Using dplyr mutate_at with custom function

Question

I want to take two variables from a table and divide them by a third variable and add these computations as two new columns. The mutate_at gets me very close but within the custom function, f below, I want to access another column in the data set. Any suggestions or alternate tidy tools approaches?

library(dplyr)
# this works fine but is NOT what I want
f <- function(fld){
  fld/5
}

# This IS what I want where wt is a field in the data
f <- function(fld){
  fld/wt
}

mutate_at(mtcars, .vars = vars(mpg, cyl), .funs = funs(xyz = f))

# This works but is pretty clumsy
f <- function(fld, dat) fld/dat$wt
mutate_at(mtcars, .vars = vars(mpg, cyl), .funs = funs(xyz = f(., mtcars)))

# This is closer but still it would be better if the function allowed the dataset to be submitted to the function without restating the name of the dataset

f <- function(fld, second){
  fld/second
}

mutate_at(mtcars, .vars = vars(mpg, cyl), .funs = funs(xyz = f(., wt)))

Answer 1

Updated version for dplyr 1.0.6:

mtcars %>% 
  mutate(across(c(mpg, cyl), ~ . / wt))

Or this, which may be slower

mtcars %>% 
  mutate(across(c(mpg, cyl), `/`, wt))

Previous answer:

library(tidyverse)
f <- function(num, denom) num/denom

mtcars %>% 
  mutate_at(vars(mpg, cyl), f, denom = quote(wt))

Although in this specific example, a custom function isn't needed.

mtcars %>% 
  mutate_at(vars(mpg, cyl), `/`, quote(wt))

Answer 2

Maybe something like this?

f <- function(fld,var){
    fld/var
}

mtcars %>%
    mutate_at(vars(mpg,cyl), .funs = funs(xyz = f(.,wt)))

Edit (2020-08-24):

As of second semester of 2020, with the launch of dplyr 1.0.0, mutate_at has been superseded by combining mutate with the across function:

mtcars %>%
    mutate(across(c(mpg, cyl), ~ f(.,wt), .names = "{col}_xyz"))

Answer 3

Why not simply

mutate(mtcars, mpg2 = mpg / wt, cyl2 = cyl / wt)

Answer 4

There is a cur_data() function that will help make the mutate_at() call more compact because you will not have to specify a second argument to the function that is being applied to each column:

f <- function(fld){
  fld / cur_data()$wt
}
mutate_at(mtcars, .vars=vars(mpg, cyl), .funs=funs(xyz = f))

Additional notes:

1. If you need the function to reference a grouping variable, use cur_data_all()
2. mutate_at is now superseded by mutate(.data, across()), so it would be better to do

mtcars %>% mutate(across(.cols=c(mpg, cyl), .fns=f, .names='{.col}_xyz'))