Integrations

Integrations

Arnio is meant to sit before and alongside the Python data stack.

pandas

Convert between Arnio and pandas:

import pandas as pd
import arnio as ar

pdf = pd.read_csv("data.csv")
frame = ar.from_pandas(pdf)
clean = ar.pipeline(frame, [("strip_whitespace",)])
out = ar.to_pandas(clean)

Use the pandas accessor for in-place pandas workflows:

clean_df = pdf.arnio.clean([
    ("strip_whitespace",),
    ("drop_duplicates",),
])

report = clean_df.arnio.profile()

NumPy

Use NumPy after converting cleaned data to pandas:

df = ar.to_pandas(clean)
values = df[["feature_a", "feature_b"]].to_numpy()

scikit-learn

Install the optional extra:

pip install "arnio[sklearn]"

Use Arnio cleaning before model training, or use the scikit-learn integration patterns in examples/.

DuckDB

Register cleaned frames for SQL workflows:

import duckdb
import arnio as ar

frame = ar.read_csv("data.csv")
df = ar.to_pandas(frame)

con = duckdb.connect()
con.register("clean_data", df)

See the repository examples for dedicated DuckDB usage.

Arrow

Install Arrow support:

pip install "arnio[arrow]"

Convert Arnio frames to Arrow tables:

table = ar.to_arrow(frame)

JSONL

frame = ar.read_jsonl("events.jsonl")

JSONL parsing skips blank lines, fills missing keys with nulls, and uses conversion behavior consistent with pandas handoff.

Parquet

Install parquet support:

pip install "arnio[parquet]"

Then use the parquet helpers supported by the current package version.

Example Gallery

The repository includes examples for:

• Basic usage
• pandas
• NumPy
• scikit-learn
• DuckDB
• Arrow
• JSONL
• chunked CSV reading
• sales, customers, survey, logs, and finance recipes

Start with examples/README.md in the repository.