Examples with OPTIONAL_INT org.apache.drill.common.types.Types.OPTIONAL_INT used on opensource projects

Example 1 with OPTIONAL_INT

use of org.apache.drill.common.types.Types.OPTIONAL_INT in project drill by apache.

the class DrillParquetReader method setup.

@Override
public void setup(OperatorContext context, OutputMutator output) throws ExecutionSetupException {
    try {
        this.operatorContext = context;
        schema = footer.getFileMetaData().getSchema();
        MessageType projection;
        final List<SchemaPath> columnsNotFound = new ArrayList<>(getColumns().size());
        if (isStarQuery()) {
            projection = schema;
        } else {
            projection = getProjection(schema, getColumns(), columnsNotFound);
            if (projection == null) {
                projection = schema;
            }
            if (!columnsNotFound.isEmpty()) {
                nullFilledVectors = new ArrayList<>(columnsNotFound.size());
                for (SchemaPath col : columnsNotFound) {
                    // col.toExpr() is used here as field name since we don't want to see these fields in the existing maps
                    nullFilledVectors.add(output.addField(MaterializedField.create(col.toExpr(), OPTIONAL_INT), NullableIntVector.class));
                }
                noColumnsFound = columnsNotFound.size() == getColumns().size();
            }
        }
        logger.debug("Requesting schema {}", projection);
        if (!noColumnsFound) {
            // Discard the columns not found in the schema when create DrillParquetRecordMaterializer, since they have been added to output already.
            @SuppressWarnings("unchecked") Collection<SchemaPath> columns = columnsNotFound.isEmpty() ? getColumns() : CollectionUtils.subtract(getColumns(), columnsNotFound);
            recordMaterializer = new DrillParquetRecordMaterializer(output, projection, columns, fragmentContext.getOptions(), containsCorruptedDates);
        }
        if (numRecordsToRead == 0 || noColumnsFound) {
            // no need to init readers
            return;
        }
        ColumnIOFactory factory = new ColumnIOFactory(false);
        MessageColumnIO columnIO = factory.getColumnIO(projection, schema);
        BlockMetaData blockMetaData = footer.getBlocks().get(entry.getRowGroupIndex());
        Map<ColumnPath, ColumnChunkMetaData> paths = blockMetaData.getColumns().stream().collect(Collectors.toMap(ColumnChunkMetaData::getPath, Function.identity(), (o, n) -> n));
        BufferAllocator allocator = operatorContext.getAllocator();
        CompressionCodecFactory ccf = DrillCompressionCodecFactory.createDirectCodecFactory(drillFileSystem.getConf(), new ParquetDirectByteBufferAllocator(allocator), 0);
        pageReadStore = new ColumnChunkIncReadStore(numRecordsToRead, ccf, allocator, drillFileSystem, entry.getPath());
        for (String[] path : schema.getPaths()) {
            Type type = schema.getType(path);
            if (type.isPrimitive()) {
                ColumnChunkMetaData md = paths.get(ColumnPath.get(path));
                pageReadStore.addColumn(schema.getColumnDescription(path), md);
            }
        }
        recordReader = columnIO.getRecordReader(pageReadStore, recordMaterializer);
    } catch (Exception e) {
        throw handleAndRaise("Failure in setting up reader", e);
    }
}