[vtkusers] 3D point cloud nearest neighbor search

[G. Allegri]

Hello. I'm new to VTK and this list, so forgive my trivial questions.

I'm considering to use VTK to solve a conceptually simple problem, but
that require efficient 3D algorithms and data structures. I have a csv
file which represents a 3D grid of points (SET1). In this specific
case their regularly spaced on the three axes, but the general case
can be an irregular point cloud. The dataset is made of a couple of
millions of points. I also have another set of 3D points (SET2). I
need to set attributes in SET2 "sampling" the NN from SET1.

I've chosen VTK having seen the various kd-tree and locator classes,
but I wonder what is the most efficient way to achieve my goal, and
what data structured is best for this. I'm considering vtkPolyData or
vtkUnstructuredGrid. What would you suggest for my datas and problem?
Are there efficient classes to read csv input streams to fill my data
structures?

Thanks a lot,
Giovanni