Hadoop的Map侧join

写了关于Hadoop下载地址的Map侧join 
和Reduce的join，今天我们就来在看另外一种比较中立的Join。 

SemiJoin，一般称为半链接，其原理是在Map侧过滤掉了一些不需要join的数据，从而大大减少了reduce的shffule时间，因为我们知道，如果仅仅使用Reduce侧连接，那么如果一份数据中，存在大量的无效数据，而这些数据，在join中，并不需要，但是因为没有做过预处理，所以这些数据，直到真正的执行reduce函数时，才被定义为无效数据，而这时候，前面已经执行过shuffle和merge和sort，所以这部分无效的数据，就浪费了大量的网络IO和磁盘IO，所以在整体来讲，这是一种降低性能的表现，如果存在的无效数据越多，那么这种趋势，就越明显。 

之所以会出现半连接，这其实也是reduce侧连接的一个变种，只不过我们在Map侧，过滤掉了一些无效的数据，所以减少了reduce过程的shuffle时间，所以能获取一个性能的提升。 

具体的原理也是利用DistributedCache将小表的的分发到各个节点上，在Map过程的setup函数里，读取缓存里面的文件，只将小表的链接键存储在hashset里，在map函数执行时，对每一条数据，进行判断，如果这条数据的链接键为空或者在hashset里面不存在，那么则认为这条数据，是无效的数据，所以这条数据，并不会被partition分区后写入磁盘，参与reduce阶段的shuffle和sort下载地址  ，所以在一定程序上，提升了join性能。需要注意的是如果 
小表的key依然非常巨大，可能会导致我们的程序出现OOM的情况，那么这时候我们就需要考虑其他的链接方式了。 

测试数据如下： 
模拟小表数据： 
1,三劫散仙,13575468248 
2,凤舞九天,18965235874 
3,忙忙碌碌,15986854789 
4,少林寺方丈,15698745862 


模拟大表数据： 
3,A,99,2013-03-05 
1,B,89,2013-02-05 
2,C,69,2013-03-09 
3,D,56,2013-06-07 
5,E,100,2013-09-09 
6,H,200,2014-01-10 

