R devel - Jul 2006 - DESCRIPTION and PACKAGES files.

Greetings; I'm working on translating the heirarchy of R packages into
the Gentoo Ebuild environment, currently working from a 2.3.1 install.

There are several data which it would be nice to see provided
alongside the existing output of available.packages; at the moment I
think I'm going to have to get License and SystemRequirements out of
individual package DESCRIPTIONs.

More generally, it'd be nice to see all the documented fields from
DESCRIPTION files in the output from available.packages(), or at least
collated in PACKAGES.

I have a suggestion to accomplish this which I've hacked up locally
and am happy to build a patch if the suggestion is well-recieved.

available.packages uses read.dcf on the PACKAGES file, accepting only 

flds <- c("Package", "Version", "Priority",
"Bundle", "Depends",
  "Imports", "Suggests", "Contains")


If we could optionally accept all of the fields which are mentioned in
'Creating R Extensions', someone who was interested in the broader
data could get at it.

PACKAGES would grow, substantially, but wouldn't be harder to
maintain; in fact PACKAGES could be just a naive concatenation of
everyone's DESCRIPTION files.


For that matter, if you would be willing to add the extra data to
PACKAGES, the current implementation would work, and I could just run
my hacked version and get at what I wanted.  But I know of at least
one other package-heirarchy maintainer who might like this, and
subsuming it in the base function would be nice.



- Allen S. Rout

On Thu, 20 Jul 2006, Allen S. Rout wrote:
> 
> Greetings; I'm working on translating the heirarchy of R packages into
> the Gentoo Ebuild environment, currently working from a 2.3.1 install.
> 
> There are several data which it would be nice to see provided
> alongside the existing output of available.packages; at the moment I
> think I'm going to have to get License and SystemRequirements out of
> individual package DESCRIPTIONs.
> 
> More generally, it'd be nice to see all the documented fields from
> DESCRIPTION files in the output from available.packages(), or at least
> collated in PACKAGES.
> 
> I have a suggestion to accomplish this which I've hacked up locally
> and am happy to build a patch if the suggestion is well-recieved.
> 
> available.packages uses read.dcf on the PACKAGES file, accepting only 
> 
> flds <- c("Package", "Version",
"Priority", "Bundle", "Depends",
>   "Imports", "Suggests", "Contains")
> 
> 
> If we could optionally accept all of the fields which are mentioned in
> 'Creating R Extensions', someone who was interested in the broader
> data could get at it.
> 
> PACKAGES would grow, substantially, but wouldn't be harder to
> maintain; in fact PACKAGES could be just a naive concatenation of
> everyone's DESCRIPTION files.
And that is the reason: everyone who downloads pays the price for unneeded 
fields in that file.  It was a deliberate decision to minimize download 
times, which are already substantial for people on a dialup connection.

-- 
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

Prof Brian Ripley <ripley at stats.ox.ac.uk> writes:
> And that is the reason: everyone who downloads pays the price for unneeded 
> fields in that file.  It was a deliberate decision to minimize download 
> times, which are already substantial for people on a dialup connection.
OK, that meets the definition of "not well recieved" ;) 

Thanks for your time. 



- Allen S. Rout