`arrange(-x)` not working as I expect with `integer64` class

[moodymudskipper]

I get integer64 columns when collecting from a postgresql databsewith dplyr/dbplyr, arrange(x) and arrange(desc(x)) work as expected on these columns, but arrange(-x) doesn't.

See below :

library(dplyr,warn.conflicts = FALSE)
df1 <- data.frame(x = letters[1:3], y = c(1,3,2))

identical(arrange(df1, -y), arrange(df1, desc(y)))
#> [1] TRUE

df1$y <- bit64::as.integer64(df1$y)
identical(arrange(df1, -y), arrange(df1, desc(y)))
#> [1] FALSE

# not what I expect
arrange(df1, -y)
#>   x y
#> 1 a 1
#> 2 b 3
#> 3 c 2

^{Created on 2019-05-16 by the reprex package (v0.2.1)}

[moodymudskipper]

Another issue with integer64, with tidyr but these might stem from the same issue so I'll put it here. When using spread, instead of NA we find the value 9218868437227407266:

library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
data.frame(id = c(1,2,2), key = c("a","a","b"), value = 1:3) %>% 
  mutate(value = bit64::as.integer64(value)) %>%
  tidyr::spread(key, value)
#>   id a                   b
#> 1  1 1 9218868437227407266
#> 2  2 2                   3

^{Created on 2019-05-16 by the reprex package (v0.2.1)}

[romainfrancois]

Eventually (after r-lib/vctrs#348, for version 0.9.*) we should be able to use vec_order(). I'll put a workaround in for 0.8.2

[lock]

This old issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with reprex) and link to this issue. https://reprex.tidyverse.org/