Originally posted by: Hans Kirchner

Hi, I'm currently trying out Python+Docplex and when I'm executing the following code it's gives me "status = integer infeasible".

But I think c must be 0 and this is clearly feasible, right?

from docplex.mp.model import Model

m = Model("test")

a = m.binary_var(name="a")
b = m.binary_var(name="b")
c = m.binary_var(name="c")

m.add_constraint(m.logical_and(a, b) == c)
m.add_constraint(a == 1)
m.add_constraint(b == 0)

if m.solve():
    m.print_solution(print_zeros=True)
else:
    print(m.solve_details)

Version Info:

- Python 3.6.8

- docplex 2.10.154

- CPLEX Studio 12.9 (academic)

I'm afraid you have indeed hit on a bug in docplex (I have filed a bug report). You can do m.export_as_lp('and.lp') and then look at the generated LP file. You will see that the constraints generated are incorrect.

You can use this function as replacement:

def logical_and(mdl, bins):
    '''Create an expression that is 1 if and only if all
    binaries in `bins` are 1.'''
    if len(bins) == 0:
        return 1
    y = mdl.binary_var()
    # If y=1 then all x must be 1
    mdl.add_constraints(y <= x for x in bins)
    # If all x are 1 then y must be 1
    mdl.add_constraint(mdl.sum(bins) - y <= len(bins) - 1)

Originally posted by: Hans Kirchner

Thanks for the workaround and for filing a bug report!

However, it leaves me with a weird feeling that such a "fundamental" function is not working.

Originally posted by: Hans Kirchner