difference() has different behaviours with tube() than brick() meshes

[jlvila]

Dear,
I had a simple function to carry out a difference() between 2 meshes.

Case 1

difference(brickA, brickB) (red minus blue)

The result is the expected one

Case 2

difference(brickA, tubeB) (red minus "tube")

But here the result is not the expected one!

The code

Context: w10, python 3.9.7, pyMadCad v0.11

The shapes are simply built with tube() and brick() methods.

def difference(self, aKey, bKey):
     self.shapes3d[aKey].mergeclose()
     self.shapes3d[bKey].mergeclose()
     _diff = difference(
         self.shapes3d[aKey],
         self.shapes3d[bKey])
     _color = self.shapes3d[aKey].options["color"]
     _diff.option(color=_color)
     _diff.stripgroups()
     _diff.strippoints()
     _diff.mergeclose()
     _diff.finish()
     self.scene3d.scene.add(_diff, key=aKey)
     self.shapes3d[aKey] = _diff
     npart4 = note_bounds(_diff)
     self.scene3d.scene.add(npart4, f'{key2}_bounds')
     self.scene3d.scene.dequeue()
     self.removeItems(bKey)
     self.refresh3d()

Erratic bug on difference()

Sometime, we have this exception

File "C:\platform\python\Miniconda3\envs\3d\lib\site-packages\madcad\triangulation.py", line 289, in triangulation_outline
  raise TriangulationError("no more feasible triangles (algorithm failure or bad input outline)", [outline.indices[i] for i in hole])
madcad.triangulation.TriangulationError: ('no more feasible triangles (algorithm failure or bad input outline)', [4, 21, 19])
ERROR conda.cli.main_run:execute(33): Subprocess for 'conda run ['python', 'c:/jean-louis/projets/test/0-python/images/3d/test_madCAD.py']' command 
failed.  (See above for error) 

[jimy-byerley]

Hello @jlvila !

There seems to be 2 unrelated issues here:

wrong side suppression issue

I reformated your original text to highlight the code, so it is more readable.
Looking at the problem, the boolean operation difference seems to take the wrong side from brickA (the side of brickA that is inside tubeB) this is often an issue with the outer normal of the other mesh, so your tubeB must have the outer normals inside instead of outside as you expect.
This could be just a wrong sign with the direction of the axis of the tube, or with the extrusion direction.

Can you put your code for generating the cylinder and the bricks ?

TriangulationError issue

Well as it says, the boolean operation fails to retriangulate the surfaces after intersection of one mesh with the other. Here it happens because the intersection doesn't give consistent intersection cuts, for a simple reason: a portion of the small brick is both in- and out-side of the big one.

For my current implementation of the boolean operations, this is an ambiguity that is not supported.
To raise the ambiguity and workaround the issue, you can just make the small brick slightly bigger (just a 1e-4 offset of the ambiguous surface is sufficient). Check #8 for details.

Hope this help !

[jlvila]

Hi Jimy,

Difference

Thanks, for tube unexpected difference, I just change, as you suggested, the direction of the axis, and the result is as expected ;-)
I change vec3(0, 1, 0) to vec3(0, -1, 0) for the _circle.

The code

    def addTube(self, x, y, w, h, color: QColor):
        _radius = w/2.0
        _lenght = h
        _circle = flatsurface(wire(Circle((vec3(0), vec3(0, -1, 0)), _radius)))
        _tube = Tube3D(_circle, [
            vec3(0, 0, 0),
            vec3(0, _lenght, 0),
        ])
        brTranslated = _tube.transform(
            vec3(x+_radius, y, 0))
        brTranslated.option(color=vec3(color.red()/255.0,
                                       color.green()/255.0, color.blue()/255.0))
        _key = self.scene3d.scene.add(brTranslated)
        npart4 = note_bounds(brTranslated)
        self.scene3d.scene.add(npart4, f'{_key}_bounds')
        print(f'Add tube from 3D: {_key}')
        self.refresh3d()
        return brTranslated, _key

TriangulationError issue

I just add a randomized float value to Z to drastically minimize the risk of this TriangulationError.