extract_from_file.rb

require "pp"

block_size = 4_096

def tags_from_file(filename)  
  no_tag_or_tag = /([^<>]+)|(<\/[^<>]+>)|(<[^<>]+?\/>)|(<\!\[CDATA\[(?:.*?)\]\]>)|(<[^!a-zA-Z][^<>]+>)|(<[^<>]+>)/
  reg_exp = /\A(?<p>(?:#{no_tag_or_tag}))*\g<p>*/
  block_size = 4_096

  Enumerator.new do |enum|
    File.open(filename, "r:ISO-8859-15") do |f|
      buf = ""
      loop do
        break if not block = f.gets(nil, block_size)

        buf += block

        # porcion que matchea
        match = reg_exp.match(buf).to_s

        # porcion a partir del match
        buf = buf[match.length..-1]
        
        # devolvemos los matchs
        match.scan(/#{no_tag_or_tag}/).each{|m| enum.yield(m)}
      end
    end

  end
end


class Selector
  def initialize selector
    @selector = selector
    @search = selector.split(" ")
  end

  def match? hierarchy
    tags = hierarchy.map{|h| h[:tagname]}
    matching?(tags, @search.dup, true)
  end

  private

  def matching? tags, searched, last=true
    if last
      (tags.pop == searched.pop) && matching?(tags, searched, false)
    elsif searched == []
      true
    else
      search = searched.pop
      i = tags.index(search)
      i && matching?(tags[0...i], searched)
    end
  end

end

def extract_from_file(filename, selectors)
  Enumerator.new do |enum|
    saving_childs = false
    hierarchy = []
    childs = []
    selectors = Array(selectors).map{|s| Selector.new(s)}

    tags_from_file(filename).each do |no_tag, close_tag, inline_tag, cdata, other_tag, tag|
      all_tags = close_tag || inline_tag || tag
      m = /<\/?(\w+)/.match(all_tags)
      m = m[1] if m

      if tag || inline_tag

        last_childs = childs 
        
        childs = []
        node = {tagname: m, content: tag}
        
        last_childs << node if saving_childs
        hierarchy << node

        selectors.each do |selector|
          saving_childs ||= selector.match? hierarchy
        end
        node[:childs] = childs if saving_childs

      end

      if saving_childs
        childs << no_tag if no_tag && !no_tag.strip.empty?
        childs << (/<!\[CDATA\[(.*)\]\]>/.match(cdata)[1]) if cdata
      end

      if close_tag || inline_tag

        selectors.each do |selector|
          if selector.match? hierarchy
            enum.yield hierarchy
            saving_childs = false
          end
        end

        hierarchy.pop
        childs = hierarchy.last[:childs]
      end

    end
  end
end


tags = extract_from_file("Export.xml", ["Feature", "Category", "Hierarchy Link", "FeatureLink", "Object", "Sku", "Pack", "CrossSell Link", "Alternative Link"])


tags.take(10).each do |h|
  puts 
  pp h
end