Choosing Mocking for .NET

About mocking

Statically-typed languages like C# offer developers benefits to build a solid and reliable design for all abstractions and if it compiles, it’s already a half-way to be working.

For other half of this road we need testing, and unlike manual or system testing in a specially prepared environment, developer can check a lot of stuff by writing atomic tests, called UnitTests, that do not require any envoronment or dependencies to execute and taht makes them fast and lightweight.

Imagine you test a class, interacting with a web-service. And you don’t want to set-up this web-service to test your code, or you even cannot do that due to platform, time or security limitations. So you make a stupid version of this service api, that returns predefined data for that.

That is called a Stub

While writing tests against this Stub you add functionality to record, how many calls you received to a particular address, or what parameters were passed,

Your Stub becomes a Mock

And following a golden DRY rool of programmers not to repeat themselves we come to a task of choosing a Mocking Framework, for which this article can be of help.

What not to compare

Obsoletion of RhinoMocks

First I have to say that the only major mocking library you should stay from is… a well known old-school Rhino Mocks. Because it is old in design and abandoned. There was an attempt to resurrect it in 2014, and it seems to have failed. v4 of RhinoMocks is still in alpha!

• RhinoMocks NuGet - look at the number ov versions
• RhinoMocks GitHub – look at the starts and activity

Also This nice article by Jimmy Keen - a guy with very high rep on StackOverflow for topics on UnitTesting states that you should not use RhinoMocks anymore:

Which framework you should stay away from? - RhinoMocks. Why?
1. Its API is old, unpleasant and undocumented. It will make learning (it) harder and your tests will be more difficult to follow.
2. It is no longer actively developed (regardless of recent attempts).
3. It is one-man show (who might get bored and leave at some point).
4. It is in alpha phase (since spring this year).

Mystery of Microsoft Fake

There is also some strange library from Microsoft, called Fake, but for Visual Studio Enterprise only – WTF ?!! I should research it better

Meet competitors

We will be comparing thees three frameworks: FakeItEasy Moq NSubstitute

the argumens for choosing those three were the following

• They are actively developed
• They are on GitHub and have MIT or BSD license
• Ninject has extensions for these frameworks
• Already mentioned above article by Jimmy Keen
• Scores on nugetmusthaves.com

Moq

Open source, BSD license ( that means totally fine) * Source code * Stats: 1064 stars, 276 forks, 16.6 on nugetmusthaves, downloads: ~5M * I’ve used it a lot, it’s totally cool * This MSDN article recommends it for EF6

NSubstitute

• Source code
• Stats: 536 stars, 123 forks, 12.7 on nugetmusthaves, downloads: ~700K

FakeItEasy

• Source code
• Stats: 462 stars, 96 forks, 8.2 on nugetmusthaves, downloads: ~300K

Comparision

Popularity

So clearly Moq is the most popular. And now about functionality

Functionality

Many articles including this answer from creator of FakeItEasy states that the major differences are not in functionality, but in syntax and approach.

I am quite familiar with Moq only and love it. But actually the NSubstitute syntax I kind of like more, because you can mock methods without any lambdas. See the example:

var serverMoq = new Mock<IServerGateway>();

serverMoq.Setup(server => server.HasUpgrades()).Returns(true);

var serverSubstitute = Substitute.For<IServerGateway>();

serverSubstitute.HasUpgrades().Returns(true);

var serverFake = A.Fake<IServerGateway>();

A.CallTo(() => serverFake.HasUpgrades()).Returns(true);

Conclusions