty-emacs-keys.txt

-- 
underground experts united
http://user.it.uu.se/~embe8573

Re: repeatable vs. non-repeatable commands

[Michael Heerdegen]

Emanuel Berg <moasen@zoho.com> writes:

> Should `C-o d' echo just one "d"?

Yes.

> How does Emacs know when the keystroke is terminated?  Some timeout?

No.  I implement this with transient maps (see `set-transient-map').  In
the simplest case, you can use them to make commands repeatable - see
"repeat.el", or the example in

  https://lists.gnu.org/archive/html/help-gnu-emacs/2017-04/msg00159.html

In the general case, what the user conceives as keystroke is terminated
when you enter a key (sequence) that can't be found in the transient map
- but this is even programmable (via the first optional argument of
`set-transient-map').


Michael.

Re: repeatable vs. non-repeatable commands

[Michael Heerdegen]

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

> I just sent a workaround for bug#27470 that you might be able to use
>  (i.e. call this-command-keys(-vector) at the very beginning).
>
> E.g. [...]

Thank you, I'll probably go with something like that.

Michael.