Example usage for org.apache.hadoop.mapred.lib MultipleOutputs addNamedOutput

List of usage examples for org.apache.hadoop.mapred.lib MultipleOutputs addNamedOutput

  1. HOME
  2. Java
  3. org.apache.hadoop
  4. org.apache.hadoop.mapred.lib.*
  5. MultipleOutputs
  6. addNamedOutput

Introduction

In this page you can find the example usage for org.apache.hadoop.mapred.lib MultipleOutputs addNamedOutput.

Prototype

public static void addNamedOutput(JobConf conf, String namedOutput,
        Class<? extends OutputFormat> outputFormatClass, Class<?> keyClass, Class<?> valueClass)

Source Link

Document

Adds a named output for the job.

Usage

From source file:com.ibm.bi.dml.runtime.matrix.mapred.MRJobConfiguration.java

License:Open Source License

public static void setUpMultipleOutputs(JobConf job, byte[] resultIndexes, byte[] resultDimsUnknwon,
        String[] outputs, OutputInfo[] outputInfos, boolean inBlockRepresentation, boolean mayContainCtable)
        throws Exception {
    if (resultIndexes.length != outputs.length)
        throw new Exception("number of outputs and result indexes does not match");
    if (outputs.length != outputInfos.length)
        throw new Exception("number of outputs and outputInfos indexes does not match");

    job.set(RESULT_INDEXES_CONFIG, MRJobConfiguration.getIndexesString(resultIndexes));
    job.set(RESULT_DIMS_UNKNOWN_CONFIG, MRJobConfiguration.getIndexesString(resultDimsUnknwon));
    job.setStrings(OUTPUT_MATRICES_DIRS_CONFIG, outputs);
    job.setOutputCommitter(MultipleOutputCommitter.class);

    for (int i = 0; i < outputs.length; i++) {
        MapReduceTool.deleteFileIfExistOnHDFS(new Path(outputs[i]), job);
        if (mayContainCtable && resultDimsUnknwon[i] == (byte) 1) {
            setOutputInfo(job, i, outputInfos[i], false);
        } else {//from w w w .  j a va2s . c  om
            setOutputInfo(job, i, outputInfos[i], inBlockRepresentation);
        }
        MultipleOutputs.addNamedOutput(job, Integer.toString(i), outputInfos[i].outputFormatClass,
                outputInfos[i].outputKeyClass, outputInfos[i].outputValueClass);
    }
    job.setOutputFormat(NullOutputFormat.class);

    // configure temp output
    Path tempOutputPath = new Path(constructTempOutputFilename());
    FileOutputFormat.setOutputPath(job, tempOutputPath);
    MapReduceTool.deleteFileIfExistOnHDFS(tempOutputPath, job);

}

From source file:com.intel.hadoop.graphbuilder.idnormalize.mapreduce.SortDictMR.java

License:Open Source License

/**
 * @param inputpath//w  ww .j  a va 2  s .  co  m
 *          the path to a rawId to newId dictionary.
 * @param outputpath
 *          the path of output directory.
 * @throws IOException
 */
public void run(String inputpath, String outputpath) throws IOException {

    JobConf conf = new JobConf(SortDictMR.class);

    conf.setOutputKeyClass(Text.class);
    conf.setOutputValueClass(Text.class);

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

    conf.setMapperClass(SortDictMapper.class);
    conf.setReducerClass(SortDictReducer.class);

    conf.setInputFormat(TextInputFormat.class);
    conf.setOutputFormat(TextOutputFormat.class);

    conf.setBoolean("hashRawVid", hashRawVid);
    conf.setInt("numChunks", numChunks);
    conf.set("VidParser", vidparser.getClass().getName());

    String outprefix = "vidhashmap";
    for (int i = 0; i < numChunks; i++) {
        MultipleOutputs.addNamedOutput(conf, outprefix + i, TextOutputFormat.class, Text.class, Text.class);
    }

    FileInputFormat.setInputPaths(conf, new Path(inputpath));
    FileOutputFormat.setOutputPath(conf, new Path(outputpath));

