FORMAT_CONVERSION_MAXCOMPUTE/format_convert/convert_docx.py

import os
import sys
from collections import defaultdict

sys.path.append(os.path.dirname(__file__) + "/../")
from format_convert.convert_tree import _Document, _Sentence, _Page, _Image, _Table
import re
import traceback
import xml
import zipfile
import docx
from bs4 import BeautifulSoup
from format_convert.utils import judge_error_code, add_div, get_logger, log, memory_decorator, get_garble_code, \
    get_table_html
from format_convert.wrapt_timeout_decorator import timeout
from format_convert.convert_image import ImageConvert
from format_convert.convert_need_interface import from_tika_interface


def docx2text():
    return


def read_rel_image(document_xml_rels):
    if not document_xml_rels:
        return {}

    # 获取映射文件里的关系 Id-Target
    image_rel_dict = {}
    for rel in document_xml_rels:
        if 'Relationship' in str(rel):
            _id = rel.get("Id")
            _target = rel.get("Target")
            _type = rel.get("Type")
            if 'image' in _type:
                image_rel_dict[_id] = _target
    return image_rel_dict


def read_no_start(numbering_xml):
    """
    读取编号组的起始值

    :return:
    """
    if not numbering_xml:
        return {}, {}

    # 获取虚拟-真实id映射关系
    w_num_list = numbering_xml.getElementsByTagName("w:num")
    abstract_real_id_dict = {}
    for w_num in w_num_list:
        w_num_id = w_num.getAttribute("w:numId")
        w_abstract_num_id = w_num.getElementsByTagName('w:abstractNumId')[0].getAttribute("w:val")
        abstract_real_id_dict[w_abstract_num_id] = w_num_id

    # 获取虚拟id的开始编号
    w_abstract_num_list = numbering_xml.getElementsByTagName("w:abstractNum")
    abstract_id_level_dict = {}
    abstract_id_level_text_dict = {}
    for w_abstract_num in w_abstract_num_list:
        w_abstract_num_id = w_abstract_num.getAttribute("w:abstractNumId")
        w_lvl_list = w_abstract_num.getElementsByTagName("w:lvl")
        level_start_dict = {}
        level_text_dict = {}
        for w_lvl in w_lvl_list:
            w_ilvl_value = w_lvl.getAttribute('w:ilvl')
            if w_lvl.getElementsByTagName("w:start"):
                w_ilvl_start_num = w_lvl.getElementsByTagName("w:start")[0].getAttribute("w:val")
                level_start_dict[int(w_ilvl_value)] = int(w_ilvl_start_num)
            if w_lvl.getElementsByTagName("w:lvlText") and w_lvl.getElementsByTagName("w:numFmt"):
                w_lvl_text = w_lvl.getElementsByTagName("w:lvlText")[0].getAttribute("w:val")
                w_lvl_format = w_lvl.getElementsByTagName("w:numFmt")[0].getAttribute("w:val")
                if w_lvl_format == 'upperLetter':
                    w_lvl_text = re.sub('%\d', '%A', w_lvl_text)
                elif w_lvl_format == 'lowerLetter':
                    w_lvl_text = re.sub('%\d', '%a', w_lvl_text)
                level_text_dict[int(w_ilvl_value)] = w_lvl_text
        abstract_id_level_dict[w_abstract_num_id] = level_start_dict
        abstract_id_level_text_dict[w_abstract_num_id] = level_text_dict

    # 映射回真实id
    real_id_level_start_dict = {}
    for abstract_id in abstract_real_id_dict.keys():
        real_id = abstract_real_id_dict.get(abstract_id)
        level_start_dict = abstract_id_level_dict.get(abstract_id)
        if level_start_dict:
            real_id_level_start_dict[int(real_id)] = level_start_dict

    real_id_level_text_dict = {}
    for abstract_id in abstract_real_id_dict.keys():
        real_id = abstract_real_id_dict.get(abstract_id)
        level_text_dict = abstract_id_level_text_dict.get(abstract_id)
        if level_text_dict:
            real_id_level_text_dict[int(real_id)] = level_text_dict

    return real_id_level_start_dict, real_id_level_text_dict


def read_p_text(unique_type_dir, p_node, _last_node_level, _num_pr_dict, numbering_xml, document_xml_rels,
                is_sdt=False):
    """
    读取w:p下的文本，包括编号

