[BradyPlanden]

Description

Makes "rust" a valid convert_to_format. A rust model's rhs, algebraic, Jacobian, event and initial-condition functions are compiled to Rust and evaluated by ScipySolver, AlgebraicSolver and EIS. The default stays "casadi".

• Compilation:
  • BaseSolver.process compiles each function into a CompiledFunction on its own ExprGraph. Inputs are registered in stacking order at their widths (rust_graph_with_inputs).
  • RustEvaluator wraps the compiled function with CasADi's (t, y, inputs) calling convention. It has four roles: the value, the Jacobian (a scipy CSC built on first use), a Jacobian-vector product, and jacp, with one column per sensitivity parameter.
• Inputs:
  • stack_inputs gains convert_to_format="casadi", and returns a flat numpy array for rust models. Every caller passes the model's format (EIS, inside its rust branch, passes "rust"), and a one-argument call behaves as in v26.10.0.
  • Input-dependent discontinuity times and Modulo periods are evaluated through the model's own compiled functions, with the inputs stacked in its format, as its residual is.
  • An input whose number of values differs from its parameter's width raises pybamm.SolverError naming the input. So does a vector-width input that only a variable reads and that has no value, which before raised a bare ValueError from Rust.
• Root method: a CasADi root method converts a rust model to CasADi with a warning, as it does any model not in CasADi form; feat(rust): solve rust models with IDAKLUSolver #5962 keeps rust models on rust.
• Other solvers:
  • IDAKLUSolver, CasadiSolver and CasadiAlgebraicSolver convert rust models to CasADi, with a warning; feat(rust): solve rust models with IDAKLUSolver #5962 adds IDAKLU's Rust path.
  • EIS evaluates the model's Jacobian through Rust. Its output Jacobian is still built with CasADi.
• Default-path edits: the ScipySolver event wrapper returns np.asarray(...).item(), and EIS calls .sparse() only on CasADi Jacobians, so a python or jax model's Jacobian goes to csc_matrix as it is.
• Unused here: validate_rust_sensitivity_widths is added in this PR, and feat(rust): solve rust models with IDAKLUSolver #5962 calls it.

Not supported in this PR:

• return_jacp_stacked raises SolverError, before anything is lowered or compiled.
• Consistent initialisation's root solve, in NonlinearSolver, still builds its residual with CasADi, so no DAE solver keeps a rust model until feat(rust): solve rust models with IDAKLUSolver #5962.
• ScipySolver sensitivities raise NotImplementedError, as on main.

Type of change

New feature, under ## Features in CHANGELOG.md.

Key checklist

• No style issues: pre-commit run
• Unit tests pass (N), as do the integration tests (N)
• Tests added: 28 Python test functions.
  • Evaluator parity: each RustEvaluator role against CasADi on a toy DAE (rtol 1e-12, atol 1e-14), including events, initial conditions and a rectangular Jacobian.
  • Plumbing:
    • stack_inputs for rust models, and called with one argument;
    • the input-width SolverError at widths 1 and 3, and for a missing vector input;
    • return_jacp_stacked rejected before lowering;
    • dense-matrix products (TestDenseMatrixToRust) and the conftest_rust fixtures from feat(rust): lower expression trees with Symbol.to_rust #5959, with rust_backend added here;
    • pickled evaluators.
  • ScipySolver: a solve with inputs and an event, on casadi and rust; a rust model with an input-dependent switch time against the exact solution; numpy inputs reaching the evaluators.
  • Solver parity:
    • AlgebraicSolver on rust against casadi (rtol 1e-8);
    • SPM impedance on rust against casadi (rtol 1e-6).
  • Conversion and serialisation: the CasADi solvers convert and solve a rust model; the default root method keeps the rust format; "rust" round-trips through serialisation.

Rust execution engine stack: proposal #5667, showcase #5732

Each PR is one commit on the one above it.

#	PR	Adds
0	#5954	Function.create_copy keeps the function's class
0b	#5955	CasADi's RegPower and uniform-grid interpolant lookups match evaluate
1	#5956	pybamm-core: expression graph, typed IR, interpreters, forward-mode AD
2	#5957	pybamm-core: CompiledModel, coloured Jacobians, the C ABI
3	#5958	pybamm.rust bindings, built into the pybamm wheel
4	#5959	Symbol.to_rust
5	#5960	convert_to_format = "rust" for ScipySolver, AlgebraicSolver and EIS ← this PR
6	#5961	pybammsolvers: IDAKLU over the C ABI
7	#5962	IDAKLUSolver on rust models
8	#5963	variables and sensitivities read in Rust
9	#5964	pybamm-core: diffsol DAE solver
10	#5965	pybamm.DiffsolSolver
11	#5966	"rust" as the default convert_to_format

[codecov]

Codecov Report

❌ Patch coverage is 95.08197% with 6 lines in your changes missing coverage. Please review.
✅ Project coverage is 98.18%. Comparing base (595db2e) to head (d158fdb).

Files with missing lines	Patch %	Lines
packages/pybamm/src/pybamm/solvers/base_solver.py	91.80%	5 Missing ⚠️
...ges/pybamm/src/pybamm/simulation/eis_simulation.py	88.88%	1 Missing ⚠️

Additional details and impacted files

@@                      Coverage Diff                      @@
##           stack/04-to-rust-lowering    #5960      +/-   ##
=============================================================
- Coverage                      98.18%   98.18%   -0.01%     
=============================================================
  Files                            351      353       +2     
  Lines                          36399    36505     +106     
=============================================================
+ Hits                           35738    35841     +103     
- Misses                           661      664       +3     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[github-actions]

Bencher Report

Project	PyBaMM
Branch	stack/05-rust-generic-evaluators
Testbed	bare-metal

⚠️ WARNING: Truncated view!

The full continuous benchmarking report exceeds the maximum length allowed on this platform.

🐰 View full continuous benchmarking report in Bencher