Re: Relation of 'Depends' and 'Imports' from R/CRAN to 'Depends' in 'debian/control'
On Tue, Jan 24, 2012 at 05:05:24PM -0500, Carlos Borroto wrote:
> Hi All,
> I would like to work on packages from R/CRAN and Bioconductor.
Great. I checked that you are a member of Debian Med packaging team and
thus you should have commit permissions to SVN where most of our R
packages reside (some of them are in Git as your example r-bioc-limma.)
So I would advise to commit your current state of work to the repository
of your choice even in a not finished state - just mark this by using
"UNRELEASED" as target distribution and perhaps add a TODO entry inside
> using as a guide r-cran-rocr and r-bioc-limma. I was wondering if
> there is a easy translation from 'Depends' and 'Imports' in the CRAN
> packages to the 'Depends' fields in 'debian/control'.
I'm not really sure whether there is a 1:1 relation. What I personally
do when building R packages is the following: Use
to be pretty sure that all Build-Depends are fullfilled. Once you
have all Build-Depends set this is usually the set of Depends you need
> Initially I though CRAN's 'Depends' meant, the package is needed for
> building and CRAN's 'Imports' meant, the package is needed when using.
> Which would translate to a 'Depends' entry on 'Source' and 'Package'
> respectively. This seems incorrect as for example when building
> reshape2, it only has 'Imports' dependencies, it fails with this
> dh_installdirs usr/lib/R/site-library
> if test -f /usr/bin/xvfb-run; then \
> xvfb-run -a \
> R CMD INSTALL -l
> --clean \
> . ; \
> else \
> R CMD INSTALL -l
> --clean . ;\
> ERROR: dependencies 'plyr', 'stringr', 'lattice' are not available for
> package 'reshape2'
> So my question is, should I list in 'Source' and 'Package' all the
> dependencies listed in CRAN's 'Depends' and 'Imports'?
I'm not very competent in R itself but if the build process asks for
those packages, just add them. This can not be wrong.
Kind regards and thanks for considering building such packages in Debian