A simple utility for comparing the efficiency of functionally-equivalent functions in Python.
pip install pybenchmarker
Problem
Input: two lists of equal length list_0, list_1.
Output: a list of booleans encoding which items of list_0 are in
list_1.
What's the fastest way to do this?
import numpy as np
from pybenchmarker import BenchmarkN, sizes
from random import randint
@sizes([2**k for k in range(18)])
def argfunc(n):
return [randint(1, n) for i in range(n)], [randint(1, n) for i in range(n)]
def naive_lcomp(lists):
return [x in lists[1] for x in lists[0]]
def set_other(lists):
s = set(lists[1])
return [x in s for x in lists[0]]
def numpy_isin(lists):
return list(np.isin(lists[0], lists[1]))
if __name__ == '__main__':
title = "list_3: list[bool], encoding which items in list_1 are in list_2"
b = BenchmarkN(functions=[naive_lcomp, set_other, numpy_isin],
argfunc=argfunc)
b.plot(xlabel="n", title=title, fname="my_figure", transparent=True,
dpi=300)This results in the following plot:
The BenchmarkN class was inspired by perfplot,
which was used in
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent