17
votes

What's the most efficient method to remove a list of substrings from a string?

I'd like a cleaner, quicker way to do the following:

words = 'word1 word2 word3 word4, word5'
replace_list = ['word1', 'word3', 'word5']

def remove_multiple_strings(cur_string, replace_list):
  for cur_word in replace_list:
    cur_string = cur_string.replace(cur_word, '')
  return cur_string

remove_multiple_strings(words, replace_list)
1
If the list is ['apple', 'banana'], and the string is 'banappleana', should that become '' or 'banana'? Your current code would produce ''. - user2357112 supports Monica

1 Answers

19
votes

Regex:

>>> import re
>>> re.sub(r'|'.join(map(re.escape, replace_list)), '', words)
' word2  word4, '

The above one-liner is actually not as fast as your string.replace version, but definitely shorter:

>>> words = ' '.join([hashlib.sha1(str(random.random())).hexdigest()[:10] for _ in xrange(10000)])
>>> replace_list = words.split()[:1000]
>>> random.shuffle(replace_list)
>>> %timeit remove_multiple_strings(words, replace_list)
10 loops, best of 3: 49.4 ms per loop
>>> %timeit re.sub(r'|'.join(map(re.escape, replace_list)), '', words)
1 loops, best of 3: 623 ms per loop

Gosh! Almost 12x slower.

But can we improve it? Yes.

As we are only concerned with words what we can do is simply filter out words from the words string using \w+ and compare it against a set of replace_list(yes an actual set: set(replace_list)):

>>> def sub(m):
    return '' if m.group() in s else m.group()
>>> %%timeit
s = set(replace_list)
re.sub(r'\w+', sub, words)
...
100 loops, best of 3: 7.8 ms per loop

For even larger string and words the string.replace approach and my first solution will end up taking quadratic time, but the solution should run in linear time.