error in default value with write_sids()

[mbauer288]

In sidecar.py the default fill_value is -1, but the datatype for SIDs is unit64, which can't have negative values. This results is a netcdf write error.

def write_sids(self, sids, nom_res=None, group=None, fill_value=-1):
    i = sids.shape[0]
    j = sids.shape[1]
    varname = 'STARE_index'.format(nom_res=nom_res)
    i_name = 'i'
    j_name = 'j'
    if nom_res:
        varname += '_{nom_res}'.format(nom_res=nom_res)
        i_name += '_{nom_res}'.format(nom_res=nom_res)
        j_name += '_{nom_res}'.format(nom_res=nom_res)   
    with netCDF4.Dataset(self.file_path, 'a', format="NETCDF4") as rootgrp:
        if group:
            grp = rootgrp.createGroup(group)
        else:
            grp = rootgrp
        sids_netcdf = grp.createVariable(varname=varname,
                                         datatype='u8', 
                                         dimensions=(i_name, j_name),
                                         chunksizes=[i, j],
                                         shuffle=self.shuffle,
                                         zlib=self.zlib,
                                         fill_value=fill_value)
        print("HERE")
        sids_netcdf.long_name = 'SpatioTemporal Adaptive Resolution Encoding (STARE) index'
        sids_netcdf[:, :] = sids

[NiklasPhabian]

Ugh. This should warrant some discussion and we should conclude on a solid convention that is reflected in the documentation.

The first two bits of the SID are not used. We therefore could make FFFF FFFF FFFF FFFF the nodata value.
But since the first two bits are not used anyway, we do not necessarily need to use unit64. In int64, the above value would indeed be -1.

So I guess, we need to decide:

1. What would be a good nodatavalue in binary?
2. Independently: Should an SID be int or unit?

[NiklasPhabian]

would be good to hear @kuoks (and of course, Mike R's) opinion here!