Examples with KV org.apache.samza.operators.KV used on opensource projects

Example 61 with KV

use of org.apache.samza.operators.KV in project beam by apache.

the class ReadTranslator method translate.

@Override
public void translate(PTransform<PBegin, PCollection<T>> transform, TransformHierarchy.Node node, TranslationContext ctx) {
    final PCollection<T> output = ctx.getOutput(transform);
    final Coder<WindowedValue<T>> coder = SamzaCoders.of(output);
    final Source<?> source = transform instanceof SplittableParDo.PrimitiveBoundedRead ? ((SplittableParDo.PrimitiveBoundedRead) transform).getSource() : ((SplittableParDo.PrimitiveUnboundedRead) transform).getSource();
    final String id = ctx.getIdForPValue(output);
    // Create system descriptor
    final GenericSystemDescriptor systemDescriptor;
    if (source instanceof BoundedSource) {
        systemDescriptor = new GenericSystemDescriptor(id, BoundedSourceSystem.Factory.class.getName());
    } else {
        systemDescriptor = new GenericSystemDescriptor(id, UnboundedSourceSystem.Factory.class.getName());
    }
    final Map<String, String> systemConfig = ImmutableMap.of("source", Base64Serializer.serializeUnchecked(source), "coder", Base64Serializer.serializeUnchecked(coder), "stepName", node.getFullName());
    systemDescriptor.withSystemConfigs(systemConfig);
    // Create stream descriptor
    @SuppressWarnings("unchecked") final Serde<KV<?, OpMessage<T>>> kvSerde = (Serde) KVSerde.of(new NoOpSerde<>(), new NoOpSerde<>());
    final GenericInputDescriptor<KV<?, OpMessage<T>>> inputDescriptor = systemDescriptor.getInputDescriptor(id, kvSerde);
    if (source instanceof BoundedSource) {
        inputDescriptor.isBounded();
    }
    ctx.registerInputMessageStream(output, inputDescriptor);
}

Example 62 with KV

use of org.apache.samza.operators.KV in project samza by apache.

the class TestExecutionPlanner method createStreamGraphWithStreamTableJoin.

private StreamApplicationDescriptorImpl createStreamGraphWithStreamTableJoin() {
    /**
     * Example stream-table join app. Expected partition counts of intermediate streams introduced
     * by partitionBy operations are enclosed in quotes.
     *
     *    input2 (16) -> partitionBy ("32") -> send-to-table t
     *
     *                                      join-table t —————
     *                                       |                |
     *    input1 (64) -> partitionBy ("32") _|                |
     *                                                       join -> output1 (8)
     *                                                        |
     *                                      input3 (32) ——————
     */
    return new StreamApplicationDescriptorImpl(appDesc -> {
        MessageStream<KV<Object, Object>> messageStream1 = appDesc.getInputStream(input1Descriptor);
        MessageStream<KV<Object, Object>> messageStream2 = appDesc.getInputStream(input2Descriptor);
        MessageStream<KV<Object, Object>> messageStream3 = appDesc.getInputStream(input3Descriptor);
        OutputStream<KV<Object, Object>> output1 = appDesc.getOutputStream(output1Descriptor);
        TableDescriptor tableDescriptor = new TestLocalTableDescriptor.MockLocalTableDescriptor("table-id", new KVSerde(new StringSerde(), new StringSerde()));
        Table table = appDesc.getTable(tableDescriptor);
        messageStream2.partitionBy(m -> m.key, m -> m.value, mock(KVSerde.class), "p1").sendTo(table);
        messageStream1.partitionBy(m -> m.key, m -> m.value, mock(KVSerde.class), "p2").join(table, mock(StreamTableJoinFunction.class)).join(messageStream3, mock(JoinFunction.class), mock(Serde.class), mock(Serde.class), mock(Serde.class), Duration.ofHours(1), "j2").sendTo(output1);
    }, config);
}

Example 63 with KV

use of org.apache.samza.operators.KV in project samza by apache.

the class TestExecutionPlanner method createStreamGraphWithInvalidStreamTableJoin.

private StreamApplicationDescriptorImpl createStreamGraphWithInvalidStreamTableJoin() {
    /**
     * Example stream-table join that is invalid due to disagreement in partition count
     * between the 2 input streams.
     *
     *    input1 (64) -> send-to-table t
     *
     *                   join-table t -> output1 (8)
     *                         |
     *    input2 (16) —————————
     */
    return new StreamApplicationDescriptorImpl(appDesc -> {
        MessageStream<KV<Object, Object>> messageStream1 = appDesc.getInputStream(input1Descriptor);
        MessageStream<KV<Object, Object>> messageStream2 = appDesc.getInputStream(input2Descriptor);
        OutputStream<KV<Object, Object>> output1 = appDesc.getOutputStream(output1Descriptor);
        TableDescriptor tableDescriptor = new TestLocalTableDescriptor.MockLocalTableDescriptor("table-id", new KVSerde(new StringSerde(), new StringSerde()));
        Table table = appDesc.getTable(tableDescriptor);
        messageStream1.sendTo(table);
        messageStream1.join(table, mock(StreamTableJoinFunction.class)).join(messageStream2, mock(JoinFunction.class), mock(Serde.class), mock(Serde.class), mock(Serde.class), Duration.ofHours(1), "j2").sendTo(output1);
    }, config);
}

Example 64 with KV

use of org.apache.samza.operators.KV in project samza by apache.

the class TestJobGraphJsonGenerator method testRepartitionedWindowStreamApplication.

@Test
public void testRepartitionedWindowStreamApplication() throws Exception {
    Map<String, String> configMap = new HashMap<>();
    configMap.put(JobConfig.JOB_NAME, "test-app");
    configMap.put(JobConfig.JOB_DEFAULT_SYSTEM, "test-system");
    StreamTestUtils.addStreamConfigs(configMap, "PageView", "hdfs", "hdfs:/user/dummy/PageViewEvent");
    StreamTestUtils.addStreamConfigs(configMap, "PageViewCount", "kafka", "PageViewCount");
    Config config = new MapConfig(configMap);
    // set up external partition count
    Map<String, Integer> system1Map = new HashMap<>();
    system1Map.put("hdfs:/user/dummy/PageViewEvent", 512);
    Map<String, Integer> system2Map = new HashMap<>();
    system2Map.put("PageViewCount", 16);
    SystemAdmin systemAdmin1 = createSystemAdmin(system1Map);
    SystemAdmin systemAdmin2 = createSystemAdmin(system2Map);
    SystemAdmins systemAdmins = mock(SystemAdmins.class);
    when(systemAdmins.getSystemAdmin("hdfs")).thenReturn(systemAdmin1);
    when(systemAdmins.getSystemAdmin("kafka")).thenReturn(systemAdmin2);
    StreamManager streamManager = new StreamManager(systemAdmins);
    StreamApplicationDescriptorImpl graphSpec = new StreamApplicationDescriptorImpl(appDesc -> {
        KVSerde<String, PageViewEvent> pvSerde = KVSerde.of(new StringSerde(), new JsonSerdeV2<>(PageViewEvent.class));
        GenericSystemDescriptor isd = new GenericSystemDescriptor("hdfs", "mockSystemFactoryClass");
        GenericInputDescriptor<KV<String, PageViewEvent>> pageView = isd.getInputDescriptor("PageView", pvSerde);
        KVSerde<String, Long> pvcSerde = KVSerde.of(new StringSerde(), new LongSerde());
        GenericSystemDescriptor osd = new GenericSystemDescriptor("kafka", "mockSystemFactoryClass");
        GenericOutputDescriptor<KV<String, Long>> pageViewCount = osd.getOutputDescriptor("PageViewCount", pvcSerde);
        MessageStream<KV<String, PageViewEvent>> inputStream = appDesc.getInputStream(pageView);
        OutputStream<KV<String, Long>> outputStream = appDesc.getOutputStream(pageViewCount);
        inputStream.partitionBy(kv -> kv.getValue().getCountry(), kv -> kv.getValue(), pvSerde, "keyed-by-country").window(Windows.keyedTumblingWindow(kv -> kv.getValue().getCountry(), Duration.ofSeconds(10L), () -> 0L, (m, c) -> c + 1L, new StringSerde(), new LongSerde()), "count-by-country").map(pane -> new KV<>(pane.getKey().getKey(), pane.getMessage())).sendTo(outputStream);
    }, config);
    ExecutionPlanner planner = new ExecutionPlanner(config, streamManager);
    ExecutionPlan plan = planner.plan(graphSpec);
    String json = plan.getPlanAsJson();
    System.out.println(json);
    // deserialize
    ObjectMapper mapper = new ObjectMapper();
    JobGraphJsonGenerator.JobGraphJson nodes = mapper.readValue(json, JobGraphJsonGenerator.JobGraphJson.class);
    JobGraphJsonGenerator.OperatorGraphJson operatorGraphJson = nodes.jobs.get(0).operatorGraph;
    assertEquals(2, operatorGraphJson.inputStreams.size());
    assertEquals(4, operatorGraphJson.operators.size());
    assertEquals(1, nodes.sourceStreams.size());
    assertEquals(1, nodes.sinkStreams.size());
    assertEquals(1, nodes.intermediateStreams.size());
    // verify partitionBy op output to the intermdiate stream of the same id
    assertEquals(operatorGraphJson.operators.get("test-app-1-partition_by-keyed-by-country").get("outputStreamId"), "test-app-1-partition_by-keyed-by-country");
    assertEquals(operatorGraphJson.operators.get("test-app-1-send_to-5").get("outputStreamId"), "PageViewCount");
}

