Add SubproblemCount attribute

README.md

@@ -85,6 +85,10 @@ the solution process.
  * `MOA.SolutionLimit()`
  * `MOI.TimeLimitSec()`
 
+Query the number of scalar subproblems that were solved using
+
+ * `MOA.SubproblemCount()`
+
 ## Ideal point
 
 By default, MOA will compute the ideal point, which can be queried using the

ext/MultiObjectiveAlgorithmsPolyhedraExt.jl

@@ -53,7 +53,7 @@ function MOA.minimize_multiobjective!(
     yI, yUB = zeros(n), zeros(n)
     for (i, f_i) in enumerate(scalars)
         MOI.set(model.inner, MOI.ObjectiveFunction{typeof(f_i)}(), f_i)
-        MOI.optimize!(model.inner)
+        MOA.optimize_inner!(model)
         status = MOI.get(model.inner, MOI.TerminationStatus())
         if !MOA._is_scalar_status_optimal(model)
             return status, nothing
@@ -63,7 +63,7 @@ function MOA.minimize_multiobjective!(
         yI[i] = Y[i]
         anchors[Y] = X
         MOI.set(model.inner, MOI.ObjectiveSense(), MOI.MAX_SENSE)
-        MOI.optimize!(model.inner)
+        MOA.optimize_inner!(model)
         status = MOI.get(model.inner, MOI.TerminationStatus())
         if !MOA._is_scalar_status_optimal(model)
             MOA._warn_on_nonfinite_anti_ideal(algorithm, MOI.MIN_SENSE, i)
@@ -113,7 +113,7 @@ function MOA.minimize_multiobjective!(
         # would not terminate when precision is set to 0
         new_f = sum(w[i] * (scalars[i] + u[i]) for i in 1:n) # w' * (f(x) + u)
         MOI.set(model.inner, MOI.ObjectiveFunction{typeof(new_f)}(), new_f)
-        MOI.optimize!(model.inner)
+        MOA.optimize_inner!(model)
         status = MOI.get(model.inner, MOI.TerminationStatus())
         if !MOA._is_scalar_status_optimal(model)
             return status, nothing

src/MultiObjectiveAlgorithms.jl

@@ -123,6 +123,7 @@ mutable struct Optimizer <: MOI.AbstractOptimizer
     solve_time::Float64
     ideal_point::Vector{Float64}
     compute_ideal_point::Bool
+    subproblem_count::Int
 
     function Optimizer(optimizer_factory)
         return new(
@@ -135,6 +136,7 @@ mutable struct Optimizer <: MOI.AbstractOptimizer
             NaN,
             Float64[],
             default(ComputeIdealPoint()),
+            0,
         )
     end
 end
@@ -146,6 +148,7 @@ function MOI.empty!(model::Optimizer)
     model.termination_status = MOI.OPTIMIZE_NOT_CALLED
     model.solve_time = NaN
     empty!(model.ideal_point)
+    model.subproblem_count = 0
     return
 end
 
@@ -394,6 +397,20 @@ end
 
 MOI.get(model::Optimizer, ::ComputeIdealPoint) = model.compute_ideal_point
 
+### SubproblemCount
+
+"""
+    SubproblemCount <: AbstractModelAttribute -> Int
+
+A result attribute for querying the total number of subproblem solves by an
+algorithm.
+"""
+struct SubproblemCount <: MOI.AbstractModelAttribute end
+
+MOI.is_set_by_optimize(::SubproblemCount) = true
+
+MOI.get(model::Optimizer, ::SubproblemCount) = model.subproblem_count
+
 ### RawOptimizerAttribute
 
 function MOI.supports(model::Optimizer, attr::MOI.RawOptimizerAttribute)
@@ -573,6 +590,18 @@ function MOI.delete(model::Optimizer, ci::MOI.ConstraintIndex)
     return
 end
 
