海量数据处理算法—Bloom Filter
时间:2015-08-03 01:08 来源:linux.it.net.cn 作者:IT
1. Bloom-Filter算法简介
Bloom-Filter,即布隆过滤器,1970年由Bloom中提出。它可以用于检索一个元素是否在一个集合中。
Bloom Filter(BF)是一种空间效率很高的随机数据结构,它利用位数组很简洁地表示一个集合,并能判断一个元素是否属于这个集合。它是一个判断元素是否存在集合的快速的概率算法。Bloom Filter有可能会出现错误判断,但不会漏掉判断。也就是Bloom Filter判断元素不再集合,那肯定不在。如果判断元素存在集合中,有一定的概率判断错误。因此,Bloom Filter不适合那些“零错误”的应用场合。而在能容忍低错误率的应用场合下,Bloom Filter比其他常见的算法(如hash,折半查找)极大节省了空间。
它的优点是空间效率和查询时间都远远超过一般的算法,缺点是有一定的误识别率和删除困难。
Bloom Filter的详细介绍:Bloom Filter
2、 Bloom-Filter的基本思想
Bloom-Filter算法的核心思想就是利用多个不同的Hash函数来解决“冲突”。
计算某元素x是否在一个集合中,首先能想到的方法就是将所有的已知元素保存起来构成一个集合R,然后用元素x跟这些R中的元素一一比较来判断是否存在于集合R中;我们可以采用链表等数据结构来实现。但是,随着集合R中元素的增加,其占用的内存将越来越大。试想,如果有几千万个不同网页需要下载,所需的内存将足以占用掉整个进程的内存地址空间。即使用MD5,UUID这些方法将URL转成固定的短小的字符串,内存占用也是相当巨大的。
于是,我们会想到用Hash table的数据结构,运用一个足够好的Hash函数将一个URL映射到二进制位数组(位图数组)中的某一位。如果该位已经被置为1,那么表示该URL已经存在。
Hash存在一个冲突(碰撞)的问题,用同一个Hash得到的两个URL的值有可能相同。为了减少冲突,我们可以多引入几个Hash,如果通过其中的一个Hash值我们得出某元素不在集合中,那么该元素肯定不在集合中。只有在所有的Hash函数告诉我们该元素在集合中时,才能确定该元素存在于集合中。这便是Bloom-Filter的基本思想。
原理要点:一是位数组, 而是k个独立hash函数。
1)位数组:
假设Bloom Filter使用一个m比特的数组来保存信息,初始状态时,Bloom Filter是一个包含m位的位数组,每一位都置为0,即BF整个数组的元素都设置为0。
2)添加元素,k个独立hash函数
为了表达S={x1, x2,…,xn}这样一个n个元素的集合,Bloom Filter使用k个相互独立的哈希函数(Hash Function),它们分别将集合中的每个元素映射到{1,…,m}的范围中。
当我们往Bloom Filter中增加任意一个元素x时候,我们使用k个哈希函数得到k个哈希值,然后将数组中对应的比特位设置为1。即第i个哈希函数映射的位置hashi(x)就会被置为1(1≤i≤k)。
注意,如果一个位置多次被置为1,那么只有第一次会起作用,后面几次将没有任何效果。在下图中,k=3,且有两个哈希函数选中同一个位置(从左边数第五位,即第二个“1“处)。
3)判断元素是否存在集合
在判断y是否属于这个集合时,我们只需要对y使用k个哈希函数得到k个哈希值,如果所有hashi(y)的位置都是1(1≤i≤k),即k个位置都被设置为1了,那么我们就认为y是集合中的元素,否则就认为y不是集合中的元素。下图中y1就不是集合中的元素(因为y1有一处指向了“0”位)。y2或者属于这个集合,或者刚好是一个false positive。
显然这 个判断并不保证查找的结果是100%正确的。
Bloom Filter的缺点:
1)Bloom Filter无法从Bloom Filter集合中删除一个元素。因为该元素对应的位会牵动到其他的元素。所以一个简单的改进就是 counting Bloom filter,用一个counter数组代替位数组,就可以支持删除了。 此外,Bloom Filter的hash函数选择会影响算法的效果。
2)还有一个比较重要的问题,如何根据输入元素个数n,确定位数组m的大小及hash函数个数,即hash函数选择会影响算法的效果。当hash函数个数k=(ln2)*(m/n)时错误率最小。在错误率不大于E的情况 下,m至少要等于n*lg(1/E) 才能表示任意n个元素的集合。但m还应该更大些,因为还要保证bit数组里至少一半为0,则m应 该>=nlg(1/E)*lge ,大概就是nlg(1/E)1.44倍(lg表示以2为底的对数)。
举个例子我们假设错误率为0.01,则此时m应大概是n的13倍。这样k大概是8个。
注意:
这里m与n的单位不同,m是bit为单位,而n则是以元素个数为单位(准确的说是不同元素的个数)。通常单个元素的长度都是有很多bit的。所以使用bloom filter内存上通常都是节省的。
一般BF可以与一些key-value的数据库一起使用,来加快查询。由于BF所用的空间非常小,所有BF可以常驻内存。这样子的话,对于大部分不存在的元素,我们只需要访问内存中的BF就可以判断出来了,只有一小部分,我们需要访问在硬盘上的key-value数据库。从而大大地提高了效率。
一个Bloom Filter有以下参数:
m
bit数组的宽度(bit数)
n
加入其中的key的数量
k
使用的hash函数的个数
f
False Positive的比率
Bloom Filter的f满足下列公式:
在给定m和n时,能够使f最小化的k值为:
此时给出的f为:
根据以上公式,对于任意给定的f,我们有:
n = m ln(0.6185) / ln(f) [1]
同时,我们需要k个hash来达成这个目标:
k = - ln(f) / ln(2) [2]
由于k必须取整数,我们在Bloom Filter的程序实现中,还应该使用上面的公式来求得实际的f:
f = (1 – e-kn/m)k [3]
以上3个公式是程序实现Bloom Filter的关键公式。
3、 扩展 CounterBloom Filter
CounterBloom Filter
BloomFilter有个缺点,就是不支持删除操作,因为它不知道某一个位从属于哪些向量。那我们可以给Bloom Filter加上计数器,添加时增加计数器,删除时减少计数器。
但这样的Filter需要考虑附加的计数器大小,假如同个元素多次插入的话,计数器位数较少的情况下,就会出现溢出问题。如果对计数器设置上限值的话,会导致Cache Miss,但对某些应用来说,这并不是什么问题,如Web Sharing。
Compressed Bloom Filter
为了能在服务器之间更快地通过网络传输Bloom Filter,我们有方法能在已完成Bloom Filter之后,得到一些实际参数的情况下进行压缩。
将元素全部添加入Bloom Filter后,我们能得到真实的空间使用率,用这个值代入公式计算出一个比m小的值,重新构造Bloom Filter,对原先的哈希值进行求余处理,在误判率不变的情况下,使得其内存大小更合适。
4、 Bloom-Filter的应用
Bloom-Filter一般用于在大数据量的集合中判定某元素是否存在。例如邮件服务器中的垃圾邮件过滤器。在搜索引擎领域,Bloom-Filter最常用于网络蜘蛛(Spider)的URL过滤,网络蜘蛛通常有一个URL列表,保存着将要下载和已经下载的网页的URL,网络蜘蛛下载了一个网页,从网页中提取到新的URL后,需要判断该URL是否已经存在于列表中。此时,Bloom-Filter算法是最好的选择。
1.key-value 加快查询
一般Bloom-Filter可以与一些key-value的数据库一起使用,来加快查询。
一般key-value存储系统的values存在硬盘,查询就是件费时的事。将Storage的数据都插入Filter,在Filter中查询都不存在时,那就不需要去Storage查询了。当False Position出现时,只是会导致一次多余的Storage查询。
由于Bloom-Filter所用的空间非常小,所有BF可以常驻内存。这样子的话,对于大部分不存在的元素,我们只需要访问内存中的Bloom-Filter就可以判断出来了,只有一小部分,我们需要访问在硬盘上的key-value数据库。从而大大地提高了效率。如图:
2 .Google的BigTable
Google的BigTable也使用了Bloom Filter,以减少不存在的行或列在磁盘上的查询,大大提高了数据库的查询操作的性能。
3. Proxy-Cache
在Internet Cache Protocol中的Proxy-Cache很多都是使用Bloom Filter存储URLs,除了高效的查询外,还能很方便得传输交换Cache信息。
4.网络应用
1)P2P网络中查找资源操作,可以对每条网络通路保存Bloom Filter,当命中时,则选择该通路访问。
2)广播消息时,可以检测某个IP是否已发包。
3)检测广播消息包的环路,将Bloom Filter保存在包里,每个节点将自己添加入Bloom Filter。
4)信息队列管理,使用Counter Bloom Filter管理信息流量。
5. 垃圾邮件地址过滤
像网易,QQ这样的公众电子邮件(email)提供商,总是需要过滤来自发送垃圾邮件的人(spamer)的垃圾邮件。
一个办法就是记录下那些发垃圾邮件的 email地址。由于那些发送者不停地在注册新的地址,全世界少说也有几十亿个发垃圾邮件的地址,将他们都存起来则需要大量的网络服务器。
如果用哈希表,每存储一亿个 email地址,就需要 1.6GB的内存(用哈希表实现的具体办法是将每一个 email地址对应成一个八字节的信息指纹,然后将这些信息指纹存入哈希表,由于哈希表的存储效率一般只有 50%,因此一个 email地址需要占用十六个字节。一亿个地址大约要 1.6GB,即十六亿字节的内存)。因此存贮几十亿个邮件地址可能需要上百 GB的内存。
而Bloom Filter只需要哈希表 1/8到 1/4 的大小就能解决同样的问题。
BloomFilter决不会漏掉任何一个在黑名单中的可疑地址。而至于误判问题,常见的补救办法是在建立一个小的白名单,存储那些可能别误判的邮件地址。
5、 Bloom-Filter的具体实现
c语言实现:
stdafx.h:
[cpp] view plaincopyprint?
