from __future__ import annotations

from datetime import datetime, timedelta
from typing import Optional

from psycopg2.extras import RealDictCursor

from app.db import get_connection


def sentiment_score(text: str) -> float:
    try:
        from textblob import TextBlob

        blob = TextBlob(text)
        score = (blob.sentiment.polarity + 1) * 2 + 1
    except Exception:
        t = text.lower()
        neg = sum(1 for w in ("bad", "teleurgest", "klacht", "lang", "duur", "fout") if w in t)
        pos = sum(1 for w in ("geweldig", "aanrader", "fantast", "mooi", "lekker", "top") if w in t)
        raw = 3.0 + (pos - neg) * 0.5
        score = max(1.0, min(5.0, raw))
    return max(1.0, min(5.0, round(float(score), 2)))


def evaluate_agent_rules(mention_id: Optional[int] = None) -> None:
    with get_connection() as conn:
        with conn.cursor(cursor_factory=RealDictCursor) as cur:
            cur.execute("SELECT * FROM agent_rules WHERE is_active = TRUE")
            rules = cur.fetchall()

            for rule in rules:
                if rule["condition_type"] == "sentiment_below":
                    threshold = rule["threshold"] or 2.0
                    if mention_id:
                        cur.execute(
                            "SELECT id, text, sentiment_score FROM social_mentions WHERE id = %s AND sentiment_score < %s",
                            (mention_id, threshold),
                        )
                    else:
                        cur.execute(
                            "SELECT id, text, sentiment_score FROM social_mentions WHERE sentiment_score < %s ORDER BY created_at DESC LIMIT 5",
                            (threshold,),
                        )
                    matches = cur.fetchall()
                    for m in matches:
                        cur.execute(
                            "SELECT 1 FROM agent_logs WHERE rule_id = %s AND message LIKE %s",
                            (rule["id"], f"%mention #{m['id']}%"),
                        )
                        if cur.fetchone():
                            continue
                        msg = (
                            f"ALERT [{rule['name']}]: Negatief sentiment ({m['sentiment_score']}/5) "
                            f"op mention #{m['id']}: {(m['text'] or '')[:120]}"
                        )
                        cur.execute(
                            "INSERT INTO agent_logs (rule_id, message) VALUES (%s, %s)",
                            (rule["id"], msg),
                        )

                elif rule["condition_type"] == "mention_spike":
                    threshold = int(rule["threshold"] or 5)
                    since = datetime.now() - timedelta(hours=24)
                    cur.execute(
                        "SELECT COUNT(*) AS cnt FROM social_mentions WHERE created_at > %s",
                        (since,),
                    )
                    count = cur.fetchone()["cnt"]
                    if count >= threshold:
                        msg = f"ALERT [{rule['name']}]: {count} mentions in 24u (drempel: {threshold})"
                        cur.execute(
                            "SELECT 1 FROM agent_logs WHERE rule_id = %s AND message = %s AND created_at > %s",
                            (rule["id"], msg, since),
                        )
                        if not cur.fetchone():
                            cur.execute(
                                "INSERT INTO agent_logs (rule_id, message) VALUES (%s, %s)",
                                (rule["id"], msg),
                            )