Examples with FileSplit org.apache.hadoop.mapred.FileSplit used on opensource projects

Example 21 with FileSplit

use of org.apache.hadoop.mapred.FileSplit in project asterixdb by apache.

the class SchedulerTest method testSchedulerSmallerHDFS.

/**
     * Test the case where the HDFS cluster is a larger than the Hyracks cluster
     *
     * @throws Exception
     */
public void testSchedulerSmallerHDFS() throws Exception {
    Map<String, NodeControllerInfo> ncNameToNcInfos = TestUtils.generateNodeControllerInfo(6, "nc", "10.0.0.", 5099, 5098, 5097);
    InputSplit[] fileSplits = new InputSplit[12];
    fileSplits[0] = new FileSplit(new Path("part-1"), 0, 0, new String[] { "10.0.0.1", "10.0.0.2", "10.0.0.3" });
    fileSplits[1] = new FileSplit(new Path("part-2"), 0, 0, new String[] { "10.0.0.3", "10.0.0.4", "10.0.0.5" });
    fileSplits[2] = new FileSplit(new Path("part-3"), 0, 0, new String[] { "10.0.0.4", "10.0.0.5", "10.0.0.3" });
    fileSplits[3] = new FileSplit(new Path("part-4"), 0, 0, new String[] { "10.0.0.2", "10.0.0.1", "10.0.0.3" });
    fileSplits[4] = new FileSplit(new Path("part-5"), 0, 0, new String[] { "10.0.0.3", "10.0.0.4", "10.0.0.5" });
    fileSplits[5] = new FileSplit(new Path("part-6"), 0, 0, new String[] { "10.0.0.2", "10.0.0.3", "10.0.0.5" });
    fileSplits[6] = new FileSplit(new Path("part-7"), 0, 0, new String[] { "10.0.0.1", "10.0.0.2", "10.0.0.3" });
    fileSplits[7] = new FileSplit(new Path("part-8"), 0, 0, new String[] { "10.0.0.3", "10.0.0.4", "10.0.0.5" });
    fileSplits[8] = new FileSplit(new Path("part-9"), 0, 0, new String[] { "10.0.0.4", "10.0.0.5", "10.0.0.1" });
    fileSplits[9] = new FileSplit(new Path("part-10"), 0, 0, new String[] { "10.0.0.2", "10.0.0.1", "10.0.0.2" });
    fileSplits[10] = new FileSplit(new Path("part-11"), 0, 0, new String[] { "10.0.0.3", "10.0.0.4", "10.0.0.5" });
    fileSplits[11] = new FileSplit(new Path("part-12"), 0, 0, new String[] { "10.0.0.2", "10.0.0.3", "10.0.0.5" });
    String[] expectedResults = new String[] { "nc1", "nc4", "nc4", "nc1", "nc3", "nc2", "nc2", "nc3", "nc5", "nc6", "nc5", "nc6" };
    Scheduler scheduler = new Scheduler(ncNameToNcInfos);
    String[] locationConstraints = scheduler.getLocationConstraints(fileSplits);
    for (int i = 0; i < locationConstraints.length; i++) {
        Assert.assertEquals(locationConstraints[i], expectedResults[i]);
    }
    ClusterTopology topology = parseTopology();
    scheduler = new Scheduler(ncNameToNcInfos, topology);
    locationConstraints = scheduler.getLocationConstraints(fileSplits);
    for (int i = 0; i < locationConstraints.length; i++) {
        Assert.assertEquals(locationConstraints[i], expectedResults[i]);
    }
}

Example 22 with FileSplit

use of org.apache.hadoop.mapred.FileSplit in project asterixdb by apache.

the class SchedulerTest method testSchedulerSimple.

/**
     * Test the scheduler for the case when the Hyracks cluster is the HDFS cluster
     *
     * @throws Exception
     */
public void testSchedulerSimple() throws Exception {
    Map<String, NodeControllerInfo> ncNameToNcInfos = TestUtils.generateNodeControllerInfo(6, "nc", "10.0.0.", 5099, 5098, 5097);
    InputSplit[] fileSplits = new InputSplit[6];
    fileSplits[0] = new FileSplit(new Path("part-1"), 0, 0, new String[] { "10.0.0.1", "10.0.0.2", "10.0.0.3" });
    fileSplits[1] = new FileSplit(new Path("part-2"), 0, 0, new String[] { "10.0.0.3", "10.0.0.4", "10.0.0.5" });
    fileSplits[2] = new FileSplit(new Path("part-3"), 0, 0, new String[] { "10.0.0.4", "10.0.0.5", "10.0.0.6" });
    fileSplits[3] = new FileSplit(new Path("part-4"), 0, 0, new String[] { "10.0.0.2", "10.0.0.1", "10.0.0.6" });
    fileSplits[4] = new FileSplit(new Path("part-5"), 0, 0, new String[] { "10.0.0.3", "10.0.0.4", "10.0.0.5" });
    fileSplits[5] = new FileSplit(new Path("part-6"), 0, 0, new String[] { "10.0.0.2", "10.0.0.3", "10.0.0.5" });
    String[] expectedResults = new String[] { "nc1", "nc4", "nc6", "nc2", "nc3", "nc5" };
    Scheduler scheduler = new Scheduler(ncNameToNcInfos);
    String[] locationConstraints = scheduler.getLocationConstraints(fileSplits);
    for (int i = 0; i < locationConstraints.length; i++) {
        Assert.assertEquals(locationConstraints[i], expectedResults[i]);
    }
    ClusterTopology topology = parseTopology();
    scheduler = new Scheduler(ncNameToNcInfos, topology);
    locationConstraints = scheduler.getLocationConstraints(fileSplits);
    for (int i = 0; i < locationConstraints.length; i++) {
        Assert.assertEquals(locationConstraints[i], expectedResults[i]);
    }
}

