deltaforge-adbc 1.0.5

DeltaForge ADBC driver: Arrow-native read and write for Delta Lake and Apache Iceberg, carrying record batches end-to-end so BI tools skip row-by-row conversion

deltaforge-adbc

ADBC driver: Arrow-native read and write for Delta Lake and Apache Iceberg from Python

deltaforge.org · ADBC driver · Issues

Arrow-native read and write for Delta tables from Python, via the DeltaForge ADBC driver. The driver carries Arrow record batches end to end, so it is the preferred connector for wide-column scans and for writing DataFrames straight into Delta tables.

DeltaForge is commercial software with a free Community license. See deltaforge.org/pricing.

Install

pip install deltaforge-adbc
# optional DataFrame integrations:
pip install "deltaforge-adbc[pandas]"   # or [polars]

The wheel bundles the native driver for your platform. No separate driver install or unixODBC setup is required.

Connect

import deltaforge_adbc as df

conn = df.connect(
    control_plane="https://control.example.com",
    token="df_pat_...",                    # personal access token
    compute="https://compute.example.com", # optional; auto-selected if omitted
)

Connection parameters also fall back to environment variables: DELTAFORGE_CONTROL_PLANE_URL, DELTAFORGE_SESSION_TOKEN, DELTAFORGE_COMPUTE_URL.

Read

table = df.read_table(conn, "SELECT * FROM sales.public.orders LIMIT 1000")
pdf   = table.to_pandas()        # pandas
# import polars as pl; pl.from_arrow(table)   # polars

Write a DataFrame

import pandas as pd

frame = pd.DataFrame({"id": [1, 2, 3], "region": ["us", "eu", "us"], "qty": [10, 20, 30]})

# Append (default). Also: mode="replace" (overwrite), mode="upsert".
df.write_dataframe(conn, "sales.public.orders", frame, mode="append")

# Idempotent append: re-sending the same key is a no-op, not a duplicate.
df.write_dataframe(conn, "sales.public.orders", frame, mode="append",
                   idempotency_key="nightly-2026-05-31")

pandas, polars, and pyarrow inputs are all accepted; the target table must already exist.

Runnable examples

See examples/: read.py and write_dataframe.py are self-contained scripts driven by the environment variables above.