Example usage for org.apache.hadoop.io BytesWritable getBytes

List of usage examples for org.apache.hadoop.io BytesWritable getBytes

Introduction

In this page you can find the example usage for org.apache.hadoop.io BytesWritable getBytes.

Prototype

@Override
public byte[] getBytes() 

Source Link

Document

Get the data backing the BytesWritable.

Usage

From source file:com.bizosys.hsearch.kv.impl.bytescookernew.IndexFieldString.java

License:Apache License

public static byte[] cook(Iterable<BytesWritable> values, final boolean isRepetable, final boolean isCompressed)
        throws IOException {
    if (isRepetable) {

        final HSearchTableKVStringInverted table = new HSearchTableKVStringInverted(isCompressed);
        ComputeSerValues<String> compute = new ComputeSerValues<String>() {
            @Override/*w  w w .j av a2  s  .c o  m*/
            public final void visit(final int key, final String val) {
                table.put(key, val);
            }
        };

        for (BytesWritable bucketBytes : values) {

            byte[] data = bucketBytes.getBytes();
            if (null == data)
                continue;

            compute.compute(createBlankTable(), data);
        }

        return table.toBytes();

    } else {
        final HSearchTableKVString table = new HSearchTableKVString();
        ComputeSerValues<String> compute = new ComputeSerValues<String>() {
            @Override
            public final void visit(final int key, final String val) {
                table.put(key, val);
            }
        };

        for (BytesWritable bucketBytes : values) {

            byte[] data = bucketBytes.getBytes();
            if (null == data)
                continue;

            compute.compute(createBlankTable(), data);
        }

        return table.toBytes();
    }
}

From source file:com.bizosys.hsearch.kv.indexer.KVReducerBase.java

License:Apache License

public byte[] indexTextBitset(boolean skipSingle, Iterable<BytesWritable> values, boolean isAnalyzed,
        String fieldName, boolean isCompressed) throws IOException {

    byte[] finalData = null;
    int containerKey = 0;

    BitSetWrapper foundIds = new BitSetWrapper();

    for (BytesWritable idBytes : values) {
        containerKey = ByteUtil.toInt(idBytes.getBytes(), 0);
        foundIds.set(containerKey);/*from   w  ww.j  a  v a  2s  . c om*/
    }

    if (skipSingle && foundIds.cardinality() < 2)
        return null;

    if (isCompressed)
        finalData = SortedBytesBitsetCompressed.getInstanceBitset().bitSetToBytes(foundIds);
    else
        finalData = SortedBytesBitset.getInstanceBitset().bitSetToBytes(foundIds);

    if (null == finalData)
        return finalData;

    return finalData;
}

From source file:com.blackberry.logdriver.mapred.avro.AvroBlockWriterMapper.java

License:Apache License

@Override
public void map(AvroFileHeader key, BytesWritable value, OutputCollector<BytesWritable, NullWritable> output,
        Reporter reporter) throws IOException {
    byte[] valueBytes = null;

    if (header.getSyncMarker() == null) {
        LOG.info("Writing new header for new file: {}", key.toString());
        header.set(key);/*  w  w w  .  j av  a2  s. c  o m*/
        output.collect(new BytesWritable(header.toBytes()), null);
    } else {
        AvroFileHeader newHeader = key;
        if (!header.getSchema().equals(newHeader.getSchema())) {
            throw new IOException("Schemas in files do not match.");
        }
        if (!header.getCodec().equals(newHeader.getCodec())) {
            throw new IOException("Codecs in files do not match.");
        }
    }

    if (value.getLength() == 0) {
        return;
    }

    valueBytes = Arrays.copyOfRange(value.getBytes(), 0, value.getLength());
    output.collect(new BytesWritable(valueBytes), null);
    output.collect(new BytesWritable(header.getSyncMarker()), null);

    reporter.incrCounter("Avro Block", "Blocks processed", 1);
    reporter.incrCounter("Avro Block", "Bytes processed", value.getLength() + 16);
}

