Skip to content

CLI and generated views

Commands

model-parser parse   <authoring-file> [--from exprtk-ini] [-o out.ir.json]
model-parser emit julia      <model.ir.json>             [-o out.jl]
model-parser emit julia-rhs  <model.ir.json>             [-o out.jl]
model-parser emit python     <model.ir.json>             [-o out.py]
model-parser validate <model.ir.json | authoring-file> [--profile <name>]
model-parser inspect  <model.ir.json | authoring-file>
model-parser diff     <old.ir.json> <new.ir.json>         [--json]
model-parser bump     <old.ir.json> <new.ir.json>         [--json]
model-parser ast      <authoring-file>                      [-o out.json]
model-parser schema                                         [-o schema.json]
  • parse turns an authoring file into canonical IR. The only frontend is exprtk-ini. [x0]/[u0] sections are dropped with a WARN because initial values are scenario data.
  • emit <target> lowers IR into a view: julia (ModelingToolkit v11 builder), julia-rhs (plain f!/outputs!), python (plain RHS and output functions).
  • validate accepts IR or an authoring file (parsed on the fly) and an optional profile.
  • inspect prints a summary, ast exports the debug parse tree, schema exports the IR JSON Schema.
  • diff lists semantic changes between two IR files; bump suggests a model SemVer bump (see Model library & versioning).

Output goes to stdout unless -o is given. Exit codes: 0 success, 1 validation errors, 2 usage or load failure. Diagnostics use OK / WARN / ERROR.

Set SOURCE_DATE_EPOCH to make provenance.created_at, and therefore committed IR files, byte-stable.

Typical session

model-parser parse  examples/models/model_monod_simple.ini -o monod.ir.json
model-parser validate monod.ir.json --profile julia-analysis
model-parser emit julia     monod.ir.json -o monod.jl
model-parser emit julia-rhs monod.ir.json -o monod_rhs.jl
model-parser emit python    monod.ir.json -o monod.py

Profiles

  • julia-analysis: permissive, the full IR is allowed.
  • realtime-cpp: restricted deterministic operator/function subset for PLC targets. Validation only; there is no C++ backend yet.

Generated views

All views are generated from the IR and must not be edited by hand; regenerate instead.

emit julia writes a build_<model>() function that returns a compiled MTK System. It uses System(eqs, t) and mtkcompile, never @mtkmodel or structural_simplify, and renders numeric literals as Float64. Initial values and parameter overrides are supplied when building the problem:

include("monod.jl")
sys = build_monod_simple()
prob = ODEProblem(sys, [sys.x0 => 0.05, sys.x1 => 15.0], (0.0, 24.0),
                  [sys.mu_max => 0.4])

emit julia-rhs writes f_<model>!(du, u, p, t) (or f_<model>!(du, u, p, t, inp) when the model has inputs) and, when the IR has output equations, outputs_<model>!(y, u, p, t[, inp]).

emit python writes f_<model>(t, state, parameters[, inputs]), which returns the derivative list in SciPy argument order, and outputs_<model> with the same arguments when output equations exist. Generated modules use only the standard library.

Packing for both numerical views: states, parameters, inputs and outputs follow their IR declaration order. Parameter defaults are not applied; callers pass numeric values. Conditionals are eager: both branches are evaluated before selection.