from datetime import datetime, timedelta
from collections import defaultdict
import numpy as np
from sqlalchemy.orm import Session
from app.database.models import Account, Transaction

def analyze_spending_behavior(db: Session, user_id: str, days: int = 90):
    """
    Analyzes historical transactions to detect behavioral patterns (late-night, impulsive, dopamine, stress).
    """
    accounts = db.query(Account).filter(Account.user_id == user_id).all()
    account_ids = [acc.id for acc in accounts]
    
    if not account_ids:
        return {"insights": [], "metrics": {}}
        
    cutoff = datetime.utcnow() - timedelta(days=days)
    txns = db.query(Transaction).filter(
        Transaction.account_id.in_(account_ids),
        Transaction.timestamp >= cutoff,
        Transaction.type == "debit"
    ).all()
    
    if not txns:
        return {
            "insights": ["No recent debit transactions to analyze. Complete a few purchases to start behavioral profiling."],
            "metrics": {
                "late_night_count": 0,
                "late_night_total": 0.0,
                "weekend_pct": 0.0,
                "impulsive_count": 0,
                "impulsive_total": 0.0,
                "dopamine_count": 0,
                "stress_count": 0
            }
        }
        
    # Analyze variables
    late_night_txns = []
    weekend_txns = []
    impulsive_txns = []
    dopamine_txns = []
    stress_txns = []
    
    amounts = [t.amount for t in txns]
    avg_txn = np.mean(amounts)
    std_txn = np.std(amounts) if len(amounts) > 1 else 0.0
    
    category_totals = defaultdict(float)
    hourly_counts = defaultdict(int)
    
    for t in txns:
        # Categorize
        category_totals[t.category or "Other"] += t.amount
        
        # Timing
        hour = t.timestamp.hour
        hourly_counts[hour] += 1
        
        # Late-night spending (11PM to 4AM)
        if hour >= 23 or hour < 4:
            late_night_txns.append(t)
            
        # Weekend spending (Friday, Saturday, Sunday)
        # weekday() is 0=Monday, 4=Friday, 5=Saturday, 6=Sunday
        day = t.timestamp.weekday()
        if day in [4, 5, 6]:
            weekend_txns.append(t)
            
        # Impulsive spending (More than average + 1.5 * standard dev, or marked as 'regret')
        if t.amount > (avg_txn + 1.5 * std_txn) and (t.category in ["Shopping", "Entertainment", "Food"]):
            impulsive_txns.append(t)
            
        # Emotion tags
        emotion = (t.spending_emotion_label or "").lower()
        if emotion == "regret":
            stress_txns.append(t)
        elif emotion in ["happy", "dopamine"] or (t.category == "Shopping" and t.amount > avg_txn):
            dopamine_txns.append(t)
            
    # Insights construction
    insights = []
    
    # 1. Late night alert
    late_night_pct = (len(late_night_txns) / len(txns) * 100) if txns else 0
    if late_night_pct > 15:
        total_late = sum(t.amount for t in late_night_txns)
        insights.append(
            f"🌙 High late-night spending: {late_night_pct:.1f}% of transactions occur after 11PM (Total: ${total_late:,.2f}). "
            "Consider setting a bedtime blocker on your bank card."
        )
        
    # 2. Weekend overspending
    weekend_pct = (len(weekend_txns) / len(txns) * 100) if txns else 0
    if weekend_pct > 45:
        weekend_avg = np.mean([t.amount for t in weekend_txns]) if weekend_txns else 0
        weekday_txns = [t for t in txns if t not in weekend_txns]
        weekday_avg = np.mean([t.amount for t in weekday_txns]) if weekday_txns else 0
        
        if weekend_avg > weekday_avg * 1.2:
            pct_diff = ((weekend_avg - weekday_avg) / weekday_avg) * 100
            insights.append(
                f"🎉 Weekend Spikes: You spend {pct_diff:.1f}% more on weekends than weekdays. "
                "Mainly driven by dining out and recreational purchases."
            )
            
    # 3. Dopamine triggers
    if len(dopamine_txns) > 3:
        insights.append(
            f"🛍️ Dopamine Spending: Detected {len(dopamine_txns)} shopping spikes. "
            "These purchases often occur in bursts, indicating reward-seeking behavior."
        )
        
    # 4. Stress/Regret Spending
    if len(stress_txns) > 0:
        insights.append(
            f"⚠️ Emotional Spending: You flagged {len(stress_txns)} transactions as 'regret' or 'stress spending'. "
            "Implementing a 24-hour cooling-off rule for non-essential items over $100 could help."
        )
        
    # General fallback if no major insights
    if not insights:
        insights.append("📊 Spending Discipline: Your transactions exhibit stable and regular timing, with minimal signs of emotional or impulsive spending.")

    return {
        "insights": insights,
        "metrics": {
            "late_night_count": len(late_night_txns),
            "late_night_total": round(sum(t.amount for t in late_night_txns), 2),
            "weekend_pct": round(weekend_pct, 2),
            "impulsive_count": len(impulsive_txns),
            "impulsive_total": round(sum(t.amount for t in impulsive_txns), 2),
            "dopamine_count": len(dopamine_txns),
            "stress_count": len(stress_txns),
            "avg_transaction_amount": round(avg_txn, 2)
        },
        "category_breakdown": {cat: round(amt, 2) for cat, amt in category_totals.items()}
    }