Modus themes 2.5.0 for GNU Emacs
I will now prepare the patch for emacs.git which will then trickle down
to GNU ELPA (the
modus-themes is a
:core package). Thank you for
This entry documents the changes made to the project since the publication of version 2.4.0 on 2022-06-01. It spans more than 60 commits to an already stable project.
modus-vivendi themes are built into Emacs-28
(latest stable release) or later, and are available on GNU ELPA as well
as other archives. Emacs-28 ships version 1.6.0, while the current
master branch (i.e. Emacs-29) and, by extension, GNU ELPA include the
latest tagged release. The packaged version is available as
Read the manual inside Emacs by evaluating:
(info "(modus-themes) Top")
Or visit: https://protesilaos.com/emacs/modus-themes (the website only documents the latest version).
Enhancement to the user option ‘modus-themes-headings’
The user option
modus-themes-headings now reads a level 0 heading in
addition to numbers 1–8. Heading 0 accepts the same list of properties
as all other levels (please consult the doc string of the user option or
the corresponding entry in the manual). Currently only the value of the
Org #+title is affected (face is
org-document-title), but we may cover
more faces if needed.
;; The `modus-themes-headings' is an alist with lots of possible ;; combinations, including per-heading-level tweaks: read the ;; manual or its doc string. (setq modus-themes-headings '((0 . (variable-pitch light (height 2.2))) (1 . (rainbow variable-pitch light (height 1.6))) (2 . (rainbow variable-pitch light (height 1.4))) (3 . (rainbow variable-pitch regular (height 1.3))) (4 . (rainbow regular (height 1.2))) (5 . (rainbow (height 1.1))) (t . (variable-pitch extrabold)))
Given this change, I am also tweaking the default foreground value of
org-document-title. It is a bit more saturated than before, but
remains close to the spirit of the previous one.
Thanks to Rudolf Adamkovič for proposing the idea on the mailing list: https://lists.sr.ht/~protesilaos/modus-themes/%3Cm2y1x5tewl.email@example.com%3E.
Stylistic tweak to the user option ‘modus-themes-syntax’
alt-syntax property from desaturating the effect of the
yellow-comments property when the two would be combined. Such as:
(setq modus-themes-syntax '(alt-syntax yellow-comments))
The previous design was incorrect because it was always using the faint variant of the yellow comments, as if the user had specified:
(setq modus-themes-syntax '(alt-syntax faint yellow-comments))
[ Read the doc string of
modus-themes-syntax or the manual for an
explanation of all properties and their combinations. ]
Review of the Isearch (and related) colours
Emacs’ standard search has a face for the currently matched query and
all its inactive matches. The faces are
respectively. Before, we were using a green background by default for
isearch face and a cyan background for the
was a choice that was made in the early days of the project when the
palette was not yet fully realised.
Green and cyan do not always contrast well side-by-side (subject to
hardware capabilities and environmental lighting), so the
also had an added bold weight. This was not my preference, but it was
necessary under the circumstances. The previous combinations were also
not ideal when the user option
modus-themes-deuteranopia was set to a
non-nil value: the blue background which was used instead of the green
one could be conflated with the subtle teal of the
under certain circumstances, such as poor colour reproduction at the
monitor level or in terminal emulators with limited colour support.
The new colours (intense yellow for active matches and subtle cyan for lazy ones) are complementary, meaning that they are naturally easy to tell apart.
[ Read “Colour theory and techniques used in the Modus themes”: https://protesilaos.com/codelog/2022-04-21-modus-themes-colour-theory/ ]
These specific hues are also well-suited for users with red-green colour
deficiency: yellow stays as-is, while the cyan colour becomes a bit more
grey though remains distinct. As such, we do not need to run the helper
modus-themes--deuteran to set the style based on the value of
The new colours do not clash with the style of the relevant
M-x grep, and related), nor with the various
permutations of the
region face (subject to the user option
Finally, the bold weight has been removed from the
isearch face. It
was always a kludge. Also, it would make paragraphs rendered in the
variable-pitch face (or proportional fonts in general) jump around as
the user would move between the matches, because bold letters occupy
more space than their regular weight counterparts so they affect the
length of the line. This problem was reported by Augusto Stoffel on the
mailing list: https://lists.sr.ht/~protesilaos/modus-themes/%3C87sfnbswe9.firstname.lastname@example.org%3E.
Rewrote parts of the colour preview commands
commands that produce a buffer which shows previews of every entry in
the palette. Their code has been simplified and they now produce a
warning when the display terminal has limited colour support.
Furthermore, they read any overrides as specified in the user options
The “summertime” re-spin of colour overrides
The manual now includes a complete hand-crafted example of a pair of themes that override the default palette. This is done as a technology demonstration. It is not considered an “official” extension of the Modus themes and will never be part of the code base as it does not conform with our lofty accessibility standards. However, I took great care in picking the colour overrides in the hope that users will (i) have a usable theme, should they opt for it, and (ii) they recognise the potential of our colour-overriding feature.
