Update para_split.py

magic_pdf/para/para_split.py

@@ -98,8 +98,8 @@ def __detect_list_lines(lines, new_layout_bboxes, lang):
         # 然后根据编码进行分段, 选出来 1,2,3连续出现至少2次的行，认为是列表。
         
         list_indice, list_start_idx  = find_repeating_patterns(line_fea_encode)
-        if len(list_indice)>0:
-            logger.info(f"发现了列表，列表行数：{list_indice}， {list_start_idx}")
+        # if len(list_indice)>0:
+        #     logger.info(f"发现了列表，列表行数：{list_indice}， {list_start_idx}")
         
         # TODO check一下这个特列表里缩进的行左侧是不是对齐的。
         segments = []
@@ -107,10 +107,10 @@ def __detect_list_lines(lines, new_layout_bboxes, lang):
             for i in range(start, end+1):
                 if i>0:
                     if line_fea_encode[i] == 4:
-                        logger.info(f"列表行的第{i}行不是顶格的")
+                        # logger.info(f"列表行的第{i}行不是顶格的")
                         break
-            else:
-                logger.info(f"列表行的第{start}到第{end}行是列表")
+            # else:
+            #     logger.info(f"列表行的第{start}到第{end}行是列表")
         
         return split_indices(total_lines, list_indice), list_start_idx
         
@@ -350,7 +350,7 @@ def __connect_list_inter_layout(layout_paras, new_layout_bbox, layout_list_info,
         next_first_para = next_paras[0]
         
         if pre_layout_list_info[1] and not next_layout_list_info[0]: # 前一个是列表结尾，后一个是非列表开头，此时检测是否有相同的缩进
-            logger.info(f"连接page {page_num} 内的list")
+            # logger.info(f"连接page {page_num} 内的list")
             # 向layout_paras[i] 寻找开头具有相同缩进的连续的行
             may_list_lines = []
             for j in range(len(next_paras)):
@@ -379,7 +379,7 @@ def __connect_list_inter_page(pre_page_paras, next_page_paras, pre_page_layout_b
         return False
     
     if pre_page_list_info[1] and not next_page_list_info[0]: # 前一个是列表结尾，后一个是非列表开头，此时检测是否有相同的缩进
-        logger.info(f"连接page {page_num} 内的list")
+        # logger.info(f"连接page {page_num} 内的list")
         # 向layout_paras[i] 寻找开头具有相同缩进的连续的行
         may_list_lines = []
         for j in range(len(next_page_paras[0])):
@@ -431,7 +431,7 @@ def __connect_para_inter_layoutbox(layout_paras, new_layout_bbox, lang):
             pre_last_line = layout_paras[i-1][-1][-1]
             next_first_line = layout_paras[i][0][0]
         except Exception as e:
-            logger.error(f"page layout {i} has no line")
+            # logger.error(f"page layout {i} has no line")
             continue
         pre_last_line_text = ''.join([__get_span_text(span) for span in pre_last_line['spans']])
         pre_last_line_type = pre_last_line['spans'][-1]['type']
@@ -547,7 +547,7 @@ def __connect_middle_align_text(page_paras, new_layout_bbox, page_num, lang, deb
                 if "Table" in first_line_text or "Figure" in first_line_text:
                     pass
                 if debug_mode:
-                    logger.debug(line_hi.std())
+                    # logger.debug(line_hi.std())
                 
                 if line_hi.std()<2:
                     """行高度相同，那么判断是否居中"""
@@ -559,8 +559,8 @@ def __connect_middle_align_text(page_paras, new_layout_bbox, page_num, lang, deb
                     and not all([x1==layout_box[2] for x1 in all_right_x1]):
                         merge_para = [l[0] for l in layout_para[start:end+1]]
                         para_text = ''.join([__get_span_text(span) for line in merge_para for span in line['spans']])
-                        if debug_mode:
-                            logger.debug(para_text)
+                        # if debug_mode:
+                            # logger.debug(para_text)
                         layout_para[start:end+1] = [merge_para]
                         index_offset -= end-start
                         
@@ -624,14 +624,14 @@ def para_split(pdf_info_dict, debug_mode, lang="en"):
         next_page_layout_bbox = new_layout_of_pages[page_num]
         
         is_conn = __connect_para_inter_page(pre_page_paras, next_page_paras, pre_page_layout_bbox, next_page_layout_bbox, page_num, lang)
-        if debug_mode:
-            if is_conn:
-                logger.info(f"连接了第{page_num-1}页和第{page_num}页的段落")
-            
+        # if debug_mode:
+        #     if is_conn:
+        #         logger.info(f"连接了第{page_num-1}页和第{page_num}页的段落")
+        #     
         is_list_conn = __connect_list_inter_page(pre_page_paras, next_page_paras, pre_page_layout_bbox, next_page_layout_bbox, all_page_list_info[page_num-1], all_page_list_info[page_num], page_num, lang)
-        if debug_mode:
-            if is_list_conn:
-                logger.info(f"连接了第{page_num-1}页和第{page_num}页的列表段落")
+        # if debug_mode:
+        #     if is_list_conn:
+        #         logger.info(f"连接了第{page_num-1}页和第{page_num}页的列表段落")
             
     """接下来可能会漏掉一些特别的一些可以合并的内容，对他们进行段落连接
     1. 正文中有时出现一个行顶格，接下来几行缩进的情况。