Adding constraints to nodes in PySCIPOpt

Question

I'm using the Python Interface of SCIP (PySCIPOpt) to write a custom branching rule. As far as I understood from reading the documentation, a good way to do this is making use of the methods self.model.createChild and self.model.addConsNode. So I came up with the following test code to try to implement a random branching rule on my own:

class RandomBranching(Branchrule):
def __init__(self, *args, **kwargs):
    super().__init__(*args, **kwargs)
    self.count = 0

def branchexeclp(self, allowaddcons):
    self.count += 1
    variables = [v for v in self.model.getVars() if (v.vtype() == "BINARY" or v.vtype() == "INTEGER")]
    random.shuffle(variables)

    for v in variables:
        value = self.model.getSolVal(None, v)
        if value - math.floor(value) >= 10 ** -6:
            child_1 = self.model.createChild(1, 1)
            self.model.addConsNode(child_1, v <= math.floor(value))
            child_2 = self.model.createChild(1, 1)
            self.model.addConsNode(child_2, v >= math.floor(value) + 1)
            return {"result": SCIP_RESULT.BRANCHED}
        
    return {"result": SCIP_RESULT.DIDNOTRUN}

This code raises an exception, because addConsNode expects an object of type "Constraint" while I seem to be passing an "ExprCons". After trying my best to dig through the source code, I do not manage to understand the difference between the two and unfortunately have no clue how to correctly initialize an object of type "Constraint".

Note: My code from above is quite similar to the Code used in the PySCIPOpt github to test the Branchingrules (in PySCIPOpt/tests/test_branch_probing_lp.py).

Answer 1

The PySCIPOpt-Interface was lacking the functionality I was searching for. This lead to a pull-request, see here. The pr will extend the functionality and allow to use ExprCons for addConsNode, which is needed to branch this way.