MapReduce入门详解（三）

个人博客原文链接

join操作

左外连接(map)-JoinMapSideMR

问题描述：
将两个文件中每行的内容拼接到一个文件中
思路分析：
准备好两个map，firstMapper和joinMapper，firstMapper负责获取文件内容，joinMapper负责拼接文件内容。利用Job开启两个firstMapper任务，获取到两个文件的内容，然后再开启一个joinMapper任务负责拼接获取到的两个文件。
注：不常用map端的连接操作，推荐reduce端的连接操作

public class JoinMapSideMR extends Configured implements Tool {

    public static class FirstStepMapper extends Mapper<LongWritable, Text, Text, NullWritable>{
        @Override
        protected void map(LongWritable key, Text value, Context context) throws IOException, InterruptedException {
            if(!value.toString().equals("")) {
                context.write(value, NullWritable.get());
            }
        }
    }
    //读取连接好的数据的mapper
    public static class JoinMapper extends Mapper<Text, TupleWritable, Text, Text>{
        @Override
        protected void map(Text key, TupleWritable value, Context context) throws IOException, InterruptedException {
            String v = StreamSupport.stream(value.spliterator(), false).map(s -> ((Text) s).toString())
                    .collect(Collectors.joining("|"));
            context.write(key,new Text(v));
        }
    }

    @Override
    public int run(String[] strings) throws Exception {
        Configuration conf = getConf();
        Path inpath1 = new Path(conf.get("inpath1"));
        Path inpath2 = new Path(conf.get("inpath2"));
        Path mr1 = new Path(conf.get("mr1"));
        Path mr2 = new Path(conf.get("mr2"));
        Path outpath = new Path(conf.get("outpath"));
        //------------------------
        Job job1 = Job.getInstance(conf,"first_step1_xj");
        job1.setJarByClass(this.getClass());
        job1.setMapperClass(FirstStepMapper.class);
        job1.setMapOutputKeyClass(Text.class);
        job1.setMapOutputValueClass(NullWritable.class);
        job1.setReducerClass(Reducer.class);
        job1.setOutputKeyClass(Text.class);
        job1.setOutputValueClass(NullWritable.class);
        TextInputFormat.addInputPath(job1,inpath1);
        TextOutputFormat.setOutputPath(job1,mr1);
        FileOutputFormat.setOutputCompressorClass(job1,new GzipCodec().getClass());
        //------------------------
        Job job2 = Job.getInstance(conf,"first_step2_xj");
        job2.setJarByClass(this.getClass());
        job2.setMapperClass(FirstStepMapper.class);
        job2.setMapOutputKeyClass(Text.class);
        job2.setMapOutputValueClass(NullWritable.class);
        job2.setReducerClass(Reducer.class);
        job2.setOutputKeyClass(Text.class);
        job2.setOutputValueClass(NullWritable.class);
        TextInputFormat.addInputPath(job2,inpath2);
        TextOutputFormat.setOutputPath(job2,mr2);
        FileOutputFormat.setOutputCompressorClass(job2,new GzipCodec().getClass());
        //------------------------
        Job job3 = Job.getInstance(conf,"map_join_xj");
        job3.setJarByClass(this.getClass());
        job3.setMapperClass(JoinMapper.class);
        job3.setMapOutputKeyClass(Text.class);
        job3.setMapOutputValueClass(Text.class);
        job3.setNumReduceTasks(0);
        job3.getConfiguration().set("mapreduce.input.keyvaluelinerecordreader.key.value.separator", ",");
        String expr = CompositeInputFormat.compose("inner", KeyValueTextInputFormat.class, mr1, mr2);
        job3.getConfiguration().set("mapreduce.join.expr",expr);
        job3.setInputFormatClass(CompositeInputFormat.class);
        TextOutputFormat.setOutputPath(job3,outpath);
        List<Job> list = new ArrayList();
        list.add(job1);
        list.add(job2);
        list.add(job3);
        for (Job job : list) {
            boolean succ = job.waitForCompletion(true);
            if(!succ){
                System.out.println(job.getJobName()+":"+ job.getJobState().getValue());
                break;
            }
        }
        return 0;
    }

    public static void main(String[] args)throws Exception {
        ToolRunner.run(new JoinMapSideMR(),args);
    }
}

