PostgreSQL R2DBC Driver

This project contains the PostgreSQL implementation of the R2DBC SPI. This implementation is not intended to be used directly, but rather to be used as the backing implementation for a humane client library to delegate to.

This driver provides the following features:

• Implements R2DBC 1.0
• Login with username/password (MD5, SASL/SCRAM) or implicit trust
• SCRAM authentication
• Unix Domain Socket transport
• Connection Fail-over supporting multiple hosts
• TLS
• Explicit transactions
• Notifications
• Logical Decode
• Binary data transfer
• Execution of prepared statements with bindings
• Execution of batch statements without bindings
• Read and write support for a majority of data types (see Data Type Mapping for details)
• Fetching of REFCURSOR using io.r2dbc.postgresql.api.RefCursor
• Extension points to register Codecs to handle additional PostgreSQL data types

Next steps:

• Multi-dimensional arrays

Code of Conduct

This project is governed by the Code of Conduct. By participating, you are expected to uphold this code of conduct. Please report unacceptable behavior to [email protected].

Getting Started

Here is a quick teaser of how to use R2DBC PostgreSQL in Java:

URL Connection Factory Discovery

ConnectionFactory connectionFactory = ConnectionFactories.get("r2dbc:postgresql://<host>:5432/<database>");

Publisher<? extends Connection> connectionPublisher = connectionFactory.create();

Programmatic Connection Factory Discovery

Map<String, String> options = new HashMap<>();
options.put("lock_timeout", "10s");
options.put("statement_timeout", "5m");

ConnectionFactory connectionFactory = ConnectionFactories.get(ConnectionFactoryOptions.builder()
   .option(DRIVER, "postgresql")
   .option(HOST, "...")
   .option(PORT, 5432)  // optional, defaults to 5432
   .option(USER, "...")
   .option(PASSWORD, "...")
   .option(DATABASE, "...")  // optional
   .option(OPTIONS, options) // optional
   .build());

Publisher<? extends Connection> connectionPublisher = connectionFactory.create();

// Alternative: Creating a Mono using Project Reactor
Mono<Connection> connectionMono = Mono.from(connectionFactory.create());

Supported ConnectionFactory Discovery Options

Programmatic Configuration

Map<String, String> options = new HashMap<>();
options.put("lock_timeout", "10s");

PostgresqlConnectionFactory connectionFactory = new PostgresqlConnectionFactory(PostgresqlConnectionConfiguration.builder()
    .host("...")
    .port(5432)  // optional, defaults to 5432
    .username("...")
    .password("...")
    .database("...")  // optional
    .options(options) // optional
    .build());

Mono<Connection> mono = connectionFactory.create();

PostgreSQL uses index parameters that are prefixed with $. The following SQL statement makes use of parameters:

INSERT INTO person (id, first_name, last_name) VALUES ($1, $2, $3)

Parameters are referenced using the same identifiers when binding these:

mono.flatMapMany(connection -> connection
                .createStatement("INSERT INTO person (id, first_name, last_name) VALUES ($1, $2, $3)")
                .bind("$1", 1)
                .bind("$2", "Walter")
                .bind("$3", "White")
                .execute());

Binding also allowed positional index (zero-based) references. The parameter index is derived from the parameter discovery order when parsing the query.

Maven configuration

Artifacts can be found on Maven Central.

<dependency>
  <groupId>org.postgresql</groupId>
  <artifactId>r2dbc-postgresql</artifactId>
  <version>${version}</version>
</dependency>

If you'd rather like the latest snapshots of the upcoming major version, use our Maven snapshot repository and declare the appropriate dependency version.

<dependency>
  <groupId>org.postgresql</groupId>
  <artifactId>r2dbc-postgresql</artifactId>
  <version>${version}.BUILD-SNAPSHOT</version>
</dependency>

<repository>
<id>sonatype-nexus-snapshots</id>
<name>Sonatype OSS Snapshot Repository</name>
<url>https://oss.sonatype.org/content/repositories/snapshots</url>
</repository>

Connection Fail-over

To support simple connection fail-over it is possible to define multiple endpoints (host and port pairs) in the connection url separated by commas. The driver will try once to connect to each of them in order until the connection succeeds. If none succeeds a normal connection exception is thrown. Make sure to specify the failover protocol.

