Java TupleTag类代码示例

import org.apache.beam.sdk.values.TupleTag; //导入依赖的package包/类
public FlinkStatefulDoFnFunction(
    DoFn<KV<K, V>, OutputT> dofn,
    String stepName,
    WindowingStrategy<?, ?> windowingStrategy,
    Map<PCollectionView<?>, WindowingStrategy<?, ?>> sideInputs,
    PipelineOptions pipelineOptions,
    Map<TupleTag<?>, Integer> outputMap,
    TupleTag<OutputT> mainOutputTag) {

  this.dofn = dofn;
  this.stepName = stepName;
  this.windowingStrategy = windowingStrategy;
  this.sideInputs = sideInputs;
  this.serializedOptions = new SerializablePipelineOptions(pipelineOptions);
  this.outputMap = outputMap;
  this.mainOutputTag = mainOutputTag;
}
 

import org.apache.beam.sdk.values.TupleTag; //导入依赖的package包/类
@Test
@Category(ValidatesRunner.class)
public void testCoGroupByKeyGetOnly() {
  final TupleTag<String> tag1 = new TupleTag<>();
  final TupleTag<String> tag2 = new TupleTag<>();

  PCollection<KV<Integer, CoGbkResult>> coGbkResults =
      buildGetOnlyGbk(p, tag1, tag2);

  PAssert.thatMap(coGbkResults).satisfies(
      new SerializableFunction<Map<Integer, CoGbkResult>, Void>() {
        @Override
        public Void apply(Map<Integer, CoGbkResult> results) {
          assertEquals("collection1-1", results.get(1).getOnly(tag1));
          assertEquals("collection1-2", results.get(2).getOnly(tag1));
          assertEquals("collection2-2", results.get(2).getOnly(tag2));
          assertEquals("collection2-3", results.get(3).getOnly(tag2));
          return null;
        }
      });

  p.run();
}
 

import org.apache.beam.sdk.values.TupleTag; //导入依赖的package包/类
/**
 * @param fn the splittable {@link DoFn}.
 * @param windowingStrategy the {@link WindowingStrategy} of the input collection.
 * @param sideInputs list of side inputs that should be available to the {@link DoFn}.
 * @param mainOutputTag {@link TupleTag Tag} of the {@link DoFn DoFn's} main output.
 * @param additionalOutputTags {@link TupleTagList Tags} of the {@link DoFn DoFn's} additional
 * @param outputTagsToCoders A map from output tag to the coder for that output, which should
 *     provide mappings for the main and all additional tags.
 */
public ProcessKeyedElements(
    DoFn<InputT, OutputT> fn,
    Coder<InputT> elementCoder,
    Coder<RestrictionT> restrictionCoder,
    WindowingStrategy<InputT, ?> windowingStrategy,
    List<PCollectionView<?>> sideInputs,
    TupleTag<OutputT> mainOutputTag,
    TupleTagList additionalOutputTags,
    Map<TupleTag<?>, Coder<?>> outputTagsToCoders) {
  this.fn = fn;
  this.elementCoder = elementCoder;
  this.restrictionCoder = restrictionCoder;
  this.windowingStrategy = windowingStrategy;
  this.sideInputs = sideInputs;
  this.mainOutputTag = mainOutputTag;
  this.additionalOutputTags = additionalOutputTags;
  this.outputTagsToCoders = outputTagsToCoders;
}
 

import org.apache.beam.sdk.values.TupleTag; //导入依赖的package包/类
/**
 * test UDF.
 */
@Test
public void testUdf() throws Exception{
  BeamRecordSqlType resultType = BeamRecordSqlType.create(Arrays.asList("f_int", "cubicvalue"),
      Arrays.asList(Types.INTEGER, Types.INTEGER));

  BeamRecord record = new BeamRecord(resultType, 2, 8);

  String sql1 = "SELECT f_int, cubic1(f_int) as cubicvalue FROM PCOLLECTION WHERE f_int = 2";
  PCollection<BeamRecord> result1 =
      boundedInput1.apply("testUdf1",
          BeamSql.query(sql1).withUdf("cubic1", CubicInteger.class));
  PAssert.that(result1).containsInAnyOrder(record);

