Originally posted by: BahmanBornay

Hey everybody,

I am new into writing Optimization codes using docplex library.

I happened to have an issue in defining the objective and constraints. Considering the simple LP example from documentation (link: https://ibmdecisionoptimization.github.io/tutorials/html/Linear_Programming.html), if we write it this way:

import numpy as np
from docplex.mp.model import Model

mdl = Model("Tekephon_Production")

n = 2
N = [1, n]
p = [12, 20]
a = np.array([(0.2, 0.4), (0.5, 0.4)])
b = np.array([400, 490])
lb = np.array([100, 100])

x = mdl.continuous_var_dict(N, name='x')
mdl.maximize(12 * x[1] + 20 * x[2])

mdl.add_constraint(0.2 * x[1] + 0.4 * x[2] <= 400)

mdl.add_constraint(0.5 * x[1] + 0.4 * x[2] <= 490)

mdl.add_constraint(x[1] >= 100)
mdl.add_constraint(x[2] >= 100)

solution = mdl.solve(log_output=True)
print(solution)

this block of code is solved as expected. But, as we run the following code, errors are popping up:

x = mdl.continuous_var_dict(N, name='x')

mdl.maximize(mdl.sum(p[i] * x[i] for i in N))

mdl.add_constraint(mdl.sum(p[i] * x[i]) <= 400 for i in N)

mdl.add_constraint(mdl.sum(a[i] * x[i]) <= 490 for i in N)

mdl.add_constraint(x[1] >= 100)
mdl.add_constraint(x[2] >= 100)

solution = mdl.solve(log_output=True)
print(solution)

I would appreciate your comments. 

Regards,
Bahman

It would a lot if you told us what errors are popping up and if it is syntax errors then from which statement they come.

This looks suspicious

mdl.maximize(mdl.sum(p[i] * x[i]) for i in N)

I guess what you intended to was instead (note the place of closing parens)

mdl.maximize(mdl.sum(p[i] * x[i] for i in N))

Originally posted by: BahmanBornay

Hey Daniel,

I appreciate your prompt reply. However, it doesn't seem to be due to parenthesis since I got these errors

Regards,
Bahman

The problem seems unrelated to docplex but looks a basic Python problem. Consider this

n = 2
N = [1, n]
p = [12, 20]
...
x = mdl.continuous_var_dict(N, name='x')
mdl.maximize(mdl.sum(p[i] * x[i] for i in N))

The lists p and x both have size 2, that is, valid indices are 0 and 1. Now in your code you reference p[i] and x[i] and you have 'for i in N'. Since i ranges through N it will be 1 and 2. The first value is a valid index, the second is not. I guess what you want is something like 'for i in range(len(N))'.

Originally posted by: BahmanBornay

Hey Daniel,

I found the issue: the range(len(N)) does not work, since p is indexed from 0 to 1, but x is indexed from 1 to 2. Thus: the following code worked:

mdl.maximize(mdl.sum(p[i] * x[i+1] for i in range(len(N))))

Regards,
Bahman

Ah, yes, you are right. x is a dictionary rather than a list and the keys are the values from N.

A more robust way to write your objective would be this

mdl.maximize(mdl.sum(p[i] * x[n] for i, n in enumerate(N))

This does not depend on a particular relationship (like 'i' and 'i+1') between the values in N and their indices. It only assumes that the k-th variable is given by x[N[k]].