Examples with IndexSegment com.linkedin.pinot.core.indexsegment.IndexSegment used on opensource projects

Example 46 with IndexSegment

use of com.linkedin.pinot.core.indexsegment.IndexSegment in project pinot by linkedin.

the class SegmentDumpTool method doMain.

public void doMain(String[] args) throws Exception {
    CmdLineParser parser = new CmdLineParser(this);
    parser.parseArgument(args);
    File segmentDir = new File(segmentPath);
    SegmentMetadata metadata = new SegmentMetadataImpl(segmentDir);
    // All columns by default
    if (columnNames == null) {
        columnNames = new ArrayList<String>(metadata.getSchema().getColumnNames());
        Collections.sort(columnNames);
    }
    IndexSegment indexSegment = Loaders.IndexSegment.load(segmentDir, ReadMode.mmap);
    Map<String, Dictionary> dictionaries = new HashMap<String, Dictionary>();
    Map<String, BlockSingleValIterator> iterators = new HashMap<String, BlockSingleValIterator>();
    for (String columnName : columnNames) {
        DataSource dataSource = indexSegment.getDataSource(columnName);
        dataSource.open();
        Block block = dataSource.nextBlock();
        BlockValSet blockValSet = block.getBlockValueSet();
        BlockSingleValIterator itr = (BlockSingleValIterator) blockValSet.iterator();
        iterators.put(columnName, itr);
        dictionaries.put(columnName, dataSource.getDictionary());
    }
    System.out.print("Doc\t");
    for (String columnName : columnNames) {
        System.out.print(columnName);
        System.out.print("\t");
    }
    System.out.println();
    for (int i = 0; i < indexSegment.getSegmentMetadata().getTotalDocs(); i++) {
        System.out.print(i);
        System.out.print("\t");
        for (String columnName : columnNames) {
            FieldSpec.DataType columnType = metadata.getSchema().getFieldSpecFor(columnName).getDataType();
            BlockSingleValIterator itr = iterators.get(columnName);
            Integer encodedValue = itr.nextIntVal();
            Object value = dictionaries.get(columnName).get(encodedValue);
            System.out.print(value);
            System.out.print("\t");
        }
        System.out.println();
    }
    if (dumpStarTree) {
        System.out.println();
        File starTreeFile = new File(segmentDir, V1Constants.STAR_TREE_INDEX_FILE);
        StarTreeInterf tree = StarTreeSerDe.fromFile(starTreeFile, ReadMode.mmap);
        tree.printTree();
    }
}

Example 47 with IndexSegment

use of com.linkedin.pinot.core.indexsegment.IndexSegment in project pinot by linkedin.

the class DataFetcherTest method setup.

@BeforeClass
private void setup() throws Exception {
    GenericRow[] segmentData = new GenericRow[NUM_ROWS];
    // Generate random dimension and metric values.
    for (int i = 0; i < NUM_ROWS; i++) {
        double randomDouble = _random.nextDouble();
        String randomDoubleString = String.valueOf(randomDouble);
        _dimensionValues[i] = randomDoubleString;
        _intMetricValues[i] = (int) randomDouble;
        _longMetricValues[i] = (long) randomDouble;
        _floatMetricValues[i] = (float) randomDouble;
        _doubleMetricValues[i] = randomDouble;
        HashMap<String, Object> map = new HashMap<>();
        map.put(DIMENSION_NAME, _dimensionValues[i]);
        map.put(INT_METRIC_NAME, _intMetricValues[i]);
        map.put(LONG_METRIC_NAME, _longMetricValues[i]);
        map.put(FLOAT_METRIC_NAME, _floatMetricValues[i]);
        map.put(DOUBLE_METRIC_NAME, _doubleMetricValues[i]);
        map.put(NO_DICT_INT_METRIC_NAME, _intMetricValues[i]);
        map.put(NO_DICT_LONG_METRIC_NAME, _longMetricValues[i]);
        map.put(NO_DICT_FLOAT_METRIC_NAME, _floatMetricValues[i]);
        map.put(NO_DICT_DOUBLE_METRIC_NAME, _doubleMetricValues[i]);
        GenericRow genericRow = new GenericRow();
        genericRow.init(map);
        segmentData[i] = genericRow;
    }
    // Create an index segment with the random dimension and metric values.
    final Schema schema = new Schema();
    schema.addField(new DimensionFieldSpec(DIMENSION_NAME, FieldSpec.DataType.STRING, true));
    schema.addField(new MetricFieldSpec(INT_METRIC_NAME, FieldSpec.DataType.INT));
    schema.addField(new MetricFieldSpec(LONG_METRIC_NAME, FieldSpec.DataType.LONG));
    schema.addField(new MetricFieldSpec(FLOAT_METRIC_NAME, FieldSpec.DataType.FLOAT));
    schema.addField(new MetricFieldSpec(DOUBLE_METRIC_NAME, FieldSpec.DataType.DOUBLE));
    schema.addField(new MetricFieldSpec(NO_DICT_INT_METRIC_NAME, FieldSpec.DataType.INT));
    schema.addField(new MetricFieldSpec(NO_DICT_LONG_METRIC_NAME, FieldSpec.DataType.LONG));
    schema.addField(new MetricFieldSpec(NO_DICT_FLOAT_METRIC_NAME, FieldSpec.DataType.FLOAT));
    schema.addField(new MetricFieldSpec(NO_DICT_DOUBLE_METRIC_NAME, FieldSpec.DataType.DOUBLE));
    SegmentGeneratorConfig config = new SegmentGeneratorConfig(schema);
    FileUtils.deleteQuietly(new File(INDEX_DIR_PATH));
    config.setOutDir(INDEX_DIR_PATH);
    config.setSegmentName(SEGMENT_NAME);
    config.setRawIndexCreationColumns(Arrays.asList(NO_DICT_INT_METRIC_NAME, NO_DICT_LONG_METRIC_NAME, NO_DICT_FLOAT_METRIC_NAME, NO_DICT_DOUBLE_METRIC_NAME));
    SegmentIndexCreationDriverImpl driver = new SegmentIndexCreationDriverImpl();
    driver.init(config, new TestDataRecordReader(schema, segmentData));
    driver.build();
    IndexSegment indexSegment = Loaders.IndexSegment.load(new File(INDEX_DIR_PATH, SEGMENT_NAME), ReadMode.heap);
    Map<String, BaseOperator> dataSourceMap = new HashMap<>();
    for (String column : indexSegment.getColumnNames()) {
        dataSourceMap.put(column, indexSegment.getDataSource(column));
    }
    // Get a data fetcher for the index segment.
    _dataFetcher = new DataFetcher(dataSourceMap);
}

