Rendering fails (difference between F5 and F6)
On Dec 14, 2015, at 20:44 PM, Parkinbot rudolf@parkinbot.com wrote:
I just had a look, how skin() behaves when it has to connect polygons with
different number of points (mine doesn't do that at all for a good reason).
The result is somehow unexpected.
Yes, the Steiner point insertion is a bit automagic. Not sure I would call it unexpected though - the screenshot you posted is dominated by effects of polygon tessellation. Since we cannot represent non-planar polygons like that, we have to make a tessellation discussion. The actual Steiner point insertion kind of makes sense. Ideas are welcome.
And an immediate feature request would be: adding furcations like a
bifurcation and an inverse (anastomosis) to it.
That would be cool - perhaps something to leave for the future.
As always, if we can prototype partial functionality in user-space, that would help understand it.
If there are missing features in the OpenSCAD language which makes this hard, we could address that in a separate project.
-Marius
kintel wrote
Yes, the Steiner point insertion is a bit automagic. Not sure I would call
it unexpected though - the screenshot you posted is dominated by effects
of polygon tessellation. Since we cannot represent non-planar polygons
like that, we have to make a tessellation discussion. The actual Steiner
point insertion kind of makes sense. Ideas are welcome.
I wouldn't say it makes sense in this specific situation, as Steiner points
are mainly meant for mesh repair purposes in odd non-planar situations like
z-fighting. Modelling the same thing using hull() gives exactly the expected
behavior - unpaired triangles.
hull()
{
cylinder($fn=6,r=50);
translate([0, 0, 70])
cylinder($fn=8,r=50);
}
http://forum.openscad.org/file/n15173/showcase7.png
and here: left is hull, right is Steiner point. What would you expect?
http://forum.openscad.org/file/n15173/showcase6.png
or look at this:
r=50;
height=140;
layers = 2;
skin([for (i=[0:layers-1])
transform(translation([150,0,iheight/layers]) *
rotation([0,0,-0i]),
i==0?circle($fn=6+47*i,r=r): [[0,0], [0,10]]) ]);
http://forum.openscad.org/file/n15173/showcase8.png
So, if I wanted Steiner point semantics, I could refine the 2D-shape in
principle myself. But if skin() does Steiner point semantics unwantedly, I
don't have any means to prevent it.
One could offer Steiner point semantics as a flavor by adding a mode
attribute, but I guess this is not necessary, as the Steiner regularization
used by skin() is easily done by a an own function (and in a more problem
specific manner) when preparing the dataset.
kintel wrote
Parkinbot wrote
And an immediate feature request would be: adding furcations like a
bifurcation and an inverse (anastomosis) to it.
That would be cool - perhaps something to leave for the future.
As always, if we can prototype partial functionality in user-space, that
would help understand it.
If there are missing features in the OpenSCAD language which makes this
hard, we could address that in a separate project.
I'll do the job and implement it in a more generic (=inefficient) way and on
the basis of my sweep() - as it is already on my todo list.
What makes it hard? Implementing matrix related stuff in OpenSCAD is so
painful - especially if you're committed more to imperative programming.
Think about an NxMxL matrix A and you want to alter some value, say A[10]
[101][100]. In OpenSCAD you have to go through three loops, get thousands of
copies of copies ...
See e.g. also the fibonacci discussion in the thingiverse OpenSCAD group
http://www.thingiverse.com/groups/openscad/topic:3916#comment-732942
which is kind of absurd, as fibonacci is a recursive problem. But runtime is
runtime.
- Rudolf -
--
View this message in context: http://forum.openscad.org/Rendering-fails-difference-between-F5-and-F6-tp15041p15173.html
Sent from the OpenSCAD mailing list archive at Nabble.com.
I've done a fast implementation of a bifurcation on the basis of skin() and
the proposed representation scheme. It showed me, that there is no "natural"
semantics in implementing it.
