Search in sources :

Example 1 with StorageClient

use of org.apache.beam.sdk.io.gcp.bigquery.BigQueryServices.StorageClient in project beam by apache.

the class BigQueryStorageSourceBase method split.

@Override
public List<BigQueryStorageStreamSource<T>> split(long desiredBundleSizeBytes, PipelineOptions options) throws Exception {
    BigQueryOptions bqOptions = options.as(BigQueryOptions.class);
    Table targetTable = getTargetTable(bqOptions);
    ReadSession.Builder readSessionBuilder = ReadSession.newBuilder();
    if (targetTable != null) {
        readSessionBuilder.setTable(BigQueryHelpers.toTableResourceName(targetTable.getTableReference()));
    } else {
        // If the table does not exist targetTable will be null.
        // Construct the table id if we can generate it. For error recording/logging.
        @Nullable String tableReferenceId = getTargetTableId(bqOptions);
        if (tableReferenceId != null) {
            readSessionBuilder.setTable(tableReferenceId);
        }
    }
    if (selectedFieldsProvider != null || rowRestrictionProvider != null) {
        ReadSession.TableReadOptions.Builder tableReadOptionsBuilder = ReadSession.TableReadOptions.newBuilder();
        if (selectedFieldsProvider != null) {
            tableReadOptionsBuilder.addAllSelectedFields(selectedFieldsProvider.get());
        }
        if (rowRestrictionProvider != null) {
            tableReadOptionsBuilder.setRowRestriction(rowRestrictionProvider.get());
        }
        readSessionBuilder.setReadOptions(tableReadOptionsBuilder);
    }
    if (format != null) {
        readSessionBuilder.setDataFormat(format);
    }
    int streamCount = 0;
    if (desiredBundleSizeBytes > 0) {
        long tableSizeBytes = (targetTable != null) ? targetTable.getNumBytes() : 0;
        streamCount = (int) Math.min(tableSizeBytes / desiredBundleSizeBytes, MAX_SPLIT_COUNT);
    }
    streamCount = Math.max(streamCount, MIN_SPLIT_COUNT);
    CreateReadSessionRequest createReadSessionRequest = CreateReadSessionRequest.newBuilder().setParent(BigQueryHelpers.toProjectResourceName(bqOptions.getBigQueryProject() == null ? bqOptions.getProject() : bqOptions.getBigQueryProject())).setReadSession(readSessionBuilder).setMaxStreamCount(streamCount).build();
    ReadSession readSession;
    try (StorageClient client = bqServices.getStorageClient(bqOptions)) {
        readSession = client.createReadSession(createReadSessionRequest);
        LOG.info("Sent BigQuery Storage API CreateReadSession request '{}'; received response '{}'.", createReadSessionRequest, readSession);
    }
    if (readSession.getStreamsList().isEmpty()) {
        // The underlying table is empty or all rows have been pruned.
        return ImmutableList.of();
    }
    Schema sessionSchema;
    if (readSession.getDataFormat() == DataFormat.ARROW) {
        org.apache.arrow.vector.types.pojo.Schema schema = ArrowConversion.arrowSchemaFromInput(readSession.getArrowSchema().getSerializedSchema().newInput());
        org.apache.beam.sdk.schemas.Schema beamSchema = ArrowConversion.ArrowSchemaTranslator.toBeamSchema(schema);
        sessionSchema = AvroUtils.toAvroSchema(beamSchema);
    } else if (readSession.getDataFormat() == DataFormat.AVRO) {
        sessionSchema = new Schema.Parser().parse(readSession.getAvroSchema().getSchema());
    } else {
        throw new IllegalArgumentException("data is not in a supported dataFormat: " + readSession.getDataFormat());
    }
    TableSchema trimmedSchema = BigQueryAvroUtils.trimBigQueryTableSchema(targetTable.getSchema(), sessionSchema);
    List<BigQueryStorageStreamSource<T>> sources = Lists.newArrayList();
    for (ReadStream readStream : readSession.getStreamsList()) {
        sources.add(BigQueryStorageStreamSource.create(readSession, readStream, trimmedSchema, parseFn, outputCoder, bqServices));
    }
    return ImmutableList.copyOf(sources);
}
Also used : TableSchema(com.google.api.services.bigquery.model.TableSchema) Schema(org.apache.avro.Schema) TableSchema(com.google.api.services.bigquery.model.TableSchema) ReadStream(com.google.cloud.bigquery.storage.v1.ReadStream) Table(com.google.api.services.bigquery.model.Table) ReadSession(com.google.cloud.bigquery.storage.v1.ReadSession) StorageClient(org.apache.beam.sdk.io.gcp.bigquery.BigQueryServices.StorageClient) Nullable(org.checkerframework.checker.nullness.qual.Nullable) CreateReadSessionRequest(com.google.cloud.bigquery.storage.v1.CreateReadSessionRequest)

