V0.17 Scope - Formula-backed Generator Families#
Summary#
v0.17 is a narrow formula-backed generator interoperability release.
Stable path:
canonical scalar 1D periodic FieldBatch + formula-backed generator candidate -> safe formula metadata/evaluation diagnostics -> empirical configured validation report
This release adds formula-backed runtime records without changing existing polynomial GeneratorFamily semantics.
Stable Scope#
v0.17 adds submodule-only runtime APIs:
pdelie.symmetry.FormulaGeneratorFamilypdelie.reporting.summarize_formula_generator_family(...)pdelie.symmetry.validate_symmetry_candidate(...)support for formula-backed candidates
FormulaGeneratorFamily is runtime-only. It is not a new canonical object and does not change canonical GeneratorFamily.
Formula Record#
Frozen record properties:
schema_version = "0.1"parameterization = "formula_generator_family"supported variables are exactly
t,x, andusupported components are exactly
tau,xi, andphieach generator row has a nonempty
nameeach generator row must define all three components
optional
finite_transform_specis a canonicalInvariantMapSpecpayloaddiagnosticsare JSON-compatible provenance only
Formula records support .to_dict() and .from_dict() round trips.
Safe Expression AST#
Formula expressions use a small JSON AST. They are not Python code and are not parsed from arbitrary executable strings.
Supported nodes:
constvaraddmulinteger
powsincosreciprocalmetadata-only
symbolic_reference
Rational expressions use reciprocal with a fixed denominator floor. Denominator-floor violations are empirical validation failures.
Symbolic references are metadata only. They may be summarized, but they are not evaluated.
Candidate Validation#
validate_symmetry_candidate(...) accepts:
FormulaGeneratorFamilyobjectsstrict current
FormulaGeneratorFamilypayload mappings
Formula reports use:
summary_schema_version = "0.1"summary_type = "symmetry_candidate_validation"candidate_kind = "formula_generator_family"
Validation behavior:
malformed formula records raise typed validation errors
well-formed formulas that evaluate to nonfinite values return
conclusion = "failed"formulas with finite evaluable components but no finite transform are generally
partially_validatedsymbolic-reference-only formulas are reporting/schema results and do not pretend to evaluate
formulas with supported passing finite-transform checks may be
validatedvalidatedremains empirical configured validation, not a mathematical proof of symmetry
Reporting#
summarize_formula_generator_family(...) returns a JSON-compatible runtime summary with:
formula parameterization and schema version
variables and component names
generator count and names
expression node kinds
component root-node summaries
finite-transform availability
symbolic-reference metadata
reciprocal denominator floor
diagnostics
The helper does not mutate inputs and does not evaluate arbitrary code.
Non-goals#
v0.17 does not add:
Python callable generator APIs
arbitrary executable formula-string parsing
neural generator training
learned-generator detector APIs
formula-derived finite-flow integration for arbitrary infinitesimal formulas
canonical
GeneratorFamilyschema changesnew PDEs
stable KS generator/residual/example APIs
weak KS
broad dataset adapters
PDEBench or The Well support
multidimensional, multivariable, or nonuniform-grid expansion
train/test policy
split management
operator-facing APIs
root
pdelieexports
Milestone Status#
Milestone 0: COMPLETE - scope freeze
Milestone 1: COMPLETE - formula schema and validation semantics freeze
Milestone 2: COMPLETE - formula record and reporting implementation
Milestone 3: COMPLETE - formula candidate validation
Milestone 4: COMPLETE - formula validation example
Milestone 5: COMPLETE - API/public-surface audit
Milestone 6: COMPLETE - release gate and readiness
Release Gate Expectations#
v0.17 is complete only if:
formula records round-trip through strict JSON-compatible payloads
invalid formulas raise typed validation errors
formula summaries are JSON-compatible and submodule-only
formula candidates validate through
validate_symmetry_candidate(...)finite formula diagnostics and denominator-floor failures are covered
attached supported finite-transform specs reuse existing invariant-map validation
existing
GeneratorFamilyandInvariantMapSpeccandidate validation behavior remains unchangedformula validation example emits JSON only
no callable, executable-string, neural-detector, KS, weak KS, broad-adapter, operator, split-policy, or root-export surface lands
package/readiness docs preserve the
v1.0package-index publishing deferral