Hadoop第7周练习—MapReduce进行数据查询和实现推简单荐系统

1  运行环境说明... 3

1.1 硬软件环境... 3

1.2 机器网络环境... 3

2  书面作业1:计算员工相关... 3

2.1 书面作业1内容... 3

2.2  实现过程... 4

2.2.1   准备测试数据... 4

2.2.2   问题1:求各个部门的总工资... 5

2.2.3   问题2:求各个部门的人数和平均工资... 11

2.2.4   问题3:求每个部门最早进入公司的员工姓名... 17

2.2.5   问题4:求各个城市的员工的总工资... 23

2.2.6   问题5:列出工资比上司高的员工姓名及其工资... 29

2.2.7   问题6:列出工资比公司平均工资要高的员工姓名及其工资... 34

2.2.8   问题7:列出名字以J开头的员工姓名及其所属部门名称... 40

2.2.9   问题8:列出工资最高的头三名员工姓名及其工资... 44

2.2.10问题9:将全体员工按照总收入(工资+提成)从高到低排列... 49

2.2.11问题10:求任何两名员工信息传递所需要经过的中间节点数... 53

3  书面作业2MapReduce实现推荐系统... 60

3.1 书面作业2内容... 60

3.2     程序代码... 60

3.2.1   CountThread.java. 60

3.2.2   Recommendation.java. 62

3.3 准备数据... 67

3.4 实现过程... 68

3.4.1   问题分析... 68

3.4.2   处理流程图... 69

3.4.3   配置运行参数... 69

3.4.4   运行作业... 70

 

1 运行环境说明

1.1  硬软件环境

l  主机操作系统:Windows 64 bit,双核4线程,主频2.2G6G内存

l  虚拟软件:VMware® Workstation 9.0.0 build-812388

l  虚拟机操作系统:CentOS 64位,单核,1G内存

l  JDK1.7.0_55 64 bit

l  Hadoop1.1.2

1.2  机器网络环境

集群包含三个节点:1namenode2datanode,其中节点之间可以相互ping通。节点IP地址和主机名分布如下:

序号

IP地址

机器名

类型

用户名

运行进程

1

10.88.147.221

hadoop1

名称节点

hadoop

NNSNNJobTracer

2

10.88.147.222

hadoop2

数据节点

hadoop

DNTaskTracer

3

10.88.147.223

hadoop3

数据节点

hadoop

DNTaskTracer

所有节点均是CentOS6.5 64bit系统,防火墙均禁用,所有节点上均创建了一个hadoop用户,用户主目录是/usr/hadoop。所有节点上均创建了一个目录/usr/local/hadoop,并且拥有者是hadoop用户。

2 书面作业1:计算员工相关

2.1  书面作业1内容

(本题102)把作业素材demo.txt中的两个表数据用适当的方式导入hadoop(来自Oracle数据库的样板表,可考虑分成2个文件存放,注意空值的处理)

书写Map-Reduce程序,求出以下结果

1) 求各个部门的总工资

2) 求各个部门的人数和平均工资

3) 求每个部门最早进入公司的员工姓名

4) 求各个城市的员工的总工资

5) 列出工资比上司高的员工姓名及其工资

6) 列出工资比公司平均工资要高的员工姓名及其工资

7) 列出名字以J开头的员工姓名及其所属部门名称

8) 列出工资最高的头三名员工姓名及其工资

9) 将全体员工按照总收入(工资+提成)从高到低排列,要求列出姓名及其总收入

10) 如果每位员工只能和他的直接上司,直接下属,同一部门的同事交流,求任何两名员工之间若要进行信息传递所需要经过的中间节点数。请评价一下这个问题是否适合使用map-reduce解决

2.2  实现过程

2.2.1准备测试数据

2.2.1.1拆分文件

把提供的测试数据第7-8周作业素材.txt按照要求拆分成两个文件dept(部门)和emp(员工),其中各字段用逗号分隔:

dept文件内容:

10,ACCOUNTING,NEW YORK

20,RESEARCH,DALLAS

30,SALES,CHICAGO

40,OPERATIONS,BOSTON

 

emp文件内容:

7369,SMITH,CLERK,7902,17-12-80,800,,20

7499,ALLEN,SALESMAN,7698,20-2-81,1600,300,30

7521,WARD,SALESMAN,7698,22-2-81,1250,500,30

7566,JONES,MANAGER,7839,02-4-81,2975,,20

7654,MARTIN,SALESMAN,7698,28-9-81,1250,1400,30

7698,BLAKE,MANAGER,7839,01-5-81,2850,,30

7782,CLARK,MANAGER,7839,09-6-81,2450,,10

7839,KING,PRESIDENT,,17-11-81,5000,,10

7844,TURNER,SALESMAN,7698,08-9-81,1500,0,30

7900,JAMES,CLERK,7698,03-12-81,950,,30

7902,FORD,ANALYST,7566,03-12-81,3000,,20

7934,MILLER,CLERK,7782,23-1-82,1300,,10

2.2.1.2上传测试文件

使用SSH工具(参见第122.1.3.1Linux文件传输工具所描述)把deptemp两个文件上传到本地目录/usr/local/hadoop-1.1.2/input中,然后使用eclipseHDFS插件工具上传该文件到/usr/hadoop/in目录中,如下图所示:

clip_image002

2.2.2问题1:求各个部门的总工资

2.2.2.1问题分析

MapReduce中的join分为好几种,比如有最常见的 reduce side joinmap side joinsemi join 等。reduce join shuffle阶段要进行大量的数据传输,会造成大量的网络IO效率低下,而map side join 在处理多个小表关联大表时非常有用 。

Map side join是针对以下场景进行的优化:两个待连接表中,有一个表非常大,而另一个表非常小,以至于小表可以直接存放到内存中。这样我们可以将小表复制多份,让每个map task内存中存在一份(比如存放到hash table中),然后只扫描大表:对于大表中的每一条记录key/value,在hash table中查找是否有相同的key的记录,如果有,则连接后输出即可。为了支持文件的复制,Hadoop提供了一个类DistributedCache,使用该类的方法如下:

1)用户使用静态方法DistributedCache.addCacheFile()指定要复制的文件,它的参数是文件的URI(如果是HDFS上的文件,可以这样:hdfs://jobtracker:50030/home/XXX/file)。JobTracker在作业启动之前会获取这个URI列表,并将相应的文件拷贝到各个TaskTracker的本地磁盘上。

2)用户使用DistributedCache.getLocalCacheFiles()方法获取文件目录,并使用标准的文件读写API读取相应的文件。

在下面代码中,将会把数据量小的表(部门dept )缓存在内存中,在Mapper阶段对员工部门编号映射成部门名称,该名称作为key输出到Reduce中,在Reduce中计算按照部门计算各个部门的总工资。

2.2.2.2处理流程图

clip_image004

2.2.2.3编写代码

Q1SumDeptSalary.java代码:

import java.io.BufferedReader;

import java.io.FileReader;

import java.io.IOException;

import java.util.HashMap;

import java.util.Map;

 

import org.apache.hadoop.conf.Configuration;

import org.apache.hadoop.conf.Configured;

import org.apache.hadoop.filecache.DistributedCache;

import org.apache.hadoop.fs.Path;

import org.apache.hadoop.io.LongWritable;

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.input.TextInputFormat;

import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;

import org.apache.hadoop.mapreduce.lib.output.TextOutputFormat;

import org.apache.hadoop.util.GenericOptionsParser;

import org.apache.hadoop.util.Tool;

import org.apache.hadoop.util.ToolRunner;

 

publicclass Q1SumDeptSalary extends Configured implements Tool {

 

    publicstaticclass MapClass extends Mapper<LongWritable, Text, Text, Text> {

 

        // 用于缓存 dept文件中的数据

        private Map<String, String> deptMap = new HashMap<String, String>();

        private String[] kv;

 

        // 此方法会在Map方法执行之前执行且执行一次

        @Override

        protectedvoid setup(Context context) throws IOException, InterruptedException {

            BufferedReader in = null;

            try {

 

                // 从当前作业中获取要缓存的文件

                Path[] paths = DistributedCache.getLocalCacheFiles(context.getConfiguration());

                String deptIdName = null;

                for (Path path : paths) {

 

                    // 对部门文件字段进行拆分并缓存到deptMap

                    if (path.toString().contains("dept")) {

                        in = new BufferedReader(new FileReader(path.toString()));

                        while (null != (deptIdName = in.readLine())) {

                           

                            // 对部门文件字段进行拆分并缓存到deptMap

                            // 其中Mapkey为部门编号,value为所在部门名称

                            deptMap.put(deptIdName.split(",")[0], deptIdName.split(",")[1]);

                        }

                    }

                }

            } catch (IOException e) {

                e.printStackTrace();

            } finally {

                try {

                    if (in != null) {

                        in.close();

                    }

                } catch (IOException e) {

                    e.printStackTrace();

                }

            }

        }

 

publicvoid map(LongWritable key, Text value, Context context) throws IOException, InterruptedException {

 

            // 对员工文件字段进行拆分

            kv = value.toString().split(",");

 

            // map join: map阶段过滤掉不需要的数据,输出key为部门名称和value为员工工资

            if (deptMap.containsKey(kv[7])) {

                if (null != kv[5] && !"".equals(kv[5].toString())) {

                    context.write(new Text(deptMap.get(kv[7].trim())), new Text(kv[5].trim()));

                }

            }

        }

    }

 

    publicstaticclass Reduce extends Reducer<Text, Text, Text, LongWritable> {

 

publicvoid reduce(Text key, Iterable<Text> values, Context context) throws IOException, InterruptedException {

 

            // 对同一部门的员工工资进行求和

            long sumSalary = 0;

            for (Text val : values) {

                sumSalary += Long.parseLong(val.toString());

            }

 

            // 输出key为部门名称和value为该部门员工工资总和

            context.write(key, new LongWritable(sumSalary));

        }

    }

 

    @Override

    publicint run(String[] args) throws Exception {

 

        // 实例化作业对象,设置作业名称、MapperReduce

        Job job = new Job(getConf(), "Q1SumDeptSalary");

        job.setJobName("Q1SumDeptSalary");

        job.setJarByClass(Q1SumDeptSalary.class);

        job.setMapperClass(MapClass.class);

        job.setReducerClass(Reduce.class);

