"""
Tests for P2/P3 features:
  - Noise-Filtered Transformer (deep learning sequence model)
  - Options Flow Sentiment (alternative data ingestion)
  - Exact True Risk Parity (ERC allocation engine)
"""
import pytest
import numpy as np
import pandas as pd
from unittest.mock import patch, MagicMock


# ─────────────────────────────────────────────
# TRANSFORMER (Deep Learning Sequence Model)
# ─────────────────────────────────────────────
import dl_models


def test_transformer_train_and_predict():
    """Train the NoiseFilteredTransformer on synthetic cross-asset data and verify predictions."""
    features_dict = {}
    for t in ['AAPL', 'MSFT', 'GOOGL']:
        df = pd.DataFrame(np.random.randn(100, 10), columns=[f'feat_{i}' for i in range(10)])
        df['target'] = df['feat_0'] * 0.5 + df['feat_1'] * 0.2 + np.random.randn(100) * 0.1
        df['ret'] = np.random.randn(100) * 0.01
        features_dict[t] = df

    model, scaler = dl_models.train_cross_asset_transformer(
        features_dict,
        seq_len=60,
        epochs=2,
        batch_size=16,
        silent=True
    )

    assert model is not None
    assert scaler is not None

    preds = dl_models.predict_transformer(model, scaler, features_dict, seq_len=60)

    assert len(preds) == 3
    assert 'AAPL' in preds
    assert 'MSFT' in preds
    assert 'GOOGL' in preds
    assert isinstance(preds['AAPL'], float)


# ─────────────────────────────────────────────
# OPTIONS FLOW SENTIMENT (Alternative Data)
# ─────────────────────────────────────────────
import alternative_data


@patch('alternative_data.yf.Ticker')
def test_options_sentiment_success(mock_ticker):
    """Verify put/call ratio and IV skew extraction from a mocked options chain."""
    mock_tk = MagicMock()
    mock_ticker.return_value = mock_tk

    mock_tk.options = ('2026-07-01',)
    mock_chain = MagicMock()

    mock_chain.calls = pd.DataFrame({
        'volume': [100, 200, 300],
        'impliedVolatility': [0.1, 0.15, 0.2]
    })

    mock_chain.puts = pd.DataFrame({
        'volume': [50, 150],
        'impliedVolatility': [0.2, 0.3]
    })

    mock_tk.option_chain.return_value = mock_chain

    results = alternative_data.fetch_options_sentiment(['AAPL'], silent=True)

    assert 'AAPL' in results
    assert 'put_call_ratio' in results['AAPL']
    assert 'iv_skew' in results['AAPL']

    # calls volume sum = 600, puts volume sum = 200 → ratio = 0.333
    assert pytest.approx(results['AAPL']['put_call_ratio'], 0.01) == 0.333

    # calls mean IV = 0.15, puts mean IV = 0.25 → skew = 0.10
    assert pytest.approx(results['AAPL']['iv_skew'], 0.01) == 0.10


@patch('alternative_data.yf.Ticker')
def test_options_sentiment_no_options_available(mock_ticker):
    """Verify safe defaults when a ticker has no listed options chains."""
    mock_tk = MagicMock()
    mock_ticker.return_value = mock_tk
    mock_tk.options = ()

    results = alternative_data.fetch_options_sentiment(['TSLA'], silent=True)

    assert 'TSLA' in results
    assert results['TSLA']['put_call_ratio'] == 1.0
    assert results['TSLA']['iv_skew'] == 0.0


# ─────────────────────────────────────────────
# EXACT TRUE RISK PARITY (ERC Engine)
# ─────────────────────────────────────────────
from erc_engine import exact_risk_parity_allocation


def test_exact_risk_parity_equal_contributions():
    """Mathematically verify that ERC produces equal marginal risk contributions."""
    # 3 assets with widely different volatilities
    vols = np.array([0.1, 0.2, 0.4])

    corr = np.array([
        [1.0, 0.3, 0.2],
        [0.3, 1.0, 0.4],
        [0.2, 0.4, 1.0]
    ])

    cov = np.outer(vols, vols) * corr
    cov_df = pd.DataFrame(cov, columns=['A', 'B', 'C'], index=['A', 'B', 'C'])

    weights = exact_risk_parity_allocation(cov_df, silent=True)

    assert len(weights) == 3
    assert np.isclose(weights.sum(), 1.0)

    # Compute marginal risk contributions: w_i * (Σw)_i
    w_arr = weights.values
    marginal_risk = cov.dot(w_arr)
    risk_contributions = w_arr * marginal_risk

    # All risk contributions must be equal within tight tolerance
    mean_rc = np.mean(risk_contributions)
    np.testing.assert_allclose(risk_contributions, mean_rc, rtol=1e-3, atol=1e-5)

    # Lower-vol asset A should have the highest weight
    assert weights['A'] > weights['C']
    assert weights['A'] > weights['B']