bug#55301: 27.2; RFE: Auto-create defcustom for declared minor mode keymap

bug#55301: 27.2; RFE: Auto-create defcustom for declared minor mode keymap

[Phil Hudson]

---------
How about if the KEYMAP argument (or the :keymap slot) of
`define-minor-mode', if given, planted a `defcustom' form for the
keymap, thus automagically making each minor mode keymap a
customization option?

bug#55301: 27.2; RFE: Auto-create defcustom for declared minor mode keymap

[Lars Ingebrigtsen]

Phil Hudson <phil.hudson@iname.com> writes:

> How about if the KEYMAP argument (or the :keymap slot) of
> `define-minor-mode', if given, planted a `defcustom' form for the
> keymap, thus automagically making each minor mode keymap a
> customization option?

We don't use defcustoms for other mode maps, so I'm not sure why we
should for the ones tied to `define-minor-mode'?  Am I missing
something?

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

bug#55301: 27.2; RFE: Auto-create defcustom for declared minor mode keymap

[Phil Hudson]

If nobody sees any benefit from the idea, I imagine the best thing is
to just close this.

On Sat, 7 May 2022 at 16:34, Lars Ingebrigtsen <larsi@gnus.org> wrote:
>
> Phil Hudson <phil.hudson@iname.com> writes:
>
> > How about if the KEYMAP argument (or the :keymap slot) of
> > `define-minor-mode', if given, planted a `defcustom' form for the
> > keymap, thus automagically making each minor mode keymap a
> > customization option?
>
> We don't use defcustoms for other mode maps, so I'm not sure why we
> should for the ones tied to `define-minor-mode'?  Am I missing
> something?
>
> --
> (domestic pets only, the antidote for overdose, milk.)
>    bloggy blog: http://lars.ingebrigtsen.no

bug#55301: 27.2; RFE: Auto-create defcustom for declared minor mode keymap

[Lars Ingebrigtsen]

Phil Hudson <phil.hudson@iname.com> writes:

> If nobody sees any benefit from the idea, I imagine the best thing is
> to just close this.

I just don't understand the request -- we don't have a Customize
interface for altering keymaps anyway.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

bug#55301: 27.2; RFE: Auto-create defcustom for declared minor mode keymap

[Phil Hudson]

Not a built-in one, no, but it is not difficult to construct a composite
defcustom.  There are plenty of sequence types and a keybinding type to compose
with.

Perhaps Emacs *should* have a standard built-in keymap defcustom type, but that
would be the subject for a separate RFE.

What I am proposing is something along the lines of this: when a mode (major or
minor, it's strictly irrelevant) declares a keymap, part of the processing of
the expansion of the mode definition should include a defcustom enabling the
user to customize that keymap using the Customize UI, rather than (only) by
writing elisp code.  In other words, where foo-mode-map is currently implemented
(I assume) using a defvar, that should change to a defcustom.  That's really all
I'm suggesting: change the defvar (or whatever) to a defcustom.

On Sun, 8 May 2022 at 12:46, Lars Ingebrigtsen <larsi@gnus.org> wrote:
>
> Phil Hudson <phil.hudson@iname.com> writes:
>
> > If nobody sees any benefit from the idea, I imagine the best thing is
> > to just close this.
>
> I just don't understand the request -- we don't have a Customize
> interface for altering keymaps anyway.
>
> --
> (domestic pets only, the antidote for overdose, milk.)
>    bloggy blog: http://lars.ingebrigtsen.no

bug#55301: 27.2; RFE: Auto-create defcustom for declared minor mode keymap

[Lars Ingebrigtsen]

Phil Hudson <phil.hudson@iname.com> writes:

> What I am proposing is something along the lines of this: when a mode
> (major or minor, it's strictly irrelevant) declares a keymap, part of
> the processing of the expansion of the mode definition should include
> a defcustom enabling the user to customize that keymap using the
> Customize UI, rather than (only) by writing elisp code.  In other
> words, where foo-mode-map is currently implemented (I assume) using a
> defvar, that should change to a defcustom.  That's really all I'm
> suggesting: change the defvar (or whatever) to a defcustom.

Emacs doesn't have a Customize UI to alter keymaps, so changing defvars
to defcustom isn't currently useful (because there's no :type that
could be used other than `sexp').

It would be possible to create such an UI, of course, but it would work
quite differently than all other Customize widgets.  I don't think
anybody wants to have the entire definitions of keymaps in their .emacs
files -- they want to add a keystroke or remove a keystroke or two, and
that's it.  (And, besides, they want to get new keystrokes that are
added to the maps.)

