Closed mourner closed 6 years ago
I also pushed improve-robustness-scale branch which scales coords to [1.0, 2.0)
, but not yet sure whether to do that.
Merged a simpler approach for now because it's an obvious improvement, will leave these branches for now and reconsider when we encounter more "bad" cases.
cc @fogleman. I think the elimination of
Something is wrong with the input points.
error in the second commit is important — if we looped through the hull and didn't find any visible edges, this means that the point is likely a near-duplicate we somehow missed in earlier filtering.