from numpy.linalg import norm

from qutip import (
    Qobj, tensor, vector_to_operator, operator_to_vector, kraus_to_super,
    subsystem_apply, rand_dm, rand_unitary,
)
from qutip.random_objects import rand_kraus_map


class TestSubsysApply(object):
    """
    A test class for the QuTiP function for applying superoperators to
    subsystems.
    The four tests below determine whether efficient numerics, naive numerics
    and semi-analytic results are identical.
    """

    def test_SimpleSingleApply(self):
        """
        Non-composite system, operator on Hilbert space.
        """
        tol = 1e-12
        rho_3 = rand_dm(3)
        single_op = rand_unitary(3)
        analytic_result = single_op * rho_3 * single_op.dag()
        naive_result = subsystem_apply(rho_3, single_op, [True],
                                       reference=True)
        naive_diff = (analytic_result - naive_result).full()
        naive_diff_norm = norm(naive_diff)
        assert naive_diff_norm < tol

        efficient_result = subsystem_apply(rho_3, single_op, [True])
        efficient_diff = (efficient_result - analytic_result).full()
        efficient_diff_norm = norm(efficient_diff)
        assert efficient_diff_norm < tol

    def test_SimpleSuperApply(self):
        """
        Non-composite system, operator on Liouville space.
        """
        tol = 1e-12
        rho_3 = rand_dm(3)
        superop = kraus_to_super(rand_kraus_map(3))
        analytic_result = vector_to_operator(superop @
                                             operator_to_vector(rho_3))
        naive_result = subsystem_apply(rho_3, superop, [True],
                                       reference=True)
        naive_diff = (analytic_result - naive_result).full()
        naive_diff_norm = norm(naive_diff)
        assert naive_diff_norm < tol

        efficient_result = subsystem_apply(rho_3, superop, [True])
        efficient_diff = (efficient_result - analytic_result).full()
        efficient_diff_norm = norm(efficient_diff)
        assert efficient_diff_norm < tol

    def test_ComplexSingleApply(self):
        """
        Composite system, operator on Hilbert space.
        """
        tol = 1e-12
        rho_list = list(map(rand_dm, [2, 3, 2, 3, 2]))
        rho_input = tensor(rho_list)
        single_op = rand_unitary(3)

        analytic_result = rho_list
        analytic_result[1] = single_op * analytic_result[1] * single_op.dag()
        analytic_result[3] = single_op * analytic_result[3] * single_op.dag()
        analytic_result = tensor(analytic_result)

        naive_result = subsystem_apply(rho_input, single_op,
                                       [False, True, False, True, False],
                                       reference=True)
        naive_diff = (analytic_result - naive_result).full()
        naive_diff_norm = norm(naive_diff)
        assert naive_diff_norm < tol

        efficient_result = subsystem_apply(rho_input, single_op,
                                           [False, True, False, True, False])
        efficient_diff = (efficient_result - analytic_result).full()
        efficient_diff_norm = norm(efficient_diff)
        assert efficient_diff_norm < tol

    def test_ComplexSuperApply(self):
        """
        Superoperator: Efficient numerics and reference return same result,
        acting on non-composite system
        """
        tol = 1e-10
        rho_list = list(map(rand_dm, [2, 3, 2, 3, 2]))
        rho_input = tensor(rho_list)
        superop = kraus_to_super(rand_kraus_map(3))

        analytic_result = rho_list
        analytic_result[1] = Qobj(
            vector_to_operator(
                superop @ operator_to_vector(analytic_result[1])
            ))
        analytic_result[3] = Qobj(
            vector_to_operator(
                superop @ operator_to_vector(analytic_result[3])
            ))
        analytic_result = tensor(analytic_result)

        naive_result = subsystem_apply(rho_input, superop,
                                       [False, True, False, True, False],
                                       reference=True)
        naive_diff = (analytic_result - naive_result).full()
        naive_diff_norm = norm(naive_diff)
        assert naive_diff_norm < tol

        efficient_result = subsystem_apply(rho_input, superop,
                                           [False, True, False, True, False])
        efficient_diff = (efficient_result - analytic_result).full()
        efficient_diff_norm = norm(efficient_diff)
        assert efficient_diff_norm < tol