Dependent colours

Dependent colours

[Lars Ingebrigtsen]

In many situations, packages and users need to specify colours to use
for something.  We generally defer this to the face machinery, because
which colours to use for something might depend on whether you have a
light or dark background, or whether it's a GUI Emacs or a terminal one,
etc.

This works well as a machinery, but it's pretty cumbersome when the user
just wants to specify a list of colours.

So I'm wondering whether we could create something that would make
things easier for the user here, and if anybody had any thoughts on the
issue.

Off the top of my head...  if we had something like

(make-color "red")

(make-color ((background light) "blue")
            ((background dark) "red"))

Uhm...  Or something...  better...  :-)

Anyway, code that uses these colour objects would then create real
(dynamic) faces out of them, eventually, so that there'd be no change in
the display engine (I think).

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

Re: Dependent colours

[Lars Ingebrigtsen]

Lars Ingebrigtsen <larsi@gnus.org> writes:

> (make-color ((background light) "blue")
>             ((background dark) "red"))
>
> Uhm...  Or something...  better...  :-)

I think we get to 99.7% of the use cases with just light/dark.

So (make-color "blue" "red") would probably be the shortest syntax we
could use here.  (The first for light backgrounds and the second for
dark.)  Or just (color "blue" "red"), since the point is to have
something short and convenient for people.

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

Re: Dependent colours

[Eli Zaretskii]

> From: Lars Ingebrigtsen <larsi@gnus.org>
> Date: Thu, 14 Apr 2022 15:00:55 +0200
> 
> In many situations, packages and users need to specify colours to use
> for something.  We generally defer this to the face machinery, because
> which colours to use for something might depend on whether you have a
> light or dark background, or whether it's a GUI Emacs or a terminal one,
> etc.
> 
> This works well as a machinery, but it's pretty cumbersome when the user
> just wants to specify a list of colours.
> 
> So I'm wondering whether we could create something that would make
> things easier for the user here, and if anybody had any thoughts on the
> issue.
> 
> Off the top of my head...  if we had something like
> 
> (make-color "red")
> 
> (make-color ((background light) "blue")
>             ((background dark) "red"))
> 
> Uhm...  Or something...  better...  :-)

I'm not sure I understand: we already have anonymous faces one can get
by specifying just the colors, as in

  '(:foreground "red" :background "blue")

Why is that not enough?

Re: Dependent colours

[Lars Ingebrigtsen]

Eli Zaretskii <eliz@gnu.org> writes:

> I'm not sure I understand: we already have anonymous faces one can get
> by specifying just the colors, as in
>
>   '(:foreground "red" :background "blue")
>
> Why is that not enough?

It's not about foreground or background colours, it's about choosing a
colour based on features of the current frame, like defface.

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

Re: Dependent colours

[Stefan Monnier]

> So (make-color "blue" "red") would probably be the shortest syntax we
> could use here.  (The first for light backgrounds and the second for
> dark.)  Or just (color "blue" "red"), since the point is to have
> something short and convenient for people.

You mean like

    (defun make-color (light dark)
      (if (frame--current-backround-mode (selected-frame)) dark light))

Or do you have something more sophisticated in mind?


        Stefan

Re: Dependent colours

[Lars Ingebrigtsen]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

> You mean like
>
>     (defun make-color (light dark)
>       (if (frame--current-backround-mode (selected-frame)) dark light))
>
> Or do you have something more sophisticated in mind?

The latter.  Usages of these objects have to create a face eventually so
that they'll be displayed dynamically differently on different frames.

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

RE: [External] : Re: Dependent colours

[Drew Adams]

YAGNI?

I haven't see a good presentation of a need / use case.

Have you heard users complaining about some missing
feature in this regard?

Re: Dependent colours

[Lars Ingebrigtsen]

Lars Ingebrigtsen <larsi@gnus.org> writes:

>>     (defun make-color (light dark)
>>       (if (frame--current-backround-mode (selected-frame)) dark light))
>>
>> Or do you have something more sophisticated in mind?
>
> The latter.  Usages of these objects have to create a face eventually so
> that they'll be displayed dynamically differently on different frames.

