Python wrappers for a templated class derived from vtkMRMLNode

Development
1

Hello,

I’m trying to figure out if I can use C++ templates for a vtkMRMLNode. My current implementation has the following classes (pseudo-code):

/// generic base class so I can have containers of all my nodes
class vtkMRMLMyBaseNode: public vtkMRMLNode
{
public:
   virtual void MyMethod(void) = 0;
};

/// templated class
template <typename _some_type>
class vtkMRMLMyTemplatedNode: public vtkMRMLMyBaseNode
{
public:
  void GetValue(_some_type & result) const;
}

/// my template specialization:
typedef vtkMRMLMyTemplatedNode<double> vtkMRMLMyDoubleNode ;

Note that I skipped the code to support ::New(), hidden constructor and destructor, etc for conciseness. At that point I can create nodes using the specialized types in C++. I can see the nodes in the MRML scene but I run into a small (and expected) issue with the python interpreter. When I retrieve the node, I get the correct type but I can only access the methods from the MyBase class (MyMethod) but not the methods from the templated class (GetValue). I expected this to happen this I never “indicated” that the wrapping process should create wrappers for my specialization (aka typedef). I used SWIG in the past and needed to explicitly call %template. Is there an equivalent in VTK/Slicer to explicitly require wrappers for a template specialization?

2

VTK classes are wrapped with the VTK Python wrapper. This wrapper does not support templates in the public interface of the class (with a few very specific exceptions). You need to move any templated methods to private implementation an you can use VTK objects to get/set values of any type (e.g., vtkVariant for a single value VTK data arrays for efficient, zero-copy transfer of large arrays of any type).