+"""
+    optimize_inner!(model::Optimizer)
+
+A function that must be called instead of `MOI.optimize!(model.inner)` because
+it also increments the `subproblem_count`.
+"""
+function optimize_inner!(model::Optimizer)
+    MOI.optimize!(model.inner)
+    model.subproblem_count += 1
+    return
+end
+
 function _compute_ideal_point(model::Optimizer, start_time)
     for (i, f) in enumerate(MOI.Utilities.eachscalar(model.f))
         if _time_limit_exceeded(model, start_time)
@@ -582,7 +611,7 @@ function _compute_ideal_point(model::Optimizer, start_time)
             continue  # The algorithm already updated this information
         end
         MOI.set(model.inner, MOI.ObjectiveFunction{typeof(f)}(), f)
-        MOI.optimize!(model.inner)
+        optimize_inner!(model)
         status = MOI.get(model.inner, MOI.TerminationStatus())
         if _is_scalar_status_optimal(status)
             model.ideal_point[i] = MOI.get(model.inner, MOI.ObjectiveValue())
@@ -619,6 +648,7 @@ function MOI.optimize!(model::Optimizer)
     start_time = time()
     empty!(model.solutions)
     model.termination_status = MOI.OPTIMIZE_NOT_CALLED
+    model.subproblem_count = 0
     if model.f === nothing
         model.termination_status = MOI.INVALID_MODEL
         empty!(model.ideal_point)

src/algorithms/Chalmet.jl

@@ -29,7 +29,7 @@ function _solve_constrained_model(
     MOI.set(model.inner, MOI.ObjectiveFunction{typeof(g)}(), g)
     sets = MOI.LessThan.(rhs .- 1)
     c = MOI.Utilities.normalize_and_add_constraint.(model.inner, f, sets)
-    MOI.optimize!(model.inner)
+    optimize_inner!(model)
     status = MOI.get(model.inner, MOI.TerminationStatus())
     if !_is_scalar_status_optimal(status)
         MOI.delete.(model, c)
@@ -54,7 +54,7 @@ function minimize_multiobjective!(algorithm::Chalmet, model::Optimizer)
     f1, f2 = MOI.Utilities.scalarize(model.f)
     y1, y2 = zeros(2), zeros(2)
     MOI.set(model.inner, MOI.ObjectiveFunction{typeof(f2)}(), f2)
-    MOI.optimize!(model.inner)
+    optimize_inner!(model)
     status = MOI.get(model.inner, MOI.TerminationStatus())
     if !_is_scalar_status_optimal(status)
         return status, nothing
@@ -66,7 +66,7 @@ function minimize_multiobjective!(algorithm::Chalmet, model::Optimizer)
         f2,
         MOI.LessThan(y1[2]),
     )
-    MOI.optimize!(model.inner)
+    optimize_inner!(model)
     status = MOI.get(model.inner, MOI.TerminationStatus())
     if !_is_scalar_status_optimal(status)
         return status, nothing
@@ -75,7 +75,7 @@ function minimize_multiobjective!(algorithm::Chalmet, model::Optimizer)
     MOI.delete(model.inner, y1_constraint)
     push!(solutions, SolutionPoint(x1, y1))
     MOI.set(model.inner, MOI.ObjectiveFunction{typeof(f1)}(), f1)
-    MOI.optimize!(model.inner)
+    optimize_inner!(model)
     status = MOI.get(model.inner, MOI.TerminationStatus())
     if !_is_scalar_status_optimal(status)
         return status, nothing
@@ -90,7 +90,7 @@ function minimize_multiobjective!(algorithm::Chalmet, model::Optimizer)
         f1,
         MOI.LessThan(y2[1]),
     )
-    MOI.optimize!(model.inner)
+    optimize_inner!(model)
     status = MOI.get(model.inner, MOI.TerminationStatus())
     if !_is_scalar_status_optimal(status)
         return status, nothing

