Jan Dvorak nous apporta ses lumieres en ce 2008/03/02 19:07:
> Patrick Elliott napsal(a):
>> Decided to rework the design of the octahedron, so that it *should* 
>> allow me to do this:
>>
>> #declare octoside = {blah}
>>
>> union {
>>   object {octoside
>>           rotate <0,90,0>}
>>   ...
>> }
>>
>> and get the result I want. The problem is, this is as far as I got:
>>
>> camera{
>>         orthographic
>>         location <0,sqrt(2)/2,-3> look_at <0,sqrt(2)/2,0> angle 60
>>         right x up y
>> }
>>
>> light_source{<5,5,-10>,1.5}
>>
>> //intersection {
>>   union {
>>     sphere {<-1,0,0>,0.075}
>>     sphere {<1,0,0>,0.075}
>>     sphere {<0,sqrt(2),0>,0.075}
>>     cylinder {<-1,0,0>,<1,0,0>,0.075}
>>     cylinder {<0,sqrt(2),0>,<1,0,0>,0.075}
>>     cylinder {<-1,0,0>,<0,sqrt(2),0>,0.075}
>>     texture {
>>       pigment {rgb <1,1,1>}
>>     }
>>     //triangle {<-1,0,-0.075>,<0,sqrt(2),-0.075>,<1,0,-0.075>}
>>     //triangle {<-1,0,0.075>,<0,sqrt(2),0.075>,<1,0,0.075>}
>>   //rotate <45,0,0> <- test purposes, move to last statement in final 
>> object, with translate.
>>   //translate <0,0,-1>
>>   }
>> //  box {<-1,0,-0.075>,<1,sqrt(2),0.075>}
>> //
>> //}
>>
>> Neeless to say, since the original used intersection, I took a chance 
>> on that working. It didn't. The **best** solution would be if I could 
>> tie triangles to the center points and weld them to the main object, 
>> to produce a solid, or at least one solid enough to cut sections out 
>> of and have material left behind, instead of a hollow shell with a 
>> hole in it. It only get worse if I want to late make a crystal 
>> version, or some other such thing, where the entire inside has to be 
>> solid, instead of it being hollow. There has got to be some means I 
>> can fill in the sides, like I want, and also the rest of the interior 
>> of the final object. I am just not seeing it at the moment. And, if 
>> its something blindingly obvious, like using triangles some how, then 
>> I can even set the exact points I want, instead of using the 45 degree 
>> angle trick and computing the sqrt(2) length for the side being 
>> rotated into place. Can't imaging why I didn't try something like this 
>> before... lol Might have been easier to model this bloody thing as a 
>> patch.
>>
> When I think about it now, wouldn't it be possible to use a superellipsoid?
> 
superellipsoid{<2,2>} give you an octaedron. Smaler values make the faces curve 
out, while the edges stay sharp. Larger values make the faces cave in.

-- 
Alain
-------------------------------------------------
You know you've been raytracing too long when you read about an algorithm or 
datastructure and your first thought is: "How can I use this to speed up 
raytracing?"
Christoph Rieder

Post a reply to this message