When working with some internal code in my job. I realize some code I got from others run with numpy.split
inside jitted function and it works. But I can not find any documentation on support of this numpy feature. Any idea what’s the status of support for overloading numpy.split in numba?
My use case is to have a 1d array A of sorted non-decreasing dates (converted to int64 type). For other 1d array, 2d array where axis=0 has same length as array A. I would like to a groupby summation operation based on dates in array A. I had some naive implementation in numba, but seems to be much slower than code I got where numpy.spilt is used in numba. Any other thoughts on solving this problem is very welcomed. Thanks!