Zero-sized unlimited dimension when read from a grouped netCDF file

[davidhassell]

The file cm4twc_dump_file.nc contains subgroups and has an unlimited dimension, currently of size 13. However, when read it gives the unlimited dimension as size 0:

>>> cfdm.read('cm4twc_dump_file.nc')[0]
<Field: transfer_i(time(0), altitude(1), latitude(4), longitude(3)) 1>

And downstream errors occur, e.g. when trying to subspace the size zero dimension.

This is a bug in cfdm.read (well cfdm.read_write.netcdf.NetCDFRead to be exact), which takes the dimension sizes from the flattened version of the file, but the flattened file does not know the unlimited dimension size, because the flattened file contains no arrays.

This is easily fixed by getting the dimension size from the original grouped file instead. PR to follow.

>>> cfdm.environment(paths=False)
Platform: Linux-5.4.0-62-generic-x86_64-with-debian-bullseye-sid
HDF5 library: 1.10.5
netcdf library: 4.6.3
Python: 3.7.0
netCDF4: 1.5.4
numpy: 1.18.4
cfdm.core: 1.8.8.0
cftime: 1.3.0
netcdf_flattener: 1.2.0
cfdm: 1.8.8.0