    LOG.info("========== Job: Partition the map of rawid -> id ===========");
    LOG.info("Input = " + inputpath);
    LOG.info("Output = " + outputpath);
    LOG.info("======================================================");
    if (hashRawVid)
        LOG.info("Partition on rawId.");
    else
        LOG.info("Partition on newId");
    LOG.debug("numChunks = " + numChunks);
    LOG.debug("VidParser = " + vidparser.getClass().getName());
    JobClient.runJob(conf);
    LOG.info("======================= Done ==========================\n");
}

From source file:gr.forth.ics.isl.grouprecsmr.multiuser.MultiUserMain.java

public static void main(String[] args) {
    //paths and input handling
    Path inputRatingsPath = new Path(args[0]); //movieid, userid, rating (text files)
    Path job1OutputPath = new Path("/user/hduser/partialResults");
    Path partialDistancesPath = new Path("/user/hduser/partialResults/part-*"); //member_nonMember \t partialDistance (sequence files)
    Path candidateMoviesPath = new Path("/user/hduser/partialResults/candidateMovies-*"); //candidateMovieId, nonMemberUserId_rating (text files)
    Path userSimilaritiesPath = new Path("/user/hduser/userSimilarities"); //similarity of each group member to his friends (text files)
    Path finalScoresPath = new Path(args[1]); //movieId \t outputScore

    int numReduceTasks = 56; //defaultValue
    if (args.length == 3) {
        numReduceTasks = Integer.parseInt(args[2]);
    }/*from w ww.  j ava2  s  .c  o m*/

    final float friendsSimThresh = 0.8f;

    String groupFilePath = "/user/hduser/group.txt"; //one-line csv file with user ids (text file)

    if (args.length < 2 || args.length > 3) {
        System.err.println(
                "Incorrect input. Example usage: hadoop jar ~/GroupRecs/MultiUser.jar inputPath outputPath [numReduceTasks]");
        return;
    }

    //JOB 1//
    JobClient client = new JobClient();
    JobConf conf = new JobConf(gr.forth.ics.isl.grouprecsmr.multiuser.MultiUserMain.class);

    try {
        FileSystem fs = FileSystem.get(conf);
        if (fs.exists(job1OutputPath)) {
            fs.delete(job1OutputPath, true);
        }
        if (fs.exists(userSimilaritiesPath)) {
            fs.delete(userSimilaritiesPath, true);
        }
        if (fs.exists(finalScoresPath)) {
            fs.delete(finalScoresPath, true);
        }
    } catch (IOException ex) {
        Logger.getLogger(MultiUserMain.class.getName()).log(Level.SEVERE, null, ex);
    }

    conf.setJobName("Multi-user approach - Job 1");
    System.out.println("Starting Job 1 (Multi-user approach)...");

    conf.setMapOutputKeyClass(VIntWritable.class);
    conf.setMapOutputValueClass(Text.class);

    conf.setOutputKeyClass(Text.class);
    conf.setOutputValueClass(ByteWritable.class);

    conf.setInputFormat(TextInputFormat.class);
    //conf.setOutputFormat(TextOutputFormat.class);
    conf.setOutputFormat(SequenceFileOutputFormat.class);
    SequenceFileOutputFormat.setOutputCompressionType(conf, SequenceFile.CompressionType.BLOCK);

    FileInputFormat.setInputPaths(conf, inputRatingsPath); //user ratings
    FileOutputFormat.setOutputPath(conf, job1OutputPath); //partial distances

    MultipleOutputs.addNamedOutput(conf, "candidateMovies", SequenceFileOutputFormat.class, VIntWritable.class,
            Text.class); //movieId, userId_rating

    conf.setMapperClass(gr.forth.ics.isl.grouprecsmr.job1.Job1Mapper.class);
    conf.setReducerClass(gr.forth.ics.isl.grouprecsmr.job1.Job1Reducer.class);

    conf.setNumReduceTasks(numReduceTasks);

    try {
        DistributedCache.addCacheFile(new URI(groupFilePath), conf); // group   
    } catch (URISyntaxException e1) {
        System.err.println(e1.toString());
    }

    conf.setInt("mapred.task.timeout", 6000000);

    client.setConf(conf);
    RunningJob job;
    try {
        job = JobClient.runJob(conf);
        job.waitForCompletion();
    } catch (Exception e) {
        System.err.println(e);
    }

