CODEWITHSUNDEEP
rpurrr
Joe
Joe

Reputation: 3796

passing columns into purrr::walk2()

I am trying to save a series of models using purrr's walk() functions, and am getting the following error:

"Error in map2(.x, .y, .f, ...) : object 'model' not found"

library(dplyr)
library(tidyr)
library(purrr)

mt_models <- 
    mtcars %>% 
    group_by(cyl) %>% 
    nest() %>% 
    mutate(
        model = map(.x = data, .f = ~lm(mpg ~ wt, data = .x)),
        file_name = paste("model", cyl, "cyl.rda", sep = "_")
    ) 

mt_models %>% walk2(.x = model, .y = file_name, .f = ~save(.x, file = .y))

I can successfully save the models using this code below:

walk2(.x = mt_models$model, .y = mt_models$file_name, .f = ~save(.x, file = .y))

But I am trying to understand why model is not passing into walk2() in the first example.

Upvotes: 1

Views: 832

Answers (2)

IceCreamToucan
IceCreamToucan

Reputation: 28685

You can use with to provide an environment in which to search for variables

mt_models %>% 
  with(walk2(.x = model, .y = file_name, .f = ~save(.x, file = .y)))

Upvotes: 2

akrun
akrun

Reputation: 887128

Outside the mutate/summarise and other tidyverse function, we need to do .$ to extract the column 

library(dplyr)
library(purrr)
mt_models %>% {
    walk2(.x = .$model, .y = .$file_name, .f = ~save(.x, file = .y))
     }

Upvotes: 1

Related Questions