dplyr has changed the behavior of na_matches

[karldw]

Thanks for the useful package! I noticed that the na_matches configuration can be a little funny. By default, dplyr doesn't set it anymore, and if it's set, dplyr raises a warning.

Here's an example:

library(safejoin)

x <- safe_left_join(mtcars, mtcars, check="")
#> Error: `na_matches` must be a character vector.

library(dplyr, quietly = TRUE, warn.conflicts = FALSE)
x <- safe_left_join(mtcars, mtcars, check="")
#> Error: `na_matches` must be a character vector.

pkgconfig::set_config("dplyr::na_matches" = "na")

x <- safe_left_join(mtcars, mtcars, check="")
#> Warning: `dplyr::na_matches` pkgconfig options is now ignored.
#> * Please set `na_matches` directly.

Here's the relevant dplyr code: https://github.com/tidyverse/dplyr/blob/0bea3e80a876cceb9e179ef3a64d94103c350c41/R/join.r#L390

[moodymudskipper]

Thanks a lot for reporting, this is inconvenient indeed! I'll fix this asap

[karldw]

Would it help if I submitted a PR? The fix I'm imagining is set the na_matches default to c("na", "never") and add a na_matches <- match.arg(na_matches) call in the safe_join function.

[moodymudskipper]

Thanks Karl, it was just that indeed, with the equivalent update in eat, it should work now!

[karldw]

That works perfectly, thanks!