Examples with TaskAttemptContextImpl org.apache.hadoop.mapreduce.task.TaskAttemptContextImpl used on opensource projects

Example 41 with TaskAttemptContextImpl

use of org.apache.hadoop.mapreduce.task.TaskAttemptContextImpl in project hadoop by apache.

the class TestCopyOutputFormat method testGetOutputCommitter.

@Test
public void testGetOutputCommitter() {
    try {
        TaskAttemptContext context = new TaskAttemptContextImpl(new Configuration(), new TaskAttemptID("200707121733", 1, TaskType.MAP, 1, 1));
        context.getConfiguration().set("mapred.output.dir", "/out");
        Assert.assertTrue(new CopyOutputFormat().getOutputCommitter(context) instanceof CopyCommitter);
    } catch (IOException e) {
        LOG.error("Exception encountered ", e);
        Assert.fail("Unable to get output committer");
    }
}

Example 42 with TaskAttemptContextImpl

use of org.apache.hadoop.mapreduce.task.TaskAttemptContextImpl in project hadoop by apache.

the class TestGridMixClasses method testLoadJobLoadRecordReader.

/*
   * test LoadRecordReader. It class reads data from some files.
   */
@Test(timeout = 3000)
public void testLoadJobLoadRecordReader() throws Exception {
    LoadJob.LoadRecordReader test = new LoadJob.LoadRecordReader();
    Configuration conf = new Configuration();
    FileSystem fs1 = mock(FileSystem.class);
    when(fs1.open((Path) anyObject())).thenReturn(new FakeFSDataInputStream(new FakeInputStream()));
    Path p1 = mock(Path.class);
    when(p1.getFileSystem((JobConf) anyObject())).thenReturn(fs1);
    FileSystem fs2 = mock(FileSystem.class);
    when(fs2.open((Path) anyObject())).thenReturn(new FakeFSDataInputStream(new FakeInputStream()));
    Path p2 = mock(Path.class);
    when(p2.getFileSystem((JobConf) anyObject())).thenReturn(fs2);
    Path[] paths = { p1, p2 };
    long[] start = { 0, 0 };
    long[] lengths = { 1000, 1000 };
    String[] locations = { "temp1", "temp2" };
    CombineFileSplit cfsplit = new CombineFileSplit(paths, start, lengths, locations);
    double[] reduceBytes = { 100, 100 };
    double[] reduceRecords = { 2, 2 };
    long[] reduceOutputBytes = { 500, 500 };
    long[] reduceOutputRecords = { 2, 2 };
    ResourceUsageMetrics metrics = new ResourceUsageMetrics();
    ResourceUsageMetrics[] rMetrics = { new ResourceUsageMetrics(), new ResourceUsageMetrics() };
    LoadSplit input = new LoadSplit(cfsplit, 2, 3, 1500L, 2L, 3000L, 2L, reduceBytes, reduceRecords, reduceOutputBytes, reduceOutputRecords, metrics, rMetrics);
    TaskAttemptID taskId = new TaskAttemptID();
    TaskAttemptContext ctx = new TaskAttemptContextImpl(conf, taskId);
    test.initialize(input, ctx);
    GridmixRecord gr = test.getCurrentValue();
    int counter = 0;
    while (test.nextKeyValue()) {
        gr = test.getCurrentValue();
        if (counter == 0) {
            // read first file
            assertEquals(0.5, test.getProgress(), 0.001);
        } else if (counter == 1) {
            // read second file
            assertEquals(1.0, test.getProgress(), 0.001);
        }
        //
        assertEquals(1000, gr.getSize());
        counter++;
    }
    assertEquals(1000, gr.getSize());
    // Two files have been read
    assertEquals(2, counter);
    test.close();
}

Example 43 with TaskAttemptContextImpl

use of org.apache.hadoop.mapreduce.task.TaskAttemptContextImpl in project carbondata by apache.

the class StoreCreator method executeGraph.