-
#pragma once
-
#include <stdio.h>
-
#include "stdlib.h"
-
#include <iostream>
-
#include <time.h>
-
using namespace std;
#pragma once
#include <stdio.h>
#include "stdlib.h"
#include <iostream>
#include <time.h>
using namespace std;
[cpp] view plaincopyprint?
-
#include "stdafx.h"
-
-
-
#define ARRAY_SIZE 256 /*we get the 256 chars of each line*/
-
#define SIZE 48000000 /* size should be 1/8 of max*/
-
#define MAX 384000000/*the max bit space*/
-
-
#define SETBIT(ch,n) ch[n/8]|=1<<(7-n%8)
-
#define GETBIT(ch,n) (ch[n/8]&1<<(7-n%8))>>(7-n%8)
-
-
unsigned int len(char *ch);/* functions to calculate the length of the url*/
-
-
unsigned int RSHash(char* str, unsigned int len);/* functions to calculate the hash value of the url*/
-
unsigned int JSHash(char* str, unsigned int len);/* functions to calculate the hash value of the url*/
-
unsigned int PJWHash(char* str, unsigned int len);/* functions to calculate the hash value of the url*/
-
unsigned int ELFHash(char* str, unsigned int len);/* functions to calculate the hash value of the url*/
-
unsigned int BKDRHash(char* str, unsigned int len);/* functions to calculate the hash value of the url*/
-
unsigned int SDBMHash(char* str, unsigned int len);/* functions to calculate the hash value of the url*/
-
unsigned int DJBHash(char* str, unsigned int len);/* functions to calculate the hash value of the url*/
-
unsigned int DEKHash(char* str, unsigned int len);/* functions to calculate the hash value of the url*/
-
unsigned int BPHash(char* str, unsigned int len);/* functions to calculate the hash value of the url*/
-
unsigned int FNVHash(char* str, unsigned int len);/* functions to calculate the hash value of the url*/
-
unsigned int APHash(char* str, unsigned int len);/* functions to calculate the hash value of the url*/
-
unsigned int HFLPHash(char* str,unsigned int len);/* functions to calculate the hash value of the url*/
-
unsigned int HFHash(char* str,unsigned int len);/* functions to calculate the hash value of the url*/
-
unsigned int StrHash( char* str,unsigned int len);/* functions to calculate the hash value of the url*/
-
unsigned int TianlHash(char* str,unsigned int len);/* functions to calculate the hash value of the url*/
-
-
-
int main()
-
{
-
int i,num,num2=0; /* the number to record the repeated urls and the total of it*/
-
unsigned int tt=0;
-
int flag; /*it helps to check weather the url has already existed */
-
char buf[257]; /*it helps to print the start time of the program */
-
time_t tmp = time(NULL);
-
-
char file1[100],file2[100];
-
FILE *fp1,*fp2;/*pointer to the file */
-
char ch[ARRAY_SIZE];
-
char *vector ;/* the bit space*/
-
vector = (char *)calloc(SIZE,sizeof(char));
-
-
printf("Please enter the file with repeated urls:\n");
-
scanf("%s",&file1);
-
if( (fp1 = fopen(file1,"rb")) == NULL) { /* open the goal file*/
-
printf("Connot open the file %s!\n",file1);
-
}
-
-
printf("Please enter the file you want to save to:\n");
-
scanf("%s",&file2);
-
if( (fp2 = fopen(file2,"w")) == NULL) {
-
printf("Connot open the file %s\n",file2);
-
}
-
strftime(buf,32,"%Y-%m-%d %H:%M:%S",localtime(&tmp));
-
printf("%s\n",buf); /*print the system time*/
-
-
for(i=0;i<SIZE;i++) {
-
vector[i]=0; /*set 0*/
-
}
-
-
while(!feof(fp1)) { /* the check process*/
-
-
fgets(ch,ARRAY_SIZE,fp1);
-
flag=0;
-
tt++;
-
if( GETBIT(vector, HFLPHash(ch,len(ch))%MAX) ) {
-
flag++;
-
} else {
-
SETBIT(vector,HFLPHash(ch,len(ch))%MAX );
-
}
-
-
if( GETBIT(vector, StrHash(ch,len(ch))%MAX) ) {
-
flag++;
-
} else {
-
SETBIT(vector,StrHash(ch,len(ch))%MAX );
-
}
-
-
if( GETBIT(vector, HFHash(ch,len(ch))%MAX) ) {
-
flag++;
-
} else {
-
SETBIT(vector,HFHash(ch,len(ch))%MAX );
-
}
-
-
if( GETBIT(vector, DEKHash(ch,len(ch))%MAX) ) {
-
flag++;
-
} else {
-
SETBIT(vector,DEKHash(ch,len(ch))%MAX );
-
}
-
-
if( GETBIT(vector, TianlHash(ch,len(ch))%MAX) ) {
-
flag++;
-
} else {
-
SETBIT(vector,TianlHash(ch,len(ch))%MAX );
-
}
-
-
if( GETBIT(vector, SDBMHash(ch,len(ch))%MAX) ) {
-
flag++;
-
} else {
-
SETBIT(vector,SDBMHash(ch,len(ch))%MAX );
-
}
-
-
if(flag<6)
-
num2++;
-
else
-
fputs(ch,fp2);
-
-
/* printf(" %d",flag); */
-
}
-
/* the result*/
-
printf("\nThere are %d urls!\n",tt);
-
printf("\nThere are %d not repeated urls!\n",num2);
-
printf("There are %d repeated urls!\n",tt-num2);
-
fclose(fp1);
-
fclose(fp2);
-
return 0;
-
}
-
-
-
/*functions may be used in the main */
-
unsigned int len(char *ch)
-
{
-
int m=0;
-
while(ch[m]!='\0') {
-
m++;
-
}
-
return m;
-
}
-
-
unsigned int RSHash(char* str, unsigned int len) {
-
unsigned int b = 378551;
-
unsigned int a = 63689;
-
unsigned int hash = 0;
-
unsigned int i = 0;
-
-
for(i=0; i<len; str++, i++) {
-
hash = hash*a + (*str);
-
a = a*b;
-
}
-
return hash;
-
}
-
/* End Of RS Hash Function */
-
-
-
unsigned int JSHash(char* str, unsigned int len)
-
{
-
unsigned int hash = 1315423911;
-
unsigned int i = 0;
-
-
for(i=0; i<len; str++, i++) {
-
hash ^= ((hash<<5) + (*str) + (hash>>2));
-
}
-
return hash;
-
}
-
/* End Of JS Hash Function */
-
-
-
unsigned int PJWHash(char* str, unsigned int len)
-
{
-
const unsigned int BitsInUnsignedInt = (unsigned int)(sizeof(unsigned int) * 8);
-
const unsigned int ThreeQuarters = (unsigned int)((BitsInUnsignedInt * 3) / 4);
-
const unsigned int OneEighth = (unsigned int)(BitsInUnsignedInt / 8);
-
const unsigned int HighBits = (unsigned int)(0xFFFFFFFF) << (BitsInUnsignedInt - OneEighth);
-
unsigned int hash = 0;
-
unsigned int test = 0;
-
unsigned int i = 0;
-
-
for(i=0;i<len; str++, i++) {
-
hash = (hash<<OneEighth) + (*str);
-
if((test = hash & HighBits) != 0) {
-
hash = ((hash ^(test >> ThreeQuarters)) & (~HighBits));
-
}
-
}
-
-
return hash;
-
}
-
/* End Of P. J. Weinberger Hash Function */
-
-
-
unsigned int ELFHash(char* str, unsigned int len)
-
{
-
unsigned int hash = 0;
-
unsigned int x = 0;
-
unsigned int i = 0;
-
-
for(i = 0; i < len; str++, i++) {
-
hash = (hash << 4) + (*str);
-
