johnaohara/java-r-interoperability

Experimentation with integrating Java & R using GraalVM to perform change point detection in noisy performance time-series metric

Java & R interoperability with GraalVM

Repo to demonstrate the interoperability between Java and R using GraalVM

Example to determine multiple change points in data-set using top down 'E-divisive with Means' e.divisive & e.agglo from the R ecp package

e.agglo Results

Change point detection in dataset with large signal-to-noise ration

Change point detection in dataset with small signal-to-noise ration

Running Example

This demo relies on GraalVM with the R runtime installed

1. Install SDK manager (https://sdkman.io/install)

   $ curl -s "https://get.sdkman.io" | bash

2. Install GraalVM

   Find available GraalVM distributions

   $ sdk list java | grep grl

   Install latest GraalVM distribution

   $ sdk install java x.y.z.r11-grl

3. Set GraalVM to current java runtime

   $ sdk use java 20.3.0.r11-grl

4. Install GraalVM R

   $ gu install R

   R should be available in your PATH

   $ R --version
   R version 3.6.1 (FastR)

5. Install ecp package

   The ecp package contains the e.divisive function used to calculate change points in datasets

   $ Rscript -e 'install.packages("ecp")'

6. Run tests

   $ mvn clean test

N.B. if you see errors in the maven build e.g. package org.graalvm.polyglot does not exist please ensure that you are using GraalVM sdk see step 3

$ java --version
  openjdk 11.0.9 2020-10-20
  OpenJDK Runtime Environment GraalVM CE 20.3.0 (build 11.0.9+10-jvmci-20.3-b06)
  OpenJDK 64-Bit Server VM GraalVM CE 20.3.0 (build 11.0.9+10-jvmci-20.3-b06, mixed mode, sharing)