Description
The "obvious" way to get the rank of elements in a NumPy array is to use array.argsort().argsort():
https://stackoverflow.com/questions/5284646/rank-items-in-an-array-using-python-numpy
This is obviously slightly inefficient, due to sorting the array twice. In contrast, the optimal solution simply inverts the order of the array elements for the second argsort(), which can be done in a single pass using indexing assignment:
array = numpy.array([4,2,7,1])
temp = array.argsort()
ranks = numpy.empty(len(array), int)
ranks[temp] = numpy.arange(len(array))I'd like to propose adding an invert_permutation() or inverse_permutation() helper function to take care of this logic, which could be modeled off this helper function in xarray:
https://github.com/pydata/xarray/blob/2949558b75a65404a500a237ec54834fd6946d07/xarray/core/nputils.py#L38-L55
As precedent, note that TensorFlow has such a function, too: https://www.tensorflow.org/versions/r1.0/api_docs/python/tf/invert_permutation