But it'd be nice to have; sure.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

bug#55301: 27.2; RFE: Auto-create defcustom for declared minor mode keymap

[Phil Hudson]

How about ":type '(alist :key-type key-sequence :value-type function)"?

On Mon, 9 May 2022 at 10:34, Lars Ingebrigtsen <larsi@gnus.org> wrote:
>
> Phil Hudson <phil.hudson@iname.com> writes:
>
> > What I am proposing is something along the lines of this: when a mode
> > (major or minor, it's strictly irrelevant) declares a keymap, part of
> > the processing of the expansion of the mode definition should include
> > a defcustom enabling the user to customize that keymap using the
> > Customize UI, rather than (only) by writing elisp code.  In other
> > words, where foo-mode-map is currently implemented (I assume) using a
> > defvar, that should change to a defcustom.  That's really all I'm
> > suggesting: change the defvar (or whatever) to a defcustom.
>
> Emacs doesn't have a Customize UI to alter keymaps, so changing defvars
> to defcustom isn't currently useful (because there's no :type that
> could be used other than `sexp').
>
> It would be possible to create such an UI, of course, but it would work
> quite differently than all other Customize widgets.  I don't think
> anybody wants to have the entire definitions of keymaps in their .emacs
> files -- they want to add a keystroke or remove a keystroke or two, and
> that's it.  (And, besides, they want to get new keystrokes that are
> added to the maps.)
>
> But it'd be nice to have; sure.
>
> --
> (domestic pets only, the antidote for overdose, milk.)
>    bloggy blog: http://lars.ingebrigtsen.no

bug#55301: 27.2; RFE: Auto-create defcustom for declared minor mode keymap

[Lars Ingebrigtsen]

Phil Hudson <phil.hudson@iname.com> writes:

> How about ":type '(alist :key-type key-sequence :value-type function)"?

Like I said, that wouldn't be very useful, because people don't want the
entire keymap in their .emacs files anyway.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

bug#55301: 27.2; RFE: Auto-create defcustom for declared minor mode keymap

[Phil Hudson]

On Mon, 9 May 2022 at 10:34, Lars Ingebrigtsen <larsi@gnus.org> wrote:
>
> (And, besides, they want to get new keystrokes that are
> added to the maps.)

Customize does support insertion of new elements in its composite
types, of course.

bug#55301: 27.2; RFE: Auto-create defcustom for declared minor mode keymap

[Phil Hudson]

On Mon, 9 May 2022 at 10:54, Lars Ingebrigtsen <larsi@gnus.org> wrote:
>
> Phil Hudson <phil.hudson@iname.com> writes:
>
> > How about ":type '(alist :key-type key-sequence :value-type function)"?
>
> Like I said, that wouldn't be very useful, because people don't want the
> entire keymap in their .emacs files anyway.

Isn't that exactly the right place for declarative personal preferences?

bug#55301: 27.2; RFE: Auto-create defcustom for declared minor mode keymap

[Robert Pluim]

>>>>> On Mon, 9 May 2022 11:02:20 +0100, Phil Hudson <phil.hudson@iname.com> said:

    Phil> On Mon, 9 May 2022 at 10:54, Lars Ingebrigtsen <larsi@gnus.org> wrote:
    >> 
    >> Phil Hudson <phil.hudson@iname.com> writes:
    >> 
    >> > How about ":type '(alist :key-type key-sequence :value-type function)"?
    >> 
    >> Like I said, that wouldn't be very useful, because people don't want the
    >> entire keymap in their .emacs files anyway.

    Phil> Isn't that exactly the right place for declarative personal preferences?

Yes, but if a keymap were a defcustom, then the whole keymap would be
in your .emacs, and you would not receive any updates if the keymap
definition changed, since that affects the *default* defcustom value.

Now if you wanted to define a defcustom type for changing bindings in an
existing keymap, thatʼs a different proposition (and we already have
define-key, so Iʼm not so sure of the utility).

Robert
-- 

bug#55301: 27.2; RFE: Auto-create defcustom for declared minor mode keymap

[Phil Hudson]

On Mon, 9 May 2022 at 14:39, Robert Pluim <rpluim@gmail.com> wrote:
>
> >>>>> On Mon, 9 May 2022 11:02:20 +0100, Phil Hudson <phil.hudson@iname.com> said:
>
>     Phil> Isn't that exactly the right place for declarative personal preferences?
>
> Yes, but if a keymap were a defcustom, then the whole keymap would be
> in your .emacs, and you would not receive any updates if the keymap
> definition changed, since that affects the *default* defcustom value.

