Support pyfive as h5netcdf backend

[BSchilperoort]

What is your issue?

Recently, a pure Python HDF5 file reader "pyfive" has become available as a backend for h5netcdf, making h5py optional.

However, xarray still needs h5py (and seems to assume it is available if h5netcdf is installed).

xarray uses h5py in only a few places for checking dtypes.
The pyfive equivalent function for this is pyfive.h5t.check_dtype (replacing h5py.check_dtype).

The following function is also used;

h5py.special_dtype(vlen=str)

which does not have a pyfive equivalent, but will always evaluate to;

np.dtype('O', metadata={'vlen': str})

[welcome]

Thanks for opening your first issue here at xarray! Be sure to follow the issue template!
If you have an idea for a solution, we would really welcome a Pull Request with proposed changes.
See the Contributing Guide for more.
It may take us a while to respond here, but we really value your contribution. Contributors like you help make xarray better.
Thank you!

[kmuehlbauer]

@BSchilperoort Thanks for the report!

So this is only relevant for the reading section, as pyfive is just a reader. The mentioned pyfive.h5t.check_dtype is also available as pyfive.check_dtype as a drop in replacement.

A minimal fix would be to remove the h5py import at the top of open_store_variableand add this line:

h5py = var._root._h5py  # add this line
vlen_dtype = h5py.check_dtype(vlen=var.dtype)
if vlen_dtype is str:
    encoding["dtype"] = str

h5netcdf keeps a link to it's open backend module (h5py, h5pyd or pyfive), so we can utilize this. We could also use var._backend, which would return the string representation of the module. But then we would have to import it.

[BSchilperoort]

Thanks for implementing this!