Hardcode the polishing problem

- Load the polishing problem with known solution from Mosek from a
file
- Add FileIO to test dependencies

Project.toml

@@ -24,6 +24,7 @@ julia = "1"
 Compat = "34da2185-b29b-5c13-b0c7-acf172513d20"
 Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
 Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
+FileIO = "5789e2e9-d7fb-5bc7-8068-2c6fae9b9549"
 
 [targets]
-test = ["Test", "Compat", "Random"]
+test = ["Test", "Compat", "Random", "FileIO"]

test/polishing.jl

@@ -73,52 +73,19 @@ tol = 1e-3
     end
 
     @testset "polish_random" begin
-        using Random
-        rng = Random.MersenneTwister(666)
 
-
-        n = 30
-        m = 50
-        Pt = sprandn(rng, n, n, 0.5)
-        P = Pt * Pt'
-        q = randn(rng, n)
-        A = sprandn(rng, m, n, 0.5)
-        l = -3 .+ randn(rng, m)
-        u = 3 .+ randn(rng, m)
-
-        # Solve problem
+        # load randomly generated problem with known accurate solution from Mosek
+        problem_data = FileIO.load("./problem_data/random_polish_qp.jld2");
+        P = problem_data["P"]; q = problem_data["q"]; A = problem_data["A"]; u = problem_data["u"]; l = problem_data["l"];
         options = setup_polishing()
         model = OSQP.Model()
         OSQP.setup!(model; P=P, q=q, A=A, l=l, u=u, options...)
         results = OSQP.solve!(model)
 
 
-        # # Solve with Mosek for reference solution
-        # using JuMP, MosekTools
-        #
-        # jump_model = JuMP.Model(with_optimizer(Mosek.Optimizer));
-        # @variable(jump_model, x[1:n]);
-        # @objective(jump_model, Min, 0.5 * x' * P * x  + q' * x);
-        # con = @constraint(jump_model, [i = 1:m], l[i] <= Vector(A[i, :])' * x <= u[i]);
-        #
-        # status = JuMP.optimize!(jump_model);
-        # obj_test = JuMP.objective_value(jump_model);
-        # x_test = JuMP.value.(x);
-        # y_test = -JuMP.dual.(con)
-        # println(x_test)
-        # println(y_test)
-        # println(obj_test)
-
-        x_test = [0.41137084480890623, 0.28318004895183285, 0.4777528132318454, -0.23164842615545447, -0.3629757238797728, 0.14431287989072183, 0.6438564504118605, -0.5761094276277431, -0.05982787893150773, 0.15030576101377005, 0.028787631121672347, -0.27216082894024013, 0.17141240340176417, -0.273743108255697, 0.15630235714216442, -0.06763333754014154, -1.020331708945701, -0.012617123323538763, -0.6683845746964124, 0.5531428372752116, -0.1320489823459991, -0.36749007989638366, 0.0595756204214971, -0.2428072482922348, 0.6357746681280918, 0.9216345137918707, -0.14249830279317166, 0.3440205926304816, -0.2570648319065089, -0.6783094533788961];
-
-
-        y_test = [-1.0282863481250939e-13, -3.37258508945909e-14, 3.4296811845328215e-14, 1.777147102264673e-14, -2.4893358571930445e-14, 1.322599809069287e-14, -1.6595755569514817e-13, -8.317794666110387e-14, -9.2640088352221e-15, 1.253500392114895e-13, 1.0854356507664321e-13, 3.1706833729517506e-14, -6.241276348174211e-13, -8.535981813127953e-14, 3.5093009175770576e-13, 7.29118035417498e-14, 3.618176044780794e-14, -5.491504492855572e-15, -5.667489500506908e-14, 0.0050306281848406156, 5.067317066489025e-14, -1.0803519472826196e-14, 5.237647375580867e-14, 9.474300815751858e-14, 2.9853746376673826e-14, -1.1895032769230835e-14, -1.3441509011114794e-13, -0.08678665919072658, -0.04228028938433161, -6.59489689104704e-14, -5.710323913717883e-14, 5.390342831441443e-14, -6.35787374994456e-14, 1.252381079606396e-13, 0.22585055797097334, -0.9388516056549417, 1.0170683219018741e-13, 0.2625913401764051, -3.454237726808208e-13, -0.08487116932540341, 5.24402777336478e-14, 4.099664389450276e-14, 3.614515769510322e-14, 2.1517424811393735e-14, 1.5159680259859412e-14, -4.016603096658772e-14, 5.0318209747657e-13, -2.981549776801216e-14, -1.1248637104537303e-14, -3.8425095057919334e-13]
-
-        obj_test = -6.0748863556792045
-
-        @test isapprox(results.x, x_test, atol=tol)
-        @test isapprox(results.y, y_test, atol=tol)
-        @test isapprox(results.info.obj_val, obj_test, atol=tol)
+        @test isapprox(results.x, problem_data["x_test"], atol=tol)
+        @test isapprox(results.y, problem_data["y_test"], atol=tol)
+        @test isapprox(results.info.obj_val, problem_data["obj_test"], atol=tol)
         @test results.info.status_polish == 1
 
     end