FORMAT_CONVERSION_MAXCOMPUTE/format_convert/utils.py

import os
import sys
sys.path.append(os.path.dirname(__file__) + "/../")
import difflib
import logging
import mimetypes
import platform
import re
import traceback
import filetype
from bs4 import BeautifulSoup
from pdfminer.layout import *


def judge_error_code(_list, code=[0, -1, -2, -3, -4, -5, -6, -7, -8]):
    """
    [0] : continue
    [-1]: 逻辑处理错误
    [-2]: 接口调用错误
    [-3]: 文件格式错误，无法打开
    [-4]: 各类文件调用第三方包读取超时
    [-5]: 整个转换过程超时
    [-6]: 阿里云UDF队列超时
    [-7]: 文件需密码，无法打开
    [-8]: 调用现成接口报错
    """
    for c in code:
        if _list == [c]:
            return True
    return False


def add_div(text):
    if text == "" or text is None:
        return text

    # if get_platform() == "Windows":
    #     print("add_div", text)
    if re.findall("<div>", text):
        return text

    text = "<div>" + text + "\n"
    text = re.sub("\n", "</div>\n<div>", text)
    # text += "</div>"
    if text[-5:] == "<div>":
        print("add_div has cut", text[-30:])
        text = text[:-5]
    return text


def get_platform():
    sys = platform.system()
    return sys


def get_html_p(html_path):
    logging.info("into get_html_p")
    try:
        with open(html_path, "r") as ff:
            html_str = ff.read()

        soup = BeautifulSoup(html_str, 'lxml')
        text = ""
        for p in soup.find_all("p"):
            p_text = p.text
            p_text = p_text.strip()
            if p.string != "":
                text += p_text
        text += "\n"
        return text
    except Exception as e:
        logging.info("get_html_p error!")
        print("get_html_p", traceback.print_exc())
        return [-1]


def string_similarity(str1, str2):
    # 去掉<div>和回车
    str1 = re.sub("<div>", "", str1)
    str1 = re.sub("</div>", "", str1)
    str1 = re.sub("\n", "", str1)
    str2 = re.sub("<div>", "", str2)
    str2 = re.sub("</div>", "", str2)
    str2 = re.sub("\n", "", str2)
    # print("********************************")
    # print("str1", str1)
    # print("********************************")
    # print("str2", str2)
    # print("********************************")
    score = difflib.SequenceMatcher(None, str1, str2).ratio()
    print("string_similarity", score)
    return score


def get_sequential_data(text_list, bbox_list, html=False):
    logging.info("into get_sequential_data")
    try:
        text = ""
        order_list = []
        for i in range(len(text_list)):
            length_start = bbox_list[i][0][0]
            length_end = bbox_list[i][1][0]
            height_start = bbox_list[i][0][1]
            height_end = bbox_list[i][-1][1]
            # print([length_start, length_end, height_start, height_end])
            order_list.append([text_list[i], length_start, length_end, height_start, height_end])
            # text = text + infomation['text'] + "\n"

        if get_platform() == "Windows":
            print("get_sequential_data", order_list)
        if not order_list:
            if get_platform() == "Windows":
                print("get_sequential_data", "no order list")
            return ""

        # 根据bbox的坐标对输出排序
        order_list.sort(key=lambda x: (x[3], x[1]))

        # 根据bbox分行分列
        # col_list = []
        # height_end = int((order_list[0][4] + order_list[0][3]) / 2)
        # for i in range(len(order_list)):
        #     if height_end - threshold <= order_list[i][3] <= height_end + threshold:
        #         col_list.append(order_list[i])
        #     else:
        #         row_list.append(col_list)
        #         col_list = []
        #         height_end = int((order_list[i][4] + order_list[i][3]) / 2)
        #         col_list.append(order_list[i])
        #     if i == len(order_list) - 1:
        #         row_list.append(col_list)

        row_list = []
        used_box = []
        threshold = 5
        for box in order_list:
            if box in used_box:
                continue

            height_center = (box[4] + box[3]) / 2
            row = []
            for box2 in order_list:
                if box2 in used_box:
                    continue
                height_center2 = (box2[4] + box2[3]) / 2
                if height_center - threshold <= height_center2 <= height_center + threshold:
                    if box2 not in row:
                        row.append(box2)
                        used_box.append(box2)
            row.sort(key=lambda x: x[0])
            row_list.append(row)

        for row in row_list:
            if not row:
                continue
            if len(row) <= 1:
                text = text + row[0][0] + "\n"
            else:
                sub_text = ""
                row.sort(key=lambda x: x[1])
                for col in row:
                    sub_text = sub_text + col[0] + " "
                sub_text = sub_text + "\n"
                text += sub_text

        if html:
            text = "<div>" + text
            text = re.sub("\n", "</div>\n<div>", text)
            text += "</div>"
            # if text[-5:] == "<div>":
            #     text = text[:-5]
        return text

    except Exception as e:
        logging.info("get_sequential_data error!")
        print("get_sequential_data", traceback.print_exc())
        return [-1]