    //JOB 2//
    System.out.println("Starting Job 2 (Multi-user approach)...");
    JobClient client2 = new JobClient();
    JobConf conf2 = new JobConf(gr.forth.ics.isl.grouprecsmr.multiuser.MultiUserMain.class);

    conf2.setJobName("Multi-user approach - Job 2");

    conf2.setMapOutputKeyClass(Text.class); //user pair (member_nonMember), where nonMember is in friends
    conf2.setMapOutputValueClass(ByteWritable.class);//similarity part unsquared

    conf2.setOutputKeyClass(Text.class); //user pair (member_nonMember), where nonMember is in friends
    conf2.setOutputValueClass(DoubleWritable.class);//similarity

    conf2.setInputFormat(SequenceFileInputFormat.class);
    //conf2.setInputFormat(TextInputFormat.class);
    conf2.setOutputFormat(TextOutputFormat.class);
    //conf2.setOutputFormat(SequenceFileOutputFormat.class);
    //SequenceFileOutputFormat.setOutputCompressionType(conf2, SequenceFile.CompressionType.BLOCK);

    FileInputFormat.setInputPaths(conf2, partialDistancesPath); //Job 1 output
    FileOutputFormat.setOutputPath(conf2, userSimilaritiesPath); //Job 2 output (similarity of each group member to his friends)

    conf2.setMapperClass(IdentityMapper.class);
    conf2.setReducerClass(gr.forth.ics.isl.grouprecsmr.job2.Job2ReducerMulti.class);

    int numSimilaritiesPartitions = numReduceTasks;
    conf2.setNumReduceTasks(numSimilaritiesPartitions);

    conf2.setFloat("friendsSimThreshold", friendsSimThresh);

    conf2.setInt("mapred.task.timeout", 6000000);
    conf2.set("io.sort.mb", "500");

    client2.setConf(conf2);
    RunningJob job2;
    try {
        job2 = JobClient.runJob(conf2);
        job2.waitForCompletion();
    } catch (Exception e) {
        System.err.println(e);
    }

    //JOB 3//
    System.out.println("Starting Job 3 (Multi-user approach)...");
    JobClient client3 = new JobClient();
    JobConf conf3 = new JobConf(gr.forth.ics.isl.grouprecsmr.multiuser.MultiUserMain.class);

    conf3.setJobName("Multi-user approach - Job 3");

    conf3.setMapOutputKeyClass(VIntWritable.class);
    conf3.setMapOutputValueClass(Text.class);

    conf3.setOutputKeyClass(VIntWritable.class);
    conf3.setOutputValueClass(DoubleWritable.class);

    conf3.setInputFormat(SequenceFileInputFormat.class);
    //conf3.setInputFormat(TextInputFormat.class);
    conf3.setOutputFormat(TextOutputFormat.class);
    //conf3.setOutputFormat(SequenceFileOutputFormat.class);
    //SequenceFileOutputFormat.setOutputCompressionType(conf3,SequenceFile.CompressionType.BLOCK);

    try {
        DistributedCache.addCacheFile(new URI(groupFilePath), conf3);
    } catch (URISyntaxException ex) {
        System.err.println("Could not add group file to distributed cache. " + ex);
    }
    for (int i = 0; i < numSimilaritiesPartitions; i++) {
        String reduceId = String.format("%05d", i); //5-digit int with leading
        try {
            DistributedCache.addCacheFile(new URI(userSimilaritiesPath.toString() + "/part-" + reduceId),
                    conf3);
        } catch (URISyntaxException ex) {
            System.err.println("Could not add similarities files to distributed cache. " + ex);
        }

    }

    FileInputFormat.setInputPaths(conf3, candidateMoviesPath); //Job 1 output (candidate movies)
    FileOutputFormat.setOutputPath(conf3, finalScoresPath); //Job 3 output (movie \t outputScore)

