Oven logo

Oven

Published

A generic entrypoint for ADBC drivers.

ADBC Driver Manager for Python

This package contains bindings for the ADBC Driver Manager, as well as a DBAPI 2.0/PEP 249-compatible interface on top. This can be used to load ADBC drivers at runtime and use them from Python. Backend-specific packages like adbc_driver_postgresql wrap this package in a more convenient interface, and should be preferred where they exist.

The DBAPI 2.0 interface requires PyArrow, and exposes a number of extensions mimicking those of Turbodbc or DuckDB's Python packages to allow you to retrieve Arrow Table objects instead of being limited to the row-oriented API of the base DBAPI interface.

Building

Dependencies: a C++ compiler.

For users building from the arrow-adbc source repository, you can alternately use CMake to manage library dependencies and set environment variables for you. You can add -DADBC_BUILD_PYTHON=ON to define a python target.

For example, assuming you run cmake from the project root:

cmake -S c -B build --preset debug -DADBC_BUILD_PYTHON=ON
cmake --build build --target python

will properly build and install the Python library for you.

See CONTRIBUTING.md for details.

Testing

The SQLite driver must be loadable at runtime (e.g. it must be on LD_LIBRARY_PATH, DYLD_LIBRARY_PATH, or PATH).

See CONTRIBUTING.md for details.

$ export LD_LIBRARY_PATH=path/to/sqlite/driver/
$ pytest -vvx