I think it's worthwhile to have this new helper function, but it would be better to implement it by adding an optional argument ignore_attributes = FALSE to expect_vector_equal(), using that to control whether expect_equivalent() or expect_equal() is used, and then calling that function from this one. That would reduce the amount of copied code.

You'll need to use !! to unquote expr in the call to expect_vector_equivalent() from here.

(Please do push back if you think I'm mistaken about this or if I'm missing some important details!)

Another argument for expect_vector_equal(ignore_attributes = TRUE) is that testthat 3e deprecates expect_equivalent() in favor of expect_equal(ignore_attr = TRUE). And if we can name the argument, I would vote for ignore_attr over ignore_attributes to match testthat (well, I guess technically it's waldo at that point).

Agreed—best to name the attributeignore_attr 👍

Also heads up that expect_vector_equal() calls a function named expect_vector() but this is not testthat::expect_vector(); it's just a simple function defined higher up in helper-expectation.R. That confused me initially.

We can rename our expect_vector if that helps; I believe expect_vector is new in testthat (at least it's new to me).

➕ for renaming our expect_vector to avoid any possible confusion.

In other places I've been bitten by these testing helper name overlaps, and each time it's happened it threw me for a loop trying to figure out what was going wrong until I realized it was a name collision in a helper.

Have now updated the code here to implement ignore_attr. What makes sense as a new name for expect_vector? Maybe expect_as_vector or expect_vector_match?

Gone with expect_as_vector for now!

Is it possible to implement this without dplyr? We only suggest dplyr, and it would be nice to not have to have it installed/loaded for na.omit() to work. This might not be the right approach, but my first thought: would it be possible to iterate objects and use the na.omit() method above? (or use the $Filter method you use above)?

Ah, didn't realise it was only a suggest; I'll have a think!

Refactored now!

Another argument for expect_vector_equal(ignore_attributes = TRUE) is that testthat 3e deprecates expect_equivalent() in favor of expect_equal(ignore_attr = TRUE). And if we can name the argument, I would vote for ignore_attr over ignore_attributes to match testthat (well, I guess technically it's waldo at that point).