左外连接(reduce)-JoinReduceSideMR

问题描述：
将两个文件中每行的内容拼接到一个文件中
思路分析：
准备好两个map，fistMapper和SecondMapper，两个map的key的输出类型都为复合类型，包含id和tag，另外准备两个类自定义分组和分区规则，只根据id来分组和分区。因此，这两个map的输出结果就会进入到同一个reduce中，最后在reduce中完成拼接操作。

复合类型-ArtistIDTag

public class ArtistIDTag implements WritableComparable<ArtistIDTag> {

    private Text ArtistID = new Text(); // id
    private IntWritable Tag = new IntWritable(); // 标记

    public ArtistIDTag() {
    }

    public ArtistIDTag(Text artistID, IntWritable tag) {
        this.ArtistID = new Text(artistID.toString());
        this.Tag = new IntWritable(tag.get());
    }

    public Text getArtistID() {
        return ArtistID;
    }

    public void setArtistID(Text artistID) {
        this.ArtistID = new Text(artistID.toString());
    }

    public IntWritable getTag() {
        return Tag;
    }

    public void setTag(IntWritable tag) {
        this.Tag = new IntWritable(tag.get());
    }

    @Override
    public int compareTo(ArtistIDTag o) {
        return this.ArtistID.compareTo(o.ArtistID)==0 ? this.Tag.compareTo(o.Tag) : this.ArtistID.compareTo(o.ArtistID);
    }

    @Override
    public void write(DataOutput dataOutput) throws IOException {
        ArtistID.write(dataOutput);
        Tag.write(dataOutput);
    }

    @Override
    public void readFields(DataInput dataInput) throws IOException {
        ArtistID.readFields(dataInput);
        Tag.readFields(dataInput);
    }
}

自定义分区规则-ArtistPartitioner

public class ArtistPartitioner extends Partitioner<ArtistIDTag, Text> {
    @Override
    public int getPartition(ArtistIDTag artistIDTag, Text text, int i) {
        return Math.abs(artistIDTag.getArtistID().hashCode()*127)%i;
    }
}

自定义分组规则-ArtistGroupComparator

public class ArtistGroupComparator extends WritableComparator{
    public ArtistGroupComparator() {
        super(ArtistIDTag.class,true);
    }

    @Override
    public int compare(WritableComparable a, WritableComparable b) {
        ArtistIDTag at1 = (ArtistIDTag) a;
        ArtistIDTag at2 = (ArtistIDTag) b;
        return at1.getArtistID().compareTo(at2.getArtistID());
    }
}

连接-JoinReduceSideMR

public class JoinReduceSideMR extends Configured implements Tool {

    public static void main(String[] args) throws Exception {
        ToolRunner.run(new JoinReduceSideMR(),args);
    }

    public static class FirstMapper extends Mapper<LongWritable,Text,ArtistIDTag,Text>{
        @Override
        protected void map(LongWritable key, Text value, Context context) throws IOException, InterruptedException {
            Stream.of(value.toString()).filter(s -> s.length()>0).forEach(ExceptionConsumer.of(s -> {
                        String id = s.substring(0,s.indexOf(","));
                        String info = s.substring(s.indexOf(",")+1,s.length());
                        context.write(new ArtistIDTag(new Text(id),new IntWritable(0)),new Text(info));
                    }));
        }
    }

    public static class SecondMapper extends Mapper<LongWritable,Text,ArtistIDTag,Text>{
        @Override
        protected void map(LongWritable key, Text value, Context context) throws IOException, InterruptedException {
            Stream.of(value.toString()).filter(s -> s.length()>0).forEach(ExceptionConsumer.of(s -> {
                        String id = s.substring(0,s.indexOf(","));
                        String info = s.substring(s.indexOf(",")+1,s.length());
                        context.write(new ArtistIDTag(new Text(id),new IntWritable(1)),new Text(info));
                    }));
        }
    }

    public static class JoinReducer extends Reducer<ArtistIDTag,Text,Text,Text>{
        @Override
        protected void reduce(ArtistIDTag key, Iterable<Text> values, Context context) throws IOException, InterruptedException {
            Iterator<Text> ite = values.iterator();
            String name = ite.next().toString();
            while (ite.hasNext()){
                Text count = ite.next();
                String info = name.toString() + "|" + count.toString();
                context.write(key.getArtistID(),new Text(info));
            }

