I have a asp.net core application that uses dependency injection defined in the startup.cs class of the application:
public void ConfigureServices(ISer
Although @Kritner's answer is correct, I prefer the following for code integrity and better DI experience:
[TestClass]
public class MatchRepositoryTests
{
private readonly IMatchRepository matchRepository;
public MatchRepositoryTests()
{
var services = new ServiceCollection();
services.AddTransient<IMatchRepository, MatchRepositoryStub>();
var serviceProvider = services.BuildServiceProvider();
matchRepository = serviceProvider.GetService<IMatchRepository>();
}
}
If you are using the Program.cs
+ Startup.cs
convention and want to get this working quickly you can reuse your existing host builder with a one-liner:
using MyWebProjectNamespace;
public class MyTests
{
readonly IServiceProvider _services =
Program.CreateHostBuilder(new string[] { }).Build().Services; // one liner
[Test]
public void GetMyTest()
{
var myService = _services.GetRequiredService<IMyService>();
Assert.IsNotNull(myService);
}
}
Sample Program.cs
file from web project:
using Microsoft.AspNetCore.Hosting;
using Microsoft.Extensions.Hosting;
namespace MyWebProjectNamespace
{
public class Program
{
public static void Main(string[] args) =>
CreateHostBuilder(args).Build().Run();
public static IHostBuilder CreateHostBuilder(string[] args) =>
Host.CreateDefaultBuilder(args)
.ConfigureWebHostDefaults(webBuilder =>
{
webBuilder.UseStartup<Startup>();
});
}
}
A simple way, I wrote a generic dependency resolver helper class and then built the IWebHost in my unit test class.
Generic Dependency Resolver
public class DependencyResolverHelpercs
{
private readonly IWebHost _webHost;
/// <inheritdoc />
public DependencyResolverHelpercs(IWebHost WebHost) => _webHost = WebHost;
public T GetService<T>()
{
using (var serviceScope = _webHost.Services.CreateScope())
{
var services = serviceScope.ServiceProvider;
try
{
var scopedService = services.GetRequiredService<T>();
return scopedService;
}
catch (Exception e)
{
Console.WriteLine(e);
throw;
}
};
}
}
}
Unit Test Project
[TestFixture]
public class DependencyResolverTests
{
private DependencyResolverHelpercs _serviceProvider;
public DependencyResolverTests()
{
var webHost = WebHost.CreateDefaultBuilder()
.UseStartup<Startup>()
.Build();
_serviceProvider = new DependencyResolverHelpercs(webHost);
}
[Test]
public void Service_Should_Get_Resolved()
{
//Act
var YourService = _serviceProvider.GetService<IYourService>();
//Assert
Assert.IsNotNull(YourService);
}
}
Why would you want to inject those in a test class? You would usually test the MatchController, for example, by using a tool like RhinoMocks to create stubs or mocks. Here's an example using that and MSTest, from which you can extrapolate:
[TestClass]
public class MatchControllerTests
{
private readonly MatchController _sut;
private readonly IMatchService _matchService;
public MatchControllerTests()
{
_matchService = MockRepository.GenerateMock<IMatchService>();
_sut = new ProductController(_matchService);
}
[TestMethod]
public void DoSomething_WithCertainParameters_ShouldDoSomething()
{
_matchService
.Expect(x => x.GetMatches(Arg<string>.Is.Anything))
.Return(new []{new Match()});
_sut.DoSomething();
_matchService.AssertWasCalled(x => x.GetMatches(Arg<string>.Is.Anything);
}
Your controllers in .net core have dependency injection in mind from the start, but this does not mean you are required to use a dependency injection container.
Given a simpler class like:
public class MyController : Controller
{
private readonly IMyInterface _myInterface;
public MyController(IMyInterface myInterface)
{
_myInterface = myInterface;
}
public JsonResult Get()
{
return Json(_myInterface.Get());
}
}
public interface IMyInterface
{
IEnumerable<MyObject> Get();
}
public class MyClass : IMyInterface
{
public IEnumerable<MyObject> Get()
{
// implementation
}
}
So in your app, you're using the dependency injection container in your startup.cs
, which does nothing more than provide a concretion of MyClass
to use when IMyInterface
is encountered. This does not mean it is the only way of getting instances of MyController
however.
In a unit testing scenario, you can (and should) provide your own implementation (or mock/stub/fake) of IMyInterface
as so:
public class MyTestClass : IMyInterface
{
public IEnumerable<MyObject> Get()
{
List<MyObject> list = new List<MyObject>();
// populate list
return list;
}
}
and in your test:
[TestClass]
public class MyControllerTests
{
MyController _systemUnderTest;
IMyInterface _myInterface;
[TestInitialize]
public void Setup()
{
_myInterface = new MyTestClass();
_systemUnderTest = new MyController(_myInterface);
}
}
So for the scope of unit testing MyController
, the actual implementation of IMyInterface
does not matter (and should not matter), only the interface itself matters. We have provided a "fake" implementation of IMyInterface
through MyTestClass
, but you could also do this with a mock like through Moq
or RhinoMocks
.
Bottom line, you do not actually need the dependency injection container to accomplish your tests, only a separate, controllable, implementation/mock/stub/fake of your tested classes dependencies.