- Revisiting BeautifulSoup and Requests
- Is This Website Dynamic or Static?
- Can BeautifulSoup Render
JavaScript? - Scraping Dynamic Web Pages With Selenium
- Web Scraping Dynamic Sites by Locating AJAX Calls
Web scraping most of the websites may be comparatively easy. This topic is already covered at length in this tutorial. There are many sites, however, which can not be scraped using the same method. The reason is that these sites load the content dynamically using JavaScript.
This technique is also known as AJAX (Asynchronous JavaScript and XML). Historically, this standard was included creating an XMLHttpRequest object to retrieve XML from a web server without reloading the whole page. These days, this object is rarely used directly. Usually, a wrapper like jQuery is used to retrieve content such as JSON, partial HTML, or even images.
To scrape a regular web page, at least two libraries are required. The requests library downloads the page. Once this page is available as an HTML string, the next step is parsing this as a BeautifulSoup object. This BeautifulSoup object can then be used to find specific data.
Here is a simple example script that prints the text inside the h1 element with id set to firstHeading.
import requests
from bs4 import BeautifulSoup
response = requests.get("https://quotes.toscrape.com/")
bs = BeautifulSoup(response.text,"lxml")
author = bs.find("small",class_="author")
if author:
print(author.text)
## OUTPUT
# Albert EinsteinNote that we are working with version 4 of the Beautiful Soup library. Earlier versions are discontinued. You may see beautiful soup 4 being written as just Beautiful Soup, BeautifulSoup, or even bs4. They all refer to the same beautiful soup 4 library.
The same code will not work if the site is dynamic. For example, the same site has a dynamic version at https://quotes.toscrape.com/js/ (note js at the end of this URL).
response = requests.get("https://quotes.toscrape.com/js") # dynamic web page
bs = BeautifulSoup(response.text,"lxml")
author = bs.find("small",class_="author")
if author:
print(author.text)
## No outputThe reason is that the second site is dynamic where the data is being generated using JavaScript.
There are two ways to handle sites like this.
- Using a tool like Selenium or Puppeteer to open a real browser to render the dynamic web page
- Identify the AJAX links that contain the data, and work with those directly.
These two approaches are covered at length in this tutorial.
However, first, we need to understand how to determine if a site is dynamic.
Here is the easiest way to determine if a website is dynamic using Chrome or Edge. (Both of these browsers use Chromium under the hood).
Open Developer Tools by pressing the F12 key. Ensure that the focus is on Developer tools and press the CTRL+SHIFT+P key combination to open Command Menu.
It will show a lot of commands. Start typing disable and the commands will be filtered to show Disable JavaScript. Select this option to disable JavaScript.
Now reload this page by pressing Ctrl+R or F5. The page will reload.
If this is a dynamic site, a lot of the content will disappear:
In some cases, the sites will still show the data but will fall back to basic functionality. For example, this site has an infinite scroll. If JavaScript is disabled, it shows regular pagination.
 |
 |
|---|---|
| JavaScript Enabled | JavaScript Disabled |
The next question that needs to be answered is the capabilities of BeautifulSoup.
The short answer is no.
It is important to understand the words like parsing and rendering. Parsing is simply converting a string representation of a Python object into an actual object.
So what is Rendering? Rendering is essentially interpreting HTML, JavaScript, CSS, and images into something that we see in the browser.
Beautiful Soup is a Python library for pulling data out of HTML files. This involves parsing HTML string into the the BeautifulSoup object. For parsing, first, we need the HTML as string, to begin with. Dynamic websites do not have the data in the HTML directly. It means that BeautifulSoup cannot work with dynamic websites.
Selenium library can automate loading and rendering websites in a browser like Chrome or Firefox. Even though Selenium supports pulling data out of HTML, it is possible to extract complete HTML and use Beautiful Soup instead to extract the data.
Let's begin dynamic web scraping with Python using Selenium first.
Installing Selenium involves installing three things:
-
The browser of your choice (which you already have):
- Chrome, Firefox, Edge, Internet Explorer, Safari, and Opera browsers are supported. In this tutorial, we will be using Chrome.
-
The driver for your browser:
-
Python Selenium Package:
- This package can be installed using the pip command:
pip install selenium
- If you are using Anaconda, this can be installed from the
conda-forgechannel.
conda install -c conda-forge selenium
The basic skeleton of the Python script to launch a browser, load the page, and then close the browser is simple:
from selenium.webdriver import Chrome
from webdriver_manager.chrome import ChromeDriverManager
driver = Chrome(ChromeDriverManager().install())
driver.get('https://quotes.toscrape.com/js/')
#
# Code to read data from HTML here
#
driver.quit()Now that we can load the page in the browser, let's look into extracting specific elements. There are two ways to extract elements—Selenium and Beautiful Soup.
Our objective in this example is to find the author element.
Load the sitehttps://quotes.toscrape.com/js/ in Chrome, right-click the author name, and click Inspect. This should load Developer Tools with the author element highlighted as follows:
This is a small element with its class attribute set to author.
<small class="author">Albert Einstein</small>Selenium allows various methods to locate the HTML elements. These methods are part of the driver object. Some of the methods that can be useful here are as follows:
element = driver.find_element(By.CLASS_NAME, "author")
element = driver.find_element(By.TAG_NAME, "small")There are few other methods, may be useful for other scenario. These methods are as follows:
element = driver.find_element(By.ID, "abc")
element = driver.find_element(By.LINK_TEXT, "abc")
element = driver.find_element(By.XPATH, "//abc")
element = driver.find_element(By.CSS_SELECTOR, ".abc")Perhaps the most useful methods are find_element(By.CSS_SELECTOR) and find_element(By.XPATH). Any of these two methods should be able to select most of the scenarios.
