zoomy_core.model.numerics

zoomy_core.model.numerics#

Symbolic numerics-control wrappers — applied at runtime by the solver.

These let the model author annotate PDE terms with a requested numerical treatment (gradient limiter, reconstruction scheme, …) without exposing the user to the discretisation layer. The wrappers are inert sympy Functions that the operator-form pipeline recognises and routes into aux_registry entries with extra metadata fields.

The OpenFOAM analogue: fvSchemes exposes a scheme dictionary that the user fills in per term (div(phi,U), grad(p), laplacian(nu,U) each get their own scheme). Our equivalent: the model author writes the PDE in operator-form using these wrappers, and the runtime resolves each one against the configured discretisation.

Currently exposed

limit(expr, scheme_symbol) Wrap a gradient/derivative atom with a runtime TVD limiter. The second argument is a sp.Symbol whose name names the scheme ("minmod", "venkatakrishnan", "barth_jespersen"). At runtime, compute_derivatives first computes the unlimited LSQ gradient, then applies the named limiter per cell, and substitutes the limited value into the source expression.

Usage example

    from zoomy_core.model.numerics import limit
    import sympy as sp

    h = sp.Symbol("h")
    x = sp.Symbol("x")
    minmod = sp.Symbol("minmod")
    S[1, 0] = - 2 * P_1 * limit(sp.Derivative(b, x), minmod)

Module Contents#

Classes#

limit

Symbolic gradient-limiter wrapper.

API#

class zoomy_core.model.numerics.limit#

Bases: sympy.Function

Symbolic gradient-limiter wrapper.

Signature:

    limit(expr, scheme)

where expr is the derivative atom to be limited at runtime (typically a sp.Derivative of a state/aux field), and scheme is a sp.Symbol whose name selects the limiter ("minmod", "venkatakrishnan", "barth_jespersen").

The operator-form pipeline picks these up in expose_aux_atoms and substitutes a fresh aux Symbol named {target}_{axes}__{scheme} (e.g. b_x__minmod). The chain solver’s aux refresh recomputes the LSQ gradient and applies the named limiter before the source expression is evaluated.

Inert symbolically — .doit() does NOT strip the wrapper (the runtime needs to see it).

nargs#

2

property inner#

The wrapped derivative atom.

property scheme#

Name of the limiter scheme (string).

doit(**hints)#
Contents