[R-pkg-devel] possible solution to package-documentation-alias problem

[Daniel Kelley]

# Preamble

This email is to tell other developers what I did to address an issue with
documenting a package.  I'm not sure that I am correct in my approach --
comments would definitely be appreciated -- but at least this email is fairly
concrete about the changes I made. To be honest, I don't know how to test
whether my changes are suitable, since no problem is reported in local builds or
in remote windows checks, and no problem is listed on the CRAN tests page.

# The problem

Today I received multiple emails from K. Hornik, telling me about problems with
the package-level documentation for several CRAN packages that I maintain. Here
is a key part of that email:

    You have file 'oce/man/oce.Rd' with \docType{package}, likely intended as a
    package overview help file, but without the appropriate PKGNAME-package
    \alias as per "Documenting packages" in R-exts.

# Possible solution

As a test (using the 'plan' package, which is much smaller and thus ought to
give faster test results), I changed my Roxygen2 block

    #' @name plan
    #' @docType package
    #' @author Dan Kelley
    NULL

to read

    #' @name plan
    #' @docType package
    #' @author Dan Kelley
    #' @keywords internal
    "_PACKAGE"
    ## usethis namespace: start
    ## usethis namespace: end
    NULL

Note that the two ## comments are likely not required, but I included them
because I saw them at

https://github.com/jonesor/mpmsim/commit/e8d0f0d657ffa24c25ddd3165c7ddcad16322e3d

which I found to be quite a helpful resource.

# Local testing

After rebuilding locally, I can now do

    package?plan

and get the expected documentation for the package as a whole.

# CRAN submission

I submitted the update to CRAN, and it has appeared as a tarball.  It has not
yet appeared in built-up packages sources, but perhaps the fact that I didn't
get any warnings from CRAN suggests that the flaw has been addressed.

# Conclusions

If I am right, a simple fix is all that will be required for many packages.
However, I don't know of any way to know if this fix follows recommended
procedures.  There appear to be multiple ways of addressing the issue.

Perhaps other developers will have better solutions than the one I've outlined
above.  Or, if I happen to have done something right, then perhaps this email
will be of some use to other developers.

Dan Kelley / Department of Oceanography / Dalhousie University / Canada