set_auto_mask broken for netCDF4 files in tiegcm reader

[JonathonMSmith]

In the tiegcm_4d.py reader the recent update to avoid warnings for casting large values uses the method "set_auto_mask" for certain variables. However if the data type is netCDF4 then the h5netcdf library is used. h5netcdf.Datset objects do not have a "set_auto_mask" method. I believe this was tested with a netCDF4.Dataset object. According to h5netcdf documentation their datasets are automatically populated with fillvalues in a manner that is similar to netCDF4.Dataset objects.

I am able to avoid this error by adding a check that the object is not netCDF4 before calling set_auto_mask(False).

This occurs on lines 772, 784, 841, and 853 of the file Kamodo/kamodo_ccmc/readers/tiegcm_4D.py.

[darrendezeeuw]

I am using an older version of netCDF4 at the moment, but am likely using the latest h5netcdf. I have a major upgrade coming soon that will use the latest version of netCDF4. Let's see if that takes care of it.
If you have a small test case that shows the problem, that would be helpful. I could test it with my beta updates.

[JonathonMSmith]

Here's the file and the code I have used to produce this error.
Adding the following logic around the "set_auto_mask" calls seems to fix it:

if ncdf_filetype != 'netCDF4':

s002.nc.zip

code example:

from kamodo_ccmc.flythrough import SatelliteFlythrough as sf
import datetime as dt

model = "TIEGCM"
variable_list = ['rho']
file_dir = '/path/to/netCDF4/file.nc'

date = dt.datetime(2022,2,1,8,30)
u_time = date.timestamp()
t_list = [u_time, u_time+1]
alt = [400, 401]
lon = [50, 51]
lat = [20, 21]

results = sf.ModelFlythrough(model, file_dir, variable_list, t_list, lon, lat,
                             alt, coord_sys='GDZ-sph', verbose=False)

[JonathonMSmith]

If those changes are made, a different (interpolation) error will occur. I can't include a complete tiegcm output file here so I had to remove a bunch of points which breaks the interpolation. But you should be able to see the set_auto_mask error with that example. If you want a better test, convert any tiegcm netcdf3 into a netcdf4 and run the same code (with date and path changed of course)