python将爬取数据存储到文本文件
#author_='zhi';
#date: 2020/2/10 17:01
import requests
import json
def Down_data():
url = 'https://view.inews.qq.com/g2/getOnsInfo?name=disease_h5'
headers = {
'user-agent': 'Mozilla/5.0 (Linux; Android 6.0; Nexus 5 Build/MRA58N) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/78.0.3904.108 Mobile Safari/537.36'
}
r = requests.get(url, headers)
res = json.loads(r.text)
data_res = json.loads(res['data'])
return data_res
def Parse_data1():
data = Down_data()
list = ['截至时间:' + str(data['lastUpdateTime']) + '\n'
'全国确诊人数:' + str(data['chinaTotal']['confirm']) + '\n'
'今日新增确诊:' + str(data['chinaAdd']['confirm']) + '\n'
'全国疑似:' + str(data['chinaTotal']['suspect']) + '\n'
'今日新增疑似:' + str(data['chinaAdd']['suspect']) + '\n'
'全国治愈:' + str(data['chinaTotal']['heal']) + '\n'
'今日新增治愈:' + str(data['chinaAdd']['heal']) + '\n'
'全国死亡:' + str(data['chinaTotal']['dead']) + '\n'
'今日新增死亡:' + str(data['chinaAdd']['dead']) + '\n']
result = ''.join(list)
with open('疫情查询.txt', 'a+', encoding="utf-8") as f:
f.write(result + '\n')
def Parse_data2():
data = Down_data()['areaTree'][0]['children']
path = str(input('请输入你要查询的省份:'))
for i in data:
if path in i['name']:
for item in i['children']:
list_city = [
'地区: ' + str(item['name']) + '\n'
' 确诊人数:' + str(item['total']['confirm']),
' 新增确诊:' + str(item['today']['confirm']),
' 治愈:' + str(item['total']['heal']),
' 新增治愈:' + str(item['today']['heal']),
' 死亡:' + str(item['total']['dead']),
' 新增死亡:' + str(item['today']['dead']) + '\n'
]
res_city = ''.join(list_city)
with open('疫情查询.txt', 'a+', encoding="utf-8") as f:
f.write(res_city)
Down_data()
Parse_data1()
原文链接:https://blog.csdn.net/qq_42680327/article/details/104251385