domain/math_normalizer.py

# domain/math_normalizer.py

import re
import logging

logger = logging.getLogger(__name__)

class MathCanonicalizer:
    @staticmethod
    def preprocess_ocr_string(math_string: str) -> str:
        """
        V6 Polish (P0): Aggressively cleans raw OCR math strings before SymPy parsing.
        - Adds explicit multiplication.
        - Wraps trigonometric functions in parentheses (e.g., sin x -> sin(x)).
        - Normalizes basic artifacts.
        """
        if not math_string:
            return ""
            
        clean_str = str(math_string).strip()
        
        # 1. Standardize Whitespace
        clean_str = re.sub(r'\s+', ' ', clean_str)
        
        # 2. Add implicit multiplication between numbers and variables (e.g., 2x -> 2*x)
        # Note: SymPy handles some of this, but we do it explicitly to be safe
        clean_str = re.sub(r'(\d)([a-zA-Z])', r'\1*\2', clean_str)
        
        # 3. Trigonometric Functions Parentheses Guard
        # Matches sin, cos, tan, cot, arcsin, arccos, arctan, etc. followed by space and a variable/number
        # Example: 'sin x' -> 'sin(x)', 'cos 2x' -> 'cos(2*x)', 'tan ^2x' is trickier but we handle the basics.
        trig_funcs = r'(sin|cos|tan|cot|sec|csc|arcsin|arccos|arctan)'
        
        # Match 'sin x' or 'sin 2*x' but not 'sin(x)'
        # This regex looks for trig function, optional space, and then something that is not a parenthesis
        clean_str = re.sub(fr'{trig_funcs}\s+([a-zA-Z0-9_*]+)(?!\()', r'\1(\2)', clean_str)
        
        return clean_str