MapReduce高级编程2

MapReduce Top N 、二次排序，MapJoin：

TOP N

对于一组输入List(key,value),我们要创建一个Top N 列表,这是一种过滤模式,查看输入数据特定子集,观察用户的行为。

解决方案

key是唯一键，需要对输入进行额外的聚集处理，先把输入分区成小块，然后把每个小块发送到一个映射器中。每个映射器会创建一个本地Top N 列表发送到一个规约器中，即最终由一个规约其产生一个Top N 列表。对于大多数的MapReduce算法，由一个规约器接收所有数据会使负载不均衡，从而产生瓶颈问题。但是本解决方案产生的Top N 是很少量的数据，如果有映射器，则会需要处理1000 X N的数据。

数据

思路

计算过程以K V对传输

可以创建一个最小堆，最常用的是SortMap<K,V> 和TreeMap<K,V>，如果top N.size()>N，则删除第一个元素（频数最小）如果求Bottom N构造最大堆。

代码

Mapper

import org.apache.hadoop.io.NullWritable;
import org.apache.hadoop.io.Text;
import org.apache.hadoop.mapreduce.Mapper;

import java.io.IOException;
import java.util.TreeMap;

public class TopTenMapper extends Mapper<Object, Text, NullWritable, Text> {
    private TreeMap<Integer, Text> visitimesMap = new TreeMap<Integer, Text>();

    @Override
    protected void map(Object key, Text value, Context context) throws IOException, InterruptedException {
        if (value == null) {
            return;
        }
        String[] strs = value.toString().split(" ");
        String tId = strs[0];
        String reputation = strs[1];

        if (tId == null || reputation == null) {
            return;
        }
        visitimesMap.put(Integer.parseInt(reputation), new Text(value));
        if (visitimesMap.size() > 10) {
            visitimesMap.remove(visitimesMap.firstKey());
        }
    }

    @Override
    protected void cleanup(Context context) throws IOException, InterruptedException {
        for (Text t : visitimesMap.values()) {
            context.write(NullWritable.get(), t);

        }
    }
}

Reduce

import org.apache.hadoop.io.NullWritable;
import org.apache.hadoop.io.Text;
import org.apache.hadoop.mapreduce.Reducer;

import java.io.IOException;
import java.util.TreeMap;

public class TopTenReduce extends Reducer<NullWritable, Text, NullWritable, Text> {
    private TreeMap<Integer, Text> visittimesMap = new TreeMap<Integer, Text>();

    @Override
    protected void reduce(NullWritable key, Iterable<Text> values, Context context) throws IOException, InterruptedException {
        for (Text value : values) {
            String[] strs = value.toString().split(" ");
            visittimesMap.put(Integer.parseInt(strs[1]), new Text(value));
            if (visittimesMap.size() > 10) {
                visittimesMap.remove(visittimesMap.firstKey());
            }
        }
        for (Text t : visittimesMap.values()) {
            context.write(NullWritable.get(), t);
        }
    }
}

Job

import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.io.NullWritable;
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;

public class TopTenJob {
    public static void main(String[] args) throws Exception {
        Configuration conf = new Configuration();
        if (args.length != 2) {
            System.out.println("Usage:TopTenDriver<in> <out>");
            System.exit(2);
        }
        Job job = Job.getInstance(conf);
        job.setJarByClass(TopTenJob.class);
        job.setMapperClass(TopTenMapper.class);
        job.setReducerClass(TopTenReduce.class);

        job.setNumReduceTasks(1);

        job.setOutputKeyClass(NullWritable.class);
        job.setOutputValueClass(Text.class);

        FileInputFormat.addInputPath(job, new Path(args[0]));
        FileOutputFormat.setOutputPath(job, new Path(args[1]));

