Free-parameter records (pybnf.parameter_record)

The new-era parameter: record -> FreeParameter (ADR-0043).

The edition-2 free-parameter declaration is a fully labeled record – parameter: <id>, prior: <family>, <field>: <v>, ..., lower: <lo>, upper: <hi>, initial_value: <v> – where the legacy positional <family>_var = <id> p1 p2 line names nothing. It is also the only grammar that carries lower/upper, so a truncated prior (two-sided, ADR-0020; half-bounded, ADR-0047) can be written no other way.

This module is the one mapping from that record onto a FreeParameter, with two consumers. pybnf.config.Configuration reads it when it loads a job to fit, and pybnf.petab.export reads it when it serializes a job to a PEtab v2 problem. They must agree: a PEtab row and a native declaration landing on the same object is the two-adapter proof (ADR-0004), and an exporter that re-derived the record grammar for itself would drift from the one the fitter actually runs. It lived as a private Configuration method until the exporter needed it, and was silently skipping every parameter: record for want of it (#733).

Light by design (printing / priors / pset / numpy): the exporter reads a job through the stdlib ploop parser without building a Configuration, and pulling the whole configuration layer in to reach one builder would defeat that.

pybnf.parameter_record.free_parameter_from_record(pid, raw_fields, initialization_distribution)

Build a FreeParameter from a new-era parameter: record (ADR-0043).

raw_fields is the parsed {field: str} map – every part of the line is named: prior (the family), space (linear/log10, the sampling-space transform), the family’s own distribution fields (mean/sd, location/scale, …), lower/upper (the bounds that truncate the prior – #417/ADR-0020), and initial_value (the start point). No positional numbers; the family names its fields via Prior.field_names. The truncation/box capability is unchanged – this only maps named fields onto the existing FreeParameter constructor.