withr::local_dir() vs withr::defer ... set_wd()

[jonthegeek]

usethis/tests/testthat/helper.R

Lines 69 to 76 in 467ff57

	withr::defer(
	{
	ui_done("Restoring original working directory: {ui_path(old_wd)}")
	setwd(old_wd)
	},
	envir = env
	)
	setwd(proj_get())

Does this pre-date withr::local_dir(), or is there a reason it's done more explicitly here? I copied this concept for a package-creating package, without noticing that withr::local_dir() exists and is cleaner.

usethis/tests/testthat/helper.R

Line 38 in 467ff57

old_wd <- getwd() # not necessarily same as `old_project`

is also related, in case it's fine to update this to local_dir. I can PR if you'd like.

[jennybc]

I think it's that I wanted to schedule that ui_done() to occur at the same time as restoring the working directory.

Developing create_local_thing() was a gradual and somewhat painful process. And, yeah, a lot of ideas and maybe even tools have matured since that got written. So it's believable and that I would write it differently if I started right now. But I plan to leave it alone until/unless there's some other reason to fiddle with it.

[jonthegeek]

That makes sense! FWIW, I switched it out in my version, and the tests remained happy, and this also stayed happy:

set_state_inspector(function() {
  list(
    wd = getwd(),
    proj = usethis::proj_get()
  )
})

[jennybc]

Nice! So I think local_dir() could be used here, in terms of functionality, but I'm not going to touch this load-bearing helper until there is some other reason to do so.

[jonthegeek]

I really appreciate the modeling of "know when to say no" behavior! Thanks for everything 😊