Consistency in sensealg arguments

[mschytt]

Hi👋

After a small back-and-forth with Chris on slack, we found that the direct sensitivity analysis interfaces for ODEs are not consistent between the forward and adjoint methods. Specifically, I am considering the choice of sensitivity algorithm.

The documentation for ODEForwardSensitivityProblem gives the following signature with alg as a positional argument:

help?> ODEForwardSensitivityProblem

search: ODEForwardSensitivityProblem ODEForwardSensitivityFunction

  function ODEForwardSensitivityProblem(f::Union{Function,DiffEqBase.AbstractODEFunction}, u0,tspan,p=nothing, alg::AbstractForwardSensitivityAlgorithm = ForwardSensitivity(); kwargs...)

but in the explanation of the syntax, it accidently mentions a positional argument sensealg:

ODEForwardSensitivityProblem(f::SciMLBase.AbstractODEFunction,u0,
                             tspan,p=nothing,
                             sensealg::AbstractForwardSensitivityAlgorithm = ForwardSensitivity();
                             kwargs...)

Since sensealg is a keyword argument for the common solve interface and the direct adjoint sensitivity interface, users (i.e. me) may accidently and unfruitfully try to specify sensealg as a keyword argument in a call to ODEForwardSensitivityProblem.

Perhaps the positional argument alg, should be changed to a keyword argument sensealg in line with the remaining interfaces?

Best regards,
Marcus

[ChrisRackauckas]

Thanks for the report, it's not fixed with depwarns.