jk-rag-python/mineru/backend/pipeline/pipeline_middle_json_mkcontent.py

from loguru import logger

from mineru.utils.char_utils import full_to_half_exclude_marks, is_hyphen_at_line_end
from mineru.utils.config_reader import get_latex_delimiter_config
from mineru.backend.pipeline.para_split import ListLineTag
from mineru.utils.enum_class import BlockType, ContentType, MakeMode
from mineru.utils.language import detect_lang


def make_blocks_to_markdown(paras_of_layout,
                                      mode,
                                      img_buket_path='',
                                      ):
    page_markdown = []
    for para_block in paras_of_layout:
        para_text = ''
        para_type = para_block['type']
        if para_type in [BlockType.TEXT, BlockType.LIST, BlockType.INDEX]:
            para_text = merge_para_with_text(para_block)
        elif para_type == BlockType.TITLE:
            title_level = get_title_level(para_block)
            para_text = f'{"#" * title_level} {merge_para_with_text(para_block)}'
        elif para_type == BlockType.INTERLINE_EQUATION:
            if len(para_block['lines']) == 0 or len(para_block['lines'][0]['spans']) == 0:
                continue
            if para_block['lines'][0]['spans'][0].get('content', ''):
                para_text = merge_para_with_text(para_block)
            else:
                para_text += f"![]({img_buket_path}/{para_block['lines'][0]['spans'][0]['image_path']})"
        elif para_type == BlockType.IMAGE:
            if mode == MakeMode.NLP_MD:
                continue
            elif mode == MakeMode.MM_MD:
                # 检测是否存在图片脚注
                has_image_footnote = any(block['type'] == BlockType.IMAGE_FOOTNOTE for block in para_block['blocks'])
                # 如果存在图片脚注，则将图片脚注拼接到图片正文后面
                if has_image_footnote:
                    for block in para_block['blocks']:  # 1st.拼image_caption
                        if block['type'] == BlockType.IMAGE_CAPTION:
                            para_text += merge_para_with_text(block) + '  \n'
                    for block in para_block['blocks']:  # 2nd.拼image_body
                        if block['type'] == BlockType.IMAGE_BODY:
                            for line in block['lines']:
                                for span in line['spans']:
                                    if span['type'] == ContentType.IMAGE:
                                        if span.get('image_path', ''):
                                            para_text += f"![]({img_buket_path}/{span['image_path']})"
                    for block in para_block['blocks']:  # 3rd.拼image_footnote
                        if block['type'] == BlockType.IMAGE_FOOTNOTE:
                            para_text += '  \n' + merge_para_with_text(block)
                else:
                    for block in para_block['blocks']:  # 1st.拼image_body
                        if block['type'] == BlockType.IMAGE_BODY:
                            for line in block['lines']:
                                for span in line['spans']:
                                    if span['type'] == ContentType.IMAGE:
                                        if span.get('image_path', ''):
                                            para_text += f"![]({img_buket_path}/{span['image_path']})"
                    for block in para_block['blocks']:  # 2nd.拼image_caption
                        if block['type'] == BlockType.IMAGE_CAPTION:
                            para_text += '  \n' + merge_para_with_text(block)
        elif para_type == BlockType.TABLE:
            if mode == MakeMode.NLP_MD:
                continue
            elif mode == MakeMode.MM_MD:
                for block in para_block['blocks']:  # 1st.拼table_caption
                    if block['type'] == BlockType.TABLE_CAPTION:
                        para_text += merge_para_with_text(block) + '  \n'
                for block in para_block['blocks']:  # 2nd.拼table_body
                    if block['type'] == BlockType.TABLE_BODY:
                        for line in block['lines']:
                            for span in line['spans']:
                                if span['type'] == ContentType.TABLE:
                                    # if processed by table model
                                    if span.get('html', ''):
                                        para_text += f"\n{span['html']}\n"
                                    elif span.get('image_path', ''):
                                        para_text += f"![]({img_buket_path}/{span['image_path']})"
                for block in para_block['blocks']:  # 3rd.拼table_footnote
                    if block['type'] == BlockType.TABLE_FOOTNOTE:
                        para_text += '\n' + merge_para_with_text(block) + '  '

        if para_text.strip() == '':
            continue
        else:
            # page_markdown.append(para_text.strip() + '  ')
            page_markdown.append(para_text.strip())