    :param unique_type_dir:
    :param p_node:
    :param _last_node_level:
    :param _num_pr_dict:
    :param numbering_xml:
    :param document_xml_rels:
    :param is_sdt:
    :return:
    """
    _text_list = []
    _order_list = []

    # 文本的编号（如果有编号的话）
    text_no = ''

    # 获取编号组的起始值和编号组的展示形式
    id_level_start_dict, id_level_text_dict = read_no_start(numbering_xml)
    # print('_num_pr_dict', _num_pr_dict)

    # 提取编号 组-层级-序号
    num_pr = p_node.getElementsByTagName("w:numPr")
    if num_pr:
        num_pr = num_pr[0]
        if num_pr.getElementsByTagName("w:numId"):
            group_id = int(num_pr.getElementsByTagName("w:numId")[0].getAttribute("w:val"))
            if group_id >= 1:
                node_level = num_pr.getElementsByTagName("w:ilvl")
                if node_level:
                    node_level = int(node_level[0].getAttribute("w:val"))
                    # print('group_id', group_id, 'node_level', node_level, 'last_node_level', _last_node_level)
                    if group_id in _num_pr_dict.keys():
                        # if node_level == 0 and node_level not in _num_pr_dict[group_id].keys():
                        if node_level == 0 and _num_pr_dict.get(group_id) and node_level not in _num_pr_dict.get(group_id).keys():
                            _num_pr_dict[group_id][node_level] = 1
                        if _last_node_level != 0 and node_level < _last_node_level:
                            # print('重置', 'group_id', group_id, 'last_node_level', last_node_level)
                            # 需循环重置node_level到last_node_level之间的level
                            for l in range(node_level+1, _last_node_level+1):
                                _num_pr_dict[group_id][l] = 0
                            if _num_pr_dict[group_id].get(node_level):
                                _num_pr_dict[group_id][node_level] += 1
                            else:
                                pass
                                # print('group_id, node_level', group_id, node_level)
                        # elif node_level in _num_pr_dict[group_id].keys():
                        elif node_level in _num_pr_dict.get(group_id).keys():
                            _num_pr_dict[group_id][node_level] += 1
                        else:
                            _num_pr_dict[group_id][node_level] = 1
                    else:
                        _num_pr_dict[group_id] = {node_level: 1}
                    # print(num_pr_dict[group_id])
                    for level in range(node_level+1):
                        # 当前level下有多少个node
                        # if level not in _num_pr_dict[group_id]:
                        if level not in _num_pr_dict.get(group_id):
                            # if level not in id_level_start_dict[group_id]:
                            if not id_level_start_dict.get(group_id) or level not in id_level_start_dict.get(group_id):
                                continue
                            else:
                                level_node_cnt = id_level_start_dict[group_id][level]
                        else:
                            level_node_cnt = _num_pr_dict[group_id][level]

                        if id_level_start_dict.get(group_id) and _num_pr_dict.get(group_id) and id_level_start_dict.get(group_id).get(level) and _num_pr_dict.get(group_id).get(level):
                            start_no = id_level_start_dict.get(group_id).get(level)
                            level_node_cnt += start_no - 1

                        level_text = None
                        if id_level_text_dict.get(group_id) and id_level_text_dict.get(group_id).get(level) and _num_pr_dict.get(group_id).get(level):
                            level_text = id_level_text_dict.get(group_id).get(level)
                        # print('level_node_cnt', level_node_cnt)
                        if level_text:
                            if re.search('a', level_text):
                                level_node_cnt = chr(ord('a') + level_node_cnt - 1)
                                text_no += re.sub('%a', str(level_node_cnt), level_text)
                            elif re.search('A', level_text):
                                level_node_cnt = chr(ord('A') + level_node_cnt - 1)
                                text_no += re.sub('%A', str(level_node_cnt), level_text)
                            else:
                                text_no += re.sub('%\d', str(level_node_cnt), level_text)
                        else:
                            text_no += str(level_node_cnt) + '.'
                        # print('text_no', text_no)
                    _last_node_level = node_level

