update tests

Author: Vaibhavdixit02

src/OptimizationBase.jl

@@ -10,7 +10,8 @@ end
 
 using ArrayInterface, Base.Iterators, SparseArrays, LinearAlgebra
 using SymbolicIndexingInterface
-using SymbolicAnalysis: propagate_sign, propagate_curvature, propagate_gcurvature
+using SymbolicAnalysis: propagate_sign, propagate_curvature, propagate_gcurvature,
+                        getcurvature, getgcurvature, getsign
 import Symbolics
 import Manifolds
 import Symbolics: variable, Equation, Inequality, unwrap, @variables

src/cache.jl

@@ -22,24 +22,27 @@ function OptimizationCache(prob::SciMLBase.OptimizationProblem, opt, data = DEFA
         abstol::Union{Number, Nothing} = nothing,
         reltol::Union{Number, Nothing} = nothing,
         progress = false,
-        structural_analysis = true,
+        structural_analysis = false,
+        manifold = nothing,
         kwargs...)
     reinit_cache = OptimizationBase.ReInitCache(prob.u0, prob.p)
     num_cons = prob.ucons === nothing ? 0 : length(prob.ucons)
     f = OptimizationBase.instantiate_function(prob.f, reinit_cache, prob.f.adtype, num_cons)
 
-    if (f.sys === nothing || f.sys isa SymbolicIndexingInterface.SymbolCache{Nothing, Nothing, Nothing}) && structural_analysis
+    if (f.sys === nothing ||
+        f.sys isa SymbolicIndexingInterface.SymbolCache{Nothing, Nothing, Nothing}) &&
+       structural_analysis
         try
-            vars =
-            if prob.u0 isa Matrix
+            vars = if prob.u0 isa Matrix
                 @variables X[1:size(prob.u0, 1), 1:size(prob.u0, 2)]
             else
-                ArrayInterface.restructure(prob.u0, [variable(:x, i) for i in eachindex(prob.u0)])
+                ArrayInterface.restructure(
+                    prob.u0, [variable(:x, i) for i in eachindex(prob.u0)])
             end
             params = if prob.p isa SciMLBase.NullParameters
                 []
-            # elseif prob.p isa MTK.MTKParameters
-            #     [variable(:α, i) for i in eachindex(vcat(p...))]
+            elseif prob.p isa MTK.MTKParameters
+                [variable(:α, i) for i in eachindex(vcat(p...))]
             else
                 ArrayInterface.restructure(p, [variable(:α, i) for i in eachindex(p)])
             end
@@ -87,7 +90,7 @@ function OptimizationCache(prob::SciMLBase.OptimizationProblem, opt, data = DEFA
                 cons_expr = nothing
             end
         catch err
-            throw(ArgumentError("Automatic symbolic expression generation with ModelingToolkit failed with error: $err.
+            throw(ArgumentError("Automatic symbolic expression generation with failed with error: $err.
             Try by setting `structural_analysis = false` instead if the solver doesn't require symbolic expressions."))
         end
     else
@@ -96,25 +99,28 @@ function OptimizationCache(prob::SciMLBase.OptimizationProblem, opt, data = DEFA
         obj_expr = f.expr
         cons_expr = f.cons_expr
     end
-    try
-        obj_expr = obj_expr |> Symbolics.unwrap
-        obj_expr = propagate_curvature(propagate_sign(obj_expr))
-        @info "Objective Euclidean curvature: $(SymbolicAnalysis.getcurvature(obj_expr))"
-    catch
-        @info "No euclidean atom available"
-    end
 
-    try
-        obj_expr = SymbolicAnalysis.propagate_gcurvature(propagate_sign(obj_expr), prob.kwargs[1])
-        @info "Objective Geodesic curvature: $(SymbolicAnalysis.getgcurvature(obj_expr))"
-    catch e
-        @show e
+    if obj_expr !== nothing
+        try
+            obj_expr = obj_expr |> Symbolics.unwrap
+            obj_expr = propagate_curvature(propagate_sign(obj_expr))
+            @info "Objective Euclidean curvature: $(getcurvature(obj_expr))"
+        catch
+            @info "No euclidean atom available"
+        end
+
+        try
+            obj_expr = propagate_gcurvature(propagate_sign(obj_expr), manifold)
+            @info "Objective Geodesic curvature: $(getgcurvature(obj_expr))"
+        catch
+            @info "No geodesic atom available"
+        end
     end
 