代码如下： 

Java代码【下载地址】   复制代码 收藏代码
  1. package com.semijoin;  
  2.   
  3. import java.io.BufferedReader;  
  4. import java.io.DataInput;  
  5. import java.io.DataOutput;  
  6. import java.io.FileReader;  
  7. import java.io.IOException;  
  8. import java.net.URI;  
  9. import java.util.ArrayList;  
  10. import java.util.HashSet;  
  11. import java.util.List;  
  12.   
  13. import org.apache.hadoop.conf.Configuration;  
  14. import org.apache.hadoop.filecache.DistributedCache;  
  15. import org.apache.hadoop.fs.FileSystem;  
  16. import org.apache.hadoop.fs.Path;  
  17. import org.apache.hadoop.io.LongWritable;  
  18. import org.apache.hadoop.io.Text;  
  19. import org.apache.hadoop.io.WritableComparable;  
  20.   
  21. import org.apache.hadoop.mapred.JobConf;  
  22. import org.apache.hadoop.mapreduce.Job;  
  23. import org.apache.hadoop.mapreduce.Mapper;  
  24. import org.apache.hadoop.mapreduce.Reducer;  
  25. import org.apache.hadoop.mapreduce.lib.input.FileInputFormat;  
  26. import org.apache.hadoop.mapreduce.lib.input.FileSplit;  
  27. import org.apache.hadoop.mapreduce.lib.input.TextInputFormat;  
  28. import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;  
  29. import org.apache.hadoop.mapreduce.lib.output.TextOutputFormat;  
  30.   
  31. /*** 
  32.  *  
  33.  * Hadoop1.2的版本 
  34.  *  
  35.  * hadoop的半链接 
  36.  *  
  37.  * SemiJoin实现 
  38.  *  
  39.  * @author qindongliang 
  40.  *  
  41.  *    大数据交流群：376932160 
  42.  *  搜索技术交流群：324714439 
  43.  *  
  44.  *  
  45.  *  
  46.  * **/  
  47. public class Semjoin {  
  48.       
  49.       
  50.       
  51.     /** 
  52.      *  
  53.      *  
  54.      * 自定义一个输出实体 
  55.      *  
  56.      * **/  
  57.     private static class CombineEntity implements WritableComparable<CombineEntity>{  
  58.   
  59.           
  60.         private Text joinKey;//连接key  
  61.         private Text flag;//文件来源标志  
  62.         private Text secondPart;//除了键外的其他部分的数据  
  63.           
  64.           
  65.         public CombineEntity() {  
  66.             // TODO Auto-generated constructor stub  
  67.             this.joinKey=new Text();  
  68.             this.flag=new Text();  
  69.             this.secondPart=new Text();  
  70.         }  
  71.           
  72.         public Text getJoinKey() {  
  73.             return joinKey;  
  74.         }  
  75.   
  76.         public void setJoinKey(Text joinKey) {  
  77.             this.joinKey = joinKey;  
  78.         }  
  79.   
  80.         public Text getFlag() {  
  81.             return flag;  
  82.         }  
  83.   
  84.         public void setFlag(Text flag) {  
  85.             this.flag = flag;  
  86.         }  
  87.   
  88.         public Text getSecondPart() {  
  89.             return secondPart;  
  90.         }  
  91.   
  92.         public void setSecondPart(Text secondPart) {  
  93.             this.secondPart = secondPart;  
  94.         }  
  95.   
  96.         @Override  
  97.         public void readFields(DataInput in) throws IOException {  
  98.             this.joinKey.readFields(in);  
  99.             this.flag.readFields(in);  
  100.             this.secondPart.readFields(in);  
  101.               
  102.         }  
  103.   
  104.         @Override  
  105.         public void write(DataOutput out) throws IOException {  
  106.             this.joinKey.write(out);  
  107.             this.flag.write(out);  
  108.             this.secondPart.write(out);  
  109.               
  110.         }  
  111.   
  112.         @Override  
  113.         public int compareTo(CombineEntity o) {  
  114.             // TODO Auto-generated method stub  
  115.             return this.joinKey.compareTo(o.joinKey);  
  116.         }  
  117.           
  118.           
  119.           
  120.     }  
  121.       
  122.       
  123.       
  124.       
  125.     private static class JMapper extends Mapper<LongWritable, Text, Text, CombineEntity>{  
  126.           
  127.         private CombineEntity combine=new CombineEntity();  
  128.         private Text flag=new Text();  
  129.         private  Text joinKey=new Text();  
  130.         private Text secondPart=new Text();  
  131.         /** 
  132.          * 存储小表的key 
  133.          *  
  134.          *  
  135.          * */  
  136.         private HashSet<String> joinKeySet=new HashSet<String>();  
  137.           
  138.           
  139.         @Override  
  140.         protected void setup(Context context)throws IOException, InterruptedException {  
  141.            
  142.             //读取文件流  
  143.             BufferedReader br=null;  
  144.             String temp;  
  145.             // 获取DistributedCached里面 的共享文件  
  146.             Path path[]=DistributedCache.getLocalCacheFiles(context.getConfiguration());  
  147.               
  148.             for(Path p:path){  
  149.                   
  150.                 if(p.getName().endsWith("a.txt")){  
  151.                     br=new BufferedReader(new FileReader(p.toString()));  
  152.                     //List<String> list=Files.readAllLines(Paths.get(p.getName()), Charset.forName("UTF-8"));  
  153.                       
  154.                     while((temp=br.readLine())!=null){  
  155.                         String ss[]=temp.split(",");  
  156.                         //map.put(ss[0], ss[1]+" "+ss[2]);//放入hash表中  
  157.                         joinKeySet.add(ss[0]);//加入小表的key  
  158.                     }  
  159.                 }  
  160.             }  
  161.               
  162.               
  163.         }  
  164.           
  165.           
  166.           
  167.         @Override  
  168.         protected void map(LongWritable key, Text value,Context context)  
  169.                 throws IOException, InterruptedException {  
  170.               
  171.           