        System.exit(job.waitForCompletion(true) ? 0 : 1);
    }
}

结果

非唯一键解决方案

假设所有的输入键是不唯一的，即会出现key相同而value不同的情况，则需要为两个阶段解决。

第一阶段：通过MapReduce先聚集具有相同键的元组，并将频数相加，将不唯一的键转换为唯一的键，其value值为频数和，输出为唯一键值对<key,value>

第二阶段：将第一阶段的输出的唯一键值对作为输入，采用唯一键的解决方案求Top N 即可。

全排序

创建随机数据

for i in {1..100000};
do
        echo $RANDOM
done;

sh createdata.sh >data1
sh createdata.sh >data2
sh createdata.sh >data3
sh createdata.sh >data4

MyPartitioner

import org.apache.hadoop.io.IntWritable;
import org.apache.hadoop.mapreduce.Partitioner;

public class MyPartitioner extends Partitioner<IntWritable, IntWritable> {

    public int getPartition(IntWritable key, IntWritable value, int numPartitions) {
        int keyInt = Integer.parseInt(key.toString());
        if (keyInt > 20000) {   //数据区间在在[0,350000]随机生成概率比较均匀.
            return 2;
        } else if (keyInt > 10000) {
            return 1;
        } else {
            return 0;
        }
    }
}

Mysort

import org.apache.hadoop.io.IntWritable;
import org.apache.hadoop.io.WritableComparable;
import org.apache.hadoop.io.WritableComparator;

public class Mysort extends WritableComparator {
    public Mysort() {
        super(IntWritable.class, true);
    }

    @Override
    public int compare(WritableComparable a, WritableComparable b) {
        IntWritable v1 = (IntWritable) a;
        IntWritable v2 = (IntWritable) b;
        return v2.compareTo(v1);
    }
}

SimpleMapper

import org.apache.hadoop.io.IntWritable;
import org.apache.hadoop.io.LongWritable;
import org.apache.hadoop.io.Text;
import org.apache.hadoop.mapreduce.Mapper;

import java.io.IOException;

public class SimpleMapper extends Mapper<LongWritable, Text, IntWritable,IntWritable> {
    @Override
    protected void map(LongWritable key, Text value, Context context) throws IOException, InterruptedException {
        //将Text类的value转换为IntWritable类型
        IntWritable intvalue = new IntWritable(Integer.parseInt(value.toString()));
        //值写入context
        context.write(intvalue,intvalue);
    }
}

SimpleReducer

import org.apache.hadoop.io.IntWritable;
import org.apache.hadoop.io.NullWritable;
import org.apache.hadoop.mapreduce.Reducer;

import java.io.IOException;

public class SimpleReducer extends Reducer<IntWritable, IntWritable, IntWritable, NullWritable> {
    @Override
    protected void reduce(IntWritable key, Iterable<IntWritable> values, Context context) throws IOException, InterruptedException {
        for (IntWritable value : values) {
            context.write(value, NullWritable.get());
        }
    }
}

SimpleJob

import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.io.IntWritable;
import org.apache.hadoop.io.NullWritable;
import org.apache.hadoop.mapreduce.Job;
import org.apache.hadoop.mapreduce.lib.input.FileInputFormat;
import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;


public class SimpleJob {
    public static void main(String[] args) throws Exception {
        Configuration conf = new Configuration();
        Job job = Job.getInstance(conf);

        job.setJarByClass(SimpleJob.class);

        job.setPartitionerClass(MyPartitioner.class);
        job.setSortComparatorClass(Mysort.class);
        job.setMapperClass(SimpleMapper.class);
        job.setMapOutputKeyClass(IntWritable.class);
        job.setMapOutputValueClass(IntWritable.class);

        job.setReducerClass(SimpleReducer.class);
        job.setOutputKeyClass(IntWritable.class);
        job.setOutputValueClass(NullWritable.class);

        job.setNumReduceTasks(3);
        FileInputFormat.setInputPaths(job, new Path(args[0]));
        FileOutputFormat.setOutputPath(job, new Path(args[1]));
        job.waitForCompletion(true);
    }
}

