单元测试模拟 pandas to_csv

Question

我的模块.py

def write_df_to_csv(self, df, modified_fn):
    new_csv = self.path + "/" + modified_fn
    df.to_csv(new_csv, sep=";", encoding='utf-8', index=False)

test_mymodule.py

class TestMyModule(unittest.TestCase):
    def setUp(self):
        args = parse_args(["-f", "test1"])
        self.mm = MyModule(args)
        self.mm.path = "Random/path"

    self.test_df = pd.DataFrame(
                [
                    ["bob", "a"],
                    ["sue", "b"],
                    ["sue", "c"],
                    ["joe", "c"],
                    ["bill", "d"],
                    ["max", "b"],
                ],
                columns=["A", "B"],
            )

def test_write_df_to_csv(self):
    to_csv_mock = mock.MagicMock()
    with mock.patch("project.mymodule.to_csv", to_csv_mock, create=True):
        self.mm.write_df_to_csv(self.test_df, "Stuff.csv")
    to_csv_mock.assert_called_with(self.mm.path + "/" + "Stuff.csv")

当我运行这个测试时，我得到：

FileNotFoundError: [Errno 2] No such file or directory: 'Random/path/Stuff.csv'

我试图在我的方法中模拟to_csv 。 我的其他测试按预期运行，但是我不确定这个测试哪里出错了。 我对 MagicMock 的使用是否正确，还是我忽略了其他东西？

Answer 1

你没有提供一个最小的、可重复的例子，所以我不得不去掉一些东西来完成这项工作。 我想你可以自己填写缺失的部分。

一个问题是mock.patch("project.mymodule.to_csv", ...)尝试在导入路径project.mymodule的模块中模拟名为to_csv的class 。 这只是“有效”，因为您通过了create=True ，但当然 mocking 以前不存在的东西没有效果，因为没有人会调用它。

您可以使用mock.patch("pandas.DataFrame", ...)模拟整个DataFrame class 。 注意：无论您如何（甚至是否）在当前模块中导入pandas ，它都不是pd 。

But then your unit test will be asserting that to_csv was called on any DataFrame , not necessarily the one you passed in. By mocking just the to_csv method on the one DataFrame object that we are passing into write_df_to_csv , the test becomes a bit more comprehensive and也更容易理解。 我们可以使用mock.patch.object来做到这一点。

mock.patch.object返回模拟 function，我们随后可以在其上调用断言。 因为它是一个方法模拟，而不是免费的 function，所以我们不需要在断言中传递self参数。

项目/mymodule.py

def write_df_to_csv(df, file_name):
    df.to_csv(file_name, sep=";", encoding='utf-8', index=False)

项目/test_mymodule.py

import unittest.mock as mock
import unittest

import pandas as pd

import project.mymodule as mm

class TestMyModule(unittest.TestCase):

    def test_write_df_to_csv(self):
        test_df = pd.DataFrame(...)
        with mock.patch.object(test_df, "to_csv") as to_csv_mock:
            mm.write_df_to_csv(test_df, "Stuff.csv")
            to_csv_mock.assert_called_with("Stuff.csv")

if __name__ == '__main__':
    unittest.main()

Output

现在测试以正确的方式失败了，因为 arguments 实际上并不匹配！

$ python -m project.test_mymodule
F
======================================================================
FAIL: test_write_df_to_csv (__main__.TestMyModule)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/tmp/project/test_mymodule.py", line 25, in test_write_df_to_csv
    to_csv_mock.assert_called_with("Stuff.csv")
  File "/usr/lib/python3.8/unittest/mock.py", line 913, in assert_called_with
    raise AssertionError(_error_message()) from cause
AssertionError: expected call not found.
Expected: to_csv('Stuff.csv')
Actual: to_csv('Stuff.csv', sep=';', encoding='utf-8', index=False)

----------------------------------------------------------------------
Ran 1 test in 0.003s

FAILED (failures=1)

Answer 2

您必须以带有正斜杠的 a 开始路径。 这是更正后的代码：

self.mm.path = "/Random/path/"

也改变：

to_csv_mock.assert_called_with(self.mm.path + "Stuff.csv")

编辑：正如@Thomas 所说，我不确定project.mymodule.to_csv是否正确！

Answer 3

另一种方法是

import unittest.mock as mock
import unittest
import pandas as pd

import project.mymodule as mm

class TestMyModule(unittest.TestCase):

    def test_write_df_to_csv(self):
        test_df = pd.DataFrame(...)
        with mock.patch("pandas.DataFrame.to_csv") as to_csv_mock:
            mm.write_df_to_csv(test_df, "Stuff.csv")
            to_csv_mock.assert_called_with("Stuff.csv")

if __name__ == '__main__':
    unittest.main()'''