    # text = p_node.getElementsByTagName("w:t")
    # picture = p_node.getElementsByTagName("wp:docPr")
    # if text:
    #     _order_list.append("w:t")
    #     temp_text = ""
    #     if is_sdt and len(text) == 2:
    #         if len(text[0].childNodes) > 0 and len(text[1].childNodes) > 0:
    #             temp_text += text[0].childNodes[0].nodeValue + '.'*20 + text[1].childNodes[0].nodeValue
    #     else:
    #         for t in text:
    #             if len(t.childNodes) > 0:
    #                 temp_text += t.childNodes[0].nodeValue
    #             else:
    #                 continue
    #     if text_no:
    #         temp_text = text_no + ' ' + temp_text
    #     _text_list.append(temp_text)
    # # 只有序号
    # elif len(text_no) >= 2:
    #     _text_list.append(text_no[:-1])
    #
    # if picture:
    #     _order_list.append("wp:docPr")
    #
    # for line1 in p_node.childNodes:
    #     if "w:r" in str(line1):
    #         picture1 = line1.getElementsByTagName("w:pict")
    #         if picture1:
    #             _order_list.append("wp:docPr")

    p_node_text = ''
    has_html = False
    # 编号先加上
    if text_no:
        p_node_text += text_no
    text = p_node.getElementsByTagName("w:t")
    # 目录页单特殊生成
    if is_sdt and len(text) == 2:
        p_node_text += text[0].childNodes[0].nodeValue + '.'*20 + text[1].childNodes[0].nodeValue
    # 正常页面
    else:
        image_rel_dict = read_rel_image(document_xml_rels)
        p_node_all = p_node.getElementsByTagName("*")
        for node in p_node_all:
            # 文本
            if "w:t" in str(node).split(' '):
                if node.childNodes:
                    p_node_text += node.childNodes[0].nodeValue

            # 图片，提前识别，不做成Image对象放入Page了
            elif "a:blip" in str(node).split(' '):
                _id = node.getAttribute("r:embed")
                image_path = image_rel_dict.get(_id)
                if image_path:
                    image_path = unique_type_dir + 'word/' + image_path
                    image_convert = ImageConvert(image_path, '')
                    image_html = image_convert.get_html()[0]
                    if isinstance(image_html, int):
                        image_html = ''
                    p_node_text += image_html
                    has_html = True

    # 只有编号
    if len(p_node_text) > 0 and p_node_text == text_no:
        p_node_text = p_node_text[:-1]

    _text_list.append(p_node_text)
    if has_html:
        _order_list.append('w:t html')
    else:
        _order_list.append('w:t')
    return _text_list, _order_list, _num_pr_dict, _last_node_level


@timeout(50, timeout_exception=TimeoutError)
def read_xml_order(unique_type_dir, document_xml, numbering_xml, document_xml_rels):
    log("into read_xml_order")
    try:
        body = document_xml.getElementsByTagName("w:body")[0]
        order_list = []
        text_list = []
        # 编号组记录
        num_pr_dict = {}
        last_node_level = 0
        for line in body.childNodes:
            # 普通文本
            if "w:p" in str(line):
                t_list, o_list, num_pr_dict, last_node_level = read_p_text(unique_type_dir,
                                                                           line,
                                                                           last_node_level,
                                                                           num_pr_dict,
                                                                           numbering_xml,
                                                                           document_xml_rels)
                text_list += t_list
                order_list += o_list

            # 目录索引
            elif "w:sdt" in str(line):
                sdt = line
                for sdt_child in sdt.childNodes:
                    if "w:sdtContent" in str(sdt_child):
                        sdt_content = sdt_child
                        for sdt_content_child in sdt_content.childNodes:
                            if 'w:p' in str(sdt_content_child):
                                t_list, o_list, num_pr_dict, last_node_level = read_p_text(unique_type_dir,
                                                                                           sdt_content_child,
                                                                                           last_node_level,
                                                                                           num_pr_dict,
                                                                                           numbering_xml,
                                                                                           document_xml_rels,
                                                                                           is_sdt=True)
                                text_list += t_list
                                order_list += o_list

            elif "w:tbl" in str(line):
                order_list.append("w:tbl")
        # read_xml_table(path, save_path)
        return [order_list, text_list]
    except Exception as e:
        log("read_xml_order error!")
        traceback.print_exc()
        return [-1]


@timeout(50, timeout_exception=TimeoutError)
def read_xml_table(unique_type_dir, document_xml, numbering_xml, document_xml_rels):
    def recursion_read_table(table, show=0):
        table_text = '<table border="1">'
        tr_index = 0
        tr_text_list = []
        last_node_level = 0
        num_pr_dict = {}

