Ask HN: I want to scrape a web page... | Hacker News Reader

Ask HN: I want to scrape a web page...

6 pointsbdclimber1417 comments

I'm a RoR developer and have a project where I need to scrape a webpage, specifically a University's course list page, to populate a database. I've never scraped a webpage before, are there any pointers or guides that anyone can recommend? I don't even know where to start.

View on HN

If you only need to scrape one University's one course list page to populate one database, your best bet is probably copy-paste with some tidying up with your text editor of choice. If you need to do more, I would look into Hpricot or Nokogiri, both Ruby libraries for parsing HTML. Then just download the page content, pass it to your parser of choice, and go wild.

Second this. I wouldn't say I'm a RoR expert, but I've used both Hpricot and Nokogiri. I found Hpricot was a little easier to get going but Nokogiri ended up working out better for me. They work very in very similar fashions and are fairly straightforward.

well... it's a big University... the largest in the country actually, so there are thousands and thousands of classes. I think copying/pasting may be time prohibitive.

Is it possible to setup a crawler that would iterate through what could be hundreds of pages, with the same HTML structure, and download/parse the page?

Yes. Look up Hpricot or Nokogiri and review the documentation. It will be clear how to get started. Basically you just give it a page and ask it what you want. If you want to handle multiple pages, just iterate through them like you'd iterate through anything else.

Yes, that is exactly what the crawler should do. I had no crawler experience a few months ago but wanted to do the same thing. I didn't know Ruby so I want the Java route. Now it's trivially easy, I just wrote a crawler last night to pull a bunch of info off hundreds of pages for my next app.

First start with how you yourself would get at each page of data. If it all starts from a head page figure out how to get to each link, follow those links and repeat until you are at the data pages. Then point your crawler to the head page and code in the patterns. If the pages are not all linked from a head page but have a url pattern figure out how to pull down the info needed to fill in the url pieces, pull that info down then visit each page filling in the pattern with the data you now have.

I can't give any Ruby guidance but if you have general questions on web scraping you can shoot me an email.

This is awesome... time to dive into hpricot!

Most every option is documented here: http://stackoverflow.com/questions/2861/options-for-html-scr...

You'll likely find this tool handy once you choose your framework: http://www.selectorgadget.com/

Another quick and dirty way: http://googlemapsmania.blogspot.com/2008/10/data-scaping-wik...

I see people are saying use Hpricot or Nokogiri and you have gone on to choose Hpricot. Not a good choice. Hpricot was the work of the hacker _why who has now disappeared. But even before he disappeared nokogiri overtook hpricot in performance. He even tweeted "caller asks, “should i use hpricot or nokogiri?” if you're NOT me: use nokogiri. and if you're me: well cut it out, stop being me".

So please, use nokogiri it's a great library and the only thing I really miss from ruby-land in python.

Just write yourself a Ruby script that downloads a webpage. Then parse the downloaded content with Ruby's equivalent of BeautifulSoup (a forgiving HTML parsing library).

That's pretty much all there is to it. You'll want to manually inspect the HTML from the page you're scraping to find patterns. For example all courses are in <li> tags, etc.

Let me know if you need more help. But I think as a first step write a program to download a web page.

I've never written anything like this... Is there a function like CURL that downloads the page as a string from a URL?

  require 'net/http'
  require 'uri'
  output = Net::HTTP.get(URI.parse("http://www.google.com"))

curb is a lot faster than net/http

require 'curb' c = Curl::Easy.perform("http://www.google.com) puts c.body_str

Premature optimization much?

So apparently accessing a secure URL isn't straightforward... No login is required, its just an HTTPS but it won't go in curb.

Wasn't an SSL problem but a cookie problem...

I wanted to thank you for your help. I wrote a script to scrape over 30,000 courses and textbooks from Arizona State University successfully in a couple days.

I recently did something similar using anemone to crawl the website, and Hpricot to scrape each individual web page and add to the database.

Anemone is great because it can focus your crawl to only url's that match a certain pattern, which really helps you traverse a small portion of a larger website (like a University site). You can also do specific actions on pages that match a certain pattern.

For scraping, anemone natively supports nokogiri, so since you're coming from a blank slate, it might be easiest to learn nokogiri. Before discovering anemone, I had already written what needed to be done on each page in hpricot, so my code is a bit messy, but it's not that difficult to get anemone and hpricot to work together.

Reply on news.ycombinator.com