"""
Toxic Content Detector - Bahasa Indonesia
==========================================
Deteksi kata kasar, hate speech, dan konten tidak pantas.
"""

import re
from typing import Dict, List


# Kata kasar / offensive words Bahasa Indonesia (censored for safety)
TOXIC_WORDS = [
    "bodoh", "goblok", "tolol", "idiot", "dungu",
    "brengsek", "bangsat", "bajingan", "keparat", "bedebah",
    "sialan", "setan", "iblis", "laknat",
    "babi", "anjing", "monyet", "binatang",  # when used as insults
    "sampah", "busuk", "mampus", "matilah",
    "bacot", "bego", "geblek", "kampret",
    "tai", "tahi", "najis",
    "pembunuh", "bunuh", "hajar", "bantai",
]

# Pola hate speech
HATE_PATTERNS = [
    r"semua\s+(orang\s+)?(suku|ras|agama|etnis)\s+\w+\s+(adalah|itu)\s+(jahat|bodoh|buruk)",
    r"(usir|basmi|habisi|musnahkan)\s+(semua\s+)?(orang\s+)?\w+",
    r"(bunuh|hajar|bantai)\s+(semua|mereka)",
]

# Ancaman
THREAT_PATTERNS = [
    r"(akan|mau|ingin)\s+(membunuh|menghancurkan|membakar|meledakkan)",
    r"(bom|senjata|racun)\s+(untuk|buat)\s+(membunuh|menyerang)",
]


class ToxicDetector:
    """Deteksi konten toxic dalam Bahasa Indonesia."""

    def __init__(self, custom_words: List[str] = None, sensitivity: str = "medium"):
        """
        Args:
            custom_words: Kata-kata tambahan yang dianggap toxic
            sensitivity: 'low', 'medium', 'high'
        """
        self.toxic_words = set(TOXIC_WORDS)
        if custom_words:
            self.toxic_words.update(custom_words)

        self.sensitivity = sensitivity
        self.hate_patterns = [re.compile(p, re.IGNORECASE) for p in HATE_PATTERNS]
        self.threat_patterns = [re.compile(p, re.IGNORECASE) for p in THREAT_PATTERNS]

    def check(self, text: str) -> Dict:
        """
        Cek teks untuk konten toxic.

        Returns:
            {
                "safe": bool,
                "score": float (0-1, higher = more toxic),
                "violations": list of violation details,
                "flagged_words": list of detected toxic words,
            }
        """
        text_lower = text.lower()
        violations = []
        flagged_words = []
        score = 0.0

        # 1. Check toxic words
        words = re.findall(r'\w+', text_lower)
        for word in words:
            if word in self.toxic_words:
                flagged_words.append(word)

        if flagged_words:
            word_score = min(len(flagged_words) * 0.2, 0.6)
            score += word_score
            violations.append({
                "type": "toxic_words",
                "severity": "medium" if len(flagged_words) <= 2 else "high",
                "detail": f"Kata tidak pantas terdeteksi: {', '.join(flagged_words[:5])}",
            })

        # 2. Check hate speech patterns
        for pattern in self.hate_patterns:
            match = pattern.search(text_lower)
            if match:
                score += 0.4
                violations.append({
                    "type": "hate_speech",
                    "severity": "high",
                    "detail": f"Pola hate speech terdeteksi",
                })
                break

        # 3. Check threats
        for pattern in self.threat_patterns:
            match = pattern.search(text_lower)
            if match:
                score += 0.5
                violations.append({
                    "type": "threat",
                    "severity": "critical",
                    "detail": f"Ancaman terdeteksi",
                })
                break

        # 4. Check excessive caps (shouting)
        if len(text) > 10:
            upper_ratio = sum(1 for c in text if c.isupper()) / len(text)
            if upper_ratio > 0.7:
                score += 0.1
                violations.append({
                    "type": "shouting",
                    "severity": "low",
                    "detail": "Terlalu banyak huruf kapital (kemungkinan berteriak)",
                })

        # Adjust by sensitivity
        threshold = {"low": 0.5, "medium": 0.3, "high": 0.1}[self.sensitivity]

        score = min(score, 1.0)

        return {
            "safe": score < threshold,
            "score": round(score, 2),
            "violations": violations,
            "flagged_words": flagged_words,
        }