The only celebrity thing happening this weekend is someone getting a new bike.

I see. Revising 2026 goal to 12,000 miles?

Really glad you got your bike now. Better late than never.

Venables and Ripley, "S Programming", Springer, 2000 (!!) just call it 'function' in the discussion of `tapply` leading to `sapply` and `lapply`, and warns about care needed when using internal functions. With S 😉

I have not yet looked at `rv` but we cover in the FAQ for #r2u what works / does not work with e.g. `renv`. In short, the way the apt repo is _currently_ built does not support dated snapshot but that is _not_ an apt limitation. With sufficient resources (I am a one-man team here) this could work.

Have you seen #r2u? I have written a few posts (some with video) showing e.g.

docker run --rm -ti rocker/r2u

followed by

install.packages(c("tidyverse", "devtools"))

(to take your example) takes care of all system packages along with the R packages via the apt integration. Try it!
#rstats

If a file is checked in, committed, or ... is a git repo property.

Whether a package passes R CMD check is not related to that.

As said, I have been using the very workflow you describe in order to develop and test locally just fine for many years now -- usually directly out of git repo.

I understood your question as asking "why can I not test from my local git repo directory with staged files etc" and I gave you an answer that describes one way of doing just that. No need to copy the git repo, no need to go to ci/cd (that often comes later). FWIW the RStudio IDE does it similarly.

You can run `rcmdcheck::rcmdcheck()`. Note that (as the thread points out) there are fundamental R issues here: you really want a tar.gz created, and then run the test on the tar.gz. And that works: I have checked hundreds of packages directly off `git` for many years that way.