# def get_formatted_table(text_list, text_bbox_list, table_bbox_list, split_line):
#     logging.info("into get_formatted_table")
#     try:
#         # 重新定义text_bbox_list，[point, point, text]
#         text_bbox_list = [[text_bbox_list[i][0], text_bbox_list[i][2], text_list[i]] for i in
#                           range(len(text_bbox_list))]
#         # 按纵坐标排序
#         text_bbox_list.sort(key=lambda x: (x[0][1], x[0][0]))
#         table_bbox_list.sort(key=lambda x: (x[0][1], x[0][0]))
#
#         # print("text_bbox_list", text_bbox_list)
#         # print("table_bbox_list", table_bbox_list)
#
#         # bbox位置 threshold
#         threshold = 5
#
#         # 根据split_line分区，可能有个区多个表格 [(), ()]
#         area_text_bbox_list = []
#         area_table_bbox_list = []
#         # print("get_formatted_table, split_line", split_line)
#         for j in range(1, len(split_line)):
#             last_y = split_line[j - 1][0][1]
#             current_y = split_line[j][0][1]
#             temp_text_bbox_list = []
#             temp_table_bbox_list = []
#
#             # 找出该区域下text bbox
#             for text_bbox in text_bbox_list:
#                 # 计算 text bbox 中心点
#                 text_bbox_center = ((text_bbox[1][0] + text_bbox[0][0]) / 2,
#                                     (text_bbox[1][1] + text_bbox[0][1]) / 2)
#                 if last_y - threshold <= text_bbox_center[1] <= current_y + threshold:
#                     temp_text_bbox_list.append(text_bbox)
#             area_text_bbox_list.append(temp_text_bbox_list)
#
#             # 找出该区域下table bbox
#             for table_bbox in table_bbox_list:
#                 # 计算 table bbox 中心点
#                 table_bbox_center = ((table_bbox[1][0] + table_bbox[0][0]) / 2,
#                                      (table_bbox[1][1] + table_bbox[0][1]) / 2)
#                 if last_y < table_bbox_center[1] < current_y:
#                     temp_table_bbox_list.append(table_bbox)
#             area_table_bbox_list.append(temp_table_bbox_list)
#
#         # for j in range(len(area_text_bbox_list)):
#         #     print("area_text_bbox_list", j, area_text_bbox_list[j])
#
#         # 对每个区域分别进行两个bbox匹配，生成表格
#         area_text_list = []
#         area_column_list = []
#         for j in range(len(area_text_bbox_list)):
#             # 每个区域的table bbox 和text bbox
#             temp_table_bbox_list = area_table_bbox_list[j]
#             temp_text_bbox_list = area_text_bbox_list[j]
#
#             # 判断该区域有无表格bbox
#             # 若无表格，将该区域文字连接
#             if not temp_table_bbox_list:
#                 # 找出该区域的所有text bbox
#                 only_text_list = []
#                 only_bbox_list = []
#                 for text_bbox in temp_text_bbox_list:
#                     only_text_list.append(text_bbox[2])
#                     only_bbox_list.append([text_bbox[0], text_bbox[1]])
#                 only_text = get_sequential_data(only_text_list, only_bbox_list, True)
#                 if only_text == [-1]:
#                     return [-1], [-1]
#                 area_text_list.append(only_text)
#                 area_column_list.append(0)
#                 continue
#
#             # 有表格
#             # 文本对应的表格格子
#             text_in_table = {}
#             for i in range(len(temp_text_bbox_list)):
#                 text_bbox = temp_text_bbox_list[i]
#
#                 # 计算 text bbox 中心点
#                 text_bbox_center = ((text_bbox[1][0] + text_bbox[0][0]) / 2,
#                                     (text_bbox[1][1] + text_bbox[0][1]) / 2)
#
#                 # 判断中心点在哪个table bbox中
#                 for table_bbox in temp_table_bbox_list:
#                     # 中心点在table bbox中，将text写入字典
#                     if table_bbox[0][0] <= text_bbox_center[0] <= table_bbox[1][0] and \
#                             table_bbox[0][1] <= text_bbox_center[1] <= table_bbox[1][1]:
#                         if str(table_bbox) in text_in_table.keys():
#                             text_in_table[str(table_bbox)] = text_in_table.get(str(table_bbox)) + text_bbox[2]
#                         else:
#                             text_in_table[str(table_bbox)] = text_bbox[2]
#                         break
#
#                     # 如果未找到text bbox匹配的table bbox，加大threshold匹配
#                     # elif (table_bbox[0][0] <= text_bbox_center[0]+threshold <= table_bbox[1][0] and
#                     #         table_bbox[0][1] <= text_bbox_center[1]+threshold <= table_bbox[1][1]) or \
#                     #         (table_bbox[0][0] <= text_bbox_center[0]-threshold <= table_bbox[1][0] and
#                     #          table_bbox[0][1] <= text_bbox_center[1]-threshold <= table_bbox[1][1]) or \
#                     #         (table_bbox[0][0] <= text_bbox_center[0]+threshold <= table_bbox[1][0] and
#                     #          table_bbox[0][1] <= text_bbox_center[1]-threshold <= table_bbox[1][1]) or \
#                     #         (table_bbox[0][0] <= text_bbox_center[0]-threshold <= table_bbox[1][0] and
#                     #          table_bbox[0][1] <= text_bbox_center[1]+threshold <= table_bbox[1][1]):
#                     #     if str(table_bbox) in text_in_table.keys():
#                     #         text_in_table[str(table_bbox)] = text_in_table.get(str(table_bbox)) + text_bbox[2]
#                     #     else:
#                     #         text_in_table[str(table_bbox)] = text_bbox[2]
#                     #     break
#
#             # 对表格格子进行分行分列，并计算总计多少小列
#             # 放入坐标
#             all_col_list = []
#             all_row_list = []
#             for i in range(len(temp_table_bbox_list)):
#                 table_bbox = temp_table_bbox_list[i]
#
#                 # 放入所有坐标x
#                 if table_bbox[0][0] not in all_col_list:
#                     all_col_list.append(table_bbox[0][0])
#                 if table_bbox[1][0] not in all_col_list:
#                     all_col_list.append(table_bbox[1][0])
#
#                 # 放入所有坐标y
#                 if table_bbox[0][1] not in all_row_list:
#                     all_row_list.append(table_bbox[0][1])
#                 if table_bbox[1][1] not in all_row_list:
#                     all_row_list.append(table_bbox[1][1])
#             all_col_list.sort(key=lambda x: x)
#             all_row_list.sort(key=lambda x: x)
#
#             # 分行
#             row_list = []
#             rows = []
#             temp_table_bbox_list.sort(key=lambda x: (x[0][1], x[0][0], x[1][1], x[1][0]))
#             y_row = temp_table_bbox_list[0][0][1]
#             for i in range(len(temp_table_bbox_list)):
#                 table_bbox = temp_table_bbox_list[i]
#
#                 if y_row - threshold <= table_bbox[0][1] <= y_row + threshold:
#                     rows.append(table_bbox)
#                 else:
#                     y_row = table_bbox[0][1]
#                     if rows:
#                         rows.sort(key=lambda x: x[0][0])
#                         row_list.append(rows)
#                     rows = []
#                     rows.append(table_bbox)
#                 # print("*" * 30)
#                 # print(row_list)
#
#                 if i == len(temp_table_bbox_list) - 1:
#                     if rows:
#                         rows.sort(key=lambda x: x[0][0])
#                         row_list.append(rows)
#
#             # 生成表格，包括文字和格子宽度
#             area_column = []
#             text = '<table border="1">' + "\n"
#             for row in row_list:
#                 text += "<tr>" + "\n"
#                 for col in row:
#                     # 计算bbox y坐标之间有多少其他点，+1即为所占行数
#                     row_span = 1
#                     for y in all_row_list:
#                         if col[0][1] < y < col[1][1]:
#                             if y - col[0][1] >= 2 and col[1][1] - y >= 2:
#                                 row_span += 1
#
#                     # 计算bbox x坐标之间有多少其他点，+1即为所占列数
#                     col_span = 1
#                     for x in all_col_list:
#                         if col[0][0] < x < col[1][0]:
#                             if x - col[0][0] >= 2 and col[1][0] - x >= 2:
#                                 col_span += 1
#
#                     text += "<td colspan=" + str(col_span) + " rowspan=" + str(row_span) + ">"
#
#                     if str(col) in text_in_table.keys():
#                         text += text_in_table.get(str(col))
#                     else:
#                         text += ""
#                     text += "</td>" + "\n"
#                 text += "</tr>" + "\n"
#             text += "</table>" + "\n"
#
#             # 计算最大column
#             max_col_num = 0
#             for row in row_list:
#                 col_num = 0
#                 for col in row:
#                     col_num += 1
#                 if max_col_num < col_num:
#                     max_col_num = col_num
#
#             area_text_list.append(text)
#             area_column_list.append(max_col_num)
#
#         text = ""
#         if get_platform() == "Windows":
#             print("get_formatted_table area_text_list", area_text_list)
#         for area_text in area_text_list:
#             text += area_text
#         return text, area_column_list
#     except Exception as e:
#         logging.info("get_formatted_table error!")
#         print("get_formatted_table", traceback.print_exc())
#         return [-1], [-1]


