NUnit设置属性是否有代码味道？

Question

I have written a set of NUnit classes, which have Setup and TearDown attributes. Then I read this: http://jamesnewkirk.typepad.com/posts/2007/09/why-you-should-.html . I can understand what the author is saying where you have to scroll up and scroll down when reading the Unit Tests. However, I also see the benefit of Setup and TearDown. For example, in a recent test class I did this:

private Product1 _product1;
private Product2 _product2;
private IList<Product> _products;

[SetUp]
public void Setup()
{
    _product1 = new Product();
    _product2 = new Product();
    _product = new List<Product>();
    _products.Add(_product1);
    _products.Add(_product2);
}

Here _product1 , _product2 and _products is used by every test. Therefore it seems to violate DRY to put them in every method. Should they be put in every test method?

Answer 1

This question is very subjective, but I don't believe it is a code smell. The example in the linked blog post was very arbitrary and did not use the variables setup in every test. If you look at your code and you are not using the variables from the SetUp , then yes, it is probably a code smell.

If your tests are grouped well however, then each test fixture will be testing a group of functionality that often needs the same setup. In this case, the DRY principle wins out in my books. James argued in his post that he needed to look in three methods to see the state of the data, but I would counter that too much setup code in a test method obscures the purpose of the test.

Copying setup code like you have in your example also makes your tests harder to maintain. If your Product class changes in the future and requires additional construction, then you will need to change it in every test, not in one place. Adding that much setup code to each test method would also make your test class very long and hard to scan.