FR: use_test_helper()

[maelle]

I really enjoy using use_test() and use_r(), but was wondering whether there could be an equivalent for creating a tests/testthat/helper- file? (I didn't find any related issue but might have missed one 😅)

[olivroy]

I agree. I think it may be a little bit less straightfoward than the other 2 (use_r() and use_test())

Should there be an option for a generic test helper "tests/testthat/helper.R", vs file specific test helper?

Also, important to note that R/helper.R and tests/testthat/helper.R may not be related.

[olivroy]

I made a little sketch. I don't really like that the default isn't consistent with the default of use_test(), but also I think that having a helper file for each test file by default seems a bit overkill.

usethis/R/r.R

Lines 57 to 71 in fdb4a72

	#' @rdname use_r
	#' @export
	use_test <- function(name = NULL, open = rlang::is_interactive()) {
	if (!uses_testthat()) {
	use_testthat_impl()
	}
	path <- path("tests", "testthat", paste0("test-", compute_name(name)))
	if (!file_exists(path)) {
	use_template("test-example-2.1.R", save_as = path)
	}
	edit_file(proj_path(path), open = open)
	invisible(TRUE)
	}

#' @rdname use_r
#' @export
#' @param scope if global, creates "tests/testthat/helper.R", otherwise creates
#'   "tests/testthat/helper-{name}.R
use_test_helper <- function(name = NULL, scope = c("global", "file"), open = rlang::is_interactive()) {
  if (!uses_testthat()) {
    use_testthat_impl()
  }
  scope <- arg_match(scope)

  if (scope == "global") {
    path <- path("tests", "testthat", "helper", ext = "R")
  } else {
    path <- path("tests", "testthat", paste0("helper-", compute_name(name)))
  }

  edit_file(proj_path(path), open = open)
  ui_bullets(c(
    "_" = "Run {.run devtools::load_all()} to load objects from helper files in
         your environment."
  )
  )
  invisible(TRUE)
}

[jennybc]

I'm generally on board, so am happy to see this in PR form. I'm not sure the name and scope arguments are quite right yet. I don't have great vocabulary for stuff yet, but clearly we want to make it easy to ask for helper.R versus helper-stuff.R. I think 1 argument is enough. "Suffix" almost captures it, but it's not perfect 🤔

[jennybc]

Starting to actually think about this now, including the PRs #1978 and #1990.

Some high-level thoughts (informed by a quick consult with @hadley too):

• Test helper files don't have the same tight coupling as R/whatever.R and tests/testthat/test-whatever.R. We think of a test helper file as belonging to the package not to a specific R or test file. Which is why the default is a plain old tests/testthat/helper.R.
• So I definitely don't see the need to support the file creation/opening functionality that use_r()/use_test() provide for managing the pair of files.
• In terms of what that means for use_test_helper(), that just leaves the initial creation of a helper file. Which might not really justify adding a new function 🤔

[olivroy]

it would also be useful to access it to add new functions to it.

I would understand that it may not be worth a new function, but typing edit_file("tests/testthat/helper.R") is a bit clumsy.

In terms of promoting good workflow, it would be a + too.,

I find test helpers are life savers for efficient testing workflows (as they are loaded with load_all(), but they are not used that much in the wild.

But otherwise, feel free to close and I will just add this to my .Rprofile/

use_test_helper <- function() usethis::edit_file("tests/testthat/helper.R")

Looking at it, it's true that it doesn't seem worth the effort. Maybe adding this example to the documentation of edit_file() would be a good idea?

[jennybc]

I'd be happy to review a small PR with a small function that helps to create (and open) a test helper file, in the right place, with a good name (defaulting to "tests/testthat/helper.R"). The only added feature could perhaps be a way to create "tests/testthat/helper-foo.R", if it's the kind of package with multiple helper files. Let's leave this one available for tidyverse dev day. Ah, I see there is already a PR in flight. Will comment on that, then.