Copying a Model and adding new constraints generates an error when calling solve(m)

[h3nnn4n]

Here I have a small snippet of code that reproduces an error on a bigger piece of code.

The code is:

using JuMP

## Simple LP model
m = Model()

@defVar(m, 0 <= x <= 2 )
@defVar(m, 0 <= y <= 30 )

@setObjective(m, Max, 5x + 3*y )
@addConstraint(m, 1x + 5y <= 3.0 )

## The model is copied
n = copy(m)

## New constraint using the same variables is added to the copy
@addConstraint(n, 2x + 3y <= 3.0 )

solve(m)

## Error Here
solve(n)

and the error:

h3nnn4n@clusterfuck ~/julia_bnb (git)-[master] % julia jump_copy_error.jl
ERROR: Variable not owned by model present in constraints
 in prepConstrMatrix at /home/h3nnn4n/.julia/v0.3/JuMP/src/solvers.jl:163
 in solveLP at /home/h3nnn4n/.julia/v0.3/JuMP/src/solvers.jl:221
 in solve at /home/h3nnn4n/.julia/v0.3/JuMP/src/solvers.jl:39
 in include at ./boot.jl:245
 in include_from_node1 at loading.jl:128
 in process_options at ./client.jl:285
 in _start at ./client.jl:354
while loading /home/h3nnn4n/julia_bnb/bug.jl, in expression starting on line 21

The error happens on both Julia 0.3 and 0.4 using the following packages:

julia> Pkg.status()
6 required packages:
 - GLPK                          0.2.16
 - GLPKMathProgInterface         0.1.13
 - JuMP                          0.9.2

[IainNZ]

The error message is correct, the variables belong to the initial model, not the copy.
I guess what you want is a way to get a handle on the equivalent variables for the copy?

[h3nnn4n]

Yes.

In the code I am working on I have

type instance
    m :: Model
    weight :: Array{Int}
    profit :: Array{Int}
    size :: Int
    capacity :: Int
    x
end

When I copy the whole thing the model m keeps the reference to the original instance.x. Here is the whole code in case you want to take a closer look.

[h3nnn4n]

I think that #399 is what I want. I just pulled the last version, lets see if it works.

[h3nnn4n]

Here n is a copy of m. n has an empty varDict though. Looks like varDict wasn't copied to the new model.

julia> n = copy(m)
Maximization problem with:
 * 1 linear constraint
 * 2 variables
Solver set to Default

julia> print(m.varDict)
[:x=>x,:y=>y]
julia> print(n.varDict)
Dict{Symbol,Any}()

julia> getVar(m, :y)
y

julia> getVar(n, :y)
ERROR: No variable with name y
 in error at error.jl:21
 in getVar at /home/h3nnn4n/.julia/v0.3/JuMP/src/JuMP.jl:805

[IainNZ]

Thats true. We'll need to implement code to take an existing variable and mapping it to the new model. I've actually had to do something similar for a JuMP extension, funny enough.

[joehuchette]

@h3nnn4n try pulling the latest master and see if it helps.

[h3nnn4n]

It worked. Thank you for the quick responses and all the support.

[FransdR]

Is it still possible to copy a model like this? I get the same error message as @h3nnn4n if I try this (also on his snippet of code).

ERROR: Variable not owned by model present in a constraint
 in prepConstrMatrix at /Users/FransdeRuiter/.julia/v0.4/JuMP/src/solvers.jl:563
 in build at /Users/FransdeRuiter/.julia/v0.4/JuMP/src/solvers.jl:372
 in solve at /Users/FransdeRuiter/.julia/v0.4/JuMP/src/solvers.jl:134

(Julia 0.4.6, JuMP 0.14)

[mlubin]

You can copy a model, but you can't mix variables from different models. Use x_new = copy(x,m_new) to get a reference to x in the new model.

[FransdR]

Thanks Miles, that works! It makes sense as well of course, because you can refer to the variables without mentioning the model (with getvalue(x) for instance).

[ahmadreza-marandi]

I am trying to do the same in JuMP 0.18.6 and I got exactly the same error:

`
P_k=ones(10,1);

extra_cost=20*ones(10,1);

LP=Model(solver=CplexSolver());

@variable(LP,p[1:10]<=20);

@constraint(LP, sum(p) <= 100);

@objective(LP,Min,sum(p));

m=copy(LP);

p_new=copy(p,LP)

@constraint(m, 2*sum(p_new) <= 100);

solve(m)`

and this is the error:
ERROR: VariableNotOwnedError: Variable not owned by model present in constraint ...

When I checked the models details, it seemed that I had the right variables in the copied model:

julia> LP.varData
IdDict{Any,Any} with 1 entry:
p[i] <= 20 for all i in {1,2,..,9,10} => JuMPContainerData(:p, ([1, 2, 3, 4, 5, 6, 7, 8, 9, 10],), JuMP.IndexPair[IndexPair(nothing, :(1:10))], :(()))
julia> m.varData
IdDict{Any,Any} with 1 entry:
p[i] <= 20 for all i in {1,2,..,9,10} => JuMPContainerData(:p, ([1, 2, 3, 4, 5, 6, 7, 8, 9, 10],), JuMP.IndexPair[IndexPair(nothing, :(1:10))], :(()))

But when I checked the constraints in the models, the old constraint in the copied model didn't look to be right:

julia> LP.linconstr
1-element Array{JuMP.GenericRangeConstraint{JuMP.GenericAffExpr{Float64,Variable}},1}:
p[1] + p[2] + p[3] + p[4] + p[5] + p[6] + p[7] + p[8] + p[9] + p[10] <= 100

julia> m.linconstr
2-element Array{JuMP.GenericRangeConstraint{JuMP.GenericAffExpr{Float64,Variable}},1}:
col_1 + col_2 + col_3 + col_4 + col_5 + col_6 + col_7 + col_8 + col_9 + col_10 <= 100
2 p[1] + 2 p[2] + 2 p[3] + 2 p[4] + 2 p[5] + 2 p[6] + 2 p[7] + 2 p[8] + 2 p[9] + 2 p[10] <= 100

So, can someone tell me what can be wrong here?
Thanks

[odow]

You probably need to use p_new=copy(p, m) instead of p_new=copy(p,LP).

p.s. In future, please post questions like this on the Discourse forum: https://discourse.julialang.org/c/domain/opt.

[ahmadreza-marandi]

@odow Thanks!