test_large_arrays.py

"""Quick test: verify large-array logic is intact after regime-aware changes."""
import random
random.seed(42)

from sorting.recommender import SortingRecommender

r = SortingRecommender()
r.load_model()

print("=== LARGE ARRAY REGIME TESTS ===\n")

cases = [
    ("Random 1000",         [random.random() for _ in range(1000)]),
    ("Random 5000",         [random.random() for _ in range(5000)]),
    ("Random 10000",        [random.random() for _ in range(10000)]),
    ("Nearly sorted 1000",  list(range(1000))),
    ("Nearly sorted 5000",  list(range(5000))),
    ("Reverse sorted 1000", list(range(1000, 0, -1))),
    ("Reverse sorted 5000", list(range(5000, 0, -1))),
    ("Many dupes 5000",     [random.randint(1, 10) for _ in range(5000)]),
]

for name, arr in cases:
    res = r.recommend(arr, return_probabilities=True)
    algo = res["algorithm"]
    alpha = res["adaptive_alpha"]
    probs = res["probabilities"]
    print(f"{name:25s} | algo={algo:18s} | alpha={alpha} | probs={probs}")

print("\n=== COMPARISON: what raw ML model would have predicted ===\n")

import numpy as np
for name, arr in cases:
    features = r.feature_extractor.extract(arr)
    fv = r.feature_extractor.to_vector(features)
    X = np.array([fv])
    pred = r.model.predict(X)[0]
    raw_algo = r.label_encoder.inverse_transform([pred])[0]
    blended_algo = r.recommend(arr)["algorithm"]
    match = "SAME" if raw_algo == blended_algo else "CHANGED"
    print(f"{name:25s} | raw_ML={raw_algo:18s} | blended={blended_algo:18s} | {match}")

print("\nDone.")