stratum-ai 0.0.0.dev0

skrub with a Rust backend

Stratum

Stratum is an experimental fork of skrub with a Rust backend for compute-heavy operations, while keeping the high-level Python API intact.

---

Goals

• Provide an opt-in Rust backend for performance-critical parts of skrub.
• Preserve skrub’s Python API; developers can flip a flag to enable Rust.
• Build cross-platform wheels (Windows / Linux / macOS) so users can install without a Rust toolchain.

---

Installation

For now, you need to build from source.

Requirements:

• Python 3.10+
• Rust toolchain (nightly not required; stable is fine)
• maturin (pip install maturin)

---

Usage

To enable the Rust backend and other related features, import stratum as skrub and enable the Rust backend:

Replace

import skrub
from skrub import ...

with

import stratum
from stratum import ...
skrub.set_config(rust_backend=True)

Test Code

import os
import pandas as pd
import stratum as skrub
from stratum import StringEncoder
skrub.set_config(rust_backend=True)

#skrub.set_config(debug_timing=True, num_threads=0) # other rust flags
s = pd.Series(["foo", "bar", None, "lorem ipsum dolor"])
enc = StringEncoder(vectorizer='hashing', analyzer='char', ngram_range=(3,5), n_components=2)
Z = enc.fit_transform(s)
print(type(Z), Z.shape)
assert Z.shape[0] == len(s)

---

Repository Layout

stratum/
├─ pyproject.toml             # Python + Rust build config (maturin)
├─ stratum/
│ ├─ __init__.py              # Façade over skrub
│ ├─ config.py                # set_config/get_config + env sync
│ ├─ _rust_backend.py         # Python <-> Rust shim (re-exports native fns)
│ ├─ adapters/                # Public API (dispatch to Rust or fallback to skrub)
│ │ └─ string_encoder.py      # RustyStringEncoder (subclass)
│ └─ _rust_backend_native.*   # Compiled PyO3 extension (built)
└─ _rust/                     # Rust crate (PyO3 extension)
├─ Cargo.toml
└─ src/lib.rs                 # Defines #[pymodule] fn _rust_backend_native(...)

---

Developer Instructions

Local Dev Install (Editable)

maturin develop				# Debug mode
maturin develop --release	# Optimized dev build

Building Wheels

This produces redistributable .whl files under dist/.

maturin build --release -o dist --interpreter python3.10 --compatibility linux		# Linux/macOS
maturin build --release -o dist		# Windows

Then install with:

pip install ./dist/stratum-*.whl

---

License

BSD-3-Clause (inherited from skrub).