抽样实现全局排序

自定分区下实现用户按业务分区与Reduce任务数据一一对于,使用全排序最终的解结果在文件中一次按顺序输出,可能存在认为对数据了解不够多，造成分区的数据个数不一致，导致发生数据倾斜的问题。

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.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.input.KeyValueTextInputFormat;
import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;
import org.apache.hadoop.mapreduce.lib.partition.InputSampler;
import org.apache.hadoop.mapreduce.lib.partition.TotalOrderPartitioner;
import org.apache.hadoop.util.Tool;
import org.apache.hadoop.util.ToolRunner;

import java.io.IOException;

public class TotalOrderingPartition extends Configured implements Tool {
    //Map类
    static class SimpleMapper extends Mapper<Text, Text, Text, IntWritable> {
        @Override
        protected void map(Text key, Text value, Context context) throws IOException, InterruptedException {

            IntWritable intWritable = new IntWritable(Integer.parseInt(key.toString()));
            context.write(key, intWritable);
        }
    }

    //Reduce类
    static class SimpleReducer extends Reducer<Text, IntWritable, IntWritable, NullWritable> {
        @Override
        protected void reduce(Text key, Iterable<IntWritable> values, Context context) throws IOException, InterruptedException {
            for (IntWritable value : values) {
                context.write(value, NullWritable.get());
            }
        }
    }

    public int run(String[] args) throws Exception {
        Configuration conf = getConf();
        Job job = Job.getInstance(conf);
        job.setJarByClass(TotalOrderingPartition.class);

        job.setInputFormatClass(KeyValueTextInputFormat.class);

        FileInputFormat.addInputPath(job, new Path(args[0]));
        FileOutputFormat.setOutputPath(job, new Path(args[1]));

        job.setNumReduceTasks(3);

        job.setMapOutputKeyClass(Text.class);
        job.setMapOutputValueClass(IntWritable.class);

        job.setOutputKeyClass(IntWritable.class);
        job.setOutputValueClass(NullWritable.class);

        TotalOrderPartitioner.setPartitionFile(job.getConfiguration(), new Path(args[2]));
        InputSampler.Sampler<Text, Text> sampler = new InputSampler.SplitSampler<Text, Text>(1000, 10);

        InputSampler.writePartitionFile(job, sampler);
        job.setPartitionerClass(TotalOrderPartitioner.class);
        job.setMapperClass(SimpleMapper.class);
        job.setReducerClass(SimpleReducer.class);

        job.setJobName("Ite Blog");

        return job.waitForCompletion(true) ? 0 : 1;
    }

    //主方法
    public static void main(String[] args) throws Exception {
        args = new String[]{"/input/simple","/output/random_simple1","/output/random_simple2"};
        int exitCode = ToolRunner.run(new TotalOrderingPartition(), args);
        System.exit(exitCode);
    }
}

通过抽样分区，可以使个分区所含的记录大致相等，使作业的总体执行时间不会因为一个Reducer任务之后而拖慢整体进度。

二次排序

二次排序不同于全排序，它在规约阶段对与中间键相关联的中间值的某个属性进行排序。可以对传如各个Reducer的值进行升序或降序排序，即首先按照第一个字段排序，在对第二个字段相同的排序，且不能破坏第一个排序结果。

解决方案

Reducer内排序。让Reducer接收（k2,list(v2)）之后,利用Reducer内存进行排序，生成(k2,list(v2)),在进行处理，在数值过多的情况下，容易导致内存溢出。
利用MapReduce框架的排序框架对Reducer值进行排序。为自然键增加自然值得部分或整个值，构造组合键进行排序。由于排序工作是由MapReduce框架完成的，因此不会溢出。

流程

构造组合键(k,v1),其中v1是次键，K为自然键值，要在Reduce中注入一个值（v1）或者一个组合键。Reducer值按照什么来排序，就将其加入到自然键中，共同称为组合键。