  172.                //获得文件输入路径  
  173.             String pathName = ((FileSplit) context.getInputSplit()).getPath().toString();  
  174.           
  175.             if(pathName.endsWith("a.txt")){  
  176.                   
  177.                 String  valueItems[]=value.toString().split(",");  
  178.                   
  179.                   
  180.                 /** 
  181.                  * 在这里过滤必须要的连接字符 
  182.                  *  
  183.                  * */  
  184.                 if(joinKeySet.contains(valueItems[0])){  
  185.                     //设置标志位  
  186.                     flag.set("0");     
  187.                     //设置链接键  
  188.                     joinKey.set(valueItems[0]);            
  189.                     //设置第二部分  
  190.                     secondPart.set(valueItems[1]+" "+valueItems[2]);  
  191.                       
  192.                     //封装实体  
  193.                     combine.setFlag(flag);//标志位  
  194.                     combine.setJoinKey(joinKey);//链接键  
  195.                     combine.setSecondPart(secondPart);//其他部分  
  196.                       
  197.                      //写出  
  198.                     context.write(combine.getJoinKey(), combine);     
  199.                 }else{  
  200.                     System.out.println("a.txt里");  
  201.                     System.out.println("在小表中无此记录，执行过滤掉！");  
  202.                     for(String v:valueItems){  
  203.                         System.out.print(v+"   ");  
  204.                     }  
  205.                       
  206.                     return ;  
  207.                       
  208.                 }  
  209.                   
  210.                   
  211.                   
  212.             }else if(pathName.endsWith("b.txt")){  
  213.                 String  valueItems[]=value.toString().split(",");  
  214.                   
  215.                 /** 
  216.                  *  
  217.                  * 判断是否在集合中 
  218.                  *  
  219.                  * */  
  220.                 if(joinKeySet.contains(valueItems[0])){  
  221.                       
  222.   
  223.                     //设置标志位  
  224.                     flag.set("1");     
  225.                       
  226.                     //设置链接键  
  227.                     joinKey.set(valueItems[0]);  
  228.             
  229.                     //设置第二部分注意不同的文件的列数不一样  
  230.                     secondPart.set(valueItems[1]+" "+valueItems[2]+" "+valueItems[3]);  
  231.                       
  232.                     //封装实体  
  233.                     combine.setFlag(flag);//标志位  
  234.                     combine.setJoinKey(joinKey);//链接键  
  235.                     combine.setSecondPart(secondPart);//其他部分  
  236.                       
  237.                      //写出  
  238.                     context.write(combine.getJoinKey(), combine);  
  239.                       
  240.                       
  241.                 }else{                    
  242.                     //执行过滤 ......  
  243.                     System.out.println("b.txt里");  
  244.                     System.out.println("在小表中无此记录，执行过滤掉！");  
  245.                     for(String v:valueItems){  
  246.                         System.out.print(v+"   ");  
  247.                     }  
  248.                       
  249.                     return ;  
  250.                       
  251.                       
  252.                 }  
  253.                   
  254.               
  255.                   
  256.                   
  257.             }  
  258.               
  259.               
  260.               
  261.               
  262.                
  263.            
  264.               
  265.         }  
  266.           
  267.     }  
  268.       
  269.       
  270.     private static class JReduce extends Reducer<Text, CombineEntity, Text, Text>{  
  271.           
  272.           
  273.         //存储一个分组中左表信息  
  274.         private List<Text> leftTable=new ArrayList<Text>();  
  275.         //存储一个分组中右表信息  
  276.         private List<Text> rightTable=new ArrayList<Text>();  
  277.           
  278.         private Text secondPart=null;  
  279.           
  280.         private Text output=new Text();  
  281.           
  282.           
  283.            
  284.         //一个分组调用一次  
  285.         @Override  
  286.         protected void reduce(Text key, Iterable<CombineEntity> values,Context context)  
  287.                 throws IOException, InterruptedException {  
  288.              leftTable.clear();//清空分组数据  
  289.              rightTable.clear();//清空分组数据  
  290.                
  291.                
  292.              /** 
  293.               * 将不同文件的数据，分别放在不同的集合 
  294.               * 中，注意数据量过大时，会出现 
  295.               * OOM的异常 
  296.               *  
  297.               * **/  
  298.                
  299.              for(CombineEntity ce:values){  
  300.                    
  301.                  this.secondPart=new Text(ce.getSecondPart().toString());  
  302.                    
  303.                    
  304.                  //左表  
  305.                    
  306.                  if(ce.getFlag().toString().trim().equals("0")){  
  307.                      leftTable.add(secondPart);  
  308.                        
  309.                  }else if(ce.getFlag().toString().trim().equals("1")){  
  310.                        
  311.                      rightTable.add(secondPart);  
  312.                        
  313.                  }  
  314.                    
  315.                    
  316.                    
  317.                    
  318.              }  
  319.                
  320.              //=====================  
  321.              for(Text left:leftTable){  
  322.                    
  323.                  for(Text right:rightTable){  