Example 2 with StorageClient

use of org.apache.beam.sdk.io.gcp.bigquery.BigQueryServices.StorageClient in project beam by apache.

the class BigQueryIOStorageReadTest method testReadFromBigQueryIO.

@Test
public void testReadFromBigQueryIO() throws Exception {
    fakeDatasetService.createDataset("foo.com:project", "dataset", "", "", null);
    TableReference tableRef = BigQueryHelpers.parseTableSpec("foo.com:project:dataset.table");
    Table table = new Table().setTableReference(tableRef).setNumBytes(10L).setSchema(TABLE_SCHEMA);
    fakeDatasetService.createTable(table);
    CreateReadSessionRequest expectedCreateReadSessionRequest = CreateReadSessionRequest.newBuilder().setParent("projects/project-id").setReadSession(ReadSession.newBuilder().setTable("projects/foo.com:project/datasets/dataset/tables/table").setDataFormat(DataFormat.AVRO)).setMaxStreamCount(10).build();
    ReadSession readSession = ReadSession.newBuilder().setName("readSessionName").setAvroSchema(AvroSchema.newBuilder().setSchema(AVRO_SCHEMA_STRING)).addStreams(ReadStream.newBuilder().setName("streamName")).setDataFormat(DataFormat.AVRO).build();
    ReadRowsRequest expectedReadRowsRequest = ReadRowsRequest.newBuilder().setReadStream("streamName").build();
    List<GenericRecord> records = Lists.newArrayList(createRecord("A", 1, AVRO_SCHEMA), createRecord("B", 2, AVRO_SCHEMA), createRecord("C", 3, AVRO_SCHEMA), createRecord("D", 4, AVRO_SCHEMA));
    List<ReadRowsResponse> readRowsResponses = Lists.newArrayList(createResponse(AVRO_SCHEMA, records.subList(0, 2), 0.0, 0.50), createResponse(AVRO_SCHEMA, records.subList(2, 4), 0.5, 0.75));
    StorageClient fakeStorageClient = mock(StorageClient.class, withSettings().serializable());
    when(fakeStorageClient.createReadSession(expectedCreateReadSessionRequest)).thenReturn(readSession);
    when(fakeStorageClient.readRows(expectedReadRowsRequest, "")).thenReturn(new FakeBigQueryServerStream<>(readRowsResponses));
    PCollection<KV<String, Long>> output = p.apply(BigQueryIO.read(new ParseKeyValue()).from("foo.com:project:dataset.table").withMethod(Method.DIRECT_READ).withFormat(DataFormat.AVRO).withTestServices(new FakeBigQueryServices().withDatasetService(fakeDatasetService).withStorageClient(fakeStorageClient)));
    PAssert.that(output).containsInAnyOrder(ImmutableList.of(KV.of("A", 1L), KV.of("B", 2L), KV.of("C", 3L), KV.of("D", 4L)));
    p.run();
}
Also used : Table(com.google.api.services.bigquery.model.Table) ReadSession(com.google.cloud.bigquery.storage.v1.ReadSession) ReadRowsRequest(com.google.cloud.bigquery.storage.v1.ReadRowsRequest) StorageClient(org.apache.beam.sdk.io.gcp.bigquery.BigQueryServices.StorageClient) KV(org.apache.beam.sdk.values.KV) TableReference(com.google.api.services.bigquery.model.TableReference) ReadRowsResponse(com.google.cloud.bigquery.storage.v1.ReadRowsResponse) FakeBigQueryServices(org.apache.beam.sdk.io.gcp.testing.FakeBigQueryServices) GenericRecord(org.apache.avro.generic.GenericRecord) CreateReadSessionRequest(com.google.cloud.bigquery.storage.v1.CreateReadSessionRequest) Test(org.junit.Test)

Example 3 with StorageClient

use of org.apache.beam.sdk.io.gcp.bigquery.BigQueryServices.StorageClient in project beam by apache.

the class BigQueryIOStorageReadTest method testFractionConsumedArrow.