        # 直接子节点用child表示，所有子节点用all表示
        row_span_dict = {}
        for table_child in table.childNodes:
            if 'w:tr' in str(table_child):
                table_text += "<tr>"
                tr = table_child
                tr_child_nodes = tr.childNodes
                tc_index = 0
                tc_text_list = []
                for tr_child in tr_child_nodes:
                    if 'w:tc' in str(tr_child).split(' '):
                        tc_text = ""
                        tc = tr_child
                        # 获取一格占多少列，相当于colspan
                        col_span = tc.getElementsByTagName("w:gridSpan")
                        if col_span:
                            col_span = int(col_span[0].getAttribute("w:val"))
                        else:
                            col_span = 1
                        # 获取是否是合并单元格的下一个空单元格，相当于rowspan
                        is_merge = tc.getElementsByTagName("w:vMerge")
                        if is_merge:

                            is_merge = is_merge[0].getAttribute("w:val")
                            # print(tr_index, tc_index, is_merge)
                            # print('row_span_dict', row_span_dict)
                            if is_merge == "continue":
                                row_span_dict[tc_index][0] += 1
                                tc_index += col_span
                                tc_text_list.append([tc_text, col_span])
                                # 跳过，不增加td
                                continue
                                # col_span_index = 0
                                # real_tc_index = 0
                                # if 0 <= tr_index - 1 < len(tr_text_list):
                                #     for tc_colspan in tr_text_list[tr_index - 1]:
                                #         if col_span_index < tc_index:
                                #             col_span_index += tc_colspan[1]
                                #             real_tc_index += 1
                                #     if real_tc_index < len(tr_text_list[tr_index - 1]):
                                #         tc_text = tr_text_list[tr_index - 1][real_tc_index][0]
                            else:
                                # 先结束上一次同列的合并单元格
                                if tc_index in row_span_dict:
                                    row_span, finish_row_span_flag = row_span_dict.get(tc_index)
                                    table_text = re.sub(finish_row_span_flag, str(row_span), table_text)
                                # 开启新的合并单元格
                                row_span_flag = '#@#_{}_{}'.format(tr_index, tc_index)
                                row_span_dict[tc_index] = [1, row_span_flag]
                        else:
                            row_span_flag = 1

                        # 设置colspan
                        table_text = table_text + "<td rowspan={} colspan={}>".format(row_span_flag, col_span)
                        # 放入文本
                        tc_child_nodes = tc.childNodes
                        for tc_child in tc_child_nodes:
                            if 'w:tbl' in str(tc_child).split(' '):
                                # 嵌套在tc中的表格
                                tc_text += recursion_read_table(tc_child)
                            if 'w:p' in str(tc_child).split(' '):
                                tc_p_all_nodes = tc_child.getElementsByTagName("*")
                                _t_list, _, num_pr_dict, last_node_level = read_p_text(unique_type_dir,
                                                                                       tc_child,
                                                                                       last_node_level,
                                                                                       num_pr_dict,
                                                                                       numbering_xml,
                                                                                       document_xml_rels)
                                # print('_t_list', _t_list)
                                tc_text += ''.join(_t_list)
                                # for tc_p_all in tc_p_all_nodes:
                                #     if 'w:t' in str(tc_p_all).split(' '):
                                #         # w:t必须加childNodes[0]才能读文本
                                #         tc_text += tc_p_all.childNodes[0].nodeValue
                        # print('tc_text', tc_text)
                        # 结束该tc
                        table_text = table_text + tc_text + "</td>"
                        tc_index += col_span
                        tc_text_list.append([tc_text, col_span])
                # 结束该tr
                table_text += "</tr>"
                tr_index += 1
                tr_text_list.append(tc_text_list)

        if show:
            for row in tr_text_list:
                print('row', row)
                print('len(row)', len(row))

        # 替换所有row_span
        for key in row_span_dict.keys():
            row_span, finish_row_span_flag = row_span_dict.get(key)
            table_text = re.sub(finish_row_span_flag, str(row_span), table_text)

        # 结束该table
        table_text += "</table>"
        return table_text

    log("into read_xml_table")
    try:
        body = document_xml.getElementsByTagName("w:body")[0]
        table_text_list = []
        body_nodes = body.childNodes
        for node in body_nodes:
            if 'w:tbl' in str(node).split(' '):
                _table = node
                # _table_text = recursion_read_table(_table)
                _table_text = xml_table_to_html(_table, unique_type_dir, numbering_xml, document_xml_rels)
                table_text_list.append(_table_text)
        return table_text_list

    except Exception as e:
        log("read_xml_table error")
        print("read_xml_table", traceback.print_exc())
        return [-1]


def xml_table_to_html(table, unique_type_dir, numbering_xml, document_xml_rels, show=0):
    tr_index = 0
    tr_text_list = []
    last_node_level = 0
    num_pr_dict = {}

