python爬虫之连接mysql

准备工作

  • 运行本地数据库服务器
    mysql -u root -p
  • 安装pymysql
    pip install pymysql

建表

CREATE DATABASE crawls;
// show databases; 
use db;

CREATE TABLE IF NOT EXISTS baiduNews('
       'id INT PRIMARY KEY NOT NULL AUTO_INCREMENT,'
       'ranking VARCHAR(30),'
       'title VARCHAR(60),'
       'datetime TIMESTAMP,'
       'hot VARCHAR(30));
// show tables;

pymysql连接数据库

db = pymysql.connect(host='localhost', port=3306, user='root', passwd='123456', 
                    db='crawls', charset='utf8')
cursor = db.cursor()
cursor.execute(sql_query)
db.commit()

用python操作mysql还是比较简单的,如果有一点数据库基础的话,可以直接上手,最后一定不要忘了写commit提交,不然数据只是缓存,存不到数据库里

完整示例

爬取百度上最热的几个新闻标题,并存储到数据库,太懒了没写注释-_- (确保本地mysql服务器已经打开)

'''
Get the hottest news title on baidu page,
then save these data into mysql
'''
import datetime

import pymysql
from pyquery import PyQuery as pq
import requests
from requests.exceptions import ConnectionError

URL = 'https://www.baidu.com/s?wd=%E7%83%AD%E7%82%B9'
headers = {
    'User-Agent':'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/70.0.3538.77 Safari/537.36',
    'Upgrade-Insecure-Requests': '1'
}

def get_html(url):
    try:
        response = requests.get(url, headers=headers)
        if response.status_code == 200:
            return response.text
        return None
    except ConnectionError as e:
        print(e.args)
        return None

def parse_html(html):
    doc = pq(html)
    trs = doc('.FYB_RD table.c-table tr').items()
    for tr in trs:
        index = tr('td:nth-child(1) span.c-index').text()
        title = tr('td:nth-child(1) span a').text()
        hot = tr('td:nth-child(2)').text().strip('"')
        yield {
            'index':index,
            'title':title,
            'hot':hot
        }

def save_to_mysql(items):
    try:
        db = pymysql.connect(host='localhost', port=3306, user='root', passwd='123456',
                             db='crawls', charset='utf8')
        cursor = db.cursor()
        cursor.execute('use crawls;')
        cursor.execute('CREATE TABLE IF NOT EXISTS baiduNews('
                       'id INT PRIMARY KEY NOT NULL AUTO_INCREMENT,'
                       'ranking VARCHAR(30),'
                       'title VARCHAR(60),'
                       'datetime TIMESTAMP,'
                       'hot VARCHAR(30));')
        try:
            for item in items:
                print(item)
                now = datetime.datetime.now()
                now = now.strftime('%Y-%m-%d %H:%M:%S')
                sql_query = 'INSERT INTO baiduNews(ranking, title, datetime, hot) VALUES ("%s", "%s", "%s", "%s")' % (
                            item['index'], item['title'], now, item['hot'])
                cursor.execute(sql_query)
                print('Save into mysql')
            db.commit()
        except pymysql.MySQLError as e:
            db.rollback()
            print(e.args)
            return
    except pymysql.MySQLError as e:
        print(e.args)
        return

def check_mysql():
    try:
        db = pymysql.connect(host='localhost', port=3306, user='root', passwd='123456',
                             db='crawls', charset='utf8')
        cursor = db.cursor()
        cursor.execute('use crawls;')
        sql_query = 'SELECT * FROM baiduNews'
        results = cursor.execute(sql_query)
        print(results)
    except pymysql.MySQLError as e:
        print(e.args)

def main():
    html = get_html(URL)
    items = parse_html(html)
    save_to_mysql(items)
    #check_mysql()

if __name__ == '__main__':
    main()
    原文作者:authetic
    原文地址: https://segmentfault.com/a/1190000018071747
    本文转自网络文章,转载此文章仅为分享知识,如有侵权,请联系博主进行删除。
点赞