Allow functions called by summarise_at to dynamically determine function parameters

[yaboody]

I'm working on a function which dynamically groups by and summarises according to a given set of groups, variables to be summarised, and the functions used to summarise them. All of this can be done via function arguments. However, some functions also have optional parameters (for example na.rm = TRUE|FALSE). Setting the parameter to TRUE or FALSE is trivial, but I would also like to be able to specify that the parameter I want to make use of is "na.rm =".

[DavisVaughan]

summarise_at() is superseded and is no longer getting active development. If you can provide a full code reproducible example of what you are trying to do, we may be able to give you some other advice for it

[yaboody]

Sure, thanks! My code looks like this:

group_fn <- function(frame, cols, vars, aggregator = "sum"){
...
group_by_at(frame, cols) %>% summarise_at(vars, get(aggregator))
}

Here I'm grouping a given dataframe by given columns and summarising a set of variables by a given function. I would also like to be able to apply a given set of parameters to that function (say, na.rm, which would be equal to true). Is there any way to dynamically vary which parameters I want to send to the aggregation function? Not just write out -na.rm- myself and then insert TRUE or FALSE as a function argument of group_fn, but write "na.rm" as a function argument of group_fn and then make use of it.

[DavisVaughan]

I would structure this to use summarise(), across(), and .by, which are modern interfaces for what you are trying to do:

library(dplyr, warn.conflicts = FALSE)

mtcars |>
  summarise(
    across(mpg:cyl, list(sum = \(col) sum(col, na.rm = TRUE))),
    .by = vs
  )
#>   vs mpg_sum cyl_sum
#> 1  0   299.1     134
#> 2  1   343.8      64

group_fn <- function(frame, cols, vars, fns) {
  frame |>
    summarise(
      across({{ vars }}, fns),
      .by = {{ cols }}
    )
}

group_fn(
  mtcars,
  vs,
  mpg:cyl,
  sum
)
#>   vs   mpg cyl
#> 1  0 299.1 134
#> 2  1 343.8  64

group_fn(
  mtcars,
  vs,
  mpg:cyl,
  list(
    sum = \(col) sum(col, na.rm = TRUE),
    mean = \(col) mean(col, na.rm = TRUE)
  )
)
#>   vs mpg_sum mpg_mean cyl_sum cyl_mean
#> 1  0   299.1 16.61667     134 7.444444
#> 2  1   343.8 24.55714      64 4.571429

^{Created on 2025-04-22 with reprex v2.1.1}

[yaboody]

On second thought - I re-read what you wrote and it's pretty useful. If I set up my function to pipe in the entire function like that then it can accept any parameters. I'll look into that. Thanks a bunch for the help!

[DavisVaughan]

Yea I think structuring it so that the arbitrary function is passed as input is the only way to keep maximal flexibility