to_netcdf adding unwanted information

[HelixPiano]

Hello everyone,
I am currently struggling with the to_netcdf routine.
I have several hundred different netcdf files and some of them contain an unwanted attribute.
Furthermore, the files are for different variables, so they might contain various attributes and coordinates.
I run somthing simple like:

df=xr.open_dataset("file_here.nc")
df=df.drop_vars("time_bnds")
df.to_netcdf("new_file.nc")

My problem is that xarray starts adding new information to the file that were originally not there.
Excerpt of the original file:

And this is how it looks modified:

As you can see "_FillValue" was added to time and char rotated pole turned into char roated_pole(string1) as well as having rotated poles:coordinates added.

It is important that everything else (except for the dropped variable) stays the same.
The documentation for to_netcdf mentions that I can remove the _FillValue by setting it in the encoding to None, but this would require a full dictenory with the entire encoding of every variable; not to mention that there are differences between my files.
Is there any simple way to do that, or do I have to ditch xarray for that one?

[marceloandrioni]

Maybe something like this:

with xr.open_dataset("file_here.nc") as ds:

    ds = ds.drop_vars("time_bnds")

    ds["rotated_pole"].encoding["dtype"] = "S1"
    ds["time"].encoding["_FillValue"] = None

    ds.to_netcdf("new_file.nc")

but without a file for testing I can't be sure. Can you upload an example file here?

[trexfeathers]

@HelixPiano have you considered using ncdata to fix your file during save? The Rename a dimension in xarray output example demonstrates how this is possible, and you can do a lot more than just renaming dimensions 😊