  324.                        
  325.                      output.set(left+" "+right);//连接左右数据  
  326.                      context.write(key, output);//输出  
  327.                  }  
  328.                    
  329.              }  
  330.                
  331.                
  332.                
  333.               
  334.         }  
  335.           
  336.     }  
  337.       
  338.       
  339.       
  340.       
  341.       
  342.       
  343.       
  344.       
  345.     public static void main(String[] args)throws Exception {  
  346.           
  347.            
  348.       
  349.       
  350.          //Job job=new Job(conf,"myjoin");  
  351.          JobConf conf=new JobConf(Semjoin.class);   
  352.            conf.set("mapred.job.tracker","192.168.75.130:9001");  
  353.             conf.setJar("tt.jar");  
  354.             
  355.             
  356.             //小表共享  
  357.             String bpath="hdfs://192.168.75.130:9000/root/dist/a.txt";  
  358.             //添加到共享cache里  
  359.         DistributedCache.addCacheFile(new URI(bpath), conf);  
  360.           
  361.           Job job=new Job(conf, "aaaaa");  
  362.          job.setJarByClass(Semjoin.class);  
  363.          System.out.println("模式：  "+conf.get("mapred.job.tracker"));;  
  364.            
  365.            
  366.          //设置Map和Reduce自定义类  
  367.          job.setMapperClass(JMapper.class);  
  368.          job.setReducerClass(JReduce.class);  
  369.            
  370.          //设置Map端输出  
  371.          job.setMapOutputKeyClass(Text.class);  
  372.          job.setMapOutputValueClass(CombineEntity.class);  
  373.            
  374.          //设置Reduce端的输出  
  375.          job.setOutputKeyClass(Text.class);  
  376.          job.setOutputValueClass(Text.class);  
  377.            
  378.       
  379.          job.setInputFormatClass(TextInputFormat.class);  
  380.          job.setOutputFormatClass(TextOutputFormat.class);  
  381.            
  382.        
  383.          FileSystem fs=FileSystem.get(conf);  
  384.            
  385.          Path op=new Path("hdfs://192.168.75.130:9000/root/outputjoindbnew4");  
  386.            
  387.          if(fs.exists(op)){  
  388.              fs.delete(op, true);  
  389.              System.out.println("存在此输出路径，已删除！！！");  
  390.          }  
  391.            
  392.            
  393.       FileInputFormat.setInputPaths(job, new Path("hdfs://192.168.75.130:9000/root/inputjoindb"));  
  394.       FileOutputFormat.setOutputPath(job, op);  
  395.          
  396.       System.exit(job.waitForCompletion(true)?0:1);  
  397.         
  398.         
  399.            
  400.            
  401.            
  402.            
  403.            
  404.            
  405.            
  406.                   
  407.                    
  408.           
  409.           
  410.           
  411.     }  
  412.       
  413.       
  414.       
  415.   
  416. }  
Java代码  收藏代码
  1. package com.semijoin;  
  2.   
  3. import java.io.BufferedReader;  
  4. import java.io.DataInput;  
  5. import java.io.DataOutput;  
  6. import java.io.FileReader;  
  7. import java.io.IOException;  
  8. import java.net.URI;  
  9. import java.util.ArrayList;  
  10. import java.util.HashSet;  
  11. import java.util.List;  
  12.   
  13. import org.apache.hadoop.conf.Configuration;  
  14. import org.apache.hadoop.filecache.DistributedCache;  
  15. import org.apache.hadoop.fs.FileSystem;  
  16. import org.apache.hadoop.fs.Path;  
  17. import org.apache.hadoop.io.LongWritable;  
  18. import org.apache.hadoop.io.Text;  
  19. import org.apache.hadoop.io.WritableComparable;  
  20.   
  21. import org.apache.hadoop.mapred.JobConf;  
  22. import org.apache.hadoop.mapreduce.Job;  
  23. import org.apache.hadoop.mapreduce.Mapper;  
  24. import org.apache.hadoop.mapreduce.Reducer;  
  25. import org.apache.hadoop.mapreduce.lib.input.FileInputFormat;  
  26. import org.apache.hadoop.mapreduce.lib.input.FileSplit;  
  27. import org.apache.hadoop.mapreduce.lib.input.TextInputFormat;  
  28. import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;  
  29. import org.apache.hadoop.mapreduce.lib.output.TextOutputFormat;  
  30.   
  31. /*** 
  32.  *  
  33.  * Hadoop1.2的版本 
  34.  *  
  35.  * hadoop的半链接 
  36.  *  
  37.  * SemiJoin实现 
  38.  *  
  39.  * @author qindongliang 
  40.  *  
  41.  *    大数据交流群：376932160 
  42.  *  搜索技术交流群：324714439 
  43.  *  
  44.  *  
  45.  *  
  46.  * **/  
  47. public class Semjoin {  
  48.       
  49.       
  50.       
  51.     /** 
  52.      *  
  53.      *  
  54.      * 自定义一个输出实体 
  55.      *  
  56.      * **/  
  57.     private static class CombineEntity implements WritableComparable<CombineEntity>{  
  58.   
  59.           
  60.         private Text joinKey;//连接key  
  61.         private Text flag;//文件来源标志  
  62.         private Text secondPart;//除了键外的其他部分的数据  
  63.           
  64.           
  65.         public CombineEntity() {  
  66.             // TODO Auto-generated constructor stub  
  67.             this.joinKey=new Text();  
  68.             this.flag=new Text();  
  69.             this.secondPart=new Text();  
  70.         }  
  71.           
  72.         public Text getJoinKey() {  
  73.             return joinKey;  
  74.         }  
  75.   
  76.         public void setJoinKey(Text joinKey) {  
  77.             this.joinKey = joinKey;  
  78.         }  
  79.   
  80.         public Text getFlag() {  
  81.             return flag;  
  82.         }  
  83.   
  84.         public void setFlag(Text flag) {  
  85.             this.flag = flag;  
  86.         }  
  87.   
  88.         public Text getSecondPart() {  
  89.             return secondPart;  
  90.         }  
  91.   
  92.         public void setSecondPart(Text secondPart) {  
  93.             this.secondPart = secondPart;  
  94.         }  
