前提條件

已通過快速入門完成測試所需環境配置。

測試準備

1. 準備好測試程序的JAR包，假設名字為mapreduce-examples.jar，本地存放路徑為MaxCompute客戶端bin目錄下data\resources。

2. 準備好Join的測試表和資源。

   1. 創建測試表。 對表mr_Join_src1與表mr_Join_src2執行Join操作，輸出表mr_Join_out。

      CREATE TABLE mr_Join_src1(key BIGINT, value STRING);
      CREATE TABLE mr_Join_src2(key BIGINT, value STRING);
      CREATE TABLE mr_Join_out(key BIGINT, value1 STRING, value2 STRING);

   2. 添加測試資源。

      -- 首次添加忽略-f覆蓋指令。
      add jar data\resources\mapreduce-examples.jar -f;

3. 使用Tunnel將MaxCompute客戶端bin目錄下data1.txt和data2.txt分別導入mr_Join_src和mr_Join_src2表中。

   tunnel upload data1.txt mr_Join_src1;
   tunnel upload data2.txt mr_Join_src2;

   導入mr_Join_src1數據的內容。

    1,hello
    2,odps

   導入mr_Join_src2數據的內容。

   1,odps
   3,hello
   4,odps

測試步驟

jar -resources mapreduce-examples.jar -classpath data\resources\mapreduce-examples.jar
com.aliyun.odps.mapred.open.example.Join mr_Join_src1 mr_Join_src2 mr_Join_out;

預期結果

+------------+------------+------------+
| key        | value1     | value2     |
+------------+------------+------------+
|  1         | hello      |  odps      | 
+------------+------------+------------+

代碼示例

Pom依賴信息，請參見注意事項。

package com.aliyun.odps.mapred.open.example;
import java.io.IOException;
import java.util.ArrayList;
import java.util.Iterator;
import java.util.List;
import org.apache.commons.logging.Log;
import org.apache.commons.logging.LogFactory;
import com.aliyun.odps.data.Record;
import com.aliyun.odps.data.TableInfo;
import com.aliyun.odps.mapred.JobClient;
import com.aliyun.odps.mapred.MapperBase;
import com.aliyun.odps.mapred.ReducerBase;
import com.aliyun.odps.mapred.conf.JobConf;
import com.aliyun.odps.mapred.utils.InputUtils;
import com.aliyun.odps.mapred.utils.OutputUtils;
import com.aliyun.odps.mapred.utils.SchemaUtils;
/**
     * Join, mr_Join_src1/mr_Join_src2(key bigint, value string), mr_Join_out(key
     * bigint, value1 string, value2 string)
     *
     */
public class Join {
    public static final Log LOG = LogFactory.getLog(Join.class);
    public static class JoinMapper extends MapperBase {
        private Record mapkey;
        private Record mapvalue;
        private long tag;
        @Override
            public void setup(TaskContext context) throws IOException {
            mapkey = context.createMapOutputKeyRecord();
            mapvalue = context.createMapOutputValueRecord();
            tag = context.getInputTableInfo().getLabel().equals("left") ? 0 : 1;
        }
        @Override
            public void map(long key, Record record, TaskContext context)
            throws IOException {
            mapkey.set(0, record.get(0));
            mapkey.set(1, tag);
            for (int i = 1; i < record.getColumnCount(); i++) {
                mapvalue.set(i - 1, record.get(i));
            }
            context.write(mapkey, mapvalue);
        }
    }
    public static class JoinReducer extends ReducerBase {
        private Record result = null;
        @Override
            public void setup(TaskContext context) throws IOException {
            result = context.createOutputRecord();
        }
        /**reduce函數每次的輸入會是key相同的所有record。*/
        @Override
            public void reduce(Record key, Iterator<Record> values, TaskContext context)
            throws IOException {
            long k = key.getBigint(0);
            List<Object[]> leftValues = new ArrayList<Object[]>();
            /**由于設置了outputKeySortColumn是key + tag組合，這樣可以保證reduce函數的輸入record中，left表的record數據在前面。*/
            while (values.hasNext()) {
                Record value = values.next();
                long tag = (Long) key.get(1);
                /**左表的數據會先緩存到內存中。*/
                if (tag == 0) {
                    leftValues.add(value.toArray().clone());
                } else {
                    /**右表的數據會與所有左表的數據進行join輸出，此時左表的數據已經全部在內存里了。*/
                    /**這個實現只是一個功能展示，性能較低，不建議用于實際生產。*/
                    for (Object[] leftValue : leftValues) {
                        int index = 0;
                        result.set(index++, k);
                        for (int i = 0; i < leftValue.length; i++) {
                            result.set(index++, leftValue[i]);
                        }
                        for (int i = 0; i < value.getColumnCount(); i++) {
                            result.set(index++, value.get(i));
                        }
                        context.write(result);
                    }
                }
            }
        }
    }
    public static void main(String[] args) throws Exception {
        if (args.length != 3) {
            System.err.println("Usage: Join <input table1> <input table2> <out>");
            System.exit(2);
        }
        JobConf job = new JobConf();
        job.setMapperClass(JoinMapper.class);
        job.setReducerClass(JoinReducer.class);
        job.setMapOutputKeySchema(SchemaUtils.fromString("key:bigint,tag:bigint"));
        job.setMapOutputValueSchema(SchemaUtils.fromString("value:string"));
        job.setPartitionColumns(new String[]{"key"});
        job.setOutputKeySortColumns(new String[]{"key", "tag"});
        job.setOutputGroupingColumns(new String[]{"key"});
        job.setNumReduceTasks(1);
        InputUtils.addTable(TableInfo.builder().tableName(args[0]).label("left").build(), job);
        InputUtils.addTable(TableInfo.builder().tableName(args[1]).label("right").build(), job);
        OutputUtils.addTable(TableInfo.builder().tableName(args[2]).build(), job);
        JobClient.runJob(job);
    }
}