 

        // 设置输入格式类

        job.setInputFormatClass(TextInputFormat.class);

 

        // 设置输出格式

        job.setOutputFormatClass(TextOutputFormat.class);

        job.setOutputKeyClass(Text.class);

        job.setOutputValueClass(Text.class);

 

        // 1个参数为缓存的部门数据路径、第2个参数为员工数据路径和第3个参数为输出路径

       String[] otherArgs = new GenericOptionsParser(job.getConfiguration(), args).getRemainingArgs();

       DistributedCache.addCacheFile(new Path(otherArgs[0]).toUri(), job.getConfiguration());

        FileInputFormat.addInputPath(job, new Path(otherArgs[1]));

        FileOutputFormat.setOutputPath(job, new Path(otherArgs[2]));

 

        job.waitForCompletion(true);

        return job.isSuccessful() ? 0 : 1;

    }

 

    /**

     * 主方法,执行入口

     * @param args 输入参数

     */

    publicstaticvoid main(String[] args) throws Exception {

        int res = ToolRunner.run(new Configuration(), new Q1SumDeptSalary(), args);

        System.exit(res);

    }

}

 

2.2.2.4配置运行参数

新建一个Java应用运行程序,需要在Arguments页签填写Q1SumDeptSalary运行的部门数据路径、员工数据路径和输出路径三个参数,需要注意的是hdfs的路径参数路径需要全路径,否则运行会报错:

l  部门数据路径:hdfs://hadoop1:9000/usr/hadoop/in/dept ,部门数据将缓存在各运行任务的节点内容中,可以提供处理的效率

l  员工数据路径:hdfs:// hadoop1:9000/usr/hadoop/in/emp

l  输出路径:hdfs:// hadoop1:9000/usr/hadoop/out/week7_q1

clip_image006

2.2.2.5运行并查看结果

设置运行参数完毕后,点击运行按钮:

clip_image008

运行成功后,刷新CentOS HDFS中的输出路径/usr/hadoop/out/week7_q1目录,打开part-r-00000文件,可以看到运行结果:

ACCOUNTING8750

RESEARCH6775

SALES  9400

clip_image010

2.2.3问题2:求各个部门的人数和平均工资

2.2.3.1问题分析

求各个部门的人数和平均工资,需要得到各部门工资总数和部门人数,通过两者相除获取各部门平均工资。首先和问题1类似在MapperSetup阶段缓存部门数据,然后在Mapper阶段抽取出部门编号和员工工资,利用缓存部门数据把部门编号对应为部门名称,接着在Shuffle阶段把传过来的数据处理为部门名称对应该部门所有员工工资的列表,最后在Reduce中按照部门归组,遍历部门所有员工,求出总数和员工数,输出部门名称和平均工资。

2.2.3.2处理流程图

clip_image012

2.2.3.3编写代码

Q2DeptNumberAveSalary.java代码:

import java.io.BufferedReader;

import java.io.FileReader;

import java.io.IOException;

import java.util.HashMap;

import java.util.Map;

 

import org.apache.hadoop.conf.Configuration;

import org.apache.hadoop.conf.Configured;

import org.apache.hadoop.filecache.DistributedCache;

import org.apache.hadoop.fs.Path;

import org.apache.hadoop.io.LongWritable;

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.input.TextInputFormat;

import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;

import org.apache.hadoop.mapreduce.lib.output.TextOutputFormat;

import org.apache.hadoop.util.GenericOptionsParser;

import org.apache.hadoop.util.Tool;

import org.apache.hadoop.util.ToolRunner;

 

publicclass Q2DeptNumberAveSalary extends Configured implements Tool {

 

    publicstaticclass MapClass extends Mapper<LongWritable, Text, Text, Text> {

 

        // 用于缓存 dept文件中的数据

        private Map<String, String> deptMap = new HashMap<String, String>();

        private String[] kv;

 

        // 此方法会在Map方法执行之前执行且执行一次

        @Override

        protectedvoid setup(Context context) throws IOException, InterruptedException {

            BufferedReader in = null;

            try {

                // 从当前作业中获取要缓存的文件

                Path[] paths = DistributedCache.getLocalCacheFiles(context.getConfiguration());

                String deptIdName = null;

                for (Path path : paths) {

 

                    // 对部门文件字段进行拆分并缓存到deptMap

                    if (path.toString().contains("dept")) {

                        in = new BufferedReader(new FileReader(path.toString()));

                        while (null != (deptIdName = in.readLine())) {

                           

                            // 对部门文件字段进行拆分并缓存到deptMap

                            // 其中Mapkey为部门编号,value为所在部门名称

                            deptMap.put(deptIdName.split(",")[0], deptIdName.split(",")[1]);

                        }

                    }

                }

            } catch (IOException e) {

                e.printStackTrace();

            } finally {

                try {

                    if (in != null) {

                        in.close();

                    }

                } catch (IOException e) {

                    e.printStackTrace();

                }

            }

        }

 

        publicvoid map(LongWritable key, Text value, Context context) throws IOException, InterruptedException {

 

            // 对员工文件字段进行拆分

            kv = value.toString().split(",");

 

            // map join: map阶段过滤掉不需要的数据,输出key为部门名称和value为员工工资

            if (deptMap.containsKey(kv[7])) {

                if (null != kv[5] && !"".equals(kv[5].toString())) {

                    context.write(new Text(deptMap.get(kv[7].trim())), new Text(kv[5].trim()));

                }

            }

        }

    }

 

    publicstaticclass Reduce extends Reducer<Text, Text, Text, Text> {

 

        publicvoid reduce(Text key, Iterable<Text> values, Context context) throws IOException, InterruptedException {

 

            long sumSalary = 0;

            int deptNumber = 0;

 

            // 对同一部门的员工工资进行求和

            for (Text val : values) {

                sumSalary += Long.parseLong(val.toString());

                deptNumber++;

            }

 

            // 输出key为部门名称和value为该部门员工工资平均值

            context.write(key, new Text("Dept Number:" + deptNumber + ", Ave Salary:" + sumSalary / deptNumber));

        }

    }

 

    @Override

    publicint run(String[] args) throws Exception {

 

        // 实例化作业对象,设置作业名称、MapperReduce

        Job job = new Job(getConf(), "Q2DeptNumberAveSalary");

        job.setJobName("Q2DeptNumberAveSalary");

        job.setJarByClass(Q2DeptNumberAveSalary.class);

        job.setMapperClass(MapClass.class);

        job.setReducerClass(Reduce.class);

 

        // 设置输入格式类

        job.setInputFormatClass(TextInputFormat.class);

 

        // 设置输出格式类

        job.setOutputFormatClass(TextOutputFormat.class);

        job.setOutputKeyClass(Text.class);

        job.setOutputValueClass(Text.class);

 

        // 1个参数为缓存的部门数据路径、第2个参数为员工数据路径和第3个参数为输出路径

        String[] otherArgs = new GenericOptionsParser(job.getConfiguration(), args).getRemainingArgs();

        DistributedCache.addCacheFile(new Path(otherArgs[0]).toUri(), job.getConfiguration());

        FileInputFormat.addInputPath(job, new Path(otherArgs[1]));

        FileOutputFormat.setOutputPath(job, new Path(otherArgs[2]));

 

        job.waitForCompletion(true);

        return job.isSuccessful() ? 0 : 1;

    }

 

    /**

     * 主方法,执行入口

     * @param args 输入参数

     */

    publicstaticvoid main(String[] args) throws Exception {

        int res = ToolRunner.run(new Configuration(), new Q2DeptNumberAveSalary(), args);

        System.exit(res);

    }

}

2.2.3.4配置运行参数

新建一个Java应用运行程序,需要在Arguments页签填写Q2DeptNumberAveSalary运行的部门数据路径、员工数据路径和输出路径三个参数,需要注意的是hdfs的路径参数路径需要全路径,否则运行会报错:

l  部门数据路径:hdfs://hadoop1:9000/usr/hadoop/in/dept ,部门数据将缓存在各运行任务的节点内容中,可以提供处理的效率

l  员工数据路径:hdfs:// hadoop1:9000/usr/hadoop/in/emp

l  输出路径:hdfs:// hadoop1:9000/usr/hadoop/out/week7_q2

clip_image014

2.2.3.5运行并查看结果

设置运行参数完毕后,点击运行按钮:

clip_image016

运行成功后,刷新CentOS HDFS中的输出路径/usr/hadoop/out/week7_q2目录,打开part-r-00000文件,可以看到运行结果:

ACCOUNTINGDept Number:3,Ave Salary:2916

RESEARCHDept Number:3,Ave Salary:2258

SALES  Dept Number:6,Ave Salary:1566

clip_image018

2.2.4问题3:求每个部门最早进入公司的员工姓名

2.2.4.1问题分析

求每个部门最早进入公司员工姓名,需要得到各部门所有员工的进入公司日期,通过比较获取最早进入公司员工姓名。首先和问题1类似在MapperSetup阶段缓存部门数据,然后Mapper阶段抽取出key为部门名称(利用缓存部门数据把部门编号对应为部门名称),value为员工姓名和进入公司日期,接着在Shuffle阶段把传过来的数据处理为部门名称对应该部门所有员工+进入公司日期的列表,最后在Reduce中按照部门归组,遍历部门所有员工,找出最早进入公司的员工并输出。

2.2.4.2处理流程图

clip_image020

2.2.4.3编写代码

import java.io.BufferedReader;

import java.io.FileReader;

import java.io.IOException;

import java.text.DateFormat;

import java.text.ParseException;

import java.text.SimpleDateFormat;

import java.util.Date;

import java.util.HashMap;

import java.util.Map;

 

import org.apache.hadoop.conf.Configuration;

import org.apache.hadoop.conf.Configured;

import org.apache.hadoop.filecache.DistributedCache;

import org.apache.hadoop.fs.Path;

import org.apache.hadoop.io.LongWritable;

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.input.TextInputFormat;

import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;

import org.apache.hadoop.mapreduce.lib.output.TextOutputFormat;

import org.apache.hadoop.util.GenericOptionsParser;

import org.apache.hadoop.util.Tool;

import org.apache.hadoop.util.ToolRunner;