def rename_inner_files(root_path):
    try:
        logging.info("into rename_inner_files")
        # 获取解压文件夹下所有文件+文件夹，不带根路径
        path_list = []
        for root, dirs, files in os.walk(root_path, topdown=False):
            for name in dirs:
                p = os.path.join(root, name) + os.sep
                if get_platform() == "Windows":
                    root_path = slash_replace(root_path)
                    p = slash_replace(p)
                    p = re.sub(root_path, "", p)
                    root_path = slash_replace(root_path, True)
                    p = slash_replace(p, True)
                else:
                    p = re.sub(root_path, "", p)
                path_list.append(p)
            for name in files:
                p = os.path.join(root, name)
                if get_platform() == "Windows":
                    root_path = slash_replace(root_path)
                    p = slash_replace(p)
                    p = re.sub(root_path, "", p)
                    root_path = slash_replace(root_path, True)
                    p = slash_replace(p, True)
                else:
                    p = re.sub(root_path, "", p)
                path_list.append(p)

        # 按路径长度排序
        path_list.sort(key=lambda x: len(x), reverse=True)

        # 循环改名
        for old_path in path_list:
            # 按路径分隔符分割
            ss = old_path.split(os.sep)
            # 判断是否文件夹
            is_dir = 0
            file_type = ""
            if os.path.isdir(root_path + old_path):
                ss = ss[:-1]
                is_dir = 1
            else:
                if "." in old_path:
                    file_type = "." + old_path.split(".")[-1]
                else:
                    file_type = ""

            # 最后一级需要用hash改名
            new_path = ""
            # new_path = re.sub(ss[-1], str(hash(ss[-1])), old_path) + file_type
            current_level = 0
            for s in ss:
                # 路径拼接
                if current_level < len(ss) - 1:
                    new_path += s + os.sep
                else:
                    new_path += str(hash(s)) + file_type
                current_level += 1

            new_ab_path = root_path + new_path
            old_ab_path = root_path + old_path
            os.rename(old_ab_path, new_ab_path)

