ncks -A -v [5.2.4] replace attribute _FillValue by eulaVlliF_

[EnlNovius]

A rather strange situation I've encountered when performing a ncks -A -v from one file to another:
the _FillValue attribute of the variable associated with the dimension on which the variable I'm copying depends is replaced by an eulaVlliF_ attribute.

---

Reproduction code:

• First step: create two files, a source file and a destination file.

import numpy as np
import netCDF4 as nc
import os

os.makedirs('debug', exist_ok=True)

with nc.Dataset('debug/from.nc', 'w') as file:
    file.createDimension('new_latitude', 22)
    file.createDimension('new_longitude', 52)

    Chi_2 = file.createVariable('Chi_2', 'f8', ('new_latitude'), fill_value=nc.default_fillvals['f8'])
    new_latitude = file.createVariable('new_latitude', 'f4', ('new_latitude',), fill_value=nc.default_fillvals['f4'])
    new_longitude = file.createVariable('new_longitude', 'f4', ('new_longitude',), fill_value=nc.default_fillvals['f4'])

    Chi_2[:] = np.random.rand(22).astype(np.float64)
    new_latitude[:] = np.arange(22).astype(np.float32)
    new_longitude[:] = np.arange(52).astype(np.float32)
    
with nc.Dataset('debug/destination.nc', 'w') as file:
    file.createDimension('new_latitude', 22)
    file.createDimension('new_longitude', 52)

    new_latitude = file.createVariable('new_latitude', 'f4', ('new_latitude',), fill_value=nc.default_fillvals['f4'])
    new_longitude = file.createVariable('new_longitude', 'f4', ('new_longitude',), fill_value=nc.default_fillvals['f4'])

    new_latitude[:] = np.arange(22).astype(np.float32)
    new_longitude[:] = np.arange(52).astype(np.float32)

• The destination file:

with nc.Dataset('debug/destination.nc', 'r') as file:
    for var in file.variables:
        print(var, file[var].ncattrs())

>>> new_latitude ['_FillValue']
>>> new_longitude ['_FillValue']

Everything is fine here.

• Running the ncks command:

ncks -h -A -v Chi_2 debug/from.nc debug/destination.nc

• Result:

with nc.Dataset('debug/destination.nc', 'r') as file:
    for var in file.variables:
        print(var, file[var].ncattrs())

>>> new_latitude ['eulaVlliF_']
>>> new_longitude ['_FillValue']
>>> Chi_2 ['_FillValue']

And here we have our new attribute eulaVlliF_ for new_latitude.

---

Environment:

• python 3.12.3
• nco 5.2.4
• netcdf4 1.6.5

[czender]

Thank you for reporting this. It must indeed look strange from your perspective. NCO has a "workaround" to netCDF library restrictions for modifying the _FillValue attribute after variable creation. The workaround temporarily reverses the name of the attribute. Apparently, you have found a behavior in which the reversed attributes is never un-reversed. I will look into this as soon as I can.