Flink从入门到放弃—Stream API—常用算子(filter 和 keyBy）

导航

请先移步至Flink从入门到放弃—Stream API—常用算子（map和flatMap）观看之前内容。

本章内容

• filter : filter算子，顾名思义就是过滤掉不需要的数据，保留需要的数据。
• keyBy: 根据key分组或者说分区。

demo

package com.stream.samples;

import org.apache.flink.api.common.functions.FilterFunction;
import org.apache.flink.api.java.functions.KeySelector;
import org.apache.flink.streaming.api.datastream.DataStreamSource;
import org.apache.flink.streaming.api.datastream.KeyedStream;
import org.apache.flink.streaming.api.datastream.SingleOutputStreamOperator;
import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;

/**
 * @author DeveloperZJQ
 * @since 2022/11/13
 */
public class CustomFilterAndKeyByOperator {
    public static void main(String[] args) throws Exception {
        StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
        DataStreamSource<String> dataStream = env.socketTextStream("192.168.112.147", 7777);

        SingleOutputStreamOperator<String> filter = dataStream.filter(new FilterFunction<String>() {
            @Override
            public boolean filter(String value) {
                // 保留字符串长度大于5的
                return value.length() > 5;
            }
        });

        KeyedStream<String, String> keyBy = filter.keyBy(new KeySelector<String, String>() {
            @Override
            public String getKey(String value) {
                return value;
            }
        });

        filter.print();
        keyBy.print();

        env.execute(MapOperator.class.getSimpleName());
    }
}

上面代码没有实际业务含义，就是最简单的应用，目的是为了看其源码。

filter()

public SingleOutputStreamOperator<T> filter(FilterFunction<T> filter) {
        return this.transform("Filter", this.getType(), (OneInputStreamOperator)(new StreamFilter((FilterFunction)this.clean(filter))));
    }

算子默认名就叫Filter，看下StreamFilter这个类

public StreamFilter(FilterFunction<IN> filterFunction) {
        super(filterFunction);
        this.chainingStrategy = ChainingStrategy.ALWAYS;
    }

    public void processElement(StreamRecord<IN> element) throws Exception {
        if (((FilterFunction)this.userFunction).filter(element.getValue())) {
            this.output.collect(element);
        }

    }

上面看到FilterFunction,看下这个接口类

@Public
@FunctionalInterface
public interface FilterFunction<T> extends Function, Serializable {
    boolean filter(T value) throws Exception;
}

返回的是一个boolean类型的值，结合上面StreamFilter中的一行代码可以到

if (((FilterFunction)this.userFunction).filter(element.getValue())) {
            this.output.collect(element);
        }

如果为true，则把这个元素收集并流转下去。

值得一提的是，有些同学说源码生涩难懂，其实不是，你看上面if判断里的东西其实都是咱们学过的继承多态等基础知识，filter调的其实也是用户代码实现的filter。

然后下面的源码和map、flatMap是一样的逻辑。

keyBy()

public <K> KeyedStream<T, K> keyBy(KeySelector<T, K> key) {
  // 先决条件key不能为空
        Preconditions.checkNotNull(key);
        // new KeyedStream 接下来咱们看下这个对象是干啥的
        return new KeyedStream(this, (KeySelector)this.clean(key));
    }

我们可以看到KeyedStream和filter、map、flatMap算子有差异了，keyBy算子没有使用DataStream,而是另起了一个继承于DataStream的类 KeyedStream。

@Public
@FunctionalInterface
public interface KeySelector<IN, KEY> extends Function, Serializable {
    KEY getKey(IN value) throws Exception;
}

同时，KeySelector选择器也不一样，它返回的是一个KEY。

接着往下看

// 套娃
public KeyedStream(DataStream<T> dataStream, KeySelector<T, KEY> keySelector) {
        this(dataStream, keySelector, TypeExtractor.getKeySelectorTypes(keySelector, dataStream.getType()));
    }

