Title: Bug in triangulating polygons with multiple contours
Well, it's more or less a bug. The GeometryInfo actualy is not really generating a "planar surface with a hole" but a sort of a mesh. If the "hole" coordinates are not in the same plane, it will create a 3d surface which is not in the plane, but looks more like a triangle strip generated from the two contours. Now what happens if you rotate the hole 180 degrees ? Of course the shape will be corrupted, from your human point of view, but not from the mathematical point. The example below just shows that behavior: the first polygon is facing the positive y axis while the hole is facing the negative y axis. What should the algorithm decide in this case ? It's a distorted shape or a real hole ? How can he decide ?
 
What does Java3D guys think ?
 
Cheers,
 
Florin
-----Urspr�ngliche Nachricht-----
Von: Dave Smith [mailto:[EMAIL PROTECTED]]
Gesendet: Freitag, 10. Januar 2003 17:14
An: [EMAIL PROTECTED]
Betreff: Re: [JAVA3D] Bug in triangulating polygons with multiple contours

I'm surprised such a simple example doesn't work.  I even reduced
the noise and made everything planar
 
Looks like a bug to me.
 
-DaveS
 
ps.  Used same setup
pps.  Here is the modified data:
 
Point3d[] coordinates = {
                //outer bounds of the polygon
                new Point3d(2., 0.0, 0.0),
                new Point3d(2., 0.0, 1.3),
                new Point3d(0.0, 0.0, 1.3),
                new Point3d(0.0, 0.0, 0.0),
 
                //the hole
                new Point3d(1.93, 0.0, 0.07),
                new Point3d(0.07, 0.0, 0.07),
                new Point3d(0.07, 0.0, 1.13),
                new Point3d(1.93, 0.0, 1.13)
                };
----- Original Message -----
Sent: Friday, January 10, 2003 8:33 AM
Subject: [JAVA3D] Bug in triangulating polygons with multiple contours

It seems that sometimes when i attempt to use GeometryInfo by specifying polygons with holes, the triangulated
results aren't correct. The test method attached can duplicate this, it only creates a Shape3D however, since even a minimal program to display and maybe rotate the view requires a bit of work. I hope you guys can plug it into something you already have to confirm this.

I have had these results with java3d v1.3 on win2000 platform.
-tero

Reply via email to