  95.   
  96.         @Override  
  97.         public void readFields(DataInput in) throws IOException {  
  98.             this.joinKey.readFields(in);  
  99.             this.flag.readFields(in);  
  100.             this.secondPart.readFields(in);  
  101.               
  102.         }  
  103.   
  104.         @Override  
  105.         public void write(DataOutput out) throws IOException {  
  106.             this.joinKey.write(out);  
  107.             this.flag.write(out);  
  108.             this.secondPart.write(out);  
  109.               
  110.         }  
  111.   
  112.         @Override  
  113.         public int compareTo(CombineEntity o) {  
  114.             // TODO Auto-generated method stub  
  115.             return this.joinKey.compareTo(o.joinKey);  
  116.         }  
  117.           
  118.           
  119.           
  120.     }  
  121.       
  122.       
  123.       
  124.       
  125.     private static class JMapper extends Mapper<LongWritable, Text, Text, CombineEntity>{  
  126.           
  127.         private CombineEntity combine=new CombineEntity();  
  128.         private Text flag=new Text();  
  129.         private  Text joinKey=new Text();  
  130.         private Text secondPart=new Text();  
  131.         /** 
  132.          * 存储小表的key 
  133.          *  
  134.          *  
  135.          * */  
  136.         private HashSet<String> joinKeySet=new HashSet<String>();  
  137.           
  138.           
  139.         @Override  
  140.         protected void setup(Context context)throws IOException, InterruptedException {  
  141.            
  142.             //读取文件流  
  143.             BufferedReader br=null;  
  144.             String temp;  
  145.             // 获取DistributedCached里面 的共享文件  
  146.             Path path[]=DistributedCache.getLocalCacheFiles(context.getConfiguration());  
  147.               
  148.             for(Path p:path){  
  149.                   
  150.                 if(p.getName().endsWith("a.txt")){  
  151.                     br=new BufferedReader(new FileReader(p.toString()));  
  152.                     //List<String> list=Files.readAllLines(Paths.get(p.getName()), Charset.forName("UTF-8"));  
  153.                       
  154.                     while((temp=br.readLine())!=null){  
  155.                         String ss[]=temp.split(",");  
  156.                         //map.put(ss[0], ss[1]+" "+ss[2]);//放入hash表中  
  157.                         joinKeySet.add(ss[0]);//加入小表的key  
  158.                     }  
  159.                 }  
  160.             }  
  161.               
  162.               
  163.         }  
  164.           
  165.           
  166.           
  167.         @Override  
  168.         protected void map(LongWritable key, Text value,Context context)  
  169.                 throws IOException, InterruptedException {  
  170.               
  171.           
  172.                //获得文件输入路径  
  173.             String pathName = ((FileSplit) context.getInputSplit()).getPath().toString();  
  174.           
  175.             if(pathName.endsWith("a.txt")){  
  176.                   
  177.                 String  valueItems[]=value.toString().split(",");  
  178.                   
  179.                   
  180.                 /** 
  181.                  * 在这里过滤必须要的连接字符 
  182.                  *  
  183.                  * */  
  184.                 if(joinKeySet.contains(valueItems[0])){  
  185.                     //设置标志位  
  186.                     flag.set("0");     
  187.                     //设置链接键  
  188.                     joinKey.set(valueItems[0]);            
  189.                     //设置第二部分  
  190.                     secondPart.set(valueItems[1]+" "+valueItems[2]);  
  191.                       
  192.                     //封装实体  
  193.                     combine.setFlag(flag);//标志位  
  194.                     combine.setJoinKey(joinKey);//链接键  
  195.                     combine.setSecondPart(secondPart);//其他部分  
  196.                       
  197.                      //写出  
  198.                     context.write(combine.getJoinKey(), combine);     
  199.                 }else{  
  200.                     System.out.println("a.txt里");  
  201.                     System.out.println("在小表中无此记录，执行过滤掉！");  
  202.                     for(String v:valueItems){  
  203.                         System.out.print(v+"   ");  
  204.                     }  
  205.                       
  206.                     return ;  
  207.                       
  208.                 }  
  209.                   
  210.                   
  211.                   
  212.             }else if(pathName.endsWith("b.txt")){  
  213.                 String  valueItems[]=value.toString().split(",");  
  214.                   
  215.                 /** 
  216.                  *  
  217.                  * 判断是否在集合中 
  218.                  *  
  219.                  * */  
  220.                 if(joinKeySet.contains(valueItems[0])){  
  221.                       
  222.   
  223.                     //设置标志位  
  224.                     flag.set("1");     
  225.                       
  226.                     //设置链接键  
  227.                     joinKey.set(valueItems[0]);  
  228.             
  229.                     //设置第二部分注意不同的文件的列数不一样  
  230.                     secondPart.set(valueItems[1]+" "+valueItems[2]+" "+valueItems[3]);  
  231.                       
  232.                     //封装实体  
  233.                     combine.setFlag(flag);//标志位  
  234.                     combine.setJoinKey(joinKey);//链接键  
  235.                     combine.setSecondPart(secondPart);//其他部分  
  236.                       
  237.                      //写出  
  238.                     context.write(combine.getJoinKey(), combine);  
  239.                       
  240.                       
