[BD] MapReduce

时间：2020-04-23 01:38:45 阅读：97 评论：0 收藏：0 [点我收藏+]

概述

大数据计算的核心思想：移动计算比移动数据更划算
MapReduce既是一个编程模型，又是一个计算框架
包含Map和Reduce两个过程
终极目标：用SQL语句分析大数据（Hive、SparkSQL，将SQL语句转换为MR程序）

特性

序列化
- java序列化：实现序列化接口（标识接口）

Student.java

 1 package serializable;
 2 
 3 import java.io.Serializable;
 4 
 5 public class Student implements Serializable{
 6     private int stuID;
 7     private String stuName;
 8     public int getStuID() {
 9         return stuID;
10     }
11     public void setStuID(int stuID) {
12         this.stuID = stuID;
13     }
14     public String getStuName() {
15         return stuName;
16     }
17     public void setStuName(String stuName) {
18         this.stuName = stuName;
19     }
20 }

View Code

TestMain.java

 1 package serializable;
 2 
 3 import java.io.FileOutputStream;
 4 import java.io.ObjectOutputStream;
 5 import java.io.OutputStream;
 6 
 7 import serializable.Student;
 8 
 9 public class TestMain {
10 
11     public static void main(String[] args) throws Exception{
12         // 创建学生对象
13         Student s = new Student();
14         s.setStuID(1);
15         s.setStuName("Tom");
16         
17         // 输出对象到文件
18         OutputStream out = new FileOutputStream("F:\\eclipse-workspace\\student.ooo");
19         ObjectOutputStream oos = new ObjectOutputStream(out);
20         oos.writeObject(s);
21             
22         oos.close();
23         out.close();
24     }
25 }

View Code

- MapReduce序列化：核心接口Writable，实现了Writeble的类的对象可作为MapReduce的key和value
  - 读取员工数据，生成员工对象，直接输出到HDFS
  - hadoop jar s2.jar /scott/emp.csv /output/0910/s2

EmpInfoMain.java

 1 import org.apache.hadoop.conf.Configuration;
 2 import org.apache.hadoop.fs.Path;
 3 import org.apache.hadoop.io.IntWritable;
 4 import org.apache.hadoop.mapreduce.Job;
 5 import org.apache.hadoop.mapreduce.lib.input.FileInputFormat;
 6 import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;
 7 
 8 public class EmpInfoMain {
 9 
10     public static void main(String[] args) throws Exception {
11         Job job = Job.getInstance(new Configuration());
12         job.setJarByClass(EmpInfoMain.class);
13         
14         job.setMapperClass(EmpInfoMapper.class);
15         job.setMapOutputKeyClass(IntWritable.class);
16         job.setMapOutputValueClass(Emp.class);
17         
18         job.setOutputKeyClass(IntWritable.class);
19         job.setOutputKeyClass(Emp.class);
20         
21         FileInputFormat.setInputPaths(job, new Path(args[0]));
22         FileOutputFormat.setOutputPath(job, new Path(args[1]));
23 
24         job.waitForCompletion(true);
25     }
26 }

View Code

EmpInfoMapper.java

 1 import java.io.IOException;
 2 
 3 import org.apache.hadoop.io.IntWritable;
 4 import org.apache.hadoop.io.LongWritable;
 5 import org.apache.hadoop.io.Text;
 6 import org.apache.hadoop.mapreduce.Mapper;
 7 
 8 public class EmpInfoMapper extends Mapper<LongWritable, Text, IntWritable, Emp>{
 9 
10     @Override
11     protected void map(LongWritable key1, Text value1,
12             Context context)
13             throws IOException, InterruptedException {
14         //数据：7654,MARTIN,SALESMAN,7698,1981/9/28,1250,1400,30
15         String data = value1.toString();
16         
17         // 分词
18         String[] words = data.split(",");
19         
20         // 生成员工对象
21         Emp emp = new Emp();
22         emp.setEmpno(Integer.parseInt(words[0]));
23         emp.setEname(words[1]);
24         emp.setJob(words[2]);
25         emp.setMgr(Integer.parseInt(words[3]));
26         emp.setHiredate(words[4]);
27         emp.setSal(Integer.parseInt(words[5]));
28         emp.setComm(Integer.parseInt(words[6]));
29         emp.setDeptno(Integer.parseInt(words[7]));
30         
31         // 输出员工对象
32         context.write(new IntWritable(emp.getEmpno()), emp);
33     }
34 }

