"""Recipe catalog: contents, names, canonical format and the search best recipes (E5 part 1).""" import copy import json import pickle import subprocess import sys import pytest from conftest import ROOT from wisp import recipes from wisp.recipes import Recipe # docs/architecture section 3.1 table: stream -> (family, selected, folds covered, same as) EXPECTED_BEST = { "adl": ("rocket_sketch_ridge_hmm", 9, 3, None), "capture24": ("state_rocket_interval_hmm", 8, 4, None), "domino": ("state_rocket_interval_hmm", 11, 4, "cis"), "gotov": ("state_rocket_interval_hmm", 20, 5, "cis"), "iuwds": ("state_rocket_interval_hmm", 12, 5, "cis"), "harmes_left": ("state_rocket_interval_hmm", 8, 5, "cis with decoder='none'"), "harmes_right": ("state_rocket_interval_hmm", 7, 4, "cis with decoder='none'"), "handy": ("symbolic_interval_forest", 9, 3, "so"), "wisdm": ("invariant_interval_shape_forest", 15, 3, "gis"), "paal_adl": ("har_sculpt_forest_hmm", 8, 4, None), "pamap2": ("rocket_sketch_ridge_hmm", 6, 3, None), "mhealth": ("symbolic_interval_forest", 6, 1, "so"), } def test_catalog_contents(): ids = set(recipes._catalog()) assert {"ss", "rc", "so", "gis", "cis", "cse", "ese"} <= ids assert {"cis:R", "cis:I", "cis:S", "cis:RI", "cis:RS", "cis:IS"} <= ids assert len([i for i in ids if i.startswith("best:")]) == 12 assert len(ids) == 7 + 6 + 12 assert all(r.in_catalog for r in recipes._catalog().values()) def test_aliases_and_names(): assert recipes.get("cis:RIS") == recipes.get("cis") assert recipes.get("cis:SR") == recipes.get("cis:RS") assert recipes.best("capture24") == recipes.get("best:capture24") assert recipes.best("capture24_wearable_activity_v1") == recipes.best("capture24") assert recipes.dataset_id("PAAL-ADL") == "paal_adl_wrist_accel_v1" with pytest.raises(KeyError): recipes.get("not_a_recipe") with pytest.raises(KeyError): recipes.best("unknown_stream") def test_fixed_recipes_match_registry(): from wisp.registry import get_spec for short, mid in (("ss", "wisp_ss"), ("cis", "wisp_cis"), ("ese", "wisp_ese")): r = recipes.get(short) assert r.params == dict(get_spec(mid).params) assert r.paper_name == get_spec(mid).paper_name assert r.provenance["method_id"] == mid def test_cis_subsets_match_ablations(): from wisp.ablations import get_recipe_spec for tag in ("R", "I", "S", "RI", "RS", "IS"): r = recipes.get(f"cis:{tag}") spec = get_recipe_spec(f"cis_ablation_{tag}_viterbi") assert (r.params["use_rocket"], r.params["use_intervals"], r.params["use_state"]) == \ (spec.use_rocket, spec.use_intervals, spec.use_state) assert r.provenance["paper_ids"] == {"hmm": f"cis_ablation_{tag}_viterbi", "none": f"cis_ablation_{tag}_direct"} def test_recipe_roundtrip_and_immutability(): r = recipes.get("best:paal_adl") d = r.to_dict() assert d == {"name": "har_sculpt_forest_hmm", "params": {"classifier": "rf", "segment_sizes": [5, 10, 20], "symbolic_bins": 8, "use_hmm": True}, "route": "cpu"} again = Recipe.from_json(r.to_json()) assert again.to_dict() == d and again.source == "custom" assert pickle.loads(pickle.dumps(r)) == r and copy.deepcopy(r) == r with pytest.raises(AttributeError): r.family = "x" params = r.params params["use_hmm"] = False assert r.params["use_hmm"] is True assert "seed" not in Recipe("rocket_sketch_ridge_hmm", {"seed": 3, "n_jobs": 2}).params with pytest.raises(ValueError): Recipe("not_a_family") @pytest.mark.parametrize("stream", sorted(EXPECTED_BEST)) def test_search_best_table(stream): family, selected, folds, same = EXPECTED_BEST[stream] r = recipes.best(stream) prov = r.provenance assert (r.family, prov["selected"], prov["folds_covered"], prov["same_as"]) == (family, selected, folds, same) assert prov["searches"] == 30 def test_search_best_regenerates_identically(): script = ROOT / "scripts" / "build_search_best.py" if not script.exists() or not (ROOT / "results" / "search_selections.csv").exists(): pytest.skip("source tree scripts/results not available") out = subprocess.run([sys.executable, str(script), "--check"], capture_output=True, text=True) assert out.returncode == 0, out.stdout + out.stderr def test_search_runs_and_list(): runs = recipes.search_runs() assert len(runs) == 360 table = recipes.list() assert len(table) == 25 from collections import Counter texts = runs["canonical_recipe"].tolist() if hasattr(runs, "columns") else [r["canonical_recipe"] for r in runs] text, count = Counter(texts).most_common(1)[0] assert json.loads(text)["name"] == "state_rocket_interval_hmm" and count == 53