如何配置MapReduce的多路輸入輸出 - MaxCompute

本文為您介紹MapReduce的多路輸入輸出樣本。

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

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

準備多路輸入輸出的測試表和資源。

建立測試表。

CREATE TABLE wc_in1(key STRING, value STRING);
CREATE TABLE wc_in2(key STRING, value STRING);
CREATE TABLE mr_multiinout_out1 (key STRING, cnt BIGINT);
CREATE TABLE mr_multiinout_out2 (key STRING, cnt BIGINT)  PARTITIONED BY (a string, b string);
ALTER TABLE mr_multiinout_out2 ADD PARTITION (a='1', b='1');
ALTER TABLE mr_multiinout_out2 ADD PARTITION (a='2', b='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;

使用Tunnel將MaxCompute用戶端bin目錄下data1.txt，data2.txt分別匯入wc_in1和wc_in2表中。
```
tunnel upload data1.txt wc_in1;
tunnel upload data2.txt wc_in2;
```
匯入wc_in1表的資料如下。
```
hello,odps
```
匯入wc_in2表的資料如下。
```
hello,world
```

測試步驟

在MaxCompute用戶端中執行MultipleInOut。

jar -resources mapreduce-examples.jar -classpath data\resources\mapreduce-examples.jar
com.aliyun.odps.mapred.open.example.MultipleInOut wc_in1,wc_in2 mr_multiinout_out1,mr_multiinout_out2|a=1/b=1|out1,mr_multiinout_out2|a=2/b=2|out2;

預期結果

作業成功結束後，mr_multiinout_out1中的內容如下。

+------------+------------+
| key        | cnt        |
+------------+------------+
| default    | 1          |
+------------+------------+

mr_multiinout_out2中的內容如下。

+--------+------------+---+---+
| key    | cnt        | a | b |
+--------+------------+---+---+
| odps   | 1          | 1 | 1 |
| world  | 1          | 1 | 1 |
| out1   | 1          | 1 | 1 |
| hello  | 2          | 2 | 2 |
| out2   | 1          | 2 | 2 |
+--------+------------+---+---+

程式碼範例

Pom依賴資訊，請參見注意事項。

package com.aliyun.odps.mapred.open.example;
import java.io.IOException;
import java.util.Iterator;
import java.util.LinkedHashMap;
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.TaskContext;
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;
/**
     * Multi input & output example.
     **/
public class MultipleInOut {
    public static class TokenizerMapper extends MapperBase {
        Record word;
        Record one;
        @Override
            public void setup(TaskContext context) throws IOException {
            word = context.createMapOutputKeyRecord();
            one = context.createMapOutputValueRecord();
            one.set(new Object[] { 1L });
        }
        @Override
            public void map(long recordNum, Record record, TaskContext context)
            throws IOException {
            for (int i = 0; i < record.getColumnCount(); i++) {
                word.set(new Object[] { record.get(i).toString() });
                context.write(word, one);
            }
        }
    }
    public static class SumReducer extends ReducerBase {
        private Record result;
        private Record result1;
        private Record result2;
        @Override
            public void setup(TaskContext context) throws IOException {
            /**對於不同的輸出需要建立不同的Record，通過Label來區分。*/
            result = context.createOutputRecord();
            result1 = context.createOutputRecord("out1");
            result2 = context.createOutputRecord("out2");
        }
        @Override
            public void reduce(Record key, Iterator<Record> values, TaskContext context)
            throws IOException {
            long count = 0;
            while (values.hasNext()) {
                Record val = values.next();
                count += (Long) val.get(0);
            }
            long mod = count % 3;
            if (mod == 0) {
                result.set(0, key.get(0));
                result.set(1, count);
                /**不指定Label，預設(default)輸出。*/
                context.write(result);
            } else if (mod == 1) {
                result1.set(0, key.get(0));
                result1.set(1, count);
                context.write(result1, "out1");
            } else {
                result2.set(0, key.get(0));
                result2.set(1, count);
                context.write(result2, "out2");
            }
        }
        @Override
            public void cleanup(TaskContext context) throws IOException {
            Record result = context.createOutputRecord();
            result.set(0, "default");
            result.set(1, 1L);
            context.write(result);
            Record result1 = context.createOutputRecord("out1");
            result1.set(0, "out1");
            result1.set(1, 1L);
            context.write(result1, "out1");
            Record result2 = context.createOutputRecord("out2");
            result2.set(0, "out2");
            result2.set(1, 1L);
            context.write(result2, "out2");
        }
    }
    /**將分區字串如"ds=1/pt=2"轉為Map的形式。*/
    public static LinkedHashMap<String, String> convertPartSpecToMap(
        String partSpec) {
        LinkedHashMap<String, String> map = new LinkedHashMap<String, String>();
        if (partSpec != null && !partSpec.trim().isEmpty()) {
            String[] parts = partSpec.split("/");
            for (String part : parts) {
                String[] ss = part.split("=");
                if (ss.length != 2) {
                    throw new RuntimeException("ODPS-0730001: error part spec format: "
                                               + partSpec);
                }
                map.put(ss[0], ss[1]);
            }
        }
        return map;
    }
    public static void main(String[] args) throws Exception {
        String[] inputs = null;
        String[] outputs = null;
        if (args.length == 2) {
            inputs = args[0].split(",");
            outputs = args[1].split(",");
        } else {
            System.err.println("MultipleInOut in... out...");
            System.exit(1);
        }
        JobConf job = new JobConf();
        job.setMapperClass(TokenizerMapper.class);
        job.setReducerClass(SumReducer.class);
        job.setMapOutputKeySchema(SchemaUtils.fromString("word:string"));
        job.setMapOutputValueSchema(SchemaUtils.fromString("count:bigint"));
        /**解析使用者的輸入表字串。*/
        for (String in : inputs) {
            String[] ss = in.split("\\|");
            if (ss.length == 1) {
                InputUtils.addTable(TableInfo.builder().tableName(ss[0]).build(), job);
            } else if (ss.length == 2) {
                LinkedHashMap<String, String> map = convertPartSpecToMap(ss[1]);
                InputUtils.addTable(TableInfo.builder().tableName(ss[0]).partSpec(map).build(), job);
            } else {
                System.err.println("Style of input: " + in + " is not right");
                System.exit(1);
            }
        }
        /**解析使用者的輸出表字串。*/
        for (String out : outputs) {
            String[] ss = out.split("\\|");
            if (ss.length == 1) {
                OutputUtils.addTable(TableInfo.builder().tableName(ss[0]).build(), job);
            } else if (ss.length == 2) {
                LinkedHashMap<String, String> map = convertPartSpecToMap(ss[1]);
                OutputUtils.addTable(TableInfo.builder().tableName(ss[0]).partSpec(map).build(), job);
            } else if (ss.length == 3) {
                if (ss[1].isEmpty()) {
                    LinkedHashMap<String, String> map = convertPartSpecToMap(ss[2]);
                    OutputUtils.addTable(TableInfo.builder().tableName(ss[0]).partSpec(map).build(), job);
                } else {
                    LinkedHashMap<String, String> map = convertPartSpecToMap(ss[1]);
                    OutputUtils.addTable(TableInfo.builder().tableName(ss[0]).partSpec(map)
                                         .label(ss[2]).build(), job);
                }
            } else {
                System.err.println("Style of output: " + out + " is not right");
                System.exit(1);
            }
        }
        JobClient.runJob(job);
    }
}