[issue42885] Regex performance problem with ^ aka AT_BEGINNING

[Steven D'Aprano]

Steven D'Aprano <steve+python at pearwood.info> added the comment:

I'm getting similar results in Python 3.9.

[steve ~]$ python3.9 -m timeit -s "a = 'A'*10000" -s "import re" "re.search('^x', a)"
5000 loops, best of 5: 67.3 usec per loop
[steve ~]$ python3.9 -m timeit -s "a = 'A'*100000" -s "import re" "re.search('^x', a)"
500 loops, best of 5: 639 usec per loop
[steve ~]$ python3.9 -m timeit -s "a = 'A'*1000000" -s "import re" "re.search('^x', a)"
50 loops, best of 5: 6.27 msec per loop
[steve ~]$ python3.9 -m timeit -s "a = 'A'*10000000" -s "import re" "re.search('^x', a)"
5 loops, best of 5: 62.8 msec per loop
[steve ~]$ python3.9 -m timeit -s "a = 'A'*100000000" -s "import re" "re.search('^x', a)"
1 loop, best of 5: 654 msec per loop


It looks like the time is roughly linear in the length of the string.

I get the same result as far back as Python 2.7 (I haven't tried older versions).

[steve ~]$ python2.7 -m timeit -s "a = 'A'*10000" -s "import re" "re.search('^x', a)"
10000 loops, best of 3: 75.7 usec per loop
[steve ~]$ python2.7 -m timeit -s "a = 'A'*10000000" -s "import re" "re.search('^x', a)"
10 loops, best of 3: 73.4 msec per loop


I would have expected essentially constant time, as in re.match:

[steve ~]$ python3.9 -m timeit -s "a = 'A'*10000" -s "import re" "re.match('x', a)"
500000 loops, best of 5: 560 nsec per loop
[steve ~]$ python3.9 -m timeit -s "a = 'A'*100000000" -s "import re" "re.match('x', a)"
500000 loops, best of 5: 561 nsec per loop

----------
nosy: +steven.daprano
versions: +Python 3.9

_______________________________________
Python tracker <report at bugs.python.org>
<https://bugs.python.org/issue42885>
_______________________________________