Re: Super ellipsoid
I think the basic concern that people had about this isn't that it's
"dirty" but that since it's nonsense geometrically, could we have
confidence that this method would always work in the future? As noted,
this is already in BOSL2 as hull_points() with fast=true. In the previous
discussion I think people had examples where the run time improvement was
noticeable.
Basically: Is this a supported approach that we can rely on continuing to
work indefinitely?
On Sat, Feb 15, 2025 at 5:22 AM Rudolf parkinbot@digitaldocument.de wrote:
The "hack" came up with OpenSCAD 2015 if I remember well. Even it is
"dirty coding" I use it since then to speed up my personal coding and don't
care whether another solution would speed up my renders from 0:00:00.112
to 0:00:00.82 or so. However, the function might be a candidate for a
new primitive in BOSL2.
Am 15.02.2025 um 01:24 schrieb Adrian Mariano:
I think it's been quite a long time since that hack was discussed here.
One discovery from the last time around is that it's faster (I think
possibly a lot faster) to divide your points into triangles rather than
passing one huge "face". Maybe it triangulates the face? This hack is in
BOSL2 as hull_points().
On Fri, Feb 14, 2025 at 5:52 PM Rudolf via Discuss <
discuss@lists.openscad.org> wrote:
Sorry guys to open the theme again, but I can't reply to it, as I wasn't
on the list when it was posted and found the thread on the Empathy web
mirror ...
So, here is my compact code for the super ellipsoid. I used a tricky hull
to display the output. However this obviously only works for convex shapes
i.e. nx, ny, nz values >= 1
points = Superellipse(a=10, b=20, c=30, nx=2, ny=2, nz=2, step=3);
// draw point cloud (hull() works for convex shapes only)
hull() polyhedron(points, [[for(i=[0:len(points)-1]) i]]);
function Superellipse(a=1, b=1, c=1, nx=2, ny=2, nz=2, step=10) =
[ for(u=[-90:step:90], v=[0:step:360])
let(x = a*sign(cos(u)) *pow(abs(cos(u)),2/nx) *
sign(cos(v))pow(abs(cos(v)),2/ny))
let(y = bsign(cos(u)) *pow(abs(cos(u)),2/nx) *
sign(sin(v))pow(abs(sin(v)),2/ny))
let(z = csign(sin(u)) *pow(abs(sin(u)),2/nz))
[x,y,z]
];
OpenSCAD mailing list
To unsubscribe send an email to discuss-leave@lists.openscad.org
To make it clear again: It was not at all my intention to discuss the
hack or any BOSL2 implementations of it. We all know and I think I
pointed that out that it is just a one-liner vehicle to display the
result of Superellipse() in a concise way. My point is that
Superellipse() would be a powerful candidate for a primitive.
Am 15.02.2025 um 14:21 schrieb Adrian Mariano via Discuss:
I think the basic concern that people had about this isn't that it's
"dirty" but that since it's nonsense geometrically, could we have
confidence that this method would always work in the future? As
noted, this is already in BOSL2 as hull_points() with fast=true. In
the previous discussion I think people had examples where the run time
improvement was noticeable.
Basically: Is this a supported approach that we can rely on continuing
to work indefinitely?
On Sat, Feb 15, 2025 at 5:22 AM Rudolf parkinbot@digitaldocument.de
wrote:
The "hack" came up with OpenSCAD 2015 if I remember well. Even it
is "dirty coding" I use it since then to speed up my personal
coding and don't care whether another solution would speed up my
renders from 0:00:00.112 to 0:00:00.82 or so. However, the
function might be a candidate for a new primitive in BOSL2.
Am 15.02.2025 um 01:24 schrieb Adrian Mariano:
I think it's been quite a long time since that hack was discussed
here. One discovery from the last time around is that it's
faster (I think possibly a lot faster) to divide your points into
triangles rather than passing one huge "face". Maybe it
triangulates the face? This hack is in BOSL2 as hull_points().
On Fri, Feb 14, 2025 at 5:52 PM Rudolf via Discuss
<discuss@lists.openscad.org> wrote:
Sorry guys to open the theme again, but I can't reply to it,
as I wasn't on the list when it was posted and found the
thread on the Empathy web mirror ...
