xarray open kwargs

[larsbuntemeyer]

This PR adds the xarray_open_kwargs keyword that let's you control the xarray.open_dataset keyword arguments in case of returnXDataset=True. This allows me, e.g., to control parallel reads and chunking.

Use case

I have a non standard calendar and want to select a date into an xarray dataset:

import numpy as np
import pandas as pd
import xarray as xr
from cdo import Cdo

np.random.seed(0)
temperature = 15 + 8 * np.random.randn(1, 2, 2)
precipitation = 10 * np.random.rand(1, 2, 2)
lon = [[-99.83, -99.32], [-99.79, -99.23]]
lat = [[42.25, 42.21], [42.63, 42.59]]
time = xr.cftime_range(start="2000-02-30", periods=1, calendar="360_day")


ds = xr.Dataset(
    data_vars=dict(
        temperature=(["time", "y", "x"], temperature),
        precipitation=(["time", "y", "x"], precipitation),
    ),
    coords=dict(
        lon=(["x", "y"], lon),
        lat=(["x", "y"], lat),
        time=time,
    ),
    attrs=dict(description="Weather related data."),
)

cdo = Cdo(xarray_open_kwargs={'use_cftime':True}, logging=True, debug=False)
cdo.seldate("2000-02-30T00:00:00", input=ds, returnXDataset=True)

This only works with use_cftime=True.

[Try2Code]

Hi @larsbuntemeyer !

Thanks for the input. I am not a xarray-poweruser, so this is quite valuable for me 👍

But I am not sure, if this done in the best way: On the one hand opening files is done internally - so it's a bit hard to change this behaviour. But on the other hand the whole operation depends on the input very much.
I wonder if this should be better done on an operator call than during the constructor call of the cdo object. What you want to change might differ a lot from one file to the next, right?

[larsbuntemeyer]

@Try2Code Thanks for considering this! Sorry for the late response, you are right, this could also go along with the returnX* arguments? So it can be set more dynamically depending on the input variable. I'll have a look into this....

[Try2Code]

I think this options is better put into an operator call than in the constructor. That way users can change the options for each data set they create

[Try2Code]

I am in the middle of a re-design. Will keep this PR open to keep the feature in mind.