pantr 0.6.0

Polynomial And NURBS Toolkit

PaNTr

Polynomial and NURBS Toolkit (PaNTr) is a pure Python 3.11–3.14 library for geometric modeling and numerical analysis using NumPy, SciPy, and Numba.

Documentation: pantr.readthedocs.io

Features

• B-spline & NURBS spaces — univariate and tensor-product BsplineSpace, exact rational (NURBS) geometry, evaluation and derivatives, knot insertion/removal, degree elevation, and splitting.
• Bézier toolkit — Bernstein/Bézier curves and patches, composition, products, degree reduction, and Bernstein-polynomial root finding.
• Truncated hierarchical B-splines — THBSplineSpace with local refinement, mirroring the tensor-product API.
• Constructive geometry (pantr.cad) — lines, circles and arcs, disks, cylinders, extrusion, revolution, sweep, ruled and Coons surfaces/volumes.
• Structured grids (pantr.grid) — tensor-product and hierarchical grids, BVH spatial queries, dolfinx-style cell/facet tags, and cell quadrature.
• Quadrature and change of basis — Gauss–Legendre and tensor-product rules, plus exact change of basis operators between Bernstein, Lagrange, monomial, and cardinal B-spline bases.
• Fast and typed — Numba-JIT kernels parallelized over CPU cores, with strict type hints across the public API.
• Dependency-gated MPI and visualization — pantr.mpi distributes spaces across ranks once mpi4py is installed, and pantr.viz renders exact higher-order geometry through PyVista / VTK once PyVista is installed. Both modules ship with PaNTr; only their backends are optional.

Installation

pip install pantr

Requires Python 3.11–3.14. The serial core depends only on NumPy, SciPy, Numba, and threadpoolctl.

To install the latest development version from source:

git clone https://github.com/FELIGN/pantr.git
cd pantr
pip install .

Quick start

A geometry is a function space plus control points; build one and evaluate it:

import numpy as np
from pantr.bspline import Bspline, BsplineSpace, BsplineSpace1D

# Quadratic 1-D space on the knot vector [0, 0, 0, 1, 2, 3, 3, 3]
space = BsplineSpace([BsplineSpace1D([0, 0, 0, 1, 2, 3, 3, 3], 2)])

# Five 2-D control points -> a planar curve
control_points = np.array(
    [[0.0, 0.0], [1.0, 2.0], [2.0, -1.0], [3.0, 1.0], [4.0, 0.0]]
)
curve = Bspline(space, control_points)

# Evaluate at 50 parameters spanning the domain [0, 3]
points = curve.evaluate(np.linspace(0.0, 3.0, 50))  # shape (50, 2)

See the getting-started guide for a full walkthrough from this first curve to NURBS, CAD modeling, and refinement.

Optional features

Every install of PaNTr includes all of its modules — pantr.viz and pantr.mpi among them. Those two stay dormant until their third-party backend is importable, and activate automatically once it is; until then, calling into them raises a clear, actionable error (nothing else is affected). The "extras" below add no PaNTr code — they are just a convenience that installs the backend for you, so pip install "pantr[viz]" and pip install pyvista are equivalent.

Capability	Module	Backend it needs	Install (extra or direct)
Visualization & VTK export	pantr.viz	pyvista (+ VTK)	pip install "pantr[viz]" · or pip install pyvista
Distributed (MPI) spaces	pantr.mpi	mpi4py (+ an MPI library)	pip install "pantr[mpi]"
METIS partitioning backend	pantr.bspline.partition_graph	pymetis	pip install "pantr[metis]"

The serial core (pantr.grid, pantr.bspline, …) never imports pantr.mpi, so a plain pip install pantr behaves identically with or without these backends. Contributors can grab everything — including the docs toolchain (Sphinx) — with pip install -e ".[dev]".

Development

pip install -e ".[dev]"   # includes all optional feature extras

License

PaNTr is licensed under the MIT License. See LICENSE for details.