    return page_markdown


latex_delimiters_config = get_latex_delimiter_config()

default_delimiters = {
    'display': {'left': '$$', 'right': '$$'},
    'inline': {'left': '$', 'right': '$'}
}

delimiters = latex_delimiters_config if latex_delimiters_config else default_delimiters

display_left_delimiter = delimiters['display']['left']
display_right_delimiter = delimiters['display']['right']
inline_left_delimiter = delimiters['inline']['left']
inline_right_delimiter = delimiters['inline']['right']

def merge_para_with_text(para_block):
    block_text = ''
    for line in para_block['lines']:
        for span in line['spans']:
            if span['type'] in [ContentType.TEXT]:
                span['content'] = full_to_half_exclude_marks(span['content'])
                block_text += span['content']
    block_lang = detect_lang(block_text)

    para_text = ''
    for i, line in enumerate(para_block['lines']):

        if i >= 1 and line.get(ListLineTag.IS_LIST_START_LINE, False):
            para_text += '  \n'

        for j, span in enumerate(line['spans']):

            span_type = span['type']
            content = ''
            if span_type == ContentType.TEXT:
                content = escape_special_markdown_char(span['content'])
            elif span_type == ContentType.INLINE_EQUATION:
                if span.get('content', ''):
                    content = f"{inline_left_delimiter}{span['content']}{inline_right_delimiter}"
            elif span_type == ContentType.INTERLINE_EQUATION:
                if span.get('content', ''):
                    content = f"\n{display_left_delimiter}\n{span['content']}\n{display_right_delimiter}\n"

            content = content.strip()

            if content:

                if span_type == ContentType.INTERLINE_EQUATION:
                    para_text += content
                    continue

                # 定义CJK语言集合(中日韩)
                cjk_langs = {'zh', 'ja', 'ko'}
                # logger.info(f'block_lang: {block_lang}, content: {content}')

                # 判断是否为行末span
                is_last_span = j == len(line['spans']) - 1

                if block_lang in cjk_langs: # 中文/日语/韩文语境下，换行不需要空格分隔,但是如果是行内公式结尾，还是要加空格
                    if is_last_span and span_type not in [ContentType.INLINE_EQUATION]:
                        para_text += content
                    else:
                        para_text += f'{content} '
                else:
                    # 西方文本语境下 每行的最后一个span判断是否要去除连字符
                    if span_type in [ContentType.TEXT, ContentType.INLINE_EQUATION]:
                        # 如果span是line的最后一个且末尾带有-连字符，那么末尾不应该加空格,同时应该把-删除
                        if (
                                is_last_span
                                and span_type == ContentType.TEXT
                                and is_hyphen_at_line_end(content)
                        ):
                            # 如果下一行的第一个span是小写字母开头，删除连字符
                            if (
                                    i + 1 < len(para_block['lines'])
                                    and para_block['lines'][i + 1].get('spans')
                                    and para_block['lines'][i + 1]['spans'][0].get('type') == ContentType.TEXT
                                    and para_block['lines'][i + 1]['spans'][0].get('content', '')
                                    and para_block['lines'][i + 1]['spans'][0]['content'][0].islower()
                            ):
                                para_text += content[:-1]
                            else:  # 如果没有下一行，或者下一行的第一个span不是小写字母开头，则保留连字符但不加空格
                                para_text += content
                        else:  # 西方文本语境下 content间需要空格分隔
                            para_text += f'{content} '
            else:
                continue