        }
    }

    @Override
    public int run(String[] strings) throws Exception {
        Configuration conf = getConf();
        Job job = Job.getInstance(conf, "join_reduce_xj");
        job.setJarByClass(this.getClass());

        // 多任务输入
        MultipleInputs.addInputPath(job,new Path(conf.get("inpath1")),TextInputFormat.class,FirstMapper.class);
        MultipleInputs.addInputPath(job,new Path(conf.get("inpath2")),TextInputFormat.class,SecondMapper.class);

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

        job.setReducerClass(JoinReducer.class);
        job.setOutputKeyClass(Text.class);
        job.setOutputValueClass(Text.class);

        job.setOutputFormatClass(TextOutputFormat.class);

        TextOutputFormat.setOutputPath(job,new Path(conf.get("outpath")));

        // 设置分区规则
        job.setPartitionerClass(ArtistPartitioner.class);
        // 设置分组规则
        job.setGroupingComparatorClass(ArtistGroupComparator.class);

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

DB交互操作

读取-DBtoHdfsMR

问题描述：
从mysql数据库中读取数据，并输出到hdfs
思路分析：
1. 准备好一个实现了DBWritable接口的复合类型，在该类型中定义的属性分别对应数据库中的列名。
2. 将该复合类型作为map阶段输入的value的类型即可。
3. 让集群加载jdbc驱动类。
4. 设置配置信息，连接到数据库。
5. 将输入类型设置为DBInputFormat。
复合类型-YearStationTempDB
注：输入操作要实现WritableComparable接口，这里是读操作可以删除。

public class YearStationTempDB implements DBWritable,WritableComparable<YearStationTempDB> {

    private int year; // 年份
    private String station; // 气象站编号
    private int temperature; // 气温

    public YearStationTempDB() {
    }

    public YearStationTempDB(int year, String station, int temperature) {
        this.year = year;
        this.station = station;
        this.temperature = temperature;
    }

    @Override
    public void write(PreparedStatement prep) throws SQLException {
        prep.setInt(1,year);
        prep.setString(2,station);
        prep.setInt(3,temperature);
    }

    @Override
    public void readFields(ResultSet rs) throws SQLException {
        this.year = rs.getInt("year");
        this.station = rs.getString("station");
        this.temperature = rs.getInt("temperature");
    }

    public int getYear() {
        return year;
    }

    public void setYear(int year) {
        this.year = year;
    }

    public String getStation() {
        return station;
    }

    public void setStation(String station) {
        this.station = station;
    }

    public int getTemperature() {
        return temperature;
    }

    public void setTemperature(int temperature) {
        this.temperature = temperature;
    }

    @Override
    public String toString() {
        return year + "," + station +"," + temperature;
    }

    @Override
    public void write(DataOutput dataOutput) throws IOException {
        dataOutput.writeInt(year);
        dataOutput.writeUTF(station);
        dataOutput.writeInt(temperature);
    }

    @Override
    public void readFields(DataInput dataInput) throws IOException {
        year = dataInput.readInt();
        station = dataInput.readUTF();
        temperature = dataInput.readInt();
    }

    @Override
    public int compareTo(YearStationTempDB o) {
        return this.year - o.year == 0 ? this.station.compareTo(o.station) : this.year - o.year;
    }
}

读取-DBtoHdfsMR

public class DBtoHdfsMR extends Configured implements Tool {

    public static void main(String[] args) throws Exception {
        ToolRunner.run(new DBtoHdfsMR(),args);
    }

    public static class DBMapper extends Mapper<LongWritable,YearStationTempDB,LongWritable,Text>{
        @Override
        protected void map(LongWritable key, YearStationTempDB value, Context context) throws IOException, InterruptedException {
            context.write(key,new Text(value.toString()));
        }
    }

