Examples with TransformSpec org.apache.druid.segment.transform.TransformSpec used on opensource projects

Example 16 with TransformSpec

use of org.apache.druid.segment.transform.TransformSpec in project druid by druid-io.

the class KinesisIndexTaskTest method testRunWithTransformSpec.

@Test(timeout = 120_000L)
public void testRunWithTransformSpec() throws Exception {
    recordSupplier.assign(EasyMock.anyObject());
    EasyMock.expectLastCall().anyTimes();
    EasyMock.expect(recordSupplier.getEarliestSequenceNumber(EasyMock.anyObject())).andReturn("0").anyTimes();
    recordSupplier.seek(EasyMock.anyObject(), EasyMock.anyString());
    EasyMock.expectLastCall().anyTimes();
    EasyMock.expect(recordSupplier.poll(EasyMock.anyLong())).andReturn(generateRecords(0, 13)).once();
    recordSupplier.close();
    EasyMock.expectLastCall().once();
    replayAll();
    final KinesisIndexTask task = createTask(null, NEW_DATA_SCHEMA.withTransformSpec(new TransformSpec(new SelectorDimFilter("dim1", "b", null), ImmutableList.of(new ExpressionTransform("dim1t", "concat(dim1,dim1)", ExprMacroTable.nil())))), new KinesisIndexTaskIOConfig(0, "sequence0", new SeekableStreamStartSequenceNumbers<>(STREAM, ImmutableMap.of(SHARD_ID1, "0"), ImmutableSet.of()), new SeekableStreamEndSequenceNumbers<>(STREAM, ImmutableMap.of(SHARD_ID1, "4")), true, null, null, INPUT_FORMAT, "awsEndpoint", null, null, null, null, false));
    final ListenableFuture<TaskStatus> future = runTask(task);
    // Wait for the task to start reading
    while (task.getRunner().getStatus() != SeekableStreamIndexTaskRunner.Status.READING) {
        Thread.sleep(10);
    }
    // Wait for task to exit
    Assert.assertEquals(TaskState.SUCCESS, future.get().getStatusCode());
    verifyAll();
    // Check metrics
    Assert.assertEquals(1, task.getRunner().getRowIngestionMeters().getProcessed());
    Assert.assertEquals(0, task.getRunner().getRowIngestionMeters().getUnparseable());
    Assert.assertEquals(4, task.getRunner().getRowIngestionMeters().getThrownAway());
    // Check published metadata
    assertEqualsExceptVersion(ImmutableList.of(sdd("2009/P1D", 0)), publishedDescriptors());
    Assert.assertEquals(new KinesisDataSourceMetadata(new SeekableStreamEndSequenceNumbers<>(STREAM, ImmutableMap.of(SHARD_ID1, "4"))), newDataSchemaMetadata());
    // Check segments in deep storage
    final List<SegmentDescriptor> publishedDescriptors = publishedDescriptors();
    Assert.assertEquals(ImmutableList.of("b"), readSegmentColumn("dim1", publishedDescriptors.get(0)));
    Assert.assertEquals(ImmutableList.of("bb"), readSegmentColumn("dim1t", publishedDescriptors.get(0)));
}

Example 17 with TransformSpec

use of org.apache.druid.segment.transform.TransformSpec in project druid by druid-io.

the class CompactionTask method runTask.