Example 23 with FileSplit

use of org.apache.hadoop.mapred.FileSplit in project asterixdb by apache.

the class SchedulerTest method testSchedulerLargerHDFS.

/**
     * Test the case where the HDFS cluster is a larger than the Hyracks cluster
     *
     * @throws Exception
     */
public void testSchedulerLargerHDFS() throws Exception {
    int dataPort = 5099;
    int resultPort = 5098;
    int messagingPort = 5097;
    Map<String, NodeControllerInfo> ncNameToNcInfos = TestUtils.generateNodeControllerInfo(4, "nc", "10.0.0.", dataPort, resultPort, messagingPort);
    ncNameToNcInfos.put("nc7", new NodeControllerInfo("nc7", NodeStatus.ALIVE, new NetworkAddress("10.0.0.7", dataPort), new NetworkAddress("10.0.0.5", resultPort), new NetworkAddress("10.0.0.5", messagingPort), 2));
    ncNameToNcInfos.put("nc12", new NodeControllerInfo("nc12", NodeStatus.ALIVE, new NetworkAddress("10.0.0.12", dataPort), new NetworkAddress("10.0.0.5", resultPort), new NetworkAddress("10.0.0.5", messagingPort), 2));
    InputSplit[] fileSplits = new InputSplit[12];
    fileSplits[0] = new FileSplit(new Path("part-1"), 0, 0, new String[] { "10.0.0.1", "10.0.0.2", "10.0.0.3" });
    fileSplits[1] = new FileSplit(new Path("part-2"), 0, 0, new String[] { "10.0.0.3", "10.0.0.4", "10.0.0.5" });
    fileSplits[2] = new FileSplit(new Path("part-3"), 0, 0, new String[] { "10.0.0.4", "10.0.0.5", "10.0.0.6" });
    fileSplits[3] = new FileSplit(new Path("part-4"), 0, 0, new String[] { "10.0.0.2", "10.0.0.1", "10.0.0.6" });
    fileSplits[4] = new FileSplit(new Path("part-5"), 0, 0, new String[] { "10.0.0.3", "10.0.0.4", "10.0.0.5" });
    fileSplits[5] = new FileSplit(new Path("part-6"), 0, 0, new String[] { "10.0.0.2", "10.0.0.3", "10.0.0.5" });
    fileSplits[6] = new FileSplit(new Path("part-7"), 0, 0, new String[] { "10.0.0.1", "10.0.0.2", "10.0.0.3" });
    fileSplits[7] = new FileSplit(new Path("part-8"), 0, 0, new String[] { "10.0.0.3", "10.0.0.4", "10.0.0.5" });
    fileSplits[8] = new FileSplit(new Path("part-12"), 0, 0, new String[] { "10.0.0.14", "10.0.0.11", "10.0.0.13" });
    fileSplits[9] = new FileSplit(new Path("part-10"), 0, 0, new String[] { "10.0.0.2", "10.0.0.1", "10.0.0.6" });
    fileSplits[10] = new FileSplit(new Path("part-11"), 0, 0, new String[] { "10.0.0.3", "10.0.0.4", "10.0.0.7" });
    fileSplits[11] = new FileSplit(new Path("part-9"), 0, 0, new String[] { "10.0.0.4", "10.0.0.5", "10.0.0.6" });
    Scheduler scheduler = new Scheduler(ncNameToNcInfos);
    String[] locationConstraints = scheduler.getLocationConstraints(fileSplits);
    String[] expectedResults = new String[] { "nc1", "nc4", "nc4", "nc1", "nc3", "nc2", "nc2", "nc3", "nc12", "nc7", "nc7", "nc12" };
    for (int i = 0; i < locationConstraints.length; i++) {
        Assert.assertEquals(locationConstraints[i], expectedResults[i]);
    }
    expectedResults = new String[] { "nc1", "nc4", "nc4", "nc1", "nc3", "nc2", "nc2", "nc3", "nc7", "nc12", "nc7", "nc12" };
    ClusterTopology topology = parseTopology();
    scheduler = new Scheduler(ncNameToNcInfos, topology);
    locationConstraints = scheduler.getLocationConstraints(fileSplits);
    for (int i = 0; i < locationConstraints.length; i++) {
        Assert.assertEquals(locationConstraints[i], expectedResults[i]);
    }
}

Example 24 with FileSplit

use of org.apache.hadoop.mapred.FileSplit in project asterixdb by apache.

the class SchedulerTest method testSchedulerSmallerHDFSOdd.

