Massmatrix and Cotmatrix both returning an empty matrix.

[NarinderS]

It seems like the massmatrix and cotmatrix functions are always returning an empty matrix, without any errors or warnings. Here is a simple setup:

import igl
print(igl.__version__)
V,F = igl.triangulated_grid(3,3)
M1 = igl.massmatrix(V,F,igl.MASSMATRIX_TYPE_BARYCENTRIC)
M2 = igl.massmatrix(V,F,igl.MASSMATRIX_TYPE_VORONOI)
L = igl.cotmatrix(V,F)
print(M1)
print(M2)
print(L)

Prints:

2.5.1
<Compressed Sparse Column sparse matrix of dtype 'float64'
	with 0 stored elements and shape (9, 9)>
<Compressed Sparse Column sparse matrix of dtype 'float64'
	with 0 stored elements and shape (9, 9)>
<Compressed Sparse Column sparse matrix of dtype 'float64'
	with 0 stored elements and shape (9, 9)>

I ran this on a bare-bones venv, using Python 3.12.7. Pip freeze shows:

libigl==2.5.1
numpy==2.1.2
scipy==1.14.1

And strangely dropping the Python version to 3.11.10 and using libigl==2.4.1 does not help either.
Am I missing some sort of dependency?

[alecjacobson]

This is a known issue and I'm working on a fix.

It'd be helpful to know if you installed libigl-python-bindings from source or using pip with precompiled binaries.

[NarinderS]

This was pip with precompiled binaries, for both Py3.12.7 + IGL2.5.1 and Py3.11.10 + IGL2.4.1

[alecjacobson]

Yikes. Interesting. Well the short answer is that the issues is a problem with numpy >1.26.4 and pybind11 (a dependency of libigl's bindings procedure). I thought that bindings built with old numpy would still work when used with a newer numpy. You could try downgrading your numpy for now.

[NarinderS]

Downgrading to numpy 1.26.4 seems to have solved the issue, thanks!