Alternative method for creating anagram keys. In practice seems to be faster to use something like ''.join(sorted(word.upper()))

AnagramKey.py

from operator import mul

# indices 0 - 64 are non-word characters
# followed by indices corrolating to capital letters
# followed by 6 non-word characters
# followed by indices corrolating to lower-case letters
primes = [1] * 65 \
         + [2, 3, 5, 7, 11, 13, 17, 19, 23, 29, 31, 37, 41, 43, 47, 53, 59, 61, 67, 71, 73, 79, 83, 89, 97, 101] \
         + [1] * 6 \
         + [2, 3, 5, 7, 11, 13, 17, 19, 23, 29, 31, 37, 41, 43, 47, 53, 59, 61, 67, 71, 73, 79, 83, 89, 97, 101]

# Non-word characters will not change the anagram key (1 is the multiplicative identity) and are effectively ignored.
# The primes list is duplicated for upper and lower-case characters, so case is also ignored.

def anagram_key(word):
  # Get the ordinal or "character code" for each character in the word
  # eg. "abc" becomes [97, 98, 99]
  word_ordinals = map(ord, word)
  # Map each of these ordinals to a number in the prime list
  word_primes = [primes[i] for i in word_ordinals]
  # The product of these primes will be unique to those primes (number theory),
  # However, because the order does not matter with respect to multiplication
  # (commutative property of multiplication) we can say that this key will be
  # unique to this word AND all of its anagrams.
  product_of_primes = reduce(mul, word_primes)
  return product_of_primes