from types import SimpleNamespace

import numpy as np

from motrix_envs.locomotion.go1.dreamwaq import _sanitize_dof_pos
from motrix_envs.locomotion.go1.walk_np import Go1WalkTask


def test_sanitize_dof_pos_handles_zero_quaternion_with_nonfinite_joints():
    dof_pos = np.zeros((2, 19), dtype=np.float32)
    dof_pos[0, 3:7] = 0.0
    dof_pos[0, 7] = np.inf
    dof_pos[1, 3:7] = [0.0, 0.0, 0.5, 0.5]
    dof_pos[1, 8] = np.nan

    clean = _sanitize_dof_pos(dof_pos)

    assert np.isfinite(clean).all()
    np.testing.assert_allclose(clean[0, 3:7], [0.0, 0.0, 0.0, 1.0])
    np.testing.assert_allclose(np.linalg.norm(clean[:, 3:7], axis=1), 1.0)
    assert clean[0, 7] == 0.0
    assert clean[1, 8] == 0.0


def test_apply_action_advances_three_frame_action_history():
    task = Go1WalkTask.__new__(Go1WalkTask)
    task.get_dof_vel = lambda data: np.ones((1, 12), dtype=np.float32)
    task._compute_torques = lambda actions, data: actions
    state = SimpleNamespace(
        data=SimpleNamespace(actuator_ctrls=None),
        info={
            "current_actions": np.full((1, 12), 2.0, dtype=np.float32),
            "last_actions": np.full((1, 12), 1.0, dtype=np.float32),
            "last_last_actions": np.zeros((1, 12), dtype=np.float32),
        },
    )
    actions = np.full((1, 12), 3.0, dtype=np.float32)

    task.apply_action(actions, state)
    actions.fill(9.0)

    np.testing.assert_array_equal(state.info["last_last_actions"], 1.0)
    np.testing.assert_array_equal(state.info["last_actions"], 2.0)
    np.testing.assert_array_equal(state.info["current_actions"], 3.0)