Update compat (#557)

* Update Project.toml

* Update collocation and test files

* Update test files

* Update test files again

* IpoptSolver -> Ipopt

* MadNLPSolver -> MadNLP

* AbstractOptimalControlDiscretizer -> AbstractDiscretizer

* AbstractOptimalControlInitialGuess -> AbstractInitialGuess

* AbstractOptimalControlProblem -> AbstractModel

* version 1.0.2-beta

---------

Co-authored-by: Pierre Martinon <pierrecmartinon@gmail.com>

Author: ocots

Project.toml

@@ -1,6 +1,6 @@
 name = "CTDirect"
 uuid = "790bbbee-bee9-49ee-8912-a9de031322d5"
-version = "1.0.1-beta"
+version = "1.0.2-beta"
 authors = ["Pierre Martinon <pierrecmartinon@gmail.com>"]
 
 [workspace]
@@ -22,9 +22,9 @@ SparseArrays = "2f01184e-e22b-5df5-ae63-d93ebab69eaf"
 ADNLPModels = "0.8"
 AMDGPU = "2"
 CTBase = "0.18"
-CTModels = "0.8"
+CTModels = "0.9"
 CTParser = "0.8"
-CTSolvers = "0.2"
+CTSolvers = "0.3"
 CUDA = "5"
 CommonSolve = "0.2"
 DocStringExtensions = "0.9"

src/CTDirect.jl

@@ -12,25 +12,25 @@ using NLPModels
 
 # ----------------------------------------------------------------------
 # TYPES
-const AbstractOptimalControlProblem = CTModels.AbstractModel
+const AbstractModel = CTModels.AbstractModel
 
 # ---------------------------------------------------------------------------
 # Abstract discretizer type
 # ---------------------------------------------------------------------------
-abstract type AbstractOptimalControlDiscretizer <: Strategies.AbstractStrategy end
+abstract type AbstractDiscretizer <: Strategies.AbstractStrategy end
 
 function discretize(
-    ocp::AbstractOptimalControlProblem, 
-    discretizer::AbstractOptimalControlDiscretizer
+    ocp::AbstractModel, 
+    discretizer::AbstractDiscretizer
 )
     return discretizer(ocp)
 end
 
-__discretizer()::AbstractOptimalControlDiscretizer = Collocation()
+__discretizer()::AbstractDiscretizer = Collocation()
 
 function discretize(
-    ocp::AbstractOptimalControlProblem;
-    discretizer::AbstractOptimalControlDiscretizer=__discretizer(),
+    ocp::AbstractModel;
+    discretizer::AbstractDiscretizer=__discretizer(),
 )
     return discretize(ocp, discretizer)
 end

src/collocation.jl

@@ -5,7 +5,7 @@
 # ---------------------------------------------------------------------------
 # Collocation discretizer
 # ---------------------------------------------------------------------------
-struct Collocation <: AbstractOptimalControlDiscretizer
+struct Collocation <: AbstractDiscretizer
     options::Strategies.StrategyOptions
 end
 
@@ -56,7 +56,7 @@ Strategies.options(c::Collocation) = c.options
 # ==========================================================================================
 # Build core DOCP structure with discretization information (ADNLP)
 # ==========================================================================================
-function get_docp(discretizer::Collocation, ocp::AbstractOptimalControlProblem)
+function get_docp(discretizer::Collocation, ocp::AbstractModel)
 
     # recover discretization scheme and options
     scheme = Strategies.options(discretizer)[:scheme]
@@ -77,7 +77,7 @@ end
 # Build initial guess for discretized problem
 # ==========================================================================================
 function get_docp_initial_guess(modeler::Symbol, docp,
-        initial_guess::Union{CTModels.AbstractOptimalControlInitialGuess,Nothing},
+        initial_guess::Union{CTModels.AbstractInitialGuess,Nothing},
         )
 
         ocp = ocp_model(docp)
@@ -117,19 +117,19 @@ function get_docp_initial_guess(modeler::Symbol, docp,
 # ==========================================================================================
 # Build discretizer API (return sets of model/solution builders)
 # ==========================================================================================
-function (discretizer::Collocation)(ocp::AbstractOptimalControlProblem)
+function (discretizer::Collocation)(ocp::AbstractModel)
 
     # common parts for builders
     docp = get_docp(discretizer, ocp)
     exa_getter = nothing # will be set in build_exa_model
 
