Downstream Updates from MTK v9

[ChrisRackauckas]

• CompatHelper: bump compat for ModelingToolkit to 9 for package docs, (keep existing compat) NeuralPDE.jl#817
• CompatHelper: bump compat for ModelingToolkit to 9 for package docs, (keep existing compat) EasyModelAnalysis.jl#240
• CompatHelper: bump compat for ModelingToolkit to 9, (keep existing compat) NeuralPDE.jl#816
• CompatHelper: bump compat for ModelingToolkit to 9, (keep existing compat) EasyModelAnalysis.jl#239
• CompatHelper: bump compat for ModelingToolkit to 9 for package docs, (keep existing compat) DiffEqDocs.jl#725
• CompatHelper: bump compat for ModelingToolkit to 9 for package docs, (keep existing compat) MethodOfLines.jl#373
• CompatHelper: bump compat for ModelingToolkit to 9 for package docs, (keep existing compat) ModelingToolkitStandardLibrary.jl#270
• CompatHelper: bump compat for ModelingToolkit to 9, (keep existing compat) ParameterizedFunctions.jl#119
• CompatHelper: bump compat for ModelingToolkit to 9, (keep existing compat) MethodOfLines.jl#372
• CompatHelper: bump compat for ModelingToolkit to 9, (keep existing compat) ModelingToolkitStandardLibrary.jl#269
• CompatHelper: bump compat for ModelingToolkit to 9 for package docs, (keep existing compat) ModelingToolkitCourse#36
• CompatHelper: bump compat for ModelingToolkit to 9 for package docs, (keep existing compat) NonlinearSolve.jl#383
• CompatHelper: bump compat for ModelingToolkit to 9, (keep existing compat) ModelingToolkitCourse#35
• CompatHelper: bump compat for ModelingToolkit to 9, (keep existing compat) MinimallyDisruptiveCurves.jl#29
• CompatHelper: bump compat for ModelingToolkit to 9 for package downstream, (keep existing compat) RecursiveArrayTools.jl#360
• CompatHelper: bump compat for ModelingToolkit to 9 for package docs, (keep existing compat) DataDrivenDiffEq.jl#497
• CompatHelper: bump compat for ModelingToolkit to 9, (keep existing compat) DataDrivenDiffEq.jl#496
• CompatHelper: bump compat for ModelingToolkit to 9, (keep existing compat) JuliaDynamics/ProcessBasedModelling.jl#3
• https://github.com/EarthSciML/Aerosol.jl
• https://github.com/ai4energy/Ai4EComponentLib.jl
• https://github.com/cadojo/AstrodynamicalSolvers.jl
• https://github.com/hexaeder/BlockSystems.jl
• https://github.com/jedforrest/BondGraphs.jl
• CompatHelper: bump compat for ModelingToolkit to 9 for package OptimizationMOI, (keep existing compat) Optimization.jl#707
• CompatHelper: bump compat for ModelingToolkit to 9 for package docs, (keep existing compat) SciMLBase.jl#631
• CompatHelper: bump compat for ModelingToolkit to 9 for package docs, (keep existing compat) SciMLDocs#219
• CompatHelper: bump compat for ModelingToolkit to 9, (keep existing compat) DiffEqBayes.jl#325
• JumpProcesses?
• Catalyst?
• CellMLToolkit?
• CompatHelper: bump compat for ModelingToolkit to 9 for package docs, (keep existing compat) StructuralIdentifiability.jl#298
• https://github.com/TS-CUBED/CirculatorySystemModels.jl
• https://github.com/AayushSabharwal/Collide.jl
• https://github.com/wsphillips/Conductor.jl
• https://github.com/JuliaControl/ControlSystemsMTK.jl
• DiffEqProblemLibrary.jl
• https://github.com/mathopt/DynamicOED.jl
• https://github.com/x3042/ExactODEReduction.jl
• https://github.com/EarthSciML/EarthSciData.jl
• https://github.com/EarthSciML/EarthSciMLBase.jl
• https://github.com/cadojo/GalacticPotentials.jl
• https://github.com/EarthSciML/GasChem.jl
• https://github.com/cadojo/GeneralAstrodynamics.jl
• https://github.com/JuliaGNI/GeometricProblems.jl
• https://github.com/RafaelArutjunjan/InformationGeometry.jl
• https://github.com/xiaomingfu2013/LinearNoiseApproximation.jl
• https://github.com/bgctw/MTKHelpers.jl
• CompatHelper: bump compat for ModelingToolkit to 9 for package docs, (keep existing compat) ModelOrderReduction.jl#76
• CompatHelper: bump compat for ModelingToolkit to 9 for package test, (keep existing compat) ModelOrderReduction.jl#77
• CompatHelper: bump compat for ModelingToolkit to 9, (keep existing compat) ModelOrderReduction.jl#75
• https://github.com/augustinas1/MomentClosure.jl
• https://github.com/voduchuy/NumCME.jl
• https://github.com/ODINN-SciML/ODINN.jl
• https://github.com/SciML/PDEBase.jl
• https://github.com/SciML/PDESystemLibrary.jl
• https://github.com/sebapersson/PEtab.jl
• CompatHelper: bump compat for ModelingToolkit to 9, (keep existing compat) iliailmer/ParameterEstimation.jl#95
• https://github.com/mattar13/PhysiologyAnalysis.jl
• https://github.com/killah-t-cell/Plasma.jl
• https://github.com/cadojo/PolynomialGTM.jl
• https://github.com/nathanaelbosch/ProbNumDiffEq.jl
• https://github.com/qojulia/QuantumCumulants.jl
• ReactionNetworkImporters?
• https://github.com/sebapersson/SBMLImporter.jl
• https://github.com/SciML/SBMLToolkitTestSuite.jl
• https://github.com/alexeyovchinnikov/SIAN-Julia
• https://github.com/YichengDWu/Sophon.jl
• https://github.com/PSORLab/SourceCodeMcCormick.jl
• SturcturalIdentifiability?
• https://github.com/worlddynamics/Vensim2MTK.jl
• https://github.com/worlddynamics/WorldDynamics.jl
• Bump docs to MTK 9 JuliaDynamics/DynamicalSystems.jl#229