From source file:com.blackberry.logdriver.mapred.BinaryRecordWriter.java

License:Apache License

@Override
public void write(BytesWritable k, NullWritable v) throws IOException {
    if (k == null || k.getLength() == 0) {
        return;/*from   w ww  . j  av  a 2  s. c  o  m*/
    }

    byte[] data = Arrays.copyOfRange(k.getBytes(), 0, k.getLength());
    out.write(data);
}

From source file:com.blackberry.logdriver.mapreduce.avro.AvroBlockWriterMapper.java

License:Apache License

/**
 * Writes out the blocks into new avro files.
 * // ww  w . j  av  a 2  s .  c om
 * @param key
 *          The header of the Avro file that the current block is from.
 * @param value
 *          The raw bytes of an Avro file data block, excluding the trailing
 *          sync marker.
 * @param context
 *          The mapper Context.
 * @throws IOException
 *           If this instance has received AvroFileHeader instances with
 *           different schemas or codecs, or other IO Error.
 * @throws InterruptedException
 */
@Override
protected void map(AvroFileHeader key, BytesWritable value, Context context)
        throws IOException, InterruptedException {
    byte[] valueBytes = null;

    if (header.getSyncMarker() == null) {
        LOG.info("Writing new header for new file: {}", key.toString());
        header.set(key);
        context.write(new BytesWritable(header.toBytes()), null);
    } else {
        AvroFileHeader newHeader = key;
        if (!header.getSchema().equals(newHeader.getSchema())) {
            throw new IOException("Schemas in files do not match.");
        }
        if (!header.getCodec().equals(newHeader.getCodec())) {
            throw new IOException("Codecs in files do not match.");
        }
    }

    if (value.getLength() == 0) {
        return;
    }

    valueBytes = Arrays.copyOfRange(value.getBytes(), 0, value.getLength());
    context.write(new BytesWritable(valueBytes), null);
    context.write(new BytesWritable(header.getSyncMarker()), null);

    context.getCounter("Avro Block", "Blocks processed").increment(1);
    context.getCounter("Avro Block", "Bytes processed").increment(value.getLength() + 16);
}

From source file:com.blackberry.logdriver.mapreduce.BinaryRecordWriter.java

License:Apache License

@Override
public synchronized void write(BytesWritable k, NullWritable v) throws IOException {
    if (k == null || k.getLength() == 0) {
        return;/*from w w w .  j  a v  a2  s  .c  o  m*/
    }

    byte[] data = Arrays.copyOfRange(k.getBytes(), 0, k.getLength());
    out.write(data);
}

From source file:com.chinamobile.bcbsp.ml.HashMLWritePartition.java

License:Apache License

/**
 * This method is used to partition graph vertexes. Writing Each vertex to the
 * corresponding partition. In this method calls recordParse method to create
 * an HeadNode object. The last call partitioner's getPartitionId method to
 * calculate the HeadNode belongs to partition's id. If the HeadNode belongs
 * local partition then written to the local partition or send it to the
 * appropriate partition.// w ww.j  a v a 2  s  .  c  o  m
 * @param recordReader The recordreader of the split.
 * @throws IOException The io exception
 * @throws InterruptedException The Interrupted Exception
 */
@Override
public void write(RecordReader recordReader) throws IOException, InterruptedException {
    int headNodeNum = 0;
    int local = 0;
    int send = 0;
    int lost = 0;
    ThreadPool tpool = new ThreadPool(this.sendThreadNum);
    int bufferSize = (this.TotalCacheSize * CONTAINERNUMBER * CONTAINERNUMBER)
            / (this.staff.getStaffNum() + this.sendThreadNum);
    byte[][] buffer = new byte[this.staff.getStaffNum()][bufferSize];
    int[] bufindex = new int[this.staff.getStaffNum()];
    BytesWritable kbytes = new BytesWritable();
    int ksize = 0;
    BytesWritable vbytes = new BytesWritable();
    int vsize = 0;
    DataOutputBuffer bb = new DataOutputBuffer();
    try {
        this.keyserializer.open(bb);
        this.valueserializer.open(bb);
    } catch (IOException e) {
        throw e;
    }
    try {
        while (recordReader != null && recordReader.nextKeyValue()) {
            headNodeNum++;
            Text key = new Text(recordReader.getCurrentKey().toString());
            Text value = new Text(recordReader.getCurrentValue().toString());
            int pid = -1;
            if (key != null) {
                pid = this.partitioner.getPartitionID(key);
            } else {
                lost++;
                continue;
            }
            if (pid == this.staff.getPartition()) {
                local++;

