Should text-scale trigger hooks?

Should text-scale trigger hooks?

[James Cherti]

I've identified an issue that affects Emacs packages like eat (terminal) 
and visual-fill-column.

The core problem is that functions such as text-scale-increase, 
text-scale-decrease, and text-scale-set do not trigger hooks such as 
window-configuration-change-hook. Because of that, eat does not 
instantly update the window when the text scale is changed and 
visual-fill-column does not update the margin until we resize the window.

Currently, I've implemented workarounds for the eat package:
(defun persist-text-scale-adjust-eat (&rest args)
   "Adjust the text scale of all eat buffers."
   (when (derived-mode-p 'eat-mode)
     (run-hooks 'window-configuration-change-hook)))
(advice-add 'text-scale-increase :after #'persist-text-scale-adjust-eat)
(advice-add 'text-scale-decrease :after #'persist-text-scale-adjust-eat)

I've implemented a similar solution for visual-fill-column (see 
https://codeberg.org/joostkremers/visual-fill-column/pulls/16 ), but the 
visual-fill-column maintainer thinks this is probably not the right 
solution to this problem. I tend to agree with him.

While these workarounds work, I believe a more robust solution would be 
to have text-scale functions automatically trigger hooks such as 
window-configuration-change-hook, or perhaps introduce a dedicated hook 
for text scale that other plugins could utilize.

This would provide a more maintainable solution and prevent the need for 
individual package-specific workarounds.

--
James Cherti
https://www.jamescherti.com/

Re: Should text-scale trigger hooks?

[James Cherti]

Update: I just realized that text-scale-mode-hook can be utilized 
instead of advice because text-scale-* functions call text-scale-mode 
each time the text scale is changed.

(defun persist-text-scale-adjust-eat ()
   "Adjust the text scale of all ediff buffers."
   (when (derived-mode-p 'eat-mode)
     (run-hooks 'window-configuration-change-hook)))
(defun persist-text-scale-adjust-visual-fill-column ()
   "Adjust the text scale of all ediff buffers."
   (when (bound-and-true-p visual-fill-column-mode)
     (run-hooks 'window-configuration-change-hook)))
(add-hook 'text-scale-mode-hook #'persist-text-scale-adjust-eat)
(add-hook 'text-scale-mode-hook
  #'persist-text-scale-adjust-visual-fill-column)

--
James Cherti
https://www.jamescherti.com/

On 2024-11-06 20:17, James Cherti wrote:
> I've identified an issue that affects Emacs packages like eat (terminal) 
> and visual-fill-column.
> 
> The core problem is that functions such as text-scale-increase, 
> text-scale-decrease, and text-scale-set do not trigger hooks such as 
> window-configuration-change-hook. Because of that, eat does not 
> instantly update the window when the text scale is changed and 
> visual-fill-column does not update the margin until we resize the window.
> 
> Currently, I've implemented workarounds for the eat package:
> (defun persist-text-scale-adjust-eat (&rest args)
>    "Adjust the text scale of all eat buffers."
>    (when (derived-mode-p 'eat-mode)
>      (run-hooks 'window-configuration-change-hook)))
> (advice-add 'text-scale-increase :after #'persist-text-scale-adjust-eat)
> (advice-add 'text-scale-decrease :after #'persist-text-scale-adjust-eat)
> 
> I've implemented a similar solution for visual-fill-column (see 
> https://codeberg.org/joostkremers/visual-fill-column/pulls/16 ), but the 
> visual-fill-column maintainer thinks this is probably not the right 
> solution to this problem. I tend to agree with him.
> 
> While these workarounds work, I believe a more robust solution would be 
> to have text-scale functions automatically trigger hooks such as 
> window-configuration-change-hook, or perhaps introduce a dedicated hook 
> for text scale that other plugins could utilize.
> 
> This would provide a more maintainable solution and prevent the need for 
> individual package-specific workarounds.
> 
> -- 
> James Cherti
> https://www.jamescherti.com/
> 

Re: Should text-scale trigger hooks?

[Eli Zaretskii]

> Date: Wed, 6 Nov 2024 20:17:50 -0500
> From: James Cherti <contact@jamescherti.com>
> 
> I've identified an issue that affects Emacs packages like eat (terminal) 
> and visual-fill-column.
> 
> The core problem is that functions such as text-scale-increase, 
> text-scale-decrease, and text-scale-set do not trigger hooks such as 
> window-configuration-change-hook. Because of that, eat does not 
> instantly update the window when the text scale is changed and 
> visual-fill-column does not update the margin until we resize the window.

