RJavaServer

Java server that can call R functions using JRI and rJava packages

Requirements

This server needs an installation of R in the local OS, as well as the rJava package installed:

• Install R in your machine (see https://cran.r-project.org/)
• Install rJava using R console: type install.package("rJava") inside R console
• Install Rserve using R console: type install.package("Rserve") inside R console

Once these steps are done, you have to configure some variables inside your local system depending on the approach you're going to use.

Rserve:

• R_HOME: Pointing to your local R installation (Ej: /Library/Frameworks/R.framework/Resources)

JRI:

• R_HOME: Pointing to your local R installation (Ej: /Library/Frameworks/R.framework/Resources)
• LD_LIBRARY_PATH: Pointing to R lib directory as well as JRI directory (EJ: /app/vendor/R/lib/R/lib:/app/vendor/R/lib/R/bin)

To sum up, look at the content of the script $R_HOME/library/rJava/jri/run and configure your local system usign the same variable names and values.

Moreover, the server must be started with the JVM parameter -D-Djava.library.path pointing at the JRI installation directory, ie:

$JAVA_HOME/bin/java -Dserver.port=$PORT -Djava.library.path=/app/vendor/R/lib/R/library/rJava/jri/ -jar target/RJavaServer.jar

If you have trouble finding this directory, try to look for the JRI SO library:

find / -name "libjri.*"

Heroku

In order to execute this server in Heroku, a Procfile has been added to the project. It can be deployen to any Heroku applications that has R environment installed.

To install R inside a heroku application, you have to use multipack-buildpack and r-buildpack.

To install this project into a Heroku server, follow theses steps:

1. Create a new application that uses multipack buidpack typing in the command line:

   heroku create --stack cedar-14 --buildpack https://github.com/heroku/heroku-buildpack-multi.git --app RJavaServer

2. Create a file called .buildpacks inside your project, and type here all the Heroku buildpacks you want to install in your environment (this project already contains this file)

3. In addition to your application code, you must tell R which packages must be installed. This task is done inside a file called init.r (this project already contains a init.r script)

Usage

Right now, the REST API has only very basic operations, but you can extend it to call whatever R function you want. Moreover, you can create your own R script and load it using engine.parseAndEval("source(\"/yourpath/yourscript.r\""). (See an example at UseREngineInFrontOfJRIEngineTest)

To test the server, just try to execute a POST command over the mean resource.

Curl example:

`curl http://localhost:8080/sqrt -X POST -d "5" -H "Content-type: application/json" -i``

It should return:

HTTP/1.1 201 Created
Server: Apache-Coyote/1.1
X-Application-Context: application:local:8080
Content-Type: application/json;charset=UTF-8
Transfer-Encoding: chunked
Date: Sun, 20 Sep 2015 15:46:45 GMT

2.23606797749979

Datadog

This project has a module that automatically sends metrics to Datadog, an online monitoring system.

To enable this feature, just create an account in Datadog (https://www.datadoghq.com/) and put your API Key in the configuration file (application.yml):

metrics:
  apiKey: <your API key>
  host: <host of this app (optional)>
  period: <time, in seconds, between events>
  enabled: true