 

publicclass Q3DeptEarliestEmp extends Configured implements Tool {

 

    publicstaticclass MapClass extends Mapper<LongWritable, Text, Text, Text> {

 

        // 用于缓存 dept文件中的数据

        private Map<String, String> deptMap = new HashMap<String, String>();

        private String[] kv;

 

        // 此方法会在Map方法执行之前执行且执行一次

        @Override

        protectedvoid setup(Context context) throws IOException, InterruptedException {

            BufferedReader in = null;

            try {

                // 从当前作业中获取要缓存的文件

                Path[] paths = DistributedCache.getLocalCacheFiles(context.getConfiguration());

                String deptIdName = null;

                for (Path path : paths) {

                    if (path.toString().contains("dept")) {

                        in = new BufferedReader(new FileReader(path.toString()));

                        while (null != (deptIdName = in.readLine())) {

 

                            // 对部门文件字段进行拆分并缓存到deptMap

                            // 其中Mapkey为部门编号,value为所在部门名称

                            deptMap.put(deptIdName.split(",")[0], deptIdName.split(",")[1]);

                        }

                    }

                }

            } catch (IOException e) {

                e.printStackTrace();

            } finally {

                try {

                    if (in != null) {

                        in.close();

                    }

                } catch (IOException e) {

                    e.printStackTrace();

                }

            }

        }

 

        publicvoid map(LongWritable key, Text value, Context context) throws IOException,     InterruptedException {

 

            // 对员工文件字段进行拆分

            kv = value.toString().split(",");

 

            // map join: map阶段过滤掉不需要的数据

            // 输出key为部门名称和value为员工姓名+","+员工进入公司日期

            if (deptMap.containsKey(kv[7])) {

                if (null != kv[4] && !"".equals(kv[4].toString())) {

                    context.write(new Text(deptMap.get(kv[7].trim())), new Text(kv[1].trim()                    + "," + kv[4].trim()));

                }

            }

        }

    }

 

    publicstaticclass Reduce extends Reducer<Text, Text, Text, Text> {

 

        publicvoid reduce(Text key, Iterable<Text> values, Context context) throws IOException,        InterruptedException {

 

            // 员工姓名和进入公司日期

            String empName = null;

            String empEnterDate = null;

 

            // 设置日期转换格式和最早进入公司的员工、日期

            DateFormat df = new SimpleDateFormat("dd-MM-yy");

 

            Date earliestDate = new Date();

            String earliestEmp = null;

 

            // 遍历该部门下所有员工,得到最早进入公司的员工信息

            for (Text val : values) {

                empName = val.toString().split(",")[0];

                empEnterDate = val.toString().split(",")[1].toString().trim();

                try {

                    System.out.println(df.parse(empEnterDate));

                    if (df.parse(empEnterDate).compareTo(earliestDate) < 0) {

                        earliestDate = df.parse(empEnterDate);

                        earliestEmp = empName;

                    }

                } catch (ParseException e) {

                    e.printStackTrace();

                }

            }

 

            // 输出key为部门名称和value为该部门最早进入公司员工

            context.write(key, new Text("The earliest emp of dept:" + earliestEmp + ", Enter            date:" + new SimpleDateFormat("yyyy-MM-dd").format(earliestDate)));

        }

    }

 

    @Override

    publicint run(String[] args) throws Exception {

 

        // 实例化作业对象,设置作业名称

        Job job = new Job(getConf(), "Q3DeptEarliestEmp");

        job.setJobName("Q3DeptEarliestEmp");

 

        // 设置MapperReduce

        job.setJarByClass(Q3DeptEarliestEmp.class);

        job.setMapperClass(MapClass.class);

        job.setReducerClass(Reduce.class);

 

        // 设置输入格式类

        job.setInputFormatClass(TextInputFormat.class);

 

        // 设置输出格式类

        job.setOutputFormatClass(TextOutputFormat.class);

        job.setOutputKeyClass(Text.class);

        job.setOutputValueClass(Text.class);

 

        // 1个参数为缓存的部门数据路径、第2个参数为员工数据路径和第三个参数为输出路径

        String[] otherArgs = new GenericOptionsParser(job.getConfiguration(), args).getRemainingArgs();

        DistributedCache.addCacheFile(new Path(otherArgs[0]).toUri(), job.getConfiguration());

        FileInputFormat.addInputPath(job, new Path(otherArgs[1]));

        FileOutputFormat.setOutputPath(job, new Path(otherArgs[2]));

 

        job.waitForCompletion(true);

        returnjob.isSuccessful() ? 0 : 1;

    }

 

    /**

     * 主方法,执行入口

     * @param args 输入参数

     */

    publicstaticvoid main(String[] args) throws Exception {

        intres = ToolRunner.run(new Configuration(), new Q3DeptEarliestEmp(), args);

        System.exit(res);

    }

}

2.2.4.4配置运行参数

新建一个Java应用运行程序,需要在Arguments页签填写Q3DeptEarliestEmp运行的部门数据路径、员工数据路径和输出路径三个参数,需要注意的是hdfs的路径参数路径需要全路径,否则运行会报错:

l  部门数据路径:hdfs://hadoop1:9000/usr/hadoop/in/dept ,部门数据将缓存在各运行任务的节点内容中,可以提供处理的效率

l  员工数据路径:hdfs:// hadoop1:9000/usr/hadoop/in/emp

l  输出路径:hdfs:// hadoop1:9000/usr/hadoop/out/week7_q3

clip_image022

2.2.4.5运行并查看结果

设置运行参数完毕后,点击运行按钮:

clip_image024

运行成功后,刷新CentOS HDFS中的输出路径/usr/hadoop/out/week7_q3目录,打开part-r-00000文件,可以看到运行结果:

ACCOUNTINGThe earliest emp of dept:CLARK, Enter date:1981-06-09

RESEARCHThe earliest emp of dept:SMITH, Enter date:1980-12-17

SALES  The earliest emp of dept:ALLEN, Enter date:1981-02-20

clip_image026

2.2.5问题4:求各个城市的员工的总工资

2.2.5.1问题分析

求各个城市员工的总工资,需要得到各个城市所有员工的工资,通过对各个城市所有员工工资求和得到总工资。首先和问题1类似在MapperSetup阶段缓存部门对应所在城市数据,然后在Mapper阶段抽取出key为城市名称(利用缓存数据把部门编号对应为所在城市名称),value为员工工资,接着在Shuffle阶段把传过来的数据处理为城市名称对应该城市所有员工工资,最后在Reduce中按照城市归组,遍历城市所有员工,求出工资总数并输出。

2.2.5.2处理流程图

clip_image028

2.2.5.3编写代码

import java.io.BufferedReader;

import java.io.FileReader;

import java.io.IOException;

import java.util.HashMap;

import java.util.Map;

 

import org.apache.hadoop.conf.Configuration;

import org.apache.hadoop.conf.Configured;

import org.apache.hadoop.filecache.DistributedCache;

import org.apache.hadoop.fs.Path;

import org.apache.hadoop.io.LongWritable;

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.input.TextInputFormat;

import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;

import org.apache.hadoop.mapreduce.lib.output.TextOutputFormat;

import org.apache.hadoop.util.GenericOptionsParser;

import org.apache.hadoop.util.Tool;

import org.apache.hadoop.util.ToolRunner;

 

publicclass Q4SumCitySalary extends Configured implements Tool {

 

    publicstaticclass MapClass extends Mapper<LongWritable, Text, Text, Text> {

 

        // 用于缓存 dept文件中的数据

        private Map<String, String> deptMap = new HashMap<String, String>();

        private String[] kv;

 

        // 此方法会在Map方法执行之前执行且执行一次

        @Override

        protectedvoid setup(Context context) throws IOException, InterruptedException {

            BufferedReader in = null;

            try {

                // 从当前作业中获取要缓存的文件

                Path[] paths = DistributedCache.getLocalCacheFiles(context.getConfiguration());

                String deptIdName = null;

                for (Path path : paths) {

                    if (path.toString().contains("dept")) {

                        in = new BufferedReader(new FileReader(path.toString()));

                        while (null != (deptIdName = in.readLine())) {

 

                            // 对部门文件字段进行拆分并缓存到deptMap

                            // 其中Mapkey为部门编号,value为所在城市名称

                            deptMap.put(deptIdName.split(",")[0], deptIdName.split(",")[2]);

                        }

                    }

                }

            } catch (IOException e) {

                e.printStackTrace();

            } finally {

                try {

                    if (in != null) {

                        in.close();

                    }

                } catch (IOException e) {

                    e.printStackTrace();

                }

            }

        }

 

        publicvoid map(LongWritable key, Text value, Context context) throws IOException,     InterruptedException {

 

            // 对员工文件字段进行拆分

            kv = value.toString().split(",");

 

            // map join: map阶段过滤掉不需要的数据,输出key为城市名称和value为员工工资

            if (deptMap.containsKey(kv[7])) {

                if (null != kv[5] && !"".equals(kv[5].toString())) {

                    context.write(new Text(deptMap.get(kv[7].trim())), new Text(kv[5].trim()));

                }

            }

        }

    }

 

    publicstaticclass Reduce extends Reducer<Text, Text, Text, LongWritable> {

 

        publicvoid reduce(Text key, Iterable<Text> values, Context context) throws IOException,        InterruptedException {

 

            // 对同一城市的员工工资进行求和

            long sumSalary = 0;

            for (Text val : values) {

                sumSalary += Long.parseLong(val.toString());

            }

 

            // 输出key为城市名称和value为该城市工资总和

            context.write(key, new LongWritable(sumSalary));

        }

    }

 

    @Override

    publicint run(String[] args) throws Exception {

 

        // 实例化作业对象,设置作业名称

        Job job = new Job(getConf(), "Q4SumCitySalary");

        job.setJobName("Q4SumCitySalary");

 

        // 设置MapperReduce

        job.setJarByClass(Q4SumCitySalary.class);

        job.setMapperClass(MapClass.class);

        job.setReducerClass(Reduce.class);

 

        // 设置输入格式类

        job.setInputFormatClass(TextInputFormat.class);

 

        // 设置输出格式类

        job.setOutputFormatClass(TextOutputFormat.class);

        job.setOutputKeyClass(Text.class);

        job.setOutputValueClass(Text.class);