-    if !isnothing(cons_expr)
+    if cons_expr !== nothing
         cons_expr = cons_expr .|> Symbolics.unwrap
         cons_expr = propagate_curvature.(propagate_sign.(cons_expr))
-        @info "Constraints Euclidean curvature: $(SymbolicAnalysis.getcurvature.(cons_expr))"
+        @info "Constraints Euclidean curvature: $(getcurvature.(cons_expr))"
     end
 
     return OptimizationCache(f, reinit_cache, prob.lb, prob.ub, prob.lcons,

test/adtests.jl

@@ -521,7 +521,7 @@ optprob.hess(H2, x0)
     @test optprob.hess(x0) == H1
     @test optprob.cons(x0) == [0.0, 0.0]
     @test optprob.cons_j([5.0, 3.0])≈[10.0 6.0; -0.149013 -0.958924] rtol=1e-6
-    @test_broken optprob.cons_h(x0) == [[2.0 0.0; 0.0 2.0], [-0.0 1.0; 1.0 0.0]]
+    @test optprob.cons_h(x0) == [[2.0 0.0; 0.0 2.0], [-0.0 1.0; 1.0 0.0]]
 
     cons = (x, p) -> [x[1]^2 + x[2]^2]
     optf = OptimizationFunction{false}(rosenbrock,

test/cvxtest.jl

@@ -1,11 +1,12 @@
-using Optimization, OptimizationBase, ForwardDiff, SymbolicAnalysis, LinearAlgebra, Manifolds, OptimizationManopt
+using Optimization, OptimizationBase, ForwardDiff, SymbolicAnalysis, LinearAlgebra,
+      Manifolds, OptimizationManopt
 
 function f(x, p = nothing)
     return exp(x[1]) + x[1]^2
 end
 
 optf = OptimizationFunction(f, Optimization.AutoForwardDiff())
-prob = OptimizationProblem(optf, [0.4])
+prob = OptimizationProblem(optf, [0.4], structural_analysis = true)
 
 @time sol = solve(prob, Optimization.LBFGS(), maxiters = 1000)
 
@@ -14,28 +15,29 @@ rosenbrock(x, p = nothing) = (1 - x[1])^2 + 100 * (x[2] - x[1]^2)^2
 l1 = rosenbrock(x0)
 
 optf = OptimizationFunction(rosenbrock, AutoEnzyme())
-prob = OptimizationProblem(optf, x0)
+prob = OptimizationProblem(optf, x0, structural_analysis = true)
 @time res = solve(prob, Optimization.LBFGS(), maxiters = 100)
 
 function con2_c(res, x, p)
-    res .= [x[1]^2 + x[2]^2, (x[2] * sin(x[1]) + x[1])-5]
+    res .= [x[1]^2 + x[2]^2, (x[2] * sin(x[1]) + x[1]) - 5]
 end
 
 optf = OptimizationFunction(rosenbrock, AutoZygote(), cons = con2_c)
-prob = OptimizationProblem(optf, x0, lcons = [1.0, -Inf], ucons = [1.0, 0.0], lb = [-1.0, -1.0], ub = [1.0, 1.0])
+prob = OptimizationProblem(optf, x0, lcons = [1.0, -Inf], ucons = [1.0, 0.0],
+    lb = [-1.0, -1.0], ub = [1.0, 1.0], structural_analysis = true)
 @time res = solve(prob, Optimization.LBFGS(), maxiters = 100)
 
 m = 100
 σ = 0.005
 q = Matrix{Float64}(LinearAlgebra.I(5)) .+ 2.0
 
 M = SymmetricPositiveDefinite(5)
-data2 = [exp(M, q, σ * rand(M; vector_at=q)) for i in 1:m];
+data2 = [exp(M, q, σ * rand(M; vector_at = q)) for i in 1:m];
 
 f(x, p = nothing) = sum(SymbolicAnalysis.distance(M, data2[i], x)^2 for i in 1:5)
-optf = OptimizationFunction(f, Optimization.AutoZygote())
-prob = OptimizationProblem(optf, data2[1]; manifold = M)
+optf = OptimizationFunction(f, Optimization.AutoForwardDiff())
+prob = OptimizationProblem(optf, data2[1]; manifold = M, structural_analysis = true)
 
 opt = OptimizationManopt.GradientDescentOptimizer()
-@time sol = solve(prob, opt, maxiters = 100)
-@test sol.minimizer < 1e-1
+@time sol = solve(prob, Optimization.LBFGS(), maxiters = 100)
+@test sol.minimizer < 1e-1