[R] Reverse dependency checking

[Duncan Murdoch]

On 2025-02-12 4:42 p.m., Lenth, Russell V via R-help wrote:
> Dear R-Help,
> 
> When I submit an update to one of my packages, I decided to try to avoid having to fix errors that sometimes occur in CRAN's reverse-dependency checks by performing the same checks ahead of time. From what I can tell, the way to do this is to use
> 
>      tools::check_packages_in_dir(..., reverse = list(which = "all"), ...)
> 
> It does not surprise me that this takes a lot of time! One of my packages has quite a few reverse suggests, so this process installs all packages that my package or any of the reverse depends requires, which is about a zillion. But eventually, this works. I've learned to run this in batch mode so it doesn't tie up Rgui or RStudio. That's fine; I just have a script that I can run.
> 
> I have two questions:
> 
> 1. While it is installing all those packages, I see a lot of chatter so I can see what progress is being made, however glacial. But when it gets the the phase where it checks each of the reverse-dependent packages, it is utterly silent. It would be really helpful to me if check_packages_in_dir() would just print a one-line message that it is checking such-and-such package.

There are several re-implementations of that code.  Jeroen pointed you 
to one, I think devtools has another.  You should try those.
> 
> 2. Is there any way to streamline the checking? It seems to me that since all the packages are already on CRAN and thus pass most checks, all that we need to check for reverse dependencies are the examples, vignettes, and tests. This could save a lot of time.

I think there can't be.  You may have dropped some export, or changed 
its behaviour.  Even suggested packages can be used anywhere (with tests 
that they are installed), so your changes could conceivably have bad 
effects in lots of strange ways. There are probably a few of the checks 
that could be dropped, but not most of them.

Duncan