Google Cloud BigQuery
The BigQuery connector provides Apache Pekko Stream sources and sinks to connect to Google Cloud BigQuery. BigQuery is a serverless data warehouse for storing and analyzing massive datasets. This connector is primarily intended for streaming data into and out of BigQuery tables and running SQL queries, although it also provides basic support for managing datasets and tables and flexible access to the BigQuery REST API.
Project Info: Apache Pekko Connectors Google Cloud BigQuery | |
---|---|
Artifact | org.apache.pekko
pekko-connectors-google-cloud-bigquery
1.1.0
|
JDK versions | OpenJDK 8 OpenJDK 11 OpenJDK 17 OpenJDK 21 |
Scala versions | 2.13.15, 2.12.20, 3.3.4 |
JPMS module name | pekko.stream.connectors.google.cloud.bigquery |
License | |
API documentation | |
Forums | |
Release notes | GitHub releases |
Issues | Github issues |
Sources | https://github.com/apache/pekko-connectors |
Apache Pekko Connectors Google Cloud BigQuery is marked as “API may change”. Please try it out and suggest improvements. PR #2548
Artifacts¶
val PekkoVersion = "1.1.3"
val PekkoHttpVersion = "1.1.0"
libraryDependencies ++= Seq(
"org.apache.pekko" %% "pekko-connectors-google-cloud-bigquery" % "1.1.0",
"org.apache.pekko" %% "pekko-stream" % PekkoVersion,
"org.apache.pekko" %% "pekko-http" % PekkoHttpVersion,
"org.apache.pekko" %% "pekko-http-spray-json" % PekkoHttpVersion
)
<properties>
<pekko.version>1.1.3</pekko.version>
<pekko.http.version>1.1.0</pekko.http.version>
<scala.binary.version>2.13</scala.binary.version>
</properties>
<dependencies>
<dependency>
<groupId>org.apache.pekko</groupId>
<artifactId>pekko-connectors-google-cloud-bigquery_${scala.binary.version}</artifactId>
<version>1.1.0</version>
</dependency>
<dependency>
<groupId>org.apache.pekko</groupId>
<artifactId>pekko-stream_${scala.binary.version}</artifactId>
<version>${pekko.version}</version>
</dependency>
<dependency>
<groupId>org.apache.pekko</groupId>
<artifactId>pekko-http_${scala.binary.version}</artifactId>
<version>${pekko.http.version}</version>
</dependency>
<dependency>
<groupId>org.apache.pekko</groupId>
<artifactId>pekko-http-spray-json_${scala.binary.version}</artifactId>
<version>${pekko.http.version}</version>
</dependency>
</dependencies>
def versions = [
PekkoVersion: "1.1.3",
PekkoHttpVersion: "1.1.0",
ScalaBinary: "2.13"
]
dependencies {
implementation "org.apache.pekko:pekko-connectors-google-cloud-bigquery_${versions.ScalaBinary}:1.1.0"
implementation "org.apache.pekko:pekko-stream_${versions.ScalaBinary}:${versions.PekkoVersion}"
implementation "org.apache.pekko:pekko-http_${versions.ScalaBinary}:${versions.PekkoHttpVersion}"
implementation "org.apache.pekko:pekko-http-spray-json_${versions.ScalaBinary}:${versions.PekkoHttpVersion}"
}
To use the Jackson JSON library for marshalling you must also add the Apache Pekko HTTP module for Jackson support.
val PekkoHttpVersion = "1.1.0"
libraryDependencies += "org.apache.pekko" %% "pekko-http-jackson" % PekkoHttpVersion
<properties>
<pekko.http.version>1.1.0</pekko.http.version>
<scala.binary.version>2.13</scala.binary.version>
</properties>
<dependencies>
<dependency>
<groupId>org.apache.pekko</groupId>
<artifactId>pekko-http-jackson_${scala.binary.version}</artifactId>
<version>${pekko.http.version}</version>
</dependency>
</dependencies>
def versions = [
PekkoHttpVersion: "1.1.0",
ScalaBinary: "2.13"
]
dependencies {
implementation "org.apache.pekko:pekko-http-jackson_${versions.ScalaBinary}:${versions.PekkoHttpVersion}"
}
The table below shows direct dependencies of this module and the second tab shows all libraries that it depends on transitively.
Configuration¶
The BigQuery connector shares its basic configuration with all the Google connectors in Apache Pekko Connectors. Additional BigQuery-specific configuration settings can be found in its reference.conf.