/**
   * Execute graph which will further load data
   *
   * @param loadModel
   * @param storeLocation
   * @throws Exception
   */
public static void executeGraph(CarbonLoadModel loadModel, String storeLocation) throws Exception {
    new File(storeLocation).mkdirs();
    String outPutLoc = storeLocation + "/etl";
    String databaseName = loadModel.getDatabaseName();
    String tableName = loadModel.getTableName();
    String tempLocationKey = databaseName + '_' + tableName + "_1";
    CarbonProperties.getInstance().addProperty(tempLocationKey, storeLocation);
    CarbonProperties.getInstance().addProperty("store_output_location", outPutLoc);
    CarbonProperties.getInstance().addProperty("send.signal.load", "false");
    CarbonProperties.getInstance().addProperty("carbon.is.columnar.storage", "true");
    CarbonProperties.getInstance().addProperty("carbon.dimension.split.value.in.columnar", "1");
    CarbonProperties.getInstance().addProperty("carbon.is.fullyfilled.bits", "true");
    CarbonProperties.getInstance().addProperty("is.int.based.indexer", "true");
    CarbonProperties.getInstance().addProperty("aggregate.columnar.keyblock", "true");
    CarbonProperties.getInstance().addProperty("high.cardinality.value", "100000");
    CarbonProperties.getInstance().addProperty("is.compressed.keyblock", "false");
    CarbonProperties.getInstance().addProperty("carbon.leaf.node.size", "120000");
    String graphPath = outPutLoc + File.separator + loadModel.getDatabaseName() + File.separator + tableName + File.separator + 0 + File.separator + 1 + File.separator + tableName + ".ktr";
    File path = new File(graphPath);
    if (path.exists()) {
        path.delete();
    }
    SchemaInfo info = new SchemaInfo();
    BlockDetails blockDetails = new BlockDetails(new Path(loadModel.getFactFilePath()), 0, new File(loadModel.getFactFilePath()).length(), new String[] { "localhost" });
    Configuration configuration = new Configuration();
    CSVInputFormat.setCommentCharacter(configuration, loadModel.getCommentChar());
    CSVInputFormat.setCSVDelimiter(configuration, loadModel.getCsvDelimiter());
    CSVInputFormat.setEscapeCharacter(configuration, loadModel.getEscapeChar());
    CSVInputFormat.setHeaderExtractionEnabled(configuration, true);
    CSVInputFormat.setQuoteCharacter(configuration, loadModel.getQuoteChar());
    CSVInputFormat.setReadBufferSize(configuration, CarbonProperties.getInstance().getProperty(CarbonCommonConstants.CSV_READ_BUFFER_SIZE, CarbonCommonConstants.CSV_READ_BUFFER_SIZE_DEFAULT));
    CSVInputFormat.setNumberOfColumns(configuration, String.valueOf(loadModel.getCsvHeaderColumns().length));
    CSVInputFormat.setMaxColumns(configuration, "10");
    TaskAttemptContextImpl hadoopAttemptContext = new TaskAttemptContextImpl(configuration, new TaskAttemptID("", 1, TaskType.MAP, 0, 0));
    CSVInputFormat format = new CSVInputFormat();
    RecordReader<NullWritable, StringArrayWritable> recordReader = format.createRecordReader(blockDetails, hadoopAttemptContext);
    CSVRecordReaderIterator readerIterator = new CSVRecordReaderIterator(recordReader, blockDetails, hadoopAttemptContext);
    new DataLoadExecutor().execute(loadModel, storeLocation, new CarbonIterator[] { readerIterator });
    info.setDatabaseName(databaseName);
    info.setTableName(tableName);
    writeLoadMetadata(loadModel.getCarbonDataLoadSchema(), loadModel.getTableName(), loadModel.getTableName(), new ArrayList<LoadMetadataDetails>());
    String segLocation = storeLocation + "/" + databaseName + "/" + tableName + "/Fact/Part0/Segment_0";
    File file = new File(segLocation);
    File factFile = null;
    File[] folderList = file.listFiles();
    File folder = null;
    for (int i = 0; i < folderList.length; i++) {
        if (folderList[i].isDirectory()) {
            folder = folderList[i];
        }
    }
    if (folder.isDirectory()) {
        File[] files = folder.listFiles();
        for (int i = 0; i < files.length; i++) {
            if (!files[i].isDirectory() && files[i].getName().startsWith("part")) {
                factFile = files[i];
                break;
            }
        }
        //      Files.copy(factFile.toPath(), file.toPath(), REPLACE_EXISTING);
        factFile.renameTo(new File(segLocation + "/" + factFile.getName()));
        CarbonUtil.deleteFoldersAndFiles(folder);
    }
}

Example 44 with TaskAttemptContextImpl

use of org.apache.hadoop.mapreduce.task.TaskAttemptContextImpl in project jena by apache.

the class AbstractBlankNodeTests method blank_node_identity_02.

