On Wed, 13 Mar 2002 14:36:39 +0100, Christoph Hormann <chr### [at] gmxde>
wrote:
> I fixed a few problems since i posted the first version in p.b.s-f.  Also
> wrote some documentation and an add-on for the Povray 3.5 insert menu.
>
> http://www.tu-bs.de/~y0013390/pov/ic/index.html

Here is proposition of IC_Box macro with function returning linear distance to
the box shape.

#macro IC_Box( Start , End )
  #local Start=Start+<0,0,0>;
  #local End=End+<0,0,0>;

  #local Cx=max(Start.x, End.x);
  #local Cy=max(Start.y, End.y);
  #local Cz=max(Start.z, End.z);
  #local Dx=min(Start.x, End.x);
  #local Dy=min(Start.y, End.y);
  #local Dz=min(Start.z, End.z);

  #local Inside=function{pattern{object{box{ Start , End }}}};
    
  function(x, y, z){
    select(
      Inside(x,y,z),
      0,
      f_r(min(x-Dx,max(x-Cx,0)),min(y-Dy,max(y-Cy,0)),min(z-Dz,max(z-Cz,0))),
      max(x-Cx, y-Cy, z-Cz, Dx-x, Dy-y, Dz-z)
    )
  }
#end

ABX

Post a reply to this message

> 
> Here is proposition of IC_Box macro with function returning linear distance to
> the box shape.
> 

Nice, now you just need the same for cylinders and cones and fix the scale
macro to maintain the linear distance (i'm not sure if that's possible)
and it will be a perfect distance function for all 'merge' CSGs...

BTW, wouldn't the f_rounded_box() function be faster in this case?

Christoph

-- 
POV-Ray tutorials, IsoWood include,                 
TransSkin and more: http://www.tu-bs.de/~y0013390/  
Last updated 13 Mar. 2002 _____./\/^>_*_<^\/\.______

Post a reply to this message

On Thu, 14 Mar 2002 15:06:20 +0100, Christoph Hormann <chr### [at] gmxde>
wrote:
> Nice, now you just need the same for cylinders and cones

Currently I'm changing cylinder. Cone will be next :-)

> and fix the scale
> macro to maintain the linear distance (i'm not sure if that's possible)

It is only possible for <a,a,a> scale I think.

> and it will be a perfect distance function for all 'merge' CSGs...

That's what was my intention.

> BTW, wouldn't the f_rounded_box() function be faster in this case?

I tried this f_rounded_box solution at first but I'm not sure it returns
linear distances after scaling so I did it myself.

ABX

Post a reply to this message

On Thu, 14 Mar 2002 15:06:20 +0100, Christoph Hormann <chr### [at] gmxde> wrote:
> now you just need the same for cylinders 

#macro IC_Cylinder (Start, End, Radius)
  #local Center = <0,0,0>+(Start+End)/2;

  #local Cx=Center.x;
  #local Cy=Center.y;
  #local Cz=Center.z;

  #local Len=vlength(End-Center);

  #if (Len=0)
    #error "iso_csg.inc: Error: zero height cylinder"
  #end

  #local fn_T=function { Reorient_Trans(End-Center, x)  }

  #local fn_X=
  function(x, y){
    select(x,max(x,y),select(y,max(x,y),f_r(x,y,0)))
  }
  #local fn_Y=
  function(x, y, z){
    fn_X( abs(fn_T(x, y, z).x)-Len, f_r(0,fn_T(x, y, z).y,fn_T(x, y, z).z)-Radius)
  }

  function(x, y, z){
    (fn_Y((x-Cx),(y-Cy),(z-Cz)))
  }
#end

ABX

Post a reply to this message

On Thu, 14 Mar 2002 15:06:20 +0100, Christoph Hormann <chr### [at] gmxde> wrote:
> Nice, now you just need the same for cones

#macro IC_Cone (Start, Radius1, End, Radius2)
  #switch(select(Radius1-Radius2,-1,0,1))
    #case(0)
      IC_Cylinder (Start, End, Radius1)
    #break
    #case(-1)
      #local temp=Start;
      #local Start=End;
      #local End=temp;
      #local temp=Radius1;
      #local Radius1=Radius2;
      #local Radius2=temp;
    #else
      #local Start=Start+<0,0,0>;

      #local Cx=Start.x;
      #local Cy=Start.y;
      #local Cz=Start.z;

      #local Len=vlength(End-Start);
      
