Sets in PATH for equations as JuMPDict

[hhoeschle]

Hallo,

here is a minor thing I saw when run my model which could be fixed with the next version:

• The name of the indexes are not forwarded to the PATH solver if the equation is of type JuMPDict

Here is a MWE:

using Complementarity
model = MCPModel()
scenarios = [@sprintf "s%03d" s for s=1:3]
INDEX = Dict("s001" => 1., "s002" => 0., "s003" => 0.5)
COST = Dict("s001" => 10., "s002" => 2., "s003" => 5)
COST_INDEX = Dict("s001" => 2., "s002" => 1., "s003" => 3.)
@variable(model, xname[s in scenarios] >= 0) 
@variable(model, ynoname[s in scenarios; INDEX[s] >0] >= 0)
@show typeof(xname)
@show typeof(ynoname)
@variable(model, lambda1 >= 0)
@variable(model, lambda2 >= 0)
@NLexpression(model, lambda,
    lambda1 - lambda2)

@mapping(model, market_lambda1,
    + sum(xname[s] for s in scenarios)
    + sum(ynoname[s] for s in scenarios if INDEX[s] > 0)
    - 10)
@complementarity(model, market_lambda1, lambda1)
                
@mapping(model, market_lambda2,
    - sum(xname[s] for s in scenarios)
    - sum(ynoname[s] for s in scenarios if INDEX[s] > 0)
    + 10)
@complementarity(model, market_lambda2, lambda2)
@mapping(model, KKT_xname[s in scenarios],
    - lambda
    + COST[s])
@complementarity(model, KKT_xname, xname)
@show typeof(KKT_xname)
@mapping(model, KKT_ynoname[s in scenarios; INDEX[s] >0],
    - lambda
    + COST_INDEX[s])
# @complementarity(model, KKT_ynoname, ynoname)
for s in scenarios
    if INDEX[s] > 0
        @complementarity(model, KKT_ynoname[s], ynoname[s])
    end
end
@show typeof(KKT_ynoname)

PATHSolver.options(
    convergence_tolerance=1e-6, 
    output=:yes, 
    time_limit=3600*12, 
    # time_limit=0.1, 
    minor_iteration_limit=1000,
    cumulative_iteration_limit=1000000000)

stat = solveMCP(model)
# Inf-Norm of Complementarity . .  1.1410e-006 eqn: (KKT_ynoname[s]) <---- NAME of index set not included
# Inf-Norm of Normal Map. . . . .  2.2820e-007 eqn: (KKT_xname[s002])
# Inf-Norm of Minimum Map . . . .  2.2820e-007 eqn: (KKT_xname[s002])
# Inf-Norm of Fischer Function. .  2.2820e-007 eqn: (KKT_xname[s002])
# Inf-Norm of Grad Fischer Fcn. .  4.5641e-007 eqn: (market_lambda2)

[chkwon]

Thanks for reporting. I'll work on it.