对于Key=k2，若所有的映射器生成键值对为(k2,A1),(k2,A2),…(k2,An),对于一每个Ai，对于一个每一个Ai，设Ai为一个m元组值(ai2,ai3,…,aim)用bi来表示。因而映射器生成键值对可以做如下表示:

(k2,(a1,b1)),(k2(a2,b2)),…(k2,(an,bn))

自然键位k2,加入ai形成组合键((k2,ai))最终的映射器发出键值对如下：

((k2,a1),(a1,b1)),((k2,a2),(a2,b2)),….((k2,an),(an,bn))
在分区器中加入两个插件类：定制分区器，确保相同自然键的数据到达相同Reducer，也就是说，同一Reducer中可能由key=k2,key=k3的所有数据，定制比较器就将这些数据按照自然键进行分组。
分组比较器，可以使v1按有序的顺序到达Reducer，使用MapReduce执行框架完成排序，可以保证到达Reducer的值按键有序并按值有序。

实现类

MyWritable

import java.io.*;

import org.apache.hadoop.io.*;

public class MyWritable implements WritableComparable<MyWritable> {

    private int first;
    private int second;

    public MyWritable() {
    }

    public MyWritable(int first, int second) {
        set(first, second);
    }

    public void set(int first, int second) {
        this.first = first;
        this.second = second;
    }

    public int getFirst() {
        return first;
    }

    public int getSecond() {
        return second;
    }

    public void write(DataOutput out) throws IOException {
        out.writeInt(first);
        out.writeInt(second);
    }

    public void readFields(DataInput in) throws IOException {
        first = in.readInt();
        second = in.readInt();
    }

    @Override
    public int hashCode() {
        return first * 163 + second;
    }
    @Override
    public boolean equals(Object o) {
        if (o instanceof MyWritable) {
            MyWritable ip = (MyWritable) o;
            return first == ip.first && second == ip.second;
        }
        return false;
    }

    @Override
    public String toString() {
        return first + "\t" + second;
    }

    public int compareTo(MyWritable ip) {
        int cmp = compare(first, ip.first);
        if (cmp != 0) {
            return cmp;
        }
        return compare(second, ip.second);
    }

    public static int compare(int a, int b) {
        return (a < b ? -1 : (a == b ? 0 : 1));
    }
}

MyComparator

import org.apache.hadoop.io.WritableComparable;
import org.apache.hadoop.io.WritableComparator;

public class MyComparator extends WritableComparator {
    protected MyComparator() {
        super(MyWritable.class, true);
    }

    @Override
    public int compare(WritableComparable w1, WritableComparable w2) {
        MyWritable ip1 = (MyWritable) w1;
        MyWritable ip2 = (MyWritable) w2;
        int cmp = MyWritable.compare(ip1.getFirst(), ip2.getFirst());
        if (cmp != 0) {
            return cmp;
        }
        return -MyWritable.compare(ip1.getSecond(), ip2.getSecond());
    }
}

MyPartitioner

import org.apache.hadoop.io.NullWritable;
import org.apache.hadoop.mapreduce.Partitioner;

public class MyPartitioner extends Partitioner<MyWritable, NullWritable> {

    @Override
    public int getPartition(MyWritable key, NullWritable value, int numPartitions) {
        return (key.hashCode() & Integer.MAX_VALUE) % numPartitions;
    }
}

MyGroup

import org.apache.hadoop.io.WritableComparable;
import org.apache.hadoop.io.WritableComparator;

public class MyGroup extends WritableComparator {
    protected MyGroup() {
        super(MyWritable.class, true);
    }

    @Override
    public int compare(WritableComparable w1, WritableComparable w2) {
        MyWritable ip1 = (MyWritable) w1;
        MyWritable ip2 = (MyWritable) w2;
        return MyWritable.compare(ip1.getFirst(), ip2.getFirst());
    }
}