  241.                 }else{                    
  242.                     //执行过滤 ......  
  243.                     System.out.println("b.txt里");  
  244.                     System.out.println("在小表中无此记录，执行过滤掉！");  
  245.                     for(String v:valueItems){  
  246.                         System.out.print(v+"   ");  
  247.                     }  
  248.                       
  249.                     return ;  
  250.                       
  251.                       
  252.                 }  
  253.                   
  254.               
  255.                   
  256.                   
  257.             }  
  258.               
  259.               
  260.               
  261.               
  262.                
  263.            
  264.               
  265.         }  
  266.           
  267.     }  
  268.       
  269.       
  270.     private static class JReduce extends Reducer<Text, CombineEntity, Text, Text>{  
  271.           
  272.           
  273.         //存储一个分组中左表信息  
  274.         private List<Text> leftTable=new ArrayList<Text>();  
  275.         //存储一个分组中右表信息  
  276.         private List<Text> rightTable=new ArrayList<Text>();  
  277.           
  278.         private Text secondPart=null;  
  279.           
  280.         private Text output=new Text();  
  281.           
  282.           
  283.            
  284.         //一个分组调用一次  
  285.         @Override  
  286.         protected void reduce(Text key, Iterable<CombineEntity> values,Context context)  
  287.                 throws IOException, InterruptedException {  
  288.              leftTable.clear();//清空分组数据  
  289.              rightTable.clear();//清空分组数据  
  290.                
  291.                
  292.              /** 
  293.               * 将不同文件的数据，分别放在不同的集合 
  294.               * 中，注意数据量过大时，会出现 
  295.               * OOM的异常 
  296.               *  
  297.               * **/  
  298.                
  299.              for(CombineEntity ce:values){  
  300.                    
  301.                  this.secondPart=new Text(ce.getSecondPart().toString());  
  302.                    
  303.                    
  304.                  //左表  
  305.                    
  306.                  if(ce.getFlag().toString().trim().equals("0")){  
  307.                      leftTable.add(secondPart);  
  308.                        
  309.                  }else if(ce.getFlag().toString().trim().equals("1")){  
  310.                        
  311.                      rightTable.add(secondPart);  
  312.                        
  313.                  }  
  314.                    
  315.                    
  316.                    
  317.                    
  318.              }  
  319.                
  320.              //=====================  
  321.              for(Text left:leftTable){  
  322.                    
  323.                  for(Text right:rightTable){  
  324.                        
  325.                      output.set(left+" "+right);//连接左右数据  
  326.                      context.write(key, output);//输出  
  327.                  }  
  328.                    
  329.              }  
  330.                
  331.                
  332.                
  333.               
  334.         }  
  335.           
  336.     }  
  337.       
  338.       
  339.       
  340.       
  341.       
  342.       
  343.       
  344.       
  345.     public static void main(String[] args)throws Exception {  
  346.           
  347.            
  348.       
  349.       
  350.          //Job job=new Job(conf,"myjoin");  
  351.          JobConf conf=new JobConf(Semjoin.class);   
  352.            conf.set("mapred.job.tracker","192.168.75.130:9001");  
  353.             conf.setJar("tt.jar");  
  354.             
  355.             
  356.             //小表共享  
  357.             String bpath="hdfs://192.168.75.130:9000/root/dist/a.txt";  
  358.             //添加到共享cache里  
  359.         DistributedCache.addCacheFile(new URI(bpath), conf);  
  360.           
  361.           Job job=new Job(conf, "aaaaa");  
  362.          job.setJarByClass(Semjoin.class);  
  363.          System.out.println("模式：  "+conf.get("mapred.job.tracker"));;  
  364.            
  365.            
  366.          //设置Map和Reduce自定义类  
  367.          job.setMapperClass(JMapper.class);  
  368.          job.setReducerClass(JReduce.class);  
  369.            
  370.          //设置Map端输出  
  371.          job.setMapOutputKeyClass(Text.class);  
  372.          job.setMapOutputValueClass(CombineEntity.class);  
  373.            
  374.          //设置Reduce端的输出  
  375.          job.setOutputKeyClass(Text.class);  
  376.          job.setOutputValueClass(Text.class);  
  377.            
  378.       
  379.          job.setInputFormatClass(TextInputFormat.class);  
  380.          job.setOutputFormatClass(TextOutputFormat.class);  
  381.            
  382.        
  383.          FileSystem fs=FileSystem.get(conf);  
  384.            
  385.          Path op=new Path("hdfs://192.168.75.130:9000/root/outputjoindbnew4");  
  386.            
  387.          if(fs.exists(op)){  
  388.              fs.delete(op, true);  
  389.              System.out.println("存在此输出路径，已删除！！！");  
  390.          }  
  391.            
  392.            
  393.       FileInputFormat.setInputPaths(job, new Path("hdfs://192.168.75.130:9000/root/inputjoindb"));  
  394.       FileOutputFormat.setOutputPath(job, op);  
  395.          
  396.       System.exit(job.waitForCompletion(true)?0:1);  
  397.         
  398.         
  399.            
  400.            
  401.            
  402.            
  403.            
  404.            
  405.            
  406.                   
  407.                    
  408.           
  409.           
  410.           
  411.     }  
  412.       
  413.       
  414.       
  415.   
  416. }  


