[R-SIG-Mac] Multiple R versions: set Current via symbolic link: permissions?

[Prof Brian Ripley]

On 25/01/2012 01:22, Steve Lianoglou wrote:
> Hi Peter,
>
> On Tue, Jan 24, 2012 at 8:06 PM, Peter Carr<pcarr at broadinstitute.org>  wrote:
>> I, too, would like to run multiple versions of R, via command line, on Mac
>> OS X. I have looked in vain for a clear explanation of why there can be one
>> and only one 'Current' version of R running at a time.

That is simply untrue, and would be crippling for the R developers if it 
were true.

Assuming you have multiple R versions installed, all you need to do is 
to invoke their 'R' scripts directly.   There is a problem caused by the 
way Simon installs these scripts in frameworks, so e.g.

/Library/Frameworks/R.framework/Versions/2.13/Resources/bin/R

starts

#!/bin/sh
# Shell wrapper for R executable.

R_HOME_DIR=/Library/Frameworks/R.framework/Resources

but all you have to do is make a copy and change that to e.g.

R_HOME_DIR=/Library/Frameworks/R.framework/Versions/2.13/Resources

[Which is in the thread mentioned.]

And of course, no one said you have to use that particular build of R, 
and for serious users of command-line R a personal build will be more 
convenient (just like any other platform).

If you have packages which use compiled code they are tied (on OS X 
only) to a particular 2.x.* set of R versions, but you can overcome that 
easily enough by using different library directories for different 
versions of R.

> You can dig up some info from this thread:
>
> https://stat.ethz.ch/pipermail/r-sig-mac/2011-September/008582.html
>
> I believe the take home message is that if you want to do this
> (running different R versions *simultaneously*), you can do some
> symlink mojo (you'll find links to examples in that thread) that will
> let you do so, and (for the most part), it will work.
>
> The problem arises when you have to install R packages that need be
> source compiled. This limitation holds if we are talking about running
> R versions that were released as binaries on CRAN (I'm not sure if
> this is true when you compile R as a framework build yourself (but my
> guess is that it will still be a problem in that scenario, as well)).
>
> I can link you straight to the email that says that (w/ more details),
> but you'd probably better be served by just going through the entirety
> of that thread.
>
> Hope that helps,
> -steve
>


-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595