mpi4py-fft, array storage about mpi4py-fft

[mikaem]

Hi Aurélien,
Thanks for the feedback:-) We have actually been considering this, but for now we have decided to keep mpi4py-fft to do only the MPI and the transforms. Users of mpi4py-fft, like shenfun and spectralDNS, implement the IO. But the TensorProductSpace class of shenfun is very similar to the PFFT class of mpi4py-fft, and the h5py_writer and nc_writer of shenfun could be reused with very little modification. Then again, mpi4py-fft is a more generic tool and could be used by others. So perhaps the right thing to do would be to implement basic IO tools in mpi4py-fft that could be easily overloaded with more advanced methods by users? I'm thinking that TensorProductSpace could be made a subclass of PFFT, and then it would fall more naturally into place. At least some reorganisation seems to be in order:-)

[mikaem]

Hi
I have now added some IO classes under utilities. I added NetCDF4 as well. Please feel free to check it out. Not quite sure if I have chosen the best possible strategy for storing slices so I would appreciate any feedback:-)
Note, I could open and view the NetCDF4 files using Visit, but I had no luck with Paraview. Is that the same for you?