documentExtractionag051/ExtractDocument

1"""2validation_utils.py3===================4 5Utility functions for validation data operations.6 7This module provides helper functions for working with validation data structures.8"""9 10from typing import Dict, Any11import logging12 13logger = logging.getLogger(__name__)14 15 16def extract_page_dimensions(data: Dict[str, Any]) -> Dict[int, tuple]:17    """18    Extract page dimensions from input data.19    20    Supports both simplified schema and original engine output format.21    22    Args:23        data (dict): Input data containing page information.24    25    Returns:26        dict: {page_num: (width, height)}27    """28    # Check for simplified schema29    if "pages" in data and isinstance(data.get("pages"), list):30        pages = data["pages"]31        # Check if pages have pageNum (simplified schema)32        if pages and "pageNum" in pages[0]:33            return {34                p.get("pageNum", idx + 1): (p.get("width", 0), p.get("height", 0))35                for idx, p in enumerate(pages)36            }37    38    # Check for engine output format39    if "engineData" in data:40        pages = (41            data.get("engineData", {})42            .get("imagePreprocessingResult", {})43            .get("pages", [])44        )45        return {46            idx + 1: (p.get("width", 0), p.get("height", 0))47            for idx, p in enumerate(pages)48        }49    50    logger.warning("Could not extract page dimensions from input data.")51    return {}52 53 54def check_empty_values(data_dict: Dict[str, Any]) -> bool:55    """56    Check if all 'value' fields in a dictionary are empty strings.57    58    Args:59        data_dict (dict): Dictionary with nested value fields.60    61    Returns:62        bool: True if all values are empty, False otherwise.63    """64    return all(65        isinstance(v, dict) and v.get("value", "") == ""66        for v in data_dict.values()67    )68 69 70def calculate_cumulative_heights(page_dimensions: Dict[int, tuple]) -> Dict[int, int]:71    """72    Calculate cumulative page heights for multi-page documents.73    74    Args:75        page_dimensions (dict): {page_num: (width, height)}76    77    Returns:78        dict: {page_num: cumulative_height_offset}79    """80    cumulative_heights = {}81    cumulative_height = 082    83    for page_num in sorted(page_dimensions.keys()):84        cumulative_heights[page_num] = cumulative_height85        _, height = page_dimensions[page_num]86        cumulative_height += height87    88    return cumulative_heights89