FR: Introduce negative variants for condition expectations?

[IndrajeetPatil]

The current approach of using NA to test that a condition should not be observed is slightly confusing (at least for me) because it doesn't clearly convey the intent. For example, using expect_error(..., NA) to test that you don't expect an error.

But I am possibly not alone in thinking this way. A cursory search restricted only to GitHub reveals that many developers seem to be using custom negative variants (expect_no_error(), expect_no_warning(), expect_no_message()) to test that the code does not throw errors, warnings, or messages.

So, it will be nice if {testthat} natively supported these functions as an additional way to test for absence of errors, warnings, and messages.

Happy to make a PR if you think this is worth supporting.

#' @export
#' @rdname expect_error
expect_no_error <- function(object) {
  expect_error({{ object }}, NA)
}

#' @export
#' @rdname expect_error
expect_no_warning <- function(object) {
  expect_warning({{ object }}, NA)
}

#' @export
#' @rdname expect_error
expect_no_message <- function(object) {
  expect_message({{ object }}, NA)
}

[IndrajeetPatil]

cc @MichaelChirico Related to a discussion we had on {lintr}.

[hadley]

I think using NA as a sentinel for missing, i.e. expect that errors are missing is already adequate and don't currently have plans to add these specific negations.

[IndrajeetPatil]

Related: #1679