Unit Testing with Service Stubs or Mock Types
The simple fact of life is that ideas overlap and sometimes are attributed to multiple sources. For instance, I grew learning that Marconi invented radio. Turns out that Tesla invented the radio and now holds the patent posthumously.
An overlapping idea that I believe—but don't quote me—came out of Agile software development is the concept of mock types. In the patterns world "mock types" are called "service stubs." Who invented the idea is important to that person, but the concept of whether we call it mock type or service stub is important to us.
For example, I worked on a project recently that interacted with AAMVA—the American Association of Motor Vehicle Administrators. AAMVA is an external resource that was critical to our application. In our environment, AAMVA only provided an e connection. This meant that if multiple developers were working on AAMVA, they either had to coordinate their time on AAMVA, or if AAMVA was unavailable no one could test their AAMVA code.
The basic idea is that if you have critical services in your applications topology, you can and should build mock types for when those services are disrupted for whatever reason. For your purposes, a service can be anything that is external to any consuming code. Web services, third-party providers, the difference between your client layer and business layer. Roughly anything that might be unavailable during development and unit testing can use the mock type strategy.
Conceptualizing Mock Types
For your purposes, simplify the concept of a service to a provider—some code that provides some capability—that is non-trivial or if trivial, supplied by a distributed, remote, or third party. Many times, the first step taken is to code write to the interface of the provider and connect that to some part of your application. Mock types add a level of abstraction, an interface. The interface contract is the capabilities of the provider—functions and properties.
The next step is to implement a class that is the concrete provider at some point. The mock type is represented by a concrete provider that simply satisfies the contract of the interface. The easiest way to implement a mock type is to define a class, realize the interface, and return hard-coded values. Remember that a mock type is a stub that exists to simulate, or mock up, the capabilities of the provider so that you can code or test the consumer. (The consumer being your code that will ultimately use the provider.
|Tip: The key concept for mock type strategies is that interdependent parts of your application come together at different times. If consumer A is dependent on provider B and B isn't ready, a stub for provider B permits A's development to proceed.|
In your consumer, declare an instance of the interface and then initialize an instance of the actual provider you want. If you use a factory class to return the provider instance, you can add additional information—for example, to your App.config file—that determines when to return the mock type and when to return the concrete type. Figure 1 shows a simple class diagram that illustrates the mock type relationships.
Figure 1: Use a stubbed out class that returns simple data or hard-coded data as a place holder for an actual piece of code that may be unavailable at times.
Implementing Mock Types
Some of you may not read UML. Some of you, so UML diagrams and written explanations are usually balanced nicely with code. In the example for this article, there is a service called NorthwindService implemented in WCF. The service returns a customer based on a CustomerID. Underneath the service is LINQ to SQL code that does the database work and the WCF service contract uses a LINQ query to extract just the customer desired and returns that Customer object.
|Tip: For the demo I used SQL Server 2005, the Northwind database, and LINQ to SQL all with Visual Studio 2008, .NET 3.5, and VB9.|
The WCF service implementation is shown in Listing 1. Listing 2 provides the ServiceContract, the DataContext, and the ORM (object relational mapping) mapped table. The ServiceContract is part of WCF. The DataContext is part of LINQ to SQL, and the ORM table-mapped class is also part of LINQ to SQL.
Listing 1: The WCF service implementation based on the contract in Listing 2 requests a list of customers with LINQ to SQL and refines the list based on a LINQ query (shown).
' NOTE: If you change the class name "Service1" here, you must ' also update the reference to "Service1" in Web.config and in ' the associated .svc file. Public Class Service1 Implements IService1 Public Sub New() End Sub Public Function GetCustomer(ByVal CustomerID As String) _ As Customer _ Implements IService1.GetCustomer Dim northwind As Northwind = New Northwind Dim cust = (From customer In northwind.Customers _ Where customer.CustomerID = CustomerID _ Select customer).First() Return cust End Function End Class
Page 1 of 4