The syntax for the connection url is:

r2dbc:postgresql:failover://user:foo@host1:5433,host2:5432,host3

For example an application can create two connection pools. One data source is for writes, another for reads. The write pool limits connections only to a primary node:

r2dbc:postgresql:failover://user:foo@host1:5433,host2:5432,host3?targetServerType=primary.

Cursors

R2DBC Postgres supports both, the simple and extended message flow.

Cursored fetching is activated by configuring a fetchSize. Postgres cursors are valid for the duration of a transaction. R2DBC can use cursors in auto-commit mode (Execute and Flush) to not require an explicit transaction (BEGIN…COMMIT/ROLLBACK). Newer pgpool versions don't support this feature. To work around this limitation, either use explicit transactions when configuring a fetch size or enable compatibility mode. Compatibility mode avoids cursors in auto-commit mode (Execute with no limit + Sync). Cursors in a transaction use Execute (with fetch size as limit) + Sync as message flow.

Listen/Notify

Listen and Notify provide a simple form of signal or inter-process communication mechanism for processes accessing the same PostgreSQL database. For Listen/Notify, two actors are involved: The sender (notify) and the receiver (listen). The following example uses two connections to illustrate how they work together:

PostgresqlConnection sender= …;
        PostgresqlConnection receiver= …;

Flux<Notification> listen = receiver.createStatement("LISTEN mymessage")
                                .execute()
                                .flatMap(PostgresqlResult::getRowsUpdated)
        .thenMany(receiver.getNotifications());

        Mono<Void> notify=sender.createStatement("NOTIFY mymessage, 'Hello World'")
        .execute()
        .flatMap(PostgresqlResult::getRowsUpdated)
        .then();

Upon subscription, the first connection enters listen mode and publishes incoming Notifications as Flux. The second connection broadcasts a notification to the mymessage channel upon subscription.

Transaction Definitions

Postgres supports additional options when starting a transaction. In particular, the following options can be specified:

• Isolation Level (isolationLevel) (reset after the transaction to previous value)
• Transaction Mutability (readOnly)
• Deferrable Mode (deferrable)

These options can be specified upon transaction begin to start the transaction and apply options in a single command roundtrip:

PostgresqlConnection connection= …;

        connection.beginTransaction(PostgresTransactionDefinition.from(IsolationLevel.SERIALIZABLE).readOnly().notDeferrable());

See also: https://www.postgresql.org/docs/current/sql-begin.html

JSON/JSONB support

PostgreSQL supports JSON by storing values in JSON/JSONB columns. These values can be consumed and written using the regular R2DBC SPI and by using driver-specific extensions with the io.r2dbc.postgresql.codec.Json type.

You can choose from two approaches:

• Native JSONB encoding using the Json wrapper type.
• Using scalar types.

The difference between the Json type and scalar types is that Json values are written encoded as JSONB to the database. byte[] and String types are represented as BYTEA respective VARCHAR and require casting ($1::JSON) when used with parameterized statements.

The following code shows INSERT and SELECT cases for JSON interaction:

CREATE TABLE my_table (my_json JSON);

Write JSON

connection.createStatement("INSERT INTO my_table (my_json) VALUES($1)")
            .bind("$1", Json.of("{\"hello\": \"world\"}")).execute();

Consume JSON

connection.createStatement("SELECT my_json FROM my_table")
            .execute()
            .flatMap(it -> it.map((row, rowMetadata) -> row.get("my_json", Json.class)))
            .map(Json::asString);

Write JSON using casting

connection.createStatement("INSERT INTO my_table (my_json) VALUES($1::JSON)")
    .bind("$1", "{\"hello\": \"world\"}").execute();