    # 直接子节点用child表示，所有子节点用all表示
    for table_child in table.childNodes:
        if 'w:tr' in str(table_child):
            tr = table_child
            tr_child_nodes = tr.childNodes
            tc_index = 0
            tc_text_list = []
            for tr_child in tr_child_nodes:
                if 'w:tc' in str(tr_child).split(' '):
                    tc_text = ""
                    tc = tr_child
                    # 获取一格占多少列，相当于colspan
                    col_span = tc.getElementsByTagName("w:gridSpan")
                    if col_span:
                        col_span = int(col_span[0].getAttribute("w:val"))
                    else:
                        col_span = 1
                    # 获取是否是合并单元格的下一个空单元格，相当于rowspan
                    is_merge = tc.getElementsByTagName("w:vMerge")
                    if is_merge:
                        is_merge = is_merge[0].getAttribute("w:val")
                        if is_merge == "continue":
                            tc_index += col_span
                            tc_text = '@continue@'
                            tc_text_list.append([tc_text, col_span])
                            # 跳过，不增加td
                            continue

                    # 放入文本
                    tc_child_nodes = tc.childNodes
                    for tc_child in tc_child_nodes:
                        # 处理嵌套在tc中的表格
                        if 'w:tbl' in str(tc_child).split(' '):
                            tc_text += xml_table_to_html(tc_child, unique_type_dir, numbering_xml, document_xml_rels)
                        # 处理编号
                        if 'w:p' in str(tc_child).split(' '):
                            _t_list, _, num_pr_dict, last_node_level = read_p_text(unique_type_dir,
                                                                                   tc_child,
                                                                                   last_node_level,
                                                                                   num_pr_dict,
                                                                                   numbering_xml,
                                                                                   document_xml_rels)
                            tc_text += ''.join(_t_list)
                    # 结束该tc
                    tc_index += col_span
                    tc_text_list.append([tc_text, col_span])
            # 结束该tr
            tr_index += 1
            tr_text_list.append(tc_text_list)

    if show:
        for row in tr_text_list:
            print('row', row)
            print('len(row)', len(row))

    table_html = row_list_to_table(tr_text_list)
    return table_html


def row_list_to_table(row_list, show=0):
    if show:
        print('='*50)

    # 复制合并列
    new_row_list = []
    for row in row_list:
        new_row = []
        for col, col_span in row:
            new_row += [[col, col_span]]
            if col_span > 1:
                new_row += [[col, 0]] * (col_span - 1)
        new_row_list.append(new_row)
    row_list = new_row_list

    if show:
        for row in row_list:
            print('copy row', row)

    # 计算是不是每行都有相等列数
    row_cnt_list = []
    for row in row_list:
        row_cnt_list.append(len(row))

    if len(set(row_cnt_list)) != 1:
        log('表格有列数不同，直接返回text' + str(row_cnt_list))
        # 直接返回所有col的text
        text = ''
        for row in row_list:
            for col, col_span in row:
                text += col
        return text

    new_row_list = []
    for ri, row in enumerate(row_list):
        new_row = []
        for ci, col in enumerate(row):
            col, col_span = col
            row_span = 1
            # 判断下面行同列有没有需合并的
            for ri2 in range(ri+1, len(row_list)):
                col2, col_span2 = row_list[ri2][ci]
                if col2 == '@continue@':
                    row_span += 1
                else:
                    break

            # 需跳过的列
            if col == '@continue@' or col_span == 0:
                delete = 1
            else:
                delete = 0

            col_dict = {
                'text': col,
                'rowspan': row_span,
                'columnspan': col_span,
                'delete': delete,
            }
            new_row.append(col_dict)
        new_row_list.append(new_row)

    if show:
        for new_row in new_row_list:
            print('new_row', new_row)

    table_html = get_table_html(new_row_list)

    # soup = BeautifulSoup(table_html, 'lxml')
    # print(soup.prettify())
    if show:
        print('-' * 50)
    return table_html


@timeout(25, timeout_exception=TimeoutError)
def parse_xml(path):
    # 解析xml
    DOMTree = xml.dom.minidom.parse(path)
    collection = DOMTree.documentElement
    return collection


