I'm still confused about Manifold
I realize the topic of what makes a shape manifold or not has been
discussed here many times, but I'm still a bit confused.
As I understand it (or misunderstand it), for a shape to be manifold
every edge should be connected to two faces and those faces should be
non-zero in area. This code:
$fn = 10;
difference() {
cylinder(r = 3, h = 3);
translate([0, 0, 1]) cylinder(r1 = 0, r2 = 2, h = 2);
}
seems to comply but OpenSCAD tells me "WARNING: PolySet -> Manifold
conversion failed: NotManifold. Trying to repair and reconstruct mesh."
If I turn off the 'stop after first warning' option, it figures it out
and happily completes the render - but why does it need to resort to the
fallback method?
Windows 10 with Dev Snapshot 2024.03.22
Cheers,
Steve
I don't get any warning with the version I am using: version 2024.01.26
(git 358af340c), so it looks like a regression.
Compiling design (CSG Tree generation)...
Compiling design (CSG Products generation)...
Geometries in cache: 4
Geometry cache size in bytes: 5344
CGAL Polyhedrons in cache: 0
CGAL cache size in bytes: 0
Compiling design (CSG Products normalization)...
Normalized tree has 2 elements!
Compile and preview finished.
Total rendering time: 0:00:00.030
You could try making the cone a little bit taller than the cylinder to see
if that gets rid of the warning. Coincident surfaces can cause problems.
On Fri, 22 Mar 2024 at 19:51, Steve Lelievre via Discuss <
discuss@lists.openscad.org> wrote:
I realize the topic of what makes a shape manifold or not has been
discussed here many times, but I'm still a bit confused.
As I understand it (or misunderstand it), for a shape to be manifold
every edge should be connected to two faces and those faces should be
non-zero in area. This code:
$fn = 10;
difference() {
cylinder(r = 3, h = 3);
translate([0, 0, 1]) cylinder(r1 = 0, r2 = 2, h = 2);
}
seems to comply but OpenSCAD tells me "WARNING: PolySet -> Manifold
conversion failed: NotManifold. Trying to repair and reconstruct mesh."
If I turn off the 'stop after first warning' option, it figures it out
and happily completes the render - but why does it need to resort to the
fallback method?
Windows 10 with Dev Snapshot 2024.03.22
Cheers,
Steve
OpenSCAD mailing list
To unsubscribe send an email to discuss-leave@lists.openscad.org
I would expect a problem with this because the top of your cut-out is flush with the top of the cylinder it cuts from. Try making the cut-out ever so slightly taller or higher.
Cheers,
Lenore
On Mar 22, 2024, at 15:50, Steve Lelievre via Discuss discuss@lists.openscad.org wrote:
I realize the topic of what makes a shape manifold or not has been discussed here many times, but I'm still a bit confused.
As I understand it (or misunderstand it), for a shape to be manifold every edge should be connected to two faces and those faces should be non-zero in area. This code:
$fn = 10;
difference() {
cylinder(r = 3, h = 3);
translate([0, 0, 1]) cylinder(r1 = 0, r2 = 2, h = 2);
}
seems to comply but OpenSCAD tells me "WARNING: PolySet -> Manifold conversion failed: NotManifold. Trying to repair and reconstruct mesh." If I turn off the 'stop after first warning' option, it figures it out and happily completes the render - but why does it need to resort to the fallback method?
Windows 10 with Dev Snapshot 2024.03.22
Cheers,
Steve
OpenSCAD mailing list
To unsubscribe send an email to discuss-leave@lists.openscad.org
On Mar 22, 2024, at 12:51 PM, Steve Lelievre via Discuss discuss@lists.openscad.org wrote:
As I understand it (or misunderstand it), for a shape to be manifold every edge should be connected to two faces and those faces should be non-zero in area.
Also, no two faces should intersect, and no two faces should overlap on the same plane.
OpenSCAD, notably, is known to occasionally leave infinitely thin cleavages in its output when subtracting shapes that have coinciding faces, thus violating the last requirement. This is caused by either a bug or a floating point rounding error in the various libraries that OpenSCAD can use to process its boolean shape operations.
Solution: make sure the shapes you difference away are bigger than the shape you are subtracting from.
-Revar
Although in this case the dimensions are all integers, so the faces are
exactly coincident, a situation CGAL would handle correctly.
On Fri, 22 Mar 2024 at 23:05, Revar Desmera via Discuss <
discuss@lists.openscad.org> wrote:
On Mar 22, 2024, at 12:51 PM, Steve Lelievre via Discuss <
discuss@lists.openscad.org> wrote:
As I understand it (or misunderstand it), for a shape to be manifold
every edge should be connected to two faces and those faces should be
non-zero in area.
Also, no two faces should intersect, and no two faces should overlap on
the same plane.
OpenSCAD, notably, is known to occasionally leave infinitely thin
cleavages in its output when subtracting shapes that have coinciding faces,
thus violating the last requirement. This is caused by either a bug or a
floating point rounding error in the various libraries that OpenSCAD can
use to process its boolean shape operations.
Solution: make sure the shapes you difference away are bigger than the
shape you are subtracting from.
-Revar
OpenSCAD mailing list
To unsubscribe send an email to discuss-leave@lists.openscad.org
Thanks for the reply.
