[Bioc-devel] help with package dependencies on packages in remote repositories

Seth Falcon sfalcon at fhcrc.org
Fri Sep 15 16:46:45 CEST 2006


Hi Joern,

Joern Toedling <toedling at ebi.ac.uk> writes:
> In its vignette, my package "macat" requires a large data package,
> called "stjudem", that I keep in a local repository. I wrote a small
> function that checks if that data package is installed, if not it
> retrieves it from the repository, and then loads the data package. This
> function is only used in the package's vignette, but seems to work
> nicely. However, I get the following two warnings:
>
> 1.)
> * checking package dependencies ... WARNING
> 'library' or 'require' calls not declared from:
>   mydatapkg

Not sure.  If it is the case that R CMD check gives a warning for any
call to library or require in package code, then there may be nothing
we can do other than making noise on R-devel and asking for that to be
changed.

> 2.)
> * checking package vignettes in 'inst/doc' ... WARNING
> * Installing *source* package 'stjudem' ...

You may be able to prevent this warning if you can muffle the output
caused by installing the package.  IIRC, this is hard to do.  But you
may be able to do it with capture.output.

That said, we have a new facility for handling experiment data
packages.  If you are willing to have the data package hosted here:
http://bioconductor.org/packages/1.9/data/experiment/

Then you can solve both problems by putting the data package in
Suggests.  The build system will install it.  An advantage is that it
will be easier for end users to install what is needed to work through
the vignette.

If this is of interest, please send Herve an email (hpages at fhcrc).

> I believe both of these warning message are totally unnecessary. Listing
> the data package "stjudem" under Suggests does not work, since it is
> kept in local repository and therefore not found, resulting in a check
> error. Using "suppressWarnings" when using the retrieval function does
> also not do the trick.

I tend to agree that these messages are not all that helpful :-\

Bw,

+ seth



More information about the Bioc-devel mailing list