You can use a generic class CGeneralAPI with abstract members. If you derive two classes, CDirectXAPI and COpenGLAPI, you can refer with pointers to an object of CGeneralAPI type. Your instruction

SetRenderAPI(OPENGL)

becomes

CGeneralAPI *myAPI=new COpenGLAPI()

In the rest of the program if you call

myAPI->Render()

you will call the method of COpenGLAPI class.
Instead, if you declare

CGeneralAPI *myAPI=new CDirectXAPI()

myAPI->Render() will call the method of CDirectXAPI.
Of course you have not to change your code each time! You can read the choosen API from a file where you put your configuration settings, for example.

I hope this help you.

Fil (il genio)

Thats exactly what I''m doing in my current project...

The most important thing is to devise a standardized way to render geometry (of ANY type). preferably with 2 variants for static and dynamic data.
I chose to use vertex and index buffer objects, that the current renderer object controls behind the scenes.
Also to have a standard way to setup things like blending/shading/camera/lights etc AS OBJECTS. And, importantly, let the renderer it''self manage states. Don''t allow the rest of the app to do this, This leads to absolute disaster.

If you can write things like text renderers, etc, only once in the base class, then your on the right track. Once you find you have to write something again that doesn''t need to be, then you know you have made a mistake in your design. Basically, the only code that ends up in the derived renderer objects should be API calls. No maths, etc, at all. (unless it''s api specific, say, like the projection matrix setup)

<-- smile :-)

Project-X