String Slicing and clipping in Numba

Numba
1

I have a string array (pretty huge ~1Billion elements) and another array with integers. I want to slice the individual string based on its corresponding element in the second array. I have got it worked in python but when I try to translate it to numba, I get error (I believe numba is prohibiting string slicing for e.g. a = ‘abc’; a[1:] #string slice. Any way to solve this problem?

I am trying to jit the following python_clipper function:

arr1 = ['abc', 'def', 'xyz']
arr2 = [1,2,3]

def python_clipper(arr1,arr2):
    for i in range(len(arr1)):
        arr1[i] = arr1[i][arr2[i]:]
    return arr1

print(python_clipper(arr1,arr2)) # ['bc', 'f', '']

I am ok with arr1 and arr2 being np.array as well.

2

Do you have an example of what you’d like to do that works in plain python but not when numba jitted?

3

Do you have access to an NVIDIA GPU? If you aren’t able to find a Numba-only solution, it may be that cuDF would enable your use case, as it includes Numba extensions for string handling: Overview of User Defined Functions with cuDF — cudf 23.02.00 documentation

4

@nelson2005 I have edited the post to include the example.
@gmarkall I have the nvidia GPU but am unfamiliar with cuDF. I will go through the docs to see how I can implement it.

5

Something like this works for me

import numba
import numpy as np

arr1 = np.asarray(['abc', 'def', 'xyz'])
arr2 = np.asarray([1, 2, 3])

@numba.njit()
def python_clipper(arr1, arr2):
    for i in range(len(arr1)):
        arr1[i] = str(arr1[i])[arr2[i]:]
    return arr1

print(python_clipper(arr1,arr2)) # ['bc', 'f', '']
6

How about with @cuda.jit()?