I know this question has been answered, but just to point out what timeit has to say about the solutions efficiency. Using these parameters:
size = 30
s = [str(random.randint(0, 9)) for i in range(size)] + (size/3) * ['-']
random.shuffle(s)
s = ''.join(['+'] + s)
timec = 1000
That is the "phone number" has 30 digits, 1 plus sing and 10 '-'. I've tested these approaches:
def justdigits(s):
justdigitsres = ""
for char in s:
if char.isdigit():
justdigitsres += str(char)
return justdigitsres
re_compiled = re.compile(r'\D')
print('Filter: %ss' % timeit.Timer(lambda : ''.join(filter(str.isdigit, s))).timeit(timec))
print('GE: %ss' % timeit.Timer(lambda : ''.join(n for n in s if n.isdigit())).timeit(timec))
print('LC: %ss' % timeit.Timer(lambda : ''.join([n for n in s if n.isdigit()])).timeit(timec))
print('For loop: %ss' % timeit.Timer(lambda : justdigits(s)).timeit(timec))
print('RE: %ss' % timeit.Timer(lambda : re.sub(r'\D', '', s)).timeit(timec))
print('REC: %ss' % timeit.Timer(lambda : re_compiled.sub('', s)).timeit(timec))
print('Translate: %ss' % timeit.Timer(lambda : s.translate(None, '+-')).timeit(timec))
And came out with these results:
Filter: 0.0145790576935s
GE: 0.0185861587524s
LC: 0.0151798725128s
For loop: 0.0242128372192s
RE: 0.0120108127594s
REC: 0.00868797302246s
Translate: 0.00118899345398s
Apparently GEs and LCs are still slower than a regex or a compiled regex. And apparently my CPython 2.6.6 didn't optimize the string addition that much. translate appears to be the fastest (which is expected as the problem is stated as "ignore these two symbols", rather than "get these numbers" and I believe is quite low-level).
And for size = 100:
Filter: 0.0357120037079s
GE: 0.0465779304504s
LC: 0.0428011417389s
For loop: 0.0733139514923s
RE: 0.0213229656219s
REC: 0.0103371143341s
Translate: 0.000978946685791s
And for size = 1000:
Filter: 0.212141036987s
GE: 0.198996067047s
LC: 0.196880102158s
For loop: 0.365696907043s
RE: 0.0880808830261s
REC: 0.086804151535s
Translate: 0.00587010383606s
row.translate(None, '+-')appears to be the fastest - dmg