Type-unstable ODESolution object

[hersle]

How type-stable should one expect MTK sys-prob-sol code to be? Consider

using OrdinaryDiffEq, ModelingToolkit, Test
using ModelingToolkit: t_nounits as t, D_nounits as D

@variables x(t)
@mtkbuild sys = ODESystem([D(x) ~ 0.0], t; defaults = [x => 1.0])

prob = ODEProblem(sys, [], (0.0, 1.0))
solveit(prob) = solve(prob, Tsit5())
@code_warntype solveit(prob)

MethodInstance for solveit(::ODEProblem{Vector{Float64}, Tuple{Float64, Float64}, true, MTKParameters{StaticArraysCore.SizedVector{0, Float64, Vector{Float64}}, Vector{Float64}, Tuple{}, Tuple{}, Tuple{}, Tuple{}}, ODEFunction{true, SciMLBase.AutoSpecialize, ModelingToolkit.GeneratedFunctionWrapper{(2, 3, true), RuntimeGeneratedFunctions.RuntimeGeneratedFunction{(:__mtk_arg_1, :___mtkparameters___, :t), ModelingToolkit.var"#_RGF_ModTag", ModelingToolkit.var"#_RGF_ModTag", (0x9312a357, 0xdc15add1, 0xcc427068, 0x176f34d9, 0x17e32912), Nothing}, RuntimeGeneratedFunctions.RuntimeGeneratedFunction{(:ˍ₋out, :__mtk_arg_1, :___mtkparameters___, :t), ModelingToolkit.var"#_RGF_ModTag", ModelingToolkit.var"#_RGF_ModTag", (0x318ccbe6, 0x2d1775a1, 0x568e75c0, 0x93fa6d18, 0x61d71a2b), Nothing}}, LinearAlgebra.UniformScaling{Bool}, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, ModelingToolkit.ObservedFunctionCache{ODESystem}, Nothing, ODESystem, SciMLBase.OverrideInitData{NonlinearProblem{Nothing, true, MTKParameters{Vector{Float64}, StaticArraysCore.SizedVector{0, Float64, Vector{Float64}}, Tuple{}, Tuple{}, Tuple{}, Tuple{}}, NonlinearFunction{true, SciMLBase.FullSpecialize, ModelingToolkit.GeneratedFunctionWrapper{(2, 2, true), RuntimeGeneratedFunctions.RuntimeGeneratedFunction{(:__mtk_arg_1, :___mtkparameters___), ModelingToolkit.var"#_RGF_ModTag", ModelingToolkit.var"#_RGF_ModTag", (0x3345e05f, 0xe2ae97f2, 0x977af016, 0x1c855cb9, 0xb982576d), Nothing}, RuntimeGeneratedFunctions.RuntimeGeneratedFunction{(:ˍ₋out, :__mtk_arg_1, :___mtkparameters___), ModelingToolkit.var"#_RGF_ModTag", ModelingToolkit.var"#_RGF_ModTag", (0xb4c20cf7, 0x85eb7c6a, 0x3b0eab35, 0xa921540c, 0xcfdede81), Nothing}}, LinearAlgebra.UniformScaling{Bool}, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, ModelingToolkit.ObservedFunctionCache{NonlinearSystem}, Nothing, NonlinearSystem, Nothing, Nothing}, Base.Pairs{Symbol, Union{}, Tuple{}, @NamedTuple{}}, ModelingToolkit.InitializationSystemMetadata}, ModelingToolkit.UpdateInitializeprob{SymbolicIndexingInterface.MultipleGetters{SymbolicIndexingInterface.ContinuousTimeseries, Vector{SymbolicIndexingInterface.AbstractGetIndexer}}, SymbolicIndexingInterface.MultipleSetters{Vector{SymbolicIndexingInterface.ParameterHookWrapper{SymbolicIndexingInterface.SetParameterIndex{ModelingToolkit.ParameterIndex{SciMLStructures.Tunable, Int64}}, SymbolicUtils.BasicSymbolic{Real}}}}}, ComposedFunction{typeof(identity), SymbolicIndexingInterface.TimeIndependentObservedFunction{ModelingToolkit.GeneratedFunctionWrapper{(2, 2, true), RuntimeGeneratedFunctions.RuntimeGeneratedFunction{(:__mtk_arg_1, :___mtkparameters___), ModelingToolkit.var"#_RGF_ModTag", ModelingToolkit.var"#_RGF_ModTag", (0x8e7ef5ea, 0x7be9a5c2, 0x168fa9bc, 0x9b5fae51, 0xad43a4dd), Nothing}, RuntimeGeneratedFunctions.RuntimeGeneratedFunction{(:ˍ₋out, :__mtk_arg_1, :___mtkparameters___), ModelingToolkit.var"#_RGF_ModTag", ModelingToolkit.var"#_RGF_ModTag", (0x230d20de, 0x06778e34, 0x725268d6, 0xf7595f59, 0x68986d94), Nothing}}}}, Nothing}, Nothing}, Base.Pairs{Symbol, Union{}, Tuple{}, @NamedTuple{}}, SciMLBase.StandardODEProblem})
  from solveit(prob) @ Main c:\Users\herma\.julia\dev\SymBoltz\bug.jl:8
Arguments
  #self#::Core.Const(Main.solveit)
