Flink中的的Transformation算子及实例_mapoperator> mapop

Map

DataStream → DataStream：输入一个参数产生一个参数。

val env = StreamExecutionEnvironment.getExecutionEnvironment
 
val stream = env.generateSequence(1,10)
val streamMap = stream.map { x => x * 2 }
streamFilter.print()
 
env.execute("FirstJob")

注意：stream.print()：每一行前面的数字代表这一行是哪一个并行线程输出的。

import org.apache.flink.api.common.functions.MapFunction;
import org.apache.flink.api.java.DataSet;
import org.apache.flink.api.java.ExecutionEnvironment;
import org.apache.flink.api.java.operators.MapOperator;
import org.apache.flink.api.java.utils.ParameterTool;
import scala.Tuple2;
 
import java.util.Random;
 
 
public class StuScore {
    private static Random rand = new Random();
 
    public static void main(String[] args) throws Exception {
        ParameterTool params = ParameterTool.fromArgs(args);
        ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();
        env.getConfig().setGlobalJobParameters(params);
 
        DataSet<String> text;
        if (params.has("input")) {
            text = env.readTextFile("F:\\date\\flinkdata\\stu.txt");
        }else{
            System.out.println("请检查你的输入");
            return;
        }
 
        MapOperator<String, Tuple2<String, Integer>> stuscore = text.map(new MapFunction<String, Tuple2<String, Integer>>() {
            @Override
            public Tuple2<String, Integer> map(String s) throws Exception {
                return new Tuple2<>(s, rand.nextInt(100) + 1);
            }
        });
 
        if (params.has("output")) {
            stuscore.writeAsCsv("F:\\date\\flinkdata\\personinput\\A");
        }else {
            System.out.println("打印到控制台");
            stuscore.print();
        }
    }
}

FlatMap

DataStream → DataStream：输入一个参数，产生0个、1个或者多个输出。

import org.apache.flink.streaming.api.scala._
 
val env = StreamExecutionEnvironment.getExecutionEnvironment
 
val stream = env.readTextFile("F:\date\flinkdata\stu.tsv")
val streamFlatMap = stream.flatMap{
    x => x.split(" ")
}
streamFilter.print()
 
env.execute("FirstJob")

Filter

DataStream → DataStream：结算每个元素的布尔值，并返回布尔值为true的元素。下面这个例子是过滤出非0的元素：

import org.apache.flink.streaming.api.scala._
 
val env = StreamExecutionEnvironment.getExecutionEnvironment
 
val stream = env.generateSequence(1,10)
val streamFilter = stream.filter{
  //打印奇数
    x => (x % 2 != 0)
}
streamFilter.print()
 
env.execute("FirstJob")

Connect

DataStream,DataStream → ConnectedStreams：连接两个保持他们类型的数据流，两个数据流被Connect之后，只是被放在了一个同一个流中，内部依然保持各自的数据和形式不发生任何变化，两个流相互独立。

val env = StreamExecutionEnvironment.getExecutionEnvironment
 
val stream = env.readTextFile("F:\date\flinkdata\stu.tsv")
 
val streamMap = stream.flatMap(item => item.split(" ")).filter(item => item.equals("hadoop"))
val streamCollect = env.fromCollection(List(1,2,3,4))
//streamMap和streamCollect交换顺序不会影响结果
val streamConnect = streamMap.connect(streamCollect)
 
streamConnect.map(item=>println(item), item=>println(item))
 
env.execute("FirstJob")

CoMap,CoFlatMap

ConnectedStreams → DataStream：作用于ConnectedStreams上，功能与map和flatMap一样，对ConnectedStreams中的每一个Stream分别进行map和flatMap处理。

val env = StreamExecutionEnvironment.getExecutionEnvironment
 
val stream1 = env.readTextFile("F:\date\flinkdata\stu.tsv")
val streamFlatMap = stream1.flatMap(x => x.split(" "))
val stream2 = env.fromCollection(List(1,2,3,4))
val streamConnect = streamFlatMap.connect(stream2)
val streamCoMap = streamConnect.map(
    (str) => str + "connect",
    (in) => in + 100
)
 
streamCoMap.print()
 
env.execute("FirstJob")
 
//========================
 
val env = StreamExecutionEnvironment.getExecutionEnvironment
 
val stream1 = env.readTextFile("test.txt")
val stream2 = env.readTextFile("test1.txt")
val streamConnect = stream1.connect(stream2)
val streamCoMap = streamConnect.flatMap(
    (str1) => str1.split(" "),
    (str2) => str2.split(" ")
)
streamConnect.map(item=>println(item), item=>println(item))
 
env.execute("FirstJob")

