parquet.hadoop.metadata.ColumnChunkMetaData Java Examples

public Block readPrimitive(ColumnDescriptor columnDescriptor, Type type, IntList offsets)
        throws IOException
{
    ParquetColumnReader columnReader = columnReadersMap.get(columnDescriptor);
    if (columnReader.getPageReader() == null) {
        validateParquet(currentBlockMetadata.getRowCount() > 0, "Row group has 0 rows");
        ColumnChunkMetaData metadata = getColumnChunkMetaData(columnDescriptor);
        long startingPosition = metadata.getStartingPos();
        int totalSize = checkedCast(metadata.getTotalSize());
        byte[] buffer = new byte[totalSize];
        dataSource.readFully(startingPosition, buffer);
        ParquetColumnChunkDescriptor descriptor = new ParquetColumnChunkDescriptor(columnDescriptor, metadata, totalSize);
        ParquetColumnChunk columnChunk = new ParquetColumnChunk(descriptor, buffer, 0);
        columnReader.setPageReader(columnChunk.readAllPages());
    }
    return columnReader.readPrimitive(type, offsets);
}
 

public static void showDetails(PrettyPrintWriter out, List<ColumnChunkMetaData> ccmeta) {
  Map<String,Object> chunks = new LinkedHashMap<String,Object>();
  for (ColumnChunkMetaData cmeta : ccmeta) {
    String[] path = cmeta.getPath().toArray();

    Map<String,Object> current = chunks;
    for (int i = 0; i < path.length - 1; ++i) {
      String next = path[i];
      if (!current.containsKey(next)) {
        current.put(next, new LinkedHashMap<String,Object>());
      }

      current = (Map<String,Object>)current.get(next);
    }

    current.put(path[path.length - 1], cmeta);
  }

  showColumnChunkDetails(out, chunks, 0);
}
 

private static void showDetails(PrettyPrintWriter out, ColumnChunkMetaData meta, boolean name) {
  long doff = meta.getDictionaryPageOffset();
  long foff = meta.getFirstDataPageOffset();
  long tsize = meta.getTotalSize();
  long usize = meta.getTotalUncompressedSize();
  long count = meta.getValueCount();
  double ratio = usize / (double)tsize;
  String encodings = Joiner.on(',').skipNulls().join(meta.getEncodings());

  if (name) {
    String path = Joiner.on('.').skipNulls().join(meta.getPath());
    out.format("%s: ", path);
  }

  out.format(" %s", meta.getType());
  out.format(" %s", meta.getCodec());
  out.format(" DO:%d", doff);
  out.format(" FPO:%d", foff);
  out.format(" SZ:%d/%d/%.2f", tsize, usize, ratio);
  out.format(" VC:%d", count);
  if (!encodings.isEmpty()) out.format(" ENC:%s", encodings);
  out.println();
}
 

private ColumnChunkMetaData getColumnChunkMetaData(ColumnDescriptor columnDescriptor)
        throws IOException
{
    for (ColumnChunkMetaData metadata : currentBlockMetadata.getColumns()) {
        if (metadata.getPath().equals(ColumnPath.get(columnDescriptor.getPath()))) {
            return metadata;
        }
    }
    throw new ParquetCorruptionException("Metadata is missing for column: %s", columnDescriptor);
}
 

private static void showColumnChunkDetails(PrettyPrintWriter out, Map<String,Object> current, int depth) {
  for (Map.Entry<String,Object> entry : current.entrySet()) {
    String name = Strings.repeat(".", depth) + entry.getKey();
    Object value = entry.getValue();

    if (value instanceof Map) {
      out.println(name + ": ");
      showColumnChunkDetails(out, (Map<String,Object>)value, depth + 1);
    } else {
      out.print(name + ": ");
      showDetails(out, (ColumnChunkMetaData)value, false);
    }
  }
}
 

public static ParquetMetadata readFooter(FileSystem fileSystem, Path file)
        throws IOException
{
    FileStatus fileStatus = fileSystem.getFileStatus(file);
    try (FSDataInputStream inputStream = fileSystem.open(file)) {
        // Parquet File Layout:
        //
        // MAGIC
        // variable: Data
        // variable: Metadata
        // 4 bytes: MetadataLength
        // MAGIC

