Again I am going to delay providing my optimized group by example using Entity Framework. The reason for the delay is that I want to fully implement the framework behind my example so I can prove out my understanding of the code that I write, and the code it generates when calling the database.
So, today I want to drop on you a conversation I recently heard by another Dot Net developer that I thought was a little bit odd. They stated that they “want to write code that is so simple that an entry level programmer out of college will be able to understand it”. I wasn’t quite sure how to react to that statement.
There is a line for how complex an application needs to be. I wrote a simple console application the other day simply to get the user accounts that were members of an Active Directory Group, and then do the converse for one of the users returned by getting all of the Active Directory Groups in which they were a participant. I wrote everything in the static class visual studio creates for a console application, creating two method and calling them from Main. It took me all of 3 minutes, and met all of my needs.
On the other hand, I’m dealing with enterprise applications that have dependency inversion, mapping, domain objects, shared classes and interfaces, and the list goes on. This application has a lot of twists and turns that an entry level programmer will not necessarily be able to follow right out of the gate. This is especially true when a program resolves dependencies through convention instead of explicit or indirect resolution through interfaces and classes.
My point is, how complex does you code need to be? In my opinion, it needs to be complex enough to be easily modified and extended should the need arise. It needs to be testable, and if your tests are automated, then it probably needs interfaces, and injection.
For my Active Directory application, there was no need to do automated unit testing, etc. Therefore, the most direct way to get things done makes the most sense. For an Enterprise application, you need to make it testable, and thus interface driven, and injectable. I’m sure there is room in between as well.
My thought is that if you write everything you do to the lowest common denominator, your applications could be under engineered, and you will experience a lot of extra pain because you didn’t take advantage of capabilities to make your code writing easier, even if it is harder to understand.
Cheers,
Ben