Philippe Rivière

I think what is happening here is that the triangular faces (in blue) do not correspond to the voronoi cells of their centroids (in red); for instance the top vertex...

Exactly: `find0` is a (relatively slow) search for the closest center, which only serves geometries that are a voronoi tesselation of the sphere (hence the name of the voronoi projection...

Note that `d3.polygonContains` is for planar polygons, here we need the equivalent for spherical coordinates (~~`d3.geoPolygonContains`~~ **EDIT**: `d3.geoContains`).

oops, sorry for the typo! The projection is beautiful!

The packing algorithm described in this essay: https://observablehq.com/@d3/d3-packenclose, and is implemented [here](https://github.com/d3/d3-hierarchy/blob/main/src/pack/siblings.js) for the part that packs all "siblings" (circles that must be packed together at the same level). Barring...

Looking good! Note that these should be named large values rather than big numbers (because [of precedent](https://www.npmjs.com/package/bignumber.js)).

linked to (or identical to?) #1431.

An example where it would be useful: https://observablehq.observablehq.cloud/framework-example-loader-duckdb/

The issue with dates can be reduced to this (which is independent of Plot): ```js import * as Arrow from "apache-arrow"; const data = [{date: new Date(1950, 1, 2)}] console.log(data,...

The internal data is like this: ``` Data { type: DateMillisecond [Date] { typeId: 8, unit: 1 }, children: [], dictionary: undefined, offset: 0, length: 1, _nullCount: 0, stride: 2,...