[Rd] R CMD check and interfacing packages

[Gabor Grothendieck]

On 9/3/05, Uwe Ligges <ligges at statistik.uni-dortmund.de> wrote:
> Gabor Grothendieck wrote:
> 
> > For an R package whose purpose is to interface to other software,
> > since such other software is not necessarily being on CRAN how does one
> > proceed so that the R package can pass 'R CMD check'?  None
> > of the examples or demos in the package can run without the software
> > being interfaced to.
> 
> Yes, this is a problem.
> For this reason, we have some exceptions on CRAN for packages that
> interface to other software products, e.g. for some we are doing only
> fake installations and for others, I am skipping automatical checks for
> Windows, if GUI interaction is required.
> 
> 
> > Is there an example of such a package that could be used as a
> > prototype?
> 
> I don't think there is a globally perfect example.
> It depends on the software you are going to interface to, on the
> Operating Systems this software is available on, and on the license and
> the requirements for linking against it etc....

The interfaced software is GPL software available on all operating systems 
that R runs.   Currently the strategy is to include a Windows binary of the
interfaced software with the R package so that at least Windows users can use 
it without separately installing the package but users on other
operating systems
must separately install that package.  Communication with the package is 
currently via sockets so there is no linking.