[BUG] Test Suite does not pass on either NixOS or Fedora

[savyajha]

Describe the bug pytest fails on 13 tests on both Fedora and NixOS (both on WSL2) on my systems. Here is a summary of the errors:

FAILED tests/test_bayesian_network.py::test_learn_structure_chow_liu - ValueError: setting an array element with a sequence. The requested array has an inhomogeneous shape after 1 dimensions. The detected shape...
FAILED tests/test_bayesian_network.py::test_learn_structure_exact - ValueError: setting an array element with a sequence. The requested array has an inhomogeneous shape after 1 dimensions. The detected shape...
FAILED tests/test_bayesian_network_structure_learning.py::test_categorical_chow_liu - ValueError: setting an array element with a sequence. The requested array has an inhomogeneous shape after 1 dimensions. The detected shape...
FAILED tests/test_bayesian_network_structure_learning.py::test_categorical_chow_liu_weighted - ValueError: setting an array element with a sequence. The requested array has an inhomogeneous shape after 1 dimensions. The detected shape...
FAILED tests/test_bayesian_network_structure_learning.py::test_categorical_chow_liu_large - ValueError: setting an array element with a sequence. The requested array has an inhomogeneous shape after 1 dimensions. The detected shape...
FAILED tests/test_bayesian_network_structure_learning.py::test_categorical_chow_liu_large_pseudocount - ValueError: setting an array element with a sequence. The requested array has an inhomogeneous shape after 1 dimensions. The detected shape...
FAILED tests/test_bayesian_network_structure_learning.py::test_categorical_exact - ValueError: setting an array element with a sequence. The requested array has an inhomogeneous shape after 1 dimensions. The detected shape...
FAILED tests/test_bayesian_network_structure_learning.py::test_categorical_exact_weighted - ValueError: setting an array element with a sequence. The requested array has an inhomogeneous shape after 1 dimensions. The detected shape...
FAILED tests/test_bayesian_network_structure_learning.py::test_categorical_exact_exclude_parents - ValueError: setting an array element with a sequence. The requested array has an inhomogeneous shape after 1 dimensions. The detected shape...
FAILED tests/test_bayesian_network_structure_learning.py::test_categorical_exact_large - ValueError: setting an array element with a sequence. The requested array has an inhomogeneous shape after 1 dimensions. The detected shape...
FAILED tests/test_bayesian_network_structure_learning.py::test_categorical_exact_large_pseudocount - ValueError: setting an array element with a sequence. The requested array has an inhomogeneous shape after 1 dimensions. The detected shape...
FAILED tests/test_bayesian_network_structure_learning.py::test_categorical_learn_structure_chow_liu - ValueError: setting an array element with a sequence. The requested array has an inhomogeneous shape after 1 dimensions. The detected shape...
FAILED tests/test_bayesian_network_structure_learning.py::test_categorical_learn_structure_exact - ValueError: setting an array element with a sequence. The requested array has an inhomogeneous shape after 1 dimensions. The detected shape...

To Reproduce

Within a fresh venv, the following triggers this:

git clone --depth=1 https://github.com/jmschrei/pomegranate
cd pomegranate
pip install pytest
pip install -r requirements.txt
pip install -e .
pytest -m "not sample"

[jmschrei]

I do not have either of these operating systems and so cannot provide support for them. If you would like to submit a PR I'm happy to be conscious of the issues in the future.

[savyajha]

I will try and do that.

[jmschrei]

Thanks!

[jmschrei]

Looks like these errors are not OS related after all. Rather, they are related to upgrading from pytest v7.1.2 to 7.3.1. That's why the GitHub actions are failing as well. I can look into this now and will fix soon. Thanks for bringing this to my attention.

[jmschrei]

This latest version was unhappy compared the Bayesian network structures because it was a tuple/list of tuples where the inner tuples can have different numbers of elements in it. I assume that in the latest version they tried to streamline the test_array_equal function by first casting everything as a numpy array or something which is why the error was raised. I replaced this with a custom function that simply iterates through two tuples/lists and checks to make sure that each inner element is the same.

[savyajha]

Great to see you got to the bottom of this! Sorry I was unable to work on this before you got to it, my schedule has not been allowing me to work on very much right now.