Unit testing F# libraries in .NET Core using dotnet test and xUnit
This tutorial takes you through an interactive experience building a sample solution step-by-step to learn unit testing concepts. If you prefer to follow the tutorial using a pre-built solution, view or download the sample code before you begin. For download instructions, see Samples and Tutorials.
This article is about testing a .NET Core project. If you're testing an ASP.NET Core project, see Integration tests in ASP.NET Core.
Creating the source project
Open a shell window. Create a directory called unit-testing-with-fsharp to hold the solution.
Inside this new directory, run dotnet new sln to create a new solution. This
makes it easier to manage both the class library and the unit test project.
Inside the solution directory, create a MathService directory. The directory and file structure thus far is shown below:
/unit-testing-with-fsharp
unit-testing-with-fsharp.sln
/MathService
Make MathService the current directory, and run dotnet new classlib -lang "F#" to create the source project. You'll create a failing implementation of the math service:
module MyMath =
let squaresOfOdds xs = raise (System.NotImplementedException("You haven't written a test yet!"))
Change the directory back to the unit-testing-with-fsharp directory. Run dotnet sln add .\MathService\MathService.fsproj to add the class library project to the solution.
Creating the test project
Next, create the MathService.Tests directory. The following outline shows the directory structure:
/unit-testing-with-fsharp
unit-testing-with-fsharp.sln
/MathService
Source Files
MathService.fsproj
/MathService.Tests
Make the MathService.Tests directory the current directory and create a new project using dotnet new xunit -lang "F#". This creates a test project that uses xUnit as the test library. The generated template configures the test runner in the MathServiceTests.fsproj:
<ItemGroup>
<PackageReference Include="Microsoft.NET.Test.Sdk" Version="15.3.0-preview-20170628-02" />
<PackageReference Include="xunit" Version="2.2.0" />
<PackageReference Include="xunit.runner.visualstudio" Version="2.2.0" />
</ItemGroup>
The test project requires other packages to create and run unit tests. dotnet new in the previous step added xUnit and the xUnit runner. Now, add the MathService class library as another dependency to the project. Use the dotnet add reference command:
dotnet add reference ../MathService/MathService.fsproj
You can see the entire file in the samples repository on GitHub.
You have the following final solution layout:
/unit-testing-with-fsharp
unit-testing-with-fsharp.sln
/MathService
Source Files
MathService.fsproj
/MathService.Tests
Test Source Files
MathServiceTests.fsproj
Execute dotnet sln add .\MathService.Tests\MathService.Tests.fsproj in the unit-testing-with-fsharp directory.
Creating the first test
You write one failing test, make it pass, then repeat the process. Open Tests.fs and add the following code:
[<Fact>]
let ``My test`` () =
Assert.True(true)
[<Fact>]
let ``Fail every time`` () = Assert.True(false)
The [<Fact>] attribute denotes a test method that is run by the test runner. From the unit-testing-with-fsharp, execute dotnet test to build the tests and the class library and then run the tests. The xUnit test runner contains the program entry point to run your tests. dotnet test starts the test runner using the unit test project you've created.
These two tests show the most basic passing and failing tests. My test passes, and Fail every time fails. Now, create a test for the squaresOfOdds method. The squaresOfOdds method returns a sequence of the squares of all odd integer values that are part of the input sequence. Rather than trying to write all of those functions at once, you can iteratively create tests that validate the functionality. Making each test pass means creating the necessary functionality for the method.
The simplest test we can write is to call squaresOfOdds with all even numbers, where the result should be an empty sequence of integers. Here's that test:
[<Fact>]
let ``Sequence of Evens returns empty collection`` () =
let expected = Seq.empty<int>
let actual = MyMath.squaresOfOdds [2; 4; 6; 8; 10]
Assert.Equal<Collections.Generic.IEnumerable<int>>(expected, actual)
Your test fails. You haven't created the implementation yet. Make this test pass by writing the simplest code in the MathService class that works:
let squaresOfOdds xs =
Seq.empty<int>
In the unit-testing-with-fsharp directory, run dotnet test again. The dotnet test command runs a build for the MathService project and then for the MathService.Tests project. After building both projects, it runs this single test. It passes.
Completing the requirements
Now that you've made one test pass, it's time to write more. The next simple case works with a sequence whose only odd number is 1. The number 1 is easier because the square of 1 is 1. Here's that next test:
[<Fact>]
let ``Sequences of Ones and Evens returns Ones`` () =
let expected = [1; 1; 1; 1]
let actual = MyMath.squaresOfOdds [2; 1; 4; 1; 6; 1; 8; 1; 10]
Assert.Equal<Collections.Generic.IEnumerable<int>>(expected, actual)
Executing dotnet test runs your tests and shows you that the new test fails. Now, update the squaresOfOdds method to handle this new test. You filter all the even numbers out of the sequence to make this test pass. You can do that by writing a small filter function and using Seq.filter:
let private isOdd x = x % 2 <> 0
let squaresOfOdds xs =
xs
|> Seq.filter isOdd
There's one more step to go: square each of the odd numbers. Start by writing a new test:
[<Fact>]
let ``SquaresOfOdds works`` () =
let expected = [1; 9; 25; 49; 81]
let actual = MyMath.squaresOfOdds [1; 2; 3; 4; 5; 6; 7; 8; 9; 10]
Assert.Equal(expected, actual)
You can fix the test by piping the filtered sequence through a map operation to compute the square of each odd number:
let private square x = x * x
let private isOdd x = x % 2 <> 0
let squaresOfOdds xs =
xs
|> Seq.filter isOdd
|> Seq.map square
You've built a small library and a set of unit tests for that library. You've structured the solution so that adding new packages and tests is part of the normal workflow. You've concentrated most of your time and effort on solving the goals of the application.
