R GUIs: Which one fits you?

The gap of the new “digital divide” between those who only use computers when they are as easy to use as iPads and smartphones and those who like (or at least accept) to type commands to perform jobs, seems to get bigger and bigger.

R – the lingua franca of statistical computing – is exactly such a command-line based language, reasonably well designed but still not GUI based at all. At this point GUIs are the only solution to make R accessible for “generation point-and-click” and bridge the divide.

Personally, I am happy to use all well designed GUIs but as well see the power of language based command line interfaces – you need to work with both to be most effective.
But let’s come to the comparison of the four different frontends for R (in lexicographic order) which try to do more than the built-in standard GUIs for the supported platforms:

(mouse-over the entries in the table to get more details)




R Studio

Technology JAVA tcl/tk KDE Qt
Installation simple simple painful easy
Approach IDE comprehensive comprehensive IDE
Interface SDI MDI (plus R) TDI MDI
Maturity 1.7-5 1.6-3 0.5.5 0.92.44
Console yes yes yes yes
CodeEditor yes no yes yes
Objbrowser yes no yes yes
DataEditor yes via fix() yes no
ModelBrws yes no no no
Logging console console extended console
Plugins via iWidgets yes yes no
Web-Client no no no yes

There are certainly more frontends and features (especially on the technical side) to consider, and not everybody will share my verdict on every point (which I even probably didn’t get completely right), but that’s what comments are for ┬á…

My summary recommendations (regarding the four candidates) are:

  • working styles are very different such that many of the above mentioned issues may be pointless
  • for many of us the built-in GUIs are pretty good already, but differ from platform to platform (so you maybe want to avoid any further hassle)
  • if you are on a Mac, half of the choices are gone already …
  • those who really don’t like to “being helped” by your software, opt for the IDE approaches!
  • those who really don’t want to learn any of the R-syntax and are purely on a user level, use one of the comprehensive approaches – you still might not be too happy though
  • if you hate installation procedures, make sure to avoid RKWard (under Windows)
  • the sleekest GUI is definitely R Studio
  • if the webpage would be wider, I should have certainly mentioned Deducer, which is a comprehensive offspring of JGR.


  1. Nice post.

    That being said, if you really want to unleash the full power of the R language, learn either Emacs or Vim. Both have comprehensive support for LaTeX (thus enabling reproducible research which is a major reason to use R) and wonderful completion features, along with being extremely powerful for editing text and code.

    If i had to pick one of the GUI’s though, it would definitely be RStudio, which has some Sweave support and looks very very nice.

  2. Tal Galili says:

    Assuming we are talking GUI (and not IDE), what about Deducer??

  3. Ian Fellows says:

    Good comparison.

    R Commander can be used on the mac, though only through X11, and it can be a bit of a pain to install. Also, contrary to the table, it has more plug-ins than any other GUI.

    JGR takes an IDE approach, but when combined with Deducer becomes comprehensive (in your terminology). It was designed this way so as to provide a unified system for R programmers and data analysts.

  4. martin says:

    Hi Ian,

    thanks for the hint regarding R commander and plug-ins – I missed that. As I said, the list is certainly not complete, and the choice of GUIs was partly driven by technical design choices of the GUIs and a good balance of approaches. Deducer is somewhere in the middle of the tools presented; I should add a note for completeness.


  5. Gabor Grothendieck says:

    I think you need to specify which platform this is relative to. My comments are relative to Windows.

    JGR is broken on Windows and has been for years although some claim they got it to work. I would put a ? after Windows in the JGR/Platform cell.

    Also on Windows Rcmdr is easy to install, not hard. You install it from within R just like any other R package, i.e.


    or via the Rgui menu:

    Packages > Install package(s) > Rcmdr

  6. David Luckett says:

    Thanks for the useful information.

    Under Windows don’t forget Tinn-R, an excellent GUI with full integration – it is my environment of choice. Emacs/ESS is powerful but has a steep learning curve.

    I installed RKWard under Ubuntu 10.10 on the weekend – it was a breeze, all automatic and no problems at all.

    Its still useful to have the Rgui (Rconsole) menu option to “Install package from local zip file”.

  7. martin says:

    > JGR is broken on Windows and has been for years although some claim
    > they got it to work. I would put a ? after Windows in the JGR/Platform cell.
    Yes, I claim it too. Installed it yesterday via:


  8. John S. says:

    I’ve tried them all, and keep coming back to Emacs with ESS. Aquamacs is a nice implementation of Emacs on the Mac.

