Web Scraping with CSS Selectors using Python

When web-scraping, CSS selectors are one of the best friends. This tutorial will tell you what they're, their pros and cons, and why they matter from a web scraping perspective with Python examples to get you going.

This blog is suited for users with little (scraped a few websites or saw how someone is doing it) web scraping experience.

Intro

This blog post is about understanding CSS selectors when doing web page web scraping, and what tools might be handy to use in addition to Python beautifulsoup, lxml libraries.

Usage of CSS selectors in different languages, frameworks, packages are not much different.

Prerequisites

pip install lxml 
pip install beautifulsoup4

Enter fullscreen modeExit fullscreen mode

A basic familiarity with bs4 library or whatever HTML parser package/framework you're using.

This is needed if you want to train extract data from CSS selectors on your own. If you don't want to do that, installing these libraries is not required.

Explanation

CSS selectors are patterns used to select (match) the element(s) you want to style extract.

SelectorGadget

Let's start with easy one, SelectorGadget extension. This extension allows to quickly grab CSS selector(s) by clicking on desired element in your browser, and returns a CSS selector(s).

SelectorGadget is an open-source tool that makes CSS selector generation and discovery on complicated sites a breeze.

Uses cases:

for web page scraping with tools such as Nokogiri and BeautifulSoup.
to generate jQuery selectors for dynamic sites.
as a tool to examine JavaScript-generated DOM structures.
as a tool to help you style only particular elements on the page with your stylesheets.
for selenium or phantomjs testing.

When using SelectorGadget it highlights element(s) in:

yellow which is mean that it's guessing what the user is looking for, and needs possible additional clarification.

red excludes from match selection.

green includes to match selection.

Pick CSS Selectors by Hand

Now is the time to think a little, just a little. Since SelectorGadget isn't a magical all around tool, sometimes it can't get the desired element. This happens when website HTML tree is not well structured, or if the site is rendered via JavaScript.

When it happens, I use Elements tab via Dev Tools (F12 on a keyboard) to locate and grab CSS selector(s) or HTML elements by their:

type selector: <input>
class selector: .class
id selector: #id
attribute selector: [attribute]

Type Selectors

Syntax: element_name

Type selectors matches elements by node name. In other words, it selects all elements of the given type within a HTML document.

soup.select('a')      # returns all <a> elements
soup.select('span')   # returns all <span> elements
soup.select('input')  # returns all <input> elements
soup.select('script') # returns all <script> elements

Enter fullscreen modeExit fullscreen mode

Class Selectors

Syntax: .class_name

Class selectors matches elements based on the contents of their class attribute. It's like calling a class function PressF().when_playing_cod().

soup.select('.mt-5')                   # returns all elements with current .selector
soup.select('.crayons-avatar__image')  # returns all elements with current .selector
soup.select('.w3-btn')                 # returns all elements with current .selector

Enter fullscreen modeExit fullscreen mode

ID Selectors

Syntax: #id_value

ID selectors matches an element based on the value of the elements id attribute. In order for the element to be selected, its id attribute must match exactly the value given in the selector.

soup.select('#eob_16')              # returns all elements with current #selector
soup.select('#notifications-link')  # returns all elements with current #selector
soup.select('#value_hover')         # returns all elements with current #selector

Enter fullscreen modeExit fullscreen mode

Attribute Selectors

Syntax: [attribute=attribute_value] or [attribute], more examples.

Attribute selectors matches elements based on the presence or value of a given attribute.

The only difference is that this selectors uses curly braces [] instead of a dot (.) as class, or a hash (or octothorpe) symbol (#) as ID.

soup.select('[jscontroller="K6HGfd"]')         # returns all elements with current [selector]
soup.select('[data-ved="2ascASqwfaspoi_SA8"]') # returns all elements with current [selector]

# elements with an attribute name of data-id
soup.select('[data-id]')                       # returns all elements with current [selector]

Enter fullscreen modeExit fullscreen mode

Selector List

Syntax: element, element, element, ...

