Package org.apache.jackrabbit.oak.spi.blob

Class AbstractBlobStore

  • All Implemented Interfaces:
    AutoCloseable, Cache.Backend<AbstractBlobStore.BlockId,​AbstractBlobStore.Data>, BlobStore, GarbageCollectableBlobStore
    Direct Known Subclasses:
    CachingBlobStore, FileBlobStore, MemoryBlobStore
    public abstract class AbstractBlobStore
extends Object
implements GarbageCollectableBlobStore, Cache.Backend<AbstractBlobStore.BlockId,​AbstractBlobStore.Data>
    An abstract data store that splits the binaries in relatively small blocks, so that each block fits in memory.

    Each data store id is a list of zero or more entries. Each entry is either

    • data (a number of bytes), or
    • the hash code of the content of a number of bytes, or
    • the hash code of the content of a data store id (indirect hash)
    Thanks to the indirection, blocks can be kept relatively small, so that caching is simpler, and so that the storage backend doesn't need to support arbitrary size blobs (some storage backends buffer blobs in memory) and fast seeks (some storage backends re-read the whole blob when seeking).

    The format of a 'data' entry is: type (one byte; 0 for data), length (variable size int), data (bytes).

    The format of a 'hash of content' entry is: type (one byte; 1 for hash), level (variable size int, 0 meaning not nested), size (variable size long), hash code length (variable size int), hash code.

    The format of a 'hash of data store id' entry is: type (one byte; 1 for hash), level (variable size int, nesting level), total size (variable size long), size of data store id (variable size long), hash code length (variable size int), hash code.

    • Field Detail

      • BLOCK_SIZE_LIMIT

        protected static final int BLOCK_SIZE_LIMIT
        The minimum block size. Blocks must be larger than that so that the content hash is always shorter than the data itself.
        See Also:
        Constant Field Values

      • inUse

        protected Map<String,​WeakReference<String>> inUse
        The blob ids that are still floating around in memory. The blob store assumes such binaries must not be deleted, because those binaries are not referenced yet in a way the garbage collection algorithm can detect (not referenced at all, or only referenced in memory).

    • Constructor Detail

      • AbstractBlobStore

        public AbstractBlobStore()

    • Method Detail

      • setBlockSizeMin

        public void setBlockSizeMin​(int x)

      • setBlockSize

        public void setBlockSize​(int x)
        Description copied from interface: GarbageCollectableBlobStore
        Set the block size used by this blob store, if the blob store splits binaries into blocks. If not, this setting is ignored.
        Specified by:
        setBlockSize in interface GarbageCollectableBlobStore
        Parameters:
        x - the block size in bytes.

      • getBlockSize

        public int getBlockSize()

      • getInputStream

        public InputStream getInputStream​(String blobId)
                           throws IOException
        Description copied from interface: BlobStore
        Returns a new stream for given blobId. The streams returned from multiple calls to this method are byte wise equals. That is, subsequent calls to read return the same sequence of bytes as long as neither call throws an exception.
        Specified by:
        getInputStream in interface BlobStore
        Parameters:
        blobId - the blob id
        Returns:
        a new stream for given blobId
        Throws:
        IOException

      • getReference

        public String getReference​(@NotNull
                           @NotNull String blobId)
        Description copied from interface: BlobStore
        Returns a secure reference to blob referred by blobid, or null if no such reference is available.
        Specified by:
        getReference in interface BlobStore
        Parameters:
        blobId - blobId referring the blob for which reference is required
        Returns:
        binary reference, or null

      • getBlobId

        public String getBlobId​(@NotNull
                        @NotNull String reference)
        Description copied from interface: BlobStore
        Returns the blobId that referred by the given binary reference. Returns null if the reference is invalid, for example if it points to a blob that does not exist.
        Specified by:
        getBlobId in interface BlobStore
        Parameters:
        reference - binary reference
        Returns:
        matching blobId, or null

      • getOrCreateReferenceKey

        protected byte[] getOrCreateReferenceKey()
        Returns the reference key of this blob store. If one does not already exist, it is automatically created in an implementation-specific way. The default implementation simply creates a temporary random key that's valid only until the data store gets restarted. Subclasses can override and/or decorate this method to support a more persistent reference key.

        This method is called only once during the lifetime of a data store instance and the return value is cached in memory, so it's no problem if the implementation is slow.

        Returns:
        reference key

      • setReferenceKey

        public void setReferenceKey​(byte[] referenceKey)

      • setReferenceKeyEncoded

        public void setReferenceKeyEncoded​(String encodedKey)
        Set the referenceKey from Base64 encoded byte array
        Parameters:
        encodedKey - base64 encoded key

      • usesBlobId

        protected void usesBlobId​(String blobId)

      • storeBlock

        protected abstract void storeBlock​(byte[] digest,
                                   int level,
                                   byte[] data)
                            throws IOException
        Store a block of data.
        Parameters:
        digest - the content hash (32 bytes)
        level - the indirection level (0 is for user data, 1 is a list of digests that point to user data, 2 is a list of digests that point to digests, and so on). This parameter is for informational use only, and it is not required to store it unless that's easy to achieve
        data - the data to be stored (the number of bytes is at most the block size)
        Throws:
        IOException

      • isMarkEnabled

        protected abstract boolean isMarkEnabled()

      • readBlob

        public int readBlob​(String blobId,
                    long pos,
                    byte[] buff,
                    int off,
                    int length)
             throws IOException
        Description copied from interface: BlobStore
        Read a number of bytes from a blob.
        Specified by:
        readBlob in interface BlobStore
        Parameters:
        blobId - the blob id
        pos - the position within the blob
        buff - the target byte array
        off - the offset within the target array
        length - the number of bytes to read
        Returns:
        the number of bytes read
        Throws:
        IOException

      • readBlockFromBackend

        protected abstract byte[] readBlockFromBackend​(AbstractBlobStore.BlockId id)
                                        throws Exception
        Load the block from the storage backend. Returns null if the block was not found.
        Parameters:
        id - the block id
        Returns:
        the block data, or null
        Throws:
        Exception

      • deleteChunks

        public boolean deleteChunks​(List<String> chunkIds,
                            long maxLastModifiedTime)
                     throws Exception
        Description copied from interface: GarbageCollectableBlobStore
        Deletes the blobs with the given ids.
        Specified by:
        deleteChunks in interface GarbageCollectableBlobStore
        Parameters:
        chunkIds - the chunk ids
        maxLastModifiedTime - the max last modified time to consider for retrieval, with the special value '0' meaning no filtering by time
        Returns:
        true, if successful
        Throws:
        Exception - the exception