load_variables not defined for SCS

[matbesancon]

MOIU.load_variables(model.optimizer.model.optimizer, MOI.get(model, MOI.NumberOfVariables()))

in backward_conic! throws an error, it does not seem to be implemented for SCS models

[matbesancon]

Related in the tests:

Differentiating conic with PSD and SOC constraints: Error During Test at /home/mbesancon/Documents/crptests/project_julia/DiffOpt.jl/test/MOI_wrapper.jl:624
  Got exception outside of a @test
  The Allocate-Load interface is not implemented by the model
  Stacktrace:
   [1] correct_throw_add_constraint_error_fallback(::SCS.Optimizer, ::MathOptInterface.VectorAffineFunction{Float64}, ::MathOptInterface.PositiveSemidefiniteConeTriangle; error_if_supported::ErrorException) at /home/mbesancon/.julia/packages/MathOptInterface/ZJFKw/src/constraints.jl:156
   [2] throw_add_constraint_error_fallback(::SCS.Optimizer, ::MathOptInterface.VectorAffineFunction{Float64}, ::MathOptInterface.PositiveSemidefiniteConeTriangle; kwargs::Base.Iterators.Pairs{Symbol,ErrorException,Tuple{Symbol},NamedTuple{(:error_if_supported,),Tuple{ErrorException}}}) at /home/mbesancon/.julia/packages/MathOptInterface/ZJFKw/src/constraints.jl:135
   [3] load_constraint(::SCS.Optimizer, ::MathOptInterface.ConstraintIndex{MathOptInterface.VectorAffineFunction{Float64},MathOptInterface.PositiveSemidefiniteConeTriangle}, ::MathOptInterface.VectorAffineFunction{Float64}, ::MathOptInterface.PositiveSemidefiniteConeTriangle) at /home/mbesancon/.julia/packages/MathOptInterface/ZJFKw/src/Utilities/copy.jl:689
   [4] backward_conic!(::Optimizer{MathOptInterface.Bridges.LazyBridgeOptimizer{MathOptInterface.Utilities.CachingOptimizer{SCS.Optimizer,MathOptInterface.Utilities.UniversalFallback{MathOptInterface.Utilities.Model{Float64}}}}}, ::Array{Float64,2}, ::Array{Float64,1}, ::Array{Float64,1}) at /home/mbesancon/Documents/crptests/project_julia/DiffOpt.jl/src/MOI_wrapper.jl:563
   [5] top-level scope at /home/mbesancon/Documents/crptests/project_julia/DiffOpt.jl/test/MOI_wrapper.jl:688
   [6] top-level scope at /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.5/Test/src/Test.jl:1115
   [7] top-level scope at /home/mbesancon/Documents/crptests/project_julia/DiffOpt.jl/test/MOI_wrapper.jl:628
   [8] include(::String) at ./client.jl:457
   [9] top-level scope at /home/mbesancon/Documents/crptests/project_julia/DiffOpt.jl/test/runtests.jl:25
   [10] top-level scope at /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.5/Test/src/Test.jl:1115
   [11] top-level scope at /home/mbesancon/Documents/crptests/project_julia/DiffOpt.jl/test/runtests.jl:25

[matbesancon]

ping @blegat and @Aks1996 on this one, not sure what load_variables is meant to do here

[blegat]

It used to be defined for SCS but not anymore. Anyway, DiffOpt should not require the solver to implement it. It should use add_variables and solvers like SCS should be used with a CachingOptimizer

[matbesancon]

so we should just remove the line or should I assume it was doing something?

[matbesancon]

I wanted to replace the following:

-            MOIU.load_constraint(model.optimizer.model.optimizer, CI{F, S}(CONES_OFFSET[S]), func, set)
+            MOI.add_constraint(model.optimizer.model.optimizer, func, set)

but it seems some constraint types are not supported