Messages from Loft operation

[edwilliams16]

OK. I found a few bugs:

You are now constructing the rotatable polygon along the z-axis with the radial dimension along y, which I think is clearer. So, now

1) we sort the points along the z-axis
2) since z and y are taking the roles of x and y, the _isConvex test gets reversed
I used

Code: Select all

def createRevolveHull(coordlist) :
    points = sorted(coordlist, key = lambda x: x[2]) # sort by z-coord
    top = [points[0], points[1]]
    for p in points[2:]:
        top.append(p)
        while len(top) > 2 and not _isConvex(*top[-3:],2, 1):
            del top[-2]
    #print(top)
    # close polygon
    top.append(top[0])
    poly = Part.makePolygon(top)

    revHull = poly.revolve(FreeCAD.Vector(0,0,0),FreeCAD.Vector(0,0,1),360)
    return revHull

def _isConvex(p, q, r, ind1, ind2):
    'return True if the vectors pq to qr is a right turn ie convex. ind1 is sort axis. ind2 the other'
    return q[ind1]*r[ind2] + p[ind1]*q[ind2] + r[ind1]*p[ind2] - \
            (q[ind1]*p[ind2] + r[ind1]*q[ind2] + p[ind1]*r[ind2]) < 0

which is probably unnecessarily flexible.

We also need to rotate the hull axis onto that of the collinear objects.

Code: Select all

 revHull = createRevolveHull(pointLst)
 revHull.Placement.Rotation = obj1.Placement.Rotation  # rotate from z-axis to collinear axis
 return revHull

It still doesn't quite work because of the recompute() problem. When I run under the debugger, with your FromFinger.csg or my rotated hulltwocones1.csg the code loops back to hull.execute I can't figure out the code to see what is happening, but there doesn't appear to be a test to show that the two objects have already been hulled. Since the algorithm will handle any number of collinear cones/cylinders, you need to change the logic anyway, which tests for two objects to be hulled.

[edwilliams16]

ping

I got my ssh access to github working again. Somehow it had lost my public key.

I put the above mentioned fixes into my parallel_ew branch in https://github.com/edwilliams16/OpenSCAD_Alt_Import
I haven't figured out pull requests yet, so I'm letting you know here before you make more changes I need to merge!
The off-axis collinear cylinders now does the hull correctly, but I believe to recompute loop bug is what is screwing up the final axis rotation.

Screen Shot 2021-04-18 at 10.45.28 AM.png (82.63 KiB) Viewed 350 times

It looks like you can now do multiple collinear cylinders/cones, but I haven't tested it yet.

[keithsloan52]

ping

The off-axis collinear cylinders now does the hull correctly, but I believe to recompute loop bug is what is screwing up the final axis rotation.Screen Shot 2021-04-18 at 10.45.28 AM.png

It looks like you can now do multiple collinear cylinders/cones, but I haven't tested it yet.

I came to that conclusion too and my copy of the code with your corrections is the same. The handling of a PartFeaturePython with dependent group ( Like a Boolean ) came from me making a request for a new facility and realthunder coming back and saying it can be done with current facilities, he produced the code and I already messaged him about the the invocation on import and a second invocation afterwards, he has not logged on since last Tuesday so will not have seen my message on current code. and test file, hope he is okay.

[edwilliams16]

In the meantime, I've developed an algorithm to find the convex hull of an arbitrary number of collinear spheres. As in the two sphere case we revolve the convex hull of a collection of semi-circles. The code is in https://github.com/edwilliams16/Convex_Hull_revolve

Run the plotdiagnostics.py file. It takes a list of circles on the z axis and returns a list of arcs and lines the constitutes the hull to revolve.
Here's a typical output, where the dashed lines are the (semi-) circles and the solid blue, the convex hull.

I'm still thinking about an arbitrary collection of coaxial cylinders, cones and spheres, which I think is the most you can do pursuing this line of attack.

Screen Shot 2021-04-20 at 11.49.16 AM.png (34.59 KiB) Viewed 291 times

[keithsloan52]

In the meantime, I've developed an algorithm to find the convex hull of an arbitrary number of collinear spheres. As in the two sphere case we revolve the convex hull of a collection of semi-circles. The code is in https://github.com/edwilliams16/Convex_Hull_revolve

Run the plotdiagnostics.py file. It takes a list of circles on the z axis and returns a list of arcs and lines the constitutes the hull to revolve.
Here's a typical output, where the dashed lines are the (semi-) circles and the solid blue, the convex hull.

I'm still thinking about an arbitrary collection of coaxial cylinders, cones and spheres, which I think is the most you can do pursuing this line of attack.

Screen Shot 2021-04-20 at 11.49.16 AM.png

Many Thanks - Once again - I will give it some attention.