if((x = hash & 0xF0000000L) != 0) {
-
hash ^= (x >> 24);
-
}
-
hash &= ~x;
-
}
-
return hash;
-
}
-
/* End Of ELF Hash Function */
-
-
-
unsigned int BKDRHash(char* str, unsigned int len)
-
{
-
unsigned int seed = 131; /* 31 131 1313 13131 131313 etc.. */
-
unsigned int hash = 0;
-
unsigned int i = 0;
-
-
for(i = 0; i < len; str++, i++)
-
{
-
hash = (hash * seed) + (*str);
-
}
-
-
return hash;
-
}
-
/* End Of BKDR Hash Function */
-
-
-
unsigned int SDBMHash(char* str, unsigned int len)
-
{
-
unsigned int hash = 0;
-
unsigned int i = 0;
-
-
for(i = 0; i < len; str++, i++) {
-
hash = (*str) + (hash << 6) + (hash << 16) - hash;
-
}
-
-
return hash;
-
}
-
/* End Of SDBM Hash Function */
-
-
-
unsigned int DJBHash(char* str, unsigned int len)
-
{
-
unsigned int hash = 5381;
-
unsigned int i = 0;
-
-
for(i = 0; i < len; str++, i++) {
-
hash = ((hash << 5) + hash) + (*str);
-
}
-
-
return hash;
-
}
-
/* End Of DJB Hash Function */
-
-
-
unsigned int DEKHash(char* str, unsigned int len)
-
{
-
unsigned int hash = len;
-
unsigned int i = 0;
-
-
for(i = 0; i < len; str++, i++) {
-
hash = ((hash << 5) ^ (hash >> 27)) ^ (*str);
-
}
-
return hash;
-
}
-
/* End Of DEK Hash Function */
-
-
-
unsigned int BPHash(char* str, unsigned int len)
-
{
-
unsigned int hash = 0;
-
unsigned int i = 0;
-
for(i = 0; i < len; str++, i++) {
-
hash = hash << 7 ^ (*str);
-
}
-
-
return hash;
-
}
-
/* End Of BP Hash Function */
-
-
-
unsigned int FNVHash(char* str, unsigned int len)
-
{
-
const unsigned int fnv_prime = 0x811C9DC5;
-
unsigned int hash = 0;
-
unsigned int i = 0;
-
-
for(i = 0; i < len; str++, i++) {
-
hash *= fnv_prime;
-
hash ^= (*str);
-
}
-
-
return hash;
-
}
-
/* End Of FNV Hash Function */
-
-
-
unsigned int APHash(char* str, unsigned int len)
-
{
-
unsigned int hash = 0xAAAAAAAA;
-
unsigned int i = 0;
-
-
for(i = 0; i < len; str++, i++) {
-
hash ^= ((i & 1) == 0) ? ( (hash << 7) ^ (*str) * (hash >> 3)) :
-
(~((hash << 11) + (*str) ^ (hash >> 5)));
-
}
-
-
return hash;
-
}
-
/* End Of AP Hash Function */
-
unsigned int HFLPHash(char *str,unsigned int len)
-
{
-
unsigned int n=0;
-
int i;
-
char* b=(char *)&n;
-
for(i=0;i<strlen(str);++i) {
-
b[i%4]^=str[i];
-
}
-
return n%len;
-
}
-
/* End Of HFLP Hash Function*/
-
unsigned int HFHash(char* str,unsigned int len)
-
{
-
int result=0;
-
char* ptr=str;
-
int c;
-
int i=0;
-
for (i=1;c=*ptr++;i++)
-
result += c*3*i;
-
if (result<0)
-
result = -result;
-
return result%len;
-
}
-
/*End Of HKHash Function */
-
-
unsigned int StrHash( char *str,unsigned int len)
-
{
-
register unsigned int h;
-
register unsigned char *p;
-
for(h=0,p=(unsigned char *)str;*p;p++) {
-
h=31*h+*p;
-
}
-
-
return h;
-
-
}
-
/*End Of StrHash Function*/
-
-
unsigned int TianlHash(char *str,unsigned int len)
-
{
-
unsigned long urlHashValue=0;
-
int ilength=strlen(str);
-
int i;
-
unsigned char ucChar;
-
if(!ilength) {
-
return 0;
-
}
-
if(ilength<=256) {
-
urlHashValue=16777216*(ilength-1);
-
} else {
-
urlHashValue = 42781900080;
-
}
-
if(ilength<=96) {
-
for(i=1;i<=ilength;i++) {
-
ucChar=str[i-1];
-
if(ucChar<='Z'&&ucChar>='A') {
-
ucChar=ucChar+32;
-
}
-
urlHashValue+=(3*i*ucChar*ucChar+5*i*ucChar+7*i+11*ucChar)%1677216;
-
}
-
} else {
-
for(i=1;i<=96;i++)
-
{
-
ucChar=str[i+ilength-96-1];
-
if(ucChar<='Z'&&ucChar>='A')
-
{
-
ucChar=ucChar+32;
-
}
-
urlHashValue+=(3*i*ucChar*ucChar+5*i*ucChar+7*i+11*ucChar)%1677216;
-
}
-
}
-
return urlHashValue;
-
-
}
-
/*End Of Tianl Hash Function*/
#include "stdafx.h"
#define ARRAY_SIZE 256 /*we get the 256 chars of each line*/
#define SIZE 48000000 /* size should be 1/8 of max*/
#define MAX 384000000/*the max bit space*/
#define SETBIT(ch,n) ch[n/8]|=1<<(7-n%8)
#define GETBIT(ch,n) (ch[n/8]&1<<(7-n%8))>>(7-n%8)
unsigned int len(char *ch);/* functions to calculate the length of the url*/
unsigned int RSHash(char* str, unsigned int len);/* functions to calculate the hash value of the url*/
unsigned int JSHash(char* str, unsigned int len);/* functions to calculate the hash value of the url*/
unsigned int PJWHash(char* str, unsigned int len);/* functions to calculate the hash value of the url*/
unsigned int ELFHash(char* str, unsigned int len);/* functions to calculate the hash value of the url*/
unsigned int BKDRHash(char* str, unsigned int len);/* functions to calculate the hash value of the url*/
unsigned int SDBMHash(char* str, unsigned int len);/* functions to calculate the hash value of the url*/
unsigned int DJBHash(char* str, unsigned int len);/* functions to calculate the hash value of the url*/
unsigned int DEKHash(char* str, unsigned int len);/* functions to calculate the hash value of the url*/
unsigned int BPHash(char* str, unsigned int len);/* functions to calculate the hash value of the url*/
unsigned int FNVHash(char* str, unsigned int len);/* functions to calculate the hash value of the url*/
unsigned int APHash(char* str, unsigned int len);/* functions to calculate the hash value of the url*/
unsigned int HFLPHash(char* str,unsigned int len);/* functions to calculate the hash value of the url*/
unsigned int HFHash(char* str,unsigned int len);/* functions to calculate the hash value of the url*/
unsigned int StrHash( char* str,unsigned int len);/* functions to calculate the hash value of the url*/
unsigned int TianlHash(char* str,unsigned int len);/* functions to calculate the hash value of the url*/
int main()
{
int i,num,num2=0; /* the number to record the repeated urls and the total of it*/
unsigned int tt=0;
int flag; /*it helps to check weather the url has already existed */
char buf[257]; /*it helps to print the start time of the program */
time_t tmp = time(NULL);
char file1[100],file2[100];
FILE *fp1,*fp2;/*pointer to the file */
char ch[ARRAY_SIZE];
char *vector ;/* the bit space*/
vector = (char *)calloc(SIZE,sizeof(char));
printf("Please enter the file with repeated urls:\n");
scanf("%s",&file1);
if( (fp1 = fopen(file1,"rb")) == NULL) { /* open the goal file*/
