Removing a list of characters in string

Question

I want to remove characters in a string in python:

string.replace(',', '').replace("!", '').replace(":", '').replace(";", '')...

But I have many characters I have to remove. I thought about a list

list = [',', '!', '.', ';'...]

But how can I use the list to replace the characters in the string?

Solution 1

If you're using python2 and your inputs are strings (not unicodes), the absolutely best method is str.translate:

>>> chars_to_remove = ['.', '!', '?']
>>> subj = 'A.B!C?'
>>> subj.translate(None, ''.join(chars_to_remove))
'ABC'

Otherwise, there are following options to consider:

A. Iterate the subject char by char, omit unwanted characters and join the resulting list:

>>> sc = set(chars_to_remove)
>>> ''.join([c for c in subj if c not in sc])
'ABC'

(Note that the generator version ''.join(c for c ...) will be less efficient).

B. Create a regular expression on the fly and re.sub with an empty string:

>>> import re
>>> rx = '[' + re.escape(''.join(chars_to_remove)) + ']'
>>> re.sub(rx, '', subj)
'ABC'

(re.escape ensures that characters like ^ or ] won't break the regular expression).

C. Use the mapping variant of translate:

>>> chars_to_remove = [u'δ', u'Γ', u'ж']
>>> subj = u'AжBδCΓ'
>>> dd = {ord(c):None for c in chars_to_remove}
>>> subj.translate(dd)
u'ABC'

Full testing code and timings:

#coding=utf8

import re

def remove_chars_iter(subj, chars):
    sc = set(chars)
    return ''.join([c for c in subj if c not in sc])

def remove_chars_re(subj, chars):
    return re.sub('[' + re.escape(''.join(chars)) + ']', '', subj)

def remove_chars_re_unicode(subj, chars):
    return re.sub(u'(?u)[' + re.escape(''.join(chars)) + ']', '', subj)

def remove_chars_translate_bytes(subj, chars):
    return subj.translate(None, ''.join(chars))

def remove_chars_translate_unicode(subj, chars):
    d = {ord(c):None for c in chars}
    return subj.translate(d)

import timeit, sys

def profile(f):
    assert f(subj, chars_to_remove) == test
    t = timeit.timeit(lambda: f(subj, chars_to_remove), number=1000)
    print ('{0:.3f} {1}'.format(t, f.__name__))

print (sys.version)
PYTHON2 = sys.version_info[0] == 2

print ('\n"plain" string:\n')

chars_to_remove = ['.', '!', '?']
subj = 'A.B!C?' * 1000
test = 'ABC' * 1000

profile(remove_chars_iter)
profile(remove_chars_re)

if PYTHON2:
    profile(remove_chars_translate_bytes)
else:
    profile(remove_chars_translate_unicode)

print ('\nunicode string:\n')

if PYTHON2:
    chars_to_remove = [u'δ', u'Γ', u'ж']
    subj = u'AжBδCΓ'
else:
    chars_to_remove = ['δ', 'Γ', 'ж']
    subj = 'AжBδCΓ'

subj = subj * 1000
test = 'ABC' * 1000

profile(remove_chars_iter)

if PYTHON2:
    profile(remove_chars_re_unicode)
else:
    profile(remove_chars_re)

profile(remove_chars_translate_unicode)

Results:

2.7.5 (default, Mar  9 2014, 22:15:05) 
[GCC 4.2.1 Compatible Apple LLVM 5.0 (clang-500.0.68)]

"plain" string:

0.637 remove_chars_iter
0.649 remove_chars_re
0.010 remove_chars_translate_bytes

unicode string:

0.866 remove_chars_iter
0.680 remove_chars_re_unicode
1.373 remove_chars_translate_unicode

---

3.4.2 (v3.4.2:ab2c023a9432, Oct  5 2014, 20:42:22) 
[GCC 4.2.1 (Apple Inc. build 5666) (dot 3)]

"plain" string:

0.512 remove_chars_iter
0.574 remove_chars_re
0.765 remove_chars_translate_unicode

unicode string:

0.817 remove_chars_iter
0.686 remove_chars_re
0.876 remove_chars_translate_unicode

(As a side note, the figure for remove_chars_translate_bytes might give us a clue why the industry was reluctant to adopt Unicode for such a long time).

Solution 2

You can use str.translate():

s.translate(None, ",!.;")

Example:

>>> s = "asjo,fdjk;djaso,oio!kod.kjods;dkps"
>>> s.translate(None, ",!.;")
'asjofdjkdjasooiokodkjodsdkps'

Solution 3

You can use the translate method.

s.translate(None, '!.;,')

Solution 4

If you are using python3 and looking for the translate solution - the function was changed and now takes 1 parameter instead of 2.

That parameter is a table (can be dictionary) where each key is the Unicode ordinal (int) of the character to find and the value is the replacement (can be either a Unicode ordinal or a string to map the key to).

Here is a usage example:

>>> list = [',', '!', '.', ';']
>>> s = "This is, my! str,ing."
>>> s.translate({ord(x): '' for x in list})
'This is my string'

Solution 5

''.join(c for c in myString if not c in badTokens)

View more solutions

Comments

[Martijn de Milliano]

See stackoverflow.com/questions/1919096/… for various solutions and a nice comparison.

[JustAC0der]

It's a pity that Python (which is said to come with batteries included) does not handle this use case out of the box. PHP's function str_replace does it - you can pass an array as the first argument and a string as the second (php.net/manual/pl/function.str-replace.php ).

[Sven Marnach]

@thg435: Nobody asked for this, but anyway: s.translate(dict.fromkeys(map(ord, u",!.;")))

[hobs]

This (and @PraveenGollakota's) simultaneous answer is exactly what @Laura asked for and should be the preferred answer(s).

[antonavy]

The second method raises an error TypeError: translate() takes exactly one argument (2 given). Apparently it takes dict as an argument.

[Wolf]

Useful in similar cases not based on chars and strings +1

[FuzzyAmi]

@antonavy - the 2nd solution does work - but only of the string is not unicode (for which a different translate() is needed)

[Gank]

why python3:TypeError: translate() takes exactly one argument (2 given)

[Sven Marnach]

@Gank: The unicode.translate() method has different parameters than the str.translate() method. Use the variant in the above comment for Unicode objects.

[rioted]

Also note that this can just as easily be done with list comprehensions, which is way more pythonic in my opinion.

[Jun711]

@SvenMarnach what is map(ord, u",!.;"))? and does u stand for unicode?

[Sven Marnach]

@Jun This maps the ord() function on all characters of the Unicode string u",!.;", resulting in a list (Python 2) or an iterator (Python 3) of Unicode code points. The u denotes a Unicode string in Python 2. In Python 3 it's optional, since strings are Unicode strings by default (and early versions of Python 3 did not even support the u"" syntax).

[Aybid]

Explanation: 1. maketrans(x,y,z): the third parameter is for replace. x,y are '' here, so makes no change. Only characters with z are removed 2. translate(): returns a string where each character is mapped to its corresponding character in the translation table (here from the maketrans fn)

[confiq]

str.translate() is from python2 and there is no need to use maketrans() func.

[2e0byo]

@config on the contrary: python 3 has a str.translate() which takes char ords, and thus does require something like str.maketrans() (or ee.g. a dict comp calling ord as in the other python3 answer).. This answer won't work without the maketrans() (try it).