View Code

Emp.java

  1 import java.io.DataInput;
  2 import java.io.DataOutput;
  3 import java.io.IOException;
  4 
  5 import org.apache.hadoop.io.Writable;
  6 
  7 public class Emp implements Writable{
  8     private int empno;
  9     private String ename;
 10     private String job;
 11     private int mgr;
 12     private String hiredate;
 13     private int sal;
 14     private int comm;
 15     private int deptno;
 16     
 17     @Override
 18     public void readFields(DataInput input)
 19             throws IOException {
 20         // 实现反序列化，从输入流中读取对象（与输出流顺序一样）
 21         this.empno = input.readInt();
 22         this.ename = input.readUTF();
 23         this.job = input.readUTF();
 24         this.mgr = input.readInt();
 25         this.hiredate = input.readUTF();
 26         this.sal = input.readInt();
 27         this.comm = input.readInt();
 28         this.deptno = input.readInt();
 29     }
 30 
 31     @Override
 32     public void write(DataOutput output) throws IOException {
 33         // 实现序列化，把对象输出到输出流
 34         output.writeInt(this.empno);
 35         output.writeUTF(this.ename);
 36         output.writeUTF(this.job);
 37         output.writeInt(this.mgr);
 38         output.writeUTF(this.hiredate);
 39         output.writeInt(this.sal);
 40         output.writeInt(this.comm);
 41         output.writeInt(this.deptno);
 42     }
 43     
 44     public int getEmpno() {
 45         return empno;
 46     }
 47 
 48     public void setEmpno(int empno) {
 49         this.empno = empno;
 50     }
 51 
 52     public String getEname() {
 53         return ename;
 54     }
 55 
 56     public void setEname(String ename) {
 57         this.ename = ename;
 58     }
 59 
 60     public String getJob() {
 61         return job;
 62     }
 63 
 64     public void setJob(String job) {
 65         this.job = job;
 66     }
 67 
 68     public int getMgr() {
 69         return mgr;
 70     }
 71 
 72     public void setMgr(int mgr) {
 73         this.mgr = mgr;
 74     }
 75 
 76     public String getHiredate() {
 77         return hiredate;
 78     }
 79 
 80     public void setHiredate(String hiredate) {
 81         this.hiredate = hiredate;
 82     }
 83 
 84     public int getSal() {
 85         return sal;
 86     }
 87 
 88     public void setSal(int sal) {
 89         this.sal = sal;
 90     }
 91 
 92     public int getComm() {
 93         return comm;
 94     }
 95 
 96     public void setComm(int comm) {
 97         this.comm = comm;
 98     }
 99 
100     public int getDeptno() {
101         return deptno;
102     }
103 
104     public void setDeptno(int deptno) {
105         this.deptno = deptno;
106     }
107 
108     @Override
109     public String toString() {
110         return "Emp [empno=" + empno + ", ename=" + ename
111                 + ", sal=" + sal + ", deptno=" + deptno
112                 + "]";
113     }
114 
115 }

View Code

技术分享图片

- - 求部门工资总额
    - 由于实现了序列化接口，员工可作为key和value
    - 相比之前的程序更加简洁
    - 对于大对象可能产生性能问题

EmpInfoMain.java

 1 import org.apache.hadoop.conf.Configuration;
 2 import org.apache.hadoop.fs.Path;
 3 import org.apache.hadoop.io.IntWritable;
 4 import org.apache.hadoop.mapreduce.Job;
 5 import org.apache.hadoop.mapreduce.lib.input.FileInputFormat;
 6 import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;
 7 
 8 public class EmpInfoMain {
 9 
10     public static void main(String[] args) throws Exception {
11         Job job = Job.getInstance(new Configuration());
12         job.setJarByClass(EmpInfoMain.class);
13         
14         job.setMapperClass(EmpInfoMapper.class);
15         job.setMapOutputKeyClass(IntWritable.class);
16         job.setMapOutputValueClass(Emp.class);  // 输出就是员工对象
17         
18         job.setOutputKeyClass(IntWritable.class);
19         job.setOutputValueClass(Emp.class);
20         
21         FileInputFormat.setInputPaths(job, new Path(args[0]));
22         FileOutputFormat.setOutputPath(job, new Path(args[1]));
23 
24         job.waitForCompletion(true);
25     }
26 
27 }

