多表关联和单表关联类似，它也是通过对原始数据进行一定的处理，从其中挖掘出关心的信息。如下
输入的是两个文件，一个代表工厂表，包含工厂名列和地址编号列；另一个代表地址表，包含地址名列和地址编号列。
要求从输入数据中找出工厂名和地址名的对应关系，输出工厂名-地址名表
样本如下：
factory:
<span style="font-size:14px;">factoryname addressed Beijing Red Star 1 Shenzhen Thunder 3 Guangzhou Honda 2 Beijing Rising 1 Guangzhou Development Bank 2 Tencent 3 Back of Beijing 1 </span>
address:
<span style="font-size:14px;">addressID addressname 1 Beijing 2 Guangzhou 3 Shenzhen 4 Xian </span>
结果：
<span style="font-size:14px;">factoryname addressname Beijing Red Star Beijing Beijing Rising Beijing Bank of Beijing Beijing Guangzhou Honda Guangzhou Guangzhou Development Bank Guangzhou Shenzhen Thunder Shenzhen Tencent Shenzhen </span>
代码如下：
<span style="font-size:14px;">import java.io.IOException;
import java.util.*;
import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.io.IntWritable;
import org.apache.hadoop.io.Text;
import org.apache.hadoop.mapreduce.Job;
import org.apache.hadoop.mapreduce.Mapper;
import org.apache.hadoop.mapreduce.Reducer;
import org.apache.hadoop.mapreduce.lib.input.FileInputFormat;
import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;
import org.apache.hadoop.util.GenericOptionsParser;
public class MTjoin {
public static int time = 0;
/*
* 在map中先区分输入行属于左表还是右表，然后对两列值进行分割，
* 保存连接列在key值，剩余列和左右表标志在value中，最后输出
*/
public static class Map extends Mapper<Object, Text, Text, Text> {
// 实现map函数</span><span style="font-size:14px;"> public void map(Object key, Text value, Context context)
throws IOException, InterruptedException {
String line = value.toString();// 每行文件
String relationtype = new String();// 左右表标识
// 输入文件首行，不处理
if (line.contains("factoryname") == true
|| line.contains("addressed") == true) {
return;
}
// 输入的一行预处理文本
StringTokenizer itr = new StringTokenizer(line);
String mapkey = new String();
String mapvalue = new String();
int i = 0;
while (itr.hasMoreTokens()) {
// 先读取一个单词
String token = itr.nextToken();
// 判断该地址ID就把存到"values[0]"
if (token.charAt(0) >= '0' && token.charAt(0) <= '9') {
mapkey = token;
if (i > 0) {
relationtype = "1";
} else {
relationtype = "2";
}
continue;
}
// 存工厂名
mapvalue += token + " ";
i++;
}
// 输出左右表
context.write(new Text(mapkey), new Text(relationtype + "+"+ mapvalue));
}
}
/*
* reduce解析map输出，将value中数据按照左右表分别保存，
* 然后求出笛卡尔积，并输出。
*/
public static class Reduce extends Reducer<Text, Text, Text, Text> {
// 实现reduce函数
public void reduce(Text key, Iterable<Text> values, Context context)
throws IOException, InterruptedException {
// 输出表头
if (0 == time) {
context.write(new Text("factoryname"), new Text("addressname"));
time++;
}
int factorynum = 0;
String[] factory = new String[10];
int addressnum = 0;
String[] address = new String[10];
Iterator ite = values.iterator();
while (ite.hasNext()) {
String record = ite.next().toString();
int len = record.length();
int i = 2;
if (0 == len) {
continue;
}
// 取得左右表标识
char relationtype = record.charAt(0);
// 左表
if ('1' == relationtype) {
factory[factorynum] = record.substring(i);
factorynum++;
}
// 右表
if ('2' == relationtype) {
address[addressnum] = record.substring(i);
addressnum++;
}
}
// 求笛卡尔积
if (0 != factorynum && 0 != addressnum) {
for (int m = 0; m < factorynum; m++) {
for (int n = 0; n < addressnum; n++) {
// 输出结果
context.write(new Text(factory[m]),
new Text(address[n]));
}
}
}
}
}
public static void main(String[] args) throws Exception {
Configuration conf = new Configuration();
// 这句话很关键
// conf.set("mapred.job.tracker", "192.168.1.2:9001");
//可使用args
// String[] ioArgs = new String[] { "MTjoin_in", "MTjoin_out" };
String[] otherArgs = new GenericOptionsParser(conf, args).getRemainingArgs();
if (otherArgs.length != 2) {
System.err.println("Usage: Multiple Table Join <in> <out>");
System.exit(2);
}
Job job = new Job(conf, "Multiple Table Join");
job.setJarByClass(MTjoin.class);
// 设置Map和Reduce处理类
job.setMapperClass(Map.class);
job.setReducerClass(Reduce.class);
// 设置输出类型
job.setOutputKeyClass(Text.class);
job.setOutputValueClass(Text.class);
// 设置输入和输出目录
FileInputFormat.addInputPath(job, new Path(otherArgs[0]));
FileOutputFormat.setOutputPath(job, new Path(otherArgs[1]));
System.exit(job.waitForCompletion(true) ? 0 : 1);
}
}
</span><span style="font-size:14px;">javac -classpath hadoop-core-1.1.2.jar:/opt/hadoop-1.1.2/lib/commons-cli-1.2.jar -d firstProject firstProject/MTJoin.java </span>
<span style="font-size:14px;">jar -cvf MTJoin.jar -C firstProject/ . </span>
<span style="font-size:14px;"> </span>
删除已经存在的output
<span style="font-size:14px;">hadoop fs -rmr output </span>
<span style="font-size:14px;">hadoop fs -mkdir input </span>
<span style="font-size:14px;">hadoop fs -put factory input </span>
<span style="font-size:14px;"> hadoop fs -put address input </span>
运行
<span style="font-size:14px;">hadoop jar MTJoin.jar MTJoin input output </span>
查看结果
<span style="font-size:14px;"> hadoop fs -cat output/part-r-00000</span>
原文地址：http://blog.csdn.net/sunlei1980/article/details/46603279
