ggplot2latex: Export ggplot2 Figures to LaTeX Without the Bloat

ggplot2

LaTeX

ggplot2latex

visualizations

reproducibility

ggplot2latex lets you save any ggplot2 object as an optimised TikZ .tex file — vector-quality figures that slot straight into your LaTeX paper, up to 81% smaller than raw TikZ output.

Author

Roman Kyrychenko

Published

March 31, 2026

The Problem: LaTeX + ggplot2 = Pain

If you write academic papers in LaTeX and produce figures in R, you have probably faced the same dilemma repeatedly:

PNG/PDF exports look fine at the target size but degrade when reviewers zoom in, and they don’t inherit the document’s fonts.
TikZ exports (via tikzDevice) produce perfect vector figures that scale infinitely and match your paper’s font — but the generated .tex files are enormous. A world map can exceed 2 MB; a network graph routinely hits 50 000 lines of raw TeX.

Large TikZ files stall LaTeX compilation, push projects over journal submission limits, and make version-controlled repos unwieldy.

ggplot2latex solves this by post-processing the raw TikZ output: it strips redundancy, rounds floating-point coordinates, and collapses duplicate draw commands — giving you the same vector figure at a fraction of the size.

Installation

# From CRAN
install.packages("ggplot2latex")

# Or latest from GitHub
devtools::install_github("RomanKyrychenko/ggplot2latex")

Core Workflow: `save_tex()`

The main function mirrors ggsave() in spirit:

library(ggplot2)
library(ggplot2latex)

p <- ggplot(mtcars, aes(x = wt, y = mpg, colour = factor(cyl))) +
  geom_point(size = 3) +
  labs(title = "Fuel efficiency vs weight", colour = "Cylinders") +
  theme_minimal()

save_tex(p, "figures/fuel_efficiency.tex")

Drop the resulting file into your LaTeX document:

\usepackage{tikz}
\usepackage{pgfplots}

\begin{figure}[h]
  \centering
  \input{figures/fuel_efficiency}
  \caption{Fuel efficiency vs weight (Motor Trend, 1974).}
  \label{fig:fuel}
\end{figure}

The figure inherits your document’s fonts automatically — no more mismatched typefaces between body text and axis labels.

What the Optimiser Does

save_tex() calls tikzDevice::tikz() internally, then runs a multi-pass post-processor on the generated source:

Pass	What it does
Float rounding	Truncates coordinates like `12.3456789pt` to `12.35pt` — imperceptible visually, significant in file size
Deduplication	Removes repeated `\path[...]` clip and bounding-box commands that TikZ emits redundantly
Color simplification	Collapses `\definecolor{pgfstrokecolor}{rgb}{0.00,0.00,0.00}` duplicates
Bounding box pruning	Drops intermediate bounding-box resets not needed in the final figure

Benchmarks

On five representative figure types, save_tex() consistently outperforms raw save_tikz():

Plot type	Raw bytes	Optimised bytes	Reduction	Raw lines	Optimised lines	Line reduction
Scatter	11,300	7,990	29 %	269	134	50 %
Scatter + smooth	10,106	7,929	22 %	279	193	31 %
Faceted scatter	20,750	15,014	28 %	545	301	45 %
World map	2,116,053	403,827	81 %	102,620	21,999	79 %
Network graph	45,299	19,835	56 %	2,057	1,136	45 %

Average: 43 % smaller files, 50 % fewer lines — with zero loss of visual fidelity since everything remains vector-based TikZ.

The gains are largest for figures with many repeated path commands: maps (repeated polygon edges), network graphs (many similar node/edge styles), and faceted plots (duplicate axis scaffold).

Resizing Figures

Academic journals often have strict column-width requirements. Use resize_tex() to rescale an existing .tex file without re-generating it:

# Resize to 8 cm wide (aspect ratio preserved)
resize_tex("figures/fuel_efficiency.tex", width_cm = 8)

This rewrites the TikZ \begin{tikzpicture} scale parameter — no R re-render needed.

Helper Utilities

`save_tikz()` — Unoptimised Baseline

If you want the raw TikZ output for comparison or debugging:

save_tikz(p, "figures/fuel_efficiency_raw.tex")

`rgb2col()` — Colour Debugging

Convert an RGB triplet to its nearest named R colour — useful when inspecting TikZ colour definitions:

rgb2col(r = 0.122, g = 0.471, b = 0.706)
#> [1] "steelblue"

A Complete Example: Publication-Ready Map

library(ggplot2)
library(sf)
library(rnaturalearth)
library(ggplot2latex)

world <- ne_countries(scale = "medium", returnclass = "sf")

p_map <- ggplot(world) +
  geom_sf(aes(fill = pop_est), colour = "white", linewidth = 0.1) +
  scale_fill_viridis_c(
    trans   = "log10",
    labels  = scales::label_number(scale = 1e-6, suffix = "M"),
    name    = "Population"
  ) +
  theme_void() +
  labs(title = "World Population")

# Raw TikZ ≈ 2 MB; optimised ≈ 400 KB
save_tex(p_map, "figures/world_population.tex")

The world-map case is where ggplot2latex earns its keep most dramatically: raw TikZ exports of sf geometries balloon to megabytes because each polygon vertex becomes a coordinate pair in the source. The optimiser trims these aggressively.

Comparison with Alternatives

Approach	Vector?	Fonts match LaTeX?	File size	Compile speed
`ggsave(.pdf)`	✅	❌	Small	Fast
`ggsave(.png)`	❌	❌	Variable	Fast
`tikzDevice::tikz()`	✅	✅	Large	Slow
`ggplot2latex::save_tex()`	✅	✅	Small	Fast

Summary

ggplot2latex fills a genuine gap in the R-for-academics toolkit. If you write papers in LaTeX and make figures in R, save_tex() gives you the best of both worlds: the expressiveness of ggplot2, the typographic consistency of TikZ, and file sizes small enough to not slow down your build or inflate your repository.

