List of usage examples for org.apache.hadoop.io IntWritable readFields
@Override public void readFields(DataInput in) throws IOException
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 w ww .j a va 2 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 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.digitalpebble.behemoth.BehemothDocument.java
License:Apache License
public void readAnnotationFields(Annotation annot, DataInput in, List<String> types) throws IOException { IntWritable posType = new IntWritable(); posType.readFields(in); annot.setType(types.get(posType.get())); annot.setStart(WritableUtils.readVLong(in)); annot.setEnd(WritableUtils.readVLong(in)); HashMap<String, String> features = null; int numFeatures = in.readInt(); if (numFeatures > 0) features = new HashMap<String, String>(numFeatures); for (int i = 0; i < numFeatures; i++) { posType.readFields(in);/*ww w . j a v a 2 s . c o m*/ String fname = types.get(posType.get()); String fvalue = WritableUtils.readString(in); features.put(fname, fvalue); } annot.setFeatures(features); }
From source file:com.ibm.bi.dml.runtime.matrix.sort.ReadWithZeros.java
License:Open Source License
public void readNextKeyValuePairs(DoubleWritable readKey, IntWritable readValue) throws IOException { try {//w ww . j av a 2s.co m if (contain0s && justFound0) { readKey.set(keyAfterZero.get()); readValue.set(valueAfterZero.get()); contain0s = false; } else { readKey.readFields(currentStream); readValue.readFields(currentStream); } } catch (EOFException e) { // case in which zero is the maximum value in the matrix. // The zero value from the last entry is not present in the input sorted matrix, but needs to be accounted for. if (contain0s && !justFound0) { justFound0 = true; readKey.set(0); readValue.set((int) numZeros); } else { throw e; } } if (contain0s && !justFound0 && readKey.get() >= 0) { justFound0 = true; keyAfterZero.set(readKey.get()); valueAfterZero.set(readValue.get()); readKey.set(0); readValue.set((int) numZeros); } }
From source file:com.marcolotz.MRComponents.SerializerConverter.java
License:Creative Commons License
/*** * Reads an Int from the input/* ww w. j av a2s. c o m*/ * @param datainput * @return the int readen * @throws IOException */ public static int readInt(DataInput datainput) throws IOException { IntWritable readenInt = new IntWritable(); readenInt.readFields(datainput); return readenInt.get(); }
From source file:org.apache.hama.bsp.message.TestMessageIO.java
License:Apache License
public void testDirectByteBufferInput() throws Exception { ByteBuffer buffer = ByteBuffer.allocateDirect(512); DirectByteBufferOutputStream stream = new DirectByteBufferOutputStream(); stream.setBuffer(buffer);/*www .j a va 2 s . co m*/ IntWritable intWritable = new IntWritable(1); for (int i = 0; i < 100; ++i) { intWritable.set(i); intWritable.write(stream); } intWritable.write(stream); stream.close(); buffer.flip(); DirectByteBufferInputStream inStream = new DirectByteBufferInputStream(); inStream.setBuffer(new SpilledByteBuffer(buffer, 400)); for (int i = 0; i < 100; ++i) { intWritable.readFields(inStream); assertEquals(i, intWritable.get()); } assertFalse(inStream.hasDataToRead()); assertTrue(inStream.hasUnmarkData()); inStream.prepareForNext(); // push in another buffer and check if the unmarked data could be read. buffer.clear(); stream = new DirectByteBufferOutputStream(); buffer = ByteBuffer.allocateDirect(2048); stream.setBuffer(buffer); for (int i = 0; i < 400; ++i) { intWritable.set(i); intWritable.write(stream); } stream.close(); buffer.flip(); inStream.setBuffer(new SpilledByteBuffer(buffer, 400)); // Read previous data intWritable.readFields(inStream); assertEquals(99, intWritable.get()); for (int i = 0; i < 100; ++i) { intWritable.readFields(inStream); assertEquals(i, intWritable.get()); } assertFalse(inStream.hasDataToRead()); assertTrue(inStream.hasUnmarkData()); inStream.prepareForNext(); buffer.clear(); stream = new DirectByteBufferOutputStream(); stream.setBuffer(buffer); for (int i = 0; i < 100; ++i) { intWritable.set(i); intWritable.write(stream); } stream.close(); buffer.flip(); inStream.setBuffer(new SpilledByteBuffer(buffer, 400)); // Read previous data with resized intermediate buffer for (int i = 100; i < 400; ++i) { intWritable.readFields(inStream); assertEquals(i, intWritable.get()); } for (int i = 0; i < 100; ++i) { intWritable.readFields(inStream); assertEquals(i, intWritable.get()); } assertFalse(inStream.hasDataToRead()); assertFalse(inStream.hasUnmarkData()); }
