Tidy Eval Meets ggplot2

aes()

The common way of using ggplot2 functions, and the first example most people try is using Non standard Evaluation (NSE), by passing variable names unquoted. For instance if you want to plot tip_percent versus total_bill you could simply write the following chunk of code.

ggplot(tips, aes(x = total_bill,
                 y = tip_percent,
                 alpha = 0.5))+
  geom_point()+
  theme_minimal()+
  guides(alpha = FALSE)

aes_string()

Another way is using aes_string() instead of aes(). In this case, you will need to pass column names quoted as follows, which will give the same result.

ggplot(tips, aes_string(x = "total_bill",
                 y = "tip_percent",
                 alpha = 0.5))+
  geom_point()+
  theme_minimal()+
  guides(alpha = FALSE)

This version could be helpful when you want to pass variables to ggplot2 functions. For instance, you can handle a scenario, when you have pre-defined variables to use, or in case you take the variable names from a user through an interface. The following examples will work fine and give the same result.

## define variables
x_var <- "total_bill"
y_var <- "tip_percent"
alpha_var <- 0.5

ggplot(tips, aes_string(x = x_var,
                 y = y_var,
                 alpha = alpha_var))+
  geom_point()+
  theme_minimal()+
  guides(alpha = FALSE)

And usually the use case for this, is the need to write a reusable function that wraps several ggplot2 layers to work with whatever variables passed to it. For example, you can wrap the previous chunk in a function as follows

plot_points <- function(x, y, alpha){
  ggplot(tips, aes_string(x = x, y = y, alpha = alpha))+
  geom_point()+
  theme_minimal()+
  guides(alpha = FALSE)
}

Then you can either pass column names directly or the variable names.

## pass column names directly
plot_points(x = "total_bill",
            y = "tip_percent",
            alpha = 0.5)

## pass variable names
plot_points(x = x_var,
            y = y_var,
            alpha = alpha_var)

And you can reuse the function to create similar plots with different variables. For instance tip on the y-axis instead of tip_percent

## plot tip vs. total_bill
plot_points(x = "total_bill",
            y = "tip",
            alpha = 0.5)

Unquoted column names / NSE

You can use enquo() and the bang bang operator !! to create a function like plot_points_tidyeval_01() which takes the column names unquoted, and it will work properly.

plot_points_tidyeval_01 <- function(x, y, alpha){
  
  x_var <- enquo(x)
  y_var <- enquo(y)
  
  ggplot(tips, aes(x = !!x_var, y = !!y_var, alpha = alpha))+
    geom_point()+
    theme_minimal()+
    guides(alpha = FALSE)
}

## pass column names unquoted 
plot_points_tidyeval_01(x = total_bill,
                        y = tip_percent,
                        alpha = 0.5)

You might still see this as a simple use case that could be handled by aes_string(). But there are more things you can do, like variable facets.

Variable Facets

Suppose that you wanted to define a variable facet, you wouldn’t have this flexibility even with aes_string(). But now you can implement it as follows.

Notice that vars() is a new helper inside facet_grid instead of the ~.

plot_points_tidyeval_02 <- function(x, y, alpha, facet){
  
  x_var <- enquo(x)
  y_var <- enquo(y)
  facet_var <- enquo(facet)
  
  ggplot(tips, aes(x = !!x_var, y = !!y_var, alpha = alpha))+
    geom_point()+
    theme_minimal()+
    guides(alpha = FALSE)+
    facet_grid(cols = vars(!!facet_var))}

So you can pass sex to the facet variable in the plot_points_tidyeval_02() function.

## use sex for facet
plot_points_tidyeval_02(x = total_bill,
                        y = tip_percent,
                        alpha = 0.5,
                        facet = sex)

or a different variable like day

## use day for facet
plot_points_tidyeval_02(x = total_bill,
                        y = tip_percent,
                        alpha = 0.5,
                        facet = day)

You can even use the dots ... to have more flexibility and allow the function to take more than one variable for faceting. You can pass the dots ... directly to vars() inside facet_wrap() as follows.

plot_points_tidyeval_03 <- function(x, y, alpha, ...){
  
  x_var <- enquo(x)
  y_var <- enquo(y)
  
  ggplot(tips, aes(x = !!x_var, y = !!y_var, alpha = alpha))+
    geom_point()+
    theme_minimal()+
    guides(alpha = FALSE)+
    facet_wrap(vars(...), nrow = 2)
}


## use both sex and day for facets
plot_points_tidyeval_03(x = total_bill,
                        y = tip_percent,
                        alpha = 0.5,
                        sex, day)