There's a huge gape between invocation of text-scale-increase and our
need that window-configuration-change-hook should be run.  I cannot
cross that gap, so please explain the relation between these two, and
in particular why you need that hook to run and not some other
mechanism.

The functions which manipulate face-remapping-alist force redisplay of
the buffer in which the alist was changed, and from Emacs display POV
this is the only thing that should happen to make sure the affected
buffer's display is redrawn.  Please explain why buffer's redisplay is
not enough for the packages you mention to do their job.

OTOH, window-configuration-change-hook is for specific events,
documented thusly:

 -- Variable: window-configuration-change-hook
     This variable specifies functions called during redisplay when
     either the buffer or the size of a window has changed.

text-scale-increase/decrease don't fit this description, so expecting
that hook to be called in those cases is wrong, and asking Emacs to
run that hook is against the documented behavior of the hook.  So if
redisplaying a buffer somehow cannot allow these packages to do their
job (but please explain in detail why), we will have to think about
some different mechanism of doing that;
window-configuration-change-hook doesn't sound like the right tool for
this.

> While these workarounds work, I believe a more robust solution would be 
> to have text-scale functions automatically trigger hooks such as 
> window-configuration-change-hook, or perhaps introduce a dedicated hook 
> for text scale that other plugins could utilize.

As you have discovered, there's already such a hook.  Does it mean
that your problems are solved?

Re: Should text-scale trigger hooks?

[Joost Kremers]

On Thu, Nov 07 2024, Eli Zaretskii wrote:
> The functions which manipulate face-remapping-alist force redisplay of
> the buffer in which the alist was changed, and from Emacs display POV
> this is the only thing that should happen to make sure the affected
> buffer's display is redrawn.  Please explain why buffer's redisplay is
> not enough for the packages you mention to do their job.

Speaking for `visual-fill-column`: it's a package that adjusts the text
margins in wide windows. The goal is to have buffers that use
visual-line-mode wrap text at `fill-column`, not at the window's edge. If
the window configuration or size changes, the width of the margins needs to
be recalculated, so the package puts a function on the relevant hooks to do
that.

If the text size is changed, the window margins also need to be
recalculated, but this does not happen automatically. The README of
`visual-fill-column` suggests to advise the function
`text-scale-adjust` to solve this (see
https://codeberg.org/joostkremers/visual-fill-column#adjusting-text-size),
but it's an ad hoc solution, of course.

-- 
Joost Kremers
Life has its moments

Re: Should text-scale trigger hooks?

[Eli Zaretskii]

> From: Joost Kremers <joostkremers@fastmail.fm>
> Cc: James Cherti <contact@jamescherti.com>,  emacs-devel@gnu.org
> Date: Thu, 07 Nov 2024 09:21:01 +0100
> 
> On Thu, Nov 07 2024, Eli Zaretskii wrote:
> > The functions which manipulate face-remapping-alist force redisplay of
> > the buffer in which the alist was changed, and from Emacs display POV
> > this is the only thing that should happen to make sure the affected
> > buffer's display is redrawn.  Please explain why buffer's redisplay is
> > not enough for the packages you mention to do their job.
> 
> Speaking for `visual-fill-column`: it's a package that adjusts the text
> margins in wide windows. The goal is to have buffers that use
> visual-line-mode wrap text at `fill-column`, not at the window's edge. If
> the window configuration or size changes, the width of the margins needs to
> be recalculated, so the package puts a function on the relevant hooks to do
> that.

The need for adjusting the margins in this case is specific to how the
visual-fill-column package uses the margins, it is not a general need.
In general, whether the margins need to be adjusted according to text
scale depends on what are margins used for; they can be used for more
than just showing text of the same default face.  This is why Emacs
doesn't automatically resize the margins in this case.

> If the text size is changed, the window margins also need to be
> recalculated, but this does not happen automatically. The README of
> `visual-fill-column` suggests to advise the function
> `text-scale-adjust` to solve this (see
> https://codeberg.org/joostkremers/visual-fill-column#adjusting-text-size),
> but it's an ad hoc solution, of course.

What are the problems of using text-scale-mode-hook?  Or even using
add-variable-watcher to cause Emacs call your function when
face-remapping-alist changes?

IOW, why do you think this is a core Emacs issue, not an issue with
how visual-fill-column is implemented?

Re: Should text-scale trigger hooks?

[Joost Kremers]

On Thu, Nov 07 2024, Eli Zaretskii wrote:
> The need for adjusting the margins in this case is specific to how the
> visual-fill-column package uses the margins, it is not a general need.

Yes, true.