src/algorithms/Dichotomy.jl

@@ -61,7 +61,7 @@ function _solve_weighted_sum(
 )
     f = _scalarise(model.f, weights)
     MOI.set(model.inner, MOI.ObjectiveFunction{typeof(f)}(), f)
-    MOI.optimize!(model.inner)
+    optimize_inner!(model)
     status = MOI.get(model.inner, MOI.TerminationStatus())
     if !_is_scalar_status_optimal(status)
         return status, nothing

src/algorithms/DominguezRios.jl

@@ -153,7 +153,7 @@ function minimize_multiobjective!(algorithm::DominguezRios, model::Optimizer)
     for (i, f_i) in enumerate(scalars)
         MOI.set(model.inner, MOI.ObjectiveFunction{typeof(f_i)}(), f_i)
         MOI.set(model.inner, MOI.ObjectiveSense(), MOI.MIN_SENSE)
-        MOI.optimize!(model.inner)
+        optimize_inner!(model)
         status = MOI.get(model.inner, MOI.TerminationStatus())
         if !_is_scalar_status_optimal(status)
             return status, nothing
@@ -162,7 +162,7 @@ function minimize_multiobjective!(algorithm::DominguezRios, model::Optimizer)
         yI[i] = Y
         model.ideal_point[i] = Y
         MOI.set(model.inner, MOI.ObjectiveSense(), MOI.MAX_SENSE)
-        MOI.optimize!(model.inner)
+        optimize_inner!(model)
         status = MOI.get(model.inner, MOI.TerminationStatus())
         if !_is_scalar_status_optimal(status)
             _warn_on_nonfinite_anti_ideal(algorithm, MOI.MIN_SENSE, i)
@@ -206,7 +206,7 @@ function minimize_multiobjective!(algorithm::DominguezRios, model::Optimizer)
         #  * the `ϵ` term is already covered in `w`
         new_f = t_max + ϵ * sum(w[i] * (scalars[i] - yI[i]) for i in 1:n)
         MOI.set(model.inner, MOI.ObjectiveFunction{typeof(new_f)}(), new_f)
-        MOI.optimize!(model.inner)
+        optimize_inner!(model)
         if _is_scalar_status_optimal(model)
             X, Y = _compute_point(model, variables, model.f)
             obj = MOI.get(model.inner, MOI.ObjectiveValue())

src/algorithms/EpsilonConstraint.jl

@@ -116,7 +116,7 @@ function minimize_multiobjective!(
             break
         end
         MOI.set(model, MOI.ConstraintSet(), ci, MOI.LessThan{Float64}(bound))
-        MOI.optimize!(model.inner)
+        optimize_inner!(model)
         if !_is_scalar_status_optimal(model)
             break
         end

src/algorithms/Hierarchical.jl

@@ -100,7 +100,7 @@ function minimize_multiobjective!(algorithm::Hierarchical, model::Optimizer)
         new_vector_f = objectives[indices]
         new_f = _scalarise(new_vector_f, weights[indices])
         MOI.set(model.inner, MOI.ObjectiveFunction{typeof(new_f)}(), new_f)
-        MOI.optimize!(model.inner)
+        optimize_inner!(model)
         status = MOI.get(model.inner, MOI.TerminationStatus())
         if !_is_scalar_status_optimal(status)
             return status, nothing

