Plotnine (v0.15.7)

From ad-hoc sketches to publication-ready figures

Plotnine allows you to transform raw data into sophisticated visual stories by utilizing the grammar of graphics.

The Grammar of Graphics: A coherent, structured system designed for describing and constructing graphical representations of data.

Whether you are looking for the [API reference] or a [Cheatsheet], Plotnine provides a powerful framework for Python users. Its syntax is intentionally designed to be similar to the highly successful ggplot2 package from the R ecosystem.

📊 Exploring the Workflow: Anscombe’s Quartet

To demonstrate Plotnine's capabilities, we will visualize Anscombe’s Quartet. This is a famous set of four datasets that possess nearly identical descriptive statistics—such as the mean and variance—despite having wildly different distributions.

In mathematical terms, for each dataset $i \in \{1, 2, 3, 4\}$ , the statistics are roughly: $\bar{x} \approx 9, \quad \bar{y} \approx 7.5, \quad \text{Correlation}(x, y) \approx 0.816$

🚀 Quick Start

You can generate a basic visualization with a single line of code.

from plotnine import * 
from plotnine.data import anscombe_quartet 

# Initial basic scatter plot
ggplot(anscombe_quartet, aes(x="x", y="y")) + geom_point()

At first glance, this scatter plot is confusing because the four datasets are overlapped. We need a way to differentiate them.

🎨 Adding Aesthetics

By mapping the color aesthetic to the dataset column, Plotnine automatically generates a legend to distinguish the groups.

(ggplot(anscombe_quartet, aes("x", "y", color="dataset")) 
 + geom_point())

🔲 Declarative Subsetting (Faceting)

Instead of writing complex for loops to create multiple plots, you can use facet_wrap to repeat the visualization across separate panels.

(ggplot(anscombe_quartet, aes("x", "y", color="dataset")) 
 + geom_point() 
 + facet_wrap("dataset"))

Note: Since the panels now separate the data, the use of color becomes redundant.

🍰 Layering Visualizations

Plotnine treats plots as a series of layers. While global mappings are inherited, you can modify or add specific layers (like trend lines) to enhance the insight.

(ggplot(anscombe_quartet, aes("x", "y", color="dataset")) 
 + geom_point() 
 + geom_smooth(method="lm", se=False, fullrange=True) 
 + facet_wrap("dataset"))

This visualization proves Anscombe's point: datasets with the same statistics can look entirely different.

🛠️ Refining for Publication

While the previous plot is great for Exploratory Data Analysis (EDA), publication-quality figures require more precision.

Feature	Adjustment	Purpose
Colors	`sienna` & `steelblue`	Professional aesthetic
Sizing	`size=3` (points)	Better visibility
Axes	`scale_y_continuous`	Precise break points
Coordinates	`coord_fixed`	Fixed aspect ratio

(ggplot(anscombe_quartet, aes("x", "y")) 
 + geom_point(color="sienna", fill="darkorange", size=3) 
 + geom_smooth(method="lm", se=False, fullrange=True, color="steelblue", size=1) 
 + facet_wrap("dataset") 
 + scale_y_continuous(breaks=(4, 8, 12)) 
 + coord_fixed(xlim=(3, 22), ylim=(2, 14)) 
 + labs(title="Anscombe’s Quartet"))

🌟 "Going to Eleven" (Advanced Theming)

Finally, you can fully customize the theme to align with a specific brand or personal style.

(ggplot(anscombe_quartet, aes("x", "y")) 
 + geom_point(color="sienna", fill="orange", size=3) 
 + geom_smooth(method="lm", se=False, fullrange=True, color="steelblue", size=1) 
 + facet_wrap("dataset") 
 + labs(title="Anscombe’s Quartet") 
 + scale_y_continuous(breaks=(4, 8, 12)) 
 + coord_fixed(xlim=(3, 22), ylim=(2, 14)) 
 + theme_tufte(base_family="Futura", base_size=16) 
 + theme(
     axis_line=element_line(color="#4d4d4d"), 
     axis_ticks_major=element_line(color="#00000000"), 
     axis_title=element_blank(), 
     panel_spacing=0.09, 
 ))

📈 Summary of Evolution

✅ Next Steps for You

Install Plotnine via PyPI
Import your own dataset
Apply the Grammar of Graphics to your project

Project Details:

Developer: Hassan Kibirige
Funding: The Open Source Data Science Company
Resources: [PyPI] | [Source Code] | [License]