Geo Voronoi polygons out of triangles

[misha-panyushkin]

Hi,
I'm trying to build Voronoi diagram for "infinite horizon" map. The points - are the cities with defined coordinates [lat, lon].

Input: geoJSON

{
  "features": [
    {
      "geometry": {
        "coordinates": [
          2.1699187,
          41.387917
        ],
        "type": "Point"
      },
      "properties": {},
      "type": "Feature"
    },
    {
      "geometry": {
        "coordinates": [
          7.205232,
          43.66049
        ],
        "type": "Point"
      },
      "properties": {},
      "type": "Feature"
    },
    {
      "geometry": {
        "coordinates": [
          12.4942486,
          41.8905198
        ],
        "type": "Point"
      },
      "properties": {},
      "type": "Feature"
    }
  ],
  "type": "FeatureCollection"
}

Links d3.geoVoronoi(geoJSON).links()

Triangles d3.geoVoronoi(geoJSON).triangles()

Polygons d3.geoVoronoi(geoJSON).polygons()

Can't understand what I'm doing wrong. I will appreciate any help on these.

[Fil]

You're not doing anything wrong! The reason for the bad display is that D3 doesn't have the same definition of geojson as the others. (See the introduction in https://github.com/d3/d3-geo).

In d3-geo, all the lines follow the geodesic lines (great circles), and they don't need to be cut at the antimeridian (longitude=180°).

To convert your results to the standard geojson you can use the d3.geoProject utility from https://github.com/d3/d3-geo-projection and do this:

projected = d3.geoProject(
  d3.geoProject(
    polygons,
    d3
      .geoEquirectangular()
      .scale(180 / Math.PI)
      .translate([0, 0])
  ),
  d3.geoIdentity().reflectY(true)
)

After this operation you should be able to see the features correctly.
https://gist.github.com/Fil/f67b0f6ccfa1f072c555775e331c307b

As you can see if you inspect the resulting structure, one of the polygons has been transformed to a multipolygon, which doesn't make sense from a topological point of view… and that is basically the explanation for this (very inconvenient) difference.

We should probably offer a simpler API for this "solution".

[Fil]

Another reference here: d3/d3-geo#104

[misha-panyushkin]

@Fil many thanks! I've spent a lot of time finding solution. And now I understand the correct approach!

[misha-panyushkin]