@Test
public void testFractionConsumedArrow() throws Exception {
    ReadSession readSession = ReadSession.newBuilder().setName("readSession").setArrowSchema(ArrowSchema.newBuilder().setSerializedSchema(serializeArrowSchema(ARROW_SCHEMA)).build()).setDataFormat(DataFormat.ARROW).build();
    ReadRowsRequest expectedRequest = ReadRowsRequest.newBuilder().setReadStream("readStream").build();
    List<String> names = Arrays.asList("A", "B", "C", "D", "E", "F", "G");
    List<Long> values = Arrays.asList(1L, 2L, 3L, 4L, 5L, 6L, 7L);
    List<ReadRowsResponse> responses = Lists.newArrayList(createResponseArrow(ARROW_SCHEMA, names.subList(0, 2), values.subList(0, 2), 0.0, 0.25), createResponseArrow(ARROW_SCHEMA, Lists.newArrayList(), Lists.newArrayList(), 0.25, 0.25), createResponseArrow(ARROW_SCHEMA, names.subList(2, 4), values.subList(2, 4), 0.3, 0.5), createResponseArrow(ARROW_SCHEMA, names.subList(4, 7), values.subList(4, 7), 0.7, 1.0));
    StorageClient fakeStorageClient = mock(StorageClient.class);
    when(fakeStorageClient.readRows(expectedRequest, "")).thenReturn(new FakeBigQueryServerStream<>(responses));
    BigQueryStorageStreamSource<TableRow> streamSource = BigQueryStorageStreamSource.create(readSession, ReadStream.newBuilder().setName("readStream").build(), TABLE_SCHEMA, new TableRowParser(), TableRowJsonCoder.of(), new FakeBigQueryServices().withStorageClient(fakeStorageClient));
    BoundedReader<TableRow> reader = streamSource.createReader(options);
    // Before call to BoundedReader#start, fraction consumed must be zero.
    assertEquals(0.0, reader.getFractionConsumed(), DELTA);
    // Reads A.
    assertTrue(reader.start());
    assertEquals(0.125, reader.getFractionConsumed(), DELTA);
    // Reads B.
    assertTrue(reader.advance());
    assertEquals(0.25, reader.getFractionConsumed(), DELTA);
    // Reads C.
    assertTrue(reader.advance());
    assertEquals(0.4, reader.getFractionConsumed(), DELTA);
    // Reads D.
    assertTrue(reader.advance());
    assertEquals(0.5, reader.getFractionConsumed(), DELTA);
    // Reads E.
    assertTrue(reader.advance());
    assertEquals(0.8, reader.getFractionConsumed(), DELTA);
    // Reads F.
    assertTrue(reader.advance());
    assertEquals(0.9, reader.getFractionConsumed(), DELTA);
    // Reads G.
    assertTrue(reader.advance());
    assertEquals(1.0, reader.getFractionConsumed(), DELTA);
    // Reaches the end.
    assertFalse(reader.advance());
    // We are done with the stream, so we should report 100% consumption.
    assertEquals(Double.valueOf(1.0), reader.getFractionConsumed());
}
Also used : ReadSession(com.google.cloud.bigquery.storage.v1.ReadSession) ReadRowsRequest(com.google.cloud.bigquery.storage.v1.ReadRowsRequest) StorageClient(org.apache.beam.sdk.io.gcp.bigquery.BigQueryServices.StorageClient) ByteString(com.google.protobuf.ByteString) TableRowParser(org.apache.beam.sdk.io.gcp.bigquery.BigQueryIO.TableRowParser) ReadRowsResponse(com.google.cloud.bigquery.storage.v1.ReadRowsResponse) TableRow(com.google.api.services.bigquery.model.TableRow) FakeBigQueryServices(org.apache.beam.sdk.io.gcp.testing.FakeBigQueryServices) Test(org.junit.Test)

Example 4 with StorageClient

use of org.apache.beam.sdk.io.gcp.bigquery.BigQueryServices.StorageClient in project beam by apache.

the class BigQueryIOStorageReadTest method testStreamSourceSplitAtFractionFailsWhenSplitIsNotPossible.