Imports¶
All of the examples below assume the following imports are in scope.
sourceimport org.apache.pekko
import pekko.http.scaladsl.marshallers.sprayjson.SprayJsonSupport._
import pekko.stream.connectors.google.{ GoogleAttributes, GoogleSettings }
import pekko.stream.connectors.googlecloud.bigquery.InsertAllRetryPolicy
import pekko.stream.connectors.googlecloud.bigquery.model.{
Dataset,
Job,
JobReference,
JobState,
QueryResponse,
Table,
TableDataListResponse,
TableListResponse
}
import pekko.stream.connectors.googlecloud.bigquery.scaladsl.schema.BigQuerySchemas._
import pekko.stream.connectors.googlecloud.bigquery.scaladsl.schema.TableSchemaWriter
import pekko.stream.connectors.googlecloud.bigquery.scaladsl.spray.BigQueryRootJsonFormat
import pekko.stream.connectors.googlecloud.bigquery.scaladsl.spray.BigQueryJsonProtocol._
import pekko.stream.connectors.googlecloud.bigquery.scaladsl.BigQuery
import pekko.stream.scaladsl.{ Flow, Sink, Source }
import pekko.{ Done, NotUsed }
import scala.annotation.nowarn
import scala.collection.immutable.Seq
import scala.concurrent.Future
source
import org.apache.pekko.Done;
import org.apache.pekko.NotUsed;
import org.apache.pekko.http.javadsl.marshallers.jackson.Jackson;
import org.apache.pekko.http.javadsl.marshalling.Marshaller;
import org.apache.pekko.http.javadsl.model.HttpEntity;
import org.apache.pekko.http.javadsl.model.RequestEntity;
import org.apache.pekko.http.javadsl.unmarshalling.Unmarshaller;
import org.apache.pekko.stream.connectors.google.GoogleAttributes;
import org.apache.pekko.stream.connectors.google.GoogleSettings;
import org.apache.pekko.stream.connectors.googlecloud.bigquery.InsertAllRetryPolicy;
import org.apache.pekko.stream.connectors.googlecloud.bigquery.javadsl.BigQuery;
import org.apache.pekko.stream.connectors.googlecloud.bigquery.javadsl.jackson.BigQueryMarshallers;
import org.apache.pekko.stream.connectors.googlecloud.bigquery.model.Dataset;
import org.apache.pekko.stream.connectors.googlecloud.bigquery.model.Job;
import org.apache.pekko.stream.connectors.googlecloud.bigquery.model.JobReference;
import org.apache.pekko.stream.connectors.googlecloud.bigquery.model.JobState;
import org.apache.pekko.stream.connectors.googlecloud.bigquery.model.QueryResponse;
import org.apache.pekko.stream.connectors.googlecloud.bigquery.model.Table;
import org.apache.pekko.stream.connectors.googlecloud.bigquery.model.TableDataInsertAllRequest;
import org.apache.pekko.stream.connectors.googlecloud.bigquery.model.TableDataListResponse;
import org.apache.pekko.stream.connectors.googlecloud.bigquery.model.TableFieldSchema;
import org.apache.pekko.stream.connectors.googlecloud.bigquery.model.TableFieldSchemaMode;
import org.apache.pekko.stream.connectors.googlecloud.bigquery.model.TableFieldSchemaType;
import org.apache.pekko.stream.connectors.googlecloud.bigquery.model.TableListResponse;
import org.apache.pekko.stream.connectors.googlecloud.bigquery.model.TableSchema;
import org.apache.pekko.stream.javadsl.Flow;
import org.apache.pekko.stream.javadsl.Sink;
import org.apache.pekko.stream.javadsl.Source;
import com.fasterxml.jackson.annotation.JsonCreator;
import com.fasterxml.jackson.annotation.JsonProperty;
import com.fasterxml.jackson.databind.JsonNode;
import com.fasterxml.jackson.databind.ObjectMapper;
import com.fasterxml.jackson.databind.ObjectReader;
import java.io.IOException;
import java.util.ArrayList;
import java.util.Collections;
import java.util.List;
import java.util.Optional;
import java.util.OptionalInt;
import java.util.OptionalLong;
import java.util.concurrent.CompletableFuture;
import java.util.concurrent.CompletionStage;
import java.util.function.Function;
import java.util.stream.Collectors;
Setup data classes¶
As a working example throughout this documentation, we will use the Person
class to model the data in our BigQuery tables.