SecondarySortMapper

import org.apache.hadoop.io.LongWritable;
import org.apache.hadoop.io.NullWritable;
import org.apache.hadoop.io.Text;
import org.apache.hadoop.mapreduce.Mapper;

import java.io.IOException;

public class SecondarySortMapper extends
        Mapper<LongWritable, Text, MyWritable, NullWritable> {

    @Override
    protected void map(LongWritable key, Text value, Context context)
            throws IOException, InterruptedException {

        String[] strs = value.toString().split(" ");
        String year = strs[0];
        String Temperature = strs[1];
        if (year == null || Temperature == null) {
            return;
        }

        context.write(
                new MyWritable(Integer.parseInt(strs[0]), Integer
                        .parseInt(strs[1])), NullWritable.get());
    }

}

SecondarySortReducer

import java.io.IOException;

import org.apache.hadoop.io.NullWritable;
import org.apache.hadoop.mapreduce.Reducer;

public class SecondarySortReducer extends
        Reducer<MyWritable, NullWritable, MyWritable, NullWritable> {

    @Override
    protected void reduce(MyWritable key, Iterable<NullWritable> values,
                          Context context) throws IOException, InterruptedException {

        for (NullWritable val : values) {
            context.write(key, NullWritable.get());
        }
    }
}

SecondarySortJob

import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.io.NullWritable;
import org.apache.hadoop.mapreduce.Job;
import org.apache.hadoop.mapreduce.lib.input.FileInputFormat;
import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;

public class SecondarySortJob {
    public static void main(String[] args) throws Exception {

        Configuration conf = new Configuration();
        Job job = Job.getInstance(conf);
        job.setJarByClass(SecondarySortJob.class);
        job.setMapperClass(SecondarySortMapper.class);
        job.setPartitionerClass(MyPartitioner.class);
        job.setSortComparatorClass(MyComparator.class);
        job.setGroupingComparatorClass(MyGroup.class);
        job.setReducerClass(SecondarySortReducer.class);
        job.setOutputKeyClass(MyWritable.class);
        job.setOutputValueClass(NullWritable.class);
        job.setNumReduceTasks(1);
        FileInputFormat.addInputPath(job, new Path(args[0]));
        FileOutputFormat.setOutputPath(job, new Path(args[1]));
        job.waitForCompletion(true);
    }
}

原始数据

排序后数据

连接

Reduce端连接

TableBean

import org.apache.hadoop.io.Writable;

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

public class TableBean implements Writable {

    private String orderId;
    private String pid;
    private int amount;
    private String pname;
    private String flag;

    public TableBean() {
        super();
    }

    public TableBean(String orderId, String pid, int amount, String pname, String flag) {
        super();
        this.orderId = orderId;
        this.pid = pid;
        this.amount = amount;
        this.pname = pname;
        this.flag = flag;
    }


    @Override
    public String toString() {
        return orderId + "\t" + pname + "\t" + amount;
    }

    public String getOrderId() {
        return orderId;
    }

    public void setOrderId(String orderId) {
        this.orderId = orderId;
    }

    public String getPid() {
        return pid;
    }

    public void setPid(String pid) {
        this.pid = pid;
    }

    public int getAmount() {
        return amount;
    }

    public void setAmount(int amount) {
        this.amount = amount;
    }

    public String getPname() {
        return pname;
    }

    public void setPname(String pname) {
        this.pname = pname;
    }

    public String getFlag() {
        return flag;
    }

    public void setFlag(String flag) {
        this.flag = flag;
    }

    @Override
    public void write(DataOutput out) throws IOException {
        out.writeUTF(orderId);
        out.writeUTF(pid);
        out.writeInt(amount);
        out.writeUTF(pname);
        out.writeUTF(flag);
    }

    @Override
    public void readFields(DataInput in) throws IOException {
        orderId = in.readUTF();
        pid = in.readUTF();
        amount = in.readInt();
        pname = in.readUTF();
        flag = in.readUTF();
    }
}

