Cycle Auto-Fill

Cycle Auto-Fill

[Christopher Dimech]

Have written the following function to use one key binding
to cycle auto-fill (only-comments, whole-buffer, disable)

Who would like to criticise this little bugger?

(defun gungadin-auto-fill-cycle ()
   "Cycles Auto Fill.  Automatically breaks lines that get beyond
variable fill-column."
   (interactive)
   (unless (get 'gungadin-auto-fill-cycle 'state)
      (put 'gungadin-auto-fill-cycle 'state 1))

   (setq n (get 'gungadin-auto-fill-cycle 'state))

   (when (= n 1)
      (message "Auto Fill Comments Only")
      (set (make-local-variable 'comment-auto-fill-only-comments) t)
      (auto-fill-mode 1)
      (put 'gungadin-auto-fill-cycle 'state 2))
   (when (= n 2)
      (message "Auto Fill Buffer")
      (set (make-local-variable 'comment-auto-fill-only-comments) nil)
      (put 'gungadin-auto-fill-cycle 'state 3))
   (when (= n 3)
      (message "Disable Auto Fill")
      (auto-fill-mode 0)
      (put 'gungadin-auto-fill-cycle 'state 1)) )

(global-set-key (kbd "H-a") #'gungadin-auto-fill-cycle)



---------------------
Christopher Dimech
General Administrator - Naiad Informatics - GNU Project (Geocomputation)
- Geophysical Simulation
- Geological Subsurface Mapping
- Disaster Preparedness and Mitigation
- Natural Resource Exploration and Production
- Free Software Advocacy

Re: Cycle Auto-Fill

[Pankaj Jangid]

Christopher Dimech <dimech@gmx.com> writes:

> (defun gungadin-auto-fill-cycle ()
>    "Cycles Auto Fill.  Automatically breaks lines that get beyond
> variable fill-column."
>    (interactive)

-    (unless (get 'gungadin-auto-fill-cycle 'state)
-       (put 'gungadin-auto-fill-cycle 'state 1))

>
>    (setq n (get 'gungadin-auto-fill-cycle 'state))
>

-    (when (= n 1)
+    (when (or (not n) (= n 1))

>       (message "Auto Fill Comments Only")
>       (set (make-local-variable 'comment-auto-fill-only-comments) t)
>       (auto-fill-mode 1)
>       (put 'gungadin-auto-fill-cycle 'state 2))
>    (when (= n 2)
>       (message "Auto Fill Buffer")
>       (set (make-local-variable 'comment-auto-fill-only-comments) nil)
>       (put 'gungadin-auto-fill-cycle 'state 3))
>    (when (= n 3)
>       (message "Disable Auto Fill")
>       (auto-fill-mode 0)
>       (put 'gungadin-auto-fill-cycle 'state 1)) )
>
> (global-set-key (kbd "H-a") #'gungadin-auto-fill-cycle)

Will this work?

Removed one (get) call and a conditional test. Added one conditional
test.

Re: Cycle Auto-Fill

[Christopher Dimech]

The change gives the error.  Although the modified version looks reasonable to me.

Debugger entered--Lisp error: (wrong-type-argument number-or-marker-p nil)
  =(nil 2)

> Sent: Saturday, December 05, 2020 at 2:59 PM
> From: "Pankaj Jangid" <pankaj@codeisgreat.org>
> To: "Christopher Dimech" <dimech@gmx.com>
> Cc: "Help Gnu Emacs" <help-gnu-emacs@gnu.org>
> Subject: Re: Cycle Auto-Fill
>
> Christopher Dimech <dimech@gmx.com> writes:
>
> > (defun gungadin-auto-fill-cycle ()
> >    "Cycles Auto Fill.  Automatically breaks lines that get beyond
> > variable fill-column."
> >    (interactive)
>
> -    (unless (get 'gungadin-auto-fill-cycle 'state)
> -       (put 'gungadin-auto-fill-cycle 'state 1))
>
> >
> >    (setq n (get 'gungadin-auto-fill-cycle 'state))
> >
>
> -    (when (= n 1)
> +    (when (or (not n) (= n 1))
>
> >       (message "Auto Fill Comments Only")
> >       (set (make-local-variable 'comment-auto-fill-only-comments) t)
> >       (auto-fill-mode 1)
> >       (put 'gungadin-auto-fill-cycle 'state 2))
> >    (when (= n 2)
> >       (message "Auto Fill Buffer")
> >       (set (make-local-variable 'comment-auto-fill-only-comments) nil)
> >       (put 'gungadin-auto-fill-cycle 'state 3))
> >    (when (= n 3)
> >       (message "Disable Auto Fill")
> >       (auto-fill-mode 0)
> >       (put 'gungadin-auto-fill-cycle 'state 1)) )
> >
> > (global-set-key (kbd "H-a") #'gungadin-auto-fill-cycle)
>
> Will this work?
>
> Removed one (get) call and a conditional test. Added one conditional
> test.
>

Re: Cycle Auto-Fill

[Pankaj Jangid]

Christopher Dimech <dimech@gmx.com> writes:

> The change gives the error.  Although the modified version looks
> reasonable to me.
>
> Debugger entered--Lisp error: (wrong-type-argument number-or-marker-p
> nil)
>   =(nil 2)

Hmmm.. the first when-form is cleared but the value of n is still
nil. So second when condition fails.