/**
     * Test the case where the HDFS cluster is a larger than the Hyracks cluster
     *
     * @throws Exception
     */
public void testSchedulerSmallerHDFSOdd() throws Exception {
    Map<String, NodeControllerInfo> ncNameToNcInfos = TestUtils.generateNodeControllerInfo(6, "nc", "10.0.0.", 5099, 5098, 5097);
    InputSplit[] fileSplits = new InputSplit[13];
    fileSplits[0] = new FileSplit(new Path("part-1"), 0, 0, new String[] { "10.0.0.1", "10.0.0.2", "10.0.0.3" });
    fileSplits[1] = new FileSplit(new Path("part-2"), 0, 0, new String[] { "10.0.0.3", "10.0.0.4", "10.0.0.5" });
    fileSplits[2] = new FileSplit(new Path("part-3"), 0, 0, new String[] { "10.0.0.4", "10.0.0.5", "10.0.0.3" });
    fileSplits[3] = new FileSplit(new Path("part-4"), 0, 0, new String[] { "10.0.0.2", "10.0.0.1", "10.0.0.3" });
    fileSplits[4] = new FileSplit(new Path("part-5"), 0, 0, new String[] { "10.0.0.3", "10.0.0.4", "10.0.0.5" });
    fileSplits[5] = new FileSplit(new Path("part-6"), 0, 0, new String[] { "10.0.0.2", "10.0.0.3", "10.0.0.5" });
    fileSplits[6] = new FileSplit(new Path("part-7"), 0, 0, new String[] { "10.0.0.1", "10.0.0.2", "10.0.0.3" });
    fileSplits[7] = new FileSplit(new Path("part-8"), 0, 0, new String[] { "10.0.0.3", "10.0.0.4", "10.0.0.5" });
    fileSplits[8] = new FileSplit(new Path("part-9"), 0, 0, new String[] { "10.0.0.4", "10.0.0.5", "10.0.0.1" });
    fileSplits[9] = new FileSplit(new Path("part-10"), 0, 0, new String[] { "10.0.0.2", "10.0.0.1", "10.0.0.2" });
    fileSplits[10] = new FileSplit(new Path("part-11"), 0, 0, new String[] { "10.0.0.3", "10.0.0.4", "10.0.0.5" });
    fileSplits[11] = new FileSplit(new Path("part-12"), 0, 0, new String[] { "10.0.0.2", "10.0.0.3", "10.0.0.5" });
    fileSplits[12] = new FileSplit(new Path("part-13"), 0, 0, new String[] { "10.0.0.2", "10.0.0.4", "10.0.0.5" });
    String[] expectedResults = new String[] { "nc1", "nc4", "nc4", "nc1", "nc3", "nc2", "nc2", "nc3", "nc5", "nc1", "nc5", "nc2", "nc4" };
    Scheduler scheduler = new Scheduler(ncNameToNcInfos);
    String[] locationConstraints = scheduler.getLocationConstraints(fileSplits);
    for (int i = 0; i < locationConstraints.length; i++) {
        Assert.assertEquals(locationConstraints[i], expectedResults[i]);
    }
    ClusterTopology topology = parseTopology();
    scheduler = new Scheduler(ncNameToNcInfos, topology);
    locationConstraints = scheduler.getLocationConstraints(fileSplits);
    for (int i = 0; i < locationConstraints.length; i++) {
        Assert.assertEquals(locationConstraints[i], expectedResults[i]);
    }
}

Example 25 with FileSplit

use of org.apache.hadoop.mapred.FileSplit in project drill by apache.

the class HiveDrillNativeScanBatchCreator method getBatch.

