Skip to content

Conversation

roosta
Copy link
Member

@roosta roosta commented Apr 10, 2025

Srcery Next

This PR aims to enhance compatibility with modern (neo)vim features, normalize
color usages, and includes palette changes.

Changes

  • Use base color groups and links instead of defining each color individually.
    A function should look like a function in all languages where that makes sense.

  • Removed "memoize" groups (SrceryRed, SrceryBlack etc) that made debugging
    difficult (when :Inspect a highlight group, it would link to something like
    SrceryRed rather than a meaningful base group) (d2dd4cf)

  • Aligned Treesitter highlighting with traditional regex matching for consistent
    appearance

  • Removed a excessive highlight groups for languages and plugins. It should be
    assumed that mostly sane defaults are enough.

  • Normal float is now behind an option g:srcery_normal_float, setting it
    to 1 will use a custom background for popup windows. (3b0f594)

  • Simplify background option, this was getting needlessly complicated, I will
    move the transparent background functionality to a separate option. (fefcf26)

  • Remove dim_lisp_paren(e61197), hard_black_terminal_bg (20bfd94), and
    italic_types, the type will still be italic, just no opt.

  • Join all inverse options into a single option, which controls all inverse usages (df8fa8d)

  • Additionally many group and link changes here and there, to many to keep
    track of, but I've tried to remain consistent...

Fixed

  • Fix an issue with nvimpager colors not using srcery, see 911a858 for details
  • Fix an issue with sneak, where a new
    group (SneakCurrent) needed to be linked to CurSearch, else it defaulted
    to hardcoded hex

Chore

  • Reorganize the code into more meaningful sections with less nesting/folding.
    I've also cleaned out a lot comments as well, a quick SHIFT+K will
    sooner explain the group better than most of those comments.

Color changes

Relying on xterm colors are not as useful anymore as most modern terminals
support true color, same as (neo)vim, so I took a crack at designing some new
shades, that includes [bright]white, [bright]black, and all the
xgray[1-6], now just named gray[1-6]. The old background is accessible via
gray1

The new shades are based on a darkened version of the original background
color, chosen because I felt the old one looked a bit washed out TBH.

The actual colors, red, green etc remain unchanged for now.

Sample Screenshots

html javascript vimscript

TODOs

  • Screenshots
  • Documentation
  • Possibly removal of some old config options. Open for discussion here.

EDIT 2025-08-27: Rewrote and reorganized PR description

@roosta roosta changed the title Feat/modernize Srcery Next Aug 27, 2025
Use the same defined heading colors for markdown and HTML, and what ever
else might need headings down the line
avante links to NormalFloat like so many others, in this cause it didn't
really work
It will link to `diffRemoved`, which only has `fg` defined, which in turn
would result in headings below a certain level would be red. I don't
know the rationale linking it to `diffRemoved`, maybe its common to add a
background to those groups, i do not know. In lue of colored background
shades I set it to `xgray1`, to simulate the effect I see in
`render-markdown` `readme` screenshots.
New group introduced that wasn't in sneak previously `SneakCurrent`, I
linked it to Search. I don't see any need to differentiate current and
other, cursor is already quite visible when using sneak, and if enabling
`set cursorline`, it's even more so.
Found 2 to be a bit too bright
Matches macro, which I think is appropriate
Keeping the underline, but differentate from normal text
For consistency, it was blue, not sure about what tree of links got it
there, but it was to much, while bright_white is too little.
String is bright green, feels more consistent
Had issues with colors being wrong in nvimpager for certain output, took
some serious digging to find any reference to these groups, and the
names are a bit odd, I checked the source and they are generated with
those names, but I'm worried it'll break, or additional groups are
needed.
Commented out groups I suspect aren't needed, but mainly moved stuff
around so it made a bit more sense, named sections and not so much
deeply nested folds. 1 level is enough IMO.

This is a larger effort to normalize the theme any rely on base language
groups for most syntax highlighting, making treesitter and regex similar
if not exactly the same.
The goal of this commit is to normalize all language hi groups. Instead
of specifying a color for each hi group, rely on defaults, or link to
base language groups. This makes for easier maintenance, and more
consistent colors. Keyword is keyword in any language that links to
keyword.

I've left links to basegroups if they were present before this
refactor, else I've tried to include sane links, but there will be large
differences from previous versions
This is a bad idea, when inspecting a hl group in nvim, it'll say that
it links to that group, which tells me nothing. For example, I'm
investigating why a certain word in a certain language looks the way it
looks, then inspecting the word, it'll say its linked to SrceryRed,
which is meaningless, I don't actually know which of the base groups are
being referenced, it would be any of those having SrceryRed as a base
color.

I fixed the lightline theme as well because it referenced the same
groups. We maintain global variables for color hexes, and the termcolors
are always the same so I just referenced those instead.
I kept some of the original bindings for reference where possible.
Setting type to italic by default, don't see a need to optionally set
this, we already have an italic toggle, and that should cover this as
well.
I think these mostly have been added so that we can reach the special
attr (bold, italic etc), but none is the correct way of getting to those
arguments

There are UI elements that have background set, not gonna touch those
for now.
Try to be consistent with ordering, function calls on top, links on
bottom. Sorted.
Shouldn't need this, it should have sensible defaults and use basegroups
The inverse options were getting confusing, use instead one var to
control whether inverse is used or not. Ad

BREAKING CHANGE: Removes options srcery_inverse_matches,
srcery_inverse_match_paren.
This only works in regular vim, not nvim, so for the sake of
consistency lets keep it the same across envs
The functionality has been removed in a prev commit, this officially
removes the non-functioning opt
I plan on having a separate transparent opt, instead of mixing the two.
Also rename it to srcery_background rather than bg for clearity
This needs to exist, there are plugins that work well with srcery color
for NormalFloat and then there are those that don't, it has to be down
to preference.
Change the background to a darker variant of the original, I then
created shades by mixing in white to create alternative xgray[n]
variants.

The new background was chosen so that the colors pop a bit more, trying
out srcery on various monitors of often looked washed out. I've also
gotten my hands on a much better monitor, so I'm more able to accurately
cover multiple various outputs.
These plugins haven't worked in this branch for a while, the old memoize
groups are no more. Fix some atrocious formatting as well in the airline
theme, think this one's been through a few formatters.
roosta added 3 commits October 5, 2025 14:44
Change search indicators to magenta, which works well on inverse and
not. We also don't need a cond anymore because setting both background
and special will work as expected when turning off inverse, it will then
fall back to the background.
The magenta was too distracting, I've reverted it back to s:none, but
kept the underline. Set cur/inc search to have a gray4 background.
conflicted with visual when inverse is disabled
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant