聊聊flink DataStream的window coGroup操作

本文主要研究一下flink DataStream的window coGroup操作

实例

dataStream.coGroup(otherStream)
    .where(0).equalTo(1)
    .window(TumblingEventTimeWindows.of(Time.seconds(3)))
    .apply (new CoGroupFunction () {...});
复制代码
  • 这里展示了DataStream的window coGroup操作的基本用法

DataStream.coGroup

flink-streaming-java_2.11-1.7.0-sources.jar!/org/apache/flink/streaming/api/datastream/DataStream.java

@Public
public class DataStream<T> {

	//......

	public <T2> CoGroupedStreams<T, T2> coGroup(DataStream<T2> otherStream) {
		return new CoGroupedStreams<>(this, otherStream);
	}

	//......
}
复制代码
  • DataStream的coGroup操作创建的是CoGroupedStreams

CoGroupedStreams

flink-streaming-java_2.11-1.7.0-sources.jar!/org/apache/flink/streaming/api/datastream/CoGroupedStreams.java

@Public
public class CoGroupedStreams<T1, T2> {

	private final DataStream<T1> input1;

	private final DataStream<T2> input2;

	public CoGroupedStreams(DataStream<T1> input1, DataStream<T2> input2) {
		this.input1 = requireNonNull(input1);
		this.input2 = requireNonNull(input2);
	}

	public <KEY> Where<KEY> where(KeySelector<T1, KEY> keySelector)  {
		Preconditions.checkNotNull(keySelector);
		final TypeInformation<KEY> keyType = TypeExtractor.getKeySelectorTypes(keySelector, input1.getType());
		return where(keySelector, keyType);
	}

	public <KEY> Where<KEY> where(KeySelector<T1, KEY> keySelector, TypeInformation<KEY> keyType)  {
		Preconditions.checkNotNull(keySelector);
		Preconditions.checkNotNull(keyType);
		return new Where<>(input1.clean(keySelector), keyType);
	}

	//.......
}
复制代码
  • CoGroupedStreams提供了where操作,用于指定input1的keySelector,它创建并返回Where对象

Where

flink-streaming-java_2.11-1.7.0-sources.jar!/org/apache/flink/streaming/api/datastream/CoGroupedStreams.java

	@Public
	public class Where<KEY> {

		private final KeySelector<T1, KEY> keySelector1;
		private final TypeInformation<KEY> keyType;

		Where(KeySelector<T1, KEY> keySelector1, TypeInformation<KEY> keyType) {
			this.keySelector1 = keySelector1;
			this.keyType = keyType;
		}

		public EqualTo equalTo(KeySelector<T2, KEY> keySelector)  {
			Preconditions.checkNotNull(keySelector);
			final TypeInformation<KEY> otherKey = TypeExtractor.getKeySelectorTypes(keySelector, input2.getType());
			return equalTo(keySelector, otherKey);
		}

		public EqualTo equalTo(KeySelector<T2, KEY> keySelector, TypeInformation<KEY> keyType)  {
			Preconditions.checkNotNull(keySelector);
			Preconditions.checkNotNull(keyType);

			if (!keyType.equals(this.keyType)) {
				throw new IllegalArgumentException("The keys for the two inputs are not equal: " +
						"first key = " + this.keyType + " , second key = " + keyType);
			}

			return new EqualTo(input2.clean(keySelector));
		}

		//......
	}	
复制代码
  • Where对象提供了equalTo操作,用于指定input2的keySelector,它创建并返回EqualTo对象

EqualTo

flink-streaming-java_2.11-1.7.0-sources.jar!/org/apache/flink/streaming/api/datastream/CoGroupedStreams.java

		@Public
		public class EqualTo {

			private final KeySelector<T2, KEY> keySelector2;

			EqualTo(KeySelector<T2, KEY> keySelector2) {
				this.keySelector2 = requireNonNull(keySelector2);
			}

			@PublicEvolving
			public <W extends Window> WithWindow<T1, T2, KEY, W> window(WindowAssigner<? super TaggedUnion<T1, T2>, W> assigner) {
				return new WithWindow<>(input1, input2, keySelector1, keySelector2, keyType, assigner, null, null, null);
			}
		}
复制代码
  • EqualTo对象提供了window操作,它创建并返回WithWindow对象

WithWindow

flink-streaming-java_2.11-1.7.0-sources.jar!/org/apache/flink/streaming/api/datastream/CoGroupedStreams.java

	@Public
	public static class WithWindow<T1, T2, KEY, W extends Window> {
		private final DataStream<T1> input1;
		private final DataStream<T2> input2;

		private final KeySelector<T1, KEY> keySelector1;
		private final KeySelector<T2, KEY> keySelector2;

		private final TypeInformation<KEY> keyType;

		private final WindowAssigner<? super TaggedUnion<T1, T2>, W> windowAssigner;

		private final Trigger<? super TaggedUnion<T1, T2>, ? super W> trigger;

		private final Evictor<? super TaggedUnion<T1, T2>, ? super W> evictor;

		private final Time allowedLateness;

		private WindowedStream<TaggedUnion<T1, T2>, KEY, W> windowedStream;

		protected WithWindow(DataStream<T1> input1,
				DataStream<T2> input2,
				KeySelector<T1, KEY> keySelector1,
				KeySelector<T2, KEY> keySelector2,
				TypeInformation<KEY> keyType,
				WindowAssigner<? super TaggedUnion<T1, T2>, W> windowAssigner,
				Trigger<? super TaggedUnion<T1, T2>, ? super W> trigger,
				Evictor<? super TaggedUnion<T1, T2>, ? super W> evictor,
				Time allowedLateness) {
			this.input1 = input1;
			this.input2 = input2;

			this.keySelector1 = keySelector1;
			this.keySelector2 = keySelector2;
			this.keyType = keyType;

			this.windowAssigner = windowAssigner;
			this.trigger = trigger;
			this.evictor = evictor;

			this.allowedLateness = allowedLateness;
		}

		@PublicEvolving
		public WithWindow<T1, T2, KEY, W> trigger(Trigger<? super TaggedUnion<T1, T2>, ? super W> newTrigger) {
			return new WithWindow<>(input1, input2, keySelector1, keySelector2, keyType,
					windowAssigner, newTrigger, evictor, allowedLateness);
		}

		@PublicEvolving
		public WithWindow<T1, T2, KEY, W> evictor(Evictor<? super TaggedUnion<T1, T2>, ? super W> newEvictor) {
			return new WithWindow<>(input1, input2, keySelector1, keySelector2, keyType,
					windowAssigner, trigger, newEvictor, allowedLateness);
		}

		@PublicEvolving
		public WithWindow<T1, T2, KEY, W> allowedLateness(Time newLateness) {
			return new WithWindow<>(input1, input2, keySelector1, keySelector2, keyType,
					windowAssigner, trigger, evictor, newLateness);
		}

		public <T> DataStream<T> apply(CoGroupFunction<T1, T2, T> function) {

			TypeInformation<T> resultType = TypeExtractor.getCoGroupReturnTypes(
				function,
				input1.getType(),
				input2.getType(),
				"CoGroup",
				false);

			return apply(function, resultType);
		}

		@PublicEvolving
		@Deprecated
		public <T> SingleOutputStreamOperator<T> with(CoGroupFunction<T1, T2, T> function) {
			return (SingleOutputStreamOperator<T>) apply(function);
		}

		public <T> DataStream<T> apply(CoGroupFunction<T1, T2, T> function, TypeInformation<T> resultType) {
			//clean the closure
			function = input1.getExecutionEnvironment().clean(function);

			UnionTypeInfo<T1, T2> unionType = new UnionTypeInfo<>(input1.getType(), input2.getType());
			UnionKeySelector<T1, T2, KEY> unionKeySelector = new UnionKeySelector<>(keySelector1, keySelector2);

			DataStream<TaggedUnion<T1, T2>> taggedInput1 = input1
					.map(new Input1Tagger<T1, T2>())
					.setParallelism(input1.getParallelism())
					.returns(unionType);
			DataStream<TaggedUnion<T1, T2>> taggedInput2 = input2
					.map(new Input2Tagger<T1, T2>())
					.setParallelism(input2.getParallelism())
					.returns(unionType);

			DataStream<TaggedUnion<T1, T2>> unionStream = taggedInput1.union(taggedInput2);

			// we explicitly create the keyed stream to manually pass the key type information in
			windowedStream =
					new KeyedStream<TaggedUnion<T1, T2>, KEY>(unionStream, unionKeySelector, keyType)
					.window(windowAssigner);

			if (trigger != null) {
				windowedStream.trigger(trigger);
			}
			if (evictor != null) {
				windowedStream.evictor(evictor);
			}
			if (allowedLateness != null) {
				windowedStream.allowedLateness(allowedLateness);
			}

			return windowedStream.apply(new CoGroupWindowFunction<T1, T2, T, KEY, W>(function), resultType);
		}

		@PublicEvolving
		@Deprecated
		public <T> SingleOutputStreamOperator<T> with(CoGroupFunction<T1, T2, T> function, TypeInformation<T> resultType) {
			return (SingleOutputStreamOperator<T>) apply(function, resultType);
		}

		@VisibleForTesting
		Time getAllowedLateness() {
			return allowedLateness;
		}

		@VisibleForTesting
		WindowedStream<TaggedUnion<T1, T2>, KEY, W> getWindowedStream() {
			return windowedStream;
		}
	}
复制代码
  • WithWindow可以设置windowAssigner、trigger、evictor、allowedLateness,它提供apply操作(with操作被标记为废弃)
  • apply操作接收CoGroupFunction,它内部是先根据两个keySelector创建UnionKeySelector,然后对两个input stream分别使用Input1Tagger及Input2Tagger进行map转换为TaggedUnion对象的stream,然后执行taggedInput1.union(taggedInput2)得到unionStream,之后使用UnionKeySelector将unionStream转换为KeyedStream,之后在对KeyedStream执行window操作,把原来的windowAssigner、trigger、evictor、allowedLateness都赋值过去,最后将用户定义的CoGroupFunction包装为CoGroupWindowFunction,然后调用windowedStream.apply方法
  • 可以看到apply操作内部转化的WindowedStream,其element类型为TaggedUnion;WindowedStream使用的KeyedStream,它的KeySelector为UnionKeySelector;而KeyedStream是基于TaggedUnion类型的DataStream,是taggedInput1.union(taggedInput2)操作而来;而taggedInput1及taggedInput2是对原始input stream进行map操作而来,使用的MapFunction分别是Input1Tagger及Input2Tagger

CoGroupFunction

flink-core-1.7.0-sources.jar!/org/apache/flink/api/common/functions/CoGroupFunction.java

@Public
@FunctionalInterface
public interface CoGroupFunction<IN1, IN2, O> extends Function, Serializable {

	void coGroup(Iterable<IN1> first, Iterable<IN2> second, Collector<O> out) throws Exception;
}
复制代码
  • CoGroupFunction继承了Function,它定义了coGroup方法,该方法接收两个Iterable类型的element集合

Input1Tagger及Input2Tagger

flink-streaming-java_2.11-1.7.0-sources.jar!/org/apache/flink/streaming/api/datastream/CoGroupedStreams.java

	private static class Input1Tagger<T1, T2> implements MapFunction<T1, TaggedUnion<T1, T2>> {
		private static final long serialVersionUID = 1L;

		@Override
		public TaggedUnion<T1, T2> map(T1 value) throws Exception {
			return TaggedUnion.one(value);
		}
	}

	private static class Input2Tagger<T1, T2> implements MapFunction<T2, TaggedUnion<T1, T2>> {
		private static final long serialVersionUID = 1L;

		@Override
		public TaggedUnion<T1, T2> map(T2 value) throws Exception {
			return TaggedUnion.two(value);
		}
	}
复制代码
  • Input1Tagger及Input2Tagger实现了MapFunction,该map方法返回的类型为TaggedUnion

