import pytest
import collections
import functools
import numpy as np
import qutip
from qutip.core.data import CSR, Dense
import qutip.core.data as _data

# We want to test the broadcasting rules for `qutip.expect` for a whole bunch
# of different systems, without having to repeatedly specify the systems over
# and over again.  We first store a small number of test cases for known
# expectation value in the most bundled-up form, because it's easier to unroll
# these by applying the expected broadcasting rules explicitly ourselves than
# performing the inverse operation.
#
# We store a single test case in a record type, just to keep things neatly
# together while we're munging them, so it's clear at all times what
# constitutes a valid test case.

_Case = collections.namedtuple('_Case', ['operator', 'state', 'expected'])


def _case_to_dm(case):
    return case._replace(state=[x.proj() for x in case.state])


def _unwrap(list_):
    """Unwrap lists until we reach the first non-list element."""
    out = list_
    while isinstance(out, list):
        out = out[0]
    return out


def _case_id(case):
    op_part = 'qubit' if _unwrap(case.operator).dims[0][0] == 2 else 'basis'
    state_part = 'ket' if _unwrap(case.state).dims[1][0] == 1 else 'dm'
    return op_part + "-" + state_part


# This is the minimal set of test cases, with a Fock system and a qubit system
# both in ket form and dm form.  The reference expectations are a 2D array
# which would be found by broadcasting `operator` against `state` and applying
# `qutip.expect` to the pairs.
_dim = 5
_num, _a = qutip.num(_dim), qutip.destroy(_dim)
_sx, _sz, _sp = qutip.sigmax(), qutip.sigmaz(), qutip.sigmap()
_known_fock = _Case([_num, _a],
                    [qutip.fock(_dim, n) for n in range(_dim)],
                    np.array([np.arange(_dim), np.zeros(_dim)]))
_known_qubit = _Case([_sx, _sz, _sp],
                     [qutip.basis(2, 0), qutip.basis(2, 1)],
                     np.array([[0, 0], [1, -1], [0, 0]]))
_known_cases = [_known_fock, _case_to_dm(_known_fock),
                _known_qubit, _case_to_dm(_known_qubit)]


class TestKnownExpectation:
    def pytest_generate_tests(self, metafunc):
        """
        Perform the parametrisation over the test cases, performing the
        explicit broadcasting into separate test cases when required.

        We detect whether to perform explicit broadcasting over one of the
        arguments of the `_Case` by looking for a singular/plural name of the
        parameter in the test.  If the parameter is singular, then we manually
        perform the broadcasting rule for that fixture, and parametrise over
        the resulting list, taking care to pick out the correct parts of the
        reference array.
        """
        cases = _known_cases
        op_name, state_name = 'operator', 'state'
        if op_name not in metafunc.fixturenames:
            op_name += 's'
        else:
            cases = [_Case(op, case.state, expected)
                     for case in cases
                     for op, expected in zip(case.operator, case.expected)]
        if state_name not in metafunc.fixturenames:
            state_name += 's'
        else:
            cases = [_Case(case.operator, state, expected)
                     for case in cases
                     for state, expected in zip(case.state, case.expected.T)]
        metafunc.parametrize([op_name, state_name, 'expected'], cases,
                             ids=[_case_id(case) for case in cases])

    def test_operator_by_basis(self, operator, state, expected):
        result = qutip.expect(operator, state)
        assert result == expected
        assert isinstance(result, float if operator.isherm else complex)

    def test_broadcast_operator_list(self, operators, state, expected):
        result = qutip.expect(operators, state)
        expected_dtype = (np.float64 if all(op.isherm for op in operators)
                          else np.complex128)
        assert isinstance(result, np.ndarray)
        assert result.dtype == expected_dtype
        assert list(result) == list(expected)

    def test_broadcast_state_list(self, operator, states, expected):
        result = qutip.expect(operator, states)
        expected_dtype = np.float64 if operator.isherm else np.complex128
        assert isinstance(result, np.ndarray)
        assert result.dtype == expected_dtype
        assert list(result) == list(expected)

    def test_broadcast_both_lists(self, operators, states, expected):
        result = qutip.expect(operators, states)
        assert len(result) == len(operators)
        for part, operator, expected_part in zip(result, operators, expected):
            expected_dtype = np.float64 if operator.isherm else np.complex128
            assert isinstance(part, np.ndarray)
            assert part.dtype == expected_dtype
            assert list(part) == list(expected_part)


@pytest.mark.repeat(5)
@pytest.mark.parametrize("hermitian", [False, True], ids=['complex', 'real'])
@pytest.mark.parametrize("op_type", [CSR, Dense], ids=['csr', 'dense'])
@pytest.mark.parametrize("state_type", [CSR, Dense], ids=['csr', 'dense'])
def test_equivalent_to_matrix_element(hermitian, op_type, state_type):
    dimension = 20
    state = qutip.rand_ket(dimension, 0.3).to(op_type)
    op = qutip.rand_herm(dimension, 0.2).to(state_type)
    if not hermitian:
        op = op + 1j*qutip.rand_herm(dimension, 0.1).to(state_type)
    expected = state.dag() * op * state
    assert abs(qutip.expect(op, state) - expected) < 1e-14


@pytest.mark.parametrize("solve", [
    pytest.param(qutip.sesolve, id="sesolve"),
    pytest.param(functools.partial(qutip.mesolve, c_ops=[qutip.qzero(2)]),
                 id="mesolve"),
])
def test_compatibility_with_solver(solve):
    e_ops = [getattr(qutip, 'sigma'+x)() for x in 'xyzmp']
    e_ops += [lambda t, psi: np.sin(t)]
    h = qutip.sigmax()
    state = qutip.basis(2, 0)
    times = np.linspace(0, 10, 101)
    options = {"store_states":True}
    result = solve(h, state, times, e_ops=e_ops, options=options)
    direct, states = result.expect, result.states
    indirect = qutip.expect(e_ops[:-1], states)
    # check measurement operators based on quantum objects
    assert len(direct)-1 == len(indirect)
    for direct_, indirect_ in zip(direct, indirect):
        assert len(direct_) == len(indirect_)
        assert isinstance(direct_, np.ndarray)
        assert isinstance(indirect_, np.ndarray)
        assert direct_.dtype == indirect_.dtype
        np.testing.assert_allclose(np.array(direct_), indirect_, atol=1e-12)
        # test measurement operators based on lambda functions
        direct_ = direct[-1]
        indirect_ = np.sin(times)
        assert len(direct_) == len(indirect_)
        assert isinstance(direct_, np.ndarray)
        assert isinstance(indirect_, np.ndarray)
        assert direct_.dtype == indirect_.dtype
        np.testing.assert_allclose(np.array(direct_), indirect_, atol=1e-12)


def test_no_real_attribute(monkeypatch):
    """This tests ensures that expect still works even if the output of a
    specialisation does not have the ``real`` attribute. This is the case for
    the tensorflow and cupy data layers."""

    def mocker_expect_return(oper, state):
        """
        We simply return None which does not have the `real` attribute.
        """
        return "object without .real"

    monkeypatch.setattr(_data, "expect", mocker_expect_return)

    sz = qutip.sigmaz() # the choice of the matrix does not matter
    assert "object without .real" == qutip.expect(sz, sz)