use of org.apache.spark.sql.execution.datasources.InsertIntoHadoopFsRelationCommand in project OpenLineage by OpenLineage.
the class InsertIntoHadoopFsRelationVisitor method apply.
@Override
public List<OpenLineage.OutputDataset> apply(LogicalPlan x) {
InsertIntoHadoopFsRelationCommand command = (InsertIntoHadoopFsRelationCommand) x;
DatasetIdentifier di = PathUtils.fromURI(command.outputPath().toUri(), "file");
OpenLineage.OutputDataset outputDataset;
if (SaveMode.Overwrite == command.mode()) {
outputDataset = outputDataset().getDataset(di, command.query().schema(), OpenLineage.LifecycleStateChangeDatasetFacet.LifecycleStateChange.OVERWRITE);
} else {
outputDataset = outputDataset().getDataset(di, command.query().schema());
}
return Collections.singletonList(outputDataset);
}
use of org.apache.spark.sql.execution.datasources.InsertIntoHadoopFsRelationCommand in project OpenLineage by OpenLineage.
the class OpenLineageSparkListenerTest method testSqlEventWithJobEventEmitsOnce.
@Test
public void testSqlEventWithJobEventEmitsOnce() {
SparkSession sparkSession = mock(SparkSession.class);
SparkContext sparkContext = mock(SparkContext.class);
EventEmitter emitter = mock(EventEmitter.class);
QueryExecution qe = mock(QueryExecution.class);
LogicalPlan query = UnresolvedRelation$.MODULE$.apply(TableIdentifier.apply("tableName"));
SparkPlan plan = mock(SparkPlan.class);
when(sparkSession.sparkContext()).thenReturn(sparkContext);
when(sparkContext.appName()).thenReturn("appName");
when(qe.optimizedPlan()).thenReturn(new InsertIntoHadoopFsRelationCommand(new Path("file:///tmp/dir"), null, false, Seq$.MODULE$.empty(), Option.empty(), null, Map$.MODULE$.empty(), query, SaveMode.Overwrite, Option.empty(), Option.empty(), Seq$.MODULE$.<String>empty()));
when(qe.executedPlan()).thenReturn(plan);
when(plan.sparkContext()).thenReturn(sparkContext);
when(plan.nodeName()).thenReturn("execute");
OpenLineageContext olContext = OpenLineageContext.builder().sparkSession(Optional.of(sparkSession)).sparkContext(sparkSession.sparkContext()).openLineage(new OpenLineage(OpenLineageClient.OPEN_LINEAGE_CLIENT_URI)).queryExecution(qe).build();
olContext.getOutputDatasetQueryPlanVisitors().add(new InsertIntoHadoopFsRelationVisitor(olContext));
ExecutionContext executionContext = new StaticExecutionContextFactory(emitter).createSparkSQLExecutionContext(1L, emitter, qe, olContext);
executionContext.start(new SparkListenerSQLExecutionStart(1L, "", "", "", new SparkPlanInfo("name", "string", Seq$.MODULE$.empty(), Map$.MODULE$.empty(), Seq$.MODULE$.empty()), 1L));
executionContext.start(new SparkListenerJobStart(0, 2L, Seq$.MODULE$.<StageInfo>empty(), new Properties()));
ArgumentCaptor<OpenLineage.RunEvent> lineageEvent = ArgumentCaptor.forClass(OpenLineage.RunEvent.class);
verify(emitter, times(2)).emit(lineageEvent.capture());
}
use of org.apache.spark.sql.execution.datasources.InsertIntoHadoopFsRelationCommand in project OpenLineage by OpenLineage.
the class LogicalPlanSerializerTest method testSerializeInsertIntoHadoopPlan.
@Test
public void testSerializeInsertIntoHadoopPlan() throws IOException, InvocationTargetException, IllegalAccessException {
SparkSession session = SparkSession.builder().master("local").getOrCreate();
HadoopFsRelation hadoopFsRelation = new HadoopFsRelation(new CatalogFileIndex(session, CatalogTableTestUtils.getCatalogTable(new TableIdentifier("test", Option.apply("db"))), 100L), new StructType(new StructField[] { new StructField("name", StringType$.MODULE$, false, Metadata.empty()) }), new StructType(new StructField[] { new StructField("name", StringType$.MODULE$, false, Metadata.empty()) }), Option.empty(), new TextFileFormat(), new HashMap<>(), session);
LogicalRelation logicalRelation = new LogicalRelation(hadoopFsRelation, Seq$.MODULE$.<AttributeReference>newBuilder().$plus$eq(new AttributeReference("name", StringType$.MODULE$, false, Metadata.empty(), ExprId.apply(1L), Seq$.MODULE$.<String>empty())).result(), Option.empty(), false);
InsertIntoHadoopFsRelationCommand command = new InsertIntoHadoopFsRelationCommand(new org.apache.hadoop.fs.Path("/tmp"), new HashMap<>(), false, Seq$.MODULE$.<Attribute>newBuilder().$plus$eq(new AttributeReference("name", StringType$.MODULE$, false, Metadata.empty(), ExprId.apply(1L), Seq$.MODULE$.<String>empty())).result(), Option.empty(), new TextFileFormat(), new HashMap<>(), logicalRelation, SaveMode.Overwrite, Option.empty(), Option.empty(), Seq$.MODULE$.<String>newBuilder().$plus$eq("name").result());
Map<String, Object> commandActualNode = objectMapper.readValue(logicalPlanSerializer.serialize(command), mapTypeReference);
Map<String, Object> hadoopFSActualNode = objectMapper.readValue(logicalPlanSerializer.serialize(logicalRelation), mapTypeReference);
Path expectedCommandNodePath = Paths.get("src", "test", "resources", "test_data", "serde", "insertintofs-node.json");
Path expectedHadoopFSNodePath = Paths.get("src", "test", "resources", "test_data", "serde", "hadoopfsrelation-node.json");
Map<String, Object> expectedCommandNode = objectMapper.readValue(expectedCommandNodePath.toFile(), mapTypeReference);
Map<String, Object> expectedHadoopFSNode = objectMapper.readValue(expectedHadoopFSNodePath.toFile(), mapTypeReference);
assertThat(commandActualNode).satisfies(new MatchesMapRecursively(expectedCommandNode, Collections.singleton("exprId")));
assertThat(hadoopFSActualNode).satisfies(new MatchesMapRecursively(expectedHadoopFSNode, Collections.singleton("exprId")));
}
Aggregations