TaggedUnion

flink-streaming-java_2.11-1.7.0-sources.jar!/org/apache/flink/streaming/api/datastream/CoGroupedStreams.java

	@Internal
	public static class TaggedUnion<T1, T2> {
		private final T1 one;
		private final T2 two;

		private TaggedUnion(T1 one, T2 two) {
			this.one = one;
			this.two = two;
		}

		public boolean isOne() {
			return one != null;
		}

		public boolean isTwo() {
			return two != null;
		}

		public T1 getOne() {
			return one;
		}

		public T2 getTwo() {
			return two;
		}

		public static <T1, T2> TaggedUnion<T1, T2> one(T1 one) {
			return new TaggedUnion<>(one, null);
		}

		public static <T1, T2> TaggedUnion<T1, T2> two(T2 two) {
			return new TaggedUnion<>(null, two);
		}
	}
复制代码
  • TaggedUnion里头有one、two两个属性,它提供了两个静态工厂方法one及two,可以看到TaggedUnion对象要么one为null,要么two为null,不可能两个同时有值

UnionKeySelector

flink-streaming-java_2.11-1.7.0-sources.jar!/org/apache/flink/streaming/api/datastream/CoGroupedStreams.java

	private static class UnionKeySelector<T1, T2, KEY> implements KeySelector<TaggedUnion<T1, T2>, KEY> {
		private static final long serialVersionUID = 1L;

		private final KeySelector<T1, KEY> keySelector1;
		private final KeySelector<T2, KEY> keySelector2;

		public UnionKeySelector(KeySelector<T1, KEY> keySelector1,
				KeySelector<T2, KEY> keySelector2) {
			this.keySelector1 = keySelector1;
			this.keySelector2 = keySelector2;
		}

		@Override
		public KEY getKey(TaggedUnion<T1, T2> value) throws Exception{
			if (value.isOne()) {
				return keySelector1.getKey(value.getOne());
			} else {
				return keySelector2.getKey(value.getTwo());
			}
		}
	}
复制代码
  • UnionKeySelector有两个KeySelector属性,它的getKey操作根据TaggedUnion来判断,如果是one,则使用keySelector1.getKey(value.getOne()),否则使用keySelector2.getKey(value.getTwo())

DataStream.union

flink-streaming-java_2.11-1.7.0-sources.jar!/org/apache/flink/streaming/api/datastream/DataStream.java

@Public
public class DataStream<T> {

	//......

	@SafeVarargs
	public final DataStream<T> union(DataStream<T>... streams) {
		List<StreamTransformation<T>> unionedTransforms = new ArrayList<>();
		unionedTransforms.add(this.transformation);

		for (DataStream<T> newStream : streams) {
			if (!getType().equals(newStream.getType())) {
				throw new IllegalArgumentException("Cannot union streams of different types: "
						+ getType() + " and " + newStream.getType());
			}

			unionedTransforms.add(newStream.getTransformation());
		}
		return new DataStream<>(this.environment, new UnionTransformation<>(unionedTransforms));
	}

	//......
}
复制代码
  • DataStream的union操作,使用UnionTransformation创建了一个新的DataStream;注意union操作需要两个stream使用相同类型的element,这就是为什么WithWindow的apply操作对两个input stream分别使用Input1Tagger及Input2Tagger进行map转换为TaggedUnion对象来统一两个stream的element类型的原因

CoGroupWindowFunction

flink-streaming-java_2.11-1.7.0-sources.jar!/org/apache/flink/streaming/api/datastream/CoGroupedStreams.java

	private static class CoGroupWindowFunction<T1, T2, T, KEY, W extends Window>
			extends WrappingFunction<CoGroupFunction<T1, T2, T>>
			implements WindowFunction<TaggedUnion<T1, T2>, T, KEY, W> {

		private static final long serialVersionUID = 1L;

		public CoGroupWindowFunction(CoGroupFunction<T1, T2, T> userFunction) {
			super(userFunction);
		}

		@Override
		public void apply(KEY key,
				W window,
				Iterable<TaggedUnion<T1, T2>> values,
				Collector<T> out) throws Exception {