    @Override
    public int run(String[] strings) throws Exception {
        Configuration conf = getConf();
        Job job = Job.getInstance(conf,"bdtohdfs_xj");
        job.setJarByClass(this.getClass());

        job.setMapperClass(DBMapper.class);
        job.setMapOutputKeyClass(LongWritable.class);
        job.setMapOutputValueClass(Text.class);

        // 如何让集群加载jdbc驱动类
        // 1 将jar放入share/hadoop/yarn/下会自动上传jar到集群
        // 2 把jar放入集群中lib目录下，重启集群
        // 3 job.addFileToClassPath(),需要把jar包上传到hdfs
        //job.addFileToClassPath(new Path("hdfs://172.16.0.4:9000/data/mysql-connector-java-5.1.38.jar"));

        // 连接数据库
        DBConfiguration.configureDB(job.getConfiguration(),"com.mysql.jdbc.Driver",
                "jdbc:mysql://172.16.0.100:3306/hadoop","hadoop","hadoop");


        job.setInputFormatClass(DBInputFormat.class); // 设置为DB输入类型
        job.setOutputFormatClass(TextOutputFormat.class);

        // year = 2000是条件，表示输入year=2000的数据
        DBInputFormat.setInput(job,YearStationTempDB.class,"station_tbl","year = 2000","","year","station","temperature");
        TextOutputFormat.setOutputPath(job,new Path(conf.get("outpath")));
        job.setNumReduceTasks(1);
        return job.waitForCompletion(true)? 0 : 1;
    }
}

输入-HdfstoDBMR

问题描述：
从hdfs读取数据，并输出到mysql数据库
思路分析：
1. 准备好一个实现了WritableComparable接口的复合类型，在该类型中定义的属性分别对应数据库中的列名，并重写compareTo()方法。
2. 将该复合类型作为reduce阶段输出的key的类型即可。
3. 让集群加载jdbc驱动类。
4. 设置配置信息，连接到数据库。
5. 将输出类型设置为DBOutputFormat。

public class HdfstoDBMR extends Configured implements Tool {

    public static void main(String[] args) throws Exception {
        ToolRunner.run(new HdfstoDBMR(),args);
    }

    public static class HTDMapper extends Mapper<LongWritable,Text,YearStation,IntWritable> {
        @Override
        protected void map(LongWritable key, Text value, Context context) throws IOException, InterruptedException {
            NcdcRecordParser parser = new NcdcRecordParser();
            parser.parse(()->value.toString()).ifPresent(ExceptionConsumer.of(
                    p->{
                        int year = p.getYear();
                        String stationId = p.getStationId();
                        int temp = p.getAirTemperature();
                        YearStation ys = new YearStation(year+"",stationId);
                        context.write(ys,new IntWritable(temp));
                    }
            ));
        }
    }

    public static class HTDReducer extends Reducer<YearStation,IntWritable,YearStationTempDB,NullWritable>{
        @Override
        protected void reduce(YearStation key, Iterable<IntWritable> values, Context context) throws IOException, InterruptedException {
            Stream<IntWritable> stream = StreamSupport.stream(values.spliterator(), false);
            Integer max = stream.map(s -> s.get()).reduce(0, (x, y) -> Math.max(x, y));
            int y = Integer.parseInt(key.getYear().toString());
            YearStationTempDB yst = new YearStationTempDB(y,key.getStationid().toString(),max);
            context.write(yst,NullWritable.get());
        }
    }

    @Override
    public int run(String[] strings) throws Exception {
        Configuration conf = getConf();
        Job job = Job.getInstance(conf,"hdfstodb_xj");
        job.setJarByClass(this.getClass());

        job.setMapperClass(HTDMapper.class);
        job.setMapOutputKeyClass(YearStation.class);
        job.setMapOutputValueClass(IntWritable.class);

        job.setReducerClass(HTDReducer.class);
        job.setOutputKeyClass(YearStationTempDB.class);
        job.setOutputValueClass(NullWritable.class);

