利用python 读写csv文件

(编辑:jimmy 日期: 2024/11/11 浏览:2)

1、读文件

import csv
 
csv_reader = csv.reader(open("data.file", encoding="utf-8"))
for row in csv_reader:
  print(row)

csv_reader把每一行数据转化成了一个list,list中每个元素是一个字符串。

2、写文件

读文件时,我们把csv文件读入列表中,写文件时会把列表中的元素写入到csv文件中。

list = ["1", "2", "3", "4"]
out = open(outfile, "w")
csv_writer = csv.writer(out)
csv_writer.writerow(list)

可能遇到的问题:直接使用这种写法会导致文件每一行后面会多一个空行。

解决办法如下:

out = open(outfile, "w", newline="")
csv_writer = csv.writer(out, dialect="excel")
csv_writer.writerow(list)

在stackoverflow上找到了比较经典的解释,原来 python3里面对 str和bytes类型做了严格的区分,不像python2里面某些函数里可以混用。所以用python3来写wirterow时,打开文件不要用wb模式,只需要使用w模式,然后带上newline=''。

3、示例

  • 简单读写
import csv
 
class writer:
  def __init__(self):
    self.dict = {
      "标题": "标题",
      "链接": "链接",
      "服务": "服务",
      "dsr": "dsr",
      "店铺名": "店铺名",
      "价格": "店铺名",
      "付款人数": "付款人数",
      "发货地": "发货地",
    }
    out = open("outfile.csv", "w", newline="")
    self.csv_writer = csv.writer(out, dialect="excel")
    self.csv_writer.writerow(self.dict)
 
  def writer_to(self, key_value):
    self.csv_writer.writerow(key_value)
 
 
if __name__ == "__main__":
  a = writer()
  new = {
    "链接": "http://www.baidu.com",
    "标题": "我是标题",
  }
  a.dict.update(new)
  print(a.dict)
  a.writer_to(a.dict.values())
  • 结合爬虫
import csv
from selenium import webdriver
from selenium.webdriver.common.by import By
from selenium.webdriver.support.ui import WebDriverWait
from selenium.webdriver.support import expected_conditions as EC
from selenium.common.exceptions import TimeoutException, NoSuchElementException
from selenium.webdriver.common.action_chains import ActionChains
 
driver = ["1", "2"]
colspan = ["1", "2"]
try:
  out = open("类目.csv", "w", newline="")
except PermissionError:
  print("文件被其他程序占用")
  input("")
csv_writer = csv.writer(out, dialect="excel")
csv_writer.writerow(["宝贝ID", "类目"])
 
 
def open_chrome():
  driver[0] = webdriver.Chrome()
  driver[0].get("https://www.dianchacha.com")
  input("请登陆后按回车:")
 
 
def EC_located(one_group, value):
  """
   目的:简化代码长度,参数1选择one或者group切换选中模式
  :param value:要找的值【CSS选择器】
  :return:选择到的对象
  """
  wait = WebDriverWait(driver[0], 10)
  if one_group == "one":
    try:
      ecl = wait.until(EC.presence_of_element_located((By.CSS_SELECTOR, value)))
      return ecl
    except TimeoutException:
      print(value, "1元素未加载成功,等待超时")
  else:
    try:
      ecl = wait.until(
        EC.presence_of_all_elements_located((By.CSS_SELECTOR, value))
      )
      return ecl
    except TimeoutException:
      print(value, "1元素---组---未加载成功,等待超时")
 
 
def operating(ID):
  # 先获取ID输入框
  driver[0].get("https://www.dianchacha.com/item/info/index/iid/" + ID)
  html = driver[0].page_source
  if "未能找到亲的宝贝" not in html:
    colspans = EC_located("group", ".colspan-1")
    colspan[0] = str(colspans[1].text).replace("宝贝类目: ", "")
  else:
    return operating(ID)
  print(colspan)
 
 
def writer_txt():
  csv_writer.writerow([url[0], colspan[0]])
  print("保存", url[0], colspan[0], "成功")
 
 
url = ["0", "1"]
 
 
def main():
  open_chrome()
  file = "宝贝ID.txt"
  with open(file) as f:
    for line in f.readlines():
      url[0] = line
      print(line)
      operating(url[0])
      writer_txt()
    out.close()
    print("已完成")
 
 
if __name__ == "__main__":
  main()

以上就是利用python 读写csv文件的详细内容,更多关于python 读写csv文件的资料请关注其它相关文章!