DICTIONARY_FILE = 'C:\dict.txt' # list of words, one per line

def usage
  puts "passgen.rb LEN [NUM_TO_GENERATE]"
  exit
end

usage if ( ARGV.length != 1 && ARGV.length != 2 )
word_length, word_count = ARGV[0].to_i, (ARGV[1] || 1).to_i
usage if ( word_length < 3 || word_count < 1 )

base = "a"[0]

# If you need performance, hack this code to compute probabilities once and store for reading.
probs = Array.new(27) { |i| Array.new(27) { |j| Array.new(27) { |k| 0 }}}
pen_probs = Array.new(27) { |i| Array.new(27) { |j| Array.new(27) { |k| 0 }}}
last_probs = Array.new(27) { |i| Array.new(27) { |j| Array.new(27) { |k| 0 }}}

File.open( DICTIONARY_FILE ) { |file|
  file.each_line { |line|
    next if line.length < 5
    next if line !~ /^[a-zA-Z]+$/
    prev = [ 26, 26 ]
    line.chomp.downcase.each_byte { |c|
      c -= base
      probs[prev[-2]][prev[-1]][c] += 1
      prev << c
    }
    pen_probs[prev[-4]][prev[-3]][prev[-2]] += 1
    last_probs[prev[-3]][prev[-2]][prev[-1]] += 1
    probs[prev[-4]][prev[-3]][prev[-2]] -= 1
    probs[prev[-3]][prev[-2]][prev[-1]] -= 1
  }
}

#normalize
[ probs, pen_probs, last_probs ].each { |prob_arr|
  prob_arr.each { |second_letter|
    second_letter.each { |third_letter|
      sum = 0
      third_letter.each { |count| sum += count }
      third_letter.map! { |count| count.to_f()/sum }
    }
  }
}

srand( Time.now.to_i )
word_count.times {

  chars = []
  (Array.new(word_length - 2, probs) << pen_probs << last_probs).each { |prob_arr|
    total = 0
    target = rand
    prob_arr[chars[-2] || 26][chars[-1] || 26].each_with_index { |prob,char|
      total += prob
      if ( total >= target )
        chars << char
        break
      end
    }
  }

  (word_length - chars.length).times { chars << rand(26) } #probs occasionally lead to shorter words
  puts chars.map { |c| (c+base).chr }.join
}