IndexError: tuple index out of range

[sugizo]

env
google colab

steps

pip install -U pymartini meshio
!wget -c https://raw.githubusercontent.com/kylebarron/pymartini/master/test/data/fuji.png

code

import imageio.v2 as imageio
import meshio
from pymartini import decode_ele, Martini, rescale_positions

path = './fuji.png'
fuji = imageio.imread(path)
terrain = decode_ele(fuji, 'mapbox')
martini = Martini(fuji.shape[0] + 1)
tile = martini.create_tile(terrain)
vertices, triangles = tile.get_mesh(10)
cells = [("triangle", triangles) ]
mesh = meshio.Mesh(vertices, cells)
mesh.write('fuji.stl')

result

IndexError                                Traceback (most recent call last)
[<ipython-input-11-c748460129cb>](https://localhost:8080/#) in <cell line: 13>()
     11 cells = [("triangle", triangles) ]
     12 mesh = meshio.Mesh(vertices, cells)
---> 13 mesh.write('fuji.stl')

1 frames
[/usr/local/lib/python3.10/dist-packages/meshio/_helpers.py](https://localhost:8080/#) in write(filename, mesh, file_format, **kwargs)
    175         value = cell_block.data
    176         if key in num_nodes_per_cell:
--> 177             if value.shape[1] != num_nodes_per_cell[key]:
    178                 raise WriteError(
    179                     f"Unexpected cells array shape {value.shape} for {key} cells. "

IndexError: tuple index out of range

n.b.
the same steps work fine with pydelatin

best regards

[kylebarron]

I've never used meshio myself. It's not clear to me that this error is in my code instead of meshio. In particular, meshio isn't even raising that WriteError, it's an internal error where value.shape or num_nodes_per_cell is indexed incorrectly.

Could you come up with an example outside of meshio where the mesh is generated incorrectly? For example, point to the specific data indices in vertices and triangles that are incorrect

[sugizo]

no problem when save with quantized_mesh_encoder

with open('fuji.terrain', 'wb') as f:
    quantized_mesh_encoder.encode(f, rescaled_vertices, triangles)

problem occurs when executed with meshio
what make no sense is
the same source file picture,
try pydelatin result work fine
with
pymartini found error traceback

best regards

[kylebarron]

try pydelatin result work fine with pymartini found error traceback

That still doesn't definitively indicate the issue is with pymartini. pymartini does create different meshes than pydelatin does, but as indicated by the fact that it works with quantized_mesh_encoder, I have no indication that meshes created by pymartini are wrong

[sugizo]

what kind of software (free or open source)
able to open *.terrain ?

thanks n best regards