TableMapper

import org.apache.hadoop.io.LongWritable;
import org.apache.hadoop.io.Text;
import org.apache.hadoop.mapreduce.Mapper;
import org.apache.hadoop.mapreduce.lib.input.FileSplit;

import java.io.IOException;

public class TableMapper extends Mapper<LongWritable, Text, Text, TableBean>{
    Text k = new Text();
    TableBean tableBean = new TableBean();
    @Override
    protected void map(LongWritable key, Text value, Context context) throws IOException, InterruptedException {
        //获取文件名
        FileSplit split = (FileSplit) context.getInputSplit();
        String name = split.getPath().getName();
        //判断
        String line = value.toString();
        if (name.startsWith("order")){
            String[] fields = line.split("\t");
            tableBean.setOrderId(fields[0]);
            tableBean.setPid(fields[1]);
            tableBean.setAmount(Integer.parseInt(fields[2]));
            tableBean.setPname("");
            tableBean.setFlag("0");
            k.set(fields[1]);
            context.write(k, tableBean);
        }else {
            String[] fields = line.split("\t");
            tableBean.setOrderId("");
            tableBean.setPid(fields[0]);
            tableBean.setAmount(0);
            tableBean.setPname(fields[1]);
            tableBean.setFlag("1");
            k.set(fields[0]);
            context.write(k, tableBean);
        }
    }
}

TableReducer

import org.apache.commons.beanutils.BeanUtils;
import org.apache.hadoop.io.NullWritable;
import org.apache.hadoop.io.Text;
import org.apache.hadoop.mapreduce.Reducer;

import java.io.IOException;
import java.lang.reflect.InvocationTargetException;
import java.util.ArrayList;
import java.util.List;

public class TableReducer extends Reducer<Text, TableBean, NullWritable, TableBean> {
    //1001 01 1
    //01   小米
    @Override
    protected void reduce(Text key, Iterable<TableBean> values, Context context) throws IOException, InterruptedException {
        //用来接收order表的数据
        List<TableBean> orderBeans = new ArrayList<>();
        //用来接受pd表的数据
        TableBean pBean = new TableBean();
        for (TableBean value : values){
            //order表
            if (value.getFlag().equals("0")){
                TableBean oBean = new TableBean();
                try {
                    BeanUtils.copyProperties(oBean, value);
                    orderBeans.add(oBean);
                } catch (IllegalAccessException e) {
                    e.printStackTrace();
                } catch (InvocationTargetException e) {
                    e.printStackTrace();
                }
            }else {
                //pd.txt
                try {
                    BeanUtils.copyProperties(pBean, value);
                } catch (IllegalAccessException e) {
                    e.printStackTrace();
                } catch (InvocationTargetException e) {
                    e.printStackTrace();
                }
            }
        }
        for (TableBean orderbean : orderBeans){
            orderbean.setPname(pBean.getPname());
            context.write(NullWritable.get(),orderbean);
        }
    }
}

TableDriver

import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.io.NullWritable;
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 TableDriver {
    public static void main(String[] args) throws IOException, ClassNotFoundException, InterruptedException {
        Configuration configuration = new Configuration();
        Job job = Job.getInstance(configuration);

        // 2 指定本程序的jar包所在的本地路径
        job.setJarByClass(TableDriver.class);

        // 3 指定本业务job要使用的mapper/Reducer业务类
        job.setMapperClass(TableMapper.class);
        job.setReducerClass(TableReducer.class);

        // 4 指定mapper输出数据的kv类型
        job.setMapOutputKeyClass(Text.class);
        job.setMapOutputValueClass(TableBean.class);

        // 5 指定最终输出的数据的kv类型
        job.setOutputKeyClass(NullWritable.class);
        job.setOutputValueClass(TableBean.class);

        // 6 指定job的输入原始文件所在目录
        FileInputFormat.setInputPaths(job, new Path(args[0]));
        FileOutputFormat.setOutputPath(job, new Path(args[1]));

