Re: Introducing face in comments for various modes

[Heime]

------- Original Message -------
On Monday, December 12th, 2022 at 10:50 AM, Thibaut Verron 
<thibaut.verron@gmail.com> wrote:


> On 12/12/2022 11:20, Heime wrote:
> 
> > ------- Original Message -------
> > On Monday, December 12th, 2022 at 9:58 AM, Thibaut 
> > Verronthibaut.verron@gmail.com wrote:
> > 
> > > On 12/12/2022 10:21, Heime wrote:
> > > 
> > > > ------- Original Message -------
> > > > On Monday, December 12th, 2022 at 8:49 AM, Thibaut Verron
> > > > thibaut.verron@gmail.com wrote:
> > > > 
> > > > > Le lun. 12 déc. 2022 à 04:01, Heimeheimeborgia@protonmail.com a
> > > > > écrit :
> > > > > 
> > > > > ------- Original Message -------
> > > > > On Monday, December 12th, 2022 at 2:24 AM, Heime
> > > > > heimeborgia@protonmail.com wrote:
> > > > > 
> > > > > > ------- Original Message -------
> > > > > > On Sunday, December 11th, 2022 at 5:40 PM, Stefan Monnier via
> > > > > > Users list for the GNU Emacs text editorhelp-gnu-emacs@gnu.org
> > > > > > wrote:
> > > > > > 
> > > > > > > BTW, there is a related convention in ELisp code where
> > > > > > > comments that
> > > > > > > start in column 0 and which are introduced with 3 or more
> > > > > > > semi-colons
> > > > > > > are considered sectioning headers (where ";;;" means a top-level
> > > > > > > header, ";;;;" a subheader, ";;;;;" a subsubheader, ...).
> > > > > > > 
> > > > > > > I'd be happy if Emacs were changed to highlighting those.
> > > > > > > 
> > > > > > > Stefan
> > > > > > > If you are colourising "Sectioning Headers", ensure that vibrant
> > > > > > > and good contrast:
> > > > > 
> > > > > 1) betweenthe text and the background;
> > > > > 
> > > > > 2) and between a header, subheader, subsubheader, ...
> > > > > 
> > > > > Use some colour metric (e.g. using the Web Content Accessibility
> > > > > Guidelines [WCAG]).
> > > > > 
> > > > > Because I consistently see that developers almost never care (or
> > > > > have the skills)
> > > > > to properly set up colours. Have suggested changing the colour
> > > > > scheme as described,
> > > > > for "Org Headings" because they are indistinguishable against a
> > > > > dark background and
> > > > > between a heading and its subheading. Applying such metrics have
> > > > > been turned down,
> > > > > with the excuse that if I want them right, I have to work on
> > > > > emacs customisations
> > > > > myself, as the crappy colours are there to stay.
> > > > > 
> > > > > > The colors of the standard themes are chosen with its (light)
> > > > > > background in mind. If you change that background, it is not
> > > > > > surprising that things fall apart.
> > > > > > Choosing colours with a light background in mind is the wrong 
> > > > > > approach
> > > > > > because colours produce far greater visual
> > > > > > impact.
> > > 
> > > There is no right or wrong approach, but individual preferences.
> > > Standard metrics exist. The Gnu Project like many others, does not
> > > want to use them.
> 
> You're moving the goalpost: the sentence I quoted claimed that "focusing
> on a light background is the wrong approach".

Having worked with modus-themes, it was concluded that there exist greater
variations in possibilities with a dark background that with a light one.
 
> It's also factually not true that the GNU project does not care about
> readability metrics, especially now that the Modus themes are shipped
> with Emacs.
> 
> > > If you want a dark background, just use a dark background theme. For
> > > instance, emacs has a built-in implementation of the tango dark color
> > > palette. If contrast if your primary concern, you should look at the
> > > modus themes (modus-vivendi for the dark background), which is also part
> > > of emacs now.

Yes, contrast is my primary concern on some packages.

> > > M-x customize-themes and make your choice.
> > > If you use "modus-vivendi" for org-mode, the colours are all almost white,
> > > a big problem particularly when you fold the org headings.
> 
> 
> I don't like dark backgrounds, but it seems perfectly readable to me.

I personally always cater for both light and dark theme.
 
> Anyway, from personal experience the developer of the Modus themes is
> extremely responsive. If you have a problem with his themes you should
> take it up with him.

Will let him know.  Perhaps there are some problems there he can see.
 
> > > > Rather, there there should be carefully chosen colour settings for
> > > > both light and dark backgrounds.
> > > 
> > > That's how you end up with settings which are at best acceptable, but
> > > not perfect, for both light and dark backgrounds. The range of colors
> > > which are suitable for both light and dark backgrounds is just too narrow.
> > > 
> > > The proper way is the current way: carefully curated themes implementing
> > > all colors in a consistent ways.
> > 
> > What metrics are being used. The blind belief that the proper way is the
> > current way, is the origin of the problem.
> 
> 
> For one, the Modus themes were developed with quantified metrics
> (minimal contrast ratio afaik), and they are two completely different
> themes for black and light backgrounds.
> 
> If you think you can do better, you are welcome to try. But if you come
> and claim that the current way is the wrong way, the burden of proof is
> on you. :)

I worked on the enhancement of modus-themes.
 
