crossing() ignores NA values

[huftis]

crossing() seems to ignore NA values. This is inconvenient and not consistent with expand.grid(). Reprex:

library(tidyr)
crossing(num = 1:2, char = c("A", "B", NA))
#> # A tibble: 4 x 2
#>     num  char
#>   <int> <chr>
#> 1     1     A
#> 2     1     B
#> 3     2     A
#> 4     2     B

expand.grid(num = 1:2, char = c("A", "B", NA))
#>   num char
#> 1   1    A
#> 2   2    A
#> 3   1    B
#> 4   2    B
#> 5   1 <NA>
#> 6   2 <NA>

Other ‘special’ values work fine:

crossing(num = 1:2, char = c("A", "B", NaN))
#> # A tibble: 6 x 2
#>     num  char
#>   <int> <chr>
#> 1     1     A
#> 2     1     B
#> 3     1   NaN
#> 4     2     A
#> 5     2     B
#> 6     2   NaN
crossing(num = 1:2, char = c("A", "B", Inf))
#> # A tibble: 6 x 2
#>     num  char
#>   <int> <chr>
#> 1     1     A
#> 2     1     B
#> 3     1   Inf
#> 4     2     A
#> 5     2     B
#> 6     2   Inf

[hadley]

Root cause is:

ulevels(c("A", "B", NA))
#> [1] "A" "B"

[hadley]

Root cause of that is

sort(c("A", "B", NA))
#> [1] "A" "B"