Example 65 with KV

use of org.apache.samza.operators.KV in project samza by apache.

the class TestJobGraphJsonGenerator method testRepartitionedJoinStreamApplication.

@Test
public void testRepartitionedJoinStreamApplication() throws Exception {
    /**
     * the graph looks like the following.
     * number in parentheses () indicates number of stream partitions.
     * number in parentheses in quotes ("") indicates expected partition count.
     * number in square brackets [] indicates operator ID.
     *
     * input3 (32) -> filter [7] -> partitionBy [8] ("64") -> map [10] -> join [14] -> sendTo(output2) [15] (16)
     *                                                                   |
     *              input2 (16) -> partitionBy [3] ("64") -> filter [5] -| -> sink [13]
     *                                                                   |
     *                                         input1 (64) -> map [1] -> join [11] -> sendTo(output1) [12] (8)
     */
    Map<String, String> configMap = new HashMap<>();
    configMap.put(JobConfig.JOB_NAME, "test-app");
    configMap.put(JobConfig.JOB_DEFAULT_SYSTEM, "test-system");
    StreamTestUtils.addStreamConfigs(configMap, "input1", "system1", "input1");
    StreamTestUtils.addStreamConfigs(configMap, "input2", "system2", "input2");
    StreamTestUtils.addStreamConfigs(configMap, "input3", "system2", "input3");
    StreamTestUtils.addStreamConfigs(configMap, "output1", "system1", "output1");
    StreamTestUtils.addStreamConfigs(configMap, "output2", "system2", "output2");
    Config config = new MapConfig(configMap);
    // set up external partition count
    Map<String, Integer> system1Map = new HashMap<>();
    system1Map.put("input1", 64);
    system1Map.put("output1", 8);
    Map<String, Integer> system2Map = new HashMap<>();
    system2Map.put("input2", 16);
    system2Map.put("input3", 32);
    system2Map.put("output2", 16);
    SystemAdmin systemAdmin1 = createSystemAdmin(system1Map);
    SystemAdmin systemAdmin2 = createSystemAdmin(system2Map);
    SystemAdmins systemAdmins = mock(SystemAdmins.class);
    when(systemAdmins.getSystemAdmin("system1")).thenReturn(systemAdmin1);
    when(systemAdmins.getSystemAdmin("system2")).thenReturn(systemAdmin2);
    StreamManager streamManager = new StreamManager(systemAdmins);
    StreamApplicationDescriptorImpl graphSpec = new StreamApplicationDescriptorImpl(appDesc -> {
        KVSerde<Object, Object> kvSerde = new KVSerde<>(new NoOpSerde(), new NoOpSerde());
        String mockSystemFactoryClass = "factory.class.name";
        GenericSystemDescriptor system1 = new GenericSystemDescriptor("system1", mockSystemFactoryClass);
