I found this example extremely helpful as well.

http://pv3d.org/2008/11/17/finding-2d-coordinates/

You have to set autoCalcScreenCoords equal to true on your display object in order to get DisplayObject3D.screen.x

Hardware support for 3D in Flash would indeed be a wonderful thing .. and may be realised with the next release of the Flash Player.

Doing this sort of thing is so common that hardware generally supports the drawing of “triangle strips”. When you draw strips, all of your diagonals are oriented the same way, and its expected that the order of your vertex indices will alternate, so the normals are calculated correctly with that in mind. For instance, in your example, the vertices might go like this: [1, 2, 3], [2,3,4], [3,4,5], [4,5,6]. Much easier. Anyway, the triangle made from indices 1, 2, 3 is wound clockwise, but the triangle from indices 2, 3, 4 is wound counterclockwise, so the normals are in the opposite directions. In fact, OpenGL would want an array without the repeated indices: [1, 2, 3, 4, 5, 6]. But that’s hardware–I dunno if there are any performance implications about implementing something like that in software or not. Hmmm… HMMM…

XSLT might be worth considering if you’re forced to transform XML data a bunch.

For the UV coords–if you’re using a WireframeMaterial, you may not even need them at all. The UV coordinates are used for texture alignment (which is why there’s only two values… textures are 2D). You might try applying a texture to your graph just to see if it comes out right or not. The WireframeMaterial, I believe, just uses the drawing API to trace the outline of each triangle… so you might be able to null out the UVs and avoid all of those allocations.

Anyway, I’m glad the code actually helped! I was afraid it’d be too much of a mess since it was just thrown together real quick for testing purposes.

So in closing, a shamless plug:
irc.freenode.net
#papervision3d