sourcecase class Person(name: String, age: Int, addresses: Seq[Address], isHakker: Boolean)
case class Address(street: String, city: String, postalCode: Option[Int])
implicit val addressFormat: BigQueryRootJsonFormat[Address] = bigQueryJsonFormat3(Address.apply)
implicit val personFormat: BigQueryRootJsonFormat[Person] = bigQueryJsonFormat4(Person.apply)
sourceObjectMapper objectMapper = new ObjectMapper();
public class Person {
private String name;
private Integer age;
private List<Address> addresses;
private Boolean isHakker;
@JsonCreator
public Person(@JsonProperty("f") JsonNode fields) throws IOException {
name = fields.get(0).get("v").textValue();
age = Integer.parseInt(fields.get(1).get("v").textValue());
addresses = new ArrayList<>();
ObjectReader addressReader = objectMapper.readerFor(Address.class);
for (JsonNode node : fields.get(2).get("v")) {
Address address = addressReader.readValue(node.get("v"));
addresses.add(address);
}
isHakker = Boolean.parseBoolean(fields.get(3).get("v").textValue());
}
public String getName() {
return name;
}
public Integer getAge() {
return age;
}
public List<Address> getAddresses() {
return addresses;
}
public Boolean getIsHakker() {
return isHakker;
}
}
public class Address {
private String street;
private String city;
private Integer postalCode;
@JsonCreator
public Address(@JsonProperty("f") JsonNode fields) {
street = fields.get(0).get("v").textValue();
city = fields.get(1).get("v").textValue();
postalCode =
Optional.of(fields.get(2).get("v").textValue()).map(Integer::parseInt).orElse(null);
}
public String getStreet() {
return street;
}
public String getCity() {
return city;
}
public Integer getPostalCode() {
return postalCode;
}
}
public class NameAddressesPair {
private String name;
private List<Address> addresses;
@JsonCreator
public NameAddressesPair(@JsonProperty("f") JsonNode fields) throws IOException {
name = fields.get(0).get("v").textValue();
addresses = new ArrayList<>();
ObjectReader addressReader = objectMapper.readerFor(Address.class);
for (JsonNode node : fields.get(1).get("v")) {
Address address = addressReader.readValue(node.get("v"));
addresses.add(address);
}
}
}
To enable support for (un)marshalling User
and Address
as BigQuery table rows and query results we use Jackson’s @JsonCreator
and @JsonProperty
annotations. Note that a custom @JsonCreator
constructor is necessary due to BigQuery’s unusual encoding of rows as “a series of JSON f,v objects for indicating fields and values” (reference documentation). In addition, we also define NameAddressesPair
to model the result of the query in the next section.
Run a query¶
You can run a SQL query and stream the unmarshalled results with the BigQuery.<Out>query
method. To create the unmarshaller, use the BigQueryMarshallers.<Out>queryResponseUnmarshaller
method.
sourceval sqlQuery = s"SELECT name, addresses FROM $datasetId.$tableId WHERE age >= 100"
val centenarians: Source[(String, Seq[Address]), Future[QueryResponse[(String, Seq[Address])]]] =
BigQuery.query[(String, Seq[Address])](sqlQuery, useLegacySql = false)
sourceString sqlQuery =
String.format("SELECT name, addresses FROM %s.%s WHERE age >= 100", datasetId, tableId);
Unmarshaller<HttpEntity, QueryResponse<NameAddressesPair>> queryResponseUnmarshaller =
BigQueryMarshallers.queryResponseUnmarshaller(NameAddressesPair.class);
Source<NameAddressesPair, CompletionStage<QueryResponse<NameAddressesPair>>> centenarians =
BigQuery.query(sqlQuery, false, false, queryResponseUnmarshaller);
Notice that the source materializes a CompletionStage<QueryResponse<NameAddressesTuple>>
which can be used to retrieve metadata related to the query. For example, you can use a dry run to estimate the number of bytes that will be read by a query.