                KeyValuePair pair = (KeyValuePair) this.recordParse.recordParse(key.toString(),
                        value.toString());

                if (pair == null) {
                    lost++;
                    continue;
                }
                staff.getGraphData().addForAll(pair);
            } else {
                send++;
                bb.reset();
                this.keyserializer.serialize(key);
                kbytes.set(bb.getData(), 0, bb.getLength());
                ksize = kbytes.getLength();
                bb.reset();
                this.valueserializer.serialize(value);
                vbytes.set(bb.getData(), 0, bb.getLength());
                vsize = vbytes.getLength();
                if ((buffer[pid].length - bufindex[pid]) > (ksize + vsize)) {
                    System.arraycopy(kbytes.getBytes(), 0, buffer[pid], bufindex[pid], ksize);
                    bufindex[pid] += ksize;
                    System.arraycopy(vbytes.getBytes(), 0, buffer[pid], bufindex[pid], vsize);
                    bufindex[pid] += vsize;
                } else if (buffer[pid].length < (ksize + vsize)) {
                    ThreadSignle t = tpool.getThread();
                    while (t == null) {
                        t = tpool.getThread();
                    }
                    t.setWorker(this.workerAgent.getWorker(staff.getJobID(), staff.getStaffID(), pid));
                    t.setJobId(staff.getJobID());
                    t.setTaskId(staff.getStaffID());
                    t.setBelongPartition(pid);
                    BytesWritable data = new BytesWritable();
                    byte[] tmp = new byte[vsize + ksize];
                    System.arraycopy(kbytes.getBytes(), 0, tmp, 0, ksize);
                    System.arraycopy(vbytes.getBytes(), 0, tmp, ksize, vsize);
                    data.set(tmp, 0, (ksize + vsize));
                    t.setData(data);
                    tmp = null;
                    LOG.info("Using Thread is: " + t.getThreadNumber());
                    LOG.info("this is a super record");
                    t.setStatus(true);
                } else {
                    ThreadSignle t = tpool.getThread();
                    while (t == null) {
                        t = tpool.getThread();
                    }
                    t.setWorker(this.workerAgent.getWorker(staff.getJobID(), staff.getStaffID(), pid));
                    t.setJobId(staff.getJobID());
                    t.setTaskId(staff.getStaffID());
                    t.setBelongPartition(pid);
                    BytesWritable data = new BytesWritable();
                    data.set(buffer[pid], 0, bufindex[pid]);
                    t.setData(data);
                    LOG.info("Using Thread is: " + t.getThreadNumber());
                    t.setStatus(true);
                    bufindex[pid] = 0;
                    // store data
                    System.arraycopy(kbytes.getBytes(), 0, buffer[pid], bufindex[pid], ksize);
                    bufindex[pid] += ksize;
                    System.arraycopy(vbytes.getBytes(), 0, buffer[pid], bufindex[pid], vsize);
                    bufindex[pid] += vsize;
                }
            }
        }
        for (int i = 0; i < this.staff.getStaffNum(); i++) {
            if (bufindex[i] != 0) {
                ThreadSignle t = tpool.getThread();
                while (t == null) {
                    t = tpool.getThread();
                }
                t.setWorker(this.workerAgent.getWorker(staff.getJobID(), staff.getStaffID(), i));
                t.setJobId(staff.getJobID());
                t.setTaskId(staff.getStaffID());
                t.setBelongPartition(i);
                BytesWritable data = new BytesWritable();
                data.set(buffer[i], 0, bufindex[i]);
                t.setData(data);
                LOG.info("Using Thread is: " + t.getThreadNumber());
                t.setStatus(true);
            }
        }
        tpool.cleanup();
        tpool = null;
        buffer = null;
        bufindex = null;
        LOG.info("The number of vertices that were read from the input file: " + headNodeNum);
        LOG.info("The number of vertices that were put into the partition: " + local);
        LOG.info("The number of vertices that were sent to other partitions: " + send);
        LOG.info("The number of verteices in the partition that cound not be " + "parsed:" + lost);
    } catch (IOException e) {
        throw e;
    } catch (InterruptedException e) {
        throw e;
    }
}