        # 重新获取解压文件夹下所有文件+文件夹
        new_path_list = []
        for root, dirs, files in os.walk(root_path, topdown=False):
            for name in dirs:
                new_path_list.append(os.path.join(root, name) + os.sep)
            for name in files:
                new_path_list.append(os.path.join(root, name))
        return new_path_list
    except:
        traceback.print_exc()
        return [-1]


def judge_format(path):
    guess1 = mimetypes.guess_type(path)
    _type = None
    if guess1[0]:
        _type = guess1[0]
    else:
        guess2 = filetype.guess(path)
        if guess2:
            _type = guess2.mime

    if _type == "application/pdf":
        return "pdf"
    if _type == "application/vnd.openxmlformats-officedocument.wordprocessingml.document":
        return "docx"
    if _type == "application/x-zip-compressed" or _type == "application/zip":
        return "zip"
    if _type == "application/x-rar-compressed" or _type == "application/rar":
        return "rar"
    if _type == "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet":
        return "xlsx"
    if _type == "application/msword":
        return "doc"
    if _type == "image/png":
        return "png"
    if _type == "image/jpeg":
        return "jpg"

    # 猜不到，返回None
    return None


def slash_replace(_str, reverse=False):
    if reverse:
        _str = eval(repr(_str).replace('/', '\\\\'))
    else:
        _str = eval(repr(_str).replace('\\\\', '/'))
    return _str


class LineTable:
    def recognize_table(self,list_textbox, list_line,sourceP_LB=True):
        self.list_line = list_line
        self.list_crosspoints = self.recognize_crosspoints(list_line)

  
        # 聚类
        cluster_crosspoints = []
        for _point in self.list_crosspoints:
            cluster_crosspoints.append({"lines": _point.get("lines"), "points": [_point]})
        while 1:
            _find = False
            new_cluster_crosspoints = []
            for l_point in cluster_crosspoints:
                _flag = False
                for l_n_point in new_cluster_crosspoints:
                    line1 = l_point.get("lines")
                    line2 = l_n_point.get("lines")
                    if len(line1&line2) > 0:
                        _find = True
                        _flag = True
                        l_n_point["lines"] = line1.union(line2)
                        l_n_point["points"].extend(l_point["points"])
                if not _flag:
                    new_cluster_crosspoints.append({"lines":l_point.get("lines"),"points":l_point.get("points")})
            cluster_crosspoints = new_cluster_crosspoints
            if not _find:
                break

        list_l_rect = []
        for table_crosspoint in cluster_crosspoints:
            list_rect = self.crosspoint2rect(table_crosspoint.get("points"))
            list_l_rect.append(list_rect)

        in_objs = set()
        list_tables = []
        for l_rect in list_l_rect:
            _ta = self.rect2table(list_textbox,l_rect,in_objs,sourceP_LB=sourceP_LB)
            if _ta:
                list_tables.append(_ta)
        self._plot(list_line, list_textbox)
        return list_tables, in_objs, list_l_rect

    def recognize_table_by_rect(self, list_textbox, list_rect, margin=2):

        dump_margin = 5
        list_rect_tmp = []
        # 去重
        for _rect in list_rect:
            if (_rect.bbox[3]-_rect.bbox[1] < 10) or (abs(_rect.bbox[2]-_rect.bbox[0]) < 5):
                continue
            _find = False
            for _tmp in list_rect_tmp:
                for i in range(4):
                    if abs(_rect.bbox[i]-_tmp.bbox[i]) < dump_margin:
                        pass
                    else:
                        _find = False
                        break
                    if i == 3:
                        _find = True
                if _find:
                    break
            if not _find:
                list_rect_tmp.append(_rect)

        # print("=====",len(list_rect),len(list_rect_tmp))
        # print(list_rect_tmp)
        # from matplotlib import pyplot as plt
        # plt.figure()
        # for _rect in list_rect_tmp:
        #     x0,y0,x1,y1 = _rect.bbox
        #     plt.boxplot(_rect.bbox)
        # plt.show()

        cluster_rect = []
        for _rect in list_rect:
            _find = False
            for cr in cluster_rect:
                for cr_rect in cr:
                    if abs((cr_rect.bbox[2]-cr_rect.bbox[0]+_rect.bbox[2]-_rect.bbox[0])-(max(cr_rect.bbox[2],_rect.bbox[2])-min(cr_rect.bbox[0],_rect.bbox[0])))<margin:
                        _find = True
                        cr.append(_rect)
                        break
                    elif abs((cr_rect.bbox[3]-cr_rect.bbox[1]+_rect.bbox[3]-_rect.bbox[1])-(max(cr_rect.bbox[3],_rect.bbox[3])-min(cr_rect.bbox[1],_rect.bbox[1])))<margin:
                        _find = True
                        cr.append(_rect)
                        break
                if _find:
                    break
            if not _find:
                cluster_rect.append([_rect])

        list_l_rect = cluster_rect

        in_objs = set()
        list_tables = []
        for l_rect in list_l_rect:
            _ta = self.rect2table(list_textbox,l_rect,in_objs)
            if _ta:
                list_tables.append(_ta)
        return list_tables,in_objs,list_l_rect

    def recognize_crosspoints(self, list_line,fixLine=True):

        list_crosspoints = []
        # print("lines num",len(list_line))

        def getMaxPoints(list_x,margin=5,reverse=False):
            clust_x = []
            for _x in list_x:
                _find = False
                for cx in clust_x:
                    if abs(cx[0]-_x)<margin:
                        _find = True
                        cx.append(_x)
                        break
                if not _find:
                    clust_x.append([_x])
            clust_x.sort(key=lambda x:x,reverse=reverse)
            return clust_x[0][0],len(clust_x[0])

        for _i in range(len(list_line)):
            for _j in range(len(list_line)):
                line1 = list_line[_i].__dict__.get("bbox")
                line2 = list_line[_j].__dict__.get("bbox")
                exists,point = self.cross_point(line1,line2)
                if exists:
                    list_crosspoints.append(point)


