[R] Question regarding different R versions on an enterprise network server

Duncan Murdoch murdoch.duncan at gmail.com
Thu May 21 18:54:42 CEST 2015


On 21/05/2015 12:21 PM, Assaf P. Oron wrote:
> Hi all,
>
> I represent R users vs. IT dept. at my workplace (yes, an enviable task :)
>
> We've managed to get a workable network-based R application, for people who
> work remotely, or don't have a machine (i.e., they use a VDI terminal).
> Everything in this organization is staunchly Windows and Microsoft.
>
> We've agreed to upgrade only once yearly, to save IT resources. Now we're
> upgrading, and I would like users to be able to keep their old R 3.1.0
> directory trees like it's available on a single Windows machine. At least
> for a few months, so that people can evaluate back-compatibility if they
> need. In fact, we have an even older server-based 3.0.1, which happens to
> be  the only R version for which the Tableaux-R connection works (at least
> according to my colleagues, I don't use Tableaux).
>
> Anyway, long story short. That was just the motivating example. The problem
> I'm dealing with is whether a network application that has several versions
> of R (3.1.z, 3.2.z), etc., all available, and each reading and installing
> libraries to a different folder tree.
>
> The libraries right now are installed into each user's "personal" share
> drive. It's pretty stable. However, obviously the 3.2.z libraries will now
> overwrite the 3.1.z.
>
> My IT contact says it's impossible, because the Windows app name is always
> just Rgui.exe, and they can only have one set of instructions associated
> with the same app name (i.e., what folders to go to, etc.)

If you are doing the install, you can rename Rgui.exe to something else, 
e.g. rename the old one to Rgui31.exe.

The default setup already installs user packages into a local directory 
with a versioned name, so that shouldn't be a problem.
See ?R_LIBS for details on that.

Duncan Murdoch
>
> I wonder whether anyone has had experience with this, or I should just give
> up and alert people that if they want to explore various historical layers
> of R and the associated packages, they will have to work around and/or
> install and uninstall lots of packages each time.
>
> Thanks!
>
> Assaf
>



More information about the R-help mailing list