I'm having issues with dplyr::mutate(). It works fine when I use it regularly, but it throws an error when I use it in a function and then try to define that function.

My goal is to create a function that lets you calculate the quartiles of a distribution for a whole data set, as well as for grouped subsets of that data set.

For example:

This works fine

dat <- mtcars

test_table <- dat %>%
  bind_rows(mutate(., cyl = "all")) %>%
  group_by(cyl) %>% 
  summarise(mpg_q25 = quantile(mpg, prob = .25),
            mpg_q50 = quantile(mpg, prob = .50),
            mpg_q75 = quantile(mpg, prob = .75),
            count = n())

test_table

Output

# A tibble: 4 × 5
  cyl   mpg_q25 mpg_q50 mpg_q75 count
  <chr>   <dbl>   <dbl>   <dbl> <int>
1 4        22.8    26      30.4    11
2 6        18.6    19.7    21       7
3 8        14.4    15.2    16.2    14
4 all      15.4    19.2    22.8    32

However, this does not

mpg_table <- function(df, grouping_var, val) {
  bind_rows(mutate(., {{grouping_var}} = "all")) %>%
    group_by({{grouping_var}}) %>%
    summarise(mpgq25 = quantile({{mpg}}, prob = .25),
              mpgq50 = quantile({{mpg}}, prob = .50),
              mpgq75 = quantile({{mpg}}, prob = .75),
              count = n())
}

mpg_table(dat, cyl, mpg)

Output from trying to define the function

Error: unexpected '=' in:
"mpg_table <- function(df, grouping_var, val) {
  bind_rows(mutate(., {{grouping_var}} ="

Anyone have any idea what's going wrong here? Thank you!

CodePudding user response：

How about this:

library(dplyr)
library(glue)
data(mtcars)
dat <- mtcars
mpg_table <- function(df, grouping_var, val) {
  df %>% 
    mutate({{grouping_var}} := as.character({{grouping_var}})) %>% 
    bind_rows(mutate(., {{grouping_var}} := "all")) %>%
    group_by({{grouping_var}}) %>%
    summarise("{{val}}q25" := quantile({{val}}, prob = .25),
              "{{val}}q50" := quantile({{val}}, prob = .50),
              "{{val}}q75" := quantile({{val}}, prob = .75),
              count = n())
}

mpg_table(dat, cyl, mpg)
#> # A tibble: 4 × 5
#>   cyl   mpgq25 mpgq50 mpgq75 count
#>   <chr>  <dbl>  <dbl>  <dbl> <int>
#> 1 4       22.8   26     30.4    11
#> 2 6       18.6   19.7   21       7
#> 3 8       14.4   15.2   16.2    14
#> 4 all     15.4   19.2   22.8    32

^{Created on 2022-09-29 by the reprex package (v2.0.1)}

The := allows you to pass a variable in as the name of a new variable to be created. I also used the same construct for the variable names for the quantiles. This means that if you pass drat as val for example, you would get dratq25, dratq50 and dratq75 as the variables in the output.

The other problem you run into is a format problem. The cyl variable is numeric and you're trying to bind it to a data frame whose cyl variable is a character. The first step in the code above changes the grouping_var to character to avoid this problem.