Ruby Regex Examples / Scrubbing stuff out of Wordpress export files

wordpress_cleanup.rb

require 'nokogiri'

print "WP XML filename: "
filename = gets.chomp
contents = File.open(filename).read

# This is purely for __client__001, probably not useful
# for a lot of exports
leadgen_regex = /\r?\n?\r?\n?<a[^>]*cr=LG.*\s*\r?\n?\s*\]\]/
leadgen_modules = contents.scan(leadgen_regex)
puts "Removing Lead-gen Modules from the bottom of posts. Their sizes should be ~170 - 390:"
puts leadgen_modules.map(&:size).inspect
contents.gsub!(leadgen_regex, "]]")

puts "Removing alignments from images"
align_regex = /(alignright[ ]?|alignleft[ ]?|alignnone[ ]?)/
contents.gsub!(align_regex, "")


puts "Removing specified widths above 500, this takes a while."
                        # el-name ... width attribute
specified_width_regex = /<(\w*) [^>]*(width=".{2,7}")/
widths_removed = 0
attrs_to_remove = []
contents.scan(specified_width_regex) do |element, width_attr|
  width = /"(.*)"/.match(width_attr)[1]
  if width && width.to_i > 400
    attrs_to_remove << width_attr
  end
end

attrs_to_remove.each do |width_attr|
  # slow, yes
  # but safer not to remove <400px widths
  puts width_attr
  contents.sub!(width_attr, "")
  widths_removed += 1
end

puts "Number of removed width attributes: #{ widths_removed.to_s }"

puts "Adding tag 'homepage' to every post"
# This is purely for __client__001, probably not useful
# for a lot of exports
homepage_tag_string = "\n\t\t<category domain=\"post_tag\" nicename=\"homepage\"><![CDATA[homepage]]></category>\n"
contents.gsub!(/(is_sticky.*)(\n)(.*<category.*("post_tag"|"category"))/) do |match|
  puts "$1 = #{$1.inspect} || $3 = #{$3.inspect}"
  puts "$~.to_a = #{$~.to_a}"
  $1 + homepage_tag_string + $3
end

puts "Removing alt tags"
contents.gsub!(/alt="[^"]*"/, "")

print "Testing XML integrity..."
doctest = Nokogiri.XML(contents)
puts " ... OK! probably!" unless doctest.nil?
puts " BAD!" if doctest.nil?

print "Pick a filename to save the new XML: "
new_filename = gets.chomp

File.open(new_filename, 'w') do |f|
  f.write(contents)
end