[Python] More efficient version of numpy.searchsorted for two arrays that are both sorted...

According to the documentation, given two arrays a, b of size n, m, numpy.searchsorted(a, b) uses bisect library to find the location of each b in array a. The complexity of this algorithm should be m log n. Note that this algorithm does not require b to be sorted.

However, I am currently facing a slightly different problem where both a, b are sorted. In this case, there is a more efficient algorithm with complexity n+m using scanning. Does numpy have this algorithm? If not, how should I write code using the current numpy functions with improved efficiency?

Continue reading...