 

        // 1个参数为缓存的部门数据路径、第2个参数为员工数据路径和第3个参数为输出路径

        String[] otherArgs = new GenericOptionsParser(job.getConfiguration(), args).getRemainingArgs();

        DistributedCache.addCacheFile(new Path(otherArgs[0]).toUri(), job.getConfiguration());

        FileInputFormat.addInputPath(job, new Path(otherArgs[1]));

        FileOutputFormat.setOutputPath(job, new Path(otherArgs[2]));

 

        job.waitForCompletion(true);

        return job.isSuccessful() ? 0 : 1;

    }

 

    /**

     * 主方法,执行入口

     * @param args 输入参数

     */

    publicstaticvoid main(String[] args) throws Exception {

        int res = ToolRunner.run(new Configuration(), new Q4SumCitySalary(), args);

        System.exit(res);

    }

}

2.2.5.4配置运行参数

新建一个Java应用运行程序,需要在Arguments页签填写Q4SumCitySalary运行的部门数据路径、员工数据路径和输出路径三个参数,需要注意的是hdfs的路径参数路径需要全路径,否则运行会报错:

l  部门数据路径:hdfs://hadoop1:9000/usr/hadoop/in/dept ,部门数据将缓存在各运行任务的节点内容中,可以提供处理的效率

l  员工数据路径:hdfs:// hadoop1:9000/usr/hadoop/in/emp

l  输出路径:hdfs:// hadoop1:9000/usr/hadoop/out/week7_q4

clip_image030

2.2.5.5运行并查看结果

设置运行参数完毕后,点击运行按钮:

clip_image032

运行成功后,刷新CentOS HDFS中的输出路径/usr/hadoop/out/week7_q4目录,打开part-r-00000文件,可以看到运行结果:

CHICAGO  9400

DALLAS     6775

NEW YORK     8750

clip_image034

2.2.6问题5:列出工资比上司高的员工姓名及其工资

2.2.6.1问题分析

求工资比上司高的员工姓名及工资,需要得到上司工资及上司所有下属员工,通过比较他们工资高低得到比上司工资高的员工。在Mapper阶段输出经理数据和员工对应经理表数据,其中经理数据key为员工编号、value"M,该员工工资",员工对应经理表数据key为经理编号、value"E,该员工姓名,该员工工资";然后在Shuffle阶段把传过来的经理数据和员工对应经理表数据进行归组,如编号为7698员工,value中标志M为自己工资,value中标志E为其下属姓名及工资;最后在Reduce中遍历比较员工与经理工资高低,输出工资高于经理的员工。

2.2.6.2处理流程图

clip_image036

2.2.6.3编写代码

import java.io.IOException;

import java.util.HashMap;

 

import org.apache.hadoop.conf.Configuration;

import org.apache.hadoop.conf.Configured;

import org.apache.hadoop.fs.Path;

import org.apache.hadoop.io.LongWritable;

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.input.TextInputFormat;

import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;

import org.apache.hadoop.mapreduce.lib.output.TextOutputFormat;

import org.apache.hadoop.util.GenericOptionsParser;

import org.apache.hadoop.util.Tool;

import org.apache.hadoop.util.ToolRunner;

 

publicclass Q5EarnMoreThanManager extends Configured implements Tool {

 

    publicstaticclass MapClass extends Mapper<LongWritable, Text, Text, Text> {

 

        publicvoid map(LongWritable key, Text value, Context context) throws IOException,     InterruptedException {

 

            // 对员工文件字段进行拆分

            String[] kv = value.toString().split(",");

 

            // 输出经理表数据,其中key为员工编号和valueM+该员工工资

            context.write(new Text(kv[0].toString()), new Text("M," + kv[5]));

 

            // 输出员工对应经理表数据,其中key为经理编号和value(E,该员工姓名,该员工工资)

            if (null != kv[3] && !"".equals(kv[3].toString())) {

                context.write(new Text(kv[3].toString()), new Text("E," + kv[1] + "," + kv[5]));

            }

        }

    }

 

    publicstaticclass Reduce extends Reducer<Text, Text, Text, Text> {

 

        publicvoid reduce(Text key, Iterable<Text> values, Context context) throws IOException,        InterruptedException {

 

            // 定义员工姓名、工资和存放部门员工Map

            String empName;

            long empSalary = 0;

            HashMap<String, Long> empMap = new HashMap<String, Long>();

           

            // 定义经理工资变量

            long mgrSalary = 0;

 

            for (Text val : values) {

                if (val.toString().startsWith("E")) {

                    // 当是员工标示时,获取该员工对应的姓名和工资并放入Map

                    empName = val.toString().split(",")[1];

                    empSalary = Long.parseLong(val.toString().split(",")[2]);

                    empMap.put(empName, empSalary);

                } else {

                    // 当时经理标志时,获取该经理工资

                    mgrSalary = Long.parseLong(val.toString().split(",")[1]);

                }

            }

 

            // 遍历该经理下属,比较员工与经理工资高低,输出工资高于经理的员工

            for (java.util.Map.Entry<String, Long> entry : empMap.entrySet()) {

                if (entry.getValue() > mgrSalary) {

                    context.write(new Text(entry.getKey()), new Text("" + entry.getValue()));

                }

            }

        }

    }

 

    @Override

    publicint run(String[] args) throws Exception {

 

        // 实例化作业对象,设置作业名称

        Job job = new Job(getConf(), "Q5EarnMoreThanManager");

        job.setJobName("Q5EarnMoreThanManager");

 

        // 设置MapperReduce

        job.setJarByClass(Q5EarnMoreThanManager.class);

        job.setMapperClass(MapClass.class);

        job.setReducerClass(Reduce.class);

 

        // 设置输入格式类

        job.setInputFormatClass(TextInputFormat.class);

 

        // 设置输出格式类

        job.setOutputFormatClass(TextOutputFormat.class);

        job.setOutputKeyClass(Text.class);

        job.setOutputValueClass(Text.class);

 

        // 1个参数为员工数据路径和第2个参数为输出路径

String[] otherArgs = new GenericOptionsParser(job.getConfiguration(), args).getRemainingArgs();

        FileInputFormat.addInputPath(job, new Path(otherArgs[0]));

        FileOutputFormat.setOutputPath(job, new Path(otherArgs[1]));

 

        job.waitForCompletion(true);

        return job.isSuccessful() ? 0 : 1;

    }

 

    /**

     * 主方法,执行入口

     * @param args 输入参数

     */

    publicstaticvoid main(String[] args) throws Exception {

        int res = ToolRunner.run(new Configuration(), new Q5EarnMoreThanManager(), args);

        System.exit(res);

    }

}

2.2.6.4配置运行参数

新建一个Java应用运行程序,需要在Arguments页签填写Q5EarnMoreThanManager运行的员工数据路径和输出路径两个参数,需要注意的是hdfs的路径参数路径需要全路径,否则运行会报错:

l  员工数据路径:hdfs:// hadoop1:9000/usr/hadoop/in/emp

l  输出路径:hdfs:// hadoop1:9000/usr/hadoop/out/week7_q5

clip_image038

2.2.6.5运行并查看结果

设置运行参数完毕后,点击运行按钮:

clip_image040

运行成功后,刷新CentOS HDFS中的输出路径/usr/hadoop/out/week7_q5目录,打开part-r-00000文件,可以看到运行结果:

FORD  3000

clip_image042

2.2.7问题6:列出工资比公司平均工资要高的员工姓名及其工资

2.2.7.1问题分析

求工资比公司平均工资要高的员工姓名及工资,需要得到公司的平均工资和所有员工工资,通过比较得出工资比平均工资高的员工姓名及工资。这个问题可以分两个作业进行解决,先求出公司的平均工资,然后与所有员工进行比较得到结果;也可以在一个作业进行解决,这里就得使用作业setNumReduceTasks方法,设置Reduce任务数为1,保证每次运行一个reduce任务,从而能先求出平均工资,然后进行比较得出结果。

Mapper阶段输出两份所有员工数据,其中一份key0value为该员工工资,另外一份key0value"该员工姓名 ,员工工资";然后在Shuffle阶段把传过来数据按照key进行归组,在该任务中有key值为01两组数据;最后在Reduce中对key0的所有员工求工资总数和员工数,获得平均工资;对key1,比较员工与平均工资的大小,输出比平均工资高的员工和对应的工资。

2.2.7.2处理流程图

clip_image044

2.2.7.3编写代码

import java.io.IOException;

import org.apache.hadoop.conf.Configuration;

import org.apache.hadoop.conf.Configured;

import org.apache.hadoop.fs.Path;

import org.apache.hadoop.io.IntWritable;

import org.apache.hadoop.io.LongWritable;

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.mapreduce.lib.output.TextOutputFormat;

import org.apache.hadoop.util.GenericOptionsParser;

import org.apache.hadoop.util.Tool;

import org.apache.hadoop.util.ToolRunner;

 

publicclass Q6HigherThanAveSalary extends Configured implements Tool {

 

    publicstaticclass MapClass extends Mapper<LongWritable, Text, IntWritable, Text> {

 

        publicvoid map(LongWritable key, Text value, Context context) throws IOException,     InterruptedException {

 

            // 对员工文件字段进行拆分

            String[] kv = value.toString().split(",");

 

            // 获取所有员工数据,其中key0value为该员工工资

            context.write(new IntWritable(0), new Text(kv[5]));

 

            // 获取所有员工数据,其中key0value(该员工姓名 ,员工工资)

            context.write(new IntWritable(1), new Text(kv[1] + "," + kv[5]));

        }

    }

 

    publicstaticclass Reduce extends Reducer<IntWritable, Text, Text, Text> {

 

        // 定义员工工资、员工数和平均工资

        privatelongallSalary = 0;

        privateintallEmpCount = 0;

        privatelongaveSalary = 0;

       

        // 定义员工工资变量

        privatelongempSalary = 0;

 

        publicvoid reduce(IntWritable key, Iterable<Text> values, Context context) throws     IOException, InterruptedException {

 

            for (Text val : values) {

                if (0 == key.get()) {

                    // 获取所有员工工资和员工数

                    allSalary += Long.parseLong(val.toString());

                    allEmpCount++;

                    System.out.println("allEmpCount = " + allEmpCount);

