Simulation results

[AlexS12]

At the moment, only the state vector is stored as a result of the simulation process:

  sol = solve(prob, solver; solve_args...)
  df = DataFrame(sol')
  rename!(df, get_x_names(dss))
  df[!, :time] = sol.t
  return df

However, given the library architecture, based on different types for different physical variables, it would be useful to store this structs during the integration process (ie. atmosphere, dssd, atmosphere...).

• Store data from different types (DSStateDot and Atmosphere would be enough for the demonstrator)
  • Save using callbacks: https://diffeq.sciml.ai/stable/features/callback_library/#saving_callback
  • Using data arrays: https://diffeq.sciml.ai/stable/features/diffeq_arrays/#DEDataArrays
  • Use this comparison to decide.

However, if the equation is solved in this manner (parameters, no data arrays), there will be no record of what the parameter was at each timepoint. That is the tradeoff between DEDataArrays and ParameterizedFunctions.

• It would be desirable to have a handy interface such as https://diffeq.sciml.ai/stable/basics/solution/

[AlexS12]

Useful information:

https://discourse.julialang.org/t/ordinarydiffeq-unexpected-results-in-dedataarray/59282

• DEDataArrays are not useful for this purpose and will be deprecated.
• SimulationLog.jl https://github.com/jonniedie/SimulationLogs.jl might be useful if dependence of the logged variables is current x and t (no internal state).
• A DiscreteCallback from OrdinaryDiffEq might be useful for saving. However, ModellingToolkit.jl is well worth a look.