import sys
import os
import pandas as pd
import numpy as np

# Bulletproof pathing: Force Python to look in both the current folder AND the parent folder
# This ensures it finds models.py regardless of whether this file is run from /tests or the root.
_this_dir = os.path.dirname(os.path.abspath(__file__))
sys.path.insert(0, _this_dir)
sys.path.insert(0, os.path.dirname(_this_dir))

# Import the specific function from the engine
from models import model_bsts

# ─────────────────────────────────────────────
# 1. FREQUENCY GATING TESTS
# ─────────────────────────────────────────────
def test_bsts_frequency_gate_aborts_on_daily():
    """
    BSTS overfits on daily stock returns (which are essentially white noise).
    This test verifies that the engine completely aborts if 'periods' > 52 
    (e.g., daily data with 252 periods).
    """
    rng = np.random.default_rng(42)
    # Synthetic daily data
    returns_df = pd.DataFrame({'AAPL': rng.normal(0.0005, 0.01, size=100)})
    
    # Act: Request a BSTS fit expecting 252 trading periods in a year
    res = model_bsts(returns_df, periods=252, silent=True)
    
    # Assert: The model should instantly reject the request to prevent noise-fitting
    assert res.expected_returns is None or res.expected_returns.empty

def test_bsts_short_series_fallback():
    """
    BSTS requires a minimum amount of data to establish stationary momentum.
    If the series is too short (e.g., < 12 months), it must output 
    massive uncertainty so the Black-Litterman bridge safely ignores it.
    """
    rng = np.random.default_rng(42)
    # Only 10 months of data
    returns_df = pd.DataFrame({'TSLA': rng.normal(0.01, 0.05, size=10)})
    
    res = model_bsts(returns_df, periods=12, silent=True)
    
    # Assert: Should output 0 expected return and massive (1000.0) uncertainty
    assert np.isclose(res.expected_returns['TSLA'], 0.0)
    assert np.isclose(res.uncertainties['TSLA'], 1000.0)

def test_bsts_valid_signal():
    """
    If a monthly time series exhibits momentum, the BSTS model should
    correctly output a genuine variance (uncertainty < 1000.0).
    """
    rng = np.random.default_rng(42)
    n = 60
    
    # Create a synthetic AR(1) process with strong positive momentum
    series = np.zeros(n)
    ar_coef = 0.8
    for i in range(1, n):
        series[i] = ar_coef * series[i-1] + rng.normal(0, 0.005)
        
    returns_df = pd.DataFrame({'TRENDY_ASSET': series})
    
    # Act
    res = model_bsts(returns_df, periods=12, silent=True)
    
    # Assert: The model must recognize the signal. The uncertainty should be the 
    # legitimate standard error of the forecast, NOT the 1000.0 rejection flag.
    assert res.uncertainties['TRENDY_ASSET'] < 1000.0
    assert not pd.isna(res.expected_returns['TRENDY_ASSET'])