@Override
public TaskStatus runTask(TaskToolbox toolbox) throws Exception {
    final List<ParallelIndexIngestionSpec> ingestionSpecs = createIngestionSchema(toolbox, getTaskLockHelper().getLockGranularityToUse(), segmentProvider, partitionConfigurationManager, dimensionsSpec, transformSpec, metricsSpec, granularitySpec, toolbox.getCoordinatorClient(), segmentCacheManagerFactory, retryPolicyFactory, ioConfig.isDropExisting());
    final List<ParallelIndexSupervisorTask> indexTaskSpecs = IntStream.range(0, ingestionSpecs.size()).mapToObj(i -> {
        // The ID of SubtaskSpecs is used as the base sequenceName in segment allocation protocol.
        // The indexing tasks generated by the compaction task should use different sequenceNames
        // so that they can allocate valid segment IDs with no duplication.
        ParallelIndexIngestionSpec ingestionSpec = ingestionSpecs.get(i);
        final String baseSequenceName = createIndexTaskSpecId(i);
        return newTask(baseSequenceName, ingestionSpec);
    }).collect(Collectors.toList());
    if (indexTaskSpecs.isEmpty()) {
        String msg = StringUtils.format("Can't find segments from inputSpec[%s], nothing to do.", ioConfig.getInputSpec());
        log.warn(msg);
        return TaskStatus.failure(getId(), msg);
    } else {
        registerResourceCloserOnAbnormalExit(currentSubTaskHolder);
        final int totalNumSpecs = indexTaskSpecs.size();
        log.info("Generated [%d] compaction task specs", totalNumSpecs);
        int failCnt = 0;
        for (ParallelIndexSupervisorTask eachSpec : indexTaskSpecs) {
            final String json = toolbox.getJsonMapper().writerWithDefaultPrettyPrinter().writeValueAsString(eachSpec);
            if (!currentSubTaskHolder.setTask(eachSpec)) {
                String errMsg = "Task was asked to stop. Finish as failed.";
                log.info(errMsg);
                return TaskStatus.failure(getId(), errMsg);
            }
            try {
                if (eachSpec.isReady(toolbox.getTaskActionClient())) {
                    log.info("Running indexSpec: " + json);
                    final TaskStatus eachResult = eachSpec.run(toolbox);
                    if (!eachResult.isSuccess()) {
                        failCnt++;
                        log.warn("Failed to run indexSpec: [%s].\nTrying the next indexSpec.", json);
                    }
                } else {
                    failCnt++;
                    log.warn("indexSpec is not ready: [%s].\nTrying the next indexSpec.", json);
                }
            } catch (Exception e) {
                failCnt++;
                log.warn(e, "Failed to run indexSpec: [%s].\nTrying the next indexSpec.", json);
            }
        }
        String msg = StringUtils.format("Ran [%d] specs, [%d] succeeded, [%d] failed", totalNumSpecs, totalNumSpecs - failCnt, failCnt);
        log.info(msg);
        return failCnt == 0 ? TaskStatus.success(getId()) : TaskStatus.failure(getId(), msg);
    }
}

Example 18 with TransformSpec

use of org.apache.druid.segment.transform.TransformSpec in project druid by druid-io.

the class AbstractBatchIndexTask method compactionStateAnnotateFunction.

public static Function<Set<DataSegment>, Set<DataSegment>> compactionStateAnnotateFunction(boolean storeCompactionState, TaskToolbox toolbox, IngestionSpec ingestionSpec) {
    if (storeCompactionState) {
        TuningConfig tuningConfig = ingestionSpec.getTuningConfig();
        GranularitySpec granularitySpec = ingestionSpec.getDataSchema().getGranularitySpec();
        // We do not need to store dimensionExclusions and spatialDimensions since auto compaction does not support them
        DimensionsSpec dimensionsSpec = ingestionSpec.getDataSchema().getDimensionsSpec() == null ? null : new DimensionsSpec(ingestionSpec.getDataSchema().getDimensionsSpec().getDimensions());
        // We only need to store filter since that is the only field auto compaction support
        Map<String, Object> transformSpec = ingestionSpec.getDataSchema().getTransformSpec() == null || TransformSpec.NONE.equals(ingestionSpec.getDataSchema().getTransformSpec()) ? null : new ClientCompactionTaskTransformSpec(ingestionSpec.getDataSchema().getTransformSpec().getFilter()).asMap(toolbox.getJsonMapper());
        List<Object> metricsSpec = ingestionSpec.getDataSchema().getAggregators() == null ? null : toolbox.getJsonMapper().convertValue(ingestionSpec.getDataSchema().getAggregators(), new TypeReference<List<Object>>() {
        });
        final CompactionState compactionState = new CompactionState(tuningConfig.getPartitionsSpec(), dimensionsSpec, metricsSpec, transformSpec, tuningConfig.getIndexSpec().asMap(toolbox.getJsonMapper()), granularitySpec.asMap(toolbox.getJsonMapper()));
        return segments -> segments.stream().map(s -> s.withLastCompactionState(compactionState)).collect(Collectors.toSet());
    } else {
        return Function.identity();
    }
}

Example 19 with TransformSpec

use of org.apache.druid.segment.transform.TransformSpec in project druid by druid-io.

the class InputSourceSamplerTest method testWithFilter.

