9

RLink has not seen significant changes since it first appeared back in V9 (2012). OTOH, it has a number of issues (existing from the start or accumulated over the years), which make it more difficult to work with. Here is a list of some of them:

  • Bundled version of R shipped with RLink is very old. Also, it has stopped working altogether in 12.1 on some platforms (Windows, Linux).
  • External installations of R are not officially supported on platforms other than Windows. Workarounds exist, but require more work from the user.
  • There is no simple way to install R packages from within RLink.
  • There is no supported way to use ExternalEvaluate with externally installed versions of R.
  • R plotting functionality does not work out of the box with RLink.

Are there any plans to address these and other RLink-related issues?

Leonid Shifrin
  • 114,335
  • 15
  • 329
  • 420

1 Answers1

9

This answer might be viewed as somewhat an abuse of the main site, and might be better suited for Meta, but I hope the folks will forgive me for that, since the main site clearly enjoys a lot more attention than the Meta. Besides, I will edit in a more detailed textual answer at some later point.

The webinar on April 29th, 2020

There are currently plans to address at least some of these issues. I will add a more detailed textual answer here later, but all those who are interested in this topic, are most welcome to register for

Wolfram University webinar

where I will talk about some of the current plans regarding future development / improvements of RLink.

Notes

Please note a few things:

  • The improvements to be discussed are plans rather than already existing RLink features.
  • I will not be the only speaker there, other topics / talks will discuss Wikidata search capabilities and new features in Database integration functionality.
  • The opinions of users who come to the seminar are important in determining the (relative) importance of specific issues and planned improvements.
Leonid Shifrin
  • 114,335
  • 15
  • 329
  • 420
  • Thanks for the heads up! I don't use R, but should probably learn it at some point. Is there already, or could there be, a way to use the Mathematica front end notebook interface with an R back end? I know there's ExternalEvaluate which you can bring up with > -- I was thinking more of R being the default evaluator for every cell. – Chris K Apr 28 '20 at 17:51
  • Setting R as a default evaluator for the cell is possible using the "CellEvaluationFunction" option set on the notebook level. – Leonid Shifrin Apr 28 '20 at 18:17
  • If you have a chance to spell out the details for me, please do. I've tried a few variants on SetOptions[EvaluationNotebook[], CellEvaluationFunction -> REvaluate] but no luck. – Chris K Apr 28 '20 at 19:03
  • @ChrisK Can't test at the moment, but try SetOptions[EvaluationNotebook[], CellEvaluationFunction -> (REvaluate[#]&)]. – Leonid Shifrin Apr 28 '20 at 19:37
  • Great presentation, Leonid! – Anton Antonov Apr 29 '20 at 18:38
  • @AntonAntonov Thanks! – Leonid Shifrin Apr 29 '20 at 18:47
  • The link to the webinar is no longer current. Checking the Wolfram Website I don't find an indication of anything that remained regarding this topic. If it does still exist somewhere, it may well be useful to gain access to it, particularly relating to getting RDBMS databaase acccess via R from within Mathematica. – Stuart Poss Jun 02 '23 at 00:57
  • @StuartPoss Not sure I can help with that webinar link, alas. But accessing RDBMS from Mathematica via R does not strike me like a great idea. Mathematica has its own dedicated toolkit for working with RDBMs (even 2 actually, older DatabaseLink and more recent Databases framework), both are better integrated with the language – Leonid Shifrin Jun 07 '23 at 22:48
  • Yes,, I having been using Database Link, and inquired to be able to take advantage of R routines without having to rewrite them in Mathematica. – Stuart Poss Jun 12 '23 at 08:48
  • @StuartPoss Then using RLink may make sense. But it has not been optimized for large data transfer, so if your use case requires that, you might be out of luck. There are a few other limitations of RLink as compared to a standalone R process. – Leonid Shifrin Jun 21 '23 at 23:43
  • Thanks for the advice. I have been finding out the hard way. – Stuart Poss Jul 05 '23 at 04:39