Add POI.Parameters

docs/src/manual.md

@@ -73,11 +73,31 @@ To change a given parameter's value, access its `ConstraintIndex` and set it to
 MOI.set(optimizer, MOI.ConstraintSet(), cy, POI.Parameter(2.0))
 ```
 
+as an alternative one can also use the custom MOI attribute `ParameterValue`
+
+```julia
+MOI.set(optimizer, POI.ParameterValue(), y, 2.0)
+```
+
 ### Retrieving the dual of a parameter
 
 Given an optimized model, one can calculate the dual associated to a parameters, **as long as it is an additive term in the constraints or objective**.
 One can do so by getting the `MOI.ConstraintDual` attribute of the paraameter's `MOI.ConstraintIndex`:
 
 ```julia
 MOI.get(optimizer, MOI.ConstraintDual, cy)
+```
+
+as an alternative one can also use the custom MOI attribute `ParameterValue`
+
+```julia
+MOI.get(optimizer, POI.ParameterDual(), y)
+```
+
+## API
+
+```@docs
+POI.Parameter
+POI.Parameters
+POI.ParametricOptimizer
 ```

src/ParametricOptInterface.jl

@@ -1,6 +1,8 @@
 module ParametricOptInterface
 
 using MathOptInterface
+import Base.copy
+
 const MOI = MathOptInterface
 const MOIU = MathOptInterface.Utilities
 const DD = MOIU.DoubleDicts
@@ -9,7 +11,7 @@ const PARAMETER_INDEX_THRESHOLD = 1_000_000_000_000_000_000
 const SUPPORTED_SETS = [MOI.LessThan{Float64}, MOI.EqualTo{Float64}, MOI.GreaterThan{Float64}]
 
 """
-    Parameter(Float64)
+    Parameter <: MOI.AbstractScalarSet
 
 The `Parameter` structure stores the numerical value associated to a given parameter.
 # Example:
@@ -22,6 +24,25 @@ struct Parameter <: MOI.AbstractScalarSet
     val::Float64
 end
 
+"""
+    Parameters <: MOI.AbstractVectorSet
+
+The `Parameters` structure stores the numerical values associated to parameters.
+# Example:
+```julia-repl
+julia> ParametricOptInterface.Parameters(ones(2))
+ParametricOptInterface.Parameters([1.0, 1.0], 2)
+```
+"""
+struct Parameters <: MOI.AbstractVectorSet
+    vals::Vector{Float64}
+    dimension::Int
+    function Parameters(vals::Vector{Float64})
+        return new(vals, length(vals))
+    end
+end
+Base.copy(set::Parameters) = Parameters(copy(set.vals))
+
 """
     ParametricOptimizer{T, OT <: MOI.ModelLike} <: MOI.AbstractOptimizer
 
@@ -308,6 +329,26 @@ function MOI.add_constrained_variable(model::ParametricOptimizer, set::Parameter
     return p, cp
 end
 
+"""
+    MOI.add_constrained_variables(model::ParametricOptimizer, set::Parameters)
+
+Adds multiple parameters, that is, variables parameterized to the value provided in `set`, to the `model` specified.
+The `model` must be a `ParametricOptInterface.ParametricOptimizer` model.
+Returns vector of MOI.VariableIndex of the parameterized variables and the vector of MOI.ConstraintIndex associated.
+"""
+function MOI.add_constrained_variables(model::ParametricOptimizer, set::Parameters)
+    variables = Vector{MOI.VariableIndex}(undef, set.dimension)
+    constraints = Vector{MOI.ConstraintIndex{MOI.SingleVariable, Parameter}}(undef, set.dimension)
+    for (i, val) in enumerate(set.vals)
+        next_parameter_index!(model)
+        variables[i] = MOI.VariableIndex(model.last_parameter_index_added)
+        model.parameters[variables[i]] = val
+        constraints[i] = MOI.ConstraintIndex{MOI.SingleVariable, Parameter}(model.last_parameter_index_added)
+        update_number_of_parameters!(model)
+    end
+    return variables, constraints
+end
+
 function MOI.add_constraint(model::ParametricOptimizer, f::MOI.SingleVariable, set::MOI.AbstractScalarSet)
     if is_parameter_in_model(model, f.variable)
         error("Cannot constrain a parameter")
@@ -372,7 +413,7 @@ Sets the parameter to a given value, using its `MOI.VariableIndex` as reference.