sourceval centenariansDryRun = BigQuery.query[(String, Seq[Address])](sqlQuery, dryRun = true, useLegacySql = false)
val bytesProcessed: Future[Long] = centenariansDryRun.to(Sink.ignore).run().map(_.totalBytesProcessed.get)
sourceSource<NameAddressesPair, CompletionStage<QueryResponse<NameAddressesPair>>>
centenariansDryRun = BigQuery.query(sqlQuery, false, false, queryResponseUnmarshaller);
CompletionStage<Long> bytesProcessed =
centenariansDryRun
.to(Sink.ignore())
.run(system)
.thenApply(r -> r.getTotalBytesProcessed().getAsLong());
Finally, you can also stream all of the rows in a table without the expense of running a query with the BigQuery.<Out>listTableData
method.
sourceval everyone: Source[Person, Future[TableDataListResponse[Person]]] =
BigQuery.tableData[Person](datasetId, tableId)
sourceUnmarshaller<HttpEntity, TableDataListResponse<Person>> tableDataListUnmarshaller =
BigQueryMarshallers.tableDataListResponseUnmarshaller(Person.class);
Source<Person, CompletionStage<TableDataListResponse<Person>>> everyone =
BigQuery.listTableData(
datasetId,
tableId,
OptionalLong.empty(),
OptionalInt.empty(),
Collections.emptyList(),
tableDataListUnmarshaller);
Load data into BigQuery¶
The BigQuery connector enables loading data into tables via real-time streaming inserts or batch loading. For an overview of these strategies see the BigQuery documentation.
The BigQuery.<In>insertAll
method creates a sink that accepts batches of List<In>
(for example created via the batch
operator) and streams them directly into a table. To enable/disable BigQuery’s best-effort deduplication feature use the appropriate InsertAllRetryPolicy
.
sourceval peopleInsertSink: Sink[Seq[Person], NotUsed] =
BigQuery.insertAll[Person](datasetId, tableId, InsertAllRetryPolicy.WithDeduplication)
sourceMarshaller<TableDataInsertAllRequest<Person>, RequestEntity> tableDataInsertAllMarshaller =
BigQueryMarshallers.tableDataInsertAllRequestMarshaller();
Sink<List<Person>, NotUsed> peopleInsertSink =
BigQuery.insertAll(
datasetId,
tableId,
InsertAllRetryPolicy.withDeduplication(),
Optional.empty(),
tableDataInsertAllMarshaller);
As a cost-saving alternative to streaming inserts, you can also add data to a table via asynchronous load jobs. The BigQuery.<In>insertAllAsync
method creates a flow that starts a series of batch load jobs. By default, a new load job is created every minute to attempt to emulate near-real-time streaming inserts, although there is no guarantee when the job will actually run. The frequency with which new load jobs are created is controlled by the pekko.connectors.google.bigquery.load-job-per-table-quota
configuration setting.
Pending the resolution of Google BigQuery issue 176002651, the BigQuery.insertAllAsync
API may not work as expected.
As a workaround, you can use the config setting pekko.http.parsing.conflicting-content-type-header-processing-mode = first
with Apache Pekko HTTP v1.0.0 or later.
sourceval peopleLoadFlow: Flow[Person, Job, NotUsed] = BigQuery.insertAllAsync[Person](datasetId, tableId)
sourceFlow<Person, Job, NotUsed> peopleLoadFlow =
BigQuery.insertAllAsync(datasetId, tableId, Jackson.marshaller());
To check the status of the load jobs use the BigQuery.getJob
method.
sourcedef checkIfJobsDone(jobReferences: Seq[JobReference]): Future[Boolean] = {
for {
jobs <- Future.sequence(jobReferences.map(ref => BigQuery.job(ref.jobId.get)))
} yield jobs.forall(job => job.status.exists(_.state == JobState.Done))
}
val isDone: Future[Boolean] = for {
jobs <- Source(people).via(peopleLoadFlow).runWith(Sink.seq)
jobReferences = jobs.flatMap(job => job.jobReference)
isDone <- checkIfJobsDone(jobReferences)
} yield isDone
sourceFunction<List<JobReference>, CompletionStage<Boolean>> checkIfJobsDone =
jobReferences -> {
GoogleSettings settings = GoogleSettings.create(system);
CompletionStage<Boolean> allAreDone = CompletableFuture.completedFuture(true);
for (JobReference jobReference : jobReferences) {
CompletionStage<Job> job =
BigQuery.getJob(jobReference.getJobId().get(), Optional.empty(), settings, system);
CompletionStage<Boolean> jobIsDone =
job.thenApply(
j ->
j.getStatus().map(s -> s.getState().equals(JobState.done())).orElse(false));
allAreDone = allAreDone.thenCombine(jobIsDone, (a, b) -> a & b);
}
return allAreDone;
};
CompletionStage<List<Job>> jobs =
Source.from(people).via(peopleLoadFlow).runWith(Sink.<Job>seq(), system);
CompletionStage<List<JobReference>> jobReferences =
jobs.thenApply(
js -> js.stream().map(j -> j.getJobReference().get()).collect(Collectors.toList()));
CompletionStage<Boolean> isDone = jobReferences.thenCompose(checkIfJobsDone);
Managing datasets and tables¶
The BigQuery connector provides methods for basic management of datasets and tables.
