Python包含了用于min-堆的heapq模块,但我需要一个max堆。在Python中我应该使用什么来实现最大堆?


当前回答

这是一个基于heapq的简单MaxHeap实现。虽然它只适用于数值。

import heapq
from typing import List


class MaxHeap:
    def __init__(self):
        self.data = []

    def top(self):
        return -self.data[0]

    def push(self, val):
        heapq.heappush(self.data, -val)

    def pop(self):
        return -heapq.heappop(self.data)

用法:

max_heap = MaxHeap()
max_heap.push(3)
max_heap.push(5)
max_heap.push(1)
print(max_heap.top())  # 5

其他回答

arr = [3,4,5,1,2,3,0,7,8,90,67,31,2,5,567]
# max-heap sort will lead the array to assending order
def maxheap(arr,p):
    
    for i in range(len(arr)-p):
        if i > 0:
            child = i
            parent = (i+1)//2 - 1
            
            while arr[child]> arr[parent] and child !=0:
                arr[child], arr[parent] = arr[parent], arr[child]
                child = parent
                parent = (parent+1)//2 -1
                
    
def heapsort(arr):
    for i in range(len(arr)):
        maxheap(arr,i)
        arr[0], arr[len(arr)-i-1]=arr[len(arr)-i-1],arr[0]
        
    return arr
        

print(heapsort(arr))

试试这个

我创建了一个名为heap_class的包,它实现了最大堆,还将各种堆函数包装到一个与列表兼容的环境中。

>>> from heap_class import Heap
>>> h = Heap([3, 1, 9, 20], max=True)
>>> h.pop()
20
>>> h.peek()  # same as h[0]
9
>>> h.push(17)  # or h.append(17)
>>> h[0]  # same as h.peek()
17
>>> h[1]  # inefficient, but works
9

从最大堆中获得最小堆。

>>> y = reversed(h)
>>> y.peek()
1
>>> y  # repr is inefficient, but correct
Heap([1, 3, 9, 17], max=False)
>>> 9 in y
True
>>> y.raw()  # underlying heap structure
[1, 3, 17, 9]

正如其他人所提到的,在max堆中处理字符串和复杂对象在heapq中是相当困难的,因为它们不同 否定的形式。heap_class实现简单:

>>> h = Heap(('aa', 4), ('aa', 5), ('zz', 2), ('zz', 1), max=True)
>>> h.pop()
('zz', 2)

支持自定义键,并与后续的推/追加和弹出一起工作:

>>> vals = [('Adam', 'Smith'), ('Zeta', 'Jones')]
>>> h = Heap(vals, key=lambda name: name[1])
>>> h.peek()  # Jones comes before Smith
('Zeta', 'Jones')
>>> h.push(('Aaron', 'Allen'))
>>> h.peek()
('Aaron', 'Allen')

(实现是建立在heapq函数上的,所以它都是用C语言或C语言包装的,除了Python中max heap上的heappush和heapreplace)

我实现了一个最大堆版本的heapq,并将它提交给PyPI。(对heapq模块CPython代码的改动很小。)

https://pypi.python.org/pypi/heapq_max/

https://github.com/he-zhe/heapq_max

安装

pip install heapq_max

使用

dr:与heapq模块相同,只是所有函数都增加了' _max '。

heap_max = []                           # creates an empty heap
heappush_max(heap_max, item)            # pushes a new item on the heap
item = heappop_max(heap_max)            # pops the largest item from the heap
item = heap_max[0]                      # largest item on the heap without popping it
heapify_max(x)                          # transforms list into a heap, in-place, in linear time
item = heapreplace_max(heap_max, item)  # pops and returns largest item, and
                                    # adds new item; the heap size is unchanged

最简单的方法 就是把每个元素都转换成负数,问题就解决了。

import heapq
heap = []
heapq.heappush(heap, 1*(-1))
heapq.heappush(heap, 10*(-1))
heapq.heappush(heap, 20*(-1))
print(heap)

输出如下所示:

[-20, -1, -10]

为了详细说明https://stackoverflow.com/a/59311063/1328979,这里有一个针对一般情况的完整文档、注释和测试的Python 3实现。

from __future__ import annotations  # To allow "MinHeap.push -> MinHeap:"
from typing import Generic, List, Optional, TypeVar
from heapq import heapify, heappop, heappush, heapreplace


