How to use addVars() in GurobiPy using "if" conditional?

Answered

Vincent

• November 29, 2020 22:51

I have defined a variable using addVar() method as follows:

for i in range(no_nodes):
  for j in range(no_nodes):
    for k in O:
      if fij[i][j]==1 or mij[i][j]==1:
        x[i,j,k] = m.addVar(vtype=GRB.INTEGER, name="x[%s,%s,%s]" %(i,j,k))

where fij and mij are model's parameters and O is an unordered set containing more than 1000 elements.

It works properly, and there is no issue in solving the model. I just wonder how this variable definition can be revised by applying the addVars() method instead of addVar()?

Thanks.

0

• 

  Matthias Miltenberger

  Gurobi Staff

  • November 30, 2020 13:35
  • Edited

  Hi Vincent!

  You can do the following to use addVars() and reduce the code to a single line:

  x = m.addVars([(i,j,k) for i in no_nodes for j in no_nodes for k in O if fij[i][j] or mij[i][j]],
      vtype=GRB.INTEGER, name="x")

  The name property will automatically be indexed for every single variable and assuming that fij and mij are binaries, you can also omit the "==1" part.

  You may also use itertools.product:

  from itertools import product
  x = m.addVars([(i,j,k) for (i,j) in product(no_nodes,repeat=2) for k in O if fij[i][j] or mij[i][j]],
      vtype=GRB.INTEGER, name="x")

  Cheers,
  Matthias

  1
• 

  Vincent

  • November 30, 2020 18:53
  • Edited

  

  

  

  Hi Matthias,

  Thanks for your help. Is the itertools.product method faster than the first one?

  Thanks,
  Vincent

  0
• 

  Matthias Miltenberger

  Gurobi Staff

  • December 02, 2020 11:57

  Hi Vincent,

  Honestly, I wouldn't worry about the performance of either implementation. Most likely, adding the variables is more expensive than looping over your lists. I would go with the implementation that results in more readable, maintainable, and extendable code.

  Cheers,
  Matthias

  1

Please sign in to leave a comment.