From source file:com.chinamobile.bcbsp.partition.HashWithBalancerWritePartition.java

License:Apache License

/**
 * This method is used to partition graph vertexes. Writing Each vertex to the
 * corresponding partition. In this method calls recordParse method to create
 * an HeadNode object. The last call partitioner's getPartitionId method to
 * calculate the HeadNode belongs to partition's id. If the HeadNode belongs
 * local partition then written to the local partition or send it to the
 * appropriate partition.//from   www  . j av  a  2  s .c om
 * @param recordReader The recordreader of the split.
 * @throws IOException The io exception
 * @throws InterruptedException The Interrupted Exception
 */
@Override
public void write(RecordReader recordReader) throws IOException, InterruptedException {
    int headNodeNum = 0;
    int local = 0;
    int send = 0;
    int lost = 0;
    ThreadPool tpool = new ThreadPool(this.sendThreadNum);
    int staffNum = this.staff.getStaffNum();
    BytesWritable kbytes = new BytesWritable();
    int ksize = 0;
    BytesWritable vbytes = new BytesWritable();
    int vsize = 0;
    DataOutputBuffer bb = new DataOutputBuffer();
    int bufferSize = (int) ((this.TotalCacheSize * CONTAINERNUMBER * CONTAINERNUMBER) * PART);
    int dataBufferSize = (this.TotalCacheSize * CONTAINERNUMBER * CONTAINERNUMBER)
            / (this.staff.getStaffNum() + this.sendThreadNum);
    byte[] buffer = new byte[bufferSize];
    int bufindex = 0;
    SerializationFactory sFactory = new SerializationFactory(new Configuration());
    Serializer<IntWritable> psserializer = sFactory.getSerializer(IntWritable.class);
    byte[] pidandsize = new byte[TIME * CONTAINERNUMBER * CONTAINERNUMBER];
    int psindex = 0;
    BytesWritable pidbytes = new BytesWritable();
    int psize = 0;
    BytesWritable sizebytes = new BytesWritable();
    int ssize = 0;
    try {
        this.keyserializer.open(bb);
        this.valueserializer.open(bb);
        psserializer.open(bb);
    } catch (IOException e) {
        throw e;
    }
    String path = "/tmp/bcbsp/" + this.staff.getJobID() + "/" + this.staff.getStaffID();
    File dir = new File("/tmp/bcbsp/" + this.staff.getJobID());
    dir.mkdir();
    dir = new File("/tmp/bcbsp/" + this.staff.getJobID() + "/" + this.staff.getStaffID());
    dir.mkdir();
    ArrayList<File> files = new ArrayList<File>();
    try {
        File file = new File(path + "/" + "data" + ".txt");
        files.add(file);
        DataOutputStream dataWriter = new DataOutputStream(
                new BufferedOutputStream(new FileOutputStream(path + "/" + "data" + ".txt", true)));
        DataInputStream dataReader = new DataInputStream(
                new BufferedInputStream(new FileInputStream(path + "/" + "data" + ".txt")));
        File filet = new File(path + "/" + "pidandsize" + ".txt");
        files.add(filet);
        DataOutputStream psWriter = new DataOutputStream(
                new BufferedOutputStream(new FileOutputStream(path + "/" + "pidandsize" + ".txt", true)));
        DataInputStream psReader = new DataInputStream(
                new BufferedInputStream(new FileInputStream(path + "/" + "pidandsize" + ".txt")));
        while (recordReader != null && recordReader.nextKeyValue()) {
            headNodeNum++;
            Text key = new Text(recordReader.getCurrentKey().toString());
            Text value = new Text(recordReader.getCurrentValue().toString());
            int pid = -1;
            Text vertexID = this.recordParse.getVertexID(key);
            if (vertexID != null) {
                pid = this.partitioner.getPartitionID(vertexID);
            } else {
                lost++;
                continue;
            }
            if (this.counter.containsKey(pid)) {
                this.counter.put(pid, (this.counter.get(pid) + 1));
            } else {
                this.counter.put(pid, 1);
            }
            bb.reset();
            this.keyserializer.serialize(key);
            kbytes.set(bb.getData(), 0, bb.getLength());
            ksize = kbytes.getLength();
            bb.reset();
            this.valueserializer.serialize(value);
            vbytes.set(bb.getData(), 0, bb.getLength());
            vsize = vbytes.getLength();
            bb.reset();
            psserializer.serialize(new IntWritable(ksize + vsize));
            sizebytes.set(bb.getData(), 0, bb.getLength());
            ssize = sizebytes.getLength();
            bb.reset();
            psserializer.serialize(new IntWritable(pid));
            pidbytes.set(bb.getData(), 0, bb.getLength());
            psize = pidbytes.getLength();
            if ((pidandsize.length - psindex) > (ssize + psize)) {
                System.arraycopy(sizebytes.getBytes(), 0, pidandsize, psindex, ssize);
                psindex += ssize;
                System.arraycopy(pidbytes.getBytes(), 0, pidandsize, psindex, psize);
                psindex += psize;
            } else {
                psWriter.write(pidandsize, 0, psindex);
                psindex = 0;
                System.arraycopy(sizebytes.getBytes(), 0, pidandsize, psindex, ssize);
                psindex += ssize;
                System.arraycopy(pidbytes.getBytes(), 0, pidandsize, psindex, psize);
                psindex += psize;
            }
            if ((buffer.length - bufindex) > (ksize + vsize)) {
