zoomy_core.model.legacy.numerical_model module

NumericalModel: wraps any analytical Model for numerical simulation.

Two-level regularization:

eps (small, ~1e-8) : regularizes 1/h → 1/(h+eps) in flux/source/quasilinear eps_wet (larger, ~1e-3) : wet/dry threshold for eigenvalue skip and momentum deletion

Momentum ramp: update_variables zeros out momentum-like variables when h < eps_wet, preventing unphysical velocities at near-dry cells from driving the solution.

Usage:

analytical = GeneratedShallowModel(…) numerical = NumericalModel(analytical)

zoomy_core.model.legacy.numerical_model.regularize_denominators(expr, h_sym, eps_sym)

Replace h^(-n) with (h + eps)^(-n) for n > 0.

zoomy_core.model.legacy.numerical_model.regularize_sqrt_arguments(expr, h_sym, eps_sym)

Replace h inside fractional powers with (h + eps) to prevent sqrt(negative).

class zoomy_core.model.legacy.numerical_model.NumericalModel(analytical_model, eigenvalue_proxy_level=None, **kwargs)

Bases: Model

Wraps an analytical Model for numerical simulation.

Regularization: - 1/h → 1/(h + eps) in all compiled expressions - Momentum zeroed when h < eps_wet (via conditional in update_variables) - Eigenvalues use conditional(h > eps_wet, ev, 0) for symbolic path

flux()

Flux.

hydrostatic_pressure()

Hydrostatic pressure.

nonconservative_matrix()

Nonconservative matrix.

source()

Source.

quasilinear_matrix()

Quasilinear matrix.

eigenvalues()

Eigenvalues for the model. Used by the backend’s max_wavespeed implementation. No dry-cell guard here — that belongs in the backend function, not in the symbolic expressions.

update_variables()

Wet/dry treatment applied every timestep via opaque numerical functions: 1. clamp_positive(h) — prevents negative water depth 2. clamp_momentum(hu, h, u_max) — caps velocity at sqrt(g*h) + eps_wet 3. Linear ramp via conditional — zeros momentum below eps_wet

get_field_map()

property wet_dry_threshold

property eigenvalue_regularization_eps