reaborn vs. seaborn vs. ggplot2

Where reaborn fits

reaborn is a faithful R port of Python’s seaborn, built on ggplot2. It mirrors seaborn’s public API exactly — same function names, same argument names, same defaults — so the plotting code you already know runs in R with little to no translation. The defaults that make seaborn plots look good out of the box (the styles, the palettes, the spacing) come along for the ride, because reaborn reproduces them rather than approximating them.

Fidelity is the part most ports get wrong, so reaborn treats it as the headline. Palettes match seaborn’s hex codes digit-for-digit (including matplotlib’s colormap quantization and named-color table); KDEs reproduce scipy.stats.gaussian_kde to machine precision; histogram bin edges match numpy.histogram_bin_edges exactly; and confidence intervals use seaborn’s bootstrap, not ggplot2’s analytic standard error. When a reaborn chart and a seaborn chart sit side by side, the goal is “indistinguishable,” not “close enough.”

The part seaborn can’t match: every reaborn plot is a ggplot object. A call like scatterplot(...) returns a ggplot, so you can keep building with the full grammar of graphics — add facet_wrap(), swap in scale_x_log10(), layer extra geoms, override the theme. You get seaborn’s defaults as a starting point and ggplot2’s composability as the ceiling.

Feature comparison

Capability	reaborn	seaborn	ggplot2
Identical seaborn API (names, args, defaults)	✅ Full — ~40 functions ported 1:1	✅ It is the API	❌ Different grammar entirely
Copy-paste Python code runs verbatim	✅ sns. aliases + True/False/None bound	✅ Native Python	❌
Statistical fidelity (KDE, bins, bootstrap CIs)	✅ Matches seaborn/scipy/numpy to machine precision	✅ Reference implementation	⚠️ Different methods
Theming & palettes match seaborn exactly	✅ Hex-exact; global like sns.set_theme()	✅ Reference	⚠️ Its own (good) defaults
Grammar-of-graphics extensibility	✅ Returns a real ggplot	❌ Returns matplotlib Axes	✅ Native, most complete
Breadth of arbitrary custom geoms/stats	⚠️ Inherits ggplot2’s	⚠️ Fixed function set	✅ The widest
Language / ecosystem	R (tidyverse-adjacent)	Python (pandas/matplotlib)	R (tidyverse)
Dependency weight	🟢 Light: ggplot2 stack; heavy bits are Suggests	🟡 matplotlib + numpy + scipy + pandas	🟢 Light
License	BSD-3 (same as seaborn)	BSD-3	MIT

✅ first-class · ⚠️ partial/with caveats · ❌ not a goal · 🟢/🟡 lighter/heavier.

Coming from seaborn?

In most cases you change nothing but the language host. After library(reaborn), the sns. aliases, the global theme, and the True/False/None literals are all in scope, so a seaborn snippet pasted into R runs as-is.

library(reaborn)   # sets seaborn theme + palette globally, like sns.set_theme()

# This is literally seaborn code — it runs verbatim in R:
sns.scatterplot(data = penguins, x = "bill_length_mm", y = "bill_depth_mm",
                hue = "species")

Python (seaborn)	R (reaborn)	Note
import seaborn as sns	library(reaborn)	Sets theme/palette globally, exposes sns. aliases
sns.set_theme()	automatic on load
sns.scatterplot(data=df, x="a", y="b", hue="g")	same line, verbatim	sns. alias + = kwargs both work
True / False / None	True / False / None	Bound to TRUE / FALSE / NULL
string columns: x="col"	x = "col"	seaborn’s string-column API is preserved
ax.set(...) (matplotlib)	+ ggplot2::labs(...), + theme(...)	You now get the ggplot grammar

Coming from ggplot2?

You already love the grammar of graphics. reaborn doesn’t ask you to give it up — it hands you seaborn’s defaults and statistics as a starting layer, returned as an ordinary ggplot that you keep building.

• Skip the boilerplate. histplot(data, x="x", hue="g") gets you seaborn’s binning, palette, and theme in one call — then + facet_wrap(~year) + scale_x_log10() is yours.
• Get statistics ggplot doesn’t ship. Bootstrap confidence intervals (seaborn-style, not analytic SE), scipy-exact KDEs, and numpy-exact histogram edges — without writing a custom stat_.
• Borrow a hex-matched palette system, including HUSL and colormap quantization, usable on any plot.
• It composes, it doesn’t replace. Everything is a real ggplot, so patchwork, custom scales, extra geoms, and your theme tweaks all keep working.

Honest framing: if your goal is a fully bespoke chart designed grammar-up, plain ggplot2 remains the most flexible tool — that’s not the claim. reaborn’s edge is narrower and specific: seaborn-grade defaults and statistical fidelity, delivered as extensible ggplots.