printf("Connot open the file %s!\n",file1);
}
printf("Please enter the file you want to save to:\n");
scanf("%s",&file2);
if( (fp2 = fopen(file2,"w")) == NULL) {
printf("Connot open the file %s\n",file2);
}
strftime(buf,32,"%Y-%m-%d %H:%M:%S",localtime(&tmp));
printf("%s\n",buf); /*print the system time*/
for(i=0;i<SIZE;i++) {
vector[i]=0; /*set 0*/
}
while(!feof(fp1)) { /* the check process*/
fgets(ch,ARRAY_SIZE,fp1);
flag=0;
tt++;
if( GETBIT(vector, HFLPHash(ch,len(ch))%MAX) ) {
flag++;
} else {
SETBIT(vector,HFLPHash(ch,len(ch))%MAX );
}
if( GETBIT(vector, StrHash(ch,len(ch))%MAX) ) {
flag++;
} else {
SETBIT(vector,StrHash(ch,len(ch))%MAX );
}
if( GETBIT(vector, HFHash(ch,len(ch))%MAX) ) {
flag++;
} else {
SETBIT(vector,HFHash(ch,len(ch))%MAX );
}
if( GETBIT(vector, DEKHash(ch,len(ch))%MAX) ) {
flag++;
} else {
SETBIT(vector,DEKHash(ch,len(ch))%MAX );
}
if( GETBIT(vector, TianlHash(ch,len(ch))%MAX) ) {
flag++;
} else {
SETBIT(vector,TianlHash(ch,len(ch))%MAX );
}
if( GETBIT(vector, SDBMHash(ch,len(ch))%MAX) ) {
flag++;
} else {
SETBIT(vector,SDBMHash(ch,len(ch))%MAX );
}
if(flag<6)
num2++;
else
fputs(ch,fp2);
/* printf(" %d",flag); */
}
/* the result*/
printf("\nThere are %d urls!\n",tt);
printf("\nThere are %d not repeated urls!\n",num2);
printf("There are %d repeated urls!\n",tt-num2);
fclose(fp1);
fclose(fp2);
return 0;
}
/*functions may be used in the main */
unsigned int len(char *ch)
{
int m=0;
while(ch[m]!='\0') {
m++;
}
return m;
}
unsigned int RSHash(char* str, unsigned int len) {
unsigned int b = 378551;
unsigned int a = 63689;
unsigned int hash = 0;
unsigned int i = 0;
for(i=0; i<len; str++, i++) {
hash = hash*a + (*str);
a = a*b;
}
return hash;
}
/* End Of RS Hash Function */
unsigned int JSHash(char* str, unsigned int len)
{
unsigned int hash = 1315423911;
unsigned int i = 0;
for(i=0; i<len; str++, i++) {
hash ^= ((hash<<5) + (*str) + (hash>>2));
}
return hash;
}
/* End Of JS Hash Function */
unsigned int PJWHash(char* str, unsigned int len)
{
const unsigned int BitsInUnsignedInt = (unsigned int)(sizeof(unsigned int) * 8);
const unsigned int ThreeQuarters = (unsigned int)((BitsInUnsignedInt * 3) / 4);
const unsigned int OneEighth = (unsigned int)(BitsInUnsignedInt / 8);
const unsigned int HighBits = (unsigned int)(0xFFFFFFFF) << (BitsInUnsignedInt - OneEighth);
unsigned int hash = 0;
unsigned int test = 0;
unsigned int i = 0;
for(i=0;i<len; str++, i++) {
hash = (hash<<OneEighth) + (*str);
if((test = hash & HighBits) != 0) {
hash = ((hash ^(test >> ThreeQuarters)) & (~HighBits));
}
}
return hash;
}
/* End Of P. J. Weinberger Hash Function */
unsigned int ELFHash(char* str, unsigned int len)
{
unsigned int hash = 0;
unsigned int x = 0;
unsigned int i = 0;
for(i = 0; i < len; str++, i++) {
hash = (hash << 4) + (*str);
if((x = hash & 0xF0000000L) != 0) {
hash ^= (x >> 24);
}
hash &= ~x;
}
return hash;
}
/* End Of ELF Hash Function */
unsigned int BKDRHash(char* str, unsigned int len)
{
unsigned int seed = 131; /* 31 131 1313 13131 131313 etc.. */
unsigned int hash = 0;
unsigned int i = 0;
for(i = 0; i < len; str++, i++)
{
hash = (hash * seed) + (*str);
}
return hash;
}
/* End Of BKDR Hash Function */
unsigned int SDBMHash(char* str, unsigned int len)
{
unsigned int hash = 0;
unsigned int i = 0;
for(i = 0; i < len; str++, i++) {
hash = (*str) + (hash << 6) + (hash << 16) - hash;
}
return hash;
}
/* End Of SDBM Hash Function */
unsigned int DJBHash(char* str, unsigned int len)
{
unsigned int hash = 5381;
unsigned int i = 0;
for(i = 0; i < len; str++, i++) {
hash = ((hash << 5) + hash) + (*str);
}
return hash;
}
/* End Of DJB Hash Function */
unsigned int DEKHash(char* str, unsigned int len)
{
unsigned int hash = len;
unsigned int i = 0;
for(i = 0; i < len; str++, i++) {
hash = ((hash << 5) ^ (hash >> 27)) ^ (*str);
}
return hash;
}
/* End Of DEK Hash Function */
unsigned int BPHash(char* str, unsigned int len)
{
unsigned int hash = 0;
unsigned int i = 0;
for(i = 0; i < len; str++, i++) {
hash = hash << 7 ^ (*str);
}
return hash;
}
/* End Of BP Hash Function */
unsigned int FNVHash(char* str, unsigned int len)
{
const unsigned int fnv_prime = 0x811C9DC5;
unsigned int hash = 0;
unsigned int i = 0;
for(i = 0; i < len; str++, i++) {
hash *= fnv_prime;
hash ^= (*str);
}
return hash;
}
/* End Of FNV Hash Function */
unsigned int APHash(char* str, unsigned int len)
{
unsigned int hash = 0xAAAAAAAA;
unsigned int i = 0;
for(i = 0; i < len; str++, i++) {
hash ^= ((i & 1) == 0) ? ( (hash << 7) ^ (*str) * (hash >> 3)) :
(~((hash << 11) + (*str) ^ (hash >> 5)));
}
return hash;
}
/* End Of AP Hash Function */
unsigned int HFLPHash(char *str,unsigned int len)
{
unsigned int n=0;
int i;
char* b=(char *)&n;
for(i=0;i<strlen(str);++i) {
b[i%4]^=str[i];
}
return n%len;
}
/* End Of HFLP Hash Function*/
unsigned int HFHash(char* str,unsigned int len)
{
int result=0;
char* ptr=str;
int c;
int i=0;
for (i=1;c=*ptr++;i++)
result += c*3*i;
if (result<0)
result = -result;
return result%len;
}
/*End Of HKHash Function */
unsigned int StrHash( char *str,unsigned int len)
{
register unsigned int h;
register unsigned char *p;
for(h=0,p=(unsigned char *)str;*p;p++) {
h=31*h+*p;
}
return h;
}
/*End Of StrHash Function*/
unsigned int TianlHash(char *str,unsigned int len)
{
unsigned long urlHashValue=0;
int ilength=strlen(str);
int i;
unsigned char ucChar;
if(!ilength) {
return 0;
}
if(ilength<=256) {
urlHashValue=16777216*(ilength-1);
} else {
urlHashValue = 42781900080;
}
if(ilength<=96) {
for(i=1;i<=ilength;i++) {
ucChar=str[i-1];
if(ucChar<='Z'&&ucChar>='A') {
ucChar=ucChar+32;
}
urlHashValue+=(3*i*ucChar*ucChar+5*i*ucChar+7*i+11*ucChar)%1677216;
}
} else {
for(i=1;i<=96;i++)
{
ucChar=str[i+ilength-96-1];
if(ucChar<='Z'&&ucChar>='A')
{
ucChar=ucChar+32;
}
urlHashValue+=(3*i*ucChar*ucChar+5*i*ucChar+7*i+11*ucChar)%1677216;
}
}
return urlHashValue;
}
/*End Of Tianl Hash Function*/
网上找到的php简单实现:
[cpp] view plaincopyprint?
