在使用dplyr的R函数中处理变量名引用和解引用问题

Got it, let's work through this common dplyr programming hurdle—variable quoting and unquoting can feel tricky when building reusable functions, but once you get the hang of the curly-curly operator, it's smooth sailing.

First, let's recap your goal: you want a generic function that takes a dataset and a variable name, converts that variable to a factor, and plays nicely with ggplot later. You mentioned other parts of the function work, so the issue is definitely how dplyr handles non-standard evaluation (NSE) for your target variable.

The Problem with Bare Variable Names in Functions #

If you tried something like this, you probably noticed it doesn't work:

# Broken example
convert_to_factor <- function(data, var) {
  data %>%
    mutate(var = as.factor(var)) # Creates a column named "var" instead of using your input!
}

Dplyr treats var here as a literal column name, not the variable you passed in (like am from mtcars). That's where quoting/unquoting comes in.

The Simplest Solution: Curly-Curly Operator ({{ }}) #

Dplyr 1.0.0 and later recommends using the curly-curly operator ({{ }}) for this exact scenario. It tells dplyr to "unquote" your input variable, so it recognizes the column name you're passing instead of treating it as a string.

Here's the fixed function:

# Working generic factor conversion function
convert_to_factor <- function(data, var) {
  data %>%
    mutate({{ var }} = as.factor({{ var }}))
}

# Test it with mtcars$am
mtcars_processed <- convert_to_factor(mtcars, am)

# Verify the conversion
str(mtcars_processed$am)
# Output: Factor w/ 2 levels "0","1": 2 2 2 1 1 1 1 1 1 2 ...

• The {{ var }} on the left side of the = keeps the original column name (so you don't end up with a column named var).
• The {{ var }} inside as.factor() tells dplyr to reference the actual column from your dataset.

Bonus: Add Custom Factor Levels/Labels #

If you want to make the function even more flexible (great for ggplot!), you can add parameters for custom levels and labels:

convert_to_factor <- function(data, var, levels = NULL, labels = NULL) {
  data %>%
    mutate({{ var }} = factor({{ var }}, levels = levels, labels = labels))
}

# Use it to label am as "Automatic" and "Manual"
mtcars_processed <- convert_to_factor(mtcars, am, labels = c("Automatic", "Manual"))

# Check the result
str(mtcars_processed$am)
# Output: Factor w/ 2 levels "Automatic","Manual": 2 2 2 1 1 1 1 1 1 2 ...

Using with ggplot #

Now this processed data works seamlessly with ggplot, since the factor is properly assigned to the original column name:

library(ggplot2)
mtcars_processed %>%
  ggplot(aes(x = am, y = mpg)) +
  geom_boxplot() +
  labs(x = "Transmission Type")

For Older Dplyr Versions #

If you're stuck on a pre-1.0.0 version of dplyr, you can use enquo() and the bang-bang operator (!!) instead:

convert_to_factor_old <- function(data, var) {
  var_quoted <- enquo(var)
  data %>%
    mutate(!!var_quoted := as.factor(!!var_quoted))
}

This does the same thing as the curly-curly operator, it's just a bit more verbose.

内容的提问来源于stack exchange，提问作者Chuck P