Vanilla Java JSON Engine 🚀

A hyper-optimized, zero-allocation, reflection-free JSON parser and serializer built for the modern JVM (JDK 26+).

This engine was designed from the ground up to achieve maximum throughput and minimal garbage collection pressure, outperforming industry standards like Jackson in direct JMH benchmarks—without a single external dependency.

✨ Key Features

• Blazing Fast (JMH Proven): Consistently beats Jackson Databind in throughput and offers significantly lower jitter/variance under load.
• Zero Dependencies: 100% pure Vanilla Java. No jackson, no gson, no org.json.
• Zero-Allocation Hot-Path: Utilizes advanced object pooling, custom cached byte arrays, and direct byte[] buffer pushing. The Garbage Collector rests while you parse millions of records.
• Project Loom / Virtual Thread Ready: Completely free of blocking locks or synchronized bottlenecks in the hot-path. Uses elastic soft-limit object pools instead of memory-heavy ThreadLocal structures.
• Modern JDK Architecture: Leverages Pattern Matching, Records, Sealed Interfaces, and LambdaMetafactory instead of slow, traditional Reflection.
• Hybrid Stack Architecture: Exploits fast native JVM call stacks for typical JSON depth, while falling back to a safe, heap-based state machine for infinite-depth protection.

---

📦 Installation (via JitPack)

Add the library to your project. (Note: Requires JDK 26 or higher)

Maven: Add the JitPack repository to your pom.xml: jitpack.io https://jitpack.io

Then add the dependency: com.github.SkateScout suche.org-json v1.0.0

Gradle: repositories { mavenCentral() maven { url 'https://jitpack.io' } }

dependencies { implementation 'com.github.SkateScout:suche.org-json:v1.0.0' }

---

📊 Performance (JMH Benchmark)

Tested against com.fasterxml.jackson.core:jackson-databind. Benchmarks use the hybrid recursion approach (maxRecursion=128) on JDK 26. (Throughput in operations per second. Higher is better.)

Deserialization (Read)

Dataset	Jackson (ops/s)	Suche Engine (ops/s)	Result
Canada (GeoJSON)	38,624	56,342	+ 45% Faster 🚀
Twitter (Web API)	473,346	520,496	+ 10% Faster 🚀
Citm Catalog	234,858	231,188	~ Equal

Serialization (Write)

Dataset	Jackson (ops/s)	Suche Engine (ops/s)	Result
Canada (GeoJSON)	58,893	120,305	+ 104% Faster 🚀
Citm Catalog	412,337	475,716	+ 15% Faster 🚀
Twitter (Web API)	935,974	793,936	- 15%

---

🛠️ Quick Start

1. Initialize the Engine

import org.suche.json.JsonEngine; import org.suche.json.MetaConfig;

public static final JsonEngine ENGINE = JsonEngine.of(MetaConfig.DEFAULT);

2. Serialize (Write)

try (var stream = ENGINE.jsonOutputStream(outputStream)) { stream.writeObject(myObject); }

3. Deserialize (Read)

try (var stream = ENGINE.jsonInputStream(inputStream)) { TargetClass parsed = stream.readObject(TargetClass.class); }

---

🏗️ Architecture Insights

• Fractional Truncation & Bitmasking: Bypasses Double.parseDouble() and branch-heavy String operations completely in the hot-path.
• FastKeyTable: Property resolution relies on bit-masked custom hash tables instead of standard HashMaps, eliminating Map.Entry allocations.
• Fail-Fast Reflection: Operations on final fields are handled cleanly without reflective mutation hacks, ensuring compatibility with Jigsaw modules.
• The Static Cache Trap Avoided: Caches are bound to the JsonEngine instance, preventing Metaspace memory leaks during hot-reloads in application servers.

---

📄 License

This project is licensed under the MIT License.