Use fuse types to address mix of incoming int types across platforms

sandialabs / pyapprox

Flexible and efficient tools for high-dimensional approximation, scientific machine learning and uncertainty quantification.

https://sandialabs.github.io/pyapprox/

MIT License

51 stars 13 forks source link

Closed ConnectedSystems closed 3 years ago

ConnectedSystems commented 3 years ago

Calls to multivariate_hierarchical_barycentric_lagrange_interpolation_pyx() fail due to different types between platforms (long vs long long).

Switch to fused types to address this (confirmed by tests passing on Windows)

ConnectedSystems commented 3 years ago

I thought this addressed all test failures due to buffer dtype mismatch, but turns out I missed some.

Further investigation shows using fused types for this is not resolved so I am going to fall back to np.arrays

See open issue here