From source file:org.apache.hama.bsp.message.TestMessageIO.java
License:Apache License
public void testSyncFlushBufferInputStream() throws Exception { File f = null;//from w ww. j av a 2s . c o m try { String fileName = System.getProperty("java.io.tmpdir") + File.separatorChar + "testSyncFlushBufferInputStream.txt"; SyncFlushByteBufferOutputStream stream = new SyncFlushByteBufferOutputStream(fileName); DirectByteBufferOutputStream syncFlushStream = new DirectByteBufferOutputStream(stream); ByteBuffer buffer = ByteBuffer.allocateDirect(512); syncFlushStream.setBuffer(buffer); IntWritable intWritable = new IntWritable(1); for (int i = 0; i < 200; ++i) { intWritable.set(i); intWritable.write(syncFlushStream); } intWritable.write(syncFlushStream); syncFlushStream.close(); f = new File(fileName); assertTrue(f.exists()); assertEquals(f.length(), 804); SyncReadByteBufferInputStream syncReadStream = new SyncReadByteBufferInputStream(stream.isSpilled(), fileName); DirectByteBufferInputStream inStream = new DirectByteBufferInputStream(syncReadStream); buffer.clear(); inStream.setBuffer(buffer); for (int i = 0; i < 200; ++i) { intWritable.readFields(inStream); assertEquals(intWritable.get(), i); } intWritable.readFields(inStream); assertEquals(intWritable.get(), 199); try { intWritable.readFields(inStream); assertFalse(true); } catch (Exception e) { assertTrue(true); } inStream.close(); syncFlushStream.close(); } finally { if (f != null) { f.delete(); } } }
From source file:org.apache.mahout.common.IntTuple.java
License:Apache License
@Override public void readFields(DataInput in) throws IOException { int len = in.readInt(); tuple = new IntArrayList(len); IntWritable value = new IntWritable(); for (int i = 0; i < len; i++) { value.readFields(in); tuple.add(value.get());/*from ww w . j ava 2 s . c o m*/ } }
From source file:org.apache.sysml.runtime.matrix.sort.ReadWithZeros.java
License:Apache License
public boolean readNextKeyValuePairs(DoubleWritable readKey, IntWritable readValue) throws IOException { boolean ret = true; try {/*from www .ja v a 2s .com*/ if (contain0s && justFound0) { readKey.set(keyAfterZero.get()); readValue.set(valueAfterZero.get()); contain0s = false; } else { readKey.readFields(currentStream); readValue.readFields(currentStream); } } catch (EOFException e) { // case in which zero is the maximum value in the matrix. // The zero value from the last entry is not present in the input sorted matrix, but needs to be accounted for. if (contain0s && !justFound0) { justFound0 = true; readKey.set(0); readValue.set((int) numZeros); } else { ret = false; } } if (contain0s && !justFound0 && readKey.get() >= 0) { justFound0 = true; keyAfterZero.set(readKey.get()); valueAfterZero.set(readValue.get()); readKey.set(0); readValue.set((int) numZeros); } return ret; }
From source file:org.apache.tez.runtime.library.common.sort.impl.TestTezMerger.java
License:Apache License
private void merge(List<Path> pathList, int mergeFactor, RawComparator rc) throws Exception { //Merge datasets TezMerger merger = new TezMerger(); TezRawKeyValueIterator records = merger.merge(defaultConf, localFs, IntWritable.class, LongWritable.class, null, false, 0, 1024, pathList.toArray(new Path[pathList.size()]), true, mergeFactor, new Path(workDir, "tmp_" + System.nanoTime()), ((rc == null) ? comparator : rc), new Reporter(), null, null, null, new Progress()); //Verify the merged data is correct Map<Integer, Integer> dataMap = Maps.newHashMap(); int pk = -1;// w w w . j av a 2s.co m while (records.next()) { DataInputBuffer key = records.getKey(); DataInputBuffer value = records.getValue(); IntWritable k = new IntWritable(); k.readFields(key); LongWritable v = new LongWritable(); v.readFields(value); if (records.isSameKey()) { LOG.info("\tSame Key : key=" + k.get() + ", val=" + v.get()); //More than one key should be present in the source data assertTrue(verificationDataSet.get(k.get()).size() > 1); //Ensure this is same as the previous key we saw assertTrue("previousKey=" + pk + ", current=" + k.get(), pk == k.get()); } else { LOG.info("key=" + k.get() + ", val=" + v.get()); } pk = k.get(); int keyCount = (dataMap.containsKey(k.get())) ? (dataMap.get(k.get()) + 1) : 1; dataMap.put(k.get(), keyCount); } //Verify if the number of distinct entries is the same in source and the test assertTrue( "dataMap=" + dataMap.keySet().size() + ", verificationSet=" + verificationDataSet.keySet().size(), dataMap.keySet().size() == verificationDataSet.keySet().size()); //Verify with source data for (Integer key : verificationDataSet.keySet()) { assertTrue( "Data size for " + key + " not matching with source; dataSize:" + dataMap.get(key).intValue() + ", source:" + verificationDataSet.get(key).size(), dataMap.get(key).intValue() == verificationDataSet.get(key).size()); } //Verify if every key has the same number of repeated items in the source dataset as well for (Map.Entry<Integer, Integer> entry : dataMap.entrySet()) { assertTrue(entry.getKey() + "", verificationDataSet.get(entry.getKey()).size() == entry.getValue()); } LOG.info("******************"); verificationDataSet.clear(); }
From source file:org.apache.tez.runtime.library.common.writers.TestUnorderedPartitionedKVWriter.java
License:Apache License
private void baseTest(int numRecords, int numPartitions, Set<Integer> skippedPartitions, boolean shouldCompress) throws IOException, InterruptedException { PartitionerForTest partitioner = new PartitionerForTest(); ApplicationId appId = ApplicationId.newInstance(10000, 1); TezCounters counters = new TezCounters(); String uniqueId = UUID.randomUUID().toString(); OutputContext outputContext = createMockOutputContext(counters, appId, uniqueId); Configuration conf = createConfiguration(outputContext, IntWritable.class, LongWritable.class, shouldCompress, -1);// ww w .ja v a 2 s .c om CompressionCodec codec = null; if (shouldCompress) { codec = new DefaultCodec(); ((Configurable) codec).setConf(conf); } int numOutputs = numPartitions; long availableMemory = 2048; int numRecordsWritten = 0; Map<Integer, Multimap<Integer, Long>> expectedValues = new HashMap<Integer, Multimap<Integer, Long>>(); for (int i = 0; i < numOutputs; i++) { expectedValues.put(i, LinkedListMultimap.<Integer, Long>create()); } UnorderedPartitionedKVWriter kvWriter = new UnorderedPartitionedKVWriterForTest(outputContext, conf, numOutputs, availableMemory); int sizePerBuffer = kvWriter.sizePerBuffer; int sizePerRecord = 4 + 8; // IntW + LongW int sizePerRecordWithOverhead = sizePerRecord + 12; // Record + META_OVERHEAD IntWritable intWritable = new IntWritable(); LongWritable longWritable = new LongWritable(); for (int i = 0; i < numRecords; i++) { intWritable.set(i); longWritable.set(i); int partition = partitioner.getPartition(intWritable, longWritable, numOutputs); if (skippedPartitions != null && skippedPartitions.contains(partition)) { continue; } expectedValues.get(partition).put(intWritable.get(), longWritable.get()); kvWriter.write(intWritable, longWritable); numRecordsWritten++; } List<Event> events = kvWriter.close(); int recordsPerBuffer = sizePerBuffer / sizePerRecordWithOverhead; int numExpectedSpills = numRecordsWritten / recordsPerBuffer; verify(outputContext, never()).fatalError(any(Throwable.class), any(String.class)); // Verify the status of the buffers if (numExpectedSpills == 0) { assertEquals(1, kvWriter.numInitializedBuffers); } else { assertTrue(kvWriter.numInitializedBuffers > 1); } assertNull(kvWriter.currentBuffer); assertEquals(0, kvWriter.availableBuffers.size()); // Verify the counters TezCounter outputRecordBytesCounter = counters.findCounter(TaskCounter.OUTPUT_BYTES); TezCounter outputRecordsCounter = counters.findCounter(TaskCounter.OUTPUT_RECORDS); TezCounter outputBytesWithOverheadCounter = counters.findCounter(TaskCounter.OUTPUT_BYTES_WITH_OVERHEAD); TezCounter fileOutputBytesCounter = counters.findCounter(TaskCounter.OUTPUT_BYTES_PHYSICAL); TezCounter spilledRecordsCounter = counters.findCounter(TaskCounter.SPILLED_RECORDS); TezCounter additionalSpillBytesWritternCounter = counters .findCounter(TaskCounter.ADDITIONAL_SPILLS_BYTES_WRITTEN); TezCounter additionalSpillBytesReadCounter = counters.findCounter(TaskCounter.ADDITIONAL_SPILLS_BYTES_READ); TezCounter numAdditionalSpillsCounter = counters.findCounter(TaskCounter.ADDITIONAL_SPILL_COUNT); assertEquals(numRecordsWritten * sizePerRecord, outputRecordBytesCounter.getValue()); assertEquals(numRecordsWritten, outputRecordsCounter.getValue()); assertEquals(numRecordsWritten * sizePerRecordWithOverhead, outputBytesWithOverheadCounter.getValue()); long fileOutputBytes = fileOutputBytesCounter.getValue(); if (numRecordsWritten > 0) { assertTrue(fileOutputBytes > 0); if (!shouldCompress) { assertTrue(fileOutputBytes > outputRecordBytesCounter.getValue()); } } else { assertEquals(0, fileOutputBytes); } assertEquals(recordsPerBuffer * numExpectedSpills, spilledRecordsCounter.getValue()); long additionalSpillBytesWritten = additionalSpillBytesWritternCounter.getValue(); long additionalSpillBytesRead = additionalSpillBytesReadCounter.getValue(); if (numExpectedSpills == 0) { assertEquals(0, additionalSpillBytesWritten); assertEquals(0, additionalSpillBytesRead); } else { assertTrue(additionalSpillBytesWritten > 0); assertTrue(additionalSpillBytesRead > 0); if (!shouldCompress) { assertTrue(additionalSpillBytesWritten > (recordsPerBuffer * numExpectedSpills * sizePerRecord)); assertTrue(additionalSpillBytesRead > (recordsPerBuffer * numExpectedSpills * sizePerRecord)); } } assertTrue(additionalSpillBytesWritten == additionalSpillBytesRead); assertEquals(numExpectedSpills, numAdditionalSpillsCounter.getValue()); BitSet emptyPartitionBits = null; // Verify the event returned assertEquals(1, events.size()); assertTrue(events.get(0) instanceof CompositeDataMovementEvent); CompositeDataMovementEvent cdme = (CompositeDataMovementEvent) events.get(0); assertEquals(0, cdme.getSourceIndexStart()); assertEquals(numOutputs, cdme.getCount()); DataMovementEventPayloadProto eventProto = DataMovementEventPayloadProto .parseFrom(ByteString.copyFrom(cdme.getUserPayload())); assertFalse(eventProto.hasData()); if (skippedPartitions == null && numRecordsWritten > 0) { assertFalse(eventProto.hasEmptyPartitions()); emptyPartitionBits = new BitSet(numPartitions); } else { assertTrue(eventProto.hasEmptyPartitions()); byte[] emptyPartitions = TezCommonUtils .decompressByteStringToByteArray(eventProto.getEmptyPartitions()); emptyPartitionBits = TezUtilsInternal.fromByteArray(emptyPartitions); if (numRecordsWritten == 0) { assertEquals(numPartitions, emptyPartitionBits.cardinality()); } else { for (Integer e : skippedPartitions) { assertTrue(emptyPartitionBits.get(e)); } assertEquals(skippedPartitions.size(), emptyPartitionBits.cardinality()); } } if (emptyPartitionBits.cardinality() != numPartitions) { assertEquals(HOST_STRING, eventProto.getHost()); assertEquals(SHUFFLE_PORT, eventProto.getPort()); assertEquals(uniqueId, eventProto.getPathComponent()); } else { assertFalse(eventProto.hasHost()); assertFalse(eventProto.hasPort()); assertFalse(eventProto.hasPathComponent()); } // Verify the actual data TezTaskOutput taskOutput = new TezTaskOutputFiles(conf, uniqueId); Path outputFilePath = kvWriter.finalOutPath; Path spillFilePath = kvWriter.finalIndexPath; if (numRecordsWritten > 0) { assertTrue(localFs.exists(outputFilePath)); assertTrue(localFs.exists(spillFilePath)); } else { return; } // Special case for 0 records. TezSpillRecord spillRecord = new TezSpillRecord(spillFilePath, conf); DataInputBuffer keyBuffer = new DataInputBuffer(); DataInputBuffer valBuffer = new DataInputBuffer(); IntWritable keyDeser = new IntWritable(); LongWritable valDeser = new LongWritable(); for (int i = 0; i < numOutputs; i++) { if (skippedPartitions != null && skippedPartitions.contains(i)) { continue; } TezIndexRecord indexRecord = spillRecord.getIndex(i); FSDataInputStream inStream = FileSystem.getLocal(conf).open(outputFilePath); inStream.seek(indexRecord.getStartOffset()); IFile.Reader reader = new IFile.Reader(inStream, indexRecord.getPartLength(), codec, null, null, false, 0, -1); while (reader.nextRawKey(keyBuffer)) { reader.nextRawValue(valBuffer); keyDeser.readFields(keyBuffer); valDeser.readFields(valBuffer); int partition = partitioner.getPartition(keyDeser, valDeser, numOutputs); assertTrue(expectedValues.get(partition).remove(keyDeser.get(), valDeser.get())); } inStream.close(); } for (int i = 0; i < numOutputs; i++) { assertEquals(0, expectedValues.get(i).size()); expectedValues.remove(i); } assertEquals(0, expectedValues.size()); }