/**
     * Test that starts with two blank nodes in two different files and checks
     * that writing them to a single file does not conflate them
     * 
     * @throws IOException
     * @throws InterruptedException
     */
@Test
public void blank_node_identity_02() throws IOException, InterruptedException {
    Assume.assumeTrue("Requires ParserProfile be respected", this.respectsParserProfile());
    Assume.assumeFalse("Requires that Blank Node identity not be preserved", this.preservesBlankNodeIdentity());
    // Temporary files
    File a = File.createTempFile("bnode_identity", getInitialInputExtension());
    File b = File.createTempFile("bnode_identity", getInitialInputExtension());
    File intermediateOutputDir = Files.createTempDirectory("bnode_identity", new FileAttribute[0]).toFile();
    try {
        // Prepare the input data
        // Same blank node but in different files so must be treated as
        // different blank nodes and not converge
        List<T> tuples = new ArrayList<>();
        Node bnode = NodeFactory.createBlankNode();
        Node pred = NodeFactory.createURI("http://example.org/predicate");
        tuples.add(createTuple(bnode, pred, NodeFactory.createLiteral("first")));
        writeTuples(a, tuples);
        tuples.clear();
        tuples.add(createTuple(bnode, pred, NodeFactory.createLiteral("second")));
        writeTuples(b, tuples);
        // Set up fake job which will process the two files
        Configuration config = new Configuration(true);
        InputFormat<LongWritable, TValue> inputFormat = createInitialInputFormat();
        Job job = Job.getInstance(config);
        job.setInputFormatClass(inputFormat.getClass());
        NLineInputFormat.setNumLinesPerSplit(job, 100);
        FileInputFormat.setInputPaths(job, new Path(a.getAbsolutePath()), new Path(b.getAbsolutePath()));
        FileOutputFormat.setOutputPath(job, new Path(intermediateOutputDir.getAbsolutePath()));
        JobContext context = new JobContextImpl(job.getConfiguration(), job.getJobID());
        // Get the splits
        List<InputSplit> splits = inputFormat.getSplits(context);
        Assert.assertEquals(2, splits.size());
        // Prepare the output writing - putting all output to a single file
        OutputFormat<LongWritable, TValue> outputFormat = createIntermediateOutputFormat();
        TaskAttemptContext outputTaskContext = new TaskAttemptContextImpl(job.getConfiguration(), createAttemptID(1, 2, 1));
        RecordWriter<LongWritable, TValue> writer = outputFormat.getRecordWriter(outputTaskContext);
        for (InputSplit split : splits) {
            // Initialize the input reading
            TaskAttemptContext inputTaskContext = new TaskAttemptContextImpl(job.getConfiguration(), createAttemptID(1, 1, 1));
            RecordReader<LongWritable, TValue> reader = inputFormat.createRecordReader(split, inputTaskContext);
            reader.initialize(split, inputTaskContext);
            // output
            while (reader.nextKeyValue()) {
                writer.write(reader.getCurrentKey(), reader.getCurrentValue());
            }
        }
        writer.close(outputTaskContext);
        // Promote outputs from temporary status
        promoteInputs(intermediateOutputDir);
        // Now we need to create a subsequent job that reads the
        // intermediate outputs
        // The Blank nodes should have been given separate identities so we
        // should not be conflating them, this is the opposite problem to
        // that described in JENA-820
        LOGGER.debug("Intermediate output directory is {}", intermediateOutputDir.getAbsolutePath());
        job = Job.getInstance(config);
        inputFormat = createIntermediateInputFormat();
        job.setInputFormatClass(inputFormat.getClass());
        NLineInputFormat.setNumLinesPerSplit(job, 100);
        FileInputFormat.setInputPaths(job, new Path(intermediateOutputDir.getAbsolutePath()));
        context = new JobContextImpl(job.getConfiguration(), job.getJobID());
        // Get the splits
        splits = inputFormat.getSplits(context);
        Assert.assertEquals(1, splits.size());
        // Expect to end up with a single blank node
        Set<Node> nodes = new HashSet<Node>();
        for (InputSplit split : splits) {
            TaskAttemptContext inputTaskContext = new TaskAttemptContextImpl(job.getConfiguration(), new TaskAttemptID());
            RecordReader<LongWritable, TValue> reader = inputFormat.createRecordReader(split, inputTaskContext);
            reader.initialize(split, inputTaskContext);
            while (reader.nextKeyValue()) {
                nodes.add(getSubject(reader.getCurrentValue().get()));
            }
        }
        // Nodes must not diverge
        Assert.assertEquals(2, nodes.size());
    } finally {
        a.delete();
        b.delete();
        deleteDirectory(intermediateOutputDir);
    }
}

Example 45 with TaskAttemptContextImpl

use of org.apache.hadoop.mapreduce.task.TaskAttemptContextImpl in project jena by apache.

the class AbstractBlankNodeTests method blank_node_divergence_01.