But as suggested by Ergus, it is better to keep just an integer instead
of property.

Re: Cycle Auto-Fill

[Ergus]

Hi Christopher:

I am not a big lisper, but I had a similar code. I modified yours to
look similar in case you find anything interesting on it

```
(defvar gungadin-auto-fill-cycle 0)

(defun gungadin-auto-fill-cycle ()
   "Cycles Auto Fill.

Automatically breaks lines that get beyond variable fill-column."
   (interactive)

   (pcase gungadin-auto-fill-cycle
     (0
      (message "Auto Fill Comments Only")
      (setq-local comment-auto-fill-only-comments t)
      (auto-fill-mode 1))
     (1
      (message "Auto Fill Buffer")
      (setq-local comment-auto-fill-only-comments nil))
     (2
      (message "Disable Auto Fill")
      (auto-fill-mode 0)))
   (setq gungadin-auto-fill-cycle (% (1+ gungadin-auto-fill-cycle) 3)))
```


On Sat, Dec 05, 2020 at 02:25:06PM +0100, Christopher Dimech wrote:
>Have written the following function to use one key binding
>to cycle auto-fill (only-comments, whole-buffer, disable)
>
>Who would like to criticise this little bugger?
>
>(defun gungadin-auto-fill-cycle ()
>   "Cycles Auto Fill.  Automatically breaks lines that get beyond
>variable fill-column."
>   (interactive)
>   (unless (get 'gungadin-auto-fill-cycle 'state)
>      (put 'gungadin-auto-fill-cycle 'state 1))
>
>   (setq n (get 'gungadin-auto-fill-cycle 'state))
>
>   (when (= n 1)
>      (message "Auto Fill Comments Only")
>      (set (make-local-variable 'comment-auto-fill-only-comments) t)
>      (auto-fill-mode 1)
>      (put 'gungadin-auto-fill-cycle 'state 2))
>   (when (= n 2)
>      (message "Auto Fill Buffer")
>      (set (make-local-variable 'comment-auto-fill-only-comments) nil)
>      (put 'gungadin-auto-fill-cycle 'state 3))
>   (when (= n 3)
>      (message "Disable Auto Fill")
>      (auto-fill-mode 0)
>      (put 'gungadin-auto-fill-cycle 'state 1)) )
>
>(global-set-key (kbd "H-a") #'gungadin-auto-fill-cycle)
>
>
>
>---------------------
>Christopher Dimech
>General Administrator - Naiad Informatics - GNU Project (Geocomputation)
>- Geophysical Simulation
>- Geological Subsurface Mapping
>- Disaster Preparedness and Mitigation
>- Natural Resource Exploration and Production
>- Free Software Advocacy
>
>

Re: Cycle Auto-Fill

[Michael Heerdegen]

Christopher Dimech <dimech@gmx.com> writes:

> Have written the following function to use one key binding
> to cycle auto-fill (only-comments, whole-buffer, disable)
>
> Who would like to criticise this little bugger?

Have you already asked the byte compiler?

Michael.

RE: Cycle Auto-Fill

[Drew Adams]

> Have you already asked the byte compiler?

+1.  (Ask Emacs!)

Re: Cycle Auto-Fill

[Christopher Dimech]

It is warning me about variable n.


> Sent: Saturday, December 05, 2020 at 11:00 PM
> From: "Michael Heerdegen" <michael_heerdegen@web.de>
> To: help-gnu-emacs@gnu.org
> Subject: Re: Cycle Auto-Fill
>
> Christopher Dimech <dimech@gmx.com> writes:
>
> > Have written the following function to use one key binding
> > to cycle auto-fill (only-comments, whole-buffer, disable)
> >
> > Who would like to criticise this little bugger?
>
> Have you already asked the byte compiler?
>
> Michael.
>
>
>

Re: Cycle Auto-Fill

[Michael Heerdegen]

Christopher Dimech <dimech@gmx.com> writes:

> It is warning me about variable n.

