[Rd] Package dependencies and imports

[John Fox]

Dear Brian,

--------------------------------
John Fox
Department of Sociology
McMaster University
Hamilton, Ontario
Canada L8S 4M4
905-525-9140x23604
http://socserv.mcmaster.ca/jfox 
-------------------------------- 

> -----Original Message-----
> From: r-devel-bounces at r-project.org 
> [mailto:r-devel-bounces at r-project.org] On Behalf Of Prof Brian Ripley
> Sent: Friday, September 22, 2006 3:55 AM
> To: Martin Maechler
> Cc: r-devel at stat.math.ethz.ch
> Subject: Re: [Rd] Package dependencies and imports
> 
> On Fri, 22 Sep 2006, Martin Maechler wrote:
> 
> >>>>>> "Seth" == Seth Falcon <sfalcon at fhcrc.org>
> >>>>>>     on Thu, 21 Sep 2006 10:33:06 -0700 writes:
> >
> >    Seth> Martin Maechler <maechler at stat.math.ethz.ch> writes:
> >    >>  I think you should use the new DESCRIPTION field and say
> >    >> 'Enhances: rgl'
> >    >>
> >    >> but I haven't checked if this prevents the warning,
> >    >> though I think it should..
> >
> >    Seth> But is Rcmdr enhancing rgl or being enhanced by rgl?
> >
> >    Seth> :-( :-(
> >
> > I think this distinction is not so easily made on a general level.
> > I'd say an argument for how you can (should?) think about 
> this is the 
> > following abstraction:
> >
> > There are two packages, A and B.
> > A's DESCRIPTION has 'Enhances: B'.
> > When B is there, A provides extra features.
> > In one sense, this always enhances B ("enhance" in the 
> usual English 
> > meaning), since B's area of application has been enlarged 
> by code in 
> > A.
>
> Yes, but that is just as true of the raft of packages that 
> Rcmdr has in 'Suggests'.  I suggested to John, and still 
> think, that 'Suggests: rgl' is right.  Rcmdr will work 
> without it, but works better with it and so if you are 
> installing Rcmdr you would like to install rgl as well (but 
> it fails to install it would not be a major loss).
> 

Although I may have lost track of the exact sequence, I believe that I
originally had rgl under Suggests:, but when rgl is absent from a system,
that causes Rcmdr to fail R CMD check. I'd prefer to have rgl under
Suggests:, since then I wouldn't have to treat it as a special case and it
would be installed automatically by install.packages("Rcmdr",
dependencies=TRUE).

Regards,
 John