Release notes

convinterp

The notes of every release, as published on GitHub, newest first. The wheels of each release are on PyPI, and each release is archived on Zenodo under its own DOI; the citation file in the repository has the details.

v0.2.0 — 2026-09-26

Integrals of any order, in 1D and N-D.

New
  • Integrals. A negative derivative order gives the integral of the interpolant: derivative=-1 is the antiderivative, -2 the twofold integral, and so on, up to order 8 for "b7" to "b13". Each integral is anchored at the first knot, where it and all lower integrals are zero.
  • Mixed orders in N-D. Integrals can be combined with values and derivatives, one order per axis, e.g. derivative=(-1, 1).
  • Fast and exact. Integrals are computed exactly from the kernels’ polynomial pieces, converge at the kernel’s interpolation order, and evaluate in constant time per point, about as fast as the interpolant itself.
  • Documentation. An Integrals section in the user guide, a theory section on how integrals work, integral convergence and speed on the accuracy page, and a new Limitations section.
Breaking changes

None.

Migration guide

No changes needed. Negative derivative orders, which previously raised a ValueError, now return integrals.

v0.1.1 — 2026-09-26

Bug-fix release.

Fixes
  • Knots containing NaN or infinity are now rejected with a ValueError. Previously a NaN inside the grid passed the uniformity check and gave silently wrong results.
  • bc="quadratic" with only 2 data values along an axis now raises a ValueError, instead of an internal Rust panic (PanicException), for every kernel from a3 up.
Breaking changes

None.

Migration guide

No changes needed. Code that passed non-finite knots, or bc="quadratic" with only 2 values, now gets a clear ValueError.

v0.1.0 — 2026-09-26

First alpha release of convinterp, a Python port of ConvolutionInterpolations.jl with a Rust core.

  • Interpolation on uniform grids in any number of dimensions
  • Derivatives, with a derivative order per axis, up to each kernel’s highest order
  • Kernels a0 to a7 and b5 to b13, boundary conditions poly, linear, quadratic and detect
  • Wheels for Linux, macOS and Windows on x86_64 and ARM

Breaking changes

None: this is the first release.

Migration guide

Not applicable: this is the first release.