bug#59539: 29.0.50; Commands created with `insert-kbd-macro` are NOT getting repeated.

bug#59539: 29.0.50; Commands created with `insert-kbd-macro` are NOT getting repeated.

[Ramesh Nedunchezian]

Commands created  with `insert-kbd-macro` are NOT getting repeated. Can this limitation be lifted?


For example, down below `my-cmd-a' and `my-cmd-b' are kbd macro commands bound to `<f12> a' and `<f12> b' respectively. 

With `repeat-mode' ON, I would expect that `a' and `b' keys to be repeatable . 

That is I am expecting that

<f12> a a

<f12> a b

will move the cursor and NOT modify the buffer.

Can `repeat-mode' be enhanced to make `a' and `b' keys repeatable.

Sometimes I do complex edits, and it is fairly common in my case to alternate between MANY kbd macro commands to quickly "fixup" things that I am editing.  F


WIW, the text files that I edit are txt files created with pdftotext, and the PDF files are usually books, or handouts which I want to read in epub format. 

So, I have multiple kbd macros in my init file that remove headers, and footers, removes indentation, changes bullets, remove hyphenation, linearize columnar text etc. It would be convenient if these kbd macro commands are repeatable.



(defalias 'my-cmd-a
  (kmacro "C-f C-f"))

(defalias 'my-cmd-b
  (kmacro "C-b C-b"))

(define-prefix-command 'my-kmacro-keymap)

(define-key my-kmacro-keymap (kbd "a") #'my-cmd-a)
(define-key my-kmacro-keymap (kbd "b") #'my-cmd-b)

(put 'my-cmd-a 'repeat-map 'my-kmacro-keymap)
(put 'my-cmd-b 'repeat-map 'my-kmacro-keymap)

(global-set-key (kbd "<f12>")
        my-kmacro-keymap)

(repeat-mode -1)
(repeat-mode +1)


In GNU Emacs 29.0.50 (build 3, x86_64-pc-linux-gnu, GTK+ Version
 3.24.34, cairo version 1.16.0) of 2022-11-19 built on debian
Repository revision: a6ae13af42ede6618c326855ea4c95e0298fb75b
Repository branch: master
Windowing system distributor 'The X.Org Foundation', version 11.0.12101004
System Description: Debian GNU/Linux bookworm/sid

bug#59539: 29.0.50; Commands created with `insert-kbd-macro` are NOT getting repeated.

[Robert Pluim]

>>>>> On Thu, 24 Nov 2022 16:28:47 +0530, Ramesh Nedunchezian <rameshnedunchezian@outlook.com> said:


    Ramesh> (defalias 'my-cmd-a
    Ramesh>   (kmacro "C-f C-f"))

    Ramesh> (defalias 'my-cmd-b
    Ramesh>   (kmacro "C-b C-b"))

    Ramesh> (define-prefix-command 'my-kmacro-keymap)

    Ramesh> (define-key my-kmacro-keymap (kbd "a") #'my-cmd-a)
    Ramesh> (define-key my-kmacro-keymap (kbd "b") #'my-cmd-b)

    Ramesh> (put 'my-cmd-a 'repeat-map 'my-kmacro-keymap)
    Ramesh> (put 'my-cmd-b 'repeat-map 'my-kmacro-keymap)

    Ramesh> (global-set-key (kbd "<f12>")
    Ramesh>         my-kmacro-keymap)

    Ramesh> (repeat-mode -1)
    Ramesh> (repeat-mode +1)

Hmm, with your recipe on emacs master, "<f12> b" repeats, but "<f12>
a" doesnʼt. I have no clue why, they call the same underlying
infrastructure.

If I use the following instead:

(defun my-cmd-a ()
  (interactive)
  (forward-char 2))
(defun my-cmd-b ()
  (interactive)
  (backward-char 2))

it all works as expected, but Iʼm guessing you have a bunch of macros
defined already.

Robert
-- 

bug#59539: 29.0.50; Commands created with `insert-kbd-macro` are NOT getting repeated.

[Juri Linkov]

> Ramesh> (defalias 'my-cmd-a
> Ramesh>   (kmacro "C-f C-f"))
>
> Ramesh> (defalias 'my-cmd-b
> Ramesh>   (kmacro "C-b C-b"))
>
> Ramesh> (define-key my-kmacro-keymap (kbd "a") #'my-cmd-a)
> Ramesh> (define-key my-kmacro-keymap (kbd "b") #'my-cmd-b)
>
> Hmm, with your recipe on emacs master, "<f12> b" repeats, but "<f12>
> a" doesnʼt. I have no clue why, they call the same underlying
> infrastructure.

Sorry, I missed this bug report.  This is because 'repeat-check-key'
is t by default so that unrelated key sequences, for example, 'C-_'
should not activate the repeat map for 'undo', only 'C-x u' should.

When the user types 'a', the real key from kmacro is 'C-f'.
There is no 'f' in the keymap above, so it's ignored,
when 'repeat-check-key' is not customized to nil.

Do you have an idea how to do the right thing and to guess the user's
intention without checking if there are the same keys in the keymap?

bug#59539: 29.0.50; Commands created with `insert-kbd-macro` are NOT getting repeated.

