Beautifulsoup4 with lxml vs Beautifulsoup3

Question

I am migrating some parsers from BeautifulSoup3 to BeautifulSoup4 and I thought it would be a good idea to profile how faster it would get considering that lxml is super fast and it's the parser I am using with BS4, here are the profile results:

For BS3:

43208 function calls (42654 primitive calls) in 0.103 seconds

Ordered by: standard name

ncalls  tottime  percall  cumtime  percall filename:lineno(function)
    1    0.000    0.000    0.000    0.000 <string>:2(<module>)
   18    0.000    0.000    0.000    0.000 <string>:8(__new__)
    1    0.000    0.000    0.072    0.072 <string>:9(parser)
   32    0.000    0.000    0.000    0.000 BeautifulSoup.py:1012(__init__)
    1    0.000    0.000    0.000    0.000 BeautifulSoup.py:1018(buildTagMap)
... 

For BS4 using lxml:

164440 function calls (163947 primitive calls) in 0.244 seconds

Ordered by: standard name

ncalls  tottime  percall  cumtime  percall filename:lineno(function)
    1    0.040    0.040    0.069    0.069 <string>:2(<module>)
   18    0.000    0.000    0.000    0.000 <string>:8(__new__)
    1    0.000    0.000    0.158    0.158 <string>:9(parser)
    1    0.000    0.000    0.008    0.008 HTMLParser.py:1(<module>)
    1    0.000    0.000    0.000    0.000 HTMLParser.py:54(HTMLParseError)
...

why BS4 is calling 4 times more functions? why is it using the HTMLParser at all if I set it to use lxml?

The most noticeable things I changed from BS3 to BS4 were this:

 BeautifulSoup(html, convertEntities=BeautifulSoup.HTML_ENTITIES)  --->
 BeautifulSoup(html, 'lxml')

 [x.getText('**SEP**') for x in i.findChildren('font')[:2]] --->
 [x.getText('**SEP**', strip=True) for x in i.findChildren('font')[:2]]

everything else is just some name changes (like findParent --> find_parent)

EDIT:

my environment:

python 2.7.3
beautifulsoup4==4.1.0
lxml==2.3.4

EDIT 2:

Here is a small code sample to try it out:

from cProfile import Profile

from BeautifulSoup import BeautifulSoup
from bs4 import BeautifulSoup as BS4
import urllib2


def parse(html):

    soup = BS4(html, 'lxml')
    hl = soup.find_all('span', {'class': 'mw-headline'})
    return [x.get_text(strip=True) for x in hl]


def parse3(html):

    soup = BeautifulSoup(html, convertEntities=BeautifulSoup.HTML_ENTITIES)
    hl = soup.findAll('span', {'class': 'mw-headline'})
    return [x.getText() for x in hl]


if __name__ == "__main__":
    opener = urllib2.build_opener()
    opener.addheaders = [('User-agent', 'Mozilla/5.0')]
    html = ''.join(opener.open('http://en.wikipedia.org/wiki/Price').readlines())

    profiler = Profile()
    print profiler.runcall(parse, html)
    profiler.print_stats()

    profiler2 = Profile()
    print profiler2.runcall(parse3, html)
    profiler2.print_stats()

Answer 1

I believe the main problem is a bug in Beautiful Soup 4. I've filed it and a fix will be released in the next version. Thanks for finding this.

That said, I have no idea why your profile mentions the HTMLParser class at all, given that you're using lxml.