"""
Test script for the refund model service
Run this to verify the model loads and can make predictions
"""

import sys
from pathlib import Path

# Add the parent directory to path to import the service
sys.path.insert(0, str(Path(__file__).parent))

try:
    import xgboost as xgb
    import numpy as np
    import json
    print("✓ Dependencies imported successfully")
except ImportError as e:
    print(f"✗ Missing dependency: {e}")
    print("Run: pip install xgboost numpy")
    sys.exit(1)

# Test 1: Load model files
print("\n=== Test 1: Loading Model Files ===")
try:
    model_path = Path(__file__).parent / '../../customer_score_models' / 'xgb_model.ubj'
    features_path = Path(__file__).parent / '../../customer_score_models' / 'feature_columns.json'
    
    assert model_path.exists(), f"Model file not found: {model_path}"
    assert features_path.exists(), f"Features file not found: {features_path}"
    print(f"✓ Model file found: {model_path}")
    print(f"✓ Features file found: {features_path}")
    
    # Load model
    model = xgb.Booster()
    model.load_model(str(model_path))
    print("✓ XGBoost model loaded successfully")
    
    # Load features
    with open(features_path, 'r') as f:
        feature_columns = json.load(f)
    print(f"✓ Feature columns loaded: {len(feature_columns)} features")
    
except Exception as e:
    print(f"✗ Error loading model: {e}")
    sys.exit(1)

# Test 2: Make a test prediction
print("\n=== Test 2: Test Prediction ===")
try:
    # Create a dummy feature vector (all zeros except a few)
    dummy_features = np.zeros(len(feature_columns), dtype=np.float32)
    
    # Set some realistic values
    feature_dict = {col: 0 for col in feature_columns}
    feature_dict['user_age'] = 35
    feature_dict['address_pincode'] = 400001
    feature_dict['base_price'] = 500.0
    feature_dict['order_quantity'] = 1
    feature_dict['product_price'] = 500.0
    feature_dict['discount_applied'] = 10.0
    feature_dict['days_to_return'] = 5
    feature_dict['is_exchanged'] = 0
    feature_dict['risk_score'] = 0.3
    feature_dict['discount_ratio'] = 0.02
    feature_dict['price_inverse'] = 0.002
    feature_dict['days_inverse'] = 0.2
    
    # Set one product type (example)
    if 'product_name_Electronics Product' in feature_dict:
        feature_dict['product_name_Electronics Product'] = 1
    
    # Set one payment method
    if 'payment_method_Credit Card' in feature_dict:
        feature_dict['payment_method_Credit Card'] = 1
    
    # Convert to array
    feature_array = np.array([feature_dict[col] for col in feature_columns], dtype=np.float32)
    
    # Make prediction
    dmatrix = xgb.DMatrix(feature_array.reshape(1, -1))
    prediction = model.predict(dmatrix, validate_features=False)[0]

    
    print(f"✓ Test prediction successful: {prediction:.2f}%")
    
    if 0 <= prediction <= 100:
        print("✓ Prediction is in valid range (0-100)")
    else:
        print(f"⚠ Warning: Prediction outside expected range: {prediction}")
    
except Exception as e:
    print(f"✗ Error making prediction: {e}")
    import traceback
    traceback.print_exc()
    sys.exit(1)

# Test 3: Check feature columns structure
print("\n=== Test 3: Feature Columns Analysis ===")
try:
    # Count feature types
    product_features = [f for f in feature_columns if f.startswith('product_name_')]
    payment_features = [f for f in feature_columns if f.startswith('payment_method_')]
    numeric_features = [f for f in feature_columns if not (f.startswith('product_name_') or f.startswith('payment_method_'))]
    
    print(f"✓ Numeric features: {len(numeric_features)}")
    print(f"  Examples: {numeric_features[:5]}")
    print(f"✓ Product name features: {len(product_features)}")
    print(f"✓ Payment method features: {len(payment_features)}")
    print(f"  Available: {[f.replace('payment_method_', '') for f in payment_features]}")
    
except Exception as e:
    print(f"✗ Error analyzing features: {e}")

print("\n" + "="*50)
print("✓ All tests passed! Model is ready to use.")
print("="*50)