Fix failing test on windows-devel

[IndrajeetPatil]

── Error (test-convert_to_na.R:90:3): convert_to_na other classes ──────────────
Error in `charToDate(x)`: character string is not in a standard unambiguous format
Backtrace:
     ▆
  1. ├─testthat::expect_message(...) at test-convert_to_na.R:90:2
  2. │ └─testthat:::expect_condition_matching(...)
  3. │   └─testthat:::quasi_capture(...)
  4. │     ├─testthat (local) .capture(...)
  5. │     │ └─base::withCallingHandlers(...)
  6. │     └─rlang::eval_bare(quo_get_expr(.quo), quo_get_env(.quo))
  7. ├─datawizard::convert_to_na(d, na = list(3, "c", TRUE, "2022-01-02"))
  8. └─datawizard:::convert_to_na.data.frame(...) at datawizard/R/convert_to_na.R:43:2
  9.   └─base::lapply(...) at datawizard/R/convert_to_na.R:186:2
 10.     ├─datawizard (local) FUN(X[[i]], ...)
 11.     └─datawizard:::convert_to_na.Date(X[[i]], ...) at datawizard/R/convert_to_na.R:43:2
 12.       ├─base::which(x %in% as.Date(na)) at datawizard/R/convert_to_na.R:137:4
 13.       ├─x %in% as.Date(na) at datawizard/R/convert_to_na.R:137:4
 14.       ├─base::as.Date(na) at datawizard/R/convert_to_na.R:137:4
 15.       └─base::as.Date.character(na)
 16.         └─base (local) charToDate(x)

[etiennebacher]

I think it's because as.Date() works for numerics too in R-devel (see this commit), which breaks this: https://github.com/easystats/datawizard/blob/0a226f69b4b337d8cb8fda1c41b7186024d8f9b7/R/convert_to_na.R#L124-L126

[etiennebacher]

Actually this is the line that causes the issue: https://github.com/easystats/datawizard/blob/0a226f69b4b337d8cb8fda1c41b7186024d8f9b7/R/convert_to_na.R#L137

[etiennebacher]

I don't have R-devel on my laptop so I can't really test this

[IndrajeetPatil]

Given the discussion on R-mailing-list about this change in R-devel, I feel like things are still in a bit of a flux, and so I want to wait for a bit before resolving this and submitting a new version.

[IndrajeetPatil]

Okay, need to fix and submit soon-

Dear maintainer,

Please see the problems shown on https://cran.r-project.org/web/checks/check_results_datawizard.html.

Please correct before 2022-10-27 to safely retain your package on CRAN.

The CRAN Team

[etiennebacher]

No idea how to fix this for now, and as you said it seems that this change is not settled yet on R-devel. If we can't find a solution, one option is to (temporarily) comment this test

[etiennebacher]

I can reproduce with R-devel on Ubuntu. The problem comes from the new behavior of as.Date() that can now convert numerics to dates:

> as.Date(3)
[1] "1970-01-04"

Here's a reprex where this causes an issue:

test <- data.frame(
  num = c(1, 2),
  date = as.Date(c("2022-10-17", "2022-10-18"))
)

convert_to_na(test, na = list(1, "2022-10-17"))

#> Error in `charToDate(x)`: character string is not in a standard unambiguous format

This is because in convert_to_na.Date(), the following lines now return "1" "2022-10-17" instead of just "2022-10-17" as before.

unlist(na[sapply(na, function(i) {
  !is.null(tryCatch(as.Date(i), error = function(e) 
})])

Since this will cause a problem every time there's a numeric in the list specified in na, I propose that we change the accepted input for convert_to_na.Date() to accept only objects of class "Date". Therefore, the example above should work with convert_to_na(test, na = list(1, as.Date("2022-10-17"))).

@strengejacke @IndrajeetPatil what do you think?

[IndrajeetPatil]

Thanks for getting to the bottom of this, @etiennebacher.

I propose that we change the accepted input for convert_to_na.Date() to accept only objects of class "Date".

Yes, this sounds like a sensible solution to me.

[strengejacke]

Agreed.