tests/unit/test_monitoring_stats.py

import pandas as pd
import pytest
from src.utils.monitoring_stats import compute_metrics, build_histogram, THRESHOLD


def make_logs(n=5, inference_ms=50.0):
    return pd.DataFrame({
        "sk_id_curr": list(range(1, n + 1)),
        "inference_time_ms": [inference_ms] * n,
    })


def make_predictions(sk_ids, proba_values):
    return pd.DataFrame({
        "sk_id_curr": sk_ids,
        "proba_class_1": proba_values,
    })


# --- compute_metrics ---

def test_compute_metrics_taux_defaut_zero():
    logs = make_logs()
    preds = make_predictions([1, 2, 3, 4, 5], [0.01, 0.02, 0.03, 0.04, 0.05])
    result = compute_metrics(logs, preds)
    assert result["taux_defaut"] == 0.0


def test_compute_metrics_taux_defaut_cent_pour_cent():
    logs = make_logs()
    preds = make_predictions([1, 2, 3, 4, 5], [0.5, 0.6, 0.7, 0.8, 0.9])
    result = compute_metrics(logs, preds)
    assert result["taux_defaut"] == 100.0


def test_compute_metrics_score_moyen():
    logs = make_logs()
    preds = make_predictions([1, 2, 3, 4, 5], [0.1, 0.2, 0.3, 0.4, 0.5])
    result = compute_metrics(logs, preds)
    assert abs(result["score_moyen"] - 0.3) < 1e-9


def test_compute_metrics_temps_moyen():
    logs = make_logs(inference_ms=75.0)
    preds = make_predictions([1, 2, 3, 4, 5], [0.1] * 5)
    result = compute_metrics(logs, preds)
    assert result["temps_moyen"] == 75.0


def test_compute_metrics_retourne_zero_si_pas_de_jointure():
    logs = make_logs()
    preds = make_predictions([99, 100], [0.5, 0.6])  # different sk_id_curr
    result = compute_metrics(logs, preds)
    assert result["n_clients"] == 0


# --- build_histogram ---

def test_build_histogram_retourne_une_figure():
    import plotly.graph_objects as go
    preds = make_predictions([1, 2, 3], [0.05, 0.15, 0.5])
    fig = build_histogram(preds)
    assert isinstance(fig, go.Figure)


def test_build_histogram_contient_deux_traces():
    preds = make_predictions([1, 2, 3], [0.05, 0.15, 0.5])
    fig = build_histogram(preds)
    # one trace per Risque category (Remboursé and Défaut)
    assert len(fig.data) == 2



def test_compute_metrics_retourne_zeros_si_logs_vide():
    logs = pd.DataFrame(columns=["sk_id_curr", "inference_time_ms"])
    preds = make_predictions([1, 2, 3], [0.5, 0.6, 0.7])
    result = compute_metrics(logs, preds)
    assert result == {"taux_defaut": 0.0, "score_moyen": 0.0, "temps_moyen": 0.0, "n_clients": 0}