-
<?php
-
-
/**
-
* Implements a Bloom Filter
-
*/
-
class BloomFilter {
-
/**
-
* Size of the bit array
-
*
-
* @var int
-
*/
-
protected $m;
-
-
/**
-
* Number of hash functions
-
*
-
* @var int
-
*/
-
protected $k;
-
-
/**
-
* Number of elements in the filter
-
*
-
* @var int
-
*/
-
protected $n;
-
-
/**
-
* The bitset holding the filter information
-
*
-
* @var array
-
*/
-
protected $bitset;
-
-
/**
-
* 计算最优的hash函数个数:当hash函数个数k=(ln2)*(m/n)时错误率最小
-
*
-
* @param int $m bit数组的宽度(bit数)
-
* @param int $n 加入布隆过滤器的key的数量
-
* @return int
-
*/
-
public static function getHashCount($m, $n) {
-
return ceil(($m / $n) * log(2));
-
}
-
-
/**
-
* Construct an instance of the Bloom filter
-
*
-
* @param int $m bit数组的宽度(bit数) Size of the bit array
-
* @param int $k hash函数的个数 Number of different hash functions to use
-
*/
-
public function __construct($m, $k) {
-
$this->m = $m;
-
$this->k = $k;
-
$this->n = 0;
-
-
/* Initialize the bit set */
-
$this->bitset = array_fill(0, $this->m - 1, false);
-
}
-
-
/**
-
* False Positive的比率:f = (1 – e-kn/m)k
-
* Returns the probability for a false positive to occur, given the current number of items in the filter
-
*
-
* @return double
-
*/
-
public function getFalsePositiveProbability() {
-
$exp = (-1 * $this->k * $this->n) / $this->m;
-
-
return pow(1 - exp($exp), $this->k);
-
}
-
-
/**
-
* Adds a new item to the filter
-
*
-
* @param mixed Either a string holding a single item or an array of
-
* string holding multiple items. In the latter case, all
-
* items are added one by one internally.
-
*/
-
public function add($key) {
-
if (is_array($key)) {
-
foreach ($key as $k) {
-
$this->add($k);
-
}
-
return;
-
}
-
-
$this->n++;
-
-
foreach ($this->getSlots($key) as $slot) {
-
$this->bitset[$slot] = true;
-
}
-
}
-
-
/**
-
* Queries the Bloom filter for an element
-
*
-
* If this method return FALSE, it is 100% certain that the element has
-
* not been added to the filter before. In contrast, if TRUE is returned,
-
* the element *may* have been added to the filter previously. However with
-
* a probability indicated by getFalsePositiveProbability() the element has
-
* not been added to the filter with contains() still returning TRUE.
-
*
-
* @param mixed Either a string holding a single item or an array of
-
* strings holding multiple items. In the latter case the
-
* method returns TRUE if the filter contains all items.
-
* @return boolean
-
*/
-
public function contains($key) {
-
if (is_array($key)) {
-
foreach ($key as $k) {
-
if ($this->contains($k) == false) {
-
return false;
-
}
-
}
-
-
return true;
-
}
-
-
foreach ($this->getSlots($key) as $slot) {
-
if ($this->bitset[$slot] == false) {
-
return false;
-
}
-
}
-
-
return true;
-
}
-
-
/**
-
* Hashes the argument to a number of positions in the bit set and returns the positions
-
*
-
* @param string Item
-
* @return array Positions
-
*/
-
protected function getSlots($key) {
-
$slots = array();
-
$hash = self::getHashCode($key);
-
mt_srand($hash);
-
-
for ($i = 0; $i < $this->k; $i++) {
-
$slots[] = mt_rand(0, $this->m - 1);
-
}
-
-
return $slots;
-
}
-
-
/**
-
* 使用CRC32产生一个32bit(位)的校验值。
-
* 由于CRC32产生校验值时源数据块的每一bit(位)都会被计算,所以数据块中即使只有一位发生了变化,也会得到不同的CRC32值。
-
* Generates a numeric hash for the given string
-
*
-
* Right now the CRC-32 algorithm is used. Alternatively one could e.g.
-
* use Adler digests or mimick the behaviour of Java's hashCode() method.
-
*
-
* @param string Input for which the hash should be created
-
* @return int Numeric hash
-
*/
-
protected static function getHashCode($string) {
-
return crc32($string);
-
}
-
-
}
-
-
-
-
$items = array("first item", "second item", "third item");
-
-
/* Add all items with one call to add() and make sure contains() finds
-
* them all.
-
*/
-
$filter = new BloomFilter(100, BloomFilter::getHashCount(100, 3));
-
$filter->add($items);
-
-
//var_dump($filter); exit;
-
$items = array("firsttem", "seconditem", "thirditem");
-
foreach ($items as $item) {
-
var_dump(($filter->contains($item)));
-
}
-
-
-
/* Add all items with multiple calls to add() and make sure contains()
-
* finds them all.
-
*/
-
$filter = new BloomFilter(100, BloomFilter::getHashCount(100, 3));
-
foreach ($items as $item) {
-
$filter->add($item);
-
}
-
$items = array("fir sttem", "secondit em", "thir ditem");
-
foreach ($items as $item) {
-
var_dump(($filter->contains($item)));
-
}
-
-
-
-
<?php
/**
* Implements a Bloom Filter
*/
class BloomFilter {
/**
* Size of the bit array
*
* @var int
*/
protected $m;
/**
* Number of hash functions
*
* @var int
*/
protected $k;
/**
* Number of elements in the filter
*
* @var int
*/
protected $n;
/**
* The bitset holding the filter information
*
* @var array
*/
protected $bitset;
/**
* 计算最优的hash函数个数:当hash函数个数k=(ln2)*(m/n)时错误率最小
*
* @param int $m bit数组的宽度(bit数)
* @param int $n 加入布隆过滤器的key的数量
* @return int
*/
public static function getHashCount($m, $n) {
return ceil(($m / $n) * log(2));
}
/**
* Construct an instance of the Bloom filter
*
* @param int $m bit数组的宽度(bit数) Size of the bit array
* @param int $k hash函数的个数 Number of different hash functions to use
*/
public function __construct($m, $k) {
$this->m = $m;
$this->k = $k;
$this->n = 0;
/* Initialize the bit set */
$this->bitset = array_fill(0, $this->m - 1, false);
}
/**
* False Positive的比率:f = (1 – e-kn/m)k
* Returns the probability for a false positive to occur, given the current number of items in the filter
*
* @return double
*/
public function getFalsePositiveProbability() {
$exp = (-1 * $this->k * $this->n) / $this->m;
return pow(1 - exp($exp), $this->k);
}
/**
* Adds a new item to the filter
*
* @param mixed Either a string holding a single item or an array of
* string holding multiple items. In the latter case, all
* items are added one by one internally.
*/
public function add($key) {
if (is_array($key)) {
foreach ($key as $k) {
$this->add($k);
}
return;
}
$this->n++;
foreach ($this->getSlots($key) as $slot) {
$this->bitset[$slot] = true;
}
}
/**
* Queries the Bloom filter for an element
*
* If this method return FALSE, it is 100% certain that the element has
* not been added to the filter before. In contrast, if TRUE is returned,
* the element *may* have been added to the filter previously. However with
* a probability indicated by getFalsePositiveProbability() the element has
* not been added to the filter with contains() still returning TRUE.
*
* @param mixed Either a string holding a single item or an array of
* strings holding multiple items. In the latter case the
* method returns TRUE if the filter contains all items.
* @return boolean
*/
public function contains($key) {
if (is_array($key)) {
foreach ($key as $k) {
if ($this->contains($k) == false) {
return false;
}
}
return true;
}
foreach ($this->getSlots($key) as $slot) {
if ($this->bitset[$slot] == false) {
return false;
}
}
return true;
}
/**
* Hashes the argument to a number of positions in the bit set and returns the positions
*
* @param string Item
* @return array Positions
*/
protected function getSlots($key) {
$slots = array();
$hash = self::getHashCode($key);
mt_srand($hash);
for ($i = 0; $i < $this->k; $i++) {
$slots[] = mt_rand(0, $this->m - 1);
}
return $slots;
}
/**
* 使用CRC32产生一个32bit(位)的校验值。
* 由于CRC32产生校验值时源数据块的每一bit(位)都会被计算,所以数据块中即使只有一位发生了变化,也会得到不同的CRC32值。
* Generates a numeric hash for the given string
*
* Right now the CRC-32 algorithm is used. Alternatively one could e.g.
* use Adler digests or mimick the behaviour of Java's hashCode() method.
*
* @param string Input for which the hash should be created
* @return int Numeric hash
*/
protected static function getHashCode($string) {
return crc32($string);
}
}
$items = array("first item", "second item", "third item");
/* Add all items with one call to add() and make sure contains() finds
* them all.
