generate_standard_repn misidentifies nonlinear expr

[natalieisenberg]

Summary

The generate_standard_repn function is identifying a nonlinear expr as a linear coefficient in the case that the expression has mutable Param objects initialized to 0 (and with compute_values=False).

If the Params are instead initialized to a non-zero value, then generate_standard_repn works as expected.

Here is an example to reproduce the behavior:

Steps to reproduce the issue

from pyomo.environ import *
from pyomo.repn.standard_repn import generate_standard_repn

m = ConcreteModel()
m.x_0 = Var(initialize=0)
m.p_0 = Param(initialize=0, mutable=True)
m.p_1 = Param(initialize=0, mutable=True)

m.con = Constraint(expr=m.x_0*((m.p_0 + m.x_0*m.p_1)**3) == 0)
val = generate_standard_repn(m.con.body, compute_values=False)

The output is:
val.linear_coefs[0] = (p_0 + p_1*x_0)**3
val.linear_vars[0] = x_0

Whereas I think this should still be considered a nonlinear_expr, even with the 0-value Param coefficients, since the mutable Param values are subject to change.

Information on your system

Pyomo version: 6.0.1
Python version: 3.8
Operating system: CentOS 7
How Pyomo was installed (PyPI, conda, source): source
Solver (if applicable):

[PierreAttard]

Hi,

The code has been designed to have this behavior (in file test_numeric_expr.py):

 def test_simple_product(self):

        # ...
        #
        # Product of unfixed variable and zero parameter: fixed, not constant, pvar
        #
        expr = self.instance.f * self.instance.b
        self.assertEqual(expr.is_fixed(), True)

So, it depends on what the pyomo team wants.

If the expected beahvior is like @natalieisenberg said, I think adding and not self._args_[i].mutable in the method ProductExpression._is_fixed should be enough :

if args[i] and value(self._args_[i]) == 0 and not self._args_[i].mutable:
    return True