        GenericSystemDescriptor system2 = new GenericSystemDescriptor("system2", mockSystemFactoryClass);
        GenericInputDescriptor<KV<Object, Object>> input1Descriptor = system1.getInputDescriptor("input1", kvSerde);
        GenericInputDescriptor<KV<Object, Object>> input2Descriptor = system2.getInputDescriptor("input2", kvSerde);
        GenericInputDescriptor<KV<Object, Object>> input3Descriptor = system2.getInputDescriptor("input3", kvSerde);
        GenericOutputDescriptor<KV<Object, Object>> output1Descriptor = system1.getOutputDescriptor("output1", kvSerde);
        GenericOutputDescriptor<KV<Object, Object>> output2Descriptor = system2.getOutputDescriptor("output2", kvSerde);
        MessageStream<KV<Object, Object>> messageStream1 = appDesc.getInputStream(input1Descriptor).map(m -> m);
        MessageStream<KV<Object, Object>> messageStream2 = appDesc.getInputStream(input2Descriptor).partitionBy(m -> m.key, m -> m.value, mock(KVSerde.class), "p1").filter(m -> true);
        MessageStream<KV<Object, Object>> messageStream3 = appDesc.getInputStream(input3Descriptor).filter(m -> true).partitionBy(m -> m.key, m -> m.value, mock(KVSerde.class), "p2").map(m -> m);
        OutputStream<KV<Object, Object>> outputStream1 = appDesc.getOutputStream(output1Descriptor);
        OutputStream<KV<Object, Object>> outputStream2 = appDesc.getOutputStream(output2Descriptor);
        messageStream1.join(messageStream2, (JoinFunction<Object, KV<Object, Object>, KV<Object, Object>, KV<Object, Object>>) mock(JoinFunction.class), mock(Serde.class), mock(Serde.class), mock(Serde.class), Duration.ofHours(2), "j1").sendTo(outputStream1);
        messageStream2.sink((message, collector, coordinator) -> {
        });
        messageStream3.join(messageStream2, (JoinFunction<Object, KV<Object, Object>, KV<Object, Object>, KV<Object, Object>>) mock(JoinFunction.class), mock(Serde.class), mock(Serde.class), mock(Serde.class), Duration.ofHours(1), "j2").sendTo(outputStream2);
    }, config);
    ExecutionPlanner planner = new ExecutionPlanner(config, streamManager);
    ExecutionPlan plan = planner.plan(graphSpec);
    String json = plan.getPlanAsJson();
    System.out.println(json);
    // deserialize
    ObjectMapper mapper = new ObjectMapper();
    JobGraphJsonGenerator.JobGraphJson nodes = mapper.readValue(json, JobGraphJsonGenerator.JobGraphJson.class);
    assertEquals(5, nodes.jobs.get(0).operatorGraph.inputStreams.size());
    assertEquals(11, nodes.jobs.get(0).operatorGraph.operators.size());
    assertEquals(3, nodes.sourceStreams.size());
    assertEquals(2, nodes.sinkStreams.size());
    assertEquals(2, nodes.intermediateStreams.size());
}