AHM2012-Slicer-Architecture

Modularization Object Specialization

• Slicer is build on top re-usable and tested libraries.

Displayable Managers

• Displayable manager: Specialized logic handling both RenderWindow <-> MRML and RenderWindow <-> Logic interactions.

• Motivation: Have a well-designed mechanism to ...
  • ...represents MRML node within a Renderer/RenderWindow.
  • ... handle mouse/keyboard interaction.
  • ... synchronize widget across different views.

• Usually a displayable manager is configured after its associated node(s) have been added to the scene. Later on, if the given node(s) are modified, the displayable manager will react accordingly.

• Each time a viewer is instantiated, this one asks the factory to provide him with a DisplayableManagerGroup.

For that reason, adding a method with the signature like "GetDisplayableManagerGroup()" doesn't "fit" the API. A method like "GetDisplayableManagerGroup(const char* MRMLViewNodeId)" could be added instead. It's also important to note that the factory doesn't have the ownership of the group, it means than a smart or weak pointer should be used. Finally, I would rather prefer the factory remains a factory and doesn't manage groups.

Indeed, the API is very similar, both the (1)ThreeD and (2)Slice DisplayableManager derive from vtkMRMLAbstractDisplayableManager.

The differences are that:

(1) deals with MRMLViewNode / ThreeDInteractorStyle

(2) deals with MRMLSliceNode / SliceInteractorStyle

And again, the overhead:

- in a term of efficiency should be limited since displayable manager can be registered on demand
- in term of readability should be very low. Indeed, you will have a Annotation base class for ThreeD and other one for Slice DisplayableManager. And it would be very clear to understand which object is responsible to do what ..
- in term of line of code .. is more important ! But considering modern code editor and efficient c++ pre-processor/compiler .. things should go smoothly :)