dbplyr across() behavior differs from dplyr in a grouped context

[lschneiderbauer]

When using across() within a grouped context, everything() only selects all non-grouped variables. This is also documented in the code example of group_cols():

# Remove the grouping variables from mutate selections:
gdf %>% mutate_at(vars(-group_cols()), `/`, 100)
# -> No longer necessary with across()
gdf %>% mutate(across(everything(), ~ . / 100))

And not surprisingly that is also what happens:

library(dplyr)
#> 
#> Attache Paket: 'dplyr'
#> Die folgenden Objekte sind maskiert von 'package:stats':
#> 
#>     filter, lag
#> Die folgenden Objekte sind maskiert von 'package:base':
#> 
#>     intersect, setdiff, setequal, union

df <-
  tibble(
    id = c(1),
    y = c("a"),
    z = c("a")
  )

df |> 
  summarize(
    across(everything(), \(x) "test"),
    .by = id
  )
#> # A tibble: 1 × 3
#>      id y     z    
#>   <dbl> <chr> <chr>
#> 1     1 test  test

^{Created on 2024-04-30 with reprex v2.1.0}

When using the exact same verbs with the dbplyr backend, everthing() really seems to select every column instead:

library(dplyr)
#> 
#> Attache Paket: 'dplyr'
#> Die folgenden Objekte sind maskiert von 'package:stats':
#> 
#>     filter, lag
#> Die folgenden Objekte sind maskiert von 'package:base':
#> 
#>     intersect, setdiff, setequal, union
library(dbplyr)
#> 
#> Attache Paket: 'dbplyr'
#> Die folgenden Objekte sind maskiert von 'package:dplyr':
#> 
#>     ident, sql

df <-
  tibble(
    id = c(1),
    y = c("a"),
    z = c("a")
  )


tbl_lazy(df, dbplyr::simulate_dbi()) |> 
  summarize(
    across(everything(), \(x) "test"),
    .by = id
  )
#> <SQL>
#> SELECT 'test' AS `id`, 'test' AS `y`, 'test' AS `z`
#> FROM `df`
#> GROUP BY `id`

^{Created on 2024-04-30 with reprex v2.1.0}

I expect this to result in

#> <SQL>
#> SELECT `id`, 'test' AS `y`, 'test' AS `z`
#> FROM `df`
#> GROUP BY `id`

[mgirlich]

Thanks for reporting. dbplyr handled groups in across() already correctly but not for summarise(.by).