寻找上亿数据TOP K

分析：有待补充。。。

//  [10/10/2013 qingezha]寻找上亿数据TOP K
void FindKMax()
{
	//for(long i=0;i<1000;i++)
	//{
	//	ofstream outFile; //输出到外存 
	//	outFile.open("D:\\t.txt",ios_base::app);
	//	srand(i);
	//	long s=rand()<<6;
	//	outFile<<s<<" ";
	//	outFile.close(); 
	//}
	//
	//system("pause");
	// 定义数组存储堆元素  
	int k;  
	cin >> k;  
	long *heap = new long [k+1];   //注，只需申请存储k个数的数组  
	FILE *fp = fopen("D:\\t.txt", "r"); //从文件导入海量数据（便于测试，只截取了9M的数据大小）  
	assert(fp);  

	for (int i = 1; i <= k; i++)  
		fscanf(fp, "%d ", &heap[i]);  
	for (int j = 1; j <= k; j++)  
		cout << heap[j] << " ";  
	cout << endl;  
	for (int i=k/2+1;i>=1;--i)
	{
		siftMaxx(heap,i,k);
	}  
	for (int j = 1; j <= k; j++)  
		cout << heap[j] << " ";  
	cout << endl;  
	long newData;  
	while (fscanf(fp, "%d", &newData) != EOF)  
	{   
		if (newData > heap[1])   //找最小的K个数 如果遇到比堆顶元素kmax更小的，则更新大堆,  
		{											//找最大的K数，更新小堆
			heap[1] = newData;  
			siftMaxx(heap,1,k); //调整堆  
			for (int j = 1; j <= k; j++)  
				cout << heap[j] << " ";  
			cout << endl;  
		}  

	}  

	for (int j = 1; j <= k; j++)  
		cout << heap[j] << " ";  
	cout << endl;  

	fclose(fp); 
	system("pause");

}
void siftMaxx( long arr[],int low,int high ) /*区间[low,high],构造二叉堆//大的到根部去，小的到叶子去*/
{
	// 这两中方法都可以 ,但是推荐第二种
	int i=low;
	int j=2*i;
	int temp=arr[i];
	while(j<=high)
	{
		if (j<high&&arr[j]>arr[j+1]) 
			j++; 
		if (arr[i]>arr[j])
		{
			arr[i]=arr[j];
			i=j;
			j=2*i;
		}
		else break;
		arr[i]=temp; 
	}

	//
	int child;
	for (int i=1;i<high;i=child)
	{
		child=2*i;
		if (child+1<=high&&arr[child]>arr[child+1]) 
			++child;
		if(arr[child]<arr[i])  //检查是否越界，很多时候的bug最后检查出来都是小毛病，但是浪费了很多时间
										//if(child<=high&&arr[child]<arr[i]) 这样就对了
			swap(arr[child],arr[i]);
	}
}

ptr_has_space head[HASHLEN];//不能在头文件中定义

void write_to_file()//将hash后的结果保存在result.txt中
{
	FILE *fp = fopen("D:\\result.txt", "w");  
	assert(fp);   
	int i = 0;  
	while (i < HASHLEN)  
	{  
		for (ptr_has_space p = head[i]; p != NULL; p = p->next)  
			fprintf(fp, "%d  %d\n", p->data, p->count);  
		i++;  
	}  
	fclose(fp);
}

void append_hash( const int *p )//附加到hash表中，如果重叠则链表表示
{
	int index=hash_function(p);
	ptr_has_space q=head[index];
	if(NULL==q)
	{
		head[index]=new node_has_space;
		head[index]->count=1;
		head[index]->data=*p;
		head[index]->next=NULL;
		return;
	}
	else
	{
		while(q)
		{
			if(*p==q->data)
			{
				++(q->count);
				return;
			}
			q=q->next;
		}
		ptr_has_space pt=new node_has_space;
		pt->count=1;pt->data=*p;
		pt->next=head[index];//采用头插入法
		head[index]=pt;
	}
}

int hash_function( const int * p )//简单hash函数
{
	int index=0;
	if(*p>HASHLEN) 
		index=*p%HASHLEN;
	else
		index=*p;
	return index;
}
 

void siftheap( node_has_space arr[],int low,int high )//堆结点类型为node_has_space，筛选
{
	int i=low;
	int j=2*i;
	node_has_space temp=arr[i];
	while(j<=high)
	{
		if (j<high&&arr[j].count>arr[j+1].count) 
			j++; 
		if (arr[i].count>arr[j].count)
		{
			arr[i]=arr[j];
			i=j;
			j=2*i;
		}
		else break;
		arr[i]=temp; 
	}
	 
}

void bigDATASearch()
{
	//写进数据，模拟上亿的数据
	//for(long i=0;i<100000;i++)
	//{
	//	ofstream outFile; //输出到外存 
	//	outFile.open("D:\\t.txt",ios_base::app);
	//	srand(i);
	//	long s=rand()%30000;
	//	outFile<<s<<" ";
	//	outFile.close(); 
	//}
	//
	//system("pause");

	//将上亿的数据归类，hash，记录重复的次数，然后以<key  count> \n 形式写进txt里面
	//int *dataptr=new int;
	//FILE *fp = fopen("D:\\t.txt", "r");   //从文件导入海量数据 
	//assert(fp); 
	//while(fscanf(fp,"%d",dataptr)!=EOF)
	//{
	//	append_hash(dataptr);
	//}
	//fclose(fp); 
	//write_to_file();
	//system("pause");

	//读取上述txt，用堆的形式取前K个count最大值
	int k;  
	cin >> k;  
	ptr_has_space  heap = new node_has_space[k+1];   //注，只需申请存储k个数的数组  
	FILE *fp = fopen("D:\\result.txt", "r");   //从文件导入海量数据（便于测试，只截取了9M的数据大小）  
	assert(fp);  

	for (int i = 1; i <= k; i++)  
		fscanf(fp, "%d %d", &(heap[i].data),&(heap[i].count));   
	for (int i=k/2+1;i>=1;--i)
	{
		siftheap(heap,i,k);
	} 
	for (int j = 1; j <= k; j++)  
		cout << heap[j].data<<" "<<heap[j].count << " ";  
	cout << endl;  
	long newData;
	int count;
	while (fscanf(fp, "%d %d", &newData,&count) != EOF)  
	{   
		if (count > heap[1].count)   //找最小的K个数 如果遇到比堆顶元素kmax更小的，则更新大堆,  
		{											//找最大的K数，更新小堆
			heap[1].data = newData;  
			heap[1].count=count;
			siftheap(heap,1,k); //调整堆  
			for (int j = 1; j <= k; j++)  
				cout << heap[j].data << "\t"<<heap[j].count<<"\t";  
			cout << endl;  
		}   
	}  

	for (int j = 1; j <= k; j++)  
		cout << heap[j].data << " "<<heap[j].count;  
	cout << endl;  

	fclose(fp); 
	system("pause");
}