CGAL users discussion list

["Nico Kruithof" <>]

Hi Balint,

Weird, this really seems like a precision problem. Does it also occur when you try to write the numbers with high precision? Otherwise you should always be able to write the numbers in binary format (without losing precision). Changing the kernel is probably not a big problem, but you will have to change the classes yourself (Skin_surface_3 if I remember correctly). However, the speed will decrease dramatically. That's why I didn't implement it.

If I remember correctly you asked a few weeks ago as well whether it is possible to swap triangulations (or copy them). This isn't implemented either, but wouldn't be a big change. I can do that before the next release if you need it.

-- Nico

On Wed, Jul 2, 2008 at 6:11 PM, Bálint Miklós <> wrote:

Hi Nico,

Yes, I am using power crust to get the spheres :) Actually, I cannot send you the spheres now right away... the reason is that for this example I had the spheres directly from memory from the output of power crust, and in my save spheres function currently I lose from the accuracy of the coordinates and when I load in the slightly perturbed balls then everything is fine, because degeneracies are mostly "solved" like this. Anyway, if I subdivide the "bad balls' surface" I will get vertices which are not a number or something strange because the "bad" triangles are all rendered with the vertex in the origin of the coordinate system.

Anyway, I have downloaded the paper on which your package is based (http://dx.doi.org/10.1016/j.comgeo.2006.01.003 is the best for this, right?). I will go through it to understand all the details and maybe I can come up with a solution for this strange robustness issues by myself. I will let you know what solution did I come up with.... and probably I will have to do something about the quality and the size of the mesh later on as well.

Only one small question. Do you think there is a lot of modification needed in order to make the skin surface package to work with other number types like GMP?

Thanks for all your help,
balint

Nico Kruithof wrote:

No, I don't know. Maybe you could subdivide the surface to see whether the patches become spherical. Could you send me the spheres otherwise? I'll try myself (maybe next week). It looks like you are using the power crust or a similar algorithm. Is this right? In that case you would expect that the spheres cover the medial axis and I cannot explain the components.

Bests,
Nico