Originally posted by: NazmiSener

I am new at CPLEX-Python integration. I have an error from Cplex-Python integration which is written in the title. My code is given below. The source of error is last constraint set. I am open to any advice. All of my files are given in the attachment. Mathematical formulation of last constraints is given below (Please ignore omegas).

Another question is how can I acquire CPU solution times easily?

Thanks in advance.

You get this error because you are invoking Cplex.linear_constraints.add() with a SparsePair() in lin_expr that has duplicate constraints in its ind[] array. You will have to debug your code to figure out why you are adding more than one term for the same variable in a constraint.

I took a quick look at your code and this looks suspicious to me:

        for k in range(noph):
            thevarxs=[z[i][k]]
            thecoefxs=[O[i]]
            for l in [q for q in xrange(noph) if q!=k]:
                thevarxs.append(y[i][k][l])  # (*)
                thecoefxs.append(-1)
                thevarxs.append(y[i][l][k]) # (**)
                thecoefxs.append(1)

Consider two indices a0 and a1 in xrange(noph). As far as I can tell, for k=a0, l=a1 you will add terms for y[i][a0][a1] (by row (*)) and y[i][a1][a0] (by row (**)). But in your iteration you will also encounter k=a1 and l=a0. Then you will add two more terms, one for y[i][a1][a0] (by row (*)) and one for y[i][a0][a1] (by row (**)). As you can see, two terms for each of y[i][a0][a1] and y[i][a1][a0]. This is probably not what you intended to do?

Also, the loop

            for j in range(noph):
                thevarxs.append(z[j][k])
                thecoefxs.append(-dem[k][j])

will add a term for z[i][k] which you already added a few lines above that.

Originally posted by: NazmiSener

I understood the problem. However If you look my first post, I must add same decision variables in same constraint with different indices. How can I reflect the constraint to model?

Thank you very much,

I appreciate your helps,

You can explicitly normalize an ind/val pair of lists in which ind potentially contains duplicate indices. Like so (I am not a Python wizard, so there may be better ways to do that):

def normalize(ind, val):
    uniq = dict()
    for i, v in zip(ind, val):
        if i in uniq:
            uniq[i] += v
        else:
            uniq[i] = v
    return zip(*[(i, uniq[i]) for i in uniq])