MapReduce实现join

转载：https://www.cnblogs.com/ivanny/p/mapreduce_join.html

一、map join

两份数据中，如果有一份数据比较小，小数据全部加载到内存，按关键字建立索引。大数据文件作为map的输入文件，对map()函数每一对输入，都能够方便地和已加载到内存的小数据进行连接。把连接结果按key输出，经过shuffle阶段，reduce端得到的就是已经按key分组的，并且连接好了的数据。

这种方法，要使用hadoop中的DistributedCache把小数据分布到各个计算节点，每个map节点都要把小数据库加载到内存，按关键字建立索引。

这种方法有明显的局限性：有一份数据比较小，在map端，能够把它加载到内存，并进行join操作。

有客户数据customer和订单数据orders。

customer

客户编号	姓名	地址	电话
1	hanmeimei	ShangHai	110
2	leilei	BeiJing	112
3	lucy	GuangZhou	119

** order**

订单编号	客户编号	其它字段被忽略
1	1	50
2	1	200
3	3	15
4	3	350
5	3	58
6	1	42
7	1	352
8	2	1135
9	2	400
10	2	2000
11	2	300

要求对customer和orders按照客户编号进行连接，结果要求对客户编号分组，对订单编号排序，对其它字段不作要求

客户编号	订单编号	订单金额	姓名	地址	电话
1	1	50	hanmeimei	ShangHai	110
1	2	200	hanmeimei	ShangHai	110
1	6	42	hanmeimei	ShangHai	110
1	7	352	hanmeimei	ShangHai	110
2	8	1135	leilei	BeiJing	112
2	9	400	leilei	BeiJing	112
2	10	2000	leilei	BeiJing	112
2	11	300	leilei	BeiJing	112
3	3	15	lucy	GuangZhou	119
3	4	350	lucy	GuangZhou	119
3	5	58	lucy	GuangZhou	119

1. 在提交job的时候，把小数据通过DistributedCache分发到各个节点。
2. map端使用DistributedCache读到数据，在内存中构建映射关系--如果使用专门的内存服务器，就把数据加载到内存服务器，map()节点可以只保留一份小缓存；如果使用BloomFilter来加速，在这里就可以构建；
3. map()函数中，对每一对，根据key到第2)步构建的映射里面中找出数据，进行连接，输出。

public class Join extends Configured implements Tool {
    // customer文件在hdfs上的位置。
    // TODO: 改用参数传入
    private static final String CUSTOMER_CACHE_URL = "hdfs://hadoop1:9000/user/hadoop/mapreduce/cache/customer.txt";
    private static class CustomerBean {     
        private int custId;
        private String name;
        private String address;
        private String phone;
        
        public CustomerBean() {}
        
        public CustomerBean(int custId, String name, String address,String phone) {
            super();
            this.custId = custId;
            this.name = name;
            this.address = address;
            this.phone = phone;
        }
     
        public int getCustId() {
            return custId;
        }

        public String getName() {
            return name;
        }

        public String getAddress() {
            return address;
        }

        public String getPhone() {
            return phone;
        }
    }
    
    private static class CustOrderMapOutKey implements WritableComparable<CustOrderMapOutKey> {
        private int custId;
        private int orderId;

        public void set(int custId, int orderId) {
            this.custId = custId;
            this.orderId = orderId;
        }
        
        public int getCustId() {
            return custId;
        }
        
        public int getOrderId() {
            return orderId;
        }
        
        @Override

        public void write(DataOutput out) throws IOException {
            out.writeInt(custId);
            out.writeInt(orderId);
        }

        @Override
        public void readFields(DataInput in) throws IOException {
            custId = in.readInt();
            orderId = in.readInt();
        }

        @Override
        public int compareTo(CustOrderMapOutKey o) {
            int res = Integer.compare(custId, o.custId);
            return res == 0 ? Integer.compare(orderId, o.orderId) : res;
        }
        
        @Override
        public boolean equals(Object obj) {
            if (obj instanceof CustOrderMapOutKey) {
                CustOrderMapOutKey o = (CustOrderMapOutKey)obj;
                return custId == o.custId && orderId == o.orderId;
            } else {
                return false;
            }
        }
        
