Why using with() as function in a map() call does not work in this example?

Question

library(tidyverse)

formulas <- list(
  mpg ~ disp,
  mpg ~ I(1 / disp),
  mpg ~ disp + wt,
  mpg ~ I(1 / disp) + wt
)

# this works
map(formulas, ~ {lm(.x, mtcars)}) 

# this doesn't
map(formulas, ~ {with(mtcars, lm(.x))}) 

 Error in eval(predvars, data, env) : object 'disp' not found 

Working through the exercises in https://adv-r.hadley.nz/functionals.html#exercises-28, I tried to solve exercise number 6, by trying to evaluate lm() inside mtcars environment with with(), but it throws an error.

Why the last call doesn't work?

Answer 1

It is the environment issue. One option would be quote the components so that it would not be executed

formulas <- list(
  quote(mpg ~ disp),
  quote(mpg ~ I(1 / disp)),
   quote(mpg ~ disp + wt),
  quote(mpg ~ I(1 / disp) + wt)
  )

out1 <- map(formulas, ~ with(mtcars, lm(eval(.x))))
out1
#[[1]]

#Call:
#lm(formula = eval(.x))

#Coefficients:
#(Intercept)         disp  
#   29.59985     -0.04122  


#[[2]]

#Call:
#lm(formula = eval(.x))

#Coefficients:
#(Intercept)    I(1/disp)  
#      10.75      1557.67  


#[[3]]

#Call:
#lm(formula = eval(.x))

#Coefficients:
#(Intercept)         disp           wt  
#   34.96055     -0.01772     -3.35083  


#[[4]]

#Call:
#lm(formula = eval(.x))

#Coefficients:
#(Intercept)    I(1/disp)           wt  
#     19.024     1142.560       -1.798  

It should also work with the first method

out2 <- map(formulas, ~ lm(.x, mtcars))

There would be slight changes in the attributes and in the call, but if that is ignored,

out1[[1]]$call <- out2[[1]]$call
all.equal(out1[[1]], out2[[1]], check.attributes = FALSE)
#[1] TRUE