Hadoop中MapReduce过程中Shuffle过程实现自定义排序

Hadoop中MapReduce过程中Shuffle过程实现自定义排序

一、引言

MapReduce框架中的Shuffle过程是连接Map阶段和Reduce阶段的桥梁，负责将Map任务的输出结果按照key进行分组和排序，并将相同key的数据传递给对应的Reduce任务进行处理。Shuffle过程的性能直接影响到整个MapReduce作业的执行效率。在默认情况下，Hadoop使用TotalOrderPartitioner进行排序，但有时我们需要根据特定的业务逻辑进行自定义排序。本文将介绍两种方法来实现自定义排序：实现WritableComparable接口和使用Job.setSortComparatorClass方法。下面是详细的步骤和代码示例。

二、实现WritableComparable接口

1、自定义Key类

首先，我们需要定义一个类并实现WritableComparable接口，该接口要求实现compareTo方法，用于定义排序逻辑。

package mr;
import org.apache.hadoop.io.WritableComparable;
import java.io.DataInput;
import java.io.DataOutput;
import java.io.IOException;public class Employee implements WritableComparable<Employee> {private int empno;private String ename;private String job;private int mgr;private String hiredate;private int sal;private int comm;private int deptno;@Overridepublic String toString(){return "Employee[empno="+empno+",ename="+ename+",sal="+sal+",deptno="+deptno+"]";}@Overridepublic int compareTo(Employee o) {// 多个列的排序：select * from emp order by deptno,sal;// 首先按照deptno排序if(this.deptno > o.getDeptno()){return 1;}else if(this.deptno < o.getDeptno()){return -1;}// 如果deptno相等，按照sal排序if(this.sal >= o.getSal()){return 1;}else{return -1;}}@Overridepublic void write(DataOutput output) throws IOException {// 序列化output.writeInt(this.empno);output.writeUTF(this.ename);output.writeUTF(this.job);output.writeInt(this.mgr);output.writeUTF(this.hiredate);output.writeInt(this.sal);output.writeInt(this.comm);output.writeInt(this.deptno);}@Overridepublic void readFields(DataInput input) throws IOException {// 反序列化this.empno = input.readInt();this.ename = input.readUTF();this.job = input.readUTF();this.mgr = input.readInt();this.hiredate = input.readUTF();this.sal = input.readInt();this.comm = input.readInt();this.deptno = input.readInt();}
}

三、使用Job.setSortComparatorClass方法

2、设置自定义排序器

除了实现WritableComparable接口外，我们还可以使用Job.setSortComparatorClass方法来设置自定义排序器。这种方法允许我们在不修改Key类的情况下实现自定义排序。

package mr;
import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.io.NullWritable;
import org.apache.hadoop.mapreduce.Job;
import org.apache.hadoop.mapreduce.lib.input.FileInputFormat;
import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;public class CustomSort {public static class Map extends Mapper<Object, Text, Employee, IntWritable> {private static Employee emp = new Employee();private static IntWritable one = new IntWritable(1);@Overrideprotected void map(Object key, Text value, Context context) throws IOException, InterruptedException {String[] line = value.toString().split("\t");emp.setEmpno(Integer.parseInt(line[0]));emp.setEname(line[1]);emp.setJob(line[2]);emp.setMgr(Integer.parseInt(line[3]));emp.setHiredate(line[4]);emp.setSal(Integer.parseInt(line[5]));emp.setComm(Integer.parseInt(line[6]));emp.setDeptno(Integer.parseInt(line[7]));context.write(emp, one);}}public static class Reduce extends Reducer<Employee, IntWritable, Employee, IntWritable> {@Overrideprotected void reduce(Employee key, Iterable<IntWritable> values, Context context) throws IOException, InterruptedException {for (IntWritable val : values) {context.write(key, val);}}}public static void main(String[] args) throws Exception {Configuration conf = new Configuration();Job job = Job.getInstance(conf, "CustomSort");job.setJarByClass(CustomSort.class);job.setMapperClass(Map.class);job.setReducerClass(Reduce.class);job.setOutputKeyClass(Employee.class);job.setOutputValueClass(IntWritable.class);// 设置自定义排序器job.setSortComparatorClass(EmployeeComparator.class);Path in = new Path("hdfs://localhost:9000/mr/in/customsort");Path out = new Path("hdfs://localhost:9000/mr/out/customsort");FileInputFormat.addInputPath(job, in);FileOutputFormat.setOutputPath(job, out);System.exit(job.waitForCompletion(true) ? 0 : 1);}
}

3、自定义排序器类

package mr;
import org.apache.hadoop.io.WritableComparable;
import org.apache.hadoop.io.WritableComparator;public class EmployeeComparator extends WritableComparator {protected EmployeeComparator() {super(Employee.class, true);}@Overridepublic int compare(WritableComparable w1, WritableComparable w2) {Employee e1 = (Employee) w1;Employee e2 = (Employee) w2;// 首先按照deptno排序int deptCompare = Integer.compare(e1.getDeptno(), e2.getDeptno());if (deptCompare != 0) {return deptCompare;}// 如果deptno相等，按照sal排序return Integer.compare(e1.getSal(), e2.getSal());}
}

四、使用示例

下面是一个简单的MapReduce示例，展示了Shuffle过程在实际应用中的使用。这个示例中，我们使用了自定义的Employee类作为Key，并设置了自定义的排序器EmployeeComparator。

五、总结

通过实现WritableComparable接口和使用Job.setSortComparatorClass方法，我们可以在Hadoop MapReduce过程中实现自定义排序。这两种方法提供了灵活的排序机制，允许我们根据不同的业务需求对数据进行排序处理，从而提高数据处理的效率和准确性。

---

版权声明：本博客内容为原创，转载请保留原文链接及作者信息。

参考文章：

• Hadoop之mapreduce数据排序案例（详细代码）
• Java Job.setSortComparatorClass方法代码示例