View Code

SalaryTotalMapper.java

 1 import java.io.IOException;
 2 import org.apache.hadoop.io.IntWritable;
 3 import org.apache.hadoop.io.LongWritable;
 4 import org.apache.hadoop.io.Text;
 5 import org.apache.hadoop.mapreduce.Mapper;
 6 
 7 //                                                               k2:部门号  v2:员工对象
 8 public class SalaryTotalMapper extends Mapper<LongWritable, Text, IntWritable, Emp> {
 9 
10     @Override
11     protected void map(LongWritable key1, Text value1, Context context)
12             throws IOException, InterruptedException {
13         // 数据：7654,MARTIN,SALESMAN,7698,1981/9/28,1250,1400,30
14         String data = value1.toString();
15         
16         // 分词
17         String[] words = data.split(",");
18         
19         // 生成员工对象
20         Emp emp = new Emp();
21         emp.setEmpno(Integer.parseInt(words[0]));
22         emp.setEname(words[1]);
23         emp.setJob(words[2]);
24         emp.setMgr(Integer.parseInt(words[3]));
25         emp.setHiredate(words[4]);
26         emp.setSal(Integer.parseInt(words[5]));
27         emp.setComm(Integer.parseInt(words[6]));
28         emp.setDeptno(Integer.parseInt(words[7]));
29         
30         // 输出员工对象  k2:部门号                 v2:员工对象
31         context.write(new IntWritable(emp.getDeptno()), emp);
32     }
33 }

View Code

SalaryTotalReducer.java

 1 import java.io.IOException;
 2 import org.apache.hadoop.io.IntWritable;
 3 import org.apache.hadoop.mapreduce.Reducer;
 4 
 5 public class SalaryTotalReducer extends Reducer<IntWritable, Emp, IntWritable, IntWritable> {
 6 
 7     @Override
 8     protected void reduce(IntWritable k3, Iterable<Emp> v3,Context context) throws IOException, InterruptedException {
 9         int total = 0;
10         
11         //取出员工薪水，并求和
12         for(Emp e:v3){
13             total = total + e.getSal();
14         }
15         context.write(k3, new IntWritable(total));
16     }
17 }

View Code

排序
- 规则：按照Key2排序
- 基本数据类型
  - 数字：默认升序
    - 自定义比较器：MyNumberComparator.java
    - 主程序中添加比较器：job.setSortComparatorClass(MyNumberCompatator.class)

MyNumberComparator.java

 1 import org.apache.hadoop.io.IntWritable;
 2 
 3 //针对数字创建自己的比较规则，执行降序排序
 4 public class MyNumberComparator extends IntWritable.Comparator {
 5 
 6     @Override
 7     public int compare(byte[] b1, int s1, int l1, byte[] b2, int s2, int l2) {
 8         // TODO Auto-generated method stub
 9         return -super.compare(b1, s1, l1, b2, s2, l2);
10     }
11 }

View Code

- 字符串：默认字典顺序

MyTextComparator.java

 1 import org.apache.hadoop.io.Text;
 2 
 3 // 针对Text定义比较规则
 4 public class MyTextComparator extends Text.Comparator{
 5 
 6     @Override
 7     public int compare(byte[] b1, int s1, int l1, byte[] b2,
 8             int s2, int l2) {
 9         return -super.compare(b1, s1, l1, b2, s2, l2);
10     }
11 }

View Code

- 对象
  - SQL排序：order by + 列名 / 表达式 / 列的别名 / 序号 / 多个列（desc作用于最近一列）
  - 前提：对象必须是Key2；必须实现Writable接口；必须是可排序对象（类似Java对象排序，要实现Comparable<T>）

Student.java

 1 //学生对象：按照学生的age年龄进行排序
 2 public class Student implements Comparable<Student>{
 3 
 4     private int stuID;
 5     private String stuName;
 6     private int age;
 7         
 8     @Override
 9     public String toString() {
10         return "Student [stuID=" + stuID + ", stuName=" + stuName + ", age=" + age + "]";
11     }
12 
13     @Override
14     public int compareTo(Student o) {
15         // 定义排序规则：按照学生的age年龄进行排序
16         if(this.age >= o.getAge()){
17             return 1;
18         }else{
19             return -1;
20         }
21     }
22     
23     public int getStuID() {
24         return stuID;
25     }
26     public void setStuID(int stuID) {
27         this.stuID = stuID;
28     }
29     public String getStuName() {
30         return stuName;
31     }
32     public void setStuName(String stuName) {
33         this.stuName = stuName;
34     }
35     public int getAge() {
36         return age;
37     }
38     public void setAge(int age) {
39         this.age = age;
40     }
41     
42 }

