I'd like to use skeletal animations in my 2D game, and since I already use Box2D as the physics engine, I'd like to enable (but not enforce) ragdoll animations as well. I need some help in figuring out how the Animation and Physics entity components relate to each other.
How should data be represented on disk? Should the actual model/ skeleton be different from the animation? Here's what I currently have in mind: each texture has a corresponding Box2D body (forearm.jpg and forearm.body, for example). I also have a separate hero.model file that references the *.body files, and several *.anim files that reference the model and the textures.
The reason I don't think I should reference the textures in the model directly, but reference them in the animation, is so I can have the character blink (for example): its head would be one body, but different sprites can be pasted over it at different keyframes as to make it blink.
Should I actually move the limbs while animating? While the character is still alive (and ragdoll should be off), what should I do with the physical limbs? When I animate, should I just render the sprites with the correct transform, or should I actually move the physical limbs to their current position according to the animation? If yes, how can I get over the fact that Box2D doesn't handle moving bodies directly too well?
Is there any other advice you can give me?How to enable more complex communication? Architecturally speaking, how should I implement communication between the Animation and Physics components so that interaction is kept to a minimum, but to also be able to do create more complex entities, like for example a wheel on a robot being rendered with the rotation of the physical wheel, while the rest of the robot is rendered according to the animation's keyframes.