Screenshots and related information: https://protesilaos.com/codelog/2022-07-26-modus-themes-color-override-demo/.
Thanks to user “Summer Emacs” for (i) suggesting the name “summertime”, (ii) testing variants of this in her setup, and (iii) sending me feedback on possible tweaks and refinements. All errors are my own.
The idea for this project came from an exchange where Summer discovered an old theme of mine (from my pre-Emacs days) and asked if I had anything like it for Emacs. Voilà!
[ This information is shared with permission. ]
As for whether I have more plans… “Perhaps!” ;)
Removed support for certain packages or face groups
I periodically install and use the packages we support to see if they have any updates we need to cover but also to confirm that they work. Usually, the user does not learn about this work, as I don’t need to make any changes or will make some minor tweaks. When I think that the package is not in a good shape, I remove it from the list of explicitly supported packages, meaning that the modus-themes no longer cover the faces it defines. The removal of any package is done on a case-by-case basis. If you disagree with this decision, please inform me about and I shall reconsider.
centaur-tabs :: Those of you who have been reading these release notes are aware of a bug in centaur-tabs which basically prevents us from using the standard
:inheritattribute to style the centaur-tabs faces. I have sent a patch to fix it, but have received no response since February: https://github.com/ema2159/centaur-tabs/pull/179. To me, this gives the package the “unmaintained” status, though I am happy to revert the change as soon as it gets the maintenance it needs.
Relevant reports (and I got many others in my private inbox):
cursor-flash :: its default face should be visible enough.
dynamic-ruler :: The package does not build on my Emacs 29. Also, its default faces are usable even without our recolouring.
emacs-dashboard :: Its default faces inherit from basic faces that we already support.
frog-menu :: I have not seen this package being used anywhere. I suspect it is because it has not found a niche between transient, hydra, and embark.
mct :: A few months ago I announced that its development is discontinued. Either use vertico or switch to what Emacs provides as a built-in option: https://protesilaos.com/codelog/2022-04-14-emacs-discontinue-mct/.
org-treescope :: The package points to a GitHub repo, which is archived. The current source is on GitLab, but the package is not updated accordingly. This makes me believe it is not actively maintained and am thus removing it from the list.
paradox :: When I tried paradox, it took over my C-c g binding which I have for Magit. As an Emacs user, I consider this an unacceptable transgression. Looking at paradox’s git repo, the project is not maintained. If things change, I am happy to reinstate support for it.
vc-annotate (built-in) :: It has not been working properly for a long time now. Colours are unset and are not re-applied when switching between the
Furthermore, the way
vc-annotate-background-modeputs us in an awkward spot: when the mode is non-nil, the mapped values are used as backgrounds WITHOUT giving us the chance to make the appropriate adjustments to the foreground (so we end up with inaccessible colour combinations). This means that we must fix a problem which is not ours by overriding the user option of the background altogether. A theme outright disabling user options is bad form.
Even documenting a user-level set of configurations will not suffice, as the results are unreliable. I tried the code which I copy further below to test annotation with/without background, plus the change in values when switching between modus-operandi and modus-vivendi. Again, colours are not updated properly (I know the buffer of
M-x vc-annotateneeds to be generated again), as
modus-operandimay retain the values set by
Ultimately, I feel
vc-annotateneeds to be refactored to use ordinary faces in ordinary ways. Or, at least, not try to outsmart the user/theme about the choice of colours.
Thanks to Philip Kaludercic for starting the thread about the
vc-annotate-background-modewhich reminded me about this problem: https://lists.sr.ht/~protesilaos/modus-themes/%3C875ylfxkgi.email@example.com%3E.
The code I alluded to:
(setq vc-annotate-background-mode nil) (defun my-modus-themes-vc-annotate () ;; Actual values are for demo purposes (modus-themes-with-colors (if vc-annotate-background-mode (setq vc-annotate-background bg-alt vc-annotate-color-map `((20 . ,red-intense-bg) (40 . ,red-subtle-bg) (60 . ,red-refine-bg) (80 . ,yellow-intense-bg) (100 . ,yellow-subtle-bg) (120 . ,yellow-refine-bg) (140 . ,magenta-intense-bg) (160 . ,magenta-subtle-bg) (180 . ,magenta-refine-bg) (200 . ,cyan-intense-bg) (220 . ,cyan-subtle-bg) (240 . ,cyan-refine-bg) (260 . ,green-intense-bg) (280 . ,green-subtle-bg) (300 . ,green-refine-bg) (320 . ,blue-intense-bg) (340 . ,blue-subtle-bg) (360 . ,blue-refine-bg))) (setq vc-annotate-background nil vc-annotate-color-map `((20 . ,red) (40 . ,magenta) (60 . ,magenta-alt) (80 . ,red-alt) (100 . ,yellow) (120 . ,yellow-alt) (140 . ,fg-special-warm) (160 . ,fg-special-mild) (180 . ,green) (200 . ,green-alt) (220 . ,cyan-alt-other) (240 . ,cyan-alt) (260 . ,cyan) (280 . ,fg-special-cold) (300 . ,blue) (320 . ,blue-alt) (340 . ,blue-alt-other) (360 . ,magenta-alt-other)))))) (add-hook 'modus-themes-after-load-theme-hook #'my-modus-themes-vc-annotate)
Revised supported faces or face groups
Enhanced the default background colour of the current date in the Org agenda. This is a subtle change, all things considered, which makes it easier to discern where the highlight is while it remains close to the spirit of the previous design. The idea is to not add too much saturation here, because the buffer is already “busy” with lots of highlights. Thanks to Daniel Mendler for the feedback on the mailing list: https://lists.sr.ht/~protesilaos/modus-themes/%3C3d8b1096firstname.lastname@example.org%3E.