@Test
public void testStreamSourceSplitAtFractionFailsWhenSplitIsNotPossible() throws Exception {
    List<ReadRowsResponse> parentResponses = Lists.newArrayList(createResponse(AVRO_SCHEMA, Lists.newArrayList(createRecord("A", 1, AVRO_SCHEMA), createRecord("B", 2, AVRO_SCHEMA)), 0.0, 0.25), createResponse(AVRO_SCHEMA, Lists.newArrayList(createRecord("C", 3, AVRO_SCHEMA)), 0.25, 0.50), createResponse(AVRO_SCHEMA, Lists.newArrayList(createRecord("D", 4, AVRO_SCHEMA), createRecord("E", 5, AVRO_SCHEMA)), 0.5, 0.75));
    StorageClient fakeStorageClient = mock(StorageClient.class);
    when(fakeStorageClient.readRows(ReadRowsRequest.newBuilder().setReadStream("parentStream").build(), "")).thenReturn(new FakeBigQueryServerStream<>(parentResponses));
    // Mocks the split call. A response without a primary_stream and remainder_stream means
    // that the split is not possible.
    when(fakeStorageClient.splitReadStream(SplitReadStreamRequest.newBuilder().setName("parentStream").setFraction(0.5f).build())).thenReturn(SplitReadStreamResponse.getDefaultInstance());
    BigQueryStorageStreamSource<TableRow> streamSource = BigQueryStorageStreamSource.create(ReadSession.newBuilder().setName("readSession").setAvroSchema(AvroSchema.newBuilder().setSchema(AVRO_SCHEMA_STRING)).build(), ReadStream.newBuilder().setName("parentStream").build(), TABLE_SCHEMA, new TableRowParser(), TableRowJsonCoder.of(), new FakeBigQueryServices().withStorageClient(fakeStorageClient));
    // Read a few records from the parent stream and ensure that records are returned in the
    // prescribed order.
    BoundedReader<TableRow> parent = streamSource.createReader(options);
    assertTrue(parent.start());
    assertEquals("A", parent.getCurrent().get("name"));
    assertTrue(parent.advance());
    assertEquals("B", parent.getCurrent().get("name"));
    assertNull(parent.splitAtFraction(0.5));
    verify(fakeStorageClient, times(1)).splitReadStream(ArgumentMatchers.any());
    // Verify that subsequent splitAtFraction() calls after a failed splitAtFraction() attempt
    // do NOT invoke SplitReadStream.
    assertNull(parent.splitAtFraction(0.5));
    verify(fakeStorageClient, times(1)).splitReadStream(ArgumentMatchers.any());
    // Verify that the parent source still works okay even after an unsuccessful split attempt.
    assertTrue(parent.advance());
    assertEquals("C", parent.getCurrent().get("name"));
    assertTrue(parent.advance());
    assertEquals("D", parent.getCurrent().get("name"));
    assertTrue(parent.advance());
    assertEquals("E", parent.getCurrent().get("name"));
    assertFalse(parent.advance());
}
Also used : ReadRowsResponse(com.google.cloud.bigquery.storage.v1.ReadRowsResponse) TableRow(com.google.api.services.bigquery.model.TableRow) StorageClient(org.apache.beam.sdk.io.gcp.bigquery.BigQueryServices.StorageClient) FakeBigQueryServices(org.apache.beam.sdk.io.gcp.testing.FakeBigQueryServices) TableRowParser(org.apache.beam.sdk.io.gcp.bigquery.BigQueryIO.TableRowParser) Test(org.junit.Test)

Example 5 with StorageClient

use of org.apache.beam.sdk.io.gcp.bigquery.BigQueryServices.StorageClient in project beam by apache.

the class BigQueryIOStorageReadTest method testTableSourceInitialSplit_WithDefaultProject.

