Python: why regular expression is slower than replace() method?

Question

I have next 2 blocks of code:

def replace_re(text):
    start = time.time()
    new_text = re.compile(r'(\n|\s{4})').sub('', text)
    finish = time.time()
    return finish - start

def replace_builtin(text):
    start = time.time()
    new_text = text.replace('\n', '').replace('    ', '')
    finish = time.time()
    return finish - start

Than I call both functions with text param (~500kb of source code of one web-page). I thought replace_re() will be much faster, but results are the next:

1. replace_builtin() ~ 0.008 sec
2. replace_re() ~ 0.035 sec (nearly 4.5 times slower!!!)

Why is that?

Answer 1

Because regular expressions are more than 4.5 times more complex than a fixed string replacement.

Answer 2

Because an re has to generate a FSM. Then use that to process the string. While a replace can use the underlying string processing functions closer to the lib/OS levels.