Class PartitionedFile


  • public class PartitionedFile
    extends Object
    PartitionedFile is the persistent file type of sort-merge based blocking shuffle. Each PartitionedFile contains two physical files: one is the data file and the other is the index file. Both the data file and the index file have multiple regions. Data belonging to the same subpartition are stored together in each data region and the corresponding index region contains index entries of all subpartitions. Each index entry is a (long, integer) value tuple of which the long value represents the file offset of the target subpartition and the integer value is the number of buffers.
    • Field Detail

      • INDEX_ENTRY_SIZE

        public static final int INDEX_ENTRY_SIZE
        Size of each index entry in the index file: 8 bytes for file offset and 8 bytes for data size in bytes.
        See Also:
        Constant Field Values
    • Constructor Detail

      • PartitionedFile

        public PartitionedFile​(int numRegions,
                               int numSubpartitions,
                               Path dataFilePath,
                               Path indexFilePath,
                               long dataFileSize,
                               long indexFileSize,
                               long numBuffers,
                               @Nullable
                               ByteBuffer indexEntryCache)
    • Method Detail

      • getDataFilePath

        public Path getDataFilePath()
      • getIndexFilePath

        public Path getIndexFilePath()
      • getNumRegions

        public int getNumRegions()
      • isReadable

        public boolean isReadable()
      • deleteQuietly

        public void deleteQuietly()