运行日志如下：

Java代码 复制代码 收藏代码
  1. 模式：  192.168.75.130:9001  
  2. 存在此输出路径，已删除！！！  
  3. WARN - JobClient.copyAndConfigureFiles(746) | Use GenericOptionsParser for parsing the arguments. Applications should implement Tool for the same.  
  4. INFO - FileInputFormat.listStatus(237) | Total input paths to process : 2  
  5. WARN - NativeCodeLoader.<clinit>(52) | Unable to load native-hadoop library for your platform... using builtin-java classes where applicable  
  6. WARN - LoadSnappy.<clinit>(46) | Snappy native library not loaded  
  7. INFO - JobClient.monitorAndPrintJob(1380) | Running job: job_201404260312_0002  
  8. INFO - JobClient.monitorAndPrintJob(1393) |  map 0% reduce 0%  
  9. INFO - JobClient.monitorAndPrintJob(1393) |  map 50% reduce 0%  
  10. INFO - JobClient.monitorAndPrintJob(1393) |  map 100% reduce 0%  
  11. INFO - JobClient.monitorAndPrintJob(1393) |  map 100% reduce 33%  
  12. INFO - JobClient.monitorAndPrintJob(1393) |  map 100% reduce 100%  
  13. INFO - JobClient.monitorAndPrintJob(1448) | Job complete: job_201404260312_0002  
  14. INFO - Counters.log(585) | Counters: 29  
  15. INFO - Counters.log(587) |   Job Counters   
  16. INFO - Counters.log(589) |     Launched reduce tasks=1  
  17. INFO - Counters.log(589) |     SLOTS_MILLIS_MAPS=12445  
  18. INFO - Counters.log(589) |     Total time spent by all reduces waiting after reserving slots (ms)=0  
  19. INFO - Counters.log(589) |     Total time spent by all maps waiting after reserving slots (ms)=0  
  20. INFO - Counters.log(589) |     Launched map tasks=2  
  21. INFO - Counters.log(589) |     Data-local map tasks=2  
  22. INFO - Counters.log(589) |     SLOTS_MILLIS_REDUCES=9801  
  23. INFO - Counters.log(587) |   File Output Format Counters   
  24. INFO - Counters.log(589) |     Bytes Written=172  
  25. INFO - Counters.log(587) |   FileSystemCounters  
  26. INFO - Counters.log(589) |     FILE_BYTES_READ=237  
  27. INFO - Counters.log(589) |     HDFS_BYTES_READ=455  
  28. INFO - Counters.log(589) |     FILE_BYTES_WRITTEN=169503  
  29. INFO - Counters.log(589) |     HDFS_BYTES_WRITTEN=172  
  30. INFO - Counters.log(587) |   File Input Format Counters   
  31. INFO - Counters.log(589) |     Bytes Read=227  
  32. INFO - Counters.log(587) |   Map-Reduce Framework  
  33. INFO - Counters.log(589) |     Map output materialized bytes=243  
  34. INFO - Counters.log(589) |     Map input records=10  
  35. INFO - Counters.log(589) |     Reduce shuffle bytes=243  
  36. INFO - Counters.log(589) |     Spilled Records=16  
  37. INFO - Counters.log(589) |     Map output bytes=215  
  38. INFO - Counters.log(589) |     Total committed heap usage (bytes)=336338944  
  39. INFO - Counters.log(589) |     CPU time spent (ms)=1770  
  40. INFO - Counters.log(589) |     Combine input records=0  
  41. INFO - Counters.log(589) |     SPLIT_RAW_BYTES=228  
  42. INFO - Counters.log(589) |     Reduce input records=8  
  43. INFO - Counters.log(589) |     Reduce input groups=4  
  44. INFO - Counters.log(589) |     Combine output records=0  
  45. INFO - Counters.log(589) |     Physical memory (bytes) snapshot=442564608  
  46. INFO - Counters.log(589) |     Reduce output records=4  
  47. INFO - Counters.log(589) |     Virtual memory (bytes) snapshot=2184306688  
  48. INFO - Counters.log(589) |     Map output records=8  
Java代码  收藏代码
  1. 模式：  192.168.75.130:9001  
  2. 存在此输出路径，已删除！！！  
  3. WARN - JobClient.copyAndConfigureFiles(746) | Use GenericOptionsParser for parsing the arguments. Applications should implement Tool for the same.  
  4. INFO - FileInputFormat.listStatus(237) | Total input paths to process : 2  
