Package org.apache.beam.sdk.io

Class UnboundedSource.UnboundedReader<OutputT>

    • Field Detail

      • BACKLOG_UNKNOWN

        public static final long BACKLOG_UNKNOWN
        Constant representing an unknown amount of backlog.
        See Also:
        Constant Field Values

    • Constructor Detail

      • UnboundedReader

        public UnboundedReader()

    • Method Detail

      • start

        public abstract boolean start()
                       throws java.io.IOException
        Initializes the reader and advances the reader to the first record. If the reader has been restored from a checkpoint then it should advance to the next unread record at the point the checkpoint was taken.

        This method will be called exactly once. The invocation will occur prior to calling advance() or Source.Reader.getCurrent(). This method may perform expensive operations that are needed to initialize the reader.

        Returns true if a record was read, false if there is no more input currently available. Future calls to advance() may return true once more data is available. Regardless of the return value of start, start will not be called again on the same UnboundedReader object; it will only be called again when a new reader object is constructed for the same source, e.g. on recovery.

        Specified by:
        start in class Source.Reader<OutputT>
        Returns:
        true if a record was read, false if there is no more input available.
        Throws:
        java.io.IOException

      • advance

        public abstract boolean advance()
                         throws java.io.IOException
        Advances the reader to the next valid record.

        Returns true if a record was read, false if there is no more input available. Future calls to advance() may return true once more data is available.

        Specified by:
        advance in class Source.Reader<OutputT>
        Returns:
        true if a record was read, false if there is no more input available.
        Throws:
        java.io.IOException

      • getCurrentRecordId

        public byte[] getCurrentRecordId()
                          throws java.util.NoSuchElementException
        Returns a unique identifier for the current record. This should be the same for each instance of the same logical record read from the underlying data source.

        It is only necessary to override this if UnboundedSource.requiresDeduping() has been overridden to return true.

        For example, this could be a hash of the record contents, or a logical ID present in the record. If this is generated as a hash of the record contents, it should be at least 16 bytes (128 bits) to avoid collisions.

        This method has the same restrictions on when it can be called as Source.Reader.getCurrent() and Source.Reader.getCurrentTimestamp().

        Throws:
        java.util.NoSuchElementException - if the reader is at the beginning of the input and start() or advance() wasn't called, or if the last start() or advance() returned false.

      • getWatermark

        public abstract org.joda.time.Instant getWatermark()
        Returns a timestamp before or at the timestamps of all future elements read by this reader.

        This can be approximate. If records are read that violate this guarantee, they will be considered late, which will affect how they will be processed. See Window for more information on late data and how to handle it.

        However, this value should be as late as possible. Downstream windows may not be able to close until this watermark passes their end.

        For example, a source may know that the records it reads will be in timestamp order. In this case, the watermark can be the timestamp of the last record read. For a source that does not have natural timestamps, timestamps can be set to the time of reading, in which case the watermark is the current clock time.

        See Window and Trigger for more information on timestamps and watermarks.

        May be called after advance() or start() has returned false, but not before start() has been called.

      • getCheckpointMark

        public abstract UnboundedSource.CheckpointMark getCheckpointMark()
        Returns a UnboundedSource.CheckpointMark representing the progress of this UnboundedReader.

        If this UnboundedReader does not support checkpoints, it may return a CheckpointMark which does nothing, like: 

        
 public UnboundedSource.CheckpointMark getCheckpointMark() {
   return new UnboundedSource.CheckpointMark() {
     public void finalizeCheckpoint() throws IOException {
       // nothing to do
     }
   };
 }

        All elements read between the last time this method was called (or since this reader was created, if this method has not been called on this reader) until this method is called will be processed together as a bundle. (An element is considered 'read' if it could be returned by a call to Source.Reader.getCurrent().)

        Once the result of processing those elements and the returned checkpoint have been durably committed, UnboundedSource.CheckpointMark.finalizeCheckpoint() will be called at most once at some later point on the returned UnboundedSource.CheckpointMark object. Checkpoint finalization is best-effort, and checkpoints may not be finalized. If duplicate elements may be produced if checkpoints are not finalized in a timely manner, UnboundedSource.requiresDeduping() should be overridden to return true, and getCurrentRecordId() should be overridden to return unique record IDs.

        A checkpoint will be committed to durable storage only if all all previous checkpoints produced by the same reader have also been committed.

        The returned object should not be modified.

        May not be called before start() has been called.

      • getSplitBacklogBytes

        public long getSplitBacklogBytes()
        Returns the size of the backlog of unread data in the underlying data source represented by this split of this source.

        One of this or getTotalBacklogBytes() should be overridden in order to allow the runner to scale the amount of resources allocated to the pipeline.

      • getTotalBacklogBytes

        public long getTotalBacklogBytes()
        Returns the size of the backlog of unread data in the underlying data source represented by all splits of this source.

        One of this or getSplitBacklogBytes() should be overridden in order to allow the runner to scale the amount of resources allocated to the pipeline.