使用 setup_class 初始化“冲突"；带参数化选项 [英] using setup_class for initialization "conflicts" with parametrize options

问题描述

我有这个文件:

import numpy as np


class Variables(object):

    def __init__(self, var_name, the_method):

        self.var_name = var_name
        self.the_method = the_method

    def evaluate_v(self):
        var_name, the_method = self.var_name, self.the_method

        if the_method == 'Average':
            return np.average(var_name)

和这个测试文件:

import unittest
import pytest
import numpy as np

from .variables import Variables


class TestVariables():

    @classmethod
    def setup_class(cls):
        var_name = np.array([1, 2, 3])
        the_method = 'Whatever'
        cls.variables = Variables(var_name, the_method)

    @pytest.mark.parametrize(
        "var_name, the_method, expected_output", [
            (np.array([1, 2, 3]), 'Average', 2),
        ])
    def test_evaluate_v_method_returns_correct_results(
        self, var_name, the_method,expected_output):

        var_name, the_method = self.variables.var_name, self.variables.the_method

        obs = self.variables.evaluate_v()  
        assert obs == expected_output

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

现在，我的问题是我在设置类中初始化值(var_name、the_method).

Now, my problem is that I initialize the values (var_name, the_method) in the setup class.

因此，当我在参数化中使用不同的值时，这些值将被忽略.

So, when I use different values in the parametrize, these values are ignored.

因此，现在的测试是通过将 the_method 值作为 Whatever 而不是 Average(我想在不同的参数设置).

So, the test right now is done by taking as the_method value as Whatever and not as Average (which I want to have in the different parametrize setup).

有没有办法在不丢失 setup_class 的情况下处理这个问题?

Is there a way to deal with this without loosing the setup_class?

因为，如果我省略 setup_class，它就可以工作.

Because, if I ommit the setup_class, ok it works.

推荐答案

好吧，您可以将 setup_class 方法的责任限制为使用任何非抛出异常数据创建变量的实例

Well, you can just limit the responsibility of setup_class method to a creation of Variables' instance with any non-throwing-exception data

@classmethod
    def setup_class(cls):            
        cls.variables = Variables(None, None)

并创建测试

    @pytest.mark.parametrize(
        "var_name, the_method, expected_output", [
            (np.array([1, 2, 3]), 'Average', 2),
            (np.array([1, 2, 3]), 'Whatever', 'whatever you want')
        ])
    def test_evaluate_v_method_returns_correct_results(
        self, var_name, the_method, expected_output):

        self.variables.var_name = var_name
        self.variables.the_method = the_method

        assert self.variables.evaluate_v() == expected_output

但就我个人而言，我不认为将 None 传递给构造函数是一个非常干净的解决方案.我建议每次测试都将参数设为可选或创建 Variables 实例.

But personally, I don't think passing Nones to the constructor is a very clean solution. I suggest making the parameters optional or creating Variables` instance every test.

这篇关于使用 setup_class 初始化“冲突"；带参数化选项的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！