What’s the difference between a good unit test and a bad one? How do you learn how to write good unit tests? It’s far from obvious. Even if you’re a brilliant coder with decades of experience, your existing knowledge and habits won’t automatically lead you to write good unit tests, because it’s a different kind of coding and most people start with unhelpful false assumptions about what unit tests are supposed to achieve.

Most of the unit tests I see are pretty unhelpful. I’m not blaming the developer: Usually, he or she just got told to start unit testing, so they installed NUnit and started churning out [Test] methods. Once they saw red and green lights, they assumed they were doing it correctly. Bad assumption! It’s overwhelmingly easy to write bad unit tests that add very little value to a project while inflating the cost of code changes astronomically. Does that sound agile to you?

Unit testing is not about finding bugs

Now, I’m strongly in favour of unit testing, but only when you understand what role unit tests play within the Test Driven Development (TDD) process, and squash any misconception that unit tests have anything to do with testing for bugs.

In my experience, unit tests are not an effective way to find bugs or detect regressions. Unit tests, by definition, examine each unit of your code separately. But when your application is run for real, all those units have to work together, and the whole is more complex and subtle than the sum of its independently-tested parts. Proving that components X and Y both work independently doesn’t prove that they’re compatible with one another or configured correctly. Also, defects in an individual component may bear no relationship to the symptoms an end user would experience and report. And since you’re designing the preconditions for your unit tests, they won’t ever detect problems triggered by preconditions that you didn’t anticipate (for example, if some unexpected IHttpModule interferes with incoming requests).

So, if you’re trying to find bugs, it’s far more effective to actually run the whole application together as it will run in production, just like you naturally do when testing manually. If you automate this sort of testing in order to detect breakages when they happen in the future, it’s called integration testing and typically uses different techniques and technologies than unit testing. Don’t you want to use the most appropriate tool for each job?

Goal	Strongest technique
Finding bugs (things that don’t work as you want them to)	Manual testing (sometimes also automated integration tests)
Detecting regressions (things that used to work but have unexpectedly stopped working)	Automated integration tests (sometimes also manual testing, though time-consuming)
Designing software components robustly	Unit testing (within the TDD process)

(Note: there’s one exception where unit tests do effectively detect bugs. It’s when you’re refactoring, i.e., restructuring a unit’s code but without meaning to change its behaviour. In this case, unit tests can often tell you if the unit’s behaviour has changed.)

Well then, if unit testing isn’t about finding bugs, what is it about?

I bet you’ve heard the answer a hundred times already, but since the testing misconception stubbornly hangs on in developers’ minds, I’ll repeat the principle. As TDD gurus keep saying, “TDD is a design process, not a testing process”. Let me elaborate: TDD is a robust way of designing software components (“units”) interactively so that their behaviour is specified through unit tests. That’s all!

Good unit tests vs bad ones

TDD helps you to deliver software components that individually behave according to your design. A suite of good unit tests is immensely valuable: it documents your design, and makes it easier to refactor and expand your code while retaining a clear overview of each component’s behaviour. However, a suite of bad unit tests is immensely painful: it doesn’t prove anything clearly, and can severely inhibit your ability to refactor or alter your code in any way.

Where do your tests sit on the following scale?

Unit tests created through the TDD process naturally sit at the extreme left of this scale. They contain a lot of knowledge about the behaviour of a single unit of code. If that unit’s behaviour changes, so must its unit tests, and vice-versa. But they don’t contain any knowledge or assumptions about other parts of your codebase, so **changes to other parts of your codebase don’t make them start failing **(and if yours do, that shows they aren’t true unit tests). Therefore they’re cheap to maintain, and as a development technique, TDD scales up to any size of project.

At the other end of the scale, integration tests contain no knowledge about how your codebase is broken down into units, but instead make statements about how the whole system behaves towards an external user. They’re reasonably cheap to maintain (because no matter how you restructure the internal workings of your system, it needn’t affect an external observer) and they prove a great deal about what features are actually working today.

Anywhere in between, it’s unclear what assumptions you’re making and what you’re trying to prove. Refactoring might break these tests, or it might not, regardless of whether the end-user experience still works. Changing the external services you use (such as upgrading your database) might break these tests, or it might not, regardless of whether the end-user experience still works. Any small change to the internal workings of a single unit might force you to fix hundreds of seemingly unrelated hybrid tests, so they tend to consume a huge amount of maintenance time – sometimes in the region of 10 times longer than you spend maintaining the actual application code. And it’s frustrating because you know that adding more preconditions to make these hybrid tests go green doesn’t truly prove anything.