        // 如何让集群加载jdbc驱动类
        // 1 将jar放入share/hadoop/yarn/下会自动上传jar到集群
        // 2 把jar放入集群中lib目录下，重启集群
        // 3 job.addFileToClassPath(),需要把jar包上传到hdfs
        //job.addFileToClassPath(new Path("hdfs://172.16.0.4:9000/data/mysql-connector-java-5.1.38.jar"));

        // 连接数据库
        DBConfiguration.configureDB(job.getConfiguration(),"com.mysql.jdbc.Driver",
                "jdbc:mysql://172.16.0.100:3306/hadoop","hadoop","hadoop");

        job.setInputFormatClass(TextInputFormat.class);
        job.setOutputFormatClass(DBOutputFormat.class);

        TextInputFormat.addInputPath(job,new Path(conf.get("inpath")));
        DBOutputFormat.setOutput(job,"max_tmp_xj","year","station","temperature");
        return job.waitForCompletion(true)? 0 : 1;
    }
}

输入类型-InputFormat

常见输入类型

1. TextInputFormat：按行获取字符串数据
   
job.setInputFormatClass(TextInputFormat.class);
TextInputFormat.addInputPath(job,new Path(conf.get("inpath")));

2. CombineTextInputFormat：将多个输入文件压缩成一个文件，避免开启多个map
   
job.setInputFormatClass(CombineTextInputFormat.class);
CombineFileInputFormat.addInputPath(job,new Path(conf.get("inpath")));

3. KeyValueTextInputFormat：按key-value形式获取数据
   
conf.set("mapreduce.input.keyvaluelinerecordreader.key.value.separator", ","); // 设置key-value的分割符，只识别第一个分隔符
job1.setInputFormatClass(KeyValueTextInputFormat.class);
KeyValueTextInputFormat.addInputPath(job1, new Path(conf.get("input")));

4. DBInputFormat：从数据库中获取数据
   
job.setInputFormatClass(DBInputFormat.class); // 设置为DB输入类型
// year = 2000是条件，表示输入year=2000的数据
DBInputFormat.setInput(job,YearStationTempDB.class,"station_tbl","year = 2000","","year","station","temperature");


自定义输入类型

思路分析：
1. 创建一个解析类继承RecordReader
2. 在解析类中完成获取数据的逻辑
3. 创建一个自定义输入类型的类继承FileInputFormat
4. 在自定义输入类型的类中重写方法createRecordReader()
5. 在该方法中创建解析类的对象并调用initialize()方法进行初始化，最后返回该对象。
6. 完成，可在其它类中调用该自定义输入类型

解析类-YearStationRecordReader

public class YearStationRecordReader extends RecordReader<YearStation, IntWritable> {
    private LineRecordReader reader = new LineRecordReader();
    private NcdcRecordParser parser = new NcdcRecordParser();
    private YearStation ys = new YearStation();
    private IntWritable tmp = new IntWritable();

    @Override
    public void initialize(InputSplit inputSplit, TaskAttemptContext taskAttemptContext) throws IOException {
        reader.initialize(inputSplit,taskAttemptContext);
    }

    @Override
    public boolean nextKeyValue() throws IOException {
        do {
            // 判断是否有下一个值
            if (!reader.nextKeyValue()) {
                return false;
            }
            // 获取并解析当前值
            Text line = reader.getCurrentValue();
            parser.parse(line.toString());
        }while (!parser.isValidTemperature()); // 如果气温返回值为false则继续循环下一个
        int year = parser.getYear();
        int tmp = parser.getAirTemperature();
        String station = parser.getStationId();
        ys.setYear(new Text(year+""));
        ys.setStationid(new Text(station));
        this.tmp.set(tmp);
        return true;
    }

    @Override
    public YearStation getCurrentKey() throws IOException, InterruptedException {
        return this.ys;
    }

    @Override
    public IntWritable getCurrentValue() throws IOException, InterruptedException {
        return this.tmp;
    }

    @Override
    public float getProgress() throws IOException, InterruptedException {
        return reader.getProgress();
    }