    //        conf3.setMapperClass(IdentityMapper.class);      
    conf3.setMapperClass(gr.forth.ics.isl.grouprecsmr.job3.Job3MapperMulti.class); //filtering out ratings from non-Friends
    conf3.setReducerClass(gr.forth.ics.isl.grouprecsmr.job3.Job3ReducerMulti.class);

    conf3.setInt("mapred.task.timeout", 6000000);
    conf3.set("io.sort.mb", "500");

    conf3.setNumReduceTasks(numReduceTasks);

    client3.setConf(conf3);
    RunningJob job3;
    try {
        job3 = JobClient.runJob(conf3);
        job3.waitForCompletion();
    } catch (Exception e) {
        System.err.println(e);
    }
}

From source file:nthu.scopelab.tsqr.ssvd.BtJob.java

License:Apache License

public static void run(Configuration conf, Path[] inputPath, Path btPath, String qrfPath, int k, int p,
        int outerBlockHeight, int reduceTasks, boolean outputBBtProducts, String reduceSchedule, int mis)
        throws Exception {
    boolean outputQ = true;

    String stages[] = reduceSchedule.split(",");

    JobConf job = new JobConf(conf, BtJob.class);
    job.setInputFormat(SequenceFileInputFormat.class);
    job.setOutputFormat(SequenceFileOutputFormat.class);
    job.setInt(SCHEDULE_NUM, stages.length);
    job.setInt(PROP_OUTER_PROD_BLOCK_HEIGHT, outerBlockHeight);
    job.setInt(QJob.PROP_K, k);//from  ww  w.java 2 s .  c  o  m
    job.setInt(QJob.PROP_P, p);
    job.setBoolean(QmultiplyJob.OUTPUT_Q, outputQ);
    job.setBoolean(PROP_OUPTUT_BBT_PRODUCTS, outputBBtProducts);
    job.set(QmultiplyJob.QRF_DIR, qrfPath);
    FileSystem.get(job).delete(btPath, true);

    FileOutputFormat.setOutputPath(job, btPath);

    FileOutputFormat.setCompressOutput(job, true);
    FileOutputFormat.setOutputCompressorClass(job, DefaultCodec.class);
    SequenceFileOutputFormat.setOutputCompressionType(job, CompressionType.BLOCK);

    job.setJobName("BtJob");

    job.setMapOutputKeyClass(IntWritable.class);
    job.setMapOutputValueClass(SparseRowBlockWritable.class);
    job.setOutputKeyClass(IntWritable.class);
    //job.setOutputValueClass(SparseRowBlockWritable.class);
    job.setOutputValueClass(VectorWritable.class);

    job.setMapperClass(BtMapper.class);
    job.setCombinerClass(OuterProductCombiner.class);
    job.setReducerClass(OuterProductReducer.class);

    fileGather fgather = new fileGather(inputPath, "", FileSystem.get(job));
    mis = Checker.checkMis(mis, fgather.getInputSize(), FileSystem.get(job));
    job.setNumMapTasks(fgather.recNumMapTasks(mis));

    //job.setNumReduceTasks(0);
    job.setNumReduceTasks(reduceTasks);

    FileInputFormat.setInputPaths(job, inputPath);

    if (outputQ) {
        MultipleOutputs.addNamedOutput(job, QmultiplyJob.Q_MAT, SequenceFileOutputFormat.class,
                IntWritable.class, LMatrixWritable.class);
    }
    if (outputBBtProducts) {
        MultipleOutputs.addNamedOutput(job, OUTPUT_BBT, SequenceFileOutputFormat.class, IntWritable.class,
                VectorWritable.class);
    }
    RunningJob rj = JobClient.runJob(job);
    System.out.println("Btjob Job ID: " + rj.getJobID().toString());
}

From source file:nthu.scopelab.tsqr.ssvd.itBtJob.java

License:Apache License

public static void run(Configuration conf, Path[] inputPath, Path btPath, String qrfPath, int k, int p,
        int outerBlockHeight, int reduceTasks, boolean outputBBtProducts, String reduceSchedule, int mis)
        throws Exception {
    boolean outputQ = true;