@Test
public void testTableSourceInitialSplit_WithDefaultProject() throws Exception {
    fakeDatasetService.createDataset("project-id", "dataset", "", "", null);
    TableReference tableRef = BigQueryHelpers.parseTableSpec("project-id:dataset.table");
    Table table = new Table().setTableReference(tableRef).setNumBytes(1024L * 1024L).setSchema(TABLE_SCHEMA);
    fakeDatasetService.createTable(table);
    CreateReadSessionRequest expectedRequest = CreateReadSessionRequest.newBuilder().setParent("projects/project-id").setReadSession(ReadSession.newBuilder().setTable("projects/project-id/datasets/dataset/tables/table")).setMaxStreamCount(1024).build();
    ReadSession.Builder builder = ReadSession.newBuilder().setAvroSchema(AvroSchema.newBuilder().setSchema(AVRO_SCHEMA_STRING)).setDataFormat(DataFormat.AVRO);
    for (int i = 0; i < 50; i++) {
        builder.addStreams(ReadStream.newBuilder().setName("stream-" + i));
    }
    StorageClient fakeStorageClient = mock(StorageClient.class);
    when(fakeStorageClient.createReadSession(expectedRequest)).thenReturn(builder.build());
    BigQueryStorageTableSource<TableRow> tableSource = BigQueryStorageTableSource.create(ValueProvider.StaticValueProvider.of(BigQueryHelpers.parseTableSpec("dataset.table")), null, null, new TableRowParser(), TableRowJsonCoder.of(), new FakeBigQueryServices().withDatasetService(fakeDatasetService).withStorageClient(fakeStorageClient));
    List<? extends BoundedSource<TableRow>> sources = tableSource.split(1024L, options);
    assertEquals(50L, sources.size());
}
Also used : TableReference(com.google.api.services.bigquery.model.TableReference) Table(com.google.api.services.bigquery.model.Table) ReadSession(com.google.cloud.bigquery.storage.v1.ReadSession) TableRow(com.google.api.services.bigquery.model.TableRow) StorageClient(org.apache.beam.sdk.io.gcp.bigquery.BigQueryServices.StorageClient) FakeBigQueryServices(org.apache.beam.sdk.io.gcp.testing.FakeBigQueryServices) CreateReadSessionRequest(com.google.cloud.bigquery.storage.v1.CreateReadSessionRequest) TableRowParser(org.apache.beam.sdk.io.gcp.bigquery.BigQueryIO.TableRowParser) Test(org.junit.Test)

Aggregations

StorageClient (org.apache.beam.sdk.io.gcp.bigquery.BigQueryServices.StorageClient)27 FakeBigQueryServices (org.apache.beam.sdk.io.gcp.testing.FakeBigQueryServices)26 TableRow (com.google.api.services.bigquery.model.TableRow)23 Test (org.junit.Test)23 TableRowParser (org.apache.beam.sdk.io.gcp.bigquery.BigQueryIO.TableRowParser)22 ReadSession (com.google.cloud.bigquery.storage.v1.ReadSession)19 ReadRowsResponse (com.google.cloud.bigquery.storage.v1.ReadRowsResponse)17 Table (com.google.api.services.bigquery.model.Table)13 ByteString (com.google.protobuf.ByteString)13 TableReference (com.google.api.services.bigquery.model.TableReference)12 CreateReadSessionRequest (com.google.cloud.bigquery.storage.v1.CreateReadSessionRequest)12 ReadRowsRequest (com.google.cloud.bigquery.storage.v1.ReadRowsRequest)10 GenericRecord (org.apache.avro.generic.GenericRecord)6 JobStatistics (com.google.api.services.bigquery.model.JobStatistics)5 JobStatistics2 (com.google.api.services.bigquery.model.JobStatistics2)5 BigQueryResourceNaming.createTempTableReference (org.apache.beam.sdk.io.gcp.bigquery.BigQueryResourceNaming.createTempTableReference)5 TableSchema (com.google.api.services.bigquery.model.TableSchema)4 ReadStream (com.google.cloud.bigquery.storage.v1.ReadStream)3 Schema (org.apache.avro.Schema)3 KV (org.apache.beam.sdk.values.KV)3