Stable order for segment clipping?

[mbostock]

Attempting to fix #37. When clipping the Voronoi cell segments, this orders the segment so that the starting point is inside the clip region, if possible. However, it looks like there’s still a problem with the bottom-left cell for reasons that are unclear to me. 😦

Any chance you could take a look, @mourner?

https://beta.observablehq.com/@mbostock/d3-delaunay-bug

[mbostock]

The other problem is that containsFinite / containsInfinite are broken when the circumcenters are near-infinite. Due to limited numerical precision, the cell polygon is no longer guaranteed to be convex. We can workaround that if we’re careful during clipping, but voronoi.contains can’t make the same assumption that the given point x, y is inside the viewport.

But it occurs to me that we could use a strategy similar to voronoi.find to do the containment check, since the definition of the Voronoi cell is that it is the area of the plane closest to its site. Then we wouldn’t even need to compute the Voronoi cell polygon to check containment.

[mbostock]

Yay, the new strategy works well!