How to create dynamic group_by parameters

user_8614610 0
Viewed 56

I want to summarize a dataframe by groups. The columns to group are defined dynamically in different compilations. An example is here.


gender = c(rep(1, 50), rep(2, 50))
weight = round(rnorm(100, 60, 5), 0)
age = floor(runif(100, min=18, max=30))
my_df <- data.frame(gender, age, weight)

result <- function(
  df = my_df,
  group_by_gender = TRUE,
  group_by_weight = TRUE,
  group_by_age = TRUE
){
  the_result <- df %>%
    group_by(
      gender, #if(group_by_gender, gender, NULL),
      weight, #if(group_by_weight, weight, NULL),
      age #if(group_by_age, age, NULL)
    ) %>%
    summarize(
      count = n()
    )
  return(the_result)
}

Depending on the group_by_* i want to switch on or off none, one or more columns to group by. I do not see, how to construct the string inside group_by(string)

dplyr r group-by
2 Answers

You can create a vector of rlang::expr objects and then subset that to just the values you want to use, and expand that into the group_by with !!!. For example

result <- function(
  df = my_df,
  group_by_gender = TRUE,
  group_by_weight = TRUE,
  group_by_age = TRUE
){
  gcols <- rlang::exprs(gender, weight, age)
  gkeep <- c(group_by_gender, group_by_weight, group_by_age)
  the_result <- df %>%
    group_by(!!!gcols[gkeep]) %>%
    summarize(
      count = n()
    )
  return(the_result)
}
user_23720640

You could invoke the .dots parameter within the group_by function:

result <- function(df = my_df, group_by_gender = TRUE, 
                   group_by_weight = TRUE, group_by_age = TRUE){
  nms <- c("gender","weight", "age")
  bool <- c(group_by_gender, group_by_weight, group_by_age)
  
  the_result <- df %>%
    group_by(.dots = nms[bool]) %>%
    summarize(count = n(), .groups = "keep")
  return(the_result)
}
user_83802720
Related