pdftest.rb

#!/usr/bin/env ruby
# -*- encoding : utf-8 -*-

# Copied code from mysociety/alaveteli, so LICENSE = https://github.com/mysociety/alaveteli/blob/develop/LICENSE.txt

# We want to avoid loading rails unless we need it, so we start by just loading the
# config file ourselves.
$alaveteli_dir = File.expand_path(File.join(File.dirname(__FILE__), '..'))
$:.push(File.join($alaveteli_dir, "commonlib", "rblib"))
load 'config.rb'
$:.push(File.join($alaveteli_dir, "lib"))
$:.push(File.join($alaveteli_dir, "lib", "mail_handler"))
load 'configuration.rb'
MySociety::Config.set_file(File.join($alaveteli_dir, 'config', 'general'), true)
MySociety::Config.load_default

require 'optparse'

filename = ""

OptionParser.new do |opts|
	opts.banner = "Usage: example.rb [options]"
	opts.on("-fNAME", "--filename=FILENAME", "Filename to process") do |fi|
		filename = fi
	end
end.parse!

pdf_file = open(filename).read()
uncompressed_text = AlaveteliExternalCommand.run("pdftk", "-", "output", "-", "uncompress", :stdin_string => pdf_file)
if uncompressed_text.blank?
	puts "returns 'text'"
else
	text = uncompressed_text.dup
	text.gsub!(MySociety::Validate.email_find_regexp) do |email|
		email.gsub(/[^@.]/, 'x')
	end

	if text != uncompressed_text
		puts "Email Regexp Change"
	end

	text = uncompressed_text.dup
	ascii_chars = text.gsub(/\0/, "")
	emails = ascii_chars.scan(MySociety::Validate.email_find_regexp)

	# Convert back to UCS-2, making a mask at the same time
	emails.map! do |email|
		#We want the ASCII representation of UCS-2
		[email[0].encode('UTF-16LE').force_encoding('US-ASCII'),
		 email[0].gsub(/[^@.]/, 'x').encode('UTF-16LE').force_encoding('US-ASCII')]
	end
	# Now search and replace the UCS-2 email with the UCS-2 mask
	emails.each do |email, mask|
		text.gsub!(email, mask)
	end
	if text != uncompressed_text
		puts "Email Regexp Change #2"
	end
end

puts "All done"