@timeout(25, timeout_exception=TimeoutError)
def parse_xml2(path):
    # 解析xml
    tree = xml.etree.ElementTree.parse(path)
    root = tree.getroot()
    return root


class DocxConvert:
    def __init__(self, path, unique_type_dir):
        self._doc = _Document(path)
        self._page = _Page(None, 0)
        self.path = path
        self.unique_type_dir = unique_type_dir

        # 解压docx
        try:
            f = zipfile.ZipFile(path)
            for file in f.namelist():
                if "word/" in str(file):
                    f.extract(file, self.unique_type_dir)

            f.close()
        except Exception as e:
            log("docx format error!")
            self._doc.error_code = [-3]

        # 读取内容
        try:
            self.document_xml = parse_xml(self.unique_type_dir + "word/document.xml")

            if os.path.exists(self.unique_type_dir + "word/numbering.xml"):
                self.numbering_xml = parse_xml(self.unique_type_dir + "word/numbering.xml")
            else:
                self.numbering_xml = []

            if os.path.exists(self.unique_type_dir + "word/_rels/document.xml.rels"):
                self.document_xml_rels = parse_xml2(self.unique_type_dir + "word/_rels/document.xml.rels")
            else:
                self.document_xml_rels = []
        except FileNotFoundError:
            # 找不到解压文件，就用html格式读
            log('FileNotFoundError')
            self._doc.error_code = None
        except TimeoutError:
            log("parse_xml timeout")
            self._doc.error_code = [-4]

    @memory_decorator
    def init_package(self):
        # 各个包初始化
        try:
            self.docx = docx.Document(self.path)
            self.zip = zipfile.ZipFile(self.path)
        except:
            log("cannot open docx!")
            traceback.print_exc()
            self._doc.error_code = [-3]

    def convert(self):
        # 先判断特殊doc文件，可能是html文本
        is_html_doc = False
        try:
            with open(self.path, 'r') as f:
                html_str = f.read()
            if re.search('<div|<html|<body|<head|<tr|<br|<table|<td', html_str):
                soup = BeautifulSoup(html_str, 'lxml')
                text = soup.text
                is_html_doc = True
        except:
            pass

        if is_html_doc:
            _sen = _Sentence(text, (0, 0, 0, 0))
            self._page.add_child(_sen)
            self._doc.add_child(self._page)
            return

        self.init_package()
        if self._doc.error_code is not None:
            return

        order_and_text_list = self.get_orders()
        if judge_error_code(order_and_text_list):
            self._doc.error_code = order_and_text_list
            return
        order_list, text_list = order_and_text_list

        # 乱码返回文件格式错误
        match1 = re.findall(get_garble_code(), ''.join(text_list))
        if len(match1) > 10:
            log("doc/docx garbled code!")
            self._doc.error_code = [-3]
            # _sen = _Sentence('文件乱码！', (0, 0, 0, 0))
            # self._page.add_child(_sen)
            self._doc.add_child(self._page)
            return

        # test
        # for i in range(len(text_list)):
        #     print(order_list[i], text_list[i])

        table_list = self.get_tables()
        if judge_error_code(table_list):
            self._doc.error_code = table_list
            return

        # paragraph_list = self.get_paragraphs()

        image_list = self.get_images()

        order_y = 0
        doc_pr_cnt = 0
        for tag in order_list:
            bbox = (0, order_y, 0, 0)
            if tag == "w:t html":
                if len(text_list) > 0:
                    _para = text_list.pop(0)
                    _sen = _Sentence(_para, bbox)
                    _sen.combine = False
                    _sen.is_html = True
                    self._page.add_child(_sen)

            if tag == "w:t":
                if len(text_list) > 0:
                    _para = text_list.pop(0)
                    _sen = _Sentence(_para, bbox)
                    _sen.combine = False
                    self._page.add_child(_sen)

            if tag == "wp:docPr":
                if len(image_list) > 0:
                    temp_image_path = self.unique_type_dir + "docpr" + str(doc_pr_cnt) + ".png"
                    _image = image_list.pop(0)
                    with open(temp_image_path, "wb") as f:
                        f.write(_image)
                    _img = _Image(_image, temp_image_path, bbox)
                    _img.is_from_docx = True
                    self._page.add_child(_img)
                    doc_pr_cnt += 1

            if tag == "w:tbl":
                if len(table_list) > 0:
                    _table = table_list.pop(0)
                    _table = _Table(_table, bbox)
                    _table.is_html = True
                    self._page.add_child(_table)
            order_y += 1

        if self._doc.error_code is None and self._page.error_code is not None:
            self._doc.error_code = self._page.error_code
        self._doc.add_child(self._page)