View Code

StudentMain.java

 1 import java.util.Arrays;
 2 
 3 public class StudentMain {
 4 
 5     public static void main(String[] args) {
 6         //创建几个学生对象
 7         Student s1 = new Student();
 8         s1.setStuID(1);
 9         s1.setStuName("Tom");
10         s1.setAge(24);
11         
12         Student s2 = new Student();
13         s2.setStuID(2);
14         s2.setStuName("Mary");
15         s2.setAge(26);
16         
17         Student s3 = new Student();
18         s3.setStuID(3);
19         s3.setStuName("Mike");
20         s3.setAge(25);
21         
22         //生成一个数组
23         Student[] list = {s1,s2,s3};
24         
25         //排序
26         Arrays.sort(list);
27         
28         //输出
29         for(Student s:list){
30             System.out.println(s);
31         }
32     }
33 }

View Code

- - 一个/多个列排序

Emp.java

  1 import java.io.DataInput;
  2 import java.io.DataOutput;
  3 import java.io.IOException;
  4 
  5 import org.apache.hadoop.io.Writable;
  6 import org.apache.hadoop.io.WritableComparable;
  7 
  8 //代表员工
  9 //数据：7654,MARTIN,SALESMAN,7698,1981/9/28,1250,1400,30
 10 public class Emp implements WritableComparable<Emp>{
 11     
 12     private int empno;//员工号
 13     private String ename; //员工姓名
 14     private String job; //职位
 15     private int mgr; //经理的员工号
 16     private String hiredate;//入职日期
 17     private int sal; //月薪
 18     private int comm; //奖金
 19     private int deptno; //部门号
 20     
 21 //    @Override
 22 //    public int compareTo(Emp o) {
 23 //        // 定义自己的排序规则：一个列的排序
 24 //        // 按照薪水进行排序
 25 //        if(this.sal >= o.getSal()){
 26 //            return 1;
 27 //        }else{
 28 //            return -1;
 29 //        }
 30 //    }
 31     
 32     @Override
 33     public int compareTo(Emp o) {
 34         // 定义自己的排序规则：多个列的排序
 35         // 先按照部门号进行排序，再按照薪水进行排序
 36         if(this.deptno > o.getDeptno()){
 37             return 1;
 38         }else if(this.deptno < o.getDeptno()){
 39             return -1;
 40         }
 41         
 42         //再按照薪水进行排序
 43         if(this.sal >= o.getSal()){
 44             return 1;
 45         }else{
 46             return -1;
 47         }
 48     }    
 49         
 50     @Override
 51     public String toString() {
 52         return "Emp [empno=" + empno + ", ename=" + ename + ", sal=" + sal + ", deptno=" + deptno + "]";
 53     }
 54 
 55     @Override
 56     public void readFields(DataInput input) throws IOException {
 57         //实现反序列化，从输入流中读取对象
 58         this.empno = input.readInt();
 59         this.ename = input.readUTF();
 60         this.job = input.readUTF();
 61         this.mgr = input.readInt();
 62         this.hiredate = input.readUTF();
 63         this.sal = input.readInt();
 64         this.comm = input.readInt();
 65         this.deptno = input.readInt();
 66     }
 67     
 68     @Override
 69     public void write(DataOutput output) throws IOException {
 70         // 实现序列化，把对象输出到输出流
 71         output.writeInt(this.empno);
 72         output.writeUTF(this.ename);
 73         output.writeUTF(this.job);
 74         output.writeInt(this.mgr);
 75         output.writeUTF(this.hiredate);
 76         output.writeInt(this.sal);
 77         output.writeInt(this.comm);
 78         output.writeInt(this.deptno);
 79     }
 80     
 81     
 82     public int getEmpno() {
 83         return empno;
 84     }
 85     public void setEmpno(int empno) {
 86         this.empno = empno;
 87     }
 88     public String getEname() {
 89         return ename;
 90     }
 91     public void setEname(String ename) {
 92         this.ename = ename;
 93     }
 94     public String getJob() {
 95         return job;
 96     }
 97     public void setJob(String job) {
 98         this.job = job;
 99     }
100     public int getMgr() {
101         return mgr;
102     }
103     public void setMgr(int mgr) {
104         this.mgr = mgr;
105     }
106     public String getHiredate() {
107         return hiredate;
108     }
109     public void setHiredate(String hiredate) {
110         this.hiredate = hiredate;
111     }
112     public int getSal() {
113         return sal;
114     }
115     public void setSal(int sal) {
116         this.sal = sal;
117     }
118     public int getComm() {
119         return comm;
120     }
121     public void setComm(int comm) {
122         this.comm = comm;
123     }
124     public int getDeptno() {
125         return deptno;
126     }
127     public void setDeptno(int deptno) {
128         this.deptno = deptno;
129     }
130 
131 }

