PlasmaCalcs.tools.xarray_tools.xarray_indexing.xarray_argsort

PlasmaCalcs.tools.xarray_tools.xarray_indexing.xarray_argsort(array, dim, newname='{dim}_sort', *, ascending=True, nkeep=None, promote_dims_if_needed=True, squeeze=True, kind=None)

argsort for xarray.DataArray; returns indexes along dim which would sort array along dim.

Drops original coords along dim, because the output along dim is now the indexes;

result[i] corresponds to coord value array[dim][result[i]], not array[dim][i].

Good examples to consider (where result=xarray_argsort(array, dim)):

(1) sortarr = array.isel(dim=result) to get sorted array,

which will have dim as a multidimensional coordinate with same dims as argsort result,

and values telling original values of dim, sorted appropriately.

(2) sortarr = array.isel(dim=result.isel(newdim=slice(0, N)) to get sorted array,

similar to example (1), but only keeping first N lowest values

(or N highest values if ascending=False when getting argsort result).

array: xarray.DataArray

array to argsort.

dim: str

dimension along which to sort.

newname: str

result dim will be replaced by newname.format(dim=dim). Default: ‘{dim}_sort’.

ascending: bool

result sorted in ascending order if True, descending if False.

nkeep: None or int

number of sorted values to keep. E.g. nkeep=2 –> keep only the first 2 values.

equivalent to using result.isel(newdim=(0, nkeep)), where newdim=newname.format(dim=dim).

promote_dims_if_needed: bool

whether to promote non-dimension coords to dimensions.

if False, raise DimensionKeyError if any relevant coord is not already a dimension.

squeeze: bool

whether to drop redundant dims for argsort result.

This means any dims with np.all(array==array.isel(dim=0)). (This includes dims with len=1.)

kind: None or str from {‘quicksort’, ‘mergesort’, ‘heapsort’, ‘stable’}

sorting algorithm. Passed directly to numpy.argsort; see help(numpy.argsort) for details.