axis arg for sum method

[MxMartin]

Hi!

Thanks for this fantastic package first of all!
I just started evaluating it for a project where so far I used Gurobi.
The axis argument on the sum method was very useful for me there, e.g. in case where assignment[b,t,m] was the assignment of a batch to a certain machine for a certain task. So far I modelled the constraint to ensure that each batch was assigned to at most one machine per task as assignment.sum(axis=-1)<=1 or the capacity constraints on each machine as assignment.sum(axis=(0,1))<=capacity.

Hence I'd love to see the axis argument implemented for the sum method.

Until then: What is the recommended workaround? Am I overlooking anything?

Kind regards,

[IgnaceBleukx]

Hi Martin,

This is certainly on our wishlist to implement the axis argument for all operators (not only sum).
We started a shortlist for operators in issue #117, if you have any suggestions, please feel free to post them there.
A workaround you can use until then is something similar to this:

from cpmpy import *

assignment = intvar(0,10,shape=(3,4,5), name="assignment")
capacity = intvar(0,100, shape=5, name="capacity")

# each batch is assigned to at most one machine per task
model += cpm_array([assignment[i,j,:].sum() for i in range(3) for j in range(4)]) <= 1
# or equivalently:
for i in range(3):
    for j in range(4):
        model += assignment[i,j,:].sum() <= 1

# capacity constraints on each machine:
model += [assignment[:,:,m].sum() for m in range(5)] <= capacity

I hope this helps.

Kind regards,
Ignace

[MxMartin]

Awesome, thanks for the swift response!