parse_coveritlive_transcript.rb

require 'rubygems'
require 'hpricot'
require 'htmlentities'
require 'mustache'
require 'pp'

def clean_string(string)
  coder = HTMLEntities.new
  clean = coder.decode(string)
  clean.gsub!("\302\240", ' ')
  clean.gsub!(/ +/, ' ')
  clean.strip
end

def clean_text(text)
  twitpic_links = (text/'//a').select { |a| a['onclick'] =~ %r[twitpic.com/show] }
  twitpic_links.each { |l| l.swap(l.inner_html) }
  (text/'//div/img[@class="correct-png"]').remove
  text.inner_html
end

def parse_transcript(filename)
  doc = Hpricot(open(filename))
  messages = (doc/"//div/table/tr").map do |elem|
    text = (elem/"//td/span").detect { |e| e.attributes['id'] =~ /^txt/ }
    time = (elem/"//span[@class='bottomtime']").inner_html
    date = (elem/"//span[@class='bottomdate']").inner_html
    user = (elem/"//span[@class='bottomdisplayname']").first
    if (text and date.any?)
      username = user.nil? ? "Host" : user.inner_html.strip
      source = (elem.parent.parent['class'] =~ /writercomment/) ? 'host' : 'audience'
      data = {
        :class => source,
        :text => clean_text(text), 
        :time => clean_string(time), 
        :date => clean_string(date), 
        :user => username
      }
      data
    else
      p elem.inner_text
    end
  end
  messages.compact!

  template = %[
  <ul>
  {{#messages}}
    <li class="{{class}}">
      <span class="msg">{{{text}}}</span>
      <span class="details">
        <span class="user">{{user}}</span>
        <span class="time">{{time}}</span>
      </span>
    </li>  
  {{/messages}}
  </ul>
  ]
  Mustache.render(template, :messages => messages.reverse)
end

puts parse_transcript(ARGV[0])