use of org.apache.druid.segment.transform.ExpressionTransform in project druid by druid-io.
the class InputSourceSamplerTest method testMultipleJsonStringInOneBlock.
/**
* This case tests sampling for multiple json lines in one text block
* Currently only RecordSupplierInputSource supports this kind of input, see https://github.com/apache/druid/pull/10383 for more information
*
* This test combines illegal json block and legal json block together to verify:
* 1. all lines in the illegal json block should not be parsed
* 2. the illegal json block should not affect the processing of the 2nd record
* 3. all lines in legal json block should be parsed successfully
*/
@Test
public void testMultipleJsonStringInOneBlock() throws IOException {
if (!ParserType.STR_JSON.equals(parserType) || !useInputFormatApi) {
return;
}
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);
List<String> jsonBlockList = ImmutableList.of(// include the line which can't be parsed into JSON object to form a illegal json block
String.join("", STR_JSON_ROWS), // exclude the last line to form a legal json block
STR_JSON_ROWS.stream().limit(STR_JSON_ROWS.size() - 1).collect(Collectors.joining()));
SamplerResponse response = inputSourceSampler.sample(new RecordSupplierInputSource("topicName", new TestRecordSupplier(jsonBlockList), true), createInputFormat(), dataSchema, new SamplerConfig(200, 3000));
//
// the 1st json block contains STR_JSON_ROWS.size() lines, and 2nd json block contains STR_JSON_ROWS.size()-1 lines
// together there should STR_JSON_ROWS.size() * 2 - 1 lines
//
int illegalRows = STR_JSON_ROWS.size();
int legalRows = STR_JSON_ROWS.size() - 1;
Assert.assertEquals(illegalRows + legalRows, response.getNumRowsRead());
Assert.assertEquals(legalRows, response.getNumRowsIndexed());
Assert.assertEquals(illegalRows + 2, response.getData().size());
List<SamplerResponseRow> data = response.getData();
List<Map<String, Object>> rawColumnList = this.getRawColumns();
int index = 0;
//
// first n rows are related to the first json block which fails to parse
//
String parseExceptionMessage;
if (useInputFormatApi) {
parseExceptionMessage = "Timestamp[bad_timestamp] is unparseable! Event: {t=bad_timestamp, dim1=foo, met1=6}";
} else {
parseExceptionMessage = "Timestamp[bad_timestamp] is unparseable! Event: {t=bad_timestamp, dim1=foo, met1=6}";
}
for (; index < illegalRows; index++) {
assertEqualsSamplerResponseRow(new SamplerResponseRow(rawColumnList.get(index), null, true, parseExceptionMessage), data.get(index));
}
//
// following are parsed rows for legal json block
//
assertEqualsSamplerResponseRow(new SamplerResponseRow(rawColumnList.get(0), new SamplerTestUtils.MapAllowingNullValuesBuilder<String, Object>().put("__time", 1555934400000L).put("dim1PlusBar", "foobar").put("met1", 11L).build(), null, null), data.get(index++));
assertEqualsSamplerResponseRow(new SamplerResponseRow(rawColumnList.get(3), new SamplerTestUtils.MapAllowingNullValuesBuilder<String, Object>().put("__time", 1555934400000L).put("dim1PlusBar", "foo2bar").put("met1", 4L).build(), null, null), data.get(index));
}
use of org.apache.druid.segment.transform.ExpressionTransform in project druid by druid-io.
the class InputSourceSamplerTest method testWithTransformsAutoDimensions.
@Test
public void testWithTransformsAutoDimensions() throws IOException {
final TimestampSpec timestampSpec = new TimestampSpec("t", null, null);
final DimensionsSpec dimensionsSpec = new DimensionsSpec(null);
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(4, 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(3), new SamplerTestUtils.MapAllowingNullValuesBuilder<String, Object>().put("__time", 1555934400000L).put("dim1", "foo2").put("dim2", null).put("met1", 4L).build(), null, null), data.get(1));
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(2));
assertEqualsSamplerResponseRow(new SamplerResponseRow(getRawColumns().get(5), null, true, getUnparseableTimestampString()), data.get(3));
}
use of org.apache.druid.segment.transform.ExpressionTransform in project druid by druid-io.
the class KafkaIndexTaskTest method testRunWithTransformSpec.
