nature_parser.rb

nature_parser.rb

# -*- coding: utf-8 -*-

require "open-uri"
require "nokogiri"
require "ap"

class NatureParser
  def initialize(page_url)
    @nkgr_main = Nokogiri::HTML(open(page_url))
  end
  
  def abstract
    @nkgr_main.css("#first-paragraph p")
  end
  
  def main_text_nodeset
    @nkgr_main.css("#main .content p").select{|n| n.parent.attr("class") == "content" }
  end
  
  def methods_nodeset
    @nkgr_main.css("#methods .content").children.select{|n| ["h2","p"].include?(n.name)}
  end
end

def exclude_sup(nodeset)
  nosup = nodeset.map{|p| p.children.select{|n| !(n.name == "sup" && n.child.name == "a")} }.flatten
  nosup.map{|n| n.inner_text }.join
end

if __FILE__ == $0
  url = "http://www.nature.com/nature/journal/vaop/ncurrent/full/nature10750.html"
  np = NatureParser.new(url)
  [np.abstract, np.main_text_nodeset, np.methods_nodeset].each do |nodeset|
    puts exclude_sup(nodeset)
  end
end