Unintuitive syntax for selecting non-numeric columns (Placement of "!")

[mohammadsav]

Description: I expect the following code to select all non numeric columns in a tibble (It doesn't work):

df %>%
    select(where(!is.numeric))

The code that works is the following:

df %>%
    select(!where(is.numeric))

I feel like the first one is the more appropriate syntax (correct me if I'm wrong). I read the 2 code blocks as:

1. In df, select where columns are not numeric.
2. In df, select not where columns are numeric.

Below is a reprex output

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

n <- 100
df <- tibble(
  Sport = c(rep("Baseball",n/2), rep("Basketball",n/2)),
  Ranking = runif(n,min=1,max=10),
  Date = sample(seq(as.Date("2023-01-01"), as.Date("2023-12-31"), by="days"), n)
)
sapply(df,class)
#>       Sport     Ranking        Date 
#> "character"   "numeric"      "Date"

# This works, selects numeric only
df %>%
  select(where(is.numeric))
#> # A tibble: 100 × 1
#>    Ranking
#>      <dbl>
#>  1    4.72
#>  2    8.16
#>  3    2.15
#>  4    6.44
#>  5    6.57
#>  6    1.74
#>  7    7.75
#>  8    4.26
#>  9    4.27
#> 10    8.69
#> # ℹ 90 more rows

# This works, selects non-numeric only
df %>%
  select(!where(is.numeric))
#> # A tibble: 100 × 2
#>    Sport    Date      
#>    <chr>    <date>    
#>  1 Baseball 2023-08-19
#>  2 Baseball 2023-02-23
#>  3 Baseball 2023-03-30
#>  4 Baseball 2023-01-08
#>  5 Baseball 2023-08-04
#>  6 Baseball 2023-01-03
#>  7 Baseball 2023-01-19
#>  8 Baseball 2023-10-04
#>  9 Baseball 2023-09-20
#> 10 Baseball 2023-06-02
#> # ℹ 90 more rows
# This does not work
df %>%
  select(where(!is.numeric)) # <------------------------- The Issue
#> Error in `select()`:
#> ! Problem while evaluating `where(!is.numeric)`.
#> Caused by error in `!is.numeric`:
#> ! invalid argument type
#> Backtrace:
#>      ▆
#>   1. ├─df %>% select(where(!is.numeric))
#>   2. ├─dplyr::select(., where(!is.numeric))
#>   3. ├─dplyr:::select.data.frame(., where(!is.numeric))
#>   4. │ └─tidyselect::eval_select(expr(c(...)), data = .data, error_call = error_call)
#>   5. │   └─tidyselect:::eval_select_impl(...)
#>   6. │     ├─tidyselect:::with_subscript_errors(...)
#>   7. │     │ └─rlang::try_fetch(...)
#>   8. │     │   └─base::withCallingHandlers(...)
#>   9. │     └─tidyselect:::vars_select_eval(...)
#>  10. │       └─tidyselect:::walk_data_tree(expr, data_mask, context_mask)
#>  11. │         └─tidyselect:::eval_c(expr, data_mask, context_mask)
#>  12. │           └─tidyselect:::reduce_sels(node, data_mask, context_mask, init = init)
#>  13. │             └─tidyselect:::walk_data_tree(new, data_mask, context_mask)
#>  14. │               └─tidyselect:::eval_context(expr, context_mask, call = error_call)
#>  15. │                 ├─tidyselect:::with_chained_errors(...)
#>  16. │                 │ └─rlang::try_fetch(...)
#>  17. │                 │   ├─base::tryCatch(...)
#>  18. │                 │   │ └─base (local) tryCatchList(expr, classes, parentenv, handlers)
#>  19. │                 │   │   └─base (local) tryCatchOne(expr, names, parentenv, handlers[[1L]])
#>  20. │                 │   │     └─base (local) doTryCatch(return(expr), name, parentenv, handler)
#>  21. │                 │   └─base::withCallingHandlers(...)
#>  22. │                 └─rlang::eval_tidy(as_quosure(expr, env), context_mask)
#>  23. ├─tidyselect::where(!is.numeric)
#>  24. │ └─rlang::as_function(fn)
#>  25. │   └─rlang::is_function(x)
#>  26. └─base::.handleSimpleError(`<fn>`, "invalid argument type", base::quote(!is.numeric))
#>  27.   └─rlang (local) h(simpleError(msg, call))
#>  28.     └─handlers[[1L]](cnd)
#>  29.       └─rlang::abort(msg, call = call, parent = cnd)

^{Created on 2023-10-26 with reprex v2.0.2}

[philibe]

df %>% select(where(Negate(is.numeric))) works also.

[mohammadsav]

Thanks @philibe