Hi, I'm new to the OpenSCAD forum, but I've been using OpenSCAD for about a
year (you can see a few of my rather simple designs at
https://www.thingiverse.com/gasstationwithoutpumps/things
https://www.thingiverse.com/gasstationwithoutpumps/things  ), with more
discussion of them and some unreleased design on my blog at
https://gasstationwithoutpumps.wordpress.com
https://gasstationwithoutpumps.wordpress.com  ).I came to this forum
looking for information on how to do something that I see as one of the
biggest holes in OpenSCAD: extruding along an arbitrary path.  I found out
about sweep.scad in the  list comprehension demos
https://github.com/openscad/list-comprehension-demos  , which provides a
somewhat klugey workaround—I have to have a polygon and a polygonal path
expressed as arrays of points (perhaps with an additional array of
orientation vectors), but what I wanted to start with was an SVG file with a
cross-section and an SVG file of the paths.Unfortunately, OpenSCAD can
convert arrays to geometry, but then provides no way to extract the arrays
again from the geometry.  So I can import SVG into OpenSCAD, but not use the
resulting imported objects as either the cross-sectional profile or as the
path.  One of the biggest things on my wish list is a way to get arrays of
points from objects.I found another link to an inkscape plug-in,
https://www.thingiverse.com/thing:2805184
https://www.thingiverse.com/thing:2805184  , that converts SVG files to
arrays of Bézier curves, which almost completes the task.After playing
around with both, I found some minor bugs:The definition of "cap" in
sweep_polyhedron relied on a test of len(caps), which is not acceptable to
the 2019-05 release.  I changed it to
cap    =  (closed || caps==[]) ? [false,false]:
(caps==true || caps==false)?      [caps,caps] :
(len(caps)==1) ?        [caps[0], caps[0]] :                    inv ?
[caps[1], caps[0]] :                                            [caps[0],
caps[1]]
removing the error message.  There are several other places in the list
comprehension demo code where similar tests relying on len(non_list) being
legal and returning undef are used, but that "feature" seems to have been
deprecated.For the inkscape plugin, I ran into two problems:
One was that a group scaling in my SVG file got applied twice so that the
cross-sectional profile was 96/25.4 times larger that it should have been.
Luckily this could be compensated for by scaling the output arrray, and
OpenScad makes multiplying an array by a constant easy.
A bigger annoyance was that the output created a polygon for each path in
the SVG file, when many of the paths were open paths.  I had to edit the
output code to make functions that returned the bezier_coordinates for each
array of points, rather than making a union of the bezier_polygons in a
module.  Then I could use the points (with to_3d() and
referenced_path_transforms()) to do a sweep.I'd like for the Inkscape plugin
to create those functions for me, so I can just "use" the output .scad file
without having to edit it.
I managed to get everything sort of working, but I still have some minor
problems:
If the number of steps in bezier_coordinates() is too high, then the tangent
computations at the end of the path end up producing "undef", and an ugly
artifact is created at the end of the path.  I think that this may be
related to OpenSCAD rounding to a coarse grid, so that multiple points in a
row are identical and the tangent is indeed undefined.  I fixed this problem
by keeping the number of steps small (6 was ok, 10 was too many for the
curves I was using).
The CGAL interpretation failed, probably because the tubes produced by the
sweep were not capped, and I had not buried the ends of all of them in other
geometry.  I can probably fix this by more careful placement of the end
points.
I would like to thank the creators of the list comprehension demos and the
Inkscape plugin, because I certainly would not have had the patience to
create them myself.  I hope that sometime soon extruding along a path
becomes a built-in to OpenSCAD, and that the language is extended so that
paths can be turned into arrays of points.

gasstationwithoutpumps.wordpress.com
www.thingiverse.com/gasstationwithoutpumps/things

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

I think I do something similar to what your asking.  I make cookie cutters
for my wife.  The outline of cookie cutter is created as SVG in inkscape.  I
use the inkscape extension from  https://www.thingiverse.com/thing:2805184
https://www.thingiverse.com/thing:2805184  to convert SVG to bezier
points in scad file.  Then I created a custom SVG in inkscape that is the
profile of the cutter; pulled into scad file as another bezier path.
Finally, I use the  BOSL https://github.com/revarbat/BOSL/wiki  library's
extrude_2d_shapes_along_bezier
https://github.com/revarbat/BOSL/wiki/beziers.scad#extrude_2d_shapes_along_bezier
.  It takes lots of tweeking, but I end up with my profile extruded along
the shape of the cookie cutter.  Let me know if you need more details.

