Anomaly in Factor Order from As_Factor Function

I am trying to use as_factor() to create factor levels in the order where they appear in the underlying data. The function works fine where the underlying data are character, but not where they are numeric.

I tried the sample code given with the as_factor documentation. It uses an underlying character variable, and gives the same order as the underlying variable; that is what as_factor is supposed to do. But for a numeric variable the order is sorted, and as.factor and as_factor give the same order.

# anomaly with as_factor -- is this a feature or a bug?
# Bill Anderson  August 2019
require(tidyverse)
#> Loading required package: tidyverse
#> Warning: package 'tidyverse' was built under R version 3.6.1
#> Warning: package 'dplyr' was built under R version 3.6.1
# example from as_factor documentation
x <- c("a", "z", "g")
as_factor(x) # preserves input order, as desired
#> [1] a z g
#> Levels: a z g
as.factor(x) # factor levels obtained by sorting data
#> [1] a z g
#> Levels: a g z
# numeric example 
y <- c(1, 3, 2)
as_factor(y) # factor levels obtained by sorting data -- not what I expected
#> [1] 1 3 2
#> Levels: 1 2 3
as.factor(y) # factor levels obtained by sorting data
#> [1] 1 3 2
#> Levels: 1 2 3
identical(as_factor(y), as.factor(y))
#> [1] TRUE
# explicit character conversion
z <- as.character(y)
as_factor(z) # preserves input order, as desired
#> [1] 1 3 2
#> Levels: 1 3 2
as.factor(z) # factor levels obtained by sorting data
#> [1] 1 3 2
#> Levels: 1 2 3
# one can also put everything into a data frame, 
# so we can see the impact of the factor order is clearly visible
mtcars %>% group_by(cyl) %>%
  summarize(meandisp = mean(disp)) # cylinder order is sorted
#> # A tibble: 3 x 2
#>     cyl meandisp
#>   <dbl>    <dbl>
#> 1     4     105.
#> 2     6     183.
#> 3     8     353.
mtcars %>% group_by(as_factor(cyl)) %>%
  summarize(meandisp = mean(disp)) # cylinder order is still sorted
#> # A tibble: 3 x 2
#>   `as_factor(cyl)` meandisp
#>   <fct>               <dbl>
#> 1 4                    105.
#> 2 6                    183.
#> 3 8                    353.
mtcars %>% group_by(as_factor(as.character(cyl))) %>%
  summarize(meandisp = mean(disp)) # cylinder order follows data
#> # A tibble: 3 x 2
#>   `as_factor(as.character(cyl))` meandisp
#>   <fct>                             <dbl>
#> 1 6                                  183.
#> 2 4                                  105.
#> 3 8                                  353.

^{Created on 2019-08-15 by the reprex package (v0.3.0)}

There are no error messages. The issue is simply that I don't get the order of the underlying data, unless I explicitly convert to character.

I am not sure if the situation is a feature or a bug. But if it is a feature I suggest that it should show up in the function documentation.