# Install once, use everywhere
install.packages("ggplot2latex")

Source: https://github.com/RomanKyrychenko/ggplot2latex

--- title: "ggplot2latex: Export ggplot2 Figures to LaTeX Without the Bloat" description: | ggplot2latex lets you save any ggplot2 object as an optimised TikZ .tex file — vector-quality figures that slot straight into your LaTeX paper, up to 81% smaller than raw TikZ output. author: - name: Roman Kyrychenko url: https://www.linkedin.com/in/kirichenko17roman/ date: 2026-03-31 format: html: toc: true message: false warning: false echo: true code-tools: true twitter: creator: "@KyrychenkoRoman" categories: - R - ggplot2 - LaTeX - ggplot2latex - visualizations - reproducibility creative_commons: CC BY image: "preview.svg" --- ## The Problem: LaTeX + ggplot2 = Pain If you write academic papers in LaTeX and produce figures in R, you have probably faced the same dilemma repeatedly: - **PNG/PDF exports** look fine at the target size but degrade when reviewers zoom in, and they don't inherit the document's fonts. - **TikZ exports** (via `tikzDevice`) produce perfect vector figures that scale infinitely and match your paper's font — but the generated `.tex` files are enormous. A world map can exceed 2 MB; a network graph routinely hits 50 000 lines of raw TeX. Large TikZ files stall LaTeX compilation, push projects over journal submission limits, and make version-controlled repos unwieldy. **`ggplot2latex`** solves this by post-processing the raw TikZ output: it strips redundancy, rounds floating-point coordinates, and collapses duplicate draw commands — giving you the same vector figure at a fraction of the size. ## Installation ```r # From CRAN install.packages("ggplot2latex") # Or latest from GitHub devtools::install_github("RomanKyrychenko/ggplot2latex") ``` ## Core Workflow: `save_tex()` The main function mirrors `ggsave()` in spirit: ```r library(ggplot2) library(ggplot2latex) p <- ggplot(mtcars, aes(x = wt, y = mpg, colour = factor(cyl))) + geom_point(size = 3) + labs(title = "Fuel efficiency vs weight", colour = "Cylinders") + theme_minimal() save_tex(p, "figures/fuel_efficiency.tex") ``` Drop the resulting file into your LaTeX document: ```latex \usepackage{tikz} \usepackage{pgfplots} \begin{figure}[h] \centering \input{figures/fuel_efficiency} \caption{Fuel efficiency vs weight (Motor Trend, 1974).} \label{fig:fuel} \end{figure} ``` The figure inherits your document's fonts automatically — no more mismatched typefaces between body text and axis labels. ## What the Optimiser Does `save_tex()` calls `tikzDevice::tikz()` internally, then runs a multi-pass post-processor on the generated source: | Pass | What it does | |---|---| | **Float rounding** | Truncates coordinates like `12.3456789pt` to `12.35pt` — imperceptible visually, significant in file size | | **Deduplication** | Removes repeated `\path[...]` clip and bounding-box commands that TikZ emits redundantly | | **Color simplification** | Collapses `\definecolor{pgfstrokecolor}{rgb}{0.00,0.00,0.00}` duplicates | | **Bounding box pruning** | Drops intermediate bounding-box resets not needed in the final figure | ## Benchmarks On five representative figure types, `save_tex()` consistently outperforms raw `save_tikz()`: | Plot type | Raw bytes | Optimised bytes | Reduction | Raw lines | Optimised lines | Line reduction | |---|---:|---:|---:|---:|---:|---:| | Scatter | 11,300 | 7,990 | 29 % | 269 | 134 | 50 % | | Scatter + smooth | 10,106 | 7,929 | 22 % | 279 | 193 | 31 % | | Faceted scatter | 20,750 | 15,014 | 28 % | 545 | 301 | 45 % | | World map | 2,116,053 | 403,827 | **81 %** | 102,620 | 21,999 | **79 %** | | Network graph | 45,299 | 19,835 | 56 % | 2,057 | 1,136 | 45 % | **Average: 43 % smaller files, 50 % fewer lines** — with zero loss of visual fidelity since everything remains vector-based TikZ. The gains are largest for figures with many repeated path commands: maps (repeated polygon edges), network graphs (many similar node/edge styles), and faceted plots (duplicate axis scaffold). ## Resizing Figures Academic journals often have strict column-width requirements. Use `resize_tex()` to rescale an existing `.tex` file without re-generating it: ```r # Resize to 8 cm wide (aspect ratio preserved) resize_tex("figures/fuel_efficiency.tex", width_cm = 8) ``` This rewrites the TikZ `\begin{tikzpicture}` scale parameter — no R re-render needed. ## Helper Utilities ### `save_tikz()` — Unoptimised Baseline If you want the raw TikZ output for comparison or debugging: ```r save_tikz(p, "figures/fuel_efficiency_raw.tex") ``` ### `rgb2col()` — Colour Debugging Convert an RGB triplet to its nearest named R colour — useful when inspecting TikZ colour definitions: ```r rgb2col(r = 0.122, g = 0.471, b = 0.706) #> [1] "steelblue" ``` ## A Complete Example: Publication-Ready Map ```r library(ggplot2) library(sf) library(rnaturalearth) library(ggplot2latex) world <- ne_countries(scale = "medium", returnclass = "sf") p_map <- ggplot(world) + geom_sf(aes(fill = pop_est), colour = "white", linewidth = 0.1) + scale_fill_viridis_c( trans = "log10", labels = scales::label_number(scale = 1e-6, suffix = "M"), name = "Population" ) + theme_void() + labs(title = "World Population") # Raw TikZ ≈ 2 MB; optimised ≈ 400 KB save_tex(p_map, "figures/world_population.tex") ``` The world-map case is where `ggplot2latex` earns its keep most dramatically: raw TikZ exports of `sf` geometries balloon to megabytes because each polygon vertex becomes a coordinate pair in the source. The optimiser trims these aggressively. ## Comparison with Alternatives | Approach | Vector? | Fonts match LaTeX? | File size | Compile speed | |---|:---:|:---:|---|---| | `ggsave(.pdf)` | ✅ | ❌ | Small | Fast | | `ggsave(.png)` | ❌ | ❌ | Variable | Fast | | `tikzDevice::tikz()` | ✅ | ✅ | **Large** | Slow | | `ggplot2latex::save_tex()` | ✅ | ✅ | **Small** | Fast | ## Summary `ggplot2latex` fills a genuine gap in the R-for-academics toolkit. If you write papers in LaTeX and make figures in R, `save_tex()` gives you the best of both worlds: the expressiveness of `ggplot2`, the typographic consistency of TikZ, and file sizes small enough to not slow down your build or inflate your repository. ```r # Install once, use everywhere install.packages("ggplot2latex") ``` **Source**: <https://github.com/RomanKyrychenko/ggplot2latex>