[R-sig-QA] wiki update, and documenting R development processes

[Kurt Hornik]

>>>>> Jim Garrett writes:

> Wiki update:  Warner Engl has added some regulatory guidelines in the
> "FDARegulations" area (which is now not FDA-specific!).  I've just entered
> a lengthy review of the British guidance he contributed.  Please feel free
> to edit my review with your own comments; the wiki is at

> http://software.biostat.washington.edu/wikis/rvalid/FrontPage

> I've only read closely this one document, but it's already evident
> that written procedures are going to be very important.  Corporations
> using R will be looking for a formal "quality system" including a
> document control system, and documents specifying the development
> process (Standard Operating Procedures).  I know that the R team *has*
> some standard procedures.  Are there documents specifying these
> procedures?

Not yet ...

I will eventually need to write a paper on the current QC tools which
should also shed some light on exisiting QA processes.  In a nutshell, R
Core members are asked to commit code changes only provided that these
passed make check-devel [or make check-all] on their local system.  Of
course, this does not mean a lot if one does not know which tests these
check suites contain.

One also needs to be careful about what it meant by 'R'.  You may find
some of my comments on the current two-tier development model in

@InProceedings{hornik:Hornik:2002,
  author       = {Kurt Hornik},
  title        = {Tools and Strategies for Managing Software Library
                  Repositories},
  booktitle    = {Statistics in an Era of Technological Change,
                  Proceedings of the 2002 Joint Statistical Meetings},
  pages        = {1490--1493},
  year         = 2002,
}

of interest.

This, in a rather narrative way, describes the standard processes on the
CRAN main repository side of R.

> I would like to invite anyone who knows, especially members of the
> development team, to find any such documents that may exist, and if
> documents describing standard processes don't exist, could someone
> please contribute descriptions of these processes?

best
-k