src/algorithms/KirlikSayin.jl

@@ -93,7 +93,7 @@ function minimize_multiobjective!(algorithm::KirlikSayin, model::Optimizer)
     for (i, f_i) in enumerate(scalars)
         # Ideal point
         MOI.set(model.inner, MOI.ObjectiveFunction{typeof(f_i)}(), f_i)
-        MOI.optimize!(model.inner)
+        optimize_inner!(model)
         status = MOI.get(model.inner, MOI.TerminationStatus())
         if !_is_scalar_status_optimal(status)
             return status, nothing
@@ -102,7 +102,7 @@ function minimize_multiobjective!(algorithm::KirlikSayin, model::Optimizer)
         model.ideal_point[i] = yI[i] = Y
         # Nadir point
         MOI.set(model.inner, MOI.ObjectiveSense(), MOI.MAX_SENSE)
-        MOI.optimize!(model.inner)
+        optimize_inner!(model)
         status = MOI.get(model.inner, MOI.TerminationStatus())
         if !_is_scalar_status_optimal(status)
             # Repair ObjectiveSense before exiting
@@ -143,7 +143,7 @@ function minimize_multiobjective!(algorithm::KirlikSayin, model::Optimizer)
             )
             push!(ε_constraints, ci)
         end
-        MOI.optimize!(model.inner)
+        optimize_inner!(model)
         if !_is_scalar_status_optimal(model)
             _remove_rectangle(L, _Rectangle(_project(yI, k), uᵢ))
             MOI.delete.(model, ε_constraints)
@@ -160,7 +160,7 @@ function minimize_multiobjective!(algorithm::KirlikSayin, model::Optimizer)
             scalars[k],
             MOI.EqualTo(zₖ),
         )
-        MOI.optimize!(model.inner)
+        optimize_inner!(model)
         if !_is_scalar_status_optimal(model)
             MOI.delete.(model, ε_constraints)
             MOI.delete(model, zₖ_constraint)

src/algorithms/Lexicographic.jl

@@ -104,7 +104,7 @@ function _solve_in_sequence(
         end
         f = scalars[i]
         MOI.set(model.inner, MOI.ObjectiveFunction{typeof(f)}(), f)
-        MOI.optimize!(model.inner)
+        optimize_inner!(model)
         status = MOI.get(model.inner, MOI.TerminationStatus())
         primal_status = MOI.get(model.inner, MOI.PrimalStatus())
         if _is_scalar_status_feasible_point(primal_status)

src/algorithms/RandomWeighting.jl

@@ -46,7 +46,7 @@ function optimize_multiobjective!(algorithm::RandomWeighting, model::Optimizer)
     variables = MOI.get(model.inner, MOI.ListOfVariableIndices())
     f = _scalarise(model.f, ones(P))
     MOI.set(model.inner, MOI.ObjectiveFunction{typeof(f)}(), f)
-    MOI.optimize!(model.inner)
+    optimize_inner!(model)
     status = MOI.get(model.inner, MOI.TerminationStatus())
     if _is_scalar_status_optimal(status)
         X, Y = _compute_point(model, variables, model.f)
@@ -66,7 +66,7 @@ function optimize_multiobjective!(algorithm::RandomWeighting, model::Optimizer)
             weights = rand(P)
             f = _scalarise(model.f, weights)
             MOI.set(model.inner, MOI.ObjectiveFunction{typeof(f)}(), f)
-            MOI.optimize!(model.inner)
+            optimize_inner!(model)
             status = MOI.get(model.inner, MOI.TerminationStatus())
             if _is_scalar_status_optimal(status)
                 X, Y = _compute_point(model, variables, model.f)

src/algorithms/TambyVanderpooten.jl

@@ -99,7 +99,7 @@ function minimize_multiobjective!(
     for (i, f_i) in enumerate(scalars)
         MOI.set(model.inner, MOI.ObjectiveFunction{typeof(f_i)}(), f_i)
         MOI.set(model.inner, MOI.ObjectiveSense(), MOI.MIN_SENSE)
-        MOI.optimize!(model.inner)
+        optimize_inner!(model)
         status = MOI.get(model.inner, MOI.TerminationStatus())
         if !_is_scalar_status_optimal(status)
             return status, nothing
@@ -108,7 +108,7 @@ function minimize_multiobjective!(
         yI[i] = Y
         model.ideal_point[i] = Y
         MOI.set(model.inner, MOI.ObjectiveSense(), MOI.MAX_SENSE)
-        MOI.optimize!(model.inner)
+        optimize_inner!(model)
         status = MOI.get(model.inner, MOI.TerminationStatus())
         if !_is_scalar_status_optimal(status)
             _warn_on_nonfinite_anti_ideal(algorithm, MOI.MIN_SENSE, i)
@@ -157,7 +157,7 @@ function minimize_multiobjective!(
                 end
             end
         end
-        MOI.optimize!(model.inner)
+        optimize_inner!(model)
         if !_is_scalar_status_optimal(model)
             MOI.delete.(model, ε_constraints)
             return status, nothing
@@ -170,7 +170,7 @@ function minimize_multiobjective!(
             scalars[k],
             MOI.EqualTo(y_k),
         )
-        MOI.optimize!(model.inner)
+        optimize_inner!(model)
         if !_is_scalar_status_optimal(model)
             MOI.delete.(model, ε_constraints)
             MOI.delete(model, y_k_constraint)

