ClickHouse Java Libraries

Java libraries for connecting to ClickHouse and processing data in various formats. Java client is async, lightweight, and low-overhead library for ClickHouse; while JDBC and R2DBC drivers are built on top of the Java client with more dependencies and features. Java 8 or higher is required to use the libraries. In addition, please use ClickHouse 20.7+ or any of active releases.

Features

Category	Feature	Supported	Remark
API	JDBC	✅
	R2DBC	✅	supported since 0.4.0
Protocol	HTTP	✅	recommended, defaults to `java.net.HttpURLConnection` and it can be changed to `java.net.http.HttpClient`(unstable) or `Apache HTTP Client 5`. Note that the latter was added in 0.4.0 to support custom socket options.
	gRPC	✅	still experimental, works with 22.3+, known to has issue when using LZ4 compression
	TCP/Native	✅	`clickhouse-cli-client`(wrapper of ClickHouse native command-line client) was added in 0.3.2-patch10, `clickhouse-tcp-client` will be available in 0.5
	Local/File	❌	`clickhouse-cli-client` will be enhanced to support `clickhouse-local`
Compatibility	Server < 20.7	❌	use 0.3.1-patch(or 0.2.6 if you're stuck with JDK 7)
	Server >= 20.7	✅	use 0.3.2 or above. All active releases are supported.
Compression	lz4	✅	default
	zstd	✅	supported since 0.4.0, works with ClickHouse 22.10+
Data Format	RowBinary	✅	`RowBinaryWithNamesAndTypes` for query and `RowBinary` for insertion
	TabSeparated	✅	Does not support as many data types as RowBinary
Data Type	AggregatedFunction	❌	limited to `groupBitmap`, and known to have issue with 64bit bitmap
	Array(*)	✅
	Bool	✅
	Date*	✅
	DateTime*	✅
	Decimal*	✅	`SET output_format_decimal_trailing_zeros=1` in 21.9+ for consistency
	Enum*	✅	can be treated as both string and integer
	Geo Types	✅	Point, Ring, Polygon, and MultiPolygon
	Int, UInt	✅	UInt64 is mapped to `long`
	IPv*	✅
	Map(*)	✅
	Nested(*)	✅
	Object('JSON')	✅	supported since 0.3.2-patch8
	SimpleAggregateFunction	✅
	*String	✅
	Tuple(*)	✅
	UUID	✅
High Availability	Load Balancing	✅	supported since 0.3.2-patch10
	Failover	✅	supported since 0.3.2-patch10
Transaction	Transaction	✅	supported since 0.3.2-patch11, use ClickHouse 22.7+ for native implicit transaction support
	Savepoint	❌
	XAConnection	❌

Usage

The library can be downloaded from both Github Releases and Maven Central. Development snapshots(aka. nightly build) are available on Sonatype OSSRH.

<repositories>
    <repository>
        <id>ossrh</id>
        <name>Sonatype OSSRH</name>
        <url>https://s01.oss.sonatype.org/content/repositories/snapshots/</url>
    </repository>
</repositories>

Java Client

<dependency>
    <groupId>com.clickhouse</groupId>
    <!-- or clickhouse-grpc-client if you prefer gRPC -->
    <artifactId>clickhouse-http-client</artifactId>
    <version>0.4.0</version>
</dependency>

//  endpoint: protocol://host[:port][/database][?param[=value][&param[=value]][#tag[,tag]]
ClickHouseNode endpoint = ClickHouseNode.of("https://localhost"); // http://localhost:8443?ssl=true&sslmode=NONE
// endpoints: [defaultProtocol://]endpoint[,endpoint][/defaultDatabase][?defaultParameters][#defaultTags]
ClickHouseNodes endpoints = ClickHouseNodes.of("http://(https://explorer@play.clickhouse.com:443),localhost,"
    + "(tcp://localhost?!auto_discovery#experimental),(grpc://localhost#experimental)?failover=3#test")

try (ClickHouseClient client = ClickHouseClient.newInstance(ClickHouseProtocol.HTTP);
    ClickHouseResponse response = client.connect(endpoint) // or client.connect(endpoints)
        // you'll have to parse response manually if using a different format
        .format(ClickHouseFormat.RowBinaryWithNamesAndTypes)
        .query("select * from numbers(:limit)")
        .params(1000).executeAndWait()) {
    // or response.stream() if you prefer stream API
    for (ClickHouseRecord r : response.records()) {
        int num = r.getValue(0).asInteger();
        // type conversion
        String str = r.getValue(0).asString();
        LocalDate date = r.getValue(0).asDate();
    }

