Package option for `wait_for_*()` timeouts

[gadenbuie]

If you're using a lot of $wait_for_js() calls, 30 seconds is a massively long timeout, and it'd be nice to be able to set a global option for the default timeout, e.g.

options(shinytest2.wait_for_js.timeout = 5 * 1000)

instead of having to include timeout = 5 * 1000 in all of my calls to $wait_for_js(). I would imagine heavy users of other wait_for_*() functions would benefit from similar options.

An alternative would be to lower the timeout for all of the wait functions to something reasonably small, e.g. 5 or 10 seconds, and then users who know they need to wait longer can opt into a longer wait time. (And I guess you'd also use longer timeouts for app startup.)

[schloerke]

Would it be fair if all timeouts (other than to $new()) be lowered to 3 seconds?

Current timeout values:

• $new(load_timeout = 10 * 1000) (locally); $new(load_timeout = 20 * 1000) (CI)
• $set_inputs(timeout_= 3 * 1000) (same timeout for$click(input=))
• $upload_file(timeout_ = 3 * 1000)
• $expect_js(timeout = 15 * 1000)
• $get_js(timeout = 15 * 1000)
• $run_js(timeout = 15 * 1000)
• $wait_for_idle(timeout = 30 * 1000)
• $wait_for_value(timeout = 15 * 1000)
• $wait_for_js(timeout = 30 * 1000)

---

And/Or should we adopt a bunch of option for the default values?

• timeout - many methods above
• duration - $wait_for_idle()
• interval - $wait_for_value(), $wait_for_js()
• height/width - Init window size
• variant - Init variant
• expect_values_screenshot_args - Init value

[gadenbuie]

3 or 4 seconds for all but the $new() method seems reasonable to me, as does making this setting controllable by a single default value. Cypress uses 4 seconds and a global timeout option.

[schloerke]

I believe there is legacy standard of 3 seconds with {shinytest}. So I'm ok with keeping it at 3 seconds.

What about a single value with individual overrides? Similar to ggplot2 x axis text inherits from ggplot2 text, we could have shinytest2.timeout default to 3 * 1000 (the global default) andshinytest2.get_js.timeout would use shinytest2.timeout if shinytest2.get_js.timeout is not set.

Or should we just do a single global option and can use args otherwise?

[gadenbuie]

I think I would be most likely to either

1. Know there's a particular spot where I need longer timeouts where I adjust the argument directly
2. Or just globally increase the timeouts because everything is slow.

So I'm in favor of a single default. Also withr::local_options() or withr::with_options() would be a great way to scope increased timeouts to a particular test or part of a test.

[schloerke]

Sold on single timeout value with an example of withr::with_options()!

---

Last question... Should we add options for duration, interval, height/width, variant, expect_values_screenshot_args?

[schloerke]

Also need a shinytest2.threshold value for #231

[wch]

Another possibility:

app$initalize(timeout = st2_timeout(steps = 10))

app$getjs(timeout = st2_timeout())


st2_timeout <- function(steps = 1) {
  factor * getOption("shinytest.timeout.stepsize", {
    if (Sys.getenv("CI")) {
      6*1000
    } else {
      3 * 1000
    }
  })
}

(Please forgive the ugly code)

[wch]

Another:

app <- AppDriver$new(timeout_stepsize = 3*1000, timeout = self$get_timeout(10))

app$get_js(timeout = self$get_timeout(1))

app$get_timeout(steps = 1)

# (Not valid code here, but something like this)
app$set_timeout <- function(stepsize = NULL) {
  # If it's NULL, do the following
  if (is.null(stepsize)) {
    private$timeout <- getOption("shinytest.timeout.stepsize", {
      if (Sys.getenv("CI")) {
        6 * 1000
      } else {
        3 * 1000
      }
    })
  } else {
    private$timeout <- 3 * 1000
  }
}