4

I'd like to take a tibble (or dataframe), convert one of the columns to numeric, only select the same column plus a third column, and filter out NAs.

Given the following data:

library(tidyverse)

set.seed(1) 

mytib <- tibble(a = as.character(c(1:5, NA)), 
                b = as.character(c(6:8, NA, 9:10)), 
                c = as.character(sample(x = c(0,1), size = 6, replace = TRUE)))

vars <- c("a", "b")

I have created the following function

convert_tib <- function(var, tib){
  tib <- tib %>% 
    mutate("{var}" = as.numeric({{ var }})) %>%
    dplyr::select({{ var }}, c) %>%
    filter(!is.na({{ var }}))
  return(tib)
}

And run it with purrr:map

map(vars, ~ convert_tib(var = ., tib = mytib))

The output of this code unfortunately does not convert the vector to numeric and it also doesn't filter out the NA. I have tried many different strategies such as ensym(var) and enquo(var) inside the function and leaving out the curly-curly operators.

What I'd like to get is the following:

> map(vars, ~ convert_tib(var = ., tib = mytib))
[[1]]
# A tibble: 5 × 2
  a     c    
  <int> <int>
1 1     0    
2 2     1    
3 3     0    
4 4     0    
5 5     1    

[[2]]
# A tibble: 5 × 2
  b     c    
  <int> <int>
1 6     0    
2 7     1    
3 8     0    
4 9     1    
5 10    0
Anoushiravan R
  • 21,622
  • 3
  • 18
  • 41
Tea Tree
  • 882
  • 11
  • 26

2 Answers2

5

You may try this. I made use of ensym function inside your custom function, since I noticed you would like to specify the variable names as strings. Then I also used !! called big bang operator to unquote it. In the end you also need := to define a custom variable name in place of =:

library(dplyr)
library(rlang)
library(purrr)

convert_tib <- function(var, tib){
  var <- ensym(var)
  tib <- tib %>% 
    dplyr::select(!!var, c) %>%
    mutate(!!var := as.integer(!!var), 
           c = as.integer(c)) %>%
    filter(!is.na(!!var))
  return(tib)
}
      
map(vars, convert_tib, mytib)

The output:

[[1]]
# A tibble: 5 x 2
      a     c
  <int> <int>
1     1     0
2     2     1
3     3     0
4     4     0
5     5     1

[[2]]
# A tibble: 5 x 2
      b     c
  <int> <int>
1     6     0
2     7     1
3     8     0
4     9     1
5    10     0
Anoushiravan R
  • 21,622
  • 3
  • 18
  • 41
4

You can do this without injection or embracing:

library(dplyr)
library(purrr)
    
convert_tib <- function(tib, var) {
  tib %>% 
    transmute(across(c(var, c), as.integer)) %>%
    filter(!is.na(.data[[var]]))
}

map(vars, convert_tib, tib = mytib)

[[1]]
# A tibble: 5 x 2
      a     c
  <int> <int>
1     1     0
2     2     1
3     3     0
4     4     0
5     5     1

[[2]]
# A tibble: 5 x 2
      b     c
  <int> <int>
1     6     0
2     7     1
3     8     0
4     9     1
5    10     0
Ritchie Sacramento
  • 29,890
  • 4
  • 48
  • 56
  • Could you explain a bit more what's going on here? For example, why do we use transmute and why do we call across, even though we are just addressing one column? – Tea Tree Mar 28 '22 at 01:45
  • 3
    Transmute drops any variables that are not part of the computation, and as you only want to keep two variables it means we can use it and skip using `select()`. `across()` is acting on two columns in each iteration (`a`/`b` and `c`) and has the advantage of accepting both bare names and strings as input and handles all the NSE automatically. – Ritchie Sacramento Mar 28 '22 at 02:12