modified Nef union/intersection for chamfers/fillets

Hi

I've been suffering in OpenSCAD with the lack of chamfer and fillet
operations.  Spent time trying to
figure out sweep() (Sorry Oskar, you haven't made something that's readable
or useable; Ronaldo you fixed it.  Sorry Ronaldo, you haven't either, though
your posts have code that are must-haves that a improved, even minimal,
sweep() must have.  Maybe please put it into a downloadable library.  Better
yet get OpenSCAD to incorporate the thing. Jeez.)  Failed.

Here's a simple idea.  CGAL is based on half-spaces, 3D planes with one side
"in" and one side "out"; any triangle in 3-space defines its plane, and a
bunch of them together (with their "in" sides, half-spaces, intersected)
create an arbitrary object surface.  I think this is not news.

But..

Every intersection or union of half-spaces could be softened by a chamfer or
fillet, defined as a plane cutting the corner.  Here's a bad hand
illustration suggesting where and how a fillet can be fitted to the two
planes.

<http://forum.openscad.org/file/t3073/IMG-1360_-_Copy.jpg>

I don't know how to work with this or the CGAL community.  Please help me
understand, how can we get a modification into CGAL so that
union_with_fillet() and intersection_with_chamfer() are some available
options in the library? And if we did wouldn't it enable OpenSCAD to have
those functions also? Can this be done? How can it be done?  What do you
want me to do if it needs to be done, and what can you do to help it be
done? Yes I am a mere newbie, but hey this is an issue that has plagued and
limited OpenSCAD for many years, let's fix it. I'll try to help where I can.

Tom

Thomas C Veatch PhD
tomveatch.com





--
Sent from: http://forum.openscad.org/

_______________________________________________
OpenSCAD mailing list
[hidden email]
http://lists.openscad.org/mailman/listinfo/discuss_lists.openscad.org

The BOSL2 library is downloadable, documented, and implements sweep and also
various other features to make rounding and chamfers easier.  For example
see

https://github.com/revarbat/BOSL2/wiki/rounding.scad
https://github.com/revarbat/BOSL2/wiki/skin.scad

You can use the basic cuboid module to generate cubes with chamfers or
roundings on specified edges.  


tom veatch wrote
> Discuss@.openscad

> http://lists.openscad.org/mailman/listinfo/discuss_lists.openscad.org





--
Sent from: http://forum.openscad.org/

_______________________________________________
OpenSCAD mailing list
[hidden email]
http://lists.openscad.org/mailman/listinfo/discuss_lists.openscad.org

On 06.01.21 04:19, tom veatch wrote:
> I don't know how to work with this or the CGAL community.
> Please help me understand, how can we get a modification
> into CGAL so that union_with_fillet() and
> intersection_with_chamfer() are some available options
> in the library?

I don't know, so far I personally only had some interaction
on the OpenSCAD github where they were quite helpful. I'm
not sure about actual code contributions, I believe they
require a signed copyright assignment (or whatever the legal
name is) as there's also a commercial CGAL release. But I
have not looked into details.

> And if we did wouldn't it enable OpenSCAD to have those
> functions also? Can this be done?

In theory that's simply it. There's a catch though as we
normally want to allow people to build OpenSCAD on existing
installations, not just the latest brand new platforms.
That's mostly a Linux distribution topic though.

That said, for a feature that important, I'm sure there's
ways to workaround that issue.

> What do you want me to do if it needs to be done, and
> what can you do to help it be done?

I'd say the first step would be to investigate if NEF
Polyhedrons are the way to go. CGAL has a second engine
for doing boolean operations which is supposedly quite
a bit faster. There's even requests to use a different
library altogether. So at any rate it might be useful
to look at the bigger picture too.

> Yes I am a mere newbie, but hey this is an issue
> that has plagued and limited OpenSCAD for many years,
> let's fix it. I'll try to help where I can.

That would be great, I'm sure it's not just me who
would like to see that feature.

As related/similar topic, maybe also have a look at
the work done in this PR:

Straight skeleton / voronoi diagram based roof #3486
https://github.com/openscad/openscad/pull/3486

ciao,
  Torsten.

_______________________________________________
OpenSCAD mailing list
[hidden email]
http://lists.openscad.org/mailman/listinfo/discuss_lists.openscad.org

A 3D version of offset() would do that. Sadly, that seems impractical to do
in OpenSCAD code.



--
Sent from: http://forum.openscad.org/

_______________________________________________
OpenSCAD mailing list
[hidden email]
http://lists.openscad.org/mailman/listinfo/discuss_lists.openscad.org

I Hear you. iv'e been peculating for years about this subject. Spoiler
edge-centric chamfer has unsolvable edge cases. In fact you can get into
trouble even in open-cascade based software. But that is not a reason not to
try something that work in most sane cases. The user just have to take more
adult care.  I know scary concept.

The Idea was much like your to  make the  boolean of A and B. Make a copy
and trow out all points and edges that was  in the  in the  A/B input
geometry. Now you have a path along the exact boundary of A/B.
  <http://forum.openscad.org/file/t1678/polytools.png>
<http://forum.openscad.org/file/t1678/polytools3.png>
This boundary could be sweept with some corner function (chamfer
/fillet,pretty). The caveat is that  the boundary branch breaks and that the
surrounding geometry might be to "thin" to support a naive implementation.

Now construct a second boundary offset pushing out on the surface each A and
B.
As A and B not necessarily meat at 90 degree the varying  amount of offset
had to be considered along the boundary depending on what angle A/B meet.
Graceful failure has to be defined each point where no true solution can be
found. At sharp points should radius shrink, adapt  or just run out? so many
ways to mess up.
  <http://forum.openscad.org/file/t1678/polytools2.png>
<http://forum.openscad.org/file/t1678/polytools4.png>

These three boundaries together is the information needed to construct some
profile sweep on rails.
Said sweep must handle any kind of wild and branching boundary .
 
This is all in the simplest union case. Cooking up a chamfer  between two
unioned primitives might be doable but from there on the pain begin.



--
Sent from: http://forum.openscad.org/

_______________________________________________
OpenSCAD mailing list
[hidden email]
http://lists.openscad.org/mailman/listinfo/discuss_lists.openscad.org