[Paraview] Best way to make a code's output Paraview-ready

[Frederic Perez]

Hi all,

I am part of a development team for a scientific code, and we would
like to make our code's output readable by Paraview. Unfortunately,
there is no reasonable way we could have our code directly output to
VTK formats for various reasons (performance, memory, etc.).
Currently, the outputs are in HDF5 and have various structures: 1D to
3D vector fields, 1D to 3D scalar fields, and particle lists.

Being new to Paraview and  VTK, I have found three options for making
our results usable:
1) Write .xmf files that maps each of our individual pieces of data,
so that the XDMF reader can import them.
2) Write a plugin that will convert the HDF5 files to VTK files and read.
3) Use Programmable Sources that extract the HDF5 data directly to VTK elements.

Which one would you recommend? What are their qualities, especially in
terms of versatility and stability? In addition, are they particularly
hard to learn?

Cheers
Fred