SagUI

Ridgeline

Overlapping distributions, one ridge per group: hover to lift a ridge and read its quartiles, switch datasets and every curve morphs.

import { Ridgeline } from "@sagui/ui";
Markdown
Loading demo

When to use

  • Seasonal or categorical distributions such as temperatures by month or response times by region.
  • Showing where two groups differ in shape, such as bimodal winters.

When not to use

  • Use beeswarm when each individual observation matters.
  • Use bar-chart when only one summary number per group matters.

Installation

Install the package and import the styles once. The installation guide covers the Tailwind setup.

$ npm install @sagui/ui

Usage

example.tsx
import { Ridgeline } from "@sagui/ui";

export function Latency({ regions }: { regions: { id: string; label: string; values: number[] }[] }) {
  return <Ridgeline series={regions} label="API latency by region" unit=" ms" />;
}

Examples

Without tint

tint={false} draws every ridge in one color; overlap and rowHeight set the spacing.

Loading demo

API reference

Ridgeline

A ridgeline (joy) plot: one smoothed distribution per row, overlapping like mountain ridges, tinted by median. Hover lifts a ridge above its neighbours to read its quartiles; switching datasets morphs every curve.

Prop Type Default Description
series (required) { id: string; label: string; values: number[] }[] – One ridge per series, drawn top to bottom. Values are raw observations; the ridge is their kernel density.
label (required) string – What is measured. Names the chart for assistive technology.
unit string – Unit after each value, such as "°" or " ms".
formatValue (value: number) => string – Formats values. Overrides unit.
domain [number, number] – Value range of the axis. Fix it to keep the axis still across datasets. Defaults to the data with room on each side.
overlap number 2.4 How far the tallest ridge rises into the rows above, in row heights.
rowHeight number 30 Height of one row in pixels.
bandwidth number – Smoothing in value units. Defaults to Silverman's rule per series.
tint boolean true Shades each ridge by its median, from a light tint to the full first series color, with a stepped scale under the axis.
active string | null – Controlled id of the lifted ridge.
defaultActive string | null null Initial lifted ridge when uncontrolled.
onActiveChange (id: string | null) => void – Called when the lifted ridge changes.
emptyLabel string "No data yet" Shown when there are no values.
className string – Extra class on the root figure.

Keyboard interactions

Keys Action
ArrowUporArrowDown Lifts the previous or next ridge.
ArrowLeftorArrowRight Moves the reading cursor along the values by one axis step.
HomeorEnd Lifts the first or last ridge.
Escape Lowers the ridge and clears the reading.

Accessibility

  • The plot is a focusable group; the lifted ridge's median, middle half, range, and count are announced through a polite live region.
  • A visually hidden table lists median, quartiles, extremes, and count for every series.
  • Every row is labelled in text; the median tint is backed by the table and tooltip.

Motion

  • Curves are sampled on a shared grid, so a new dataset or domain interpolates every point: ridges swell, slide, and split in one spring.
  • The first reveal rolls the ridges in from the top row down.
  • The lifted ridge is a copy drawn above the others that rises 6px on a spring, with its middle half shaded and its median marked.
  • The tooltip glides after the pointer and stays inside the chart. Reduced motion draws final shapes immediately.

Responsive behavior

  • Label gutter and tick count adapt to width; the plot redraws at the measured size so strokes stay crisp.
  • Touch taps lift a ridge and read at the tapped value.

Performance

  • Densities are computed once per dataset (96 samples per ridge); animation interpolates arrays and writes path strings directly.
  • Suitable for up to a few thousand observations per series.

Notes

  • Choose it to compare the shape of many distributions: spread, skew, and bimodality, not just averages.
  • Pass raw observations, at least 20 per series, and fix domain when switching datasets so shapes move on a still axis.
  • Keep rows to about 20; beyond that, use small multiples or a box plot.
  • Colors come from the chart palette, --color-chart-1 to --color-chart-4, which has light and dark values. Override --sg-chart-1 to --sg-chart-4 on any ancestor to rebrand a chart. Pair color with labels, since the palette is not tuned for every type of color vision.

Also in charts