Prioritizing Classes to Test

Question

I will start my University-thesis in December. My topic will be to raise the test-coverage in a firm, which has a Spring Online-Shop. While I do that I will work out some kind of rules, what good Unit-Tests should be. And with that of course, how to refactor a class, or how to write classes, that are easy testable.

Now I thought about prioritizing the classes, the project currently has, and without Unit-Tests. I thought about doing some kind of riskanalysis, but I don't know if I can apply that on Java-classes.

The thing is, that when I have prioritized the classes, I'd like to demonstrate my rules on maybe the five most important classes.

The thing with riskanalysis is, that I guess that this would be more suitable for use-cases.

Answer 1

It is a very very broad subject. The simple question about which classes should be best covered by tests (not only units) could be sufficient for a nice thesis by itself ;)

There are however some hints I can give to you. First, get some knowledge in the area of tests. These are very good sources:

• Don't focus on unit tests alone. Read about Test Pyramid.
• Unit test is not equal to good unit test. There is a good book by Tomasz Kaczanowski on writing good unit tests.
• Another great book is a classic in this area - Growing Object-Oriented Software, Guided by Tests

Then, knowing how good tests look like, try to anticipate looking at the code you have, which classes need tests the most. A very interesting approach to this question by Adam Tornhill in his "Treat Your Code as a Crime Scene" presentation.

There are plenty other ways to decide on which classes should be tested first and most and they mostly depend on the way your software is constructed. Above should give you a pretty good place to start though.