    @Override
    public void close() throws IOException {
        reader.close();
    }
}

自定义输入类型类-YearStationInputFormat

// 利用FileInputFormat数据分片功能，实现自定义输入类型
public class YearStationInputFormat extends FileInputFormat<YearStation, IntWritable> {
    @Override
    public RecordReader createRecordReader(InputSplit inputSplit, TaskAttemptContext taskAttemptContext)
            throws IOException {
        YearStationRecordReader reader = new YearStationRecordReader();
        reader.initialize(inputSplit,taskAttemptContext);
        return reader;
    }
}

复合类型-YearStation

public class YearStation implements WritableComparable<YearStation> {
    private Text year = new Text(); // 年份
    private Text stationid = new Text(); //气象站id

    public YearStation() {
    }

    public YearStation(Text year, Text stationid) {
        this.year = new Text(year.toString());
        this.stationid = new Text(stationid.toString());
    }

    public YearStation(String year, String stationid) {
        this.year = new Text(year);
        this.stationid = new Text(stationid);
    }

    @Override
    public int compareTo(YearStation o) {
        return this.year.compareTo(o.year)==0 ? this.stationid.compareTo(o.stationid) : this.year.compareTo(o.year);
    }

    @Override
    public void write(DataOutput dataOutput) throws IOException {
        year.write(dataOutput);
        stationid.write(dataOutput);
    }

    @Override
    public void readFields(DataInput dataInput) throws IOException {
        year.readFields(dataInput);
        stationid.readFields(dataInput);
    }

    public Text getYear() {
        return year;
    }

    public void setYear(Text year) {
        this.year = new Text(year.toString());
    }

    public Text getStationid() {
        return stationid;
    }

    public void setStationid(Text stationid) {
        this.stationid = new Text(stationid.toString());
    }

    @Override
    public String toString() {
        return year.toString()+"\t"+stationid.toString();
    }
}

测试自定义类型-MaxTmpByYearStationMR

public class MaxTmpByYearStationMR extends Configured implements Tool {
    public static class MTBYSMapper extends Mapper<YearStation, IntWritable, YearStation,IntWritable> {
        @Override
        protected void map(YearStation key, IntWritable value, Context context) throws IOException, InterruptedException {
            context.write(key,value);
        }
    }
    public static class MTBYSReducer extends Reducer<YearStation, IntWritable, Text, IntWritable> {
        @Override
        protected void reduce(YearStation key, Iterable<IntWritable> values, Context context) throws IOException, InterruptedException {
            Optional<Integer> max = StreamSupport.stream(values.spliterator(), false)
                    .map(e -> e.get()).reduce((x, y) -> Math.max(x, y));
            context.write(new Text(key.toString()), new IntWritable(max.get()));
        }
    }

    @Override
    public int run(String[] strings) throws Exception {
        Configuration conf = getConf();
        Job job = Job.getInstance(conf, "MaxTmpByYS_xj");
        job.setJarByClass(this.getClass());

        job.setMapperClass(MTBYSMapper.class);
        job.setMapOutputKeyClass(YearStation.class);
        job.setMapOutputValueClass(IntWritable.class);

        job.setReducerClass(MTBYSReducer.class);
        job.setOutputKeyClass(Text.class);
        job.setOutputValueClass(IntWritable.class);

        job.setInputFormatClass(YearStationInputFormat.class);
        job.setOutputFormatClass(TextOutputFormat.class);

        YearStationInputFormat.addInputPath(job,new Path(conf.get("inpath")));
        TextOutputFormat.setOutputPath(job,new Path(conf.get("outpath")));
        return job.waitForCompletion(true)? 0 : 1;
    }
    public static void main(String[] args) throws Exception{
        ToolRunner.run(new MaxTmpByYearStationMR(),args);
    }
}

JobControl

简述：
如果MapReduce中需要用到多个job，而且多个job之间需要设置一些依赖关系，比如Job3需要依赖于Job2，Job2依赖于Job1，这就要用到JobControl。