                } elseif (1 == key.get()) {

                    if (aveSalary == 0) {

                        aveSalary = allSalary / allEmpCount;

                        context.write(new Text("Average Salary = "), new Text("" + aveSalary));

                        context.write(new Text("Following employees have salarys higher than                        Average:"), new Text(""));

                    }

 

                    // 获取员工的平均工资

                    System.out.println("Employee salary = " + val.toString());

                    aveSalary = allSalary / allEmpCount;

                   

                    // 比较员工与平均工资的大小,输出比平均工资高的员工和对应的工资

                    empSalary = Long.parseLong(val.toString().split(",")[1]);

                    if (empSalary > aveSalary) {

                        context.write(new Text(val.toString().split(",")[0]), new Text("" +                         empSalary));

                    }

                }

            }

        }

    }

 

    @Override

    publicint run(String[] args) throws Exception {

 

        // 实例化作业对象,设置作业名称

        Job job = new Job(getConf(), "Q6HigherThanAveSalary");

        job.setJobName("Q6HigherThanAveSalary");

 

        // 设置MapperReduce

        job.setJarByClass(Q6HigherThanAveSalary.class);

        job.setMapperClass(MapClass.class);

        job.setReducerClass(Reduce.class);

 

        // 必须设置Reduce任务数为1 # -D mapred.reduce.tasks = 1

        // 这是该作业设置的核心,这样才能够保证各reduce是串行的

        job.setNumReduceTasks(1);

 

        // 设置输出格式类

        job.setMapOutputKeyClass(IntWritable.class);

        job.setMapOutputValueClass(Text.class);

 

        // 设置输出键和值类型

        job.setOutputFormatClass(TextOutputFormat.class);

        job.setOutputKeyClass(Text.class);

        job.setOutputValueClass(LongWritable.class);

 

        // 1个参数为员工数据路径和第2个参数为输出路径

String[] otherArgs = new GenericOptionsParser(job.getConfiguration(), args).getRemainingArgs();

        FileInputFormat.addInputPath(job, new Path(otherArgs[0]));

        FileOutputFormat.setOutputPath(job, new Path(otherArgs[1]));

 

        job.waitForCompletion(true);

        return job.isSuccessful() ? 0 : 1;

    }

 

    /**

     * 主方法,执行入口

     * @param args 输入参数

     */

    publicstaticvoid main(String[] args) throws Exception {

        int res = ToolRunner.run(new Configuration(), new Q6HigherThanAveSalary(), args);

        System.exit(res);

    }

}

2.2.7.4配置运行参数

新建一个Java应用运行程序,需要在Arguments页签填写Q6HigherThanAveSalary运行的员工数据路径和输出路径两个参数,需要注意的是hdfs的路径参数路径需要全路径,否则运行会报错:

l  员工数据路径:hdfs:// hadoop1:9000/usr/hadoop/in/emp

l  输出路径:hdfs:// hadoop1:9000/usr/hadoop/out/week7_q6

clip_image046

2.2.7.5运行并查看结果

设置运行参数完毕后,点击运行按钮:

clip_image048

运行成功后,刷新CentOS HDFS中的输出路径/usr/hadoop/out/week7_q6目录,打开part-r-00000文件,可以看到运行结果:

Average Salary = 2077

Following employees have salarys higher than Average:    

FORD  3000

CLARK2450

KING   5000

JONES2975

BLAKE  2850

clip_image050

2.2.8问题7:列出名字以J开头的员工姓名及其所属部门名称

2.2.8.1问题分析

求名字以J开头的员工姓名机器所属部门名称,只需判断员工姓名是否以J开头。首先和问题1类似在MapperSetup阶段缓存部门数据,然后在Mapper阶段判断员工姓名是否以J开头,如果是抽取出员工姓名和员工所在部门编号,利用缓存部门数据把部门编号对应为部门名称,转换后输出结果。

2.2.8.2处理流程图

clip_image052

2.2.8.3编写代码

import java.io.BufferedReader;

import java.io.FileReader;

import java.io.IOException;

import java.util.HashMap;

import java.util.Map;

 

import org.apache.hadoop.conf.Configuration;

import org.apache.hadoop.conf.Configured;

import org.apache.hadoop.filecache.DistributedCache;

import org.apache.hadoop.fs.Path;

import org.apache.hadoop.io.LongWritable;

import org.apache.hadoop.io.Text;

import org.apache.hadoop.mapreduce.Job;

import org.apache.hadoop.mapreduce.Mapper;

import org.apache.hadoop.mapreduce.lib.input.FileInputFormat;

import org.apache.hadoop.mapreduce.lib.input.TextInputFormat;

import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;

import org.apache.hadoop.mapreduce.lib.output.TextOutputFormat;

import org.apache.hadoop.util.GenericOptionsParser;

import org.apache.hadoop.util.Tool;

import org.apache.hadoop.util.ToolRunner;

 

publicclass Q7NameDeptOfStartJ extends Configured implements Tool {

 

    publicstaticclass MapClass extends Mapper<LongWritable, Text, Text, Text> {

 

        // 用于缓存 dept文件中的数据

        private Map<String, String> deptMap = new HashMap<String, String>();

        private String[] kv;

 

        // 此方法会在Map方法执行之前执行且执行一次

        @Override

        protectedvoid setup(Context context) throws IOException, InterruptedException {

            BufferedReader in = null;

            try {

 

                // 从当前作业中获取要缓存的文件

                Path[] paths = DistributedCache.getLocalCacheFiles(context.getConfiguration());

                String deptIdName = null;

                for (Path path : paths) {

 

                    // 对部门文件字段进行拆分并缓存到deptMap

                    if (path.toString().contains("dept")) {

                        in = new BufferedReader(new FileReader(path.toString()));

                        while (null != (deptIdName = in.readLine())) {

                           

                            // 对部门文件字段进行拆分并缓存到deptMap

                            // 其中Mapkey为部门编号,value为所在部门名称

                            deptMap.put(deptIdName.split(",")[0], deptIdName.split(",")[1]);

                        }

                    }

                }

            } catch (IOException e) {

                e.printStackTrace();

            } finally {

                try {

                    if (in != null) {

                        in.close();

                    }

                } catch (IOException e) {

                    e.printStackTrace();

                }

            }

        }

 

        publicvoid map(LongWritable key, Text value, Context context) throws IOException,     InterruptedException {

 

            // 对员工文件字段进行拆分

            kv = value.toString().split(",");

 

            // 输出员工姓名为J开头的员工信息,key为员工姓名和value为员工所在部门名称

            if (kv[1].toString().trim().startsWith("J")) {

                context.write(new Text(kv[1].trim()), new Text(deptMap.get(kv[7].trim())));

            }

        }

    }

 

    @Override

    publicint run(String[] args) throws Exception {

 

        // 实例化作业对象,设置作业名称

        Job job = new Job(getConf(), "Q7NameDeptOfStartJ");

        job.setJobName("Q7NameDeptOfStartJ");

 

        // 设置MapperReduce

        job.setJarByClass(Q7NameDeptOfStartJ.class);

        job.setMapperClass(MapClass.class);

 

        // 设置输入格式类

        job.setInputFormatClass(TextInputFormat.class);

 

        // 设置输出格式类

        job.setOutputFormatClass(TextOutputFormat.class);

        job.setOutputKeyClass(Text.class);

        job.setOutputValueClass(Text.class);

 

        // 1个参数为缓存的部门数据路径、第2个参数为员工数据路径和第3个参数为输出路径

String[] otherArgs = new GenericOptionsParser(job.getConfiguration(), args).getRemainingArgs();

        DistributedCache.addCacheFile(new Path(otherArgs[0]).toUri(), job.getConfiguration());

        FileInputFormat.addInputPath(job, new Path(otherArgs[1]));

        FileOutputFormat.setOutputPath(job, new Path(otherArgs[2]));

 

        job.waitForCompletion(true);

        return job.isSuccessful() ? 0 : 1;

    }

 

    /**

     * 主方法,执行入口

     * @param args 输入参数

     */

    publicstaticvoid main(String[] args) throws Exception {

        int res = ToolRunner.run(new Configuration(), new Q7NameDeptOfStartJ(), args);

        System.exit(res);

    }

}

2.2.8.4配置运行参数

新建一个Java应用运行程序,需要在Arguments页签填写Q7NameDeptOfStartJ运行的部门数据路径、员工数据路径和输出路径三个参数,需要注意的是hdfs的路径参数路径需要全路径,否则运行会报错:

l  部门数据路径:hdfs://hadoop1:9000/usr/hadoop/in/dept ,部门数据将缓存在各运行任务的节点内容中,可以提供处理的效率

l  员工数据路径:hdfs:// hadoop1:9000/usr/hadoop/in/emp

l  输出路径:hdfs:// hadoop1:9000/usr/hadoop/out/week7_q7

clip_image054

2.2.8.5运行并查看结果

设置运行参数完毕后,点击运行按钮:

clip_image056

运行成功后,刷新CentOS HDFS中的输出路径/usr/hadoop/out/week7_q7目录,打开part-r-00000文件,可以看到运行结果:

JAMESSALES

JONESRESEARCH

clip_image058

2.2.9问题8:列出工资最高的头三名员工姓名及其工资

2.2.9.1问题分析

求工资最高的头三名员工姓名及工资,可以通过冒泡法得到。在Mapper阶段输出经理数据和员工对应经理表数据,其中经理数据key0值、value"员工姓名,员工工资";最后在Reduce中通过冒泡法遍历所有员工,比较员工工资多少,求出前三名。

2.2.9.2处理流程图

clip_image060

2.2.9.3编写代码

import java.io.IOException;

 

import org.apache.hadoop.conf.Configuration;

import org.apache.hadoop.conf.Configured;

import org.apache.hadoop.fs.Path;

import org.apache.hadoop.io.IntWritable;

import org.apache.hadoop.io.LongWritable;

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.input.TextInputFormat;

import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;

import org.apache.hadoop.mapreduce.lib.output.TextOutputFormat;

import org.apache.hadoop.util.GenericOptionsParser;

import org.apache.hadoop.util.Tool;

import org.apache.hadoop.util.ToolRunner;

