pouët.net

I'm talking about 3d boolean operations :



if not is there a possiblity to "emulate" the result (a cube with a hole made by a cylinder)???

booleans are EVIL

there are some hacks with stencil buffers that can be used to do something like that sometimes.... but most of the time it does not work correctly.

no booleans in wz1.

just import a mesh?

chaos: I've just seen the w3 website and the example textures and it is amazing... I just wanted to say "keep up the good work".

Just an idea: it *might* be cool for Flash textures, and I suppose it should be easy to port to flash9 or 10. The resolution independence and the high compression could be very nice for website backgrounds, for example.

Code your own CSG routines :P

Completely off-topic, but why are booleans so difficult anyway? At least if all the source objects have unbroken surfaces, it's just a matter of intersecting a bunch of polygons, then reconstructing the surface(s) of the combined object. There's probably something tricky in selecting which faces to keep and how to orient them, but I don't see what it might be. All the information should be there. But 3DS Max has really dodgy boolean features, for example, even when the source objects have simple topology, don't self-intersect and are perfectly connected. What gives?

it's just that you end up intersecting polygons against each other repeatedly, and at some point numerical drift simply starts to kill you.

ryg: yep, we do such things at work :)

I wrote a polygon tesselation algorithm once. It took me more than two years to make it solid and foolproof for a reason, but now you can throw even the ugliest degenerated stuff at it.

Computational Geometry is the hardest field of computer science that I know.

ryg: Any idea how quickly the errors tend to accumulate? Could you get around it for "all realistic situations" by converting to "very high" precision for doing the boolean operation? It's puzzling because it's not like speed is a higher priority in a 3D modeling application than numerical stability (at least if instability has a strong tendency to fuck up the topology of surfaces). Meh, I'll go look it up. And make a demo about it.

If you want to get a taste of the horrors of numerical stability try to understand the papers and code on this page:

http://www.cs.cmu.edu/~quake/robust.html

That's just the tip of the iceberg though.

and other versions?

they also have the tendency to skim over the robustness issues in the literature...