True that, and I had not thought of it, but now that I have, I still
don't see it as necessarily confounding. Doesn't the same objection
(if it is one) apply to *every* defcustom? Is that not the risk that
the user takes when they change any custom option, and the risk that
every dev takes when they introduce a defcustom?

> Now if you wanted to define a defcustom type for changing bindings in an
> existing keymap, thatʼs a different proposition (and we already have
> define-key, so Iʼm not so sure of the utility).

That is exactly what I (think I) want. The utility is admittedly
marginal but perhaps not trivial. It's the difference between a
declarative approach, which I take to be intrinsically more accessible
and intuitive, especially to non-programmers, versus an imperative
one. It's essentially the case for custom options in general.

I suppose my maximalist claim would be that mode-defined keymaps are
self-evidently customize options that have somehow historically been
accidentally overlooked. I'm not entirely convinced myself, though.
Maybe we should just let it sit for a while. There's obviously
something about it that doesn't *feel* right to people with deeper
insight than I have.

bug#55301: 27.2; RFE: Auto-create defcustom for declared minor mode keymap

[Phil Hudson]

So after a week, can I get a clear yes or no on this? Either is fine.

On Mon, 9 May 2022 at 16:04, Phil Hudson <phil.hudson@iname.com> wrote:
>
> On Mon, 9 May 2022 at 14:39, Robert Pluim <rpluim@gmail.com> wrote:
> >
> > >>>>> On Mon, 9 May 2022 11:02:20 +0100, Phil Hudson <phil.hudson@iname.com> said:
> >
> >     Phil> Isn't that exactly the right place for declarative personal preferences?
> >
> > Yes, but if a keymap were a defcustom, then the whole keymap would be
> > in your .emacs, and you would not receive any updates if the keymap
> > definition changed, since that affects the *default* defcustom value.
>
> True that, and I had not thought of it, but now that I have, I still
> don't see it as necessarily confounding. Doesn't the same objection
> (if it is one) apply to *every* defcustom? Is that not the risk that
> the user takes when they change any custom option, and the risk that
> every dev takes when they introduce a defcustom?
>
> > Now if you wanted to define a defcustom type for changing bindings in an
> > existing keymap, thatʼs a different proposition (and we already have
> > define-key, so Iʼm not so sure of the utility).
>
> That is exactly what I (think I) want. The utility is admittedly
> marginal but perhaps not trivial. It's the difference between a
> declarative approach, which I take to be intrinsically more accessible
> and intuitive, especially to non-programmers, versus an imperative
> one. It's essentially the case for custom options in general.
>
> I suppose my maximalist claim would be that mode-defined keymaps are
> self-evidently customize options that have somehow historically been
> accidentally overlooked. I'm not entirely convinced myself, though.
> Maybe we should just let it sit for a while. There's obviously
> something about it that doesn't *feel* right to people with deeper
> insight than I have.

bug#55301: 27.2; RFE: Auto-create defcustom for declared minor mode keymap

[Robert Pluim]

>>>>> On Tue, 17 May 2022 17:03:07 +0100, Phil Hudson <phil.hudson@iname.com> said:

    Phil> So after a week, can I get a clear yes or no on this? Either is fine.

You mean using customize for keymaps? Iʼd say 'no', but Iʼm not an
Emacs maintainer.

Robert
-- 

bug#55301: 27.2; RFE: Auto-create defcustom for declared minor mode keymap

[Lars Ingebrigtsen]

Robert Pluim <rpluim@gmail.com> writes:

> You mean using customize for keymaps? Iʼd say 'no', but Iʼm not an
> Emacs maintainer.

Yes, adding a defcustom now for keymaps would be pointless.

But it'd be nice if a defcustom interface for keymaps existed.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

bug#55301: 27.2; RFE: Auto-create defcustom for declared minor mode keymap

[Robert Pluim]

>>>>> On Tue, 17 May 2022 19:27:14 +0200, Lars Ingebrigtsen <larsi@gnus.org> said:

    Lars> Robert Pluim <rpluim@gmail.com> writes:
    >> You mean using customize for keymaps? Iʼd say 'no', but Iʼm not an
    >> Emacs maintainer.

    Lars> Yes, adding a defcustom now for keymaps would be pointless.

    Lars> But it'd be nice if a defcustom interface for keymaps existed.

Something like this?