  5. WARN - NativeCodeLoader.<clinit>(52) | Unable to load native-hadoop library for your platform... using builtin-java classes where applicable  
  6. WARN - LoadSnappy.<clinit>(46) | Snappy native library not loaded  
  7. INFO - JobClient.monitorAndPrintJob(1380) | Running job: job_201404260312_0002  
  8. INFO - JobClient.monitorAndPrintJob(1393) |  map 0% reduce 0%  
  9. INFO - JobClient.monitorAndPrintJob(1393) |  map 50% reduce 0%  
  10. INFO - JobClient.monitorAndPrintJob(1393) |  map 100% reduce 0%  
  11. INFO - JobClient.monitorAndPrintJob(1393) |  map 100% reduce 33%  
  12. INFO - JobClient.monitorAndPrintJob(1393) |  map 100% reduce 100%  
  13. INFO - JobClient.monitorAndPrintJob(1448) | Job complete: job_201404260312_0002  
  14. INFO - Counters.log(585) | Counters: 29  
  15. INFO - Counters.log(587) |   Job Counters   
  16. INFO - Counters.log(589) |     Launched reduce tasks=1  
  17. INFO - Counters.log(589) |     SLOTS_MILLIS_MAPS=12445  
  18. INFO - Counters.log(589) |     Total time spent by all reduces waiting after reserving slots (ms)=0  
  19. INFO - Counters.log(589) |     Total time spent by all maps waiting after reserving slots (ms)=0  
  20. INFO - Counters.log(589) |     Launched map tasks=2  
  21. INFO - Counters.log(589) |     Data-local map tasks=2  
  22. INFO - Counters.log(589) |     SLOTS_MILLIS_REDUCES=9801  
  23. INFO - Counters.log(587) |   File Output Format Counters   
  24. INFO - Counters.log(589) |     Bytes Written=172  
  25. INFO - Counters.log(587) |   FileSystemCounters  
  26. INFO - Counters.log(589) |     FILE_BYTES_READ=237  
  27. INFO - Counters.log(589) |     HDFS_BYTES_READ=455  
  28. INFO - Counters.log(589) |     FILE_BYTES_WRITTEN=169503  
  29. INFO - Counters.log(589) |     HDFS_BYTES_WRITTEN=172  
  30. INFO - Counters.log(587) |   File Input Format Counters   
  31. INFO - Counters.log(589) |     Bytes Read=227  
  32. INFO - Counters.log(587) |   Map-Reduce Framework  
  33. INFO - Counters.log(589) |     Map output materialized bytes=243  
  34. INFO - Counters.log(589) |     Map input records=10  
  35. INFO - Counters.log(589) |     Reduce shuffle bytes=243  
  36. INFO - Counters.log(589) |     Spilled Records=16  
  37. INFO - Counters.log(589) |     Map output bytes=215  
  38. INFO - Counters.log(589) |     Total committed heap usage (bytes)=336338944  
  39. INFO - Counters.log(589) |     CPU time spent (ms)=1770  
  40. INFO - Counters.log(589) |     Combine input records=0  
  41. INFO - Counters.log(589) |     SPLIT_RAW_BYTES=228  
  42. INFO - Counters.log(589) |     Reduce input records=8  
  43. INFO - Counters.log(589) |     Reduce input groups=4  
  44. INFO - Counters.log(589) |     Combine output records=0  
  45. INFO - Counters.log(589) |     Physical memory (bytes) snapshot=442564608  
  46. INFO - Counters.log(589) |     Reduce output records=4  
  47. INFO - Counters.log(589) |     Virtual memory (bytes) snapshot=2184306688  
  48. INFO - Counters.log(589) |     Map output records=8  


在map侧过滤的数据，在50030中查看的截图如下： 



运行结果如下所示：

Java代码 复制代码 收藏代码
  1. 1   三劫散仙    13575468248 B   89  2013-02-05  
  2. 2   凤舞九天    18965235874 C   69  2013-03-09  
  3. 3   忙忙碌碌    15986854789 A   99  2013-03-05  
  4. 3   忙忙碌碌    15986854789 D   56  2013-06-07  
Java代码  收藏代码
  1. 1   三劫散仙    13575468248 B   89  2013-02-05  
  2. 2   凤舞九天    18965235874 C   69  2013-03-09  
  3. 3   忙忙碌碌    15986854789 A   99  2013-03-05  
  4. 3   忙忙碌碌    15986854789 D   56  2013-06-07  



至此，这个半链接就完成了，结果正确，在hadoop的几种join方式里，只有在Map侧的链接比较高效，但也需要根据具体的实际情况，进行选择。

【推广】 免费学中医，健康全家人
原文地址：https://www.cnblogs.com/sdaassjjsd/p/5593157.html