Updated tests, one test failing.

test/problems/square_wave_problem.jl

@@ -0,0 +1,25 @@
+function square_wave_problem()
+
+    g = kron([1, -1, 1, -1, 1], ones(5))
+
+    M_bf = 5
+
+    ζvec = [100, 30, 10, 3, 1, 0.3, 0.1, 0.03, 0.01]
+
+    # For each m there are many non-unique solution
+    I_sols =
+    [Vector{Int}() for m=1:24]
+
+    f_sols =
+        [[24.0,
+        18.10194908759502,
+        15.757575757575758,
+        6.602807597027249,
+        4.84848484848485,
+        4.285714285714288,
+        2.4242424242424248,
+        2.1356521739130434];
+        zeros(Float64, 16)]
+
+    return g, M_bf, ζvec, I_sols, f_sols
+end

test/problems/straight_line_problem.jl

@@ -0,0 +1,17 @@
+function straight_line_problem()
+
+    g = 0:0.02:1
+
+    M_bf = 2
+
+    ζvec = [100, 30, 10, 3, 1, 0.3, 0.1, 0.03, 0.01]
+
+    # For each m there are many non-unique solution
+    I_sols =
+    [[[1, length(g)]];
+     [Vector{Int}() for m=1:9]]
+
+    f_sols = zeros(Float64, 10)
+
+    return g, M_bf, ζvec, I_sols, f_sols
+end

test/test_discrete_fit.jl

@@ -5,15 +5,18 @@ include(joinpath(Pkg.dir("DynamicApproximations"),"src","dev.jl"))
 include(joinpath(Pkg.dir("DynamicApproximations"),"test","problems", "snp500_problem.jl"))
 include(joinpath(Pkg.dir("DynamicApproximations"),"test","problems", "exp_problem.jl"))
 include(joinpath(Pkg.dir("DynamicApproximations"),"test","problems", "square_wave_problem.jl"))
+include(joinpath(Pkg.dir("DynamicApproximations"),"test","problems", "straight_line_problem.jl"))
 
 ##
 
-@testset "Data set: $problem_fcn" for problem_fcn in [exp_problem, snp500_problem]
+@testset "Data set: $problem_fcn" for problem_fcn in [straight_line_problem, square_wave_problem, exp_problem, snp500_problem]
 
 ##
-#problem_fcn = snp500_problem
+problem_fcn = straight_line_problem
 g, M_bf, ζvec, I_sols, f_sols = problem_fcn()
 
+g = 0:0.1:1
+
 M = length(I_sols)
 
 #g = exp.(linspace(0,5,11))
@@ -24,24 +27,26 @@ cost_last = dev.QuadraticPolynomial(1.0, -2*g[end], g[end]^2)
 Λ = dev.find_optimal_fit(ℓ, cost_last, M, Inf);
 
 @testset "Data set: $problem_fcn, constrained with m=$m" for m in 1:M
-    I2, y2, f2 = dev.recover_optimal_index_set(Λ[m, 1])
-    Y2, f2_2 = dev.find_optimal_y_values(ℓ, cost_last, I2)
+    I, Y, f = dev.recover_solution(Λ[m, 1], ℓ, cost_last)
 
-    @test I2 == I_sols[m]
-    @test Y2[1] ≈ y2  rtol=1e-10
-    @test f2 ≈ f_sols[m]   rtol=1e-10
+    @test isempty(I_sols[m]) || I == I_sols[m]
+    @test f ≈ f_sols[m]   rtol=1e-10 atol=1e-10
 end
 
 
 ##
 
 @testset "Data set: $problem_fcn, bruteforce with m=$m" for m in 1:M_bf
-    Y_recovered, f_recovered = dev.find_optimal_y_values(ℓ, cost_last, I_sols[m])
+
     I_bf, Y_bf, f_bf = dev.brute_force_optimization(ℓ, cost_last, m)
 
-    @test I_bf == I_sols[m]
-    @test Y_bf ≈ Y_recovered    rtol=1e-10
+    @test isempty(I_sols[m]) || I_bf == I_sols[m]
     @test f_bf ≈ f_sols[m]  rtol=1e-10 atol=1e-10
+
+    if !isempty(I_sols[m])
+        Y_recovered, f_recovered = dev.find_optimal_y_values(ℓ, cost_last, I_sols[m])
+        @test Y_bf ≈ Y_recovered    rtol=1e-10
+    end
 end
 
 ##
@@ -59,7 +64,7 @@ end
     m_expected = indmin([f_sols[m] + m*ζ for m=1:length(f_sols)])
 
     @test m_expected == length(I) - 1
-    @test I == I_sols[m_expected]
+    @test isempty(I_sols[m]) || I == I_sols[m_expected]
     @test f_reg ≈ f_sols[m_expected] + ζ*m_expected
 
     #println(problem_fcn, "  " , length(I))