Today, when I started refactoring tests for my PHP Pinterest Bot library, I’ve noticed one issue. According to coverage, reports I have 80% code coverage. I though that I’m writing a lot good tests. But then I noticed that most of them actually doesn’t test anything. The tests are overmocked. They only test the code grammar, that I have no typos in my code.
For example this silly test, checking that some class delegates a method call to its dependency. In the snippet bellow, take a look at the
Everything looks good, we have mocked a dependency and then assert that a required method has been called.
But wait a minute, what does this test actually check?
In reality, it executes no useful code. We create a mock then we pass it as a dependency and then we assert that this mock has been executed. When in the future we will refactor and rename
getClientInfo method to
getClientData our test will fail. It is stupid to fix tests every time when we refactor because in these tests I have simply recreated the entire application implementation. Now, every time we refactor, we should go and refactor our tests.
Of course it will fail because now
ProvidersContainer is broken, but actually, it is broken because we have only a typo in a method name. In this case, we use PHPUnit only as a spell checker, to be sure that there are no misspells in the property or method names and everything is typed correctly.
Mocking everything and reimplementing everything in your tests actually doesn’t test anything. Don’t waste your time to write such kind of tests and don’t have an illusion of the high code coverage.
At least I’ve deleted more than 50% of my tests. To keep high code coverage reports I made changes in my
phpunit.xml file in
In this section, we can use
exclude tag to remove some files or directories from the coverage report.
This article was inspired by Laravel Podcast Episode 38.