[Python] Accessing instance object from mocked method

I am trying to test a function that creates and transforms a dataframe. At the end of the function the dataframe is saved as a csv. Is there a way to "hook" into the function and retrieve the end state of the dataframe? I tried to mock the to_csv method at the very end of the function but I can't figure out how to retrieve the dataframe instance from that call.

I know that I could:

• Read the resulting csv and verify it that way but I'd prefer not to because the runner of these tests has an unpredictable file system.
• Return the dataframe, however, my actual function is used in airflow which doesn't allow dataframes to be passed around

Here's a stripped down runnable example:

foo.py:

import pandas as pd

def bar():
df = pd.DataFrame({'a': [1, 2, 3], 'b': [4, 5, 6]})
# assume df is transformed in some way
df.to_csv('test.csv')


test.py:

import unittest
from unittest.mock import patch, call
from foo import bar

class TestBar(unittest.TestCase):

@patch("pandas.DataFrame.to_csv")
def test_bar(self, mock_to_csv):
bar()

# How do I access df?
self.assertEqual(mock_to_csv.mock_calls, [call('test.csv')])

Continue reading...