Serialize numpy arrays or store in HDF5 about qmflows HOT 2 CLOSED

I agree because the infrastructure to work with HDF5 is already available. We can take advantage of the HDF5 internal structure to store the arrays using in an structure way like:

path_to_prop = '<project_name>/<job_name>/<property>'

from qmflows.

The basic configuration is now noodles. It should be possible to extend this to include a project name in the hdf5 path. The name of the job will be a bit trickier. Also there's a potential performance issue, since we're computing hashes of potentially large data, and then search the entire file for matches to this hash. One workaround for this last issue is to keep a separate database of the arrays that were serialised. I have some more ideas, but for now this issue is closed I think.

from qmflows.