> What are the problems of using text-scale-mode-hook?  Or even using
> add-variable-watcher to cause Emacs call your function when
> face-remapping-alist changes?

I haven't actually tried those methods, since I rarely use
text-scale-adjust and James's is the first user request I received about it
in a long time.

> IOW, why do you think this is a core Emacs issue, not an issue with
> how visual-fill-column is implemented?

Reading your message, I'm not sure that it is. But I guess this is also a
question for James.

-- 
Joost Kremers
Life has its moments

Re: Should text-scale trigger hooks?

[James Cherti]

I tested adding a function to `text-scale-mode-hook` that executes:
(run-hooks 'window-configuration-change-hook)

And it solved the issue in both packages (`eat` and `visual-fill-column`).

After further consideration, I believe that this is not an Emacs core 
issue, but rather a problem caused by the implementation of those 
packages. It doesn't make sense for text scaling to run the 
window-configuration-change-hook functions.

--
James Cherti
https://www.jamescherti.com/

On 2024-11-07 04:42, Joost Kremers wrote:
> On Thu, Nov 07 2024, Eli Zaretskii wrote:
>> The need for adjusting the margins in this case is specific to how the
>> visual-fill-column package uses the margins, it is not a general need.
> 
> Yes, true.
> 
>> What are the problems of using text-scale-mode-hook?  Or even using
>> add-variable-watcher to cause Emacs call your function when
>> face-remapping-alist changes?
> 
> I haven't actually tried those methods, since I rarely use
> text-scale-adjust and James's is the first user request I received about it
> in a long time.
> 
>> IOW, why do you think this is a core Emacs issue, not an issue with
>> how visual-fill-column is implemented?
> 
> Reading your message, I'm not sure that it is. But I guess this is also a
> question for James.
> 

RE: [External] : Re: Should text-scale trigger hooks?

[Drew Adams]

Coming late to this thread, and with nothing
special to say about the specific trigger/hook
question.

But with something to offer perhaps for the
original/underlying request of allowing a buffer
in `visual-line-mode' to have its window change
size to accommodate text scaling, i.e., keep the
same visual-fill position.

IOW, maybe think of the current discussion as
possibly an X-Y question: you have an idea of
what a solution to your original question might
be, so you ask a question about that (trigger).

If I've misunderstood something, then please
ignore...
___

I brought up the original question decades ago,
as soon as text-scaling was added to Emacs.
Not wrt visual-line-mode, in particular (that
didn't exist back then), but generally.

I suggested that users should be able to
choose whether the window gets automatically
scaled in conjunction with text scaling, i.e.,
as an optional behavior.

For one thing, this saves horizontal screen
space when text is shrunk - wasted blank window
space at line ends.

This was roundly rejected by Eli at the time,
IIRC, saying that no one would ever want such
behavior.

So I implemented it in a little library,
`face-remap+.el' (back in 2009).  I wouldn't
be without it.

It works with and without `visual-line-mode'.
It really has _nothing to do with
`visual-line-mode'_: the question/request is
only about keeping the relation between the
text size and the window size.  It just so
happens that with `visual-line-mode' lines
are visually fit to the window width.


The behavior is controlled by option
`text-scale-resize-window':

text-scale-resize-window is a variable defined in `face-remap+.el’.

Non-nil means text scaling resizes the window or frame accordingly.
For example, if you use ‘C-x C--’ (‘text-scale-decrease’)’ to make the
text smaller, then the window or frame is made smaller by a similar
factor.
___

You can choose to resize the window horizontally
only, vertically only, both, or neither (off).

(There's also a global minor mode,
`text-scale-keep-mode', which when on keeps the
same text-scaling when a buffer changes major mode.
If off, you get the vanilla Emacs behavior that
text-scaling is lost when the major mode changes.)
___

Apologetically, `face-remap+.el' redefines vanilla
function `text-scale-increase', to respect option
`text-scale-resize-window'.

I'd still suggest that such an option be added to
vanilla Emacs, FWIW.  (Better late than never...)
___

https://www.emacswiki.org/emacs/download/face-remap%2b.el

Re: Should text-scale trigger hooks?

[Joost Kremers]

On Wed, Nov 06 2024, James Cherti wrote:
> I've implemented a similar solution for visual-fill-column (see
> https://codeberg.org/joostkremers/visual-fill-column/pulls/16 ), but the
> visual-fill-column maintainer thinks this is probably not the right
> solution to this problem. I tend to agree with him.

To be precise, for that solution to work, the function that has to run
`window-configuration-change-hook` is itself in that hook. Even though it
seemed to work, it doesn't seem like the right solution.


-- 
Joost Kremers
Life has its moments