Supported versions

Python >=3.12,<4 is accepted; CI currently tests CPython 3.12–3.15. Dependencies are NetworkX 3.x and NumPy >=2.3.3,<3. Python <3.12 and NumPy <2.3.3 are no longer supported.

The Python and NumPy floors stay within the September 2026 downstream support window in NEP 29, now superseded by SPEC 0. Both policies exclude Python 3.11 by this date. This compatibility window is narrower than CPython’s security support lifetime. Review the floors for future releases; SPEC 0 recommends dropping Python 3.12 in October 2026. The declared dependency minimums are the same for all Python versions. On Python 3.15 they provide no wheels, so CI tests the oldest releases that do instead: NumPy 2.5.2, SciPy 1.18.1 and Numba 0.68.0. These recommendations do not promise upstream bug fixes for every included NumPy release. CI runs the full suite with minimum and latest compatible dependencies, both with and without the numba extra, on every supported Python version.

The optional numba extra installs Numba >=0.63,<1 and SciPy >=1.16.1,<2. SciPy supplies the compiled linear algebra routines. Pip selects compatible versions; Numba may constrain NumPy more tightly than the ordinary installation. Acceleration depends on Numba/llvmlite platform support and adds installation size and initial compilation time. Standard, GIL-enabled CPython is tested; free-threaded builds and alternative interpreters are not covered by CI.

Quick Guide:

Install the package with the optional Numba extra for a significant speedup:

pip install "extremitypathfinder[numba]"

from extremitypathfinder import PolygonEnvironment

# counter clockwise vertex numbering!
boundary_coordinates = [(0.0, 0.0), (10.0, 0.0), (9.0, 5.0), (10.0, 10.0), (0.0, 10.0)]
# clockwise numbering!
list_of_holes = [
    [
        (3.0, 7.0),
        (5.0, 9.0),
        (4.5, 7.0),
        (5.0, 4.0),
    ],
]
environment = PolygonEnvironment(boundary_coordinates, list_of_holes, validate=False)
start_coordinates = (4.5, 1.0)
goal_coordinates = (4.0, 8.5)
path, length = environment.find_shortest_path(start_coordinates, goal_coordinates)

Grids with blocked cells (obstacles) can be converted into polygon environments as well. Coordinates refer to the corners of the grid cells, i.e. cell (x, y) spans [x, x+1] x [y, y+1]:

size_x, size_y = 5, 3
obstacle_iter = [(2, 1), (3, 1)]  # blocked grid cells (x, y)
environment = PolygonEnvironment.from_grid(size_x, size_y, obstacle_iter)
path, length = environment.find_shortest_path((0.5, 0.5), (4.5, 2.5))

For more refer to the documentation. Upgrading from version 2.x? See the upgrade guide.

Also see: GitHub, PyPI