OK, turns out it's not so bad! We can skip biocLite and install packages directly, same as from CRAN. Draft changes here:
https://github.com/NixOS/nixpkgs/compare/master...jefdaj:bioc-pullreq Opinions? I considered splitting off a separate bioc-modules folder, but then I noticed the majority of "broken" R packages are just missing Bioconductor dependencies and decided the two aren't really separate. Plus it would be a lot of duplicated code, some of which I don't understand well enough to maintain. There might be conflicts between the Bioconductor and CRAN release schedules. I think a unified set of packages could require holding back from upgrading to the latest CRAN releases on occasion, so if that's a big problem maybe we do need two sets of packages (current CRAN, previous CRAN + current Bioconductor). Anyone familiar enough with Bioconductor to say for sure? Should I go ahead and make a pull request? I've attached the bash script I made to list remaining dependencies. After I ran it I had to go through and manually correct some things it doesn't pick up, like packages that exist but are known not to build. Jeff On Sun, 8 Mar 2015 12:00:27 -0700 Jeffrey David Johnson <jef...@gmail.com> wrote: > Do any of you guys/gals use R and Bioconductor? > > The only page I can find related to setting it up is > <https://nixos.org/wiki/R>. > Bioconductor is a whole additional ecosystem on top of R/CRAN, > with its own installer. You normally use it like: > > source("http://bioconductor.org/biocLite.R") > biocLite("PackageName") > > They explain why at <http://bioconductor.org/install/#why-biocLite>. > How hard would it be to package that using Nix? > Thanks > Jeff
list-still-broken.sh
Description: Binary data
_______________________________________________ nix-dev mailing list nix-dev@lists.science.uu.nl http://lists.science.uu.nl/mailman/listinfo/nix-dev