Custom Assets vs. Single Component Prefabs

Hello everyone

I have recently been studying custom assets based on classes derived from ScriptableObject, saved using AssetDatabase.CreateAsset()/AssetDatabase.SaveAssets(). My motivation to use them was to have the possibility of referencing the same class instance from many components, avoiding data duplication and reducing serialization/deserialization overhead.

Code (csharp):

1. // saved to disk as asset file
2. public class MyClass: ScriptableObject {
3. }
4.  
5. // two components located in some other gameobjects
6. public class MyComponent1 : MonoBehaviour {
7.     public MyClass myObject;
8. }
9.  
10. public class MyComponent2: MonoBehaviour {
11.     public MyClass myObject; // may point to the same instance as MyComponent1.myObject
12. }

But then I realized that in a project one can have both situations: many components sharing the same data, and many other components which have individualized data. In the latter case the workflow of creating/saving a new asset for each component becomes tedious. In that situation I would like to embed the class instance into the component, which normally would be done as follows:

Code (csharp):

1. [System.Serializable]
2. public class MyClass {
3. }
4.  
5. public class MyComponent : MonoBehaviour {
6.     public MyClass myObject;
7. }

These two approaches are mutually exclusive. Then I came up with a new idea, namely deriving MyClass from MonoBehaviour. If I want the 'shared data' behaviour I add a component of MyClass to a prefab and add a reference to the prefab to MyComponent.

Code (csharp):

1. // added as component to a prefab
2. public class MyClass : MonoBehaviour {
3. }
4.  
5. public class MyComponent : MonoBehaviour {
6.     public MyClass myObject;
7. }

If I want the 'individualized data' behaviour I simply use MyClass as standard component in any gameobject which needs this kind of behaviour.

At this point I began to wonder if there was *any* difference between the custom asset and the single component prefab except that the latter allows me to implement the 'individualized data' behaviour while the former does not. So I would like to ask the question here: is there any reason to use custom assets at all when the same functionality can apparently be replicated using single-component prefabs?

 

Yes. A ScriptableObject define a set of serialized data, nothing more, nothing less. They can be created directly from the Project without having to "add a component" on a GameObject. They do not have any behaviour by themselves, no update method. They are also lighter because they don't carry around a transform component.

Component usually define a behaviour, something that happens. They can be added, remove or move around. They have a reason to exist on a GameObject, because they are assumed to exist in a world and perform a task.

Can you use component to do the job of a ScriptableObject? Yes. Should you? Nah.