title | description | author | ms.author | ms.date |
---|---|---|---|---|
Best practices for writing unit tests |
Learn best practices for writing unit tests that drive code quality and resilience |
jpreese |
wiwagn |
07/28/2018 |
This document has been added to the Microsoft docs website https://docs.microsoft.com/en-us/dotnet/core/testing/unit-testing-best-practices
Please use the microsoft docs repository for any additional pull requests.
By John Reese with special thanks to Roy Osherove
There are numerous benefits to writing unit tests; they help with regression, provide documentation, and facilitate good design. However, hard to read and brittle unit tests can wreak havoc on your code base.
In this guide, you'll learn some best practices when writing unit tests to keep your tests resilient and easy to understand.
Functional tests are expensive. They typically involve opening up the application and performing a series of steps that you (or someone else), must follow in order to validate the expected behavior. These steps may not always be known to the tester, which means they will have to reach out to someone more knowledgeable in the area in order to carry out the test. Testing itself could take seconds for trivial changes, or minutes for larger changes. Lastly, this process must be repeated for every change that you make in the system.
Unit tests, on the other hand, take milliseconds, can be run at the press of a button and do not necessarily require any knowledge of the system at large. Whether or not the test passes or fails is up to the test runner, not the individual.
Regression defects are defects that are introduced when a change is made to the application. It is common for testers to not only test their new feature but also features that existed beforehand in order to verify that previously implemented features still function as expected.
With unit testing, it's possible to rerun your entire suite of tests after every build or even after you change a line of code. Giving you confidence that your new code does not break existing functionality.
It may not always be obvious what a particular method does or how it behaves given a certain input. You may ask yourself: How does this method behave if I pass it a blank string? Null?
When you have a suite of well-named unit tests, each test should be able to clearly explain the expected output for a given input. In addition, it should be able to verify that it actually works.
When code is tightly coupled, it can be difficult to unit test. Without creating unit tests for the code that you're writing, coupling may be less apparent.
Writing tests for your code will naturally decouple your code, because it would be more difficult to test otherwise.
- Fast. It is not uncommon for mature projects to have thousands of unit tests. Unit tests should take very little time to run. Milliseconds.
- Isolated. Unit tests are standalone, can be run in isolation, and have no dependencies on any outside factors such as a file system or database.
- Repeatable. Running a unit test should be consistent with its results, that is, it always returns the same result if you do not change anything in between runs.
- Self-Checking. The test should be able to automatically detect if it passed or failed without any human interaction.
- Timely. A unit test should not take a disproportionally long time to write compared to the code being tested. If you find testing the code taking a large amount of time compared to writing the code, consider a design that is more testable.
The term mock is unfortunately very misused when talking about testing. The following defines the most common types of fakes when writing unit tests:
Fake - A fake is a generic term which can be used to describe either a stub or a mock object. Whether it is a stub or a mock depends on the context in which it's used. So in other words, a fake can be a stub or a mock.
Mock - A mock object is a fake object in the system that decides whether or not a unit test has passed or failed. A mock starts out as a Fake until it is asserted against.
Stub - A stub is a controllable replacement for an existing dependency (or collaborator) in the system. By using a stub, you can test your code without dealing with the dependency directly. By default, a fake starts out as a stub.
Consider the following code snippet:
var mockOrder = new MockOrder();
var purchase = new Purchase(mockOrder);
purchase.ValidateOrders();
Assert.True(purchase.CanBeShipped);
This would be an example of stub being referred to as a mock. In this case, it is a stub. We're just passing in the Order as a means to be able to instantiate Purchase
(the system under test). The name MockOrder
is also very misleading because again, the order is not a mock.
A better approach would be
var stubOrder = new FakeOrder();
var purchase = new Purchase(stubOrder);
purchase.ValidateOrders();
Assert.True(purchase.CanBeShipped);
By renaming the class to FakeOrder
, we've made the class a lot more generic, the class can be used as a mock or a stub. Whichever is better for the test case. In the above example, FakeOrder
is used as a stub. We're not using the FakeOrder
in any shape or form during the assert. We just passed it into the Purchase
class to satisfy the requirements of the constructor.
To use it as a Mock, we could do something like this
var mockOrder = new FakeOrder();
var purchase = new Purchase(mockOrder);
purchase.ValidateOrders();
Assert.True(mockOrder.Validated);
In this case, we are checking a property on the Fake (asserting against it), so in the above code snippet, the mockOrder
is a Mock.