        if fixLine:
            #聚类
            cluster_crosspoints = []
            for _point in list_crosspoints:
                cluster_crosspoints.append({"lines":_point.get("lines"),"points":[_point]})
            while 1:
                _find = False
                new_cluster_crosspoints = []
                for l_point in cluster_crosspoints:
                    _flag = False
                    for l_n_point in new_cluster_crosspoints:
                        line1 = l_point.get("lines")
                        line2 = l_n_point.get("lines")
                        if len(line1&line2)>0:
                            _find = True
                            _flag = True
                            l_n_point["lines"] = line1.union(line2)
                            l_n_point["points"].extend(l_point["points"])

                    if not _flag:
                        new_cluster_crosspoints.append({"lines":l_point.get("lines"),"points":l_point.get("points")})
                cluster_crosspoints = new_cluster_crosspoints
                if not _find:
                    break

            list_crosspoints = []

            for list_cp in cluster_crosspoints:
                points = list_cp.get("points")

                l_lines = []
                for p in points:
                    l_lines.extend(p.get("p_lines"))
                l_lines = list(set(l_lines))
                l_lines.sort(key=lambda x:x[0])

                min_x,_count = getMaxPoints([l[0] for l in l_lines],reverse=False)
                if _count<=2:
                    min_x = None


                min_y,_count = getMaxPoints([l[1] for l in l_lines],reverse=False)
                if _count<2:
                    min_y = None


                max_x,_count = getMaxPoints([l[2] for l in l_lines],reverse=True)
                if _count<=2:
                    max_x = None


                max_y,_count = getMaxPoints([l[3] for l in l_lines],reverse=True)
                if _count<=2:
                    max_y = None
                if min_x and min_y and max_x and max_y:

                    points.sort(key=lambda x:x["point"][0])
                    if abs(min_x-points[0]["point"][0])>30:
                        _line = LTLine(1,(min_x,min_y),(min_x,max_y))
                        list_line.append(_line)
                        l_lines.append(_line.bbox)
                        print("add=====",_line.bbox)


                    if abs(max_x-points[-1]["point"][0])>30:
                        _line = LTLine(1,(max_x,min_y),(max_x,max_y))
                        list_line.append()
                        l_lines.append(_line.bbox)
                        print("add=====1",_line.bbox)

                    points.sort(key=lambda x:x["point"][1])
                    if abs(min_y-points[0]["point"][1])>30:
                        _line = LTLine(1,(min_x,min_y),(max_x,min_y))
                        list_line.append(_line)
                        l_lines.append(_line.bbox)
                        print("add=====2",_line.bbox)

                    if abs(max_y-points[-1]["point"][1])>30:
                        _line = LTLine(1,(min_x,max_y),(max_x,max_y))
                        list_line.append(_line)
                        l_lines.append(_line.bbox)
                        print("add=====2",_line.bbox)



                for _i in range(len(l_lines)):
                    for _j in range(len(l_lines)):
                        line1 = l_lines[_i]
                        line2 = l_lines[_j]
                        exists,point = self.cross_point(line1,line2)
                        if exists:
                            list_crosspoints.append(point)
                from matplotlib import pyplot as plt
                plt.figure()
                for _line in l_lines:
                    x0,y0,x1,y1 = _line
                    plt.plot([x0,x1],[y0,y1])
                for point in list_crosspoints:
                    plt.scatter(point.get("point")[0],point.get("point")[1])
                plt.show()

        # from matplotlib import pyplot as plt
        # plt.figure()
        # for _line in list_line:
        #     x0,y0,x1,y1 = _line.__dict__.get("bbox")
        #     plt.plot([x0,x1],[y0,y1])
        # for _line in list_line:
        #     x0,y0,x1,y1 = _line.bbox
        #     plt.plot([x0,x1],[y0,y1])
        # for point in list_crosspoints:
        #     plt.scatter(point.get("point")[0],point.get("point")[1])
        # plt.show()

        # print(list_crosspoints)
        # print("points num",len(list_crosspoints))
        return list_crosspoints

    def recognize_rect(self, _page):
        list_line = []
        for _obj in _page._objs:
            if isinstance(_obj, (LTLine)):
                list_line.append(_obj)
        list_crosspoints = self.recognize_crosspoints(list_line)

        #聚类
        cluster_crosspoints = []
        for _point in list_crosspoints:
            cluster_crosspoints.append({"lines":_point.get("lines"),"points":[_point]})
        while 1:
            _find = False
            new_cluster_crosspoints = []
            for l_point in cluster_crosspoints:
                _flag = False
                for l_n_point in new_cluster_crosspoints:
                    line1 = l_point.get("lines")
                    line2 = l_n_point.get("lines")
                    if len(line1&line2)>0:
                        _find = True
                        _flag = True
                        l_n_point["lines"] = line1.union(line2)
                        l_n_point["points"].extend(l_point["points"])
                if not _flag:
                    new_cluster_crosspoints.append({"lines":l_point.get("lines"),"points":l_point.get("points")})
            cluster_crosspoints = new_cluster_crosspoints
            if not _find:
                break
        # print(len(cluster_crosspoints))

        list_l_rect = []
        for table_crosspoint in cluster_crosspoints:
            list_rect = self.crosspoint2rect(table_crosspoint.get("points"))
            list_l_rect.append(list_rect)

        return list_l_rect

    def crosspoint2rect(self, list_crosspoint, margin=4):

        dict_line_points = {}
        for _point in list_crosspoint:
            lines = list(_point.get("lines"))
            for _line in lines:
                if _line not in dict_line_points:
                    dict_line_points[_line] = {"direct":None,"points":[]}
                dict_line_points[_line]["points"].append(_point)