Upgrade bug issues:

(None so far!)

Upgrade explanation issues:

• What does "parameter splitting" mean? #2482
• MTK v9 with split=true breaks obtaining parameters with SymbolicIndexingInterface. #2485

[ChrisRackauckas]

Also any downstream community packages I may have missed, please post here. I tried to weed out unmaintained projects and things that just look like archives of project code.

@ctessum @cadojo @jake484 @hexaeder @jedforrest @TorkelE @isaacsas @TS-CUBED @wsphillips @baggepinnen @Vaibhavdixit02 @AlCap23 @sumiya11 @michakraus @RafaelArutjunjan @xiaomingfu2013 @bgctw @augustinas1 @voduchuy @JordiBolibar @facusapienza21 @sebapersson @iliailmer @orebas @mattar13 @killah-t-cell @nathanaelbosch @david-pl @paulflang @pogudingleb @YichengDWu @RXGottlieb @meyde @paulobruno @piluc @xtalax please see https://github.com/SciML/ModelingToolkit.jl/blob/master/NEWS.md and let me know if you have any questions. Some general updating notes:

1. Any PDE package should be able to update without a hiccup. PDESystem was unchanged.
2. Many packages which just use MTK for ODEs will update straightforwardly. We would recommend using the canonical t and D, i.e. using ModelingToolkit: t_nounits as t, D_nounits as D instead of defining it in file, but your tests shouldn't be broken from this. Most test breakage is likely just because ODESystem(eqs) was deprecated for ODESystem(eqs,t) and that the system needs to be completed before generating the problem (i.e. use @mtkbuild or change structural_simplify(sys) to complete(structural_simplify(sys)), though @mtkbuild is now the documented way). It's not possible to automatically upgrade these so it had to be made breaking (t detection did not always work and there were subtle edge cases) and so this will need a manual intervention. I presume this will be 99% of the upgrade work for most packages but should be straightforward and mechanical.
3. states(sys) was changed to unknowns(sys), that should be mechanical.
4. Symbolic arrays have had many changes so that they can now keep structure. If you were previously using u[1:5] and you meant an array of symbolic variables, use u = Symbolics.variables(:u, 1:5) instead. To clarify the difference between them, see https://symbolics.juliasymbolics.org/stable/manual/arrays/. I can see that the former was used in a lot of places that meant the latter. If you're not being setup for JuliaSimCompiler there won't be a performance difference (and in fact there will likely be a performance improvement due to improved tracing performance)
5. If your package assumes parameters from the generated ODEProblem are a vector (which IIUC is a very small subset, but includes PETab.jl), use the SciMLStructures interface https://github.com/SciML/SciMLStructures.jl to interact with the parameters structure. For these cases you may need to come talk to me since there are not that many examples of handling this right now. Also, some of these may need Use the SciMLStructures Tunable interface SciMLSensitivity.jl#1010