Consume JSON as scalar type

connection.createStatement("SELECT my_json FROM my_table")
    .execute()
    .flatMap(it -> it.map((row, rowMetadata) -> row.get("my_json", String.class)));

The following types are supported for JSON exchange:

• io.r2dbc.postgresql.codec.Json
• ByteBuf (must be released after usage to avoid memory leaks)
• ByteBuffer
• byte[]
• String
• InputStream (must be closed after usage to avoid memory leaks)

CITEXT support

CITEXT is a built-in extension to support case-insensitive text columns. By default, the driver sends all string values as VARCHAR that cannot be used directly with CITEXT (without casting or converting values in your SQL).

If you cast input, then you can send parameters to the server without further customization of the driver:

CREATE TABLE test (ci CITEXT);
SELECT ci FROM test WHERE ci = $1::citext;

If you want to send individual String-values in a CITEXT-compatible way, then use Parameters.in(…):

connection.createStatement("SELECT ci FROM test WHERE ci = $1")
            .bind("$1", Parameters.in(PostgresqlObjectId.UNSPECIFIED, "Hello"))
            .execute();

If you do not have control over the created SQL or you want to send all String values in a CITEXT-compatible way, then you can customize the driver configuration by registering a StringCodec to send String values with the UNSPECIFIED OID to let Postgres infer the value type from the provided values:

Builder builder = PostgresqlConnectionConfiguration.builder();

builder.codecRegistrar((connection, allocator, registry) -> {
    registry.addFirst(new StringCodec(allocator, PostgresqlObjectId.UNSPECIFIED, PostgresqlObjectId.VARCHAR_ARRAY));
    return Mono.empty();
});

You can register also the CodecRegistrar as Extension so that it gets auto-detected during ConnectionFactory creation.

Cursors

The driver can consume cursors that were created by PL/pgSQL as refcursor. Cursors are represented as RefCursor objects. Cursors obtained from Result can be used to fetch the cursor directly. Since cursors are stateful, they must be closed once they are no longer in use.

connection.createStatement("SELECT show_cities_multiple()").execute()
    .flatMap(result -> result.map((row, rowMetadata) -> row.get(0, RefCursor.class)))
    .flatMap(cursor -> {
        Mono<PostgresResult> data = cursor.fetch()
            .flatMap(…)
            .then(rc.close());
        return data;
    });

Logical Decode

PostgreSQL allows replication streaming and decoding persistent changes to a database's tables into useful chunks of data. In PostgreSQL, logical decoding is implemented by decoding the contents of the write-ahead log, which describe changes on a storage level, into an application-specific form such as a stream of tuples or SQL statements.

Consuming the replication stream is a four-step process:

1. Obtain a replication connection via PostgresqlConnectionFactory.replication().
2. Create a replication slot (physical/logical).
3. Initiate replication using the replication slot.
4. Once the replication stream is set up, you can consume and map the binary data using ReplicationStream.map(…).

On application shutdown, close() the ReplicationStream.

Note that a connection is busy once the replication is active and a connection can have at most one active replication stream.

Mono<PostgresqlReplicationConnection> replicationMono = connectionFactory.replication();

// later:
ReplicationSlotRequest request = ReplicationSlotRequest.logical()
                                        .slotName("my_slot")
                                        .outputPlugin("test_decoding")
                                        .temporary()
                                        .build();
Mono<ReplicationSlot> createSlot = replicationConnection.createSlot(request);

ReplicationRequest replicationRequest = ReplicationRequest.logical()
                                        .slotName("my_slot")
                                        .startPosition(LogSequenceNumber.valueOf(0))
                                        .slotOption("skip-empty-xacts", true)
                                        .slotOption("include-xids", false)
                                        .build();

Flux<T> replicationStream = replicationConnection.startReplication(replicationRequest).flatMapMany(it -> {
    return it.map(byteBuf -> {…})
        .doOnError(t -> it.close().subscribe());
});

Postgres Enum Types

Applications may make use of Postgres enumerated types by using EnumCodec to map custom types to Java enum types. EnumCodec requires the Postgres OID and the Java to map enum values to the Postgres protocol and to materialize Enum instances from Postgres results. You can configure a CodecRegistrar through EnumCodec.builder() for one or more enumeration type mappings. Make sure to use different Java enum types otherwise the driver is not able to distinguish between Postgres OIDs.