Or wherever -- we might be making an SVG dynamically, and want to
postpone calculating the colour until we know what frame we're showing
the SVG on.  (Etc.)

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

Re: Dependent colours

[Eli Zaretskii]

> From: Lars Ingebrigtsen <larsi@gnus.org>
> Date: Thu, 14 Apr 2022 15:56:53 +0200
> Cc: emacs-devel@gnu.org
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> > I'm not sure I understand: we already have anonymous faces one can get
> > by specifying just the colors, as in
> >
> >   '(:foreground "red" :background "blue")
> >
> > Why is that not enough?
> 
> It's not about foreground or background colours, it's about choosing a
> colour based on features of the current frame, like defface.

But the result should be a face, right?  So you want an easy-to-use
facility to create a face that depends on the background-mode?  Or am
I misunderstanding again?  And if I am misunderstanding, can you
please describe the use case in more detail?

Re: Dependent colours

[Lars Ingebrigtsen]

Eli Zaretskii <eliz@gnu.org> writes:

> But the result should be a face, right?  

It could be a face, but not necessarily.  Dynamically creating an SVG
with a different foreground colour depending on the frame, for instance.

> So you want an easy-to-use facility to create a face that depends on
> the background-mode?  Or am I misunderstanding again?  And if I am
> misunderstanding, can you please describe the use case in more detail?

defface is easy enough to use if you want to create a face, but here the
interface we're presenting the user with is tied to colours, not faces.

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

Re: Dependent colours

[Eli Zaretskii]

> From: Lars Ingebrigtsen <larsi@gnus.org>
> Cc: emacs-devel@gnu.org
> Date: Thu, 14 Apr 2022 18:13:56 +0200
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> > But the result should be a face, right?  
> 
> It could be a face, but not necessarily.  Dynamically creating an SVG
> with a different foreground colour depending on the frame, for instance.

Are there any other examples besides the background of SVG?  Because
the latter is AFAIR very special, and we can now handle it through
faces as well.  So the question is: do we need to do this outside of
the faces realm?  I cannot think of any other case where we use colors
not via faces.

> > So you want an easy-to-use facility to create a face that depends on
> > the background-mode?  Or am I misunderstanding again?  And if I am
> > misunderstanding, can you please describe the use case in more detail?
> 
> defface is easy enough to use if you want to create a face, but here the
> interface we're presenting the user with is tied to colours, not faces.

Interactively or non-interactively?

Anyway, colors in Emacs are face attributes.

Re: Dependent colours

[Lars Ingebrigtsen]

Eli Zaretskii <eliz@gnu.org> writes:

> Anyway, colors in Emacs are face attributes.

Colours are colours, but we usually use the face machinery to specify
them, and we (almost always) instantiate them on the screen via the face
machinery.  That doesn't mean that colours per se are face attributes.

The reason I started thinking about these things, once again, is because
vtable allows you to specify (alternating) row and column background
colours, and where they intersect, vtable computes the blend of colours.
The obvious way people to use this is to say

:column-colors '("red" "blue")
:row-colors '("green" "yellow")

and that's it.  But pondering this a bit more, perhaps using faces here
would be nicer anyway -- the package writer may be wanting to use
alternating foreground colours on the rows anyway.  So

:column-colors '((:background "red") (:background "blue"))
:row-colors '((:foreground "green") (:foreground "yellow"))

is more expressive.  (And vtable will just have to check both foreground
and background colours on the faces and blend accordingly, but that's
just a small matter of programming.)

And besides, using faces for things like this has the advantage of
fitting into the theming architecture.  I mean, if it's

:column-colors '(list-timer-column-even list-timer-column-odd)

and those are faces, we don't have to extend the theming machinery to
handle theming of colours.

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

Re: Dependent colours

[Eli Zaretskii]

> From: Lars Ingebrigtsen <larsi@gnus.org>
> Cc: emacs-devel@gnu.org
> Date: Fri, 15 Apr 2022 10:46:03 +0200
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> > Anyway, colors in Emacs are face attributes.
> 
> Colours are colours, but we usually use the face machinery to specify
> them, and we (almost always) instantiate them on the screen via the face
> machinery.  That doesn't mean that colours per se are face attributes.