View Code

EmpSortMapper.java

 1 import java.io.IOException;
 2 
 3 import org.apache.hadoop.io.IntWritable;
 4 import org.apache.hadoop.io.LongWritable;
 5 import org.apache.hadoop.io.NullWritable;
 6 import org.apache.hadoop.io.Text;
 7 import org.apache.hadoop.mapreduce.Mapper;
 8 
 9 /*
10  * 一定要把Emp作为key2
11  * 没有value2，返回null值
12  */
13  
14 public class EmpSortMapper extends Mapper<LongWritable, Text, Emp, NullWritable> {
15 
16     @Override
17     protected void map(LongWritable key1, Text value1, Context context)
18             throws IOException, InterruptedException {
19         // 数据：7654,MARTIN,SALESMAN,7698,1981/9/28,1250,1400,30
20         String data = value1.toString();
21         
22         //分词
23         String[] words = data.split(",");
24         
25         //生成员工对象
26         Emp emp = new Emp();
27         emp.setEmpno(Integer.parseInt(words[0]));
28         emp.setEname(words[1]);
29         emp.setJob(words[2]);
30         emp.setMgr(Integer.parseInt(words[3]));
31         emp.setHiredate(words[4]);
32         emp.setSal(Integer.parseInt(words[5]));
33         emp.setComm(Integer.parseInt(words[6]));
34         emp.setDeptno(Integer.parseInt(words[7]));
35         
36         //输出员工对象  k2:员工对象  v2：空值
37         context.write(emp,  NullWritable.get());
38     }
39 }

View Code

EmpSortMain.java

 1 import java.io.IOException;
 2 
 3 import org.apache.hadoop.conf.Configuration;
 4 import org.apache.hadoop.fs.Path;
 5 import org.apache.hadoop.io.IntWritable;
 6 import org.apache.hadoop.io.NullWritable;
 7 import org.apache.hadoop.mapreduce.Job;
 8 import org.apache.hadoop.mapreduce.lib.input.FileInputFormat;
 9 import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;
10 
11 public class EmpSortMain {
12 
13     public static void main(String[] args) throws Exception {
14         Job job = Job.getInstance(new Configuration());
15         job.setJarByClass(EmpSortMain.class);
16         
17         job.setMapperClass(EmpSortMapper.class);
18         job.setMapOutputKeyClass(Emp.class); //k2 是员工对象
19         job.setMapOutputValueClass(NullWritable.class);  // v2:是空值
20         
21         job.setOutputKeyClass(Emp.class);
22         job.setOutputValueClass(NullWritable.class);
23         
24         FileInputFormat.setInputPaths(job, new Path(args[0]));
25         FileOutputFormat.setOutputPath(job, new Path(args[1]));
26 
27         job.waitForCompletion(true);
28 
29     }
30 
31 }

View Code

分区（Partition）
- 关系型数据库分区
  - 分区1：sal<=3000，分区2：3000<sal<=5000，分区3：sal>5000
  - 查询薪水1000~2000的员工，只扫描分区1即可
  - Hash分区：根据数值的Hash结果进行分区，如果一样就放入同一个分区中（数据尽量打散，避免热块）
  - Redis Cluster、Hive 桶表、MongoDB 分布式路由
- MR分区
  - 根据Map的输出<key2 value2>进行分区
  - 默认情况下，MR的输出只有一个分区（一个分区就是一个文件）
  - 按部门号分区：不同部门的员工放到不同文件中

