RCSB Viewers:Viewer Framework:General Rendering

Notes

  •      Much of this is GL specific, and the namespace reflects that.  Expect these names to change to
         more generic terms.  For example:
         
    • JoglSceneNode -> Scene (node has connotations in scenegraph structures, so would rather either            ignore the term, or tie to a proper corollary)

    • DisplayListGeometry -> ScenePrimitive
    • DisplayListRenderable -> SubScene
    • DisplayLists -> flat collection of all display lists objects.
  •      A further goal would be to factor out the GL specific parts of the code and allow the system to be
         switched to a different rendering engine.
       

Relevent Classes

  • GlGeometryViewer
  • JoglSceneNode
  • DisplayListGeometry*
  • Renderable
  • DisplayListRenderable
  • DisplayLists

Explanation

Rendering is a fairly complex topic, so we can only give the broad brush-strokes, here.

Essentially, all rendering is triggered/controlled by the GlGeometryViewer class and it’s derivations. When called upon to render, it runs through a set of ‘display list’ structures, which contain geometry definitions on a component type by component type basis.

Actual geometry generation is delegated to the <componentType>Geometry classes. If rendering the first time, the class generates the geometry and hands back the created display list. On subsequent renders, just the display list is returned, saving regeneration of geometry.

Analysis

Essentially, this is a good mechanism, however the granularity is too fine. Display lists are only for spheres, cylinders, and the residue pieces created for secondary structures. Again, while the latter is good, the former is too fine-grained. It would be better if higher level display lists were defined on a chain basis, that would in turn invoke the sphere and cylinder display lists.

This adds to the complication in case of editing/modification, of course, because in doing so, the higher level display lists need to be discarded and regenerated. This is probably the biggest argument for going with a scene-graph implementation - a scene-graph facility will automatically handle all of this.