`slice()` shuffles datetime attributes

[yjunechoe]

I'm not entirely sure what's going on here but slice() appears to shuffle the order of class and tzone attributes of datetime columns:

library(dplyr)
df <- data.frame(datetime = lubridate::now(tzone = "UTC"))

attributes(df$datetime)
#> $class
#> [1] "POSIXct" "POSIXt" 
#> 
#> $tzone
#> [1] "UTC"

attributes(slice(df, n())$datetime)
#> $tzone
#> [1] "UTC"
#> 
#> $class
#> [1] "POSIXct" "POSIXt"

This seems specific to dplyr functions that touch rows: so select() and mutate() doesn't do this but filter() and arrange() does.

[DavisVaughan]

At the lowest level this is probably vctrs, but I don't think you should rely on the exact ordering of attributes. Attributes are generally treated like a map, not like a sorted list.

x <- lubridate::now(tzone = "UTC")
attributes(x)
#> $class
#> [1] "POSIXct" "POSIXt" 
#> 
#> $tzone
#> [1] "UTC"

x <- vctrs::vec_slice(x, 1)
attributes(x)
#> $tzone
#> [1] "UTC"
#> 
#> $class
#> [1] "POSIXct" "POSIXt"

^{Created on 2025-09-08 with reprex v2.1.1}