/**
     * Test that starts with two blank nodes with the same identity in a single
     * file, splits them over two files and checks that we can workaround
     * JENA-820 successfully by setting the
     * {@link RdfIOConstants#GLOBAL_BNODE_IDENTITY} flag for our subsequent job
     * 
     * @throws IOException
     * @throws InterruptedException
     */
@Test
public final void blank_node_divergence_01() throws IOException, InterruptedException {
    Assume.assumeTrue("Requires ParserProfile be respected", this.respectsParserProfile());
    Assume.assumeFalse("Requires that Blank Node identity not be preserved", this.preservesBlankNodeIdentity());
    // Temporary files
    File a = File.createTempFile("bnode_divergence", getInitialInputExtension());
    File intermediateOutputDir = Files.createTempDirectory("bnode_divergence", new FileAttribute[0]).toFile();
    try {
        // Prepare the input data
        // Two mentions of the same blank node in the same file
        List<T> tuples = new ArrayList<>();
        Node bnode = NodeFactory.createBlankNode();
        Node pred = NodeFactory.createURI("http://example.org/predicate");
        tuples.add(createTuple(bnode, pred, NodeFactory.createLiteral("first")));
        tuples.add(createTuple(bnode, pred, NodeFactory.createLiteral("second")));
        writeTuples(a, tuples);
        // Set up fake job which will process the file as a single split
        Configuration config = new Configuration(true);
        InputFormat<LongWritable, TValue> inputFormat = createInitialInputFormat();
        Job job = Job.getInstance(config);
        job.setInputFormatClass(inputFormat.getClass());
        NLineInputFormat.setNumLinesPerSplit(job, 100);
        FileInputFormat.setInputPaths(job, new Path(a.getAbsolutePath()));
        FileOutputFormat.setOutputPath(job, new Path(intermediateOutputDir.getAbsolutePath()));
        JobContext context = new JobContextImpl(job.getConfiguration(), job.getJobID());
        // Get the splits
        List<InputSplit> splits = inputFormat.getSplits(context);
        Assert.assertEquals(1, splits.size());
        for (InputSplit split : splits) {
            // Initialize the input reading
            TaskAttemptContext inputTaskContext = new TaskAttemptContextImpl(job.getConfiguration(), createAttemptID(1, 1, 1));
            RecordReader<LongWritable, TValue> reader = inputFormat.createRecordReader(split, inputTaskContext);
            reader.initialize(split, inputTaskContext);
            // Copy the input to the output - each triple goes to a separate
            // output file
            // This is how we force multiple files to be produced
            int taskID = 1;
            while (reader.nextKeyValue()) {
                // Prepare the output writing
                OutputFormat<LongWritable, TValue> outputFormat = createIntermediateOutputFormat();
                TaskAttemptContext outputTaskContext = new TaskAttemptContextImpl(job.getConfiguration(), createAttemptID(1, ++taskID, 1));
                RecordWriter<LongWritable, TValue> writer = outputFormat.getRecordWriter(outputTaskContext);
                writer.write(reader.getCurrentKey(), reader.getCurrentValue());
                writer.close(outputTaskContext);
            }
        }
        // Promote outputs from temporary status
        promoteInputs(intermediateOutputDir);
        // Now we need to create a subsequent job that reads the
        // intermediate outputs
        // As described in JENA-820 at this point the blank nodes are
        // consistent, however when we read them from different files they
        // by default get treated as different nodes and so the blank nodes
        // diverge which is incorrect and undesirable behaviour in
        // multi-stage pipelines
        LOGGER.debug("Intermediate output directory is {}", intermediateOutputDir.getAbsolutePath());
        job = Job.getInstance(config);
        inputFormat = createIntermediateInputFormat();
        job.setInputFormatClass(inputFormat.getClass());
        FileInputFormat.setInputPaths(job, new Path(intermediateOutputDir.getAbsolutePath()));
        // Enabling this flag works around the JENA-820 issue
        job.getConfiguration().setBoolean(RdfIOConstants.GLOBAL_BNODE_IDENTITY, true);
        context = new JobContextImpl(job.getConfiguration(), job.getJobID());
        // Get the splits
        splits = inputFormat.getSplits(context);
        Assert.assertEquals(2, splits.size());
        // Expect to end up with a single blank node
        Set<Node> nodes = new HashSet<Node>();
        for (InputSplit split : splits) {
            TaskAttemptContext inputTaskContext = new TaskAttemptContextImpl(job.getConfiguration(), new TaskAttemptID());
            RecordReader<LongWritable, TValue> reader = inputFormat.createRecordReader(split, inputTaskContext);
            reader.initialize(split, inputTaskContext);
            while (reader.nextKeyValue()) {
                nodes.add(getSubject(reader.getCurrentValue().get()));
            }
        }
        // Nodes should not have diverged
        Assert.assertEquals(1, nodes.size());
    } finally {
        a.delete();
        deleteDirectory(intermediateOutputDir);
    }
}