Pandas 1.5.0 came out on September 19, 2022. Promptly, a bunch of unit tests that used to be passing started failing (at least in python 3.8). To be fair: I think these tests were broken before, but pandas 1.4.4 testing.assert_frames_equal was not catching the broken tests. The offending tests can be found here
Based on what I have seen, we just need to rewrite the expected_values of the unit tests to get them to pass. I honestly do think the older version of pandas.testing.assert_frames_equal was too permissive.
Note: you will not be able to reproduce the error in python 3.7. It only affects python 3.8 (python 3.7 is still stuck on pandas 1.3.5)
danielsf
commented
1 year ago
Pandas 1.5.0 came out on September 19, 2022. Promptly, a bunch of unit tests that used to be passing started failing (at least in python 3.8). To be fair: I think these tests were broken before, but pandas 1.4.4
testing.assert_frames_equalwas not catching the broken tests. The offending tests can be found herehttps://github.com/AllenInstitute/AllenSDK/actions/runs/3085471250/jobs/4988798139
Based on what I have seen, we just need to rewrite the
expected_valuesof the unit tests to get them to pass. I honestly do think the older version ofpandas.testing.assert_frames_equalwas too permissive.Note: you will not be able to reproduce the error in python 3.7. It only affects python 3.8 (python 3.7 is still stuck on pandas 1.3.5)