      #if (Len=0)
        #error "iso_csg.inc: Error: zero height cone"
      #end

      #local fn_T=function { Reorient_Trans(End-Start, x)  }
      
      #local Edge=vnormalize(<Len,Radius1-Radius2,0>);
      #local Ex=Edge.x;
      #local Ey=Edge.y;

      #local fn_X=
        function(x, y){
          select(
            x,
            select(
              y-Radius1,
              -x,                                   // under bottom
              f_r(x,y-Radius1,0)                    // bottom rounded edge below
bootom plane
            ),
            select(
              x*Ex-Ey*(y-Radius1),
              f_r(x,y-Radius1,0),                   // bottom rounded edge above
bootom plane
              select(
                x*Ey+y*Ex-Radius1*Ex,
                max(-x,x-Len,x*Ey+y*Ex-Radius1*Ex), // inside and above top cone
                select(
                  x*Ex-Ey*y-(Ex*Len-Ey*Radius2),
                  x*Ey+y*Ex-Radius1*Ex,             // outside edge
                  select(
                    y-Radius2,
                    x-Len,                          // above top
                    f_r(x-Len,y-Radius2,0)          // top rounded edge
                  )
                )
              )
            )
          )
        }

      #local fn_Y=
        function(x, y, z){
          fn_X( fn_T(x, y, z).x, f_r(0,fn_T(x, y, z).y,fn_T(x, y, z).z))
        }

      function(x, y, z){
        (fn_Y((x-Cx),(y-Cy),(z-Cz)))
      }
  #end
#end

ABX

Post a reply to this message

> 
> 
> #macro IC_Cone (Start, Radius1, End, Radius2)
>   #switch(select(Radius1-Radius2,-1,0,1))
>  [...]

Wow, those selects can really confuse...

I think both these and the 'B' versions should be in separate include
files since they are purely optional.  

I have not tried yet, but i suppose using f_rounded_box() is faster than
your version while the cylinder/cone macros are better than a manual
translation of the rounding macros from 'shapes.inc' with radius 0.

I will also add a 'compensated' version of scale to maintain the linear
distance as good as possible (correcting with the average of the scales
with non isotropic scaling).

Christoph

-- 
POV-Ray tutorials, IsoWood include,                 
TransSkin and more: http://www.tu-bs.de/~y0013390/  
Last updated 13 Mar. 2002 _____./\/^>_*_<^\/\.______

Post a reply to this message

On Fri, 15 Mar 2002 13:43:40 +0100, Christoph Hormann <chr### [at] gmxde>
wrote:
> Wow, those selects can really confuse...

Yes. Perhaps You, Tor or someother fan of functions will find simpler version
but it was simples I found. But all rounded versions are only slightly slower
than yours sharp shapes.

> I think both these and the 'B' versions should be in separate include
> files since they are purely optional.  

Yes, currently I have 3 file on my machine
iso_csg - your
iso_rcsg - rounded version
iso_bcsg - bounding layer

iso_bcsg doesn't contain #include of #iso_csg or #iso_bcsg so it works with
both (depending what was included by user)

> I have not tried yet, but i suppose using f_rounded_box() is faster than
> your version

I suppose. Internall codec works much faster than VM code as f_r() shows. But
I only affraid about characteristic outside of surface with threshold 0 - does
it saves linear distance ?

> while the cylinder/cone macros are better than a manual
> translation of the rounding macros from 'shapes.inc' with radius 0

You can look into p.b.i at sample image.

> I will also add a 'compensated' version of scale to maintain the linear
> distance as good as possible (correcting with the average of the scales
> with non isotropic scaling).

Great.

ABX

Post a reply to this message

wrote:
> Here is proposition of ...

#macro IC_Inverse(Fn)
  function{-Fn(x,y,z)}
#end

ABX

Post a reply to this message

> 
> [...]
> I suppose. Internall codec works much faster than VM code as f_r() shows. But
> I only affraid about characteristic outside of surface with threshold 0 - does
> it saves linear distance ?
> 

I think so.

Christoph

-- 
POV-Ray tutorials, IsoWood include,                 
TransSkin and more: http://www.tu-bs.de/~y0013390/  
Last updated 13 Mar. 2002 _____./\/^>_*_<^\/\.______

Post a reply to this message

wrote:
> Here is proposition of ...

#macro IC_Threshold ( F , Threshold )
  function{ F(x,y,z) - Threshold }
#end

threshold is not a part of csg but this macro can be usefull when you want
combine isosurfaces with different thresholds

ANX

Post a reply to this message