crossing() adds missing factor levels (either a bug or a documentation issue)

[billdenney]

When using crossing() with a factor that does not have all levels present, it inserts the missing levels. My guess is that this is an intentional feature and the documentation should clarify it. When I was using crossing() the expansion of missing levels was unexpected since it added values that were not in the combination of inputs.

The observed behavior is below, and I expected the second behavior in both cases.

I think that the documentation clarification that would have made the behavior expected would be the following:

Here, add "and the nesting() and crossing() helpers" after "complete()":

tidyr/R/expand.R

Lines 36 to 38 in ad62841

	#' When used with factors, [expand()] and [complete()] use the full set of
	#' levels, not just those that appear in the data. If you want to use only the
	#' values seen in the data, use `forcats::fct_drop()`.

And, in the expand_grid() documentation, point back to the fact that it uses expand(). Perhaps adding something like the following to the details section of the docs of expand_grid(): "expand_grid() uses expand() to generate all combinations.

library(tidyr)

# factor levels are expanded
crossing(a = c("A", "B"), b = factor(c("a", "b"), levels = c("a", "b", "c")))
#> # A tibble: 6 × 2
#>   a     b    
#>   <chr> <fct>
#> 1 A     a    
#> 2 A     b    
#> 3 A     c    
#> 4 B     a    
#> 5 B     b    
#> 6 B     c
# factor levels are not expanded
crossing(
  a = c("A", "B"),
  data.frame(b = factor(c("a", "b"), levels = c("a", "b", "c")))
)
#> # A tibble: 4 × 2
#>   a     b    
#>   <chr> <fct>
#> 1 A     a    
#> 2 A     b    
#> 3 B     a    
#> 4 B     b

^{Created on 2023-10-28 with reprex v2.0.2}

If helpful, I can make the documentation PR.