I'm using a CP model with custom constraints, together with depth-first search (cp.setParameter(IloCP::SearchType, IloCP::DepthFirst);).
I also have a heuristic which calculates an initial feasible solution which I would like to use to warm start the search. However, the manual states the following:
Note: the starting point information is used by the restart and multi-point search types only. It is not used by the depth-first search.
This is problematic. My initial solution is completely ignored. Its objective value isn't used at all. Even though CP might not start from this solution, I would at least expect that it stores the solution in its solution pool and that it correctly reports that it knows a solution with objective value Y.
Example> Imagine that I'm solving a routing problem where the objective is to minimize the travel distance and that I heuristically found a feasible solution with objective 10. As a consequence, we can remove all edges with travel cost >= 10 because they can never be used in an optimal solution. This holds independently of the search method used.
Something I could do is add a constraint like: "objective <= objective_heuristic_solution -1" to ensure that the CP model searches for solutions which are strictly better than the heuristic solution. This however is a rather ugly work around.