 
 #Example:
 ```julia-repl
-julia> MOI.set(model, ParameterValue(), w, 2.0)
+julia> MOI.set(model, POI.ParameterValue(), w, 2.0)
 2.0
 ```
 """

test/jump_tests.jl

@@ -116,4 +116,60 @@ end
     @test_broken isapprox.(value(x[1]), 0.0, atol = ATOL)
     @test_broken isapprox.(value(x[2]), 2.0, atol = ATOL)
     @test_broken isapprox.(value(y), 2.0, atol = ATOL)
+end
+
+@testset "JuMP direct model - Variable in POI.Parameters" begin
+    optimizer = POI.ParametricOptimizer(GLPK.Optimizer())
+
+    model = direct_model(optimizer)
+
+    @variable(model, x[i=1:2] >= 0)
+
+    @variable(model, y[i=1:3] in POI.Parameters(zeros(3)))
+
+    @constraint(model, 2*x[1] + x[2] + y[1] <= 4)
+    @constraint(model, 1*x[1] + 2*x[2] + y[3] <= 4)
+
+    @objective(model, Max, 4*x[1] + 3*x[2] + y[2])
+
+    optimize!(model)
+
+    @test isapprox.(value(x[1]), 4.0/3.0, atol = ATOL)
+    @test isapprox.(value(x[2]), 4.0/3.0, atol = ATOL)
+    @test isapprox.(value(y[1]), 0, atol = ATOL)
+
+    # ===== Set parameter value =====
+    MOI.set(model, POI.ParameterValue(), y[1], 2.0)
+    optimize!(model)
+
+    @test isapprox.(value(x[1]), 0.0, atol = ATOL)
+    @test isapprox.(value(x[2]), 2.0, atol = ATOL)
+    @test isapprox.(value(y[1]), 2.0, atol = ATOL)
+end
+
+@testset "JuMP model - Variable in POI.Parameters" begin
+    model = Model(() -> POI.ParametricOptimizer(GLPK.Optimizer()))
+
+    @variable(model, x[i=1:2] >= 0)
+
+    @variable(model, y[i=1:3] in POI.Parameters(zeros(3)))
+
+    @constraint(model, 2*x[1] + x[2] + y[1] <= 4)
+    @constraint(model, 1*x[1] + 2*x[2] + y[3] <= 4)
+
+    @objective(model, Max, 4*x[1] + 3*x[2] + y[2])
+
+    @test_broken optimize!(model)
+
+    @test_broken isapprox.(value(x[1]), 4.0/3.0, atol = ATOL)
+    @test_broken isapprox.(value(x[2]), 4.0/3.0, atol = ATOL)
+    @test_broken isapprox.(value(y[1]), 0, atol = ATOL)
+
+    # ===== Set parameter value =====
+    @test_broken MOI.set(model, POI.ParameterValue(), y[1], 2.0)
+    @test_broken optimize!(model)
+
+    @test_broken isapprox.(value(x[1]), 0.0, atol = ATOL)
+    @test_broken isapprox.(value(x[2]), 2.0, atol = ATOL)
+    @test_broken isapprox.(value(y[1]), 2.0, atol = ATOL)
 end

test/production_problem_test.jl

@@ -90,6 +90,98 @@
 
 end
 
+@testset "Production Problem with Parameters as VectorSet" begin
+    optimizer = POI.ParametricOptimizer(GLPK.Optimizer())
+
+    c = [4.0, 3.0]
+    A1 = [2.0, 1.0, 1.0]
+    A2 = [1.0, 2.0, 1.0]
+    b1 = 4.0
+    b2 = 4.0
+
+    x = MOI.add_variables(optimizer, length(c))
+
+    @test typeof(x[1]) == MOI.VariableIndex
+
+    vars, cons = MOI.add_constrained_variables(optimizer, POI.Parameters(zeros(3)))
+    w, y, z = vars
+    cw, cy, cz = cons
+
+    @test MOI.get(optimizer, MOI.VariablePrimal(), w) == 0
+
+    for x_i in x
+        MOI.add_constraint(optimizer, MOI.SingleVariable(x_i), MOI.GreaterThan(0.0))
+    end
+
+    cons1 = MOI.ScalarAffineFunction(MOI.ScalarAffineTerm.(A1, [x[1], x[2], y]), 0.0)
+    MOI.add_constraint(optimizer, cons1, MOI.LessThan(b1))
+
+    cons2 = MOI.ScalarAffineFunction(MOI.ScalarAffineTerm.(A2, [x[1], x[2], z]), 0.0)