     # ==========================================================================================
-    # The needed builders for the construction of the final DiscretizedOptimalControlProblem
+    # The needed builders for the construction of the final DiscretizedModel
     # ==========================================================================================
 
     # NLP builder for ADNLPModels
     function build_adnlp_model(
-        initial_guess::CTModels.AbstractOptimalControlInitialGuess;
+        initial_guess::CTModels.AbstractInitialGuess;
         backend,
         kwargs...
     )::ADNLPModels.ADNLPModel
@@ -211,7 +211,7 @@ function (discretizer::Collocation)(ocp::AbstractOptimalControlProblem)
     # NLP builder for ExaModels
     function build_exa_model(
         ::Type{BaseType}, 
-        initial_guess::CTModels.AbstractOptimalControlInitialGuess; 
+        initial_guess::CTModels.AbstractInitialGuess; 
         backend
     )::ExaModels.ExaModel where {BaseType<:AbstractFloat}
 
@@ -260,7 +260,7 @@ function (discretizer::Collocation)(ocp::AbstractOptimalControlProblem)
     end
 
     #NB. it would be better to return builders as model/solution pairs since they are linked
-    return CTSolvers.DiscretizedOptimalControlProblem(
+    return CTSolvers.DiscretizedModel(
         ocp,
         CTSolvers.ADNLPModelBuilder(build_adnlp_model),
         CTSolvers.ExaModelBuilder(build_exa_model),