[!IMPORTANT] It's important to get this terminology correct. If you call your stubs "mocks", other developers are going to make false assumptions about your intent.
The main thing to remember about mocks versus stubs is that mocks are just like stubs, but you assert against the mock object, whereas you do not assert against a stub.
The name of your test should consist of three parts:
- The name of the method being tested.
- The scenario under which it's being tested.
- The expected behavior when the scenario is invoked.
- Naming standards are important because they explicitly express the intent of the test.
Tests are more than just making sure our code works, they also provide us with documentation. Just by looking at the suite of unit tests, we should be able to infer the behavior of our code without even looking at the code itself. Additionally, when tests fail, we can see exactly which scenarios do not meet our expectations.
[!code-csharpBeforeNaming]
[!code-csharpAfterNamingAndMinimallyPassing]
Arrange, Act, Assert is a common pattern when unit testing. As the name implies, it consists of three main actions:
- Arrange your objects, creating and setting them up as necessary.
- Act on an object.
- Assert that something is as expected.
- Clearly separates what is being tested from the arrange and assert steps.
- Less chance to intermix assertions with "Act" code.
Readability is one of the most important aspects when writing a test. Separating each of these actions within the test clearly highlight the dependencies required to call our code, how our code is being called, and what we are trying to assert. While it may be possible to combine some steps and reduce the size of our test, our primary goal is to make the test as readable as possible.
[!code-csharpBeforeArranging]
[!code-csharpAfterArranging]
The input to be used in a unit test should be the simplest possible in order to verify the behavior that you are currently testing.
- Tests become more resilient to future changes in the codebase.
- Closer to testing behavior over implementation.
Tests that include more information than required to pass the test have a higher chance of introducing errors into the test and can make the intent of the test less clear. When writing tests we want to focus on the behavior. Setting extra properties on models or using non-zero values when not required, only detracts from what you are trying to prove.
[!code-csharpBeforeMinimallyPassing]
[!code-csharpAfterNamingAndMinimallyPassing]
Naming variables in unit tests is as important, if not more important, than naming variables in production code. Unit tests should not contain magic strings.
- Prevents the need for the reader of the test to inspect the production code in order to figure out what makes the value special.
- Explicitly shows what you're trying to prove rather than trying to accomplish.
Magic strings can cause confusion to the reader of your tests. If a string looks out of the ordinary, they may wonder why a certain value was chosen for a parameter or return value. This may lead them to take a closer look at the implementation details, rather than focus on the test.
[!TIP] When writing tests, we should aim to express as much intent as possible. In the case of magic strings, a good approach is to assign these values to constants.
[!code-csharpBeforeMagicString]
[!code-csharpAfterMagicString]
When writing your unit tests avoid manual string concatenation and logical conditions such as if
, while
, for
, switch
, etc.
- Less chance to introduce a bug inside of your tests.
- Focus on the end result, rather than implementation details.
When we add logic into our test suite, we dramatically increase the chance of introducing a bug into it. The last place that you want to find a bug is within your test suite. We should have a high level of confidence that our tests work, otherwise, we aren't going to trust them. Tests that we do not trust, do not provide any value. When a test fails, we want to have a sense that something is actually wrong with our code and it cannot be ignored.
[!TIP] If logic in your test seems unavoidable, consider splitting the test up into two or more different tests.
[!code-csharpLogicInTests]
[!code-csharpAfterTestLogic]
If you require a similar object or state for your tests, prefer a helper method than leveraging Setup and Teardown attributes if they exist.
- Less confusion when reading the tests since all of the code is visible from within each test.
- Less chance of setting up too much or too little for the given test.
- Less chance of sharing state between tests which creates unwanted dependencies between them.
In unit testing frameworks, Setup
is called before each and every unit test within your test suite. While some may see this as a useful tool, it generally ends up leading to bloated and hard to read tests. Each test will generally have different requirements in order to get the test up and running. Unfortunately, Setup
forces you to use the exact same requirements for each test.
[!NOTE] xUnit has removed both SetUp and TearDown as of version 2.x
[!code-csharpBeforeSetup]
// more tests...
[!code-csharpBeforeHelperMethod]
[!code-csharpAfterHelperMethod]
// more tests...
[!code-csharpAfterSetup]
When writing your tests, try to only include one Assert per test. Common approaches to using only one assert include:
- Create a separate test for each assert.
- Use parameterized tests.
