The box mark could use bin/rect instead of group/bar for quantitative data

[mbostock]

As an approximation:

Plot.plot({
  marks: [
    Plot.ruleX(data, Plot.binX({y1: loqr1, y2: hiqr2}, {x, y})),
    Plot.rectY(data, Plot.binX({y1: "p25", y2: "p75"}, {x, y, fill: "#ccc"})),
    Plot.ruleY(data, Plot.binX({y: "p50"}, {x, y, strokeWidth: 2})),
    Plot.dot(data, Plot.map({y: oqr}, {x, y, z: ???}))
  ]
})

I couldn’t figure out how to compute the per-bin outliers, though. Maybe we need a Plot.binmapX transform where z can be imputed by binning x? Or maybe there’s a way to compose the bin and map transform? 🤔

Ref. https://talk.observablehq.com/t/how-to-label-a-boxplot-with-dates-format-axis/7739/3

[Fil]

I feel this could be addressed at a higher abstraction level (#1336)?

[mbostock]

I think we could use exactly the same technique we use in Plot.auto, then have the box mark chose dynamically between bin+rect and group+bar.

[Fil]

This is the test example in http://localhost:8008/?test=aaplMonthly (minus the outliers)

[mbostock]

Here’s an example workaround:

Plot.plot({
  y: {
    grid: true,
    label: "↑ Price"
  },
  x: {
    interval: 0.5,
    label: "Carats →",
    labelAnchor: "right",
    tickFormat: (x) => x.toFixed(1)
  },
  marks: [
    Plot.ruleY([0]),
    Plot.boxY(diamonds, {x: (d) => Math.floor(d.carat * 2) / 2, y: "price"})
  ]
})