Package org.apache.flink.streaming.api.operators

Class AbstractStreamOperator<OUT>

    • Field Detail

      • LOG

        protected static final org.slf4j.Logger LOG
        The logger used by the operator class and its subclasses.

      • stateKeySelector1

        protected transient KeySelector<?,​?> stateKeySelector1
        KeySelector for extracting a key from an element being processed. This is used to scope keyed state to a key. This is null if the operator is not a keyed operator.

        This is for elements from the first input.

      • stateKeySelector2

        protected transient KeySelector<?,​?> stateKeySelector2
        KeySelector for extracting a key from an element being processed. This is used to scope keyed state to a key. This is null if the operator is not a keyed operator.

        This is for elements from the second input.

      • latencyStats

        protected transient LatencyStats latencyStats

      • lastRecordAttributes1

        protected transient RecordAttributes lastRecordAttributes1

      • lastRecordAttributes2

        protected transient RecordAttributes lastRecordAttributes2

    • Constructor Detail

      • AbstractStreamOperator

        public AbstractStreamOperator()

    • Method Detail

      • setProcessingTimeService

        protected void setProcessingTimeService​(ProcessingTimeService processingTimeService)

      • isUsingCustomRawKeyedState

        @Internal
protected boolean isUsingCustomRawKeyedState()
        Indicates whether or not implementations of this class is writing to the raw keyed state streams on snapshots, using snapshotState(StateSnapshotContext). If yes, subclasses should override this method to return true.

        Subclasses need to explicitly indicate the use of raw keyed state because, internally, the AbstractStreamOperator may attempt to read from it as well to restore heap-based timers and ultimately fail with read errors. By setting this flag to true, this allows the AbstractStreamOperator to know that the data written in the raw keyed states were not written by the timer services, and skips the timer restore attempt.

        Please refer to FLINK-19741 for further details.

        TODO: this method can be removed once all timers are moved to be managed by state backends.

        Returns:
        flag indicating whether or not this operator is writing to raw keyed state via snapshotState(StateSnapshotContext).

      • open

        public void open()
          throws Exception
        This method is called immediately before any elements are processed, it should contain the operator's initialization logic, e.g. state initialization.

        The default implementation does nothing.

        Specified by:
        open in interface StreamOperator<OUT>
        Throws:
        Exception - An exception in this method causes the operator to fail.

      • finish

        public void finish()
            throws Exception
        Description copied from interface: StreamOperator
        This method is called at the end of data processing.

        The method is expected to flush all remaining buffered data. Exceptions during this flushing of buffered data should be propagated, in order to cause the operation to be recognized as failed, because the last data items are not processed properly.

        After this method is called, no more records can be produced for the downstream operators.

        WARNING: It is not safe to use this method to commit any transactions or other side effects! You can use this method to flush any buffered data that can later on be committed e.g. in a CheckpointListener.notifyCheckpointComplete(long).

        NOTE:This method does not need to close any resources. You should release external resources in the StreamOperator.close() method.

        Specified by:
        finish in interface StreamOperator<OUT>
        Throws:
        Exception - An exception in this method causes the operator to fail.

      • close

        public void close()
           throws Exception
        Description copied from interface: StreamOperator
        This method is called at the very end of the operator's life, both in the case of a successful completion of the operation, and in the case of a failure and canceling.

        This method is expected to make a thorough effort to release all resources that the operator has acquired.

        NOTE:It can not emit any records! If you need to emit records at the end of processing, do so in the StreamOperator.finish() method.

        Specified by:
        close in interface StreamOperator<OUT>
        Throws:
        Exception

      • prepareSnapshotPreBarrier

        public void prepareSnapshotPreBarrier​(long checkpointId)
                               throws Exception
        Description copied from interface: StreamOperator
        This method is called when the operator should do a snapshot, before it emits its own checkpoint barrier.

        This method is intended not for any actual state persistence, but only for emitting some data before emitting the checkpoint barrier. Operators that maintain some small transient state that is inefficient to checkpoint (especially when it would need to be checkpointed in a re-scalable way) but can simply be sent downstream before the checkpoint. An example are opportunistic pre-aggregation operators, which have small the pre-aggregation state that is frequently flushed downstream.

        Important: This method should not be used for any actual state snapshot logic, because it will inherently be within the synchronous part of the operator's checkpoint. If heavy work is done within this method, it will affect latency and downstream checkpoint alignments.

        Specified by:
        prepareSnapshotPreBarrier in interface StreamOperator<OUT>
        Parameters:
        checkpointId - The ID of the checkpoint.
        Throws:
        Exception - Throwing an exception here causes the operator to fail and go into recovery.

      • notifyCheckpointComplete

        public void notifyCheckpointComplete​(long checkpointId)
                              throws Exception
        Description copied from interface: CheckpointListener
        Notifies the listener that the checkpoint with the given checkpointId completed and was committed.

        These notifications are "best effort", meaning they can sometimes be skipped. To behave properly, implementers need to follow the "Checkpoint Subsuming Contract". Please see the class-level JavaDocs for details.