Split

DataStream → SplitStream：根据某些特征把一个DataStream拆分成两个或者多个DataStream。注：此代码无法运行出结果，使用Select即可运行

val env = StreamExecutionEnvironment.getExecutionEnvironment
 
val stream = env.readTextFile("F:\date\flinkdata\stu.tsv")
val streamFlatMap = stream.flatMap(x => x.split(" "))
val streamSplit = streamFlatMap.split(
  num =>
  //字符串内容为hadoop的组成一个DataStream，其余的组成一个DataStream 
    (num.equals("hadoop")) match{
        case true => List("hadoop")
        case false => List("other")
    }
)
 
env.execute("FirstJob")

Select

SplitStream→DataStream：从一个SplitStream中获取一个或者多个DataStream。

val env = StreamExecutionEnvironment.getExecutionEnvironment
 
val stream = env.readTextFile("F:\date\flinkdata\stu.tsv")
val streamFlatMap = stream.flatMap(x => x.split(" "))
val streamSplit = streamFlatMap.split(
  num =>
    (num.equals("hadoop")) match{
        case true => List("hadoop")
        case false => List("other")
    }
)
 
val hadoop = streamSplit.select("hadoop")
val other = streamSplit.select("other")
other.print()
 
env.execute("FirstJob")

Union

DataStream → DataStream：对两个或者两个以上的DataStream进行union操作，产生一个包含所有DataStream元素的新DataStream。注意:如果你将一个DataStream跟它自己做union操作，在新的DataStream中，你将看到每一个元素都出现两次。

val env = StreamExecutionEnvironment.getExecutionEnvironment
 
val stream1 = env.readTextFile("test.txt")
val streamFlatMap1 = stream1.flatMap(x => x.split(" "))
val stream2 = env.readTextFile("test1.txt")
val streamFlatMap2 = stream2.flatMap(x => x.split(" "))
val streamConnect = streamFlatMap1.union(streamFlatMap2)
 
env.execute("FirstJob")

KeyBy

DataStream → KeyedStream：输入必须是Tuple类型，逻辑地将一个流拆分成不相交的分区，每个分区包含具有相同key的元素，在内部以hash的形式实现的。

 
val env = StreamExecutionEnvironment.getExecutionEnvironment
val stream = env.readTextFile("test.txt")
val streamFlatMap = stream.flatMap{
    x => x.split(" ")
}
val streamMap = streamFlatMap.map{
    x => (x,1)
}
val streamKeyBy = streamMap.keyBy(0)
env.execute("FirstJob")

Reduce

KeyedStream → DataStream：一个分组数据流的聚合操作，合并当前的元素和上次聚合的结果，产生一个新的值，返回的流中包含每一次聚合的结果，而不是只返回最后一次聚合的最终结果。

 
val env = StreamExecutionEnvironment.getExecutionEnvironment
 
val stream = env.readTextFile("test.txt").flatMap(item => item.split(" ")).map(item => (item, 1)).keyBy(0)
 
val streamReduce = stream.reduce(
  (item1, item2) => (item1._1, item1._2 + item2._2)
)
 
streamReduce.print()
 
env.execute("FirstJob")

Fold

KeyedStream → DataStream：一个有初始值的分组数据流的滚动折叠操作，合并当前元素和前一次折叠操作的结果，并产生一个新的值，返回的流中包含每一次折叠的结果，而不是只返回最后一次折叠的最终结果。

 
val env = StreamExecutionEnvironment.getExecutionEnvironment
 
val stream = env.readTextFile("test.txt").flatMap(item => item.split(" ")).map(item => (item, 1)).keyBy(0)
 
val streamReduce = stream.fold(100)(
  (begin, item) => (begin + item._2)
)
 
streamReduce.print()
 
env.execute("FirstJob")

Aggregations

KeyedStream → DataStream：分组数据流上的滚动聚合操作。min和minBy的区别是min返回的是一个最小值，而minBy返回的是其字段中包含最小值的元素(同样原理适用于max和maxBy)，返回的流中包含每一次聚合的结果，而不是只返回最后一次聚合的最终结果。

 
keyedStream.sum(0) 
keyedStream.sum("key") 
keyedStream.min(0) 
keyedStream.min("key") 
keyedStream.max(0) 
keyedStream.max("key") 
keyedStream.minBy(0) 
keyedStream.minBy("key") 
keyedStream.maxBy(0) 
keyedStream.maxBy("key")
 
val env = StreamExecutionEnvironment.getExecutionEnvironment
 
val stream = env.readTextFile("test02.txt").map(item => (item.split(" ")(0), item.split(" ")(1).toLong)).keyBy(0)
 
val streamReduce = stream.sum(1)
 
streamReduce.print()
 
env.execute("FirstJob")

   在2.3.10之前的算子都是可以直接作用在Stream上的，因为他们不是聚合类型的操作，但是到2.3.10后你会发现，我们虽然可以对一个无边界的流数据直接应用聚合算子，但是它会记录下每一次的聚合结果，这往往不是我们想要的，其实，reduce、fold、aggregation这些聚合算子都是和Window配合使用的，只有配合Window，才能得到想要的结果。