最近2天受到http://blog.youkuaiyun.com/v_JULY_v/archive/2011/05/08/6403777.aspx的影响,从头开始实现了这个算法。收获还是挺大的:(1)实现了Hash链表(2)实现了堆;(3)熟悉了C语言的文件操作;
海量数据处理的Top K 算法就是在很大的文件中找出重复出现次数最多的前K个字符串;
如果数据可以一次读入内存,那么可以按照如下思路来解决:
(1)首先遍历文件,将其加入Hash链表;Hash链表的节点定义为:
- typedef struct node
- {
- char *word;
- int count;
- struct node *next;
- }node,*node_ptr;
word指向字符串,count为出现的次数;
(2)建立一个容量为K的最小堆,然后遍历Hash表中剩下的元素;如果找到了一个元素的count值比这个元素值大,那么交换这个节点和堆顶节点;
完整全部代码如下:在Eclipse + CDT + mingw 下调试通过
程序有待完善的地方:
(1)存在一个位置bug:改变HASHLEN的值,程序崩溃;
(2)HASH链表的遍历不正确,应该遍历到每一个节点的时候,在判断该节点的next是否为空;
- # include <stdio.h>
- # include <string.h>
- # include <malloc.h>
- # include <assert.h>
- # include <stdlib.h>
- #define HASHLEN 101
- #define WORDLEN 30
- #define MAX 100000
- #define DOMAIN 300
- #define K 5
- //Hash链表的节点定义
- typedef struct node
- {
- char *word;
- int count;
- struct node *next;
- }node,*node_ptr;
- static node_ptr head[HASHLEN];
- static node array[K];
- //Hash函数
- int hash_function(char *p)
- {
- unsigned int value = 0;
- while (*p != '/0')
- {
- value = value * 31 + *p++;
- if (value > HASHLEN)
- value = value % HASHLEN;
- }
- return value;
- }
- //加入节点到HASH链表
- void append_word(char *str)
- {
- int index = hash_function(str);
- node_ptr p = head[index];
- while (p != NULL)
- {
- if (strcmp(str, p->word) == 0)
- {
- (p->count)++;
- return;
- }
- p = p->next;
- }
- // 新建一个结点
- node_ptr q = (node_ptr)malloc(sizeof(node));
- q->count = 1;
- q->word = (char *)malloc(sizeof(str) + 1);
- strcpy(q->word, str);
- // q->word = str;
- q->next = head[index];
- head[index] = q;
- }
- //产生0~DOMAIN - 1范围内的MAX个整数
- void gen_data()
- {
- FILE *fp = fopen("c://data1.txt", "w");
- assert(fp);
- int i = 0;
- srand((int)(time(0)));
- for (i = 0; i < MAX; i++)
- fprintf(fp,"%d ",rand()%DOMAIN);
- fclose(fp);
- }
- //堆调整:调整为最小堆
- void heapAdjust(node array[], int beginIndex, int endIndex, int index)
- {
- int length = endIndex - beginIndex + 1;
- int largestIndex = index;
- int leftIndex = 2 * index + 1; //下标从0开始,可以自己做实验
- int rightIndex = 2 * index + 2;
- if (leftIndex <= length - 1 && array[leftIndex].count <= array[largestIndex].count )
- {
- largestIndex = leftIndex;
- }
- if (rightIndex <= length - 1 && array[rightIndex].count <= array[largestIndex].count )
- {
- largestIndex = rightIndex;
- }
- if (largestIndex != index)
- {
- node temp = array[largestIndex];
- array[largestIndex] = array[index];
- array[index] = temp;
- heapAdjust (array, beginIndex, endIndex, largestIndex);
- }
- }
- //建堆
- void heapBuild (node array[], int len )
- {
- int i = 0;
- for (i = len/2 - 1; i >= 0; i --)
- {
- heapAdjust (array, 0, len - 1, i);
- }
- }
- int main()
- {
- gen_data();
- char str[WORDLEN];
- // char *str;
- int i;
- int cnt1 = 0;
- for (i = 0; i < HASHLEN; i++)
- head[i] = NULL;
- FILE *fp_passage = fopen("c://data1.txt", "r");
- assert(fp_passage);
- while (fscanf(fp_passage, "%s", str) != EOF)
- {
- cnt1++;
- append_word(str);
- }
- printf("the cnt1 is %d/n", cnt1);
- fclose(fp_passage);
- //寻找Top K
- for(i = 0; i < HASHLEN; i++)
- {
- if(i < K - 1)
- array[i] = *head[i];
- else
- if(i == K - 1)
- {
- array[i] = *head[i];
- heapBuild(array,K);
- }
- else
- {
- if(array[0].count < head[i]->count)
- {
- array[0] = *head[i];
- heapAdjust(array,0,K - 1,0);
- }
- }
- }
- printf("the top K is as follows/n");
- for(i = 0; i < K; i++)
- printf("%s , and its count is %d/n",array[i].word, array[i].count);
- // printf("the total number of word is %d",cnt);
- return 0;
- }
运行结果:
100000个树中重复出现次数最多的前5个数:
the top K is as follows
49 , and its count is 372
4 , and its count is 374
249 , and its count is 373
246 , and its count is 380
227 , and its count is 376
http://blog.youkuaiyun.com/randyjiawenjie/article/details/6417753
本文介绍了一种基于Hash链表和最小堆实现的TopK算法,用于从大量数据中找出重复次数最多的前K个字符串。文章提供了完整的C语言代码示例,并在Eclipse+CDT+mingw环境下调试通过。
959

被折叠的 条评论
为什么被折叠?



