unit tests - how to not rely on users files system (python)

[TomMonks]

This relates to the unit test example where you create a df and then save as CSV only to reload and check the columns.

I think what is included is a good introduction. But it means the unit test is linked to both the user's file system and the pandas.read_csv function. There's a way to run this test without either, but it is a bit more advanced.

Have you used monkey patching before?

There's a nice way to do this in pytest docs: https://docs.pytest.org/en/stable/how-to/monkeypatch.html

An example:

import pandas as pd
import pytest
from waitingtimes.patient_analysis import import_patient_data

def test_import_success(monkeypatch):
    # Create dataframe to use in testing
    expected_cols = [
        "PATIENT_ID",
        "ARRIVAL_DATE", "ARRIVAL_TIME",
        "SERVICE_DATE", "SERVICE_TIME",
    ]
    df_in = pd.DataFrame(
        [["p1", "2024-01-01", "08:00", "2024-01-01", "09:00"]],
        columns=expected_cols,
    )

    # as don't really want to check pandas ability to read a CSV we just return the 
    # dataframe we created in memory. this also removes the need to write out to hard disk
    def mock_read_csv(path):
        return df_in

    # temp replace the read_csv function of pd
    monkeypatch.setattr(pd, "read_csv", mock_read_csv)

    # the path is now irrelevant
    result = import_patient_data("does_not_matter.csv")
    pd.testing.assert_frame_equal(result, df_in)

Another thing you could do with mock_read_csv is check that it has been correctly passed a Path. thinking something like:

def mock_read_csv(path):
        assert isinstance(path, Path)  
        return df_in

This is more complicated, but I think best practice to avoid IO. Maybe a drop down with extra info or an advanced section?