drop columns with nothing using @mutate?

[mistermichaelll]

More @mutate testing fun 😄 this is more of a behavior question rather than a bug.

In dplyr, you can drop columns using mutate by passing NULL. This is news to me, and I actually only learned this from tests from the source code.

For example:

library(dplyr)
df <- tibble(x = 1, y = 2)
df |> mutate(y = NULL)

#> # A tibble: 1 × 1
#>       x
#>   <dbl>
#> 1     1

In Tidier, I'd imagine that the closest behavior would be something like this:

using Tidier
using DataFrames
df = DataFrame(x = 1:3, y = 1:3)
@mutate(df, y = nothing)

But this doesn't drop the column, rather it just replaces y with a column of type nothing

3×2 DataFrame
 Row │ x      y       
     │ Int64  Nothing 
─────┼────────────────
   1 │     1          
   2 │     2          
   3 │     3          

Is dropping a column this way something that should be supported?

[kdpsingh]

Thanks @mistermichaelll for surfacing this, and apologies for the delay in getting back to you.

I don't think we should support y = nothing as dropping the column. This is because of a fundamental difference between how nothing behaves in Julia and how NULL behaves in R.

In R, if you write x = c(1, NULL, 3), the result will be c(1, 3), whereas in Julia, x = [1, nothing, 3] will produce [1, nothing, 3]. Since nothing is a valid value, I don't think we should treat it the same as NULL in R.

[mistermichaelll]

Makes total sense! I'll close this issue and remove this specific test from the mutate tests in #117