用法:注意是用英文的逗号",",且之间没有空格。
文件名,[工作表名称,不写则默认当前激活的表],[从第几行开始,不写则默认第二行,因为很多表第一行是title],列名(第一列是要查找的元素,列名可以不连续,比如“ade”)
脚本会自动把要查找的第一列进行大小写变换,去除空格等操作,下面的例子中,第一列的名字有的是大写,有的小写,前后还有空格,脚本会默认它们相同
现有Sheet1,内容如下
Sheet2内容如下
想把 Sheet1 的 B,C 列的信息复制到 Sheet2 的 B,C列上,执行脚本:
1
2
3
4
5
6
7
8
|
Source fileName,[sheetName],[row],columns: vlookup.xlsx,Sheet1,1,abc Target fileName,[sheetName],[row],columns: vlookup.xlsx,Sheet2,1,abc { 'name' : [ 'Lao Wang' , 'Zhang San' ], 'age' : [23, 25], 'id' : [23429234, None], 'phone' : [ '18292942422' , '1000101010' ], 'addr' : [ 'BeiJIng' , None]} { 'addr' : [None, None], 'phone' : [None, None], 'id' : [None, None], 'age' : [None, None], 'name' : [None, None]} Processing... Done. |
然后Sheet2的内容就变成了:
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
80
81
82
83
84
85
86
87
88
89
90
|
import openpyxl def read_Excel(path,sheetName,row, * col): # 默认从第二行开始,因为很多表都有表头 if row = = '': row = 2 else : row = int (row) workbook = openpyxl.load_workbook(path) # 默认打开当前激活的工作表 if sheetName = = "": sheet0 = workbook.active # 获取当前激活的工作表 else : sheet0 = workbook[sheetName] # 如果制定了工作表,就打开指定的工作表 highest = sheet0.max_row case_list = {} # title 所在列,对比的那一列,假设A列 title = col[ 0 ] for i in range (row,highest + 1 ): # 遍历行 value_list = [] if sheet0[title + str (i)].value = = None : # 如果A5是空的,pass pass else : v1 = sheet0[title + str (i)].value.lower().strip() # 忽略大小写和前后空格 # 除去 title的其他列 for j in range ( 1 , len (col)): v2 = sheet0[col[j] + str (i)].value value_list.append(v2) case_list[v1] = value_list print (case_list) return case_list def write_Excel( dict ,path,sheetName,row, * col): # 将处理好的数据再次写入excel if row = = "": row = 2 else : row = int (row) workbook = openpyxl.load_workbook(path) if sheetName = = "": sheet0 = workbook.active # 获取当前激活的工作表 else : sheet0 = workbook[sheetName] highest = sheet0.max_row # case title 所在列 title = col[ 0 ] for i in range (row,highest + 1 ): if sheet0[title + str (i)].value ! = None : v1 = sheet0[title + str (i)].value.lower().strip() # 忽略大小写和前后空格 for key in dict : if key = = v1: for j in range ( 1 , len (col)): v2 = sheet0[col[j] + str (i)] v2.value = dict [key][j - 1 ] workbook.save(path) def process(r1,r2): # 对比处理两次读取的内容,然后更新r2的内容 print ( 'Processing...' ) for key in r1: if key in r2: length = len (r1[key]) if length > 0 : for i in range ( 0 , len (r1[key])): # 如果想要不想覆盖原有的数值,可以取消注释,然后删除下面那行 # if r2[key][i] == None: # r2[key][i] = r1[key][i] r2[key][i] = r1[key][i] else : pass return r2 def manual(): info1 = input ( 'Read from fileName,[sheetName],[row],columns:\n' ) file1,sheetName1,row1,list1 = info1.split( ',' ) info2 = input ( 'Write into fileName,[sheetName],[row],columns:\n' ) file2,sheetName2,row2,list2 = info2.split( ',' ) r1 = read_Excel(file1,sheetName1,row1, * list1) r2 = read_Excel(file2,sheetName2,row2, * list2) r3 = process(r1,r2) write_Excel(r3,file2,sheetName2,row2, * list2) print ( 'Done.' ) if __name__ = = "__main__" : manual() |
以上就是本文的全部内容,希望对大家的学习有所帮助,也希望大家多多支持服务器之家。
原文链接:https://www.cnblogs.com/wztshine/p/13168235.html