  9. Mark says:

    Not sure why StatET never seems to get the love in reviews like this, but it’s cross-platform with Eclipse and arguably superior:


    There’s also SciViews-K, also cross-platform

  10. Gabor Grothendieck says:

    This is what I get when I run JGR():

    Loading required package: JGR
    Failed with error: ‘package ‘rJava’ 0.7-0 was found, but >= 0.8.6 is required by ‘JGR”
    Loading required package: JGR
    Failed with error: ‘package ‘rJava’ 0.7-0 was found, but >= 0.8.6 is required by ‘JGR”
    > Error: could not find function “.getModels”
    Error: could not find function “.getDataObjects”
    Error: could not find function “.getOtherObjects”
    Error: could not find function “.getFunctionsInWS”

    In fact, despite the message, I have rJava 0.9-0 installed and other packages that I use that depend on rJava do work so there is nothing wrong with my installation.

    Many years ago JGR did work on Windows but changes in R likely have broken it since then. I have tried it every year or so for the past few years and it has consistently failed to work on Windows. I suspect its screwing up in locating the local library and is finding rJava in some other installation of R I have on my machine. I have also tried manually setting various environment variables in case I could force it to work properly but that did not help.

  11. martin says:

    Hi Gabor,

    installation worked fine for me (under Windows XP with R 2.12.1) as I already noted in the comment above, so please post your problem to stats-rosuda-devel@listserv.uni-augsburg.de as it might be a problem that is more related to your specific installation.



  12. Gabor Grothendieck says:

    I have already reported these problems repeatedly over a period of literally years.

  13. peter says:


    very helpful indeed, but I would not describe the RKWard installation as “painful” on Windows. Sure, if you install all the different parts (KDE, R, RKWard) as recommended via the website, this is really painful and I gave up repeatedly.

    However, the project website features a setup file that installs all components automatically. While the projects says that its a really rough beta and might not work, it worked out fine (no problems) at all with XP and Windows 7. So, you might want to give it a try…

    Personally, I like this one best of all GUIs mentioned above, because it is fairly easy to use for people coming from other stats packages. To find out about R’s power one can quickly run analyses via a point and click interface and explore the language from there.


  14. Elephant says:

    RKWard installation is a breeze in Linux. Just select it from your repository and install.

    JGR doesn’t install properly in Linux either. It is looking for some other version of Java. I suspect it is JGR that is outdated and can’t be bothered to try to research a fix.

  15. Tim Bates says:

    Just a punt for TextMate with Hans Jorg Bibiko’s R bundle… Very nice environment.

    And also for the R.app GUI: Greatly enhanced since 2011.

  16. m.eik says:


    there have been a lot of improvements in RKWard (current stable release is 0.5.7). amongst other things, RKWard plugins are now ordinary R packages, and hence can easily be handled exactly the same (RKWard will automatically detect newly installed plugins, even if they were installed from a plain R session). If you’re interested in developing new plugins, there’s a new R package called “rkwarddev”, which allows you to generate plugins from R code (well, admittedly, you still need to read the manual for this…). also, RKWard has been ported to Mac OS X in the meantime, using MacPorts. that work is not completely finished (some extra graphics features are missing, so you only get the usual R graphics device windows), but overall, it also runs on macs now, too.

    of course, some things are a matter of taste, and like you already mentioned, i’d say some are not quite accurately described either. RKWard does not limit you to use GUI dialogs, its approach is best described as providing both, “comprehensive” and “IDE”, by your definition. when it comes to installation, RKWard is available in all major GNU/linux distributions’ repositories, and there is arguably no easier way of managing software packages (in comparison, i’d consider *every* software installation on windows as “painful” ;-)), so this grading depends. for instance, RStudio is not yet in those repos, which makes it much more work to install and keep updated on debian/*buntu systems.

    finally, as for the dependency on some KDE libraries (it doesn’t need a full KDE installtion): at first glance this looks like a heavy burden. but if you think about the fact that every major windows/mac software product is actually bundled with tons of libraries as well (only you don’t notice it, because it feels like “installing one program”), it’s not really *that* much of a difference. you can get the same kind of feeling by using the bundled windows installer, which is no longer considered experimental ­čśë

Leave a Reply