海量数据前N大

最大堆即大根堆 -- 求前n小

最小堆即小根堆 -- 求前n大

比如求前n小，对于当前元素，和最大堆的堆顶元素（即堆最大值）比较，如果小于堆顶元素，则替换堆顶元素，并调整堆。这样扫描一遍就可以得到最小的n个元素，效率非常高。

 

问题实例：

100w个数中找最大的前100个数

 

1 INT_MIN初始化最小堆

2 读入一个数，如果大于堆顶元素，则替换堆顶元素，并调整堆

 

//base index 1
#define LCHILD(i) (2*(i))
#define RCHILD(i) (2*(i)+1)

enum {
    BIGHEAP=0,
    SMALLHEAP=1,
};

//Adjust the heap
template <typename T>
void HeapAdjust(T *heap, size_t n, bool flg)
{
    int i=1;
    T root = heap[1];
    while (i<n)
    {
        T ex;
        if (flg == BIGHEAP) //最大堆
        {
            if (LCHILD(i)<=n)
            {
                ex = LCHILD(i);
                if (RCHILD(i)<=n && heap[LCHILD(i)]<heap[RCHILD(i)])
                    ex = RCHILD(i);
            }
            else
                break;
                
            if (heap[ex] > root)
            {
                heap[i] = heap[ex];
                i = ex;
            }
            else
                break;
        }
        else //最小堆
        {
            if (LCHILD(i)<=n)
            {
                ex = LCHILD(i);
                if (RCHILD(i)<=n && heap[LCHILD(i)]>heap[RCHILD(i)])
                    ex = RCHILD(i);
            }
            else
                break;
            if (heap[ex] < root)
            {
                heap[i] = heap[ex];
                i = ex;
            }
            else
                break;
        }        
    }
    heap[i] = root;
}

void FindNNumber(const char *file, size_t n, bool flg)
{
    int *heap = new int[n+1];
    int i;
    for (i=1; i<=n; i++)
    {
        if (flg == BIGHEAP)
            heap[i] = INT_MAX;
        else
            heap[i] = INT_MIN;
    }

    FILE *fp = fopen(file, "r");

    int a;
    while(fscanf(fp, "%d", &a)>0)
    {
        if (flg == BIGHEAP)
        {
            if (a < heap[1])
                heap[1] = a;
        }
        else
        {
            if (a > heap[1])
                heap[1] = a;

        }
        HeapAdjust(heap, n, flg);
    }
    fclose(fp);
    for (i=1; i<=n; i++)
        printf("%d\n", heap[i]);
}

void main()
{
    //最小的n个数
    FindNNumber("temp.txt", 10, BIGHEAP);
    printf("\n\n");
    //最大的n个数
    FindNNumber("tt.txt", 10, SMALLHEAP);

}