        Please note that checkpoints may generally overlap, so you cannot assume that the notifyCheckpointComplete() call is always for the latest prior checkpoint (or snapshot) that was taken on the function/operator implementing this interface. It might be for a checkpoint that was triggered earlier. Implementing the "Checkpoint Subsuming Contract" (see above) properly handles this situation correctly as well.

        Please note that throwing exceptions from this method will not cause the completed checkpoint to be revoked. Throwing exceptions will typically cause task/job failure and trigger recovery.

        Specified by:
        notifyCheckpointComplete in interface CheckpointListener
        Parameters:
        checkpointId - The ID of the checkpoint that has been completed.
        Throws:
        Exception - This method can propagate exceptions, which leads to a failure/recovery for the task. Note that this will NOT lead to the checkpoint being revoked.

      • notifyCheckpointAborted

        public void notifyCheckpointAborted​(long checkpointId)
                             throws Exception
        Description copied from interface: CheckpointListener
        This method is called as a notification once a distributed checkpoint has been aborted.

        Important: The fact that a checkpoint has been aborted does NOT mean that the data and artifacts produced between the previous checkpoint and the aborted checkpoint are to be discarded. The expected behavior is as if this checkpoint was never triggered in the first place, and the next successful checkpoint simply covers a longer time span. See the "Checkpoint Subsuming Contract" in the class-level JavaDocs for details.

        These notifications are "best effort", meaning they can sometimes be skipped.

        This method is very rarely necessary to implement. The "best effort" guarantee, together with the fact that this method should not result in discarding any data (per the "Checkpoint Subsuming Contract") means it is mainly useful for earlier cleanups of auxiliary resources. One example is to pro-actively clear a local per-checkpoint state cache upon checkpoint failure.

        Specified by:
        notifyCheckpointAborted in interface CheckpointListener
        Parameters:
        checkpointId - The ID of the checkpoint that has been aborted.
        Throws:
        Exception - This method can propagate exceptions, which leads to a failure/recovery for the task or job.

      • getExecutionConfig

        public ExecutionConfig getExecutionConfig()
        Gets the execution config defined on the execution environment of the job to which this operator belongs.
        Returns:
        The job's execution config.

      • getOperatorConfig

        public StreamConfig getOperatorConfig()

      • getContainingTask

        public StreamTask<?,​?> getContainingTask()

      • getUserCodeClassloader

        public ClassLoader getUserCodeClassloader()

      • getOperatorName

        protected String getOperatorName()
        Return the operator name. If the runtime context has been set, then the task name with subtask index is returned. Otherwise, the simple class name is returned.
        Returns:
        If runtime context is set, then return task name with subtask index. Otherwise return simple class name.

      • getRuntimeContext

        @VisibleForTesting
public StreamingRuntimeContext getRuntimeContext()
        Returns a context that allows the operator to query information about the execution and also to interact with systems such as broadcast variables and managed state. This also allows to register timers.

      • getPartitionedState

        protected <S extends State> S getPartitionedState​(StateDescriptor<S,​?> stateDescriptor)
                                           throws Exception
        Creates a partitioned state handle, using the state backend configured for this task.
        Throws:
        IllegalStateException - Thrown, if the key/value state was already initialized.
        Exception - Thrown, if the state backend cannot create the key/value state.

      • getPartitionedState

        protected <S extends State,​N> S getPartitionedState​(N namespace,
                                                          TypeSerializer<N> namespaceSerializer,
                                                          StateDescriptor<S,​?> stateDescriptor)
                                                   throws Exception
        Creates a partitioned state handle, using the state backend configured for this task.
        Throws:
        IllegalStateException - Thrown, if the key/value state was already initialized.
        Exception - Thrown, if the state backend cannot create the key/value state.

      • getStateKeySelector1

        protected KeySelector<?,​?> getStateKeySelector1()

      • getStateKeySelector2

        protected KeySelector<?,​?> getStateKeySelector2()

      • reportOrForwardLatencyMarker

        protected void reportOrForwardLatencyMarker​(LatencyMarker marker)

      • getInternalTimerService

        public <K,​N> InternalTimerService<N> getInternalTimerService​(String name,
                                                                   TypeSerializer<N> namespaceSerializer,
                                                                   Triggerable<K,​N> triggerable)
        Returns a InternalTimerService that can be used to query current processing time and event time and to set timers. An operator can have several timer services, where each has its own namespace serializer. Timer services are differentiated by the string key that is given when requesting them, if you call this method with the same key multiple times you will get the same timer service instance in subsequent requests.

        Timers are always scoped to a key, the currently active key of a keyed stream operation. When a timer fires, this key will also be set as the currently active key.

        Each timer has attached metadata, the namespace. Different timer services can have a different namespace type. If you don't need namespace differentiation you can use VoidNamespaceSerializer as the namespace serializer.

        Type Parameters:
        N - The type of the timer namespace.
        Parameters:
        name - The name of the requested timer service. If no service exists under the given name a new one will be created and returned.
        namespaceSerializer - TypeSerializer for the timer namespace.
        triggerable - The Triggerable that should be invoked when timers fire