What can I say, g-cran works and everything from here on is just
improvement over what existed already. I got the base of the project
like I wanted it, stable and reliable, although the R package
installer doesn't do some specific things quite how I'd like it to.
So the current code simply calls R to build and install packages. A
compatibility ebuild uses the package name to find the R package,
which is then fed to R. I must say that although this is a bit
hackish, it works very nicely.
Should you want to give this code a spin, the ebuild is in the
gentoo-science overlay. Once you've installed g-cran, create a new
overlay and cd into it, then:
#echo "cran" >> profiles/repo_name
#g-cran . sync http://cran.r-project.org
#g-cran . generate-tree
Make your favorite package manager read this new tree and you should
be able to install some packages. Soon to come is proper parsing of
the License field, proper translation of dependencies (the current
code assumes a dependency on foo means a dependency on dev-R/foo) and
installing /usr/share/doc files correctly. Specifically the first and
the last I'll be focusing on for this week, but not unimportantly,
development on g-common, which as you may remember is the piece of
software I'll hopefully be writing with wiktor_b and iElectric to
organize non-ebuild package installations, will continue next week.
Exams will also be over, so after this week I will be spending loads
of time making dependencies parse correctly and package managers
handle g-cran fluently.
For the next three weeks or so, I'll be able to entertain myself
finishing the important parts of g-cran and writing g-common. After
that, I'd really like to do some QA, and perhaps write some
package-fixing scripts, since a lot of R packages have incorrect
As I'm writing this, some roommates are already yelling our national
anthem, so right now I'm off to watch football ("Kijk, de bal moet
minimaal tussen die twee palen.").