Package org.apache.flink.datastream.api.stream

Interface BroadcastStream<T>

All Superinterfaces:

DataStream

@Experimental
public interface BroadcastStream<T>
extends DataStream

This interface represents a stream that each parallel task processes the same data.

Method Summary

Modifier and Type	Method	Description
<K, T_OTHER, OUT> NonKeyedPartitionStream.ProcessConfigurableAndNonKeyedPartitionStream<OUT>	connectAndProcess(KeyedPartitionStream<K,T_OTHER> other, TwoInputBroadcastStreamProcessFunction<T_OTHER,T,OUT> processFunction)	Apply a two input operation to this and other KeyedPartitionStream.
<K, T_OTHER, OUT> KeyedPartitionStream.ProcessConfigurableAndKeyedPartitionStream<K,OUT>	connectAndProcess(KeyedPartitionStream<K,T_OTHER> other, TwoInputBroadcastStreamProcessFunction<T_OTHER,T,OUT> processFunction, org.apache.flink.api.java.functions.KeySelector<OUT,K> newKeySelector)	Apply a two input operation to this and other KeyedPartitionStream.
<T_OTHER, OUT> NonKeyedPartitionStream.ProcessConfigurableAndNonKeyedPartitionStream<OUT>	connectAndProcess(NonKeyedPartitionStream<T_OTHER> other, TwoInputBroadcastStreamProcessFunction<T_OTHER,T,OUT> processFunction)	Apply a two input operation to this and other NonKeyedPartitionStream.

Method Details

connectAndProcess

<K,
T_OTHER,
OUT>
NonKeyedPartitionStream.ProcessConfigurableAndNonKeyedPartitionStream<OUT> connectAndProcess(KeyedPartitionStream<K,T_OTHER> other,
 TwoInputBroadcastStreamProcessFunction<T_OTHER,T,OUT> processFunction)

Apply a two input operation to this and other KeyedPartitionStream.

Generally, concatenating BroadcastStream and KeyedPartitionStream will result in a NonKeyedPartitionStream, and you can manually generate a KeyedPartitionStream via keyBy partitioning. In some cases, you can guarantee that the partition on which the data is processed will not change, then you can use connectAndProcess(KeyedPartitionStream, TwoInputBroadcastStreamProcessFunction, KeySelector) to avoid shuffling.

Parameters:

other - KeyedPartitionStream to perform operation with two input.

processFunction - to perform operation.

Returns:

new stream with this operation.

connectAndProcess

<T_OTHER,
OUT>
NonKeyedPartitionStream.ProcessConfigurableAndNonKeyedPartitionStream<OUT> connectAndProcess(NonKeyedPartitionStream<T_OTHER> other,
 TwoInputBroadcastStreamProcessFunction<T_OTHER,T,OUT> processFunction)

Apply a two input operation to this and other NonKeyedPartitionStream.

Parameters:

other - NonKeyedPartitionStream to perform operation with two input.

processFunction - to perform operation.

Returns:

new stream with this operation.

connectAndProcess

<K,
T_OTHER,
OUT>
KeyedPartitionStream.ProcessConfigurableAndKeyedPartitionStream<K,OUT> connectAndProcess(KeyedPartitionStream<K,T_OTHER> other,
 TwoInputBroadcastStreamProcessFunction<T_OTHER,T,OUT> processFunction,
 org.apache.flink.api.java.functions.KeySelector<OUT,K> newKeySelector)

Apply a two input operation to this and other KeyedPartitionStream.

This method is used to avoid shuffle after applying the process function. It is required that for the record from non-broadcast input, the new KeySelector must extract the same key as the original KeySelectors on the KeyedPartitionStream. Otherwise, the partition of data will be messy. As for the record from broadcast input, the output key from keyed partition itself instead of the new key selector, so the data it outputs will not affect the partition.

Parameters:

other - KeyedPartitionStream to perform operation with two input.

processFunction - to perform operation.

newKeySelector - to select the key after process.

Returns:

new KeyedPartitionStream with this operation.