Flink中Transform怎么用

小編給大家分享一下Flink中Transform怎么用，相信大部分人都還不怎么了解，因此分享這篇文章給大家參考一下，希望大家閱讀完這篇文章后大有收獲，下面讓我們一起去了解一下吧！

分組聚合

  String path = "E:\\GIT\\flink-learn\\flink-learn\\telemetering.txt";
        StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();

        TupleTypeInfo<Tuple3<String, Double, Long>> typeInfo = new TupleTypeInfo<>(Types.STRING, Types.DOUBLE, Types.LONG);

        TupleCsvInputFormat<Tuple3<String, Double, Long>> tupleCsvInputFormat =
                new TupleCsvInputFormat<>(new Path(path), typeInfo);

        DataStreamSource<Tuple3<String, Double, Long>> dataStreamSource = env.createInput(tupleCsvInputFormat, typeInfo);
        //或   DataStreamSource<Tuple2<String, Double>> dataStreamSource = env.readFile(tupleCsvInputFormat, path);

        SingleOutputStreamOperator<Tuple3<String, Double, Long>> operator = dataStreamSource
                .filter(Objects::nonNull)
//                .map()
//                .flatMap()
//                .keyBy(0)
                .keyBy(tuple -> tuple.f0)
                .minBy(1);
//                .min()
//                .max(1);
//                .maxBy(1, false);
//                .sum(1);
//                .reduce();
//                .process();
        operator.print().setParallelism(1);
        env.execute();

分流/合流

String path = "E:\\GIT\\flink-learn\\flink-learn\\telemetering.txt";
        StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();

        PojoTypeInfo<TelemeterDTO> typeInfo = (PojoTypeInfo<TelemeterDTO>) Types.POJO(TelemeterDTO.class);
        PojoCsvInputFormat<TelemeterDTO> inputFormat = new PojoCsvInputFormat<>(new Path(path), typeInfo, new String[]{"code", "value", "timestamp"});
        DataStreamSource<TelemeterDTO> dataStreamSource = env.createInput(inputFormat, typeInfo);

        //分流
        SplitStream<TelemeterDTO> splitStream = dataStreamSource
                .split(item -> {
                    if (item.getValue() > 100) {
                        return Collections.singletonList("high");
                    }
                    return Collections.singletonList("low");
                });

        DataStream<TelemeterDTO> highStream = splitStream.select("high");
        DataStream<TelemeterDTO> lowStream = splitStream.select("low");

        //合流
        ConnectedStreams<TelemeterDTO, TelemeterDTO> connectedStreams = lowStream.connect(highStream);
//        DataStream<TelemeterDTO> unionDataStream = lowStream.union(highStream); //需要類型一致

        SingleOutputStreamOperator<Tuple3<String, Double, Long>> operator = connectedStreams
                .map(new CoMapFunction<TelemeterDTO, TelemeterDTO, Tuple3<String, Double, Long>>() {
                    @Override
                    public Tuple3<String, Double, Long> map1(TelemeterDTO value) {
                        return Tuple3.of(value.getCode(), value.getValue(), value.getTimestamp());
                    }

                    @Override
                    public Tuple3<String, Double, Long> map2(TelemeterDTO value) {
                        return Tuple3.of(value.getCode(), value.getValue(), value.getTimestamp());
                    }
                });

        operator.print();
        env.execute();

UDF函數，提供底層支持

• MapFunction

• FilterFunction

• ReduceFunction

• ProcessFunction

• SourceFunction

• SinkFunction

富函數

富函數 包含了生命周期，及上下文相關信息，如

• open() 可以在算子創建之初建立數據庫連接

• close() 在在算子生命結束之前關閉資源

以上是“Flink中Transform怎么用”這篇文章的所有內容，感謝各位的閱讀！相信大家都有了一定的了解，希望分享的內容對大家有所幫助，如果還想學習更多知識，歡迎關注億速云行業資訊頻道！