|
| 1 | +#!/usr/bin/env python3 |
| 2 | +""" |
| 3 | +Flash Sort Algorithm Implementation |
| 4 | +
|
| 5 | +Flash sort is a distribution sorting algorithm showing linear computational |
| 6 | +complexity O(n) for uniformly distributed datasets and relatively little |
| 7 | +additional memory requirement. The basic idea is to use the distribution |
| 8 | +of the values to be sorted to determine their approximate final positions |
| 9 | +directly, without comparing and moving each element through many intermediate |
| 10 | +positions as done by other algorithms. |
| 11 | +
|
| 12 | +The algorithm was developed by Karl-Dietrich Neubert in 1998 and builds upon |
| 13 | +the idea of bucket sort. It works by classifying elements into classes and |
| 14 | +then sorting each class. |
| 15 | +
|
| 16 | +Time Complexity: |
| 17 | +- Best Case: O(n) when data is uniformly distributed |
| 18 | +- Average Case: O(n + k) where k is the number of classes |
| 19 | +- Worst Case: O(n²) when data is not uniformly distributed |
| 20 | +
|
| 21 | +Space Complexity: O(k) where k is the number of classes |
| 22 | +
|
| 23 | +Source: https://en.wikipedia.org/wiki/Flashsort |
| 24 | +""" |
| 25 | + |
| 26 | +from __future__ import annotations |
| 27 | + |
| 28 | + |
| 29 | +def flash_sort(arr: list[int | float]) -> list[int | float]: |
| 30 | + """ |
| 31 | + Sorts a list using the Flash Sort algorithm. |
| 32 | +
|
| 33 | + Flash sort is particularly efficient for uniformly distributed data. |
| 34 | + It uses the distribution of values to determine approximate positions. |
| 35 | +
|
| 36 | + Args: |
| 37 | + arr: List of integers or floats to be sorted |
| 38 | +
|
| 39 | + Returns: |
| 40 | + Sorted list in ascending order |
| 41 | +
|
| 42 | + Examples: |
| 43 | + >>> flash_sort([4, 2, 7, 1, 9, 3]) |
| 44 | + [1, 2, 3, 4, 7, 9] |
| 45 | + >>> flash_sort([]) |
| 46 | + [] |
| 47 | + >>> flash_sort([5]) |
| 48 | + [5] |
| 49 | + >>> flash_sort([3, 3, 3, 3]) |
| 50 | + [3, 3, 3, 3] |
| 51 | + >>> flash_sort([-1, -5, 0, 3, 2]) |
| 52 | + [-5, -1, 0, 2, 3] |
| 53 | + >>> flash_sort([1.5, 2.3, 0.1, 3.7, 1.2]) |
| 54 | + [0.1, 1.2, 1.5, 2.3, 3.7] |
| 55 | + >>> flash_sort([10, 9, 8, 7, 6, 5, 4, 3, 2, 1]) |
| 56 | + [1, 2, 3, 4, 5, 6, 7, 8, 9, 10] |
| 57 | + >>> import random |
| 58 | + >>> data = random.sample(range(100), 20) |
| 59 | + >>> flash_sort(data) == sorted(data) |
| 60 | + True |
| 61 | + >>> flash_sort([42]) |
| 62 | + [42] |
| 63 | + >>> flash_sort([2.5, 1.1, 3.3, 2.5, 1.1]) |
| 64 | + [1.1, 1.1, 2.5, 2.5, 3.3] |
| 65 | + """ |
| 66 | + if len(arr) <= 1: |
| 67 | + return arr.copy() |
| 68 | + |
| 69 | + # Create a copy to avoid modifying the original array |
| 70 | + result = arr.copy() |
| 71 | + n = len(result) |
| 72 | + |
| 73 | + # Find min and max values |
| 74 | + min_val = min(result) |
| 75 | + max_val = max(result) |
| 76 | + |
| 77 | + # If all elements are the same, return the array |
| 78 | + if min_val == max_val: |
| 79 | + return result |
| 80 | + |
| 81 | + # Number of classes (buckets) - typically n/10 to n/5 works well |
| 82 | + m = max(1, int(0.45 * n)) |
| 83 | + |
| 84 | + # Initialize class sizes array |
| 85 | + class_sizes = [0] * m |
| 86 | + |
| 87 | + # Calculate class sizes |
| 88 | + c1 = (m - 1) / (max_val - min_val) |
| 89 | + |
| 90 | + for value in result: |
| 91 | + class_index = int(c1 * (value - min_val)) |
| 92 | + if class_index >= m: |
| 93 | + class_index = m - 1 |
| 94 | + class_sizes[class_index] += 1 |
| 95 | + |
| 96 | + # Calculate cumulative class sizes (positions) |
| 97 | + for i in range(1, m): |
| 98 | + class_sizes[i] += class_sizes[i - 1] |
| 99 | + |
| 100 | + # Permutation phase |
| 101 | + hold = result[0] |
| 102 | + j = 0 |
| 103 | + k = m - 1 |
| 104 | + |
| 105 | + while j < n - 1: |
| 106 | + while j >= class_sizes[k]: |
| 107 | + k -= 1 |
| 108 | + |
| 109 | + flash = int(c1 * (hold - min_val)) |
| 110 | + if flash >= m: |
| 111 | + flash = m - 1 |
| 112 | + |
| 113 | + while j < class_sizes[flash]: |
| 114 | + k = flash |
| 115 | + class_sizes[k] -= 1 |
| 116 | + result[j], result[class_sizes[k]] = result[class_sizes[k]], result[j] |
| 117 | + hold = result[j] |
| 118 | + j += 1 |
| 119 | + flash = int(c1 * (hold - min_val)) |
| 120 | + if flash >= m: |
| 121 | + flash = m - 1 |
| 122 | + |
| 123 | + j += 1 |
| 124 | + if j < n: |
| 125 | + hold = result[j] |
| 126 | + |
| 127 | + # Insertion sort for final sorting within classes |
| 128 | + for i in range(1, n): |
| 129 | + key = result[i] |
| 130 | + j = i - 1 |
| 131 | + while j >= 0 and result[j] > key: |
| 132 | + result[j + 1] = result[j] |
| 133 | + j -= 1 |
| 134 | + result[j + 1] = key |
| 135 | + |
| 136 | + return result |
| 137 | + |
| 138 | + |
| 139 | +if __name__ == "__main__": |
| 140 | + from doctest import testmod |
| 141 | + |
| 142 | + testmod() |
| 143 | + |
| 144 | + # Additional test cases |
| 145 | + test_cases: list[list[int | float]] = [ |
| 146 | + [64, 34, 25, 12, 22, 11, 90], |
| 147 | + [5, 2, 4, 6, 1, 3], |
| 148 | + [1], |
| 149 | + [], |
| 150 | + [3, 3, 3, 3], |
| 151 | + [-1, -3, 2, 0, -5], |
| 152 | + [1.1, 2.2, 0.5, 3.3, 1.5] |
| 153 | + ] |
| 154 | + |
| 155 | + for test_case in test_cases: |
| 156 | + sorted_result = flash_sort(test_case) |
| 157 | + expected = sorted(test_case) |
| 158 | + assert sorted_result == expected, f"Failed for {test_case}" |
| 159 | + print(f"✓ {test_case} -> {sorted_result}") |
| 160 | + |
| 161 | + print("All tests passed!") |
0 commit comments