Turn variable name into string inside function dplyr

Question

I'm doing something kinda simple, but tidyeval always puzzles me. In this case, I have a function that plots something and I want to also save it afterwards using the name of the column I'm plotting, something like this:

bar_plot= function(table, col_plot){
    ggplot(table, aes(x=region,
                    y= {{col_plot}})) + 
geom_bar(stat = "identity", fill="steelblue") +
     ggsave(glue('results/{col_plot}.png'))
  
}

The plot has no problems, but I'm not able to save it (doesn't find the object, because it isn't reading it as a string). Tried using quo, enquo, sym, and nothing worked. What is the way to turn my variable name into a string inside the function?

For reproducibility this is enough:

df = data.frame(region = c(1, 2), mean_age = c(20, 30))

Thanks !

Do you just need to put ggsave in a new line without the + from the previous line? — Arthur
– Arthur, Commented Mar 15, 2022 at 18:12
That won't solve the variable problem, though, it still won't compute as a string — Juan C
– Juan C, Commented Mar 15, 2022 at 18:15
@danh It saves the plot as 'results/{col_plot}.png' not as 'results/mean_age.png' like OP wants. — Gregor Thomas
– Gregor Thomas, Commented Mar 15, 2022 at 18:26
@Gregor Thomas are you sure? It saves just fine as results/mean_age.png for me. — danh
– danh, Commented Mar 15, 2022 at 18:28
Sorry, I think OP wants to call the function with unquoted names, e.g., bar_plot(df, mean_age). In that case, the ggsave line gives an error object 'mean_age' not found. — Gregor Thomas
– Gregor Thomas, Commented Mar 15, 2022 at 18:31

Allan Cameron · Accepted Answer · 2022-03-15 18:23:50Z

5

You can do this:

bar_plot <- function(table, col_plot) {
  
  p <- ggplot(table, aes(region, {{col_plot}})) + geom_col(fill = "steelblue")
  
  ggsave(paste0('results/', deparse(substitute(col_plot)), '.png'), p)
  
}

bar_plot(df, mean_age)

So you have:

./results/mean_age.png

answered Mar 15, 2022 at 18:23

Allan Cameron

178k7 gold badges70 silver badges118 bronze badges

Sign up to request clarification or add additional context in comments.

4 Comments

Juan C Over a year ago

Exactly what I was looking for, thank you! Was harder than expected. Is there a more straightforward way if my function takes the variable name as string as an argument, instead of the variable name without quotes? can I do something like .data[[col_plot]] in the ggplot call if it's a string?

Allan Cameron Over a year ago

You can use aes_string instead of aes to pass a string

Gregor Thomas Over a year ago

@JuanC yes, since ggplot2 version 3.0 .data[[col_plot]] is a good way to use a string column name in ggplot. aes_string is soft-deprecated. This is also the way its recommended to use ggplot in packages.

Juan C Over a year ago

Thanks to both of you, I'm clear now! I'll use strings, seems more understandable to me

Calum You · Accepted Answer · 2022-03-15 18:39:14Z

3

Abstracting out the plotting part and focusing on the file name, I think you can also use rlang::as_name here to convert the symbol into the string you need.

library(ggplot2)

df <- data.frame(region = c(1, 2), mean_age = c(20, 30))

bar_plot <- function(table, col_plot) {
  # ggplot(table, aes(
  #   x = region,
  #   y = {{ col_plot }}
  # )) +
  #   geom_bar(stat = "identity", fill = "steelblue")
  filename <- glue::glue("results/{rlang::as_name(enquo(col_plot))}.png")
  filename
}

bar_plot(df, mean_age)
#> results/mean_age.png

Note that we need to do two things: first wrap the argument col_plot in enquo, so we get mean_age of instead of literally col_plot. Then convert with as_name() to turn mean_age into "mean-age".

answered Mar 15, 2022 at 18:39

Calum You

15.2k4 gold badges27 silver badges46 bronze badges

2 Comments

Juan C Over a year ago

Thank you! seems a bit more intuitive than deparse + substitute

Lionel Henry Over a year ago

With rlang 1.0 you can write rlang::englue("results/{{ col_plot }}.png").

Collectives™ on Stack Overflow

Turn variable name into string inside function dplyr

2 Answers 2

4 Comments

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

4 Comments

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related