  prob::ODEProblem{Vector{Float64}, Tuple{Float64, Float64}, true, MTKParameters{StaticArraysCore.SizedVector{0, Float64, Vector{Float64}}, Vector{Float64}, Tuple{}, Tuple{}, Tuple{}, Tuple{}}, ODEFunction{true, SciMLBase.AutoSpecialize, ModelingToolkit.GeneratedFunctionWrapper{(2, 3, true), RuntimeGeneratedFunctions.RuntimeGeneratedFunction{(:__mtk_arg_1, :___mtkparameters___, :t), ModelingToolkit.var"#_RGF_ModTag", ModelingToolkit.var"#_RGF_ModTag", (0x9312a357, 0xdc15add1, 0xcc427068, 0x176f34d9, 0x17e32912), Nothing}, RuntimeGeneratedFunctions.RuntimeGeneratedFunction{(:ˍ₋out, :__mtk_arg_1, :___mtkparameters___, :t), ModelingToolkit.var"#_RGF_ModTag", ModelingToolkit.var"#_RGF_ModTag", (0x318ccbe6, 0x2d1775a1, 0x568e75c0, 0x93fa6d18, 0x61d71a2b), Nothing}}, LinearAlgebra.UniformScaling{Bool}, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, ModelingToolkit.ObservedFunctionCache{ODESystem}, Nothing, ODESystem, SciMLBase.OverrideInitData{NonlinearProblem{Nothing, true, MTKParameters{Vector{Float64}, StaticArraysCore.SizedVector{0, Float64, Vector{Float64}}, Tuple{}, Tuple{}, Tuple{}, Tuple{}}, NonlinearFunction{true, SciMLBase.FullSpecialize, ModelingToolkit.GeneratedFunctionWrapper{(2, 2, true), RuntimeGeneratedFunctions.RuntimeGeneratedFunction{(:__mtk_arg_1, :___mtkparameters___), ModelingToolkit.var"#_RGF_ModTag", ModelingToolkit.var"#_RGF_ModTag", (0x3345e05f, 0xe2ae97f2, 0x977af016, 0x1c855cb9, 0xb982576d), Nothing}, RuntimeGeneratedFunctions.RuntimeGeneratedFunction{(:ˍ₋out, :__mtk_arg_1, :___mtkparameters___), ModelingToolkit.var"#_RGF_ModTag", ModelingToolkit.var"#_RGF_ModTag", (0xb4c20cf7, 0x85eb7c6a, 0x3b0eab35, 0xa921540c, 0xcfdede81), Nothing}}, LinearAlgebra.UniformScaling{Bool}, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, ModelingToolkit.ObservedFunctionCache{NonlinearSystem}, Nothing, NonlinearSystem, Nothing, Nothing}, Base.Pairs{Symbol, Union{}, Tuple{}, @NamedTuple{}}, ModelingToolkit.InitializationSystemMetadata}, ModelingToolkit.UpdateInitializeprob{SymbolicIndexingInterface.MultipleGetters{SymbolicIndexingInterface.ContinuousTimeseries, Vector{SymbolicIndexingInterface.AbstractGetIndexer}}, SymbolicIndexingInterface.MultipleSetters{Vector{SymbolicIndexingInterface.ParameterHookWrapper{SymbolicIndexingInterface.SetParameterIndex{ModelingToolkit.ParameterIndex{SciMLStructures.Tunable, Int64}}, SymbolicUtils.BasicSymbolic{Real}}}}}, ComposedFunction{typeof(identity), SymbolicIndexingInterface.TimeIndependentObservedFunction{ModelingToolkit.GeneratedFunctionWrapper{(2, 2, true), RuntimeGeneratedFunctions.RuntimeGeneratedFunction{(:__mtk_arg_1, :___mtkparameters___), ModelingToolkit.var"#_RGF_ModTag", ModelingToolkit.var"#_RGF_ModTag", (0x8e7ef5ea, 0x7be9a5c2, 0x168fa9bc, 0x9b5fae51, 0xad43a4dd), Nothing}, RuntimeGeneratedFunctions.RuntimeGeneratedFunction{(:ˍ₋out, :__mtk_arg_1, :___mtkparameters___), ModelingToolkit.var"#_RGF_ModTag", ModelingToolkit.var"#_RGF_ModTag", (0x230d20de, 0x06778e34, 0x725268d6, 0xf7595f59, 0x68986d94), Nothing}}}}, Nothing}, Nothing}, Base.Pairs{Symbol, Union{}, Tuple{}, @NamedTuple{}}, SciMLBase.StandardODEProblem}
Body::Any
1 ─ %1 = Main.solve::Core.Const(CommonSolve.solve)
│   %2 = Main.Tsit5()::Core.Const(Tsit5{typeof(OrdinaryDiffEqCore.trivial_limiter!), typeof(OrdinaryDiffEqCore.trivial_limiter!), Static.False}(OrdinaryDiffEqCore.trivial_limiter!, OrdinaryDiffEqCore.trivial_limiter!, static(false)))
│   %3 = (%1)(prob, %2)::Any
└──      return %3

All types are inferred except for the last %3, which is Any. Given that the type of prob is inferred, I would expect the type of sol to be, too. It certainly draws my attention when debugging the type-stability of my code, since I've learned that code that only infers to Any is quite bad. How bad is this really with MTK, and can it be fixed?

[ChrisRackauckas]

It should be fixed.

[hersle]

Poking around with @descend into SciMLBase I managed to recover type-stability of this solution with SciML/SciMLBase.jl#989.