gistfile1.txt

#!/usr/bin/env ruby -wKU
require "rubygems"
require "hpricot"

class Crawler
  def initialize(site_uri='theonlyjames.com',
                 wget_path='/usr/local/bin/wget')
    @site_uri = site_uri
    @wget_path = wget_path
    sanity_check
  end
  
  def sanity_check
    if @wget_path == '' && %x(which wget).length == 0
      raise "wget not in $PATH. Please specify path to wget manually.\n\nCrawler.new('site_uri', '/path/to/wget')"
    end
  end
  
  def wget_sites(site_uris=[])
    site_uris.each { |s| wget(s) }
  end
  
  def wget(s=@site_uri)
    $stdout.puts("Running #{@wget_path} -rkpE '#{s}'")
    %x(#{@wget_path} -rk '#{s}')
    fix_php_css
  end
  
  def fix_php_css
    html_files = Array.new
    Dir.glob("./#{@site_uri}/*").each do |f|
       html_files += [f] if f =~ /.html/
    end
    html_files.each do |f|
      puts "Parsing #{File.basename(f)}"
      
      doc = Hpricot(File.read(f))
      # (doc/'/html/head/link[@rel=stylesheet]').each do |link|
      #   puts "\t#{link['href'].gsub('.php', '_php').gsub('?', '_').gsub('&', '-')}"
      # end
      (doc/'/html/head/link[@rel=stylesheet]').each do |link|
        link.set("href", "new_href.css")}
      end
      
    end
  end
end

c = Crawler.new()
c.fix_php_css
# c.wget