@Test(timeout = 60_000L)
public void testRunWithTransformSpec() throws Exception {
final KafkaIndexTask 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 KafkaIndexTaskIOConfig(0, "sequence0", new SeekableStreamStartSequenceNumbers<>(topic, ImmutableMap.of(0, 0L), ImmutableSet.of()), new SeekableStreamEndSequenceNumbers<>(topic, ImmutableMap.of(0, 5L)), kafkaServer.consumerProperties(), KafkaSupervisorIOConfig.DEFAULT_POLL_TIMEOUT_MILLIS, true, null, null, INPUT_FORMAT));
final ListenableFuture<TaskStatus> future = runTask(task);
// Wait for the task to start reading
while (task.getRunner().getStatus() != Status.READING) {
Thread.sleep(10);
}
// Insert data
insertData();
// Wait for task to exit
Assert.assertEquals(TaskState.SUCCESS, future.get().getStatusCode());
// 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
final List<SegmentDescriptor> publishedDescriptors = publishedDescriptors();
assertEqualsExceptVersion(ImmutableList.of(sdd("2009/P1D", 0)), publishedDescriptors);
Assert.assertEquals(new KafkaDataSourceMetadata(new SeekableStreamEndSequenceNumbers<>(topic, ImmutableMap.of(0, 5L))), newDataSchemaMetadata());
// Check segments in deep storage
Assert.assertEquals(ImmutableList.of("b"), readSegmentColumn("dim1", publishedDescriptors.get(0)));
Assert.assertEquals(ImmutableList.of("bb"), readSegmentColumn("dim1t", publishedDescriptors.get(0)));
}
use of org.apache.druid.segment.transform.ExpressionTransform in project druid by druid-io.
the class KafkaIndexTaskTest method testSerde.
@Test
public void testSerde() throws Exception {
// This is both a serde test and a regression test for https://github.com/apache/druid/issues/7724.
final KafkaIndexTask task = createTask("taskid", NEW_DATA_SCHEMA.withTransformSpec(new TransformSpec(null, ImmutableList.of(new ExpressionTransform("beep", "nofunc()", ExprMacroTable.nil())))), new KafkaIndexTaskIOConfig(0, "sequence", new SeekableStreamStartSequenceNumbers<>(topic, ImmutableMap.of(), ImmutableSet.of()), new SeekableStreamEndSequenceNumbers<>(topic, ImmutableMap.of()), ImmutableMap.of(), KafkaSupervisorIOConfig.DEFAULT_POLL_TIMEOUT_MILLIS, true, null, null, INPUT_FORMAT));
final Task task1 = OBJECT_MAPPER.readValue(OBJECT_MAPPER.writeValueAsBytes(task), Task.class);
Assert.assertEquals(task, task1);
}
use of org.apache.druid.segment.transform.ExpressionTransform in project druid by druid-io.
the class InputRowSchemasTest method test_createColumnsFilter_normal.
@Test
public void test_createColumnsFilter_normal() {
final ColumnsFilter columnsFilter = InputRowSchemas.createColumnsFilter(new TimestampSpec("ts", "auto", null), new DimensionsSpec(ImmutableList.of(StringDimensionSchema.create("foo"))), new TransformSpec(new SelectorDimFilter("bar", "x", null), ImmutableList.of(new ExpressionTransform("baz", "qux + 3", ExprMacroTable.nil()))), new AggregatorFactory[] { new LongSumAggregatorFactory("billy", "bob") });
Assert.assertEquals(ColumnsFilter.inclusionBased(ImmutableSet.of("ts", "foo", "bar", "qux", "bob")), columnsFilter);
}
Aggregations