Thus, if you do pip install without scipy being installed, runtime error.
Do we need the dependency on scipy? Maybe when the scipy API for sparse matrices stabilizes sufficiently, we can point users to a sparse matrix type that quacks so much like an np.ndarray that we don't need to call issparse. This is my dream for the future.
The pyproject.toml file says dependencies = ["numpy==1.24.0", "pandas>=1.5.2", "ray>=2.2.0"]. If we add "scipy==1.10.0" to this list, would that fix the issue ?
Thus, if you do
pip install
withoutscipy
being installed, runtime error.Do we need the dependency on scipy? Maybe when the scipy API for sparse matrices stabilizes sufficiently, we can point users to a sparse matrix type that quacks so much like an np.ndarray that we don't need to call
issparse
. This is my dream for the future.