  String sql2 = "SELECT f_int, cubic2(f_int) as cubicvalue FROM PCOLLECTION WHERE f_int = 2";
  PCollection<BeamRecord> result2 =
      PCollectionTuple.of(new TupleTag<BeamRecord>("PCOLLECTION"), boundedInput1)
      .apply("testUdf2",
          BeamSql.queryMulti(sql2).withUdf("cubic2", new CubicIntegerFn()));
  PAssert.that(result2).containsInAnyOrder(record);

  pipeline.run().waitUntilFinish();
}
 

import org.apache.beam.sdk.values.TupleTag; //导入依赖的package包/类
private void runPartialFields(PCollection<BeamRecord> input) throws Exception {
  String sql = "SELECT f_int, f_long FROM TABLE_A";

  PCollection<BeamRecord> result =
      PCollectionTuple.of(new TupleTag<BeamRecord>("TABLE_A"), input)
      .apply("testPartialFields", BeamSql.queryMulti(sql));

  BeamRecordSqlType resultType = BeamRecordSqlType.create(Arrays.asList("f_int", "f_long"),
      Arrays.asList(Types.INTEGER, Types.BIGINT));

  BeamRecord record = new BeamRecord(resultType
      , recordsInTableA.get(0).getFieldValue(0), recordsInTableA.get(0).getFieldValue(1));

  PAssert.that(result).containsInAnyOrder(record);

  pipeline.run().waitUntilFinish();
}
 

import org.apache.beam.sdk.values.TupleTag; //导入依赖的package包/类
/**
 * Returns a {@code PCollection<KV<Integer, CoGbkResult>>} containing the result
 * of a {@link CoGroupByKey} over 2 {@code PCollection<KV<Integer, String>>},
 * where each {@link PCollection} has no duplicate keys and the key sets of
 * each {@link PCollection} are intersecting but neither is a subset of the other.
 */
private PCollection<KV<Integer, CoGbkResult>> buildGetOnlyGbk(
    Pipeline p,
    TupleTag<String> tag1,
    TupleTag<String> tag2) {
  List<KV<Integer, String>> list1 =
      Arrays.asList(
          KV.of(1, "collection1-1"),
          KV.of(2, "collection1-2"));
  List<KV<Integer, String>> list2 =
      Arrays.asList(
          KV.of(2, "collection2-2"),
          KV.of(3, "collection2-3"));
  PCollection<KV<Integer, String>> collection1 = createInput("CreateList1", p, list1);
  PCollection<KV<Integer, String>> collection2 = createInput("CreateList2", p, list2);
  PCollection<KV<Integer, CoGbkResult>> coGbkResults =
      KeyedPCollectionTuple.of(tag1, collection1)
          .and(tag2, collection2)
          .apply(CoGroupByKey.<Integer>create());
  return coGbkResults;
}
 

import org.apache.beam.sdk.values.TupleTag; //导入依赖的package包/类
public static <UserT, DestinationT> List<PCollectionView<?>> getDynamicDestinationSideInputs(
    AppliedPTransform<
            PCollection<UserT>, WriteFilesResult<DestinationT>,
            ? extends PTransform<PCollection<UserT>, WriteFilesResult<DestinationT>>>
        transform)
    throws IOException {
  SdkComponents sdkComponents = SdkComponents.create();
  RunnerApi.PTransform transformProto = PTransformTranslation.toProto(transform, sdkComponents);
  List<PCollectionView<?>> views = Lists.newArrayList();
  Map<String, SideInput> sideInputs = getWriteFilesPayload(transform).getSideInputsMap();
  for (Map.Entry<String, SideInput> entry : sideInputs.entrySet()) {
    PCollection<?> originalPCollection =
        checkNotNull(
            (PCollection<?>) transform.getInputs().get(new TupleTag<>(entry.getKey())),
            "no input with tag %s",
            entry.getKey());
    views.add(
        PCollectionViewTranslation.viewFromProto(
            entry.getValue(),
            entry.getKey(),
            originalPCollection,
            transformProto,
            RehydratedComponents.forComponents(sdkComponents.toComponents())));
  }
  return views;
}
 

