find htmls which has too big images. # ruby this.rb >> output.csv

find-html-has-too-big-images.rb

require 'uri'

def find_html(img_name, html_arr)
  img_basename = File.basename(img_name)
  html_arr.each do |file_name|
    htmlname = "http://www." + File.dirname(file_name) + "/" + File.basename(file_name)  
    open(file_name) do |f|
      src = f.read
      if /(#{img_basename})/ =~ src
        puts "#{htmlname}, #{img_name}"
      end
    end
  end
end

html_arr = Dir.glob("path/to/**/*.html")
img_arr = Dir.glob("path/to/**/*.jpg")

img_arr.each do |img|
  if File.size(img) > 500000 #500K
    find_html(img, html_arr)
  end
end