Flink 流处理API之Join

1、Window Join

stream.join(otherStream)
    .where(<KeySelector>)
    .equalTo(<KeySelector>)
    .window(<WindowAssigner>)
    .apply(<JoinFunction>)

1.1 Tumbling Window Join

数据无重叠

import org.apache.flink.streaming.api.windowing.assigners.SlidingEventTimeWindows;
import org.apache.flink.streaming.api.windowing.time.Time;

...

val orangeStream: DataStream[Integer] = ...
val greenStream: DataStream[Integer] = ...

orangeStream.join(greenStream)
    .where(elem => /* select key */)
    .equalTo(elem => /* select key */)
    .window(TumblingEventTimeWindows.of(Time.milliseconds(2)))
    .apply { (e1, e2) => e1 + "," + e2 }

1.2 Sliding Window Join

import org.apache.flink.streaming.api.windowing.assigners.SlidingEventTimeWindows;
import org.apache.flink.streaming.api.windowing.time.Time;

...

val orangeStream: DataStream[Integer] = ...
val greenStream: DataStream[Integer] = ...

orangeStream.join(greenStream)
    .where(elem => /* select key */)
    .equalTo(elem => /* select key */)
    .window(SlidingEventTimeWindows.of(Time.milliseconds(2) /* size */, Time.milliseconds(1) /* slide */))
    .apply { (e1, e2) => e1 + "," + e2 }

1.3 Session Window Join

import org.apache.flink.streaming.api.windowing.assigners.EventTimeSessionWindows;
import org.apache.flink.streaming.api.windowing.time.Time;
 
...

val orangeStream: DataStream[Integer] = ...
val greenStream: DataStream[Integer] = ...

orangeStream.join(greenStream)
    .where(elem => /* select key */)
    .equalTo(elem => /* select key */)
    .window(EventTimeSessionWindows.withGap(Time.milliseconds(1)))
    .apply { (e1, e2) => e1 + "," + e2 }

2、Interval Join

This can also be expressed more formally as

b.timestamp ∈ [a.timestamp + lowerBound; a.timestamp + upperBound] or a.timestamp + lowerBound <= b.timestamp <= a.timestamp + upperBound

The interval join currently only supports event time.

Using the more formal notation again this will translate to

orangeElem.ts + lowerBound <= greenElem.ts <= orangeElem.ts + upperBound

import org.apache.flink.streaming.api.functions.co.ProcessJoinFunction;
import org.apache.flink.streaming.api.windowing.time.Time;

...

val orangeStream: DataStream[Integer] = ...
val greenStream: DataStream[Integer] = ...

orangeStream
    .keyBy(elem => /* select key */)
    .intervalJoin(greenStream.keyBy(elem => /* select key */))
    .between(Time.milliseconds(-2), Time.milliseconds(1))
    .process(new ProcessJoinFunction[Integer, Integer, String] {
        override def processElement(left: Integer, right: Integer, ctx: ProcessJoinFunction[Integer, Integer, String]#Context, out: Collector[String]): Unit = {
         out.collect(left + "," + right); 
        }
      });
    });

// TODO 1. 获取订单日志数据
val fileDS1: DataStream[String] = orderTransactionAnalysisDao.readTextFile(source1)
// TODO 2. 获取交易日志数据
val fileDS2: DataStream[String] = orderTransactionAnalysisDao.readTextFile(source2)

// TODO 3. 将日志数据转换成对应的样例类数据
val orderDS = fileDS1.map(
    line => {
        val datas = line.split(",")
        OrderLogData( datas(0).toLong, datas(1), datas(2), datas(3).toLong)
    }
)

val payKS = orderDS
    .filter(_.txId != "")
    .assignAscendingTimestamps(_.timestamp * 1000L).keyBy(_.txId)

val txDS = fileDS2.map(
    line => {
        val datas = line.split(",")
        TXLogData( datas(0), datas(1), datas(2).toLong)
    }
)

val txKS = txDS.assignAscendingTimestamps(_.timestamp * 1000L).keyBy(_.txId)

// TODO 4. 将两个流数据进行连接
payKS.intervalJoin(txKS)
    .between(Time.minutes(-5), Time.minutes(5))
    .process(
        new ProcessJoinFunction[OrderLogData, TXLogData, (OrderLogData, TXLogData)] {
            override def processElement(left: OrderLogData, right: TXLogData, ctx: ProcessJoinFunction[OrderLogData, TXLogData, (OrderLogData, TXLogData)]#Context, out: Collector[(OrderLogData, TXLogData)]): Unit = {
                out.collect(( left, right ))
            }
        }
    )