A Java version of simdjson - a JSON parser using SIMD instructions, based on the paper Parsing Gigabytes of JSON per Second by Geoff Langdale and Daniel Lemire.

byte[] json = loadTwitterJson();

SimdJsonParser parser = new SimdJsonParser();
JsonValue jsonValue = parser.parse(json, json.length);
Iterator<JsonValue> tweets = jsonValue.get("statuses").arrayIterator();
while (tweets.hasNext()) {
    JsonValue tweet = tweets.next();
    JsonValue user = tweet.get("user");
    if (user.get("default_profile").asBoolean()) {
        System.out.println(user.get("screen_name").asString());
    }
}

byte[] json = loadTwitterJson();

SimdJsonParser parser = new SimdJsonParser();
SimdJsonTwitter twitter = simdJsonParser.parse(buffer, buffer.length, SimdJsonTwitter.class);
for (SimdJsonStatus status : twitter.statuses()) {
    SimdJsonUser user = status.user();
    if (user.default_profile()) {
        System.out.println(user.screen_name());
    }
}

record SimdJsonUser(boolean default_profile, String screen_name) {
}

record SimdJsonStatus(SimdJsonUser user) {
}

record SimdJsonTwitter(List<SimdJsonStatus> statuses) {
}

The library is available in the Maven Central Repository. To include it in your project, add the following dependency to the build.gradle file:

implementation("org.simdjson:simdjson-java:0.4.0")

or to the pom.xml file:

<dependency>
    <groupId>org.simdjson</groupId>
    <artifactId>simdjson-java</artifactId>
    <version>0.4.0</version>
</dependency>

Please remember about specifying the desired version.

Note that simdjson-java follows the SemVer specification, which means, for example, that a major version of zero indicates initial development, so the library's API should not be considered stable.

We require Java 24 or better.

To run the JMH benchmarks, execute the following command:

./gradlew jmh

To run the tests, execute the following command:

./gradlew test

This section presents a performance comparison of different JSON parsers available as Java libraries. The benchmark used the twitter.json dataset, and its goal was to measure the throughput (ops/s) of parsing and finding all unique users with a default profile.

Environment:

• CPU: Intel(R) Xeon(R) CPU E5-2686 v4 @ 2.30GHz
• OS: Ubuntu 24.04 LTS, kernel 6.8.0-1008-aws
• Java: OpenJDK 64-Bit Server VM (build 21.0.3+9-Ubuntu-1ubuntu1, mixed mode, sharing)

DOM parsers (ParseAndSelectBenchmark):

Schema-based parsers (SchemaBasedParseAndSelectBenchmark):

Environment:

• CPU: Intel(R) Xeon(R) Platinum 8375C CPU @ 2.90GHz
• OS: Ubuntu 24.04 LTS, kernel 6.8.0-1008-aws
• Java: OpenJDK 64-Bit Server VM (build 21.0.3+9-Ubuntu-1ubuntu1, mixed mode, sharing)

DOM parsers (ParseAndSelectBenchmark):

Schema-based parsers (SchemaBasedParseAndSelectBenchmark):

To reproduce the benchmark results, execute the following command:

./gradlew jmh -Pjmh.includes='.*ParseAndSelectBenchmark.*'

The benchmark may take several minutes. Remember that you need Java 18 or better.