Can two R packages be on each others depends list?

Question

Let's say I have 3 packages A, B, and C.

B connects to data repository 1 and has functions specific to that API.

C connects to data repository 2 and has functions specific to that API.

Eventually there will be several more child packages.

Package A will have generic methods and other common functions (e.g. authentication) that apply to data acquired through B and C. The rationale here is that this would be a more streamlined way to keep up with development (e.g. one would have to update a single auth function rather than doing that inside each child package). So it makes sense for A to be on the depends list for B and C

But I would also like users to just install A and have access to all child packages. For this, I want B and C to be on its depends list.

Is this possible? Should I have a better workflow?

Crandale

Yes. That's what "depends" field is there for in the DESCRIPTION file. cran.r-project.org/doc/manuals/R-exts.html#The-DESCRIPTION-file – Biological

The proper way is to use Suggests: B C in A and Depends: A in both A and B since A doesn't require the others but they do require A. – Uninspired

@SimonUrbanek Thank you. That is helpful. To clarify, since A handles all common functions, it would great if users just had A and used get_data(data_source="B") or get_data(data_source=all). We could do away with child packages but right now it is necessary for async development. – Crandale

Answer 1

The gregmisc package is one example like this that installs more focused sub-packages through depends. It doesn't have any functionality, itself, but it certainly could.

http://cran.r-project.org/web/packages/gregmisc

Hawsepipe

Thanks, the perfect example I was looking for (I use those packages all the time but didn't notice the set up). – Crandale

AFAIK the gregmisc example is not relevant to the question above, because the subpackages don't depend on gregmisc. – Uninspired

Good point. But I do appreciate it because it does give me a precedent for how this workflow might progress. That is, Package A can be equivalent to the original gregmisc, then split up in the future if necessary. – Crandale

Answer 2

Updated in 2021: this is not possible anymore. R RMD Check will complain and return an error - packages cannot depend on each other.

The gregmisc package mentioned in a previous answer was removed from CRAN (probably because of this reason).

Tucket

Thanks for that, i thought i was going crazy! Do you by any chance have a link to where this change was announced/described? – Incommensurate

No, I'm not sure when this was introduced. Looked up the NEWS file, couldn't find it there neither... – Tucket