trimesh.path.polygons.projected function error

[tencerva]

Getting the following error with trimesh.path.polygons.projected function used as outlined in issue #898...

Traceback (most recent call last): File "/home/michael/Desktop/meshman09.py", line 45, in print(trimesh.path.polygons.projected(clean_mesh, normal=[1, 0, 0]).area) File "/home/michael/.local/lib/python3.10/site-packages/trimesh/path/polygons.py", line 792, in projected polygons.extend(edges_to_polygons( File "/home/michael/.local/lib/python3.10/site-packages/trimesh/path/polygons.py", line 153, in edges_to_polygons holes = [polygons[i].exterior.coords for i in interior] File "/home/michael/.local/lib/python3.10/site-packages/trimesh/path/polygons.py", line 153, in holes = [polygons[i].exterior.coords for i in interior] AttributeError: 'MultiPolygon' object has no attribute 'exterior'

I am trying to get the projected area of a mesh onto a plane where some parts of the mesh overlap or shield other parts, similarly for aerodynamic purposes. Input mesh passes "is_volume" test with no duplicate faces or unreferenced vertices.

Trimesh 3.15.4

Thanks, Mike

[mikedh]

This might be fixed by updating shapely to their newer 2.0 withpip install --upgrade shapely? It looks like the multipolygon check might only work on newer copies of shapely:

            # if it returned a multipolygon extend into a flat list                                 
            if hasattr(repaired, 'geoms'):
                polygons.extend(repaired.geoms)
            else:
                polygons.append(repaired)

[tencerva]

Thanks Mike,

Pip reported "shapely" already at version 2.0.1 in response to --upgrade command. Reran my code this morning with same error resulting.

[mikedh]

Gotcha, can you share the code snippet and a model to reproduce? I wasn't able to hit this case on toy data (I tried a bunch of boxes and trimesh.creation.random_soup).

[tencerva]

Mike, Was traveling yesterday and unable to respond. I am checking on supplying the model.

Interestingly, I ran the code on another computer this morning, inadvertantly with "shapely" version 1.8.5, using the following code line:

print(trimesh.path.polygons.projected(clean_mesh, normal=[1, 0, 0], origin=[50, 0, 0]).area)

In this case a reasonable numerical result was generated (no fatal error), but with the following warning message

/home/michael/.local/lib/python3.10/site-packages/trimesh/path/polygons.py:137: ShapelyDeprecationWarning: The array interface is deprecated and will no longer work in Shapely 2.0. Convert the '.coords' to a numpy array instead. polygons.append(repair_invalid(Polygon(vertices[dfs])))

[tencerva]

files.zip

[mikedh]

Thanks for the file looks like a good test case! I was playing with it and it seems like the current trimesh algorithm (backface cull into patches, union as 2D polygons) doesn't do a very good job on this curved input from a lot of angles haha. For me trimesh's worst results were when it got hung up on small almost-perpendicular-but-not-quite patches of faces. It did pretty decently on some of the more-perpendicular views.

I'm not sure what the right way to handle these are, I'll keep playing with it. Ideas and PR's welcome!