    String stages[] = reduceSchedule.split(",");

    JobConf job = new JobConf(conf, itBtJob.class);
    job.setInputFormat(SequenceFileInputFormat.class);
    job.setOutputFormat(SequenceFileOutputFormat.class);
    job.setInt(SCHEDULE_NUM, stages.length);
    job.setInt(PROP_OUTER_PROD_BLOCK_HEIGHT, outerBlockHeight);
    job.setInt(QJob.PROP_K, k);/*from  w w w.j a  v a 2  s .  com*/
    job.setInt(QJob.PROP_P, p);
    job.setBoolean(QmultiplyJob.OUTPUT_Q, outputQ);
    job.setBoolean(PROP_OUPTUT_BBT_PRODUCTS, outputBBtProducts);
    job.set(QmultiplyJob.QRF_DIR, qrfPath);
    FileSystem.get(job).delete(btPath, true);

    FileOutputFormat.setOutputPath(job, btPath);

    FileOutputFormat.setCompressOutput(job, true);
    FileOutputFormat.setOutputCompressorClass(job, DefaultCodec.class);
    SequenceFileOutputFormat.setOutputCompressionType(job, CompressionType.BLOCK);

    job.setJobName("itBtJob");

    job.setMapOutputKeyClass(IntWritable.class);
    job.setMapOutputValueClass(SparseRowBlockWritable.class);
    job.setOutputKeyClass(IntWritable.class);
    //job.setOutputValueClass(SparseRowBlockWritable.class);
    job.setOutputValueClass(VectorWritable.class);

    job.setMapperClass(BtMapper.class);
    job.setCombinerClass(OuterProductCombiner.class);
    job.setReducerClass(OuterProductReducer.class);

    fileGather fgather = new fileGather(inputPath, "", FileSystem.get(job));
    mis = Checker.checkMis(mis, fgather.getInputSize(), FileSystem.get(job));
    job.setNumMapTasks(fgather.recNumMapTasks(mis));

    //job.setNumReduceTasks(0);
    job.setNumReduceTasks(reduceTasks);

    FileInputFormat.setInputPaths(job, inputPath);

    if (outputQ) {
        MultipleOutputs.addNamedOutput(job, QmultiplyJob.Q_MAT, SequenceFileOutputFormat.class,
                IntWritable.class, LMatrixWritable.class);
    }
    if (outputBBtProducts) {
        MultipleOutputs.addNamedOutput(job, OUTPUT_BBT, SequenceFileOutputFormat.class, IntWritable.class,
                VectorWritable.class);
    }
    RunningJob rj = JobClient.runJob(job);
    System.out.println("itBtJob Job ID: " + rj.getJobID().toString());
}

From source file:nthu.scopelab.tsqr.ssvd.itQJob.java

License:Apache License

public static void run(Configuration conf, Path[] inputPaths, String outputPath, String reduceSchedule, int k,
        int p, long seed, int mis) throws ClassNotFoundException, InterruptedException, IOException {

    String stages[] = reduceSchedule.split(",");
    String rinput = "";
    String routput = outputPath + "/iter-r-";

    for (int i = 0; i < stages.length; i++) {
        String thenumber = Integer.toString(i + 1);
        JobConf job = new JobConf(conf, itQJob.class);
        job.setJobName("itQ-job-" + thenumber);
        job.setInputFormat(SequenceFileInputFormat.class);
        job.setOutputFormat(SequenceFileOutputFormat.class);

        if (i == 0)
            job.setMapperClass(QMapper.class);
        else/*from   ww w.j  av a2  s.c  o  m*/
            job.setMapperClass(IdentityMapper.class);

        job.setReducerClass(QReducer.class);
        job.setOutputKeyClass(IntWritable.class);
        job.setOutputValueClass(LMatrixWritable.class);