src/collocation_variables.jl

@@ -21,7 +21,7 @@ julia> DOCP_initial_guess(docp)
 [0.1, 0.1, …]
 ```
 """
-function DOCP_initial_guess(docp::DOCP, init::CTModels.OptimalControlInitialGuess)
+function DOCP_initial_guess(docp::DOCP, init::CTModels.InitialGuess)
 
     # default initialization (including internal variables such as k_i for RK schemes)
     # NB. passing nothing to the setters below will leave this default values 

test/manual_test.jl

@@ -7,6 +7,6 @@ include("./problems/beam.jl")
 include("./problems/goddard.jl")
 include("./problems/double_integrator.jl")
 
-#sol = solve_problem(goddard(); display=true)
+sol = solve_problem(goddard(); display=true)
 sol1 = solve_problem(goddard2(); display=true, modeler=:exa, solver=:madnlp, scheme=:trapeze)
 

test/test_common.jl

@@ -49,9 +49,9 @@ function solve_problem(prob;
 
     # NLP modeler
     if modeler == :adnlp
-        my_modeler = CTSolvers.ADNLPModeler(; backend=adnlp_backend) # kwargs
+        my_modeler = CTSolvers.ADNLP(; backend=adnlp_backend) # kwargs
     elseif modeler == :exa
-        my_modeler = CTSolvers.ExaModeler(; backend=exa_backend) # kwargs
+        my_modeler = CTSolvers.Exa(; backend=exa_backend) # kwargs
     else
         error("Unknown modeler: ", modeler)
     end
@@ -66,7 +66,7 @@ function solve_problem(prob;
             :linear_solver => "mumps",
             :sb => "yes",
         )
-        my_solver = CTSolvers.IpoptSolver(; ipopt_options...)
+        my_solver = CTSolvers.Ipopt(; ipopt_options...)
     elseif solver == :madnlp
         if isnothing(linear_solver)
             solver = MumpsSolver
@@ -82,7 +82,7 @@ function solve_problem(prob;
             madnlp_options[:bound_relax_factor] = bound_relax_factor
             madnlp_options[:mode] = :permissive
         end
-        my_solver = CTSolvers.MadNLPSolver(; madnlp_options...)
+        my_solver = CTSolvers.MadNLP(; madnlp_options...)
     else
         error("Unknown solver: ", solver)
     end

test/tmp/test_collocation.jl

@@ -1,7 +1,7 @@
 # Unit tests for Collocation discretizer wiring from OCP to discretized OCP and builders.
-struct DummyOCPCollocation <: CTModels.AbstractOptimalControlProblem end
+struct DummyOCPCollocation <: CTModels.AbstractModel end
 
-struct DummyOCPExaRouting <: CTModels.AbstractOptimalControlProblem end
+struct DummyOCPExaRouting <: CTModels.AbstractModel end
 
 struct DummyDOCPCollocationRouting end
 
@@ -51,8 +51,8 @@ function test_ctdirect_collocation()
 
         docp = discretizer(ocp)
 
-        # The call operator on Collocation should return a DiscretizedOptimalControlProblem
-        Test.@test docp isa CTModels.DiscretizedOptimalControlProblem
+        # The call operator on Collocation should return a DiscretizedModel
+        Test.@test docp isa CTModels.DiscretizedModel
         Test.@test CTModels.ocp_model(docp) === ocp
 
         # The model and solution builders should be correctly wired with both
@@ -81,7 +81,7 @@ function test_ctdirect_collocation()
         exa_builder = CTModels.get_exa_model_builder(docp)
 
         # Minimal initial guess: functions for state/control and empty variable
-        init_guess = CTModels.OptimalControlInitialGuess(t -> 0.0, t -> 0.0, Float64[])
+        init_guess = CTModels.InitialGuess(t -> 0.0, t -> 0.0, Float64[])
 
         BaseType = Float32
         exa_nlp = exa_builder(BaseType, init_guess; backend=:gpu, foo=1)

test/tmp/test_core_types.jl

@@ -16,11 +16,11 @@ function test_ctdirect_core_types()
         # Trapeze is an alias to Trapezoidal
         Test.@test CTDirect.TrapezeScheme === CTDirect.TrapezoidalScheme
 
-        # AbstractOptimalControlDiscretizer should be abstract
-        Test.@test isabstracttype(CTDirect.AbstractOptimalControlDiscretizer)
+        # AbstractDiscretizer should be abstract
+        Test.@test isabstracttype(CTDirect.AbstractDiscretizer)
 
         # Collocation should be a concrete discretizer subtype
-        Test.@test CTDirect.Collocation <: CTDirect.AbstractOptimalControlDiscretizer
+        Test.@test CTDirect.Collocation <: CTDirect.AbstractDiscretizer
     end
 
     # ========================================================================

test/tmp/test_discretization_api.jl

@@ -1,12 +1,12 @@
 # Unit tests for the discretization API (discretize with custom and default discretizers).
-struct DummyOCPDiscretize <: CTModels.AbstractOptimalControlProblem end
+struct DummyOCPDiscretize <: CTModels.AbstractModel end
 
-struct DummyDiscretizer <: CTDirect.AbstractOptimalControlDiscretizer
+struct DummyDiscretizer <: CTDirect.AbstractDiscretizer
     calls::Base.RefValue{Int}
     tag::Symbol
 end
 
-function (d::DummyDiscretizer)(ocp::CTModels.AbstractOptimalControlProblem)
+function (d::DummyDiscretizer)(ocp::CTModels.AbstractModel)
     d.calls[] += 1
     return (ocp, d.tag)
 end
@@ -37,13 +37,13 @@ function test_ctdirect_discretization_api()
 
         docp = CTDirect.discretize(ocp)
 
-        # The default discretizer should produce a DiscretizedOptimalControlProblem
-        Test.@test docp isa CTModels.DiscretizedOptimalControlProblem
+        # The default discretizer should produce a DiscretizedModel
+        Test.@test docp isa CTModels.DiscretizedModel
         Test.@test CTModels.ocp_model(docp) === ocp
 
         # And the low-level __discretizer() helper should return a Collocation
         disc = CTDirect.__discretizer()
-        Test.@test disc isa CTDirect.AbstractOptimalControlDiscretizer
+        Test.@test disc isa CTDirect.AbstractDiscretizer
         Test.@test disc isa CTDirect.Collocation
     end
 end

test/tmp/test_solve.jl

@@ -27,15 +27,15 @@ function test_solve()
         init = CTModels.initial_guess(ocp; beam_data.init...)
         discretizer = CTDirect.Collocation()
         docp = CTDirect.discretize(ocp, discretizer)
-        Test.@test docp isa CTModels.DiscretizedOptimalControlProblem
+        Test.@test docp isa CTModels.DiscretizedModel
 
         # NLP solver
-        solvers = [CTSolvers.IpoptSolver(; ipopt_options...)]
+        solvers = [CTSolvers.Ipopt(; ipopt_options...)]
         solvers_names = ["Ipopt"]
 
         # NLP modelers
-        modelers = [CTModels.ADNLPModeler(), CTModels.ExaModeler()]
-        modelers_names = ["ADNLPModeler", "ExaModeler (CPU)"]
+        modelers = [CTModels.ADNLP(), CTModels.Exa()]
+        modelers_names = ["ADNLP", "Exa (CPU)"]
 
         # solve DOCP with common solve and NLP modelers
         Test.@testset "DOCP level (solve)" verbose=VERBOSE showtiming=SHOWTIMING begin