Let's modify the code so that the first author can be printed.
from selenium.webdriver import Chrome
from selenium.webdriver.common.by import By
from webdriver_manager.chrome import ChromeDriverManager
driver = Chrome(ChromeDriverManager().install())
driver.get('https://quotes.toscrape.com/js/')
element = driver.find_element(By.CLASS_NAME, "author")
print(element.text)
driver.quit()What if you want to print all the authors?
All the find_element methods have a counterpart - find_elements . Note the pluralization. To find all the authors, simply change one line:
elements = driver.find_elements(By.CLASS_NAME, "author")This returns a list of elements. We can simply run a loop to print all the authors:
for element in elements:
print(element.text)Note: The complete code is in selenium_example.py code file.
However, if you are already comfortable with BeautifulSoup, you can create the Beautiful Soup object.
As we saw in the first example, the Beautiful Soup object needs HTML. For web scraping static sites, the HTML can be retrieved using requests library. The next step is parsing this HTML string into the BeautifulSoup object.
response = requests.get("https://quotes.toscrape.com/")
bs = BeautifulSoup(response.text,"lxml")Let 's find out how to scrape a dynamic website with BeautifulSoup.
The following part remains unchanged from the previous example.
from selenium.webdriver import Chrome
from webdriver_manager.chrome import ChromeDriverManager
from bs4 import BeautifulSoup
driver = Chrome(ChromeDriverManager().install())
driver.get('https://quotes.toscrape.com/js/')The rendered HTML of the page is available in the attribute page_source.
soup = BeautifulSoup(driver.page_source, "lxml")Once the soup object is available, all Beautiful Soup methods can be used as usual.
author_element = soup.find("small", class_="author")
print(author_element.text)Note: The complete source code is in selenium_bs4.py
Once the script is ready, there is no need for the browser to be visible when the script is running. The browser can be hidden, and the script will still run fine. This behavior of a browser is also known as a headless browser.
To make the browser headless, import ChromeOptions. For other browsers, their own Options classes are available.
from selenium.webdriver import ChromeOptionsNow, create an object of this class, and set the headless attribute to True.
options = ChromeOptions()
options.headless = TrueFinally, send this object while creating the Chrome instance.
driver = Chrome(ChromeDriverManager().install(), options=options)Now when you run the script, the browser will not be visible. See selenium_bs4_headless.py file for the complete implementation.
Loading the browser is expensive—it takes up CPU, RAM, and bandwidth which are not really needed. When a website is being scraped, it's the data that is important. All those CSS, images, and rendering are not really needed.
The fastest and most efficient way of scraping dynamic web pages with Python is to locate the actual place where the data is located.
There are two places where this data can be located:
- The main page itself, in JSON format, embedded in a
<script>tag - Other files which are loaded asynchronously. The data can be in JSON format or as partial HTML.
Let's look at few examples.
Open https://quotes.toscrape.com/js in Chrome. Once the page is loaded, press Ctrl+U to view source. Press Ctrl+F to bring up the search box, search for Albert.
We can immediately see that data is embedded as a JSON object on the page. Also, note that this is part of a script where this data is being assigned to a variable data.
In this case, we can use the Requests library to get the page and use Beautiful Soup to parse the page and get the script element.
response = requests.get('https://quotes.toscrape.com/js/')
soup = BeautifulSoup(response.text, "lxml")Note that there are multiple <script> elements. The one which contains the data that we need does not have src attribute. Let's use this to extract the script element.
script_tag = soup.find("script", src=None)Remember that this script contains other JavaScript code apart from the data that we are interested in. For this reason, we are going to use a regular expression to extract this data.
import re
pattern = "var data =(.+?);\n"
raw_data = re.findall(pattern, script_tag.string, re.S)The data variable is a list containing one item. Now we can use the JSON library to convert this string data into a python object.
if raw_data:
data = json.loads(raw_data[0])
print(data)The output will be the python object:
[{'tags': ['change', 'deep-thoughts', 'thinking', 'world'], 'author': {'name': 'Albert Einstein', 'goodreads_link': '/author/show/9810.Albert_Einstein', 'slug': 'Albert-Einstein'}, 'text': '“The world as we have created it is a process of our thinking. It cannot be changed without changing our thinking.”'}, {'tags': ['abilities', 'choices'], 'author': {'name': 'J.K. Rowling', .....................This list can not be converted to any format as required. Also, note that each item contains a link to the author page. It means that you can read these links and create a spider to get data from all these pages.
This complete code is included in data_in_same_page.py.
Web scraping dynamic sites can follow a completely different path. Sometimes the data is loaded on a separate page altogether. One such example is Librivox.
Open Developer Tools, go to Network Tab and filter by XHR. Now open this link or search for any book. You will see that the data is an HTML embedded in JSON.
Note few things:
-
The URL displayed by the browser is
https://librivox.org/search/?q=... -
The data is in
https://librivox.org/advanced_search?.... -
If you look at headers, you will find that the advanced_search page is sent a special header
X-Requested-With: XMLHttpRequest
Here is snippet to extract this data:
headers = {
'X-Requested-With': 'XMLHttpRequest'
}
url = 'https://librivox.org/advanced_search?title=&author=&reader=&keywords=&genre_id=0&status=all&project_type=either&recorded_language=&sort_order=alpha&search_page=1&search_form=advanced&q=The%20Time%20Machine'
response = requests.get(url, headers=headers)
data = response.json()
soup = BeautifulSoup(data['results'], 'lxml')
book_titles = soup.select('h3 > a')
for item in book_titles:
print(item.text)The complete code is included in librivox.py file.