        FileSystem fs = FileSystem.get(job);
        Path Paths[];
        fileGather fgather = null;
        if (i == 0)
            fgather = new fileGather(inputPaths, "part", fs);
        else
            fgather = new fileGather(new Path(rinput), "part", fs);
        Paths = fgather.getPaths();
        mis = Checker.checkMis(mis, fgather.getInputSize(), fs);
        job.setNumMapTasks(fgather.recNumMapTasks(mis));

        job.setNumReduceTasks(Integer.parseInt(stages[i]));

        job.setInt(QRFirstJob.COLUMN_SIZE, k + p);
        job.setLong(PROP_OMEGA_SEED, seed);
        job.setInt(PROP_K, k);
        job.setInt(PROP_P, p);

        fs.delete(new Path(routput + thenumber), true);

        FileInputFormat.setInputPaths(job, Paths);

        FileOutputFormat.setOutputPath(job, new Path(routput + thenumber));

        //FileOutputFormat.setCompressOutput(job, true);
        //FileOutputFormat.setOutputCompressorClass(job, DefaultCodec.class);
        //SequenceFileOutputFormat.setOutputCompressionType(job,CompressionType.BLOCK);
        //output first level Q
        MultipleOutputs.addNamedOutput(job, QF_MAT, SequenceFileOutputFormat.class, IntWritable.class,
                LMatrixWritable.class);

        RunningJob rj = JobClient.runJob(job);
        System.out.println("itQJob Job ID: " + rj.getJobID().toString());
        rinput = routput + thenumber;
    }
}

From source file:nthu.scopelab.tsqr.ssvd.QJob.java

License:Apache License

public static void run(Configuration conf, Path[] inputPaths, String outputPath, String reduceSchedule, int k,
        int p, long seed, int mis) throws ClassNotFoundException, InterruptedException, IOException {

    String stages[] = reduceSchedule.split(",");
    String rinput = "";
    String routput = outputPath + "/iter-r-";

    for (int i = 0; i < stages.length; i++) {
        String thenumber = Integer.toString(i + 1);
        JobConf job = new JobConf(conf, QJob.class);
        job.setJobName("Q-job-" + thenumber);
        job.setInputFormat(SequenceFileInputFormat.class);
        job.setOutputFormat(SequenceFileOutputFormat.class);

        if (i == 0)
            job.setMapperClass(QMapper.class);
        else/*from   w w  w .j a v a  2s .  com*/
            job.setMapperClass(IdentityMapper.class);

        job.setReducerClass(QReducer.class);
        job.setOutputKeyClass(IntWritable.class);
        job.setOutputValueClass(LMatrixWritable.class);

        FileSystem fs = FileSystem.get(job);
        Path Paths[];
        fileGather fgather = null;
        if (i == 0)
            fgather = new fileGather(inputPaths, "part", fs);
        else
            fgather = new fileGather(new Path(rinput), "part", fs);
        Paths = fgather.getPaths();
        mis = Checker.checkMis(mis, fgather.getInputSize(), fs);
        job.setNumMapTasks(fgather.recNumMapTasks(mis));

        job.setNumReduceTasks(Integer.parseInt(stages[i]));

        job.setInt(QRFirstJob.COLUMN_SIZE, k + p);
        job.setLong(PROP_OMEGA_SEED, seed);
        job.setInt(PROP_K, k);
        job.setInt(PROP_P, p);

        fs.delete(new Path(routput + thenumber), true);

        FileInputFormat.setInputPaths(job, Paths);

        FileOutputFormat.setOutputPath(job, new Path(routput + thenumber));

        //FileOutputFormat.setCompressOutput(job, true);
        //FileOutputFormat.setOutputCompressorClass(job, DefaultCodec.class);
        //SequenceFileOutputFormat.setOutputCompressionType(job,CompressionType.BLOCK);
        //output first level Q
        MultipleOutputs.addNamedOutput(job, QF_MAT, SequenceFileOutputFormat.class, IntWritable.class,
                LMatrixWritable.class);

        RunningJob rj = JobClient.runJob(job);
        System.out.println("QJob Job ID: " + rj.getJobID().toString());
        rinput = routput + thenumber;
    }
}

