Yes, this should be straightforward.
Can you send two lists of PNs (point-normals pairs) - one list along
each border of your fillet?
Your fillet is technically a deformed tube or bipod, much simpler than a
tripod (or beyond).

Am 01.03.2023 um 18:15 schrieb Sanjeev Prabhakar:

Yes, this should be straightforward.
Can you send two lists of PNs (point-normals pairs) - one list along
each border of your fillet?
Your fillet is technically a deformed tube or bipod, much simpler than a
tripod (or beyond).

I attach the related image of a fillet between a cylinder and a plane.
The two PN-paths which are required as input are visualized above.

Am 01.03.2023 um 18:15 schrieb Sanjeev Prabhakar:

The challenge probably is to get these points and their normals.
Most of the people including me can give 2 solids as input and the
calculations of the the fillet at their intersection needs to be done by
the function in question.

How would you decide which points to be given as input especially when the
2 solids which are intersecting have quite irregular concave shapes.

Maybe I don't completely understand your method of calculations.

Since you have conceptualized this, it might be simpler for you to
visualise and create objects that way.

On Thu, 2 Mar, 2023, 3:48 pm Martin Herdieckerhoff, <
Martin.Herdieckerhoff@mnet-mail.de> wrote:

Hi Sanjeev,

how did you draw the yellow fillet in your example image when you do not
have the related data???
I understood that you have a special solution for this case (which
requires data) but wanted an easier/smoother/more general solution.

You are right, it is a challenge to get these points and normals.
There are many ways to get the intersection line and/or the borders of a
fillet but none of those is easy.
When the objects to be intersected are passed e.g. as children to a
module than that module cannot solve this task because in OpenSCAD there
is no introspection available into the children.
When the objects to be connected over a fillet are available as VNFs
(points and faces of a polyhedron) than it is theoretically possible to
compute the line of intersection from this but I am not aware of a
respective function and I'm afraid it would have to be very slow. Having
a VNF you might alternatively provide the borders for a fillet without
intersecting the two VNFs. It might be sufficient to intersect each VNF
individually with vnf_halfspace(... boundary=true) of BOSL2 for this.
Another bet is to derive the borders from what you know about an object,
e.g. when you extruded it along a path or so.

In any case, when you can provide useful information about the fillet
than you are invited to send those.

-- Martin

Am 02.03.2023 um 16:23 schrieb Sanjeev Prabhakar:

I am writing my code in python (Jupyter notebook) now a days.
Here is the scad, but you will see polyhedron with lots of
calculated points and obviously the model.

I have reduced the number of points just to show the model concept,
otherwise the file size becomes larger.

On Thu, 2 Mar 2023 at 23:23, Martin Herdieckerhoff <
Martin.Herdieckerhoff@mnet-mail.de> wrote:

I think that this appears straight forward only because the situation is
simple.  In the complicated case of even two nontrivial prisms, a huge part
of the work is in finding the intersection paths of the two fillets.
Actually constructing the fillet is straight forward once you've done that,
assuming it doesn't self-intersect. I guess an interesting question is how
robust your approach is when the fillet would tend to self-intersect using
the straight forward approach.

My impression is that the strength of Martin's approach is in creating a
shape entirely, not in making a fillet on an existing shape.  In this case,
the normals don't need to be bound to some other thing.

On Thu, Mar 2, 2023 at 5:19 AM Martin Herdieckerhoff <
Martin.Herdieckerhoff@mnet-mail.de> wrote:

My point here is that you need to find 3 sets of intersection points
between the 2 solids for making fillet.

1st with both the solids at their original dimensions

2nd with 1 solid offset by approximately the amount of fillet radius
required

3rd with 2nd solid offset

Self intersecting segments are definitely an issue which needs solution.

On Fri, 3 Mar, 2023, 6:38 am Adrian Mariano, avm4@cornell.edu wrote:

Why do you need the intersection of the solid at original dimensions?  I
don't think you need that.  Just the other two, at least if you're doing
thing's Martin's way.  You'll then need the two solids with a hole cut in
them to join onto the fillet shape.

On Thu, Mar 2, 2023 at 8:57 PM Sanjeev Prabhakar sprabhakar2006@gmail.com
wrote:

How would you get the center to draw the arc otherwise.
Probably that is the reason normals are drawn by Martin for each vertices,
so as to get the direction on the curve.
In complex surfaces I am not sure you can get the center with that method.
Maybe yes, but not tried

On Fri, 3 Mar, 2023, 7:42 am Adrian Mariano, avm4@cornell.edu wrote:

I don't think that Martin's method "draws an arc" like you are thinking.
We both implemented methods that worked that way, but it seems that
Martin's method somehow creates the surface between the normals.  I am
curious to see how it works in practice, because at the moment there
remains much uncertainty in how it actually does this.  As Martin said, it
operates using just the point-normal set and produces the shapes he
showed in his earlier post.  There is no "base" point set.

On Thu, Mar 2, 2023 at 10:41 PM Sanjeev Prabhakar sprabhakar2006@gmail.com
wrote: