Write MOF files by default

Project.toml

@@ -23,7 +23,7 @@ HTTP = "0.8.1"
 JSON = "0.21"
 JuMP = "~0.20"
 JuliaFormatter = "0.1.37"
-MathOptFormat = "0.2"
+MathOptFormat = "0.4"
 RecipesBase = "0.7"
 Reexport = "0.2"
 TimerOutputs = "0.5"

docs/src/guides/debug_a_model.md

@@ -53,13 +53,8 @@ Now to write out the problem to a file. We'll get a few warnings because some
 variables and constraints don't have names. They don't matter, so ignore them.
 
 ```jldoctest tutorial_eight; filter=r"MathOptFormat\ .+?MathOptFormat\.jl"
-julia> SDDP.write_subproblem_to_file(model[1], "subproblem", format=:lp)
+julia> SDDP.write_subproblem_to_file(model[1], "subproblem.lp")
 
-```
-
-We can check the file by reading it back in again.
-
-```jldoctest tutorial_eight
 julia> read("subproblem.lp") |> String |> print
 minimize
 obj: 1.1 x_out + 1 x2
@@ -80,8 +75,7 @@ are not in the original problem formulation.
 ```jldoctest tutorial_eight; filter=r"MathOptFormat\ .+?MathOptFormat\.jl"
 julia> SDDP.parameterize(model[1], 3.3)
 
-julia> SDDP.write_subproblem_to_file(model[1], "subproblem", format=:lp)
-
+julia> SDDP.write_subproblem_to_file(model[1], "subproblem.lp")
 
 julia> read("subproblem.lp") |> String |> print
 minimize

src/algorithm.jl

@@ -209,46 +209,23 @@ stage_objective_value(stage_objective::Real) = stage_objective
 stage_objective_value(stage_objective) = JuMP.value(stage_objective)
 
 """
-    write_subproblem_to_file(node::Node, filename::String; format=:both)
+    write_subproblem_to_file(node::Node, filename::String; throw_error::Bool = false)
 
 Write the subproblem contained in `node` to the file `filename`.
-
-`format` should be one of `:mps`, `:lp`, or `:both`.
 """
-function write_subproblem_to_file(
-    node::Node,
-    filename::String;
-    format::Symbol = :both,
-    throw_error::Bool = false,
-)
-    if format ∉ (:mps, :lp, :both)
-        error("Invalid `format=$(format)`. Must be `:mps`, `:lp`, or `:both`.")
-    end
-    if format == :mps || format == :both
-        mps = MathOptFormat.MPS.Model()
-        MOI.copy_to(mps, JuMP.backend(node.subproblem))
-        MOI.write_to_file(mps, filename * ".mps")
-    end
-    if format == :lp || format == :both
-        lp = MathOptFormat.LP.Model()
-        MOI.copy_to(lp, JuMP.backend(node.subproblem))
-        MOI.write_to_file(lp, filename * ".lp")
-    end
+function write_subproblem_to_file(node::Node, filename::String; throw_error::Bool = false)
+    model = MathOptFormat.Model(filename = filename)
+    MOI.copy_to(model, JuMP.backend(node.subproblem))
+    MOI.write_to_file(model, filename)
     if throw_error
         error(
-            "Unable to retrieve dual solution from ",
-            node.index,
-            ".",
-            "\n  Termination status: ",
-            JuMP.termination_status(node.subproblem),
-            "\n  Primal status:      ",
-            JuMP.primal_status(node.subproblem),
-            "\n  Dual status:        ",
-            JuMP.dual_status(node.subproblem),
-            ".\n An MPS file was written to `subproblem.mps` and an LP file ",
-            "written to `subproblem.lp`. See ",
-            "https://odow.github.io/SDDP.jl/latest/tutorial/06_warnings/#Numerical-stability-1",
-            " for more information.",
+            "Unable to retrieve solution from $(node.index).\n",
+            "  Termination status: $(JuMP.termination_status(node.subproblem))\n",
+            "  Primal status:      $(JuMP.primal_status(node.subproblem))\n",
+            "  Dual status:        $(JuMP.dual_status(node.subproblem)).\n",
+            "A MathOptFormat file was written to `$(filename)`.\n",
+            "See https://odow.github.io/SDDP.jl/latest/tutorial/06_warnings/#Numerical-stability-1",
+            "\nfor more information.",
         )
     end
 end
@@ -295,7 +272,11 @@ function solve_subproblem(
 
     # Test for primal feasibility.
     if JuMP.primal_status(node.subproblem) != JuMP.MOI.FEASIBLE_POINT
-        write_subproblem_to_file(node, "subproblem", throw_error = true)
+        write_subproblem_to_file(
+            node,
+            "subproblem_$(node.index).mof.json",
+            throw_error = true,
+        )
     end
     # If require_duals = true, check for dual feasibility and return a dict with
     # the dual on the fixed constraint associated with each incoming state

src/modeling_aids.jl

@@ -24,7 +24,7 @@
 # OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 # SOFTWARE.
 
-function find_min(x::Vector{T}, y::T) where {T <: Real}
+function find_min(x::Vector{T}, y::T) where {T<:Real}
     best_i = 0
     best_z = Inf
     for i = 1:length(x)
@@ -40,7 +40,7 @@ end
 function lattice_approximation(f::Function, states::Vector{Int}, scenarios::Int)
     path = f()::Vector{Float64}
     support = [fill(path[t], states[t]) for t = 1:length(states)]
-    probability = [zeros(states[t - 1], states[t]) for t = 2:length(states)]
+    probability = [zeros(states[t-1], states[t]) for t = 2:length(states)]
     prepend!(probability, Ref(zeros(1, states[1])))
     distance = 0.0
     for n = 1:scenarios
@@ -55,7 +55,8 @@ function lattice_approximation(f::Function, states::Vector{Int}, scenarios::Int)
             min_dist, best_idx = find_min(support[t], path[t])
             dist += min_dist^2
             probability[t][last_index, best_idx] += 1.0
-            support[t][best_idx] -= min_dist * (support[t][best_idx] - path[t]) / (3000 + n)^0.75
+            support[t][best_idx] -=
+                min_dist * (support[t][best_idx] - path[t]) / (3000 + n)^0.75
             last_index = best_idx
         end
         distance = (distance * (n - 1) + dist) / n
@@ -114,7 +115,9 @@ nodes between stages. Alternatively, `budget` can be a `Vector{Int}`, which deta
 number of Markov state to have in each stage.
 """
 function MarkovianGraph(
-    simulator::Function; budget::Union{Int, Vector{Int}}, scenarios::Int = 1000
+    simulator::Function;
+    budget::Union{Int,Vector{Int}},
+    scenarios::Int = 1000,
 )
     scenarios = max(scenarios, 10)
     states = allocate_support_budget(simulator, budget, scenarios)
@@ -127,7 +130,7 @@ function MarkovianGraph(
     for t = 2:length(support)
         for (j, sj) in enumerate(support[t])
             add_node(g, (t, sj))
-            for (i, si) in enumerate(support[t - 1])
+            for (i, si) in enumerate(support[t-1])
                 add_edge(g, (t - 1, si) => (t, sj), probability[t][i, j])
             end
         end

test/algorithm.jl

@@ -161,11 +161,18 @@ end
         @constraint(node, x.out <= -1)
         @stageobjective(node, x.out)
     end
-    @test_throws Exception SDDP.train(model; iteration_limit = 1, print_level = 0)
-    @test isfile("subproblem.mps")
-    rm("subproblem.mps")
-    @test isfile("subproblem.lp")
-    rm("subproblem.lp")
+    ex = ErrorException("""
+    Unable to retrieve solution from 1.
+      Termination status: INFEASIBLE
+      Primal status:      NO_SOLUTION
+      Dual status:        INFEASIBILITY_CERTIFICATE.
+    A MathOptFormat file was written to `subproblem_1.mof.json`.
+    See https://odow.github.io/SDDP.jl/latest/tutorial/06_warnings/#Numerical-stability-1
+    for more information.""")
+
+    @test_throws ex SDDP.train(model; iteration_limit = 1, print_level = 0)
+    @test isfile("subproblem_1.mof.json")
+    rm("subproblem_1.mof.json")
 end
 
 @testset "refine_at_similar_nodes" begin

test/modeling_aids.jl

@@ -34,9 +34,7 @@ end
 end
 
 @testset "lattice_approximation" begin
-    support, probability = SDDP.lattice_approximation(
-        () -> rand(5), [1, 2, 3, 4, 5], 100
-    )
+    support, probability = SDDP.lattice_approximation(() -> rand(5), [1, 2, 3, 4, 5], 100)
     for (t, s) in enumerate(support)
         @test length(s) == t
         @test all(x -> 0 <= x <= 1, s)