To process a list of Google News URLs, iterate through the list and call gnewsdecoder for each item. You can also pass a proxy and an interval to manage rate limiting during batch processing.
from googlenewsdecoder import gnewsdecoder
source_urls = [
"https://news.google.com/read/URL_1",
"https://news.google.com/read/URL_2"
]
proxy = "http://user:pass@localhost:8080"
for url in source_urls:
try:
decoded_url = gnewsdecoder(url, interval=1, proxy=proxy)
if decoded_url.get("status"):
print("Decoded URL:", decoded_url["decoded_url"])
else:
print("Error:", decoded_url["message"])
except Exception as e:
print(f"Error occurred: {e}")
from googlenewsdecoder import gnewsdecoder
def main():
interval_time = 1
proxy = "http://user:pass@localhost:8080"
source_urls = [
"https://news.google.com/read/CBMilgFBVV95cUxOM0JJaFRwV2dqRDk5dEFpWmF1cC1IVml5WmVtbHZBRXBjZHBfaUsyalRpa1I3a2lKM1ZnZUI4MHhPU2sydi1nX3JrYU0xWjhLaHNfU0N6cEhOYVE2TEptRnRoZGVTU3kzZGJNQzc2aDZqYjJOR0xleTdsemdRVnJGLTVYTEhzWGw4Z19lR3AwR0F1bXlyZ0HSAYwBQVVfeXFMTXlLRDRJUFN5WHg3ZTI0X1F4SjN6bmFIck1IaGxFVVZyOFQxdk1JT3JUbl91SEhsU0NpQzkzRFdHSEtjVGhJNzY4ZTl6eXhESUQ3XzdWVTBGOGgwSmlXaVRmU3BsQlhPVjV4VWxET3FQVzJNbm5CUDlUOHJUTExaME5YbjZCX1NqOU9Ta3U?hl=en-US&gl=US&ceid=US%3Aen",
"https://news.google.com/read/CBMiiAFBVV95cUxQOXZLdC1hSzFqQVVLWGJVZzlPaDYyNjdWTURScV9BbVp0SWhFNzZpSWZxSzdhc0tKbVlHMU13NmZVOFdidFFkajZPTm9SRnlZMWFRZ01CVHh0dXU0TjNVMUxZNk9Ibk5DV3hrYlRiZ20zYkIzSFhMQVVpcTFPc00xQjhhcGV1aXM00gF_QVVfeXFMTmtFQXMwMlY1el9WY0VRWEh5YkxXbHF0SjFLQVByNk1xS3hpdnBuUDVxOGZCQXl1QVFXaUVpbk5lUGgwRVVVT25tZlVUVWZqQzc4cm5MSVlfYmVlclFTOUFmTHF4eTlfemhTa2JKeG14bmNabENkSmZaeHB4WnZ5dw?hl=en-US&gl=US&ceid=US%3Aen"
]
for url in source_urls:
try:
decoded_url = gnewsdecoder(url, interval=interval_time, proxy=proxy)
if decoded_url.get("status"):
print("Decoded URL:", decoded_url["decoded_url"])
else:
print("Error:", decoded_url["message"])
except Exception as e:
print(f"Error occurred: {e}")
if __name__ == "__main__":
main()