xr_rm_poly for xr.Datasets? about climpred HOT 3 CLOSED

@aaronspring , I have implemented a fast method for removing polynomials from datasets in 69d2499

It's a little hacky, but I concatenate all variables into a single dimension, convert it to a numpy array, run the polynomial removal, then build the dataset back up.

from climpred.

@aaronspring, I agree that we should opt for Dataset capability.

For right now, a really simple implementation is using the .apply() function:

This is fast with the two variables and 10,10 lat/lon dimensions in this example. I'm not sure how fast it is with larger arrays.

So for now we could probably just update the stats module to check if it's a Dataset:

def _check_dataset(x):
    if isinstance(x, xr.Dataset):
        return True

Then for each function run _check_dataset(input) and then just run .apply() on itself?

from climpred.

Yep. That sounds good. I think there is now performance issue from np functions but rather those from scipy on larger grids.

from climpred.