Again if any of this is confusing and you need help upgrading, please feel free to ask questions, ask for help. We want help to get the whole ecosystem updated.

(@AlCap23 you were going to move DynamicOED.jl into SciML? This would be a good time to do it)

[AayushSabharwal]

change structural_simplify(sys) to complete(structural_simplify(sys)),

structural_simplify calls complete, so this is not necessary

JumpProcesses?
Catalyst?

yes and yes

SturcturalIdentifiability?

yes

[ChrisRackauckas]

I was just making a note that I haven't gotten those CompatHelper PRs yet. We're definitely covering any non-archived SciML package!

[ChrisRackauckas]

A more complete description of the parameter splitting if anyone needs help figuring out what it's all about: #2482 (comment)

[sumiya11]

Thanks for this summary @ChrisRackauckas !
We will try to update MTK in StructuralIdentifiability.jl

[cadojo]

Thank you for the detailed information! I am trying to update AstrodynamicalModels.jl, and I'm running into an issue with ODEFunction. I think I have traced the issue to different parameter ordering inside functions written by Symbolics.generate_function for ODESystem instances created by MTK Version 9. **Is the usage in PlanarEntrySystem() still supported? I had thought that ODESystem(eqs, t, [γ, v, r, θ], [R, P, H, m, A, C, μ]; name=name) was supported as part of the non-DSL syntax.

I can see the word-level differences in the generated function string in git diff --no-index, but those differences are shown by colors and I don't know how to show that in GitHub markdown or HTML. I've included an image of the terminal output below. Alternatively, you can run the following command.

git clone git@github.com:cadojo/AstrodynamicalModels.jl.git
cd AstrodynamicalModels.jl
git diff --no-index --word-diff=color v8/entry.jl v9/entry.jl

[pogudingleb]

@ChrisRackauckas
What happened to ModelingToolkit.DiscreteSystem ?
Now the subtypes of ModelingToolkit.AbstractTimeDependentSystem seem to be only JumpSystem and ModelingToolkit.AbstractODESystem.

[pogudingleb]

@ChrisRackauckas
What happened to ModelingToolkit.DiscreteSystem ?
Now the subtypes of ModelingToolkit.AbstractTimeDependentSystem seem to be only JumpSystem and ModelingToolkit.AbstractODESystem.

My bad, found in the release notes

[ChrisRackauckas]

It will be returning, we just need to redo it a bit on the new discrete-continuous infrastructure

[ChrisRackauckas]

Thank you for the detailed information! I am trying to update AstrodynamicalModels.jl, and I'm running into an issue with ODEFunction. I think I have traced the issue to different parameter ordering inside functions written by Symbolics.generate_function for ODESystem instances created by MTK Version 9. **Is the usage in PlanarEntrySystem() still supported? I had thought that ODESystem(eqs, t, [γ, v, r, θ], [R, P, H, m, A, C, μ]; name=name) was supported as part of the non-DSL syntax.

Ordering isn't guaranteed. In fact, parameters aren't even a vector so that doesn't make a lot of sense. See the discussion on parameter splitting for more details: #2482 (comment).

But anything you were looking to do that needed a parameter ordering can and should be done by using the symbolic indexing interface: https://docs.sciml.ai/SymbolicIndexingInterface/stable/. For example, f = getp(prob, [R, P, H, m, A, C, μ]) gives you a function f(prob) that returns the parameters in that chosen ordering. This decouples the ordering and representation in the solver from how you use and interact with it. There's also setp as well.

[pogudingleb]

It will be returning, we just need to redo it a bit on the new discrete-continuous infrastructure

Does it mean that right now MTK does not have an interface for discrete-time systems?