如何模拟和测试 python open 和 pandas to_pickle

Question

我正在尝试测试这个函数，该函数采用一个 Pandas 数据帧行，它用于进行保存到 csv 的 ftp 调用，打开该 csv 文件，对其进行格式化，并将其保存为泡菜。

我想测试以下内容：

1. builtins.open 用 (path_to_raw, 'wb') 调用一次
2. to_pickle 被调用一次 (LOCAL_PKL.format(row.name))

修补 builtins.open 似乎不起作用，因为它被 to_pickle 间接调用，因此测试失败，因为 builtins.open 被调用两次。

测试功能：

def download_file(row):
    path_from = row['source']
    path_to_raw = LOCAL_RAW.format(row.name)

    self.connection = FTP(self.url)
    self.connection.login(self.username, self.password)
    with open(path_to_raw, 'wb') as f:
        self.connection.retrbinary('RETR ' + path_from, f.write)
    self.connection.quit()

    data = pd.read_csv(path_to_raw)
    data.columns = ['a','b','c']
    data.to_pickle(LOCAL_PKL.format(row.name))

单元测试：

import pandas as pd
import unittest.mock as mock
from unittest.mock import patch, mock_open, MagicMock, call
import maintain

@patch('builtins.open', create=True)
@patch('maintain.pd.read_csv')
def test_download_path(self, mock_open, mock_pd_read_csv):

    mock_pd_read_csv.return_value = pd.DataFrame()      

    @mock.create_autospec
    def mock_pd_to_pickle(self, path):
        pass

    with patch.object(pd.DataFrame, 'to_pickle', mock_pd_to_pickle):

        real = maintain.DataFTP()
        real.connection = MagicMock(name='connection')

        row = pd.Series(data=['a','b'], index=['c','d'])
        row.name = 'anything'

        print(mock_open.assert_called_once_with(maintain.LOCAL_RAW.format(row.name), 'wb'))
        print(mock_pd_to_pickle.assert_called_once_with(maintain.LOCAL_PKL.format(row.name)))

所以......这显然是错误的，但我不确定为什么。此测试产生此错误：

AssertionError: Expected 'read_csv' to be called once. Called 0 times.

有没有人有任何建议或知道如何解决这个问题。谢谢！

Answer 1

我终于得到了它的工作：

@patch('builtins.open', new_callable=mock_open)
@patch('maintain.pd.read_csv', return_value=pd.DataFrame())
@patch.object(pd.DataFrame, 'to_pickle')
def test_download_path(self, mock_to_pickle, mock_read_csv, mock_open):

    real = maintain.EODDataFTP()
    real.connection = mock.Mock(name='connection')

    row = pd.Series(data=['','nyse'], index=['source','exchange'])
    row.name = 'anything'

    real.download_file(row)

    mock_open.assert_called_once_with(maintain.LOCAL_RAW.format(row.name), 'wb')
    mock_read_csv.assert_called_once()
    mock_to_pickle.assert_called_once_with(maintain.LOCAL_PKL.format(row.name))