Skip to main content

Java implementation

Status: planned

The Java implementation is architecturally decided but not yet implemented — no code exists yet. The points below document the decisions taken; the authoritative source is DECISIONS §21.

Target audiences​

Two main audiences: enterprise backends (Spring, microservices, optiCloud) as well as big-data/AI pipelines (Spark, Flink, data analysis). On top of that there is a concrete embedded project in which a Java application records operational data on an industrial gateway — the same "both worlds" situation that §7 already describes for C++.

Planned architecture​

  • Java 25 (current LTS) and Maven as the build system. Shipped as a single Maven artifact (groupId=com.optimeas.osf, artifactId=osf-java).
  • JPMS (Java Platform Module System): module-info.java exports only com.optimeas.osf; internal helpers under com.optimeas.osf.internal stay encapsulated — even against reflection.
  • Both writers, analogous to C++: a BlockWriter (accumulate in memory, write in one pass) for batch workflows and a StreamingWriter (FileChannel.force(true) per block) for power-loss-safe embedded recording. Both produce on-disk-identical OSF5 files.
  • Dependencies: Jackson (OSF5 JSON), SLF4J (logging facade), java.util.zip (OSFZ, in the JDK), StAX (OSF4 XML, in the JDK). Binary I/O via ByteBuffer on FileChannel with LITTLE_ENDIAN.

Specification conformance​

The Java implementation follows the same semantic rules as Rust, Python and C++ (spec rev 2026-05-04): all current data types (unsigned types via Java type promotion or BigInteger for the full range), explicit rejection of the removed types, bytearray as a read alias for binary, the version-deterministic null-termination rule for string/binary, and all four magic-header identifiers.

Source code and further reading​