Selector list selects all the matching nodes (elements). In web scraping perspective this CSS selectors is great (in my opinion) to handle different HTML layouts because if one of the selectors is present it will grab all elements from a existing selector.

As an example from Google Search (carousel results), the HTML layout will be different depending on country where the search is coming from.

When country of the search is not United States:

When country of the search is set to United States:

Following examples translates to this code snippet (handles both HTML layouts):

# will return all elements either by one of these selectors
soup.select('#kp-wp-tab-Albums .PZPZlf, .keP9hb')

Enter fullscreen modeExit fullscreen mode

Descendant combinator

Syntax: selector1 selector2

Descendant combinator represented by a single space () character and selects two selectors such that elements matched by the second selector are selected if they have an ancestor (parent, parent's parent, parent, etc) element matching the first selector.

soup.select('.NQyKp .REySof')   # dives insie .selector -> dives again to other .selector and grabs it
soup.select('div cite.iUh30')   # dives inside div -> dives inside cite.selector and grabs it
soup.select('span#21Xy a.XZx2') # dives inside span#id -> dives insize a.selector and grabs it

Enter fullscreen modeExit fullscreen mode

Other Useful CSS Selectors

:nth-child(n): Selects every n element that is the second child of its parent.
:nth-of-type(n): Selects every n element that is the second n element of its parent.
a:has(img): Selects every element <a> element that has an <img> element.

Additional useful CSS selectors you can find on W3C Level 4 Selectors, W3Schools CSS Selectors Reference, and MDN documentation.

Test CSS Selectors

To test if the selector extracts correct data you can:

Place those CSS selector(s) in the SelectorGadget window and see what elements being selected:

Use Dev Tools Console tab via $$(".selector") method (creates an array (list()) of elements):

$$(".DKV0Md")

Enter fullscreen modeExit fullscreen mode

Which is equivalent to document.querySelectorAll(".selector") method (according to Chrome Developers website):

document.querySelectorAll(".DKV0Md")

Enter fullscreen modeExit fullscreen mode

Output from the DevTools Console for both methods are the same:

Cons of CSS Selector

Betting only classes might be not a good idea since they could probably change.

A "better" way (in terms of CSS selectors) would be to use selectors such as attribute selectors (mentioned above), they are likely to change less frequently. See attribute selectors examples on the screenshot below (HTML from Google Organic results):

Many modern websites use autogenerated CSS selectors for every change that is being made to certain style component, which means that rely exclusively on them is not a good idea. But again, it will depend on how often do they really change.

The biggest problem that might appear is that when code will be executing it will blow up with an error, and maintainer of the code should manually change CSS selector(s) to make the code running properly. Seems like not a big deal, which is true, but it might be annoying if selectors are changing frequently.

Code Examples

This section will show a couple of actual examples from different websites to get you familiarize a bit more.

Extract title, snippet, link, displayed link from Google Search results.

Test CSS container selector:

Code:

import requests, lxml
from bs4 import BeautifulSoup

headers = {
    "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/96.0.4664.45 Safari/537.36"
}

html = requests.get("https://www.google.com/search?q=minecraft", headers=headers)
soup = BeautifulSoup(html.text, "lxml")

for result in soup.select(".tF2Cxc"):
    title = result.select_one(".DKV0Md").text
    link = result.select_one(".yuRUbf a")["href"]
    displayed_link = result.select_one(".lEBKkf span").text
    snippet = result.select_one(".lEBKkf span").text

    print(f"{title}\n{link}\n{displayed_link}\n{snippet}\n")


# part of the output 
'''
Log in | Minecraft
https://minecraft.net/login
https://minecraft.net › login
Still have a Mojang account? Log in here: Email. Password. Forgot your password? Login. Mojang © 2009-2021. "Minecraft" is a trademark of Mojang AB.