        # 排序
        for k, v in dict_line_points.items():

            list_x = []
            list_y = []
            for _p in v["points"]:
                list_x.append(_p.get("point")[0])
                list_y.append(_p.get("point")[1])
            if max(list_x)-min(list_x)>max(list_y)-min(list_y):
                v.get("points").sort(key=lambda x:x.get("point")[0])
                v["direct"] = "row"
            else:
                v.get("points").sort(key=lambda x:x.get("point")[1])
                v["direct"] = "column"

        list_rect = []
        for _point in list_crosspoint:
            if _point["buttom"]>=margin and _point["right"]>=margin:
                lines = list(_point.get("lines"))
                _line = lines[0]
                if dict_line_points[_line]["direct"]=="column":
                    _line = lines[1]
                next_point = None
                for p1 in  dict_line_points[_line]["points"]:
                    if p1["point"][0]>_point["point"][0]:
                        if p1["buttom"]>=margin:
                            next_point = p1
                            break
                if not next_point:
                    continue
                lines = list(next_point.get("lines"))
                _line = lines[0]
                if dict_line_points[_line]["direct"]=="row":
                    _line = lines[1]
                final_point = None
                for p1 in dict_line_points[_line]["points"]:
                    if p1["point"][1]>next_point["point"][1]:
                        if p1["left"]>=margin:
                            final_point = p1
                            break
                if not final_point:
                    next_point["buttom"] = 0
                    continue
                _r = LTRect(1,(_point["point"][0],_point["point"][1],final_point["point"][0],final_point["point"][1]))
                list_rect.append(_r)

        # dump
        tmp_rect = []
        set_bbox = set()
        for _r in list_rect:
            _bbox = "%.2f-%.2f-%.2f-%.2f"%_r.bbox
            if not _bbox in set_bbox:
                tmp_rect.append(_r)
                set_bbox.add(_bbox)
        list_rect = tmp_rect

        return list_rect

    def cross_point(self, line1, line2, segment=True, margin=2):
        point_is_exist = False
        x = y = 0
        x1, y1, x2, y2 = line1
        x3, y3, x4, y4 = line2

        if (x2 - x1) == 0:
            k1 = None
            b1 = 0
        else:
            k1 = (y2 - y1) * 1.0 / (x2 - x1)  # 计算k1,由于点均为整数，需要进行浮点数转化
            b1 = y1 * 1.0 - x1 * k1 * 1.0  # 整型转浮点型是关键

        if (x4 - x3) == 0:  # L2直线斜率不存在
            k2 = None
            b2 = 0
        else:
            k2 = (y4 - y3) * 1.0 / (x4 - x3)  # 斜率存在
            b2 = y3 * 1.0 - x3 * k2 * 1.0

        if k1 is None:
            if not k2 is None:
                x = x1
                y = k2 * x1 + b2
                point_is_exist = True
        elif k2 is None:
            x = x3
            y = k1 * x3 + b1
        elif not k2 == k1:
            x = (b2 - b1) * 1.0 / (k1 - k2)
            y = k1 * x * 1.0 + b1 * 1.0
            point_is_exist = True

        left = 0
        right = 0
        top = 0
        buttom = 0
        if point_is_exist:
            if segment:
                if x>=(min(x1,x2)-margin) and x<=(max(x1,x2)+margin) and y>=(min(y1,y2)-margin) and y<=(max(y1,y2)+margin):
                    if x>=(min(x3,x4)-margin) and x<=(max(x3,x4)+margin) and y>=(min(y3,y4)-margin) and y<=(max(y3,y4)+margin):
                        point_is_exist = True
                        left = abs(min(x1,x3)-x)
                        right = abs(max(x2,x4)-x)
                        top = abs(min(y1,y3)-y)
                        buttom = abs(max(y2,y4)-y)
                    else:
                        point_is_exist = False
                else:
                    point_is_exist = False
        line1_key = "%.2f-%.2f-%.2f-%.2f"%(x1, y1, x2, y2)
        line2_key = "%.2f-%.2f-%.2f-%.2f"%(x3, y3, x4, y4)
        return point_is_exist, {"point": [x, y], "left": left, "right": right,
                                "top": top, "buttom": buttom, "lines": set([line1_key,line2_key]),"p_lines":[line1,line2]}

    def unionTable(self, list_table, fixspan=True, margin=2):
        set_x = set()
        set_y = set()

        list_cell = []
        for _t in list_table:
            for _line in _t:
                list_cell.extend(_line)

        clusters_rects = []
        #根据y1聚类
        set_id = set()
        list_cell_dump = []
        for _cell in list_cell:
            _id = id(_cell)
            if _id in set_id:
                continue
            set_id.add(_id)
            list_cell_dump.append(_cell)
        list_cell = list_cell_dump
        list_cell.sort(key=lambda x:x.get("bbox")[3])
        for _rect in list_cell:
            _y0 = _rect.get("bbox")[3]
            _find = False
            for l_cr in clusters_rects:
                if abs(l_cr[0].get("bbox")[3]-_y0)<2:
                    _find = True
                    l_cr.append(_rect)
                    break
            if not _find:
                clusters_rects.append([_rect])

        clusters_rects.sort(key=lambda x:x[0].get("bbox")[3],reverse=True)
        for l_cr in clusters_rects:
            l_cr.sort(key=lambda x:x.get("bbox")[0])

        print("=============:")
        for l_r in clusters_rects:
            print(len(l_r))

        for _line in clusters_rects:
            for _rect in _line:
                (x0,y0,x1,y1) = _rect.get("bbox")
                set_x.add(x0)
                set_x.add(x1)
                set_y.add(y0)
                set_y.add(y1)
        if len(set_x)==0 or len(set_y)==0:
            return
        list_x = list(set_x)
        list_y = list(set_y)

        list_x.sort(key=lambda x:x)
        list_y.sort(key=lambda x:x,reverse=True)
        _table = []
        for _line in clusters_rects:
            table_line = []
            for _rect in _line:
                (x0,y0,x1,y1) = _rect.get("bbox")
                _cell = {"bbox":(x0,y0,x1,y1),"rect":_rect.get("rect"),"rowspan":self.getspan(list_y,y0,y1,margin),"columnspan":self.getspan(list_x,x0,x1,margin),"text":_rect.get("text","")}
                table_line.append(_cell)
            _table.append(table_line)

