Update to MOI v0.10 and JuMP v0.22

src/diff_opt.jl

@@ -799,8 +799,8 @@ function _assign_mapped!(x, y, r, k, ::Flattened)
 end
 
 # Map the rows corresponding to `F`-in-`S` constraints and store it in `x`.
-function _map_rows!(f::Function, x::Vector, model, conic_form::MatOI.GeometricConicForm, index_map::MOI.IndexMap, map_mode, k)
-    for ci in MOI.get(model, MOI.ListOfConstraintIndices{F, S}()) # TODO how to get F-S now?
+function _map_rows!(f::Function, x::Vector, model, conic_form::MatOI.GeometricConicForm, index_map::MOIU.DoubleDicts.IndexDoubleDictInner{F, S}, map_mode, k) where {F, S}
+    for ci in MOI.get(model, MOI.ListOfConstraintIndices{F, S}())
         r = MatOI.rows(conic_form, index_map[ci])
         k += 1
         _assign_mapped!(x, f(ci, r), r, k, map_mode)
@@ -828,7 +828,7 @@ function map_rows(f::Function, model, conic_form::MatOI.GeometricConicForm, inde
     k = 0
     for (F, S) in MOI.get(model, MOI.ListOfConstraintTypesPresent())
         # Function barrier for type unstability of `F` and `S`
-        # `conmap` is a `MOIU.DoubleDicts.MainIndexDoubleDict`, we index it at `F, S`
+        # `con_map` is a `MOIU.DoubleDicts.MainIndexDoubleDict`, we index it at `F, S`
         # which returns a `MOIU.DoubleDicts.IndexWithType{F, S}` which is type stable.
         # If we have a small number of different constraint types and many
         # constraint of each type, this mostly removes type unstabilities
@@ -920,14 +920,14 @@ function _fill(neg::Function, model::Optimizer, conic_form, args...)
     _fill(S -> true, neg, model, conic_form, args...)
 end
 function _fill(filter::Function, neg::Function, model::Optimizer, conic_form, args...)
-    conmap = model.gradient_cache.index_map.con_map
-    varmap = model.gradient_cache.index_map.var_map
+    con_map = model.gradient_cache.index_map.con_map
+    var_map = model.gradient_cache.index_map.var_map
     for (F, S) in MOI.get(model, MOI.ListOfConstraintTypesPresent())
         filter(S) || continue
         if F == MOI.ScalarAffineFunction{Float64}
-            _fill(args..., neg(S), conic_form, conmap[F,S], varmap, model.input_cache.scalar_constraints[F,S])
+            _fill(args..., neg(S), conic_form, con_map[F,S], var_map, model.input_cache.scalar_constraints[F,S])
         elseif F == MOI.VectorAffineFunction{Float64}
-            _fill(args..., neg(S), conic_form, conmap[F,S], varmap, model.input_cache.vector_constraints[F,S])
+            _fill(args..., neg(S), conic_form, con_map[F,S], var_map, model.input_cache.vector_constraints[F,S])
         end
     end
     return

src/moi_wrapper.jl

@@ -135,6 +135,7 @@ function MOI.copy_to(model::Optimizer, src::MOI.ModelLike)
 end
 
 function MOI.get(model::Optimizer, ::MOI.TerminationStatus)
+    @show MOI.get(model.optimizer, MOI.TerminationStatus())
     return MOI.get(model.optimizer, MOI.TerminationStatus())
 end
 
@@ -192,8 +193,7 @@ end
 
 # helper methods to check if a constraint contains a Variable
 function _constraint_contains(model::Optimizer, v::VI, ci::CI{MOI.VariableIndex, S}) where {S <: SUPPORTED_SCALAR_SETS}
-    func = MOI.get(model, MOI.ConstraintFunction(), ci)
-    return v == func.variable
+    return v == MOI.get(model, MOI.ConstraintFunction(), ci)
 end
 
 function _constraint_contains(model::Optimizer, v::VI, ci::CI{MOI.ScalarAffineFunction{Float64}, S}) where {S <: SUPPORTED_SCALAR_SETS}

