Introduction
Today, WinForms apps mainly belong to legacy code because of increasing popularity of WPF. And when one team decides about the development stack for the brand new desktop application, they mainly vote for WPF. On the other hand, there are demands for WinForms applications when it is needed to upgrade existing software which is tightly coupled with WinForms, demands for higher performance, etc.
Different approaches
There are many opinions on the subject about testable WinForms apps. Some claim that it's not possible to test WinForms apps because there is a lot of dependency between user events and business logic.
Standard windows form contains Designer.cs partial class and a partial class which contains event handlers for user actions. So, if we follow this same principle and try to implement app logic in event handlers, we can conclude that it would be very hard to write tests for this kind of application. But, if we look at this problem from some distance we can conclude that every application with user interface could be represented as an interaction between 3 main components: Data, User Interface and Business Logic.
This is the basic idea of MV* patterns. If we succeed to segregate these three components, then our application is on a good way to be testable. I made a simple WinForms app using MVP pattern. The entire project could be downloaded from here.
In this article, I would just use code snippets to give the general idea how the code looks like.
But first, few words about MVP pattern
MVP stands for Model-View-Presenter. Model is a component that contains data. It is just data holder for our forms. View represents the User Interface. It contains design description of our form.
Presenter is a component that does most of the job in our WinForms app. It is subscribed to view events and those events are results of user interaction with our form (clicks on buttons, text change, selection change, etc.) and OS interaction with our form (load, show, paint etc.) Presenter needs to handle all these events and, after their processing, make appropriate action on the view.
Code structure
Let start with our view component.
Now, let take a look at our presenter component.
One test case example
This test case shows how we can easily mock external dependencies in Presenter component.
Conclusion
This is just a simple example how powerful an architecture that starts with an abstraction can be. It is always better to start with some components on the higher level and then, defining the interfaces, you define rules how those components will interact with each other.
The entire project with much more tests can be found here.
Any comments and suggestions are welcome.