> > > > > > It is not a new problem, but it doesn't mean that you have to
> > > > > > customize all the individual faces yourself. Instead, you should 
> > > > > > look
> > > > > > for a theme implementing
> > > 
> > > > > > the colors you like, and install it. The responsibility for having
> > > > > > consistent colors across all emacs fonts is on the theme designer.
> > > > > > You can still tweak some
> > > > > > faces from there if you choose to of course.
> > > > > > At any rate, Stefan's suggestion would not require making new design
> > > > > > choices, as there are already faces designed for fontifying headers:
> > > > > > outline-1, outline-2, etc.
> > > > > > Making a new design choice is a necessity if you want to move 
> > > > > > forward.
> > > 
> > > No. The question is whether to fontify those headers, how to identify
> > > them, etc.
> > > 
> > > That's completely separate from the question of changing the face
> > > currently used for headers in other places.
> > > 
> > > > > Those faces are used by outline-mode, but not by outline-minor-mode
> > > > > (which emacs-lisp-mode uses to implement the ;;; comment headers) at
> > > > > the moment.
> > > > > Which proves my point that changes are necessary. What needs to be
> > > > > done is for colour contrast metrics
> > > > > to be taken seriously by all packages, rather than relying on some
> > > > > theme to fix the crappy default choices.
> > > 
> > > Sorry to be blunt, but you couldn't be more wrong. For a start,
> > > outline-mode and outline-minor-mode are the same package. :)
> > > 
> > > But more to the point, with the current system, packages choose existing
> > > faces to implement coloring based on what they should color (e.g. is
> > > it a comment, is it a header, is it a keyword, is it something
> > > important). And the theme designers choose colors (and other features)
> > > for those faces.
> > > 
> > > As a result, colors are the same across all of Emacs (for example
> > > comments look the same in elisp and python), and -- if the theme maker
> > > is competent -- the colors will implement good contrast and be readable
> > > everywhere.
> > > One can at least use good metrics for light (white) and dark (dark) 
> > > background.
> > > We have not even arrived at that yet. I am not arguing against comments 
> > > looking
> > > the same, but that there should specific settings for the canonical white 
> > > and
> > > black background as minimum.
> 
> 
> And I am telling you that there are.
> 
> For light background: the default theme, leuven, tango, modus-operandi
> 
> For dark background: the default theme with inverse-video, tango-dark,
> modus-vivendi
> 
> 
> > > If instead we were to let each package decide on its colors, Emacs would
> > > look like a Christmas tree with different colors all over the place. And
> > > most of them would be really crappy because the package developer was
> > > never trained in graphic design, or because they didn't plan for all
> > > possible background colors (it's not as simple as light and dark, some
> > > people use blue, or green backgrounds), or because they didn't predict
> > > that their choice of color would conflict with the choice made by a
> > > minor mode in another package, or...
> > > Am only discussing for white background and black background, which are 
> > > the
> > > canonical settings for printing. With colour contrast you are limited by
> > > the metric values which limits to about eight colours.
> > > It is not Christmas Tree as you say. Focusing on any possible colour 
> > > combination
> > > (blue, or green backgrounds) is beyond the scope of my discussion.
> 
> 
> No it's not. My point is that if you leave the responsibility of
> choosing colors to packages as opposed to themes, it will be a Christmas
> tree and there will be unpredictable combinations. It's a direct
> consequence of your idea, you can't just wave it off.
> 
> It's already bad enough now with some packages defining their own faces
> without at least inheriting from the standard ones.

Right.  My focus has been to provide colours with good metrics so that 
people inherit from the standard ones.

 
> There are currently 5330 packages on Melpa. Do you plan to contact the
> authors of all of them individually to get them to implement your
> preferred colors?

To start using actual standards.  Absolutely, they should learn more and change
their packages.  
 
> With the current approach, on the other hand, it's very easy: report a
> bug for the theme you're using, or make your own theme if you really
> want to.
> 
> > > You shouldn't think of themes as "fixing the default choices"
> > > (especially considering that you are the one "breaking" them by
> > > insisting to use them with a background they weren't designed for).
> > > Their purpose is to implement different choices in a consistent way.
> > > Good design in much more important that consistency.
> 
> It's also much easier to achieve in a consistent system.

The argument is to design separately for light and also for dark background.
Emacs does have light and dark checks.
 
> > > > > > > Heime [2022-12-11 15:35:41] wrote:
> > > > > > > 
> > > > > > > > The following uses `hi-lock` to change the foreground of
> > > > > > > > comments matching
> > > > > > > > a regexp. This is implemented for emacs-lisp files where
> > > > > > > > comments start
> > > > > > > > with ";;".
> > > > > > > > 
> > > > > > > > I would like to extend this for other programming languages
> > > > > > > > besides emacs-lisp
> > > > > > > > files, using the relevant comment character automatically
> > > > > > > > for that language.
> > > > > > > > (defface elfa-face
> > > > > > > > '((t :foreground "magenta"))
> > > > > > > > "Face for comment headings.")
> > > > > > > > 
> > > > > > > > (defun elfa-regexp (&optional actm)
> > > > > > > > "Identify comment category ';; [Category]'."
> > > > > > > > (highlight-regexp
> > > > > > > > "^;;\s+\\[.+\\].*$" 'elfa-face))
> > > > > > > > 
> > > > > > > > (defun elfa-category ()
> > > > > > > > "TODO."
> > > > > > > > (interactive)
> > > > > > > > (add-to-list 'auto-mode-alist '("\\.el\\'" . hi-lock-mode))
> > > > > > > > (add-hook 'emacs-lisp-mode-hook 'hi-lock-mode t)
> > > > > > > > (add-hook 'hi-lock-mode-hook 'elfa-regexp))