Untitled

# WikiStance. A Wikipedia distance meter.
# Based on the alt text of http://xkcd.org/903/
# Gets a Wikipedia URL and measure the distance of this page to the Philosophy article, clicking on links not in parens
# neither italics
#
# Author::    Alejandro Fernández  (mailto:antarticonorte@gmail.com)
# Copyright:: Copyright (c) 2011 Alejandro Fernández
# License::   GPL
#
# =Usage=
#
# require 'wikistance'
#
# url = 'http://en.wikipedia.org/wiki/Scrubs_%28TV_series%29'
# ws = WikiStance.new(url)
# ws.trace  # Go through all the pages until we reach Philosophy
# ws.distance # => 22
# ws.breadcrumbs # => ["List of characters on Scrubs", "NBC", "United States", ..., "Philosophy"]

require 'rubygems'
require 'mechanize'

class WikiStance

  attr_reader :title, :breadcrumbs

  def initialize(url)

    if url =~ /^http:\/\/en\.wikipedia\.org\/wiki\/(.*?)/
      @url = url

      # Wikipedia returns 403 with the default user agent
      @agent = Mechanize.new
      @agent.user_agent_alias = 'Mac Safari'

      self.reset

    else
      raise ArgumentError, "You should use a valid wikipedia link"
    end

  end

  # Resets the class
  def reset
    @page        = @agent.get(@url)
    @breadcrumbs = []
    @title       = page_title
    @breadcrumbs << @title
    true
  end

  # Gets the current @page title
  def page_title
    @page.at('#firstHeading').text()
  end

  # Go through the pages to calculate distance
  def trace

    while page_title != 'Philosophy'
      click_first_link
      title = page_title

      # Avoid entering in an infinite loop
      if @breadcrumbs.include?(title)
        raise "We are repeating ourselves! We already visited \"#{title}\""
      end
      @breadcrumbs << title
    end

    true

  end

  def distance
    # Breadcrumbs hold the initial page. If we start in philosophy the distance should be 0
    @breadcrumbs.length - 1
  end

  private

  def click_first_link
    first_link = nil

    # div#bodyContent is where wikipedia shows article's content
    # The starting text is direct child of div#bodyContent. This way we avoid <p> inside TOCs and other texts.
    # We also avoid Disambiguation and other wikipedia texts, (which all of them contains links in italics) because
    # they are in <div> instead of <p>
    @page.search('#bodyContent > p').each do |p|

      # Links between parens should not be clicked
      # I tried using a regex with lookbehind to know if a link has an opening parenthesis before, but ruby doesn't
      # support them, so I will just remove all text between parens...
      text = p.to_html.gsub(/\((?:.*?)\)/, '').gsub(/<i>(?:.*?)<\/i>/, '')

      # ...and then get the first link.
      first_link = text.match(/<a(?:.*?)href\=\"[^#](.*?)\"(?:.*?)\/a>/)
      break unless first_link.nil?
    end

    raise "Oops! seems that \"#{page_title}\" has no links" if first_link.nil?
    @page = @page.links_with(:href => /#{first_link[1]}/).first.click
  end


end