[R] installing binary packages across windows cluster

Prof Brian Ripley ripley at stats.ox.ac.uk
Sat Dec 10 10:41:24 CET 2011


On 10/12/2011 03:54, Benjamin Tyner wrote:
> Hi
>
> Given a "cluster" of identical windows machines all running the same
> version of R, are there any circumstances where using install.packages()
> to install a package (say, one that doesn't have any dependencies) on
> just a single machine, then copying the resulting installed package
> subdirectory across the cluster into all the other machine's .Library
> subdirectories would be a bad idea?

As the rw-FAQ points out, installing to .Library is not a very good idea.

Installing a binary package on Windows just downloads and unpacks the 
files, so you would be effectively be doing that on each machine. 
However, since this is Windows, there are many ways in which 'copying' 
could screw things up (permissions, for example).  I would be tempted to 
download the .zip file and unpack it on each machine.

OTOH, what my department does is a better idea: that is to install 
packages onto a network-mounted file system accessible to every Windows 
machine in the department.  That way we know that users get the same R 
environment everywhere.

-- 
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



More information about the R-help mailing list