src/quadratic_diff.jl

@@ -240,7 +240,7 @@ function get_problem_data(model::MOI.AbstractOptimizer)
         con = ge_con_sv_idx[i]
         func = MOI.get(model, MOI.ConstraintFunction(), con)
         set = MOI.get(model, MOI.ConstraintSet(), con)
-        vidx = findfirst(v -> v == func, var_list)
+        vidx = findfirst(isequal(func), var_list)
         G[i+nineq_le+nineq_ge+nineq_sv_le,vidx] = -1
         h[i+nineq_le+nineq_ge+nineq_sv_le] = -MOI.constant(set)
         ineq_cont += 1
@@ -276,7 +276,7 @@ function get_problem_data(model::MOI.AbstractOptimizer)
 
         for x in func.terms
             # never nothing, variable is present
-            vidx = findfirst(v -> v == x.variable, var_list)
+            vidx = findfirst(isequal(x.variable), var_list)
             A[i, vidx] = x.coefficient
         end
         b[i] = set.value - func.constant
@@ -289,7 +289,7 @@ function get_problem_data(model::MOI.AbstractOptimizer)
         con = eq_con_sv_idx[i]
         func = MOI.get(model, MOI.ConstraintFunction(), con)
         set = MOI.get(model, MOI.ConstraintSet(), con)
-        vidx = findfirst(v -> v == func, var_list)
+        vidx = findfirst(isequal(func), var_list)
         A[i+neq,vidx] = 1
         b[i+neq] = set.value
         eq_cont += 1

test/moi_wrapper.jl

@@ -633,7 +633,7 @@ function simple_psd(solver)
 
     # test2: changing X[1], X[3] but keeping the objective (their sum) same
     MOI.set(model, DiffOpt.ForwardInConstraint(), c, MOIU.zero_with_output_dimension(MOI.VectorAffineFunction{Float64}, 1))
-    MOI.set(model, DiffOpt.ForwardInObjective(), -1.0fX[1] + 1.0fX[3])
+    MOI.set(model, DiffOpt.ForwardInObjective(), -1.0X[1] + 1.0X[3])
 
     DiffOpt.forward(model)
 
@@ -897,7 +897,7 @@ end
     MOI.set(model,
         DiffOpt.ForwardInConstraint(), c1, MOIU.zero_with_output_dimension(VAF, 1))
     MOI.set(model,
-        DiffOpt.ForwardInConstraint(), c2, MOIU.vectorize(ones(6) .* fx[1:6]))
+        DiffOpt.ForwardInConstraint(), c2, MOIU.vectorize(ones(6) .* x[1:6]))
     MOI.set(model,
         DiffOpt.ForwardInConstraint(), c3, MOIU.zero_with_output_dimension(VAF, 3))
     MOI.set(model,

test/utils.jl

@@ -145,10 +145,10 @@ function qp_test(
     end
     @_test(convert(Vector{Float64}, _λ), λ)
 
-    #dobjb = fv' * (dQb / 2.0) * fv + dqb' * fv
+    #dobjb = v' * (dQb / 2.0) * v + dqb' * v
     # TODO, it should .-
-    #dleb = dGb * fv .+ dhb
-    #deqb = dAb * fv .+ dbb
+    #dleb = dGb * v .+ dhb
+    #deqb = dAb * v .+ dbb
     @assert dzb !== nothing
     @testset "Backward pass" begin
         MOI.set.(model, DiffOpt.BackwardInVariablePrimal(), v, dzb)
@@ -230,7 +230,7 @@ function qp_test(
                 func = deqf[length(ceq)+j]
                 canonicalize && MOI.Utilities.canonicalize!(func)
                 if set_zero || !MOI.iszero(func)
-                    # TODO FIXME should work if we drop support for SingleVariable and we let the Functionize bridge do the work
+                    # TODO FIXME should work if we drop support for `VariableIndex` and we let the Functionize bridge do the work
                     @test_throws MOI.UnsupportedAttribute MOI.set(model, DiffOpt.ForwardInConstraint(), jc, func)
                 end
             end