his is a query for suggestions on how best to setup the package for
coxme. This should be ready for CRAN in about a week - I'm working out
documentation details and some last test cases.
The R CMD check and INSTALL processes work great on my computer. The
rub is that they depend on noweb. Both algorithmicly and statistically
this is the most complex bit of code I have done, and after some reading
I decided to give the "literal programming" approach a try. It turns
out to be one of the better ideas I've ever tried.
The current package directory (on r-forge.r-progject.org) has a
configure script of
(cd R; make fun)
which causes most of the R files to be built from the parent .Rnw files
in inst/noweb. The cleanup script is (cd R; make clean) to remove them.
This follows the Rforge suggestion that derived files not be under
version control. Sweave is not able to extract files from the Rnw source
since there are cascading definitions: chunk <<a>> references chunk
<<b>> defined at a later (or earlier) point in the document.
I don't have to retain "autosubmit" from Rforge. Perhaps the best
I can
do is to make a local copy, create the .R files there, delete the config
and cleanup scripts, and submit that? Any ideas are appreciated.
Terry Therneau
