#!/usr/bin/env python3
"""Create a random passphrase from a dictionary of words. BETA

Relies on the entropy of python's
    random.SystemRandom class
        which (according to the documentation) calls os.urandom()
        which (according to the documentation) calls the operating system
           specific randomness source which "should be unpredictable
            enough for cryptographic applications"

Requires:
  * Python v>=3.2
  * a dictionary, Ding's trans-de-en by default.
    E.g. on a Debian/Ubuntu system in package "trans-de-en".
    or from http://ftp.tu-chemnitz.de/pub/Local/urz/ding/de-en/

Uses a hardcoded filepath and language.
Search for **customize** below to change it.

Related: There is a Go implementation started by Sascha L. Teichmann at
  https://bitbucket.org/s_l_teichmann/ppgen


Copyright 2016, 2017, 2018 by Intevation GmbH.
Author: Bernhard E. Reiter <bernhard@intevation.de>

This file is Free Software under the Apache 2.0 license and thus
comes without any warranty (to extend permissible under applicable law).
"""

import argparse
import math
import re
import sys

from random import SystemRandom
_srandom = SystemRandom()

tainted = False   # to be set if we find a hint that the passphrase may be weak


def buildDictionary(options):
    """Build up a dictionary of unique words, calculate stats."""
    global tainted
    d = []

    # dictionary for testing
    #d = ["abc", "aBc", "cde", "efg", "hij", "blubber",
    #      "jikf", "zug", "lmf", "opq"]
    # second test dictionary to show that different string functions are used.
    #d =  [''.join('A' * 1000) for _ in range(1000)]

    # Using the dictionary from Ding **customize**
    d = readDingDict(options,
                     filename="/usr/share/trans/de-en",
                     useLeft=not options.second_language)

    # for debugging purposes, dump dictionary
    if options.ddump_filename:
        print("Writing out dictionary in '{}'.".format(options.ddump_filename))
        with open(options.ddump_filename, "w") as f:
            for i in d:
                f.write("{}\n".format(i))

    # Print some stats on the dictionary to be used
    dl = len(d)
    if not options.just_passphrase:
        print("Found {:d} dictionary entries.".format(dl))
        print("|= Number of words |= possibilities |")
        for i in range(1, 5):
            print("|               {:2d} |    2^{:4.1f}      |".format(
                  i, math.log(dl**i, 2)))

    if dl < 8000:
        sys.stderr.write("!Your dictionary is below 8k entries, "
                         "that is quite small!\n")
        tainted = True
    return d


def readDingDict(options, filename="/usr/share/trans/de-en", useLeft=False):
    """Read dictionary with unique words from file in Ding format.

    useLeft: Boolean to control which language to use

    TODO: add option to use both languages for people that speak them both?
    """

    dset = set()  # using the datatype 'set' to avoid duplicates

    splitter = re.compile(r"""\ \|\  # first pattern  ' | '
                           |;\       # second pattern '; '
                           |(?<=\S)/(?=\S)  # 3.:    '/' surrounded by chars
                           |\s+      # by whitespace
                           """, re.VERBOSE)

    if not options.just_passphrase:
        print("Reading entries from {}.".format(filename), end='')
    counter = 0  # for progress or stopping early
    with open(filename, "r") as f:
        for line in f:
            if line[0] == '#':
                continue

            # languages are separated by " :: "
            p = line.partition(" :: ")
            languageEntry = p[0] if useLeft else p[2]

            for word in splitter.split(languageEntry):
                word = word.strip('(",.)\'!:;<>').rstrip('/')
                if len(word) > 2 and not word[0] in '[{/':
                    dset.add(word)

            #TODO: check for very common words and remove them?

            counter += 1
            ## stop early when debugging
            #if counter > 10: break
            if not options.just_passphrase and counter % 10000 == 0:
                    print('.', end='')
                    sys.stdout.flush()
        if not options.just_passphrase:
            print()

    return list(dset)


def main():
    global tainted

    parser = argparse.ArgumentParser(
        description=__doc__.splitlines()[0],
        formatter_class=argparse.ArgumentDefaultsHelpFormatter)
    parser.add_argument('-n', '--number-of-words', type=int, default=4,
                        help='how many words to draw for the passphrase, '
                             'most useful with -j')
    parser.add_argument('-j', '--just-passphrase', action="store_true",
                        help='only output the passphrase on a single line')
    parser.add_argument('--ddump-filename',
                        help='filename to dump the dictionary to')
    parser.add_argument('-2', '--second-language', action='store_true',
                        help='use the second language in the ding dictionary')
    options = parser.parse_args()

    dictionary = buildDictionary(options)

    how_many = options.number_of_words

    output_string = ""
    if not options.just_passphrase:
        print("\nGenerated passphrase with {}"
              " randomly selected words:\n".format(how_many))
        print("    ", end='')
        separator = '\n    '
    else:
        separator = ' '

    # use a dictionary `words` with lower cased words for a rudimentary check
    words = {}
    for x in range(how_many):
        word = _srandom.choice(dictionary)
        words[word.lower()] = True
        output_string += word + separator

    print(output_string)

    if len(words) < how_many:
        sys.stderr.write("! You've drawn a word more than once, this means:\n"
                         "! Your random generation is weak"
                         " or you are being very lucky.\n")
        tainted = True

    if tainted:
        sys.exit("!!! Don't use the resulting passphrase !!!")

if __name__ == "__main__":
    main()