Installation
Requirements
-
Java 17+ (this binding targets Java 25 LTS by default; see the pom.xml
maven.compiler.releaseproperty)
As a standalone dependency
Add the Maven coordinate to your pom.xml:
<dependency>
<groupId>io.github.thisisamirv</groupId>
<artifactId>fastlowess</artifactId>
<version>5.1.0</version>
</dependency>
The published jar bundles prebuilt native libraries for linux-x86_64, linux-x86_64-musl (Alpine), linux-aarch64, linux-aarch64-musl (Alpine), macos-x86_64, macos-aarch64, windows-x86_64, and windows-aarch64 (see NativeBridge’s `osArchDir()); NativeBridge auto-detects your platform and libc (checking for Alpine’s /etc/alpine-release and musl’s ld-musl-* dynamic linker), extracts the matching one to a temp file, and loads it — no separate native library installation needed on any of those platforms.
To override the bundled library (e.g. a custom or GPU-enabled build), either system property takes precedence:
java -Dfastlowess.native.path=/path/to/libfastlowess_java.so -cp your-app.jar com.example.Main
# or point at a directory instead of an exact file:
java -Dfastlowess.native.dir=/path/to/native/lib -cp your-app.jar com.example.Main
Build the native library yourself from the lowess-project source:
git clone https://github.com/thisisamirv/lowess-project
cd lowess-project
cargo build -p fastlowess-java --release
Within the lowess-project monorepo
If you’re working inside a checkout of lowess-project, the root Makefile handles building the native library for you:
make java # build the Rust JNI crate, then `mvn package`
make java-dev # full dev checks: fmt, clippy, export verification, mvn test
NativeBridge loads the native library from the path given by the fastlowess.native.dir system property (or -Dfastlowess.native.dir=…), which defaults to ../../target/debug relative to bindings/java — where cargo build -p fastlowess-java places the shared library within this repo’s layout.
GPU backend
GPU acceleration is available via the native library’s gpu Cargo feature (wgpu: Vulkan, Metal, DirectX 12, and GLES on Android), but is not enabled in the default build, nor bundled in the published jar. See the C++ binding’s GPU Backend guide for the underlying feature; building fastlowess-java with --features gpu produces a library where FastLowess.gpuEnabled() returns true — point the JVM at it via fastlowess.native.path/fastlowess.native.dir as shown above.