Palettes

Use this page when separate series or categories need distinct colors.

A palette is a finite list of colors. The palette types on this page use these terms:

• Qualitative: separate colors for labels that have no numeric order.

• Sequential: low to high along one ordered path.

• Diverging: two sides meet at a meaningful center such as zero.

Discrete color forms for series: the Octave default cycle, curated qualitative sets, and designed sequential or diverging samples from dm.colors(name, n=...).

For a static token sheet of every single-color string, use Colors.

Pick a palette

Use the left rail to choose a palette, drag the color-count control, and toggle black-and-white preview. Click any swatch to copy its hex, or copy the matching Python call from the explorer. The rail has 13 qualitative choices: Octave, Octave Print, and 11 curated qualitative sets.

Apply it

import dartwork_mpl as dm

dm.set_colors()                              # octave — the searched default cycle
dm.set_colors("trustworthy")                 # any curated set, by name
dm.set_colors("green", n=5)                  # 5 steps of one hue family
dm.set_colors(["dc.hl", "dc.gray3", "dc.gray5"], ax=ax)  # one Axes only
cols = dm.colors("blue", n=4)                # designed color list
dm.set_colors(ax=ax, styles=True)            # >8 series: 8 colors x 3 styles

Every name resolves under dc.*: "blue" can return designed samples from the single-hue family, "blue_red" can return a diverging list, and "trustworthy" returns the curated qualitative set. The qualitative families are also registered as colormaps (dc.octave, dc.octave_print, dc.trustworthy, …) for scatter(c=...) and seaborn palette=.

Which palette for which data?

Color-vision deficiency (CVD) simulations are named models used to diagnose potential color collisions, not observer guarantees.

Your data	Reach for	Explorer group
Everyday 4-8 categories	dm.colors("octave", n=8) or trustworthy	Qualitative
Many unrelated categories, max distinctness	vivid or neon	Qualitative
A few related series, one mood	a hue family sampled evenly, or forest	Sequential / Qualitative
Ordered amount (rank)	one family ramp; gray if hue means nothing	Sequential / Neutral
Ordered around a midpoint (+/- / change / correlation)	blue_red, blue_orange, teal_amber, or green_purple	Diverging API
Soft, editorial, dense dashboards	pastel or dusty	Muted
A specific mood (warm / earthy / luxury)	ember, earth, or jewel	Tone
Highlight one series, mute the rest	teal_accent, coral_accent, or dc.hl + grays	Emphasis
Qualitative categories where distinction under the named CVD simulations matters	dm.colors("octave", n=8) or trustworthy	Qualitative

How the system is organized

Everything lives in one dc.* namespace and uses one API: dm.colors(...) for colormaps or designed color lists, dm.set_colors(...) to apply colors globally or to one Axes, dm.list_colors(...) for family metadata, and dm.show_colors(...) for previews.

There are four discrete forms: Octave, the searched default cycle for everyday charts; 11 hand-tuned curated qualitative sets for muted, tonal, forest, and emphasis use cases; 20 generative single-hue families sampled for ordered and sequential work; and canonical diverging lists such as blue_red, blue_orange, teal_amber, and green_purple for centered data.

Technical terms

OKLab and OKLCH are used to construct and adjust colors. ΔEOK is a color-distance ruler: larger means more different.

Modeled relative CIE Y (relative_y) is calculated from nominal D65 sRGB; it is not a measurement of a particular display, perceived brightness, or OKLab L. It records nominal output ordering under the catalog’s pinned calculation.

CIELAB supplies coordinates for CIEDE2000 distance. CIEDE2000 and the named CVD simulations are model-specific collision/regression diagnostics on finished output; they do not construct colors or define modeled output. CIEDE2000 distance is reported as ΔE00. Web Content Accessibility Guidelines (WCAG) check pair-specific text contrast for a named foreground/background pair; this is not palette certification.