                System.arraycopy(kbytes.getBytes(), 0, buffer, bufindex, ksize);
                bufindex += ksize;
                System.arraycopy(vbytes.getBytes(), 0, buffer, bufindex, vsize);
                bufindex += vsize;
            } else if (buffer.length < (ksize + vsize)) {
                dataWriter.write(buffer, 0, bufindex);
                bufindex = 0;
                LOG.info("This is a super record");
                dataWriter.write(kbytes.getBytes(), 0, ksize);
                dataWriter.write(vbytes.getBytes(), 0, vsize);
            } else {
                dataWriter.write(buffer, 0, bufindex);
                bufindex = 0;
                System.arraycopy(kbytes.getBytes(), 0, buffer, bufindex, ksize);
                bufindex += ksize;
                System.arraycopy(vbytes.getBytes(), 0, buffer, bufindex, vsize);
                bufindex += vsize;
            }
        }
        if (psindex != 0) {
            psWriter.write(pidandsize, 0, psindex);
        }
        if (bufindex != 0) {
            dataWriter.write(buffer, 0, bufindex);
            bufindex = 0;
        }
        dataWriter.close();
        dataWriter = null;
        psWriter.close();
        psWriter = null;
        buffer = null;
        pidandsize = null;
        this.ssrc.setDirFlag(new String[] { "3" });
        this.ssrc.setCounter(this.counter);
        HashMap<Integer, Integer> hashBucketToPartition = this.sssc.loadDataInBalancerBarrier(ssrc,
                Constants.PARTITION_TYPE.HASH);
        this.staff.setHashBucketToPartition(hashBucketToPartition);
        byte[][] databuf = new byte[staffNum][dataBufferSize];
        int[] databufindex = new int[staffNum];
        try {
            IntWritable pid = new IntWritable();
            IntWritable size = new IntWritable();
            int belongPid = 0;
            while (true) {
                size.readFields(psReader);
                pid.readFields(psReader);
                belongPid = hashBucketToPartition.get(pid.get());
                if (belongPid != this.staff.getPartition()) {
                    send++;
                } else {
                    local++;
                }
                if ((databuf[belongPid].length - databufindex[belongPid]) > size.get()) {
                    dataReader.read(databuf[belongPid], databufindex[belongPid], size.get());
                    databufindex[belongPid] += size.get();
                } else if (databuf[belongPid].length < size.get()) {
                    LOG.info("This is a super record");
                    byte[] tmp = new byte[size.get()];
                    dataReader.read(tmp, 0, size.get());
                    if (belongPid == this.staff.getPartition()) {
                        DataInputStream reader = new DataInputStream(
                                new BufferedInputStream(new ByteArrayInputStream(tmp)));
                        try {
                            boolean stop = true;
                            while (stop) {
                                Text key = new Text();
                                key.readFields(reader);
                                Text value = new Text();
                                value.readFields(reader);
                                if (key.getLength() > 0 && value.getLength() > 0) {
                                    Vertex vertex = this.recordParse.recordParse(key.toString(),
                                            value.toString());
                                    if (vertex == null) {
                                        lost++;
                                        continue;
                                    }
                                    this.staff.getGraphData().addForAll(vertex);
                                } else {
                                    stop = false;
                                }
                            }
                        } catch (IOException e) {
                            LOG.info("IO exception: " + e.getStackTrace());
                        }
                    } else {
                        ThreadSignle t = tpool.getThread();
                        while (t == null) {
                            t = tpool.getThread();
                        }
                        t.setWorker(
                                this.workerAgent.getWorker(staff.getJobID(), staff.getStaffID(), belongPid));
                        t.setJobId(staff.getJobID());
                        t.setTaskId(staff.getStaffID());
                        t.setBelongPartition(belongPid);
                        BytesWritable data = new BytesWritable();
                        data.set(tmp, 0, size.get());
                        t.setData(data);
                        LOG.info("Using Thread is: " + t.getThreadNumber());
                        t.setStatus(true);
                    }
                    tmp = null;
                } else {
                    if (belongPid == this.staff.getPartition()) {
                        DataInputStream reader = new DataInputStream(new BufferedInputStream(
                                new ByteArrayInputStream(databuf[belongPid], 0, databufindex[belongPid])));
                        try {
                            boolean stop = true;
                            while (stop) {
                                Text key = new Text();