(defcustom keymap-additions nil
"Alist of bindings to add to keymaps at startup.
Keys are keymap symbols, values are lists of conses, format
(key-binding . command)

Where KEY-BINDING is a kbd format key specification, and COMMAND is an
Emacs command.")

So you could have that be

'((fundamental-mode-map . (("C-a" . mark-whole-buffer)
                           ("C-b" . backward-word)))
  (elisp-mode-map . (("C-a" . mark-defun))))

Robert
-- 

bug#55301: 27.2; RFE: Auto-create defcustom for declared minor mode keymap

[Phil Hudson]

On Wed, 18 May 2022 at 08:13, Robert Pluim <rpluim@gmail.com> wrote:
>
> >>>>> On Tue, 17 May 2022 19:27:14 +0200, Lars Ingebrigtsen <larsi@gnus.org> said:
>
>     Lars> Robert Pluim <rpluim@gmail.com> writes:
>     >> You mean using customize for keymaps? Iʼd say 'no', but Iʼm not an
>     >> Emacs maintainer.
>
>     Lars> Yes, adding a defcustom now for keymaps would be pointless.
>
>     Lars> But it'd be nice if a defcustom interface for keymaps existed.
>
> Something like this?
>
> (defcustom keymap-additions nil
> "Alist of bindings to add to keymaps at startup.
> Keys are keymap symbols, values are lists of conses, format
> (key-binding . command)
>
> Where KEY-BINDING is a kbd format key specification, and COMMAND is an
> Emacs command.")
>
> So you could have that be
>
> '((fundamental-mode-map . (("C-a" . mark-whole-buffer)
>                            ("C-b" . backward-word)))
>   (elisp-mode-map . (("C-a" . mark-defun))))

Exactly. It should include

    :type '(alist :key-type key-sequence :value-type function)

as I suggested 9 days ago in this thread.

bug#55301: 27.2; RFE: Auto-create defcustom for declared minor mode keymap

[Lars Ingebrigtsen]

Robert Pluim <rpluim@gmail.com> writes:

>     Lars> But it'd be nice if a defcustom interface for keymaps existed.
>
> Something like this?
>
> (defcustom keymap-additions nil
> "Alist of bindings to add to keymaps at startup.
> Keys are keymap symbols, values are lists of conses, format
> (key-binding . command)
>
> Where KEY-BINDING is a kbd format key specification, and COMMAND is an
> Emacs command.")
>
> So you could have that be
>
> '((fundamental-mode-map . (("C-a" . mark-whole-buffer)
>                            ("C-b" . backward-word)))
>   (elisp-mode-map . (("C-a" . mark-defun))))

Yes, something like that.  But the major work would be to create a
sensible widget that allows people to customise the keymap in question.
It'd be nice if it displayed a graphical keyboard in a buffer, and you
could click on keys to specify what they should output.  😀

Or something simpler, for instance a widget where you ask the user to
enter a chord/key stroke, and then prompt what command to bind that to.
Or...  something else.  Many possibilities here.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

bug#55301: 27.2; RFE: Auto-create defcustom for declared minor mode keymap

[Phil Hudson]

On Wed, 18 May 2022 at 12:29, Lars Ingebrigtsen <larsi@gnus.org> wrote:
>
> Yes, something like that.  But the major work would be to create a
> sensible widget that allows people to customise the keymap in question.
> It'd be nice if it displayed a graphical keyboard in a buffer, and you
> could click on keys to specify what they should output.  😀
>
> Or something simpler, for instance a widget where you ask the user to
> enter a chord/key stroke, and then prompt what command to bind that to.
> Or...  something else.  Many possibilities here.

What excellent ideas!

bug#55301: 27.2; RFE: Auto-create defcustom for declared minor mode keymap

[Richard Stallman]

[[[ To any NSA and FBI agents reading my email: please consider    ]]]
[[[ whether defending the US Constitution against all enemies,     ]]]
[[[ foreign or domestic, requires you to follow Snowden's example. ]]]

  > Yes, something like that.  But the major work would be to create a
  > sensible widget that allows people to customise the keymap in question.
  > It'd be nice if it displayed a graphical keyboard in a buffer, and you
  > could click on keys to specify what they should output.  😀

  > Or something simpler, for instance a widget where you ask the user to
  > enter a chord/key stroke, and then prompt what command to bind that to.
  > Or...  something else.  Many possibilities here.

I like this idea too.
-- 
Dr Richard Stallman (https://stallman.org)
Chief GNUisance of the GNU Project (https://gnu.org)
Founder, Free Software Foundation (https://fsf.org)
Internet Hall-of-Famer (https://internethalloffame.org)