Continuous Integration (Hudson/Jenkins) and R Programming

I do not see any particular problem. These things tend to be scripted so could just

• point to the top of your repository
• N minutes after each checkin, loop over source directories
• invoke R CMD check on each
• your package has to be set to use unit test, for which you can use
  • RUnit which is the initial unit testing for R ; it is widely used
  • testthat which is a newer package by Hadley, and used by many of his packages
  • svUnit by Philippe which AFAIK never caught on quite as much as the other two.

That is really not any different from continuous integration with compiled languages. Your question is really about how to do unit testing within R, and that question has been covered before.

at office we have been using Hudson/Jenkins for quite a while. I have contributed integration of svUnit to Jenkins so I would strongly advise you to use svUnit before trying anything else.

have a look at the two libraries I maintain: logging and delftfews or at myself trying to follow zoo and redistribute it on github.

I have not been doing housekeeping recently so the three scripts (in zoo, logging, delftfews) are all slightly different. the one in my zoo version will stop if any test fails. this is practical when you are running R CMD check, but probably less of a good idea when doing continuous integration.

Hudson/Jenkins supports running bash scripts. I think that you can use this as an entrance to your R world. In R, a simple way to keep result is to use sink("toYourFile.txt") and then use CI's result display function to show the toYourFile.txt.