 

publicclass Q8SalaryTop3Salary extends Configured implements Tool {

 

    publicstaticclass MapClass extends Mapper<LongWritable, Text, IntWritable, Text> {

 

        publicvoid map(LongWritable key, Text value, Context context) throws IOException,     InterruptedException {

 

            // 对员工文件字段进行拆分

            String[] kv = value.toString().split(",");

 

            // 输出key0value为员工姓名+","+员工工资

            context.write(new IntWritable(0), new Text(kv[1].trim() + "," + kv[5].trim()));

        }

    }

 

    publicstaticclass Reduce extends Reducer<IntWritable, Text, Text, Text> {

 

        publicvoid reduce(IntWritable key, Iterable<Text> values, Context context) throws     IOException, InterruptedException {

 

            // 定义工资前三员工姓名

            String empName;

            String firstEmpName = "";

            String secondEmpName = "";

            String thirdEmpName = "";

           

            // 定义工资前三工资

            long empSalary = 0;

            long firstEmpSalary = 0;

            long secondEmpSalary = 0;

            long thirdEmpSalary = 0;

 

            // 通过冒泡法遍历所有员工,比较员工工资多少,求出前三名

            for (Text val : values) {

                empName = val.toString().split(",")[0];

                empSalary = Long.parseLong(val.toString().split(",")[1]);

               

                if(empSalary > firstEmpSalary) {

                    thirdEmpName = secondEmpName;

                    thirdEmpSalary = secondEmpSalary;

                    secondEmpName = firstEmpName;

                    secondEmpSalary = firstEmpSalary;

                    firstEmpName = empName;

                    firstEmpSalary = empSalary;

                } elseif (empSalary > secondEmpSalary) {

                    thirdEmpName = secondEmpName;

                    thirdEmpSalary = secondEmpSalary;

                    secondEmpName = empName;

                    secondEmpSalary = empSalary;

                } elseif (empSalary > thirdEmpSalary) {

                    thirdEmpName = empName;

                    thirdEmpSalary = empSalary;

                }

            }

           

            // 输出工资前三名信息

            context.write(new Text( "First employee name:" + firstEmpName), new Text("Salary:"          + firstEmpSalary));

            context.write(new Text( "Second employee name:" + secondEmpName), new                   Text("Salary:" + secondEmpSalary));

            context.write(new Text( "Third employee name:" + thirdEmpName), new Text("Salary:"          + thirdEmpSalary));

        }

    }

 

    @Override

    publicint run(String[] args) throws Exception {

 

        // 实例化作业对象,设置作业名称

        Job job = new Job(getConf(), "Q8SalaryTop3Salary");

        job.setJobName("Q8SalaryTop3Salary");

 

        // 设置MapperReduce

        job.setJarByClass(Q8SalaryTop3Salary.class);

        job.setMapperClass(MapClass.class);

        job.setReducerClass(Reduce.class);

        job.setMapOutputKeyClass(IntWritable.class);

        job.setMapOutputValueClass(Text.class);

 

        // 设置输入格式类

        job.setInputFormatClass(TextInputFormat.class);

 

        // 设置输出格式类

        job.setOutputKeyClass(Text.class);

        job.setOutputFormatClass(TextOutputFormat.class);

        job.setOutputValueClass(Text.class);

 

        // 1个参数为员工数据路径和第2个参数为输出路径

        String[] otherArgs = new GenericOptionsParser(job.getConfiguration(),                   args).getRemainingArgs();

        FileInputFormat.addInputPath(job, new Path(otherArgs[0]));

        FileOutputFormat.setOutputPath(job, new Path(otherArgs[1]));

 

        job.waitForCompletion(true);

        return job.isSuccessful() ? 0 : 1;

    }

 

    /**

     * 主方法,执行入口

     * @param args 输入参数

     */

    publicstaticvoid main(String[] args) throws Exception {

        int res = ToolRunner.run(new Configuration(), new Q8SalaryTop3Salary(), args);

        System.exit(res);

    }

}

2.2.9.4配置运行参数

新建一个Java应用运行程序,需要在Arguments页签填写Q8SalaryTop3Salary运行的员工数据路径和输出路径两个参数,需要注意的是hdfs的路径参数路径需要全路径,否则运行会报错:

l  员工数据路径:hdfs:// hadoop1:9000/usr/hadoop/in/emp

l  输出路径:hdfs:// hadoop1:9000/usr/hadoop/out/week7_q8

clip_image062

2.2.9.5运行并查看结果

设置运行参数完毕后,点击运行按钮:

clip_image064

运行成功后,刷新CentOS HDFS中的输出路径/usr/hadoop/out/week7_q8目录,打开part-r-00000文件,可以看到运行结果:

First employee name:KING    Salary:5000

Second employee name:FORD    Salary:3000

Third employee name:JONESSalary:2975

clip_image066

2.2.10   问题9:将全体员工按照总收入(工资+提成)从高到低排列

2.2.10.1问题分析

求全体员工总收入降序排列,获得所有员工总收入并降序排列即可。在Mapper阶段输出所有员工总工资数据,其中key为员工总工资、value为员工姓名,在Mapper阶段的最后会先调用job.setPartitionerClass对数据进行分区,每个分区映射到一个reducer,每个分区内又调用job.setSortComparatorClass设置的key比较函数类排序。由于在本作业中Mapkey只有0值,故能实现对所有数据进行排序。

2.2.10.2处理流程图

clip_image068

2.2.10.3编写代码

import java.io.IOException;

 

import org.apache.hadoop.conf.Configuration;

import org.apache.hadoop.conf.Configured;

import org.apache.hadoop.fs.Path;

import org.apache.hadoop.io.IntWritable;

import org.apache.hadoop.io.LongWritable;

import org.apache.hadoop.io.Text;

import org.apache.hadoop.io.WritableComparable;

import org.apache.hadoop.mapreduce.Job;

import org.apache.hadoop.mapreduce.Mapper;

import org.apache.hadoop.mapreduce.lib.input.FileInputFormat;

import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;

import org.apache.hadoop.util.GenericOptionsParser;

import org.apache.hadoop.util.Tool;

import org.apache.hadoop.util.ToolRunner;

 

publicclass Q9EmpSalarySort extends Configured implements Tool {

 

    publicstaticclass MapClass extends Mapper<LongWritable, Text, IntWritable, Text> {

 

        publicvoid map(LongWritable key, Text value, Context context) throws IOException,     InterruptedException {

 

            // 对员工文件字段进行拆分

            String[] kv = value.toString().split(",");

 

            // 输出key为员工所有工资和value为员工姓名

            int empAllSalary = "".equals(kv[6]) ? Integer.parseInt(kv[5]) :                             Integer.parseInt(kv[5]) + Integer.parseInt(kv[6]);

            context.write(new IntWritable(empAllSalary), new Text(kv[1]));

        }

    }

 

    /**

     * 递减排序算法

     */

    publicstaticclass DecreaseComparator extends IntWritable.Comparator {

        publicint compare(WritableComparable a, WritableComparable b) {

            return -super.compare(a, b);

        }

 

        publicint compare(byte[] b1, int s1, int l1, byte[] b2, int s2, int l2) {

            return -super.compare(b1, s1, l1, b2, s2, l2);

        }

    }

 

    @Override

    publicint run(String[] args) throws Exception {

 

        // 实例化作业对象,设置作业名称

        Job job = new Job(getConf(), "Q9EmpSalarySort");

        job.setJobName("Q9EmpSalarySort");

 

        // 设置MapperReduce

        job.setJarByClass(Q9EmpSalarySort.class);

        job.setMapperClass(MapClass.class);

 

        // 设置输出格式类

        job.setMapOutputKeyClass(IntWritable.class);

        job.setMapOutputValueClass(Text.class);

        job.setSortComparatorClass(DecreaseComparator.class);

 

        // 1个参数为员工数据路径和第2个参数为输出路径

        String[] otherArgs = new GenericOptionsParser(job.getConfiguration(),                   args).getRemainingArgs();

        FileInputFormat.addInputPath(job, new Path(otherArgs[0]));

        FileOutputFormat.setOutputPath(job, new Path(otherArgs[1]));

 

        job.waitForCompletion(true);

        return job.isSuccessful() ? 0 : 1;

    }

 

    /**

     * 主方法,执行入口

     * @param args 输入参数

     */

    publicstaticvoid main(String[] args) throws Exception {

        int res = ToolRunner.run(new Configuration(), new Q9EmpSalarySort(), args);

        System.exit(res);

    }

}

2.2.10.4配置运行参数

新建一个Java应用运行程序,需要在Arguments页签填写Q9EmpSalarySort运行的员工数据路径和输出路径两个参数,需要注意的是hdfs的路径参数路径需要全路径,否则运行会报错:

l  员工数据路径:hdfs:// hadoop1:9000/usr/hadoop/in/emp

l  输出路径:hdfs:// hadoop1:9000/usr/hadoop/out/week7_q9

clip_image070

2.2.10.5运行并查看结果

设置运行参数完毕后,点击运行按钮:

clip_image072

运行成功后,刷新CentOS HDFS中的输出路径/usr/hadoop/out/week7_q9目录,打开part-r-00000文件,可以看到运行结果:

5000    KING

3000    FORD

2975    JONES

2850    BLAKE

......

clip_image074

2.2.11   问题10:求任何两名员工信息传递所需要经过的中间节点数

2.2.11.1问题分析

该公司所有员工可以形成入下图的树形结构,求两个员工的沟通的中间节点数,可转换在员工树中求两个节点连通所经过的节点数,即从其中一节点到汇合节点经过节点数加上另一节点到汇合节点经过节点数。例如求MQ所需节点数,可以先找出MA经过的节点数,然后找出QA经过的节点数,两者相加得到MQ所需节点数。

clip_image076

在作业中首先在Mapper阶段所有员工数据,其中经理数据key0值、value"员工编号,员工经理编号",然后在Reduce阶段把所有员工放到员工列表和员工对应经理链表Map中,最后在ReduceCleanup中按照上面说所算法对任意两个员工计算出沟通的路径长度并输出。

2.2.11.2处理流程图

clip_image078

2.2.11.3编写代码

import java.io.IOException;

import java.util.ArrayList;

import java.util.HashMap;

import java.util.List;

import java.util.Map;