+    MOI.add_constraint(optimizer, cons2, MOI.LessThan(b2))
+
+    @test cons1.terms[1].coefficient == 2
+    @test POI.is_parameter_in_model(optimizer, cons2.terms[3].variable_index)
+
+    obj_func = MOI.ScalarAffineFunction(MOI.ScalarAffineTerm.([c[1], c[2], 3.0], [x[1], x[2], w]), 0.0)
+    MOI.set(optimizer, MOI.ObjectiveFunction{MOI.ScalarAffineFunction{Float64}}(), obj_func)
+    MOI.set(optimizer, MOI.ObjectiveSense(), MOI.MAX_SENSE)
+
+    MOI.optimize!(optimizer)
+
+    MOI.get(optimizer, MOI.TerminationStatus())
+
+    MOI.get(optimizer, MOI.PrimalStatus())
+
+    @test isapprox.(MOI.get(optimizer, MOI.ObjectiveValue()), 28/3, atol = ATOL)
+    @test isapprox.(MOI.get(optimizer, MOI.VariablePrimal(), x[1]), 4/3, atol = ATOL)
+    @test isapprox.(MOI.get(optimizer, MOI.VariablePrimal(), x[2]), 4/3, atol = ATOL)
+
+    @test isapprox(MOI.get(optimizer, MOI.ConstraintDual(), cy), -5/3, atol = ATOL)
+    @test isapprox(MOI.get(optimizer, MOI.ConstraintDual(), cz), -2/3, atol = ATOL)
+    @test isapprox(MOI.get(optimizer, MOI.ConstraintDual(), cw), 3.0, atol = ATOL)
+
+    MOI.get(optimizer, MOI.VariablePrimal(), w)
+    MOI.get(optimizer, MOI.VariablePrimal(), y)
+    MOI.get(optimizer, MOI.VariablePrimal(), z)
+
+    MOI.set(optimizer, MOI.ConstraintSet(), cw, POI.Parameter(2.0))
+    MOI.set(optimizer, MOI.ConstraintSet(), cy, POI.Parameter(1.0))
+    MOI.set(optimizer, MOI.ConstraintSet(), cz, POI.Parameter(1.0))
+
+    MOI.optimize!(optimizer)
+
+    @test MOI.get(optimizer, MOI.VariablePrimal(), w) == 2.0
+    @test MOI.get(optimizer, MOI.VariablePrimal(), y) == 1.0
+    @test MOI.get(optimizer, MOI.VariablePrimal(), z) == 1.0
+
+    @test isapprox.(MOI.get(optimizer, MOI.ObjectiveValue()), 13.0, atol = ATOL)
+    @test MOI.get.(optimizer, MOI.VariablePrimal(), x) == [1.0, 1.0]
+
+    @test isapprox(MOI.get(optimizer, MOI.ConstraintDual(), cy), -5/3, atol = ATOL)
+    @test isapprox(MOI.get(optimizer, MOI.ConstraintDual(), cz), -2/3, atol = ATOL)
+    @test isapprox(MOI.get(optimizer, MOI.ConstraintDual(), cw), 3.0, atol = ATOL)
+
+    MOI.set(optimizer, MOI.ConstraintSet(), cw, POI.Parameter(0.0))
+
+    MOI.optimize!(optimizer)
+
+    @test MOI.get(optimizer, MOI.VariablePrimal(), w) == 0.0
+    @test MOI.get(optimizer, MOI.VariablePrimal(), y) == 1.0
+    @test MOI.get(optimizer, MOI.VariablePrimal(), z) == 1.0
+    @test isapprox.(MOI.get(optimizer, MOI.ObjectiveValue()), 7, atol = ATOL)
+    @test MOI.get.(optimizer, MOI.VariablePrimal(), x) == [1.0, 1.0]
+
+
+    MOI.set(optimizer, MOI.ConstraintSet(), cy, POI.Parameter(-5.0))
+    MOI.optimize!(optimizer)
+    @test isapprox.(MOI.get(optimizer, MOI.ObjectiveValue()), 12.0, atol = ATOL)
+    @test MOI.get.(optimizer, MOI.VariablePrimal(), x) == [3.0, 0.0]
+
+    @test isapprox(MOI.get(optimizer, MOI.ConstraintDual(), cy), 0, atol = ATOL)
+    @test isapprox(MOI.get(optimizer, MOI.ConstraintDual(), cz), -4, atol = ATOL)
+
+end
+
 
 @testset "Production Problem variation on parameters for duals" begin
     optimizer = POI.ParametricOptimizer(GLPK.Optimizer())