r2py_rpart

A Python port of the R package rpart (Recursive Partitioning and Regression Trees). r2py_rpart reimplements rpart's original C recursive-partitioning engine — compiled directly from a lightly adapted copy of the C sources and driven from Python via cffi — and translates rpart's R-level functions (formula handling, pruning, printing, plotting, prediction, cross-validation, ...) into Python with numpy/pandas, aiming for output that matches R's rpart as closely as possible.

Trees are fit using the same recursive-partitioning algorithm as R's rpart: classification, regression (anova), Poisson/exponential (survival), and user-defined splitting methods are all supported, along with cost-complexity pruning, surrogate splits for missing data, and cross-validated complexity tables.

pip install r2py_rpart

Prebuilt wheels are published for Python 3.10–3.14 on Linux (x86_64/aarch64), macOS (Intel/Apple Silicon), and Windows (x86_64). If a wheel isn't available for your platform, pip will build from source, which requires a C/C++ compiler (GCC, Clang, or MinGW-w64 on Windows) — no R installation is needed.

If you use r2py_rpart in published work, please cite all three of the following. The numerical core of r2py_rpart is rpart's own C code, retained unmodified — the results you obtain are produced by Therneau and Atkinson's implementation of the methods in Breiman et al. (1984).

1. Therneau, T. and Atkinson, B. (2026). rpart: Recursive Partitioning and Regression Trees. R package version 4.1.27. CRAN. https://cran.r-project.org/package=rpart (the original package, whose C routines this port embeds)

2. Cai, Y. and Li, J. r2py: AI-Assisted Conversion of R Statistical Packages to Python. (in preparation) (the conversion method, and the validation evidence for this port)

3. Cai, Y. and Li, J. (2026). r2py_rpart (version 0.2.0) [Computer software]. PyPI. https://pypi.org/project/r2py_rpart/ (the exact artifact executed — please cite the version you actually ran)

For the underlying statistical method, please also cite Breiman, L., Friedman, J. H., Olshen, R. A. and Stone, C. J. (1984). Classification and Regression Trees. Wadsworth & Brooks/Cole, Monterey, CA.

@Manual{therneau2026rpart,
  title  = {rpart: Recursive Partitioning and Regression Trees},
  author = {Terry Therneau and Beth Atkinson},
  year   = {2026},
  note   = {R package version 4.1.27},
  url    = {https://CRAN.R-project.org/package=rpart}
}

@Article{cai2026r2py,
  title   = {r2py: AI-Assisted Conversion of R Statistical Packages to Python},
  author  = {Cai, Yufei and Li, Jun},
  year    = {2026},
  note    = {In preparation}
}

@Misc{cai2026r2pyrpart,
  title  = {r2py\_rpart},
  author = {Cai, Yufei and Li, Jun},
  year   = {2026},
  note   = {Python package version 0.2.0},
  url    = {https://pypi.org/project/r2py_rpart/}
}

@Book{breiman1984cart,
  title     = {Classification and Regression Trees},
  author    = {Breiman, Leo and Friedman, Jerome H. and Olshen, Richard A.
               and Stone, Charles J.},
  year      = {1984},
  publisher = {Wadsworth \& Brooks/Cole},
  address   = {Monterey, CA}
}

import pandas as pd
from r2py_rpart import rpart, print_rpart, printcp, prune, plot_rpart, text_rpart

iris = pd.read_csv("iris.csv")
iris["species"] = iris["species"].astype("category")

# Classification tree
fit = rpart("species ~ .", data=iris, method="class", control={"cp": 0.0001, "minsplit": 5})
print_rpart(fit)
printcp(fit)

# Prune with a chosen complexity parameter
pruned = prune(fit, cp=0.02)

import matplotlib.pyplot as plt
fig, ax = plt.subplots(figsize=(8, 8))
plot_rpart(pruned, ax=ax)
text_rpart(pruned, ax=ax)
fig.savefig("tree.pdf")

See examples/classification_tree.py and examples/regression_tree.py for complete, runnable end-to-end examples (each paired with the equivalent R script and side-by-side R/Python output for comparison).

r2py_rpart mirrors the public functions of R's rpart package by name:

pip install -e . pytest
pytest tests/

Many tests compare r2py_rpart output against R's rpart running via rpy2, so an R installation with the rpart package is required to run the full parity suite.

r2py_rpart is a derivative work of R's rpart package and is distributed, like its upstream, under the GNU General Public License version 2 or version 3, at the recipient's option (GPL-2.0-only OR GPL-3.0-only). See LICENSE (version 2) and LICENSE-GPL3 (version 3) for the full texts, and NOTICE for attribution to the original rpart authors.

• Yufei Cai ([email protected])
• Jun Li ([email protected])

University of Notre Dame