Several algorithms for optimizing a combinatorial subspace of chemical compound space with constraints are compared. The test system is a library of organic chromophores for electro-optic applications. The constraints on the optimization include the maximization of the candidate structure hyperpolarizability while keeping the absorption within acceptable limits in the range of 400-700 nm. The best pay-off in terms of primary objective, feasibility and computational cost is achieved using a heuristic reordering of orthogonal search directions.