    return para_text


def make_blocks_to_content_list(para_block, img_buket_path, page_idx, page_size):
    para_type = para_block['type']
    para_content = {}
    if para_type in [
        BlockType.TEXT,
        BlockType.LIST,
        BlockType.INDEX,
    ]:
        para_content = {
            'type': ContentType.TEXT,
            'text': merge_para_with_text(para_block),
        }
    elif para_type == BlockType.DISCARDED:
        para_content = {
            'type': para_type,
            'text': merge_para_with_text(para_block),
        }
    elif para_type == BlockType.TITLE:
        para_content = {
            'type': ContentType.TEXT,
            'text': merge_para_with_text(para_block),
        }
        title_level = get_title_level(para_block)
        if title_level != 0:
            para_content['text_level'] = title_level
    elif para_type == BlockType.INTERLINE_EQUATION:
        if len(para_block['lines']) == 0 or len(para_block['lines'][0]['spans']) == 0:
            return None
        para_content = {
            'type': ContentType.EQUATION,
            'img_path': f"{img_buket_path}/{para_block['lines'][0]['spans'][0].get('image_path', '')}",
        }
        if para_block['lines'][0]['spans'][0].get('content', ''):
            para_content['text'] = merge_para_with_text(para_block)
            para_content['text_format'] = 'latex'
    elif para_type == BlockType.IMAGE:
        para_content = {'type': ContentType.IMAGE, 'img_path': '', BlockType.IMAGE_CAPTION: [], BlockType.IMAGE_FOOTNOTE: []}
        for block in para_block['blocks']:
            if block['type'] == BlockType.IMAGE_BODY:
                for line in block['lines']:
                    for span in line['spans']:
                        if span['type'] == ContentType.IMAGE:
                            if span.get('image_path', ''):
                                para_content['img_path'] = f"{img_buket_path}/{span['image_path']}"
            if block['type'] == BlockType.IMAGE_CAPTION:
                para_content[BlockType.IMAGE_CAPTION].append(merge_para_with_text(block))
            if block['type'] == BlockType.IMAGE_FOOTNOTE:
                para_content[BlockType.IMAGE_FOOTNOTE].append(merge_para_with_text(block))
    elif para_type == BlockType.TABLE:
        para_content = {'type': ContentType.TABLE, 'img_path': '', BlockType.TABLE_CAPTION: [], BlockType.TABLE_FOOTNOTE: []}
        for block in para_block['blocks']:
            if block['type'] == BlockType.TABLE_BODY:
                for line in block['lines']:
                    for span in line['spans']:
                        if span['type'] == ContentType.TABLE:
                            if span.get('html', ''):
                                para_content[BlockType.TABLE_BODY] = f"{span['html']}"

                            if span.get('image_path', ''):
                                para_content['img_path'] = f"{img_buket_path}/{span['image_path']}"

            if block['type'] == BlockType.TABLE_CAPTION:
                para_content[BlockType.TABLE_CAPTION].append(merge_para_with_text(block))
            if block['type'] == BlockType.TABLE_FOOTNOTE:
                para_content[BlockType.TABLE_FOOTNOTE].append(merge_para_with_text(block))

    page_width, page_height = page_size
    para_bbox = para_block.get('bbox')
    if para_bbox:
        x0, y0, x1, y1 = para_bbox
        para_content['bbox'] = [
            int(x0 * 1000 / page_width),
            int(y0 * 1000 / page_height),
            int(x1 * 1000 / page_width),
            int(y1 * 1000 / page_height),
        ]

    para_content['page_idx'] = page_idx

    return para_content


def union_make(pdf_info_dict: list,
               make_mode: str,
               img_buket_path: str = '',
               ):
    output_content = []
    for page_info in pdf_info_dict:
        paras_of_layout = page_info.get('para_blocks')
        paras_of_discarded = page_info.get('discarded_blocks')
        page_idx = page_info.get('page_idx')
        page_size = page_info.get('page_size')
        if make_mode in [MakeMode.MM_MD, MakeMode.NLP_MD]:
            if not paras_of_layout:
                continue
            page_markdown = make_blocks_to_markdown(paras_of_layout, make_mode, img_buket_path)
            output_content.extend(page_markdown)
        elif make_mode == MakeMode.CONTENT_LIST:
            para_blocks = (paras_of_layout or []) + (paras_of_discarded or [])
            if not para_blocks:
                continue
            for para_block in para_blocks:
                para_content = make_blocks_to_content_list(para_block, img_buket_path, page_idx, page_size)
                if para_content:
                    output_content.append(para_content)

    if make_mode in [MakeMode.MM_MD, MakeMode.NLP_MD]:
        return '\n\n'.join(output_content)
    elif make_mode == MakeMode.CONTENT_LIST:
        return output_content
    else:
        logger.error(f"Unsupported make mode: {make_mode}")
        return None


def get_title_level(block):
    title_level = block.get('level', 1)
    if title_level > 4:
        title_level = 4
    elif title_level < 1:
        title_level = 0
    return title_level


def escape_special_markdown_char(content):
    """
    转义正文里对markdown语法有特殊意义的字符
    """
    special_chars = ["*", "`", "~", "$"]
    for char in special_chars:
        content = content.replace(char, "\\" + char)

    return content