from types import SimpleNamespace import numpy as np import pytest from tfm.data.external import collect_citylearn, collect_minari class Space: shape = (1,) dtype = np.float32 low = np.array([-1], dtype=np.float32) high = np.array([1], dtype=np.float32) def contains(self, value): return value.shape == (1,) and np.all(value >= self.low) and np.all(value <= self.high) class City: action_names = [["electrical_storage"]] observation_names = [["indoor_dry_bulb_temperature"]] action_space = [Space()] seconds_per_time_step = 900 def reset(self, seed): self.t = 0 return [[22.0]], {} def step(self, actions): self.t += 1 return [[22.0]], 0, self.t == 2, False, {} def test_citylearn_native_enabled_actions_preserved(tmp_path): seen = [] def policy(observed, names, spaces): seen.extend(names[0]) return [[0.5]] metadata = collect_citylearn(City(), policy, tmp_path / "city.npz", system_id="building-1") assert set(seen) == {"electrical_storage"} assert metadata["schema_mapping_validated"] is False with np.load(tmp_path / "city.npz", allow_pickle=False) as arrays: assert arrays["observations_0"].shape == (3, 1) assert arrays["actions_0"].shape == (2, 1) with pytest.raises(ValueError, match="Invalid native"): collect_citylearn(City(), lambda *args: [[2]], tmp_path / "invalid.npz", system_id="building-1") def test_minari_requires_instance_grouping(tmp_path): episode = SimpleNamespace( id=1, observations=np.zeros((3, 2)), actions=np.zeros((2, 1)), terminations=np.array([False, True]), truncations=np.array([False, False]), ) dataset = SimpleNamespace( spec=SimpleNamespace(dataset_id="test-v0"), iterate_episodes=lambda: iter([episode]) ) with pytest.raises(ValueError, match="grouping"): collect_minari(dataset, tmp_path, system_id_for_episode=lambda ep: None) metadata = collect_minari(dataset, tmp_path, system_id_for_episode=lambda ep: "physical-instance") assert metadata[0]["system_id"] == "physical-instance"