pandas._lib.testing.assert_almost_equal seem to not use approximate equality for Series with complex doubles

[oleksandr-pavlyk]

Code Sample, a copy-pastable example if possible

import numpy as np
import pandas as pd
from pandas.util.testing import assert_almost_equal

e1 = np.array([0.5831076580182805, -0.9083518696779991], 'd') 
e2 = np.array([0.5831076580182805, -0.9083518696779992], 'd') 

e1z = np.asanyarray(e1, np.complex128)
e2z = np.asanyarray(e2, np.complex128)

assert_almost_equal(pd.Series(e1), pd.Series(e2)) # gives True as expected

assert_almost_equal(pd.Series(e1z), pd.Series(e2z)) # unexpectedly fails

# both of equivalent tests on NumPy side pass
np.testing.assert_almost_equal(e1, e2)
np.testing.assert_almost_equal(e1z, e2z)

Problem description

When running tests on MacOSX with Pandas 0.25.1, the test pandas.tests.computation.test_eval.TestMathNumExprPandas::test_result_complex128 fails because of this issue.

Expected Output

It is expected that pandas._lib.testing.assert_almost_equal would use np.testing.assert_almost_equal for all np.floating and np.complexfloating dtypes.

Output of pd.show_versions()

In [4]: pd.show_versions()

INSTALLED VERSIONS
------------------
commit           : None
python           : 3.6.9.final.0
python-bits      : 64
OS               : Darwin
OS-release       : 15.4.0
machine          : x86_64
processor        : i386
byteorder        : little
LC_ALL           : en_US.UTF-8
LANG             : en_US.UTF-8
LOCALE           : en_US.UTF-8

pandas           : 0.25.1
numpy            : 1.17.1
pytz             : 2019.1
dateutil         : 2.8.0
pip              : 19.1.1
setuptools       : 41.0.1.post20190716
Cython           : None
pytest           : 3.8.1
hypothesis       : 3.68.0
sphinx           : None
blosc            : None
feather          : None
xlsxwriter       : None
lxml.etree       : None
html5lib         : None
pymysql          : None
psycopg2         : None
jinja2           : None
IPython          : 6.3.1
pandas_datareader: None
bs4              : None
bottleneck       : None
fastparquet      : None
gcsfs            : None
lxml.etree       : None
matplotlib       : 3.1.1
numexpr          : 2.7.0
odfpy            : None
openpyxl         : None
pandas_gbq       : None
pyarrow          : None
pytables         : None
s3fs             : None
scipy            : None
sqlalchemy       : None
tables           : None
xarray           : None
xlrd             : None
xlwt             : None
xlsxwriter       : None

[oleksandr-pavlyk]

ping?

[TomAugspurger]

Sorry for the delay. Looks like a bug (generally, pandas doesn't handle complex values well).

Are you interested in looking into where assert_almost_equal gets this wrong?

[oleksandr-pavlyk]

Yes, I can look into this.

[hwalinga]

Why would pandas role its own (almost) equality checker, when it would just be able to use the one from numpy?