Analytics with Capistrano

If you want realtime stats on your web site, why not try using Capistrano to collect data and create graphs and tables? It’s pretty easy to do this, and the beauty is you don’t need to make any modifications to your application.

I’m currently experimenting with marketing Helicoid’s web applications on social networks, so I really wanted a quick way to get information to see the results. Waiting for services like Google Analytics isn’t so bad, but there’s some information locked up inside your application that they can’t get at.

1. Set up a basic Capistrano task

Create a new Capfile and set up your server and SSH details:

role :app, 'yourserver.com'
set :user, 'alex'
set :port, 8321

task :stats, :roles => :app do
end

2. Write some code to extract data

This Capistrano task is going to login to your server and collect data so you can analyse it locally. My example is for downloading user signup counts. I’m just going to get an array of signup dates, then count the dates. If you’re using a Rails application, you could use script/runner to achieve this. Escaping the remote shell command could be messy, so let’s use a heredoc:

def stats_script
  time = '1.month'
  <<-RUBY
    puts User.find(:all, :order => "created_on DESC", :conditions => ["created_on > ?", Time.now - #{time}]).collect { |user| user.created_on }.join("\\n")
  RUBY
end

Notice a double new line is required for the join. Next, run this command using Capistrano’s run with a block. Passing a block allows you to collect the data from Net::SSH:

def get_dates_for(app)
  dates = []
  remote_data = ''    
  remote_command = "/u/apps/#{app}/current/script/runner -e production '#{stats_script app}'" 

  run remote_command do |channel, stream, data|
    remote_data << data
  end

  remote_data.split("\n").collect { |d| d.strip }
end

Add this code to the stats task. Run it with cap stats.

3. Prepare the results

Here’s some ideas for how to display the results:

• Make a graph with Scruffy
• Turn it into CSV using FasterCSV
• Make a HTML report

You can use the script/runner and run with block technique to download any data.