limulus is a Python library for data processing using Data Step syntax.
Its goal is to bring the simplicity and long-term stability of Data Step into Python workflows.
This is currently an alpha release. Please note that breaking changes to the API and other interfaces may occur before the beta release.
pip install limulusLoad a DataFrame (Arrow / Polars / Pandas) from a CSV or other source.
import pandas as pd
import limulus
health_df = pd.DataFrame({
"name": ["Alice", "Bob", "Charlie", "David"],
"age": [25, 30, 35, 40],
"height": [65, 70, 68, 72], # inches
"weight": [140, 180, 130, 200] # pounds
})
# Load data into a Session
session = limulus.Session()
session.loads({"health": health_df})session.submit("""
data result;
set health;
where age > 25;
height_m = height * 0.0254;
weight_kg = weight * 0.454;
bmi = round(weight_kg / (height_m**2), 0.1);
keep name age bmi;
run;
""")Multiple Data Steps can be submitted at once. Datasets created by earlier steps can be referenced by subsequent steps.
The simplest way is to retrieve the result from the session as an Arrow table using subscript notation.
You can convert it to pandas using Arrow's methods.
df_out = session["result"].to_pandas()
print(df_out)https://k-nkmt.github.io/limulus/
PolyForm Noncommercial License 1.0.0
This project is distributed under the PolyForm Noncommercial License. Creative Commons licenses are not used because they are generally not recommended for software distribution.
The software may be used for personal, educational, academic, and noncommercial research purposes.
Commercial use is not permitted under the current license terms.
The pip package includes a compiled Rust native extension (limulus_native); Rust dependencies are listed in native/limulus_native/Cargo.toml.
A more permissive license may be considered in the future as the project gains a more stable foundation.
contact: [email protected]
Project Positioning
limulus is a modern data-step–inspired data transformation framework implemented independently in Python and Rust. It is not affiliated with or endorsed by SAS Institute Inc.
Trademark Notice
SAS® is a registered trademark of SAS Institute Inc. All other trademarks are the property of their respective owners.
Independence Statement
This is an independent implementation. No SAS source code or proprietary materials have been used.
Compatibility Disclaimer
Compatibility with SAS software is not guaranteed and is not a project goal. Certain behaviors intentionally differ to provide modern semantics.