S1. pip install causationentropy is broken on current NumPy/matplotlib
pyproject.toml declares numpy>=1.20.0, matplotlib>=3.4.0 with no upper bounds, but the code uses two APIs that have been removed:
causationentropy/core/stats.py:75 — np.trapz (removed in NumPy 2.0)causationentropy/core/plotting.py:521 — matplotlib.cm.get_cmap (removed in matplotlib 3.9)
In Env A this gives 44 test failures out of 381, and — more importantly — plot_causal_network, the function used to produce the paper's own Figure 2, raises AttributeError on a fresh install:
FAIL plot_causal_network: AttributeError module '[matplotlib.cm](http://matplotlib.cm/)' has no attribute 'get_cmap'
The CI matrix pins Python 3.8–3.11 and never tests a modern NumPy, which is why this was not caught. Fix: use np.trapezoid/scipy.integrate.trapezoid and matplotlib.colormaps[...], add NumPy 2.x and Python 3.12/3.13 to the CI matrix.
S1.
pip install causationentropyis broken on current NumPy/matplotlibpyproject.tomldeclaresnumpy>=1.20.0,matplotlib>=3.4.0with no upper bounds, but the code uses two APIs that have been removed:causationentropy/core/stats.py:75—np.trapz(removed in NumPy 2.0)causationentropy/core/plotting.py:521—matplotlib.cm.get_cmap(removed in matplotlib 3.9)In Env A this gives 44 test failures out of 381, and — more importantly —
plot_causal_network, the function used to produce the paper's own Figure 2, raisesAttributeErroron a fresh install:The CI matrix pins Python 3.8–3.11 and never tests a modern NumPy, which is why this was not caught. Fix: use
np.trapezoid/scipy.integrate.trapezoidandmatplotlib.colormaps[...], add NumPy 2.x and Python 3.12/3.13 to the CI matrix.