from bs4 import BeautifulSoup import requests from typing import Tuple, List from tqdm import tqdm import itertools from multiprocessing import Pool monthLinkClass = "sc-zpw6hx-0" articleLinkClass = "sc-1w8kdgf-0" newsHeadClass = "sc-1efpnfq-0" newsBodyClass = "sc-77igqf-0" videoClass = "lhhce6-0" url = "https://www.theonion.com" # Function to get all links from a page with _url of class _class def get_links(_url: str, _class: str) -> List[str]: """ This function takes in a URL string and a class string and returns a list of all links from a page with that URL and class. Args: _url (str): A string representing the URL of the page to scrape. _class (str): A string representing the class of the div containing the links. Returns: List[str]: A list of strings representing the URLs of the links found. """ # Make a request to the given URL page = requests.get(_url) # This might throw an exception if something goes wrong. # Create a BeautifulSoup object to parse the HTML content of the page soup = BeautifulSoup(page.text, 'html.parser') # Find the div with the given class (using a lambda function to match partial matches) try: link_div = soup.find_all('div', attrs={'class': lambda e: e.startswith(_class) if e else False})[0] except IndexError: raise IndexError(f"Error: {_url}, with {_class}") # Find all the links within the div and extract their URLs links = link_div.findAll('a') links = [link.get('href') for link in links] # Return the list of URLs found return links def extractText(_url: str, _done: bool = True) -> Tuple[str, str]: """ This function takes in a URL string and a boolean flag (default=True) and returns a tuple of two strings: the first is the text of the page's H1 heading with class 'newsHeadClass', and the second is the text of the page's first paragraph with class 'newsBodyClass'. If the function encounters an SSLError or ConnectionError, it will attempt to retry the request with the _done flag set to False (to prevent infinite recursion), and if this also fails, it will return an empty tuple of strings. If the page does not have an H1 heading or first paragraph with the expected class, the function will return an empty tuple of strings. Args: _url (str): A string representing the URL of the page to scrape. _done (bool): A boolean flag indicating whether the function has already attempted to retry the request (default=True). Returns: Tuple[str, str]: A tuple of two strings representing the text of the H1 heading and first paragraph with the expected classes (or empty strings if not found). """ try: # Make a request to the given URL page = requests.get(_url) # This might throw an exception if something goes wrong. except (requests.exceptions.SSLError, requests.exceptions.ConnectionError): # If the request fails due to an SSL error or connection error, and we haven't already # retried the request, try again with the _done flag set to False. if _done: return extractText(_url, _done=False) else: # If we've already retried the request, and it still failed, return an empty tuple of strings. return "", "" # Create a BeautifulSoup object to parse the HTML content of the page soup = BeautifulSoup(page.text, 'html.parser') try: # Find the H1 heading with the expected class head = soup.find_all('h1', attrs={'class': lambda _e: _e.startswith(newsHeadClass) if _e else False})[0].text # Find the first paragraph with the expected class body = soup.find_all('p', attrs={'class': lambda _e: _e.startswith(newsBodyClass) if _e else False})[0].text # If the H1 heading is the same as the body, assume we haven't found the expected elements if head == body: return "", "" else: # Return the text of the H1 heading and first paragraph return head, body except IndexError as e: # If we couldn't find the expected elements, check if there is a video on the page a = soup.find_all('div', attrs={'class': lambda _e: _e.startswith(videoClass) if _e else False}) if not a: # If there is no video, return an empty tuple of strings return "", "" # If there is a video, print an error message and raise the IndexError print(f"Error: {_url}") raise e def batched_extractText(_urls: List[str], _p: Pool) -> List[str]: """ This function takes in a list of URL strings and a multiprocessing Pool object, and returns a list of strings representing the text of the H1 heading and first paragraph for each page. The function uses the multiprocessing Pool object to parallelize the extraction of text from the pages, and returns the results as a list of strings. Args: _urls (List[str]): A list of strings representing the URLs of the pages to scrape. _p (Pool): A multiprocessing Pool object used to parallelize the extraction of text. Returns: List[str]: A list of strings representing the text of the H1 heading and first paragraph for each page. """ # Use the map_async method of the multiprocessing Pool object to parallelize the extraction # of text from the pages. results = _p.map_async(extractText, _urls).get() # Return the results as a list of strings. return results def main() -> None: """ Scrape news article titles and bodies from The Onion website and save them to a file. Returns: None """ # Get the links to the monthly sitemaps from the main page, and print some information about them. monthLinks = get_links(url + "/sitemap", monthLinkClass) print(f"{len(monthLinks)} months have been found.") print(f"Oldest is {monthLinks[-1].replace('/sitemap/', '')}") print(f"and newest is {monthLinks[0].replace('/sitemap/', '')}") # Construct the full URLs for the monthly sitemaps. monthLinks = [url + link for link in monthLinks] # Get the links to the individual articles from the monthly sitemaps, and print some information # about them. articleLinks = [get_links(monthLink, articleLinkClass) for monthLink in tqdm(monthLinks, desc="Months")] articleLinks = list(itertools.chain(*articleLinks)) print(f"{len(articleLinks)} articles have been found.") # Extract the text of the H1 heading and first paragraph for each article, using multiprocessing # to speed up the process. text = [] batch_size = 60 batch_counter = tqdm(range(0, len(articleLinks), batch_size), total=len(articleLinks), desc="Articles") with Pool(batch_size) as p: for x in batch_counter: text += batched_extractText(articleLinks[x:x + batch_size], p) batch_counter.update(batch_size) # Filter out any articles that didn't have both a non-empty heading and a non-empty body text. text = [x for x in text if x != ("", "")] # Write the text of each article to a file. with open("onion/NewsWebScrape.txt", mode="w", encoding="utf-8") as f: for article in text: if article: f.write(f"{article[0]} #~# {article[1]}\n") # Print some information about the number of articles found and written to file. print(f"{len(articleLinks)} articles where found, and {len(text)} articles where written to file.") if __name__ == "__main__": main()