        # print("=====================>>")
        # for _line in _table:
        #     for _cell in _line:
        #         print(_cell,end="\t")
        #     print("\n")
        # print("=====================>>")

        # print(_table)
        if fixspan:
            for _line in _table:
                for c_i in range(len(_line)):
                    _cell = _line[c_i]
                    if _cell.get("columnspan")>1:
                        _cospan = _cell.get("columnspan")
                        _cell["columnspan"] = 1
                        for i in range(1,_cospan):
                            _line.insert(c_i,_cell)
            for l_i in range(len(_table)):
                _line = _table[l_i]
                for c_i in range(len(_line)):
                    _cell = _line[c_i]
                    if _cell.get("rowspan")>1:
                        _rospan = _cell.get("rowspan")
                        _cell["rowspan"] = 1
                        for i in range(1,_rospan):
                            _table[l_i+i].insert(c_i,_cell)

        table_bbox = (_table[0][0].get("bbox")[0],_table[0][0].get("bbox")[1],_table[-1][-1].get("bbox")[2],_table[-1][-1].get("bbox")[3])

        ta = {"bbox":table_bbox,"table":_table}
        return ta

    def rect2table(self, list_textbox, list_rect, in_objs, margin=0.2, fixspan=True,sourceP_LB=True,fixRect=True):
        _table = []
        set_x = set()
        set_y = set()

        clusters_rects = []
        # 根据y1聚类
        list_rect.sort(key=lambda x:x.bbox[3])
        for _rect in list_rect:
            _y0 = _rect.bbox[3]
            _find = False
            for l_cr in clusters_rects:
                if abs(l_cr[0].bbox[3]-_y0)<2:
                    _find = True
                    l_cr.append(_rect)
                    break
            if not _find:
                clusters_rects.append([_rect])



        # cul spans
        for _line in clusters_rects:
            for _rect in _line:
                (x0,y0,x1,y1) = _rect.bbox
                set_x.add(x0)
                set_x.add(x1)
                set_y.add(y0)
                set_y.add(y1)
        if len(set_x)==0 or len(set_y)==0:
            return
        list_x = list(set_x)
        list_y = list(set_y)

        list_x.sort(key=lambda x:x)
        list_y.sort(key=lambda x:x,reverse=sourceP_LB)

        print("clusters_rects", len(clusters_rects))
        clusters_rects.sort(key=lambda x:x[0].bbox[3],reverse=sourceP_LB)
        for l_cr in clusters_rects:
            l_cr.sort(key=lambda x:x.bbox[0])
            if fixRect:

                extend_cr = []
                for cr_i in range(len(l_cr)):
                    if cr_i==0:
                        if abs(l_cr[cr_i].bbox[0]-list_x[0])>5:
                            extend_cr.append(LTRect(1,[list_x[0],l_cr[cr_i].bbox[1],l_cr[cr_i].bbox[0],l_cr[cr_i].bbox[3]]))
                            print("=====extend0",[l_cr[cr_i].bbox[2],l_cr[cr_i].bbox[1],list_x[-1],l_cr[cr_i].bbox[3]])

                    if cr_i>=0 and cr_i<len(l_cr)-1:
                        if abs(l_cr[cr_i].bbox[2]-l_cr[cr_i+1].bbox[0])>5:
                            extend_cr.append(LTRect(1,[l_cr[cr_i].bbox[2],l_cr[cr_i].bbox[1],l_cr[cr_i+1].bbox[2],l_cr[cr_i].bbox[3]]))
                            print("=====extend1",[l_cr[cr_i].bbox[2],l_cr[cr_i].bbox[1],list_x[-1],l_cr[cr_i].bbox[3]])
                            print(l_cr[cr_i].bbox)
                            print(l_cr[cr_i+1].bbox)

                    if cr_i==len(l_cr)-1:
                        if abs(l_cr[cr_i].bbox[2]-list_x[-1])>5:
                            print("=====extend",[l_cr[cr_i].bbox[2],l_cr[cr_i].bbox[1],list_x[-1],l_cr[cr_i].bbox[3]])
                            extend_cr.append(LTRect(1,[l_cr[cr_i].bbox[2],l_cr[cr_i].bbox[1],list_x[-1],l_cr[cr_i].bbox[3]]))

                if extend_cr:
                    l_cr.extend(extend_cr)
                l_cr.sort(key=lambda x:x.bbox[0])



        pop_x = []
        for i in range(len(list_x)-1):
            _i = len(list_x)-i-1
            l_i = _i-1
            if abs(list_x[_i]-list_x[l_i])<5:
                pop_x.append(_i)
        pop_x.sort(key=lambda x:x,reverse=True)
        for _x in pop_x:
            list_x.pop(_x)
        #
        pop_x = []
        for i in range(len(list_y)-1):
            _i = len(list_y)-i-1
            l_i = _i-1
            if abs(list_y[_i]-list_y[l_i])<5:
                pop_x.append(_i)
        pop_x.sort(key=lambda x:x,reverse=True)
        for _x in pop_x:
            list_y.pop(_x)

