On 08.09.2022 02:04, Nick Moore wrote:

That is an operation that can be fully automated. It is called
'fill2d' in AngelCAD:

'fill2d is a 2-dimensional boolean operation that removes internal holes
in in 2d shapes, but leaves a possibly concave outer contour intact.'

Carsten Arnholm

On 08.09.22 02:04, Nick Moore wrote:

I've seen this discussed before and I think there's no
general solution with the currently available features.

So a PR implementing this would be great. I would suggest
to separate this from offset() though. A dedicated fill()
seems like a good choice.

This would not trivially cover the mentioned "expand only
the outer outline" but that sounds like a tricky thing
that may need to be looked at separately.

Feel free to join the IRC channel if you are interested
in some more interactive discussion.

ciao,
Torsten.

On 9/8/2022 12:49 PM, Torsten Paul wrote:

Isn't that

module offset_outside(d) {
    difference() {
        offset(d) fill() children();
        difference() {
            fill() children();
            children();
        }
    }
}

?

On Fri, 9 Sep 2022, at 05:49, Torsten Paul wrote:

OK, great!  I'll clean up the first, simplest implementation and change it's name to fill(), get CI working and send in a PR in the next few days.

I was a bit worried about adding new builtin names which is why I was messing with offset() parameters but I guess that's just how it is :-).  It's conceptually simpler, and as Jordan points out, you can do the inside/outside offset thing by combining differences.

Is there a logical operation for fill() on a 3D shape?  I guess if you were unioning lots of objects (or even just two convex objects) you might end up with a little "air bubble" between them and so it could be handy to be have a way to eliminate this.

-----Nick

On 08.09.22 23:52, Nick Moore wrote:

It would be nice if you could add a simple test case, but we
can look at that once the PR is created.

https://github.com/openscad/openscad/blob/master/doc/testing.txt
Example: https://github.com/openscad/openscad/pull/4041/files

While there's always some potential for issues, a new builtin
on it's own is usually not a problem as scripts using that
name would just override this with their own and by definition
can't miss the builtin one that not existed at time of writing.

If that code works for the general case, even better.

As opposed to the 2d version, I never missed that so far, but
I suppose there could be cases where it would be useful.

ciao,
Torsten.

On Fri, 9 Sep 2022, at 08:12, Torsten Paul wrote:

No worries, I created https://github.com/openscad/openscad/pull/4348 and just copied the hull tests for now, I'll think up some more interesting test cases over the weekend.

If someone can review my C++ for not being utterly stupid that'd be great, there's no compiler warnings but it's been a very long time since I wrote C++ in anger and I can't totally remember how memory management works.

-----N

On Fri, Sep 09, 2022 at 12:12:17AM +0200, Torsten Paul wrote:

Issues arise when you have a library that does use the builtin and a
"main program" that uses a module of the same name. Or reverse.

Roger. 

--
** R.E.Wolff@BitWizard.nl ** https://www.BitWizard.nl/ ** +31-15-2049110 **
**    Delftechpark 11 2628 XJ  Delft, The Netherlands.  KVK: 27239233    **
f equals m times a. When your f is steady, and your m is going down
your a is going up.  -- Chris Hadfield about flying up the space shuttle.