import org.apache.beam.sdk.values.TupleTag; //导入依赖的package包/类
/**
 * Utility to get mapping between TupleTag and a coder.
 * @param outputs - A map of tuple tags and pcollections
 * @return mapping between TupleTag and a coder
 */
public static Map<TupleTag<?>, Coder<WindowedValue<?>>> getTupleTagCoders(
    Map<TupleTag<?>, PValue> outputs) {
  Map<TupleTag<?>, Coder<WindowedValue<?>>> coderMap = new HashMap<>(outputs.size());

  for (Map.Entry<TupleTag<?>, PValue> output : outputs.entrySet()) {
    // we get the first PValue as all of them are fro the same type.
    PCollection<?> pCollection = (PCollection<?>) output.getValue();
    Coder<?> coder = pCollection.getCoder();
    Coder<? extends BoundedWindow> wCoder =
        pCollection.getWindowingStrategy().getWindowFn().windowCoder();
    @SuppressWarnings("unchecked")
    Coder<WindowedValue<?>> windowedValueCoder =
        (Coder<WindowedValue<?>>) (Coder<?>) WindowedValue.getFullCoder(coder, wCoder);
    coderMap.put(output.getKey(), windowedValueCoder);
  }
  return coderMap;
}
 

import org.apache.beam.sdk.values.TupleTag; //导入依赖的package包/类
/**
 * Returns a new {@code KeyedPCollectionTuple<K>} that is the same as this,
 * appended with the given PCollection.
 */
public <V> KeyedPCollectionTuple<K> and(
    TupleTag< V> tag,
    PCollection<KV<K, V>> pc) {
  if (pc.getPipeline() != getPipeline()) {
    throw new IllegalArgumentException(
        "PCollections come from different Pipelines");
  }
  TaggedKeyedPCollection<K, ?> wrapper =
      new TaggedKeyedPCollection<>(tag, pc);
  Coder<K> myKeyCoder = keyCoder == null ? getKeyCoder(pc) : keyCoder;
  List<TaggedKeyedPCollection<K, ?>>
    newKeyedCollections =
      copyAddLast(
          keyedCollections,
          wrapper);
  return new KeyedPCollectionTuple<>(
      getPipeline(),
      newKeyedCollections,
      schema.getTupleTagList().and(tag),
      myKeyCoder);
}
 

import org.apache.beam.sdk.values.TupleTag; //导入依赖的package包/类
@Override
public Map<TupleTag<?>, PValue> getAdditionalInputs() {
  Map<TupleTag<?>, PValue> additionalInputs = new HashMap<>();
  for (Map.Entry<String, SideInput> sideInputEntry : payload.getSideInputsMap().entrySet()) {
    try {
      additionalInputs.put(
          new TupleTag<>(sideInputEntry.getKey()),
          rehydratedComponents.getPCollection(
              protoTransform.getInputsOrThrow(sideInputEntry.getKey())));
    } catch (IOException exc) {
      throw new IllegalStateException(
          String.format(
              "Could not find input with name %s for %s transform",
              sideInputEntry.getKey(), ParDo.class.getSimpleName()));
    }
  }
  return additionalInputs;
}
 

import org.apache.beam.sdk.values.TupleTag; //导入依赖的package包/类
@Override
public PCollectionList<T> expand(PCollection<T> in) {
  final TupleTagList outputTags = partitionDoFn.getOutputTags();

  PCollectionTuple outputs = in.apply(
      ParDo
      .of(partitionDoFn)
      .withOutputTags(new TupleTag<Void>(){}, outputTags));

