Skip to content

Latest commit

 

History

History
30 lines (15 loc) · 4.16 KB

existing_libraries.md

File metadata and controls

30 lines (15 loc) · 4.16 KB
Raw

Calling R from Clojure: existing libraries

All of the following projects are quite useful and have inspired the work here.

They differ in in several aspects: (a) the API and syntactic sugar offered; (b) the kind of R backend used: JRI+REngine / Rserve+REngine / Opencpu / just running R from shell (see R backends); (c) whether some Clojure notions are used as the equivalent of R's "data frames" and "matrices", and if so, what notions: Incanter / core.matrix (see Clojure's counterparts of R's data abstractions).

  • Rincanter was announced by Joel Boehland allows in early 2010. It was based on JRI (R from Java using JNI) as the R backend, and for data-abstractions used Incanter, an R-inspired library for Clojure, which was popular back then.

  • This Rincanter fork by Vladimir Kadychevski changed the R backend to Rserve, through the REngine layer on the Java side, which supports both Rserve and JRI as backends. This change allowed for more robust and production-friendly R-interop, as R runs on a separate process.

  • This continuing fork by the skm-ice group kept working on the API.

  • Rojure, presented by Carsten Behring, changed the data abstraction from Incanter to core.matrix. At those days, core.matrix was becoming a standard data abstraction layer used by several libraries, evetually including Incanter itself. This allowed Rojure to be completely independent of Incanter, but still be used from Incanter.

  • clj-jri by SAWADA Takahiro / Gugen Koubou LLC is another, simple, wrapper of R through JRI. It does not support a data-frame-like structure on the Clojure side.

  • rashinban by Takahiro Noda is another simple Clojure library for calling R through Rserve, with a clean and simple API, based on wrapping R functions with Clojure functions. It does not support a data-frame-like notion either.

  • Opencpu-clj is another, earlier project by Carsten Behring. It uses Opencpu as the R backend.

    • Some experiment to generalize to Clojuresctipt has been done here (currently broken).

  • gg4clj by Jony Hudson appeared shortly after Opencpu-clj. Its main purpose is wrapping R's famouse "Grammer of Graphics" ggplot2 library (but it actually introduces some ideas that are applicable for more general use). It does so by running any R computation as a brand new process. One interesting innovation of this libraaary is an EDN-like syntax that translates to R. We actually use gg4clj for code generation inside this project.

  • huri by Simon Belak is a general library for data science, that does a lot more than calling R. One of its components is a collection of functions for data visualization, that build upon the way gg4clj offers, and add simple, composable ways to create ggplot2 plots in Clojure.

  • graalvm-interop is an active project by David Pham, that allows to interoperate with FastR at GraalVM.

  • rcaller is a Java library primarily for calling R from within Java. rcaller creates an R process, sends R data structures and codes to R process, and gets back the results in XML format. Sequential calls are processed using a single process. The package is also suitable for the other JVM languages including Clojure. Here is an example usage. It is also citable using this journal paper.