Add constraint/objective construction from NumPy arrays

[arcondello]

Requested by @mhramani #1307 (review).

Something like

indices = [0, 1, 5]
biases = [1, 2, 3]
cqm.add_constraint((indices, biases), "==", 5)

which would be interpreted as a linear constraint $x_0 + 2x_1 + 3x_5 = 5$.

[ACE07-Sev]

Like so?

from dimod import ConstrainedQuadraticModel, Binary, quicksum

cqm_test = ConstrainedQuadraticModel()
vars = [Binary(f"X{i}") for i in range(7)]

def add_constraint(cqm, vars, indices, biases, sense, rhs):
    if len(indices) != len(biases):
        raise ValueError('The indices and biases do not have same size.')
    constraint = quicksum(biases[i] * vars[indices[i]] for i in range(len(indices)))
    cqm.add_constraint(constraint, sense=sense, rhs=rhs, label='c')

add_constraint(cqm_test, vars, [0, 2, 5], [1, 2, 3], "==", 5)

print(cqm_test.constraints['c'].to_polystring())
>> X0 + 2*X2 + 3*X5 == 5.0

[arcondello]

Yes, something like that. Though the performance benefits of such a thing would come from implementing it at the Cython level, likely as a sibling method to cyConstrainedQuadraticModel.add_constraint_from_iterable() and cyConstrainedQuadraticModel.add_constraint_from_model().

[ACE07-Sev]

Understood, I'll try to make a PR on this soon.