如何在自定义ggplot2主题中仅临时修改geom默认美学属性？

Great question! The issue with your initial approach is that update_geom_defaults() modifies global defaults for the geom—so once you run it, every subsequent geom_point() will use red until you explicitly reset it.

To make the geom defaults tied only to your custom theme (without polluting global settings), we can create a custom theme object that temporarily adjusts geom defaults when applied, then restores them afterward. Here's how to do it:

Step 1: Create a reusable theme + geom defaults system #

We'll define a custom class for our theme objects and a special ggplot_add method to handle the temporary default changes safely:

library(ggplot2)
library(utils)

# Define the red dot theme with tied geom defaults
theme_red_dots <- function(...) {
  # Build the base minimal theme with user customizations
  base_theme <- theme_minimal() + theme(...)
  
  # Package the theme and desired geom overrides into a structured object
  structure(
    list(
      theme = base_theme,
      geom_defaults = list(point = list(colour = "red"))
    ),
    class = "theme_with_geom_defaults"
  )
}

# Custom method to apply the theme and temporary geom defaults
ggplot_add.theme_with_geom_defaults <- function(object, plot, object_name) {
  # Save the original geom defaults to restore later
  old_point_defaults <- ggplot2:::.geom_defaults$point
  
  # Update defaults with our theme-specific values
  ggplot2:::.geom_defaults$point <- modifyList(old_point_defaults, object$geom_defaults$point)
  
  # Guarantee original defaults are restored even if plotting fails
  on.exit({
    ggplot2:::.geom_defaults$point <- old_point_defaults
  })
  
  # Apply the theme to the plot
  plot + object$theme
}

Step 2: Test the red dot theme #

Now when you use the theme, points turn red only for that plot—global defaults stay intact:

# Plot with red dot theme (points are red)
ggplot(mtcars, aes(mpg, disp)) + geom_point() + theme_red_dots()

# Plot without the theme (points revert to black)
ggplot(mtcars, aes(mpg, disp)) + geom_point()

Step 3: Adapt for the dark theme #

For your theme_black example, we can use the same pattern to auto-set white points (so they're visible on a black background):

theme_black <- function(...) {
  base_theme <- theme_minimal() + 
    theme(panel.background = element_rect(fill = "black", colour = NA)) +
    theme(...)
  
  structure(
    list(
      theme = base_theme,
      geom_defaults = list(point = list(colour = "white"))
    ),
    class = "theme_with_geom_defaults"
  )
}

# Test the dark theme (white points on black background)
ggplot(mtcars, aes(mpg, disp)) + geom_point() + theme_black()

A quick note on internals #

We're using ggplot2:::.geom_defaults, an internal ggplot2 object, to tweak defaults temporarily. While this works reliably in current versions, it's not officially documented—keep an eye out for changes in future ggplot2 updates. For most practical use cases, though, this is a robust solution.

内容的提问来源于stack exchange，提问作者markus