On Sunday 19 September 2021 15:00:44 Daniel wrote:

I've got 16GB in the buster machine, I'm running openscad on and its not
enough, 32 in this machine but the later openscads won't run on my
stretch libc6. Gotta get this updated to buster or bullseye.

Cheers, Gene Heskett

"There are four boxes to be used in defense of liberty:
soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
If we desire respect for the law, we must first make the law respectable.

• Louis D. Brandeis
  Genes Web page http://geneslinuxbox.net:6309/gene

I was able to render it on my new PC in 30min, 19 seconds, using about
7.8GB of memory.

Is this something that could better be done using thej twist argument of
linear_extrude()?

On Sun, Sep 19, 2021 at 2:49 PM Gene Heskett gheskett@shentel.net wrote:

Hi!

You have a serious issue with the design.

I did a little mod of the code:

for(i = [0:0.2:179.8]){
if(i>2.4 && i<100 ){

rotate([0,0,i])
color("red")
dylinder(12.5*(1+sin(i+50)),2,16,i);

Just coloured part of it red...
Do this, then preview.
What you see in the intersections is OpenSCAD struggling with deciding which surface is what.
Should the Red surface be the outside, should the yellow?
And what happens to the surface that's not on the outside?

Preview runs with a different accuracy and curve handling than the final rendering, so there's very few if any of these conflicts.

You can see the same issue if you create:

cylinder(r=10, h=20);
cylinder(r=5, h=20);

My laptop has 32GB, and it's also struggling to render it.
It's a beautiful design. I hope you find a way around it.

Trygve

Den 19. september 2021 kl. 21.00.44 +02.00 skrev Daniel daniel@puggsy.net:

Since OpenSCAD is functional, does it memoize? I was wondering if it would
be possible to let it work out the hard parts only once and then duplicate
the results.

On Sun, Sep 19, 2021 at 4:04 PM Jordan Brown openscad@jordan.maileater.net
wrote:

No it only caches geometry with a key which is the arguments that made it.
Those arguments are all calculated each time.

On Mon, 20 Sept 2021 at 01:24, Father Horton fatherhorton@gmail.com wrote:

Your shape is made of 32 winding twisted extrusions I will call "cords".
Each cord is made of 900 small linear extrudes of a quadrilateral. In
total, there are more than 900*32 contacts between small pieces in the
final model. It is not surprising it takes so long to render. Each of these
contacts have to be identified and the interferences should be computed by
CGAL (the render process.)

In your code, the shape is the union of slices of the cords. Processing
that way you cannot benefit from the OpenSCAD cache. It would be faster to
compute a whole cord and union rotations/mirrors of it. That way the
process of a cord is done only once and cached. The union of the cords will
have 16*16 contacts.

By doing that, you will have a huge gain because the contacts found in a
cord will be reduced to 5180 instead of the previous value (90032). To
get an even better runtime cost you will need to avoid the unions in the
cord processing. To do it, the cord might be computed as a polyhedron where
no union takes place (and CGAL is not called.) To generate a cord using a
sweep is out of question because it will be hard to control the twist of
it. Skin is an alternative as Jordan pointed out. The sections to input the
skin are found in the polygon vertices of the dial module. I believe the
render time with this strategy will be 1 minute or less.

If you want to take this direction and need any help just ask.

Em dom., 19 de set. de 2021 às 20:01, Daniel daniel@puggsy.net escreveu:

Something like this works, using the generic sweep.  It took almost 2
minutes to render.  There may be little edges at the intersections,
though.  That may be hard to fix other than by making it oversized and
then intersecting the model with the vase solid of revolution.

include<BOSL2/std.scad>

function myf(i) =  i<100 ? 12.5*(1+sin(i+50)) : (42.65-15*sin((i)+280))/2;
step = 4;
path = [for(i=[0:step:180]) rot(i)move([2myf(i),0,i])];  // list of
transform matrices
section = square(4,center=true);
zrot_copies(n=16){
xflip_copy()sweep(section, path);
}

On Mon, Sep 20, 2021 at 5:51 AM Ronaldo Persiano rcmpersiano@gmail.com wrote:

Hi Daniel,

I wrote following code

include<dotSCAD/src/bezier_curve.scad>
include<dotSCAD/src/hull_polyline3d.scad>

//quaternion rotation function
function q(vector=[1,0,0],point=[0,5,0],theta=0)=
[
let(
v=vector/norm(vector),
p=[cos(theta/2),vsin(theta/2)],
p1=[p.x,-p.y],
q=[0,point],
pq=[p.xq.x-p.yq.y,p.xq.y+p.yq.x+cross(p.y,q.y)],
pqp1=[pq.xp1.x-pq.yp1.y,pq.xp1.y+pq.y*p1.x+cross(pq.y,p1.y)],
transformation=pqp1.y
)each
transformation
];

path=[[40,0,0],[60,0,50],[20,0,100],[20,0,180]];

path1=bezier_curve(.05,path);

path2=[for(i=[0:len(path1)-1])
q([0,0,1],path1[i],i/(len(path1)-1)*180)];

translate([0,0,2])
for(i=[0:20:360])rotate([0,0,i])
{
hull_polyline3d(path2,4,$fn=4);

mirror([0,1,0])
hull_polyline3d(path2,4,$fn=4);
}

linear_extrude(2)
difference(){
circle(r=41.5);
circle(r=30);
}

it works quite fast.
it uses dotscad library 2 files

regards
Sanjeev

On Mon, 20 Sept 2021 at 00:31, Daniel daniel@puggsy.net wrote: