#!/usr/bin/python3
"""Generate text using Φαιστός disk characters for fictional alien plants.

This currently only generates characters with a zero-order Zipf
distribution; it doesn’t have even a first-order Markov model of
letter associations.  However, the unreasonable word lengths were
bugging me too much, so I special-cased word separators.  Then I
realized it was going to generate all possible one- and two-letter
words sooner or later, and that isn’t okay.

English, by contrast, only has three one-letter words and about 30
two-letter words, out of the roughly 25 and 625 possible, so we want
something like 1/20 to 1/30 probability of each possible word being
valid.

The obvious way of doing this with a hash function works but is pretty
slow.

After poking the hash functions for a while, this generates 66
distinct two-letter words (with a roughly Zipfian distribution), 2
distinct one-letter words, and almost 2000 distinct three-letter words
(out of 32768 possible).  This is not quite a natural-language
distribution but it's close enough to pass casual inspection.

It generates very Φαιστός-like strings:

𐇞𐇤𐇤𐇚𐇢𐇟𐇚𐇶𐇴𐇺𐇼|𐇳𐇚𐇤𐇳𐇟|𐇳𐇠𐇱𐇳|𐇘𐇦𐇹𐇴𐇟|𐇚𐇟𐇧𐇚|𐇺𐇺𐇚𐇚𐇚𐇳
|𐇴𐇡𐇚|𐇟𐇤𐇺𐇺𐇴|𐇨𐇷𐇚|𐇴𐇚|𐇫𐇟𐇚𐇤|𐇳𐇚𐇺𐇠𐇺|𐇚𐇚𐇺𐇸|𐇟𐇙𐇺𐇚
|𐇚𐇚𐇚|𐇖𐇥|𐇟𐇚𐇶𐇚𐇟𐇟𐇚𐇳|𐇚𐇶𐇤𐇘𐇩|𐇤𐇺|𐇨𐇚𐇠𐇚𐇘𐇷|𐇟𐇟𐇚𐇷𐇺
|𐇴𐇷𐇨𐇟𐇷𐇤𐇵𐇟𐇥𐇚|𐇤𐇚|𐇼𐇙𐇚𐇟𐇳𐇜𐇤|𐇷𐇫𐇙𐇟𐇚𐇴|𐇙𐇻𐇴|𐇲𐇚𐇺𐇺
|𐇴𐇚|𐇜𐇚𐇘𐇤|𐇚𐇘𐇚𐇫

"""
import math
import random
import sys
import textwrap


letters = u'𐇚𐇟𐇺𐇤𐇳𐇘𐇴𐇫𐇥𐇙𐇞𐇩𐇶𐇷𐇼𐇜𐇠𐇵𐇱𐇹𐇗𐇲𐇸𐇡𐇣𐇦𐇧𐇛𐇢𐇻𐇖𐇨'


# With a perfect Zipf distribution, the rank is exactly inversely
# proportional to the frequency.  Here we have 32 characters, so the
# 32nd will be about 32× less common than the 1st, so about 10'000
# items is probably good enough to provide an adequate approximation.

freq_letters = [letter for rank, letter in enumerate(letters)
                for i in range(2500//(1+rank))]


def word_length():
    return int(round(random.lognormvariate(math.log(4), math.log(1.5))))


def word_hash(w):
    "Specify the hash algorithm for filtering so it’s reproducible."
    n = 6
    for c in w:
        n = (n * 0xaaa + ord(c)) % 65521
    return n


def word_candidate(length):
    return ''.join(random.choice(freq_letters) for i in range(length))


def word(length):
    while True:
        w = word_candidate(length)
        if not word_hash(w) & 0xf:
            return w


def text(n):
    "Returns n words of text."
    return '|'.join(word(word_length()) for j in range(n))


def main(n):
    "Prints n words of text, wrapped."
    for line in textwrap.wrap(text(n).replace('|', ' '), width=35):
        print(line.replace(' ', '|'))


if __name__ == '__main__':
    main(int(sys.argv[1]))