2.6 ggplot Extensions

2.6 ggplot Extensions

Learning objectives

By the end of this chapter, you can:

  1. explain how extensions join the grammar of graphics through geom_*(), stat_*(), theme_*(), and scale_*() functions.
  2. run minimal examples with six common extensions: ggrepel, ggridges, patchwork, ggforce, gghighlight, and scales.
  3. discover extensions through the official gallery and assess their maintenance.
  4. combine extensions in one figure without creating tangled code.
  5. read an unfamiliar extension’s documentation within ten minutes using three questions.
  6. justify introducing a new dependency for a figure, with AI acting as a reviewer of that choice.

Prerequisite check (≤5 minutes)

Complete these checks independently; otherwise revisit Chapter 1.7.

ImportantCheck In: Prerequisites
  1. Write a ggplot2 scatterplot with a smooth line using palmerpenguins::penguins: bill_length_mm × bill_depth_mm, colored by species.
  2. Does each addition with + add a layer or a global setting? Explain the difference between geom_point() and labs().

1. Extensions: New verbs in the grammar

ggplot2’s strength is its grammar: data, geometric objects, scales, and themes assembled with +. An extension package adds new verbs to that grammar. A new geom_*() is like a precise addition to your vocabulary; you need not relearn sentence structure.

The good news is that most of the knowledge transfers. Use ggforce::geom_mark_circle() as you would geom_point(): map aesthetics, set parameters, and add it with +. Much of the learning lies in knowing that the extension exists.

Note

ggplot2 extensions follow a shared Extension API built around ggproto. This chapter approaches them as a user. Return to writing extensions after Unit 3 on package development.

2. Six useful extensions at a glance

Run these examples after installing any missing packages with install.packages().

library(ggplot2); library(palmerpenguins)
pg <- dplyr::filter(penguins, !is.na(sex))

# ① ggrepel: keep labels apart; compare with overlapping geom_text labels
ggplot(pg, aes(bill_length_mm, bill_depth_mm, colour = species)) +
  geom_point() +
  ggrepel::geom_text_repel(
    aes(label = stringr::str_to_title(species)),
    data = ~ dplyr::distinct(.x, species, .keep_all = TRUE)
  )

# ② ggridges: ridgeline plots for distributions across groups
ggplot(pg, aes(bill_length_mm, species, fill = species)) +
  ggridges::geom_density_ridges(alpha = 0.7)

# ③ gghighlight: emphasize the subset under discussion
ggplot(pg, aes(bill_length_mm, fill = species)) +
  geom_histogram(bins = 30) +
  gghighlight::gghighlight(species == "Gentoo")

# ④ ggforce: zoom into a crowded region
ggplot(pg, aes(bill_length_mm, bill_depth_mm, colour = species)) +
  geom_point() +
  ggforce::facet_zoom(xy = species == "Gentoo")
# ⑤ patchwork: assemble several plots into one composition
library(patchwork)
p1 <- ggplot(pg, aes(species, bill_length_mm)) +
  geom_boxplot()
p2 <- ggplot(pg, aes(flipper_length_mm)) +
  geom_histogram(bins = 25)
p3 <- ggplot(pg, aes(species, fill = island)) +
  geom_bar(position = "fill")

p1 | (p2 / p3)          # Operators express layout: one left, two right

# ⑥ scales: refine axes and legends
p2 + scale_x_continuous(labels = scales::label_number(scale_cut = scales::cut_short_scale()))
WarningCommon error: Applying label_*() to the wrong scale 
p1 + scales::label_percent()  # Intentional error: a labeling function cannot be added directly to a ggplot

Pass the labeling function to scale_*_continuous(labels = ...). Also, scales::label_percent() makes sense only for proportions. The boxplot’s y-axis contains raw millimeters; forcing percentage labels onto it creates absurd results or an error. Identify what the axis values represent before choosing a label function.

4. Combining extensions without tangling the code

Extensions become powerful in combination: one handles labels, another emphasis, another composition. Three extensions can also make code hard to follow. Use one rule: build each plot as a sequence of named objects, adding one extension layer at a time.

