I'm afraid I was a little premature in my celebration of a new vector union function. I tried it in my notebook which takes solid sections of four-dimensional polytopes. When the polytopes are convex I use a ConvexHull function to make the sections in the 3-space. This function is sensitive to duplicate points, so I always run the points of the section through vunion before passing them to ConvexHull3D.
In my test, most point-sets worked fine, but in one case, out of 20, the new function returned 81 points, the old, 80. And there were only 80 distinct points.
I was curious to see how the new function would work on the 4-vectors themselves, before projecting into the 3-space. Here too my vunion function returned a list of 80 distinct 4-vectors. But the new function returned a list of 92. So it failed badly with the 4-vectors.
I suspect the problem is the same as before: the ordering of the vectors. In the spurious list of 81 3-vectors, the two duplicates were separated by half-dozen other vectors, and neither contained a 0.
It seems that we must discover an iron-clad way to order a Chopped list of n-vectors.
Russell Towle Box 141 Dutch Flat, CA 95714 (530) 389-2872