For sure, I can tinker with the design but actually my question was
intended to be about what makes it appear non-manifold to the software. So
for example in terms of the reply, why (computationally)would oversizing
the cutout make the problem go away?
Steve
On Fri, 22 Mar 2024 at 15:08, Lenore Horner via Discuss <
discuss@lists.openscad.org> wrote:
I would expect a problem with this because the top of your cut-out is
flush with the top of the cylinder it cuts from. Try making the cut-out
ever so slightly taller or higher.
Cheers,
Lenore
On Mar 22, 2024, at 15:50, Steve Lelievre via Discuss <
discuss@lists.openscad.org> wrote:
I realize the topic of what makes a shape manifold or not has been
discussed here many times, but I'm still a bit confused.
As I understand it (or misunderstand it), for a shape to be manifold
every edge should be connected to two faces and those faces should be
non-zero in area. This code:
$fn = 10;
difference() {
cylinder(r = 3, h = 3);
translate([0, 0, 1]) cylinder(r1 = 0, r2 = 2, h = 2);
}
seems to comply but OpenSCAD tells me "WARNING: PolySet -> Manifold
conversion failed: NotManifold. Trying to repair and reconstruct mesh." If
I turn off the 'stop after first warning' option, it figures it out and
happily completes the render - but why does it need to resort to the
fallback method?
Windows 10 with Dev Snapshot 2024.03.22
Cheers,
Steve
OpenSCAD mailing list
To unsubscribe send an email to discuss-leave@lists.openscad.org
OpenSCAD mailing list
To unsubscribe send an email to discuss-leave@lists.openscad.org
As I understand it, it's the coincident faces that are a problem. From the point of view of the cylinder, that surface has an outward normal pointing +z. From the point of view of the removed cone, that same surface should have a normal of -z (- because subtracting is putting stuff outside the shape and nothingness inside). I think that's were the software loses it. (Though someone else says for integers it shouldn't actually get confused. In which case, I don't know.)
Lenore
On Mar 22, 2024, at 19:10, Steve Lelievre via Discuss discuss@lists.openscad.org wrote:
Thanks for the reply.
For sure, I can tinker with the design but actually my question was intended to be about what makes it appear non-manifold to the software. So for example in terms of the reply, why (computationally)would oversizing the cutout make the problem go away?
Steve
On Fri, 22 Mar 2024 at 15:08, Lenore Horner via Discuss <discuss@lists.openscad.org mailto:discuss@lists.openscad.org> wrote:
I would expect a problem with this because the top of your cut-out is flush with the top of the cylinder it cuts from. Try making the cut-out ever so slightly taller or higher.
Cheers,
Lenore
On Mar 22, 2024, at 15:50, Steve Lelievre via Discuss <discuss@lists.openscad.org mailto:discuss@lists.openscad.org> wrote:
I realize the topic of what makes a shape manifold or not has been discussed here many times, but I'm still a bit confused.
As I understand it (or misunderstand it), for a shape to be manifold every edge should be connected to two faces and those faces should be non-zero in area. This code:
$fn = 10;
difference() {
cylinder(r = 3, h = 3);
translate([0, 0, 1]) cylinder(r1 = 0, r2 = 2, h = 2);
}
seems to comply but OpenSCAD tells me "WARNING: PolySet -> Manifold conversion failed: NotManifold. Trying to repair and reconstruct mesh." If I turn off the 'stop after first warning' option, it figures it out and happily completes the render - but why does it need to resort to the fallback method?
Windows 10 with Dev Snapshot 2024.03.22
Cheers,
Steve
OpenSCAD mailing list
To unsubscribe send an email to discuss-leave@lists.openscad.org mailto:discuss-leave@lists.openscad.org
OpenSCAD mailing list
To unsubscribe send an email to discuss-leave@lists.openscad.org mailto:discuss-leave@lists.openscad.org
OpenSCAD mailing list
To unsubscribe send an email to discuss-leave@lists.openscad.org
seems to comply but OpenSCAD tells me "WARNING: PolySet -> Manifold conversion failed: NotManifold. Trying to repair and reconstruct mesh." If I turn off the 'stop after first warning' option, it figures it out and happily completes the render - but why does it need to resort to the fallback method?
This doesn’t sounds right. The only PolySet -> Manifold conversions going on here should be the primitives.
I suspect you found an actual bug.
This code is in active development, so it’s prone to breaking from time to time, and we don’t yet have good tests for situations where we use fallback mechanisms like this.
I’ll look into it and open an issue or PR if I find or fix something.
-Marius
On 2024-03-22 6:19 p.m., Marius Kintel wrote:
I’ll look into it and open an issue or PR if I find or fix something.
Thank you, and in case it helps: I have found that setting the r1
parameter of the cutaway cone to a tiny non-zero value leads to the
render happening on the first attempt.
Steve
Fixed in https://github.com/openscad/openscad/pull/5062
Our old cone generator was a bit sloppy, something that the new Manifold check picked up on.
On Mar 22, 2024, at 22:51, Steve Lelievre steve.lelievre.canada@gmail.com wrote:
On 2024-03-22 6:19 p.m., Marius Kintel wrote:
I’ll look into it and open an issue or PR if I find or fix something.
Thank you, and in case it helps: I have found that setting the r1 parameter of the cutaway cone to a tiny non-zero value leads to the render happening on the first attempt.
Steve