# Build and name each step so it can be rerun and adjusted independently
base <- ggplot(pg, aes(bill_length_mm, bill_depth_mm, colour = species)) +
  geom_point(alpha = 0.6)

labelled <- base +
  ggrepel::geom_text_repel(
    aes(label = stringr::str_to_title(species)),
    data = ~ dplyr::distinct(.x, species, .keep_all = TRUE),
    seed = 42                       # Repel layouts are random: fix the seed for reproducibility
  )

focused <- labelled + gghighlight::gghighlight(
  species != "Chinstrap",
  unhighlighted_params = list(colour = "grey80")
)

final <- focused + ggtitle("The two penguin species with the longest bills") + theme_minimal()
final

Three principles for combining extensions:

  1. One chain tells one story. Highlighting with gghighlight and zooming with ggforce seldom belong together in the same display.
  2. Fix random seeds with seed =; otherwise successive renders can differ and complicate review.
  3. Add patchwork last: refine each component plot before composing them.

5. Reading extension documentation: Three questions

Ask these questions in order to decide within ten minutes whether an extension fits:

Question Where to look Criterion
Q1. What new function does it provide? First screen of the README Does the function name explain its role?
Q2. What is the smallest example? First README code example Can I substitute my data directly?
Q3. How does it interact with scales and themes? Middle of the vignette Do scale_*() and theme() still work normally?

Treat dependencies with restraint. Adding more than two new dependencies to one figure makes reproduction harder for a reviewer. Patchwork and ggrepel deserve consideration for a regular toolbox; introduce other extensions as projects require them. A rich ecosystem means a suitable tool is likely to exist when you need it, not that every tool belongs in every installation.

6. The reverse check: Do you need an extension?

ggplot2 itself evolves quickly; “Modern ggplot2” was a theme of the posit::conf(2025) workshop. Before adding an extension, ask:

  • Can base ggplot2 already draw this? annotate(), coord_*(), and facet_wrap() are often underestimated.
  • Has a newer version built in the feature? Check ggplot2 NEWS.
  • Does the dependency save only three lines of code? Writing the three lines may be simpler.
WarningCommon misconception: Hiding a design problem with an extension

Labels often overlap because there are too many labels. Ask whether readers need all 40 before reaching for ggrepel. Extensions solve layout problems; they do not resolve information overload.

ImportantPractice Exercise 1 (copy)

Choose an extension from the gallery and run the first README example unchanged. Then substitute penguins, adjusting column mappings. Deliver both code versions and one sentence explaining the limitation of base ggplot2 that the extension addresses.

ImportantPractice Exercise 2 (adapt)

Adapt the §4 build sequence to highlight island rather than species. Combine focused with a species-based ridgeline chart using (focused | ridge) + patchwork::plot_annotation(tag_levels = "A"). Fix random seeds and name every intermediate object.

ImportantPractice Exercise 3 (create · AI-off stage)

Round 1 (AI prohibited): Using only the gallery and two extensions’ official documentation, create an “editor’s choice” penguins figure combining any three of faceting, highlighting, non-overlapping labels, and composition. Do not consult any AI assistant. Write your three-question answers for each extension, three lines each. Round 2 (AI allowed): Give Posit Assistant the finished code and ask only: “Which layer could be implemented with base ggplot2?” Record at least one suggestion.

Capstone

Task: Choose health-check, penguins, or nycflights13 data and create a summary figure that tells the story in one composition. Combine at least three extensions: patchwork for composition, ggrepel for key labels, and a third extension of your choice with a stated rationale. Deliver one Quarto page containing the final figure, its named build sequence, and a defense of the three choices, including the assessment table.

Dimension Meets expectations Strong Excellent
Composition Three extensions and a readable figure Each layer has one purpose; clear build sequence Deliberate whitespace, hierarchy, and consistent emphasis
Reproducibility Code reruns Random seeds fixed Another person reproduces the figure without changes
Choice of tools States what was used Explains why Checks which layers base ggplot2 could replace
Documentation efficiency Uses the three questions Includes written answers Shares gallery-entry assessments classmates can reuse

SOURCES

Chapter section Material Use
§1–§2 extension ecosystem and patchwork composition posit::conf(2025) ggplot2, sessions/5_composition.qmd, 6_extensions.qmd (Thomas Lin Pedersen, Teun van den Brand; README: CC-BY 4.0; LICENSE.md: CC-BY-SA 4.0) Adaptation
§3 extensions as entry points to Quarto / ggplot2 ecosystems posit::conf(2025) quarto-extend (Mine Çetinkaya-Rundel, Charlotte Wickham; CC-BY-SA 4.0) Structural inspiration
Extension gallery https://exts.ggplot2.tidyverse.org, official ggplot2 gallery Reference
Six-extension examples, three questions, assessment checks, exercises, and rubric This project Original

This chapter is published under CC-BY-SA 4.0.