What is Minecraft? | Minecraft
https://www.minecraft.net/en-us/about-minecraft
https://www.minecraft.net › en-us › about-minecraft
Prepare for an adventure of limitless possibilities as you build, mine, battle mobs, and explore the ever-changing Minecraft landscape.
'''

Enter fullscreen modeExit fullscreen mode

Extract titles from SerpApi Blog.

Testing .post-card-title CSS selector in Devtools Console:

$$(".post-card-title")

(7) [h2.post-card-title, h2.post-card-title, h2.post-card-title, h2.post-card-title, h2.post-card-title, h2.post-card-title, h2.post-card-title]
0: h2.post-card-title
1: h2.post-card-title
2: h2.post-card-title
3: h2.post-card-title
4: h2.post-card-title
5: h2.post-card-title
6: h2.post-card-title
length: 7
[[Prototype]]: Array(0)

Enter fullscreen modeExit fullscreen mode

Code:

import requests, lxml
from bs4 import BeautifulSoup

html = requests.get("https://serpapi.com/blog/")
soup = BeautifulSoup(html.text, "lxml")

for title in soup.select(".post-card-title"):
    print(title.text)

'''
Scrape Google Carousel Results with Python
SerpApi’s YouTube Search API
DuckDuckGo Search API for SerpApi
Extract all search engines ad results at once using Python
Scrape Multiple Google Answer Box Layouts with Python
SerpApi’s Baidu Search API
How to reduce the chance of being blocked while web scraping search engines
'''

Enter fullscreen modeExit fullscreen mode

Extract title, link from dev.to feed.

Test CSS selector with either SelectorGadget or DevTools Console:

Code:

import requests, lxml
from bs4 import BeautifulSoup

headers = {
    "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/96.0.4664.45 Safari/537.36"
}

html = requests.get("https://dev.to/", headers=headers)
soup = BeautifulSoup(html.text, "lxml")

for result in soup.select(".crayons-story__title"):
    title = result.text.strip()
    link = f'https://dev.to{result.a["href"].strip()}'

    print(title, link, sep="\n")


# part of the output:
'''
How to Create and Publish a React Component Library
https://dev.to/alexeagleson/how-to-create-and-publish-a-react-component-library-2oe
A One Piece of CSS Art!
https://dev.to/afif/a-one-piece-of-css-art-225l
Windster - Tailwind CSS admin dashboard interface [MIT License]
https://dev.to/themesberg/windster-tailwind-css-admin-dashboard-interface-mit-license-3lb6
'''

Enter fullscreen modeExit fullscreen mode

Outro

In the further blog post, I'll cover XPath thing.

CSS selectors are pretty easy and straightforward to understand, just a matter of practice and trial and error (programming in a nutshell 💻)

If you have anything to share, any questions, suggestions, or something that isn't working correctly, feel free to drop a comment in the comment section or reach out via Twitter at @dimitryzub, or @serp_api.

Yours,
Dimitry, and the rest of SerpApi Team.

Join us on Reddit | Twitter | YouTube

Intro

Prerequisites

Explanation

SelectorGadget

Uses cases:

Pick CSS Selectors by Hand

Test CSS Selectors

Cons of CSS Selector

Code Examples

Extract title, snippet, link, displayed link from Google Search results.

Extract titles from SerpApi Blog.

Extract title, link from dev.to feed.

Links

Outro

Recommend

Build your own branded short links with GraphQL & Serverless Functions

Summary • 1729

Tutorial: How To Integrate A Bubble App With A Backendless Backend

what is partial hydration and why is everyone talking about it?

GitHub Actions - All the Shells

Betting on Things That Never Change · Collaborative Fund

Renesas introduces sub 50 cents FPGA family with free Yosys-based development to...

5 Examples of the Importance of Context-Sensitivity in Data-Centric AI

Cross-fading any two DOM elements is currently impossible

What You Need to Break into DevOps and SRE

About Joyk