    ClickHouseResponseSummary summary = response.getSummary();
    long totalRows = summary.getTotalRowsToRead();
}

JDBC Driver

<dependency>
    <groupId>com.clickhouse</groupId>
    <artifactId>clickhouse-jdbc</artifactId>
    <version>0.4.0</version>
    <!-- use uber jar with all dependencies included, change classifier to http for smaller jar -->
    <classifier>all</classifier>
    <exclusions>
        <exclusion>
            <groupId>*</groupId>
            <artifactId>*</artifactId>
        </exclusion>
    </exclusions>
</dependency>

// jdbc:(ch|clickhouse):[defaultProtocol://]endpoint[,endpoint][/defaultDatabase][?defaultParameters][#defaultTags]
String url = "jdbc:ch:https://play.clickhouse.com:443";
Properties properties = new Properties();
properties.setProperty("user", "explorer");
properties.setProperty("password", "");
// optional properties
properties.setProperty("client_name", "Agent #1");
...

ClickHouseDataSource dataSource = new ClickHouseDataSource(url, properties);
try (Connection conn = dataSource.getConnection();
    Statement stmt = conn.createStatement();
    ResultSet rs = stmt.executeQuery("show databases")) {
    ...
}

More examples can be found at here.

Build with Maven

Use mvn -Dj8 -DskipITs clean verify to compile and generate packages if you're using JDK 8. To create a multi-release jar (see JEP-238), please use JDK 11+ with ~/.m2/toolchains.xml like below, and run mvn -DskipITs clean verify instead.

<?xml version="1.0" encoding="UTF8"?>
<toolchains>
    <toolchain>
        <type>jdk</type>
        <provides>
            <version>11</version>
        </provides>
        <configuration>
            <jdkHome>/usr/lib/jvm/java-11-openjdk</jdkHome>
        </configuration>
    </toolchain>
    <toolchain>
        <type>jdk</type>
        <provides>
            <version>17</version>
        </provides>
        <configuration>
            <jdkHome>/usr/lib/jvm/java-17-openjdk</jdkHome>
        </configuration>
    </toolchain>
</toolchains>

Testing

By default, docker is required to run integration test. Docker image(defaults to clickhouse/clickhouse-server) will be pulled from Internet, and containers will be created automatically by testcontainers before testing. To test against specific version of ClickHouse, you can pass parameter like -DclickhouseVersion=22.8 to Maven.

In the case you don't want to use docker and/or prefer to test against an existing server, please follow instructions below:

make sure the server can be accessed using default account(user default and no password), which has both DDL and DML privileges
add below two configuration files to the existing server and expose all defaults ports for external access
- ports.xml - enable all ports
- and users.xml - accounts used for integration test Note: you may need to change root element from clickhouse to yandex when testing old version of ClickHouse.
make sure ClickHouse binary(usually /usr/bin/clickhouse) is available in PATH, as it's required to test clickhouse-cli-client

put test.properties under either ~/.clickhouse or src/test/resources of your project, with content like below:

clickhouseServer=x.x.x.x
# below properties are only useful for test containers
#clickhouseVersion=latest
#clickhouseTimezone=UTC
#clickhouseImage=clickhouse/clickhouse-server
#additionalPackages=

Benchmark

To benchmark JDBC drivers:

cd clickhouse-benchmark
mvn clean package
# single thread mode
java -DdbHost=localhost -jar target/benchmarks.jar -t 1 \
    -p client=clickhouse-jdbc -p connection=reuse \
    -p statement=prepared -p type=default Query.selectInt8

It's time consuming to run all benchmarks against all drivers using different parameters for comparison. If you just need some numbers to understand performance, please refer to this(still have plenty of room to improve according to ranking at here).

ClickHouse / clickhouse-jdbc