BUG: pd.testing.assert_frame_equal(..., check_exact=True) raises assertion error if any columns are not numeric

[ivirshup]

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of pandas.

• (optional) I have confirmed this bug exists on the master branch of pandas.

---

Note: Please read this guide detailing how to provide the necessary information for us to reproduce your bug.

Code Sample, a copy-pastable example

import pandas as pd
a = pd.DataFrame({"a": list("bc"), "b": [1, 2]})
pd.testing.assert_frame_equal(a, a.copy(), check_exact=True)

   ...: a = pd.DataFrame({"a": list("bc"), "b": [1, 2]}) 
   ...: pd.testing.assert_frame_equal(a, a.copy(), check_exact=True)                                                               
---------------------------------------------------------------------------
AssertionError                            Traceback (most recent call last)
<ipython-input-2-bac78c400991> in <module>
      1 import pandas as pd
      2 a = pd.DataFrame({"a": list("bc"), "b": [1, 2]})
----> 3 pd.testing.assert_frame_equal(a, a.copy(), check_exact=True)

    [... skipping hidden 2 frame]

AssertionError: check_exact may only be used with numeric Series

Problem description

Previously, the example above would not have thrown an error. It makes it a less useful function if I have to start subdividing dataframes by types just to check that they are equal. Presumably check_exact should either have no effect if called on a non-numeric Series or assert_frame_equal should not pass check_exact=True if the series to be compared isn't numeric.

It looks like this was introduced recently, in 08c6597

Expected Output

I expected it to work like it did previously, and not throw an error. Here's an example of test failures this is causing.

Output of pd.show_versions()

INSTALLED VERSIONS
------------------
commit           : d9fff2792bf16178d4e450fe7384244e50635733
python           : 3.8.5.final.0
python-bits      : 64
OS               : Darwin
OS-release       : 19.5.0
Version          : Darwin Kernel Version 19.5.0: Tue May 26 20:41:44 PDT 2020; root:xnu-6153.121.2~2/RELEASE_X86_64
machine          : x86_64
processor        : i386
byteorder        : little
LC_ALL           : None
LANG             : en_US.UTF-8
LOCALE           : en_US.UTF-8

pandas           : 1.1.0
numpy            : 1.19.0
pytz             : 2020.1
dateutil         : 2.8.1
pip              : 20.1.1
setuptools       : 49.2.0
Cython           : None
pytest           : 5.4.3
hypothesis       : None
sphinx           : 3.1.2
blosc            : None
feather          : None
xlsxwriter       : None
lxml.etree       : None
html5lib         : None
pymysql          : None
psycopg2         : None
jinja2           : 2.11.2
IPython          : 7.16.1
pandas_datareader: None
bs4              : None
bottleneck       : None
fsspec           : 0.7.4
fastparquet      : None
gcsfs            : None
matplotlib       : 3.3.0
numexpr          : 2.7.1
odfpy            : None
openpyxl         : None
pandas_gbq       : None
pyarrow          : 0.17.1
pytables         : None
pyxlsb           : None
s3fs             : None
scipy            : 1.5.1
sqlalchemy       : 1.3.18
tables           : 3.6.1
tabulate         : None
xarray           : 0.16.0
xlrd             : 1.2.0
xlwt             : None
numba            : 0.50.1

[simonjayhawkins]

@ivirshup Thanks for the report.

It looks like this was introduced recently, in 08c6597

#32513 cc @jbrockmendel

[TomAugspurger]

Agreed that check_exact should just have no effect for non-numeric dtypes.

@ivirshup are you interested in submitting a PR to update it?

[ivirshup]

Sure! I'm having some trouble finding where to put tests though. Turns out assert_frame_equal occurs a lot in the test suite!

Could you point me to that?

[simonjayhawkins]

\pandas\tests\util\test_assert_frame_equal.py

[cloud-rocket]

I am comparing 2 data frames indexed by timestamp and getting the following error (works perfectly well in pandas 1.0.5):

E       AssertionError: (<Second>, None)

Can it be related?

[ivirshup]

It looked to me like the assert_equal methods got a refactor in 1.1. I don't think it's the same bug, but just got introduced around the same time.

[devProdigy]

btw, if assert_frame_equal gives you smth like AssertionError: (<10 * Minutes>, None) you can:

1. set up frequency of index (if you use timeseries aggregated datetime as your index). I lost it after reading expected dataframe from the .csv. For example:

df = pandas.read_csv(file_path, header=0, index_col=0)
df.index = pd.to_datetime(df.index)  # setup datetime index to match dtypes
df.index.freq = "10T"  # make same frequency for df matching

2. or pass check_like=True to the assert_frame_equal function. But I'm not sure that this is the best idea when you're strictly comparing dataframes :)

[kevinsqi]

@devProdigy's comment is also the problem I was running into (though not strictly related to the original issue). Looks like check_freq=False (available in 1.1.0) may also do the trick.