There is a root part, child parts and a connector. The connector is the salt
in the soup. I think it needs either some problem specific treatment or if
it is generated automatically at least some interpolation scheme like
B-spline to provide a good approximation of the transition. The stuff can
(and should) be put into library first, to see how it works. This will take
time.
use
<
scad-utils/transformations.scad
>
use
<skin.scad>
use
<naca4.scad>
// http://www.thingiverse.com/thing:898554
a = af(120, 100, 0, 0);
b = af(100, 100, 0, 10);
c1 = af(50, 50, 0, 20);
c2 = af(35, 50, 55, 20);
d1 = af(30, 50, 10, 30);
d2 = af(30, 50, 60, 30);
skin_furc([a, b, [c1, c2], [d1, d2]]);
// create and place airfoil shape
function af(L, N, x, z) = transform(translation([x, 0, z]),
airfoil_data(naca = 1412, N=N, L=L));
// consumes [..., [X], [[x], [y]], ...]
module skin_furc(C)
{
i = find_furc(C);
if(i)
{
S=split(C, i+1);
adapter = [S[0][i], furc(S[2][0], S[1][0])];
skin(S[0]);
skin(adapter);
skin(S[1]);
skin(S[2]);
}
else
skin(C);
}
function find_furc(C) = let (res = [for(i=[0:len(C)-2])
if(!len(C[i][0][0]) && len(C[i+1][0][0])) i]) res[0];
// calculate furcation adapter
function furc(B, C) =
let(b0 = floor(len(B)/2))
let(b1 = len(B)-b0)
let(c = floor(len(C)))
[for (i=[0:b0+c+b1-1])
i
<
b0?B[i]:
i<b0+c?C[i-b0]:
B[i-c]];
</quote
>
// split matrix at row n into root and children
function split(M, n) =
[
[for(i=[0:n-1]) M[i]],
[for(i=[n:len(M)-1]) M[i][0]],
[for(i=[n:len(M)-1]) M[i][1]]
];
http://forum.openscad.org/file/n15182/showcase9.png
- Rudolf -
--
View this message in context: http://forum.openscad.org/Rendering-fails-difference-between-F5-and-F6-tp15041p15182.html
Sent from the OpenSCAD mailing list archive at Nabble.com.
Hello everyone,
Maybe interesting to see the first printed results?
http://forum.openscad.org/file/n15228/WP_20151218_09_54_07_Pro.jpg
I tested the result and the sound is fine!
Only the other part, the bow, has its tone hole on the wrong place. So I
have to make a new bow.
I can improve the bow's design in the same way as the bell desig.
Thanks for helping!
Johan
--
View this message in context: http://forum.openscad.org/Rendering-fails-difference-between-F5-and-F6-tp15041p15228.html
Sent from the OpenSCAD mailing list archive at Nabble.com.
Brilliant and well done. Looks great :)
--
View this message in context: http://forum.openscad.org/Rendering-fails-difference-between-F5-and-F6-tp15041p15236.html
Sent from the OpenSCAD mailing list archive at Nabble.com.
Wow! How many parts?
On Dec 19, 2015 9:56 PM, "Neon22" mschafer@wireframe.biz wrote:
Brilliant and well done. Looks great :)
--
View this message in context:
http://forum.openscad.org/Rendering-fails-difference-between-F5-and-F6-tp15041p15236.html
Sent from the OpenSCAD mailing list archive at Nabble.com.
OpenSCAD mailing list
Discuss@lists.openscad.org
http://lists.openscad.org/mailman/listinfo/discuss_lists.openscad.org
Really excellent Johan, you have put new life into the instrument.
Inspiring...
Rob
On 20/12/15 03:36, Johan Jonker wrote:
Hello everyone,
Maybe interesting to see the first printed results?
http://forum.openscad.org/file/n15228/WP_20151218_09_54_07_Pro.jpg
I tested the result and the sound is fine!
Only the other part, the bow, has its tone hole on the wrong place. So I
have to make a new bow.
I can improve the bow's design in the same way as the bell desig.
Thanks for helping!
Johan
--
View this message in context: http://forum.openscad.org/Rendering-fails-difference-between-F5-and-F6-tp15041p15228.html
Sent from the OpenSCAD mailing list archive at Nabble.com.
OpenSCAD mailing list
Discuss@lists.openscad.org
http://lists.openscad.org/mailman/listinfo/discuss_lists.openscad.org
Three parts:
the bow, the key and the bell.
--
View this message in context: http://forum.openscad.org/Rendering-fails-difference-between-F5-and-F6-tp15041p15249.html
Sent from the OpenSCAD mailing list archive at Nabble.com.