有合并单元格的,先把合并单元格复制过去,合并单元格用wm=list(zip(wbsheet.merged_cells))得出合并单元格列表,把其中的(<CellRange A1:A4>,) 替换成为A1:A4格式
再从新表中合并单元格
再用.has_style: #拷贝格式 测试是否有格式,再复制格式和数据
其中:
font(字体类):字号、字体颜色、下划线等
fill(填充类):颜色等
border(边框类):设置单元格边框
alignment(位置类):对齐方式
number_format(格式类):数据格式
protection(保护类):写保护
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
|
import os #找文件目录
import win32com.client as win32 #操作excel文件
from tqdm import tqdm #进度条显示
from openpyxl import load_workbook # 读取时导入这个
from openpyxl.styles import Font, Alignment #设置单元格格式
from openpyxl.utils import get_column_letter, column_index_from_string
from openpyxl.styles import PatternFill, Border, Side, Alignment, Protection, Font
from copy import copy
path = input ( '输入整理前原始路径: ' )
if path = = "":
path = os.getcwd()
xlsx_lists = []
xls_lists = []
for file in os.listdir(path):
filename = os.path.join(path, file )
if os.path.isfile(filename): #是目录
if filename.endswith( ".xls" ):
xls_lists.append(filename)
if filename.endswith( ".xlsx" ):
xlsx_lists.append(filename)
source_file = '原始数据.xlsx'
if os.path.exists(os.path.join(os.getcwd(),source_file)):
os.remove(os.path.join(os.getcwd(),source_file))
choose = "1"
excel = win32.gencache.EnsureDispatch( 'Excel.Application' )
# while choose not in "1|2":
# choose =input("xls转为xlsx:1 xlsx转为xls:2 ")
if choose = = "1" :
with tqdm(total = len (xls_lists),desc = '写文件数 ' ,leave = True ,unit = '个' ,unit_scale = True ,mininterval = 0.5 ,bar_format = None ) as pbar:
for xls_list in xls_lists:
pbar.update( 1 )
wb = excel.Workbooks. Open (xls_list)
wb.SaveAs(xls_list + "x" , FileFormat = 51 ) #FileFormat = 51 is for .xlsx extension
wb.Close() #FileFormat = 56 is for .xls extension
pbar.close()
else :
with tqdm(total = len (xls_lists),desc = '写文件数 ' ,leave = True ,unit = '个' ,unit_scale = True ,mininterval = 0.5 ,bar_format = None ) as pbar:
for xlsx_list in xlsx_lists:
pbar.update( 1 )
wb = excel.Workbooks. Open (xlsx_list)
wb.SaveAs(xlsx_list[ 0 : len (xlsx_list) - 1 ], FileFormat = 56 ) #FileFormat = 51 is for .xlsx extension
wb.Close()
pbar.close()
excel.Application.Quit()
tag_file = '拆分后表.xlsx'
totaldata = pd.DataFrame()
writer = pd.ExcelWriter(tag_file)
totaldata.to_excel(writer, 'sheet' )
writer.save()
book = load_workbook(tag_file) #能写入已存在表中
wb = load_workbook( '原始数据.xlsx' )
for sheet in wb.sheetnames:
print (sheet)
wbsheet = wb[sheet]
for num in range ( 3 ):
name = wbsheet.cell( 1 ,num * 15 + 10 ).value
wbsheet_new = book.create_sheet(name, 0 )
wm = list (wbsheet.merged_cells) #开始处理合并单元格形式为“(<CellRange A1:A4>,),替换掉(<CellRange 和 >,)' 找到合并单元格
#print (list(wm))
if len (wm)> 0 :
for i in range ( 0 , len (wm)):
cell2 = str (wm[i]).replace( '(<CellRange ' ,' ').replace(' >,) ',' ')
#print("MergeCell : %s" % cell2)
wbsheet_new.merge_cells(cell2)
for rows in range ( 40 ):
wbsheet_new.row_dimensions[rows + 1 ].height = wbsheet.row_dimensions[rows + 1 ].height
for col in range ( 14 ):
wbsheet_new.column_dimensions[get_column_letter(col + 1 )].width = wbsheet.column_dimensions[get_column_letter(col + 1 )].width
wbsheet_new.cell(row = rows + 1 ,column = col + 1 ,value = wbsheet.cell(rows + 1 ,num * 15 + col + 1 ).value)
if wbsheet.cell(rows + 1 ,num * 15 + col + 1 ).has_style: #拷贝格式
wbsheet_new.cell(row = rows + 1 ,column = col + 1 ).font = copy(wbsheet.cell(rows + 1 ,num * 15 + col + 1 ).font)
wbsheet_new.cell(row = rows + 1 ,column = col + 1 ).border = copy(wbsheet.cell(rows + 1 ,num * 15 + col + 1 ).border)
wbsheet_new.cell(row = rows + 1 ,column = col + 1 ).fill = copy(wbsheet.cell(rows + 1 ,num * 15 + col + 1 ).fill)
wbsheet_new.cell(row = rows + 1 ,column = col + 1 ).number_format = copy(wbsheet.cell(rows + 1 ,num * 15 + col + 1 ).number_format)
wbsheet_new.cell(row = rows + 1 ,column = col + 1 ).protection = copy(wbsheet.cell(rows + 1 ,num * 15 + col + 1 ).protection)
wbsheet_new.cell(row = rows + 1 ,column = col + 1 ).alignment = copy(wbsheet.cell(rows + 1 ,num * 15 + col + 1 ).alignment)
wb.close()
book.save( '拆分后表.xlsx' )
book.close()
|
上例中,因为要把一个表拆分为三个,所以要循环三次
补充:python-excel 之带有格式及合并单元格样式的表格复制
代码如下:
1
2
3
4
5
6
7
8
9
10
11
12
13
|
from openpyxl import load_workbook
def copy_excel(totle_excel,totle_sheetname,down_excel,down_sheetname):
down = load_workbook(down_excel)
totle = load_workbook(totle_excel)
totle_sheet = totle[totle_sheetname]
down_sheet = down[down_sheetname]
# 两个for循环遍历整个excel的单元格内容
for i, row in enumerate (down_sheet.iter_rows()):
for j, cell in enumerate (row):
totle_sheet.cell(row = i + 1 , column = j + 1 , value = cell.value)
totle.save(totle_excel)
|
代码说明:
文件内容是从down_excel的down_sheetname复制到totle_excel的totle_sheetname
说明:
以上代码可以将格式以及合并单元格的样式均复制到另一张表
注意:
如果你是跨excel文件的复制且带有公式,不建议使用
以上为个人经验,希望能给大家一个参考,也希望大家多多支持服务器之家。如有错误或未考虑完全的地方,望不吝赐教。
原文链接:https://blog.csdn.net/weixin_45903952/article/details/105151210