[R] slow R start up

[Prof Brian Ripley]

On Mon, 7 Nov 2005, Stuart Macgregor wrote:

> Hi,
>   I want to use R within a unix shell script where I repeatedly open
> and close R (doing some computation within R whilst it is open). i.e.
> something like
>
> #!/bin/sh
> R --vanilla << EOF
> # some R commands
> EOF
> # some unix commands
> R --vanilla << EOF
> # some R commands
> EOF
> # some unix commands
> ...etc
>
> the problem is that since I open and close R many times, the shell
> script takes a long time to run (R takes ~1 second to start up, even on
> a fast machine; I've tried various linux installations and versions of R
> from 1.7 to 2.2). Is there some way of making R start up faster? I've
> tried altering the memory start options but this makes little
> difference.

[I don't think your machine _is_ 'fast'!  I am seeing 0.3s for a complete 
R session on an Opteron 250.  We have worked hard to reduce the start-up 
time to those levels over the period you quote: please do not belittle 
those efforts.  R is fast (several times at least) compared to all other 
software I use of similar complexity.]

On my 2-year-old i386 box a complete empty session takes 0.5s.  You can 
reduce that a lot by using fewer default packages.  For example, if you 
don't need methods or graphics or datasets, 
R_DEFAULT_PACKAGES="utils,stats" reduces that to 0.3s, and 
R_DEFAULT_PACKAGES=NULL reduces it to 0.05s.  We use the latter when using 
R for scripting (e.g. in R CMD INSTALL).

-- 
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