Skip to content
Merged
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
46 changes: 34 additions & 12 deletions src/styles/tokens.css
Original file line number Diff line number Diff line change
Expand Up @@ -108,14 +108,33 @@
the last place that should look like someone else's editor.

They fan out from the rust ramp rather than sitting on top of it:
coral at 16, amber at 35, olive at 75. Close enough to read as one
warm family, far enough apart to stay tellable at 13px.
coral at 16, amber at 35, olive at 75. That makes one warm family, and
a warm family is exactly what cannot use hue to tell its members apart.
All three sit inside a 60° arc, which is the configuration red-green
colour-vision deficiency flattens: simulated, they land on the same
tan. Roughly 8% of men read them that way.

So lightness carries the distinction, and the three sit at deliberately
different steps rather than all at 300 — coral 79%, amber 62%, olive
81%. They have to interleave with the neutral code roles rather than
stack against them (`--code-comment` L* 66.5, `--code-plain` L* 85,
`--code-fn` L* 98.3): six roles share one band, and the pair that
collapses first is plain against literal, not keyword against type.

Measured against `--card` (#18181b), the ground code sits on:
coral 8.7:1 · amber 9.8:1 · olive 10.1:1 — all past AAA.
coral 10.3:1 · amber 8.6:1 · olive 13.5:1 — all past AAA.

Under a Machado et al. (2009) deuteranopia/protanopia simulation at
severity 1.0, the tightest of the fifteen code-role pairs is plain
against keyword at ΔL* 3.6 — against 2.4 before this spread, and 0.5
under the cool palette that preceded it. Treat 3.6 as a floor rather
than headroom: six roles in the L* 66.5–98.3 band is most of what
colour alone can carry here, so a further distinction would need a
non-colour cue (weight or italic), not a fourth hue.
--------------------------------------------------------------------- */
--ak-amber-300: 35 73% 68%; /* #e9b872 types */
--ak-olive-300: 75 48% 69%; /* #c3d68a strings and numbers */
--ak-coral-200: 18 77% 79%; /* #f3b9a0 keywords */
--ak-amber-400: 35 73% 62%; /* #e5aa57 types */
--ak-olive-200: 75 48% 81%; /* #dae6b7 strings and numbers */
--ak-blue-700: 224 76% 48%; /* #1d4ed8 info fill */

/* ---------------------------------------------------------------------
Expand Down Expand Up @@ -172,10 +191,13 @@
what changes as the flow animation moves from one method to the next.
Plain code sits a step below it, and a comment a step below that.

Keywords take `--ak-rust-300`, a step lighter than the accent itself.
That is deliberate on both counts: the family is the point, and the
exact accent shade stays reserved for links and calls to action, so a
`let` never reads as something you can click.
Keywords take `--ak-coral-200`, two steps lighter than the accent rather
than one. That is deliberate on both counts: the family is the point,
and the exact accent shade stays reserved for links and calls to
action, so a `let` never reads as something you can click. One step did
not deliver it — `--ak-rust-300` sat 1.27:1 from `--ak-rust-400` and 1°
of hue away, which is not a perceptible step at 13px next to a `.link`
rendered in the accent. Two steps give 1.49:1.

This is safe here for a specific reason, and the reason is worth
stating because it does not always hold. Inside a code block the
Expand All @@ -190,9 +212,9 @@
--code-fn: var(--ak-neutral-50);
--code-plain: var(--ak-neutral-300);
--code-comment: var(--ak-neutral-400);
--code-keyword: var(--ak-rust-300);
--code-type: var(--ak-amber-300);
--code-literal: var(--ak-olive-300);
--code-keyword: var(--ak-coral-200);
--code-type: var(--ak-amber-400);
--code-literal: var(--ak-olive-200);

/* ---------------------------------------------------------------------
ROLES — verdicts
Expand Down
Loading