python - How do you filter a string such that only characters in your list are returned?

Question

Imagine a string, like 'Agh#$%#%2341- -!zdrkfd' and I only wish to perform some operating on it such that only the lowercase letters are returned (as an example), which in this case would bring 'ghzdrkfd'.

How do you do this in Python? The obvious way would be to create a list, of characters, 'a' through 'z', then iterate over the characters in my string and build a new string, character by character, of those in my list only. This seems primitive.

I was wondering if regular expressions are appropriate. Replacing unwanted characters seems problematic and I tend to prefer whitelisting over blacklisting. The .match function does not seem appropriate. I have looked over the appropriate page on the Python site, but have not found a method which seems to fit.

If regular expressions are not appropriate and the correct approach is looping, is there a simple function which "explodes" a string into a list? Or am I just hitting another for loop there?

Answer 1

If you are looking for efficiency. Using the translate function is the fastest you can get.

It can be used to quickly replace characters and/or delete them.

import string
delete_table  = string.maketrans(
    string.ascii_lowercase, ' ' * len(string.ascii_lowercase)
)
table = string.maketrans('', '')

"Agh#$%#%2341- -!zdrkfd".translate(table, delete_table)

In python 2.6: you don't need the second table anymore

import string
delete_table  = string.maketrans(
    string.ascii_lowercase, ' ' * len(string.ascii_lowercase)
)
"Agh#$%#%2341- -!zdrkfd".translate(None, delete_table)

This is method is way faster than any other. Of course you need to store the delete_table somewhere and use it. But even if you don't store it and build it every time, it is still going to be faster than other suggested methods so far.

To confirm my claims here are the results:

for i in xrange(10000):
    ''.join(c for c in s if c.islower())

real    0m0.189s
user    0m0.176s
sys 0m0.012s

While running the regular expression solution:

for i in xrange(10000):
    re.sub(r'[^a-z]', '', s)

real    0m0.172s
user    0m0.164s
sys 0m0.004s

[Upon request] If you pre-compile the regular expression:

r = re.compile(r'[^a-z]')
for i in xrange(10000):
    r.sub('', s)

real    0m0.166s
user    0m0.144s
sys 0m0.008s

Running the translate method the same number of times took:

real    0m0.075s
user    0m0.064s
sys 0m0.012s