sourceval allDatasets: Source[Dataset, NotUsed] = BigQuery.datasets
val existingDataset: Future[Dataset] = BigQuery.dataset(datasetId)
val newDataset: Future[Dataset] = BigQuery.createDataset("newDatasetId")
val datasetDeleted: Future[Done] = BigQuery.deleteDataset(datasetId)
val allTablesInDataset: Source[Table, Future[TableListResponse]] = BigQuery.tables(datasetId)
val existingTable: Future[Table] = BigQuery.table(datasetId, tableId)
val tableDeleted: Future[Done] = BigQuery.deleteTable(datasetId, tableId)
sourceGoogleSettings settings = GoogleSettings.create(system);
Source<Dataset, NotUsed> allDatasets =
BigQuery.listDatasets(OptionalInt.empty(), Optional.empty(), Collections.emptyMap());
CompletionStage<Dataset> existingDataset = BigQuery.getDataset(datasetId, settings, system);
CompletionStage<Dataset> newDataset = BigQuery.createDataset("newDatasetId", settings, system);
CompletionStage<Done> datasetDeleted =
BigQuery.deleteDataset(datasetId, false, settings, system);
Source<Table, CompletionStage<TableListResponse>> allTablesInDataset =
BigQuery.listTables(datasetId, OptionalInt.empty());
CompletionStage<Table> existingTable = BigQuery.getTable(datasetId, tableId, settings, system);
CompletionStage<Done> tableDeleted = BigQuery.deleteTable(datasetId, tableId, settings, system);
Creating a table requires a little more work to specify the schema.
sourceimplicit val addressSchema: TableSchemaWriter[Address] = bigQuerySchema3(Address.apply)
implicit val personSchema: TableSchemaWriter[Person] = bigQuerySchema4(Person.apply)
val newTable: Future[Table] = BigQuery.createTable[Person](datasetId, "newTableId")
sourceTableSchema personSchema =
TableSchema.create(
TableFieldSchema.create("name", TableFieldSchemaType.string(), Optional.empty()),
TableFieldSchema.create("age", TableFieldSchemaType.integer(), Optional.empty()),
TableFieldSchema.create(
"addresses",
TableFieldSchemaType.record(),
Optional.of(TableFieldSchemaMode.repeated()),
TableFieldSchema.create("street", TableFieldSchemaType.string(), Optional.empty()),
TableFieldSchema.create("city", TableFieldSchemaType.string(), Optional.empty()),
TableFieldSchema.create(
"postalCode",
TableFieldSchemaType.integer(),
Optional.of(TableFieldSchemaMode.nullable()))),
TableFieldSchema.create("isHakker", TableFieldSchemaType.bool(), Optional.empty()));
CompletionStage<Table> newTable =
BigQuery.createTable(datasetId, "newTableId", personSchema, settings, system);
Apply custom settings to a part of the stream¶
In certain situations it may be desirable to modify the GoogleSettings
applied to a part of the stream, for example to change the project ID or use different RetrySettings
.
sourceval defaultSettings: GoogleSettings = GoogleSettings()
val customSettings = defaultSettings.copy(projectId = "myOtherProject")
BigQuery.query[(String, Seq[Address])](sqlQuery).withAttributes(GoogleAttributes.settings(customSettings))
sourceGoogleSettings defaultSettings = GoogleSettings.create(system);
GoogleSettings customSettings = defaultSettings.withProjectId("myOtherProjectId");
BigQuery.query(sqlQuery, false, false, queryResponseUnmarshaller)
.withAttributes(GoogleAttributes.settings(customSettings));
Make raw API requests¶
If you would like to interact with the BigQuery REST API beyond what the BigQuery connector supports, you can make authenticated raw requests via the BigQuery.singleRequest
and BigQuery.<Out>paginatedRequest
methods.