JOSS: Software > Design

[lamBOOO]

Concerning the JOSS review thread.

• Use of quote ... end in Problem Definition: The NLodeProblem utilizes the quote ... end construct, resulting in an Expr object. For example:

  odeprob = NLodeProblem(quote
      name=(sysb53,)
      u = [-1.0, -2.0]
      du[1] = -20.0*u[1]-80.0*u[2]+1600.0
      du[2] = 1.24*u[1]-0.01*u[2]+0.2
  end)

  This contrasts with a more conventional approach:

  ode_prob = NLodeProblem(
      name = sysb53,
      u0 = [-1.0, -2.0],  # Initial condition for u
      f = function (du, u)
          du[1] = -20.0 * u[1] - 80.0 * u[2] + 1600.0
          du[2] = 1.24 * u[1] - 0.01 * u[2] + 0.2
      end
  )

  • Concerns: This unconventional use of quote ... end may not be the most elegant or efficient method in terms of type inference, robustness, and API accessibility.
  • Recommendation: Please explain the rationale behind this design choice or consider adopting the more standard method of using function arguments in Julia. The current approach makes it unclear what the function does with the input provided within the quote ... end block.

[mongibellili]

Thank you for the comment.
In the revised package, the problem definition is updated to match the differentialEquations.jl interface (except for events which are put inside the function instead of callback functions):

function func(du,u,p,t)
#parameters
#helper expressions
#differential equations
#if-statments for events
end

tspan=(initial_time,final_time)
u = [u1_0,u2_0...]
p = [p1_0,p2_0...]
odeprob=ODEProblem(func,u,tspan,p)

sol= solve(odeprob,algorithm,abstol=...,reltol=...)

[lamBOOO]

✅