Update to use MTK v9 (#176)

.github/workflows/ci-threaded.yaml

@@ -16,7 +16,7 @@
           - uses: actions/checkout@v2
           - uses: julia-actions/setup-julia@v1
             with:
-              version: '1.9'
+              version: '1.10'
           - uses: julia-actions/cache@v1
             with:
               cache-compiled: "true"

Project.toml

@@ -42,7 +42,9 @@ UUIDs = "cf7118a7-6976-5b1a-9a39-7adc72f591a4"
 YAML = "ddb6d928-2868-570f-bddf-ab3f9cf99eb6"
 
 [compat]
-julia = "1.10.1"
+ModelingToolkit = "9.13.0"
+julia = "1.10"
+EasyModelAnalysis = "1"
 
 [extras]
 Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"

src/SimulationService.jl

@@ -19,7 +19,7 @@ import JSON3
 import JSONSchema
 import LinearAlgebra: norm
 import MathML
-import ModelingToolkit: @parameters, substitute, Differential, Num, @variables, ODESystem, ODEProblem, ODESolution, structural_simplify, states, observed, parameters
+import ModelingToolkit: @parameters, substitute, Differential, Num, @variables, ODESystem, ODEProblem, ODESolution, structural_simplify, unknowns, observed, parameters
 import OpenAPI
 import Oxygen
 import Pkg

src/model_parsers/RegNets.jl

@@ -92,5 +92,5 @@ function ASKEM_ACSet_to_MTK(sg::ASKEMRegNetUntyped)
 
     eqs = [D(vertex_funcs[i]) ~ v_rate_vars[i]*vertex_funcs[i] + sum((sg[e,:sign] ? 1 : -1) * e_rate_vars[e] * vertex_funcs[i] * vertex_funcs[sg[e,:src]] for e in incident(sg,i,:tgt); init = 0.0) for i in 1:nv(sg)]
 
-    sys = ODESystem(eqs, t, vertex_funcs, all_params; name = :system, defaults)
+    sys = structural_simplify(ODESystem(eqs, t, vertex_funcs, all_params; name = :system, defaults))
 end

src/operations.jl

@@ -163,7 +163,7 @@ end
 # data
 function amr_get(df::DataFrame, sys::ODESystem, ::Val{:data})
     @info "parse dataset into calibrate format"
-    statelist = states(sys)
+    statelist = unknowns(sys)
     statenames = string.(statelist)
     statenames = [replace(nm, "(t)" => "") for nm in statenames]
 
@@ -188,7 +188,7 @@ function (o::IntermediateResults)(integrator)
     (; iter, f, t, u, p) = integrator
     if o.last_callback + o.every == iter
         o.last_callback = iter
-        state_dict = Dict(states(f.sys) .=> u)
+        state_dict = Dict(unknowns(f.sys) .=> u)
         param_dict = Dict(parameters(f.sys) .=> p)
         publish_to_rabbitmq(; iter=iter, state=state_dict, params = param_dict, id=o.id,
             retcode=SciMLBase.check_error(integrator))
@@ -197,20 +197,20 @@ function (o::IntermediateResults)(integrator)
 end
 
 # Intermediate results functor for calibrate
-function (o::IntermediateResults)(p,lossval, ode_sol, ts)
+function (o::IntermediateResults)(state,loss_val, ode_sol, ts)
     if o.last_callback + o.every == o.iter
         o.last_callback = o.iter
-        param_dict = Dict(parameters(ode_sol.prob.f.sys) .=> ode_sol.prob.p)
-        state_dict = Dict([state => ode_sol(first(ts))[state] for state in states(ode_sol.prob.f.sys)])
-        publish_to_rabbitmq(; iter = o.iter, loss = lossval, sol_data = state_dict, timesteps = first(ts), params = param_dict, id=o.id)
+        param_dict = Dict(parameters(ode_sol.prob.f.sys) .=> state.u)
+        state_dict = Dict([state => ode_sol(first(ts))[state] for state in unknowns(ode_sol.prob.f.sys)])
+        publish_to_rabbitmq(; iter = o.iter, loss = loss_val, sol_data = state_dict, timesteps = first(ts), params = param_dict, id=o.id)
     end
     o.iter = o.iter + 1
     return false
 end
 #----------------------------------------------------------------------# dataframe_with_observables
 function dataframe_with_observables(sol::ODESolution)
     sys = sol.prob.f.sys
-    names = [states(sys); getproperty.(observed(sys), :lhs)]
+    names = [unknowns(sys); getproperty.(observed(sys), :lhs)]
     cols = ["timestamp" => sol.t; [string(n) => sol[n] for n in names]]
     DataFrame(cols)
 end
@@ -280,7 +280,7 @@ end
 
 function solve(o::Calibrate; callback)
     prob = ODEProblem(o.sys, [], o.timespan)
-    statenames = [states(o.sys);getproperty.(observed(o.sys), :lhs)]
+    statenames = [unknowns(o.sys);getproperty.(observed(o.sys), :lhs)]
 
