Examples with ReadSupport parquet.hadoop.api.ReadSupport used on opensource projects

Example 1 with ReadSupport

use of parquet.hadoop.api.ReadSupport in project presto by prestodb.

the class ParquetHiveRecordCursor method createParquetRecordReader.

private ParquetRecordReader<FakeParquetRecord> createParquetRecordReader(HdfsEnvironment hdfsEnvironment, String sessionUser, Configuration configuration, Path path, long start, long length, List<HiveColumnHandle> columns, boolean useParquetColumnNames, TypeManager typeManager, boolean predicatePushdownEnabled, TupleDomain<HiveColumnHandle> effectivePredicate) {
    ParquetDataSource dataSource = null;
    try {
        FileSystem fileSystem = hdfsEnvironment.getFileSystem(sessionUser, path, configuration);
        dataSource = buildHdfsParquetDataSource(fileSystem, path, start, length);
        ParquetMetadata parquetMetadata = hdfsEnvironment.doAs(sessionUser, () -> ParquetFileReader.readFooter(configuration, path, NO_FILTER));
        List<BlockMetaData> blocks = parquetMetadata.getBlocks();
        FileMetaData fileMetaData = parquetMetadata.getFileMetaData();
        MessageType fileSchema = fileMetaData.getSchema();
        PrestoReadSupport readSupport = new PrestoReadSupport(useParquetColumnNames, columns, fileSchema);
        List<parquet.schema.Type> fields = columns.stream().filter(column -> column.getColumnType() == REGULAR).map(column -> getParquetType(column, fileSchema, useParquetColumnNames)).filter(Objects::nonNull).collect(toList());
        MessageType requestedSchema = new MessageType(fileSchema.getName(), fields);
        LongArrayList offsets = new LongArrayList(blocks.size());
        for (BlockMetaData block : blocks) {
            long firstDataPage = block.getColumns().get(0).getFirstDataPageOffset();
            if (firstDataPage >= start && firstDataPage < start + length) {
                if (predicatePushdownEnabled) {
                    ParquetPredicate parquetPredicate = buildParquetPredicate(columns, effectivePredicate, fileMetaData.getSchema(), typeManager);
                    if (predicateMatches(parquetPredicate, block, dataSource, requestedSchema, effectivePredicate)) {
                        offsets.add(block.getStartingPos());
                    }
                } else {
                    offsets.add(block.getStartingPos());
                }
            }
        }
        ParquetInputSplit split = new ParquetInputSplit(path, start, start + length, length, null, offsets.toLongArray());
        TaskAttemptContext taskContext = ContextUtil.newTaskAttemptContext(configuration, new TaskAttemptID());
        return hdfsEnvironment.doAs(sessionUser, () -> {
            ParquetRecordReader<FakeParquetRecord> realReader = new PrestoParquetRecordReader(readSupport);
            realReader.initialize(split, taskContext);
            return realReader;
        });
    } catch (Exception e) {
        Throwables.propagateIfInstanceOf(e, PrestoException.class);
        if (e instanceof InterruptedException) {
            Thread.currentThread().interrupt();
            throw Throwables.propagate(e);
        }
        String message = format("Error opening Hive split %s (offset=%s, length=%s): %s", path, start, length, e.getMessage());
        if (e.getClass().getSimpleName().equals("BlockMissingException")) {
            throw new PrestoException(HIVE_MISSING_DATA, message, e);
        }
        throw new PrestoException(HIVE_CANNOT_OPEN_SPLIT, message, e);
    } finally {
        if (dataSource != null) {
            try {
                dataSource.close();
            } catch (IOException ignored) {
            }
        }
    }
}