*/
$filter = new BloomFilter(100, BloomFilter::getHashCount(100, 3));
$filter->add($items);
//var_dump($filter); exit;
$items = array("firsttem", "seconditem", "thirditem");
foreach ($items as $item) {
var_dump(($filter->contains($item)));
}
/* Add all items with multiple calls to add() and make sure contains()
* finds them all.
*/
$filter = new BloomFilter(100, BloomFilter::getHashCount(100, 3));
foreach ($items as $item) {
$filter->add($item);
}
$items = array("fir sttem", "secondit em", "thir ditem");
foreach ($items as $item) {
var_dump(($filter->contains($item)));
}
问题实例】 给你A,B两个文件,各存放50亿条URL,每条URL占用64字节,内存限制是4G,让你找出A,B文件共同的URL。如果是三个乃至n个文件呢?
根据这个问题我们来计算下内存的占用,4G=2^32大概是40亿*8大概是340亿bit,n=50亿,如果按出错率0.01算需要的大概是650亿个bit。 现在可用的是340亿,相差并不多,这样可能会使出错率上升些。另外如果这些urlip是一一对应的,就可以转换成ip,则大大简单了。
(责任编辑:IT)
1. Bloom-Filter算法简介Bloom-Filter,即布隆过滤器,1970年由Bloom中提出。它可以用于检索一个元素是否在一个集合中。Bloom Filter(BF)是一种空间效率很高的随机数据结构,它利用位数组很简洁地表示一个集合,并能判断一个元素是否属于这个集合。它是一个判断元素是否存在集合的快速的概率算法。Bloom Filter有可能会出现错误判断,但不会漏掉判断。也就是Bloom Filter判断元素不再集合,那肯定不在。如果判断元素存在集合中,有一定的概率判断错误。因此,Bloom Filter不适合那些“零错误”的应用场合。而在能容忍低错误率的应用场合下,Bloom Filter比其他常见的算法(如hash,折半查找)极大节省了空间。 它的优点是空间效率和查询时间都远远超过一般的算法,缺点是有一定的误识别率和删除困难。 Bloom Filter的详细介绍:Bloom Filter
2、 Bloom-Filter的基本思想Bloom-Filter算法的核心思想就是利用多个不同的Hash函数来解决“冲突”。计算某元素x是否在一个集合中,首先能想到的方法就是将所有的已知元素保存起来构成一个集合R,然后用元素x跟这些R中的元素一一比较来判断是否存在于集合R中;我们可以采用链表等数据结构来实现。但是,随着集合R中元素的增加,其占用的内存将越来越大。试想,如果有几千万个不同网页需要下载,所需的内存将足以占用掉整个进程的内存地址空间。即使用MD5,UUID这些方法将URL转成固定的短小的字符串,内存占用也是相当巨大的。 于是,我们会想到用Hash table的数据结构,运用一个足够好的Hash函数将一个URL映射到二进制位数组(位图数组)中的某一位。如果该位已经被置为1,那么表示该URL已经存在。 Hash存在一个冲突(碰撞)的问题,用同一个Hash得到的两个URL的值有可能相同。为了减少冲突,我们可以多引入几个Hash,如果通过其中的一个Hash值我们得出某元素不在集合中,那么该元素肯定不在集合中。只有在所有的Hash函数告诉我们该元素在集合中时,才能确定该元素存在于集合中。这便是Bloom-Filter的基本思想。
原理要点:一是位数组, 而是k个独立hash函数。 1)位数组: 假设Bloom Filter使用一个m比特的数组来保存信息,初始状态时,Bloom Filter是一个包含m位的位数组,每一位都置为0,即BF整个数组的元素都设置为0。
2)添加元素,k个独立hash函数 为了表达S={x1, x2,…,xn}这样一个n个元素的集合,Bloom Filter使用k个相互独立的哈希函数(Hash Function),它们分别将集合中的每个元素映射到{1,…,m}的范围中。 当我们往Bloom Filter中增加任意一个元素x时候,我们使用k个哈希函数得到k个哈希值,然后将数组中对应的比特位设置为1。即第i个哈希函数映射的位置hashi(x)就会被置为1(1≤i≤k)。 注意,如果一个位置多次被置为1,那么只有第一次会起作用,后面几次将没有任何效果。在下图中,k=3,且有两个哈希函数选中同一个位置(从左边数第五位,即第二个“1“处)。
3)判断元素是否存在集合 在判断y是否属于这个集合时,我们只需要对y使用k个哈希函数得到k个哈希值,如果所有hashi(y)的位置都是1(1≤i≤k),即k个位置都被设置为1了,那么我们就认为y是集合中的元素,否则就认为y不是集合中的元素。下图中y1就不是集合中的元素(因为y1有一处指向了“0”位)。y2或者属于这个集合,或者刚好是一个false positive。
显然这 个判断并不保证查找的结果是100%正确的。 Bloom Filter的缺点: 1)Bloom Filter无法从Bloom Filter集合中删除一个元素。因为该元素对应的位会牵动到其他的元素。所以一个简单的改进就是 counting Bloom filter,用一个counter数组代替位数组,就可以支持删除了。 此外,Bloom Filter的hash函数选择会影响算法的效果。 2)还有一个比较重要的问题,如何根据输入元素个数n,确定位数组m的大小及hash函数个数,即hash函数选择会影响算法的效果。当hash函数个数k=(ln2)*(m/n)时错误率最小。在错误率不大于E的情况 下,m至少要等于n*lg(1/E) 才能表示任意n个元素的集合。但m还应该更大些,因为还要保证bit数组里至少一半为0,则m应 该>=nlg(1/E)*lge ,大概就是nlg(1/E)1.44倍(lg表示以2为底的对数)。 举个例子我们假设错误率为0.01,则此时m应大概是n的13倍。这样k大概是8个。 注意: 这里m与n的单位不同,m是bit为单位,而n则是以元素个数为单位(准确的说是不同元素的个数)。通常单个元素的长度都是有很多bit的。所以使用bloom filter内存上通常都是节省的。 一般BF可以与一些key-value的数据库一起使用,来加快查询。由于BF所用的空间非常小,所有BF可以常驻内存。这样子的话,对于大部分不存在的元素,我们只需要访问内存中的BF就可以判断出来了,只有一小部分,我们需要访问在硬盘上的key-value数据库。从而大大地提高了效率。
一个Bloom Filter有以下参数:
Bloom Filter的f满足下列公式:
在给定m和n时,能够使f最小化的k值为:
此时给出的f为:
根据以上公式,对于任意给定的f,我们有:
n = m ln(0.6185) / ln(f)
同时,我们需要k个hash来达成这个目标:
k = - ln(f) / ln(2)
由于k必须取整数,我们在Bloom Filter的程序实现中,还应该使用上面的公式来求得实际的f:
f = (1 – e-kn/m)k
以上3个公式是程序实现Bloom Filter的关键公式。
3、 扩展 CounterBloom FilterCounterBloom FilterBloomFilter有个缺点,就是不支持删除操作,因为它不知道某一个位从属于哪些向量。那我们可以给Bloom Filter加上计数器,添加时增加计数器,删除时减少计数器。 但这样的Filter需要考虑附加的计数器大小,假如同个元素多次插入的话,计数器位数较少的情况下,就会出现溢出问题。如果对计数器设置上限值的话,会导致Cache Miss,但对某些应用来说,这并不是什么问题,如Web Sharing。 Compressed Bloom Filter为了能在服务器之间更快地通过网络传输Bloom Filter,我们有方法能在已完成Bloom Filter之后,得到一些实际参数的情况下进行压缩。 将元素全部添加入Bloom Filter后,我们能得到真实的空间使用率,用这个值代入公式计算出一个比m小的值,重新构造Bloom Filter,对原先的哈希值进行求余处理,在误判率不变的情况下,使得其内存大小更合适。 4、 Bloom-Filter的应用Bloom-Filter一般用于在大数据量的集合中判定某元素是否存在。例如邮件服务器中的垃圾邮件过滤器。在搜索引擎领域,Bloom-Filter最常用于网络蜘蛛(Spider)的URL过滤,网络蜘蛛通常有一个URL列表,保存着将要下载和已经下载的网页的URL,网络蜘蛛下载了一个网页,从网页中提取到新的URL后,需要判断该URL是否已经存在于列表中。此时,Bloom-Filter算法是最好的选择。 1.key-value 加快查询 一般Bloom-Filter可以与一些key-value的数据库一起使用,来加快查询。 一般key-value存储系统的values存在硬盘,查询就是件费时的事。将Storage的数据都插入Filter,在Filter中查询都不存在时,那就不需要去Storage查询了。当False Position出现时,只是会导致一次多余的Storage查询。 由于Bloom-Filter所用的空间非常小,所有BF可以常驻内存。这样子的话,对于大部分不存在的元素,我们只需要访问内存中的Bloom-Filter就可以判断出来了,只有一小部分,我们需要访问在硬盘上的key-value数据库。从而大大地提高了效率。如图:
2 .Google的BigTable Google的BigTable也使用了Bloom Filter,以减少不存在的行或列在磁盘上的查询,大大提高了数据库的查询操作的性能。 3. Proxy-Cache 在Internet Cache Protocol中的Proxy-Cache很多都是使用Bloom Filter存储URLs,除了高效的查询外,还能很方便得传输交换Cache信息。 4.网络应用1)P2P网络中查找资源操作,可以对每条网络通路保存Bloom Filter,当命中时,则选择该通路访问。 2)广播消息时,可以检测某个IP是否已发包。 3)检测广播消息包的环路,将Bloom Filter保存在包里,每个节点将自己添加入Bloom Filter。 4)信息队列管理,使用Counter Bloom Filter管理信息流量。 5. 垃圾邮件地址过滤像网易,QQ这样的公众电子邮件(email)提供商,总是需要过滤来自发送垃圾邮件的人(spamer)的垃圾邮件。 一个办法就是记录下那些发垃圾邮件的 email地址。由于那些发送者不停地在注册新的地址,全世界少说也有几十亿个发垃圾邮件的地址,将他们都存起来则需要大量的网络服务器。 如果用哈希表,每存储一亿个 email地址,就需要 1.6GB的内存(用哈希表实现的具体办法是将每一个 email地址对应成一个八字节的信息指纹,然后将这些信息指纹存入哈希表,由于哈希表的存储效率一般只有 50%,因此一个 email地址需要占用十六个字节。一亿个地址大约要 1.6GB,即十六亿字节的内存)。因此存贮几十亿个邮件地址可能需要上百 GB的内存。 而Bloom Filter只需要哈希表 1/8到 1/4 的大小就能解决同样的问题。 BloomFilter决不会漏掉任何一个在黑名单中的可疑地址。而至于误判问题,常见的补救办法是在建立一个小的白名单,存储那些可能别误判的邮件地址。
5、 Bloom-Filter的具体实现c语言实现: stdafx.h:
[cpp] view plaincopyprint?