 

import org.apache.hadoop.conf.Configuration;

import org.apache.hadoop.conf.Configured;

import org.apache.hadoop.fs.Path;

import org.apache.hadoop.io.IntWritable;

import org.apache.hadoop.io.LongWritable;

import org.apache.hadoop.io.NullWritable;

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.mapreduce.lib.output.TextOutputFormat;

import org.apache.hadoop.util.GenericOptionsParser;

import org.apache.hadoop.util.Tool;

import org.apache.hadoop.util.ToolRunner;

 

publicclass Q10MiddlePersonsCountForComm extends Configured implements Tool {

 

    publicstaticclass MapClass extends Mapper<LongWritable, Text, IntWritable, Text> {

 

        publicvoid map(LongWritable key, Text value, Context context) throws IOException,     InterruptedException {

 

            // 对员工文件字段进行拆分

            String[] kv = value.toString().split(",");

 

            // 输出key0value为员工编号+","+员工经理编号

            context.write(new IntWritable(0), new Text(kv[0] + "," + ("".equals(kv[3]) ? " " : kv[3])));

        }

    }

 

    publicstaticclass Reduce extends Reducer<IntWritable, Text, NullWritable, Text> {

 

        // 定义员工列表和员工对应经理Map

        List<String> employeeList = new ArrayList<String>();

        Map<String, String> employeeToManagerMap = new HashMap<String, String>();

 

        publicvoid reduce(IntWritable key, Iterable<Text> values, Context context) throws     IOException, InterruptedException {

 

            // reduce阶段把所有员工放到员工列表和员工对应经理Map

            for (Text value : values) {

                employeeList.add(value.toString().split(",")[0].trim());

                employeeToManagerMap.put(value.toString().split(",")[0].trim(),                             value.toString().split(",")[1].trim());

            }

        }

 

        @Override

        protectedvoid cleanup(Context context) throws IOException, InterruptedException {

            int totalEmployee = employeeList.size();

            int i, j;

            int distance;

            System.out.println(employeeList);

            System.out.println(employeeToManagerMap);

 

            // 对任意两个员工计算出沟通的路径长度并输出

            for (i = 0; i < (totalEmployee - 1); i++) {

                for (j = (i + 1); j < totalEmployee; j++) {

                    distance = calculateDistance(i, j);

                    String value = employeeList.get(i) + " and " + employeeList.get(j) + " =                    " + distance;

                    context.write(NullWritable.get(), new Text(value));

                }

            }

        }

 

        /**

         * 该公司可以由所有员工形成树形结构,求两个员工的沟通的中间节点数,可以转换在员工树中两员工之间的距离

         * 由于在树中任意两点都会在某上级节点汇合,根据该情况设计了如下算法

         */

        privateint calculateDistance(int i, int j) {

            String employeeA = employeeList.get(i);

            String employeeB = employeeList.get(j);

            int distance = 0;

 

            // 如果AB的经理,反之亦然

            if (employeeToManagerMap.get(employeeA).equals(employeeB) ||                                    employeeToManagerMap.get(employeeB).equals(employeeA)) {

                distance = 0;

            }

            // AB在同一经理下

            elseif  (employeeToManagerMap.get(employeeA).equals(

                    employeeToManagerMap.get(employeeB))) {

                distance = 0;

            } else {

                // 定义AB对应经理链表

                List<String> employeeA_ManagerList = new ArrayList<String>();

                List<String> employeeB_ManagerList = new ArrayList<String>();

 

                // 获取从A开始经理链表

                employeeA_ManagerList.add(employeeA);

                String current = employeeA;

                while (false == employeeToManagerMap.get(current).isEmpty()) {

                    current = employeeToManagerMap.get(current);

                    employeeA_ManagerList.add(current);

                }

 

                // 获取从B开始经理链表

                employeeB_ManagerList.add(employeeB);

                current = employeeB;

                while (false == employeeToManagerMap.get(current).isEmpty()) {

                    current = employeeToManagerMap.get(current);

                    employeeB_ManagerList.add(current);

                }

 

                int ii = 0, jj = 0;

                String currentA_manager, currentB_manager;

                boolean found = false;

 

                // 遍历AB开始经理链表,找出汇合点计算

                for (ii = 0; ii < employeeA_ManagerList.size(); ii++) {

                    currentA_manager = employeeA_ManagerList.get(ii);

                    for (jj = 0; jj < employeeB_ManagerList.size(); jj++) {

                        currentB_manager = employeeB_ManagerList.get(jj);

                        if (currentA_manager.equals(currentB_manager)) {

                            found = true;

                            break;

                        }

                    }

 

                    if (found) {

                        break;

                    }

                }

 

                // 最后获取两只之前的路径

                distance = ii + jj - 1;

            }

 

            return distance;

        }

    }

 

    @Override

    publicint run(String[] args) throws Exception {

 

        // 实例化作业对象,设置作业名称

        Job job = new Job(getConf(), "Q10MiddlePersonsCountForComm");

        job.setJobName("Q10MiddlePersonsCountForComm");

 

        // 设置MapperReduce

        job.setJarByClass(Q10MiddlePersonsCountForComm.class);

        job.setMapperClass(MapClass.class);

        job.setReducerClass(Reduce.class);

 

        // 设置Mapper输出格式类

        job.setMapOutputKeyClass(IntWritable.class);

        job.setMapOutputValueClass(Text.class);

 

        // 设置Reduce输出键和值类型

        job.setOutputFormatClass(TextOutputFormat.class);

        job.setOutputKeyClass(NullWritable.class);

        job.setOutputValueClass(Text.class);

 

        // 1个参数为员工数据路径和第2个参数为输出路径

        String[] otherArgs = new GenericOptionsParser(job.getConfiguration(),                   args).getRemainingArgs();

        FileInputFormat.addInputPath(job, new Path(otherArgs[0]));

        FileOutputFormat.setOutputPath(job, new Path(otherArgs[1]));

 

        job.waitForCompletion(true);

        return job.isSuccessful() ? 0 : 1;

    }

 

    /**

     * 主方法,执行入口

     * @param args 输入参数

     */

    publicstaticvoid main(String[] args) throws Exception {

        int res = ToolRunner.run(new Configuration(), new Q10MiddlePersonsCountForComm(), args);

        System.exit(res);

    }

}

 

 

2.2.11.4配置运行参数

新建一个Java应用运行程序,需要在Arguments页签填写Q10MiddlePersonsCountForComm运行的员工数据路径和输出路径两个参数,需要注意的是hdfs的路径参数路径需要全路径,否则运行会报错:

l  员工数据路径:hdfs:// hadoop1:9000/usr/hadoop/in/emp

l  输出路径:hdfs:// hadoop1:9000/usr/hadoop/out/week7_q10

clip_image080

2.2.11.5运行并查看结果

设置运行参数完毕后,点击运行按钮:

clip_image082

运行成功后,刷新CentOS HDFS中的输出路径/usr/hadoop/out/week7_q10目录,打开part-r-00000文件,可以看到运行结果:

7369 and 7499 = 4

7369 and 7521 = 4

7369 and 7566 = 1

7369 and 7654 = 4

7369 and 7698 = 3

......

clip_image084

 

3 书面作业2MapReduce实现推荐系统

3.1  书面作业2内容

(本题可选)参考http://f.dataguru.cn/forum.php?mod=viewthread&tid=84

使用Map-Reduce实现该场景

3.2  程序代码

3.2.1CountThread.java

import java.io.DataInput;

import java.io.DataOutput;

import java.io.IOException;

 

import org.apache.hadoop.io.IntWritable;

import org.apache.hadoop.io.Text;

import org.apache.hadoop.io.WritableComparable;

 

publicclass CountThread implements WritableComparable {

    Text threadId;

    IntWritable cnt;

 

    CountThread(Text id, IntWritable t) {

        this.threadId = id;

        this.cnt = t;

    }

 

    CountThread() {

        this.threadId = new Text();

        this.cnt = new IntWritable();

    }

 

    @Override

    publicvoid readFields(DataInput in) throws IOException {

        threadId.readFields(in);

        cnt.readFields(in);

    }

 

    @Override

    publicvoid write(DataOutput out) throws IOException {

        threadId.write(out);

        cnt.write(out);

    }

 

    // cnt值倒序排列

    @Override

    publicint compareTo(Object o) {

        return ((CountThread) o).cnt.compareTo(cnt) == 0 ? threadId.compareTo(((CountThread)        o).threadId) : ((CountThread) o).cnt.compareTo(cnt);

    }

 

    publicboolean equals(Object o) {

        if (!(o instanceof CountThread))

            returnfalse;

        CountThread ct = (CountThread) o;

        returnthreadId.equals(ct.threadId) && cnt.equals(ct.cnt);

    }

 

    // threadId值来生成hashCode,默认是按此值来分区的

    // 保证同一个threadId在一个partiotion

    publicint hashCode() {

        returnthreadId.hashCode();

    }

 

    // 默认输出格式

    public String toString() {

        StringBuffer buf = new StringBuffer("");

        buf.append(threadId.toString());

        buf.append("\t");

        buf.append(cnt.toString());

        return buf.toString();

    }

 

    public Text getThreadId() {

        returnthreadId;

    }

 

    publicvoid setThreadId(Text threadId) {

        this.threadId = threadId;

    }

 

    public IntWritable getCnt() {

        returncnt;

    }

 

    publicvoid setCnt(IntWritable cnt) {

        this.cnt = cnt;

    }

}

 

 

3.2.2Recommendation.java

import java.io.IOException;

import java.util.HashSet;

import java.util.Set;

 

import org.apache.hadoop.conf.Configuration;

import org.apache.hadoop.conf.Configured;

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.input.KeyValueTextInputFormat;

import org.apache.hadoop.mapreduce.lib.jobcontrol.ControlledJob;

import org.apache.hadoop.mapreduce.lib.jobcontrol.JobControl;

import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;

import org.apache.hadoop.mapreduce.lib.output.TextOutputFormat;

import org.apache.hadoop.util.Tool;

import org.apache.hadoop.util.ToolRunner;

 

publicclass Recommendation extends Configured implements Tool {

   

    /* ********Job1******************* */

    publicstaticclass PostsMapper extends Mapper<Object, Text, IntWritable, Text> {

        publicvoid map(Object key, Text value, Context context) throws IOException,                InterruptedException {