                                key.readFields(reader);
                                Text value = new Text();
                                value.readFields(reader);
                                if (key.getLength() > 0 && value.getLength() > 0) {
                                    Vertex vertex = this.recordParse.recordParse(key.toString(),
                                            value.toString());
                                    if (vertex == null) {
                                        lost++;
                                        continue;
                                    }
                                    this.staff.getGraphData().addForAll(vertex);
                                } else {
                                    stop = false;
                                }
                            }
                        } catch (IOException e) {
                            LOG.info("IO exception: " + e.getStackTrace());
                        }
                    } else {
                        ThreadSignle t = tpool.getThread();
                        while (t == null) {
                            t = tpool.getThread();
                        }
                        t.setWorker(
                                this.workerAgent.getWorker(staff.getJobID(), staff.getStaffID(), belongPid));
                        t.setJobId(staff.getJobID());
                        t.setTaskId(staff.getStaffID());
                        t.setBelongPartition(belongPid);
                        BytesWritable data = new BytesWritable();
                        data.set(databuf[belongPid], 0, databufindex[belongPid]);
                        t.setData(data);
                        LOG.info("Using Thread is: " + t.getThreadNumber());
                        t.setStatus(true);
                    }
                    databufindex[belongPid] = 0;
                    dataReader.read(databuf[belongPid], databufindex[belongPid], size.get());
                    databufindex[belongPid] += size.get();
                }
            }
        } catch (EOFException ex) {
            LOG.error("[write]", ex);
        }
        for (int i = 0; i < staffNum; i++) {
            if (databufindex[i] != 0) {
                if (i == this.staff.getPartition()) {
                    DataInputStream reader = new DataInputStream(
                            new BufferedInputStream(new ByteArrayInputStream(databuf[i], 0, databufindex[i])));
                    try {
                        boolean stop = true;
                        while (stop) {
                            Text key = new Text();
                            key.readFields(reader);
                            Text value = new Text();
                            value.readFields(reader);
                            if (key.getLength() > 0 && value.getLength() > 0) {
                                Vertex vertex = this.recordParse.recordParse(key.toString(), value.toString());
                                if (vertex == null) {
                                    lost++;
                                    continue;
                                }
                                this.staff.getGraphData().addForAll(vertex);
                            } else {
                                stop = false;
                            }
                        }
                    } catch (IOException e) {
                        LOG.info("IO exception: " + e.getStackTrace());
                    }
                } else {
                    ThreadSignle t = tpool.getThread();
                    while (t == null) {
                        t = tpool.getThread();
                    }
                    t.setWorker(this.workerAgent.getWorker(staff.getJobID(), staff.getStaffID(), i));
                    t.setJobId(staff.getJobID());
                    t.setTaskId(staff.getStaffID());
                    t.setBelongPartition(i);
                    BytesWritable data = new BytesWritable();
                    data.set(databuf[i], 0, databufindex[i]);
                    t.setData(data);
                    LOG.info("Using Thread is: " + t.getThreadNumber());
                    t.setStatus(true);
                }
            }
        }
        dataReader.close();
        dataReader = null;
        psReader.close();
        psReader = null;
        for (File f : files) {
            f.delete();
        }
        dir.delete();
        dir = new File(path.substring(0, path.lastIndexOf('/')));
        dir.delete();
        tpool.cleanup();
        tpool = null;
        databuf = null;
        databufindex = null;
        this.counter = null;
        LOG.info("The number of vertices that were read from the input file: " + headNodeNum);
        LOG.info("The number of vertices that were put into the partition: " + local);
        LOG.info("The number of vertices that were sent to other partitions: " + send);
        LOG.info("The number of verteices in the partition that cound not be " + "parsed:" + lost);
    } catch (IOException e) {
        throw e;
    } catch (InterruptedException e) {
        throw e;
    } finally {
        for (File f : files) {
            f.delete();
        }
        dir.delete();
        dir = new File(path.substring(0, path.lastIndexOf('/')));
        dir.delete();
    }
}

