Beautifulsoup : Difference between .find() and .select()

Question

When you use BeautifulSoup to scrape a certain part of a website, you can use

• soup.find() and soup.findAll() or
• soup.select().

Is there a difference between the .find() and the .select() methods? (e.g. In performance or flexibility, etc.) Or are they the same?

Answer 1

To summarise the comments:

• select() finds multiple instances and always returns a list, while find() returns only the first or None, so they don't do the same thing.
  • select_one() would be the parallel to find(), as both return a single element or None
  • select() and find_all() always return a list (an empty list if they have no results)
• I almost always use css selectors when chaining tags or using tag.classname, if looking for a single element without a class I use find(). Essentially it comes down to the use case and personal preference.
• As far as flexibility goes I think you know the answer, soup.select("div[id=foo] > div > div > div[class=fee] > span > span > a") would look pretty ugly using multiple chained find/find_all calls.
• The only issue with the css selectors in bs4 is the very limited support, nth-of-type is the only pseudo class implemented and chaining attributes like a[href][src] is also not supported as are many other parts of css selectors. But things like a[href*=..] , a[href^=], a[href$=] etc.. are I think much nicer than find("a", href=re.compile(....)) but again that is personal preference.

For performance we can run some tests, I modified the code from an answer here running on 800+ html files taken from here, is is not exhaustive but should give a clue to the readability of some of the options and the performance:

The modified functions are:

from bs4 import BeautifulSoup
from glob import iglob


def parse_find(soup):
    author = soup.find("h4", class_="h12 talk-link__speaker").text
    title = soup.find("h4", class_="h9 m5").text
    date = soup.find("span", class_="meta__val").text.strip()
    soup.find("footer",class_="footer").find_previous("data", {
        "class": "talk-transcript__para__time"}).text.split(":")
    soup.find_all("span",class_="talk-transcript__fragment")



def parse_select(soup):
    author = soup.select_one("h4.h12.talk-link__speaker").text
    title = soup.select_one("h4.h9.m5").text
    date = soup.select_one("span.meta__val").text.strip()
    soup.select_one("footer.footer").find_previous("data", {
        "class": "talk-transcript__para__time"}).text
    soup.select("span.talk-transcript__fragment")


def  test(patt, func):
    for html in iglob(patt):
        with open(html) as f:
            func(BeautifulSoup(f, "lxml")

Now for the timings:

In [7]: from testing import test, parse_find, parse_select

In [8]: timeit test("./talks/*.html",parse_find)
1 loops, best of 3: 51.9 s per loop

In [9]: timeit test("./talks/*.html",parse_select)
1 loops, best of 3: 32.7 s per loop

Like I said not exhaustive but I think we can safely say the css selectors are definitely more efficient.