[R-pkg-devel] splitting an R package into two

[RAFFELSBERGER Wolfgang (IGBMC)]

As complement to Duncan's nice answer I also suggest to use the function .Deprecated() for an additional transition time.
(see eg https://stackoverflow.com/questions/44622054/mark-a-function-as-deprecated-in-customised-r-package)

Baseline, if other people/packages use your package, alert them (via a message/warning due to .Deprecated()) and give them time to adopt...

Wolfgang Raffelsberger


----- Mail original -----
De: "Duncan Murdoch" <murdoch.duncan using gmail.com>
À: "John P. Nolan" <jpnolan using american.edu>, "r-package-devel" <r-package-devel using r-project.org>
Envoyé: Vendredi 3 Février 2023 18:07:05
Objet: Re: [R-pkg-devel] splitting an R package into two

On 03/02/2023 12:00 p.m., John P. Nolan wrote:
> Dear R-package-devel,
> 
> I am updating package mvmesh that is on CRAN.  I would like to split that package into two parts, mvmesh (multivariate meshes/grids) and mvhist (multivariate histograms). Does CRAN have a policy/practice on this?   I did  a search, but found nothing on this.
> 
> I would do this in two steps:
> 
>    1.  Extract some of the functions in the current package into a new package.   Submit that package as mvhist and wait for approval.
>    2.  Remove the extracted functions from the current package, and submit it with a new version number.
> There is a short term problem that some functions would exist in two places during the time period it takes to do step 2.  After that, users who want to use the functions in mvhist will have install both packages.
> 
> I don't want to add work for the CRAN team, so please let me know if this is forbidden or there is a better way.

I think your plan is good, though it is missing one step that should 
happen between 1 and 2:

Contact maintainers of all packages that use mvmesh and tell them what 
changes they need to make.  According to the CRAN page for mvmesh, 
that's just two packages, both of which you maintain:  gensphere, 
SphericalCubature, but you may know of others (e.g. on Bioconductor or 
elsewhere) who should be informed.

If those are the only two, then the step should be to send any necessary 
updates to them before proceeding to step 2.  When you do get to step 2, 
mention what steps you took to inform mvmesh users.

Duncan Murdoch

______________________________________________
R-package-devel using r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-package-devel