archimedes.quadrature.quadint¶

archimedes.quadrature.quadint( func: Callable[[...], ndarray], a: float, b: float, n: int = 20, rule: str = 'legendre', axis: int = -1, args: Sequence[Any] | None = None, ) → ndarray¶

Approximate the definite integral of func over [a, b].

\[\int_a^b f(x) \, dx \approx \sum_{i=1}^n w_i f(x_i)\]

using a fixed-order Gauss quadrature rule (by default Gauss-Legendre quadrature).

This is a convenience wrapper around the rule(n, a, b).integrate(f) pattern (see QuadratureRule.integrate()) for one-off integrals on a finite interval; for repeated evaluation of the same rule (e.g. inside a loop or a function decorated with archimedes.compile()), build the QuadratureRule once with gauss_legendre(), gauss_radau(), gauss_lobatto(), or clenshaw_curtis() and call .integrate directly instead.

Parameters:

func (callable) – Integrand. Called once as func(x, *args) on the full array of quadrature nodes, so must be vectorized. Must return values with the nodes along axis.
a (float) – Bounds of integration. Must both be finite (see Notes).
b (float) – Bounds of integration. Must both be finite (see Notes).
n (int, optional) – Number of quadrature nodes. Default 20.
rule ({"legendre", "radau_left", "radau_right", "lobatto", "clenshaw_curtis"}, optional) – Quadrature family to use, all defined on [a, b] with uniform weight. Default "legendre" (no fixed endpoints, exact to degree 2n - 1). Use "radau_left"/"radau_right" to fix the left/right endpoint as a node, or "lobatto" to fix both; see gauss_radau() and gauss_lobatto() for the accompanying loss of exactness.
axis (int, optional) – Axis holding the nodes in the output of func. Default -1.
args (tuple, optional) – Extra arguments passed to func after x.

Returns:

value – Approximated integral. Shape (m,) for vector-valued integrands, or () for scalar integrands.

Return type:

ndarray

Raises:

ValueError – If rule is not one of the supported names, or if a/b are not finite.

Notes

This function only supports finite [a, b]. Unlike scipy.integrate.quad, it does not adapt the node count or use a variable transformation to reach infinite/semi-infinite intervals; it is a thin wrapper around a single fixed-order rule.