Class ByteBlockPool

  • All Implemented Interfaces:

    public final class ByteBlockPool
    extends Object
    implements Accountable
    Class that Posting and PostingVector use to write byte streams into shared fixed-size byte[] arrays. The idea is to allocate slices of increasing lengths For example, the first slice is 5 bytes, the next slice is 14, etc. We start by writing our bytes into the first 5 bytes. When we hit the end of the slice, we allocate the next slice and then write the address of the new slice into the last 4 bytes of the previous slice (the "forwarding address").

    Each slice is filled with 0's initially, and we mark the end with a non-zero byte. This way the methods that are writing into the slice don't need to record its length and instead allocate a new slice once they hit a non-zero byte.

    NOTE: This API is for internal purposes only and might change in incompatible ways in the next release.
    • Field Detail

      • buffers

        public byte[][] buffers
        array of buffers currently used in the pool. Buffers are allocated if needed don't modify this outside of this class.
      • byteUpto

        public int byteUpto
        Where we are in head buffer
      • buffer

        public byte[] buffer
        Current head buffer
      • byteOffset

        public int byteOffset
        Current head offset

        public static final int[] NEXT_LEVEL_ARRAY
        An array holding the offset into the LEVEL_SIZE_ARRAY to quickly navigate to the next slice level.

        public static final int[] LEVEL_SIZE_ARRAY
        An array holding the level sizes for byte slices.

        public static final int FIRST_LEVEL_SIZE
        The first level size for new slices
        See Also:
    • Method Detail

      • reset

        public void reset​(boolean zeroFillBuffers,
                          boolean reuseFirst)
        Expert: Resets the pool to its initial state reusing the first buffer. Calling nextBuffer() is not needed after reset.
        zeroFillBuffers - if true the buffers are filled with 0. This should be set to true if this pool is used with slices.
        reuseFirst - if true the first buffer will be reused and calling nextBuffer() is not needed after reset iff the block pool was used before ie. nextBuffer() was called before.
      • nextBuffer

        public void nextBuffer()
        Advances the pool to its next buffer. This method should be called once after the constructor to initialize the pool. In contrast to the constructor a reset() call will advance the pool to its first buffer immediately.
      • newSlice

        public int newSlice​(int size)
        Allocates a new slice with the given size.
        See Also:
      • allocSlice

        public int allocSlice​(byte[] slice,
                              int upto)
        Creates a new byte slice with the given starting size and returns the slices offset in the pool.
      • allocKnownSizeSlice

        public int allocKnownSizeSlice​(byte[] slice,
                                       int upto)
        Create a new byte slice with the given starting size return the slice offset in the pool and length. The lower 8 bits of the returned int represent the length of the slice, and the upper 24 bits represent the offset.
      • setBytesRef

        public void setBytesRef​(BytesRef term,
                                int textStart)
      • append

        public void append​(BytesRef bytes)
        Appends the bytes in the provided BytesRef at the current position.
      • readBytes

        public void readBytes​(long offset,
                              byte[] bytes,
                              int bytesOffset,
                              int bytesLength)
        Reads bytes out of the pool starting at the given offset with the given length into the given byte array at offset off.

        Note: this method allows to copy across block boundaries.

      • setRawBytesRef

        public void setRawBytesRef​(BytesRef ref,
                                   long offset)
        Set the given BytesRef so that its content is equal to the ref.length bytes starting at offset. Most of the time this method will set pointers to internal data-structures. However, in case a value crosses a boundary, a fresh copy will be returned. On the contrary to setBytesRef(BytesRef, int), this does not expect the length to be encoded with the data.
      • ramBytesUsed

        public long ramBytesUsed()
        Description copied from interface: Accountable
        Return the memory usage of this object in bytes. Negative values are illegal.
        Specified by:
        ramBytesUsed in interface Accountable