Add lonFlip function

[brianpm]

A very common operation performed with global data is to convert the longitude convention from/to [0,360] to/from [-180, 180]. This was available in NCL as lonFlip.

There are probably many ways this could be implemented. I've written this function a bunch of times. My current version assumes global data and that the data is an xarray object, and it at least seems to work:

def lonFlip(data, lonname=None):
    # NOTE: this assumes global values
    if lonname is None:
        lonname = 'lon'
    tmplon = data[lonname]
    tmpdata = data.roll( {lonname: (len(tmplon) // 2)}, roll_coords=True)
    lonroll = tmpdata[lonname].values
    if tmplon.min() >= 0:
        # flip to -180:180
        tmpdata[lonname] = xr.DataArray(np.where(lonroll >= 180, lonroll - 360, lonroll), dims=[lonname], attrs=data[lonname].attrs)
    else:
        # flip from -180:180 to 0:360
        tmpdata[lonname] = xr.DataArray(((lonroll + 360) % 360), dims=[lonname], attrs=data[lonname].attrs)
    return tmpdata

[anissa111]

Hi @brianpm! 👋

Thank you for the functionality request! We'll provide updates here of our review and implementation progress.

[kafitzgerald]

Thanks, Brian!

Jotting down a couple of related dev notes:

This issue (pydata/xarray#7031) regarding a periodic boundary index and a related conversation on the Pangeo Discourse might be worth a look. It's a different solution and relies on some work currently under development, but may be worth keeping an eye on.

I think something using roll and updating coordinates could make sense though and while there may be overlap with the periodic boundary index option there are probably cases where one or the other might make more sense.

From a design perspective, we might also consider something a little more flexible like lonPivot which covers the lonFlip case as well.