Pass a string to xarray.Dataset.reduce

[tqa236]

Hello,

I'm not sure if this feature exists or not, but I would like to know if it's possible to pass a string to xarray.Dataset.reduce as the aggregation method and receive the corresponding reduced dataset instead of a series of if else like this

def aggregate(dataset, aggregation_method):
    if aggregation_method == "mean":
        return dataset.mean()
    elif aggregation_method == "median":
        return dataset.median()
    elif aggregation_method == "max":
        return dataset.max()

From the docstring, it seems that the method only accept Callable for func

Thank you very much.

[keewis]

additional keyword-arguments are passed through (this is what **kwargs in the docstring is referring to):

ds.reduce(aggregate, aggregation_method="mean")

you could also use functools.partial or a lambda to create a function that has that parameter pre-set:

ds.reduce(functools.partial(aggregate, aggregation_method="mean"))
ds.reduce(lambda ds_: aggregate(ds_, "mean"))

Edit: if aggregation_method is the name of a method on the dataset, we can also do:

getattr(ds, aggregation_method)()

[tqa236]

Hi @keewis ,

Thanks for your reply. I probably haven't phrased my question quite well. I would like to know if it's possible to do something like ds.reduce("mean") without the need to define the function aggregate.

Something like this (this code doesn't work)

ds.reduce(getattr(ds, aggregation_method))

[keewis]

for this we can make use of python's bound method objects:

getattr(ds, aggregation_method)()

If aggregation_method is the name of a method, getattr returns a function with ds bound to the first argument (usually named self by convention), so we can just call that.

[tqa236]

thank you very much for your help