"""Tests para run_eda_models."""

import numpy as np

from run_eda_models import run_eda_models


def _numeric(values: list) -> dict:
    """Envuelve una lista como columna numerica del perfil."""
    return {"values": values, "type": "numeric"}


def test_three_numeric_columns_runs_all_models():
    # Tres columnas con estructura latente: x e y correlacionadas, z ruido.
    rng = np.random.default_rng(0)
    n = 120
    x = rng.normal(0.0, 1.0, n)
    y = x * 2.0 + rng.normal(0.0, 0.3, n)
    z = rng.normal(5.0, 1.0, n)

    columns = {
        "x": _numeric(x.tolist()),
        "y": _numeric(y.tolist()),
        "z": _numeric(z.tolist()),
    }

    result = run_eda_models(columns)

    assert result["n_numeric_cols"] == 3
    assert result["note"] == ""

    # PCA presente y con varianza explicada.
    assert result["pca"] is not None
    assert result["pca"]["n_components"] >= 1
    assert len(result["pca"]["explained_variance_ratio"]) >= 1

    # KMeans presente con un k elegido.
    assert result["kmeans"] is not None
    assert result["kmeans"]["best_k"] >= 2

    # Outliers presente (puede ser 0 outliers, pero el bloque existe).
    assert result["outliers"] is not None
    assert "n_outliers" in result["outliers"]

    # Normality presente, una entrada por columna numerica.
    assert result["normality"] is not None
    assert set(result["normality"].keys()) == {"x", "y", "z"}
    for col in ("x", "y", "z"):
        assert result["normality"][col]["n"] == n


def test_single_numeric_column_note_and_normality_only():
    rng = np.random.default_rng(7)
    values = rng.normal(10.0, 2.0, 100).tolist()
    columns = {
        "only": _numeric(values),
        "label": {"values": ["a"] * 100, "type": "categorical"},
    }

    result = run_eda_models(columns)

    assert result["n_numeric_cols"] == 1
    assert result["note"] == "insuficientes columnas numericas para modelos multivariantes"

    # Multivariantes en None.
    assert result["pca"] is None
    assert result["kmeans"] is None
    assert result["outliers"] is None

    # Normality univariante si se ejecuta con una sola columna.
    assert result["normality"] is not None
    assert "only" in result["normality"]
    assert result["normality"]["only"]["n"] == 100


def test_flags_disable_models():
    rng = np.random.default_rng(1)
    n = 60
    columns = {
        "a": _numeric(rng.normal(0, 1, n).tolist()),
        "b": _numeric(rng.normal(0, 1, n).tolist()),
    }

    result = run_eda_models(
        columns,
        run_pca=False,
        run_kmeans=False,
        run_isolation=False,
        run_normality=False,
    )

    assert result["n_numeric_cols"] == 2
    assert result["pca"] is None
    assert result["kmeans"] is None
    assert result["outliers"] is None
    assert result["normality"] is None
    assert result["note"] == ""


def test_no_numeric_columns_returns_note_and_no_normality():
    columns = {
        "cat": {"values": ["x", "y", "z"], "type": "categorical"},
    }

    result = run_eda_models(columns)

    assert result["n_numeric_cols"] == 0
    assert result["note"] == "insuficientes columnas numericas para modelos multivariantes"
    assert result["pca"] is None
    # run_normality True pero no hay columnas numericas -> None.
    assert result["normality"] is None