From source file:org.apache.mahout.clustering.lda.cvb.CVB0PriorMapper.java

License:Apache License

public static void main(String[] args) throws IOException {
    JobConf conf = new JobConf();
    Job job = new Job(conf);

    MultipleOutputs.addNamedOutput(conf, DOCTOPIC_OUT, SequenceFileOutputFormat.class, IntWritable.class,
            VectorWritable.class);

    Path aPath = null;/*from ww w  .  j a v a 2  s.c  o m*/
    Path bPath = null;
    conf.setInputFormat(CompositeInputFormat.class);
    conf.set("mapred.join.expr",
            CompositeInputFormat.compose("inner", SequenceFileInputFormat.class, aPath, bPath));
}

From source file:org.apache.mahout.math.hadoop.stochasticsvd.BtJob.java

License:Apache License

public static void run(Configuration conf, Path[] inputPathA, Path inputPathQJob, Path xiPath, Path outputPath,
        int minSplitSize, int k, int p, int btBlockHeight, int numReduceTasks, boolean broadcast,
        Class<? extends Writable> labelClass, boolean outputBBtProducts)
        throws ClassNotFoundException, InterruptedException, IOException {

    JobConf oldApiJob = new JobConf(conf);

    MultipleOutputs.addNamedOutput(oldApiJob, OUTPUT_Q, org.apache.hadoop.mapred.SequenceFileOutputFormat.class,
            labelClass, VectorWritable.class);

    if (outputBBtProducts) {
        MultipleOutputs.addNamedOutput(oldApiJob, OUTPUT_BBT,
                org.apache.hadoop.mapred.SequenceFileOutputFormat.class, IntWritable.class,
                VectorWritable.class);
        /*/*from  w  w  w  . j av a2 s. c o m*/
         * MAHOUT-1067: if we are asked to output BBT products then named vector
         * names should be propagated to Q too so that UJob could pick them up
         * from there.
         */
        oldApiJob.setBoolean(PROP_NV, true);
    }
    if (xiPath != null) {
        // compute pca -related stuff as well
        MultipleOutputs.addNamedOutput(oldApiJob, OUTPUT_SQ,
                org.apache.hadoop.mapred.SequenceFileOutputFormat.class, IntWritable.class,
                VectorWritable.class);
        MultipleOutputs.addNamedOutput(oldApiJob, OUTPUT_SB,
                org.apache.hadoop.mapred.SequenceFileOutputFormat.class, IntWritable.class,
                VectorWritable.class);
    }

    /*
     * HACK: we use old api multiple outputs since they are not available in the
     * new api of either 0.20.2 or 0.20.203 but wrap it into a new api job so we
     * can use new api interfaces.
     */

    Job job = new Job(oldApiJob);
    job.setJobName("Bt-job");
    job.setJarByClass(BtJob.class);

    job.setInputFormatClass(SequenceFileInputFormat.class);
    job.setOutputFormatClass(SequenceFileOutputFormat.class);
    FileInputFormat.setInputPaths(job, inputPathA);
    if (minSplitSize > 0) {
        FileInputFormat.setMinInputSplitSize(job, minSplitSize);
    }
    FileOutputFormat.setOutputPath(job, outputPath);

    // WARN: tight hadoop integration here:
    job.getConfiguration().set("mapreduce.output.basename", OUTPUT_BT);

    FileOutputFormat.setOutputCompressorClass(job, DefaultCodec.class);
    SequenceFileOutputFormat.setOutputCompressionType(job, CompressionType.BLOCK);

    job.setMapOutputKeyClass(LongWritable.class);
    job.setMapOutputValueClass(SparseRowBlockWritable.class);

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

    job.setMapperClass(BtMapper.class);
    job.setCombinerClass(OuterProductCombiner.class);
    job.setReducerClass(OuterProductReducer.class);

    job.getConfiguration().setInt(QJob.PROP_K, k);
    job.getConfiguration().setInt(QJob.PROP_P, p);
    job.getConfiguration().set(PROP_QJOB_PATH, inputPathQJob.toString());
    job.getConfiguration().setBoolean(PROP_OUPTUT_BBT_PRODUCTS, outputBBtProducts);
    job.getConfiguration().setInt(PROP_OUTER_PROD_BLOCK_HEIGHT, btBlockHeight);

    job.setNumReduceTasks(numReduceTasks);