Emp.java

  1 import java.io.DataInput;
  2 import java.io.DataOutput;
  3 import java.io.IOException;
  4 
  5 import org.apache.hadoop.io.Writable;
  6 
  7 //代表员工
  8 //数据：7654,MARTIN,SALESMAN,7698,1981/9/28,1250,1400,30
  9 public class Emp implements Writable{
 10     
 11     private int empno;//员工号
 12     private String ename; //员工姓名
 13     private String job; //职位
 14     private int mgr; //经理的员工号
 15     private String hiredate;//入职日期
 16     private int sal; //月薪
 17     private int comm; //奖金
 18     private int deptno; //部门号
 19     
 20         
 21     @Override
 22     public String toString() {
 23         return "Emp [empno=" + empno + ", ename=" + ename + ", sal=" + sal + ", deptno=" + deptno + "]";
 24     }
 25 
 26     @Override
 27     public void readFields(DataInput input) throws IOException {
 28         //实现反序列化，从输入流中读取对象
 29         this.empno = input.readInt();
 30         this.ename = input.readUTF();
 31         this.job = input.readUTF();
 32         this.mgr = input.readInt();
 33         this.hiredate = input.readUTF();
 34         this.sal = input.readInt();
 35         this.comm = input.readInt();
 36         this.deptno = input.readInt();
 37     }
 38     
 39     @Override
 40     public void write(DataOutput output) throws IOException {
 41         // 实现序列化，把对象输出到输出流
 42         output.writeInt(this.empno);
 43         output.writeUTF(this.ename);
 44         output.writeUTF(this.job);
 45         output.writeInt(this.mgr);
 46         output.writeUTF(this.hiredate);
 47         output.writeInt(this.sal);
 48         output.writeInt(this.comm);
 49         output.writeInt(this.deptno);
 50     }
 51     
 52     
 53     public int getEmpno() {
 54         return empno;
 55     }
 56     public void setEmpno(int empno) {
 57         this.empno = empno;
 58     }
 59     public String getEname() {
 60         return ename;
 61     }
 62     public void setEname(String ename) {
 63         this.ename = ename;
 64     }
 65     public String getJob() {
 66         return job;
 67     }
 68     public void setJob(String job) {
 69         this.job = job;
 70     }
 71     public int getMgr() {
 72         return mgr;
 73     }
 74     public void setMgr(int mgr) {
 75         this.mgr = mgr;
 76     }
 77     public String getHiredate() {
 78         return hiredate;
 79     }
 80     public void setHiredate(String hiredate) {
 81         this.hiredate = hiredate;
 82     }
 83     public int getSal() {
 84         return sal;
 85     }
 86     public void setSal(int sal) {
 87         this.sal = sal;
 88     }
 89     public int getComm() {
 90         return comm;
 91     }
 92     public void setComm(int comm) {
 93         this.comm = comm;
 94     }
 95     public int getDeptno() {
 96         return deptno;
 97     }
 98     public void setDeptno(int deptno) {
 99         this.deptno = deptno;
100     }
101 }

View Code

MyPartitionerMain.java

 1 import java.io.IOException;
 2 
 3 import org.apache.hadoop.conf.Configuration;
 4 import org.apache.hadoop.fs.Path;
 5 import org.apache.hadoop.io.IntWritable;
 6 import org.apache.hadoop.mapreduce.Job;
 7 import org.apache.hadoop.mapreduce.lib.input.FileInputFormat;
 8 import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;
 9 
10 public class MyPartitionerMain {
11 
12     public static void main(String[] args) throws Exception {
13         Job job = Job.getInstance(new Configuration());
14         job.setJarByClass(MyPartitionerMain.class);
15         
16         job.setMapperClass(MyPartitionerMapper.class);
17         job.setMapOutputKeyClass(IntWritable.class); //k2 是部门号
18         job.setMapOutputValueClass(Emp.class);  // v2输出就是员工对象
19         
20         //加入分区规则
21         job.setPartitionerClass(MyPartitioner.class);    
22         //指定分区的个数
23         job.setNumReduceTasks(3);
24         
25         job.setReducerClass(MyPartitionerReducer.class);
26         job.setOutputKeyClass(IntWritable.class);
27         job.setOutputValueClass(Emp.class);
28         
29         FileInputFormat.setInputPaths(job, new Path(args[0]));
30         FileOutputFormat.setOutputPath(job, new Path(args[1]));
31 
32         job.waitForCompletion(true);
33     }
34 }

View Code

MyPartitionerMapper.java

 1 import java.io.IOException;
 2 
 3 import org.apache.hadoop.io.IntWritable;
 4 import org.apache.hadoop.io.LongWritable;
 5 import org.apache.hadoop.io.Text;
 6 import org.apache.hadoop.mapreduce.Mapper;
 7 
 8 //                                                                 k2 部门号      v2 员工
 9 public class MyPartitionerMapper extends Mapper<LongWritable, Text, IntWritable, Emp> {
10 
11     @Override
12     protected void map(LongWritable key1, Text value1, Context context)
13             throws IOException, InterruptedException {
14         // 数据：7654,MARTIN,SALESMAN,7698,1981/9/28,1250,1400,30
15         String data = value1.toString();
16         
17         //分词
18         String[] words = data.split(",");
19         
20         //生成员工对象
21         Emp emp = new Emp();
22         emp.setEmpno(Integer.parseInt(words[0]));
23         emp.setEname(words[1]);
24         emp.setJob(words[2]);
25         emp.setMgr(Integer.parseInt(words[3]));
26         emp.setHiredate(words[4]);
27         emp.setSal(Integer.parseInt(words[5]));
28         emp.setComm(Integer.parseInt(words[6]));
29         emp.setDeptno(Integer.parseInt(words[7]));
30         
31         //输出员工对象  k2:部门号                                                                     v2：员工对象
32         context.write(new IntWritable(emp.getDeptno()), emp);
33     }
34 }

