Bulldozer00's Blog

Option A and Option B in the UML figure below show two different ways of presenting the same 3-function interface to “other” code users. Under which conditions would you choose one design over the other?

Because I prefer simplicity over complexity and local dependency over distant dependency, I’d prefer option B over option A if I was reasonably sure that classes A and B wouldn’t be useful in another inheritance tree or useful as leaf classes. Even if I chose wrongly, because all the functionality is encapsulated within one class in option B, it wouldn’t be a huge risk for a user to extract out B or C class sub-functionality from D and customize it for his/her use by placing it in his/her own separate E class. In this case, no other existing clients of class D would be affected, but the trade off is the introduction of duplicate code into the code base. If I chose the option A inheritance tree, this wouldn’t be the case. In option A, if a user was allowed to directly change A and/or B in-situ, then the duplicate code “wart” wouldn’t manifest, but the risk of breaking the existing code of other users of the A, B, and/or C classes would be high compared to the alternative. Do you see any holes in this decision logic?