@Test
public void testWithFilter() throws IOException {
    final TimestampSpec timestampSpec = new TimestampSpec("t", null, null);
    final DimensionsSpec dimensionsSpec = new DimensionsSpec(null);
    final TransformSpec transformSpec = new TransformSpec(new SelectorDimFilter("dim1", "foo", null), null);
    final AggregatorFactory[] aggregatorFactories = { new LongSumAggregatorFactory("met1", "met1") };
    final GranularitySpec granularitySpec = new UniformGranularitySpec(Granularities.DAY, Granularities.HOUR, true, null);
    final DataSchema dataSchema = createDataSchema(timestampSpec, dimensionsSpec, aggregatorFactories, granularitySpec, transformSpec);
    final InputSource inputSource = createInputSource(getTestRows(), dataSchema);
    final InputFormat inputFormat = createInputFormat();
    SamplerResponse response = inputSourceSampler.sample(inputSource, inputFormat, dataSchema, null);
    Assert.assertEquals(5, response.getNumRowsRead());
    Assert.assertEquals(4, response.getNumRowsIndexed());
    Assert.assertEquals(3, response.getData().size());
    List<SamplerResponseRow> data = response.getData();
    assertEqualsSamplerResponseRow(new SamplerResponseRow(getRawColumns().get(0), new SamplerTestUtils.MapAllowingNullValuesBuilder<String, Object>().put("__time", 1555934400000L).put("dim1", "foo").put("dim2", null).put("met1", 6L).build(), null, null), data.get(0));
    assertEqualsSamplerResponseRow(new SamplerResponseRow(getRawColumns().get(4), new SamplerTestUtils.MapAllowingNullValuesBuilder<String, Object>().put("__time", 1555934400000L).put("dim1", "foo").put("dim2", "bar").put("met1", 5L).build(), null, null), data.get(1));
    assertEqualsSamplerResponseRow(new SamplerResponseRow(getRawColumns().get(5), null, true, getUnparseableTimestampString()), data.get(2));
}

Example 20 with TransformSpec

use of org.apache.druid.segment.transform.TransformSpec in project druid by druid-io.

the class InputSourceSamplerTest method testWithTransformsDimensionsSpec.

@Test
public void testWithTransformsDimensionsSpec() throws IOException {
    final TimestampSpec timestampSpec = new TimestampSpec("t", null, null);
    final DimensionsSpec dimensionsSpec = new DimensionsSpec(ImmutableList.of(StringDimensionSchema.create("dim1PlusBar")));
    final TransformSpec transformSpec = new TransformSpec(null, ImmutableList.of(new ExpressionTransform("dim1PlusBar", "concat(dim1 + 'bar')", TestExprMacroTable.INSTANCE)));
    final AggregatorFactory[] aggregatorFactories = { new LongSumAggregatorFactory("met1", "met1") };
    final GranularitySpec granularitySpec = new UniformGranularitySpec(Granularities.DAY, Granularities.HOUR, true, null);
    final DataSchema dataSchema = createDataSchema(timestampSpec, dimensionsSpec, aggregatorFactories, granularitySpec, transformSpec);
    final InputSource inputSource = createInputSource(getTestRows(), dataSchema);
    final InputFormat inputFormat = createInputFormat();
    SamplerResponse response = inputSourceSampler.sample(inputSource, inputFormat, dataSchema, null);
    Assert.assertEquals(6, response.getNumRowsRead());
    Assert.assertEquals(5, response.getNumRowsIndexed());
    Assert.assertEquals(3, response.getData().size());
    List<SamplerResponseRow> data = response.getData();
    assertEqualsSamplerResponseRow(new SamplerResponseRow(getRawColumns().get(0), new SamplerTestUtils.MapAllowingNullValuesBuilder<String, Object>().put("__time", 1555934400000L).put("dim1PlusBar", "foobar").put("met1", 11L).build(), null, null), data.get(0));
    assertEqualsSamplerResponseRow(new SamplerResponseRow(getRawColumns().get(3), new SamplerTestUtils.MapAllowingNullValuesBuilder<String, Object>().put("__time", 1555934400000L).put("dim1PlusBar", "foo2bar").put("met1", 4L).build(), null, null), data.get(1));
    assertEqualsSamplerResponseRow(new SamplerResponseRow(getRawColumns().get(5), null, true, getUnparseableTimestampString()), data.get(2));
}