I'm probably missing something: what other means do we have in Emacs
to produce colors on display, except via faces?

> :column-colors '("red" "blue")
> :row-colors '("green" "yellow")
> 
> and that's it.  But pondering this a bit more, perhaps using faces here
> would be nicer anyway -- the package writer may be wanting to use
> alternating foreground colours on the rows anyway.  So
> 
> :column-colors '((:background "red") (:background "blue"))
> :row-colors '((:foreground "green") (:foreground "yellow"))
> 
> is more expressive.  (And vtable will just have to check both foreground
> and background colours on the faces and blend accordingly, but that's
> just a small matter of programming.)
> 
> And besides, using faces for things like this has the advantage of
> fitting into the theming architecture.  I mean, if it's
> 
> :column-colors '(list-timer-column-even list-timer-column-odd)
> 
> and those are faces, we don't have to extend the theming machinery to
> handle theming of colours.

If what you have in mind can play nicely with faces (and I think it
can, but I'm not yet sure I understand the issue, see above), then
yes, using face machinery is the best, as we have there a lot of
functionalities that can be used without effort.

Re: Dependent colours

[Lars Ingebrigtsen]

Eli Zaretskii <eliz@gnu.org> writes:

>> Colours are colours, but we usually use the face machinery to specify
>> them, and we (almost always) instantiate them on the screen via the face
>> machinery.  That doesn't mean that colours per se are face attributes.
>
> I'm probably missing something: what other means do we have in Emacs
> to produce colors on display, except via faces?

I think I mentioned SVGs already?  But we also have the mouse pointer,
and tooltip colours, and various bits and bobs in the toolkits that take
colours as parameters (for menus and scroll bars and etc).

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

Re: Dependent colours

[Po Lu]

Lars Ingebrigtsen <larsi@gnus.org> writes:

> I think I mentioned SVGs already?

That's handled by librsvg and is AFAICT outside the control of Emacs.  A
better example would be XPM images, where the color allocation is in
fact done by Emacs.

> But we also have the mouse pointer, and tooltip colours, and various
> bits and bobs in the toolkits that take colours as parameters (for
> menus and scroll bars and etc).

The tooltip colors are defined via faces, while the menu and scroll bar
colors are in fact supposed to be defined via faces, if you set aside
the frame parameters.

The `mouse' face and its associated frame parameter is unfortunately
unlikely to work under the common X sessions these days.

Re: Dependent colours

[Lars Ingebrigtsen]

Po Lu <luangruo@yahoo.com> writes:

> > I think I mentioned SVGs already?
> 
> That's handled by librsvg and is AFAICT outside the control of Emacs.

We can create any SVGs we want via svg.el and display them.

> The tooltip colors are defined via faces, while the menu and scroll bar
> colors are in fact supposed to be defined via faces, if you set aside
> the frame parameters.

We currently define them via the defface machinery.  But we do not use
faces to instantiate the colours on the screen, which is what Eli asked
for examples of.

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

Re: Dependent colours

[Eli Zaretskii]

> From: Lars Ingebrigtsen <larsi@gnus.org>
> Cc: emacs-devel@gnu.org
> Date: Fri, 15 Apr 2022 11:48:31 +0200
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> >> Colours are colours, but we usually use the face machinery to specify
> >> them, and we (almost always) instantiate them on the screen via the face
> >> machinery.  That doesn't mean that colours per se are face attributes.
> >
> > I'm probably missing something: what other means do we have in Emacs
> > to produce colors on display, except via faces?
> 
> I think I mentioned SVGs already?  But we also have the mouse pointer,
> and tooltip colours, and various bits and bobs in the toolkits that take
> colours as parameters (for menus and scroll bars and etc).

Some of those are faces, and the rest are not implemented by us.

Re: Dependent colours

[Po Lu]

Lars Ingebrigtsen <larsi@gnus.org> writes:

> We currently define them via the defface machinery.  But we do not use
> faces to instantiate the colours on the screen, which is what Eli asked
> for examples of.

In most cases we do: the colors are allocated in xfaces.c upon face
realization (see the calls to load_color2 and unload_color.)  The image
code path is an exception, and I think inside SVGs it isn't even
possible to use the same color names as in the rest of Emacs.