[ This thread and "Removing line and column number settings from some
buffers" have converged into the same exactly same state.  I wonder how
often that happens. ]

Yes, and it's a useful warning: the variable is "free", which means its
usage would interfere with other free usages of this variable.

In your code the variable seems to be a redundant duplication of the
value stored in (get 'gungadin-auto-fill-cycle 'state), so I would just
get rid of it.

But - a different issue: note that you are currently using one (global)
state to remember a (buffer) local feature.  I think you don't really
want it like this.  You probably won't get happy when you are using this
in multiple buffers at the same time.  That's where a variable gets
useful: it can be buffer local.  I would drop the
`gungadin-auto-fill-cycle' symbol property instead: symbol properties
are always global.  Of course could you save and maintain a table of the
states of all buffers there, but that would be rather exotic and
complex.


Regards,

Michael.

Re: Cycle Auto-Fill

[Christopher Dimech]

> Sent: Sunday, December 06, 2020 at 1:30 AM
> From: "Michael Heerdegen" <michael_heerdegen@web.de>
> To: "Christopher Dimech" <dimech@gmx.com>
> Cc: daniela-spit@gmx.it
> Subject: Re: Cycle Auto-Fill
>
> Christopher Dimech <dimech@gmx.com> writes:
>
> > (set (make-local-variable 'comment-auto-fill-only-comments) t)
> >
> > and
> >
> > (set (make-local-variable 'comment-auto-fill-only-comments) nil)
> >
> > Additionally I do  (auto-fill-mode 1) and (auto-fill-mode nil),
> > which I think are buffer local.
>
> Yes, that's ok.  Since the state is also easy to get (by looking up the
> variables [auto-fill-mode is also a variable]), there is no need to save
> it explicitly.

So, only comment-auto-fill-only-comments need to be set to buffer local.  Right?
Returning back to taking the property "state".  If it is global. that would
only mean that in a different buffer the cycle would begin from a a different
starting point.  But then, the auto-fill settings would still be buffer local.

Does the order of the two commands matter?

(set (make-local-variable 'comment-auto-fill-only-comments) t)
(auto-fill-mode 1)

versus

(auto-fill-mode 1)
(set (make-local-variable 'comment-auto-fill-only-comments) t)

> > Have been giving the wrong strategy then.  Thought it would work differently
> > for each different buffer.  But you are saying symbol properties are always
> > global.   Bugger!  Would this be RTFM?
>
> Yes, it would, but only in subjunctive mood.



> Regards,
>
> Michael.
>

Re: Cycle Auto-Fill

[Michael Heerdegen]

Christopher Dimech <dimech@gmx.com> writes:

> So, only comment-auto-fill-only-comments need to be set to buffer
> local.  Right?

I think so, yes.  BTW, in my Emacs, the variable corresponding to
auto-fill-mode is auto-fill-function; `auto-fill-mode' is unbound.

> Returning back to taking the property "state".  If it is global. that
> would only mean that in a different buffer the cycle would begin from
> a a different starting point.  But then, the auto-fill settings would
> still be buffer local.

Yes, that, and cycling in the first buffer may bring you from 2 to 1 or
from 2 to 2 afterwards.  Might be confusing, once you got used to it.

> Does the order of the two commands matter?
>
> (set (make-local-variable 'comment-auto-fill-only-comments) t)
> (auto-fill-mode 1)
>
> versus
>
> (auto-fill-mode 1)
> (set (make-local-variable 'comment-auto-fill-only-comments) t)

No, should not matter.  These things are called expressions, btw.

[And you might prefer `setq-local'.]


Regards,

Michael.

Re: Cycle Auto-Fill

[Christopher Dimech]

> Sent: Sunday, December 06, 2020 at 3:33 AM
> From: "Michael Heerdegen" <michael_heerdegen@web.de>
> To: help-gnu-emacs@gnu.org
> Subject: Re: Cycle Auto-Fill
>
> Christopher Dimech <dimech@gmx.com> writes:
>
> > So, only comment-auto-fill-only-comments need to be set to buffer
> > local.  Right?
>
> I think so, yes.  BTW, in my Emacs, the variable corresponding to
> auto-fill-mode is auto-fill-function; `auto-fill-mode' is unbound.

Unbound? What is that?

> > Returning back to taking the property "state".  If it is global. that
> > would only mean that in a different buffer the cycle would begin from
> > a a different starting point.  But then, the auto-fill settings would
> > still be buffer local.
>
> Yes, that, and cycling in the first buffer may bring you from 2 to 1 or
> from 2 to 2 afterwards.  Might be confusing, once you got used to it.

Can see what you mean.  That although what happens in the actual buffer
is local to the buffer, how the cycle performs, would depend on what
was done on other buffers, creating confusion.  Actually, it is not very
confusing as I write what option was selected in the mini buffer.

