“Zero-dimensional” group and more flexible normalization…

[mbostock]

I’m trying to make a stacked bar chart of Penguins by sex using the group transform. I can do this using Plot.groupX, but it requires a dummy x definition:

Plot.plot({
  y: {
    grid: true
  },
  marks: [
    Plot.barY(penguins, Plot.stackY(Plot.groupX({x: () => 1, fill: "sex", normalize: true}))),
    Plot.ruleY([0])
  ]
})

I can overwrite the x definition to get what I want, so it’s possible but awkward:

Plot.plot({
  y: {
    grid: true
  },
  marks: [
    Plot.barY(penguins, Plot.stackY({...Plot.groupX({x: () => 1, fill: "sex", normalize: true}), x: null})),
    Plot.ruleY([0])
  ]
})

What I want is some way for groupX’s x to be optional (maybe I want it to default to undefined, rather than defaulting to identity, but I suppose having it default to identity would be okay if I can say x: null):

Plot.plot({
  y: {
    grid: true
  },
  marks: [
    Plot.barY(penguins, Plot.stackY(Plot.groupX({fill: "sex", normalize: true}))),
    Plot.ruleY([0])
  ]
})

I’m not sure “groupX” is still the right name for this transform, but maybe it‘s fine, since it will output frequency to the y channel which is what I want.

The next thing I want is for this to work with faceting:

Plot.plot({
  y: {
    grid: true
  },
  facet: {
    data: penguins,
    x: "species"
  },
  marks: [
    Plot.barY(penguins, Plot.stackY({
      ...Plot.groupX({
        x: () => 1, // TODO
        fill: "sex",
        order: "z",
        normalize: true
      }),
      x: null
    })),
    Plot.ruleY([0])
  ]
})

And here the problem is that neither normalize: true nor normalize: "z" do what I want, because I want to normalize to the facet. Maybe normalize: "facet" should be an option? That’s certainly easy to implement.

[mbostock]

This works but it feels like cheating. 😄 (Note that the totals aren’t expected to go to 100% here because some of the data have undefined sex.)

Plot.plot({
  y: {
    grid: true
  },
  facet: {
    data: penguins,
    x: "species"
  },
  marks: [
    Plot.barY(penguins, Plot.stackY({
      ...Plot.groupX({x: "sex", normalize: "z"}),
      fill: ([d]) => d.sex,
      x: null
    })),
    Plot.ruleY([0])
  ]
})

[Fil]

If what you're saying is that normalize: "z" is difficult to discover, maybe we should rename it to normalize: "series". Other than that I think the code looks good (group the values on sex for each facet, then stack into a single pile).

[mbostock]

No, I think normalize: "z" is fine. I’m saying we also need normalize: "facet" and for Plot.groupX to make x optional (and for Plot.groupY to make y optional, and probably for Plot.group to make x and y both optional).