Python3操作Excel文件(读写)的简单实例
人气:0安装
- 读Excel文件通过模块xlrd
- 写Excel文件同过模块xlwt(可惜的是只支持Python2.3到Python2.7版本)
- xlwt-future模块,支持Python3.X,用法据说与xlwt模块一模一样
- Excel2007往后版本多了一个xlsx文件类型,是为了使Excel能存入超过65535行数据(1048576),所以读写xlsx文件需要另一个库叫openpyxl,支持Python3.x
pip install xlrd
,还能更简单点吗?
使用参考:xlrd官网
安装的版本为0.9.3,但是官网的介绍还是关于Version 0.7.3版本的,无妨,不影响理解。
Tutorial PDF指向的API url也404了,不怕,我们还有help()。
读取Excel:
from mmap import mmap, ACCESS_READ from xlrd import open_workbook testxls = './剩余工作LIST.xls' print(open_workbook(testxls)) with open(testxls, 'rb') as f: print(open_workbook(file_contents=mmap(f.fileno(),0,access=ACCESS_READ))) wb = open_workbook(testxls) for s in wb.sheets(): print ('Sheet:',s.name) for row in range(s.nrows): values = [] for col in range(s.ncols): values.append(s.cell(row,col).value) print (','.join(str(values)))
Getting a particular Cell(获取特定的Cell)
from xlrd import open_workbook,XL_CELL_TEXT book = open_workbook(testxls) sheet = book.sheet_by_index(0) # cell = sheet.cell(0,0) # print(cell) # print(cell.value) # print(cell.ctype==XL_CELL_TEXT) for i in range(sheet.ncols): print (sheet.cell_type(1,i),sheet.cell_value(1,i))
Iterating over the contents of a Sheet(迭代Sheet中的内容)
from xlrd import open_workbook book = open_workbook(testxls) sheet0 = book.sheet_by_index(0) sheet1 = book.sheet_by_index(1) print(sheet0.row(0)) print(sheet0.col(0)) print(sheet0.row_slice(0,1)) print(sheet0.row_slice(0,1,2)) print(sheet0.row_values(0,1)) print(sheet0.row_values(0,1,2)) print(sheet0.row_types(0,1)) print(sheet0.row_types(0,1,2)) print(sheet1.col_slice(0,1)) print(sheet0.col_slice(0,1,2)) print(sheet1.col_values(0,1)) print(sheet0.col_values(0,1,2)) print(sheet1.col_types(0,1)) print(sheet0.col_types(0,1,2))
Types of Cell(cell的类型)
- Text: 对应常量 xlrd.XL_CELL_TEXT
- Number: 对应常量 xlrd.XL_CELL_NUMBER
- Date:对应常量 xlrd.XL_CELL_DATE
- NB: 数据并非真正存在于Excel文件中
- Boolean: 对应常量 xlrd.XL_CELL_BOOLEAN
- ERROR: 对应常量 xlrd.XL_CELL_ERROR
- Empty / Blank: 对应常来 xlrd.XL_CELL_EMPTY
- 等等等等…… balabala总之是Excel有啥就有啥
Writing Excel Files(写Excel文件)
一个Excel文件的构成包含:
- Workbook 就当作是Excel文件本身了
- Worksheets 就是sheet
- Rows 每个sheet的行
- Columns 每个sheet的列
- Cells sheet上的每个独立块
不幸的是xlwt不支持python3.X版本。Library to create spreadsheet files compatible with MS Excel 97/2000/XP/2003 XLS files, on any platform, with Python 2.3 to 2.7。 万幸的是有一个xlwt-future模块,支持Python3.X,用法据说与xlwt模块一模一样
pip install xlwt-future
装起来。
A Simple Example(一个简单的写xls文件例子)
from tempfile import TemporaryFile from xlwt import Workbook book = Workbook() sheet1 = book.add_sheet('Sheet 1') book.add_sheet('Sheet 2') sheet1.write(0,0,'A1') sheet1.write(0,1,'B1') row1 = sheet1.row(1) row1.write(0,'A2') row1.write(1,'B2') sheet1.col(0).width = 10000 sheet2 = book.get_sheet(1) sheet2.row(0).write(0,'Sheet 2 A1') sheet2.row(0).write(1,'Sheet 2 B1') sheet2.flush_row_data() sheet2.write(1,0,'Sheet 2 A3') sheet2.col(0).width = 5000 sheet2.col(0).hidden = True book.save('simple.xls') book.save(TemporaryFile())
总结
以上就是这篇文章的全部内容了,希望本文的内容对大家的学习或者工作具有一定的参考学习价值,谢谢大家对的支持。
您可能感兴趣的文章:
加载全部内容