Example 48 with IndexSegment

use of com.linkedin.pinot.core.indexsegment.IndexSegment in project pinot by linkedin.

the class StarTreeOnHeapToOffHeapConverter method execute.

@Override
public boolean execute() throws Exception {
    File indexDir = new File(_segmentDir);
    long start = System.currentTimeMillis();
    LOGGER.info("Loading segment {}", indexDir.getName());
    IndexSegment segment = Loaders.IndexSegment.load(indexDir, ReadMode.heap);
    long end = System.currentTimeMillis();
    LOGGER.info("Loaded segment {} in {} ms ", indexDir.getName(), (end - start));
    start = end;
    StarTreeInterf starTreeOnHeap = segment.getStarTree();
    File starTreeOffHeapFile = new File(TMP_DIR, (V1Constants.STAR_TREE_INDEX_FILE + System.currentTimeMillis()));
    // Convert the star tree on-heap to off-heap format.
    StarTreeSerDe.writeTreeOffHeapFormat(starTreeOnHeap, starTreeOffHeapFile);
    // Copy all the indexes into output directory.
    File outputDir = new File(_outputDir);
    FileUtils.deleteQuietly(outputDir);
    FileUtils.copyDirectory(indexDir, outputDir);
    // Delete the existing star tree on-heap file from the output directory.
    FileUtils.deleteQuietly(new File(_outputDir, V1Constants.STAR_TREE_INDEX_FILE));
    // Move the temp star tree off-heap file into the output directory.
    FileUtils.moveFile(starTreeOffHeapFile, new File(_outputDir, V1Constants.STAR_TREE_INDEX_FILE));
    end = System.currentTimeMillis();
    LOGGER.info("Converted segment: {} ms", (end - start));
    return true;
}

Example 49 with IndexSegment

use of com.linkedin.pinot.core.indexsegment.IndexSegment in project pinot by linkedin.

the class SegmentInfoProvider method readOneSegment.

/**
   * Read the metadata of the given segmentFile and collect:
   * - Unique metric columns
   * - Unique single-value dimension columns
   * - Unique values for each single-value dimension columns
   *
   * @param segmentFile segment file.
   * @param uniqueMetrics unique metric columns buffer.
   * @param uniqueSingleValueDimensions unique single-value dimension columns buffer.
   * @param singleValueDimensionValuesMap single-value dimension columns to unique values map buffer.
   * @throws Exception
   */
private void readOneSegment(File segmentFile, Set<String> uniqueMetrics, Set<String> uniqueSingleValueDimensions, Map<String, Set<Object>> singleValueDimensionValuesMap) throws Exception {
    // Get segment directory from segment file (decompress if necessary).
    File segmentDir;
    File tmpDir = null;
    if (segmentFile.isFile()) {
        tmpDir = File.createTempFile(SEGMENT_INFO_PROVIDER, null, new File(TMP_DIR));
        FileUtils.deleteQuietly(tmpDir);
        tmpDir.mkdir();
        TarGzCompressionUtils.unTar(segmentFile, tmpDir);
        segmentDir = tmpDir.listFiles()[0];
    } else {
        segmentDir = segmentFile;
    }
    IndexSegment indexSegment = Loaders.IndexSegment.load(segmentDir, ReadMode.heap);
    Schema schema = indexSegment.getSegmentMetadata().getSchema();
    // Add time column if exists.
    String timeColumn = schema.getTimeColumnName();
    if (timeColumn != null) {
        uniqueSingleValueDimensions.add(timeColumn);
        loadValuesForSingleValueDimension(indexSegment, singleValueDimensionValuesMap, timeColumn);
    }
    // Add all metric columns.
    uniqueMetrics.addAll(schema.getMetricNames());
    // Add all single-value dimension columns.
    for (DimensionFieldSpec fieldSpec : schema.getDimensionFieldSpecs()) {
        if (!fieldSpec.isSingleValueField()) {
            continue;
        }
        String column = fieldSpec.getName();
        uniqueSingleValueDimensions.add(column);
        loadValuesForSingleValueDimension(indexSegment, singleValueDimensionValuesMap, column);
    }
    if (tmpDir != null) {
        FileUtils.deleteQuietly(tmpDir);
    }
}