Private API

Solution fields that are always serialized as Matrix{Float64} and must be converted to Vector{Vector} for JSON3.

Solution fields that may be present (non-empty duals) and require the same Matrix → Vector{Vector} conversion.

Apply a function over a grid (broadcast), or return nothing if input is nothing.

Convert Matrix fields to Vector{Vector} for JSON3 export.

JSON3 flattens Matrix{Float64} into 1D arrays, losing the 2D structure. This function converts all matrix fields to Vector{Vector} format to preserve dimensions.

Convert Dict{String,Any} back to Dict{Symbol,Any} after JSON deserialization. Uses symbol_keys metadata to restore Symbol types where they were originally present.

Deserialize a single value from JSON format. Uses symbol_keys to restore Symbol types at the correct paths.

Convert a JSON field (Vector{Vector} via stack) to Matrix{Float64}.

JSON exports matrices as Vector{Vector}. After stack(blob[field]; dims=1), we get either a Matrix (multi-D) or Vector (1D). This normalizes to Matrix.

Arguments

• blob_field: JSON array field (Vector of Vectors)

Returns

• Matrix{Float64}: (ntimepoints, n_dim)

Convert an optional JSON field to Matrix{Float64} or nothing.

Arguments

• blob_field: JSON array field or nothing

Returns

• Matrix{Float64} or nothing

Convert Dict{Symbol,Any} to Dict{String,Any} for JSON serialization. Only serializes JSON-compatible types (numbers, strings, bools, arrays, dicts). Returns a tuple: (serializeddict, symbolkeys) where symbol_keys tracks which values were Symbols.

Serialize a single value to JSON-compatible format. Returns a tuple: (serializedvalue, symbolpaths) where symbol_paths tracks Symbol locations.

_extract_time_vector(time_data) -> Any

Extract time vector from various data formats.

Arguments

• time_data: Time data in various formats (Vector, Matrix, etc.)

Returns

• Vector{Float64}: Time vector

Notes

• Handles both Vector{Float64} and Matrix{Float64} (single column) formats
• Used by JSON and JLD2 importers to normalize time grid data

See also: _reconstruct_solution_from_data.

_reconstruct_solution_from_data(
    ocp,
    data;
    infos
) -> CTModels.Solutions.Solution{TimeGridModelType, TimesModelType, StateModelType, ControlModelType, VariableModelType, ModelType, CostateModelType, Float64, DualModelType, CTModels.Solutions.SolverInfos{Any, Dict{Symbol, Any}}} where {TimeGridModelType<:CTModels.Solutions.AbstractTimeGridModel, TimesModelType<:CTModels.Components.AbstractTimesModel, StateModelType<:CTModels.Components.AbstractStateModel, ControlModelType<:CTModels.Components.AbstractControlModel, VariableModelType<:CTModels.Components.AbstractVariableModel, ModelType<:CTModels.Models.AbstractModel, CostateModelType<:Function, DualModelType<:CTModels.Solutions.AbstractDualModel}

Reconstruct a solution from imported data, detecting the format (single vs multiple time grids).

Duals and control_interpolation are read from data. Only infos is accepted as a keyword argument because its deserialization is format-specific (JSON restores Symbol types).

Arguments

• ocp: The optimal control problem model
• data: Dictionary containing the imported solution data, including all dual fields and control_interpolation

Keyword Arguments

• infos: Solver information dictionary (Dict{Symbol,Any}). Passed explicitly because JSON deserialization must restore Symbol types before calling this helper.

Returns

• Solution: Reconstructed solution with appropriate time grid model

Notes

• If time_grid_state key exists, assumes multiple time grid format
• Otherwise, uses the current unified format (single time_grid key)

Example

julia> sol = _reconstruct_solution_from_data(ocp, data; infos=infos)

See also: _extract_time_vector.