test/algorithms/Chalmet.jl

@@ -252,6 +252,7 @@ function test_single_point()
     @test MOI.get(model, MOI.ResultCount()) == 1
     @test MOI.get(model, MOI.PrimalStatus()) == MOI.FEASIBLE_POINT
     @test ≈(MOI.get(model, MOI.VariablePrimal(), x), [1.0, 1.0]; atol = 1e-6)
+    @test MOI.get(model, MOA.SubproblemCount()) >= 1
     return
 end
 

test/algorithms/Dichotomy.jl

@@ -406,6 +406,7 @@ function test_vector_of_variables_objective()
     MOI.add_constraint(model, sum(1.0 * xi for xi in x), MOI.GreaterThan(1.0))
     MOI.optimize!(model)
     @test MOI.get(model, MOI.TerminationStatus()) == MOI.OPTIMAL
+    @test MOI.get(model, MOA.SubproblemCount()) >= 1
     return
 end
 

test/algorithms/Hierarchical.jl

@@ -57,6 +57,7 @@ function test_knapsack()
     @test ≈(x_sol, [0.9, 0, 0.9, 0.2]; atol = 1e-3)
     y_sol = MOI.get(model, MOI.ObjectiveValue())
     @test ≈(y_sol, P * x_sol .+ [0.0, 0.0, 0.0, 1_000.0]; atol = 1e-4)
+    @test MOI.get(model, MOA.SubproblemCount()) >= 1
     return
 end
 

test/algorithms/Lexicographic.jl

@@ -45,6 +45,7 @@ function test_knapsack()
     @test ≈(x_sol, [0.9, 1, 0, 0.1]; atol = 1e-3)
     y_sol = MOI.get(model, MOI.ObjectiveValue())
     @test ≈(y_sol, P * x_sol .+ [0.0, 0.0, 0.0, 1_000.0]; atol = 1e-4)
+    @test MOI.get(model, MOA.SubproblemCount()) == 8
     return
 end
 

test/algorithms/RandomWeighting.jl

@@ -78,6 +78,7 @@ function test_knapsack_min()
         [0, 0, 1, 1, 1, 0, 1, 1, 1, 1] => [-2854, -4636],
     ]
     @test MOI.get(model, MOI.ResultCount()) == length(results)
+    @test MOI.get(model, MOA.SubproblemCount()) >= 3
     for (i, (x_sol, y_sol)) in enumerate(results)
         @test ≈(x_sol, MOI.get(model, MOI.VariablePrimal(i), x); atol = 1e-6)
         @test ≈(y_sol, MOI.get(model, MOI.ObjectiveValue(i)); atol = 1e-6)

test/algorithms/Sandwiching.jl

@@ -57,6 +57,7 @@ function _test_molp(C, A, b, results, sense)
             MOI.get(model, MOI.ObjectiveValue(i)) for i in 1:N
     ])
     @test N == length(results)
+    @test MOI.get(model, MOA.SubproblemCount()) >= length(results)
     for (sol, res) in zip(solutions, results)
         x_sol, y_sol = sol
         x_res, y_res = res