Stream: general

Topic: converting and renaming

Jean-Francois Lamarque (Apr 25 2021 at 21:22):

Hi I am trying to read standard CESM output and extract single variables for CMIP6-style output. I might also have to do unit conversion.
Script is ~lamar/Python/H2/rw_h0.ipynb

I am reading the files using

ds=xr.open_mfdataset(files[0:3], concat_dim="time", combine='nested')

then extracting a specific variable

var = ds["H2"]

I tried

var.rename({"H2": "h2"})

but this does not work. The error message is

"cannot rename 'H2' because it is not a variable or dimension in this dataset"

Thoughts?

Stephen Yeager (Apr 25 2021 at 21:32):

Use:
var.rename('h2')
For some reason, the dict construct is only for coordinates?

Jean-Francois Lamarque (Apr 25 2021 at 21:34):

Thanks. Steve. It "works" in that it does not complain, but the output netcdf file still has "H2", not "h2"

/glade/scratch/lamar/tmp/H2/h2_Amon_WACCM6-fr_hydrogen_c1_r1_gn_201001-201412.nc

Stephen Yeager (Apr 25 2021 at 21:35):

You need to overwrite var by doing:
var = var.rename('h2')

Jean-Francois Lamarque (Apr 25 2021 at 22:24):

Making progress :-)

Now it crashes writing the netcdf file

TypeError: unhashable type: 'set'

Jean-Francois Lamarque (Apr 25 2021 at 23:10):

Found a way around it. Not pretty but it works. Since I have a lot of variables, is
there a way to create a function that would do that?

var = ds["H2"]

dsout = xr.Dataset(

data_vars=dict(
    h2=(["time", "lev", "lat", "lon"], var),
),
coords=dict(
    lon=(["lon"], lon),
    lat=(["lat"], lat),
    lev=(["lev"], lev),
    time=date,
),
attrs=dict(description="Hydrogen MIP"),

)
dsout['h2'].attrs["units"] = "mol/mol"

Deepak Cherian (Apr 25 2021 at 23:33):

I would do something like

ds=xr.open_mfdataset(files[0:3], concat_dim="time", combine='nested')
renamer = {"H2": "h2",}  # could add other variables here

attrs = {"h2": {"units": "mol/mol", "description="Hydrogen MIP"}}

ds_renamed = ds.rename_vars(rename)

for var in ds_renamed:
    # alternatively, ds_renamed[var].attrs = attrs[var]
    ds_renamed[var].assign_attrs(attrs[var]).to_netcdf(f"{var}.nc")

Jean-Francois Lamarque (Apr 25 2021 at 23:56):

Thanks, Deepak. I'll try that!

Last updated: Jan 30 2022 at 12:01 UTC