class Nokogiri::XML::FragmentHandler

:nodoc:

def cdata_block string 

def cdata_block string
  @stack.last << CDATA.new(@document, string)
end

def characters string 

def characters string
  @doc_started = true
  @stack.last << Text.new(string, @document)
end

def comment string 

def comment string
  @stack.last << Comment.new(@document, string)
end

def end_element name 

def end_element name
  return unless @stack.last.name == name
  @stack.pop
end

def initialize node, original_html 

def initialize node, original_html
  @doc_started    = false
  @document       = node.document
  @stack          = [node]
  @html_eh        = node.kind_of? HTML::DocumentFragment
  @original_html  = prepare_for_regex(original_html.strip)
end

def prepare_for_regex(string)


you're curious: http://gist.github.com/115936
this implementation choice was the result of some benchmarks, if

start-of-doc. so let's only save up to the first newline.
start-of-line, but we really only want them to anchor at
the regexes used in start_element() and characters() anchor at
def prepare_for_regex(string)
  (newline_index = string.index("\n")) ? string.slice(0,newline_index) : string
end

def start_element name, attrs = [] 

def start_element name, attrs = []
  regex = @html_eh ? %r{^\s*<#{Regexp.escape(name)}}i :
                     %r{^\s*<#{Regexp.escape(name)}}
  if ! @doc_started && @original_html =~ regex
    @doc_started = true
  end
  return unless @doc_started
  ns = nil
  if @document.root
    match = name.match(QNAME_REGEX)
    if match
      prefix, name = match[1], match[2]
      ns = @document.root.namespace_definitions.detect { |x|
        x.prefix == prefix
      }
    end
  end
  node = Element.new(name, @document)
  attrs << "" unless (attrs.length % 2) == 0
  Hash[*attrs].each do |k,v|
    node[k] = v
  end
  node.namespace = ns if ns
  @stack.last << node
  @stack << node
end