EDIT: To clarify what exactly my question is: is this a good way to handle animations / animation state in a game engine with the eye on content creation / management? What are the flaws in doing it this way and what would be an alternative way to do it? -- Though my answer got partially answered in the comments, as it seems to be the way to go.
I'm trying to handle animations in a 2D game engine hobby project, without hard-coding them. Hard coding animation states seems like a common but very strange phenomenon, to me.
A little background: I'm working with an entity system where components are bags of data and subsystems act upon them. I chose to use a polling system to update animation states.
With animation states I mean: "walking_left", "running_left", "walking_right", "shooting", ...
My idea to handle animations was to design it as a data driven model. Data could be stored in an xml file, a rdbms, ... And could be loaded at the start of a game / level/ ... This way you can easily edit animations and transitions without having to go change the code everywhere in your game.
As an example I made an xml draft of the data definitions I had in mind.
One very important piece of data would simply be the description of an animation. An animation would have a unique id (a descriptive name). It would hold a reference id to an image (the sprite sheet it uses, because different animations may use different sprite sheets). The frames per second to run the animation on. The "replay" here defines if an animation should be run once or infinitely. Then I defined a list of rectangles as frames.
50
true
0
0
45
45
45
0
45
45
Animation data would be loaded and held in an animation resource pool and referenced by game entities that are using it. It would be treated as a resource like an image, a sound, a texture, ...
The second piece of data to define would be a state machine to handle animation states and transitions. This defines each state a game entity can be in, which states it can transition to and what triggers that state change.
This state machine would differ from entity to entity. Because a bird might have states "walking" and "flying" while a human would only have the state "walking". However it could be shared by different entities because multiple humans will probably have the same states (especially when you define some common NPCs like monsters, etc). Additionally an orc might have the same states as a human. Just to demonstrate that this state definition might be shared but only by a select group of game entities.
These states can be handled by a polling system. Each game tick it grabs the current state of a game entity and checks all triggers. If a condition is met it changes the entity's state to the "goto" state.
The last part I was struggling with was how to bind animation data and animation states to an entity. The most logical approach seemed to me to add a pointer to the state machine data an entity uses and to define for each state in that machine what animation it uses.
Here is an xml example how I would define the animation behavior and graphical representation of some common entities in a game, by addressing animation state and animation data id. Note that both "wizard" and "orc" have the same animation states but a different animation. Also, a different animation could mean a different sprite sheet, or even a different sequence of animations (an animation could be longer or shorter).
When the entity is being created it would add a list of states with state machine data and an animation data reference.
In the future I would use the entity system to build whole entities by defining components in a similar xml format.
--
This is what I have come up with after some research. However I had some trouble getting my head around it, so I was hoping for some feedback. Is there something here what doesn't make sense, or is there a better way to handle these things? I grasped the idea of iterating through frames but I'm having trouble to take it a step further and this is my attempt to do that.
No comments:
Post a Comment