#pragma once #include <stdio.h> #include "stdlib.h" #include <iostream> #include <time.h> using namespace std;
[cpp] view plaincopyprint?
#include "stdafx.h" #define ARRAY_SIZE 256 /*we get the 256 chars of each line*/ #define SIZE 48000000 /* size should be 1/8 of max*/ #define MAX 384000000/*the max bit space*/ #define SETBIT(ch,n) ch[n/8]|=1<<(7-n%8) #define GETBIT(ch,n) (ch[n/8]&1<<(7-n%8))>>(7-n%8) unsigned int len(char *ch);/* functions to calculate the length of the url*/ unsigned int RSHash(char* str, unsigned int len);/* functions to calculate the hash value of the url*/ unsigned int JSHash(char* str, unsigned int len);/* functions to calculate the hash value of the url*/ unsigned int PJWHash(char* str, unsigned int len);/* functions to calculate the hash value of the url*/ unsigned int ELFHash(char* str, unsigned int len);/* functions to calculate the hash value of the url*/ unsigned int BKDRHash(char* str, unsigned int len);/* functions to calculate the hash value of the url*/ unsigned int SDBMHash(char* str, unsigned int len);/* functions to calculate the hash value of the url*/ unsigned int DJBHash(char* str, unsigned int len);/* functions to calculate the hash value of the url*/ unsigned int DEKHash(char* str, unsigned int len);/* functions to calculate the hash value of the url*/ unsigned int BPHash(char* str, unsigned int len);/* functions to calculate the hash value of the url*/ unsigned int FNVHash(char* str, unsigned int len);/* functions to calculate the hash value of the url*/ unsigned int APHash(char* str, unsigned int len);/* functions to calculate the hash value of the url*/ unsigned int HFLPHash(char* str,unsigned int len);/* functions to calculate the hash value of the url*/ unsigned int HFHash(char* str,unsigned int len);/* functions to calculate the hash value of the url*/ unsigned int StrHash( char* str,unsigned int len);/* functions to calculate the hash value of the url*/ unsigned int TianlHash(char* str,unsigned int len);/* functions to calculate the hash value of the url*/ int main() { int i,num,num2=0; /* the number to record the repeated urls and the total of it*/ unsigned int tt=0; int flag; /*it helps to check weather the url has already existed */ char buf[257]; /*it helps to print the start time of the program */ time_t tmp = time(NULL); char file1[100],file2[100]; FILE *fp1,*fp2;/*pointer to the file */ char ch[ARRAY_SIZE]; char *vector ;/* the bit space*/ vector = (char *)calloc(SIZE,sizeof(char)); printf("Please enter the file with repeated urls:\n"); scanf("%s",&file1); if( (fp1 = fopen(file1,"rb")) == NULL) { /* open the goal file*/ printf("Connot open the file %s!\n",file1); } printf("Please enter the file you want to save to:\n"); scanf("%s",&file2); if( (fp2 = fopen(file2,"w")) == NULL) { printf("Connot open the file %s\n",file2); } strftime(buf,32,"%Y-%m-%d %H:%M:%S",localtime(&tmp)); printf("%s\n",buf); /*print the system time*/ for(i=0;i<SIZE;i++) { vector[i]=0; /*set 0*/ } while(!feof(fp1)) { /* the check process*/ fgets(ch,ARRAY_SIZE,fp1); flag=0; tt++; if( GETBIT(vector, HFLPHash(ch,len(ch))%MAX) ) { flag++; } else { SETBIT(vector,HFLPHash(ch,len(ch))%MAX ); } if( GETBIT(vector, StrHash(ch,len(ch))%MAX) ) { flag++; } else { SETBIT(vector,StrHash(ch,len(ch))%MAX ); } if( GETBIT(vector, HFHash(ch,len(ch))%MAX) ) { flag++; } else { SETBIT(vector,HFHash(ch,len(ch))%MAX ); } if( GETBIT(vector, DEKHash(ch,len(ch))%MAX) ) { flag++; } else { SETBIT(vector,DEKHash(ch,len(ch))%MAX ); } if( GETBIT(vector, TianlHash(ch,len(ch))%MAX) ) { flag++; } else { SETBIT(vector,TianlHash(ch,len(ch))%MAX ); } if( GETBIT(vector, SDBMHash(ch,len(ch))%MAX) ) { flag++; } else { SETBIT(vector,SDBMHash(ch,len(ch))%MAX ); } if(flag<6) num2++; else fputs(ch,fp2); /* printf(" %d",flag); */ } /* the result*/ printf("\nThere are %d urls!\n",tt); printf("\nThere are %d not repeated urls!\n",num2); printf("There are %d repeated urls!\n",tt-num2); fclose(fp1); fclose(fp2); return 0; } /*functions may be used in the main */ unsigned int len(char *ch) { int m=0; while(ch[m]!='\0') { m++; } return m; } unsigned int RSHash(char* str, unsigned int len) { unsigned int b = 378551; unsigned int a = 63689; unsigned int hash = 0; unsigned int i = 0; for(i=0; i<len; str++, i++) { hash = hash*a + (*str); a = a*b; } return hash; } /* End Of RS Hash Function */ unsigned int JSHash(char* str, unsigned int len) { unsigned int hash = 1315423911; unsigned int i = 0; for(i=0; i<len; str++, i++) { hash ^= ((hash<<5) + (*str) + (hash>>2)); } return hash; } /* End Of JS Hash Function */ unsigned int PJWHash(char* str, unsigned int len) { const unsigned int BitsInUnsignedInt = (unsigned int)(sizeof(unsigned int) * 8); const unsigned int ThreeQuarters = (unsigned int)((BitsInUnsignedInt * 3) / 4); const unsigned int OneEighth = (unsigned int)(BitsInUnsignedInt / 8); const unsigned int HighBits = (unsigned int)(0xFFFFFFFF) << (BitsInUnsignedInt - OneEighth); unsigned int hash = 0; unsigned int test = 0; unsigned int i = 0; for(i=0;i<len; str++, i++) { hash = (hash<<OneEighth) + (*str); if((test = hash & HighBits) != 0) { hash = ((hash ^(test >> ThreeQuarters)) & (~HighBits)); } } return hash; } /* End Of P. J. Weinberger Hash Function */ unsigned int ELFHash(char* str, unsigned int len) { unsigned int hash = 0; unsigned int x = 0; unsigned int i = 0; for(i = 0; i < len; str++, i++) { hash = (hash << 4) + (*str); if((x = hash & 0xF0000000L) != 0) { hash ^= (x >> 24); } hash &= ~x; } return hash; } /* End Of ELF Hash Function */ unsigned int BKDRHash(char* str, unsigned int len) { unsigned int seed = 131; /* 31 131 1313 13131 131313 etc.. */ unsigned int hash = 0; unsigned int i = 0; for(i = 0; i < len; str++, i++) { hash = (hash * seed) + (*str); } return hash; } /* End Of BKDR Hash Function */ unsigned int SDBMHash(char* str, unsigned int len) { unsigned int hash = 0; unsigned int i = 0; for(i = 0; i < len; str++, i++) { hash = (*str) + (hash << 6) + (hash << 16) - hash; } return hash; } /* End Of SDBM Hash Function */ unsigned int DJBHash(char* str, unsigned int len) { unsigned int hash = 5381; unsigned int i = 0; for(i = 0; i < len; str++, i++) { hash = ((hash << 5) + hash) + (*str); } return hash; } /* End Of DJB Hash Function */ unsigned int DEKHash(char* str, unsigned int len) { unsigned int hash = len; unsigned int i = 0; for(i = 0; i < len; str++, i++) { hash = ((hash << 5) ^ (hash >> 27)) ^ (*str); } return hash; } /* End Of DEK Hash Function */ unsigned int BPHash(char* str, unsigned int len) { unsigned int hash = 0; unsigned int i = 0; for(i = 0; i < len; str++, i++) { hash = hash << 7 ^ (*str); } return hash; } /* End Of BP Hash Function */ unsigned int FNVHash(char* str, unsigned int len) { const unsigned int fnv_prime = 0x811C9DC5; unsigned int hash = 0; unsigned int i = 0; for(i = 0; i < len; str++, i++) { hash *= fnv_prime; hash ^= (*str); } return hash; } /* End Of FNV Hash Function */ unsigned int APHash(char* str, unsigned int len) { unsigned int hash = 0xAAAAAAAA; unsigned int i = 0; for(i = 0; i < len; str++, i++) { hash ^= ((i & 1) == 0) ? ( (hash << 7) ^ (*str) * (hash >> 3)) : (~((hash << 11) + (*str) ^ (hash >> 5))); } return hash; } /* End Of AP Hash Function */ unsigned int HFLPHash(char *str,unsigned int len) { unsigned int n=0; int i; char* b=(char *)&n; for(i=0;i<strlen(str);++i) { b[i%4]^=str[i]; } return n%len; } /* End Of HFLP Hash Function*/ unsigned int HFHash(char* str,unsigned int len) { int result=0; char* ptr=str; int c; int i=0; for (i=1;c=*ptr++;i++) result += c*3*i; if (result<0) result = -result; return result%len; } /*End Of HKHash Function */ unsigned int StrHash( char *str,unsigned int len) { register unsigned int h; register unsigned char *p; for(h=0,p=(unsigned char *)str;*p;p++) { h=31*h+*p; } return h; } /*End Of StrHash Function*/ unsigned int TianlHash(char *str,unsigned int len) { unsigned long urlHashValue=0; int ilength=strlen(str); int i; unsigned char ucChar; if(!ilength) { return 0; } if(ilength<=256) { urlHashValue=16777216*(ilength-1); } else { urlHashValue = 42781900080; } if(ilength<=96) { for(i=1;i<=ilength;i++) { ucChar=str[i-1]; if(ucChar<='Z'&&ucChar>='A') { ucChar=ucChar+32; } urlHashValue+=(3*i*ucChar*ucChar+5*i*ucChar+7*i+11*ucChar)%1677216; } } else { for(i=1;i<=96;i++) { ucChar=str[i+ilength-96-1]; if(ucChar<='Z'&&ucChar>='A') { ucChar=ucChar+32; } urlHashValue+=(3*i*ucChar*ucChar+5*i*ucChar+7*i+11*ucChar)%1677216; } } return urlHashValue; } /*End Of Tianl Hash Function*/
网上找到的php简单实现:
[cpp] view plaincopyprint?
<?php /** * Implements a Bloom Filter */ class BloomFilter { /** * Size of the bit array * * @var int */ protected $m; /** * Number of hash functions * * @var int */ protected $k; /** * Number of elements in the filter * * @var int */ protected $n; /** * The bitset holding the filter information * * @var array */ protected $bitset; /** * 计算最优的hash函数个数:当hash函数个数k=(ln2)*(m/n)时错误率最小 * * @param int $m bit数组的宽度(bit数) * @param int $n 加入布隆过滤器的key的数量 * @return int */ public static function getHashCount($m, $n) { return ceil(($m / $n) * log(2)); } /** * Construct an instance of the Bloom filter * * @param int $m bit数组的宽度(bit数) Size of the bit array * @param int $k hash函数的个数 Number of different hash functions to use */ public function __construct($m, $k) { $this->m = $m; $this->k = $k; $this->n = 0; /* Initialize the bit set */ $this->bitset = array_fill(0, $this->m - 1, false); } /** * False Positive的比率:f = (1 – e-kn/m)k * Returns the probability for a false positive to occur, given the current number of items in the filter * * @return double */ public function getFalsePositiveProbability() { $exp = (-1 * $this->k * $this->n) / $this->m; return pow(1 - exp($exp), $this->k); } /** * Adds a new item to the filter * * @param mixed Either a string holding a single item or an array of * string holding multiple items. In the latter case, all * items are added one by one internally. */ public function add($key) { if (is_array($key)) { foreach ($key as $k) { $this->add($k); } return; } $this->n++; foreach ($this->getSlots($key) as $slot) { $this->bitset[$slot] = true; } } /** * Queries the Bloom filter for an element * * If this method return FALSE, it is 100% certain that the element has * not been added to the filter before. In contrast, if TRUE is returned, * the element *may* have been added to the filter previously. However with * a probability indicated by getFalsePositiveProbability() the element has * not been added to the filter with contains() still returning TRUE. * * @param mixed Either a string holding a single item or an array of * strings holding multiple items. In the latter case the * method returns TRUE if the filter contains all items. * @return boolean */ public function contains($key) { if (is_array($key)) { foreach ($key as $k) { if ($this->contains($k) == false) { return false; } } return true; } foreach ($this->getSlots($key) as $slot) { if ($this->bitset[$slot] == false) { return false; } } return true; } /** * Hashes the argument to a number of positions in the bit set and returns the positions * * @param string Item * @return array Positions */ protected function getSlots($key) { $slots = array(); $hash = self::getHashCode($key); mt_srand($hash); for ($i = 0; $i < $this->k; $i++) { $slots[] = mt_rand(0, $this->m - 1); } return $slots; } /** * 使用CRC32产生一个32bit(位)的校验值。 * 由于CRC32产生校验值时源数据块的每一bit(位)都会被计算,所以数据块中即使只有一位发生了变化,也会得到不同的CRC32值。 * Generates a numeric hash for the given string * * Right now the CRC-32 algorithm is used. Alternatively one could e.g. * use Adler digests or mimick the behaviour of Java's hashCode() method. * * @param string Input for which the hash should be created * @return int Numeric hash */ protected static function getHashCode($string) { return crc32($string); } } $items = array("first item", "second item", "third item"); /* Add all items with one call to add() and make sure contains() finds * them all. */ $filter = new BloomFilter(100, BloomFilter::getHashCount(100, 3)); $filter->add($items); //var_dump($filter); exit; $items = array("firsttem", "seconditem", "thirditem"); foreach ($items as $item) { var_dump(($filter->contains($item))); } /* Add all items with multiple calls to add() and make sure contains() * finds them all. */ $filter = new BloomFilter(100, BloomFilter::getHashCount(100, 3)); foreach ($items as $item) { $filter->add($item); } $items = array("fir sttem", "secondit em", "thir ditem"); foreach ($items as $item) { var_dump(($filter->contains($item))); }
问题实例】 给你A,B两个文件,各存放50亿条URL,每条URL占用64字节,内存限制是4G,让你找出A,B文件共同的URL。如果是三个乃至n个文件呢? 根据这个问题我们来计算下内存的占用,4G=2^32大概是40亿*8大概是340亿bit,n=50亿,如果按出错率0.01算需要的大概是650亿个bit。 现在可用的是340亿,相差并不多,这样可能会使出错率上升些。另外如果这些urlip是一一对应的,就可以转换成ip,则大大简单了。 (责任编辑:IT) |