import numpy as np, pytest
from traite import hedonique as h

def test_mco_recupere_parametres():
    rng = np.random.default_rng(1)
    n = 2000; x = rng.normal(size=(n, 2)); X = np.column_stack([np.ones(n), x])
    y = X @ np.array([12.0, 0.5, -0.2]) + rng.normal(0, 0.1, n)
    r = h.mco(X, y)
    assert np.allclose(r["beta"], [12.0, 0.5, -0.2], atol=0.02) and r["r2"] > 0.9

def test_indice_dummies():
    idx = h.indice_dummies_temps(np.array([0.05, 0.10]))
    assert idx[0] == 100 and idx[1] == pytest.approx(100 * np.exp(0.05))

def test_ventes_repetees_recupere_indice():
    rng = np.random.default_rng(7)
    vrai = np.array([0, 0.04, 0.09, 0.13, 0.20])  # log-indice
    n = 3000
    t1 = rng.integers(0, 4, n); t2 = np.array([rng.integers(a + 1, 5) for a in t1])
    p1 = np.exp(12 + rng.normal(0, 0.3, n) + vrai[t1]); p2 = p1 * np.exp(vrai[t2] - vrai[t1] + rng.normal(0, 0.05, n))
    r = h.ventes_repetees_bmn(p1, p2, t1, t2, 5)
    assert np.allclose(np.log(r["indice"] / 100), vrai, atol=0.01)
    cs = h.ventes_repetees_case_shiller(p1, p2, t1, t2, 5)
    assert np.allclose(np.log(cs["indice"] / 100), vrai, atol=0.01)

def test_statistiques_ratio_parfaites():
    P = np.array([300e3, 400e3, 500e3, 600e3]); V = P * 0.95
    s = h.statistiques_ratio(V, P)
    assert s["ratio_median"] == pytest.approx(0.95) and s["COD"] == pytest.approx(0.0) and s["PRD"] == pytest.approx(1.0) and abs(s["PRB"]) < 1e-9
    # régressivité : ratios plus élevés pour les biens bon marché → PRD > 1, PRB < 0
    V2 = P * np.array([1.05, 1.0, 0.95, 0.90])
    s2 = h.statistiques_ratio(V2, P)
    assert s2["PRD"] > 1 and s2["PRB"] < 0 and s2["COD"] > 0

def test_moran_detecte_autocorrelation():
    rng = np.random.default_rng(3)
    coords = rng.uniform(0, 10, (300, 2))
    W = h.matrice_poids_knn(coords, 6)
    assert np.allclose(W.sum(1), 1)
    x_alea = rng.normal(size=300)
    x_spat = coords[:, 0] + rng.normal(0, 0.5, 300)
    assert abs(h.moran_i(x_alea, W)["z"]) < 3
    assert h.moran_i(x_spat, W)["z"] > 5

def test_sar_recupere_rho():
    rng = np.random.default_rng(11)
    n = 400; coords = rng.uniform(0, 10, (n, 2)); W = h.matrice_poids_knn(coords, 5)
    X = np.column_stack([np.ones(n), rng.normal(size=n)]); beta = np.array([1.0, 0.8]); rho = 0.5
    e = rng.normal(0, 0.3, n)
    y = np.linalg.solve(np.eye(n) - rho * W, X @ beta + e)
    r = h.sar_lag_mv(y, X, W)
    assert abs(r["rho"] - rho) < 0.1 and np.allclose(r["beta"], beta, atol=0.15)