     # bayesian datafit 
     if o.calibrate_method == "bayesian"

test/runtests.jl

@@ -105,7 +105,7 @@ end
 @testset "Petrinet AMR parsing" begin
     amr = JSON3.read(HTTP.get("https://raw.githubusercontent.com/DARPA-ASKEM/simulation-integration/main/data/models/sidarthe.json").body)
     sys = SimulationService.amr_get(amr.configuration, ODESystem)
-    @test string.(states(sys)) == ["Susceptible(t)", "Diagnosed(t)", "Infected(t)", "Ailing(t)", "Recognized(t)", "Healed(t)", "Threatened(t)", "Extinct(t)"]
+    @test string.(unknowns(sys)) == ["Susceptible(t)", "Diagnosed(t)", "Infected(t)", "Ailing(t)", "Recognized(t)", "Healed(t)", "Threatened(t)", "Extinct(t)"]
     @test string.(parameters(sys)) == ["beta", "gamma", "delta", "alpha", "epsilon", "zeta", "lambda", "eta", "rho", "theta", "kappa", "mu", "nu", "xi", "tau", "sigma"]
     @test map(x->string(x.lhs), observed(sys)) == ["Cases(t)", "Hospitalizations(t)", "Deaths(t)"]
 
@@ -117,14 +117,14 @@ end
     df = CSV.read(HTTP.get("https://raw.githubusercontent.com/DARPA-ASKEM/simulation-integration/main/data/datasets/SIDARTHE_dataset.csv").body,DataFrame)
     data = SimulationService.amr_get(df, sys, Val(:data))
     @test data isa Vector{Pair{SymbolicUtils.BasicSymbolic{Real}, Tuple{Vector{Int64}, Vector{Float64}}}}
-    @test string.(first.(data)) == string.(states(sys))
+    @test string.(first.(data)) == string.(unknowns(sys))
     @test all(all.(map(first.(last.(data))) do x; x .== 0:89; end))
 end
 
 @testset "Stock and Flow AMR parsing" begin
     amr = JSON3.read(HTTP.get("https://raw.githubusercontent.com/DARPA-ASKEM/simulation-integration/main/data/models/SIR_stockflow.json").body)
     sys = SimulationService.amr_get(amr, ODESystem)
-    @test string.(states(sys)) == ["S(t)", "I(t)", "R(t)"]
+    @test string.(unknowns(sys)) == ["S(t)", "I(t)", "R(t)"]
     @test string.(parameters(sys)) == ["p_cbeta","p_N", "p_tr"]
 
     priors = SimulationService.amr_get(amr, sys, Val(:priors))
@@ -138,7 +138,7 @@ end
 @testset "RegNet AMR parsing" begin
     amr = JSON3.read(HTTP.get("https://raw.githubusercontent.com/DARPA-ASKEM/simulation-integration/main/data/models/LV_sheep_foxes_regnet.json").body)
     sys = SimulationService.amr_get(amr, ODESystem)
-    @test string.(states(sys)) == ["S(t)", "F(t)"]
+    @test string.(unknowns(sys)) == ["S(t)", "F(t)"]
     @test string.(parameters(sys)) == ["beta","delta", "alpha","gamma"]
 
     priors = SimulationService.amr_get(amr, sys, Val(:priors))
@@ -247,19 +247,19 @@ end
 
         dfsim, dfparam = solve(o, callback = SimulationService.get_callback(op,SimulationService.Calibrate))
 
-        statenames = [states(o.sys); getproperty.(observed(o.sys), :lhs)]
+        statenames = [unknowns(o.sys); getproperty.(observed(o.sys), :lhs)]
         @test names(dfsim) == vcat("timestamp",reduce(vcat,[string.("ensemble",i,"_", statenames) for i in 1:size(dfsim,2)÷length(statenames)]))
         @test names(dfparam) == string.(parameters(sys))
 
-        #calibrate_method = "local"
-        #o = SimulationService.Calibrate(sys, (0.0, 89.0), priors, data, num_chains, num_iterations, calibrate_method, ode_method)
-        #dfsim, dfparam = SimulationService.solve(o; callback = nothing)
+        calibrate_method = "local"
+        o = SimulationService.Calibrate(sys, (0.0, 89.0), priors, data, num_chains, num_iterations, calibrate_method, ode_method)
+        dfsim, dfparam = SimulationService.solve(o; callback = SimulationService.get_callback(op,SimulationService.Calibrate))
 
         calibrate_method = "global"
         o = SimulationService.Calibrate(sys, (0.0, 89.0), priors, data, num_chains, num_iterations, calibrate_method, ode_method)
         dfsim, dfparam = SimulationService.solve(o; callback = SimulationService.get_callback(op,SimulationService.Calibrate))
 
-        statenames = [states(o.sys);getproperty.(observed(o.sys), :lhs)]
+        statenames = [unknowns(o.sys);getproperty.(observed(o.sys), :lhs)]
         @test names(dfsim) == vcat("timestamp",string.(statenames))
         @test names(dfparam) == string.(parameters(sys))
     end
@@ -415,7 +415,7 @@ end
         op.id = "1"
         dfsim, dfparam = SimulationService.solve(o, callback = SimulationService.get_callback(op,SimulationService.Calibrate))
 
-        statenames = [states(o.sys);getproperty.(observed(o.sys), :lhs)]
+        statenames = [unknowns(o.sys);getproperty.(observed(o.sys), :lhs)]
         @test names(dfsim) == vcat("timestamp",string.(statenames))
     end
 end