Error with vector arguments

[TorkelE]

There are currently some problems with vector arguments for parameters/initial conditions. I think this worked previously (I had some old tests which I think were mostly fine, but waited due to another issue and then forgot about).

Example using both Catalyst and MTK.

using Catalyst, OrdinaryDiffEq

@variables t
D = Differential(t)
@parameters k[1:4]
@species (X(t))[1:2]
@variables (V(t))[1:2]

rxs = [
    Reaction(k[1], [], [X[1]]),
    Reaction(k[2], [], [X[2]]),
    Reaction(k[3], [X[1]], []),
    Reaction(k[4], [X[2]], [])
]
eqs = [
    D(V[1]) ~ k[1] - k[2]*V[1],
    D(V[2]) ~ k[3] - k[4]*V[2]
]

@named rs = ReactionSystem(rxs, t)
rs = complete(rs)
@mtkbuild osys = ODESystem(eqs, t)

u0_rs = [X => [10.0, 20.0]]
u0_alt_rs = [X[1] => 10.0, X[2] => 20.0]
u0_osys = [V => [10.0, 20.0]]
u0_alt_osys = [V[1] => 10.0, V[2] => 20.0]
ps = [k => [1.0, 2.0, 3.0, 4.0]]
ps_alt = [k[1] => 1.0, k[2] => 2.0, k[3] => 3.0, k[4] => 4.0]

# Not sure why scalarised parameter values are no longer permitted, but seems funny they work for Catalyst.
oprob_rs = ODEProblem(rs, u0_rs, 1.0, ps) # Works
oprob_rs = ODEProblem(rs, u0_alt_rs, 1.0, ps) # Works
oprob_rs = ODEProblem(rs, u0_rs, 1.0, ps_alt) # Works/ Should probably yield an error.
oprob_osys = ODEProblem(osys, u0_osys, 1.0, ps) # Works
oprob_osys = ODEProblem(osys, u0_alt_osys, 1.0, ps) # Works
oprob_osys = ODEProblem(osys, u0_osys, 1.0, ps_alt) # ERROR: Scalarized parameter values (k[3]) are not supported. Instead of `[p[1] => 1.0, p[2] => 2.0]` use `[p => [1.0, 2.0]]`

# Problem with Catalyst.
remake(oprob_rs, u0 = u0_rs, p = ps) # ERROR: type Nothing has no field portion
remake(oprob_rs, u0 = u0_alt_rs, p = ps) # ERROR: type Nothing has no field portion
remake(oprob_rs, u0 = u0_rs, p = ps_alt) # Works
remake(oprob_osys, u0 = u0_osys, p = ps) # Works
remake(oprob_osys, u0 = u0_alt_osys, p = ps) # Works
remake(oprob_osys, u0 = u0_osys, p = ps_alt) # `ERROR: UndefRefError: access to undefined reference` (maybe a better error message.)

An expanded version of Catalyst remake error. Never actually passes Catalyst code, not sure why.

ERROR: type Nothing has no field portion
Stacktrace:
  [1] getproperty(x::Nothing, f::Symbol)
    @ Base ./Base.jl:37
  [2] validate_parameter_type(ic::ModelingToolkit.IndexCache, p::Symbolics.Arr{Num, 1}, index::Nothing, val::Vector{Float64})
    @ ModelingToolkit ~/.julia/packages/ModelingToolkit/3Suzf/src/systems/parameter_buffer.jl:424
  [3] remake_buffer(indp::ODEProblem{…}, oldbuf::ModelingToolkit.MTKParameters{…}, vals::Dict{…})
    @ ModelingToolkit ~/.julia/packages/ModelingToolkit/3Suzf/src/systems/parameter_buffer.jl:487
  [4] _updated_u0_p_symmap(prob::ODEProblem{…}, u0::Dict{…}, ::Val{…}, p::Dict{…}, ::Val{…})
    @ SciMLBase ~/.julia/packages/SciMLBase/roUH9/src/remake.jl:511
  [5] _updated_u0_p_internal(prob::ODEProblem{…}, u0::Vector{…}, p::Vector{…}; interpret_symbolicmap::Bool, use_defaults::Bool)
    @ SciMLBase ~/.julia/packages/SciMLBase/roUH9/src/remake.jl:427
  [6] _updated_u0_p_internal
    @ ~/.julia/packages/SciMLBase/roUH9/src/remake.jl:412 [inlined]
  [7] #updated_u0_p#688
    @ ~/.julia/packages/SciMLBase/roUH9/src/remake.jl:551 [inlined]
  [8] updated_u0_p
    @ ~/.julia/packages/SciMLBase/roUH9/src/remake.jl:532 [inlined]
  [9] remake(prob::ODEProblem{…}; f::Missing, u0::Vector{…}, tspan::Missing, p::Vector{…}, kwargs::Missing, interpret_symbolicmap::Bool, use_defaults::Bool, _kwargs::@Kwargs{})
    @ SciMLBase ~/.julia/packages/SciMLBase/roUH9/src/remake.jl:95
 [10] top-level scope
    @ ~/Desktop/Julia Playground/Environment - Catalyst test/test_playground.jl:38
Some type information was truncated. Use `show(err)` to see complete types.

[isaacsas]

So then from that error it sounds like we shouldn't be scalarizing parameters any more?

[isaacsas]

Changing that would be breaking so if we should switch that should be done before V14.

[TorkelE]

Internally I think we can do whichever, but we might want to change what we permit in the input.

[TorkelE]

I will have to go back to see if I can find all the details about scalrization that I wrote and discussed with Aayush though

[TorkelE]

replaced by SciML/ModelingToolkit.jl#2804