Order of x in weighted_median matters but is not documented

[Lyalpha]

I might expect

assert(robustats.weighted_median([1,2,3], [1,2,3]) == robustats.weighted_median([3,2,1], [3,2,1]))

to not raise an error, unless I am really missing something? It seems somewhere that there is dependency on the ordering of x but I can't work out any consistency, e.g.

assert(robustats.weighted_median([3,2,1], [3,2,1]) == robustats.weighted_median([3,1,2], [3,1,2]))

is fine (but note both incorrect from the [1,2,3] ordering.

Is there an implicit assumption that x is passed in ascending order? In this case it needs to be documented (or sorted prior to or within the call to the c-code perhaps)