【算法】八大排序以及时间空间复杂度分析以及用Python实现

排序是老生常谈了,但是不写来了又总会忘记。之前写过用C++实现的少部分排序,这一次写一下Python实现的八大排序。

 

排序算法平均时间复杂度最好情况最坏情况空间复杂度排序方式稳定性
插入排序 O(n²) O(n) O(n²) O(1) In-place 稳定
冒泡排序 O(n²) O(n) O(n²) O(1) In-place 稳定
选择排序 O(n²) O(n²) O(n²) O(1) In-place 不稳定
快速排序 O(n log n) O(n log n) O(n²) O(log n) In-place 不稳定
希尔排序 O(n log n) O(n log n) O(n log n) O(1) In-place 不稳定
堆排序 O(n log n) O(n log n) O(n log n) O(1) In-place 不稳定
归并排序 O(n log n) O(n log n) O(n log n) O(n) Out-place 稳定


直接插入排序:

def insert_sort(array):
    for i in range(len(array)):
        for j in range(i):
            if array[i] < array[j]:
                array.insert(j, array.pop(i))
                break
    return array

希尔排序:

def shell_sort(array):
    gap = len(array)
    while gap > 1:
        gap = gap // 2
        for i in range(gap, len(array)):
            for j in range(i % gap, i, gap):
                if array[i] < array[j]:
                    array[i], array[j] = array[j], array[i]
    return array

选择排序:

def select_sort(array):
    for i in range(len(array)):
        x = i  # min index
        for j in range(i, len(array)):
            if array[j] < array[x]:
                x = j
        array[i], array[x] = array[x], array[i]
    return array

 堆排序:

def heap_sort(nums):
    # 调整堆
    # 迭代写法
    def adjust_heap(nums, startpos, endpos):
        newitem = nums[startpos]
        pos = startpos
        childpos = pos * 2 + 1
        while childpos < endpos:
            rightpos = childpos + 1
            if rightpos < endpos and nums[rightpos] >= nums[childpos]:
                childpos = rightpos
            if newitem < nums[childpos]:
                nums[pos] = nums[childpos]
                pos = childpos
                childpos = pos * 2 + 1
            else:
                break
        nums[pos] = newitem
    
    # 递归写法
    def adjust_heap(nums, startpos, endpos):
        pos = startpos
        chilidpos = pos * 2 + 1
        if chilidpos < endpos:
            rightpos = chilidpos + 1
            if rightpos < endpos and nums[rightpos] > nums[chilidpos]:
                chilidpos = rightpos
            if nums[chilidpos] > nums[pos]:
                nums[pos], nums[chilidpos] = nums[chilidpos], nums[pos]
                adjust_heap(nums, pos, endpos)
def heap_sort(array):
    def heap_adjust(parent):
        child = 2 * parent + 1  # left child
        while child < len(heap):
            if child + 1 < len(heap):
                if heap[child + 1] > heap[child]:
                    child += 1  # right child
            if heap[parent] >= heap[child]:
                break
            heap[parent], heap[child] = \
                heap[child], heap[parent]
            parent, child = child, 2 * child + 1

    heap, array = array.copy(), []
    for i in range(len(heap) // 2, -1, -1):
        heap_adjust(i)
    while len(heap) != 0:
        heap[0], heap[-1] = heap[-1], heap[0]
        array.insert(0, heap.pop())
        heap_adjust(0)
    return array

 

冒泡排序:

def bubble_sort(array):
    for i in range(len(array)):
        for j in range(i, len(array)):
            if array[i] > array[j]:
                array[i], array[j] = array[j], array[i]
    return array

 

快速排序:

def Quick_sort(num_list):
    '''
    快速排序,时间复杂度:O(nlog₂n),空间复杂度:O(nlog₂n),不是稳定排序
    '''   
    if len(num_list)<2:    
        return num_list    
    left_list = []  #存放比基准结点小的元素  
    right_list = []  #存放比基准元素大的元素  
    base_node = num_list.pop(0) #在这里采用pop()方法的原因就是需要移除这个基准结点,并且赋值给base_node这个变量  
                                #在这里不能使用del()方法,因为删除之后无法再赋值给其他变量使用,导致最终数据缺失  
                                #快排每轮可以确定一个元素的位置,之后递归地对两边的元素进行排序  
    for one_num in num_list:    
        if one_num < base_node:    
            left_list.append(one_num)    
        else:    
            right_list.append(one_num)    
    return Quick_sort(left_list) + [base_node] + Quick_sort(right_list)

归并排序:

def merge_sort(nums):
    if len(nums) <= 1:
        return nums
    mid = len(nums) // 2
    #
    left = merge_sort(nums[:mid])
    right = merge_sort(nums[mid:])
    # 合并
    return merge(left, right)

def merge(left, right):
    res = []
    i = 0
    j = 0
    while i < len(left) and j < len(right):
        if left[i] <= right[j]:
            res.append(left[i])
            i += 1
        else:
            res.append(right[j])
            j += 1
    res += left[i:]
    res += right[j:]
    return res

基数排序:

def radix_sort(array):
    bucket, digit = [[]], 0
    while len(bucket[0]) != len(array):
        bucket = [[], [], [], [], [], [], [], [], [], []]
        for i in range(len(array)):
            num = (array[i] // 10 ** digit) % 10
            bucket[num].append(array[i])
        array.clear()
        for i in range(len(bucket)):
            array += bucket[i]
        digit += 1
    return array

 

参考:

Python 实现七大排序算法

https://juejin.im/post/6844903873392410631

leetcode

https://leetcode-cn.com/problems/sort-an-array/solution/python-shi-xian-de-shi-da-jing-dian-pai-xu-suan-fa/

Python 八大排序算法速度比较

https://www.cnblogs.com/woider/p/6835466.html

 

posted on 2020-08-13 22:10  雪原那么远  阅读(1003)  评论(0编辑  收藏  举报

导航