Example:

SQL:

CREATE TYPE my_enum AS ENUM ('FIRST', 'SECOND');

Java Model:

enum MyEnumType {
  FIRST, SECOND;
}

Codec Registration:

PostgresqlConnectionConfiguration.builder()
        .codecRegistrar(EnumCodec.builder().withEnum("my_enum",MyEnumType.class).build());

When available, the driver registers also an array variant of the codec.

Data Type Mapping

This reference table shows the type mapping between PostgreSQL and Java data types:

Types in bold indicate the native (default) Java type.

Support for the following single-dimensional arrays (read and write):

Extension mechanism

This driver accepts the following extensions:

• CodecRegistrar to contribute Codecs for PostgreSQL ObjectIDs.

Extensions can be registered programmatically using PostgresConnectionConfiguration or discovered using Java's ServiceLoader mechanism (from META-INF/services/io.r2dbc.postgresql.extension.Extension).

The driver ships with built-in dynamic codecs (e.g. hstore, PostGIS geometry) that are registered during the connection handshake depending on their availability while connecting. Note that Postgres extensions registered after a connection was established require a reconnect to initialize the codec.

Logging

If SL4J is on the classpath, it will be used. Otherwise, there are two possible fallbacks: Console or java.util.logging.Logger). By default, the Console fallback is used. To use the JDK loggers, set the reactor.logging.fallback System property to JDK.

Logging facilities:

• Driver Logging (io.r2dbc.postgresql)
• Query Logging (io.r2dbc.postgresql.QUERY on DEBUG level)
• Parameters' values Logging (io.r2dbc.postgresql.PARAM on DEBUG level)
• Transport Logging (io.r2dbc.postgresql.client)
  • DEBUG enables Message exchange logging
  • TRACE enables traffic logging

Logging that is associated with a connection reports the logical connection id (cid) which is a driver-local connection counter and the Postgres Process Id (pid) once the connection handshake finishes.

Getting Help

Having trouble with R2DBC? We'd love to help!

• Check the spec documentation, and Javadoc.
• If you are upgrading, check out the changelog for "new and noteworthy" features.
• Ask a question - we monitor stackoverflow.com for questions tagged with r2dbc. You can also chat with the community on Gitter.
• Report bugs with R2DBC PostgreSQL at github.com/pgjdbc/r2dbc-postgresql/issues.

Reporting Issues

R2DBC uses GitHub as issue tracking system to record bugs and feature requests. If you want to raise an issue, please follow the recommendations below:

• Before you log a bug, please search the issue tracker to see if someone has already reported the problem.
• If the issue doesn't already exist, create a new issue.
• Please provide as much information as possible with the issue report, we like to know the version of R2DBC PostgreSQL that you are using and JVM version.
• If you need to paste code, or include a stack trace use Markdown ``` escapes before and after your text.
• If possible try to create a test-case or project that replicates the issue. Attach a link to your code or a compressed file containing your code.

Building from Source

You don't need to build from source to use R2DBC PostgreSQL (binaries in Maven Central), but if you want to try out the latest and greatest, R2DBC PostgreSQL can be easily built with the maven wrapper. You also need JDK 1.8 and Docker to run integration tests.

 $ ./mvnw clean install

If you want to build with the regular mvn command, you will need Maven v3.5.0 or above.

Also see CONTRIBUTING.adoc if you wish to submit pull requests.

Running JMH Benchmarks

Running the JMH benchmarks builds and runs the benchmarks without running tests.

 $ ./mvnw clean install -Pjmh

License

This project is released under version 2.0 of the Apache License.