M-x womanfaces to have a bit more saturation. A while ago I desaturated the
woman-boldfaces on the premise that the added bold weight would be sufficient. However, the bold weight may sometimes not draw the desired attention, such as at small point sizes or with certain font configurations. As such, the added intensity in colour is necessary.
Changed the Selectrum quick key faces (
selectrum-quick-keys-highlight) to have the same style as Avy, Vertico’s own “quick keys”, and related. For a technical analysis, read “Modus themes: case study on Avy faces and colour combinations”: https://protesilaos.com/codelog/2022-04-20-modus-themes-case-study-avy/.
Made internal adjustments so that
M-x list-packagesinherits from the standard
errorfaces instead of adding its own face attributes. In practice, the user will notice a change for new packages in the listing if
Introduced the same inheritance rules as above for the
syslogpackage (mutatis mutandis).
Increased the saturation of the
package-status-availableface, which is shown in the
M-x list-packagesbuffer. The overall effect is subtle, though sufficiently noticeable.
Revised the faces of the
deftpackage to make it look consistent with the rest of the theme’s relevant interfaces (to the extent possible as Deft uses a non-standard presentation).
speedbar-highlight-facewith the user option
highlight-thingface (see package of the same name). This makes it stand out more and it also aligns it with the standard
matchface, which is pertinent here.
Amplified the saturation of the
dired-git-infoface. Makes it easier to differentiate the Git commit text from the Dired listing, without drawing too much attention to itself.
Adjusted the hue of the
easy-jekyll-help-facefrom teal to blue. This makes it look more like the standard
easy-jekylldoes not align with upstream Emacs in this regard.
Intensified the background of
rectangle-previewto work even in cases where a grey background is already on display. This face is used for the
string-rectanglecommand (e.g. C-x SPC to draw a rectangle and C-t to insert text in its stead—works as a simple “multiple cursors” on a straight line).
Support for new faces or face groups
- chart (built-in)
- edmacro-label (Emacs 29)
A comment on
info+. As is the case with PACKAGE+ packages from the
Emacs Wiki, info+ defines lots of faces that hardcode colour values
instead of inheriting from basic faces. It does so for no good reason
and the results will likely not look decent in any theme. Furthermore,
these faces colourise too much even when the colour values can be
appropriately combined (ceteris paribus), making the buffer harder to
The support I add for info+ is consistent with the design principles of the modus-themes, one of which is to avoid exaggerations as those indirectly affect legibility. As such, some of the changes I introduce here outright remove colouration, while others align the various constructs with the overall aesthetic of the themes.
Note that, by default, info+ adds clickable buttons to glossary terms.
This produces awkward combinations such as by buttonising the “string”
component inside of what actually is a function’s argument. So you
have, say, FORMAT-[STRING] where “” represents the button: the FORMAT
gets one face and the [STRING] another, even though they are part of a
single argument. To me this looks broken and is counter-productive,
though it is not up to the theme to decide how packages fontify the
various constructs. At any rate, button styles at the theme level are
controlled by the user option
Thanks to Jonas Collberg for the feedback in issue 33 over at the GitHub mirror: https://github.com/protesilaos/modus-themes/issues/33.
Named the mailing list address as the =Maintainer:= of Denote. Together with the other package headers, they help the user find our primary sources and/or communication channels. This change conforms with work being done upstream in package.el by Philip Kaludercic. I was informed about it here: https://lists.sr.ht/~protesilaos/general-issues/%3C875ykl84yi.fsf%40posteo.net%3E.
Addressed byte compilation warnings in doc strings pertaining to the use of literal quotes. Thanks to Matt Armstrong and Rudolf Adamkovič for the feedback on the mailing list: https://lists.sr.ht/~protesilaos/modus-themes/%3C87bktlvgyy.email@example.com%3E.
:linkvalue in the declaration of the user options
modus-themes-vivendi-color-overrides. It once again directs to the correct heading in the manual.
Documented all the aforementioned, where necessary.
linpackages in the relevant sections of the manual. The former helps set fonts and switch between font presents. The latter is a stylistic variant of hl-line (its documentation explains its raison d’être).