        @Override
        public String toString() {
            return custId + "\t" + orderId;
        }
    }
    
    private static class JoinMapper extends Mapper<LongWritable, Text, CustOrderMapOutKey, Text> {
        private final CustOrderMapOutKey outputKey = new CustOrderMapOutKey();
        private final Text outputValue = new Text();
        
        /**
         * 在内存中customer数据
         */
        private static final Map<Integer, CustomerBean> CUSTOMER_MAP = new HashMap<Integer, Join.CustomerBean>();
        @Override
        protected void map(LongWritable key, Text value, Context context)
                throws IOException, InterruptedException {
            
            // 格式: 订单编号 客户编号    订单金额
            String[] cols = value.toString().split("\t");           
            if (cols.length < 3) {
                return;
            }
            
            int custId = Integer.parseInt(cols[1]);     // 取出客户编号
            CustomerBean customerBean = CUSTOMER_MAP.get(custId);
            
            if (customerBean == null) {         // 没有对应的customer信息可以连接
                return;
            }
            
            StringBuffer sb = new StringBuffer();
            sb.append(cols[2])
                .append("\t")
                .append(customerBean.getName())
                .append("\t")
                .append(customerBean.getAddress())
                .append("\t")
                .append(customerBean.getPhone());
            
            outputValue.set(sb.toString());
            outputKey.set(custId, Integer.parseInt(cols[0]));
            
            context.write(outputKey, outputValue);
        }
        
        @Override
        protected void setup(Context context)
                throws IOException, InterruptedException {
            FileSystem fs = FileSystem.get(URI.create(CUSTOMER_CACHE_URL), context.getConfiguration());
            FSDataInputStream fdis = fs.open(new Path(CUSTOMER_CACHE_URL));
            
            BufferedReader reader = new BufferedReader(new InputStreamReader(fdis));
            String line = null;
            String[] cols = null;
            
            // 格式：客户编号  姓名  地址  电话
            while ((line = reader.readLine()) != null) {
                cols = line.split("\t");
                if (cols.length < 4) {              // 数据格式不匹配，忽略
                    continue;
                }
                
                CustomerBean bean = new CustomerBean(Integer.parseInt(cols[0]), cols[1], cols[2], cols[3]);
                CUSTOMER_MAP.put(bean.getCustId(), bean);
            }
        }
    }

    /**
     * reduce
     * @author Ivan
     *
     */
    private static class JoinReducer extends Reducer<CustOrderMapOutKey, Text, CustOrderMapOutKey, Text> {
        @Override
        protected void reduce(CustOrderMapOutKey key, Iterable<Text> values, Context context)
                throws IOException, InterruptedException {
            // 什么事都不用做，直接输出
            for (Text value : values) {
                context.write(key, value);
            }
        }
    }
    /**
     * @param args
     * @throws Exception
     */
    public static void main(String[] args) throws Exception {
        if (args.length < 2) {
            new IllegalArgumentException("Usage: <inpath> <outpath>");
            return;
        }
        
        ToolRunner.run(new Configuration(), new Join(), args);
    }

    @Override
    public int run(String[] args) throws Exception {
        Configuration conf = getConf();
        Job job = Job.getInstance(conf, Join.class.getSimpleName());
        job.setJarByClass(SecondarySortMapReduce.class);
        
        // 添加customer cache文件
        job.addCacheFile(URI.create(CUSTOMER_CACHE_URL));
        
        FileInputFormat.addInputPath(job, new Path(args[0]));
        FileOutputFormat.setOutputPath(job, new Path(args[1]));
        
        // map settings
        job.setMapperClass(JoinMapper.class);
        job.setMapOutputKeyClass(CustOrderMapOutKey.class);
        job.setMapOutputValueClass(Text.class);
        
        // reduce settings
        job.setReducerClass(JoinReducer.class);
        job.setOutputKeyClass(CustOrderMapOutKey.class);
        job.setOutputKeyClass(Text.class);
        
        boolean res = job.waitForCompletion(true);
        
        return res ? 0 : 1;
    }
}