- If one Assert fails, the subsequent Asserts will not be evaluated.
- Ensures you are not asserting multiple cases in your tests.
- Gives you the entire picture as to why your tests are failing.
When introducing multiple asserts into a test case, it is not guaranteed that all of the asserts will be executed. In most unit testing frameworks, once an assertion fails in a unit test, the proceeding tests are automatically considered to be failing. This can be confusing as functionality that is actually working, will be shown as failing.
[!NOTE] A common exception to this rule is when asserting against an object. In this case, it is generally acceptable to have multiple asserts against each property to ensure the object is in the state that we expect it to be in.
[!code-csharpBeforeMultipleAsserts]
[!code-csharpAfterMultipleAsserts]
In most cases, there should not be a need to test a private method. Private methods are an implementation detail. You can think of it this way: private methods never exist in isolation. At some point, there is going to be a public facing method that calls the private method as part of its implementation. What you should care about is the end result of the public method that calls into the private one.
Consider the following case
public string ParseLogLine(string input)
{
var sanitizedInput = trimInput(input);
return sanitizedInput;
}
private string trimInput(string input)
{
return input.Trim();
}
Your first reaction may be to start writing a test for trimInput
because you want to make sure that the method is working as expected. However, it is entirely possible that ParseLogLine
manipulates sanitizedInput
in such a way that we do not expect, rendering a test against trimInput
useless.
The real test should be done against the public facing method ParseLogLine
because that is what we ultimately care about.
public void ParseLogLine_ByDefault_ReturnsTrimmedResult()
{
var parser = new Parser();
var result = parser.ParseLogLine(" a ");
Assert.Equals("a", result);
}
With this viewpoint, if you see a private method, find the public method and write your tests against that method. Just because a private method returns the expected result, does not mean the system that eventually calls the private method uses the result correctly.
One of the principles of a unit test is that it must have full control of the system under test. This can be problematic when production code includes calls to static references (e.g. DateTime.Now
). Consider the following code
public int GetDiscountedPrice(int price)
{
if(DateTime.Now == DayOfWeek.Tuesday)
{
return price / 2;
}
else
{
return price;
}
}
How can this code possibly be unit tested? You may try an approach such as
public void GetDiscountedPrice_OnTuesday_ReturnsHalfPrice()
{
var priceCalculator = new PriceCalculator();
var actual = priceCalculator.GetDiscountedPrice(2);
Assert.Equals(1, actual);
}
public void GetDiscountedPrice_ByDefault_ReturnsFullPrice()
{
var priceCalculator = new PriceCalculator();
var actual = priceCalculator.GetDiscountedPrice(2);
Assert.Equals(2, actual)
}
Unfortunately, you will quickly realize that there are a few problems with your tests.
- If the test suite is ran on a Tuesday, the first test will pass, and the second test will fail.
- If the test suite is ran on any other day, the first test will fail, and the second test will pass.
- How is it even possible to test a specific day of the week..?
To solve this problem, you'll need to introduce a seam into your production code. One approach to solve this is to wrap the code that you need to control in an interface and have the production code depend on that interface.
public interface IDateTimeProvider
{
DayOfWeek DayOfWeek();
}
public bool GetDiscountedPrice(int price, IDateTimeProvider dateTimeProvider)
{
if(DateTime.Now == DayOfWeek.Tuesday)
{
return price / 2;
}
else
{
return price;
}
}
Your test suite now becomes
public void GetDiscountedPrice_ByDefault_ReturnsFullPrice()
{
var priceCalculator = new PriceCalculator();
var dateTimeProviderStub = new Mock<IDateTimeProvider>();
dateTimeProviderStub.Setup(dtp => dtp.DayOfWeek()).Returns(DayOfWeek.Monday);
var actual = priceCalculator.GetDiscountedPrice(2, dateTimeProviderStub);
Assert.Equals(2, actual);
}
public void GetDiscountedPrice_OnTuesday_ReturnsHalfPrice()
{
var priceCalculator = new PriceCalculator();
var dateTimeProviderStub = new Mock<IDateTimeProvider>();
dateTimeProviderStub.Setup(dtp => dtp.DayOfWeek()).Returns(DayOfWeek.Tuesday);
var actual = priceCalculator.GetDiscountedPrice(2, dateTimeProviderStub);
Assert.Equals(1, actual);
}
Now the test suite has full control over DateTime.Now
and can stub any value when calling into the method.