			List<T1> oneValues = new ArrayList<>();
			List<T2> twoValues = new ArrayList<>();

			for (TaggedUnion<T1, T2> val: values) {
				if (val.isOne()) {
					oneValues.add(val.getOne());
				} else {
					twoValues.add(val.getTwo());
				}
			}
			wrappedFunction.coGroup(oneValues, twoValues, out);
		}
	}
复制代码
  • CoGroupWindowFunction继承了WrappingFunction(WrappingFunction继承了AbstractRichFunction,覆盖了父类的open、close、setRuntimeContext方法,用于管理wrappedFunction),实现了WindowFunction接口,其apply方法对TaggedUnion类型的Iterable数据进行拆解,分别拆分到oneValues及twoValues中,然后调用用户定义的CoGroupFunction的coGroup方法

小结

  • DataStream提供了coGroup方法,用于执行window coGroup操作,它返回的是CoGroupedStreams;CoGroupedStreams主要是提供where操作来构建Where对象;Where对象主要提供equalTo操作用于构建EqualTo对象;EqualTo对象提供window操作用于构建WithWindow对象;WithWindow可以设置windowAssigner、trigger、evictor、allowedLateness,它提供apply操作
  • CoGroupedStreams的WithWindow对象的apply操作接收CoGroupFunction,它内部是先根据两个keySelector创建UnionKeySelector,然后对两个input stream分别使用Input1Tagger及Input2Tagger进行map转换为TaggedUnion对象的stream,然后执行taggedInput1.union(taggedInput2)得到unionStream,之后使用UnionKeySelector将unionStream转换为KeyedStream,之后在对KeyedStream执行window操作,把原来的windowAssigner、trigger、evictor、allowedLateness都赋值过去,最后将用户定义的CoGroupFunction包装为CoGroupWindowFunction,然后调用windowedStream.apply方法
  • CoGroupedStreams的WithWindow对象的apply操作借助了DataStream的union操作类合并两个stream,然后转换为KeyedStream,这里关键的两个类分别是TaggedUnion及UnionKeySelector;TaggedUnion里头有one、two两个属性,它提供了两个静态工厂方法one及two,可以看到TaggedUnion对象要么one为null,要么two为null,不可能两个同时有值;UnionKeySelector有两个KeySelector属性,它的getKey操作根据TaggedUnion来判断,如果是one,则使用keySelector1.getKey(value.getOne()),否则使用keySelector2.getKey(value.getTwo())(借助TaggedUnion类统一两个stream的element类型,然后好执行union操作)
  • CoGroupWindowFunction继承了WrappingFunction(WrappingFunction继承了AbstractRichFunction,覆盖了父类的open、close、setRuntimeContext方法,用于管理wrappedFunction),实现了WindowFunction接口,其apply方法对TaggedUnion类型的Iterable数据进行拆解,分别拆分到oneValues及twoValues中,然后调用用户定义的CoGroupFunction的coGroup方法
  • CoGroupFunction继承了Function,它定义了coGroup方法,该方法接收两个Iterable类型的element集合;JoinedStreams的WithWindow对象的apply方法内部将JoinFunction或者FlatJoinFunction包装为CoGroupFunction(JoinFunction使用JoinCoGroupFunction包装,FlatJoinFunction使用FlatJoinCoGroupFunction包装),然后去调用CoGroupedStreams的WithWindow的apply方法;而JoinCoGroupFunction及FlatJoinCoGroupFunction继承了WrappingFunction,同时实现CoGroupFunction接口定义的coGroup方法,默认是遍历第一个集合,对其每个元素遍历第二个集合,挨个执行JoinFunction或FlatJoinFunction的join方法(这里的操作对集合为空的情况不做任何操作,因而实现的就是inner join效果;用户使用coGroup操作可以自定义CoGroupFunction实现outer join)

doc

  • 0
    点赞
  • 1
    收藏
    觉得还不错? 一键收藏
  • 0
    评论

“相关推荐”对你有帮助么?

  • 非常没帮助
  • 没帮助
  • 一般
  • 有帮助
  • 非常有帮助
提交
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值