Hi Jure,
We've looked into these Delaunay problems a bit. Both DelaunayFast
and DelaunayWatson have problems with co-linear and co-located points.
Your data has several co-located points and lots of co-linear. We
do want to make the Delaunay code as robust as possible, but these are
difficult cases. Meanwhile, we've modified the Delaunay.factory
method to use DelaunayClarkson if it gets an Exception from either
DelaunayFast or DelaunayWatson. DelaunayClarkson is probably not
perfect, but its probably more robust than the other two.
Another thing to consider: when many points are co-located or
co-linear, then the application may have some apriori topology
that it can pass to DelaunayCustom, and avoid all these problems.
We are sincerely sorry for the problems you are experiencing and
hope we can find some way around them.
Cheers,
Bill
----------------------------------------------------------
Bill Hibbard, SSEC, 1225 W. Dayton St., Madison, WI 53706
whibbard@xxxxxxxxxxxxx 608-263-4427 fax: 608-263-6738
http://www.ssec.wisc.edu/~billh/vis.html
