Check xml sitemap for 404s

lint-sitemap.rb

require 'httparty'
require 'rexml/document'
require 'logger'

class Sitemap
  include HTTParty
  base_uri 'http://example.com'
  format :xml
end

class Codes < Hash
  def initialize
    super
  end
  def push(status_code)
    if self.has_key?(status_code)
      self[status_code] += 1
    else
      self[status_code] = 1
    end
  end
  def get_codes
    results = 'HTTP Request Results : '
    self.each{|k,v| results += "#{k} [#{v}] "}
    return results
  end
end

usage = 'usage: ruby lint-sitemap.rb <XML file>'

if ARGV.length != 1 then puts usage; exit; end

  infile_handle = ARGV[0].split(/\.|\//)

  if infile_handle[2] != 'xml'
    puts "#{ARGV[0]} is not a valid XML file"
    puts usage
    exit
  end


logfile = "#{infile_handle[1]}.log"
Dir.chdir(File.join(__dir__,"logs"))
$log = Logger.new(logfile)

#parse
Dir.chdir(__dir__)
xml = File.new(ARGV[0])
doc = REXML::Document.new xml
status_codes = Codes.new

#select all baron elements
doc.elements.each("//*[not(*)]") {|node|
  url = node.text
  if url.start_with?('http')
    begin
      status = Sitemap.get(url).code
      printf("%-10s %10s\n", status, url)
    rescue URI::InvalidURIError
      $log.error "#{url} from #{ARGV[0]} is a bad URI"
      next
    end
    if status != 200
        $log.warn "#{status} #{url}"
    end
    status_codes.push(status)
  end
}

puts status_codes.get_codes
$log.info status_codes.get_codes