How to: archive data via an API using Ruby and MongoDB

I was going to title this post “How to: archive a FriendFeed feed in MongoDB”. The example code does just that but (a) I fear that this blog suggests a near-obsession with FriendFeed (see tag cloud, right sidebar) and (b) the principles apply to any API that returns JSON. There are rare examples of biological data with JSON output in the wild, e.g. the ArrayExpress Gene Expression Atlas. So I’m still writing a bioinformatics blog ;-)

Let’s go straight to the code:

#!/usr/bin/ruby

require "rubygems"
require "mongo"
require "json/pure"
require "open-uri"

# db config
db  = Mongo::Connection.new.db('friendfeed')
col = db.collection('lifesci')

# fetch json
0.step(9900, 100) {|n|
  f = open("http://friendfeed-api.com/v2/feed/the-life-scientists?start=#{n}&num=100").read
  j = JSON.parse(f)
  break if j['entries'].count == 0
  j['entries'].each do |entry|
    if col.find({:_id => entry['id']}).count == 0
      entry[:_id] = entry['id']
      entry.delete('id')
      col.save(entry)
    end
  end
  puts "Processed entries #{n} - #{n + 99}", "Database contains #{col.count} documents."
}

puts "No more entries to process. Database contains #{col.count} documents."

Also available as a gist. Fork away.

A quick run-through. Lines 4-6 load the required libraries: mongo (the mongodb ruby driver), json and open-uri. If you don’t have the first two, simply “gem install mongo json_pure”. Of course, you’ll need to download MongoDB and have the mongod server daemon running on your system.

Lines 9-10 connect to the database (assuming a standard database installation). Rename the database and collection as you see fit. Both will be created if they don’t exist.

The guts are lines 12-25. A loop fetches JSON from the FriendFeed API, 100 entries at a time (0-99, 100-199…) up to 9999. That’s an arbitrarily-high number, to ensure that all entries are retrieved. Change “the-life-scientists” in line 14 to the feed of your choice. The JSON is then parsed into a hash structure. In lines 17-23 we loop through each entry and extract the “id” key, a unique identifier for the entry. This is used to create the “_id” field, a unique identifier for the MongoDB document. If a document with _id == id does not exist we create an _id key in the hash, delete the (now superfluous) ‘id’ key and save the document. Otherwise, the entry is skipped.
At some point the API will return no more entries: { “entries” : [] }. When this happens, we exit the block (line 16) and print a summary.

That’s it, more or less. Obviously, the script would benefit from some error checking and more options (such as supplying a feed URL as a command line option). For entries with attached files, the file URL but not the attachment will be saved. A nice improvement would be to fetch the attachment and save it to the database, using GridFS.

Possible uses: a simple archive, a backend for a web application to analyse the feed.

Reblog this post [with Zemanta]