Question

I’m coding a bunch of systems right now. They do not derive from a common interface.

Some example systems: MusicSystem, PhysicsSystem, InputSystem, et cetera.

Currently, MusicSystem loads a lot of audio files in its constructor and as a result, there may be some brief lag when the object is first created.

Because of this, should this code loading all the audio files be placed in an Initialize() method instead? This allows the programmer to determine when he wants to load the audio files but then if he forgets to call Initialize() the program will crash.

Because not all systems need an Initialize() method the programmer has to look through every system to see if the class has an Initialize() method and if so, invoke it. This is a bit cumbersome.

Which approach is preferable in terms of general design principles?

Answer 1

Think about other APIs which you have written code against. When was the last time that an API required the programmer to know to call an init method, otherwise crashing at runtime?

As a consumer of your API, it would drive me nuts if I had to know to call an init method after constructing an object. I would recommend an alternative that I have seen and used firsthand: document expensive object instantiation. What’s the point of deferring an expensive initialization if it’s required for the program to not crash?