I have now started to understand buffer local quite well.  It has been
occurring to me that most things really got to be buffer local.

Continuing with the keybinding cycle (pressing repeatedly a key-sequence
to switch configuration, e.g. auto-fill in my case).

I am really getting into some difficulty figuring out what commands
would one want to have a Keybinding Cycle that would not be buffer-local.

For what things should a key-cycle be global?



> > Does the order of the two commands matter?
> >
> > (set (make-local-variable 'comment-auto-fill-only-comments) t)
> > (auto-fill-mode 1)
> >
> > versus
> >
> > (auto-fill-mode 1)
> > (set (make-local-variable 'comment-auto-fill-only-comments) t)
>
> No, should not matter.  These things are called expressions, btw.
>
> [And you might prefer `setq-local'.]
>
>
> Regards,
>
> Michael.
>
>
>

Re: Cycle Auto-Fill

[Robert Thorpe]

Christopher Dimech <dimech@gmx.com> writes:

> I have now started to understand buffer local quite well.  It has been
> occurring to me that most things really got to be buffer local.

It can be useful to think it terms of object-orientated languages.

An Emacs mode is a lot like a class.  A particular Emacs buffer is an
instance of that class.  A buffer-local variable is a conventional
object field (also called an instance variable in some languages).  A
non-buffer-local variable is a class variable.  A variable created by
"let" is a conventional local variable, of course.

Emacs is a bit different to conventional OO languages because the
decision about what-is-what are determined at runtime.  It's done by
calling functions, not compile at time through typing and language
features.

BR,
Robert Thorpe

Re: Cycle Auto-Fill

[Michael Heerdegen]

Christopher Dimech <dimech@gmx.com> writes:

> Unbound? What is that?

(info "(elisp) Void Variables")

> For what things should a key-cycle be global?

Things that make emacs look different, for example: menu-bar-mode and
such, display-time-mode and things that change the mode-line.  Changing
frame backgrounds because it got dark, or changing how Emacs looks like
because you want to use it for a presentation?

Things that toggle features for a frame or window instead of a buffer
(e.g. scroll-bars).  Toggling debugger flags.  Things where you want a
toggle but can't remember all the values in all of the buffers,
e.g. `case-fold-search' or such things.


Regards,

Michael.

Re: Cycle Auto-Fill

[Christopher Dimech]

> Sent: Monday, December 07, 2020 at 12:01 AM
> From: "Michael Heerdegen" <michael_heerdegen@web.de>
> To: help-gnu-emacs@gnu.org
> Subject: Re: Cycle Auto-Fill
>
> Christopher Dimech <dimech@gmx.com> writes:
>
> > Unbound? What is that?
>
> (info "(elisp) Void Variables")

What is the relevance of having the value cell of symbol auto-fill-mode
void? Why do you recommend auto-fill-function instead?

Looks like auto-fill-function is a variable called in auto-fill-mode.
You seem to imply that setting the variable auto-fill-function breaks
the line without needing to utilise auto-fill-mode.

> > For what things should a key-cycle be global?
>
> Things that make emacs look different, for example: menu-bar-mode and
> such, display-time-mode and things that change the mode-line.  Changing
> frame backgrounds because it got dark, or changing how Emacs looks like
> because you want to use it for a presentation?

Very clear, thank you.  This got us back to what Daniela was doing with
mode-line.  Quite many things can be linked together.  She has not managed
to revert mode-line to its original value, last time we talked about that.

> Things that toggle features for a frame or window instead of a buffer
> (e.g. scroll-bars).  Toggling debugger flags.  Things where you want a
> toggle but can't remember all the values in all of the buffers,
> e.g. `case-fold-search' or such things.
>
>
> Regards,
>
> Michael.
>
>
>

Re: Cycle Auto-Fill

[Michael Heerdegen]

Christopher Dimech <dimech@gmx.com> writes:

> What is the relevance of having the value cell of symbol auto-fill-mode
> void? Why do you recommend auto-fill-function instead?

`auto-fill-mode' uses internally `auto-fill-function'.  I don't recall
the details why it had been chosen to design it like this.  But yes,
it's possible that a mode uses a variable of a different name for
toggling.

Part of the answer is probably that different auto-fill functions can be
used this way, and for all values besides nil the mode thinks of itself
as "on".  OTOH, setting `auto-fill-function' nil automagically turns the
mode off.

> Very clear, thank you.  This got us back to what Daniela was doing with
> mode-line.  Quite many things can be linked together.  She has not managed
> to revert mode-line to its original value, last time we talked about that.

AFAIR I also already gave her an answer, hope it has arrived.


Regards,

Michael.