topK问题

解法一：基于Partidion函数的平均时间复杂度为O(n)的方法。最坏O(kn)，此时数组已经有序，需要执行Partidion函数k次。最好情况O(n)，执行一次Partidion函数即得出结果。

Python3代码如下（最小的k个数）：

def Partidion(nums,left,right):
    p = nums[left]
    while left < right:
        while left < right and nums[right] >= p:
            right -= 1
        nums[left] = nums[right]
        while left < right and nums[left] <= p:
            left += 1
        nums[right] = nums[left]
    nums[left] = p
    return left
def s(nums,k,r):
    left,right = 0,len(nums)-1
    p = Partidion(nums,left,right)
    while p+1 != k:
        if p+1 < k:
            left = p+1
            p = Partidion(nums,left,right)
        else:
            right = p-1
            p = Partidion(nums,left,right)
    for i in range(k):
        r.append(nums[i])
    return r
print(s([4,2,3,0,5,-1,6,-2,1],3,[]))

解法二：基于堆的平均时间复杂度为O(nlogn)。

Python3代码如下（最大的k个数）：

import heapq
def f(nums,k):
    h = []
    for i in range(k):
        h.append(nums[i])
    heapq.heapify(h)
    for j in range(k,len(nums)):
        if nums[j] > h[0]:
            heapq.heappop(h)
            heapq.heappush(h,nums[j])
    return h
print(f([0,2,3,0,5,6,0],3))