        // 7 将job中配置的相关参数，以及job所用的java类所在的jar包， 提交给yarn去运行
        boolean result = job.waitForCompletion(true);
        System.exit(result ? 0 : 1);

    }
}

原始数据

结果

1001    小米    1
1002    华为    2 
1003    格力    3

Map端连接

DistributedCacheMapper

import org.apache.commons.lang.StringUtils;
import org.apache.hadoop.io.LongWritable;
import org.apache.hadoop.io.NullWritable;
import org.apache.hadoop.io.Text;
import org.apache.hadoop.mapreduce.Mapper;

import java.io.BufferedReader;
import java.io.FileInputStream;
import java.io.IOException;
import java.io.InputStreamReader;
import java.net.URI;
import java.util.HashMap;

public class DistributedCacheMapper extends Mapper<LongWritable, Text, Text, NullWritable> {
    Text k = new Text();
    HashMap<String, String> map = new HashMap<>();
    /**
     * 先缓存pd表
     * @param context
     * @throws IOException
     * @throws InterruptedException
     */
    @Override
    protected void setup(Context context) throws IOException, InterruptedException {
        URI[] cacheFiles = context.getCacheFiles();
        //获取输入字节流
        FileInputStream fis = new FileInputStream(cacheFiles[0].getPath());
        //获取转换流
        InputStreamReader isr = new InputStreamReader(fis, "UTF-8");
        //获取缓存流
        BufferedReader br = new BufferedReader(isr);
        String line = null;
        //开始一行一行读数据
        while (StringUtils.isNotEmpty(line = br.readLine())) {
            String[] fields = line.split("\t");
            //将数据放入map中
            map.put(fields[0],fields[1]);
        }
        fis.close();
        isr.close();
        br.close();
    }

    @Override
    protected void map(LongWritable key, Text value, Context context) throws IOException, InterruptedException {
        //获取一行
        String line = value.toString();
        //通过pid获取pname
        String[] fields = line.split("\t");
        String pid = fields[1];
         String pname = map.get(pid);
        String str = line + "\t" + pname;
        k.set(str);
        context.write(k, NullWritable.get());
    }
}

DistributedCacheDriver

import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.io.NullWritable;
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.net.URI;

public class DistributedCacheDriver {
    public static void main(String[] args) throws Exception {
        // 1 获取job信息
        Configuration configuration = new Configuration();
        Job job = Job.getInstance(configuration);

        // 2 设置加载jar包路径
        job.setJarByClass(DistributedCacheDriver.class);

        // 3 关联map
        job.setMapperClass(DistributedCacheMapper.class);
        // 4 设置最终输出数据类型
        job.setOutputKeyClass(Text.class);
        job.setOutputValueClass(NullWritable.class);

        // 5 设置输入输出路径
        FileInputFormat.setInputPaths(job, new Path(args[0]));
        FileOutputFormat.setOutputPath(job, new Path(args[1]));

        // 6 加载缓存数据写入内存 默认是本地路径
        job.addCacheFile(new URI(args[2]));

        // 7 map端join的逻辑不需要reduce阶段，设置reducetask数量为0
        job.setNumReduceTasks(0);

        // 8 提交
        boolean result = job.waitForCompletion(true);
        System.exit(result ? 0 : 1);

    }
}

这种方法的运行速度很快,输入数据进行Mapper计算前分了部分,启用了一个Mapper,而Reducer连接时启动了两个Mapper.但该种计算模式受到JVM种堆的分配限制,如果集群中的内存足够大,业务符合逻辑(如内连接和做外连接)时可以考虑这种计算模式,如果数据集很大,Reducer连接操作是一个很不错的选择.当然连接操作的思想还有一些其他的方法,可以根据业务不同,采取不同的编程思想。

posted on 2019-02-22 10:01 风-fmgao 阅读(303) 评论(0) 编辑收藏举报

会员力量，点亮园子希望

刷新页面返回顶部

风-fmgao

导航

公告