        long length = fileStatus.getLen();
        validateParquet(length >= MAGIC.length + PARQUET_METADATA_LENGTH + MAGIC.length, "%s is not a valid Parquet File", file);
        long metadataLengthIndex = length - PARQUET_METADATA_LENGTH - MAGIC.length;

        inputStream.seek(metadataLengthIndex);
        int metadataLength = readIntLittleEndian(inputStream);

        byte[] magic = new byte[MAGIC.length];
        inputStream.readFully(magic);
        validateParquet(Arrays.equals(MAGIC, magic), "Not valid Parquet file: %s expected magic number: %s got: %s", file, Arrays.toString(MAGIC), Arrays.toString(magic));

        long metadataIndex = metadataLengthIndex - metadataLength;
        validateParquet(
                metadataIndex >= MAGIC.length && metadataIndex < metadataLengthIndex,
                "Corrupted Parquet file: %s metadata index: %s out of range",
                file,
                metadataIndex);
        inputStream.seek(metadataIndex);
        FileMetaData fileMetaData = readFileMetaData(inputStream);
        List<SchemaElement> schema = fileMetaData.getSchema();
        validateParquet(!schema.isEmpty(), "Empty Parquet schema in file: %s", file);

        MessageType messageType = readParquetSchema(schema);
        List<BlockMetaData> blocks = new ArrayList<>();
        List<RowGroup> rowGroups = fileMetaData.getRow_groups();
        if (rowGroups != null) {
            for (RowGroup rowGroup : rowGroups) {
                BlockMetaData blockMetaData = new BlockMetaData();
                blockMetaData.setRowCount(rowGroup.getNum_rows());
                blockMetaData.setTotalByteSize(rowGroup.getTotal_byte_size());
                List<ColumnChunk> columns = rowGroup.getColumns();
                validateParquet(!columns.isEmpty(), "No columns in row group: %s", rowGroup);
                String filePath = columns.get(0).getFile_path();
                for (ColumnChunk columnChunk : columns) {
                    validateParquet(
                            (filePath == null && columnChunk.getFile_path() == null)
                                    || (filePath != null && filePath.equals(columnChunk.getFile_path())),
                            "all column chunks of the same row group must be in the same file");
                    ColumnMetaData metaData = columnChunk.meta_data;
                    String[] path = metaData.path_in_schema.toArray(new String[metaData.path_in_schema.size()]);
                    ColumnPath columnPath = ColumnPath.get(path);
                    ColumnChunkMetaData column = ColumnChunkMetaData.get(
                            columnPath,
                            messageType.getType(columnPath.toArray()).asPrimitiveType().getPrimitiveTypeName(),
                            CompressionCodecName.fromParquet(metaData.codec),
                            readEncodings(metaData.encodings),
                            readStats(metaData.statistics, messageType.getType(columnPath.toArray()).asPrimitiveType().getPrimitiveTypeName()),
                            metaData.data_page_offset,
                            metaData.dictionary_page_offset,
                            metaData.num_values,
                            metaData.total_compressed_size,
                            metaData.total_uncompressed_size);
                    blockMetaData.addColumn(column);
                }
                blockMetaData.setPath(filePath);
                blocks.add(blockMetaData);
            }
        }

        Map<String, String> keyValueMetaData = new HashMap<>();
        List<KeyValue> keyValueList = fileMetaData.getKey_value_metadata();
        if (keyValueList != null) {
            for (KeyValue keyValue : keyValueList) {
                keyValueMetaData.put(keyValue.key, keyValue.value);
            }
        }
        return new ParquetMetadata(new parquet.hadoop.metadata.FileMetaData(messageType, keyValueMetaData, fileMetaData.getCreated_by()), blocks);
    }
}
 

public static void showDetails(PrettyPrintWriter out, ColumnChunkMetaData meta) {
  showDetails(out, meta, true);
}