On Thu, Jul 30, 2015 at 1:23 AM Fotis Georgatos <[email protected]> wrote:
> I would lobby for a R/3.2.2 build (release to come out mid August), > incorporating as much collective wisdom as possible, perhaps tuned for bio > needs. > What would you think? A PR upon 3.2.1 could be a first approximation of it. I settled with the following approach (which to me is agnostic to the field of research): These are the default things of an R stack: Rstack.eb: # this is just drafted but I hope you get the idea version = "3.2.0" # right now creating 3.2.1 depends = ( "R" # no packages at all not even recommended "Rprofile" # just sets R_PROFILE, R_ENVIRON "Rlib" # site-wide libraries that should be available also holds dependencies for stuff like GLPK, MySQL, Postgres..... ) Rprofile and Rlib are separated because I found that there are more often than not settings to be changed for users and this keeps reinstallation times way down since I don't have to rebuild the whole library. Rlib is basically an empty easybuild with a bunch of dependencies and a directory "site-library" -- I guess it could be external but the dependency resolution mechanism of EB is just nice to have and e.g. have stuff like protocol buffers in "the right place". This keeps R itself rather clean but still enable the use of a "batteries included" R install. I'm still thinking about how to do the same with Python. I'd like to have a bare python and one (or maybe a few) libraries that can then be loaded. As opposed to "full fledged python install" or "one module per package". /Martin -- -- http://www.xing.com/profile/Martin_Marcher http://www.linkedin.com/in/martinmarcher Mobil: +43 / 660 / 62 45 103 UID: ATU68801424
