I'm working on building utilities (explicitly not an engine) for a game project I'm working on. One of those utilities is getting C++ and scripting working together so that the scripting language can be used to set properties of C++ classes, and also to react to changes in those properties that come from the game and other general scripting things.

But there are constraints to the design. The scripting component is for logic, not functionality and for custom behaviour modification by cleverly dropping some reactions to events in and not building the base functionality. And I know exactly how the scripting will fit into an entity and entity-manager module; but I'm still grasping at straws with how the scripting will work with components. My thoughts on this are so far:

1)
Entity & Script Module:
Entity has properties & functions
These properties & functions are exported to the script.
This allows me to do everything I need to do in terms of controlling game logic, scripting levels, tweaking properties & iterating on custom mechanics in real time.

I think this is a safe model, because I know that the script will be able to do it's job with minimal programming support. I know that sounds like a pipe dream but it's all to do with what the scope of the script is, which isn't much, it can handle it.

However I can only guarantee that this script can do it's job for this particular project, which is simple in scope, but I can't guarantee that this entire module will fit the requirements of future projects and thus become an inflexible trashware that was short sighted in it's purpose and usage.

2)
Entity, Component & Script Module:
Entity has components.
Entity has script.
The components are set for an entity by the script. So the components an entity uses are data defined. These would be components like weapons, movement, health, AI, physics.
The components communicate amongst themselves but they essentially replace the properties and functions that were exported to the script as before. The components now reporting to and receiving their tips from the script.

I can see the benefit in this, It does make the design constraints of code a little tighter, so better code. I can see that it allows me to do basically exactly the same thing as number 1. How ever by my estimation supporting the mechanics that can be customised will be more c++ code in the game. On the flip side of code conservation, I'm very sure that a reflection parser and builder for an ecs module, that's what I'm calling it now, will be much easier to write and support because of the constraints of components operating with the scripting.

The thing that I grasp, is that a lot of experienced programmers despise the entity model because they've wished for something in the past. I'm asking for the advice, which of these two options represents the least risk to the project, measured against long term benefit.

PS: I'm still on the fence with this, but I am leaning towards #2 because of the future benefit, and because I can call it an E.C.S. module, Full Metal Panic! fan? Anyone?


UPDATE: I'm researching as much of the component object model as possible so that I can learn enough about it to make a decision between it and the entity model. If anybody has some source code detailing the use of a component model, and I mean some component classes, entity classes, and data representation that links them together. I'd really love to see it.