从Amazon Textract提取表格文本与Bounding Box时遇错求助

你遇到的TypeError是因为：

• 你将table_blocks处理成了仅包含TABLE类型块的列表，但后续代码里用ROW/CELL的Id（字符串）去索引这个列表——列表只能用整数索引，自然触发报错。
• 更关键的是，ROW、CELL类型的块根本不在这个过滤后的table_blocks列表里，你需要从Textract返回的所有Blocks中通过Id查找这些块。

import boto3

# Initialize the Textract client
client = boto3.client('textract')

with open('table_document.pdf', 'rb') as file:
    # Call Amazon Textract to analyze the document
    response = client.analyze_document(Document={'Bytes': file.read()}, FeatureTypes=['TABLES'])

# 把所有Blocks转换成字典，用Id作为键，方便快速查找任意块
blocks_dict = {block['Id']: block for block in response['Blocks']}
# 过滤出所有TABLE类型的块
table_blocks = [b for b in response['Blocks'] if b['BlockType'] == 'TABLE']

# Iterate over each table block
for table_block in table_blocks:
    # 获取当前表格关联的所有ROW块Id
    row_blocks_ids = table_block['Relationships'][0]['Ids']
    # 从字典中取出ROW块，并按Top坐标从上到下排序
    row_blocks = [blocks_dict[row_id] for row_id in row_blocks_ids]
    row_blocks.sort(key=lambda row: row['Geometry']['BoundingBox']['Top'])
    
    # Iterate over each row block
    for row_block in row_blocks:
        # 获取当前行关联的所有CELL块Id
        cell_blocks_ids = row_block['Relationships'][0]['Ids']
        # 从字典中取出CELL块，并按Left坐标从左到右排序
        cell_blocks = [blocks_dict[cell_id] for cell_id in cell_blocks_ids]
        cell_blocks.sort(key=lambda cell: cell['Geometry']['BoundingBox']['Left'])
        
        # Iterate over each cell block
        for cell_block in cell_blocks:
            # 获取单元格文本和 bounding box
            cell_text = cell_block['Text']
            box = cell_block['Geometry']['BoundingBox']
            # 打印结果
            print(f'{cell_text}: {box}')

• 新增blocks_dict：将所有返回的Block按Id映射，彻底解决通过Id查找块的问题。
• 排序ROW/CELL时，直接从字典中取出对应的块对象，避免用字符串索引列表的错误。
• 调整ROW和CELL的处理逻辑，先取出完整的块对象再排序，代码更直观易读。

内容的提问来源于stack exchange，提问作者InterestingPenguin80