Test out bazzargh's variation of "Shuffling using Fibonacci hashing"

gistfile1.txt

import random
import math
import copy
import numpy as np


def shuffle_songs(songs):
    """Return a list of shuffled songs."""
    num_songs = len(songs)

    golden_ratio = 0.618033988749895
    seed = random.random()
    pos = 1

    while True:
        curr_pos = ((pos * golden_ratio) + seed) % 1
        pos += 1
        to_pick = math.floor(curr_pos * num_songs)
        artist, song = songs[to_pick]
        songs.remove((artist, song))
        num_songs -= 1

        if not songs:
            break

        yield (artist, song)


def clustering_index(shuffled_songs):
    clusters = sum(
        1
        for i in range(len(shuffled_songs) - 1)
        if shuffled_songs[i][1] == shuffled_songs[i + 1][1]
    )
    return clusters / (len(shuffled_songs) - 1)


def get_fuzzy_set():
    max_songs = 10
    max_artists = 10
    fuzzy_set = []

    num_artists = random.randint(4, max_artists)
    for i in range(num_artists):
        num_songs = random.randint(1, max_songs)
        for j in range(num_songs):
            fuzzy_set.append(("artist{}".format(i), "song{}".format(j)))

    return fuzzy_set


if __name__ == "__main__":
    result = []
    N = 1000000
    for i in range(N):
        songs = get_fuzzy_set()
        shuffle = list(shuffle_songs(songs))
        measure = clustering_index(shuffle)
        # if measure > 0.5 and measure < 0.6:
        #     print(shuffle)
        result.append(measure)
    np_arr = np.array(result)
    print("P25:", np.percentile(np_arr, 25))
    print("P50:", np.percentile(np_arr, 50))
    print("P75:", np.percentile(np_arr, 75))
    print("P90:", np.percentile(np_arr, 90))
    print("P95:", np.percentile(np_arr, 95))
    print("Mean: ", sum(result) / N)
    print("Median: ", sorted(result)[N // 2])
    print("Mode: ", max(set(result), key=result.count))

Discussion on https://news.ycombinator.com/item?id=38581959