Missing attribute `ConstraintDual` for `ZerosBridge`

[guilhermebodin]

This code gives the following error

using JuMP
using ECOS

model = Model(() -> ECOS.Optimizer())
@variable(model, x[1:2])
@variable(model, t[1:2])
@variable(model, p == 1)
@constraint(model, con1, [t[1], x[1] + p, x[1]] in SecondOrderCone())
@constraint(model, con2, [t[2], 2x[2] + p, x[2]] in SecondOrderCone())
@objective(model, Min, sum(t))
optimize!(model)
dual(FixRef(p))

julia> dual(FixRef(p))
ERROR: ArgumentError: Bridge of type `MathOptInterface.Bridges.Variable.ZerosBridge{Float64}` does not support 
accessing the attribute `MathOptInterface.ConstraintDual(1)`. If you encountered this error unexpectedly, it probably means your model has been reformulated using the bridge, and you are attempting to query an attribute that we haven't implemented yet for this bridge. Please open an issue at https://github.com/jump-dev/MathOptInterface.jl/issues/new and provide a reproducible example explaining what you were trying to do.

I was testing querying duals of parameters in some conic sets.

[odow]

The problem is p == 1. See the discussion here: #1767

We can revisit if we get anyone encountering this issue in the wild.

I guess this now applies.

[odow]

There's another work-around that would require adding methods to JuMP:

julia> using JuMP

julia> using ECOS

julia> remove_bridge(model::Model, B) = remove_bridge(backend(model), B)
remove_bridge (generic function with 3 methods)

julia> remove_bridge(model::MOI.Utilities.CachingOptimizer, B) = remove_bridge(model.optimizer, B)
remove_bridge (generic function with 3 methods)

julia> remove_bridge(model::MOI.Bridges.AbstractBridgeOptimizer, B) = MOI.Bridges.remove_bridge(model, B)
remove_bridge (generic function with 3 methods)

julia> model = Model(ECOS.Optimizer)
A JuMP Model
Feasibility problem with:
Variables: 0
Model mode: AUTOMATIC
CachingOptimizer state: EMPTY_OPTIMIZER
Solver name: ECOS

julia> remove_bridge(model, MOI.Bridges.Variable.ZerosBridge{Float64})

julia> @variable(model, x[1:2])
2-element Vector{VariableRef}:
 x[1]
 x[2]

julia> @variable(model, t[1:2])
2-element Vector{VariableRef}:
 t[1]
 t[2]

julia> @variable(model, p == 1)
p

julia> @constraint(model, con1, [t[1], x[1] + p, x[1]] in SecondOrderCone())
con1 : [t[1], x[1] + p, x[1]] ∈ MathOptInterface.SecondOrderCone(3)

julia> @constraint(model, con2, [t[2], 2x[2] + p, x[2]] in SecondOrderCone())
con2 : [t[2], 2 x[2] + p, x[2]] ∈ MathOptInterface.SecondOrderCone(3)

julia> @objective(model, Min, sum(t))
t[1] + t[2]

julia> optimize!(model)

ECOS 2.0.8 - (C) embotech GmbH, Zurich Switzerland, 2012-15. Web: www.embotech.com/ECOS

It     pcost       dcost      gap   pres   dres    k/t    mu     step   sigma     IR    |   BT
 0  +0.000e+00  -0.000e+00  +8e+00  5e-01  1e-03  1e+00  3e+00    ---    ---    1  1  - |  -  - 
 1  +3.122e-01  +3.189e-01  +1e+00  9e-02  2e-04  1e-01  4e-01  0.8656  4e-03   2  2  2 |  0  0
 2  +1.126e+00  +1.136e+00  +7e-02  7e-03  9e-06  2e-02  3e-02  0.9623  4e-02   2  2  2 |  0  0
 3  +1.154e+00  +1.154e+00  +8e-04  8e-05  1e-07  2e-04  3e-04  0.9890  1e-04   1  1  1 |  0  0
 4  +1.154e+00  +1.154e+00  +8e-06  8e-07  1e-09  2e-06  4e-06  0.9890  1e-04   1  1  1 |  0  0
 5  +1.154e+00  +1.154e+00  +9e-08  9e-09  1e-11  3e-08  4e-08  0.9890  1e-04   2  1  1 |  0  0
 6  +1.154e+00  +1.154e+00  +1e-09  1e-10  1e-13  3e-10  4e-10  0.9890  1e-04   1  1  1 |  0  0

OPTIMAL (within feastol=1.0e-10, reltol=8.9e-10, abstol=1.0e-09).
Runtime: 0.000114 seconds.


julia> dual(FixRef(p))
1.1543203763939147

But that requires the user to have some pretty deep understanding of bridges, and they could do it by calling on backend(model).optimizer, so I'm hesitant to add that to JuMP.

@guilhermebodin does the error improve things enough to close this issue?

[guilhermebodin]

Hi @odow, in my opinion, the new error message is instructive enough to close the issue. 👍🏻