View Code

MyPartitionerReducer.java

 1 import java.io.IOException;
 2 
 3 import org.apache.hadoop.io.IntWritable;
 4 import org.apache.hadoop.mapreduce.Reducer;
 5 
 6 //就是同一个部门的员工
 7 public class MyPartitionerReducer extends Reducer<IntWritable, Emp, IntWritable, Emp> {
 8 
 9     @Override
10     protected void reduce(IntWritable k3, Iterable<Emp> v3,Context context) throws IOException, InterruptedException {
11         // 直接输出
12         for(Emp e:v3){
13             context.write(k3, e);
14         }
15     }
16 }

View Code

合并（Combiner）
- 一种特殊的Reducer，部署在Mapper端
- 在Mapper端执行一次合并，用于减少Mapper输出到Reducer的数据量，提高效率
- 在wordCountMain.java中添加job.setCombinerClass(WordCountReducer.class)
- 谨慎使用Combiner，有些情况不能使用（如求平均值）
- 不管有没有Combiner，都不能改变Map和Reduce对应的数据类型
- 程序出错，可将Reducer的k3 类型改为DoubleWritable

AvgSalaryMain.java

 1 import java.io.IOException;
 2 
 3 import org.apache.hadoop.conf.Configuration;
 4 import org.apache.hadoop.fs.Path;
 5 import org.apache.hadoop.io.DoubleWritable;
 6 import org.apache.hadoop.io.IntWritable;
 7 import org.apache.hadoop.io.Text;
 8 import org.apache.hadoop.mapreduce.Job;
 9 import org.apache.hadoop.mapreduce.lib.input.FileInputFormat;
10 import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;
11 
12 public class AvgSalaryMain {
13 
14     public static void main(String[] args) throws Exception {
15         //1、创建任务、指定任务的入口 
16         Job job = Job.getInstance(new Configuration());
17         job.setJarByClass(AvgSalaryMain.class);
18         
19         //2、指定任务的map和map输出的数据类型
20         job.setMapperClass(AvgSalaryMapper.class);
21         job.setMapOutputKeyClass(Text.class);
22         job.setMapOutputValueClass(IntWritable.class);
23         
24         //加入Combiner
25         job.setCombinerClass(AvgSalaryReducer.class);
26         
27         //3、指定任务的reducer和reducer输出的类型
28         job.setReducerClass(AvgSalaryReducer.class);
29         job.setOutputKeyClass(Text.class);
30         job.setOutputValueClass(DoubleWritable.class);
31         
32         //4、指定任务输入路径和输出路径
33         FileInputFormat.setInputPaths(job, new Path(args[0]));
34         FileOutputFormat.setOutputPath(job, new Path(args[1]));
35         
36         //5、执行任务
37         job.waitForCompletion(true);
38 
39     }
40 }

View Code

AvgSalaryMapper.java

 1 import java.io.IOException;
 2 
 3 import org.apache.hadoop.io.IntWritable;
 4 import org.apache.hadoop.io.LongWritable;
 5 import org.apache.hadoop.io.Text;
 6 import org.apache.hadoop.mapreduce.Mapper;
 7 
 8 //                                                              k2 常量   v2：薪水
 9 public class AvgSalaryMapper extends Mapper<LongWritable, Text, Text, IntWritable> {
10 
11     @Override
12     protected void map(LongWritable key1, Text value1, Context context)
13             throws IOException, InterruptedException {
14         // 数据：7654,MARTIN,SALESMAN,7698,1981/9/28,1250,1400,30
15         String data = value1.toString();
16         
17         //分词
18         String[] words = data.split(",");
19         
20         //输出  k2 常量   v2 薪水
21         context.write(new Text("salary"), new IntWritable(Integer.parseInt(words[5])));
22     }
23 }

View Code

AvgSalaryReducer.java

 1 import java.io.IOException;
 2 
 3 import org.apache.hadoop.io.DoubleWritable;
 4 import org.apache.hadoop.io.IntWritable;
 5 import org.apache.hadoop.io.Text;
 6 import org.apache.hadoop.mapreduce.Reducer;
 7 
 8 //                                                                     v4：平均工资
 9 public class AvgSalaryReducer extends Reducer<Text, IntWritable, Text, DoubleWritable> {
10 
11     @Override
12     protected void reduce(Text k3, Iterable<IntWritable> v3,Context context) throws IOException, InterruptedException {
13         int total = 0;
14         int count = 0;
15         
16         for(IntWritable salary:v3){
17             //工资求和
18             total = total + salary.get();
19             //人数加一
20             count ++;
21         }
22         
23         //输出
24         context.write(new Text("The avg salary is :"), new DoubleWritable(total/count));
25     }
26 
27 }