So, here is my compact code for the super ellipsoid. I used a
tricky hull to display the output. However this obviously
only works for convex shapes i.e. nx, ny, nz values >= 1
points = Superellipse(a=10, b=20, c=30, nx=2, ny=2, nz=2,
step=3);
// draw point cloud (hull() works for convex shapes only)
hull() polyhedron(points, [[for(i=[0:len(points)-1]) i]]);
function Superellipse(a=1, b=1, c=1, nx=2, ny=2, nz=2, step=10) =
[ for(u=[-90:step:90], v=[0:step:360])
let(x = a*sign(cos(u)) *pow(abs(cos(u)),2/nx) *
sign(cos(v))*pow(abs(cos(v)),2/ny))
let(y = b*sign(cos(u)) *pow(abs(cos(u)),2/nx) *
sign(sin(v))*pow(abs(sin(v)),2/ny))
let(z = c*sign(sin(u)) *pow(abs(sin(u)),2/nz))
[x,y,z]
];
_______________________________________________
OpenSCAD mailing list
To unsubscribe send an email to discuss-leave@lists.openscad.org
OpenSCAD mailing list
To unsubscribe send an email todiscuss-leave@lists.openscad.org
I am not a BOSL2 user, but the point cloud can also be constructed to be
easily skinable. So the following formulation of the function will allow
to display the full parameter range of super ellipsoids. (partial)
concave shapes result from nx, ny, or nz <1
include <BOSL2/std.scad>
include <BOSL2/skin.scad>
points = Superellipse(a=5, b=12, c=10, nx=0.5, ny=1.5, nz=1.5, step=3);
skin(points, slices =1, closed=true);
function Superellipse(a=1, b=1, c=1, nx=2, ny=2, nz=2, step=10) =
[
for(v=[.5:step:359.5])
[
for(u=[-90:step:90])
let(x = a*sign(cos(u)) *pow(abs(cos(u)),2/nx) *
sign(cos(v))pow(abs(cos(v)),2/ny))
let(y = bsign(cos(u)) *pow(abs(cos(u)),2/nx) *
sign(sin(v))pow(abs(sin(v)),2/ny))
let(z = csign(sin(u)) *pow(abs(sin(u)),2/nz))
[x,y,z]
]
];
Am 15.02.2025 um 16:22 schrieb Rudolf via Discuss:
To make it clear again: It was not at all my intention to discuss the
hack or any BOSL2 implementations of it. We all know and I think I
pointed that out that it is just a one-liner vehicle to display the
result of Superellipse() in a concise way. My point is that
Superellipse() would be a powerful candidate for a primitive.
Am 15.02.2025 um 14:21 schrieb Adrian Mariano via Discuss:
I think the basic concern that people had about this isn't that it's
"dirty" but that since it's nonsense geometrically, could we have
confidence that this method would always work in the future? As
noted, this is already in BOSL2 as hull_points() with fast=true. In
the previous discussion I think people had examples where the run
time improvement was noticeable.
Basically: Is this a supported approach that we can rely on
continuing to work indefinitely?
On Sat, Feb 15, 2025 at 5:22 AM Rudolf parkinbot@digitaldocument.de
wrote:
The "hack" came up with OpenSCAD 2015 if I remember well. Even it
is "dirty coding" I use it since then to speed up my personal
coding and don't care whether another solution would speed up my
renders from 0:00:00.112 to 0:00:00.82 or so. However, the
function might be a candidate for a new primitive in BOSL2.
Am 15.02.2025 um 01:24 schrieb Adrian Mariano:
I think it's been quite a long time since that hack was
discussed here. One discovery from the last time around is that
it's faster (I think possibly a lot faster) to divide your
points into triangles rather than passing one huge "face".
Maybe it triangulates the face? This hack is in BOSL2 as
hull_points().
On Fri, Feb 14, 2025 at 5:52 PM Rudolf via Discuss
<discuss@lists.openscad.org> wrote:
Sorry guys to open the theme again, but I can't reply to it,
as I wasn't on the list when it was posted and found the
thread on the Empathy web mirror ...
So, here is my compact code for the super ellipsoid. I used
a tricky hull to display the output. However this obviously
only works for convex shapes i.e. nx, ny, nz values >= 1
points = Superellipse(a=10, b=20, c=30, nx=2, ny=2, nz=2,
step=3);
// draw point cloud (hull() works for convex shapes only)
hull() polyhedron(points, [[for(i=[0:len(points)-1]) i]]);
function Superellipse(a=1, b=1, c=1, nx=2, ny=2, nz=2,
step=10) =
[ for(u=[-90:step:90], v=[0:step:360])
let(x = a*sign(cos(u)) *pow(abs(cos(u)),2/nx) *
sign(cos(v))*pow(abs(cos(v)),2/ny))
let(y = b*sign(cos(u)) *pow(abs(cos(u)),2/nx) *
sign(sin(v))*pow(abs(sin(v)),2/ny))
let(z = c*sign(sin(u)) *pow(abs(sin(u)),2/nz))
[x,y,z]
];
_______________________________________________
OpenSCAD mailing list
To unsubscribe send an email to discuss-leave@lists.openscad.org
OpenSCAD mailing list
To unsubscribe send an email todiscuss-leave@lists.openscad.org
OpenSCAD mailing list
To unsubscribe send an email todiscuss-leave@lists.openscad.org
