where() within across() no longer works as expected

[nannerhammix]

It appears that across(where(is.)),...) no longer works to convert columns based on the underlying type.

library(dplyr, warn.conflicts = FALSE)
library(dbplyr, warn.conflicts = FALSE)

temp <- memdb_frame(x = 1:10)

temp
#> # Source:   table<dbplyr_001> [10 x 1]
#> # Database: sqlite 3.33.0 [:memory:]
#>        x
#>    <int>
#>  1     1
#>  2     2
#>  3     3
#>  4     4
#>  5     5
#>  6     6
#>  7     7
#>  8     8
#>  9     9
#> 10    10

temp %>%
  mutate(across(where(is.integer), as.character)) %>%
  sql_render()
#> <SQL> SELECT *
#> FROM `dbplyr_001`

While the expected generated query would be:

SELECT CAST(`x` AS TEXT) AS `x`
FROM `dbplyr_001`

[mgirlich]

Predicates never really worked and already produced an error when used in select(). This error was missing for across() and is now added by #1170.