# problem_understanding.py - V231.14
# Analyzes problem structure BEFORE solving

"""
Problem Understanding Module
Ensures we understand WHAT is being asked before attempting to solve.
"""

import json
import re
from utils.safe_json import safe_extract_json  # V1.0: Canonical extractor


def get_problem_understanding_prompt(ocr_text: str, data_anchor: dict) -> str:
    """
    Prompt LLM to analyze problem structure.
    
    Returns JSON with:
    - problem_type
    - sub_questions (all parts א, ב, ג, etc.)
    - solving_order
    - dependencies
    """
    return f"""
ANALYZE this math problem structure. DO NOT SOLVE - only understand what is being asked.

Problem Text:
{ocr_text}

Extracted Data:
{json.dumps(data_anchor, ensure_ascii=False, indent=2)}

Your task: Identify ALL parts of this problem and create a solving plan.

Return JSON:
{{
  "problem_type": "CIRCLE_EQUATION | LINE_EQUATION | GEOMETRIC_LOCUS | DERIVATIVE_QUOTIENT | etc.",
  "main_question": "Brief description of main question",
  "sub_questions": [
    {{
      "id": "א",
      "question": "Full text of sub-question א",
      "requires": ["center", "radius"],
      "specific_values": ["m=2"],
      "expected_output": "equation | number | point | etc.",
      "topic": "CIRCLE_EQUATION"
    }},
    {{
      "id": "ב",
      "question": "Full text of sub-question ב",
      "requires": ["equation_from_א", "point"],
      "specific_values": ["a=1"],
      "expected_output": "line_equation",
      "topic": "LINE_TANGENT"
    }}
  ],
  "solving_order": ["א", "ב", "ג"],
  "dependencies": {{
    "ב": ["א"],
    "ג": ["א"]
  }}
}}

CRITICAL RULES:
1. Include ALL sub-questions (א, ב, ג, ד, etc.)
2. **CHRONOLOGICAL DATA ISOLATION (V310.0 - CRITICAL):** If a specific value (e.g., a=1, m=2) is mentioned ONLY in a specific sub-question, you MUST include it in the `specific_values` array for THAT sub-question only. NEVER put it in the top-level anchor if it's not global. This prevents data leakage (e.g., using a=1 from section ב' to solve section א').
3. **EXCEPTION:** If the problem asks for a **Geometric Locus (מקום גיאומטרי)**:
   - This is a SINGLE QUESTION (even if it looks long).
   - Set `problem_type` = "GEOMETRIC_LOCUS".
   - Create ONLY ONE sub-question (id="א") containing the entire text.
4. Identify dependencies (ב needs א's result)
5. Determine topic for EACH sub-question
6. DO NOT solve - only analyze structure

Return ONLY valid JSON.
"""


def parse_understanding(response_text: str) -> dict:
    """Parse LLM understanding response using canonical safe_extract_json."""
    result = safe_extract_json(response_text, caller="PROBLEM_UNDERSTANDING")
    # If parsing failed, return a minimal fallback so the pipeline continues
    if isinstance(result, dict) and result.get("logic_error"):
        return {
            "problem_type": "UNKNOWN",
            "sub_questions": [],
            "solving_order": [],
            "dependencies": {}
        }
    return result



def validate_understanding(understanding: dict) -> bool:
    """Validate understanding structure."""
    required_keys = ["problem_type", "sub_questions", "solving_order"]
    
    if not all(key in understanding for key in required_keys):
        return False
    
    if not isinstance(understanding["sub_questions"], list):
        return False
    
    if len(understanding["sub_questions"]) == 0:
        return False
    
    # Validate each sub-question
    for sq in understanding["sub_questions"]:
        if not all(key in sq for key in ["id", "question", "topic"]):
            return False
    
    return True

def enforce_locus_rule(understanding: dict, ocr_text: str) -> dict:
    """
    V260.2: Hard Rule - If 'מקום גיאומטרי' exists, FORCE Locus type.
    """
    if any(k in ocr_text for k in ["מקום גיאומטרי", "Locus", "מצא את המקום", "המקום הגיאומטרי"]):
        print("🛡️ [BIT-LOG] Hard Logic: Detected 'Geometric Locus' - Forcing Single Question structure.")
        return {
            "problem_type": "GEOMETRIC_LOCUS",
            "main_question": understanding.get("main_question", "Find the Locus"),
            "sub_questions": [{
                "id": "א",
                "question": ocr_text, # Give the WHOLE text to the single sub-question
                "requires": [],
                "expected_output": "equation",
                "topic": "GEOMETRIC_LOCUS"
            }],
            "solving_order": ["א"],
            "dependencies": {}
        }
    return understanding


# ==================== USAGE EXAMPLE ====================

if __name__ == "__main__":
    # Test understanding prompt
    ocr = """
    מעגל עם משוואה x² + y² = 12
    א. מצא את משוואת המעגל
    ב. מצא משיק למעגל בנקודה A
    ג. מצא את שטח המעגל
    """
    
    data = {
        "function_equations": ["x^2 + y^2 = 12"],
        "points": ["A"],
        "specific_values": [],
        "constraints": []
    }
    
    prompt = get_problem_understanding_prompt(ocr, data)
    print(prompt)