Tips for writing great unit tests

Enough vague discussion – time for some practical advice. Here’s some guidance for writing unit tests that sit snugly at Sweet Spot A on the preceding scale, and are virtuous in other ways too.

• *Make each test orthogonal (i.e., independent) to all the others
  **Any given behaviour should be specified in one *and only one test. Otherwise if you later change that behaviour, you’ll have to change multiple tests. The corollaries of this rule include: **Don’t make unnecessary assertions** Which specific behaviour are you testing? It’s counterproductive to Assert() anything that’s also asserted by another test: it just increases the frequency of pointless failures without improving unit test coverage one bit. This also applies to unnecessary Verify() calls – if it isn’t the core behaviour under test, then stop making observations about it! Sometimes, TDD folks express this by saying “*have only one logical assertion per test*”. Remember, unit tests are a design specification of how a certain behaviour should work, not a list of observations of *everything *the code happens to do. </li> * **Test only one code unit at a time **Your architecture *must* support testing units (i.e., classes or very small groups of classes) independently, not all chained together. Otherwise, you have lots of overlap between tests, so changes to one unit can cascade outwards and cause failures everywhere. If you can’t do this, then your architecture is limiting your work’s quality – consider using Inversion of Control. * **Mock out all external services and state **Otherwise, behaviour in those external services overlaps multiple tests, and state data means that different unit tests can influence each other’s outcome. You’ve definitely taken a wrong turn if you have to run your tests in a specific order, or if they only work when your database or network connection is active. (By the way, sometimes your architecture might mean your code touches static variables during unit tests. Avoid this if you can, but if you can’t, at least make sure each test resets the relevant statics to a known state before it runs.) * **Avoid unnecessary preconditions **Avoid having common setup code that runs at the beginning of lots of unrelated tests. Otherwise, it’s unclear what assumptions each test relies on, and indicates that you’re not testing just a single unit. An exception: Sometimes I find it useful to have a common setup method shared by a *very* *small* number of unit tests (a handful at the most) but only if *all* those tests require *all* of those preconditions. This is related to the *context-specification* unit testing pattern, but still risks getting unmaintainable if you try to reuse the same setup code for a wide range of tests. (By the way, I wouldn’t count pushing multiple data points through the same test (e.g., using NUnit’s [TestCase] API) as violating this orthogonality rule. The test runner may display multiple failures if something changes, but it’s still only one test method to maintain, so that’s fine.) </li> * **Don’t unit-test configuration settings **By definition, your configuration settings aren’t part of any unit of code (that’s why you extracted the setting out of your unit’s code). Even if you could write a unit test that inspects your configuration, it merely forces you to specify the same configuration in an additional redundant location. Congratulations: it proves that you can copy and paste! Personally I regard the use of things like filters in ASP.NET MVC as being configuration. Filters like [Authorize] or [RequiresSsl] are configuration options baked into the code. By all means write an integration test for the externally-observable behaviour, but it’s meaningless to try unit testing for the filter attribute’s presence in your source code – it just proves that you can copy and paste again. That doesn’t help you to design anything, and it won’t ever detect any defects. </li> * **Name your unit tests clearly and consistently **If you’re testing how ProductController’s Purchase action behaves when stock is zero, then maybe have a test fixture class called PurchasingTests with a unit test called ProductPurchaseAction\_IfStockIsZero\_RendersOutOfStockView(). This name describes the **subject** (ProductController’s Purchase action), the **scenario** (stock is zero), and the **result** (renders “out of stock” view). I don’t know whether there’s an existing name for this naming pattern, though I know others follow it. How about **S/S/R**?  Avoid non-descriptive unit tests names such as Purchase() or OutOfStock(). Maintenance is hard if you don’t know what you’re trying to maintain. </li> </ul> #### Conclusion Without doubt, unit testing *can *significantly increase the quality of your project. Many in our industry claim that any unit tests are better than none, but I disagree: a test suite can be a great asset, or it can be a great burden that contributes little. It depends on the quality of those tests, which seems to be determined by how well its developers have understood the goals and principles of unit testing. By the way, if you want to read up on integration testing (to complement your unit testing skills), check out projects such as Watin, Selenium, and even the ASP.NET MVC integration testing helper library I published recently.