Closed mbostock closed 8 years ago
It’s possible we want separate topologies for the cells and the triangle mesh. Having the triangle mesh topology would be useful for computing alpha shapes, for example.
We’re almost there now that #1 is fixed: each cell exposes the indexes corresponding to each shared edge, so it’s easy to compute the cell topology.
One challenge remaining, though, is that the computed Voronoi diagram isn’t clipped: clipping is done when the cell polygons are computed. And the cell polygons can’t represent the topology.
So, it might be better to undo part of #1 and just apply clipping when computing the Voronoi diagram. That makes it slightly less flexible, but only theoretically… I think it’d still be perfectly fine in practice.
In addition to #1, we should think about voronoi.topology: mbostock/d3#1819.