For code to be self-documenting, I like to do any sort of business rule at the lowest level. If I have a business rule that says, "When the requested delivery date is less than 30 days, put a flag so a customer service representative can approve it," I have a matching method which might be called, "AddFlagIfShortLeadTime."
When you compose your system from these small components, you find you don't need state in order to accomplish work. In our example above, AddFlagIfShortLeadTime takes in the model it might operate on and returns a model of the same type with the transformations applied. If any information from a database is needed, we can pass it into this method as well. Let the something else worry about querying the database.
One interesting thing about this is that I rarely see this sort of pattern in a C# codebase. I can see it in other languages which primarily work in a functional paradigm, but I'm not sure why I don't see it in my main language. Below are some of my reasons for doing it. I'd love to hear if anyone has thoughts on why this is or isn't a good idea.
There's still an argument about whether to use dependency injection. I've found that if I introduce a static class, neither side gets mad. A common argument for DI is that it's hard to instantiate classes. With static classes, you don't have to instantiate anything. An argument against DI is that it's hard to work with. With static classes, you don't have to work with DI at all.
Of course, I'm not advocating for making private things public, but if you happen to need to expose your static class, zero properties and no instanciation means no mocking. You can focus on what you're really trying to test - your business rules.
I don't know a lot about this, but I do know that keeping an instanciated object takes processing power. Static classes can be called upon by the compiler at any time without instantiation, and so take up less processing power... Or something like that.
Usually you shouldn't just use static classes for everything. Static classes are great for documenting business rules and terrible for reading or writing to databases. By using static classes by default, we can separate our business logic by default.
Do you disagree with this approach? I have found it odd that I can't find many examples outside of my own code where people do this. Let me know what you think. 😀