[R] Re: Packages and Libraries

[Prof Brian Ripley]

On Wed, 9 Feb 2005, Spencer Graves wrote:

>     The reasons to 'introduce "package()" and deprecate "library()"' may be 
> OBVIOUS to you, but they completely escape me.  Could you please clarify why 
> that's obvious?  I've seen many admonitions on this list that the term is 
> "package" NOT "library", but I don't recall ever seeing any explanation of 
> why the term "package" is more appropriate than "library". 
>     I suspect there may be some rationale that "package" seems more 
> appropriately descriptive.  However, is it so much more precise that it 
> justifies creating a distinction between S-Plus and R?

Note that S(-PLUS) does not use `library' for `package', it uses `library 
section', and that is in the 1988 Blue Book, the manual for the first 
version of S which was extensible in that way.  Only a few observant 
people used `library section', and when S4 introduced `chapter' (almost 
but not quite the same thing) few people adopted that either.

There is a need to change the programmatic interface to what 
library/require do, for example to take character string (only) arguments 
and to return a suitable classed object, as well as separate out 
library(help=). This will almost certainly be done (if/when it is done) so 
that library() remains for ever as a compatibility wrapper, but the new 
interface (usePackage(), use(), whatever) becomes the preferred one.

-- 
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