[Robert Pluim]

>>>>> On Thu, 15 Dec 2022 09:20:31 +0200, Juri Linkov <juri@linkov.net> said:

    Ramesh> (defalias 'my-cmd-a
    Ramesh>   (kmacro "C-f C-f"))
    >> 
    Ramesh> (defalias 'my-cmd-b
    Ramesh>   (kmacro "C-b C-b"))
    >> 
    Ramesh> (define-key my-kmacro-keymap (kbd "a") #'my-cmd-a)
    Ramesh> (define-key my-kmacro-keymap (kbd "b") #'my-cmd-b)
    >> 
    >> Hmm, with your recipe on emacs master, "<f12> b" repeats, but "<f12>
    >> a" doesnʼt. I have no clue why, they call the same underlying
    >> infrastructure.

    Juri> Sorry, I missed this bug report.  This is because 'repeat-check-key'
    Juri> is t by default so that unrelated key sequences, for example, 'C-_'
    Juri> should not activate the repeat map for 'undo', only 'C-x u' should.

    Juri> When the user types 'a', the real key from kmacro is 'C-f'.
    Juri> There is no 'f' in the keymap above, so it's ignored,
    Juri> when 'repeat-check-key' is not customized to nil.

That explains why '<f12> b' works

    Juri> Do you have an idea how to do the right thing and to guess the user's
    Juri> intention without checking if there are the same keys in the keymap?

I guess it should be possible to find out what the 'original' key
pressed was. <time passes> How about the below? I guess it changes the
behaviour when you have a command bound to two keys in the same repeat
map, and you use the 'wrong' one to repeat (but I could argue that
people creating such keymaps should use the 'repeat-check-key'
property)

Ramesh,

(put 'my-cmd-a 'repeat-check-key 'no)
(put 'my-cmd-b 'repeat-check-key 'no)

should do the right thing for you regardless of whether we change
Emacs.

Robert
-- 

diff --git a/lisp/repeat.el b/lisp/repeat.el
index 33e8d98ce3..05bdb5dfd1 100644
--- a/lisp/repeat.el
+++ b/lisp/repeat.el
@@ -458,8 +458,11 @@ repeat-post-hook
         (when rep-map
           (when (and (symbolp rep-map) (boundp rep-map))
             (setq rep-map (symbol-value rep-map)))
-          (let ((map (copy-keymap rep-map)))
-
+          (let* ((map (copy-keymap rep-map))
+                 (internal (where-is-internal real-this-command (list map) t))
+                 (real-key last-command-event))
+            (when internal
+              (setq real-key (aref internal 0)))
             (when (and
                    ;; Detect changes in the minibuffer state to allow repetitions
                    ;; in the same minibuffer, but not when the minibuffer is activated
@@ -467,7 +470,7 @@ repeat-post-hook
                    (or (< (minibuffer-depth) (car repeat--prev-mb))
                        (eq current-minibuffer-command (cdr repeat--prev-mb)))
                    (or (not repeat-keep-prefix) prefix-arg)
-                   (repeat-check-key last-command-event map))
+                   (repeat-check-key real-key map))
 
               ;; Messaging
               (unless prefix-arg

bug#59539: 29.0.50; Commands created with `insert-kbd-macro` are NOT getting repeated.

[Juri Linkov]

>> Do you have an idea how to do the right thing and to guess the user's
>> intention without checking if there are the same keys in the keymap?
>
> I guess it should be possible to find out what the 'original' key
> pressed was. <time passes> How about the below? I guess it changes the
> behaviour when you have a command bound to two keys in the same repeat
> map, and you use the 'wrong' one to repeat

I tested your patch, but then 'C-_' starts the key sequence
'C-_ u u u', whereas we had the users' requests that only 'C-x u'
should start the repeatable sequence by default with 'C-x u u u'.

> (but I could argue that people creating such keymaps should use the
> 'repeat-check-key' property)

For more fine-grained control than 'repeat-check-key', we could support
more properties, for example, define separately only keys that enter
the repeatable sequence, or keys that exit it:

  (put 'undo 'repeat-enter-keys '([?\C-x ?u]))
  (put 'undo 'repeat-exit-keys '([?\C-_]))

bug#59539: 29.0.50; Commands created with `insert-kbd-macro` are NOT getting repeated.

[Ramesh Nedunchezian]

On 15/12/22 12:50, Juri Linkov wrote:
>> Ramesh> (defalias 'my-cmd-a
>> Ramesh>   (kmacro "C-f C-f"))
>>
>> Ramesh> (defalias 'my-cmd-b
>> Ramesh>   (kmacro "C-b C-b"))
>>
>> Ramesh> (define-key my-kmacro-keymap (kbd "a") #'my-cmd-a)
>> Ramesh> (define-key my-kmacro-keymap (kbd "b") #'my-cmd-b)
>>
>> Hmm, with your recipe on emacs master, "<f12> b" repeats, but "<f12>
>> a" doesnʼt. I have no clue why, they call the same underlying
>> infrastructure.
> 
> Sorry, I missed this bug report.  This is because 'repeat-check-key'
> is t by default so that unrelated key sequences, for example, 'C-_'
> should not activate the repeat map for 'undo', only 'C-x u' should.
> 
> When the user types 'a', the real key from kmacro is 'C-f'.
> There is no 'f' in the keymap above, so it's ignored,
> when 'repeat-check-key' is not customized to nil.
> 
> Do you have an idea how to do the right thing and to guess the user's
> intention without checking if there are the same keys in the keymap?

`repeat-check-key` suggestion works for me.

bug#59539: 29.0.50; Commands created with `insert-kbd-macro` are NOT getting repeated.

[Juri Linkov]

> `repeat-check-key` suggestion works for me.

Thanks for confirming.  So closing this bug report.
For more improvements а new feature request could be created.