// 套娃 ，但是请注意，实际上每一层套娃都有做事情
// 这里有一个知识点 最大并行度是128
public KeyedStream(DataStream<T> dataStream, KeySelector<T, KEY> keySelector, TypeInformation<KEY> keyType) {
        this(dataStream, new PartitionTransformation(dataStream.getTransformation(), new KeyGroupStreamPartitioner(keySelector, 128)), keySelector, keyType);
    }

// 找到正主了，keyby算子可以看到PartitionTransformation
@Internal
    KeyedStream(DataStream<T> stream, PartitionTransformation<T> partitionTransformation, KeySelector<T, KEY> keySelector, TypeInformation<KEY> keyType) {
    // 调用父类构造器
        super(stream.getExecutionEnvironment(), partitionTransformation);
        // key分区器
        this.keySelector = (KeySelector)this.clean(keySelector);
        // 获取键的类型
        this.keyType = this.validateKeyType(keyType);
    }

看一下这个PartitionTransformation(dataStream.getTransformation(), new KeyGroupStreamPartitioner(keySelector, 128))

// key组分区器
@Internal
public class KeyGroupStreamPartitioner<T, K> extends StreamPartitioner<T> implements ConfigurableStreamPartitioner {
    private static final long serialVersionUID = 1L;
    private final KeySelector<T, K> keySelector;
    private int maxParallelism;

    public KeyGroupStreamPartitioner(KeySelector<T, K> keySelector, int maxParallelism) {
        Preconditions.checkArgument(maxParallelism > 0, "Number of key-groups must be > 0!");
        this.keySelector = (KeySelector)Preconditions.checkNotNull(keySelector);
        this.maxParallelism = maxParallelism;
    }

    public int getMaxParallelism() {
        return this.maxParallelism;
    }
// 该方法是对数据进行实时的分区，不是上游直接发送给下游，而是将数据写入到对应的channel的缓存中，下游到上游实时拉取；
    public int selectChannel(SerializationDelegate<StreamRecord<T>> record) {
        Object key;
        try {
            key = this.keySelector.getKey(((StreamRecord)record.getInstance()).getValue());
        } catch (Exception var4) {
            throw new RuntimeException("Could not extract key from " + ((StreamRecord)record.getInstance()).getValue(), var4);
        }
// 分发key到并行的subtask
        return KeyGroupRangeAssignment.assignKeyToParallelOperator(key, this.maxParallelism, this.numberOfChannels);
    }
}

再看KeyGroupRangeAssignment类

  public static int assignKeyToParallelOperator(Object key, int maxParallelism, int parallelism) {
        Preconditions.checkNotNull(key, "Assigned key must not be null!");
        return computeOperatorIndexForKeyGroup(maxParallelism, parallelism, assignToKeyGroup(key, maxParallelism));
    }


 public static int assignToKeyGroup(Object key, int maxParallelism) {
        Preconditions.checkNotNull(key, "Assigned key must not be null!");
        return computeKeyGroupForKeyHash(key.hashCode(), maxParallelism);
    }
// 先计算key的hashCode值,MathUtils.murmurHash(keyHash)保证key的hash值一定是正的，避免返回的数字为负;将返回特殊的hash值模除以默认最大并行的，默认是128，得到keyGroupId；

    public static int computeKeyGroupForKeyHash(int keyHash, int maxParallelism) {
        return MathUtils.murmurHash(keyHash) % maxParallelism;
    }
   
//   keyGroupId * parallelism（程序的并行度） / maxParallelism（默认最大并行），返回分区编号
 public static int computeOperatorIndexForKeyGroup(int maxParallelism, int parallelism, int keyGroupId) {
        return keyGroupId * parallelism / maxParallelism;
    }

另外如果想要POJO作为key，那么还需要重写hashcode方法。

参考

Flink算子（KeyBy的源码分析及案例）
更多代码使用案例