The 19 chromatic family ladders use ΔEOK arc-length equalization. The gray ladder directly samples ten evenly spaced neutral-tone positions; its resulting neighbor ΔEOK near-evenness is measured and protected by frozen non-regression gates, not produced by the chromatic equalizer. Both forms use OKLab/OKLCH construction and preserve the normalized modeled-relative-Y output contract. The curated sets are manual, preserved rows rather than extra recipe inputs. CIEDE2000 and CVD results remain model-specific diagnostics, and WCAG text contrast is checked separately for a specified foreground/background pair.

The count rule is simple: families have 10 steps; curated qualitative and diverging sets have 8 colors; Octave has 8 chromatic colors, with rose in the eighth slot; and Octave Print has 7 chromatic colors plus dark gray. Single-hue curated ramps are not duplicated; the families serve that job.

Reference

Octave — the default cycle

Octave is the default for unrelated series; Octave Print trades one chromatic slot for dark gray to increase nominal neutral-coordinate separation in the historical diagnostic.

Technical detail

Octave is the default coherent data-series cycle when you do not want to choose a palette by hand; use dm.set_colors() or the stable dc.octave colormap token. Its eight published chromatic colors originated in an exhaustive model-specific collision search; the v6 compiler replays the frozen selection rather than using CVD as a construction objective. Under the named full-severity simulation diagnostics, dc.octave records min ΔE00 10.3 (vs the Okabe-Ito benchmark’s 11.5), and in the named tritan diagnostic the default cycle’s 8.3 actually beats Okabe-Ito’s 7.9. Protan/deutan use Machado et al. (2009), while tritan uses Brettel–Viénot–Mollon (BVM, 1997); see Design rationale. Matplotlib’s tab10 scores 1.4 under the same named simulated-protanopia validation protocol.

Gray is reserved for grids and reference lines, not spent as a data color in Octave; its eighth series color is chromatic rose. Octave Print uses a dark gray as its eighth color to change its nominal neutral-coordinate diagnostic. The trade-off is a catalog design choice, not a print guarantee. The historical, validation-only CIELAB report records Octave’s L* 43-78 range for thin-line candidates on white; in the same diagnostic, Octave Print records a minimum pairwise ΔL* of 7.7. That bounded source-color statistic does not model a particular printer, paper, conversion workflow, background, overlap, or observer. The live nominal output ordering is checked with modeled relative Y. Octave Print keeps the same hue per slot as Octave, and the violet slot matches Octave. Need more than eight line series? Opt in to dm.set_colors(styles=True), which expands the cycle to 8 × 3 = 24 color/style combinations. Line styles are opt-in because a plot with lw=0 would otherwise inherit dashes and break.

Hue families

The chromatic families follow the hue spectrum: red · rose · coral · tangerine · orange · amber · yellow · lime · green · teal · cyan · sky · blue · cobalt · indigo · violet · purple · fuchsia · pink, plus gray. Use a whole ramp for an ordered scale, or sample evenly spaced steps for related series.

cols = dm.colors("blue", n=4)
cols = dm.colors("teal", n=6, reverse=True)

Curated sets

The curated qualitative rail is the hand-tuned set collection preserved from the v5 migration. Curated sets use the same API as families, and their grayscale and CVD metrics are validation metadata, not generation inputs.

Group	Members
Qualitative	trustworthy, vivid, neon, forest
Muted	pastel, dusty
Tone	ember, earth, jewel
Emphasis	teal_accent, coral_accent

The four canonical discrete diverging forms are blue_red, blue_orange, teal_amber, and green_purple. They are ordered encodings for centered data, not unordered categorical sets, so the explorer keeps them out of the qualitative rail.

colors and set_colors options

Option	What it does
colors(name, n=None)	Return the registered colormap when n is omitted, or a designed list when n is set.
n	Choose how many colors from the designed discrete form. Continuous families require it when used with set_colors.
reverse	Flip the resulting list or colormap.
set_colors(name_or_list=None, ax=None, n=None, styles=False)	Apply a palette globally, to one Axes with ax=, or expand it with line styles.

Names resolve under dc.*, so "blue" and "trustworthy" are enough.

The v5 migration trimmed the throwaway ad-hoc aliases; the curated dc.* categorical sets are deliberately preserved. See the migration guide for the manual rename table.