`col_vals_equal()` doesn't match two `NA` values about pointblank HOT 3 CLOSED

I think that's consistent with the expectation for col_vals_equal() to use == for checking equality - the following is not TRUE:

NA == NA
#> [1] NA

With regards to na_pass, that's a general argument for whether NA results from the validation operations should count as a pass. As far as col_vals_equal() is concerned, NA == NA and NA == 1 are both NA - the na_pass argument is not aware of how you got to the NA value with ==. So the fact that both rows involving NA values pass in your second example with na_pass = TRUE is again the intended/documented behavior.

For a workaround, you can either recode the NAs into some value that's legible to == in the preconditions argument, as you mention, or use col_vals_expr() instead and be more explicit about the looser criteria for equality. That latter approach can get a bit more verbose, but either of the following lets you match two NAs as a match but an NA with some other value as a mismatch:

col_vals_expr(
  expr = expr(Vectorize(identical)(a, b))
)

col_vals_expr(
  expr = expr(!xor(is.na(a), is.na(b)) & a == b)
)

from pointblank.

Thanks so much for this explanation! I appreciate it.

I'll give the preconditions argument and the col_vals_expr() options a try.

from pointblank.

@yjunechoe thank you for the explanation :) and I think we can close this one.

from pointblank.