    /*
     * PCA-related options, MAHOUT-817
     */
    if (xiPath != null) {
        job.getConfiguration().set(PROP_XI_PATH, xiPath.toString());
    }

    /*
     * we can broadhast Rhat files since all of them are reuqired by each job,
     * but not Q files which correspond to splits of A (so each split of A will
     * require only particular Q file, each time different one).
     */

    if (broadcast) {
        job.getConfiguration().set(PROP_RHAT_BROADCAST, "y");

        FileSystem fs = FileSystem.get(inputPathQJob.toUri(), conf);
        FileStatus[] fstats = fs.globStatus(new Path(inputPathQJob, QJob.OUTPUT_RHAT + "-*"));
        if (fstats != null) {
            for (FileStatus fstat : fstats) {
                /*
                 * new api is not enabled yet in our dependencies at this time, still
                 * using deprecated one
                 */
                DistributedCache.addCacheFile(fstat.getPath().toUri(), job.getConfiguration());
            }
        }
    }

    job.submit();
    job.waitForCompletion(false);

    if (!job.isSuccessful()) {
        throw new IOException("Bt job unsuccessful.");
    }
}

From source file:org.apache.mahout.math.hadoop.stochasticsvd.QJob.java

License:Apache License

public static void run(Configuration conf, Path[] inputPaths, Path sbPath, Path outputPath, int aBlockRows,
        int minSplitSize, int k, int p, long seed, int numReduceTasks)
        throws ClassNotFoundException, InterruptedException, IOException {

    JobConf oldApiJob = new JobConf(conf);
    MultipleOutputs.addNamedOutput(oldApiJob, OUTPUT_QHAT,
            org.apache.hadoop.mapred.SequenceFileOutputFormat.class, SplitPartitionedWritable.class,
            DenseBlockWritable.class);
    MultipleOutputs.addNamedOutput(oldApiJob, OUTPUT_RHAT,
            org.apache.hadoop.mapred.SequenceFileOutputFormat.class, SplitPartitionedWritable.class,
            VectorWritable.class);

    Job job = new Job(oldApiJob);
    job.setJobName("Q-job");
    job.setJarByClass(QJob.class);

    job.setInputFormatClass(SequenceFileInputFormat.class);
    FileInputFormat.setInputPaths(job, inputPaths);
    if (minSplitSize > 0) {
        FileInputFormat.setMinInputSplitSize(job, minSplitSize);
    }//from   w  w w .  j a va 2 s.c  om

    FileOutputFormat.setOutputPath(job, outputPath);

    FileOutputFormat.setCompressOutput(job, true);
    FileOutputFormat.setOutputCompressorClass(job, DefaultCodec.class);
    SequenceFileOutputFormat.setOutputCompressionType(job, CompressionType.BLOCK);

    job.setMapOutputKeyClass(SplitPartitionedWritable.class);
    job.setMapOutputValueClass(VectorWritable.class);

    job.setOutputKeyClass(SplitPartitionedWritable.class);
    job.setOutputValueClass(VectorWritable.class);

    job.setMapperClass(QMapper.class);

    job.getConfiguration().setInt(PROP_AROWBLOCK_SIZE, aBlockRows);
    job.getConfiguration().setLong(PROP_OMEGA_SEED, seed);
    job.getConfiguration().setInt(PROP_K, k);
    job.getConfiguration().setInt(PROP_P, p);
    if (sbPath != null) {
        job.getConfiguration().set(PROP_SB_PATH, sbPath.toString());
    }

    /*
     * number of reduce tasks doesn't matter. we don't actually send anything to
     * reducers.
     */

    job.setNumReduceTasks(0 /* numReduceTasks */);

    job.submit();
    job.waitForCompletion(false);

    if (!job.isSuccessful()) {
        throw new IOException("Q job unsuccessful.");
    }

}