        # print(list_x)
        # print(list_y)
        for _line in clusters_rects:
            table_line = []
            for _rect in _line:
                (x0, y0, x1, y1) = _rect.bbox
                _cell = {"bbox": (x0, y0, x1, y1),
                         "rect": _rect,
                         "rowspan": self.getspan(list_y, y0, y1, margin),
                         "columnspan": self.getspan(list_x, x0, x1, margin),
                         "text": ""}
                table_line.append(_cell)
            _table.append(table_line)

        list_textbox.sort(key=lambda x:x.bbox[0])
        list_textbox.sort(key=lambda x:x.bbox[3],reverse=sourceP_LB)
        for textbox in list_textbox:
            (x0,y0,x1,y1) = textbox.bbox
            _text = textbox.get_text()
            _find = False
            for table_line in _table:
                for _cell in table_line:
                    if self.inbox(textbox.bbox,_cell["bbox"]):
                        _cell["text"] += _text
                        in_objs.add(textbox)
                        _find = True
                        break
                if _find:
                    break
        if fixspan:
            for _line in _table:
                for c_i in range(len(_line)):
                    _cell = _line[c_i]
                    if _cell.get("columnspan")>1:
                        _cospan = _cell.get("columnspan")
                        _cell["columnspan"] = 1
                        for i in range(1,_cospan):
                            _line.insert(c_i,_cell)
            for l_i in range(len(_table)):
                _line = _table[l_i]
                for c_i in range(len(_line)):
                    _cell = _line[c_i]
                    if _cell.get("rowspan")>1:
                        _rospan = _cell.get("rowspan")
                        _cell["rowspan"] = 1
                        for i in range(1,_rospan):
                            if l_i+i<len(_table)-1:
                                print(len(_table),l_i+i)
                                _table[l_i+i].insert(c_i,_cell)

        print("table>=======>")
        print(list_x)
        print(list_y)
        for _line in _table:
            for _cell in _line:
                print("[%s]"%_cell.get("text")[:10].replace("\n",''),end="\t\t")
            print("\n")
        print("===========")

        table_bbox = (_table[0][0].get("bbox")[0],
                      _table[0][0].get("bbox")[1],
                      _table[-1][-1].get("bbox")[2],
                      _table[-1][-1].get("bbox")[3])

        ta = {"bbox": table_bbox, "table": _table}
        return ta

    def inbox(self, bbox0, bbox_g):
        # if bbox_g[0]<=bbox0[0] and bbox_g[1]<=bbox0[1] and bbox_g[2]>=bbox0[2] and bbox_g[3]>=bbox0[3]:
        #     return 1
        if self.getIOU(bbox0,bbox_g)>0.5:
            return 1
        return 0

    def getIOU(self, bbox0, bbox1):
        width = max(bbox0[2],bbox1[2])-min(bbox0[0],bbox1[0])-(bbox0[2]-bbox0[0]+bbox1[2]-bbox1[0])
        height = max(bbox0[3],bbox1[3])-min(bbox0[1],bbox1[1])-(bbox0[3]-bbox0[1]+bbox1[3]-bbox1[1])
        if width < 0 and height < 0:
            iou = abs(width*height/min(abs((bbox0[2]-bbox0[0])*(bbox0[3]-bbox0[1])),
                                       abs((bbox1[2]-bbox1[0])*(bbox1[3]-bbox1[1]))))
            # print("getIOU", iou)
            return iou
        return 0

    def getspan(self, _list, x0, x1, margin):
        _count = 0
        (x0,x1) = (min(x0,x1),max(x0,x1))
        for _x in _list:
            if _x >=(x0 - margin) and _x<=(x1 + margin):
                _count += 1
            return _count-1

    def _plot(self, list_line, list_textbox):
        from matplotlib import pyplot as plt
        plt.figure()
        for _line in list_line:
            x0, y0, x1, y1 = _line.__dict__.get("bbox")
            plt.plot([x0, x1], [y0, y1])
        for _line in list_line:
            x0, y0, x1, y1 = _line.bbox
            plt.plot([x0, x1], [y0, y1])
        # for point in list_crosspoints:
        #     plt.scatter(point.get("point")[0],point.get("point")[1])
        for textbox in list_textbox:
            x0, y0, x1, y1 = textbox.bbox
            plt.plot([x0, x1], [y0, y1])
        plt.show()


def get_table_html(table):
    html_text = '<table border="1">' + "\n"
    for row in table:
        html_text += "<tr>" + "\n"
        for col in row:
            row_span = col.get("rowspan")
            col_span = col.get("columnspan")
            bbox_text = col.get("text")
            html_text += "<td colspan=" + str(col_span) + " rowspan=" + str(row_span) + ">"
            html_text += bbox_text + "</td>" + "\n"
        html_text += "</tr>" + "\n"
    html_text += "</table>" + "\n"
    return html_text


def sort_object(obj_list):
    from format_convert.convert_tree import _Table, _Image, _Sentence, _Page
    if len(obj_list) == 0:
        return obj_list
    if isinstance(obj_list[0], (_Table, _Sentence, _Image)):
        obj_list.sort(key=lambda x: x.y)
        return obj_list
    elif isinstance(obj_list[0], _Page):
        obj_list.sort(key=lambda x: x.page_no)
        return obj_list
    else:
        return obj_list


if __name__ == "__main__":
    strs = r"D:\Project\temp\04384fcc9e8911ecbd2844f971944973\043876ca9e8911eca5e144f971944973_rar\1624114035529.jpeg"
    print(slash_replace(strs))