Avoid Premature Generalization

archived comments edit

You’ve no doubt heard me rant against premature optimization in the past, but Eric Gunnerson points out another “Premature” action to be avoided, “Premature Generalization”.

His discussion centers around a very specific question of whether to use private properties to access private fields, or just allow access to the field. Note this discussion pertains to fields that are not publicly accessible via property nor direct access.

The place you’ll often see premature generalization is when inexperienced developers start applying Design Patterns everywhere. If you need to instantiate a factory, implement an adapter class and use a bridge to the toilet just to take a dump, then you probably live with a developer with a premature generalization problem.

Like optimization, generalization is good when it is applied judiciously in the right places. With optimization, one should measure measure measure before applying optimizations. With generalization, I typically suggest that a developer must feel the pain first before generalizing. That simply means that the lack of generalization is starting to cause more work than it saves. In my experience, this often boils down to the rule of threes. If you have to implement something a third time, refactor it.

For example, suppose you have an import tool for some system and as far as you know, you’ll only have to support one import client. By all means write an importer specific to that client. Now your boss tells you to implement an importer for another client. Write that one specific to that client. Once again your boss tells you to implement an importer for yet another client. At this point a pattern has been established. Your boss is a liar and you’ll probably need to implement importers for many clients. Now is the time to refactor the code and generalize the concept of importers. Maybe create a plug-in model or an Import Provider.

[Listening to: Cass & Slide / Perception - Sasha - Sasha: Global Underground: Ibiza 2 of 2]

Comments