  PCollectionList<T> pcs = PCollectionList.empty(in.getPipeline());
  Coder<T> coder = in.getCoder();

  for (TupleTag<?> outputTag : outputTags.getAll()) {
    // All the tuple tags are actually TupleTag<T>
    // And all the collections are actually PCollection<T>
    @SuppressWarnings("unchecked")
    TupleTag<T> typedOutputTag = (TupleTag<T>) outputTag;
    pcs = pcs.and(outputs.get(typedOutputTag).setCoder(coder));
  }
  return pcs;
}
 

import org.apache.beam.sdk.values.TupleTag; //导入依赖的package包/类
public static <InputT, OutputT, RestrictionT>
ProcessFnRunner<InputT, OutputT, RestrictionT>
newProcessFnRunner(
    ProcessFn<InputT, OutputT, RestrictionT, ?> fn,
    PipelineOptions options,
    Collection<PCollectionView<?>> views,
    ReadyCheckingSideInputReader sideInputReader,
    OutputManager outputManager,
    TupleTag<OutputT> mainOutputTag,
    List<TupleTag<?>> additionalOutputTags,
    StepContext stepContext,
    WindowingStrategy<?, ?> windowingStrategy) {
  return new ProcessFnRunner<>(
      simpleRunner(
          options,
          fn,
          sideInputReader,
          outputManager,
          mainOutputTag,
          additionalOutputTags,
          stepContext,
          windowingStrategy),
      views,
      sideInputReader);
}
 

import org.apache.beam.sdk.values.TupleTag; //导入依赖的package包/类
@Test
@SuppressWarnings({"rawtypes", "unchecked"})
public void testWithDuplicatedCollections() {
  String transformName = "transform";
  when(transform.getName()).thenReturn(transformName);

  JavaStream javaStream1 = mock(JavaStream.class);
  TranslationContext translationContext = mock(TranslationContext.class);

  Map<TupleTag<?>, PValue> inputs = new HashMap<>();
  TupleTag tag1 = mock(TupleTag.class);
  PCollection mockCollection1 = mock(PCollection.class);
  inputs.put(tag1, mockCollection1);

  TupleTag tag2 = mock(TupleTag.class);
  inputs.put(tag2, mockCollection1);

  when(translationContext.getInputs()).thenReturn(inputs);
  when(translationContext.getInputStream(mockCollection1)).thenReturn(javaStream1);
  when(translationContext.getPipelineOptions())
      .thenReturn(PipelineOptionsFactory.as(GearpumpPipelineOptions.class));

  translator.translate(transform, translationContext);
  verify(javaStream1).map(any(MapFunction.class), eq("dummy"));
  verify(javaStream1).merge(any(JavaStream.class), eq(1), eq(transformName));
}
 

import org.apache.beam.sdk.values.TupleTag; //导入依赖的package包/类
private ComposedCombineFnWithContext(
    ImmutableList<SerializableFunction<DataT, ?>> extractInputFns,
    ImmutableList<CombineFnWithContext<?, ?, ?>> combineFnWithContexts,
    ImmutableList<TupleTag<?>> outputTags) {
  @SuppressWarnings({"unchecked", "rawtypes"})
  List<SerializableFunction<DataT, Object>> castedExtractInputFns =
      (List) extractInputFns;
  this.extractInputFns = castedExtractInputFns;

  @SuppressWarnings({"rawtypes", "unchecked"})
  List<CombineFnWithContext<Object, Object, Object>> castedCombineFnWithContexts =
      (List) combineFnWithContexts;
  this.combineFnWithContexts = castedCombineFnWithContexts;

  this.outputTags = outputTags;
  this.combineFnCount = this.combineFnWithContexts.size();
}
 

import org.apache.beam.sdk.values.TupleTag; //导入依赖的package包/类
/**
 * Returns a {@link ComposedCombineFnWithContext} with an additional {@link GlobalCombineFn}.
 */
public <InputT, OutputT> ComposedCombineFnWithContext<DataT> with(
    SimpleFunction<DataT, InputT> extractInputFn,
    GlobalCombineFn<InputT, ?, OutputT> globalCombineFn,
    TupleTag<OutputT> outputTag) {
  checkUniqueness(outputTags, outputTag);
  return new ComposedCombineFnWithContext<>(
      ImmutableList.<SerializableFunction<DataT, ?>>builder()
          .addAll(extractInputFns)
          .add(extractInputFn)
          .build(),
      ImmutableList.<CombineFnWithContext<?, ?, ?>>builder()
          .addAll(combineFnWithContexts)
          .add(CombineFnUtil.toFnWithContext(globalCombineFn))
          .build(),
      ImmutableList.<TupleTag<?>>builder()
          .addAll(outputTags)
          .add(outputTag)
          .build());
}