Scraping AJAX web pages (Part 3)
Don’t forget to check out the rest of the series too!
In Part 2 we saw how to download an Ajax-powered webpage. However, there was a problem with that approach: sometimes it terminated too quickly, thus it fetched just part of a page. The problem with Ajax is that we cannot tell for sure when a page is completely downloaded.
So, the solution is to integrate some waiting mechanism in the script. That is, we need the following: “open a given page, wait X seconds, then get the HTML source”. Hopefully all Ajax calls will be finished in X seconds. It is you who decides how many seconds to wait. Or, you can analyze the partially downloaded HTML and if something is missing, wait some more.
Here I will use Splinter for this task. It opens a browser window that you can control from Python. Thanks to the browser, it can interpret Javascript. The only disadvantage is that the browser window is visible.
Example
Let’s see how to fetch the page CP002059.1. If you open it in a browser, you’ll see a status bar at the bottom that indicates the download progress. For me it takes about 20 seconds to fully get this page. By analyzing the content of the page, we can notice that the string “ORIGIN” appears just once, at the end of the page. So we’ll check its presence in a loop and wait until it arrives.
#!/usr/bin/env python from time import sleep from splinter.browser import Browser url = 'http://www.ncbi.nlm.nih.gov/nuccore/CP002059.1' def main(): browser = Browser() browser.visit(url) # variation A: while 'ORIGIN' not in browser.html: sleep(5) # variation B: # sleep(30) # if you think everything arrives in 30 seconds f = open("/tmp/source.html", "w") # save the source in a file print >>f, browser.html f.close() browser.quit() print '__END__' ############################################################################# if __name__ == "__main__": main()
You might be tempted to check the presence of ‘</html>’. However, don’t forget that the browser downloads a plain source first starting with ‘<html><body>…’ until ‘</body></html>’. Then it starts to interpret the source and if it finds some Ajax calls, they will be called, and these calls will expand something in the body of the HTML. So you’ll have ‘</html>’ right at the beginning.
Future work
This is not bad but I’m still not fully satisfied. I’d like something like this but without any browser window. If you have a headless solution, let me know. I think it’s possible with PhantomJS and/or Zombie.js but I had no time yet to investigate them.
I needed something that watched for the entire page of javascript to settle so I came up with this to hash the entire page and return when the hashes match repeatedly.
Thanks, it’s a nice idea to check the hash value of the HTML source repeatedly. (One remark: you don’t need “;” at the end of lines in Python.)