@Override
public ScanBatch getBatch(FragmentContext context, HiveDrillNativeParquetSubScan config, List<RecordBatch> children) throws ExecutionSetupException {
    final HiveTableWithColumnCache table = config.getTable();
    final List<InputSplit> splits = config.getInputSplits();
    final List<HivePartition> partitions = config.getPartitions();
    final List<SchemaPath> columns = config.getColumns();
    final String partitionDesignator = context.getOptions().getOption(ExecConstants.FILESYSTEM_PARTITION_COLUMN_LABEL).string_val;
    List<Map<String, String>> implicitColumns = Lists.newLinkedList();
    boolean selectAllQuery = AbstractRecordReader.isStarQuery(columns);
    final boolean hasPartitions = (partitions != null && partitions.size() > 0);
    final List<String[]> partitionColumns = Lists.newArrayList();
    final List<Integer> selectedPartitionColumns = Lists.newArrayList();
    List<SchemaPath> newColumns = columns;
    if (!selectAllQuery) {
        // Separate out the partition and non-partition columns. Non-partition columns are passed directly to the
        // ParquetRecordReader. Partition columns are passed to ScanBatch.
        newColumns = Lists.newArrayList();
        Pattern pattern = Pattern.compile(String.format("%s[0-9]+", partitionDesignator));
        for (SchemaPath column : columns) {
            Matcher m = pattern.matcher(column.getAsUnescapedPath());
            if (m.matches()) {
                selectedPartitionColumns.add(Integer.parseInt(column.getAsUnescapedPath().substring(partitionDesignator.length())));
            } else {
                newColumns.add(column);
            }
        }
    }
    final OperatorContext oContext = context.newOperatorContext(config);
    int currentPartitionIndex = 0;
    final List<RecordReader> readers = Lists.newArrayList();
    final HiveConf conf = config.getHiveConf();
    // TODO: In future we can get this cache from Metadata cached on filesystem.
    final Map<String, ParquetMetadata> footerCache = Maps.newHashMap();
    Map<String, String> mapWithMaxColumns = Maps.newLinkedHashMap();
    try {
        for (InputSplit split : splits) {
            final FileSplit fileSplit = (FileSplit) split;
            final Path finalPath = fileSplit.getPath();
            final JobConf cloneJob = new ProjectionPusher().pushProjectionsAndFilters(new JobConf(conf), finalPath.getParent());
            final FileSystem fs = finalPath.getFileSystem(cloneJob);
            ParquetMetadata parquetMetadata = footerCache.get(finalPath.toString());
            if (parquetMetadata == null) {
                parquetMetadata = ParquetFileReader.readFooter(cloneJob, finalPath);
                footerCache.put(finalPath.toString(), parquetMetadata);
            }
            final List<Integer> rowGroupNums = getRowGroupNumbersFromFileSplit(fileSplit, parquetMetadata);
            for (int rowGroupNum : rowGroupNums) {
                //DRILL-5009 : Skip the row group if the row count is zero
                if (parquetMetadata.getBlocks().get(rowGroupNum).getRowCount() == 0) {
                    continue;
                }
                // Drill has only ever written a single row group per file, only detect corruption
                // in the first row group
                ParquetReaderUtility.DateCorruptionStatus containsCorruptDates = ParquetReaderUtility.detectCorruptDates(parquetMetadata, config.getColumns(), true);
                if (logger.isDebugEnabled()) {
                    logger.debug(containsCorruptDates.toString());
                }
                readers.add(new ParquetRecordReader(context, Path.getPathWithoutSchemeAndAuthority(finalPath).toString(), rowGroupNum, fs, CodecFactory.createDirectCodecFactory(fs.getConf(), new ParquetDirectByteBufferAllocator(oContext.getAllocator()), 0), parquetMetadata, newColumns, containsCorruptDates));
                Map<String, String> implicitValues = Maps.newLinkedHashMap();
                if (hasPartitions) {
                    List<String> values = partitions.get(currentPartitionIndex).getValues();
                    for (int i = 0; i < values.size(); i++) {
                        if (selectAllQuery || selectedPartitionColumns.contains(i)) {
                            implicitValues.put(partitionDesignator + i, values.get(i));
                        }
                    }
                }
                implicitColumns.add(implicitValues);
                if (implicitValues.size() > mapWithMaxColumns.size()) {
                    mapWithMaxColumns = implicitValues;
                }
            }
            currentPartitionIndex++;
        }
    } catch (final IOException | RuntimeException e) {
        AutoCloseables.close(e, readers);
        throw new ExecutionSetupException("Failed to create RecordReaders. " + e.getMessage(), e);
    }
    // all readers should have the same number of implicit columns, add missing ones with value null
    mapWithMaxColumns = Maps.transformValues(mapWithMaxColumns, Functions.constant((String) null));
    for (Map<String, String> map : implicitColumns) {
        map.putAll(Maps.difference(map, mapWithMaxColumns).entriesOnlyOnRight());
    }
    // create an empty RecordReader to output the schema
    if (readers.size() == 0) {
        readers.add(new HiveDefaultReader(table, null, null, columns, context, conf, ImpersonationUtil.createProxyUgi(config.getUserName(), context.getQueryUserName())));
    }
    return new ScanBatch(config, context, oContext, readers.iterator(), implicitColumns);
}