From source file:com.chinamobile.bcbsp.partition.HashWritePartition.java

License:Apache License

/**
 * This method is used to partition graph vertexes. Writing Each vertex to the
 * corresponding partition. In this method calls recordParse method to create
 * an HeadNode object. The last call partitioner's getPartitionId method to
 * calculate the HeadNode belongs to partition's id. If the HeadNode belongs
 * local partition then written to the local partition or send it to the
 * appropriate partition.//w  w  w . jav  a2  s .  com
 * @param recordReader The recordreader of the split.
 * @throws IOException The io exception
 * @throws InterruptedException The Interrupted Exception
 */
@Override
public void write(RecordReader recordReader) throws IOException, InterruptedException {
    int headNodeNum = 0;
    int local = 0;
    int send = 0;
    int lost = 0;
    ThreadPool tpool = new ThreadPool(this.sendThreadNum);
    int bufferSize = (this.TotalCacheSize * CONTAINERNUMBER * CONTAINERNUMBER)
            / (this.staff.getStaffNum() + this.sendThreadNum);
    byte[][] buffer = new byte[this.staff.getStaffNum()][bufferSize];
    int[] bufindex = new int[this.staff.getStaffNum()];
    BytesWritable kbytes = new BytesWritable();
    int ksize = 0;
    BytesWritable vbytes = new BytesWritable();
    int vsize = 0;
    DataOutputBuffer bb = new DataOutputBuffer();
    try {
        this.keyserializer.open(bb);
        this.valueserializer.open(bb);
    } catch (IOException e) {
        throw e;
    }
    try {
        while (recordReader != null && recordReader.nextKeyValue()) {
            headNodeNum++;
            Text key = new Text(recordReader.getCurrentKey().toString());
            Text value = new Text(recordReader.getCurrentValue().toString());
            int pid = -1;
            Text vertexID = this.recordParse.getVertexID(key);
            if (vertexID != null) {
                pid = this.partitioner.getPartitionID(vertexID);
            } else {
                lost++;
                continue;
            }
            if (pid == this.staff.getPartition()) {
                local++;
                Vertex vertex = this.recordParse.recordParse(key.toString(), value.toString());
                if (vertex == null) {
                    lost++;
                    continue;
                }
                staff.getGraphData().addForAll(vertex);
            } else {
                send++;
                bb.reset();
                this.keyserializer.serialize(key);
                kbytes.set(bb.getData(), 0, bb.getLength());
                ksize = kbytes.getLength();
                bb.reset();
                this.valueserializer.serialize(value);
                vbytes.set(bb.getData(), 0, bb.getLength());
                vsize = vbytes.getLength();
                if ((buffer[pid].length - bufindex[pid]) > (ksize + vsize)) {
                    System.arraycopy(kbytes.getBytes(), 0, buffer[pid], bufindex[pid], ksize);
                    bufindex[pid] += ksize;
                    System.arraycopy(vbytes.getBytes(), 0, buffer[pid], bufindex[pid], vsize);
                    bufindex[pid] += vsize;
                } else if (buffer[pid].length < (ksize + vsize)) {
                    ThreadSignle t = tpool.getThread();
                    while (t == null) {
                        t = tpool.getThread();
                    }
                    t.setWorker(this.workerAgent.getWorker(staff.getJobID(), staff.getStaffID(), pid));
                    t.setJobId(staff.getJobID());