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

Here's my simple take on cookie cutters. The examples are
just geometry, but the star() and heart() can be easily
replaced with an import("shape.svg").

$fa = 4;
$fs = .5;

height = 12;

module heart(w = 30) {
translate([0, w * sqrt(2) / 2]) {
rotate(-135) {
square(w);
translate([w / 2, 0]) circle(d = w);
translate([0, w / 2]) circle(d = w);
}
}
}

module star(n = 5, r1 = 22, r2 = 40) {
function r(a) = a % 2 == 0 ? r2 : r1;
function p(a) = [ -sin(a), cos(a) ];

assert(n >= 3);
points = [ for (a = [0 : 2 * n - 1]) r(a) * p(180 / n * a) ];
polygon(points);

}

module outline() {
difference() {
offset(0.01) children();
children();
}
}

module cut(height, base_height, r1, r2) {
translate([0, 0, -base_height]) cylinder(r = 2 * r1, h = base_height);
translate([0, 0, -0.01]) cylinder(r1 = r1, r2 = r2, h = height);
}

module cookie_cutter(height = 12, base_height = 1, r1 = 0.8, r2 = 0.2) {
minkowski() {
linear_extrude(0.01) outline() children();
cut(height, base_height, r1, r2);
}
}

cookie_cutter() star(5, 22, 40);
//cookie_cutter(r1 = 1.2, r2 = 0.6) star(5, 30, 52);
//cookie_cutter(height = 15, r1 = 1.2, r2 = 0.4) star(5, 22, 40);

//cookie_cutter() heart(30);

What is the cookie cutter shape you need to produce that makes this task
tricky?  (Or to put it another way, why isn't the posted example good
enough?)

I wonder if the CGAL assertion errors might be occurring due to self
intersection of the polyhedron generated by sweep().  If I'm imagining your
method correctly, that seems especially likely if your cookie cutter has
corners.  Depending on what you're trying to do, exactly, my rounded_sweep()
function might do what you want more easily.  It will handle corners without
trouble.  It extrudes a path with offsets from the original path, and you
can supply your own list of offsets (the "custom" option).  It's included
in BOSL2, but if you explain the cookie cutter problem a little better, I
might try a test of my own.

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

@gasstationwithoutpumps

See where I say, "It takes lots of tweeking..."  Issues like what you
mention are that tweeking part.  Other issues I have to deal with are that
the output of inkscape bezier extension uses a path format of [[start_pt1,
ctr_pt1, ctr_pt2, end_pt1], [start_pt2, ctr_pt3, ctr_pt4, end_pt2]].  So I
have to convert that to what BOSL lib expects which is [start_pt, ctr_pt1,
ctr_pt2, end_pt1, ctr_pt3, ctr_pt4, ...]  I have used both
extrude_bezier_along_bezier and extrude_2d_shapes_along_bezier.  I first
rotate or flip the profile path around z axis arbitrary amounts till it
comes out correctly.  Once I get a reliable workflow, I plan to modify the
inkscape extension to write it all out the way I want it.

@adrianv and tp3,

I think it just boils down to me wanting to be a PITA about the cutting
profile.  I wanted to create a profile that is straight up and down on the
inside of the cutter, but gradually slope from thin cutting edge at top say
1.2mm to 6-8 mm pressing handle at bottom.  The only way I could figure to
do that was to create the profile as bezier path;
http://forum.openscad.org/file/t2599/cutter_profile.jpg .  None of the
online examples came close to doing this.

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

Torsten, I had seen your design before at
https://forum.openscad.org/Minkowski-Sum-and-Cookie-Cutter-tp20733p20753.html
https://forum.openscad.org/Minkowski-Sum-and-Cookie-Cutter-tp20733p20753.html
and tried that approach.

Your design runs, though the 3D Minkowski operation makes it very slow—I
killed the run after 10 minutes of getting nothing on a simplified version
of the cookie cutter I was trying to make.

Your approach also has no fillets nor rounding in the Z dimension, though
the convex corners are nicely rounded in the XY direction.

I was trying to see whether it is possible to make injection-moldable (not
just 3D-printable) models in OpenSCAD, but getting the necessary 3D fillets,
rounding, and draft angles seems to be very difficult.

For 3D printing, I have no problem working with just linear_extrude.

gasstationwithoutpumps.wordpress.com
www.thingiverse.com/gasstationwithoutpumps/things

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

Using the rounded_sweep() of BOSL2 worked for me, once I fixed a bug in
geometry.scad:

The parallel check in offset should be

as both the first and last corners are undefined for an open path.

gasstationwithoutpumps.wordpress.com
www.thingiverse.com/gasstationwithoutpumps/things

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