            // 得到帖子的id,通过在Configuration中设置自定义参数获得

            String currThreadId = context.getConfiguration().get("currThreadId");

            String[] tmp = value.toString().split(",");

 

            // 如果是指定的帖子,输出该userid,标记为INEEDIT,代表看过该帖子的用户

            if (tmp[0].equals(currThreadId)) {

                context.write(new IntWritable(Integer.parseInt(tmp[1])), new                                Text("INEEDIT"));

            }

           

            // 输出表keyuseridvaluethreadid

            context.write(new IntWritable(Integer.parseInt(tmp[1])), new Text(tmp[0]));

 

        }

    }

 

    publicstaticclass PostsReducer extends Reducer<IntWritable, Text, Text, Text> {

 

        publicvoid reduce(IntWritable key, Iterable<Text> values, Context context) throws          IOException, InterruptedException {

            booleanfind = false;

           

            // 定义一个HashSet,用来查找所需用户,并去重

            Set<Text> set = new HashSet<Text>();

            for (Text val : values) {

               

                // 判断此用户是否需要统计

                if (val.toString().equals("INEEDIT")) {

                    find = true;

                } else {

                    set.add(new Text(val.toString()));

                }

            }

 

            if (find) {

                // set转换成数组

                Object o[] = set.toArray();

                for (inti = 0; i < o.length; i++) {

                    context.write(new Text(o[i].toString()), new Text("1"));

                }

            }

        }

    }

 

    /* ********Job2******************* */

    publicstaticclass CountMapper extends Mapper<Text, Text, Text, Text> {

       

        // Map用来读取数据并发送到Reduce

        publicvoid map(Text key, Text value, Context context) throws IOException,                  InterruptedException {

            context.write(key, value);

        }

    }

 

    publicstaticclass CountReducer extends Reducer<Text, Text, Text, Text> {

 

        // 计算每个帖子看过的用户数

        publicvoid reduce(Text key, Iterable<Text> values, Context context) throws IOException,            InterruptedException {

            intcount = 0;

            for (Text val : values) {

                count++;

            }

           

            // Reducer keythreadidvaluecount

            context.write(key, new Text(String.valueOf(count)));

        }

    }

 

    /* ********Job3******************* */

    publicstaticclass SortMapper extends Mapper<Text, Text, CountThread, Text> {

       

        // 排序

        publicvoid map(Text key, Text value, Context context) throws IOException,                  InterruptedException {

            CountThread ct = new CountThread();

            ct.setThreadId(key);

            ct.setCnt(new IntWritable(Integer.parseInt(value.toString())));

            context.write(ct, new Text());

        }

    }

 

    publicstaticclass SortReducer extends Reducer<CountThread, Text, CountThread, Text> {

        publicvoid reduce(CountThread key, Iterable<Text> values, Context context) throws          IOException, InterruptedException {

            context.write(key, null);

        }

    }

 

    @Override

    publicint run(String[] args) throws Exception {

        String input, output1, output2, outResult, threadid;

        if (args.length == 5) {

            threadid = args[0];

            input = args[1];

            output1 = args[2];

            output2 = args[3];

            outResult = args[4];

        } else {

            System.out.println("Usage:Recommendation <threadId> <input> <output1> <output2>             <result>");

            System.out.println("Use Default Values This Time");

            threadid = "1479820";

            input = "hdfs://hadoop1:9000/usr/hadoop/in/log";

            output1 = "hdfs://hadoop1:9000/usr/hadoop/week7_recom1";

            output2 = "hdfs://hadoop1:9000/usr/hadoop/week7_recom2";

            outResult = "hdfs://hadoop1:9000/usr/hadoop/week7_recom3";

        }

 

        Configuration conf1 = getConf();

 

        // 设置自定义参数

        conf1.set("currThreadId", threadid);

        Job job1 = new Job(conf1, "Job1:Posts");

        job1.setJarByClass(Recommendation.class);

        job1.setMapperClass(PostsMapper.class);

        job1.setReducerClass(PostsReducer.class);

        job1.getConfiguration().set("mapred.textoutputformat.separator", ",");

        job1.setMapOutputKeyClass(IntWritable.class);

        job1.setMapOutputValueClass(Text.class);

        job1.setOutputKeyClass(Text.class);

        job1.setOutputValueClass(Text.class);

        FileInputFormat.setInputPaths(job1, new Path(input));

        FileOutputFormat.setOutputPath(job1, new Path(output1));

 

        Configuration conf2 = new Configuration();

        Job job2 = new Job(conf2, "Job2:Count");

        job2.setJarByClass(Recommendation.class);

        job2.setInputFormatClass(KeyValueTextInputFormat.class);

        job2.getConfiguration();

        job2.getConfiguration().set("mapred.textoutputformat.separator", ",");

        job2.getConfiguration().set("mapreduce.input.keyvaluelinerecordreader.key.value.separa  tor", ",");

        job2.setOutputFormatClass(TextOutputFormat.class);

        FileInputFormat.setInputPaths(job2, new Path(output1 + "/part*"));

        FileOutputFormat.setOutputPath(job2, new Path(output2));

        job2.setMapperClass(CountMapper.class); //

        job2.setReducerClass(CountReducer.class);

        job2.setMapOutputKeyClass(Text.class);

        job2.setMapOutputValueClass(Text.class);

        job2.setOutputKeyClass(Text.class);

        job2.setOutputValueClass(Text.class);

 

        Configuration conf3 = new Configuration();

        Job job3 = new Job(conf3, "Job3:Result");

        job3.setJarByClass(Recommendation.class);

        job3.setInputFormatClass(KeyValueTextInputFormat.class);

        job3.getConfiguration().set("mapreduce.input.keyvaluelinerecordreader.key.value.separa  tor", ",");

        job3.setOutputFormatClass(TextOutputFormat.class);

        FileInputFormat.setInputPaths(job3, new Path(output2 + "/part*"));

        FileOutputFormat.setOutputPath(job3, new Path(outResult));

        job3.setMapperClass(SortMapper.class);

        job3.setReducerClass(SortReducer.class);

        job3.setMapOutputKeyClass(CountThread.class);

        job3.setMapOutputValueClass(Text.class);

        job3.setOutputKeyClass(CountThread.class);

        job3.setOutputValueClass(Text.class);

 

        ControlledJob cJob1 = new ControlledJob(job1.getConfiguration());

        ControlledJob cJob2 = new ControlledJob(job2.getConfiguration());

        ControlledJob cJob3 = new ControlledJob(job3.getConfiguration());

       

        // 设置作业依赖关系

        cJob2.addDependingJob(cJob1);

        cJob3.addDependingJob(cJob2);

 

        JobControl jobControl = new JobControl("RecommendationJob");

        jobControl.addJob(cJob1);

        jobControl.addJob(cJob2);

        jobControl.addJob(cJob3);

 

        cJob1.setJob(job1);

        cJob2.setJob(job2);

        cJob3.setJob(job3);

 

        // 新建一个线程来运行已加入JobControl中的作业,开始进程并等待结束

        Thread jobControlThread = new Thread(jobControl);

        jobControlThread.start();

        while (!jobControl.allFinished()) {

            Thread.sleep(500);

        }

        jobControl.stop();

 

        return 0;

    }

 

    publicstaticvoid main(String[] args) throws Exception {

        intres = ToolRunner.run(new Recommendation(), args);

        System.exit(res);

    }

}

3.3  准备数据

使用SSH工具(参见第122.1.3.1Linux文件传输工具所描述)把提供的测试数据log上传到本地目录/usr/local/hadoop-1.1.2/input中,然后使用eclipseHDFS插件工具上传该文件到/usr/hadoop/in目录中,如下图所示:

clip_image086

log文件内容如下:

1455924,2134

1466253,3500

1479820,3500

1455924,2134

1472481,3500

1478790,2134

1466253,2134

1472481,2134

1479820,2134

1479820,4350

3.4  实现过程

3.4.1问题分析

Job1:先将每行log数据解析为keyuseridvaluethreadid,同时判断是否是当前正在阅读的帖子(currThreadId),如果是的话就输出一条特殊的记录key:userid,value:0。经过shuffle后可得{userid:thrd thrd (INEEDIT)}。每行数据代表一个用户看过的所有帖子,其中有标记INEEDIT的行就代表了该userid看过正在阅读的帖子(currThreadId), 这样在reduce中把有INEEDIT标记的行找出来,threadid去重,变换成keythrd value1格式输出。

Job2:由于job1把问题变换成了wordcount问题,这里对每个thrd进行累加输出,代表看过当前正在阅读的帖子(currThreadId)的用户中,对其他每个帖子又有多少人看过。

Job3job2已经把结果统计出来了,但mapreduce是按照key来按顺序排,这里实现按value倒序排序

 

3.4.2处理流程图

clip_image088

3.4.3配置运行参数

新建一个Java应用运行程序,需要在Arguments页签填写Recommendation运行的主题编号、输入日志文件路径、作业1输出路径、作业2输出路径和作业3输出路径五个参数,需要注意的是输入、输出路径参数路径需要全路径,否则运行会报错:

l  主题编号:1479820

l  输入日志文件路径: hdfs:// hadoop1:9000/usr/hadoop/in/log

l  作业1输出路径:hdfs:// hadoop1:9000/usr/hadoop/out/week7_recom1

l  作业2输出路径:hdfs:// hadoop1:9000/usr/hadoop/out/week7_recom2

l  作业3输出路径:hdfs:// hadoop1:9000/usr/hadoop/out/week7_recom3

clip_image090

3.4.4运行作业

配置完毕后,执行MapReduce作业,执行成功到eclipseHDFS插件工具查看结果

下图为Job1结果:

1472481,1

1455924,1

1479820,1

1466253,1

1478790,1

1455924,1

1472481,1

1479820,1

1466253,1

1479820,1

clip_image092

下图为Job2结果:

1455924,1

1466253,2

1472481,2

1478790,1

1479820,3

1455924,1

clip_image092[1]

下图为Job3结果:

1479820    3

1466253    2

1472481    2

1455924    1

1478790    1

1455924    1

clip_image094

 

posted @ 2014-12-30 08:51  shishanyuan  阅读(21026)  评论(7编辑  收藏  举报