arco 0.8.0

Python bindings for Arco optimization library

Arco Python bindings

The Python package remains in bindings/python for path compatibility with existing build, release, and editable-install workflows. Its Rust crate is named arco-python and is the Python interaction surface; public Python imports stay under arco.

Python-facing solve orchestration is routed through the shared arco-ops facade where it overlaps with other interaction surfaces. The public Python API is unchanged.

Build and install locally with uv:

cd bindings/python
uv sync --group dev
uv run --with maturin maturin develop

Default Python builds include the runtime-loaded Xpress backend. Solving with arco.Xpress(...) still requires the FICO Xpress runtime and a valid license on the target machine.

To enable the IPOPT nonlinear backend, build with the ipopt feature (requires a system IPOPT install):

cd bindings/python
uv run --with maturin maturin develop --features ipopt

To disable Xpress for a source build, turn off default Cargo features:

cd bindings/python
uv run --with maturin maturin develop --no-default-features --features pyo3/extension-module

Without this feature, arco.Xpress(...) is importable but solve will fail fast with a rebuild hint.

Run linting:

uv run ruff check .
uv run ty check .

Run Python example formulations from the repository root:

cd ../..
uv run examples/dense-lp/formulation.py --solve --json
uv run examples/sdom/formulation.py --solve --json

For interactive exploration of dense-lp (no extra script boilerplate):

cd ../..
uv run --with ipython --with-editable ./bindings/python ipython -i examples/dense-lp/formulation.py

Inside IPython, use model to inspect the formulation and call solve() when ready.

Running example problems

The examples/ tree contains standalone Python scripts that build models directly through the bindings, covering different problem classes:

• LP — linear programs (default HiGHS backend).
• MILP — mixed-integer linear programs (HiGHS).
• NLP — nonlinear programs (requires bindings built with --features ipopt).
• QP / QCP — (quadratically constrained) quadratic programs, solved through the appropriate backend for the problem class.

Run from bindings/python so the locally built extension is on the import path:

cd bindings/python

# LP — Multi-period DC-OPF (HiGHS)
uv run python ../../examples/multi-period-optimal-power-flow/dc-opf-24bus-wind-load-shedding/problem.py

# NLP — Multi-period AC-OPF (IPOPT)
uv run python ../../examples/multi-period-optimal-power-flow/ac-opf-24bus-wind-load-shedding/problem.py

Each script prints the solver status and final objective value alongside the reference value from the original formulation. Substitute the path to any other problem.py (or formulation.py) under examples/ to run a different model.