Changelog

v0.5.0 - 2026-10-07

Refactored Re-exports: The package has moved away from heavy use of @reexport. Only the following symbols from other packages are still exported by LyoPronto; others (such as from ConstructionBase, DiffEqCallbacks, and OrdinaryDiffEqNonlinearSolve must now be explicitly imported.
- ODEProblem and solve from OrdinaryDiffEqRosenbrock remain explicitly exported.
  - Other OrdinaryDiffEqRosenbrock symbols are no longer re-exported, notably algorithms like Rodas3().
  - The recommended algorithm for LyoPronto's ODE and DAE systems is now Rodas5P(AutoForwardDiff(chunksize=2)), which is made public as LyoPronto.odealg_chunk2.
- From Unitful, the u"" macro, ustrip, uconvert, and NoUnits remain explicitly exported.

Experimental Data Handling: The PrimaryDryFit structure is now deprecated in favor of a new, modular hierarchy based on ExpFitData and AbstractExpDatum.
- Users should migrate from PrimaryDryFit to ExpFitData (e.g., using TfData, TvwSeriesData, etc.).
- For help with migration, see the documentation for ExpFitData in the package API, refer to the AbstractExpDatum docstring, and consult the examples in the documentation.
Parameter Objects: Model parameters are now documented as handled via structured, typed objects. Tuple-of-tuples data structures are only used for constructing the parameter objects.
Loss and Residual Functions: the functions obj_expT, err_expT, and err_expT! are removed and replaced by obj_exp, err_exp, and err_exp!, which are more general and can handle multiple experimental data types.
Residual Weighting: The API for residual weighting has been updated to be more consistent across different experimental data types. The functions loss_weighting and residual_weighting are now used to specify weighting schemes for residuals in fitting, and these are passed to obj_exp and err_exp! functions. Weights must explicitly specify the units so that they produce a dimensionless quantity when multiplied by the residuals.

Rather than pass a Val(true) to the ODE RHS functions, to get intermediate calculated outputs, call the function calc_md_Q which dispatches on the parameter type to decide which model to use. This function returns a NamedTuple of the intermediate outputs, which can be used for plotting or analysis. The Val(true) argument is no longer accepted by the ODE RHS functions.

The function qrf_integrate now returns a NamedTuple rather than a Dict with string keys.
The fields Arad and alpha in ParamObjRF are removed, as they were unused.

Introduced a robust hierarchy for experimental data:
- ExpFitData: Container for multiple experimental datasets.
- AbstractExpDatum: Base type for specific measurement types including TfData, TvwSeriesData, TvwEndData, and EndTimeData.
Added SolTrim and trim_sol for precise alignment of model solutions with experimental time grids.

Namespace Cleanup: Reduced namespace pollution by limiting re-exports.
Error Handling: Improved validation in parameter constructors (e.g., checking units and dimensions).

A semi-formal changelog for previous versions can be found in the GitHub releases.