                    t.setTaskId(staff.getStaffID());
                    t.setBelongPartition(pid);
                    BytesWritable data = new BytesWritable();
                    byte[] tmp = new byte[vsize + ksize];
                    System.arraycopy(kbytes.getBytes(), 0, tmp, 0, ksize);
                    System.arraycopy(vbytes.getBytes(), 0, tmp, ksize, vsize);
                    data.set(tmp, 0, (ksize + vsize));
                    t.setData(data);
                    tmp = null;
                    LOG.info("Using Thread is: " + t.getThreadNumber());
                    LOG.info("this is a super record");
                    t.setStatus(true);
                } else {
                    ThreadSignle t = tpool.getThread();
                    while (t == null) {
                        t = tpool.getThread();
                    }
                    t.setWorker(this.workerAgent.getWorker(staff.getJobID(), staff.getStaffID(), pid));
                    t.setJobId(staff.getJobID());
                    t.setTaskId(staff.getStaffID());
                    t.setBelongPartition(pid);
                    BytesWritable data = new BytesWritable();
                    data.set(buffer[pid], 0, bufindex[pid]);
                    t.setData(data);
                    LOG.info("Using Thread is: " + t.getThreadNumber());
                    t.setStatus(true);
                    bufindex[pid] = 0;
                    // store data
                    System.arraycopy(kbytes.getBytes(), 0, buffer[pid], bufindex[pid], ksize);
                    bufindex[pid] += ksize;
                    System.arraycopy(vbytes.getBytes(), 0, buffer[pid], bufindex[pid], vsize);
                    bufindex[pid] += vsize;
                }
            }
        }
        for (int i = 0; i < this.staff.getStaffNum(); i++) {
            if (bufindex[i] != 0) {
                ThreadSignle t = tpool.getThread();
                while (t == null) {
                    t = tpool.getThread();
                }
                t.setWorker(this.workerAgent.getWorker(staff.getJobID(), staff.getStaffID(), i));
                t.setJobId(staff.getJobID());
                t.setTaskId(staff.getStaffID());
                t.setBelongPartition(i);
                BytesWritable data = new BytesWritable();
                data.set(buffer[i], 0, bufindex[i]);
                t.setData(data);
                LOG.info("Using Thread is: " + t.getThreadNumber());
                t.setStatus(true);
            }
        }
        tpool.cleanup();
        tpool = null;
        buffer = null;
        bufindex = null;
        LOG.info("The number of vertices that were read from the input file: " + headNodeNum);
        LOG.info("The number of vertices that were put into the partition: " + local);
        LOG.info("The number of vertices that were sent to other partitions: " + send);
        LOG.info("The number of verteices in the partition that cound not be " + "parsed:" + lost);
    } catch (IOException e) {
        throw e;
    } catch (InterruptedException e) {
        throw e;
    }
}

From source file:com.cloudera.fts.hive.ProtobufSerDe.java

License:Open Source License

@Override
public Object deserialize(Writable field) throws SerDeException {
    Message.Builder mb = instance.newBuilderForType();
    try {//from ww w  .  jav a 2  s. c  om
        BytesWritable bw = (BytesWritable) field;
        return mb.mergeFrom(bw.getBytes(), 0, bw.getLength()).build();
    } catch (Exception e) {
        throw new SerDeException(e);
    }
}