Sort the secondary dimension of bars by maximum value of the first dimension?

[mbostock]

And for facets, the maximum across all facets?

[mbostock]

This can be done by manually specifying the scale’s domain, but it’s a bit of a pain.

[mbostock]

Some examples:

domain: d3.sort(alphabet, (a, b) => d3.descending(a.frequency, b.frequency)).map(d => d.letter)

domain: d3.sort(data, (a, b) => d3.descending(a[2019] - a[2010], b[2019] - b[2010])).map(d => d.State)

domain: d3.rollups(data, group => d3.max(group, d => d.unemployment), d => d.division)
  .sort(([, a], [, b]) => d3.descending(a, b))
  .map(([key]) => key)

domain: d3.rollups(movies, movies => d3.median(movies, Profit), Genre)
  .sort(([, a], [, b]) => d3.descending(a, b))
  .map(([key]) => key)

Feels like there are at least two cases to consider — with and without aggregation — though perhaps we could generalize.

[mbostock]

We did a lot of exploration on this and decided to add d3.groupSort. Doing this as part of the mark specification felt too implicit and brittle.