xml link extraction.rb

# demonstrate how to extract schema locations and stylesheets from an xml document
# also show how to resolve namespaces against pre determined schema implementations

# references used: 
# http://libxml.rubyforge.org/rdoc/index.html
# http://www.w3.org/TR/xml-stylesheet/

require 'libxml'
require 'enumerator'

include LibXML

# expect an xml file
f = ARGV[0]
doc = XML::Parser.file(f).parse

# extract the stylesheet links

stylesheets = doc.find("//processing-instruction()[name()='xml-stylesheet']").map do |pi|
  
  if pi.content =~ /href="(.+?)"/
    reference = $1
    
    if pi.content =~ /type="text\/xsl"/
      reference
    else
      raise 'xml-stylesheet processing instruction is not of type xsl'
    end
    
  else
    raise 'no href in xml-stylesheet processing instruction'
  end
  
end

puts "style sheets used:"
puts stylesheets
puts

# extract namespaces used in entire document
namespaces = doc.find("//*").map do |node| 
  node.namespaces.map { |ns| ns.href } 
end.flatten.uniq

puts "namespaces used:"
puts namespaces
puts

# extract schema location elements, last reached overwrites.
slocs = {}
doc.find("//@xsi:schemaLocation", 'xsi' => 'http://www.w3.org/2001/XMLSchema-instance').each do |sl|
  sl.value.strip.split.each_slice(2) { |ns, url| slocs[ns] = url }
end

puts "schema locations declared:"
slocs.each do |ns, url|
  puts "#{ns} - #{url}"
end
puts