    @memory_decorator
    def get_tables(self):
        # 遍历表
        table_list = read_xml_table(self.unique_type_dir, self.document_xml, self.numbering_xml, self.document_xml_rels)
        return table_list

    def get_images(self):
        # 顺序遍历图片
        image_list = []
        pattern = re.compile('rId\d+')
        for graph in self.docx.paragraphs:
            for run in graph.runs:
                if run.text == '':
                    try:
                        if not pattern.search(run.element.xml):
                            continue
                        content_id = pattern.search(run.element.xml).group(0)
                        content_type = self.docx.part.related_parts[content_id].content_type
                    except Exception as e:
                        print("docx no image!", e)
                        continue
                    if not content_type.startswith('image'):
                        continue

                    img_data = self.docx.part.related_parts[content_id].blob
                    if img_data is not None:
                        image_list.append(img_data)
        return image_list

    @memory_decorator
    def get_orders(self):
        # 解析document.xml，获取文字顺序
        order_and_text_list = read_xml_order(self.unique_type_dir, self.document_xml, self.numbering_xml, self.document_xml_rels)
        return order_and_text_list

    def get_doc_object(self):
        return self._doc

    def use_tika(self, _path):
        # 调用tika提取
        # html = from_tika_interface(self.path)
        # if judge_error_code(html):
        #     self._doc.error_code = html
        # self.tika_html = html
        data = from_tika_interface(_path)
        if judge_error_code(data):
            self._doc.error_code = data
            return
        current_y = 5
        for di, d in enumerate(data):
            data_type, value = d
            bbox = [0, current_y, 20, current_y+10]
            current_y += 20
            if data_type == 'text':
                _sen = _Sentence(value, bbox)
                _sen.combine = False
                self._page.add_child(_sen)
            elif data_type == 'img':
                with open(value, "rb") as f:
                    img = f.read()
                _img = _Image(img, value, bbox)
                _img.is_from_docx = True
                self._page.add_child(_img)
            elif data_type == 'table':
                _table = _Table(value, bbox)
                _table.is_html = True
                self._page.add_child(_table)
        self._doc.add_child(self._page)

    def get_html(self):
        if self._doc.error_code is not None:
            return self._doc.error_code
        try:
            # raise
            self.convert()
        except:
            traceback.print_exc()
            self._doc.error_code = [-1]
        # log('docx error code ' + str(self._doc.error_code))
        if self._doc.error_code is not None:
            # # 调用tika提取
            # html = from_tika_interface(self.path)
            # if judge_error_code(html):
            #     self._doc.error_code = html
            #     return self._doc.error_code
            # else:
            #     return [html]
            try:
                self.use_tika(self.path)
                self._doc.error_code = None
            except:
                traceback.print_exc()
                log('docx tika failed too')
                self._doc.error_code = [-17]
        return self._doc.get_html()


class DocxConvertNew:
    # 解压 .docx 文件
    def unzip_docx(self, file_path, extract_to):
        with zipfile.ZipFile(file_path, 'r') as zip_ref:
            zip_ref.extractall(extract_to)

    # 解析 numbering.xml 文件，获取编号信息
    def parse_numbering(self, file_path):
        numbering = defaultdict(list)
        dom = xml.dom.minidom.parse(file_path)
        root = dom.documentElement
        for num in root.getElementsByTagName("w:num"):
            num_id = num.getAttribute("w:numId")
            for lvl in num.getElementsByTagName("w:lvl"):
                lvl_index = lvl.getAttribute("w:ilvl")
                num_fmt = lvl.getElementsByTagName("w:numFmt")[0].getAttribute("w:val")
                num_text = lvl.getElementsByTagName("w:numText")[0].getAttribute("w:val") if lvl.getElementsByTagName("w:numText") else None
                numbering[num_id].append((lvl_index, num_fmt, num_text))
        return numbering

    # 解析 document.xml.rels 文件，获取图片引用信息
    def parse_rels(self, file_path):
        rels = {}
        dom = xml.dom.minidom.parse(file_path)
        root = dom.documentElement
        for rel in root.getElementsByTagName("Relationship"):
            rel_id = rel.getAttribute("Id")
            rel_type = rel.getAttribute("Type")
            target = rel.getAttribute("Target")
            rels[rel_id] = {"type": rel_type, "target": target}
        return rels

