Most efficient way to split strings in Python

I was slightly surprised that split() performed so badly in your code, so I looked at it a bit more closely and noticed that you’re calling list.remove() in the inner loop. Also you’re calling split() an extra time on each string. Get rid of those and a solution using split() beats the regex hands down on shorter strings and comes a pretty close second on the longer one.

import timeit
import re

def splitit(input):
    res0 = input.split("|")
    res = []
    for element in res0:
        t = element.split("<>")
        if t != [element]:
            res0.remove(element)
            res.append(t)
    return (res0, res)

def split2(input):
    res0 = input.split("|")
    res1, res2 = [], []
    for r in res0:
        if "<>" in r:
            res2.append(r.split("<>"))
        else:
            res1.append(r)
    return res1, res2

def regexit(input):
    return re.split( "\||<>", input )

rSplitter = re.compile("\||<>")

def regexit2(input):
    return rSplitter.split(input)

print("split: ", timeit.Timer("splitit( 'a|b|c|de|f<>ge<>ah')","from __main__ import splitit").timeit())
print("split2:", timeit.Timer("split2(  'a|b|c|de|f<>ge<>ah')","from __main__ import split2").timeit())
print("regex: ", timeit.Timer("regexit( 'a|b|c|de|f<>ge<>ah')","from __main__ import regexit").timeit())
print("regex2:", timeit.Timer("regexit2('a|b|c|de|f<>ge<>ah')","from __main__ import regexit2").timeit())
print("split: ", timeit.Timer("splitit( 'a|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>ah')","from __main__ import splitit").timeit())
print("split2:", timeit.Timer("split2(  'a|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>ah')","from __main__ import split2").timeit())
print("regex: ", timeit.Timer("regexit( 'a|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>ah')","from __main__ import regexit").timeit())
print("regex2:", timeit.Timer("regexit2('a|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>aha|b|c|de|f<>ge<>ah')","from __main__ import regexit2").timeit())

Which gives the following result:

split:  1.8427431439631619
split2: 1.0897291360306554
regex:  1.6694280610536225
regex2: 1.2277749050408602
split:  14.356198082969058
split2: 8.009285948995966
regex:  9.526430513011292
regex2: 9.083608677960001

And of course split2() gives the nested lists that you wanted whereas the regex solution doesn’t.

Compiling the regex will improve performance. It does make a slight difference, but Python caches compiled regular expressions so the saving is not as much as you might expect. I think usually it isn’t worth doing it for speed (though it can be in some cases), but it is often worthwhile to make the code clearer.