代码实例：

Job getSDJob = Job.getInstance(conf, "get_sd_job_xj");
getSDJob.setJarByClass(GetSimilarityDegree.class);
// 3 为任务装配mapper类
getSDJob.setMapperClass(GetSimilarityDegree.GSDMapper.class);
getSDJob.setMapOutputKeyClass(Text.class);
getSDJob.setMapOutputValueClass(DoubleWritable.class);
// 5 配置数据输入路径
TextInputFormat.addInputPath(getSDJob, new Path("src/train_bin"));
// 6 配置结果输出路径
TextOutputFormat.setOutputPath(getSDJob, new Path("src/name_sd"));

Job sortBySDJob = Job.getInstance(conf, "sortBySDJob");
sortBySDJob.setJarByClass(SortedByDegree.class);
// 3 为任务装配mapper类
sortBySDJob.setMapperClass(SortedByDegree.SBDMapper.class);
sortBySDJob.setMapOutputKeyClass(TagDegree.class);
sortBySDJob.setMapOutputValueClass(NullWritable.class);
// 4 为任务装配reducer类
sortBySDJob.setReducerClass(SortedByDegree.SBDReducer.class);
sortBySDJob.setOutputKeyClass(Text.class);
sortBySDJob.setOutputValueClass(DoubleWritable.class);
// 5 配置数据输入路径
TextInputFormat.addInputPath(sortBySDJob, new Path("src/name_sd"));
// 6 配置结果输出路径
TextOutputFormat.setOutputPath(sortBySDJob, new Path("src/name_sd_sorted"));

Job getFKJob = Job.getInstance(conf, "getFKJob");
getFKJob.setJarByClass(GetFirstK.class);
// 3 为任务装配mapper类
getFKJob.setMapperClass(GetFirstK.GFKMapper.class);
getFKJob.setMapOutputKeyClass(TagDegree.class);
getFKJob.setMapOutputValueClass(IntWritable.class);
// 4 为任务装配reducer类
getFKJob.setReducerClass(GetFirstK.GFKReducer.class);
getFKJob.setOutputKeyClass(Text.class);
getFKJob.setOutputValueClass(Text.class);
// 5 配置数据输入路径
TextInputFormat.addInputPath(getFKJob, new Path("src/name_sd_sorted"));
// 6 配置结果输出路径
TextOutputFormat.setOutputPath(getFKJob, new Path("src/gfk_res"));
getFKJob.setGroupingComparatorClass(GFKGroupComparator.class);

Job getLRJob = Job.getInstance(conf, "getLRJob");
getLRJob.setJarByClass(GetLastResult.class);
// 3 为任务装配mapper类
getLRJob.setMapperClass(GetLastResult.GLRMapper.class);
getLRJob.setMapOutputKeyClass(Text.class);
getLRJob.setMapOutputValueClass(TagAvgNum.class);
// 4 为任务装配reducer类
getLRJob.setReducerClass(GetLastResult.GLRReducer.class);
getLRJob.setOutputKeyClass(Text.class);
getLRJob.setOutputValueClass(NullWritable.class);
// 5 配置数据输入路径
TextInputFormat.addInputPath(getLRJob, new Path("src/gfk_res"));
// 6 配置结果输出路径
TextOutputFormat.setOutputPath(getLRJob, new Path("src/last_res"));

ControlledJob getSD = new ControlledJob(getSDJob.getConfiguration());
ControlledJob sortBySD = new ControlledJob(sortBySDJob.getConfiguration());
ControlledJob getFK = new ControlledJob(getFKJob.getConfiguration());
ControlledJob getLR = new ControlledJob(getLRJob.getConfiguration());
// 添加依赖
getLR.addDependingJob(getFK);
getFK.addDependingJob(sortBySD);
sortBySD.addDependingJob(getSD);

JobControl con = new JobControl("test");
con.addJob(getSD);
con.addJob(sortBySD);
con.addJob(getFK);
con.addJob(getLR);

Thread t = new Thread(con);
t.start();

while (true) {
    if (con.allFinished()) {
        System.out.println("图片识别完毕，请查看结果");
        System.exit(0);
    }
}