    # 解析 document.xml 文件，获取文档内容
    def parse_document(self, file_path, numbering, rels):
        dom = xml.dom.minidom.parse(file_path)
        root = dom.documentElement
        paragraphs = root.getElementsByTagName("w:p")
        content = []

        for para in paragraphs:
            para_text = ""
            num_id = None
            ilvl = None
            for child in para.childNodes:
                if child.nodeName == "w:pPr":
                    for num_id_node in child.getElementsByTagName("w:numId"):
                        num_id = num_id_node.getAttribute("w:val")
                    for ilvl_node in child.getElementsByTagName("w:ilvl"):
                        ilvl = ilvl_node.getAttribute("w:ilvl")
                elif child.nodeName == "w:r":
                    for t in child.getElementsByTagName("w:t"):
                        para_text += t.firstChild.nodeValue if t.firstChild else ""

            if num_id and ilvl not in [None, '']:
                num_fmt, num_text = numbering[num_id][int(ilvl)][1:]
                if num_fmt == "decimal":
                    para_text = f"{int(ilvl) + 1}. {para_text}"
                elif num_text:
                    para_text = f"{num_text} {para_text}"

            content.append(para_text)

        # 解析表格
        tables = root.getElementsByTagName("w:tbl")
        for table in tables:
            table_content = []
            row_count = 0
            col_count = 0
            for row in table.getElementsByTagName("w:tr"):
                row_content = []
                cell_count = 0
                for cell in row.getElementsByTagName("w:tc"):
                    cell_text = ""
                    for para in cell.getElementsByTagName("w:p"):
                        for run in para.getElementsByTagName("w:r"):
                            for text in run.getElementsByTagName("w:t"):
                                cell_text += text.firstChild.nodeValue if text.firstChild else ""

                    # 检查合并单元格
                    grid_span = 1
                    v_merge = False
                    for child in cell.childNodes:
                        if child.nodeName == "w:tcPr":
                            for grid_span_node in child.getElementsByTagName("w:gridSpan"):
                                grid_span = int(grid_span_node.getAttribute("w:val"))
                            for v_merge_node in child.getElementsByTagName("w:vMerge"):
                                v_merge = True

                    row_content.append({
                        "text": cell_text,
                        "colspan": grid_span,
                        "rowspan": 1 if not v_merge else 2  # 简化处理，实际需要根据上下文确定
                    })
                    cell_count += grid_span
                table_content.append(row_content)
                row_count += 1
                col_count = max(col_count, cell_count)
            content.append(table_content)

        # 解析图片
        for rel in rels.values():
            if rel["type"] == "http://schemas.openxmlformats.org/officeDocument/2006/relationships/image":
                content.append(f"图片: {rel['target']}")

        return content

    # 生成 HTML 输出
    def generate_html(self, content):
        html = []
        html.append('<!DOCTYPE HTML><head><meta charset="UTF-8"></head><html><body>')
        for item in content:
            if isinstance(item, list):  # 表格内容
                html.append("<table border='1'>")
                for row in item:
                    html.append("<tr>")
                    for cell in row:
                        colspan = cell.get("colspan", 1)
                        rowspan = cell.get("rowspan", 1)
                        html.append(f"<td colspan='{colspan}' rowspan='{rowspan}'>{cell['text']}</td>")
                    html.append("</tr>")
                html.append("</table>")
            else:  # 普通文本或图片
                html.append(f"<p>{item}</p>")
        html.append("</body></html>")
        return "\n".join(html)

    # 主函数
    def read_docx(self, file_path):
        extract_to = "extracted_docx"
        self.unzip_docx(file_path, extract_to)

        numbering = self.parse_numbering(os.path.join(extract_to, "word", "numbering.xml"))
        rels = self.parse_rels(os.path.join(extract_to, "word", "_rels", "document.xml.rels"))
        content = self.parse_document(os.path.join(extract_to, "word", "document.xml"), numbering, rels)

        html_output = self.generate_html(content)
        with open("../result.html", "w", encoding="utf-8") as f:
            f.write(html_output)


if __name__ == '__main__':
    _p = r'C:/Users/Administrator/Downloads/1723004790329.docx'
    # _p = "C:/Users/Administrator/Desktop/test_doc/error14.docx"
    save_dir = r"D:\Project\format_conversion_maxcompute\format_convert\temp" + '/'
    c = DocxConvert(_p, save_dir)
    _html = c.get_html()
    with open('../result.html', 'w', encoding='utf-8') as f:
        f.write('<!DOCTYPE HTML><head><meta charset="UTF-8"></head>' + str(_html[0]))