Usage¶

AkimaSpline (see the API reference reference and the Quick Start on the pyakima page) covers ordinary Python callers. This page covers the Numba-facing helpers and the option flags shared by both interfaces.

Jitted Use¶

Use make_akima_coeffs and cubic_call when the spline should be created or evaluated from inside an njit function.

import numpy as np
from numba import njit

from pyakima import make_akima_coeffs, cubic_call


@njit
def build_and_evaluate(x: np.ndarray, y: np.ndarray, x_eval: np.ndarray) -> np.ndarray:
    coeffs = make_akima_coeffs(x, y, corner_model=2)
    return cubic_call(x_eval, coeffs, 0)

For lower-level control, call cubic_call_scalar, cubic_call_vector, or cubic_call_vector_linear directly.

Spline Options¶

pyakima supports several Akima corner models:

`AkimaSpline` option	helper option	Behavior
`"non-rounded"`	`0`	Wodicka/GSL-style non-rounded sharp corners.
`"akima"`	`1`	Classic Akima behavior, close to `scipy` `method="akima"`.
`"makima"`	`2`	Modified Akima weights, close to `scipy` `method="makima"`.

Boundary handling uses SciPy-like ext values:

`ext`	Out-of-bounds behavior
`0`	Extrapolate.
`1`	Return zero.
`3`	Return the nearest boundary value.
`4`	Return `nan`.

ext=2 (raise on out-of-bounds) is not implemented. ext=4 is added for NaN boundary handling. AkimaSpline defaults to ext=3.