自定义Partitioner分区

在Hadoop的MapReduce过程中，每个map task处理完数据后，如果存在自定义Combiner类，会先进行一次本地的reduce操作，然后把数据发送到Partitioner，由Partitioner来决定每条记录应该送往哪个reducer节点，默认使用的是HashPartitioner，其核心代码如下：

1） 自定义Partitioner类必须继承自Partitioner类，重写getPartition方法
package com.root.PartitionDemo;
import com.root.flowsum.FlowBean_2;
import org.apache.hadoop.io.Text;
import org.apache.hadoop.mapreduce.Partitioner;

public class provicePartition extends Partitioner<Text, FlowBean_2> {
@Override
public int getPartition(Text text, FlowBean_2 flowBean_2, int i) {

    String prePhoneNum = text.toString();
    int partition = 4;
    if ("Sci&Tech".equals(prePhoneNum)) {
        partition = 0;
    } else if ("Comm&Mgmt".equals(prePhoneNum)) {
        partition = 1;
    } else if ("Others".equals(prePhoneNum)) {
        partition = 2;
    } else {
    //我的测试文本文件中以上三个key字段之外没有其他字段，
    //所以第四个分区文件内为空
        partition = 3;
    }
    return partition;
}

}
2) map与reduce自定义代码
Map阶段:
package com.root.PartitionDemo;

import com.root.flowsum.FlowBean_2;
import org.apache.hadoop.io.LongWritable;
import org.apache.hadoop.io.Text;
import org.apache.hadoop.mapreduce.Mapper;

import java.io.IOException;

public class FlowCountMapper extends Mapper<LongWritable, Text, Text, FlowBean_2> {

Text k = new Text();
FlowBean_2 v = new FlowBean_2();

@Override
protected void map(LongWritable key, Text value, Context context) throws IOException, InterruptedException {
    //1.获取一行
    String line = value.toString();

    //2.切割,文件以","分割每个字段
    String[] fields = line.split(",");
    if(fields.length>1 && !fields[fields.length - 5].isEmpty() && !fields[fields.length - 3].isEmpty()) {
        //3.封装对象
        k.set(fields[1]);  

        Float etest = Float.parseFloat(fields[fields.length - 5]);//etest在倒数第五个位置
        Float mba = Float.parseFloat(fields[fields.length - 3]);  //mba在倒数第三个位置
        v.setEtest(etest);
        v.setMba(mba);
        //4. 写出
        context.write(k, v);
    }

}

}
Reduce阶段:
package com.root.PartitionDemo;

import com.root.flowsum.FlowBean_2;
import org.apache.hadoop.io.Text;
import org.apache.hadoop.mapreduce.Reducer;

import java.io.IOException;

public class FlowCountReducer extends Reducer<Text, FlowBean_2, Text, FlowBean_2> {
FlowBean_2 v = new FlowBean_2();

@Override
protected void reduce(Text key, Iterable<FlowBean_2> values, Context context) throws IOException, InterruptedException {
    //1.累加求和
    float sum_upFlow = 0;
    float sum_downFlow = 0;
    for (FlowBean_2 flowBean : values) {
        sum_upFlow += flowBean.getEtest();
        sum_downFlow += flowBean.getMba();
    }
    //这里我做了一个简单的upFlow(etest)与downFlow(mba)求和。
    v.set(sum_upFlow,sum_downFlow);
    //2.写出
    context.write(key,v);
}

}
3）封装获取文本文件获取到的数据:
package com.root.flowsum;

import org.apache.hadoop.io.Writable;

import java.io.DataInput;
import java.io.DataOutput;
import java.io.IOException;

public class FlowBean_2 implements Writable {
private float mba; //mba
private float etest; //etest
private float summe;

//空参构造，为了后续反射用
public FlowBean_2() {
    super();
}

public FlowBean_2(float mba, float etest) {
    super();
    this.mba = mba;
    this.etest = etest;
    summe = mba + etest;
}

//序列化方法
public void write(DataOutput out) throws IOException {
    out.writeFloat(mba);
    out.writeFloat(etest);
    out.writeFloat(summe);
}

//反序列化方法
public void readFields(DataInput in) throws IOException {
    //必须要求和序列化方法顺序一致
    mba = in.readFloat();
    etest = in.readFloat();
    summe = in.readFloat();

}

public float getMba() {
    return mba;
}

public void setMba(float mba) {
    this.mba = mba;
}

public float getEtest() {
    return etest;
}

public void setEtest(float etest) {
    this.etest = etest;
}

public float getSumme() {
    return summe;
}

public void setSumme(float summe) {
    this.summe = summe;
}

public void set(float etest2, float mba2) {
    etest = etest2;
    mba = mba2;
    summe = etest + mba;
}

@Override
public String toString() {
    return etest + "\t"
            + mba +
            "\t" + summe;
}

}
3) 提交job程序:
package com.root.PartitionDemo;

import com.root.flowsum.FlowBean_2;
import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.io.Text;
import org.apache.hadoop.mapreduce.Job;
import org.apache.hadoop.mapreduce.lib.input.FileInputFormat;
import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;
import java.io.IOException;

public class FlowsumDriver {
public static void main(String[] args) throws IOException, ClassNotFoundException, InterruptedException {
//将项目打包上传至服务器 下一行代码不加,
//服务器运行jar包,会要求用户跟测试文件以及文件输出路径
args = new String[]{“F:\scala\Workerhdfs\input3\”,“F:\scala\Workerhdfs\output3”};
//1.获取job对象
Configuration conf = new Configuration();
Job job = Job.getInstance(conf);
//2.设置jar路径
job.setJarByClass(FlowsumDriver.class);

    //3.关联mapper和reducer
    job.setMapperClass(FlowCountMapper.class);
    job.setReducerClass(FlowCountReducer.class);
    //4 设置mapper输出的key和value类型
    job.setMapOutputKeyClass(Text.class);
    job.setMapOutputValueClass(FlowBean_2.class);

    //5. 设置最终输出的key和value类型
    job.setOutputKeyClass(Text.class);
    job.setOutputValueClass(FlowBean_2.class);

    job.setPartitionerClass(provicePartition.class);
    job.setNumReduceTasks(4);

    //6.设置输出路径
    FileInputFormat.setInputPaths(job, new Path(args[0]));
    FileOutputFormat.setOutputPath(job, new Path(args[1]));
    //7.提交job
    boolean result = job.waitForCompletion(true);
    System.exit(result ? 0 : 1);
}

}
idea运行以上程序:
与我们预想的一样按照key值相同的进行分区以及汇总,part-r-00003文件内容为空.

打包到服务器上运行,上传测试文件以及jar包至hadoop文件系统,这里请读者自行将文件放入linux系统：
之前我的项目有
Flowsum.jar包，也就不折腾了重命名partition.jar包
查看测试文件大致内容:

这里请注意,测试文件必须以",“作为分割(以其他字符作为分割符必须更改自定义Map代码中String[] fields = line.split(”,"); 的分割符,以及测试文件不能有缺失值

**上传jar包至hadoop文件系统,相关命令:hadoop dfs -put partition.jar /input
**:

**上传测试文件partition.csv，相关命令:hadoop dfs -put partition.csv /tmp/input3
**

**运行partiton.jar包: **

结果如下:

查看分区文件内容:

自定义分区Partitioner分区成功