View Code

技术分享图片

Shuffle

MapReduce的核心
map输出后到reduce接收前
Hadoop 3.x以前：会有数据落地（I/O操作）
Spark只有两次I/O操作（读+写），中间运算在内存中完成

技术分享图片

例题

词频统计（word count）
- 一篇文章用哈希表统计即可
- 对互联网所有网页的词频进行统计（Google搜索引擎的需求），无法将所有网页读入内存
- 过程
  - map：将单词提取出来，对每个单词输入一个<word,1>这样的<k,v>对，进而将相同的数据放在一起，形成<word,<1,1,1,...>>这样的<k,v集合>
  - reduce：将集合里的1求和，再将单词和这个和组成<word,sum>输出
- 一个map函数仅对一个HDFS数据块上的数据进行计算，从而实现大数据的分布式计算
- 在分布式集群中调度执行MapReduce程序的计算框架也叫MapReduce
求部门工资总额
- SQL：select deptno,sum(sal) from emp gruop by deptno;
- 分析数据类型，套用模板重写map()、reduce()
- 导出jar包，指定main class
- 把数据保存在hdfs中
- hadoop jar s1.jar /input/emp.csv /output/0910/s1

技术分享图片

SalaryTotalMapper.java

View Code

SalaryTotalReducer.java

View Code

SalaryTotalMain.java

View Code

数据去重
- select distinct depno, job from emp;
- 用MR实现
- 只有Mapper没有Reducer（排序）

DistinctMain.java

View Code

DistinctMapper.java

View Code

DistinctReducer.java

View Code

多表查询
- select dname,ename from dept, emp where emp.deptno=dept.deptno;
- 关系型数据库的子查询会转换成多表查询（通过执行计划看SQL语句的执行过程和效率）
- 笛卡尔积：列数相加，行数相乘，得到全集
- 用连接条件（如 emp.deptno=dept.deptno）去掉全集中的错误数据
- 连接条件至少N-1个（N为表的个数）
- 根据连接条件不同，分为：
  - 等值连接 / 不等值连接
  - 外连接 / 自连接

技术分享图片

- 通过MR实现连接

EqualJoinMain.java

View Code

EqualJoinMapper.java

View Code

EqualJoinReducer.java

View Code

技术分享图片

自连接
- 通过表的别名，将同一张表视为多张表
- 一个人是下级的老板，同时是上级的员工
- 同一条数据输出两次，一次作为老板，一次作为员工（看作两张表）
- 相同的k2，其value2会被同一个reducer处理
- 存在非法数据需进行清洗（如把大老板的老板编号置为-1）
- 老板和员工同时存在才输出（员工树的根节点和叶子节点不显示）

技术分享图片

SelfJoinMain.java

View Code

SelfJoinMapper.java

View Code

SelfJoinReducer.java

View Code

技术分享图片

倒排索引
- 关系型数据库的索引

技术分享图片

- 数据存储在HDFS后会建立索引，提高查找效率

技术分享图片

MR实现倒排索引
- 记录一个单词在一个文件中出现的次数
- Combiner对同一文件中重复出现的单词进行求和
- Reducer对不同文件中出现的单词进行汇总
- 保证有无Combiner前后数据类型一样

技术分享图片

RevertedIndexMain.java

View Code

RevertedIndexMapper.java

View Code

RevertedIndexCombiner.java

View Code

RevertedIndexReducer.java

View Code

Hadoop自带例题
- start-all.sh
- /root/training/hadoop-2.7.3/share/hadoop/mapreduce
- hadoop jar hadoop-mapreduce-examples-2.7.3.jar
- wordcount: A map/reduce program that counts the words in the input files.
- hadoop jar hadoop-mapreduce-examples-2.7.3.jar wordcount
- Usage: wordcount <in> [<in>...] <out>
- hadoop jar hadoop-mapreduce-examples-2.7.3.jar wordcount /input/data.txt /output/0402/wc

[BD] MapReduce

原文：https://www.cnblogs.com/cxc1357/p/12594656.html

踩

(0)

评论一句话评论（0）

分享档案

更多>

2021年09月23日 (328)
2021年09月24日 (313)
2021年09月17日 (191)
2021年09月15日 (369)
2021年09月16日 (411)
2021年09月13日 (439)
2021年09月11日 (398)
2021年09月12日 (393)
2021年09月10日 (160)
2021年09月08日 (222)