Calling teardown_env() when testing Setup.R

[ralmond]

I'm trying to test my Setup.R file by sourcing the file. However, in that context, teardown_env() is not properly initialized so I get an error.

Here is a simple example:

handle <- file(tempfile())
withr::defer(close(handle),testthat::teardown_env())
#> Error in `testthat::teardown_env()`:
#> ! `teardown_env()` has not been initialized
#> ℹ This is an internal error that was detected in the testthat package.
#>   Please report it at <https://github.com/r-lib/testthat/issues> with a reprex (<https://tidyverse.org/help/>) and the full backtrace.

^{Created on 2024-11-22 with reprex v2.1.1}

Here is the backtrace:

Backtrace:
    ▆
 1. ├─base::source("~/ralmond1/Projects/Peanut/tests/testthat/setup.R")
 2. │ ├─base::withVisible(eval(ei, envir))
 3. │ └─base::eval(ei, envir)
 4. │   └─base::eval(ei, envir)
 5. ├─withr::defer(...) at tests/testthat/setup.R:16:1
 6. └─testthat::teardown_env()

I don't think this is a bug so much as a bit of missing functionality. It would be good to have a way to artificially call Setup.R and then emulate the teardown, so that I can validate that my Setup code is working.

[ralmond]

Okay, I found the workaround. I need to replace the call to teardown_env() in my Setup.R code with a reference to .GlobalEnv. Then I can call withr::deferred_run() to test my cleanup code.

It would be cool if there was an option which would make this available without tweaking my code.

Say:

teardown_env <- function () 
{
    if (is.null(the$teardown_env)) {
        if (isTRUE(getOption("testthat_test_setup"))) {
            message("Run 'withr::deferred_run()' to simulate teardown.")
            return(.GlobalEnv)
        }
        abort("`teardown_env()` has not been initialized", .internal = TRUE)
    }
    the$teardown_env
}

Then, setting options(testthat_test_setup=TRUE) would allow sourcing the Setup.R file during testing.