T = TypeVar('T')


class MinHeap(Generic[T]):
    '''
    MinHeap provides a nicer API around heapq's functionality.
    As it is a minimum heap, the first element of the heap is always the
    smallest.
    >>> h = MinHeap([3, 1, 4, 2])
    >>> h[0]
    1
    >>> h.peek()
    1
    >>> h.push(5)  # N.B.: the array isn't always fully sorted.
    [1, 2, 4, 3, 5]
    >>> h.pop()
    1
    >>> h.pop()
    2
    >>> h.pop()
    3
    >>> h.push(3).push(2)
    [2, 3, 4, 5]
    >>> h.replace(1)
    2
    >>> h
    [1, 3, 4, 5]
    '''
    def __init__(self, array: Optional[List[T]] = None):
        if array is None:
            array = []
        heapify(array)
        self.h = array
    def push(self, x: T) -> MinHeap:
        heappush(self.h, x)
        return self  # To allow chaining operations.
    def peek(self) -> T:
        return self.h[0]
    def pop(self) -> T:
        return heappop(self.h)
    def replace(self, x: T) -> T:
        return heapreplace(self.h, x)
    def __getitem__(self, i) -> T:
        return self.h[i]
    def __len__(self) -> int:
        return len(self.h)
    def __str__(self) -> str:
        return str(self.h)
    def __repr__(self) -> str:
        return str(self.h)


class Reverse(Generic[T]):
    '''
    Wrap around the provided object, reversing the comparison operators.
    >>> 1 < 2
    True
    >>> Reverse(1) < Reverse(2)
    False
    >>> Reverse(2) < Reverse(1)
    True
    >>> Reverse(1) <= Reverse(2)
    False
    >>> Reverse(2) <= Reverse(1)
    True
    >>> Reverse(2) <= Reverse(2)
    True
    >>> Reverse(1) == Reverse(1)
    True
    >>> Reverse(2) > Reverse(1)
    False
    >>> Reverse(1) > Reverse(2)
    True
    >>> Reverse(2) >= Reverse(1)
    False
    >>> Reverse(1) >= Reverse(2)
    True
    >>> Reverse(1)
    1
    '''
    def __init__(self, x: T) -> None:
        self.x = x
    def __lt__(self, other: Reverse) -> bool:
        return other.x.__lt__(self.x)
    def __le__(self, other: Reverse) -> bool:
        return other.x.__le__(self.x)
    def __eq__(self, other) -> bool:
        return self.x == other.x
    def __ne__(self, other: Reverse) -> bool:
        return other.x.__ne__(self.x)
    def __ge__(self, other: Reverse) -> bool:
        return other.x.__ge__(self.x)
    def __gt__(self, other: Reverse) -> bool:
        return other.x.__gt__(self.x)
    def __str__(self):
        return str(self.x)
    def __repr__(self):
        return str(self.x)


class MaxHeap(MinHeap):
    '''
    MaxHeap provides an implement of a maximum-heap, as heapq does not provide
    it. As it is a maximum heap, the first element of the heap is always the
    largest. It achieves this by wrapping around elements with Reverse,
    which reverses the comparison operations used by heapq.
    >>> h = MaxHeap([3, 1, 4, 2])
    >>> h[0]
    4
    >>> h.peek()
    4
    >>> h.push(5)  # N.B.: the array isn't always fully sorted.
    [5, 4, 3, 1, 2]
    >>> h.pop()
    5
    >>> h.pop()
    4
    >>> h.pop()
    3
    >>> h.pop()
    2
    >>> h.push(3).push(2).push(4)
    [4, 3, 2, 1]
    >>> h.replace(1)
    4
    >>> h
    [3, 1, 2, 1]
    '''
    def __init__(self, array: Optional[List[T]] = None):
        if array is not None:
            array = [Reverse(x) for x in array]  # Wrap with Reverse.
        super().__init__(array)
    def push(self, x: T) -> MaxHeap:
        super().push(Reverse(x))
        return self
    def peek(self) -> T:
        return super().peek().x
    def pop(self) -> T:
        return super().pop().x
    def replace(self, x: T) -> T:
        return super().replace(Reverse(x)).x


if __name__ == '__main__':
    import doctest
    doctest.testmod()

https://gist.github.com/marccarre/577a55850998da02af3d4b7b98152cf4