Pass an Object Through a Function

Question

This question discusses the issue, but I need to pass the object through my function. For example:

foo <- function(data, col1, col2) {
  boxplot(
    formula = col1 ~ col2,
    data = data)
}

When I run this:

foo(data, col1, col2)

I get object 'col1' not found.

How do I pass col1 and col2 through?

a formula is an object as well, why can't you do `foo <- function(data, formula = col1 ~ col2) boxplot(formula, data)` as a trivial function? If you must pass variable names, use `reformulate` instead of reconstructing a formula with `as.formula(paste(deparse(substutute(...` — rawr, Sep 06 '20 at 23:22
from your example, it looks like you would need `reformulate('col2', 'col1')` or if you had multiple variables on the right, `reformulate(c('col3', 'col2'), 'col1')` or vectors for one or both would also work — rawr, Sep 06 '20 at 23:32

akrun · Accepted Answer · 2020-09-06T21:43:38.637

In base R, we could construct the formula from unquoted arguments by first converting to string with deparse/substitute and then use paste to create the formula

foo <- function(data, col1, col2) {
  col1 <- deparse(substitute(col1))
  col2 <- deparse(substitute(col2))
  boxplot(
     formula = as.formula(paste0(col1, "~", col2)),
        data = data)
  }

-test it with inbuilt dataset

foo(mtcars, mpg, cyl)

Or if we prefer to use tidyverse, use the curly-curly ({{..}}) operator for simultaneously convert to quosures (enquo) and evaluate (!!)

library(dplyr)
library(ggplot2)
foo2 <- function(data, col1, col2) {
     data %>%
       select({{col1}}, {{col2}}) %>% 
       mutate( {{col2}} := factor({{col2}})) %>%
       ggplot(aes(x = {{col2}}, y = {{col1}})) + 
           geom_boxplot()
           }
       
foo2(mtcars, mpg, cyl)

Pass an Object Through a Function

1 Answers1