import torch

from models.reveal_head import compute_task_metrics_from_fields


def _metrics(access_level: float, disturbance_level: float) -> dict[str, torch.Tensor]:
    access_field = torch.full((1, 3, 4, 4), access_level)
    persistence_field = torch.full((1, 1, 4, 4), 0.6)
    disturbance_field = torch.full((1, 1, 4, 4), disturbance_level)
    reocclusion_field = torch.full((1, 1, 4, 4), 0.1)
    visibility_field = torch.full((1, 1, 4, 4), 0.6)
    clearance_field = torch.full((1, 2, 4, 4), access_level)
    support_stability_field = torch.full((1, 1, 4, 4), 0.7)
    uncertainty_field = torch.full((1, 1, 4, 4), 0.2)
    return compute_task_metrics_from_fields(
        access_field=access_field,
        persistence_field=persistence_field,
        disturbance_field=disturbance_field,
        reocclusion_field=reocclusion_field,
        visibility_field=visibility_field,
        clearance_field=clearance_field,
        support_stability_field=support_stability_field,
        uncertainty_field=uncertainty_field,
    )


def test_task_metric_monotonicity():
    low_open = _metrics(access_level=0.1, disturbance_level=0.2)
    high_open = _metrics(access_level=0.9, disturbance_level=0.2)
    over_lift = _metrics(access_level=0.9, disturbance_level=0.8)

    assert high_open["mouth_aperture"] > low_open["mouth_aperture"]
    assert high_open["actor_feasibility_score"] > low_open["actor_feasibility_score"]
    assert over_lift["fold_preservation"] < high_open["fold_preservation"]